Editorial

The price of health

The burden of non-communicable diseases people with diabetes to fully participate and operate
(NCDs) cannot be understated. Chronic diseases, efficiently in the workforce. However, quantifying the
including cardiovascular diseases, cancers, chronic total indirect cost is difficult due to not only limited

Ikon Images/Patrick George/

respiratory diseases, and diabetes, are responsible for availability of data but also due to differing criteria

Science Photo Library

41 million deaths, almost three-quarters of all deaths applied. For example, some definitions also encompass
globally. Even more alarming is that 17 million deaths productivity loss from relatives of the patient who
occur before the age of 70 years and of those, 86% occur drop out from either school or work and not just the
in low-income and middle-income countries (LMICs). individual. For more facts on NCDs see
However, NCD burden on an individual or population In recognition of the magnitude of NCDs and mental https://www.who.int/news-
room/fact-sheets/detail/
level cannot be captured by mortality and morbidity health, Sustainable Development Goal (SDG) 3.4 aims noncommunicable-diseases#:~:
statistics alone. Along with the overwhelming loss to reduce premature NCD mortality by a third by 2030. text=Noncommunicable%20
diseases%20(NCDs)%20kill%20
of life comes a financial toll with huge consequences This is a commendable yet very ambitious target given 41,%2D%20and%20
on individuals and families, health systems, and the increasing prevalence of NCDs in recent years middle%2Dincome%20countries

governments. and the large scale public health campaigns needed. For more on global costs of
NCDs see Articles
The estimated global cost of all preventable NCDs and It is of great concern that projections of global economic Lancet Glob Health 2023;
mental health conditions for 2020–30 is US$301·8 billion. burden of diabetes and its complications in adults 11: e32–39
For more on the American
Of that, type 2 diabetes accounts for 2% of preventable show that even if countries meet the SDG target, the Diabetes Association Report see
cases but 9% of costs. In the USA, on average, people economic burden in 2030 will be 61% higher than in Diabetes Care 2024; 47: 26–43

with diagnosed diabetes have medical expenditures
that are more than double what would be expected for
people without diabetes. This expenditure is mostly due
to increasing inpatient hospital stays and an increase
in price of prescription medications such as insulin and
other glucose-lowering agents. A new American Diabetes
Association Report found that after adjusting for inflation,
the total cost of insulin and other medications to manage
blood glucose increased by 26% from 2017 to 2022.
Inadequate insurance coverage and high out of pocket
expenses cause substantial distress and can lead to people
changing to cheaper medication or stopping medicines
altogether. These issues can be greater in LMICs, where
almost 80% of people with diabetes live but fewer than
one in ten people receive coverage of guideline-based
comprehensive diabetes treatment.
Increases in direct costs of diabetes are well known to
be detrimental to health-care economies, but often the
impact of indirect costs is unrecognised. Missing work
or reduced work productivity due to ill health alongside
a reduced workforce participation due to disability or
premature mortality are responsible for almost a quarter
of the total estimated national cost of diabetes in the
USA. In other countries such as Brazil, indirect costs
accounted for the highest proportion of diabetes costs.
This highlights the importance of supportive working
environments and workplace policies that can allow
2015. Therefore, prevention is key and by investing For more on patients’
perspective about the cost of
in and implementing cost-effective strategies, which diabetes management see
can range from reformulation policies for healthier Mayo Clinic Proceedings 2021;
5: 898–906
food and drinks to providing preventative foot care
For more on the state of
for people with diabetes, not only will mortality and diabetes coverage in LMICs see
Articles Lancet Healthy Longev
morbidity improve, but so will economic benefit.
2021; 2: e340–51
In 2017, the World Health Assembly endorsed the For more on direct and indirect
NCD Best Buys, a package of 16 affordable and evidence costs of diabetes in Brazil see
Ann Glob Health 2022; 88: 14
based NCD interventions aimed to prevent and manage
For more on global economic
key NCDs. A paper from the NCD Countdown 2030 burden of diabetes in adults:
collaborators reported that an additional US$18 billion projections from 2015 to 2030
see Diabetes Care 2018;
spent annually on a package of interventions, fully 5: 963–70
aligned with the NCD best buys, could save For more on NCD Countdown
2030 see Health Policy
39 million lives, and generate an average net economic Lancet 2022; 399: 1266–78
benefit of US$2·7 trillion by 2030. Although this would For more on Heath4Life see
be a struggle to finance on domestic resources alone https://cdn.who.int/media/docs/
default-source/unitaf/uniatf-
for some countries, initiatives such as the Health4Life mptf-flyer_071021_final.pdf
Fund (the UN NCD and mental health multipartner trust
fund established by WHO, UNDP, and UNICEF) grant
crucial funds that enable a stepped-up NCD country led,
inclusive management.
The problem is clear. NCDs are growing and the
associated costs are enormous. It is essential that
a unified approach between governments, interna­
tional organisations, and pharmaceutical companies
is adopted if the burdens and costs of NCDs are to be
mitigated. ■ The Lancet Diabetes & Endocrinology

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 83

 Comment
Aspirin and diabetes prevention among healthy older
adults—practice-changing or hypothesis-generating?
The global prevalence of diabetes—the vast majority
of which is type 2 diabetes—increases steeply with
age, affecting approximately one in four adults aged
75–79 years.1 As such, diabetes is an exceedingly
Godruma/Getty Images
common condition among older adults. Although
younger populations generally experience the highest
risk of adverse diabetes-related outcomes, owing
Published Online to greater cumulative exposure, older adults with
December 21, 2023
https://doi.org/10.1016/
incident diabetes face an increased risk of premature
S2213-8587(23)00363-7 death, cardiovascular disease, multimorbidity, and
See Articles page 98 functional disability.2 These observations, in the
context of worldwide population ageing, highlight
the need for improved preventive strategies in this
high-risk group.
Systemic inflammation appears to influence the onset
and progression of diabetes.3 Hence, immunomodulatory
therapies targeting this pathophysiological pathway have
been proposed for diabetes prevention and treatment.
Aspirin is a widely available, low-cost, and well-studied
compound with established anti-inflammatory proper­
ties; however, its potential in the prevention of diabetes
among older adults is uncertain.
In The Lancet Diabetes & Endocrinology, Sophia Zoungas
and colleagues4 report the findings of a post-hoc analysis
of the ASPREE (Aspirin in Reducing Events in the Elderly)
trial that aimed to address this important global public
health question. ASPREE was a randomised, placebo-
controlled trial designed to evaluate whether low-dose
enteric-coated aspirin (100 mg daily) would prolong
disability-free survival among healthy older adults.
Participants aged 70 years or older (≥65 years if Black or
Hispanic ethnicity in the USA) were enrolled in Australia
and the USA between March 10, 2010, and Dec 24, 2014.
In the previously reported primary analysis,5 aspirin
did not prolong 5-year disability-free survival when
compared with placebo. Aspirin was associated with
a 14% increased rate of all-cause mortality, driven
by cancer-related death, and a 38% increased rate of
major bleeding.5,6 In an ASPREE substudy,7 aspirin was
additionally shown to be associated with a higher rate
of serious falls. These findings, combined with others
from contemporary aspirin trials, subsequently led to
international recommendations against the routine use
84
of aspirin among adults aged 60 years and older and
who were free from cardiovascular disease.8 However,
the substantial enrolment of ASPREE, the long duration,
and the high rate of study drug adherence provide
a unique opportunity to advance understanding of
aspirin-related glycaemic effects.
In this analysis by Zoungas and colleagues,4 including
16 209 ASPREE participants, nearly one in 15 developed
incident diabetes over a median follow-up of 4·7 years.
Compared with placebo, treatment with daily low-dose
aspirin was associated with a 15% reduction in the rate of
incident diabetes (hazard ratio 0·85 [95% CI 0·75–0·97];
p=0·013), with greater absolute benefits among men
compared with women. Aspirin also significantly slowed
the rate of increase of fasting plasma glucose (FPG). The
findings were consistent across all major subgroups
(including race and frailty status) as well as in sensitivity
analyses evaluating different definitions of incident
diabetes and competing risk of all-cause death. Consistent
with the main ASPREE analysis, aspirin therapy was
associated with a 44% increased rate of major bleeding
(hazard ratio 1·44 [95% CI 1·21–1·72]; p<0·0001),
primarily gastrointestinal bleeding.
Zoungas and colleagues4 should be congratulated
for a well-performed, timely, and balanced analysis
addressing an important question relevant to
clinical and research efforts involving older adults.
However, several limitations should be highlighted.
Foremost, ASPREE included predominantly White
participants from two high-income countries, limiting
generalisability to other populations. Furthermore,
incident diabetes in this analysis was defined by either
self-report, initiation of glucose-lowering medications,
or a single FPG level of 7·0 mmol/L or more. The absence
of serial FPG testing and insufficient capture of other
diagnostic tests for diabetes, including HbA1c, highlights
the potential for misclassification. Finally, data on
C-reactive protein and other markers of inflammation
were not available, which limited mechanistic insights.
Taken together, findings from the ASPREE pro­
gramme emphasise a foundational concept in the care
of all, but especially older adults, which is net benefit.
In this analysis, the absolute difference between the
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

aspirin and placebo groups in the rate of incident
diabetes (–2·1 per 1000 person-years) was offset by the
absolute difference in the rate of major bleeding events
(2·5 per 1000 person-years). As such, when considering
the increased rate of all-cause death observed in the
primary trial, these findings do not provide sufficient
rationale for a change in routine practice. However, this
effort amplifies the need for vigilant diabetes screening
and treatment across the lifespan and extends findings
from previous studies3 supporting the potential role of
inflammation-targeted strategies in primary diabetes
prevention. For aspirin, precision approaches might
identify subpopulations that are likely to derive the
maximal net cardiometabolic benefit (eg, those with
the highest risk of incident diabetes and cardiovascular
disease but the lowest risk of bleeding and progressive
cancer).9 Appreciation of these wider benefits might
also inform decision-making in selected secondary
prevention populations for whom reappraisal of the
role of lifelong aspirin therapy has been suggested.10
VRA reports fees for consultancy from Applied Therapeutics, Pfizer, Novo
Nordisk, Sanofi, Mediflix, and Fractyl; fees from contracts from Applied
Therapeutics, Eli Lilly, Novo Nordisk, Sanofi, and Fractyl; and their spouse is an
employee of Johnson & Johnson. JWO declares no competing interests.
Association of glycaemic index and glycaemic load with
type 2 diabetes and related conditions in prospective studies
There is debate over the nature of the carbohydrate-
containing foods that should be recommended
to reduce the risk of type 2 diabetes and
related conditions. In particular, it is not well
established whether glycaemic index (GI) is
a rel­evant dietary factor for the prevention of
chronic diseases in the general population. In
The Lancet Diabetes & Endocrinology, David Jenkins and
colleagues1 report a meta-analysis of large prospective
cohorts (≥100 000 participants) on the associations
between GI and glycaemic load (GL) and the incidence
of type 2 diabetes, cardiovascular disease, diabetes-
related cancers, and all-cause mortality. This study was
done in response to a WHO-sponsored meta-analysis
that concluded that, in relation to the associations of
low GI and GL with the incidence of type 2 diabetes
and other crucial clinical outcomes in observational
studies, the available evidence “is graded as low or very
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Comment
John W Ostrominski, *Vanita R Aroda
varoda@bwh.harvard.edu
Cardiovascular Division (JWO) and Division of Endocrinology (JWO, VRA), Brigham
and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USA
1 Ong KL, Stafford LK, McLaughlin SA, et al. Global, regional, and national
burden of diabetes from 1990 to 2021, with projections of prevalence to
2050: a systematic analysis for the Global Burden of Disease Study
2021. Lancet 2023; 402: 203–34.
2 ElSayed NA, Aleppo G, Aroda VR, et al. 13. Older adults: standards of care in
diabetes-2023. Diabetes Care 2023; 46 (suppl 1): S216–29.
3 Rohm TV, Meier DT, Olefsky JM, Donath MY. Inflammation in obesity,
diabetes, and related disorders. Immunity 2022; 55: 31–55.
4 Zoungas S, Zhou Z, Owen AJ, et al. Daily low dose aspirin and incident type 2
diabetes in community-dwelling healthy older adults: a post-hoc analysis of
efficacy and safety in the ASPREE randomised placebo-controlled
trial. Lancet Diabetes Endocrinol 2023; published online Dec 21.
https://doi.org/10.1016/S2213-8587(23)00327-3.
5 McNeil JJ, Woods RL, Nelson MR, et al. Effect of aspirin on disability-free
survival in the healthy elderly. N Engl J Med 2018; 379: 1499–508.
6 McNeil JJ, Nelson MR, Woods RL, et al. Effect of aspirin on all-cause mortality
in the healthy elderly. N Engl J Med 2018; 379: 1519–28.
7 Barker AL, Morello R, Thao LTP, et al. Daily low-dose aspirin and risk of
serious falls and fractures in healthy older people: a substudy of the ASPREE
randomized clinical trial. JAMA Intern Med 2022; 182: 1289–97.
8 Fegers-Wustrow I, Gianos E, Halle M, Yang E. Comparison of American and
European guidelines for primary prevention of cardiovascular
disease: JACC Guideline Comparison. J Am Coll Cardiol 2022; 79: 1304–13.
9 Cofer LB, Barrett TJ, Berger JS. Aspirin for the primary prevention of
cardiovascular disease: time for a platelet-guided approach.
Arterioscler Thromb Vasc Biol 2022; 42: 1207–16.
10 Jacobsen AP, Raber I, McCarthy CP, et al. Lifelong aspirin for all in the
secondary prevention of chronic coronary syndrome: still sacrosanct or is
reappraisal warrranted? Circulation 2020; 142: 1579–90.
low quality as a result of high risk of bias, imprecision,
Paul Rapson/Science Photo Library
and inconsistencies”.2
Previous meta-analyses showed significant associa­
tions between the habitual consumption of low GI and
GL foods and a reduced incidence of diabetes and related
conditions.3,4 However, there are two features that make
this study novel and scientifically relevant. First, the See Articles page 107
selection of large cohorts allowed for sufficient statistical
power and minimised the variability seen with small
study assessments. Second, direct comparisons within
the same cohorts were made between the health benefits
associated with low GI and GL and those associated with
high fibre and wholegrain intake.
Jenkins and colleagues1 provide strong and high-quality
evidence that a low GI diet is associated with a reduced
risk of type 2 diabetes, cardiovascular disease, diabetes-
related cancers, and all-cause morta­ lity, and that the
magnitude of the risk reduction is similar to the widely
85
 Articles

Association of glycaemic index and glycaemic load with

type 2 diabetes, cardiovascular disease, cancer, and all-cause
mortality: a meta-analysis of mega cohorts of more than
100 000 participants
David J A Jenkins, Walter C Willett, Salim Yusuf, Frank B Hu, Andrea J Glenn, Simin Liu, Andrew Mente, Victoria Miller, Shrikant I Bangdiwala,
Hertzel C Gerstein, Sabina Sieri, Pietro Ferrari, Alpa V Patel, Marjorie L McCullough, Loïc Le Marchand, Neal D Freedman, Erikka Loftfield,
Rashmi Sinha, Xiao-Ou Shu, Mathilde Touvier, Norie Sawada, Shoichiro Tsugane, Piet A van den Brandt, Kerem Shuval, Tauseef Ahmad Khan,
Melanie Paquette, Sandhya Sahye-Pudaruth, Darshna Patel, Teenie Fei Yi Siu, Korbua Srichaikul, Cyril W C Kendall, John L Sievenpiper on behalf of
Clinical Nutrition & Risk Factor Modification Centre Collaborators*

Summary
Background There is debate over whether the glycaemic index of foods relates to chronic disease. We aimed to assess Lancet Diabetes Endocrinol
the associations between glycaemic index (GI) and glycaemic load (GL) and type 2 diabetes, cardiovascular disease, 2024; 12: 107–18

diabetes-related cancers, and all-cause mortality. See Comment page 85

*Collaborators listed at the end
of the Article
Methods We did a meta-analysis of large cohorts (≥100 000 participants) identified from the Richard Doll Consortium.
We searched the Cochrane Library, MEDLINE, PubMed, Embase, Web of Science, and Scopus for cohorts that Department of Nutritional
Sciences (Prof D J A Jenkins MD,
prospectively examined associations between GI or GL and chronic disease outcomes published from database A J Glenn PhD, T A Khan MBBS,
inception to Aug 4, 2023. Full-article review and extraction of summary estimates data were conducted by M Paquette MSc,
three independent reviewers. Primary outcomes were incident type 2 diabetes, total cardiovascular disease (including S Sahye-Pudaruth MPH,
mortality), diabetes-related cancers (ie, bladder, breast, colorectal, endometrial, hepatic, pancreatic, and non-Hodgkin D Patel BA, T F Y Siu BSc,
K Srichaikul MD,
lymphoma), and all-cause mortality. We assessed comparisons between the lowest and highest quantiles of GI and C W C Kendall PhD,
GL, adjusting for dietary factors, and pooling their most adjusted relative risk (RR) estimates using a fixed-effects Prof J L Sievenpiper MD) and
model. We also assessed associations between diets high in fibre and whole grains and the four main outcomes. The Department of Medicine
(Prof D J A Jenkins MD,
study protocol is registered with PROSPERO, CRD42023394689.
Prof J L Sievenpiper MD),
Temerty Faculty of Medicine,
Findings From ten prospective large cohorts (six from the USA, one from Europe, two from Asia, and one international), University of Toronto, Toronto,
we identified a total of 48 studies reporting associations between GI or GL and the outcomes of interest: 34 (71%) on ON, Canada; Li Ka Shing
Knowledge Institute
various cancers, nine (19%) on cardiovascular disease, five (10%) on type 2 diabetes, and three (6%) on all-cause
(Prof D J A Jenkins,
mortality. Consumption of high GI foods was associated with an increased incidence of type 2 diabetes (RR 1·27 Prof J L Sievenpiper), Clinical
[95% CI 1·21–1·34]; p<0·0001), total cardiovascular disease (1·15 [1·11–1·19]; p<0·0001), diabetes-related Nutrition & Risk Factor
cancer (1·05 [1·02–1·08]; p=0·0010), and all-cause mortality (1·08 [1·05–1·12]; p<0·0001). Similar associations were Modification Centre
(Prof D J A Jenkins, M Paquette,
seen between high GL and diabetes (RR 1·15 [95% CI 1·09–1·21]; p<0·0001) and total cardiovascular
S Sahye-Pudaruth, D Patel,
disease (1·15 [1·10–1·20]; p<0·0001). Associations between diets high in fibre and whole grains and the four main T F Y Siu, K Srichaikul,
outcomes were similar to those for low GI diets. Prof J L Sievenpiper), and
Division of Endocrinology and
Metabolism (Prof D J A Jenkins,
Interpretation Dietary recommendations to reduce GI and GL could have effects on health outcomes that are similar
Prof J L Sievenpiper),
to outcomes of recommendations to increase intake of fibre and whole grain. St Michael’s Hospital, Toronto,
ON, Canada; Toronto 3D
Funding Banting and Best and the Karuna Foundation. Knowledge Synthesis and
Clinical Trials Unit, Toronto,
ON, Canada (Prof D J A Jenkins,
Copyright © 2023 World Health Organization. Published by Elsevier Ltd. All rights reserved. A J Glenn, T A Khan,
C W C Kendall,
Introduction use of high-fibre and wholegrain carbohydrate foods for Prof J L Sievenpiper);
Department of Epidemiology
The International Diabetes Federation predicts that the management of type 2 diabetes, cardiovascular disease, (Prof W C Willett MD,
global prevalence of diabetes will increase from and cancer.1 However, glycaemic index (GI) was not Prof F B Hu PhD) and
537 million people to approximately 643 million people considered to be a relevant dietary factor for the Department of Nutrition
in 2030 and 783 million people by 2045, with 75% of prevention of chronic diseases, a conclusion endorsed (Prof W C Willett, Prof F B Hu,
A J Glenn), Harvard T H Chan
adults with diabetes living in rapidly westernising low- from the executive summary of the WHO report.1,2 School of Public Health,
income and middle-income countries. However, dietary GI is the degree to which 50 g of carbohydrate in a food Boston, MA, USA; Population
recommendations, including carbohydrate food intake, raises blood glucose concentrations postprandially, as Health Research Institute
for people with diabetes are debated. In 2019, The Lancet a percentage of a standard 50 g of glucose or 50 g of (Prof S Yusuf DPhil, V Miller PhD,
S I Bangdiwala PhD,
published a WHO-sponsored report that supported the starch in white bread,3 whereas glycaemic load (GL) is

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 107

 Articles
Prof H C Gerstein MD),
Department of Medicine
(Prof S Yusuf, V Miller,
Prof A Mente PhD), Department
of Health Research Methods,
Evidence, and Impact
(Prof A Mente, S I Bangdiwala),
McMaster University and
Hamilton Health Sciences,
Hamilton, ON, Canada;
Department of Medicine,
Channing Division of Network
Medicine, Brigham and
Women’s Hospital and Harvard
Medical School, Boston, MA,
USA (Prof F B Hu); Center for
Global Cardiometabolic Health,
Department of Epidemiology,
Department of Medicine, and
Department of Surgery, Brown
University, Providence, RI, USA
(Prof S Liu ScD); Epidemiology
and Prevention Unit,
Fondazione IRCCS Istituto
Nazionale dei Tumori, Milan,
Italy (S Sieri PhD); Nutrition
and Metabolism Branch,
International Agency for
Research on Cancer, Lyon,
France (P Ferrari PhD);
Department of Population
Science, American Cancer
Society, Atlanta, GA, USA
(A V Patel PhD,
M L McCullough ScD);
Epidemiology Program,
University of Hawaii Cancer
Center, Honolulu, HI, USA
(Prof L Le Marchand MD);
Division of Cancer
Epidemiology and Genetics,
National Cancer Institute,
National Institutes of Health,
Bethesda, MD, USA
(N D Freedman PhD,
E Loftfield PhD, R Sinha PhD);
Department of Medicine,
Vanderbilt University Medical
Center, Nashville, TN, USA
(Prof X-O Shu PhD); Université
Sorbonne Paris Nord and
Université Paris Cité, INSERM,
INRAE, CNAM, Center of
Research in Epidemiology and
Statistics, Nutritional
Epidemiology Research Team,
Bobigny, France
(M Touvier PhD); French
Network for Nutrition and
Cancer Research, Jouy-en-Josas,
France (M Touvier); Division of
Cohort Research, National
Cancer Center Institute for
Cancer Control, Tokyo, Japan
(Prof N Sawada PhD,
Prof S Tsugane MD);
International University of
Health and Welfare Graduate
School of Public Health, Tokyo,
Japan (Prof S Tsugane); GROW
School for Oncology and
108
Research in context
Evidence before this study
This study was done in response to the WHO-sponsored
meta-analysis published in The Lancet in 2019, which concluded
that GI and GL had little relevance to incidence of chronic disease
or mortality. The report concluded that although evidence for
low GI and GL was reviewed, there was little consistency
observed in benefit on mortality or non-communicable disease
incidence from observational assessments (very low to
moderate certainty evidence), and little to no improvement in
cardiometabolic risk factors observed in randomised controlled
trials (very low to high certainty evidence). This conclusion was
made despite data in the appendix suggesting significant
associations with diabetes, mortality from cardiovascular
disease, mortality from stroke, and incidence of breast cancer.
Other large cohort assessments and meta-analyses have
indicated an association between high GI diets and an increased
risk of diabetes and cardiovascular disease. Additionally,
a 2021 meta-analysis of clinical trials showed the beneficial
effects of low GI and GL diets on glycaemic control and
cardiometabolic risk factors in people with diabetes. In view of
the importance of carbohydrate foods as the major nutrient in
all diets globally, its classification is important to ensure that the
greatest health benefits can be derived from dietary advice.
Furthermore, randomised controlled trials have suggested that
slowing carbohydrate absorption by the use of acarbose reduces
the incidence of diabetes.
We identified large cohorts (>100 000 participants) from
the Richard Doll Consortium, as a consistent source of
information, assessing associations between GI or GL and
one or more or the following outcomes: diabetes incidence,
cardiovascular disease and its components, diabetes-related
cancers, and all-cause mortality. To allow direct comparison
with GI and GL, we identified cohorts that also reported data on
the total impact of the food eaten, achieved through
multiplication of the GI by the total amount of
carbohydrate in the food or diet.4 Given that GI was
developed to identify foods with a low and high potential
of raising blood glucose concentrations, exclusion of this
food characteristic seems counterintuitive. Systematic
reviews and meta-analyses published in 2022 have
suggested that both high GI and GL are associated
with an increased risk of chronic diseases, such as
type 2 diabetes and cardiovascular disease.5 Therefore,
we aimed to assess the associations between both GI
and GL and the incidence of type 2 diabetes, related
diseases, and mortality using large cohorts of at least
100 000 participants. We examined whether low GI diets
had potential benefits in people with type 2 diabetes,
total cardiovascular disease (ie, cardiovascular disease,
coronary heart disease and myocardial infarction, stroke,
mortality from cardiovascular disease, mortality from
coronary heart disease and myocardial infarction, and
mortality from stroke), and cancers positively associated
fibre and whole grains. We searched Cochrane Library,
MEDLINE, PubMed, Embase, Web of Science, and Scopus,
with no language restrictions, from database inception to
Aug 4, 2023, to identify relevant cohort associations. We
analysed data using the generic inverse variance method with
fixed-effects and random-effects models, with additional
analyses to assess the certainty of the evidence. We also
attempted to use data that had adjusted each key carbohydrate
component by the two other carbohydrate quality markers,
so that truly independent associations could be assessed.
Added value of the study
This study has added value in showing that the risk of diabetes
and related conditions is reduced by the consumption of low GI
food and that these associations are similar to the widely
accepted benefits of diets high in fibre and whole grains. These
findings justify the combination of GI with fibre and whole
grains in dietary recommendations to reduce the risk of
diabetes and related chronic diseases.
Implications of all the available evidence
These findings have the potential to increase the value of dietary
advice, if included in recommendations for the treatment of
individuals at risk of cardiometabolic diseases. The findings
could also increase the value of dietary advice for the food
industry and agriculture in improving the quality of
carbohydrate foods globally. This study might encourage further
work on the effects of postprandial glucose fluctuations (mean
ambulatory glucose excursions); for example, on measures of
oxidative stress in relation to pancreatic dysfunction and,
ultimately, incidence of diabetes. These findings might also
encourage testing of GI and GL in long-term trials, such as after
successful diabetes remission induced by weight loss, to
establish clinical usefulness more directly.
with diabetes (ie, diabetes-related cancers). We also
investigated whether these GI and GL associations were
similar to those observed with diets high in fibre or
whole grain by comparing associations directly in the
same cohorts.
Methods
Search strategy and selection criteria
We did a meta-analysis of large cohorts
(≥100 000 participants) identified from the Richard Doll
Consortium. We selected large cohorts with high
participant numbers to provide consistent results, gain
statistical power, minimise extreme variability that can
result from small study assessments, examine less
common outcomes, and allow comparisons with fibre
and whole grains as accepted indicators of carbohydrate
quality. We searched the Cochrane Library, MEDLINE,
PubMed, Embase, Web of Science, and Scopus, with no
language restrictions, for cohorts that prospectively
examined associations between GI or GL and diabetes,
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

cardiovascular disease, diabetes-related cancers, and all-
cause mortality, published between database inception,
and Aug 4, 2023. Search terms included GI, GL, fibre,
and whole grain, with one or more of the following:
diabetes, cardiovascular disease (ie, coronary heart
disease, myocardial infarction, stroke), or cancer, both
incidence and related deaths, and all-cause mortality. For
Embase, we used the following search terms: (glycaemic
index OR glycaemic load OR fibre OR wholegrain) AND
(cardiovascular disease OR coronary artery disease OR
myocardial infarction OR stroke OR cardiovascular death
OR mortality OR cancer OR type 2 diabetes). A full list of
search terms can be found in the appendix (pp 8–14).
Data on fibre and whole grain were only included if data
on GI or GL had been measured in the same cohort
(appendix pp 147–148).
Full-article review and data extraction of summary
estimates were conducted by three independent
reviewers (TFYS, XW, and FL), with disagreements
reconciled through consensus.
The quality of the prospective cohorts was assessed
with the Newcastle-Ottawa Scale (NOS).6 A score of up
to 9 was awarded on the basis of cohort selection
(ie, representativeness, selection of non-exposed cohort,
exposure assessment, and outcome not present at base­
line), ascertainment of outcome (ie, follow-up length,
adequacy of follow-up, and outcome assessment),
and comparability (controlling for one prespecified
primary variable [age] and several secondary confound­
ing variables, depending on the outcome; appendix
pp 124–125, 199). These confounding variables were
selected on the basis of their association with risk of
type 2 diabetes, cardiovascular disease, cancer, and all-
cause mortality.
We followed the Cochrane Handbook for Systematic
Reviews and Interventions,7 and reported the results
according to Meta-analysis of Observational Assessments
in Epidemiology8 and PRISMA guidelines. Ethics approval
was not required for this analysis of published data.
Outcomes
The primary outcome of interest was the relative
risk (RR) from the most adjusted model in which authors
had attempted to control their findings for all available
relevant covariates at their disposal (appendix pp 149–154).
The four primary outcomes were incident type 2 diabetes,
total cardiovascular disease (including incidence of
coronary heart disease, myocardial infarction, and stroke,
and mortality from cardiovascular disease), cancers with
an increased risk associated with diabetes (ie, bladder,
breast, colorectal, endometrial, hepatic, pancreatic, and
non-Hodgkin lymphoma),9 and all-cause mortality.
Secondary outcomes were the compo­ nents
of cardiovascular disease (ie, coronary heart disease
and myocardial infarction, stroke, mortality from
cardiovascular disease, mortality from coronary heart
disease and myocardial infarction, and mortality from
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Articles
stroke), cancers unrelated to diabetes (ie, biliary, brain, Developmental Biology, and
head and neck, oesophageal, kidney, lung, melanoma, Department of Epidemiology,
Care and Public Health
myeloid leukaemia, ovarian, stomach, and thyroid),10 and Research Institute–School for
the individual type of cancer (including prostate). Public Health and Primary Care,
Maastricht University Medical
Data analysis Centre, Maastricht,
Netherlands
Food GI values used in food frequency questionnaires (Prof P A van den Brandt PhD);
were derived from the international GI tables published The Cooper Institute, Dallas,
by Atkinson and colleagues11 and were supplemented TX, USA (K Shuval PhD); College
with local food tables specifically for Asian food frequency of Pharmacy and Nutrition,
University of Saskatchewan,
questionnaires. For all cohorts, GI was measured against SK, Canada (C W C Kendall)
the glucose scale (where glucose was equivalent to 100), Correspondence to:
except for the PURE cohort, in which the GI for bread David J A Jenkins, Department
was converted to a glucose GI for use in the meta-analyses. of Nutritional Sciences, Temerty
GL values were dependent on GI values, given that GL is Faculty of Medicine, University of
Toronto, Toronto, ON M5S 1A8,
the total effect of the food eaten achieved through Canada
multiplication of the GI by the total amount of david.jenkins@utoronto.ca
carbohydrate in the food or diet, divided by 100. Where For the International Diabetes
cohorts reported sex separately (eg, the Japan Public Federation see https://
Health Center-based Prospective Study, Shanghai diabetesatlas.org/
cohorts, and the Harvard Nurses and Health Professional For the Richard Doll Consortium
see https://www.
cohorts), sex-specific data were also included from
richarddollconsortium.org/
cohorts with a combined participant number of both
See Online for appendix
sexes of at least 100 000 participants.
We assessed comparisons between the lowest and
highest quantiles using the most adjusted model. In the
case of multiple assessments, the most recent
assessment was used. Log-transformed RRs were pooled
with a fixed-effects model to estimate the average true
effect size in these cohorts,12 and the results were
expressed as pooled RRs with 95% CIs. We also reported
a random-effects model separately, which assumed that
the collection of assessments represented a random
sample of a larger population. For the random-effects
results, we applied the DeSimonian and Laird random-
effects model13 when the number of assessments
exceeded five, and the Sidik-Jonkman adjustment14 to
address the uncertainty in tau-squared estimation for
five or fewer assessments.
Where available, we also compared the associations
between GI and GL and the four main outcomes with
those between dietary fibre and whole grains and these
outcomes, if examined in the same cohorts used for
GI or GL analyses. For these comparisons, GI and GL
associations were reversed to allow the lowest quantile
versus the highest quantile to be compared with the
potential benefit of a diet high in fibre and whole grain.
Heterogeneity between the assessments was assessed
with the Cochran Q statistic (p<0·1) and was quantified
by I². We interpreted an I² value of at least 50% as
substantial heterogeneity (with p<0·1).15 Sources of
heterogeneity were explored as a subgroup analysis for
the main four outcomes. Subgroup analyses were
assessed by sex, number of participants, follow-up
duration, average age, race or ethnicity, continent,
study quality (NOS), and funding source. We used
a meta-regression to test for differences within
109
 Articles

If at least ten cohort comparisons were available,

2673 records identified publication bias was assessed by visual inspection of
514 Embase
354 MEDLINE the funnel plots and with Begg’s and Egger’s tests
63 Cochrane (with p<0·1 considered to be significant).18,19 The Duval
572 PubMed
494 Web of Science
and Tweedie trim and fill method was applied to adjust
665 Scopus for funnel plot asymmetry by imputing the missing
11 manual cohort data.20
We estimated a linear dose–response relationship using
2464 excluded a one-stage random-effects model, with restricted maxi­
1486 duplicates mum likelihood methods assuming a linear function.21
308 no outcomes of interest
447 no exposure of interest The non-linear dose response model was assessed by use
73 letters, protocols, or abstracts of restricted cubic splines, with three knots according
66 reviews to Harrell’s recommended percentiles (ie, 10%, 50%,
21 subgroup analyses
38 cohort too small (<100 000 participants) and 90%).22 Departure from linearity was assessed with
8 randomised controlled trials Wald’s test.
4 case-control or cross-sectional studies
4 animal studies The overall certainty of the associations were assessed
9 wrong population with two approaches, NutriGrade23 and GRADE.24,25
NutriGrade was used to address the criticism that
GRADE might not adequately assess the potential value
209 articles assessed for full
eligibility of prospective cohort dietary assessments.26 For consis­
tency, we modified the NutriGrade publication bias
domain to align with GRADE.
138 excluded
47 no GI or GL data*
All data were analysed with Review Manager (version 5.4),
27 no exposure of interest the Nordic Cochrane Centre, the Cochrane Collaboration,
24 alternate publications and STATA (version 16.1). The study protocol is registered
15 subgroup analyses
7 no outcomes of interest with PROSPERO, CRD42023394689.
8 letters, protocols, or abstracts
5 case-control studies
3 reported fewer than three quantiles
Role of the funding source
2 reviews The funders of the study had no role in study design,
data collection, data analysis, data interpretation, or
writing the report.
71 studies included in the analysis
10 cohorts
48 GI or GL studies Results
35 fibre or wholegrain
studies from cohorts
From ten prospective large cohorts of the Richard Doll
with GI or GL Consortium, we identified a total of 48 GI or GL studies
assessments† reporting the outcomes of interest, of which 101 GI
associations and 105 GL associations were reported
Figure 1: PRISMA flow diagram (figure 1; appendix pp 15–28, 158–163). Nine assessments
*To match data on fibre or wholegrain exposure. †12 of these studies overlap
with GI or GL studies.
reported sex separately. Diet assessment was com­
pleted at baseline for all 48 studies.27–74 14 (29%)
categories, stratifying by cohort size, follow-up, NOS, studies reassessed diet after baseline (appendix
mean age, and location.7 We also performed a continuous pp 15–28).28,30,32,35,37,46–48,51,53,64,67,69,74 Cohort size ranged
meta-regression analysis to assess the heterogeneity from approximately 104 000 to 567 000 participants.
between studies, using the continuous measures of Six (60%) cohorts came from the USA (Harvard cohorts
cohort size (number of participants), duration of follow- HPFS–NHS I–NHS II, CPS II, MEC, NIH-AARP,
up, and mean age as covariates. We computed E-values (the PLCO, and WHI), one (10%) from Europe (Greece,
minimum strength of association that an unmeasured Spain, Italy, France, Germany, Netherlands, the UK,
confounder needs to have with both the exposure and the Denmark, Sweden, and Norway; EPIC), two (20%) from
outcome to fully explain away the observed association Asia (China and Japan; SWHS–SMHS and JPHC), and
between exposure and outcome) for the point estimate one (10%) was an international cohort (PURE). Among
and the lower confidence interval of GI and GL with the ten prospective cohorts reporting associations
all four major outcomes to ascertain the effect of between GI or GL and the outcomes of interest,
a potential unmeasured or uncontrolled confounder on nine cohorts (34 [71%] studies) reported data on
the relationship between exposure and outcome.16 various cancer outcomes,27,29–34,36,37,41,42,44–46,48–54,56–60,62,63,65–68,71,73
We considered that any value less than 1·2 indicated five cohorts (nine [19%] studies) on cardiovascular
average confounding that could potentially explain away disease outcomes,35,39,40,47,61,69,70,72,74 five cohorts (five [10%]
the results.17 studies) on the incidence of type 2 diabetes,28,38,43,55,64

110 www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

 Articles

Number of Number of RR (95% CI) p value

observations cohorts

Type 2 diabetes 7 4 1·27 (1·21–1·34) <0·0001

Total cardiovascular disease 19 5 1·15 (1·11–1·19) <0·0001
Cardiovascular disease 1 1 1·14 (1·02–1·27) 0·019
Coronary heart disease 5 4 1·09 (1·01–1·17) 0·022
Stroke 3 3 1·17 (1·06–1·28) 0·001
Cardiovascular disease mortality 5 3 1·14 (1·08–1·21) <0·0001
Coronary heart disease mortality 2 1 1·33 (1·14–1·55) 0·0004
Stroke mortality 3 2 1·29 (1·10–1·50) 0·0015
Diabetes-related cancer 38 8 1·05 (1·02–1·08) 0·001
Colorectal cancer 8 5 1·11 (1·03–1·18) 0·0032
Breast cancer 8 6 1·05 (1·01–1·09) 0·016
Total mortality 5 3 1·08 (1·05–1·12) <0·0001

0·50 1·00 2·00

Favours higher GI Favours lower GI

Figure 2: GI exposure and type 2 diabetes, total cardiovascular disease, diabetes-related cancers, and mortality
Pooled estimates are expressed as RRs, where black squares represents the pooled RR of GI exposure, with corresponding 95% CIs. Estimates were the comparison
between extreme quantiles. The p values represent the analysis performed with the generic inverse variance method with a fixed-effects model (where p<0·05 is
significant). RR=relative risk. GI=glycaemic index.

and three cohorts (three [6%] studies) on all-cause
mortality.39,40,72 Mean population age was 56 years (SD 8)
and mean follow-up was 12·6 years (SD 5·4). These
cohorts formed the basis of our reported analyses,
in which the lowest GI quantile mean was 58 and
the highest was 67, with a relatively small range on the
glucose scale (where glucose was equivalent to 100) and
the bread scale (where GI bread was equivalent to 100;
appendix pp 15–28). The lowest GL quantile mean
was 116 and the highest was 180 (appendix p 28).
Associations with fixed-effects and random-effects
models were generally similar (appendix pp 29–105).
In the maximally adjusted model, the most commonly
used variables were age, sex, adiposity, smoking, family
history, energy intake, physical activity, and alcohol
intake (appendix pp 149–154). For the four main out­
comes, only 12 (17%) of 69 GI assessments controlled for
fibre and no fibre assessments controlled for GI; and
only 12 (17%) GI assessments controlled for wholegrain
and no wholegrain assessments controlled for GI.
Cohorts were assessed (NOS) to be of high quality,
receiving a score of 7 or higher (appendix pp 124–125).
Consumption of high GI foods was associated with an
increased incidence of type 2 diabetes and diabetes-
related diseases (figure 2; appendix pp 32, 36–37, 68–70).
The RRs were 1·27 (95% CI 1·21–1·34; p<0·0001; I²=71%)
for type 2 diabetes (figures 2 and 3), 1·15 (1·11–1·19;
p<0·0001; I²=35%) for total cardiovascular disease, and
1·05 (1·02–1·08; p=0·0010; I²=23%) for diabetes-related
cancers (figure 2; appendix pp 32, 36–37, 68–70).
GL had similar associations with type 2 diabetes
(RR 1·15 [95% CI 1·09–1·21]; p<0·0001; I²=60%,
appendix p 34) and total cardiovascular
disease (1·15 [1·10–1·20]; p<0·0001; I²=38%; appendix
pp 40–41). However, for cancer, these associations with
GL were either not significant or showed significant
inverse associations, possibly the result of an unmeasured
or uncontrolled confounder on the exposure–outcome
relationship (appendix pp 74–76).
GI showed a significant positive association with all-
cause mortality (RR 1·08 [95% CI 1·05–1·12; p<0.0001;
I²=90%), with a significant association observed in
women only (1·11 [1·06–1·16]; I²=0%; appendix p 64).
GL showed a similar direction of association with
all-cause mortality; however, this association was not
significant (appendix p 66). The associations between GI
and type 2 diabetes, mortality from cardio­ vascular
disease, and all-cause mortality also showed significant
dose–response relationships, which strength­ ened the
certainty of the evidence as assessed by GRADE
(appendix pp 110–123).
The subgroup analysis showed some variation when
stratifying by cohort size, follow-up, NOS, mean age, and
location. Statistical interaction for GI indicated a high RR
for smaller cohort size for type 2 diabetes and total
cardiovascular disease, whereas a larger cohort size
for all-cause mortality. A high RR was also observed
for a shorter follow-up duration for type 2 diabetes, but
a longer duration for total cardiovascular disease. For GL,
interaction was limited with a high RR for a smaller
cohort size and shorter follow-up duration for type 2
diabetes only. There were no interactions for other
comparisons (appendix pp 251–266). These interactions
might not have clinical relevance because the differences
between subgroups are minor, whereas the effect
estimate was in the same direction.
Apart from the association between GI or GL and
diabetes-related cancers and between GL and all-cause
mortality, all E-values for the point estimate and lower
confidence interval of GI and GL with all four major

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 111

 Articles

Cases Total Weight (%) RR (95% CI)

Female
SWHS 1608 64 227 9·4% 1·21 (1·03–1·43)
JPHC 500 36 864 2·2% 1·14 (0·81–1·60)
NHS II 4515 90 411 21·6% 1·20 (1·08–1·34)
NHS I 7400 74 248 36·6% 1·44 (1·33–1·56)
WHI 11 009 73 495 10·4% 0·99 (0·85–1·16)
Subtotal (95% CI) 25 032 339 245 80·2% 1·27 (1·20–1·34)
Heterogeneity: χ²=20·45, df=4 (p=0·0004); I²=80%
Test for overall effect: Z=8·40 (p<0·0001)
Male
JPHC 690 27 769 3·1% 1·19 (0·90–1·58)
HPFS 3112 40 498 16·7% 1·30 (1·15–1·47)
Subtotal (95% CI) 3802 68 267 19·8% 1·28 (1·15–1·44)
Heterogeneity: χ²=0·31, df=1 (p=0·58); I²=0%
Test for overall effect: Z=4·32 (p<0·0001)
Total (95% CI) 28 834 407 512 100% 1·27 (1·21–1·34)
Heterogeneity: χ²=20·78, df=6 (p=0·0020); I²=71%
Test for overall effect: Z=9·44 (p<0·0001) 0·5 0·7 1 1·5 2·0
Test for subgroup differences: χ²=0·02, df=1 (p=0·90), I²=0%
Favours higher GI Favours lower GI

Figure 3: Association between GI exposure and risk of type 2 diabetes by sex

All results are presented as RRs with 95% CIs. The diamonds represents the pooled risk estimate. Estimates were the comparison between extreme quantiles. The
p values represent the analysis performed with the generic inverse variance method with a fixed-effects model (where p<0·05 is significant). Heterogeneity between
studies was tested with the Cochran Q statistic (χ²) at a significance level of p<0·10 and quantified by the I² statistic, where an I² value of at least 50% was interpreted
as substantial heterogeneity. GI=glycaemic index. RR=risk ratio.

outcomes exceeded 1·2 (appendix p 266). For the specific
associations with E-values less than 1·2, confounders
could potentially explain some of the observed associations.
The E-values exceeding 1·2 for all other associations
indicate the robustness of the associations, given that
confounders would not be able to explain the observed
associations.
No evidence of publication bias was found by visual
inspection or formal testing with the Begg’s and Egger’s
test for the association between GI or GL and total
cardiovascular disease or diabetes-related cancers where
there were more than ten observations (appendix
pp 106–108). Furthermore, use of GRADE and
NutriGrade indicated reasonable confidence in the data
(appendix pp 126–146). For GRADE, GI associations
were considered to be low grade for diabetes and total
cardiovascular disease. A low grade was found for all-
cause mortality and diabetes-related cancers. The
components of total cardiovascular disease were also
low grade, except for mortality from cardiovascular
disease, which was mod­erate grade. For NutriGrade,
associations between GI and diabetes, total
cardiovascular disease, all-cause mortality, and diabetes-
related cancers were considered to be moderate grade.
The components of total cardiovascular disease were
also moderate grade.
Ten cohorts (HPFS–NHS I–NHS II, NIH-AARP, PLCO,
WHI, EPIC, SWHS–SMHS, JPHC, CPS II, PURE, and
MEC) provided data on GI or GL associations for at least
one outcome and nine cohorts (HPFS–NHS I–NHS II,
NIH-AARP, PLCO, WHI, EPIC, SWHS–SMHS, JPHC,
PURE, and MEC) also had data on fibre, whole grain, or
both (appendix pp 158–216).29–31,33,38,52,53,60,63,67,71,74–97 To allow
direct comparison with fibre and whole grain, GI
associations were reversed and explored as low versus
high GI for the outcomes (appendix pp 217–248).
Associations of the four main outcomes were similar in
the comparison of low GI with high fibre and whole grain
within the same cohorts (figures 4, 5; appendix
pp 219, 221, 223, 225). With the exception of cancer in the
fibre comparison, these comparisons were restricted to
only one or two comparisons. Among the GI comparisons,
five (22%) of 23 controlled for fibre and no fibre com­
parisons controlled for GI. Among the GI comparisons,
seven (50%) of 14 controlled for wholegrain and no
wholegrain comparisons controlled for GI.
Discussion
We found that a high GI diet was associated with type 2
diabetes, cardiovascular disease, diabetes-related cancers,
and all-cause mortality. Additionally, consumption of
high fibre and whole grains showed reductions in these
outcomes that were similar to the associations observed
for low GI.
Despite how only 12 (17%) of the 69 GI main outcome
assessments controlled for fibre, controlling for fibre or
not made little difference to GI assessments (appendix
pp 155–157). Our findings supplement previous reports
that GI and GL are important predictors of health
outcomes, alongside fibre and whole grains as indicators
of carbohydrate quality.2 For cancer, the positive
association between GI and diabetes-related cancers
could be predicted if GI was related to diabetes.
However, the non-significant association between GL and

112 www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

 Articles

Exposure Number of Number of RR (95% CI) p value

comparisons cohorts

Type 2 diabetes
GI 0·69 (0·64–0·75)
1 1 0·029
Fibre 0·80 (0·72–0·89)
Total cardiovascular disease
GI 0·93 (0·86–1·00)
3 2 0·47
Fibre 0·89 (0·83–0·96)
Diabetes-related cancer
GI 0·94 (0·91–0·97)
17 7 0·67
Fibre 0·95 (0·91–0·98)
Mortality
GI 0·88 (0·83–0·93)
2 1 0·0055
Fibre 0·79 (0·75–0·83)
Total
GI 0·90 (0·88–0·92)
23 7 0·22
Fibre 0·88 (0·86–0·91)

0·50 1·00 2·00

Favours lower GI and higher fibre Favours higher GI and lower fibre

Figure 4: Morbidity and mortality outcomes related to GI vs fibre exposure

Pooled estimates are expressed as RRs, where black squares represents the pooled RR of GI exposure and white circles represents the pooled RR of fibre exposure,
with corresponding 95% CIs. Estimates were the comparison between extreme quantiles. Pooled analyses were performed with the generic inverse variance method
with a fixed-effects model. The p values represent the test for group differences between GI and fibre exposure for the respective outcome (where p<0·05 is
significant). Total cardiovascular disease includes the combination of coronary heart disease and stroke. Five (22%) of 23 GI comparisons controlled for fibre exposure
and no fibre comparisons controlled for GI exposure. GI=glycaemic index. RR=relative risk.

Exposure Number of Number of RR (95% CI) p value

comparisons cohorts

Type 2 diabetes
GI 0·78 (0·74–0·82)
4 2 0·19
Whole grain 0·74 (0·71–0·78)
Total cardiovascular disease
GI 0·86 (0·80–0·92)
5 2 0·32
Whole grain 0·89 (0·85–0·94)
Diabetes-related cancer
GI 0·91 (0·85–0·97)
4 2 0·41
Whole grain 0·88 (0·84–0·92)
Mortality
GI 0·75 (0·67–0·83)
1 1 <0·0001
Whole grain 1·03 (0·93–1·14)
Total
GI 0·83 (0·80–0·85)
14 4 0·58
Whole grain 0·84 (0·82–0·86)

0·50 1·00 2·00

Favours lower GI Favours higher GI

and higher whole grain and lower whole grain
Figure 5: Morbidity and mortality outcomes related to GI vs wholegrain exposure
Pooled estimates are expressed as RRs, where black squares represents the pooled RR of GI exposure and white circles represents the pooled RR of wholegrain exposure,
with corresponding 95% CIs. Estimates were the comparison between extreme quantiles. Pooled analyses were performed with the generic inverse variance method with
a fixed-effects model. The p values represent the test for group differences between GI and fibre exposure for the respective outcome (where p<0·05 is significant). Total
cardiovascular disease includes the combination of cardiovascular disease, coronary heart disease, stroke, and mortality from cardiovascular disease. Seven (50%) of
14 GI comparisons controlled for wholegrain exposure and no wholegrain exposure comparisons controlled for GI exposure. GI=glycaemic index. RR=relative risk.

diabetes-related cancers observed in this analysis might
indicate the absence of an association or could be
explained by unmeasured or unintended confounding
not seen with the other outcomes.
Data from the large cohorts, albeit relatively sparse,
indicate that the certainty of evidence for low GI and GL
is similar to that reported for fibre and whole grains for
diabetes and cardiovascular disease.
GL was developed to capture the total effect of GI on
postprandial blood glucose concentrations. However,
foods high in carbohydrates also tend to have a high GI;
therefore, associations with disease outcomes tend to be

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 113

 Articles
more similar than expected from the originally intended
definition of GL.
From a mechanistic standpoint, the significant
relationships between GI and health outcomes observed in
this meta-analysis might be explained by increased
postprandial glycaemic excursions of high GI foods, which
can cause oxidative stress and free radical generation.98
High GI meals might initially raise blood glucose and
insulin concentrations, but later trigger insulin resistance
with increases in the concentration of free fatty acids.
Ludwig and Ebbeling99 suggested that a diet high in GL
increased adiposity with associated insulin resistance.
Ceriello and colleagues100 have shown that pulses of glucose
in clamp assessments reduce flow-mediated vasodilation
and increase the urinary output of nitrotyrosine, an
oxidative product that indicates oxidative stress. These
effects were pre­vented with an infusion of the antioxidant,
vitamin C. Furthermore, Monnier and colleagues101 have
shown that, in partic­ipants with diabetes on continuous
glucose monitoring, glucose pulses, as mean ambulatory
glucose excursions, increase the urinary output of
8-iso-prostaglandin F2α, a further marker of oxidative
damage. Free radical gen­eration has been suggested to
increase both inflammation and insulin resistance to
produce oxidised cholesterol, resulting in foam cells,
plaque formation, and an increased risk of cardiovascular
disease and possibly can­ cer through inflammation,
DNA damage, and mutation.98,101–104
Early assessments by Salmerón and colleagues105 in 1997
suggested the independent and additive effects of high GL
and low cereal fibre in increasing the risk of type 2
diabetes in men and, in particular, women. Subsequent
meta-analyses have substantiated these conclusions.1,5,106
Similarly, early assessments by Liu and colleagues47 in 2000
showed that GL was associated with coronary heart disease
in women, with a high BMI indicating a marked
association between GL with coronary heart disease. These
findings have been supported by subsequent cohort
assessments, including the EPIC cohort.61 In the EPIC
study, an association between GL and coronary heart
disease was observed in 338 325 participants, especially
among those with a BMI above 25 kg/m² (HR 1·22
[95% CI 1·07–1·40]).61 These associations seem to have
international relevance, given that a study of the inter­
national PURE cohort also showed a strong association
between GI and high BMI for both mortality from
cardiovascular disease and all-cause mortality.40 In
further prospective cohort assessments published in 2022,
the Shanghai Men’s and Women’s Health Study
(59 770 men and 74 735 women) reported that diets high in
GI and GL were associated with an increased risk of
mortality from cardiovascular disease in Chinese adults.72
Interestingly, the adverse associations of high GI diets for
total mortality, cardiovascular disease, and cancer were
observed mostly in women.
The 2022 meta-analysis by Dwivedi and colleagues5
also showed similar associations, especially between GL
114
and coronary heart disease. A further comprehensive
systematic review of meta-analyses published by Miller
and colleagues107 in 2022 showed that GI was strongly
associated with diabetes and cardiovascular disease.
Importantly, a meta-analysis of randomised clinical
trials has shown that risk factors for diabetes and
cardiovascular disease were significantly improved with
a low GI diet.108 Of note, cohort assessments reported in
the literature show a small range of low GI and GL
values, possibly related to health-conscious individuals
showing interest in being involved in health-related
assessments.40 Furthermore, two randomised controlled
trials, STOP NIDDM109 and ACE,110 which investigated
acarbose, a glycosylase inhibitor that pharmacologically
reduces the rate of digestion of starch and disaccharides
and effectively creates a diet low in GI and GL, showed
a reduction in the risk of incident type 2 diabetes
without other dietary changes.
Our study has several limitations. This meta-analysis
was limited by few assessments in some analyses.
Absence of assessment or absence of positive effect
might have contributed to paucity of reported studies
for some outcomes, despite the fact that studies are
expected to document cancers, MI, stroke, and CVD
deaths as serious adverse events. This limitation could
potentially result in an inaccurate estimation of the
heterogeneity between studies, assessment of
publication bias, and uncontrolled or unmeasured
confounding factors. The robustness of the findings for
some outcomes is limited due to small differences in
RR. Such differences could result from confounders for
which full adjustments were not possible. Although
similar to GI findings for diabetes and cardiovascular
disease, GL findings differed for all-cause mortality and
diabetes-related cancer. The associations between GL
and cancer require further investigation before
conclusions can be drawn. Additionally, the direct
comparison of disease outcomes related to GI with
fibre and wholegrain exposure only included one or two
cohorts for most outcomes. Furthermore, possible
differences by sex need to be evaluated in future
studies. Differences by sex have been observed but
diabetes, one of our main outcomes, does not show
a difference by sex with our limited data. With larger
numbers, this important topic should be explored. With
more complete data, it would also be useful to control
CVD and cancer outcomes for diabetes, although this
might be seen as an over-adjustment.
Low GI diets were associated with a reduced incidence
of diabetes, cardiovascular disease, diabetes-related
cancers, and all-cause mortality in large international
cohorts. Use of the same cohorts to compare the protective
effects of low GI diets with fibre and whole grains showed
similar associations in the reduction of these diseases.
Therefore, despite the controversy, advantages for chronic
disease might exist for diets low in GI that are similar to
those associated with increased consumption of fibre and
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

whole grains. More complete assessments in large
cohorts, possibly using the harmonised federated meta
analysis approach, are essential. Long-term randomised
controlled trials with high compliance would be desirable.
Nevertheless, the current evidence from large prospective
cohort assessments and clinical trials of glycosylase
inhibitors109,110 support the importance of reducing
postprandial glycaemic excursions and, thus, the
importance of GI and GL, together with fibre and whole
grains, as markers of carbohydrate quality in dietary
recommendations.
Clinical Nutrition & Risk Factor Modification Centre Collaborators
Bashyam Balachandran, Andreea Zurbau, Xunan Wang, Fred Liang,
Wanning Yang
Collaborator affiliations
Department of Nutritional Sciences, Temerty Faculty of Medicine,
University of Toronto, Toronto, ON, Canada (B Balachandran, A Zurbau,
X Wang, F Liang, W Yang); Clinical Nutrition & Risk Factor Modification
Centre (B Balachandran, A Zurbau, X Wang, F Liang, W Yang),
St Michael’s Hospital, Toronto, ON, Canada; Toronto 3D Knowledge
Synthesis and Clinical Trials Unit, Toronto, ON, Canada (A Zurbau).
Contributors
DJAJ and WCW conceptualised this study. DJAJ, WCW, SY, FBH, AJG,
SL, AM, VM, SIB, HCG, SS, PF, AVP, MLM, LLM, NDF, EL, RS, X-OS,
MT, NS, ST, PAvdB, and KeS approved the concept. DJAJ, WCW, and
SY planned the research approach. MP, SS-P, DP, TFYS, CWCK,
and KoS accessed the raw data. MP, SS-P, DP, TFYS, and CWCK did the
literature search. DJAJ, WCW, SY, FBH, AJG, SL, AM, VM, SIB, HCG,
SS, PF, AVP, MLM, LLM, NDF, EL, RS, X-OS, MT, NS, ST, PAvdB, KeS,
and KoS verified the data for their own cohorts. TAK and JLS planned
and organised the statistical approach. DJAJ, TAK, MP, SS-P, DP, TYFS,
and KoS responded to reviewers’ comments. DJAJ, WCW, SY, FBH, AJG,
SL, AM, VM, SIB, HCG, SS, PF, AVP, MLM, LLM, NDF, EL, RS, X-OS,
MT, NS, ST, PAvdB, and KeS approved the responses to reviewers’
comments. DJAJ rewrote the manuscript. All authors had full access to
all the data in the study and had final responsibility for the decision
to submit for publication.
Declaration of interests
DJAJ reports research grants from the Soy Nutrition Institute and the
Canadian Institutes of Health Research (CIHR); in-kind supplies for
trials as a research support from the Almond Board of California,
Walnut Council of California, the Peanut Institute, Barilla, Unilever,
Unico, Primo, Loblaw Companies, Quaker (Pepsico), Pristine Gourmet,
Bunge Limited, Kellogg Canada, and WhiteWave Foods; payment or
honoraria for lectures or presentations from Nutritional Fundamentals
for Health–Nutramedica, Saint Barnabas Medical Center Rutgers
New Jersey Medical School, The University of Chicago, 2020 China
Glycemic Index International Conference, Atlantic Pain Conference,
Academy of Life Long Learning; and travel support from NUTS 2022
and the 40th International Symposium on Diabetes and Nutrition.
DJAJ is a co-chair of the International Carbohydrate Quality Consortium
(ICQC) and has been invited by the International Diabetes Federation to
join the committee on diabetes treatment and to take the lead in writing
the dietary guidelines for the treatment of diabetes. His wife,
Alexandra L Jenkins, is a director and partner of INQUIS Clinical
Research for the Food Industry; his two daughters, Wendy Jenkins and
Amy Jenkins, have published a vegetarian book that promotes the use of
the foods described here (Jenkins WM, Jenkins AE, Jenkins AL,
Brydson C. The portfolio diet for cardiovascular disease risk reduction.
London: Elsevier, 2019); and his sister, Caroline Brydson, received
funding through a grant from the St Michael’s Hospital Foundation to
develop a cookbook for one of his studies. AJG reports travel support,
honoraria, or both from Vinasoy and the Academy of Nutrition and
Dietetics; and a CIHR fellowship. SL reports consulting payments and
honoraria (or promises of the same) for scientific presentations or
reviews at numerous venues, including (but not limited to) the National
Institute of Health, Fred Hutchinson Cancer Center, University of
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Articles
Missouri, Harvard University, University of Buffalo, Universidade
Federal de São Paulo, and Guangdong Provincial Hospital and Academy
of Medical Science. SL also reports honoraria from Twin Digital Health;
compensation for serving on the data safety and monitoring board for
several trials, including the SELECT trial sponsored by Novo Nordisk;
royalties from UpToDate; and an honorarium from the American
Society for Nutrition for his duties as Associate Editor. VM reports
grants from the CIHR. HCG holds the McMaster–Sanofi Population
Health Institute Chair in Diabetes Research and Care. HCG reports
research grants from Sanofi, Eli Lilly, AstraZeneca, Novo Nordisk,
Merck, Abbott, Hanmi, and Boehringer Ingelheim; honoraria for
speaking from Sanofi, Eli Lilly, AstraZeneca, Novo Nordisk, Zuellig,
DKSH Pharma, and Jiangsu Hanson; and consulting fees from Abbott,
Kowa Research Institute, Carbon Brand, and Biolinq. CWCK reports
grants from the Advanced Food Materials Network, Agriculture and
Agri-Foods Canada, Almond Board of California, Barilla, CIHR, Canola
Council of Canada, International Nut and Dried Fruit Council,
International Tree Nut Council Research and Education Foundation,
Loblaw Brands, the Peanut Institute, Pulse Canada, and Unilever;
in-kind research support from the Almond Board of California, Barilla,
California Walnut Commission, Kellogg Canada, Loblaw Companies,
Nutrartis, Quaker (PepsiCo), the Peanut Institute, Primo, Unico,
Unilever, and WhiteWave Foods–Danone; travel support from the
Barilla, California Walnut Commission, Canola Council of Canada,
General Mills, International Nut and Dried Fruit Council, International
Pasta Organization, Loblaw Brands, Nutrition Foundation of Italy,
Oldways Preservation Trust, Paramount Farms, the Peanut Institute,
Pulse Canada, Sun-Maid, Tate & Lyle, Unilever, and White Wave
Foods–Danone; and consulting fees from Lantmannen. CWCK has
served on the scientific advisory board for the McCormick Science
Institute and Oldways Preservation Trust; and is a founding member of
the ICQC, Executive Board Member of the Diabetes and Nutrition Study
Group, and Director of Glycemic Consulting and the Toronto 3D
Knowledge Synthesis and Clinical Trials foundation. JLS reports
research grants from the CIHR, Canada Foundation for Innovation,
Ontario Research Fund, American Society for Nutrition, International
Nut and Dried Fruit Council Foundation, National Honey Board,
Institute for the Advancement of Food and Nutrition Sciences, Pulse
Canada, The United Soybean Board, The Tate and Lyle Nutritional
Research Fund (University of Toronto), The Glycemic Control and
Cardiovascular Disease in Type 2 Diabetes Fund established by
the Alberta Pulse Growers (University of Toronto), Nutrition Trialists
Network Fund established by an inaugural donation from the Calorie
Control Council (University of Toronto), Diabetes Canada, and Quaker
Oats Center of Excellence; in-kind research support from the Almond
Board of California, California Walnut Commission, Peanut Institute,
Barilla, Unilever–Upfield, Unico–Primo, Loblaw Companies, Quaker,
Kellogg Canada, WhiteWave Foods–Danone, Nutrartis, and Dairy
Farmers of Canada; and speaker fees, honoraria, or both from Dairy
Farmers of Canada, FoodMinds, International Sweeteners Association,
Nestlé, Abbott, General Mills, Nutrition Communications, International
Food Information Council, Calorie Control Council, International
Glutamate Technical Committee, Arab Beverages Assocation, and
Phynova. JLS reports ad hoc consulting arrangements with Perkins Coie,
Tate & Lyle, Danone, Inquis Clinical Research, and Brightseed; has
served as a scientific board member of the European Fruit Juice
Association and the Soy Nutrition Institute; is on the Clinical Practice
Guidelines Expert Committees of Diabetes Canada, European
Association for the Study of Diabetes, Canadian Cardiovascular Society,
and Obesity Canada–Canadian Association of Bariatric Physicians and
Surgeons; is a member of the ICQC, Institute for the Advancement of
Food and Nutrition Sciences, Canadian Nutrition Society; is executive
board member of the Diabetes and Nutrition Study Group of the
European Association for the Study of Diabetes; and is director of the
Toronto 3D Knowledge Synthesis and Clinical Trials foundation. His
wife is an employee of AB InBev. All other authors declare no competing
interests. Where authors are identified as personnel of the International
Agency for Research on Cancer or WHO, the authors alone are
responsible for the views expressed in this article and they do not
necessarily represent the decisions, policy or views of the International
Agency for Research on Cancer or WHO.
115
 Articles
Data sharing
The authors will share their database with those who request it, either
for verification or for collaborative purposes, via email request to the
corresponding author.
Acknowledgments
We acknowledge the assistance of our students and volunteers,
including Kaviya Devaraja, Ayan Agarwal, Sahba Eskandari, Jiakun Lin,
Jasmine Shirvani, Yilin Bi, Gursharan Singh Marwaha, Ayush Patel,
Rasbir Marwaha, Nima Behravan, Jad Fadlallah, Nishant Patel,
Khosrow Khodabandehlou, Aleksandra Bourdine, Laiba Adnan,
Minji Jinny Kim, Maria Giuliana Giugovaz, and Raphaël Blanc.
References
1 Reynolds A, Mann J, Cummings J, Winter N, Mete E, Te Morenga L.
Carbohydrate quality and human health: a series of systematic
reviews and meta-analyses. Lancet 2019; 393: 434–45.
2 WHO. Carbohydrate intake for adults and children: WHO guideline.
Geneva: World Health Organization, 2023.
3 Jenkins DJ, Wolever TM, Taylor RH, et al. Glycemic index of foods:
a physiological basis for carbohydrate exchange. Am J Clin Nutr
1981; 34: 362–66.
4 Mirrahimi A, Chiavaroli L, Srichaikul K, et al. The role of glycemic
index and glycemic load in cardiovascular disease and its risk
factors: a review of the recent literature. Curr Atheroscler Rep 2014;
16: 381.
5 Dwivedi AK, Dubey P, Reddy SY, Clegg DJ. Associations of glycemic
index and glycemic load with cardiovascular disease: updated evidence
from meta-analysis and cohort Studies. Curr Cardiol Rep 2022;
24: 141–61.
6 Wells GA, Shea B, O’Connell D, et al. The Newcastle–Ottawa Scale
(NOS) for assessing the quality of nonrandomised studies in
meta-analyses. Ottawa: Ottawa Hospital Research Institute, 2021.
7 Higgins JPT, Thomas J, Chandler J, et al (eds). Cochrane handbook
for systematic reviews of interventions, 2nd edn. Chichester:
John Wiley & Sons, 2019.
8 Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of
observational studies in epidemiology: a proposal for reporting.
Meta-analysis Of Observational Studies in Epidemiology (MOOSE)
group. JAMA 2000; 283: 2008–12.
9 Vigneri P, Frasca F, Sciacca L, Pandini G, Vigneri R. Diabetes and
cancer. Endocr Relat Cancer 2009; 16: 1103–23.
10 Giovannucci E, Harlan DM, Archer MC, et al. Diabetes and cancer:
a consensus report. Diabetes Care 2010; 33: 1674–85.
11 Atkinson FS, Foster-Powell K, Brand-Miller JC. International tables
of glycemic index and glycemic load values: 2008. Diabetes Care
2008; 31: 2281–83.
12 Rice K, Higgins JP, Lumley T. A re-evaluation of fixed effect (s)
meta-analysis. J R Stat Soc Ser A Stat Soc 2018; 181: 205–27.
13 DerSimonian R, Laird N. Meta-analysis in clinical trials.
Control Clin Trials 1986; 7: 177–88.
14 Sidik K, Jonkman JN. A simple confidence interval for meta-
analysis. Stat Med 2002; 21: 3153–59.
15 Guyatt GH, Oxman AD, Kunz R, et al. GRADE guidelines: 7. Rating
the quality of evidence--inconsistency. J Clin Epidemiol 2011;
64: 1294–302.
16 VanderWeele TJ, Ding P. Sensitivity analysis in observational
research: introducing the E-value. Ann Intern Med 2017; 167: 268–74.
17 Potischman N, Weed DL. Causal criteria in nutritional
epidemiology. Am J Clin Nutr 1999; 69: 1309S–14S.
18 Begg CB, Mazumdar M. Operating characteristics of a rank
correlation test for publication bias. Biometrics 1994; 50: 1088–101.
19 Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-
analysis detected by a simple, graphical test. BMJ 1997; 315: 629–34.
20 Duval S, Tweedie R. Trim and fill: a simple funnel-plot-based
method of testing and adjusting for publication bias in
meta-analysis. Biometrics 2000; 56: 455–63.
21 Crippa A, Discacciati A, Bottai M, Spiegelman D, Orsini N.
One-stage dose-response meta-analysis for aggregated data.
Stat Methods Med Res 2019; 28: 1579–96.
22 Harrell FE. Regression modeling strategies: with applications to
linear models, logistic regression, and survival analysis. New York
(NY): Springer-Verlag, 2001.
116
23 Schwingshackl L, Knüppel S, Schwedhelm C, et al. Perspective:
NutriGrade: a scoring system to assess and judge the meta-evidence
of randomized controlled trials and cohort studies in nutrition
research. Adv Nutr 2016; 7: 994–1004.
24 Balshem H, Helfand M, Schünemann HJ, et al. GRADE
guidelines: 3. Rating the quality of evidence. J Clin Epidemiol 2011;
64: 401–06.
25 Schünemann H, Brożek J, Guyatt G, Oxman A. GRADE Handbook
for grading the quality of evidence and the strength of
recommendations using the GRADE approach. https://gdt.
gradepro.org/app/handbook/handbook.html
(accessed Feb 23, 2023).
26 Tobias DK, Wittenbecher C, Hu FB. Grading nutrition evidence:
where to go from here? Am J Clin Nutr 2021; 113: 1385–87.
27 Abe SK, Inoue M, Sawada N, et al. Glycemic index and glycemic
load and risk of colorectal cancer: a population-based cohort study
(JPHC Study). Cancer Causes Control 2016; 27: 583–93.
28 Bhupathiraju SN, Tobias DK, Malik VS, et al. Glycemic index,
glycemic load, and risk of type 2 diabetes: results from 3 large
US cohorts and an updated meta-analysis. Am J Clin Nutr 2014;
100: 218–32.
29 Coleman HG, Kitahara CM, Murray LJ, et al. Dietary carbohydrate
intake, glycemic index, and glycemic load and endometrial cancer
risk: a prospective cohort study. Am J Epidemiol 2014; 179: 75–84.
30 Cui X, Rosner B, Willett WC, Hankinson SE. Dietary fat, fiber, and
carbohydrate intake in relation to risk of endometrial cancer.
Cancer Epidemiol Biomarkers Prev 2011; 20: 978–89.
31 Cust AE, Slimani N, Kaaks R, et al. Dietary carbohydrates, glycemic
index, glycemic load, and endometrial cancer risk within the
European Prospective Investigation into Cancer and Nutrition
cohort. Am J Epidemiol 2007; 166: 912–23.
32 Farvid MS, Eliassen AH, Cho E, Chen WY, Willett WC. Adolescent
and early adulthood dietary carbohydrate quantity and quality in
relation to breast cancer risk. Cancer Epidemiol Biomarkers Prev
2015; 24: 1111–20.
33 Fedirko V, Lukanova A, Bamia C, et al. Glycemic index, glycemic
load, dietary carbohydrate, and dietary fiber intake and risk of liver
and biliary tract cancers in Western Europeans. Ann Oncol 2013;
24: 543–53.
34 George SM, Mayne ST, Leitzmann MF, et al. Dietary glycemic
index, glycemic load, and risk of cancer: a prospective cohort study.
Am J Epidemiol 2009; 169: 462–72.
35 Halton TL, Willett WC, Liu S, et al. Low-carbohydrate-diet score and
the risk of coronary heart disease in women. N Engl J Med 2006;
355: 1991–2002.
36 Hartman TJ, McCullough ML, Hodge JM, Gaudet MM, Wang Y,
Gapstur SM. Dietary energy density, glycemic load, glycemic index,
and risk for endometrial cancer in the CPS-II nutrition cohort.
Cancer Epidemiol Biomarkers Prev 2018; 27: 113–15.
37 Holmes MD, Liu S, Hankinson SE, Colditz GA, Hunter DJ,
Willett WC. Dietary carbohydrates, fiber, and breast cancer risk.
Am J Epidemiol 2004; 159: 732–39.
38 Hopping BN, Erber E, Grandinetti A, Verheus M, Kolonel LN,
Maskarinec G. Dietary fiber, magnesium, and glycemic load alter
risk of type 2 diabetes in a multiethnic cohort in Hawaii. J Nutr
2010; 140: 68–74.
39 Huang HL, Abe SK, Sawada N, et al. Dietary glycemic index,
glycemic load and mortality: Japan Public Health Center-based
prospective study. Eur J Nutr 2021; 60: 4607–20.
40 Jenkins DJA, Dehghan M, Mente A, et al. Glycemic index, glycemic
load, and cardiovascular disease and mortality. N Engl J Med 2021;
384: 1312–22.
41 Jiao L, Flood A, Subar AF, Hollenbeck AR, Schatzkin A,
Stolzenberg-Solomon R. Glycemic index, carbohydrates, glycemic
load, and the risk of pancreatic cancer in a prospective cohort study.
Cancer Epidemiol Biomarkers Prev 2009; 18: 1144–51.
42 Jin Q, Hart PA, Shi N, et al. Dietary patterns of insulinemia,
inflammation and glycemia, and pancreatic cancer risk: findings
from the Women’s Health Initiative.
Cancer Epidemiol Biomarkers Prev 2021; 30: 1229–40.
43 Jin Q, Shi N, Aroke D, et al. Insulinemic and inflammatory dietary
patterns show enhanced predictive potential for type 2 diabetes risk
in postmenopausal women. Diabetes Care 2021; 44: 707–14.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

44 Jonas CR, McCullough ML, Teras LR, Walker-Thurmond KA,
Thun MJ, Calle EE. Dietary glycemic index, glycemic load, and risk
of incident breast cancer in postmenopausal women.
Cancer Epidemiol Biomarkers Prev 2003; 12: 573–77.
45 Kabat GC, Shikany JM, Beresford SA, et al. Dietary carbohydrate,
glycemic index, and glycemic load in relation to colorectal cancer
risk in the Women’s Health Initiative. Cancer Causes Control 2008;
19: 1291–98.
46 Li HL, Yang G, Shu XO, et al. Dietary glycemic load and risk of
colorectal cancer in Chinese women. Am J Clin Nutr 2011; 93: 101–07.
47 Liu S, Willett WC, Stampfer MJ, et al. A prospective study of dietary
glycemic load, carbohydrate intake, and risk of coronary heart
disease in US women. Am J Clin Nutr 2000; 71: 1455–61.
48 Luo X, Sui J, Yang W, et al. Type 2 diabetes prevention diet and
hepatocellular carcinoma risk in US men and women.
Am J Gastroenterol 2019; 114: 1870–77.
49 Meinhold CL, Dodd KW, Jiao L, et al. Available carbohydrates,
glycemic load, and pancreatic cancer: is there a link? Am J Epidemiol
2010; 171: 1174–82.
50 Michaud DS, Liu S, Giovannucci E, Willett WC, Colditz GA,
Fuchs CS. Dietary sugar, glycemic load, and pancreatic cancer risk
in a prospective study. J Natl Cancer Inst 2002; 94: 1293–300.
51 Michaud DS, Fuchs CS, Liu S, Willett WC, Colditz GA,
Giovannucci E. Dietary glycemic load, carbohydrate, sugar, and
colorectal cancer risk in men and women.
Cancer Epidemiol Biomarkers Prev 2005; 14: 138–47.
52 Mongiovi JM, Freudenheim JL, Moysich KB, McCann SE. Glycemic
index, glycemic load, and risk of ovarian cancer in the Prostate,
Lung, Colorectal and Ovarian (PLCO) cohort. J Nutr 2021;
151: 1597–608.
53 Nimptsch K, Kenfield S, Jensen MK, et al. Dietary glycemic index,
glycemic load, insulin index, fiber and whole-grain intake in relation
to risk of prostate cancer. Cancer Causes Control 2011; 22: 51–61.
54 Nöthlings U, Murphy SP, Wilkens LR, Henderson BE, Kolonel LN.
Dietary glycemic load, added sugars, and carbohydrates as risk
factors for pancreatic cancer: the Multiethnic Cohort Study.
Am J Clin Nutr 2007; 86: 1495–501.
55 Oba S, Nanri A, Kurotani K, et al. Dietary glycemic index, glycemic
load and incidence of type 2 diabetes in Japanese men and women:
the Japan Public Health Center-based Prospective Study. Nutr J
2013; 12: 165.
56 Patel AV, McCullough ML, Pavluck AL, Jacobs EJ, Thun MJ,
Calle EE. Glycemic load, glycemic index, and carbohydrate intake in
relation to pancreatic cancer risk in a large US cohort.
Cancer Causes Control 2007; 18: 287–94.
57 Qi H, Xia D, Xu X. Dietary glycemic index, glycemic load, and renal
cancer risk: findings from prostate, lung, colorectal, and ovarian
cancer trial. Front Nutr 2023; 10: 1073373.
58 Romieu I, Ferrari P, Rinaldi S, et al. Dietary glycemic index and
glycemic load and breast cancer risk in the European Prospective
Investigation into Cancer and Nutrition (EPIC). Am J Clin Nutr
2012; 96: 345–55.
59 Shikany JM, Flood AP, Kitahara CM, et al. Dietary carbohydrate,
glycemic index, glycemic load, and risk of prostate cancer in the
Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial
(PLCO) cohort. Cancer Causes Control 2011; 22: 995–1002.
60 Shikany JM, Redden DT, Neuhouser ML, et al. Dietary glycemic
load, glycemic index, and carbohydrate and risk of breast cancer in
the Women’s Health Initiative. Nutr Cancer 2011; 63: 899–907.
61 Sieri S, Agnoli C, Grioni S, et al. Glycemic index, glycemic load,
and risk of coronary heart disease: a pan-European cohort study.
Am J Clin Nutr 2020; 112: 631–43.
62 Sun JW, Zheng W, Li HL, et al. Dietary glycemic load, glycemic
index, and carbohydrate intake on the risk of lung cancer among
men and women in Shanghai. Nutr Cancer 2018; 70: 671–77.
63 Tao J, Jatoi A, Crawford J, et al. Role of dietary carbohydrates on risk
of lung cancer. Lung Cancer 2021; 155: 87–93.
64 Villegas R, Liu S, Gao YT, et al. Prospective study of dietary
carbohydrates, glycemic index, glycemic load, and incidence of
type 2 diabetes mellitus in middle-aged Chinese women.
Arch Intern Med 2007; 167: 2310–16.
65 Vogtmann E, Li HL, Shu XO, et al. Dietary glycemic load, glycemic
index, and carbohydrates on the risk of primary liver cancer among
Chinese women and men. Ann Oncol 2013; 24: 238–44.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Articles
66 Watanabe Y, Katagiri R, Goto A, et al. Dietary glycemic index,
glycemic load, and endometrial cancer risk: The Japan Public
Health Center-based Prospective Study. Cancer Sci 2021;
112: 3682–90.
67 Wen W, Shu XO, Li H, et al. Dietary carbohydrates, fiber, and breast
cancer risk in Chinese women. Am J Clin Nutr 2009; 89: 283–89.
68 Xu X, Shivappa N. Dietary glycemic index, glycemic load and risk of
bladder cancer: a prospective study. Eur J Nutr 2021; 60: 1041–48.
69 Yu D, Zhang X, Shu XO, et al. Dietary glycemic index, glycemic
load, and refined carbohydrates are associated with risk of stroke:
a prospective cohort study in urban Chinese women. Am J Clin Nutr
2016; 104: 1345–51.
70 Yu D, Shu XO, Li H, et al. Dietary carbohydrates, refined grains,
glycemic load, and risk of coronary heart disease in Chinese adults.
Am J Epidemiol 2013; 178: 1542–49.
71 Zamora-Ros R, Rinaldi S, Tsilidis KK, et al. Energy and
macronutrient intake and risk of differentiated thyroid carcinoma
in the European Prospective Investigation into Cancer and
Nutrition study. Int J Cancer 2016; 138: 65–73.
72 Zhao LG, Li HL, Liu DK, et al. Dietary glycemic index, glycemic
load, and cause-specific mortality: two population-based prospective
cohort studies. Eur J Clin Nutr 2022; 76: 1142–49.
73 Howarth NC, Murphy SP, Wilkens LR, Henderson BE, Kolonel LN.
The association of glycemic load and carbohydrate intake with
colorectal cancer risk in the Multiethnic Cohort Study.
Am J Clin Nutr 2008; 88: 1074–82.
74 Oh K, Hu FB, Cho E, et al. Carbohydrate intake, glycemic index,
glycemic load, and dietary fiber in relation to risk of stroke in
women. Am J Epidemiol 2005; 161: 161–69.
75 AlEssa HB, Cohen R, Malik VS, et al. Carbohydrate quality and
quantity and risk of coronary heart disease among US women and
men. Am J Clin Nutr 2018; 107: 257–67.
76 AlEssa HB, Bhupathiraju SN, Malik VS, et al. Carbohydrate quality
and quantity and risk of type 2 diabetes in US women.
Am J Clin Nutr 2015; 102: 1543–53.
77 Cho E, Spiegelman D, Hunter DJ, Chen WY, Colditz GA,
Willett WC. Premenopausal dietary carbohydrate, glycemic index,
glycemic load, and fiber in relation to risk of breast cancer.
Cancer Epidemiol Biomarkers Prev 2003; 12: 1153–58.
78 Daniel CR, Park Y, Chow WH, Graubard BI, Hollenbeck AR,
Sinha R. Intake of fiber and fiber-rich plant foods is associated with
a lower risk of renal cell carcinoma in a large US cohort.
Am J Clin Nutr 2013; 97: 1036–43.
79 Farvid MS, Cho E, Eliassen AH, Chen WY, Willett WC. Lifetime
grain consumption and breast cancer risk. Breast Cancer Res Treat
2016; 159: 335–45.
80 Ferrari P, Rinaldi S, Jenab M, et al. Dietary fiber intake and risk of
hormonal receptor-defined breast cancer in the European
Prospective Investigation into Cancer and Nutrition study.
Am J Clin Nutr 2013; 97: 344–53.
81 He X, Wu K, Zhang X, et al. Dietary intake of fiber, whole grains
and risk of colorectal cancer: an updated analysis according to food
sources, tumor location and molecular subtypes in two large US
cohorts. Int J Cancer 2019; 145: 3040–51.
82 Hu Y, Ding M, Sampson L, et al. Intake of whole grain foods and
risk of type 2 diabetes: results from three prospective cohort
studies. BMJ 2020; 370: m2206.
83 Hu Y, Willett WC, Manson JAE, Rosner B, Hu FB, Sun Q. Intake of
whole grain foods and risk of coronary heart disease in US men and
women. BMC Med 2022; 20: 192.
84 Hullings AG, Sinha R, Liao LM, Freedman ND, Graubard BI,
Loftfield E. Whole grain and dietary fiber intake and risk of
colorectal cancer in the NIH-AARP Diet and Health Study cohort.
Am J Clin Nutr 2020; 112: 603–12.
85 Katagiri R, Goto A, Sawada N, et al. Dietary fiber intake and total
and cause-specific mortality: the Japan Public Health Center-based
prospective study. Am J Clin Nutr 2020; 111: 1027–35.
86 Lam TK, Cross AJ, Freedman N, et al. Dietary fiber and grain
consumption in relation to head and neck cancer in the
NIH-AARP Diet and Health Study. Cancer Causes Control 2011;
22: 1405–14.
87 Lan T, Park Y, Colditz GA, et al. Adolescent plant product intake in
relation to later prostate cancer risk and mortality in the NIH-AARP
diet and health study. J Nutr 2021; 151: 3223–31.
117
 Articles
88 Li J, Glenn AJ, Yang Q, et al. Dietary protein sources, mediating
biomarkers, and incidence of type 2 diabetes: findings from the
Women’s Health Initiative and the UK Biobank. Diabetes Care 2022;
45: 1742–53.
89 Luo J, Xu X. Dietary fiber intake and the risk of bladder cancer in
the Prostate, Lung, Colorectal and Ovarian (PLCO) cohort.
Carcinogenesis 2020; 41: 478–82.
90 Navarro SL, Neuhouser ML, Cheng TD, et al. The interaction
between dietary fiber and fat and risk of colorectal cancer in the
Women’s Health Initiative. Nutrients 2016; 8: 779.
91 Otani T, Iwasaki M, Ishihara J, Sasazuki S, Inoue M, Tsugane S.
Dietary fiber intake and subsequent risk of colorectal cancer: the
Japan Public Health Center-based prospective study. Int J Cancer
2006; 119: 1475–80.
92 Park SY, Wilkens LR, Kolonel LN, Henderson BE, Le Marchand L.
Inverse associations of dietary fiber and menopausal hormone
therapy with colorectal cancer risk in the Multiethnic Cohort Study.
Int J Cancer 2016; 139: 1241–50.
93 Perez-Cornago A, Crowe FL, Appleby PN, et al. Plant foods, dietary
fibre and risk of ischaemic heart disease in the European
Prospective Investigation into Cancer and Nutrition (EPIC) cohort.
Int J Epidemiol 2021; 50: 212–22.
94 Shin A, Li H, Shu XO, Yang G, Gao YT, Zheng W. Dietary intake of
calcium, fiber and other micronutrients in relation to colorectal
cancer risk: results from the Shanghai Women’s Health Study.
Int J Cancer 2006; 119: 2938–42.
95 Swaminathan S, Dehghan M, Raj JM, et al. Associations of cereal
grains intake with cardiovascular disease and mortality across
21 countries in Prospective Urban and Rural Epidemiology study:
prospective cohort study. BMJ 2021; 372: m4948.
96 Willett WC, Hunter DJ, Stampfer MJ, et al. Dietary fat and fiber in
relation to risk of breast cancer. An 8-year follow-up. JAMA 1992;
268: 2037–44.
97 Park Y, Brinton LA, Subar AF, Hollenbeck A, Schatzkin A.
Dietary fiber intake and risk of breast cancer in postmenopausal
women: the National Institutes of Health-AARP Diet and Health
Study. Am J Clin Nutr 2009; 90: 664–71.
98 Ceriello A, Quagliaro L, Piconi L, et al. Effect of postprandial
hypertriglyceridemia and hyperglycemia on circulating adhesion
molecules and oxidative stress generation and the possible role of
simvastatin treatment. Diabetes 2004; 53: 701–10.
99 Ludwig DS, Ebbeling CB. The carbohydrate-insulin model of
obesity: beyond “calories in, calories out”. JAMA Intern Med 2018;
178: 1098–103.
118
100 Ceriello A, Quagliaro L, Catone B, et al. Role of hyperglycemia in
nitrotyrosine postprandial generation. Diabetes Care 2002;
25: 1439–43.
101 Monnier L, Mas E, Ginet C, et al. Activation of oxidative stress by
acute glucose fluctuations compared with sustained chronic
hyperglycemia in patients with type 2 diabetes. JAMA 2006;
295: 1681–87.
102 Dhalla NS, Temsah RM, Netticadan T. Role of oxidative stress in
cardiovascular diseases. J Hypertens 2000; 18: 655–73.
103 Wolever TM, Gibbs AL, Mehling C, et al. The Canadian Trial of
Carbohydrates in Diabetes (CCD), a 1-y controlled trial of low-
glycemic-index dietary carbohydrate in type 2 diabetes: no effect on
glycated hemoglobin but reduction in C-reactive protein.
Am J Clin Nutr 2008; 87: 114–25.
104 Reuter S, Gupta SC, Chaturvedi MM, Aggarwal BB. Oxidative
stress, inflammation, and cancer: how are they linked?
Free Radic Biol Med 2010; 49: 1603–16.
105 Salmerón J, Manson JE, Stampfer MJ, Colditz GA, Wing AL,
Willett WC. Dietary fiber, glycemic load, and risk of non-insulin-
dependent diabetes mellitus in women. JAMA 1997; 277: 472–77.
106 Livesey G, Taylor R, Livesey HF, et al. Dietary glycemic index and
load and the risk of type 2 diabetes: a systematic review and
updated meta-analyses of prospective cohort studies. Nutrients
2019; 11: 1280.
107 Miller V, Micha R, Choi E, Karageorgou D, Webb P, Mozaffarian D.
Evaluation of the quality of evidence of the association of foods and
nutrients with cardiovascular disease and diabetes: a systematic
review. JAMA Netw Open 2022; 5: e2146705.
108 Chiavaroli L, Lee D, Ahmed A, et al. Effect of low glycaemic index
or load dietary patterns on glycaemic control and cardiometabolic
risk factors in diabetes: systematic review and meta-analysis of
randomised controlled trials. BMJ 2021; 374: n1651.
109 Chiasson JL, Josse RG, Gomis R, Hanefeld M, Karasik A, Laakso M.
Acarbose treatment and the risk of cardiovascular disease and
hypertension in patients with impaired glucose tolerance:
the STOP-NIDDM trial. JAMA 2003; 290: 486–94.
110 Holman RR, Coleman RL, Chan JCN, et al. Effects of acarbose on
cardiovascular and diabetes outcomes in patients with coronary
heart disease and impaired glucose tolerance (ACE): a randomised,
double-blind, placebo-controlled trial. Lancet Diabetes Endocrinol
2017; 5: 877–86.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

Advancing precision medicine in type 2 diabetes
An emerging consensus in the diabetes field postulates
that the traditional subclassification of diabetes into
type 1 diabetes, type 2 diabetes, gestational diabetes,
monogenic or syndromic diabetes, or diabetes sec­
ondary to underlying pancreatic disease is outdated.
Global efforts such as the Precision Medicine in Diabetes
Initiative1 acknowledge that the current system does
not fully capture the underlying pathophysiology
that gives rise to various forms of diabetes. Its recent
systematic review emphasises that not recognising
the multiple molecular pathways that contribute to
diabetes pathogenesis hinders disease management
and therapeutic development.2
This issue is particularly vexing for type 2 diabetes.3
Type 2 diabetes is largely a diagnosis of exclusion,
such that patients whose hyperglycaemia is not
caused by autoimmunity, single-gene mutations,
or pre-existing pancreatic disease are given the
type 2 diabetes label, regardless of the mechanisms
by which their hyperglycaemia arose. Consequently,
treatment algorithms seldom consider the pathways
affected, and therapeutic choices are not tailored to
the individual’s presentation. To fill this gap, multiple
groups have endeavoured to elucidate and characterise
the clinically evident heterogeneity in type 2 diabetes,
developing various techniques that define categories
of type 2 diabetes on the basis of phenotypic or genetic
similarities, or a combination of both.
An early and widely replicated approach was introduced
by Ahlqvist and colleagues,4 who used k-means clustering
on simple clinical characteristics available in a large cohort
of Scandinavian participants with recent diabetes onset
to define five novel subgroups (severe autoimmune
diabetes, severe insulin-deficient diabetes, severe insulin-
resistant diabetes, moderate obesity-related diabetes,
and moderate age-related diabetes). Importantly, they
and others showed that these subgroups had differential
incidence of diabetic complications and response to
therapy.4,5 The IMI-DIRECT investigators similarly used
soft clustering and 32 clinical variables measured in
a deeply phenotyped cohort to generate four diabetes
archetypes (insulin-deficient diabetes, obese and insulin-
sensitive diabetes, obese and insulin-resistant diabetes,
and global severe diabetes),6 with relevance to disease
progression and intensity of treatment.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Comment
The above approaches use clinical variables,
which are subject to change and can reflect either
disease progression or response to treatment: thus
the assignment of individuals to a specific cluster
Carol and Mike Werner/Science Photo Library
is contingent on the metabolic state in which the
measurements took place, and might not be stable over
time.7 One way to overcome this limitation is to leverage
the immutability of germline genetics to generate clusters
that are anchored in genetics but informed by physiology,
which yields process-specific polygenic scores by which
genetic variants associated with type 2 diabetes are Published Online
December 21, 2023
grouped by the metabolic features they affect.8 https://doi.org/10.1016/
A second limitation of these approaches concerns S2213-8587(23)00384-4
the assignment of individuals to discrete categories, See Articles page 119
rather than using the quantitative information inherent
to the predictive variables to place each person on
a continuous multidimensional map, producing
an individualised risk score. Dennis and colleagues5
suggested the latter approach was preferable, and Nair
and colleagues9 implemented a data dimensionality
reduction tree algorithm that reduced nine phenotypic
variables (HbA1c, BMI, total cholesterol, HDL-cholesterol,
triglycerides, alanine aminotransferase, creatinine,
systolic blood pressure, and diastolic blood pressure)
into a non-linear tree structure, with branches of the
tree showing different phenotypic characteristics, inci­
dence of complications, and response to treatment.
Building on this approach, in The Lancet Diabetes
& Endocrinology, Martin Schön and colleagues10 con­
structed similar trees on the basis of the continuous
distributions of the same nine simple clinical factors in
the German Diabetes Study, comprising 927 participants
with type 2 diabetes of recent onset, and they examined
how this method stratifies pathophysiological compo­
nents of dysglycaemia as well as diabetic complications.
This cohort is very well phenotyped, including an
appropriate run-in period without glucose-lowering
medication and with a controlled diet and exercise level.
Additionally, there is availability of both intravenous
glucose tolerance tests and mixed-meal tolerance tests
to evaluate β-cell function, and the cohort includes
a longitudinal component of decent duration (5 years).
The authors found that complex dynamic measures of
insulin secretion and sensitivity were distributed across
the branches of the tree, with lower insulin sensitivity
87
 Comment
correlating with hepatic steatosis, pro-inflammatory
biomarkers, and cardiovascular disease risk, whereas
lower insulin secretion was correlated with higher need
for insulin therapy and a higher risk of peripheral and
autonomic neuropathy. In a different cohort, mortality
was higher in the branch with highest insulin resistance.
The authors documented the stability of insulin sensitivity
but found that insulin secretion had a proclivity to decline
over time. About 20% of participants changed their
position across the tree over the 5 years of follow-up. As
might be expected, the previously defined phenotypic
cluster of severe insulin-resistant diabetes mapped to
the low insulin sensitivity branch of the tree, whereas
low numbers precluded accurate mapping of the severe
insulin-deficient diabetes cluster.
This report validates tree-based dimensionality reduc­
tion in a non-British European context, thus expanding
its generalisability. It thereby highlights the value of
prediction models based on continuous distributions
of easily accessible routine clinical variables, rather than
categorical assignments to discrete subtypes. Its novelty
lies in illustrating how sophisticated measures of glucose
homeostasis vary across the tree map, and its clinical
relevance resides on its ability to associate specific diabetic
complications to different topographical locations in
the tree. It builds on previous knowledge by assigning
previously described clusters to specific regions of the
tree, and showing that incident complications appear
independently of HbA1c trajectories.
The authors acknowledge some key limitations.
First, the study focused on participants with recent-
onset diabetes and excluded those with decompen­sated
diabetes, and thus does not capture the full spectrum
of diabetes, leading to low numbers in the extreme
branches of the tree. Second, it only included participants
of European descent. Third, it lacks genetic data, hindering
placement of process-specific polygenic clusters on this
phenotypic tree. Nevertheless, this study represents
Enhancing global access to diabetes medicines: policy
lessons from the HIV response
Published Online One of the biggest global health success stories is
December 18, 2023
https://doi.org/10.1016/
the response to the HIV pandemic over the past
S2213-8587(23)00359-5 three decades. Scientists, clinicians, activists, policy
88
a major step forward in applying a more precise metabolic
definition to people with type 2 diabetes, and in showing
why it might matter in practice. In this manner, one can
envision a more individualised approach to secondary
prevention and surveillance based on where patients map
onto a robust and reproducible physiological tree.
I am supported by US National Institutes of Health (NIH) grants NHLBI K24
HL157960, NIDDK U54 DK118612, NIGMS R01 GM117163, NIDDK UM1
DK126185, NIDDK R01 DK123019, and NIDDK R01 DK132299. I have received
investigator-initiated grant support from Novo Nordisk, and speaker’s honoraria
from Novo Nordisk and Merck for research presentations over which I had full
control of content. I have also received a consulting honorarium from
AstraZeneca.
Jose C Florez
jcflorez@mgh.harvard.edu
Department of Medicine and Center for Genomic Medicine, Massachusetts
General Hospital, Programs in Metabolism and Medical & Population Genetics,
Broad Institute Department of Medicine, Harvard Medical School, Boston,
MA 02114, USA
1 Chung WK, Erion K, Florez JC, et al. Precision medicine in diabetes:
a consensus report from the American Diabetes Association (ADA) and the
European Association for the Study of Diabetes (EASD). Diabetes Care 2020;
43: 1617–35.
2 Tobias DK, Merino J, Ahmad A, et al. Second international consensus report
on gaps and opportunities for the clinical translation of precision diabetes
medicine. Nat Med 2023; 29: 2438–57.
3 Misra S, Wagner R, Ozkan B, et al. Precision subclassification of type 2
diabetes: a systematic review. Commun Med 2023; 3: 138.
4 Ahlqvist E, Storm P, Karajamaki A, et al. Novel subgroups of adult-onset
diabetes and their association with outcomes: a data-driven cluster analysis
of six variables. Lancet Diabetes Endocrinol 2018; 6: 361–99.
5 Dennis JM, Shields BM, Henley WE, Jones AG, Hattersley AT. Disease
progression and treatment response in data-driven subgroups of type 2
diabetes compared with models based on simple clinical features: an
analysis using clinical trial data. Lancet Diabetes Endocrinol 2019; 7: 442–51.
6 Wesolowska-Andersen A, Brorsson CA, Bizzotto R, et al. Four groups of
type 2 diabetes contribute to the etiological and clinical heterogeneity in
newly diagnosed individuals: an IMI DIRECT study. Cell Rep Med 2022;
3: 100477.
7 Zaharia OP, Strassburger K, Strom A, et al. Risk of diabetes-associated
diseases in subgroups of patients with recent-onset diabetes:
a 5-year follow-up study. Lancet Diabetes Endocrinol 2019; 7: 684–94.
8 Udler MS, Kim J, von Grotthuss M, et al. Type 2 diabetes genetic loci
informed by multi-trait associations point to disease mechanisms and
subtypes: a soft clustering analysis. PLoS Med 2018; 15: e1002654.
9 Nair ATN, Wesolowska-Andersen A, Brorsson C, et al. Heterogeneity in
phenotype, disease progression and drug response in type 2 diabetes.
Nat Med 2022; 28: 982–88.
10 Schön M, Prystupa K, Mori T, et al. Analysis of type 2 diabetes
heterogeneity with a tree-like representation: insights from the
prospective German Diabetes Study and the LURIC cohort.
Lancet Diabetes Endocrinol 2023; published online Dec 21. https://doi.org/
S2213-8587(23)00329-7.
makers, and pharmaceutical companies turned a deadly
disease into a treatable disease by ensuring near-
universal access to antiretroviral therapies (ARTs). The
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
 Comment
correlating with hepatic steatosis, pro-inflammatory
biomarkers, and cardiovascular disease risk, whereas
lower insulin secretion was correlated with higher need
for insulin therapy and a higher risk of peripheral and
autonomic neuropathy. In a different cohort, mortality
was higher in the branch with highest insulin resistance.
The authors documented the stability of insulin sensitivity
but found that insulin secretion had a proclivity to decline
over time. About 20% of participants changed their
position across the tree over the 5 years of follow-up. As
might be expected, the previously defined phenotypic
cluster of severe insulin-resistant diabetes mapped to
the low insulin sensitivity branch of the tree, whereas
low numbers precluded accurate mapping of the severe
insulin-deficient diabetes cluster.
This report validates tree-based dimensionality reduc­
tion in a non-British European context, thus expanding
its generalisability. It thereby highlights the value of
prediction models based on continuous distributions
of easily accessible routine clinical variables, rather than
categorical assignments to discrete subtypes. Its novelty
lies in illustrating how sophisticated measures of glucose
homeostasis vary across the tree map, and its clinical
relevance resides on its ability to associate specific diabetic
complications to different topographical locations in
the tree. It builds on previous knowledge by assigning
previously described clusters to specific regions of the
tree, and showing that incident complications appear
independently of HbA1c trajectories.
The authors acknowledge some key limitations.
First, the study focused on participants with recent-
onset diabetes and excluded those with decompen­sated
diabetes, and thus does not capture the full spectrum
of diabetes, leading to low numbers in the extreme
branches of the tree. Second, it only included participants
of European descent. Third, it lacks genetic data, hindering
placement of process-specific polygenic clusters on this
phenotypic tree. Nevertheless, this study represents
Enhancing global access to diabetes medicines: policy
lessons from the HIV response
Published Online One of the biggest global health success stories is
December 18, 2023
https://doi.org/10.1016/
the response to the HIV pandemic over the past
S2213-8587(23)00359-5 three decades. Scientists, clinicians, activists, policy
88
a major step forward in applying a more precise metabolic
definition to people with type 2 diabetes, and in showing
why it might matter in practice. In this manner, one can
envision a more individualised approach to secondary
prevention and surveillance based on where patients map
onto a robust and reproducible physiological tree.
I am supported by US National Institutes of Health (NIH) grants NHLBI K24
HL157960, NIDDK U54 DK118612, NIGMS R01 GM117163, NIDDK UM1
DK126185, NIDDK R01 DK123019, and NIDDK R01 DK132299. I have received
investigator-initiated grant support from Novo Nordisk, and speaker’s honoraria
from Novo Nordisk and Merck for research presentations over which I had full
control of content. I have also received a consulting honorarium from
AstraZeneca.
Jose C Florez
jcflorez@mgh.harvard.edu
Department of Medicine and Center for Genomic Medicine, Massachusetts
General Hospital, Programs in Metabolism and Medical & Population Genetics,
Broad Institute Department of Medicine, Harvard Medical School, Boston,
MA 02114, USA
1 Chung WK, Erion K, Florez JC, et al. Precision medicine in diabetes:
a consensus report from the American Diabetes Association (ADA) and the
European Association for the Study of Diabetes (EASD). Diabetes Care 2020;
43: 1617–35.
2 Tobias DK, Merino J, Ahmad A, et al. Second international consensus report
on gaps and opportunities for the clinical translation of precision diabetes
medicine. Nat Med 2023; 29: 2438–57.
3 Misra S, Wagner R, Ozkan B, et al. Precision subclassification of type 2
diabetes: a systematic review. Commun Med 2023; 3: 138.
4 Ahlqvist E, Storm P, Karajamaki A, et al. Novel subgroups of adult-onset
diabetes and their association with outcomes: a data-driven cluster analysis
of six variables. Lancet Diabetes Endocrinol 2018; 6: 361–99.
5 Dennis JM, Shields BM, Henley WE, Jones AG, Hattersley AT. Disease
progression and treatment response in data-driven subgroups of type 2
diabetes compared with models based on simple clinical features: an
analysis using clinical trial data. Lancet Diabetes Endocrinol 2019; 7: 442–51.
6 Wesolowska-Andersen A, Brorsson CA, Bizzotto R, et al. Four groups of
type 2 diabetes contribute to the etiological and clinical heterogeneity in
newly diagnosed individuals: an IMI DIRECT study. Cell Rep Med 2022;
3: 100477.
7 Zaharia OP, Strassburger K, Strom A, et al. Risk of diabetes-associated
diseases in subgroups of patients with recent-onset diabetes:
a 5-year follow-up study. Lancet Diabetes Endocrinol 2019; 7: 684–94.
8 Udler MS, Kim J, von Grotthuss M, et al. Type 2 diabetes genetic loci
informed by multi-trait associations point to disease mechanisms and
subtypes: a soft clustering analysis. PLoS Med 2018; 15: e1002654.
9 Nair ATN, Wesolowska-Andersen A, Brorsson C, et al. Heterogeneity in
phenotype, disease progression and drug response in type 2 diabetes.
Nat Med 2022; 28: 982–88.
10 Schön M, Prystupa K, Mori T, et al. Analysis of type 2 diabetes
heterogeneity with a tree-like representation: insights from the
prospective German Diabetes Study and the LURIC cohort.
Lancet Diabetes Endocrinol 2023; published online Dec 21. https://doi.org/
S2213-8587(23)00329-7.
makers, and pharmaceutical companies turned a deadly
disease into a treatable disease by ensuring near-
universal access to antiretroviral therapies (ARTs). The
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

introduction and dissemination of ART averted an
estimated 25 million deaths between 1996 and 2022.
Although a range of factors contributed to the success
of the HIV response, these achievements would have
been unimaginable without concerted global efforts
to lower prices for ARTs.1 Given the growing global
burden of chronic diseases, lessons from HIV should be
considered to improve accessibility and affordability of
medicines for diabetes, a deadly metabolic disorder of
contemporary importance.
An estimated 6·7 million people die due to diabetes
and its complications every year.2 Global diabetes
prevalence has tripled since 2000 and 537 million
adults are affected, 81% of whom live in low-income
and middle-income countries (LMICs).2 However, only
38% of people with diabetes in LMICs receive any form
of treatment, 23% achieve glycaemic control, and
approximately 5% get all treatments needed to prevent
cardiovascular disease.3 Although behavioural diabetes
prevention interventions are important to slow incidence
(as was the case for HIV), averting deaths and irreversible
complications from diabetes is crucial and will inevitably
require access to quality health care and a reliable supply
of affordable medicines. However, both newer diabetes
medicines (GLP-1 receptor agonists and SGLT2 inhibitors)
and off-patent formulations (particularly of insulin) are
unaffordable for the majority of people with diabetes,
especially those residing in LMICs.4
Global policies that enhanced equitable access to ARTs
can offer insight. First, pooled procurement of patented
and generic drugs nationally and regionally lowered
drug prices by leveraging economies of scale, increasing
procurement efficiency, and reducing transaction costs.5
The Global Fund for Aids, Tuberculosis, and Malaria
successfully implemented a mechanism that enabled
procurement agents to manage negotiation and tenders
on behalf of a recipient country.5 Multipartner trust funds
for non-communicable diseases (NCDs), such as the UN’s
Health4Life fund, could play a similar role as facilitators
between procurement agents and countries. Additionally,
such trust funds could be used to provide catalytic
funding to incentivise and improve country-led pooled
procurement programmes, which have worked for other
NCDs. For example, in India, a network of more than
250 cancer centres independently launched a pooled
procurement initiative that led to median savings of
82% for expensive anti-cancer medicines.6
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Comment
Second, pharmaceutical companies played a pivotal
role in increasing ART access by providing up to 90%
discounted prices in LMICs.1 Similar actions for diabetes
medicines could also be a competitive business strategy
(rather than exclusively altruistically motivated). The
demand from more than 400 million people with
Sezeryadigar/Getty Images
diabetes in LMICs creates a strong case to prioritise large
market reach over high per-person prices.2 Such price
differentiation could thus be a profitable pricing strategy
yielding improved access to life-saving medications in
LMICs. For more on deaths caused by
HIV see https://www.state.gov/
Third, licenses, patent pools, and patent waivers wp-content/uploads/2022/11/
decreased ART prices by enabling LMICs to manufacture PEPFAR-Latest-Global-Results_
December-2022.pdf
medications without paying full royalties. Although
mustering political support for patent waivers has
been difficult beyond HIV, voluntary licenses that grant
companies permission to produce or distribute patented
drugs might be a viable strategy to increase the supply
and affordability of diabetes medications in LMICs. In
September, 2023, Novo Nordisk (one of the three main For more on the Novo Nordisk
partnership with Aspen
insulin producers worldwide) contracted the South Pharmacare see https://www.
Africa-based company Aspen Pharmacare to produce novonordisk.com/news-and-
media/latest-news/new-
16 million vials of human insulin in 2024 for export partnership-to-supply-human-
to African countries through a low-cost government insulin.html
tender system that is capped at US$3 per vial. Similar
approaches are conceivable for patented diabetes
medicines via voluntary licenses.
Achieving net cost savings in LMICs would require
drastic price reductions for GLP-1 receptor agonists,
while the price of SGLT2 inhibitors would only need to be
reduced by around one-fifth.7 Substantial improvements
in medication affordability and accessibility are thus
within reach and could help curb the immense global
economic burden of diabetes.8 Notably, reducing com­
plications from diabetes in LMICs not only requires
affordable glucose-lowering medicines, but also requires
access to anti-hypertensive treatments and statins.9
Hence, global policies aiming to reduce drug prices
need to be integrated across different NCDs, including For more on the
Health4Life fund see
cardiovascular disease. https://cdn.who.int/media/
Ensuring the affordability of medicines alone will not docs/default-source/unitaf/
uniatf-mptf-flyer_071021_
solve the diabetes pandemic. Additional steps, such final.pdf
as strengthening primary health care and surveillance
systems and training health service providers, will be
necessary to achieve country-level targets for diabetes.10
However, access to existing and novel medicines is key
for a successful global public health response.1 Making
89
 Comment
RUNSTUDIO/Getty Images
For more on drugs approved in
Japan in 2010 see https://www.
mhlw.go.jp/shingi/2010/06/dl/
s0602-3a.pdf
For more on drugs approved
in Japan in 2023 see
https://www.mhlw.go.jp/
content/12404000/001167629.
pdf
90
diabetes medicines affordable globally is feasible and
discourse about the best ways to achieve this goal is
overdue.
CAB was supported by the Dutch Research Council’s Rubicon programme (project
number 452022313). MKA reported a past grant from Merck & Co to Emory
University and past consulting fees from Bayer and Eli Lilly and Company outside
the scope of the submitted work. All other authors declare no competing interests.
*Felix Teufel, Caroline A Bulstra, Justine I Davies,
Mohammed K Ali
felix.teufel@emory.edu
Hubert Department of Global Health, Rollins School of Public Health,
Emory University, Atlanta, GA 30322, USA (FT, MKA); Emory Global Diabetes
Research Center of the Woodruff Health Sciences Center and Emory University,
Atlanta, GA, USA (FT, MKA); Health Systems Innovation Lab, Department of
Global Health and Population, Harvard T H Chan School of Public Health,
Harvard University, Boston, MA, USA (CAB); Heidelberg Institute of Global
Health, Heidelberg University Medical Center, Heidelberg, Germany (CAB);
Institute of Applied Health Research, University of Birmingham, Birmingham,
UK (JID); MRC/Wits Rural Public Health and Health Transitions Research Unit,
Faculty of Health Sciences, School of Public Health, University of the
Witwatersrand, Johannesburg, South Africa (JID); Centre for Global Surgery,
Department of Global Health, Stellenbosch University, Cape Town, South Africa
(JID); Department of Family and Preventive Medicine, School of Medicine,
Emory University, Atlanta, GA, USA (MKA)
1 WHO. Accelerating Access Initiative: widening access to care and support
for people living with HIV/AIDS: progress report, June 2002. Geneva, World
Health Organization, 2002.
Anti-obesity drugs, eating disorders, and thinness among
Japanese young women
Recently, GLP-1 receptor agonists, which were originally
developed for the treatment of diabetes, have attracted
much attention because of their weight loss effects in
people with obesity. In Japan, GLP-1 receptor agonists
were approved for the treatment of diabetes in
January, 2010, and in March, 2023, they were approved
for the treatment of obesity.
Since then, the number of cases of inappropriate
prescriptions for weight loss in healthy people has
increased globally, including in Japan. Physicians have
expressed particular concern about the rapid increase
in the number of inappropriate prescriptions among
Japanese women with a BMI of less than 25 kg/m².3
Some adverse events, including gastrointestinal
disorders and hypoglycaemia,1 have been reported
by this population. Many of these inappropriate
prescriptions are provided by doctors who are not
specialists in diabetes or other metabolic diseases.
To compound this issue, private, for-profit medical
institutions often prescribe drugs to patients who
2 Sun H, Saeedi P, Karuranga S, et al. IDF Diabetes Atlas: global, regional and
country-level diabetes prevalence estimates for 2021 and projections for
2045. Diabetes Res Clin Pract 2022; 183: 109119.
3 Flood D, Seiglie JA, Dunn M, et al. The state of diabetes treatment coverage
in 55 low-income and middle-income countries: a cross-sectional study of
nationally representative, individual-level data in 680 102 adults.
Lancet Healthy Longev 2021; 2: e340–51.
4 Chow CK, Ramasundarahettige C, Hu W, et al. Availability and affordability
of essential medicines for diabetes across high-income, middle-income,
and low-income countries: a prospective epidemiological study.
Lancet Diabetes Endocrinol 2018; 6: 798–808.
5 Kim SW, Skordis-Worrall J. Can voluntary pooled procurement reduce the
price of antiretroviral drugs? a case study of Efavirenz. Health Policy Plan
2017; 32: 516–26.
6 Pramesh CS, Sengar M, Patankar S, et al. A National Cancer Grid pooled
procurement initiative, India. Bull World Health Organ 2023; 101: 587–94.
7 Global Health & Population Project on Access to Care for Cardiometabolic
Diseases (HPACC). Expanding access to newer medicines for people with
type 2 diabetes in low-income and middle-income countries: a cost-
effectiveness and price target analysis. Lancet Diabetes Endocrinol 2021;
9: 825–36.
8 Bommer C, Sagalova V, Heesemann E, et al. Global economic burden of
diabetes in adults: projections from 2015 to 2030. Diabetes Care 2018;
41: 963–70.
9 Basu S, Flood D, Geldsetzer P, et al. Estimated effect of increased diagnosis,
treatment, and control of diabetes and its associated cardiovascular
risk factors among low-income and middle-income countries:
a microsimulation model. Lancet Glob Health 2021; 9: e1539–52.
10 Gregg EW, Buckley J, Ali MK, et al. Improving health outcomes of people
with diabetes: target setting for the WHO Global Diabetes Compact.
Lancet 2023; 401: 1302–12.
want to lose weight, and refuse to see patients who
have side-effects, claiming that they are not specialists
in the field.1 Diabetes specialists are demanding
regulations to deal with cases of side-effects caused by
such inappropriate prescriptions and the shortage of
GLP-1 receptor agonists due to prescriptions written
for dieting. In April, 2023, the Japan Diabetes Society
highlighted the widespread misuse of GLP-1 receptor
agonists. In June, 2023, four pharmaceutical companies,
Novo Nordisk, AstraZeneca, Sanofi, and Eli Lilly Japan,
raised the same concern. The situation remained
unchanged, and on July 28, 2023, the Japanese
Government issued a warning to municipalities,
medical institutions, and pharmacies nationwide
regarding the shortage of GLP-1 receptor agonists for
people with diabetes and obesity.
According to the latest National Health and Nutrition
Survey, the prevalence of women in Japan who are
underweight (BMI <18·5 kg/m²) is 20·7% among
women aged 20–29 years, and 11·5% among women
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
 Comment
RUNSTUDIO/Getty Images
For more on drugs approved in
Japan in 2010 see https://www.
mhlw.go.jp/shingi/2010/06/dl/
s0602-3a.pdf
For more on drugs approved
in Japan in 2023 see
https://www.mhlw.go.jp/
content/12404000/001167629.
pdf
90
diabetes medicines affordable globally is feasible and
discourse about the best ways to achieve this goal is
overdue.
CAB was supported by the Dutch Research Council’s Rubicon programme (project
number 452022313). MKA reported a past grant from Merck & Co to Emory
University and past consulting fees from Bayer and Eli Lilly and Company outside
the scope of the submitted work. All other authors declare no competing interests.
*Felix Teufel, Caroline A Bulstra, Justine I Davies,
Mohammed K Ali
felix.teufel@emory.edu
Hubert Department of Global Health, Rollins School of Public Health,
Emory University, Atlanta, GA 30322, USA (FT, MKA); Emory Global Diabetes
Research Center of the Woodruff Health Sciences Center and Emory University,
Atlanta, GA, USA (FT, MKA); Health Systems Innovation Lab, Department of
Global Health and Population, Harvard T H Chan School of Public Health,
Harvard University, Boston, MA, USA (CAB); Heidelberg Institute of Global
Health, Heidelberg University Medical Center, Heidelberg, Germany (CAB);
Institute of Applied Health Research, University of Birmingham, Birmingham,
UK (JID); MRC/Wits Rural Public Health and Health Transitions Research Unit,
Faculty of Health Sciences, School of Public Health, University of the
Witwatersrand, Johannesburg, South Africa (JID); Centre for Global Surgery,
Department of Global Health, Stellenbosch University, Cape Town, South Africa
(JID); Department of Family and Preventive Medicine, School of Medicine,
Emory University, Atlanta, GA, USA (MKA)
1 WHO. Accelerating Access Initiative: widening access to care and support
for people living with HIV/AIDS: progress report, June 2002. Geneva, World
Health Organization, 2002.
Anti-obesity drugs, eating disorders, and thinness among
Japanese young women
Recently, GLP-1 receptor agonists, which were originally
developed for the treatment of diabetes, have attracted
much attention because of their weight loss effects in
people with obesity. In Japan, GLP-1 receptor agonists
were approved for the treatment of diabetes in
January, 2010, and in March, 2023, they were approved
for the treatment of obesity.
Since then, the number of cases of inappropriate
prescriptions for weight loss in healthy people has
increased globally, including in Japan. Physicians have
expressed particular concern about the rapid increase
in the number of inappropriate prescriptions among
Japanese women with a BMI of less than 25 kg/m².3
Some adverse events, including gastrointestinal
disorders and hypoglycaemia,1 have been reported
by this population. Many of these inappropriate
prescriptions are provided by doctors who are not
specialists in diabetes or other metabolic diseases.
To compound this issue, private, for-profit medical
institutions often prescribe drugs to patients who
2 Sun H, Saeedi P, Karuranga S, et al. IDF Diabetes Atlas: global, regional and
country-level diabetes prevalence estimates for 2021 and projections for
2045. Diabetes Res Clin Pract 2022; 183: 109119.
3 Flood D, Seiglie JA, Dunn M, et al. The state of diabetes treatment coverage
in 55 low-income and middle-income countries: a cross-sectional study of
nationally representative, individual-level data in 680 102 adults.
Lancet Healthy Longev 2021; 2: e340–51.
4 Chow CK, Ramasundarahettige C, Hu W, et al. Availability and affordability
of essential medicines for diabetes across high-income, middle-income,
and low-income countries: a prospective epidemiological study.
Lancet Diabetes Endocrinol 2018; 6: 798–808.
5 Kim SW, Skordis-Worrall J. Can voluntary pooled procurement reduce the
price of antiretroviral drugs? a case study of Efavirenz. Health Policy Plan
2017; 32: 516–26.
6 Pramesh CS, Sengar M, Patankar S, et al. A National Cancer Grid pooled
procurement initiative, India. Bull World Health Organ 2023; 101: 587–94.
7 Global Health & Population Project on Access to Care for Cardiometabolic
Diseases (HPACC). Expanding access to newer medicines for people with
type 2 diabetes in low-income and middle-income countries: a cost-
effectiveness and price target analysis. Lancet Diabetes Endocrinol 2021;
9: 825–36.
8 Bommer C, Sagalova V, Heesemann E, et al. Global economic burden of
diabetes in adults: projections from 2015 to 2030. Diabetes Care 2018;
41: 963–70.
9 Basu S, Flood D, Geldsetzer P, et al. Estimated effect of increased diagnosis,
treatment, and control of diabetes and its associated cardiovascular
risk factors among low-income and middle-income countries:
a microsimulation model. Lancet Glob Health 2021; 9: e1539–52.
10 Gregg EW, Buckley J, Ali MK, et al. Improving health outcomes of people
with diabetes: target setting for the WHO Global Diabetes Compact.
Lancet 2023; 401: 1302–12.
want to lose weight, and refuse to see patients who
have side-effects, claiming that they are not specialists
in the field.1 Diabetes specialists are demanding
regulations to deal with cases of side-effects caused by
such inappropriate prescriptions and the shortage of
GLP-1 receptor agonists due to prescriptions written
for dieting. In April, 2023, the Japan Diabetes Society
highlighted the widespread misuse of GLP-1 receptor
agonists. In June, 2023, four pharmaceutical companies,
Novo Nordisk, AstraZeneca, Sanofi, and Eli Lilly Japan,
raised the same concern. The situation remained
unchanged, and on July 28, 2023, the Japanese
Government issued a warning to municipalities,
medical institutions, and pharmacies nationwide
regarding the shortage of GLP-1 receptor agonists for
people with diabetes and obesity.
According to the latest National Health and Nutrition
Survey, the prevalence of women in Japan who are
underweight (BMI <18·5 kg/m²) is 20·7% among
women aged 20–29 years, and 11·5% among women
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

20 years or older. Japan has the lowest obesity rate
in the Organisation for Economic Co-operation and
Development at 4·6%,2 and the highest proportion of
women who are underweight.3 The increase in the rate
of low-birthweight babies has been reported as one
of the consequences of the prevalence of people who
are underweight.3 Measures to address the prevalence
of being underweight among young women in Japan
have become a public health priority. The Japanese
Government has been raising awareness of the health
risks associated with thinness and providing nutritional
education, but has not achieved sufficient results.
There are several factors contributing to the trend
of young women in Japan seeking to lose weight.
Firstly, there is a strong desire for thinness among
young women, influenced by the fact that east Asian
countries such as Japan are classified as collectivist
cultures in which conformity to social norms can
more substantially affect food behaviours than in
individualistic countries.4 In Japanese society, the
media has embraced the western appearance and
thinness, which is different from the appearance and
body shape of typical Japanese people, and young
women can feel conflicted between this so-called ideal
and their reality.5 Because of the collectivist emphasis
on conformity to social norms in Japan, young women
can see discrepancies between their body shapes and
the media’s ideal of unrealistic thinness—lowering
both self-esteem and body esteem.4 To achieve this
thinness, some young Japanese women have resorted
to pathological weight management practices such
as skipping meals, compulsive exercise, vomiting,
and the use of laxatives.5 Notably, many women who
fall into the underweight category in terms of BMI
perceive themselves as being of normal weight.6 In
addition, although a higher BMI is usually associated
with a stronger desire to lose weight, research has
shown that among young Japanese women, a lower
BMI is linked to a stronger desire for thinness.6
Furthermore, an increase in eating disorders (eg,
anorexia nervosa, bulimia nervosa, and binge eating
disorder) has recently been reported in Japan.7
In particular, the prevalence of anorexia nervosa in
young women has been reported to have increased
from 2·5 per 100 000 people in the 1980s to more
than 10 per 100 000 people in the 2000s and 2010s.7
The recent increase in the incidence of eating disorders
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Comment
among Japanese women has been reported to be
attributable to environmental changes, such as the
increased availability of high-calorie foods as society
simultaneously demands thinness, and increased social
stress, such as gender-based discrimination in the
workplace as more women enter the workforce.7
Japanese society has failed to fully understand and
address the fact that emaciation among young women
has become a public health issue, and that the culture
and climate of idealising emaciation has driven young
women into a corner. Additionally, social media has
added to the pressure on women to become thinner. In
the USA and other western countries, a body positivity
movement has emerged in recent years, in which
people accept all body types, including their own. This
movement is thought to be contributing to an improved
sense of body image in young women;8 however, it has
not yet gained traction in Japan, and both traditional
and social media predominantly focus on extreme
thinness, exposing young women to an unrealistic body
image.8 In addition, media advertisements encourage
disordered eating behaviours such as fad diets or
supplements. Moreover, some young women in Japan
are resorting to invasive appearance-altering practices
such as cosmetic procedures as they struggle with the
conflict between western and eastern values.8 East Asia
is known for its strong interest in cosmetic surgery, with
China, Japan, and South Korea ranking third, fourth, and
fifth, respectively, among the top five countries with
the highest number of plastic surgeons. Japan ranks
third in terms of the number of performed cosmetic
procedures, following the USA and Brazil.9 Only recently
has this distortion of the body image of young Japanese
women been acknowledged. In April, 2023, a call for
research support of up to ¥400 million (US$2·7 million)
was issued under a programme directly administered by
Japan’s Cabinet Office,10 and the project entitled Jyosei No
Body Image To Kenko Kaizen No Tameno Kenkyu Kaihatsu
(Development of Social Technologies that Promote the Value
of Body Image as a New Kind of Beauty), under principal
investigator Yoshifumi Tamura, was accepted and started.
This government programme, Strategic Innovation
Promotion, supports the research and development
of educational and intervention programmes in
Japan to promote body image transformation among
the younger generation, and requires applicants to
establish cooperative frameworks with industries and
91
 Comment
other relevant entities related to body image, such as
cosmetics, clothing, and food products.10
The strong desire for thinness among young women
in Japan due to cultural and social pressures and the
adoption of invasive approaches for beauty and weight
loss contributes to the inappropriate use of GLP-1
receptor agonists in Japan and is becoming a public
health concern. The widespread misuse of these drugs
might enhance an obsession with thinness and result
in further negative consequences even for those who
currently do not take them. In addition, there is concern
that some young women who use or wish to use GLP-1
receptor agonists for weight loss might already be
struggling with eating disorders.3 These developments
pose additional public health risks in Japan, where the
number of patients with eating disorders is increasing,
yet access to appropriate treatment is limited by
a shortage of specialists who can treat eating disorders.9
With even stronger anti-obesity drugs expected to
be released soon, the Government of Japan should
strengthen regulations to prevent the spread of
inappropriate prescriptions for anti-obesity drugs.
SK declares no competing interests. TK has received grants from Sumitomo
Pharma and Otsuka Pharma; royalties or licenses from Sumitomo Pharma and
FRONTEO; consulting fees from TechDoctor and FRONTEO; speaker’s honoraria
from Sumitomo Pharma, Boehringer Ingelheim, Takeda, Astellas, Meiji Seika,
and Janssen; and stock from i2medical and TechDoctor.
The role of medical students in humanitarian responses to
armed conflict
Published Online The Russia–Ukraine and Israel–Palestine armed conflicts
January 3, 2024
https://doi.org/10.1016/
have been widely covered by the media in 2023. However,
S2213-8587(23)00366-2 on a global scale, humanity is no stranger to violence with
many countries currently experiencing armed conflict,
terrorism, political unrest, or ethnic violence.1 There
is great demand for medical professional expertise to
assist with these humanitarian crises. Compounding the
pre-conflict medical workforce shortages often present,
conflict can often displace health-care workers internally
and across borders, interrupts medical education, and
causes disproportionate loss of life among working-age
individuals who could form future health workforces.2,3
One potential approach to addressing this shortage lies in
unlocking the latent potential of medical students, which
in our opinion are a predominantly young, energetic,
92
Shotaro Kinoshita, *Taishiro Kishimoto
tkishimoto@keio.jp
Graduate School of Interdisciplinary Information Studies, The University of Tokyo,
Tokyo, Japan (SK); Hills Joint Research Laboratory for Future Preventive Medicine
and Wellness, Keio University School of Medicine, Tokyo 106-0041, Japan (SK, TK)
1 Nikkei Medical. The “GLP-1 diet” for cosmetic purposes: health hazards and
responses. July 21, 2023. https://medical.nikkeibp.co.jp/all/weekly/
images/20230721_weekly.pdf (accessed Oct 28, 2023; in Japanese).
2 Organisation for Economic Co-operation and Development. OECD Health
Statistics 2023. July 3, 2023 https://www.oecd.org/health/health-data.htm
(accessed Nov 26, 2023).
3 Nakanishi K, Saijo Y, Yoshioka E, et al. Severity of low pre-pregnancy body
mass index and perinatal outcomes: the Japan Environment and Children’s
Study. BMC Pregnancy Childbirth 2022; 22: 121.
4 Chisuwa N, O’Dea JA. Body image and eating disorders amongst Japanese
adolescents. A review of the literature. Appetite 2010; 54: 5–15.
5 Kowner R. When ideals are too “far off’: physical self-ideal discrepancy and
body dissatisfaction in Japan. Genet Soc Gen Psychol Monogr 2004;
130: 333–61.
6 Yasuda T. Desire for thinness among young Japanese women from the
perspective of objective and subjective ideal body shape. Sci Rep 2023;
13: 14129.
7 Nakai Y, Nin K, Goel NJ. The changing profile of eating disorders and related
sociocultural factors in Japan between 1700 and 2020: a systematic
scoping review. Int J Eat Disord 2021; 54: 40–53.
8 Ando K, Giorgianni FE, Danthinne ES, Rodgers RF. Beauty ideals, social
media, and body positivity: a qualitative investigation of influences on
body image among young women in Japan. Body Image 2021; 38: 358–69.
9 Yim SH, Schmidt U. The effectiveness and cultural adaptations of
psychological interventions for eating disorders in East Asia: a systematic
scoping review. Int J Eat Disord 2023; published online Sept 19. https://doi.
org/10.1002/eat.24061.
10 National Research and Development Agency, National Institute of
Biomedical Innovation, Health and Nutrition. Cross-ministerial strategic
innovation promotion program: construction of inclusive community
platforms, guidelines for publicly invited applications for FY2024.
April, 2023. https://www.nibiohn.go.jp/sip3-housetsu/files/sip3housetsu_
koubo_guidelines_202304.pdf (accessed Oct 28, 2023; in Japanese).
and altruistic group with developing medical expertise,
uniquely placed to assist in this context.
Responses by medical students to crises is no
for­
eign concept, documented during conflicts in
Bosnia and Herzegovina (1991–95), Nagorno-Karabakh
(2020 to present), and Russia–Ukraine (2022 to
present).2,4,5 Students have collected casualty data,
educated the public, coordinated physicians and
researchers to produce war-related academic publi­
cations, performed medical history taking, and prepared
battlefield first aid kits.2,4,5 Those who received additional
specific train­ ing provided psychosocial support to
refugee children.4 It is important to note, however, with
the rise of non-communicable diseases (NCDs) globally,
especially in regions of humanitarian crises, there is an
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
 Comment
other relevant entities related to body image, such as
cosmetics, clothing, and food products.10
The strong desire for thinness among young women
in Japan due to cultural and social pressures and the
adoption of invasive approaches for beauty and weight
loss contributes to the inappropriate use of GLP-1
receptor agonists in Japan and is becoming a public
health concern. The widespread misuse of these drugs
might enhance an obsession with thinness and result
in further negative consequences even for those who
currently do not take them. In addition, there is concern
that some young women who use or wish to use GLP-1
receptor agonists for weight loss might already be
struggling with eating disorders.3 These developments
pose additional public health risks in Japan, where the
number of patients with eating disorders is increasing,
yet access to appropriate treatment is limited by
a shortage of specialists who can treat eating disorders.9
With even stronger anti-obesity drugs expected to
be released soon, the Government of Japan should
strengthen regulations to prevent the spread of
inappropriate prescriptions for anti-obesity drugs.
SK declares no competing interests. TK has received grants from Sumitomo
Pharma and Otsuka Pharma; royalties or licenses from Sumitomo Pharma and
FRONTEO; consulting fees from TechDoctor and FRONTEO; speaker’s honoraria
from Sumitomo Pharma, Boehringer Ingelheim, Takeda, Astellas, Meiji Seika,
and Janssen; and stock from i2medical and TechDoctor.
The role of medical students in humanitarian responses to
armed conflict
Published Online The Russia–Ukraine and Israel–Palestine armed conflicts
January 3, 2024
https://doi.org/10.1016/
have been widely covered by the media in 2023. However,
S2213-8587(23)00366-2 on a global scale, humanity is no stranger to violence with
many countries currently experiencing armed conflict,
terrorism, political unrest, or ethnic violence.1 There
is great demand for medical professional expertise to
assist with these humanitarian crises. Compounding the
pre-conflict medical workforce shortages often present,
conflict can often displace health-care workers internally
and across borders, interrupts medical education, and
causes disproportionate loss of life among working-age
individuals who could form future health workforces.2,3
One potential approach to addressing this shortage lies in
unlocking the latent potential of medical students, which
in our opinion are a predominantly young, energetic,
92
Shotaro Kinoshita, *Taishiro Kishimoto
tkishimoto@keio.jp
Graduate School of Interdisciplinary Information Studies, The University of Tokyo,
Tokyo, Japan (SK); Hills Joint Research Laboratory for Future Preventive Medicine
and Wellness, Keio University School of Medicine, Tokyo 106-0041, Japan (SK, TK)
1 Nikkei Medical. The “GLP-1 diet” for cosmetic purposes: health hazards and
responses. July 21, 2023. https://medical.nikkeibp.co.jp/all/weekly/
images/20230721_weekly.pdf (accessed Oct 28, 2023; in Japanese).
2 Organisation for Economic Co-operation and Development. OECD Health
Statistics 2023. July 3, 2023 https://www.oecd.org/health/health-data.htm
(accessed Nov 26, 2023).
3 Nakanishi K, Saijo Y, Yoshioka E, et al. Severity of low pre-pregnancy body
mass index and perinatal outcomes: the Japan Environment and Children’s
Study. BMC Pregnancy Childbirth 2022; 22: 121.
4 Chisuwa N, O’Dea JA. Body image and eating disorders amongst Japanese
adolescents. A review of the literature. Appetite 2010; 54: 5–15.
5 Kowner R. When ideals are too “far off’: physical self-ideal discrepancy and
body dissatisfaction in Japan. Genet Soc Gen Psychol Monogr 2004;
130: 333–61.
6 Yasuda T. Desire for thinness among young Japanese women from the
perspective of objective and subjective ideal body shape. Sci Rep 2023;
13: 14129.
7 Nakai Y, Nin K, Goel NJ. The changing profile of eating disorders and related
sociocultural factors in Japan between 1700 and 2020: a systematic
scoping review. Int J Eat Disord 2021; 54: 40–53.
8 Ando K, Giorgianni FE, Danthinne ES, Rodgers RF. Beauty ideals, social
media, and body positivity: a qualitative investigation of influences on
body image among young women in Japan. Body Image 2021; 38: 358–69.
9 Yim SH, Schmidt U. The effectiveness and cultural adaptations of
psychological interventions for eating disorders in East Asia: a systematic
scoping review. Int J Eat Disord 2023; published online Sept 19. https://doi.
org/10.1002/eat.24061.
10 National Research and Development Agency, National Institute of
Biomedical Innovation, Health and Nutrition. Cross-ministerial strategic
innovation promotion program: construction of inclusive community
platforms, guidelines for publicly invited applications for FY2024.
April, 2023. https://www.nibiohn.go.jp/sip3-housetsu/files/sip3housetsu_
koubo_guidelines_202304.pdf (accessed Oct 28, 2023; in Japanese).
and altruistic group with developing medical expertise,
uniquely placed to assist in this context.
Responses by medical students to crises is no
for­
eign concept, documented during conflicts in
Bosnia and Herzegovina (1991–95), Nagorno-Karabakh
(2020 to present), and Russia–Ukraine (2022 to
present).2,4,5 Students have collected casualty data,
educated the public, coordinated physicians and
researchers to produce war-related academic publi­
cations, performed medical history taking, and prepared
battlefield first aid kits.2,4,5 Those who received additional
specific train­ ing provided psychosocial support to
refugee children.4 It is important to note, however, with
the rise of non-communicable diseases (NCDs) globally,
especially in regions of humanitarian crises, there is an
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

increas­ing need amongst conflict-affected populations
for specialist physicians, in addition to more traditional
requirements—eg, trauma surgeons.6 Additionally, tech­
nological advance­ments, such as the expanded use of
telemedicine since the COVID-19 pandemic, enable new
models of medical aid projects that medical students are
uniquely placed to facilitate.
In this Comment we present a cate­gorisation of how
medical students around the world can assist with
conflict-related humanitarian crises (table).
Final-year medical students can use their clinical skills
in local health-care settings, including performing medi­
cal interviews, venepuncture, intravenous cannu­lation,
and non-operative injury care, which allows nurses and
doctors to attend to urgent or more complex cases.
Those trained in mental health can assist with psycholog­
ical counselling—a historically under-recognised need.
Time consuming yet relatively fundamental tasks such
as diabetes management education (including dietary
management, glucose checking, medication manage­
ment, and complication screening), could potentially be
performed by even early year students. Universities in
conflict-prone regions might consider curricula equipping
medical students, and thus future doctors, for scenarios
encountered during armed hostilities.
Medical students of any year can educate the public on
life-saving skills, including basic life support, wound care,
and essential concepts (eg, acute stress management),
thereby equipping civilians to potentially save lives when
health infrastructure is inaccessible or destroyed.4 Effective
NCD care involves addressing health system inequities and
meeting unmet health needs of vulnerable populations
in a culturally appropriate manner. Community health
workers are respected local community members who
have received brief education and training to support the
local health system, predominantly in under-resourced
countries. Medical students can be trained to perform
the health promotion and disease prevention tasks of
community health workers with cultural competency.8
Because prevention and management of diabetes, for
instance, benefits from knowledge of diet and practices
unique to each community, local medical students with
this locality-specific knowledge are well placed to provide
patient education.
Having knowledge of medical language, local and
expatriate medical students are well-equipped to act as
medical interpreters. One example is TeleHelp Ukraine
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Comment
Examples of roles
Students in the affected region
Clinical and procedural Medical interviews; procedural skills (venepunctures, intravenous cannulations,
skills and non-operative injury care); mental health first aid
Public education Basic life support (eg, cardiopulmonary resuscitation); basic wound care; acute
stress management; community health worker tasks
Students in the affected region and out of the affected region
Medical interpreter and Interpretation during consultations (in-country international aid mobile clinic or
translator via telemedicine); translation of physician treatment notes and investigation
results (eg, radiology reports)
Students out of the country
Facilitating access to Creating and coordinating telemedicine platform, in which physicians can:
foreign medical provide consultations to patients directly (eg, TeleHelp Ukraine), offer second
expertise opinions to local practitioners, deliver educational sessions (eg, roundtable,
webinar, and question and answer), train generalists and nurses in particular
technical skills, and provide mental health peer-counselling for health
practitioners on the ground
External philanthropic Founding charities; traditional fundraising seeking monetary donations;
projects organising transfer of medical supplies from health-care institutions to conflict
zone (eg, hospital consumables, first aid items, and medications)
Table: Categorisation of roles for medical students in conflict-related humanitarian crises
which, initiated by medical students and faculties
of Stanford University, USA, provides free telehealth
consultations to civilians affected by the Russia-Ukraine
conflict in the Russia-Ukraine war. In the experience
of TeleHelp Ukraine, medical students who can speak
Ukrainian and English, many of whom are second and
third generation Ukrainian migrants residing in North
America, act as both interpreters during consultations
and translators for consultation notes and investigation
results (eg, radiology reports). Having the relevant lan­
guage fluency is not limited to locals and expatriates
but also subsequent generations of migrant families.
Connection with the native country is a strong motiva­
tional force for international aid projects as also demon­
strated by Telehelp Ukraine, whereby most founders are
first to third generation Ukrainian medical students in
the USA.
In areas of humanitarian crises, with the destruction
of health facilities, casualties, internal displacement,
and increasing NCDs including diabetes, there is
a press­ing need for easy access to medical specialists.3
Via telemedicine, international specialists can provide
remote consultations into regions of hostilities directly
with patients and supporting local practitioners.3,7
Medical students can act to connect local and inter­
national stakeholders because their medical training
and knowl­edge of health-care workflow allows them
to organise telemedicine platforms and identify For more on TeleHelp Ukraine
medical issues in conflict zones. Humanitarian actions see https://telehelpukraine.com/
93
 Comment
might extend to founding charities, seeking monetary
dona­tions, and organising transfer of surplus or near
expiration medical supplies from health-care institutions
to the conflict zone (eg, hospital consumables, first aid
supplies, and essential medications such as insulin9). For
instance, the Ukrainian Students Association at Stanford
University, comprised of medical student volunteers, has
delivered over 600 000 doses of insulin to people with
diabetes in Ukraine.9
Although medical students might be interested in
actively assisting during humanitarian crises, ministries
of health and universities must defend ongoing medical
education in areas of active humanitarian crises as
a primary responsibility is to maintain the future of the
country’s health workforce. Having not completed their
training, medical students might also be less resilient
to psychological trauma than more experienced junior
doctors. Therefore, deployment of medical students in
conflict response must be done cautiously to prevent
compounding psychological burden. Although medical
students occupy a unique position, trainees of other
health disciplines might be better placed to address
specific workforce gaps and should not be neglected (eg,
nursing students trained in wound care and psychology
students trained in psychotrauma counselling).
Involvement of medical students in humanitarian
crisis response is a unique concept previously employed
in conflict settings. However, a scarcity of published
94
literature calls for further exploration, which can
motivate universities, health-care institutions, and
students themselves to realise the latent potential
of medical students in a crisis setting to make up for
systemic deficiencies to help local people in need.
We declare no competing interests.
*Ji Won Susie Yoo, Rahul D Barmanray
jiyoo@student.unimelb.edu.au
Department of Medicine (JWSY, RDB) and Department of Diabetes &
Endocrinology (RDB), The Royal Melbourne Hospital, Melbourne, VIC 3050,
Australia; TeleHelp Україна (TeleHelp Ukraine), Stanford University, Palo Alto,
CA, USA (JWSY)
1 The World Factbook. Field listing – disputes – international. https://www.
cia.gov/the-world-factbook/field/disputes-international/ (accessed
Oct 24, 2023).
2 Srichawla BS, Khazeei Tabari MA, Găman MA, Munoz-Valencia A,
Bonilla-Escobar FJ. War on Ukraine: impact on Ukrainian medical students.
Int J Med Stud 2022; 10: 15–17.
3 Boulle P, Kehlenbrink S, Smith J, Beran D, Jobanputra K. Challenges
associated with providing diabetes care in humanitarian settings.
Lancet Diabetes Endocrinol 2019; 7: 648–56.
4 Gluncić V, Pulanić D, Prka M, Marusíc A, Marusíc M. Curricular and
extracurricular activities of medical students during war, Zagreb University
School of Medicine, 1991-1995. Acad Med 2001; 76: 82–87.
5 Ibrahimli A. A medical student’s volunteering experience during the second
Nagorno-Karabakh War. Int J Med Stud 2021; 9: 312–13.
6 Ndubuisi NE. Noncommunicable diseases prevention in low- and middle-
income countries: an overview of health in all policies (HiAP). Inquiry 2021;
58: 46958020927885.
7 Zhou C, Crawford A, Serhal E, Kurdyak P, Sockalingam S. The impact of
project ECHO on participant and patient outcomes: a systematic review.
Acad Med 2016; 91: 1439–61.
8 Rawal L, Jubayer S, Choudhury SR, Islam SMS, Abdullah AS. Community
health workers for non-communicable diseases prevention and control
in Bangladesh: a qualitative study. Glob Health Res Policy 2020; 6: 1.
9 Ukraine Support Alliance at Stanford. Our work. https://usas.stanford.edu/
our-work/ (accessed Oct 24, 2023).
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

Profile
Rafik Jacob: Caring for the most vulnerable
Growing up in Egypt, back in the
early 1970s, Rafik Jacob was struck by
the challenges faced by families who
had children with intellectual and
developmental disabilities. “There
were not a lot of resources around.
Parents had to figure every­thing out
for themselves. The whole family
was affected. It was difficult for them
to make plans or go on vacation.
They could not enjoy life to its fullest
potentia,” said Jacob, who is now
Medical Director of the Jacksonville
Program for Adults with Intellectual and
Developmental Disabilities (PAIDD),
at the University of Florida College of
Medicine (Jacksonville, FL, USA).
Jacob’s closest childhood friend had
a brother with autism. “Of course,
there were triumphs and good times,
but life in that household was far
harder than it needed to be. It was
not just them. It was common to see
families trying to navigate a maze
of obstacles in their effort to care
for a loved one with an intellectual
and developmental disabilitiy.
They were not given the kind of
support that they deserved and the
stigmatisation made everything
worse. People were still using terms
like ‘mentally retarded’”, Jacob told
The Lancet Diabetes & Endocrinology. “It
made me think about what I wanted
to do with my life; what could I do to
make a difference?”
After taking a medical degree at Ain
Shams University (Cairo, Egypt), Jacob
practised privately in Egypt. The move
to the USA came in 2009. A residency
in internal medicine at Brown
University (RI, USA) followed; the
holistic nature of the specialty
appealed to Jacob. He then applied for
a position at the University of Florida.
“When I went for the interview, the
chief of division told me about their
vision to establish a programme to
take care of adults with intellectual and
developmental disabilities. It mapped
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
completely onto my own objectives”,
recalled Jacob. “It was a wonderful
stroke of luck—the best thing that has
happened in my professional career.”
Jacob started as co-director at PAIDD
in 2013. His clinical work is roughly
divided one-third to the patient and
two-thirds to their family. “We help
people get the funding and extra
support that they need, as well as
some kind of respite”, said Jacob.
He stressed that such work is only
possible in an environment in which
reimbursement is not the central
concern. People with intellectual and
developmental disabilities require
more time, around 40 min for each
appointment, and they are typically
only covered by Medicaid. “More effort
for less money—not something most
private practices are very keen on”,
said Jacob. “The university supports
our programme, so we can promote
health outcomes without worrying
about financial incentives, which is
why I decided to become a doctor in
the first place.”
In his free time, Jacob volunteers at
a clinic for underinsured and uninsured
Americans. He also attempts to interest
medical students and residents in
working with people with intellectual
and developmental disabilities. “My
goal is for young physicians to feel
like they have enough infor­mation
and understanding to be able to treat
this group of patients”, he said. “I try
to publish on the common conditions
that affect people with intellectual and
developmental disabilities .”
Diabetes is a major focus. Individuals
with an intellectual and developmental
disability are more likely to develop
diabetes and more likely to experience
complications. Treating these patients
can be challenging. “You have to take
into account the support system”,
explained Jacob. “People may not be
able to depend on themselves 100%,
in terms of following a diet or doing
In Focus
exercise, so you should be able to think
imaginatively how best to treat them.
They may not like needles—it’s okay
so find options that do not require
injections.” He cautioned against
the assumption that intellectual and
developmental disabilities necessarily
equate to worse outcomes. “Keep
an open mind”, advised Jacob. “Do
not think, ‘this person is not fully
functional so we have to accept that
their blood sugar control is going to
be much worse than someone without
an intellectual and developmental
disability’.”
US physical activity guidelines for
all adults recommend 150–300 min of
moderate exercise, or 75–150 min
of intensive exercise, every week.
Jacob is leading a research project
in which people with intellectual
and developmental disabilities and
diabetes and their caregivers engage
in regular exercise programmes. The
preliminary results are encouraging.
“We think if the caregiver is involved,
it will improve outcomes for both
parties”, said Jacob.
All of which is in keeping with
Jacob’s ethos of connectivity.
“Medicine is not just about the
patient in front of you. It is about
their family and the society they
are living in”, he said. “Diabetes is
a good example. Why do people
with intellectual and developmental
disabilities have a greater prevalence
of diabetes? Why are they more likely
to receive worse care? What does
it mean to have an intellectual and
developmental disability and diabetes
in Egypt compared with the USA? You
cannot look at the patient to get an
answer to these questions. You have
to examine what the medical system
is doing, and how society values
and respects its most vulnerable
members.”
Talha Burki
95
 In Focus
Book
The fight to put food on the table: from past to present
Roughly halfway through Stuffed:
A History of Good Food and Hard Times in
Britain, Pen Vogler likens supermarkets
to the protagonist in Goethe’s
The Sorcerer’s Apprentice. “The novice
knows how to cast a spell to get his
broom to bring water—but not the
magic to end it”, writes Vogler. “His
attempt to stop the enchanted broom
simply makes two brooms working at
double the speed to bring water, which
floods the kitchen floor.” Vogler asks
Published Online the reader to think of the broom as the
January 5, 2024 business model and the water as cheap
https://doi.org/10.1016/
S2213-8587(23)00389-3 food. “Each company feels forced to
Stuffed: A History of Good compete by making the food cheaper,
Food and Hard Times in Britain but the only way they can profit from
Pen Vogler cheap food is to sell more of it than
Atlantic Books, 2023
pp 453, £22∙00
customers need. It is flooding the
ISBN 9781838955748 market with too much food”, she
explains. “The supermarkets are the
hapless apprentices, implicated in
a disastrous system they have initiated
but can’t control.”
It is a fun passage and it is typical
that Vogler would choose a comic
scene from high literature to make
a serious point. But the analogy is
imperfect. Two pages earlier, we learned
that by the end of the 20th century,
the four biggest supermarket chains
handled 60% of all grocery sales in the
UK. “Their profitability hinged, to a large
extent, on their ability to abandon
traditional wholesalers, and squeeze
suppliers directly”, notes Vogler. Far
from being at the mercy of the market,
supermarkets are the market.
The author closes by recommending
that the government intervene
with legislation, although she
does not explain what form this
legislation should take, and implies
that supermarkets would welcome
such measures. Indeed, while Vogler
is generally clear-sighted, crisply
96
describing the historical purpose of
free trade, she occasionally betrays
a slightly naive attitude towards big
business. Food producers may well
have “moral consciences”, as Vogler
states in the conclusion to the book,
but they also have fiduciary duties to
their shareholders, which are likely
to be far more powerful. “Businesses
must plan for future as well as current
customers”, writes Vogler, hinting
that it is not in their interest to wreck
the health of the nation. But she had
previously outlined how, by fostering
a taste for sugar in youngsters, the
food industry helps to shape the habits
of the next generation of adults.
Vogler’s previous book was Scoff,
a breezy history of the relationship
between food and class. Stuffed is
similarly structured: a succession
of short chapters, each centred
on a different food or beverage
representative of an idea or event.
The chapter on geese, for example,
examines the enclosures—the fencing
off of previously common land (geese
do not do well unless they can range).
There are evocative accounts of ancient
feasts, customs of hospitality, and
allotments. The Christmas pudding
section is a highlight. Previously known
as plum pudding, until immortalised
by Charles Dickens in A Christmas Carol,
the dish is comprised of ingredients
gathered from all over the world.
Vogler introduces the chapter with the
words, “how the sweet, fruity, spiced,
alcohol-soaked, globe-shaped King’s
Empire Christmas Pudding of 1928
fudged Britain’s conflict between free
trade and Empire”.
The overarching theme of Stuffed is
that what we eat and how much of it
is available is conditioned by economic,
historical, and political factors over
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
which the individual has very little
influence. Vogler is determined to
go further than this, however. She
attempts to draw a direct connection
between the events she describes and
the food crises that prevail in modern
Britain. Her efforts are unconvincing.
Vogler is better at drawing parallels
than providing analysis. The conclusion
makes explicit an argument that is
barely present in the rest of the book,
for example. There is not a great deal
in the book on the past few decades
or on the geopolitical trends that drive
the obesity epidemic. It is strange that
Vogler does not include a chapter on
turkey twizzlers, chicken nuggets,
chips, chocolates, or any kind of ultra-
processed food. As far as I could tell,
none of the 26 chapters focuses on
food or drink that would not have been
available in the 17th century.
Nonetheless, as an episodic history
of food in what is now the UK, Stuffed
does work. Vogler is such a skilled
writer, she can even make 14 pages on
the turnip seem interesting. Refracting
history through a specific food has its
advantages. Vogler uses salt beef to
examine the problems of keeping
troops supplied during the Napoleonic
and Crimean Wars. The approach
brings an immediacy to the story
and allows the author to sketch the
experiences of those doing the fighting
and those doing the organising. Stuffed
amply demonstrates how food can
tell us something meaningful about
particular eras and cultures. Perhaps it
is a point that we intuitively recognise.
After all, when people return from
holiday, one of the first questions they
are asked is always, “what was the
food like?”
Talha Burki

Better testing and labelling of drugs for people with obesity
On Nov 28, 2023, several obesity
advocacy organisations issued a joint
statement calling on the US Food
and Drug Administration (FDA) to
stipulate that people with obesity
must be included in drug trials. They
also asked the FDA to mandate that
any obesity-related differences in the
effectiveness or safety of a drug should
be detailed on its label. “Exclusion
criteria routinely bar people at higher
bodyweights from participating in
[drug] trials”, noted the statement.
“Yet it is well known drug kinetics can
materially change in larger bodies...
that puts people with obesity, who
make up 42% of the US population,
at unnecessary risk of poorer health
outcomes and adverse events,
including death.”
Caroline Apovian is Co-Director of
the Center for Weight Management
and Wellness at Brigham and
Women’s Hospital (Boston, MA,
USA) and former president of the
Obesity Society, one of the signatories
to the joint statement. She told
The Lancet Diabetes & Endocrinology
that the call for inclusion and the
call for accurate drug labelling are
intertwined. “How can you expect
to know how a medication works in
people with higher bodyweights if you
are deliberately excluding these people
from your clinical trials?”, she asked.
At the centre of the issue are
lipophilic drugs, which exhibit
a greater affinity for adipose tissue,
resulting in a much larger volume
of distribution and lower serum
concentration in people with obesity,
reducing the level of drug circulating
in the bloodstream. This in turn can
delay or even prevent the drug from
becoming effective. Unless prescribers
are able to take into account how
a drug works in people with obesity,
there is the risk of underdosing,
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
pre­mature abandonment of a med­
ication wrongly thought to be inef­
fective, or the continuation of
a genuinely ineffective medication. All
of which could result in people with
obesity essentially going untreated.
Drugs for cancer, depression,
schizophrenia, and pain relief have
been found to work differently in
people with obesity. Some oral
contraceptives become less effective
as body mass increases. However,
in this instance, the FDA has taken
action. FDA guidance on hormonal
drugs intended to prevent pregnancy
recommends that clinical trials
include women with BMIs of 30 and
above, and that the labelling of such
drugs should spell out whether there
are adequate data on their use in
populations with obesity.
Mark Kelson, Professor of Statistics
for Health at the University of
Exeter (Exeter, UK), pointed out that
randomised clinical trials are run on
the basis that the randomisation
process more or less equates to
generalisability. “It is never a truly
random sample”, he explained. “You
are typically taking people who have
the time and inclination to participate
and who are from a certain area. The
idea is that the process is rigorous
enough to ensure that the results of
the trial will translate to people who
were not involved.” Systematically
excluding entire populations disrupts
this model, however. “I would start to
worry about the integrity of the entire
process”, said Kelson.
Besides, simply being involved in
a clinical trial brings benefits of its own.
“It does not matter whether you are in
the control group or the treatment
group, participating is good for your
medical care”, said Kelson. “We should
be opening this opportunity to
as many people as possible.” The
In Focus
inclusion of individuals with obesity is
Westend61/Getty Images
a matter of equity. “People with higher
bodyweights are under-represented
in trials”, said Apovian. “They do not
tend to visit their doctor as frequently
and they are more likely to be from
marginalised groups.” Kelson believes Published Online
that drug companies would be willing January 4, 2024
https://doi.org/10.1016/
to accommodate adjustments in line S2213-8587(23)00387-X
with those in the joint statement. For the obesity advocacy
“It is in the industry’s interest to organisations joint statement
make sure that their medications are see https://www.obesityaction.
org/statement-drug-approval-
optimised for everyone to whom they labeling/
are prescribed”, he said. “They are keen
on robust studies that generate solid
data.”
This raises the question of why the
joint statement was necessary in the
first place. Apovian suggested that
apathy could be to blame. “Clinical
trials are a labour intensive process.
Making sure that your patient
population is representative requires
even more time and effort”, she said.
Kelson urged those running clinical
trials to continually scrutinise the
inclusion criteria and question any
assumptions. “Without a medical
rationale, it is sheer laziness to bar
anyone with a BMI above 30. If they
are going to be given the drug, then
they need to be allowed into the trial”,
he said. Nonetheless, he cautioned
against imposing quotas. “I would
not like to see a situation where we
are demanding that the composition
of the trial population matches the
composition of the general population;
under-recruitment is a huge issue
and we do not want to put up any
more barriers”, he said. In which case,
preventing trials from unnecessarily
excluding people with obesity would
have the additional advantage of
increasing the pool of those eligible for
recruitment.
Talha Burki
97
 Corrections

Correction to Lancet
Diabetes Endocrinol
2023; 11: 42–57
Battelino T, Alexander CM, Amiel SA,
et al. Continuous glucose monitoring
and metrics for clinical trials: an
international consensus statement.
Lancet Diabetes Endocrinol 2023;
11: 42–57—In this Review, some
values in table 3 should have read 3·9.
This correction has been made to the
online version as of Jan 23, 2024.

Correction to Lancet
Diabetes Endocrinol
2023; 11: 915–25
Taylor PN, Collins KS, Lam A, et al. Published Online
C-peptide and metabolic outcomes December 11, 2023
https://doi.org/10.1016/
in trials of disease modifying therapy S2213-8587(23)00381-9
in new-onset type 1 diabetes: an
individual participant meta-analysis.
Lancet Diabetes Endocrinol 2023;
11: 915–25—In this Article, the
x axis for figures 2A–F should have
been labelled from the smallest
percentage to the largest. For
figure 2G, the colour key and
graph should have been labelled as
>76% preservation red, 48–76% green,
22–47% blue, and <22% purple. For
figure 3B, the colour key should have
been >0·92 red, 0·64–0·92 blue,
0·44–0·63 green, and <0·44 purple,
and the appendix has been corrected.
These corrections have been made to
the online version as of Dec 11, 2023.

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 e12

 Articles
Lancet Diabetes Endocrinol
2024; 12: 98–106
Published Online
December 21, 2023
https://doi.org/10.1016/
S2213-8587(23)00327-3
See Comment page 84
School of Public Health and
Preventive Medicine, Monash
University, Melbourne, VIC,
Australia (Prof S Zoungas PhD,
Z Zhou PhD, A J Owen PhD,
A J Curtis PhD,
Prof R L Woods PhD,
S G Orchard PhD,
Prof J J McNeil PhD, J Ryan PhD,
Prof R Wolfe PhD);
Sam and Ann Barshop Institute,
UT Health San Antonio,
San Antonio, TX, USA
(Prof S E Espinoza MD);
Geriatrics Research, South
Texas Veterans Health Care
System, San Antonio, TX, USA
(Prof S E Espinoza); Department
of Pharmacy Practice and
Science, College of Pharmacy
(Prof M E Ernst PharmD), and
Department of Family
Medicine, Carver College of
Medicine (Prof M E Ernst),
University of Iowa, Iowa City,
IA, USA; Department of
Medicine, Geriatrics Division,
Hennepin HealthCare and
Berman Centre for Clinical
Research, Hennepin Healthcare
Research Institute,
Minneapolis, MN, USA
(Prof A M Murray MD); Menzies
Institute for Medical Research,
University of Tasmania,
Hobart, TAS, Australia
(Prof M R Nelson PhD); School
of Public Health, Curtin
University, Perth, WA, Australia
(Prof C M Reid PhD)
Correspondence to:
Prof Sophia Zoungas, School of
Public Health and Preventive
Medicine, Monash University,
Melbourne, VIC 3004, Australia
sophia.zoungas@monash.edu
98
Daily low-dose aspirin and incident type 2 diabetes in
community-dwelling healthy older adults: a post-hoc
analysis of efficacy and safety in the ASPREE randomised
placebo-controlled trial
Sophia Zoungas, Zhen Zhou, Alice J Owen, Andrea J Curtis, Sara E Espinoza, Michael E Ernst, Robyn L Woods, Suzanne G Orchard, John J McNeil,
Anne M Murray, Mark R Nelson, Christopher M Reid, Joanne Ryan, Rory Wolfe
Summary
Background Inflammation has been implicated in the pathogenesis of diabetes. This study investigated the randomised
treatment effect of low-dose aspirin on incident type 2 diabetes and fasting plasma glucose (FPG) concentrations
among older adults.
Methods ASPREE was a double-blind, placebo-controlled trial of daily oral low-dose aspirin. The study population
included community-dwelling individuals aged 70 years or older (≥65 years for US minority ethnic groups) in the USA
and Australia who were free of cardiovascular disease, independence-limiting physical disability, or dementia. For the
post-hoc analysis, we excluded participants with diabetes at baseline or with incomplete or missing incident diabetes
data during follow-up. Participants were randomly assigned 1:1 to oral 100 mg daily enteric-coated aspirin or placebo.
Incident diabetes was defined as self-reported diabetes, commencement of glucose-lowering medication, or a FPG
concentration of 7·0 mmol/L or more assessed at annual follow-up visits among participants with no diabetes at
baseline. We used Cox proportional hazards models and mixed-model repeated measures to assess the effect of
aspirin on incident diabetes and FPG concentrations in the intention-to-treat population. We assessed major bleeding
in participants who had taken at least one dose of study medication.
Findings Between March 10, 2010, and Dec 24, 2014, a total of 16 209 participants were included (8086 [49·9%]
randomly assigned to aspirin and 8123 [50·1%] randomly assigned to placebo). During a median follow-up of
4·7 years (IQR 3·6–5·7), 995 (in 6·1% individuals) incident cases of type 2 diabetes were recorded (459 in the aspirin
group and 536 in the placebo group). Compared with placebo, the aspirin group had a 15% reduction in risk of
incident diabetes (hazard ratio 0·85 [95% CI 0·75 to 0·97]; p=0·013) and a slower rate of increase in FPG concentration
at year 5 (between-group difference estimate –0·048 mmol/L [95% CI –0·079 to –0·018]; p=0·0017). Major bleeding
(major gastrointestinal bleeding, intracranial bleeding, and clinically significant bleeding at other sites) occurred in
510 (3·2%) of 16 104 participants (300 [3·7%] in the aspirin group and 210 [2·6%] in the placebo group). Compared
with placebo, the aspirin group had a 44% increase in risk of major bleeding (hazard ratio 1·44 [95% CI 1·21 to 1·72];
p<0·0001).
Interpretation Aspirin treatment reduced the incidence of type 2 diabetes and slowed the increase in FPG concentration
but increased major bleeding among community-dwelling older adults. Given the increasing prevalence of
type 2 diabetes among older adults, the potential for anti-inflammatory agents such as aspirin to prevent type 2 diabetes
or improve glucose levels warrants further study with a comprehensive assessment of all potential safety events of
interest.
Funding US National Institute on Aging, US National Cancer Institute, National Health and Medical Research Council
of Australia, Monash University, and the Victorian Cancer Agency
Copyright © 2023 Elsevier Ltd. All rights reserved.
Introduction in five people aged older than 70 years affected.
The world is undergoing a demographic transition Projections for the US population suggest that there has
towards an older population. Older adults (aged 65 years been a 4·5-fold increase in the number of cases of
and older) are at high risk of developing type 2 diabetes diabetes in people older than 75 years (compared with
and the glucometabolic derangements that precede it. a 2-fold increase in the total US population) between
The 2021 International Diabetes Federation report1 on 2005 and 2050.2
the global prevalence of diabetes estimated that The mechanisms that underpin type 2 diabetes in older
537 million people had diabetes in 2021, with almost one age can include loss of muscle mass and declines in
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

Research in context
Evidence before this study
Diabetes is a global public health problem and one of the major
causes of disability and mortality in populations of adults
65 years or older. Inflammation has been implicated in the
pathogenesis of diabetes, and medications with anti-
inflammatory properties, such as aspirin, might be beneficial for
preventing incident type 2 diabetes among older people. We
searched PubMed for randomised studies published between
database inception and Nov 1, 2022, that investigated the
potential effect of preventive aspirin on incident diabetes in any
age group. Two large randomised studies were identified: the
Physicians’ Health Study, which reported that low-dose aspirin
(325 mg every other day) followed by self-selected aspirin use
decreased the risk of incident diabetes by 14% in healthy men
(mean age 54 [SD 9] years) initially given aspirin for 5 years and
then followed for up to 22 years; and the Women’s Health Study,
which reported that low-dose aspirin (100 mg every other day)
did not decrease the risk of incident diabetes in women older
than 45 years given aspirin for 10 years. However, the
Physicians’ Health Study and the Women’s Health Study were
both conducted more than 20 years ago, used alternate daily
dosing of aspirin, and mostly included adults aged 50–60 years.
insulin sensitivity and secretion, and can differ from
those observed at younger ages including those younger
han 45 years and those aged 45–65 years.3,4 Experimental
and epidemiological data have suggested that subclinical
inflammation might contribute to metabolic diseases,
insulin resistance, and type 2 diabetes.5–7 For example,
increases in the concentrations of inflammatory markers,
such as C-reactive protein (CRP), interleukin-6, and
fibrinogen, have been associated with poor health
outcomes in older adults.8–10
A new diagnosis of type 2 diabetes might have
substantial implications for older adults, who often have
other medical conditions, functional decline, poly­
pharmacy, other treatment costs, and a high risk of
adverse effects from drugs. Anti-inflammatory and anti-
platelet agents, such as aspirin, have been proposed to
improve glucose handling and insulin resistance.11
However, results from randomised trials of aspirin for
diabetes prevention have not been consistent and
randomised trials exclusively in older adults are
absent.12–15 Given that small decreases in the risk of
type 2 diabetes can result in substantial benefits at a
global population level, simple, low-cost, and safe
preventive approaches are urgently needed.16,17
The Aspirin in Reducing Events in the Elderly
(ASPREE) trial was a primary prevention trial that
investigated whether a daily use of 100 mg of enteric-
coated aspirin would prolong the healthy life span of
older adults (older than 70 years).18–20 The trial was
conducted in Australia and the USA and recruited
19 114 older people from community settings. The
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Articles
Added value of this study
This study is the first analysis of a more contemporary
randomised trial assessing the effect of low-dose aspirin in
preventing type 2 diabetes in an older community-dwelling
population (>70 years for participants in non-minority
ethnic groups and ≥65 years for participants in
minority race and ethnic groups). We found that use of
100 mg enteric-coated aspirin daily reduced the risk of
incident diabetes by 15% and significantly slowed the rate of
increase in fasting blood glucose among participants without
diabetes between baseline and year 5. However, aspirin also
increased the risk of major bleeding (primarily
gastrointestinal bleeding).
Implications of all the available evidence
Our results are similar to those of the Physicians’ Health Study,
which showed a 14% reduction in type 2 diabetes risk with low-
dose aspirin use. Given the increasing prevalence of diabetes
around the world, the potential for anti-inflammatory agents,
such as aspirin, to prevent or delay incident type 2 diabetes or
improve glucose levels warrants further study, including
assessment of all potential safety events.
primary endpoint, survival free from dementia and
persistent physical disability, did not differ significantly
in the aspirin group compared with the placebo group
after a median of 4·7 years of follow-up. However, the
risks of major bleeding and death from cancer were
higher in the aspirin group than in the placebo group,
raising concerns about the safety of aspirin use as
a preventive therapy for healthy older people.
Using the comprehensive data collected in the ASPREE
trial, we sought to further investigate the randomised
treatment effects of aspirin on incident diabetes and
fasting plasma glucose (FPG) concentrations in a large,
community-based cohort of older adults. We hypothesised
that treatment of healthy older adults with 100 mg daily
of enteric-coated oral aspirin would not reduce incident
diabetes or slow the increase in FPG concentration over
time when compared with treatment with placebo.
Methods
Study design and participants
The study design, rationale, and principal findings of the
ASPREE trial have been previously detailed.18–21 In brief,
ASPREE was an international, prospective, double-blind,
randomised placebo-controlled trial examining whether
daily oral low-dose aspirin (enteric-coated 100mg) versus
a matching placebo would extend the trial’s primary
endpoint of survival that is free from dementia and
persistent physical disability in community-dwelling
older adults. In this study, we report a post-hoc analysis
of the ASPREE trial examining the effect of low-dose
aspirin on incident diabetes. The trial included
99
 Articles
19 114 participants aged 70 years or older (≥65 years
among participants in US minority race or ethnic groups)
with no previous cardiovascular events, dementia, or
independence-limiting physical disability. Participants
were recruited between March 10, 2010, and Dec 24, 2014,
in Australia (87% of participants) and the USA (13%).18–21
To be eligible for the ASPREE trial, participants did not
have any serious intercurrent illness that was likely to
cause death within the next 5 years, did not have
a condition known to be associated with a high risk of
major bleeding, had a score of 78 or more for the
Modified Mini-Mental State Examination test, and had
no major physical disability, defined as severe difficulty
in any one of the six basic activities of daily living
(bathing, dressing, toileting, transferring, walking, and
feeding).22 The trial was conducted in accordance with
the criteria of the International Conference on
Harmonisation for the conduct of clinical trials. The
institutional review board at each participating institution
approved the trial and all the participants provided
written informed consent. The study was registered with
ClinicalTrials.gov, NCT01038583.
For this post-hoc study, two additional exclusion criteria
were applied: participants with diabetes at baseline (self-
reported, use of glucose-lowering medication, or FPG
≥7·0 mmol/L) and participants with incomplete or
missing incident diabetes data during follow-up—ie,
those who did not self-report diabetes and did not have
FPG concentration and glucose-lowering medication
data collected at all annual follow-up visits.
Randomisation and masking
Participants who met the eligibility criteria at a screening
visit were enrolled in a 4-week placebo run-in phase for
compliance checking. Participants who took 80% or
more of the placebo pills during a 4-week run-in phase
were randomly assigned, in a 1:1 ratio, to receive daily
low-dose aspirin (enteric-coated aspirin 100 mg) or
matching placebo. Participants were randomly assigned
remotely by study staff via the ASPREE web portal
according to a computer-generated randomisation
schedule in a ratio of 1:1 to receive aspirin or matching
placebo. Randomisation was stratified for general
practice in Australia, for regional site in the USA, and
for age (65–69 years, 70–79 years, and ≥80 years).
Randomisation was blocked within strata, using variable-
sized blocks of sizes two, four, or six. ASPREE
participants, staff, and trial investigators were masked to
study group allocation during follow-up.
Procedures
Recruitment of ASPREE participants commenced on
March 10, 2010, and ended on Dec 24, 2014. The
interventional phase of the trial was ended on
June 12, 2017, by the US National Institute on Aging
(funder) after data reviewed by the data safety and
monitoring board showed similar rates of the primary
100
endpoint in the two groups that made it very unlikely
that continuation of the trial until its scheduled end date
of Dec 31, 2017, would show a significant treatment effect
for the primary endpoint. Participants were followed up
through annual in-person study visits with study staff at
their local clinic and a range of physical and cognitive
health measures were done, pathology samples collected,
and questionnaires on health behaviours, health events,
and medications were completed.
Outcomes
The primary outcome of this post-hoc analysis was
incident diabetes. The presence of diabetes was deter­
mined on the basis of self-report of diabetes in response
to a specific question about diabetes, use of glucose-
lowering medication, or FPG concentration on annual
clinical testing of 7·0 mmol/L or higher (American
Diabetes Association [ADA]23 and WHO criteria24).
Participants were asked to bring all currently used
prescription medications or a list of these to their
baseline and annual study visits. When this request was
not possible, medication use was self-reported and,
subsequently, confirmed via review of primary care
practice records, where possible. Blood samples for
measurement of FPG concentrations were collected
from participants after an overnight fast at baseline and
each annual follow-up visit in a clinic or local pathology
centre. Self-reported diabetes was captured at baseline
and follow-up visits. The timing of incident diabetes
was the date of the annual follow-up visit at which
a participant first had any one of self-reported diabetes,
commencement of glucose-lowering medication, or
FPG of 7 mmol/L or more.
The secondary outcome was the change in FPG
concentration over time. FPG concentrations were
recorded by collecting a single FPG concentration
annually from baseline to the end of follow-up or to the
initiation of any glucose-lowering medication, whichever
occurred first.
Major bleeding was the safety outcome (a prespecified
endpoint in the ASPREE trial) and was defined as the
composite of major gastrointestinal bleeding, intracranial
bleeding, or clinically significant bleeding at other sites
(defined as bleeding that led to transfusion, hospital­
isation, prolongation of hospitalisation, surgery, or
death). Non-serious safety events were not collected and
thus could not be reported for this analysis.
Statistical analysis
The sample size of the ASPREE trial was originally
determined for the primary trial endpoint of survival free
of both dementia and persistent physical disability. With
995 diabetes cases recorded during the follow-up in
16 209 participants (in the main analysis cohort), our
post-hoc study of incident diabetes had at least 80% power
to detect hazard ratios (HRs) less than 0·84 or greater
than 1·19.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
 Articles

All statistical analyses were conducted on an intention- secondary outcomes. All statistical tests were two-sided,
to-treat basis except for the safety analysis for major and we considered a p value of 0·05 or less to be
bleeding, which included all randomly assigned statistically significant. Analyses were conducted using
participants who had taken at least one dose of study Stata (SE 17·0).
medication, as assessed by pill counts. The number of
events and incidence rates (events per 1000 person-years) Role of the funding source
for diabetes in the aspirin and placebo groups were The funders of the ASPREE study had no role in study
calculated separately. We used Cox proportional hazards design, data collection, data analysis, data interpretation,
regression models with Efron’s method of tie handling to or writing of this report.
calculate cause-specific HRs with 95% CIs comparing
incident diabetes between aspirin and placebo groups in Results
the main analysis cohort. We assessed the proportional Between March 10, 2010 and Dec 24, 2014,
hazards assumption by Schoenfeld residual tests and 83 376 participants were screened by telephone with
found no violations. Cumulative incidence curves for 23 163 entering the run-in phase. A total of
incident diabetes were plotted, considering the 19 114 participants were randomly assigned, with
competing risk of death. We investigated effect 9525 assigned to receive aspirin and 9589 assigned to
modification by baseline characteristics, which were sex, receive placebo (figure 1). Of 19 114 participants,
median age (<74 and ≥74 years), country of residence, 2045 (10·7%) with diabetes at baseline were excluded.
race and ethnicity, previous regular aspirin use, BMI, Participants who had incomplete incident diabetes
statin use, smoking, hypertension, frailty,25 and
prediabetes status (ADA23 and WHO24 criteria), by adding
83 376 participants were screened by telephone
an interaction term between the randomisation group
and stratifying variable.
To investigate the treatment effect on the change in FPG 23 163 included in the run-in
concentration at each annual visit (year 1 to 5), we used
mixed-model repeated measures, including randomised 4049 excluded
treatment, the year of glucose assessment (0 [baseline] to 5), 2453 ineligible
and year-by-treatment interaction. Restricted maximum 1518 unwilling to continue
78 had other or unknown reasons
likelihood estimation with a Kenward-Roger correction
and an unstructured variance-covariance matrix generated
parameter estimates with 95% CIs. We also estimated the 19 114 were randomly assigned
least-squares means and corresponding 95% CIs at each
timepoint for within-group change in FPG concentration
and differences between groups. 9525 assigned to receive aspirin 9589 assigned to receive placebo
Four sensitivity analyses were done for the primary
outcome of incident diabetes. First, we considered more
liberal criteria for diabetes data completeness than the 1439 excluded 1466 excluded
1024 had diabetes at 1021 had diabetes at
main analysis. Of 860 participants who were originally baseline* baseline*
excluded due to incomplete data on self-reported diabetes, 415 did not self-report 445 did not self-report
diabetes and had no diabetes and had no
commencement of glucose-lowering medication, or medication and FPG medication and FPG
having an FPG of 7 mmol/L or more during follow- data collected at all data collected at all
up, 799 had at least self-reported not having diabetes at all annual visits annual visits

visits and were included for a sensitivity analysis. Second,

a more objective definition of incident diabetes than in 8086 included in primary analysis 8123 included in primary analysis
the main analysis was assessed by excluding self-reported 2161 did not have FPG measured at 2157 did not have their FPG measured at
their last visit their last visit
incident diabetes from the main analysis. Third, the 367 died 321 died
intermittent nature of the ascertainment of incident 37 withdrew from the trial 31 withdrew from the trial
diabetes was accounted for in analyses using discrete
time proportional hazards regression models instead
46 excluded due to not 59 excluded due to not
of Cox models. Fourth, we used the Fine-Gray sub- taking any study taking any study
distribution hazard model to further calculate the medication medication
sub-distribution HR (sHR) for incident diabetes between
aspirin and placebo groups, where all-cause death was 8040 included in safety analysis 8064 included in safety analysis
considered a potential competing event.26
No imputation of missing data or adjustment for Figure 1: Trial profile
multiple comparisons of the primary outcome were FPG=fasting plasma glucose. *Presence of diabetes at baseline was defined as self-report, being on
considered. There is no allowance for multiplicity for the glucose-lowering medication, or having an FPG concentration of 7 mmol/L or more (≥126 mg/dL).

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 101

 Articles

data at any annual visits (415 in the aspirin group and

Aspirin (n=8086) Placebo (n=8123)
445 in the placebo group) were also omitted, leaving
Median age at baseline (IQR), years 74·0 (71·7–77·7) 73·9 (71·6–77·5) 16 209 participants included in the main analysis cohort.
Sex, n (%) Of these, 8086 were in the aspirin group and 8123 in
Female 4622 (57·2%) 4634 (57·0%) the placebo group, and there was a median of
Male 3464 (42·8%) 3489 (43·0%) 4·7 (IQR 3·6–5·7) years of follow-up. The participants
Race or ethnicity, n (%) in each treatment group were evenly matched for mean
White 7568 (93·6%) 7611 (93·7%) FPG concentrations, age, and other cardiovascular
Black 277 (3·4%) 256 (3·2%) disease risk factors at baseline (table 1). The median age
Hispanic or Latino 146 (1·8%) 144 (1·8%) was 73·9 (IQR 71·7–77·6) years, mean weight was
Other 95 (1·2%) 112 (1·4%) 76·1 (SD 14·5; IQR 65·8–85·1) kg, mean BMI was
Country, n (%) 27·8 (4·6; 24·7–30·3) kg/m², and mean FPG
Australia 7264 (89·8%) 7330 (90·2%) concentration of the trial population was 5·3 (0·6;
USA 822 (10·2%) 793 (9·8%) 4·9–5·6) mmol/L. Hypertension was present in
Education (>12 years of education), n (%) 4414 (54·6%) 4434 (54·6%) 11 796 (72·8%) of 16 209 participants and dyslipidaemia
Mean weight (SD), kg 76·0 (14·5) 76·3 (14·5) was present in 10 496 (64·8%) of 16 209 participants at
Mean BMI (SD), kg/m² 27·8 (4·5) 27·8 (4·6) baseline. Previous regular aspirin use was self-reported
Mean abdominal circumference (SD), cm by 1582 (9·8%) of 16 209 participants at baseline. The
Male 101·2 (10·4) 101·4 (10·4) baseline characteristics of participants with missing
Female 92·4 (12·5) 92·6 (12·7) data (860 [4·5%] of 19 114) were similar to those included
Hypertension, n (%) 5866 (72·5%) 5930 (73·0%) in the main analysis cohort, except for a greater
Dyslipidaemia, n (%) 5192 (64·2%) 5304 (65·3%) proportion of participants residing in the USA in the
Chronic kidney disease, n (%) 1895 (23·4%) 1877 (23·1%) missing data cohort (appendix p 1).
Smoking status, n (%) At the final in-trial follow-up visit, concomitant
Non-smoker 4549 (56·3%) 4564 (56·2%) medication use, including anti-hypertensive drug classes,
Previous smoker 3259 (40·3%) 3282 (40·4%)
statins, other lipid-lowering medications, and non-
Current smoker 278 (3·4%) 277 (3·4%)
steroidal anti-inflammatory drugs remained well
Frailty, n (%)
balanced between the randomised aspirin and placebo
groups. Compared with baseline, the use of all
None 4902 (60·6%) 4967 (61·1%)
medications had increased modestly at the final visit,
Pre-frail 3029 (37·5%) 3015 (37·1%)
more so in participants who developed diabetes than in
Frail 155 (1·9%) 141 (1·7%)
those who did not (appendix p 2).
Previous regular aspirin use 800 (9·9%) 782 (9·6%)
During the entire follow-up, 11 787 (72·7%) of
Current use of medications, n (%)
16 209 participants had an average adherence rate of
Angiotensin-converting-enzyme inhibitor 1228 (15·2%) 1240 (15·3%)
50% or more and 8368 (51·6%) of 16 209 had an average
Angiotensin II receptor blocker 1944 (24·0%) 2003 (24·7%)
adherence rate of 85% or more by pill count in both
Diuretics 1435 (17·7%) 1394 (17·2%)
treatment groups (appendix p 3).
β-blocker 610 (7·5%) 584 (7·2%)
During the follow-up, 995 (6·1% of the study
Calcium channel blocker 1321 (16·3%) 1256 (15·5%)
population) incident diabetes events were reported,
Statin 2290 (28·3%) 2278 (28·0%)
including 459 in the aspirin group (12·7 [11·5–13·9] cases
Other lipid-lowering agents 339 (4·2%) 314 (3·9%) per 1000 person-years) and 536 in the placebo group
Non-steroidal anti-inflammatory drugs 1263 (15·6) 1254 (15·4) (14·8 [13·6–16·1] cases per 1000 person-years;
Mean fasting plasma glucose (SD), mmol/L 5·27 (0·57) 5·28 (0·58) table 2). The types of glucose-lowering medications used
In the race or ethnicity row, the other category includes Aboriginal or Torres Strait Islander, Native American, mixed by participants during the follow-up consisted mostly of
race, Native Hawaiian or Pacific Islander, and those who were not Hispanic and who did not state their ethnicity or race. oral agents, particularly metformin, DPP-4 inhibitors,
Hypertension was defined as being on treatment for high blood pressure or having a blood pressure of more than
140/90 mm Hg at study entry. Dyslipidaemia was defined as taking cholesterol-lowering medications or having a
and sulfonylureas (appendix p 4).
serum cholesterol of 212 mg/dL or more (≥5 mmol/L in Australia) and 240 mg/dL or more (≥6·2 mmol/L in the USA) or Compared with placebo, assignment to aspirin resulted
low-density lipoprotein cholesterol of more than 160 mg/dL (>4·1 mmol/L). Chronic kidney disease was defined as in a 15% reduction in risk of incident diabetes (HR 0·85
estimated glomerular filtration rate less than 60 mL/min per 1·73m² or urinary albumin-to-creatinine ratio
of 3 mg/mmol or more. Previous regular aspirin use was self-reported regular use of aspirin immediately before first
[95% CI 0·75–0·97]; table 2). The difference in the rate of
baseline visit with a 4-week run-in before random assignment to study medication. Pre-frail included anyone with one diabetes was –2·1 per 1000 person-years in the total study
or two criteria and frail included anyone with three or more criteria of the adapted Fried frailty criteria,25 including population, and –3·9 per 1000 person-years for male
bodyweight, strength, exhaustion, walking speed, and physical activity. participants and –0·8 per 1000 person-years for female
Table 1: Baseline characteristics of participants in the full analysis set randomly assigned to aspirin or participants. A divergence in case numbers beginning
placebo after the second year of aspirin or placebo treatment was
shown in curves that plotted the cumulative incidence of
diabetes (figure 2). The risk reduction was generally
consistent in all pre-specified subgroups considered,

102 www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

 Articles

including sex, age, country of residence, race and

Aspirin (n=8086) Placebo (n=8123) Hazard ratio p value
ethnicity, previous regular aspirin use, BMI, statin use, (95% CI)
smoking, hypertension, frailty, and prediabetes status at
Total 12·7 (11·5–13·9); 459 14·8 (13·6–16·1); 536 0·85 (0·75–0·97) 0·013
baseline (all p values for interaction >0·10; table 2).
Sex
Mean FPG concentration was 5·3 (SD 0·6) mmol/L at
Female 11·2 (9·9–12·7); 236 12·0 (10·6–13·6); 251 0·93 (0·78–1·12) 0·17
baseline and increased over time in both assigned
Male 14·6 (12·8–16·7); 223 18·5 (16·5–20·8); 285 0·78 (0·66–0·94) ..
treatment and placebo groups. The mean FPG
Age
concentrations increased from baseline to year 5 in both
<74 years 12·5 (11·0–14·2); 225 14·7 (13·0–16·6); 266 0·85 (0·71–1·01) 0·93
treatment groups, with least squares mean change of
≥74 years 12·8 (11·3–14·6); 234 14·8 (13·2–16·7); 270 0·86 (0·72–1·02) ..
0·061 (95% CI 0·040 to 0·083) mmol/L in the aspirin
group and of 0·109 (0·088 to 0·131) mmol/L in the Country

placebo group. In addition, at year 5, the difference in Australia 11·7 (10·6–13·0); 378 14·3 (13·1–15·7); 464 0·82 (0·71–0·94) 0·11

the least squares mean change between aspirin and USA 19·8 (15·9–24·6); 81 18·2 (14·5–22·9); 72 1·10 (0·80–1·51) ..
placebo groups was –0·048 mmol/L (–0·079 to –0·018; Race
p=0·0017; appendix p 5). The mean FPG concentration White 11·8 (10·7–13·0); 400 14·2 (13·0–15·5); 485 0·83 (0·72–0·94) 0·42
increased less in the aspirin group than in the placebo Black 30·7 (22·2–42·3); 37 24·0 (16·3–35·2); 26 1·26 (0·76–2·08) ..
group at all annual follow-up visits (figure 3). Hispanic or Latino 25·2 (15·4–41·1); 16 26·0 (15·9–42·4); 16 0·96 (0·48–1·91) ..
The safety analysis set included 16 104 participants, Other 14·2 (6·4–31·5); 6 18·9 (9·8–36·4); 9 0·76 (0·27–2·13) ..
with 8040 in the aspirin group and 8064 in the Previous regular aspirin use
placebo group. Major bleeding occurred in 300 (3·7%) No 12·5 (11·3–13·8); 404 14·3 (13·1–15·7); 465 0·87 (0·76–0·99) 0·45
par­ticipants in the aspirin group and 210 (2·6%) part­ Yes 13·9 (10·7–18·1); 55 18·5 (14·7–23·3); 71 0·75 (0·53–1·07) ..
icipants in the placebo group, with incidence rates of BMI
8·2 events per 1000 person-years in the aspirin group <25 kg/m² 6·3 (5·0–8·1); 66 8·9 (7·3–11·0); 90 0·71 (0·52–0·98) 0·11
and 5·7 events per 1000 person-years in the placebo 25–29.9 kg/m² 10·9 (9·4–12·6); 178 13·7 (12·0–15·6); 226 0·79 (0·65–0·96) ..
group. Compared with placebo, aspirin treatment ≥30 kg/m² 22·9 (20·0–26·2); 215 22·6 (19·7–25·8); 216 1·01 (0·84–1·22) ..
increased the risk of major bleeding by 44% (HR 1·44 Statin use
[95% CI 1·21–1·72]; p<0·0001) including major No 11·2 (10·0–12·5); 292 13·2 (11·9–14·7); 347 0·84 (0·72–0·98) 0·72
gastrointestinal bleeding, intracranial bleeding, and Yes 16·5 (14·1–19·1); 167 18·7 (16·2–21·6); 189 0·88 (0·72–1·09) ..
other clinically significant bleeding (table 3). Smoking
When analyses were repeated for incident diabetes to Never 11·7 (10·3–13·2); 240 12·7 (11·2–14·3); 260 0·91 (0·77–1·09) 0·57
additionally include participants who did not have Past 13·6 (11·8–15·6); 197 16·9 (14·9–19·2); 248 0·80 (0·66–0·97) ..
incident diabetes based on self-report alone (n=799) or Current 18·5 (12·2–28·2); 22 23·9 (16·5–34·6); 28 0·78 (0·44–1·36) ..
exclude participants with incident diabetes based on self- Hypertension
report alone (n=161), the results were similar to those in No 8·3 (6·7–10·3); 83 10·4 (8·6–12·6); 103 0·79 (0·59–1·06) 0·55
the full analysis set (HRs 0·86 [95% CI 0·76–0·97] for Yes 14·3 (12·9–15·8); 376 16·4 (14·9–18·0); 433 0·87 (0·76–1·00) ..
aspirin vs 0·85 [95% CI 0·75–0·97] for placebo). When Frailty
a discrete time proportional hazards regression model None 10·4 (9·2–11·9); 229 13·2 (11·7–14·7); 292 0·79 (0·66–0·94) 0·28
was used instead of a Cox model, the results were also Pre-frail 16·0 (14·1–18·3); 219 17·0 (14·9–19·3); 229 0·95 (0·79–1·14) ..
similar (0·86 [0·76–0·97]). Finally, in analyses taking
Frail 16·2 (9·0–29·3); 11 25·1 (15·1–41·6); 15 0·65 (0·30–1·42) ..
account of all cause death as a competing risk, the results
Prediabetes (American Diabetes Association criteria: FPG ≥5·6 mmol/L)
were unchanged (sHR 0·85 [0·75–0·97]).
No 5·0 (4·2–5·9); 139 6·7 (5·8–7·8); 188 0·74 (0·59–0·92) 0·12
Yes 39·1 (35·0–43·7); 310 42·9 (38·6–47·7); 339 0·91 (0·78–1·06) ..
Discussion
Prediabetes (WHO criteria: FPG ≥6·1 mmol/L)
In this analysis of data from a large-scale, two-country,
No 8·2 (7·3–9·3); 275 9·4 (8·4–10·5); 314 0·87 (0·74–1·03) 0·69
randomised trial of community-dwelling healthy older
Yes 77·1 (66·5–89·5); 174 92·5 (80·9–105·8); 213 0·84 (0·68–1·02) ..
adults without previous cardiovascular events, dementia,
or independence-limiting physical disability, the use Data are in rate per 1000 person-years; n, unless stated. Each variable, except total, shows a p value for interaction.
FPG=fasting plasma glucose.
of low-dose aspirin (100 mg per day for a median of
4·7 years) reduced the risk of incident diabetes and Table 2: Hazard ratios of incident diabetes for aspirin versus placebo groups in the main analysis set and
slowed the age-associated increase in FPG concentrations. by subgroups defined by baseline characteristics
The effects did not significantly differ by sex, age, BMI,
country of recruitment, statin use, or previous regular
use of aspirin; however, the largest absolute effects were Given the proposed role of chronic subclinical See Online for appendix
observed among male participants. The effects were also inflammation in the development of insulin resistance
robust in sensitivity analyses considering minor changes or deficiency, which can increase an individual’s
to the definition of incident diabetes and the competing susceptibility to glucometabolic disorders, it is important
risk of death. to further understand any effect of aspirin on incident

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 103

 Articles

100 Placebo given aspirin over 10 years (rate ratio 1·01 [95% CI
Aspirin 0·91–1·11]). Of note, these trials12,15 were conducted more
10
Cumulative incidence of diabetes (%)
than 20 years ago and in younger participants, whose use
of background preventive therapies, such as statins, and
8 HR 0·85 (95% CI 0·75–0·97);
log rank p=0·013
absolute risk would have substantially differed from that
of our trial population. It is also of interest that we
6 observed greater absolute treatment effects in male
than in female participants, suggesting possible sex
4 differences in responses. Indeed, previous studies have
proposed that variation in the bioavailability of low-dose
2 aspirin (which can differ by gender, formulation, dosing
frequency and timing, or bodyweight) might explain
0 differences in efficacy, particularly when used for the
0 1 2 3 4 5 6 prevention of vascular disease.27–29
Time since randomisation (years)
Number at risk
Preclinical and clinical studies have provided a good
Placebo 8123 8013 7803 6805 5138 3154 1102 rationale for targeting inflammation to improve the action
Aspirin 8086 7996 7804 6824 5151 3171 1112 of insulin and improve blood glucose levels.11 Moreover,
Figure 2: Cumulative incidence of diabetes in participants randomised to studies of salicylate treatment have reported anti-
aspirin or placebo inflammatory effects at both low and high doses,
The plots were constructed from the cumulative incidence of diabetes predicted albeit through different mechanisms of action. In an
using separate models in participants assigned to aspirin (red solid line) and
experimental model of cantharidin-induced acute inflam­
participants assigned to placebo (blue solid line), taking into account the
competing risk of death. Data are not shown after year 6 because only a small mation, low-dose aspirin (75 mg) administered for 10 days
number of participants reached year 7. HR=hazard ratio. to healthy men inhibited innate immune-mediated
responses by reducing total leukocyte as well as neutrophil
5·40 Placebo
and macrophage accumulation in skin blisters. These
Aspirin effects were dependent on 15-epi-lipoxin A4 synthesis and
Mean fasting plasma glucose (mmol/L)

signalling, which triggered anti-adhesive nitric oxide

release, a crucial determinant of extravascular leukocyte
5·35 accumulation and inhibition of prostacyclin production
through the cyclooxygenase pathway.30 In contrast, high-
dose salicylates, such as salsalate (the non-acetylated
dimer of salicylic acid), have been shown to be potent
5·30
inhibitors of IκB kinase and the NF-κB cascade.31 In
studies of high-risk populations with prediabetes or
populations with diabetes, high-dose salsalate improved
5·25 insulin sensitivity, as well as increased insulin release and
0 1 2 3 4 5 peripheral glucose disposal.32–35 A 4-week randomised,
Time since randomisation (years) placebo-controlled study of 4 g/day of salsalate in people
Figure 3: The effect of aspirin on change in mean fasting plasma glucose over with obesity reported a 14% reduction in fasting glucose as
time well as reductions in C-peptide levels.35 Similarly, a
The changes in raw mean fasting glucose with aspirin assignment (red solid line) 12-week randomised, placebo-controlled study of up to
versus placebo assignment (blue solid line). Error bars indicate 95% CIs. Fasting
4 g/day of salsalate in people with impaired fasting glucose
glucose was set as missing at annual visits at which glucose-lowering
medication use was reported. or impaired glucose tolerance reported a 6% reduction in
fasting glucose and reduced adipose tissue NF-κB activity.
diabetes using a contemporary cohort of older adults Mean salicylate concentrations correlated inversely with
who can be at greater risk due to age alone. Our finding the change in fasting glucose concentrations, suggesting
of a 15% reduction in risk of incident diabetes was very that reductions in glucose were related to an individuals’
similar to that of the Physicians’ Health Study,12 which salicylic acid levels.36 Future work using the biobanked
reported that aspirin (randomised treatment and then samples of the ASPREE cohort might help to explain the
self-use) reduced the risk of incident diabetes by putative mechanisms for our findings.
14% (HR 0·86 [95% CI 0·77–0·97]) in healthy men aged Although our findings are from a post-hoc analysis of
40–84 years (mean age 54 years) followed up for up to a trial designed to address primary and secondary
22 years. In contrast, the Women’s Health Study15 outcomes that did not include diabetes, it is worth noting
reported that 100 mg aspirin every other day did not that our analyses of changes in FPG over time were
modify the risk of incident diabetes in women older than consistent with our analysis of incident diabetes. As has
45 years (61% in women aged 45–54 years, 29% in women been reported by other observational studies of ageing
aged 55–64 years, and 10% in women aged ≥65 years) populations, the FPG concentration trajectory increased

104 www.thelancet.com/diabetes-endocrinology Vol 12 February 2024


modestly over time in both treatment groups, but the
increase was lower with aspirin treatment. Although
population studies suggest FPG concentrations increase
with age,37 the rate of increase appears to slow in people
older than 60 years, with analysis of US National Health
and Nutrition Examination Survey data suggesting an
increase in the order of 0·0056 mmol/L (0·1 mg/dL) per
year of age between age 60 and 70 years.38 Additional
studies characterising the phenotypic and genetic risk for
the development of diabetes in older populations will
be important to clarify any future role of targeting
inflammation and to balance the risks and benefits of
aspirin use.
We also report an increased risk of major bleeding,
which is an important consideration for older
community-dwelling individuals. The magnitude of this
increased risk was commensurate with that reported by
other low-dose aspirin trials, including the ASCEND trial
in patients with type 2 diabetes, but higher than that
reported by the ARRIVE trial in people with a low-to-
moderate cardiovascular risk.39,40 Of particular importance
for future studies will be the comprehensive assessment
and mitigation of bleeding risk and any other safety
issues associated with use of anti-inflammatory
agents.18,19,41,42 Indeed, novel formulations of aspirin that
could enhance the beneficial anti-platelet and anti-
inflammatory effects but reduce gastrointestinal toxicity
are currently being evaluated.43,44
One limitation of this post-hoc analysis was the
definition of diabetes, which was based on self-reporting,
commencement of glucose-lowering medication, or
a single FPG measurement and did not include a second
confirmatory blood test within 2 weeks. In addition,
HbA1c concentrations and oral glucose tolerance testing,
which were not routinely tested in all participants, could
not be included in the definition. However, the repeated
assessment of FPG concentrations over time, with the
results being communicated directly to the general
practitioners providing care to the study participants
gives some assurance that cases were not missed or
misclassified.
Our study also had several strengths. All ASPREE
participants had no previous cardiovascular events,
dementia, or independence-limiting physical disability at
trial entry, and were selected as relatively healthy
ambulatory community-dwelling participants. Therefore,
our data do not allow conclusions as to whether aspirin
prevents diabetes among other populations or older
adults in high-risk racial or ethnic groups, including
those likely to be most vulnerable to the risks of incident
diabetes and its consequences. Nonetheless, the large
sample size of older participants, the long duration of
follow-up to 4·7 years with detailed annual data capture,
the large number of incident diabetes cases, and high
adherence to treatment (more than 70% of participants
were taking ≥50% of the study pills on average in both
groups during the entire follow-up) gives us confidence
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Articles
Aspirin (n=8040) Placebo (n=8064) Hazard ratio p value
(95% CI)
Total major bleeding 8·2 (7·3–9·2); 300 5·7 (5·0–6·5); 210 1·44 (1·21–1·72) <0·0001
Subgroups
Gastrointestinal bleeding 3·6 (3·1–4·3); 133 2·1 (1·7–2·6); 77 .. ..
Intracranial bleeding 2·2 (1·8–2·7); 81 1·5 (1·1–1·9); 54 .. ..
Other bleeding sites 2·4 (1·9–2·9); 86 2·1 (1·7–2·7); 79 .. ..
Data are in rate per 1000 person-years; n, unless stated.
Table 3: Major bleeding events in aspirin versus placebo group in the safety population
that the effects were robust.
Subgroup analyses might have modest power to detect
differences in treatment effects, such as the differences
by sex, country of residence (Australia or USA), and BMI
category. The qualitative differences in point estimates
observed need to be further studied to understand where
the true effects lie and whether these effects are enduring
with longer follow-up of the trial cohort.
In this post-hoc analysis, aspirin treatment appeared to
reduce incident diabetes and slow the increase in FPG
concentration but also increased major bleeding in
initially healthy adults aged 65 years or older when
compared with placebo. Given the increasing prevalence
of diabetes around the world, the potential for anti-
inflammatory agents such as aspirin to prevent or delay
incident diabetes or improve glucose levels warrants
further study.
Contributors
All authors contributed to the study concept and design. SZ, ZZ, AJO,
AJC, RW, and RLW contributed to the acquisition, analysis, or
interpretation of data. ZZ and RW contributed to statistical analysis.
SZ drafted the manuscript. All authors contributed to the critical
revision of the manuscript for important intellectual content. RW, RLW,
JJM, and AMM contributed to administrative, technical, or material
support. RW and ZZ verified the data. All authors had full access to all
the data in the study and had final responsibility for the decision to
submit for publication.
Declaration of interests
MRN reported receiving a meeting honorarium from Bayer and trial
product in ASPREE provided by Bayer and Australian National Health
and Medical Research Council grant support for STAREE. CMR reported
being funded through an Australian National Health and Medical
Research Council Principal Research Fellowship. SZ has received
payment to their institution from Eli Lilly Australia, Boehringer-
Ingelheim, Merck Sharp & Dohme Australia, AstraZeneca, Novo
Nordisk, Sanofi, and Servier for consultancy work outside the submitted
work. All other authors declare no competing interests.
Data sharing
Requests for data access can be made via the ASPREE principal
investigators, with details for applications provided at https://aspree.org/
aus/for-researchers/ or https://aspree.org/usa/for-researchers/.
Acknowledgments
The Aspirin in Reducing Events in the Elderly (ASPREE) trial was
supported by the US National Institute on Aging and the National
Cancer Institute at the National Institutes of Health (U01AG029824),
the National Health and Medical Research Council of Australia (334047
and 1127060), and Monash University and the Victorian Cancer Agency.
We thank the ASPREE participants who gave their precious time to
participate in the ASPREE trial; registered general practitioners;
endorsing organisations; and all members of the ASPREE team.
105
 Articles
References
1 International Diabetes Federation. Diabetes atlas. 2021. https://
diabetesatlas.org/idfawp/resource-files/2021/07/IDF_Atlas_10th_
Edition_2021.pdf (accessed Nov 8, 2021).
2 Narayan KM, Boyle JP, Geiss LS, Saaddine JB, Thompson TJ.
Impact of recent increase in incidence on future diabetes burden:
US, 2005–2050. Diabetes Care 2006; 29: 2114–16.
3 Meneilly GS, Tessier D. Diabetes in elderly adults.
J Gerontol A Biol Sci Med Sci 2001; 56: M5–13.
4 Lee PG, Halter JB. The pathophysiology of hyperglycemia in older
adults: clinical considerations. Diabetes Care 2017; 40: 444–52.
5 Greenfield JR, Campbell LV. Relationship between inflammation,
insulin resistance and type 2 diabetes: ‘cause or effect’?.
Curr Diabetes Rev 2006; 2: 195–211.
6 Temelkova-Kurktschiev T, Siegert G, Bergmann S, et al. Subclinical
inflammation is strongly related to insulin resistance but not to
impaired insulin secretion in a high risk population for diabetes.
Metabolism 2002; 51: 743–49
7 Oguntibeju OO. Type 2 diabetes mellitus, oxidative stress and
inflammation: examining the links.
Int J Physiol Pathophysiol Pharmacol 2019; 11: 45–63.
8 Harris TB, Ferrucci L, Tracy RP, et al. Associations of elevated
interleukin-6 and C-reactive protein levels with mortality in the
elderly. Am J Med 1999; 106: 506–12.
9 Danesh J, Lewington S, Thompson SG, et al. Plasma fibrinogen
level and the risk of major cardiovascular diseases and nonvascular
mortality: an individual participant meta-analysis. JAMA 2005;
294: 1799–809.
10 Kaptoge S, Di Angelantonio E, Lowe G, et al. C-reactive protein
concentration and risk of coronary heart disease, stroke, and
mortality: an individual participant meta-analysis. Lancet 2010;
375: 132–40.
11 Rumore MM, Kim KS. Potential role of salicylates in type 2 diabetes.
Ann Pharmacother 2010; 44: 1207–21.
12 Hayashino Y, Hennekens CH, Kurth T. Aspirin use and risk of
type 2 diabetes in apparently healthy men. Am J Med 2009;
122: 374–79.
13 Kim SH, Liu A, Ariel D, et al. Effect of salsalate on insulin action,
secretion, and clearance in nondiabetic, insulin-resistant
individuals: a randomized, placebo-controlled study. Diabetes Care
2014; 37: 1944–50.
14 Bratusch-Marrain PR, Vierhapper H, Komjati M, Waldhäusl WK.
Acetyl-salicylic acid impairs insulin-mediated glucose utilization
and reduces insulin clearance in healthy and non-insulin-
dependent diabetic man. Diabetologia 1985; 28: 671–76.
15 Pradhan AD, Cook NR, Manson JE, Ridker PM, Buring JE.
A randomized trial of low-dose aspirin in the prevention of clinical
type 2 diabetes in women. Diabetes Care 2009; 32: 3–8.
16 Shi BY. The importance and strategy of diabetes prevention.
Chronic Dis Transl Med 2016; 2: 204–07.
17 Carris NW, Magness RR, Labovitz AJ. Prevention of diabetes mellitus
in patients with prediabetes. Am J Cardiol 2019; 123: 507–12.
18 McNeil JJ, Nelson MR, Woods RL, et al. Effect of aspirin on all-cause
mortality in the healthy elderly. N Engl J Med 2018; 379: 1519–28.
19 McNeil JJ, Wolfe R, Woods RL, et al. Effect of aspirin on
cardiovascular events and bleeding in the healthy elderly.
N Engl J Med 2018; 379: 1509–18.
20 McNeil JJ, Woods RL, Nelson MR, et al. Effect of aspirin on
disability-free survival in the healthy elderly. N Engl J Med 2018;
379: 1499–508.
21 ASPREE Investigator Group. Study design of Aspirin in Reducing
Events in the Elderly (ASPREE): a randomized, controlled trial.
Contemp Clin Trials 2013; 36: 555–64.
22 Ryan J, Woods RL, Britt C, et al. Normative performance of healthy
older individuals on the Modified Mini-Mental State (3MS)
examination according to ethno-racial group, gender, age, and
education level. Clin Neuropsychol 2019; 33: 779–97.
23 American Diabetes Association. Classification and diagnosis of
diabetes: standards of medical care in diabetes—2021. Diabetes Care
2021; 44 (suppl 1): S15–33.
106
24 WHO. Classification of diabetes mellitus; 2019. https://apps.who.
int/iris/rest/bitstreams/1233344/retrieve (accessed June 29, 2023).
25 Espinoza SE, Woods RL, Ekram ARMS, et al. The effect of low-dose
aspirin on frailty phenotype and frailty index in community-
dwelling older adults in the Aspirin in Reducing Events in the
Elderly study. J Gerontol A Biol Sci Med Sci 2022; 77: 2007–14.
26 Fine JP, Gray RJ. A proportional hazards model for the
subdistribution of a competing risk. J Am Stat Assoc 1999;
94: 496–509.
27 Peace A, McCall M, Tedesco T, et al. The role of weight and enteric
coating on aspirin response in cardiovascular patients.
J Thromb Haemost 2010; 8: 2323–25.
28 Rocca B, Fox KAA, Ajjan RA, et al. Antithrombotic therapy and
body mass: an expert position paper of the ESC Working Group
on Thrombosis. Eur Heart J 2018; 39: 1672–1686f.
29 Rothwell PM, Cook NR, Gaziano JM, et al. Effects of aspirin on
risks of vascular events and cancer according to bodyweight and
dose: analysis of individual patient data from randomised trials.
Lancet 2018; 392: 387–99.
30 Morris T, Stables M, Hobbs A, et al. Effects of low-dose aspirin on
acute inflammatory responses in humans. J Immunol 2009;
183: 2089–96.
31 Shoelson SE, Lee J, Yuan M. Inflammation and the IKK beta/I
kappa B/NF-kappa B axis in obesity- and diet-induced insulin
resistance. Int J Obes (Lond) 2003; 27 (suppl 3): S49–52.
32 Manrique C, Lastra G, Palmer J, Gardner M, Sowers JR. Aspirin
and diabetes mellitus: revisiting an old player.
Ther Adv Cardiovasc Dis 2008; 2: 37–42.
33 Lastra G, Whaley-Connell A. Diabetes: aspirin and prevention of
diabetes still a topic of debate. Nat Rev Endocrinol 2009; 5: 365–66.
34 Faghihimani E, Aminorroaya A, Rezvanian H, Adibi P,
Ismail-Beigi F, Amini M. Reduction of insulin resistance and
plasma glucose level by salsalate treatment in persons with
prediabetes. Endocr Pract 2012; 18: 826–33.
35 Fleischman A, Shoelson SE, Bernier R, Goldfine AB. Salsalate
improves glycemia and inflammatory parameters in obese young
adults. Diabetes Care 2008; 31: 289–94.
36 Goldfine AB, Conlin PR, Halperin F, et al. A randomised trial of
salsalate for insulin resistance and cardiovascular risk factors in
persons with abnormal glucose tolerance. Diabetologia 2013;
56: 714–23.
37 Chia CW, Egan JM, Ferrucci L. Age-related changes in glucose
metabolism, hyperglycemia, and cardiovascular risk. Circ Res 2018;
123: 886–904.
38 Pani LN, Korenda L, Meigs JB, et al. Effect of aging on A1C levels in
individuals without diabetes: evidence from the Framingham
Offspring Study and the National Health and Nutrition
Examination Survey 2001–2004. Diabetes Care 2008; 31: 1991–96.
39 Bowman L, Mafham M, Wallendszus K, et al. Effects of aspirin for
primary prevention in persons with diabetes mellitus. N Engl J Med
2018; 379: 1529–39.
40 Gaziano JM, Brotons C, Coppolecchia R, et al. Use of aspirin to
reduce risk of initial vascular events in patients at moderate risk of
cardiovascular disease (ARRIVE): a randomised, double-blind,
placebo-controlled trial. Lancet 2018; 392: 1036–46.
41 Mahady SE, Margolis KL, Chan A, et al. Major GI bleeding in older
persons using aspirin: incidence and risk factors in the ASPREE
randomised controlled trial. Gut 2021; 70: 717–24.
42 Cloud GC, Williamson JD, Thao LTP, et al. Low-dose aspirin and
the risk of stroke and intracerebral bleeding in healthy older people:
secondary analysis of a randomized clinical trial. JAMA Netw Open
2023; 6: e2325803.
43 Cryer B, Bhatt DL, Lanza FL, Dong JF, Lichtenberger LM,
Marathi UK. Low-dose aspirin-induced ulceration is attenuated by
aspirin-phosphatidylcholine: a randomized clinical trial.
Am J Gastroenterol 2011; 106: 272–77.
44 Angiolillo DJ, Bhatt DL, Lanza F, et al. Bioavailability of aspirin in
fasted and fed states of a novel pharmaceutical lipid aspirin
complex formulation. J Thromb Thrombolysis 2020; 49: 337–43.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

aspirin and placebo groups in the rate of incident
diabetes (–2·1 per 1000 person-years) was offset by the
absolute difference in the rate of major bleeding events
(2·5 per 1000 person-years). As such, when considering
the increased rate of all-cause death observed in the
primary trial, these findings do not provide sufficient
rationale for a change in routine practice. However, this
effort amplifies the need for vigilant diabetes screening
and treatment across the lifespan and extends findings
from previous studies3 supporting the potential role of
inflammation-targeted strategies in primary diabetes
prevention. For aspirin, precision approaches might
identify subpopulations that are likely to derive the
maximal net cardiometabolic benefit (eg, those with
the highest risk of incident diabetes and cardiovascular
disease but the lowest risk of bleeding and progressive
cancer).9 Appreciation of these wider benefits might
also inform decision-making in selected secondary
prevention populations for whom reappraisal of the
role of lifelong aspirin therapy has been suggested.10
VRA reports fees for consultancy from Applied Therapeutics, Pfizer, Novo
Nordisk, Sanofi, Mediflix, and Fractyl; fees from contracts from Applied
Therapeutics, Eli Lilly, Novo Nordisk, Sanofi, and Fractyl; and their spouse is an
employee of Johnson & Johnson. JWO declares no competing interests.
Association of glycaemic index and glycaemic load with
type 2 diabetes and related conditions in prospective studies
There is debate over the nature of the carbohydrate-
containing foods that should be recommended
to reduce the risk of type 2 diabetes and
related conditions. In particular, it is not well
established whether glycaemic index (GI) is
a rel­evant dietary factor for the prevention of
chronic diseases in the general population. In
The Lancet Diabetes & Endocrinology, David Jenkins and
colleagues1 report a meta-analysis of large prospective
cohorts (≥100 000 participants) on the associations
between GI and glycaemic load (GL) and the incidence
of type 2 diabetes, cardiovascular disease, diabetes-
related cancers, and all-cause mortality. This study was
done in response to a WHO-sponsored meta-analysis
that concluded that, in relation to the associations of
low GI and GL with the incidence of type 2 diabetes
and other crucial clinical outcomes in observational
studies, the available evidence “is graded as low or very
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Comment
John W Ostrominski, *Vanita R Aroda
varoda@bwh.harvard.edu
Cardiovascular Division (JWO) and Division of Endocrinology (JWO, VRA), Brigham
and Women’s Hospital, Harvard Medical School, Boston, MA 02115, USA
1 Ong KL, Stafford LK, McLaughlin SA, et al. Global, regional, and national
burden of diabetes from 1990 to 2021, with projections of prevalence to
2050: a systematic analysis for the Global Burden of Disease Study
2021. Lancet 2023; 402: 203–34.
2 ElSayed NA, Aleppo G, Aroda VR, et al. 13. Older adults: standards of care in
diabetes-2023. Diabetes Care 2023; 46 (suppl 1): S216–29.
3 Rohm TV, Meier DT, Olefsky JM, Donath MY. Inflammation in obesity,
diabetes, and related disorders. Immunity 2022; 55: 31–55.
4 Zoungas S, Zhou Z, Owen AJ, et al. Daily low dose aspirin and incident type 2
diabetes in community-dwelling healthy older adults: a post-hoc analysis of
efficacy and safety in the ASPREE randomised placebo-controlled
trial. Lancet Diabetes Endocrinol 2023; published online Dec 21.
https://doi.org/10.1016/S2213-8587(23)00327-3.
5 McNeil JJ, Woods RL, Nelson MR, et al. Effect of aspirin on disability-free
survival in the healthy elderly. N Engl J Med 2018; 379: 1499–508.
6 McNeil JJ, Nelson MR, Woods RL, et al. Effect of aspirin on all-cause mortality
in the healthy elderly. N Engl J Med 2018; 379: 1519–28.
7 Barker AL, Morello R, Thao LTP, et al. Daily low-dose aspirin and risk of
serious falls and fractures in healthy older people: a substudy of the ASPREE
randomized clinical trial. JAMA Intern Med 2022; 182: 1289–97.
8 Fegers-Wustrow I, Gianos E, Halle M, Yang E. Comparison of American and
European guidelines for primary prevention of cardiovascular
disease: JACC Guideline Comparison. J Am Coll Cardiol 2022; 79: 1304–13.
9 Cofer LB, Barrett TJ, Berger JS. Aspirin for the primary prevention of
cardiovascular disease: time for a platelet-guided approach.
Arterioscler Thromb Vasc Biol 2022; 42: 1207–16.
10 Jacobsen AP, Raber I, McCarthy CP, et al. Lifelong aspirin for all in the
secondary prevention of chronic coronary syndrome: still sacrosanct or is
reappraisal warrranted? Circulation 2020; 142: 1579–90.
low quality as a result of high risk of bias, imprecision,
Paul Rapson/Science Photo Library
and inconsistencies”.2
Previous meta-analyses showed significant associa­
tions between the habitual consumption of low GI and
GL foods and a reduced incidence of diabetes and related
conditions.3,4 However, there are two features that make
this study novel and scientifically relevant. First, the See Articles page 107
selection of large cohorts allowed for sufficient statistical
power and minimised the variability seen with small
study assessments. Second, direct comparisons within
the same cohorts were made between the health benefits
associated with low GI and GL and those associated with
high fibre and wholegrain intake.
Jenkins and colleagues1 provide strong and high-quality
evidence that a low GI diet is associated with a reduced
risk of type 2 diabetes, cardiovascular disease, diabetes-
related cancers, and all-cause morta­ lity, and that the
magnitude of the risk reduction is similar to the widely
85
 Comment
accepted benefits of high fibre and wholegrain intake.
Nevertheless, the evidence for GL associations was not as
equally robust as that for GI. In fact, although low dietary
GL was also significantly associated with a reduced risk
of type 2 diabetes (in women only) and cardiovascular
diseases, it was not significantly associated with all-cause
mortality and diabetes-related cancers.1
This finding is somehow surprising given that
GL—a parameter combining the value of GI with the total
amount of carbohydrate consumed—is a better predictor
than GI of the postprandial glucose response, which
represents an important marker of health risk.5 Although
the reasons for this finding are not completely clear,
a possible explanation is that a diet with high GL usually
contains a large amount of high GI foods (eg, bread or
rice)5 and, in most low GL diets, these foods are replaced
by non-carbohydrate foods (eg, processed and red meat),
which are associated with an even higher risk of type 2
diabetes, cardiovascular diseases, and some cancers.6 If
this interpretation is correct, it might suggest that the
most appropriate measure for reducing dietary GL is the
replacement of high GI foods with low GI ones, rather
than with foods of animal origin.
The major factor underlying the association between
low GI and beneficial health outcomes is the reduction of
glycaemic excursion. This mechanism has been shown in
a randomised controlled trial in people without diabetes,
in which 24 h blood glucose profiles were measured
by continuous glucose monitoring after a 3-month
intervention with diets differing only in GI and GL.7
Remarkably, the difference in the postprandial glucose
response between participants receiving a high versus
low GI diet increased over time. This observation was due
to the ensuing glucose toxicity from the high GI and GL
diet that impaired both insulin sensitivity and insulin
secretion, two key pathogenetic mechanisms of type 2
diabetes.
One frequent criticism of the inclusion of low GI foods
among the recommended choices for the prevention
of type 2 diabetes and related conditions is that most
low GI foods (eg, legumes, vegetables, fruit, and some
cereals) are also rich in fibre; therefore, the potential
clinical benefits of low GI diets reported in observational
studies might be due to their fibre content. Accordingly,
it is not clear whether manufactured foods with a low GI
and a low fibre content are associated with the same long-
term benefits conferred by plant-based foods with low
86
GIs.8 The meta-analysis by Jenkins and colleagues1 partially
clarifies this, given that 17% of assessments controlled for
fibre and showed little difference to analyses performed
without controlling for fibre. However, future studies are
needed to fully elucidate this issue.
Another aspect that deserves further investigation
is the influence of sex in modulating the associations
between GI and GL and health outcomes. As reported
in the meta-analysis,1 the associations between a low
GI diet and a reduced risk of cardiovascular disease,
cancer, and total mortality were seen mostly in women.
Additionally, a randomised controlled trial published in
2023 showed that a low GI diet reduced postprandial
glucose response in women, but not in men.9
In conclusion, the meta-analysis by Jenkins and
colleagues1 clearly shows that the clinical benefits of
a low GI diet are similar to those achieved by a diet high
in fibre and wholegrain. These findings support the
inclusion of GI as a marker of carbohydrate quality in
dietary recommendations for the adult population.
GR is a member of the Scientific Advisory Board of the Nutrition Foundation
of Italy and of the Istituto Nutrizionale Carapelli Foundation and a member of
the Health and Wellbeing Advisory Board of the Barilla GeR Fratelli Company.
OV declares no competing interests.
*Gabriele Riccardi, Olga Vaccaro
gabriele.riccardi@unina.it
Diabetes, Nutrition and Metabolism Unit, Department of Clinical Medicine and
Surgery, Federico II University, Naples 80131, Italy
1 Jenkins DJA, Willett WC, Yusuf S, et al. Association of glycaemic index and
glycaemic load with type 2 diabetes, cardiovascular disease, cancer,
and all-cause mortality: a meta-analysis of mega cohorts of more than
100 000 participants. Lancet Diabetes Endocrinol 2024; 12: 107–18.
2 Reynolds A, Mann J, Cummings J, Winter N, Mete E, Te Morenga L.
Carbohydrate quality and human health: a series of systematic reviews and
meta-analyses. Lancet 2019; 393: 434–45.
3 Livesey G, Taylor R, Livesey HF, et al. Dietary glycemic index and load and
the risk of type 2 diabetes: a systematic review and updated meta-analyses
of prospective cohort studies. Nutrients 2019; 11: 1280.
4 Dwivedi AK, Dubey P, Reddy SY, Clegg DJ. Associations of glycemic index
and glycemic load with cardiovascular disease: updated evidence from
meta-analysis and cohort studies. Curr Cardiol Rep 2022; 24: 141–61.
5 Dehghan M, Mente A, Zhang X, et al. Associations of fats and carbohydrate
intake with cardiovascular disease and mortality in 18 countries from
five continents (PURE): a prospective cohort study. Lancet 2017;
390: 2050–62.
6 Bao J, Atkinson F, Petocz P, Willett WC, Brand-Miller JC. Prediction of
postprandial glycemia and insulinemia in lean, young, healthy adults:
glycemic load compared with carbohydrate content alone. Am J Clin Nutr
2011; 93: 984–96.
7 Bergia RE, Giacco R, Hjorth T, et al. Differential glycemic effects of
low- versus high-glycemic index Mediterranean-style eating patterns in
adults at risk for type 2 diabetes: the MEDGI-Carb randomized controlled
trial. Nutrients 2022; 14: 706.
8 Mann J. Dietary carbohydrate: relationship to cardiovascular disease
and disorders of carbohydrate metabolism. Eur J Clin Nutr 2007;
61 (suppl 1): S100–11.
9 Vitale M, Costabile G, Bergia RE, et al. The effects of Mediterranean diets
with low or high glycemic index on plasma glucose and insulin profiles are
different in adult men and women: data from MEDGI-Carb randomized
clinical trial. Clin Nutr 2023; 42: 2022–28.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
 Review
Lancet Diabetes Endocrinol
2024; 12: 132–48
*Joint first authors
Department of Health
Promotion and Policy,
University of Massachusetts
Amherst, Amherst, MA, USA
(J Hartmann-Boyce DPhil);
Nuffield Department of
Primary Care Health Sciences
(J Hartmann-Boyce,
E Morris MRCGP), Department
for Continuing Education
(I Onakpoya DPhil), and Medical
Sciences Division
(J Morgan BMBCh,
R Lynch BMBCh), University of
Oxford, Oxford, UK; Diabetes
Research Centre, University of
Leicester, UK (P Highton PhD,
L O’Mahoney PhD, S Seidu MD,
Prof K Khunti FMedSci);
Warwick, UK (K Rees PhD);
National Institute for Health
and Care Research Newcastle In
Vitro Diagnostics Co-operative,
Newcastle University,
Newcastle, UK (J Suklan PhD);
Liverpool Reviews and
Implementation Group,
University of Liverpool,
Liverpool, UK (F Curtis PhD);
Medical Research Council
Epidemiology Unit, University
of Cambridge, Cambridge, UK
(L Kudlek MSc); WHO,
Los Angeles, CA, USA
(S Marpadga MSc)
Correspondence to:
Jamie Hartmann-Boyce,
Department of Health
Promotion and Policy, University
of Massachusetts Amherst,
Amherst, MA 01003, USA
jhartmannboy@umass.edu
For the study protocol see
https://osf.io/hevpj/
132
The impact of the COVID-19 pandemic and associated
disruptions in health-care provision on clinical outcomes in
people with diabetes: a systematic review
Jamie Hartmann-Boyce*, Patrick Highton*, Karen Rees, Igho Onakpoya, Jana Suklan, Ffion Curtis, Lauren O’Mahoney, Elizabeth Morris,
Laura Kudlek, Jessica Morgan, Rosie Lynch, Sanjana Marpadga, Samuel Seidu, Kamlesh Khunti
The COVID-19 pandemic triggered disruptions to health care and lifestyles that could conceivably impact diabetes
management. We set out to identify the impact of disruptions caused by COVID-19 on clinical outcomes in people
with diabetes. We performed a systematic review of the available literature in the MEDLINE and OVID databases
from Jan 1, 2020, to June 7, 2023, and included 138 studies (n>1 000 000 people). All but five studies were judged to be
at some risk of bias. All studies compared prepandemic with pandemic periods. All-cause mortality (six studies) and
diabetes-related mortality (13 studies) showed consistent increases, and most studies indicated increases in sight loss
(six studies). In adult and mixed samples, data generally suggested no difference in diabetic ketoacidosis frequency or
severity, whereas in children and adolescents most studies showed increases with some due to new-onset diabetes
(69 studies). Data suggested decreases in hospital admissions in adults but increases in diabetes-related admissions
to paediatric intensive care units (35 studies). Data were equivocal on diabetic foot ulcer presentations (nine studies),
emergency department admissions (nine studies), and overall amputation rates (20 studies). No studies investigated
renal failure. Where reported, the impact was most pronounced for females, younger people, and racial and ethnic
minority groups. Further studies are needed to investigate the longer-term impact of the pandemic and the on
potential differential impacts, which risk further exacerbating existing inequalities within people with diabetes.
Introduction diabetes, both in those with pre-existing diabetes and in
Diabetes is a well documented risk factor for COVID-19 those presenting with diabetes for the first time during
severity and mortality.1 However, there is limited research the pandemic.
on the indirect impacts of the COVID-19 pandemic and
associated prevention measures (lockdowns, social Methods
distancing, and closure or limitation of certain industries, This systematic review was commissioned by WHO
including those related to health-care delivery, food and WHO member states to address key questions and
access, and exercise) on clinical outcomes in people with provide high-quality, evidence-informed information
diabetes. During the pandemic, routine diabetes care regarding COVID-19 and diabetes. A protocol was
was influenced by various factors including: limited or preregistered on Open Science Framework.
altered health-care provision due to infection prevention
measures; staff shortages due to staff time off for illness Search strategy and selection criteria
and reallocation of resources; patient concerns about We searched two electronic databases (MEDLINE and
visiting health-care sites due to infection risk; and the OVID) on June 7, 2023, for articles published between
impacts of the pandemic on disease self-management,2 Jan 1, 2020, and June 7, 2023, combining terms related to
nutritional habits, physical activity,3 and mental health.4 diabetes (eg, diabetes, diabetic), SARS-CoV-2 (eg, COVID,
Limited health-care access during the pandemic could coronavirus), and clinical outcomes (eg, mortality,
cause a delayed surge in health-care use post pandemic hospital admission), published in any language. We also
due to inadequate disease management and prevention, screened reference lists of included studies. Results
with serious health and economic implications. This were screened in duplicate using Covidence with
phenomenon has been documented during previous disagreements resolved via discussion or referral to
natural disasters, which impacted management of a third reviewer.
diabetes complications including diabetic foot,5 diabetic Our inclusion criteria were defined using the PICOS
retinopathy,6 and diabetic ketoacidosis.7 (Population, Exposure, Comparator, and Outcomes)
To our knowledge, there is no comprehensive framework. For population, we included people
systematic review of the evidence relating to the impact diagnosed with any type of diabetes (not including
of COVID-19-related disruptions on clinical outcomes in prediabetes), with no limitations by age, disease severity,
people with diabetes. Investigating this impact is vital to or duration. Our exposure of interest was COVID-19-
ensure that any negative consequences are minimised, associated prevention measures, often described by
and to inform responses to future pandemics or natural authors as the pandemic period, excluding studies that
disasters. We aimed to synthesise and describe the only measured the direct impact of COVID-19 infection,
current body of evidence relating to disruptions caused compared with prepandemic or postpandemic periods
by COVID-19 on clinical outcomes in people with as defined by authors. We were interested in studies
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
 Review

NOS Country Effect Supporting data

Bello-Chavolla and colleagues19 4
Hernandez-Vasquez and
colleagues20
Kleibert and colleagues21
Laliotis and colleagues22
Lee and colleagues23
Lozano-Corona and colleagues24 4
Lv 202224
McCoy and colleagues14
Raknes and colleagues26
Sekowski and colleagues27
Silverio-Murillo and colleagues28 5
Todd and colleagues29
Yao and colleagues30
NOS=Newcastle-Ottawa Scale.
direction
Mexico ↑ 148 437 diabetes-related deaths (177 per 100 000 inhabitants) in 2020 compared with an average of 101 496 deaths in
2017–2019 (125 per 100 000 inhabitants), a 41·6% increase.
3 Peru ↑ 92% increase in the average number of monthly deaths in people admitted to hospital with type 2 diabetes as cause for
admission.
4 Poland ↑ Increase in percentage of patients who died during hospitalisation for diabetic foot ulcers (from 3·9% to 4·8%, p=0·03).
6 UK ↑ Using national registry data, an excess 4108 deaths due to diabetes are estimated during 2020 compared with before the
pandemic.
7 USA ↑ Using US Centers for Disease Control and Prevention data to model mortality, it is reported that deaths from diabetes increased
during compared with before the pandemic; an estimated 24 700 excess deaths were attributable to diabetes due to the indirect
impacts of the pandemic (95% prediction interval 15 900–33 300).
Mexico ↑ Increase (not statistically significant) in percentage of patients who died during hospitalisation due to diabetic foot (13% vs 12%).
4 USA ↑ Increase of more than 30% during the pandemic, from 106·8 per 100 000 persons in 2019 to 144·1 in 2020 and 148·3 in 2021,
using data from 4 243 254 US adults.
4 USA ↑ Deaths with diabetes as primary cause increased (incident rate ratio 1·08, 95% CI 1·01–1·15; not statistically significant after
adjusting for multiple comparisons), data from 20 054 people in Minnesota with diabetes.
4 Norway ↑ Using national data, death rates were higher than predicted for diabetes mellitus (4·1%, 95% CI 2·1–3·4).
8 Poland ↑ Compared with 2019, the in-hospital mortality rate in 2020 increased by 66·7% among patients hospitalised with type 1
diabetes and by 48·5% among patients hospitalised with type 2 diabetes.
Mexico ↑ Using state mortality records, a 36·8% increase in diabetes-related mortality was reported during the pandemic (2020)
compared with prepandemic (2019; p<0·001).
6 USA ↑ Using Pennsylvania mortality data, an estimated 41% more deaths than expected due to diabetes were reported during
compared with before the pandemic (the authors state that diabetes was one of the top three greatest proportional increases in
non-COVID-19 causes of death during the pandemic).
7 USA ↑ Excess mortality was modelled based on national data, giving the estimates that from March, 2020, to March, 2022, deaths that
had diabetes as one of multiple causes of death were 47·6% higher than expected and those with diabetes as an underlying
cause of death were 18·4% higher than expected.

Table 1: Diabetes-related mortality during versus before the COVID-19 pandemic

reporting the following outcomes where they were not
solely attributable to COVID-19 infection: mortality;
hospital admissions; diabetic ketoacidosis; amputations;
sight loss; emergency admissions; foot ulcer presen­
tations; severe hypoglycaemic events; and renal failure.
We excluded letters, editorials, case series, and case
reports but otherwise did not restrict studies based on
their design, language, or geographical location. We
included systematic reviews meeting the above criteria.
Data extraction and analysis
One reviewer appraised and extracted data on study
design, study setting, sample characteristics, time
periods of interest, analysis methods, outcome
measures, and results; a second (PH or JH-B) checked
them. Discrepancies were resolved through discussion.
We relied on the outcome measures reported by
primary studies, and did not attempt to recategorise
outcomes. This means there is overlap between some
of the outcomes in our review. For example, where a
study reported diabetes-related hospital admissions
and did not break this down further (eg, by type, such
as diabetic ketoacidosis), it was included in the diabetes-
related hospital admissions outcome. If a study
reported hospitalisations for diabetic ketoacidosis but
not overall diabetes-related hospital admissions, it was
included in the diabetic ketoacidosis section. If it
reported both, it was included in diabetes-related
hospital admissions (admissions including but not
limited to diabetic ketoacidosis) and in the diabetic
ketoacidosis outcome (including diabetic ketoacidosis
admissions only).
Critical appraisal
Primary studies were appraised using the Newcastle-
Ottawa Scale (NOS).8 Systematic reviews were assessed
using the A Measurement Tool to Assess Systematic
Reviews (AMSTAR) 2 checklist.9 Studies were not
excluded based on appraisal outcomes.
Data synthesis
Heterogeneity in study design and outcomes precluded
meta-analysis. Data were narratively synthesised by
clinical outcome, with effect direction plots used where
more than six studies contributed data to an outcome, in
accordance with Synthesis Without Meta-analysis
guidance10 and the Cochrane Handbook.11,12 Based on this
guidance, we judged effect direction based on point
estimates (where available) rather than on statistical
significance. Where point estimates were not available
but some other indication of direction of effect was given
(eg, absolute numbers or narrative description), we used
this to inform our categorisation. Where results were
described as not statistically significant, and no other

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 133

 Review

reviews, including AMSTAR 2 scores (range 9–13). All

11 046 records identified
11 026 database searching
20 other sources
25 records removed before screening
25 duplicate records
11 021 records screened
10 485 records excluded (irrelevant articles)
536 reports sought for retrieval
536 reports assessed for eligibility
398 reports excluded
157 ineligible outcomes
125 ineligible comparator
32 ineligible patient population
55 ineligible exposure
29 ineligible setting
138 studies included in review
reviews were judged to have two or three critical
weaknesses.
54 studies were conducted exclusively in people with
type 1 diabetes, and four exclusively in people with type 2
diabetes. The remainder were conducted in mixed
populations or did not specify diabetes type.
Studies were conducted in North America (39 studies),
Western Europe (39 studies), Asia (17 studies),
Eastern Europe (14 studies), South America (four studies),
Egypt (one study), and Australia (one study). The
remainder incorporated multiple regions.
All studies compared prepandemic with pandemic
periods. Of the 138 included studies, the following
outcomes were covered: all-cause (six studies) and
diabetes-related (13 studies) mortality; all-cause
(five studies) and diabetes-related (30 studies) hospital
admissions; diabetes-related emergency hospital visits
(nine studies); diabetic ketoacidosis (69 studies); severe
hypoglycaemia (four studies); amputations (20 studies);
foot ulcer incidence (nine studies); and sight loss
(six studies). No studies reported data on renal failure.
Findings are summarised by outcome and effect
direction in figure 2.

All-cause mortality in people with diabetes

Figure 1: PRISMA 2020 flow diagram
Records are of new systematic reviews that included searches of databases and
registers only.
information was provided with which to judge an effect
direction, we categorised this as no difference and note
the absence of further information in table 1. Systematic
review findings were narratively synthesised alongside
data on each outcome.
Role of the funding source
WHO approved the review protocol and our decision to
submit for publication. NIHR had no role in study
design, analysis, interpretation, writing, or the decision
to submit for publication.
Results
Included studies
We screened 11 021 records, 20 of which were identified
through screening reference lists of included studies, and
the remainder through database searching. The most
common reasons for exclusion at full text stage were that
studies did not report on any of our outcomes of interest,
or did not compare to prepandemic or postpandemic
periods. We included 138 publications (133 primary
studies and five systematic reviews; figure 1). The
See Online for appendix appendix (pp 1–24) contains key characteristics of the
primary studies, including NOS scores (range 2–8). Of
the 133 primary studies, 58 had NOS scores lower than 5;
only five had NOS scores of 8, indicating they were not
judged to be at any risk of bias. The appendix (p 25)
contains key characteristics of five the included systematic
Out of the six studies that examined all-cause mortality,
all but one found increases during the pandemic. Only
one study13 (NOS 7) excluded COVID-19 deaths from
analysis. This study used data from the National Diabetes
Audit in England comparing the period July 3–Oct 15, 2021
(n=3 218 570) with the equivalent 15-week period in 2019
(n=2 973 645), and found an increase in non-COVID-
related deaths from 2019 to 2021 of 11% (95% CI 9–13).
The unadjusted incidence rate ratio (RR) for 2021
compared with 2019 was 1·026 (95% CI 1·009–1·043);
this was unchanged after adjustment for age, sex,
ethnicity, socioeconomic deprivation, and diabetes type.
Both unadjusted and adjusted mortality rates were
higher in those who did not receive eight annual care
processes during the pandemic (foot surveillance, urine
albumin, BMI, blood pressure, smoking, cholesterol,
HbA1c, and serum creatine) compared with those who
received all eight during the pandemic.13
The remaining studies did not exclude deaths from
COVID-19 from their analysis; four found increases
during compared with before the pandemic. McCoy and
colleagues14 (NOS 4, n=20 054) used registry data from
people with diabetes in Minnesota, USA to compare all-
cause mortality in 2018–19 with that in 2020, and reported
an unadjusted incidence RR of 1·30 (95% CI 1·26–1·33).
Torre and colleagues15 (NOS 5, n not reported) used
national statistics from Italy comparing 2020 with 2019
and comparing people with and without diabetes. People
with diabetes recorded a two times higher rate of excess
mortality in 2020 compared with those without diabetes
(20·4% vs 10·2%). McAlister and colleagues16 (NOS 8)

134 www.thelancet.com/diabetes-endocrinology Vol 12 February 2024


conducted a population-based retrospective cohort study
in Canada and found that within people with diabetes
attending a health-care visit, death rates per 1000 visits
were higher during the pandemic (1·4 in 30 days and
5·5 in 90 days following the visit) than in the year before
(1·2 in 30 days and 4·5 in 90 days following the visit).
Khaydarova and colleagues17 (NOS 3) reviewed medical
records in Uzbekistan and found mortality among people
with type 2 diabetes increased compared with
prepandemic (4·3% in 2020 vs 2·8% in 2019). Lastly,
Harun and colleagues18 (NOS 6) examined deaths (any
cause) in people with diabetic foot ulcers in one hospital
in Saudi Arabia, and found a non-statistically significant
decrease.
Diabetes-related mortality
13 studies compared diabetes-related mortality during
(2020–21) versus before (2006 to February, 2020) the
pandemic (table 1). We used data on diabetes-related
mortality as reported by the study authors. Definitions
were rarely provided therefore the underlying definition
might have varied between studies. All studies reporting
this measure detected increases during compared with
before the pandemic.
Three of these studies also analysed data by subgroups.
Lv and colleagues24 (NOS 4), which used US population-
wide data, reported that the relative rise in diabetes-related
mortality was most pronounced in younger adults (aged
25–44 years). Older adults showed the smallest difference
in mortality between before and during the pandemic. The
sharpest rise was found in the Hispanic population,
followed by non-Hispanic Black, non-Hispanic Asian, and
non-Hispanic Native American and Alaskan, whereas
the smallest rise was found in the non-Hispanic
White population. The mortality increase among the
Hispanic population was nearly three times that in
Decrease during compared
with before pandemic
Outcome
All-cause mortality
Diabetes-related mortality
All-cause hospital admissions
Diabetes-related hospital admissions
Emergency (ER or A&E) admissions
Severe hypoglycaemic events
Amputations
Sight loss
Diabetic foot ulcer presentations
DKA in children/adolescent samples
DKA in adult/mixed samples
10 5 1
N studies
Figure 2: Number of studies showing increase, decrease, or no difference in outcomes during compared with before the pandemic
For diabetes-related hospital admissions, cross hatching indicates PICU admissions. For diabetic ketoacidosis in child or adolescent samples, cross hatching indicates new-onset diabetes. A&E=accident
and emergency department. ER=emergency room. PICU=paediatric intensive care unit.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Review
the non-Hispanic White population. Excess rates of
diabetes-related mortality were 30% for men and
40% for women.
McCoy and colleagues14 (NOS 4) found no age-related
differences, but found that the proportion of diabetes-
related deaths in 2020 in people with diabetes belonging
to a racial or ethnic minority group was greater than in
the prepandemic period (p<0·001). By contrast, Todd and
colleagues29 (NOS 6) reported that although excess
mortality due to other causes varied by ethnicity, it did
not appear to do so for diabetes, and that excess mortality
was higher in older age groups.
All-cause hospital admissions in people with diabetes
Five studies provided data on all-cause hospital admissions;
three noted a decline and two observed no difference. Via
a survey in Italy, Bossi and colleagues31 (NOS 3) interviewed
947 people with diabetes. Compared with November, 2019,
most participants (80%) in November, 2020, indicated
their number of hospitalisations that year was unchanged,
whereas 6% indicated they had decreased and
13% indicated they had increased. Nowak and colleagues32
(NOS 4) performed a retrospective evaluation of acute
hospitalisations in people with type 1 diabetes aged
15–17 years in a hospital in Poland, and found no
statistically significant difference in incidence between the
prepandemic (23 patients in 2018–19) and pandemic
period (16 patients in 2020–21).
Tehrani and colleagues33 (NOS 4) also examined
hospital admissions in children and adolescents with
type 1 diabetes in one clinic in Iran, and noted
a statistically significant decline during the pandemic
(p=0·005). Yoon and colleagues34 (NOS 6) used data from
a cohort of people with diabetes judged to be at high risk
within a Veterans Affairs medical centre in the USA, and
reported a mean decrease of 0·23 (95% CI –0·28 to –0·19)
No difference Increase during compared
with before pandemic
PICU admissions
New onset
1 5 10 15 20 25 30 35 40 45
N studies
135
 Review

in all-cause hospitalisations per patient per quarter in the
early-pandemic phase compared with the prepandemic
phase. Hammersen and colleagues35 (NOS 6) used data
from a large German registry of people with type 1
diabetes and noted fewer hospitalisations during
compared with before the pandemic.
Diabetes-related hospital admissions
30 studies evaluated diabetes-related hospital admissions
(table 2). Findings were mixed, with some showing
increases and others showing decreases relative to the
prepandemic period. Much of this variation was explained
by population and setting, and hence, table 2 is divided
into studies in children (where increases were common)
and in adults (where most studies showed no difference
or an increase).
Three studies analysed trends in diabetes-related
hospitalisation by population groups. Loh and
colleagues61 (NOS 6, in children with diabetes) looked at
the difference in admissions to hospital due to a new
onset or previous diabetes diagnosis between boys and
girls aged under 18 years, and reported that in 2020,
more girls (62·5%) were admitted to hospital due to
diabetes diagnosis (new-onset or already known disease),
whereas in 2019, sex was nearly evenly distributed. Moin
and colleagues58 (NOS 6, in mixed samples in hospitals
across Ontario) reported no differences in trends based
on age, socioeconomic status, sex, or comorbidities.
Sekowski and colleagues27 (NOS 8), using national
registry data from adults in Poland, noted an overall
decrease in the number of diabetes-related hospitali­
sations from 2019 to 2020. This decrease was steeper in
people with type 2 than type 1 diabetes, in women than in
men, in non-insulin-dependent diabetes than in insulin-
dependent diabetes, in people with type 1 diabetes aged
20–39 years compared with other age groups, and in
people with type 2 diabetes aged 40–49 years or 80 years
and older compared with other age groups.
Emergency admissions in people with diabetes
Findings from the nine studies with data on emergency
hospital admissions in people with diabetes during

NOS Country Effect Supporting data

direction
Children and adolescents
Breinig and colleagues36 4 France ↑ National database including all children aged 7 days to 18 years admitted in PICU in March–May, 2020, compared with the same
period in 2019. Diabetes admissions increased significantly (60%, IRR 2·54 [95% CI 1·30–5·25], p=0·003).
Choudhary and 4 USA ↔ Children and adolescents with type 1 diabetes seen at least once at the large urban paediatric teaching hospital (or smaller local
colleagues37 hospital) during the study period; no change detected from 2019 to 2020.
Goldman and 5 Israel ↑ Compared data on newly diagnosed young people (aged <18 years) with type 1 diabetes in March–June, 2020 (during lockdown) and
colleagues38 the same periods in 2017–2019. ICU admissions were higher in 2020 than in previous years: 34·2% versus 25·7%, 17·6%, and 15·8%
(p=0·001).
Guemes and colleagues39 6 Spain ↑ During the pandemic, 7 of 10 patients presenting to the emergency department with type 1 diabetes were admitted to the PICU,
compared with 0 of 10 in timeframe-matched controls from 2018 and 2019 (p=0·003).
Hammersen and 4 Germany ↓ Patients with type 1 diabetes younger than 18 years with diabetes duration more than 3 months with visits in 231 centres in 2020;
colleagues40 relative risk for hospitalisation was lower compared with the preceding year, most notably in the second lockdown month (risk ratio
0·52 [0·46–0·58], p< 0·001) compared with 2019.
Jafari and colleagues41 5 USA ↑ In 175 paediatric patients with type 1 diabetes presenting with diabetic ketoacidosis, the proportion requiring PICU admission
increased from 34·2% to 54·6%.
Kiral and colleagues42 4 Türkiye ↑ Data from multi-centre PICUs; for 917 children younger than 18 years with type 1 diabetes, PICU stays increased during the pandemic
(further data not provided).
LahTomulic and 4 Croatia ↑ Average number of children with diabetic ketoacidosis admitted to the PICU, 2011–2019 (4·1) and 2020 (11·0).
colleagues43
Mameli and colleagues44 6 Italy ↑ In children in Lombardy, Italy with type 1 diabetes, frequency of PICU admission reached its highest value in the first wave (10·0% of
new-onset type 1 diabetes) compared with the second wave (8·3%) of 2020, and also compared with the same months of the
previous years (7·3% in 2019, 3·7% in 2018, and 6·9% in 2017).
McGlacken-Byrne and 6 UK ↑ Hospitals from the North Central London Paediatric Diabetes Network: PICU diabetic ketoacidosis admissions were 23·5% during the
colleagues45 first COVID- 19 wave compared with 6·7% prepandemic.
Miller and colleagues46 7 USA ↑ Retrospective chart review of 132 children with diabetes admitted to two hospitals for diabetes-related causes: 3·08% in 2018,
3·54% in 2019 (p=0·012), 4·73% in 2020 (p=0·077).
Pietrzak and colleagues47 4 Poland ↑ Data from regional diabetes centres showed frequency of ICU admissions among type 1 diabetes cases increased between prepandemic
and during the pandemic (13·6% vs 17·4%, p=0·63), and significantly more children required assisted ventilation (1·3% vs 3·4%,
p=0·037).
Ponmani and 5 UK and ↑ Children aged 6 months to 16 years presenting to 49 emergency departments with diabetic ketoacidosis; observed increases in
colleagues48 Ireland admissions to intensive care from 38 to 72 (89%) during compared with before the pandemic.
Salmi and colleagues49 6 Finland ↑ In one hospital district, incidence of PICU admission due to new-onset type 1 diabetes increased from 2·89 per 100 000 person-years
in 2016–2019 to 9·35 per 100 000 person-years in 2020 (IRR 3·24 [1·80–5·83], p<0·001).
Zee-Cheng and 5 USA ↔ Database from 77 PICUs: prepandemic Q1 1072 (4·6%) cases due to diabetes, during-pandemic Q1 1054 (4·6%); p>0·99.
colleagues50
(Table 2 continues on next page)

136 www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

 Review

NOS Country Effect Supporting data

direction
(Continued from previous page)
Adults or mixed samples
Cassell and colleagues51 4 USA Mixed National health-care billing database. Estimated IRR (ratio of observed and predicted cases per diagnosis per month, with monthly
predictions made by using trends in previous years): 0·6–0·9 March–May, 2020; 0·9–1·1 June–Dec, 2020; 0·6–0·9 Jan, 2021;
0·9–1·1 Feb, 2021; 1·1–1·4 March–April, 2021; 0·9–1·1 May–June, 2021.
Dayal and colleagues52 5 India ↑ Hospital admissions of new cases of type 1 diabetes during lockdown compared with before: 19 in April, 2019–March, 2020:
4 in April, 2020 (p value not reported).
Grudziaz-Sekowska and 3 Poland ↓ National registry data on people with type 1 diabetes; compared with the prepandemic period, there was a significant decrease
colleagues53 (−27% compared with 2019; p value not reported) in the number of diabetes-related hospitalisations.
Guimaraes and 4 Brazil ↓ National data in adults aged 20 years or older; there was a statistically significant decrease (p<0·001) in hospitalisations due to
colleagues54 diabetes (–24·0% [–25·0% to –22·0%]) when comparing Jan–Feb, 2020 (prepandemic) with March, 2020–May, 2021.
Harun and colleagues18 6 Saudi ↓ Hospital admissions in 358 people with diabetic foot disease in one hospital non-significantly decreased during (2·07 [0·60–3·54])
Arabia compared with before the pandemic (2·29 [0·50–4·08], p=0·289).
Hernandez-Vasquez and 3 Peru ↓ In adults aged 18 years or older from a multihospital database, there was a positive linear slope from 2017 to March, 2020. The
colleagues22 difference in the slope after March 2020 was −2·4 admissions per month (–32·56 to 27·70; p<0·872) in the number of cases.
Kim and colleagues55 6 USA ↓ In data from more than 3 million people with chronic illnesses, admissions for diabetes decreased 31% during spring, 2020, and
36% during summer, 2020, compared with the same periods in 2017–19.
Kleibert and colleagues23 4 Poland ↔ The number of diabetic foot ulcer-related hospitalisations extracted from national health fund data did not change significantly in
2020 in comparison with 2017–19 (averaged; 13 375 vs 14 383, p=0·17).
Lui and colleagues56 5 Hong ↓ Total hospitalisation rate for all diabetes: IRR study period versus inter-year control 0·73 (0·09–0·78; p<0·001); IRR study period
Kong versus intra-year control 0·77 (0·02–0·83; p<0·001).
Michalowsky and 5 Germany ↓ In adults older than 65 years, diabetes hospital admissions were 0·6 in March, 2019, versus 0·3 in March, 2020 (–49%); 0·6 in
colleagues57 April, 2019, versus 0·4 in April, 2020 (–39%); and 0·6 in May, 2019, versus 0·3 in May, 2020 (–50%; p values not reported).
Moin and colleagues58 6 Canada ↓ Ontario, Canada hospitals showed a 16% decrease in hospital admissions for diabetes complications (majority type 2 diabetes).
Reschen and colleagues59 4 UK ↔ For one UK National Health Service trust, number of diabetes (non-COVID) admissions were 917 (22%) prepandemic and 558 (22%)
in the first pandemic wave (p=0·8).
Santana and colleagues60 3 Brazil ↓ In one paediatric hospital, 143 patients with type 1 diabetes were evaluated, of whom 12 (8·4%) needed hospitalisation because of
diabetes during the pandemic. In the same period prepandemic, the rate was 10·0%.
Sekowski and 8 Poland ↓ Using national registry data, among patients with type 1 diabetes, hospitalisation rate per 100 000 decreased from 74·6 in 2019 to
colleagues28 53·0 in 2020 (−28·9%). An even greater drop was observed among patients with type 2 diabetes, from 99·4 in 2019 to
61·6 in 2020 (−38·0%).
Yoon and colleagues34 6 USA ↓ Using data from a cohort of people with diabetes judged to be at high risk within a US Veterans Affairs medical centre, a mean
decrease was reported of 0·002 (–0·002 to –0·001) in diabetes-related hospitalisations per patient per quarter in the early pandemic
phase compared with the prepandemic phase.
ICU=intensive care unit. IRR=incident rate ratio. NOS=Newcastle-Ottawa Scale. PICU=paediatric intensive care unit.

Table 2: Diabetes-related hospital admissions during versus before the COVID-19 pandemic

compared with before the pandemic were mixed (table 3).
Two studies did not detect any difference, four studies
detected decreases, and three detected increases.
Diabetic ketoacidosis
Four systematic reviews evaluated associations between
the pandemic and diabetic ketoacidosis. Three found
increases in diabetic ketoacidosis and severe diabetic
ketoacidosis in children or new-onset type 1 diabetes
during compared with before the pandemic (appendix
p 25).
AlFayez and colleagues66 (AMSTAR 2 10/16; critical
domains 5/7) combined data from 20 studies, and found
the risks of diabetic ketoacidosis and severe diabetic
ketoacidosis increased during compared with before the
pandemic (diabetic ketoacidosis RR 1·35 [95% CI
1·20–1·53], I²=71%; severe diabetic ketoacidosis RR 1·76
[1·33–2·33], I²=44%). In those with new-onset diabetes,
the risk of diabetic ketoacidosis was 44% higher (RR 1·44
[1·26–1·65], I²=64%).
Elgenidy and colleagues67 (AMSTAR 2 9/16; critical
domains 4/7) analysed data across 24 studies in children
with newly diagnosed type 1 diabetes and found
a statistically significant increase in the risk of diabetic
ketoacidosis (RR 1·41 [1·19–1·67]) and severe
diabetic ketoacidosis (RR 1·66 [1·30–2·11], p<0·01,
I²=86%) during the pandemic compared with before.
No significant differences were detected among people
with pre-existing type 1 diabetes (RR 1·07 [0·79–1·46])
or mixed patients (RR 1·04 [0·84–1·29]).
Rahmati and colleagues68 (AMSTAR 2 9/16; critical
domains 4/7) analysed data from 21 studies on new-onset
type 1 diabetes in children. Meta-analyses showed
increased incidence of diabetic ketoacidosis (RR 0·064
[0·043–0·084], I²=3%) and severe diabetic ketoacidosis
(RR 0·049 [0·029–0·066], I²=14%) during compared with
before the pandemic.
O’Mahoney and colleagues69 (AMSTAR 2 9/16;
critical domains 5/7) did not statistically synthesise
data on diabetic ketoacidosis, but found no clear

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 137

 Review

NOS Country Effect Supporting data

Aubert and colleagues62
Bossi and colleagues31
Caruso and colleagues5
Celona and colleagues63
Giannouchos and colleagues64 3
McCalister and colleagues16
Moin and colleagues58
Sarikaya and colleagues65
Yoon and colleagues34
NOS=Newcastle-Ottawa Scale. RR=rate ratio.
direction
4 USA ↔ In veterans aged 65 years and older with type 2 diabetes, there was no evidence of a difference in hypoglycaemia-related
or hyperglycaemia-related emergency room visits or hospitalisations during the pandemic compared with
July, 2018–February, 2020. Hypoglycaemia: March, 2020 RR 1·01 (95% CI 0·54–1·09); April–November, 2020
RR 0·86 (0·68–1·09). Hyperglycaemia: March, 2020 RR 1·02 (0·66–1·58); April–November, 2020 RR 0·95 (0·80–1·11).
3 Italy ↔ Self-reported emergency department access during compared with before the pandemic in people with diabetes: unchanged
78%, decreased 9%, increased 13%.
5 Italy ↑ Patients admitted for diabetic foot ulcer emergency: 19 (76%) in 2020, ten (26%) in 2019 (p<0·001).
6 USA ↑ Analysis of medical record data from a large hospital, for young adults (aged 20–26 years) with chronic health conditions
presenting at the emergency department, reports that the proportion of emergency department encounters requiring
emergency hospital admissions increased from 41·8% to 61·1% during the pandemic among people with diabetes.
USA ↓ Emergency department visits for diabetes: January–March: 131 in 2019 versus 109 in 2020 April: 31 in 2019 versus 19 in 2020
May–August: 138 in 2019 versus 81 in 2020
8 Canada ↓ Emergency department visits for any cause, per 1000 healthcare visits: in 30 days after visit, 108·0 before pandemic,
83·6 during pandemic (p<0·001); in 90 days after visit, 299·5 before pandemic, 242·2 during pandemic (p<0·01).
6 Canada ↓ 18% decrease in emergency department admissions for acute diabetes complications in the first 6 months of the pandemic
compared with the two previous 6-month periods.
5 Türkiye ↑ In 89 children with type 1 diabetes, an increase in frequency of admissions to emergency services was noted during the
pandemic compared with previously.
6 USA ↓ Using data from a cohort of people with diabetes judged to be at high risk within a US Veterans Affairs medical centre, a mean
decrease was reported of 0·22 (95% CI –0·24 to –0·21) in emergency department visits per patient per quarter in the early
pandemic phase compared with the prepandemic phase.

Table 3: Emergency hospital admissions in people with diabetes during versus before the COVID-19 pandemic

evidence of a difference during compared with before
the pandemic.
A further 69 primary studies32,35,37,38,40–49,52,53,56,59,61,70–120
included diabetic ketoacidosis data (table 4). Of these,
13 were in adult or mixed cohorts; three of these found
a decrease in occurrence or severity of diabetic
ketoacidosis during compared with prepandemic,
two found an increase, and the remainder found no clear
evidence of a difference in either outcome. However,
most studies restricted to children and adolescent
samples (47 of 56) found that diabetic ketoacidosis
occurrence or severity increased during the pandemic
compared with before, whereas four found a reduction
and six found no clear difference. Of the 30 studies in
people with new-onset diabetes, 28 found increased
occurrence or severity of diabetic ketoacidosis during
compared with before the pandemic.
Three studies analysed trends based on socio­
demographic characteristics; Misra and colleagues70
(NOS 7, national dataset from England) also analysed by
diabetes type, and found that diabetic ketoacidosis
emergency hospital admissions were lower in people with
pre-existing type 1 diabetes during compared with before
the pandemic, but higher in people with type 2 diabetes
and in people with newly diagnosed diabetes. Increases
were particularly pronounced in those of non-
White ethnicities. Alassaf and colleagues71 (NOS 3,
137 with type 1 diabetes) found no significant difference in
age at diagnosis (p=0·914) or monthly income (p=0·254)
when comparing patients presenting with diabetic
ketoacidosis in new-onset type 1 diabetes during the
pandemic with the preceding year,. Female patients had
higher risk of presenting with diabetic ketoacidosis during
the prepandemic year (p=0·047). Monkemoller and
colleagues72 (NOS 6, data from 216 diabetes centres in
Germany) found that the largest increase from
prepandemic to during the pandemic was seen in children
aged 6 years and younger (141·6% increase), followed by
those aged 12–18 years (91·8% increase) and 6–11 years
(52·3% increase). Children aged 6 years and younger also
saw the largest increase in severe diabetic ketoacidosis
incidence (97%), followed by those aged 6–11 years (33·1%)
and 12–18 years (19·1%). The percentage increase during
compared with before the pandemic was similar between
men (88·2%) and women (81·3%) for diabetic
ketoacidosis, but severe diabetic ketoacidosis saw a steeper
rise in women (52·6%) than in men (40·4%). Children
with a migration background (definition not provided)
saw steeper percentage increases in diabetic ketoacidosis
(92·4% vs 80·8%) and severe diabetic ketoacidosis (62·0%
vs 36·5%) during compared with before the pandemic in
comparison with those without a migration background.
Severe hypoglycaemic events
Four studies provided data on occurrence of severe
hypoglycaemic events during compared with before
the pandemic; findings were mixed. Alsalman and
colleagues73 (NOS 2) conducted interviews in a sample of
164 paediatric patients with type 1 diabetes from a Saudi
university hospital; they reported that the need to visit the
emergency department for hypoglycaemia reduced
significantly during lockdown (p=0·001). Rabbone and
colleagues74 (NOS 6) also reported a decrease (statistical
significance not tested) in incidence of severe

138 www.thelancet.com/diabetes-endocrinology Vol 12 February 2024


Children and adolescents
Abdou and colleagues76
Al Agha and colleagues77
Alaqeel and colleagues78
Alassaf and colleagues71
Alsalman and colleagues73
Ansar and colleagues79
Baechle and colleagues80
Basamatur and colleagues81
Birkebaek and colleagues82
Boboc and colleagues83
Bogale and colleagues84
Botelho and colleagues85
Chambers and colleagues86
Cherubini and colleagues87
Choudhary and colleagues37
Danne and colleagues88
d’Annunzio and colleagues89
Dayal and colleagues52
Dovc and colleagues90
Dzygalo and colleagues91
Fathi and colleagues92
Goldman and colleagues38
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Review
NOS Country Effect Supporting data
direction
4 Egypt ↑ 324 children with type 1 diabetes, hypokalaemia, as complication of diabetic ketoacidosis management: 18% of patients
admitted during the pandemic developed hypokalaemia as a complication, in comparison with 6·7% of patients before the
pandemic, which was statistically significant (p=0·001).
2 Saudi Arabia ↓ 150 children with type 1 diabetes; percentage of patients who experienced diabetic ketoacidosis decreased from 35% before
the lockdown to 11% during the lockdown.
4 Saudi Arabia ↑ In 260 children with type 1 diabetes, diabetic ketoacidosis was higher in 2020 than in 2019 (83% vs 73%; RR 1·15 [95% CI
1·04–1·26], p=0·05), after adjusting for age and sex. Diabetic ketoacidosis frequency among children with new-onset type 1
diabetes was higher in 2020 than in 2019 (26% vs 13%, p≤0·001).
3 Jordan ↑ 137 children and adolescents with type 1 diabetes, from March 2, 2020, to March 1, 2021, compared with those diagnosed with
type 1 diabetes in the year preceding the pandemic. The percentage of patients diagnosed with diabetic ketoacidosis as first
presentation of type 1 diabetes during the pre-pandemic year was 35% compared with 52% during the pandemic year
(p=0·049).
2 Saudi Arabia ↓ 164 paediatric patients with type 1 diabetes self-reported a decrease in need for ICU admission due to diabetic ketoacidosis
during compared with before the pandemic.
2 USA ↑ Children with type 1 diabetes and type 2 diabetes (n not reported); mean monthly rates of severe diabetic ketoacidosis
(pH≤7·1) increased from 0·85 prepandemic to 2·14 during the pandemic (p<0·0001).
8 Germany ↑ National registry data of 30 840 children and adolescents with new-onset type 1 diabetes; incidence of diabetic ketoacidosis
during the pandemic was higher than predicted (2020 IRR 1.34 [1·23–1·46]; 2021 IRR 1·37 [1·26–1·49]).
4 UK ↑ In children with type 1 diabetes, median diabetic ketoacidosis referrals in March–July in the 5 preceding years was 12
(range 11–20), and in Mar–July 2020 it was 31 (p value not reported).
4 New ↑ 104 290 children with type 1 diabetes; adjusted observed prevalence of diabetic ketoacidosis at diagnosis of type 1 diabetes
Zealand, was 39·4% (95% CI 34·0–45·6) in 2020 and 38·9% (33·6–45·0) in 2021, significantly higher than the predicted prevalence of
Australia, 32·5% (27·8–37·9) for 2020 and 33·0% (28·3–38·5) for 2021 (p<0·0001 for both years).
USA, Europe
4 Romania ↑ 459 patients at a children’s hospital; proportion of diabetic ketoacidosis type 1 diabetes onset increased during the pandemic
with 67%, from 39% in the pre-pandemic group to 66% in the pandemic group (OR 2·98 [1·97–4·49], p<0·0001). In the
pandemic group, a higher percentage of people with diabetic ketoacidosis developed the severe form compared with the
prepandemic group (42% vs 27%; OR 1·99 [1·13–3·51], p=0·016).
3 USA ↔ 412 children and adolescents with new type 1 diabetes in a tertiary referral centre; percentages of diabetic ketoacidosis
diagnoses at admission were very similar between the prepandemic and postpandemic groups (47% vs 48%), as were the
severity percentages (13% vs 14% mild diabetic ketoacidosis, 33% vs 31% moderate or severe diabetic ketoacidosis; p=0·89).
3 Portugal ↑ Newly diagnosed type 1 diabetes from a single paediatric hospital; 54 cases (73%) of non-diabetic ketoacidosis prepandemic,
14 (48%) during the pandemic.
5 USA ↔ 412 admitted for new-onset type 1 or type 2 diabetes; “percentage of diabetic ketoacidosis diagnoses at admission were very
similar” (no data provided).
6 Italy ↑ Newly diagnosed type 1 diabetes at 47 diabetes centres in Italy; during all periods of extreme and partial restrictions, the
frequency of severe diabetic ketoacidosis was significantly higher in 2020 than in 2017–2019. March–May, 2020 versus
2017–19: OR 2·31 (1·34–4·01); June–September, 2020 versus 2017–19: OR 1·68 (1·00–2·84); October–December 2020 versus
2017–19: OR 1·54 (0·93–2·57).
4 USA ↔ More than 1600 patients with type 1 diabetes at a large teaching hospital; no change observed in diabetic ketoacidosis
hospitalisation frequency (271 admissions in 2019, 270 in 2020).
5 Multiple ↑ 28 892 people aged 21 years and younger with type 1 diabetes; diabetic ketoacidosis rates were reported to have increased
significantly in May–June (p=0·034) and August–September (p=0·007), 2020, compared with 2019.
6 Italy ↑ Paediatric hospitals in one region of Italy (99 people with newly diagnosed type 1 diabetes); frequency of diabetic ketoacidosis
at admission was statistically significantly higher during the pandemic compared with the period immediately beforehand
(61% vs 38%, p=0·03).
5 India ↑ Children with type 1 diabetes and severe diabetic ketoacidosis admissions. April 2019–March 2020: 2 of 12 admissions severe;
April 2020: 3 of 3 admissions severe (p value not reported).
6 Slovenia ↔ 326 children, adolescents, and young adults with type 1 diabetes; no diabetic ketoacidosis severe events or hospitalisations
reported prepandemic or during the pandemic.
5 Poland ↑ 86 children with type 1 diabetes. Diabetic ketoacidosis at the time of type 1 diabetes diagnosis: 2019 cohort 40%; 2020 cohort
53% (p=0·276). Diabetic ketoacidosis severity at the time of type 1 diabetes diagnosis: 2019 cohort six severe cases (12%) of 52;
2020 cohort 11 severe cases (32%) of 34 (p=0·0262).
5 USA ↑ 93 patients with new type 1 diabetes at PICU; more severe acidosis during than before the pandemic (pH 7·10 vs 7·17,
p=0·044).
5 Israel ↑ Individuals younger than 18 years who were newly diagnosed with type 1 diabetes between March 15 (coinciding with the
commencement of the first nationwide lockdown period) and June 30, 2020; and during the same periods in 2019, 2018,
and 2017. Diabetic ketoacidosis incidence was 58·2%, significantly higher than in 2019 (aOR 2·18 [1·31–3·60], p=0·003),
2018 (aOR 2·05 [1·26–3·34], p=0.004), and 2017 (aOR 1·79 [1·09–2·93], p=0·022).
(Table 4 continues on next page)
139
 Review
(Continued from previous page)
Grudziaz-Sekowska and
colleagues53
Hammersen and colleagues40
Han and colleagues93
Hawkes and colleagues94
Ho and colleagues95
Jafari and colleagues41
Jalilova and colleagues96
Kaya and colleagues97
Kiral and colleagues42
Knip and colleagues98
LahTomulic and colleagues43
Lawrence and colleagues99
Lee and colleagues100
Leiva-Gea and colleagues101
Loh and colleagues61
Mameli and colleagues44
Mangus and colleagues102
McCluskey and colleagues103
McGlacken-Byrne and
colleagues45
Miller and colleagues46
140
NOS Country Effect Supporting data
direction
3 Poland ↑ Children and adolescents with type 1 diabetes; compared to the prepandemic period there was a reported substantial increase
(22% compared with 2019) in the number of diabetic ketoacidosis cases.
4 Germany ↔ People with type 1 diabetes younger than 18 years and with a diabetes duration longer than 3 months with visits in 231 German
centers; diabetic ketoacidosis rates in 2020 were not significantly different from those in 2019.
3 Korea ↑ Rate of paediatric patients with diabetic ketoacidosis admitted to the emergency room in 2020 (0·459%) was more than twice
the mean rate of 0·206% for the first 6 months of each of 2017–20.
6 USA ↑ In 73 children with type 1 diabetes, diabetic ketoacidosis presentations in newly diagnosed type 1 diabetes in March–July, 2017,
2018, and 2019 was 38%, and in March–July, 2020 was 45% (p=0·3). Severe diabetic ketoacidosis presentations in newly
diagnosed type 1 diabetes in March–July, 2017, 2018, and 2019 was 12%, and in March–July, 2020 was 15% (p=0·4).
4 Canada ↑ From two children’s hospitals, the frequency of diabetic ketoacidosis at type 1 diabetes onset was significantly higher in the
pandemic period (68·2% in 2020 vs 45·6% in 2019, p<0·001) and incidence of severe diabetic ketoacidosis was also higher
(27·1% in 2020 vs 13·2% in 2019, p=0·01).
5 USA ↓ Using pH as a severity measure, data from one hospital found special cause variation with a downward shift in the mean pH on
diabetic ketoacidosis presentation from 7·2 in 2019 to 7·1 in 2020 for all patients.
6 Türkiye ↑ In 199 children with new-onset type 1 diabetes, it was reported that although the rate of diabetic ketoacidosis was similar, the
rate of severe diabetic ketoacidosis during the pandemic was higher than in previous years (30% in 2020 vs 46% in 2019,
24% in 2018, 19% in 2017, and 17% in 2016; p=0·027), and that durations of symptoms were longer in the pandemic period
than in the previous years (p>0·05)
4 Türkiye ↑ Increased frequency of diabetic ketoacidosis in new-onset type 1 diabetes during (n=44, 68%)compared with before the
pandemic (n=79, 41%); no statistically significant difference in frequency of severe diabetic ketoacidosis.
4 Türkiye ↑ The percentage of children with new-onset type 1 diabetes presenting with diabetic ketoacidosis was higher during the
pandemic (p<0·0001) than before the pandemic. The incidence of severe diabetic ketoacidosis was also higher during the
pandemic (p<0·0001) and was higher among children with new-onset type 1 diabetes (p<0·0001).
6 Finland ↑ Data from the Finnish Pediatric Diabetes Register; children and adolescents with type 1 diabetes, pandemic cohort
228 of 741 (30·8% [27·6–34·2]), control cohort 448 of 1982 (22·6% [20·8–24·5], p<0·001). Severe ketoacidosis p=0·009.
4 Croatia ↑ A statistically significant increase was reported in the proportion of children with new-onset type 1 diabetes presenting with
diabetic ketoacidosis at the PICU during the pandemic.
4 Australia ↑ A statistically significant increase was reported in the frequency of children and adolescents presenting with severe diabetic
ketoacidosis at onset of type 1 diabetes during the pandemic. OR of during compared with before the pandemic 16·7 (95% CI
2·0–194·7; n=11 during pandemic period, 6-10 per year prepandemic).
4 South Korea ↑ Newly diagnosed people with type 1 or type 2 diabetes, aged younger than 18 years: proportion of diabetic ketoacidosis was
higher in 2020 than in 2018–2019 (60.8% vs 39·0%; p=0·038), with more severe initial presentation in 2020 than in
2018–2019 (p=0·040). Newly diagnosed people with type 1 diabetes in 2020 had 2·286 times higher odds (1·190–4·390;
p=0·013) of diabetic ketoacidosis compared with those in 2018–2019; this was not the case for type 2 diabetes.
5 Spain ↑ New onset in paediatric type 1 diabetes. During 2020–2021, the number of patients with diabetic ketoacidosis increased
significantly by 12% (95% CI 7·2–20·4). In this period, 48% of patients with new-onset cases presented diabetic ketoacidosis
(26% mild, 38% moderate, and 36% severe), whereas during the 2015–2019 period, a lower percentage of diabetic ketoacidosis
was reported, 36% (33% mild, 32% moderate, and 34% severe). A higher percentage of patients with moderate and severe
diabetic ketoacidosis was seen in this period, although this increase was not significant.
6 Germany ↑ Children and adolescents who were admitted to hospital during the first lockdown of the COVID-19-pandemic from
March 15 to October, 2020, with new-onset or known diabetes and poor metabolic control, compared with the same period in
2019. The incidence of diabetic ketoacidosis was 50% (pre-lockdown) versus 66·7% (lockdown). Binary logistic regression
calculated a 1·65 times higher odds ratio (95% CI 0·58–4·74) for being admitted with diabetic ketoacidosis in 2020 compared
with 2019, adjusted for age, gender, and BMI.
6 Italy ↑ New paediatric patients (aged 0–17 years) with type 1 diabetes onset in the Lombardy region. Frequency of diabetic
ketoacidosis was higher in 2020 compared with 2017–19 but did not reach statistical significance: 45·3% vs 34·4% in 2017,
42% in 2018, and 34·4% in 2019 (p=0·07).
7 USA ↑ Data from 14 emergency departments; adjusted odds of diabetic ketoacidosis were higher during the pandemic (0·52%) than
prepandemic (0·16%; OR 2·40 [2.07–2·78]).
4 USA ↑ There were 1936 PICU admissions for children with diabetic ketoacidosis in 2020 and 1795 admissions per year to those same
PICUs in 2018–19. The difference between 2020 admissions and 2018–19 admissions was not different from zero before school
closure, and significantly higher than zero after school closure, but was significantly increased in 2020 at more than 30 days
after school closure (p=0·039).
6 UK ↑ Newly diagnosed type 1 diabetes in London hospitals. Children presented more frequently with diabetic ketoacidosis during
the first COVID-19 wave compared with prepandemic (prepandemic: mild 13%, moderate 6·7%, severe 10%; first COVID-19
wave: mild 5·9%, moderate 24%, severe 47%; p=0·002).
7 USA ↑ Retrospective chart review of 132 children with diabetes admitted to two hospitals for diabetes-related causes; rates of
new-onset diabetes presenting with diabetic ketoacidosis increased from 0·24% in 2018 to 0·96% in 2020 (p=0·0014).
(Table 4 continues on next page)
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

(Continued from previous page)
Monkemoller and
colleagues72
Nagl and colleagues104
Ordooei and colleagues105
Pelletier and colleagues106
Pietrzak and colleagues47
Pillai and colleagues107
Ponmani and colleagues48
Rabbone and colleagues74
Rivero-Martin and
colleagues108
Rusak and colleagues109
Salmi and colleagues49
Sellers and colleagues110
Vorgucin and colleagues111
Zubkiewicz-Kucharska and
colleagues112
Adults and mixed samples
Ayoub and colleagues113
Elbarbary and colleagues114
Godsey and colleagues115
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Review
NOS Country Effect Supporting data
direction
6 Germany ↑ Data of 532 patients with new-onset type 1 diabetes from 216 diabetes centres showed that the risk for diabetic ketoacidosis
increased by 84·7% and the risk for severe diabetic ketoacidosis increased by 45·3% compared with the years 2018 and 2019.
4 Austria ↑ National dataset of all newly diagnosed children with type 1 diabetes aged under 15 years. During the lockdown period caused
by the COVID-19 pandemic in 2020, a significant increase of 17·2% was observed in onset-diabetic ketoacidosis prevalence
(p=0·022) from 42·1% during non-lockdown time periods from the five previous years to 59·3%. 20% of children had severe
diabetic ketoacidosis at type 1 diabetes diagnosis during the lockdown period, compared with 14% during the comparison
period.
5 Iran ↑ Dataset from one hospital of children with diabetic ketoacidosis, with significantly more severe diabetic ketoacidosis cases
reported during the pandemic than before (35·7% vs 21·2%), whereas the frequency of moderate diabetic ketoacidosis cases
was not noticeably different in the two periods (37·1% vs 36·4%).
4 USA ↑ Paediatric admissions in 39 US hospitals for diabetic ketoacidosis between January and June 2020 were above the model
95% CI for previous years.
4 Poland ↑ In regional paediatric diabetes centres there were 521 diabetic ketoacidosis cases noted in 2019–2020 (37·5%), and 826 in
2020–2021 (49·4%). The increase was significant both at the patient level (p<0·0001) and at the observation region level (mean
increase across the regions 11·1% [SD 7·2%], p=0·0001). At the national level (considering all new-onset type 1 diabetes cases
diagnosed during each month), the major increase was noted in February, 2020, denoting a shift in diabetic ketoacidosis
incidence (p=0·00008, significant even after considering Bonferroni-adjusted threshold of 0·0022). A month later, a nationwide
lockdown was introduced, during which the observed diabetic ketoacidosis incidence reached its peak (60%). Diabetic
ketoacidosis severity overall, n=1332 (p=0·0089) (difference between prepandemic and during pandemic); mild 2019–2020:
288 (55·4%), 2020–2021: 392 (47·8%), difference –7·8%; moderate 2019–2020: 125 (24·5%), 2020–2021: 246 (29·9%),
difference +5·4%.
5 USA ↑ From one children’s hospital, diabetic ketoacidosis at diagnosis in 2018: eight (24%); 2019: nine (35%); 2020: nine (60%; p=0·048).
5 UK and ↑ Children aged 6 months to 16 years presenting to 49 emergency departments with diabetic ketoacidosis; increases were
Ireland observed in children presenting with new-onset diabetes in diabetic ketoacidosis (395 in March 1, 2020–Feb 28, 2021, to 566
in March 1, 2019–Feb 28, 2020; 43% increase) and severe diabetic ketoacidosis (141 to 252, 79% increase).
6 Italy ↓ Children with type 1 diabetes across 53 centres: 86 (41% of those presenting with type 1 diabetes) had diabetic ketoacidosis at
onset of type 1 diabetes in February–April, 2019, and 61 (38%) in February–April, 2020 (p=0·08). There were 22 diabetic
ketoacidosis episodes in patients with established diabetes in February–April, 2019, and 13 in February–April, 2020 (p value not
reported) Proportion of diabetic ketoacidosis patients with severe diabetic ketoacidosis at onset February–April, 2019,
was 31 (36%); that at onset February–April, 2020 was 27 (44%; p=0·03).
6 Spain ↑ New-onset type 1 diabetes data from seven hospitals showed differences with respect to previous years, with ketoacidosis
present in 39·5% of patients in 2018 and 26·5% in 2019 compared with 52·5% in 2020 (p<0·01).
6 Poland ↑ Patients with type 1 diabetes in one paediatric centre, comparing the same period in 2020 with 2019: number with diabetic
ketoacidosis was 35·2% in 2019 and 47·5% in 2020 (p=0·005); number with severe diabetic ketoacidosis was 14·1% in 2019
and 18·4% in 2020 (p=0·118).
6 Finland ↑ Children on national diabetes registry with type 1 diabetes. Number of children with severe diabetic ketoacidosis (blood
pH<7·10) was 20 (9%) in 2016–19 and 13 (16%) in 2020 (p=0·10).
5 Canada ↑ Eight paediatric centres, new-onset type 1 diabetes. The percentage of children presenting with diabetic ketoacidosis was
significantly higher in 2020 compared with the same period in 2019 (55·0% [143/260] versus 36·4% [86/236]; p<0·0001).
In addition, there was also an increase in the percentage of children presenting with severe diabetic ketoacidosis in the 2020
period compared with the same period in 2019 (69 [48%] of 143 vs 29 [33%] of 860, p=0·044).
7 Serbia ↑ Data on type 1 diabetes from four regional centres. Diabetic ketoacidosis frequency during the pre-COVID-19 period 2017–19
(45 [34%] of 132) is statistically different from diabetic ketoacidosis frequency during the period 2020–21 (42 [42%] of 99;
z=−6·28, p<0·001). Mild diabetic ketoacidosis: 2017–19 19 (14%), 2020–21 14 (14%). Moderate diabetic ketoacidosis:
prepandemic 10 (8%), during pandemic 13 (13%). Severe diabetic ketoacidosis: 2017–19 16 (12%), 2020–21 15 (15%).
5 Poland ↔ Type 1 diabetes registry. In the first 4 months of 2020, diabetic ketoacidosis was associated with the diagnosis of diabetes in
36·67% of patients, similarly to previous years (p=0·69), including 2019 (p=0·94). Diabetic ketoacidosis was present in 50% of
patients diagnosed with type 1 diabetes in March and April, 2020; nevertheless, its incidence was similar to previous years
(p=0·36), including 2019 (p=0·75).
6 Saudi Arabia ↔ Using 285 medical records from one emergency department, not statistically significant relationship between the number of
patients with diabetic ketoacidosis and the period of visitation to the emergency department.
3 75 countries ↔ Health-care professionals reported increases in new-onset diabetes with diabetic ketoacidosis during compared with before the
pandemic: 44 (15%) of 297 reported higher incidence, 253 (85%) did not.
3 USA ↔ Overall, the authors report no statistically significant difference in ICU admissions with diabetic ketoacidosis during the
pandemic compared to prepandemic years. They also projected the number of admissions during the pandemic period had the
pandemic not happened, and compared this with observed rates during the pandemic. They report a possible reduction of
22% (p=0·014).
(Table 4 continues on next page)
141
 Review

NOS Country Effect Supporting data

(Continued from previous page)
Hammersen and colleagues35
Lavik and colleagues116
Lockhart and colleagues117
Lui and colleagues56
Misra and colleagues118
Misra and colleagues70
Nowak and colleagues32
Raucci and colleagues119
Reschen and colleagues59
Wallett and colleagues120
aOR=adjusted odds ratio. ICU=intensive care unit. IRR=incident risk ratio. NOS=Newcastle-Ottawa score. OR=odds ratio. PICU=paediatric intensive care unit. RR=rate ratio.
direction
6 Germany ↓ Data from a large German registry of people with type 1 diabetes; event rates for diabetic ketoacidosis before and during the
pandemic did not differ significantly (p=0·91).
4 USA ↑ From seven hospitals, mixed children and adults; the percentage of total patients with type 1 diabetes (established or newly
diagnosed) presenting in diabetic ketoacidosis month over month in 2019 versus 2020 saw a statistically significant (p=0·001)
increase during the pandemic.
2 Ireland ↔ Young Adult clinic, patients aged 16–24 years with type 1 diabetes; no clear evidence of difference in one hospital
(four reported patients with diabetic ketoacidosis in the pre-lockdown period and three in the lockdown period; p=0·70).
5 Hong Kong ↔ In adults with any type of diabetes, diabetic ketoacidosis hospitalisation rate in January–April, 2019 was 95, in
October, 2019–January, 2020 was 118, and in January–April, 2020 was 96 (p value not reported).
8 UK ↓ In 108 adults with type 1 diabetes, diabetic ketoacidosis cases in February–May, 2017–19 on average were 44 (32–59), and in
February–May, 2020 were 43 (31–58).
7 UK ↑ National dataset of types 1 and 2 diabetes; compared with preceding years, diabetic ketoacidosis rates were 6% (95% CI 4–9;
p<0·0001) higher in the first wave of the pandemic (n=8048), 6% (3–8; p<0·0001) higher in the post-first wave (n=8260), and
7% (4–9; p<0·0001) higher in the second wave (n=9610).
4 Poland ↔ In one hospital, the incidence of diabetic ketoacidosis episodes did not statistically differ between prepandemic (39%, n=9) and
during the pandemic (31%, n=5).
4 Italy ↔ There were 11 cases of ketoacidosis in the prepandemic period: four cases in patients with known type 1 diabetes and
seven cases in patients with new-onset diabetes. In the pandemic there were three cases of ketoacidosis, all new-onset
diabetes.
4 UK ↔ Emergency department admissions in one hospital for diabetic ketoacidosis in adults: 23 prepandemic, 27 in the first pandemic
wave (p=0·58).
4 England, UK ↔ All patients treated for diabetic ketoacidosis at hospital system: “No significant difference in the severity of diabetic
ketoacidosis at presentation (median for COVID-positive, COVID-negative, and pre-COVID groups): pH (7·15 vs 7·2 vs 7·2),
bicarbonate (11·4 mmol/L vs 11 mmol/L vs 13·3 mmol/L), glucose (25·85 mmol/L vs 30·9 mmol/L vs 29·1 mmol/L), lactate
(2·7 mmol/L vs 3·2 mmol/L vs 2·8 mmol/L), serum osmolality (314·6 mmol/L vs 323·1 mmol/L vs 316·2 mmol/L). There was
also no significant difference between the groups for the duration of diabetic ketoacidosis (from the time of admission to
resolution of diabetic ketoacidosis: 12·5 h vs 14·9 h vs 17·9 h for COVID-positive, COVID-negative, and pre-COVID groups,
respectively.” p values not reported.

Table 4: Diabetic ketoacidosis incidence and severity during versus before the COVID-19 pandemic

hypoglycaemia at 53 type 1 diabetes centres in Italy, from
10 cases during lockdown in 2020 compared with 13 in
the same period in 2019. Ruan and colleagues75 (NOS 6),
using data from one hospital trust in England, reported
that the incidence of severe hypoglycaemia was
significantly higher during waves 1 (February–June, 2020)
and 2 (September, 2020–April, 2021) of the pandemic
compared with the prepandemic period (11·7% and
11·5% vs 10·3%). Hammersen and colleagues35 (NOS 6),
using data from a large German registry of people with
type 1 diabetes, reported that rates were similar during
compared with before the pandemic.35
during the pandemic was primarily in those older than
65 years. The authors stated that COVID-19 may have
been a competing endpoint in the study because this
decline was mainly observed in older patients.
Presentations with foot ulcers
Nine studies presented data on patients presenting
with diabetic foot ulcer disease.23,122,129,130,133,136–139 Of the
four studies judged to be at lowest risk of bias,
three showed an increase and one showed no clear
difference; the four judged to be at higher risk of bias
showed a decrease (appendix p 27).

Amputations Sight loss

20 studies provided data on amputations
(appendix p 26).5,23,26,120–135 Of the nine that did not
differentiate by amputation type, three showed an
increase, two showed a decrease, and four indicated no
difference. Within major amputations, nine of 12 studies
showed an increase, two showed a decrease, and
one showed no difference. The pattern was reversed for
minor amputations, with three of the four studies showing
a decrease and the remainder showing an increase.
Valhabjji and colleagues121 (NOS 8, national dataset from
England) noted that the decline in amputations observed
One systematic review by Im and colleagues140
(AMSTAR 2 8/16; critical domains 4/7; appendix p 25)
analysed the impact of delayed antivascular endothelial
growth factor injections on visual acuity in various retinal
diseases, including diabetic macular oedema. They
found a statistically significant decrease in visual acuity
across all diseases, including when restricting just to
studies in people with diabetic macular oedema.
Six primary studies also reported mixed findings
(appendix p 28);141–146 three reported an increase in sight
loss, two did not detect a difference, and one found

142 www.thelancet.com/diabetes-endocrinology Vol 12 February 2024


a decrease. There were no clear patterns across the
six studies.
Discussion
This systematic review found generally worse clinical
outcomes in people with diabetes during the pandemic
compared with prepandemic rates. All-cause and diabetes-
related mortality showed consistent increases during
compared with before the pandemic, and most studies
indicated increases in major amputations and sight loss.
In adult and mixed samples, data generally suggested no
difference in overall amputations or in diabetic ketoacidosis
frequency or severity during compared with before the
pandemic, whereas in children and adolescents most
studies showed increases in diabetic ketoacidosis. This
increase was more pro­nounced in people with new-onset
diabetes compared with those with a previous history of
diabetes, but data suggest overall increases observed were
not solely driven by new-onset diabetes in this population.
Within people with new-onset type 1 diabetes, the
proportion with diabetic ketoacidosis was increased
compared with previous years (ie, the increase does not
seem to have been solely driven by observed increases in
type 1 diabetes diagnoses during this period). Hospital
admissions data suggested decreases in the adult
population with diabetes, but increases in diabetes-related
admissions to paediatric intensive care units. Data were
equivocal on diabetic foot ulcer presentations and
emergency department admissions.
Few (nine of 138) studies analysed trends based on
sociodemographic characteristics. Where reported, data
suggested the impact of the pandemic on diabetes
outcomes was worse for females (four studies) and racial
and ethnic minority groups (four studies). Data were
mixed on age trends, but single studies suggested
outcomes might be worse for younger adults and for
younger children.147
Strengths of our study include the comprehensive
search strategy and large number of included studies.
The limitations are that all but five studies were judged to
be at risk of bias, reporting was often poor—including no
differentiation between diabetes types—and we
were unable to rule out publication bias. Some studies
were very small, meaning they were underpowered but
also that they could result in extreme point estimates.
There might be some overlap between studies (eg, a study
using a national dataset might overlap with a study using
a regional dataset in the same country), and some of the
primary studies are included in systematic reviews. Due
to heterogeneity in reported outcome data, it was not
possible to complete meta-analyses. Additionally, the
included study timepoints varied between comparing
either immediately prepan­ demic or the same time
point in the preceding year(s), making interpretation
challenging. Similarly, due to the challenging nature of
defining a postpandemic period (considering that
incidence is ongoing), no studies have compared during
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Review
the pandemic with postpandemic and, therefore, some
downstream consequences of changes in health-care use
are likely to be underestimated. Different countries
incorporated different lengths and degrees of lockdowns,
resulting in varying impacts on health-care provision and
access. We found no eligible randomised controlled trials
for inclusion in the review, and so cannot comment on
any preventative efforts. Finally, most studies were in
high-income countries, and their findings might not be
generalisable to low-income or middle-income countries.
Interpreting some health-care use outcomes is chal­
lenging. For instance, some studies observed reductions
in the number of severe hypoglycaemic events, which
could be due to either lower occurrence or avoidance of
treatment due to the pandemic. Given the negative impact
of the pandemic on factors such as medication
adherence,148 lifestyle behaviours,149 mental health,150 and
disease self-management,151 it is likely that these events
were occurring at the same (or a greater) rate, but that they
were less likely to result in admission or hospitalisation
due to COVID-19 pressures on the health-care system.
This would be consistent with observed increases in
diabetic ketoacidosis severity in some studies; patients
might not have accessed health care (whether by choice or
not) until their condition deteriorated, resulting in greater
long-term health-care use. However, it is also possible that
recent acceleration in the use of digital diabetes man­
agement might have improved diabetic control; a recent
systematic review found improved glycaemic control in
people with type 1 diabetes during lockdown attributed to
increased use of continuous glucose monitors,152 which
would be expected to reduce hypoglycaemia-related events
or admissions. Other studies have reported changes to
underlying risk factors during the pandemic, including
increased bodyweight and increases in blood pressure and
lipids, which might also contribute to meaningful
increases in diabetes-related complications.153,154
These findings have important implications for health-
care service provision during future pandemics, and for
prioritising patient care during the recovery phase.155
People considered at higher risk should be prioritised, but
health care should be provided to all people with diabetes
as required. Primary data is limited, but suggestions
include: using telemedicine and digital health platforms to
provide patient care and facilitate self-monitoring and peer
support; providing patients with emergency medication
kits, including extended diabetes supplies and medicines;
webinars about diabetes management during emergen­
cies; home-based exercise and structured education
programmes; mental health support through virtual
counselling and helplines for people with diabetes feeling
overwhelmed; nutritional support through easy, diabetes-
friendly recipes using non-perishable items; community
volunteers to support tasks like medication pickups;
flexible medication delivery options from pharmacies; and
emergency preparedness workshops teaching people with
diabetes how to prepare for emergencies in collaboration
143
 Review
with local health departments.156 Without such efforts,
risks include downstream worsening of health outcomes
and a delayed surge in health-care use. The full impact of
the observed reduction in health-care use will become even
more apparent in the coming years. Future research must
investigate the impact of the pandemic on long-term
condition incidence, management, and outcomes in order
to support future preventative strategies, particularly in
underserved communities.157
In summary, this systematic review found that the
COVID-19 pandemic and associated disruptions in
health-care provision were associated with mixed trends
in hospitalisation and generally increased diabetes-related
morbidity and mortality. Future research should focus on
investigating the longer-term impact of the pandemic.
Further studies are also needed to understand potential
differential impacts of these disruptions, which risk
further exacerbating existing inequalities within people
with diabetes.
Contributors
JHB and KK conceived the review. JHB, PH, KR, JS, FC, LOM, EM, LK,
SS, and KK designed and approved the protocol. PH, KR, IO, JS, FC,
LOM, LK, JM, RL, SM, EM, and JHB contributed to screening. PH, KR,
IO, JS, FC, LOM, LK, JM, RL, SM, EM, and JHB contributed to data
extraction and appraisal. JHB and PH led the write-up of the review and
verified the data in the review. All authors edited and approved the final
manuscript, had full access to all the data in the review, and were
responsible for the decision to submit for publication.
Declaration of interests
SS reports receiving grants from Sanofi, AstraZeneca, Novo Nordisk,
Boehringer Ingelheim, and Servier; speaker honoraria from Boehringer
Ingelheim, Lilly, MSD, Novartis, Novo Nordisk, Janssen, Amgen, Sanofi,
AstraZeneca, Abbott, and Menarini; advisory board honoraria from
Boehringer Ingelheim, Lilly, Novo Nordisk, AstraZeneca, and Abbott;
and support for attending meetings or travel from AstraZeneca, Abbott,
and Menarini. KK reports acting as a consultant or speaker for or
receiving grants for investigator-initiated studies from AstraZeneca,
Abbott, Amgen, Bayer, Novartis, Novo Nordisk, Roche, Servier,
Sanofi-Aventis, Lilly, MSD, Boehringer Ingelheim, Oramed
Pharmaceuticals, and Applied Therapeutics; and being chair of the
Scientific Advisory Group for Emergencies Ethnicity Subgroup. All other
authors declare no competing interests.
Data sharing
Completed data extraction sheets are available upon reasonable request
to the corresponding author (note all data have been published so are
also in the public domain).
Acknowledgments
WHO commissioned and financially supported this work. This research
was also funded by the NIHR Applied Research Collaboration Oxford
and Thames Valley at Oxford Health NHS Foundation Trust. KK, PH,
LO, JS, and SIS are supported by the NIHR Applied Research
Collaboration East Midlands, MedTech In Vitro Diagnostic Co-operative,
and NIHR Leicester Biomedical Research Centre. EM is supported by a
Wellcome Trust Clinical Doctoral Fellowship. The views expressed here
are those of the authors and do not necessarily represent those of the
funders. We thank Nia Roberts, subject librarian at the University of
Oxford, for facilitating our searches.
References
1 Abdi A, Jalilian M, Sarbarzeh PA, Vlaisavljevic Z. Diabetes and
COVID-19: a systematic review on the current evidences.
Diabetes Res Clin Pract 2020; 166: 108347.
2 Mukona DM, Zvinavashe M. Self-management of diabetes mellitus
during the COVID-19 pandemic: recommendations for a resource
limited setting. Diabetes Metab Syndr 2020; 14: 1575–78.
144
3 Ruiz-Roso MB, Knott-Torcal C, Matilla-Escalante DC, et al.
COVID-19 lockdown and changes of the dietary pattern and
physical activity habits in a cohort of patients with type 2 diabetes
mellitus. Nutrients 2020; 12: 2327.
4 Alessi J, De Oliveira GB, Franco DW, et al. Mental health in the era
of COVID-19: prevalence of psychiatric disorders in a cohort of
patients with type 1 and type 2 diabetes during the social distancing.
Diabetol Metab Syndr 2020; 12: 1–10.
5 Caruso P, Longo M, Signoriello S, et al. Diabetic foot problems
during the COVID-19 pandemic in a tertiary care center: the
emergency among the emergencies. Diabetes Care 2020;
43: e123–24.
6 Chatziralli I, Dimitriou E, Kazantzis D, Machairoudia G,
Theodossiadis G, Theodossiadis P. Effect of COVID-19-associated
lockdown on patients with diabetic retinopathy. Cureus 2021;
13: e14831.
7 Lawrence C, Seckold R, Smart C, et al. Increased paediatric
presentations of severe diabetic ketoacidosis in an Australian
tertiary centre during the COVID-19 pandemic. Diabetic Med 2021;
38: e14417.
8 Peterson J, Welch V, Losos M, Tugwell P. The Newcastle-Ottawa
scale (NOS) for assessing the quality of nonrandomised studies in
meta-analysis. Ottawa: Ottawa Hospital Research Institute,
2011: 1–12.
9 Shea BJ, Reeves BC, Wells G, et al. AMSTAR 2: a critical appraisal
tool for systematic reviews that include randomised or
non-randomised studies of healthcare interventions, or both.
BMJ 2017; 358: j4008.
10 Campbell M, McKenzie JE, Sowden A, et al. Synthesis without
meta-analysis (SWiM) in systematic reviews: reporting guideline.
BMJ 2020; 368: l6890.
11 Cochrane Training. Cochrane Handbook Chapter 12: Synthesizing
and presenting findings using other methods. 2021. https://
training.cochrane.org/handbook/current/chapter-12 (accessed
June 11, 2023).
12 Boon MH, Thomson H. The effect direction plot revisited:
application of the 2019 Cochrane Handbook guidance on alternative
synthesis methods. Res Synth Methods 2021; 12: 29–33.
13 Valabhji J, Barron E, Gorton T, et al. Associations between
reductions in routine care delivery and non-COVID-19-related
mortality in people with diabetes in England during the COVID-19
pandemic: a population-based parallel cohort study.
Lancet Diabetes Endocrinol 2022; 10: 561–70.
14 McCoy RG, Mullan AF, Jeffery MM, Bucks CM, Clements CM,
Campbell RL. Excess all-cause and cause-specific mortality among
people with diabetes during the COVID-19 pandemic in
Minnesota: population-based study. J Gen Intern Med 2022;
37: 3228–31.
15 Torre E, Colombo GL, Di Matteo S, et al. Economic impact of
COVID-19 lockdown on Italian NHS: focus on diabetes mellitus.
Clinicoecon Outcomes Res 2021; 13: 503–18.
16 McAlister FA, Hsu Z, Dong Y, Tsuyuki RT, van Walraven C,
Bakal JA. Frequency and type of outpatient visits for patients with
cardiovascular ambulatory-care sensitive conditions during the
COVID-19 pandemic and subsequent outcomes: a retrospective
cohort study. J Am Heart Assoc 2023; 12: e027922.
17 Khaydarova F, Alieva A, Berdikulova D, Alimova N, Khalilova D,
Tojieva I. IDF21-0090 Structure of mortality among patients with
diabetes mellitus in the Republic of Uzbekistan during COVID-19
pandemia. Diabetes Mellit 2022; 25: 322–26.
18 Harun RT, Almohammadi AA, Alnashri MM, Alsamiri S,
Alkhatieb M. Impact of COVID-19 crisis in the management of
diabetic foot patients in King Abdulaziz University Hospital,
Jeddah, Saudi Arabia. Cureus 2023; 15: e36613.
19 Bello-Chavolla OY, Antonio-Villa NE, Fermin-Martinez CA, et al.
Diabetes-related excess mortality in Mexico: a comparative analysis
of national death registries between 2017–2019 and 2020.
Diabetes Care 2022; 45: 2957–66.
20 Hernandez-Vasquez A, Barrenechea-Pulache A,
Portocarrero-Bonifaz A, Rojas-Roque C, Gamboa-Unsihuay JE.
Multimorbidity analysis and hospitalizations for diabetes before and
after lockdown due to the COVID-19 pandemic in Peru.
Prev Med Rep 2022; 28: 101884.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

21 Kleibert M, Mrozikiewicz-Rakowska B, Bak PM, Balut D,
Zielinski J, Czupryniak L. Breakdown of diabetic foot ulcer care
during the first year of the pandemic in Poland: a retrospective
national cohort study. Int J Environ Res Public Health 2022; 19: 3827.
22 Laliotis I, Stavropoulou C, Ceely G, Brett G, Rushton R. Excess
deaths by cause and place of death in England and Wales during the
first year of COVID-19. Health Econ 2023; 32: 1982–2005.
23 Lee W-E, Park SW, Weinberger DM, et al. Direct and indirect
mortality impacts of the COVID-19 pandemic in the United States,
March 1, 2020 to January 1, 2022. Elife 2023; 12: e77562.
24 Lv F, Gao X, Huang AH, et al. Excess diabetes mellitus-related
deaths during the COVID-19 pandemic in the United States.
EClinicalMedicine 2022; 54: 101671.
25 Lozano-Corona R, Reyes-Monroy JA, Lara-Gonzalez V,
Anaya-Ayala JE, Dardik A, Hinojosa CA. Revascularization prevents
amputation among patients with diabetic foot during the COVID-19
era. Vascular 2022; 31: 729–36.
26 Raknes G, Strom MS, Sulo G, Overland S, Roelants M,
Juliusson PB. Lockdown and non-COVID-19 deaths: cause-
specific mortality during the first wave of the 2020 pandemic in
Norway: a population-based register study. BMJ Open 2021;
11: e050525.
27 Sekowski K, Grudziaz-Sekowska J, Gorynski P, Pinkas J,
Jankowski M. Epidemiological analysis of diabetes-related
hospitalization in Poland before and during the COVID-19
pandemic, 2014–2020. Int J Environ Res Public Health 2022; 19: 10030.
28 Silverio-Murillo A, Balmori de la Miyar JR, Martínez-Alfaro A.
Non-COVID-19 deaths in times of pandemic. J Public Health 2023;
45: e196–203.
29 Todd M, Scheeres A. Excess mortality from non-COVID-19 causes
during the COVID-19 pandemic in Philadelphia, Pennsylvania,
2020–2021. Am J Public Health 2022; 112: 1800–03.
30 Yao XI, Han L, Sun Y, He D, Zhao S, Ran J. Temporal variation of
excess deaths from diabetes during the COVID-19 pandemic in the
United States. J Infect Public Health 2023; 16: 483–89.
31 Bossi CB, d’Oro LC, Derosa G, et al. COVID-19 pandemic impact
on people with diabetes: results from a large representative sample
of Italian older adults. Prim Care Diabetes 2022; 16: 650–57.
32 Nowak Z, Gawlik J, Wedrychowicz A, Nazim J, Starzyk J.
The incidence and causes of acute hospitalizations and emergency
room visits in adolescents with type 1 diabetes mellitus prior to
and during the COVID-19 pandemic: a single-centre experience.
Pediatr Endocrinol Diabetes Metab 2023; 29: 22–29.
33 Tehrani TH, Razavi Z, Salimi S, Farahi H, Bazmamoun H,
Soltanian AR. Impact of coronavirus disease 2019 outbreak on
children and adolescents with type 1 diabetes mellitus.
J Res Health Sci 2021; 21: e00534.
34 Yoon J, Chen C, Chao S, Wong E, Rosland A-M. Adherence to
diabetes medications and health care use during the COVID-19
pandemic among high-risk patients. J Am Board Fam Med 2023;
36: 289–302.
35 Hammersen J, Tittel SR, Khodaverdi S, et al. Metabolic control
during the first two years of the COVID-19 pandemic in pediatric
patients with type 1 diabetes: results from the German DPV
initiative. Acta Diabetol 2023; 60: 757–66.
36 Breinig S, Mortamet G, Brossier D, et al. Impact of the French
national lockdown on admissions to 14 pediatric intensive care
units during the 2020 COVID-19 pandemic—a retrospective
multicenter study. Front Pediatr 2021; 9: 764583.
37 Choudhary A, Adhikari S, White PC. Impact of the COVID-19
pandemic on management of children and adolescents with type 1
diabetes. BMC Pediatr 2022; 22: 124.
38 Goldman S, Pinhas-Hamiel O, Weinberg A, et al. Alarming increase
in ketoacidosis in children and adolescents with newly diagnosed
type 1 diabetes during the first wave of the COVID-19 pandemic in
Israel. Pediatr Diabetes 2022; 23: 10–18.
39 Güemes M, Storch-de-Gracia P, Enriquez SV, Martín-Rivada Á,
Brabin AG, Argente J. Severity in pediatric type 1 diabetes mellitus
debut during the COVID-19 pandemic. J Pediatr Endocrinol Metab
2020; 33: 1601–03.
40 Hammersen J, Reschke F, Tittel SR, et al. Metabolic control during
the SARS-CoV-2 lockdown in a large German cohort of pediatric
patients with type 1 diabetes: results from the DPV initiative.
Pediatr Diabetes 2022; 23: 351–61.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Review
41 Jafari K, Koves I, Rutman L, Brown JC. Impact of the COVID-19
pandemic on the severity of diabetic ketoacidosis presentations in
a tertiary pediatric emergency department. Pediatr Qual Saf 2022;
7: e502.
42 Kiral E, Kirel B, Havan M, et al. Increased severe cases and
new-onset type 1 diabetes among children presenting with diabetic
ketoacidosis during first year of COVID-19 pandemic in Turkey.
Front Pediatr 2022; 10: 926013.
43 Lah Tomulic K, Matko L, Verbic A, et al. Epidemiologic
characteristics of children with diabetic ketoacidosis treated in
a pediatric intensive care unit in a 10-year-period: single centre
experience in Croatia. Medicina 2022; 58: 638.
44 Mameli C, Scaramuzza A, Macedoni M, et al. Type 1 diabetes onset
in Lombardy region, Italy, during the COVID-19 pandemic:
the double-wave occurrence. EClinicalMedicine 2021; 39: 101067.
45 McGlacken-Byrne SM, Drew SEV, Turner K, Peters C, Amin R.
The SARS-CoV-2 pandemic is associated with increased severity of
presentation of childhood onset type 1 diabetes mellitus: a multi-
centre study of the first COVID-19 wave. Diabetic Med 2021;
38: e14640.
46 Miller A, Joseph S, Badran A, Umpaichitra V, Bargman R, Chin VL.
Increased rates of hospitalized children with type 1 and type 2
diabetes mellitus in central Brooklyn during the COVID-19
pandemic. Int J Pediatr 2023; 2023: 4580809.
47 Pietrzak I, Michalak A, Seget S, et al. Diabetic ketoacidosis
incidence among children with new-onset type 1 diabetes in Poland
and its association with COVID-19 outbreak—two-year cross-
sectional national observation by PolPeDiab study group.
Pediatr Diabetes 2022; 23: 944–55.
48 Ponmani C, Nijman RG, Roland D, et al. Children presenting with
diabetes and diabetic ketoacidosis to emergency departments during
the COVID-19 pandemic in the UK and Ireland: an international
retrospective observational study. Arch Dis Child 2023; 108: 799–807.
49 Salmi H, Heinonen S, Hastbacka J, et al. New-onset type 1 diabetes
in Finnish children during the COVID-19 pandemic. Arch Dis Child
2022; 107: 180–85.
50 Zee-Cheng JE, McCluskey CK, Klein MJ, et al. Changes in pediatric
ICU utilization and clinical trends during the coronavirus
pandemic. Chest 2021; 160: 529–37.
51 Cassell K, Zipfel CM, Bansal S, Weinberger DM. Trends in
non-COVID-19 hospitalizations prior to and during the COVID-19
pandemic period, United States, 2017–2021. Nat Commun 2022;
13: 5930.
52 Dayal D, Gupta S, Raithatha D, Jayashree M. Missing during
COVID-19 lockdown: children with new-onset type 1 diabetes. 2020;
109: 2144–46.
53 Grudziaz-Sekowska J, Sekowski K, Kobuszewski B. Healthcare
utilization and adherence to treatment recommendations among
children with type 1 diabetes in Poland during the COVID-19
pandemic. Int J Environ Res Public Health 2022; 19: 4798.
54 Guimaraes RA, Policena GM, Paula H, et al. Analysis of the
impact of coronavirus disease 19 on hospitalization rates for chronic
non-communicable diseases in Brazil. PLoS One 2022; 17: e0265458.
55 Kim Y, Gordon A, Rowerdink K, Herrera Scott L, Chi W. The impact
of the COVID-19 pandemic on health care utilization among
insured individuals with common chronic conditions. Med Care
2022; 60: 673–79.
56 Lui DTW, Lee CH, Chow WS, et al. A territory-wide study on the
impact of COVID-19 on diabetes-related acute care.
J Diabetes Investig 2020; 11: 1303–06.
57 Michalowsky B, Hoffmann W, Bohlken J, Kostev K. Effect of the
COVID-19 lockdown on disease recognition and utilisation of
healthcare services in the older population in Germany:
a cross-sectional study. Age Ageing 2021; 50: 317–25.
58 Moin JS, Troke N, Plumptre L, Anderson GM. Impact of the
COVID-19 pandemic on diabetes care for adults with type 2
diabetes in Ontario, Canada. Can J Diabetes 2022; 46: 715–21.
59 Reschen ME, Bowen J, Novak A, et al. Impact of the COVID-19
pandemic on emergency department attendances and acute
medical admissions. BMC Emerg Med 2021; 21: 143.
60 Santana YE, Liberatore RDRJ. Teleconsultation for pediatric
patients with type 1 diabetes mellitus during the COVID-19
pandemic: experience of a university hospital in Brazil.
J Pediatr (Rio J) 2022; 98: 587–89.
145
 Review
61 Loh C, Weihe P, Kuplin N, Placzek K, Weihrauch-Bluher S.
Diabetic ketoacidosis in pediatric patients with type 1- and type 2
diabetes during the COVID-19 pandemic. Metabolism 2021;
122: 154842.
62 Aubert CE, Henderson JB, Kerr EA, Holleman R, Klamerus ML,
Hofer TP. Type 2 diabetes management, control and outcomes
during the COVID-19 pandemic in older US veterans:
an observational study. J Gen Intern Med 2022; 37: 870–77.
63 Celona CA, Jackman K, Smaldone A. Emergency department use
by young adults with chronic illness before and during the
COVID-19 pandemic. J Emerg Nurs 2023; 49: 755–64.
64 Giannouchos TV, Biskupiak J, Moss MJ, Brixner D, Andreyeva E,
Ukert B. Trends in outpatient emergency department visits during
the COVID-19 pandemic at a large, urban, academic hospital
system. Am J Emerg Med 2021; 40: 20–26.
65 Sarikaya E, Cicek D, Gok E, et al. The effect of the COVID-19
pandemic on metabolic control in children with type 1 diabetes:
a single-center experience. J Pediatr Endocrinol Metab 2022;
35: 191–95.
66 Alfayez OM, Aldmasi KS, Alruwais NH, et al. Incidence of diabetic
ketoacidosis among pediatrics with type 1 diabetes prior to and
during COVID-19 pandemic: a meta-analysis of observational
studies. Front Endocrinol 2022; 13: 856958.
67 Elgenidy A, Awad AK, Saad K, et al. Incidence of diabetic
ketoacidosis during COVID-19 pandemic: a meta-analysis of
124,597 children with diabetes. Pediatr Res 2022; 93: 1149–60.
68 Rahmati M, Keshvari M, Mirnasuri S, et al. The global impact of
COVID-19 pandemic on the incidence of pediatric new-onset type 1
diabetes and ketoacidosis: a systematic review and meta-analysis.
J Med Virol 2022; 94: 5112–27.
69 O’Mahoney LL, Highton PJ, Kudlek L, et al. The impact of the
COVID-19 pandemic on glycaemic control in people with diabetes:
a systematic review and meta-analysis. Diabetes Obes Metab 2022;
24: 1850–60.
70 Misra S, Barron E, Vamos E, et al. Temporal trends in emergency
admissions for diabetic ketoacidosis in people with diabetes in
England before and during the COVID-19 pandemic:
a population-based study. Lancet Diabetes Endocrinol 2021;
9: 671–80.
71 Alassaf A, Gharaibeh L, Ibrahim S, et al. Effect of COVID-19
pandemic on presentation and referral patterns of newly diagnosed
children with type 1 diabetes in a developing country.
J Pediatr Endocrinol Metab 2022; 35: 859–66.
72 Monkemoller K, Kamrath C, Hammersen J, et al. Is it possible to
prevent diabetic ketoacidosis at diagnosis of pediatric type 1
diabetes? Lessons from the COVID-19 pandemic.
Monatsschr Kinderheilkd 2021; 169: 451–60 (in German).
73 Alsalman AA, Aldossari MR, Alomani ZD, et al. Impact of
coronavirus disease lockdown on children with type 1 diabetes
mellitus in Al-Khobar, Saudi Arabia. Cureus 2022; 14: e21350.
74 Rabbone I, Schiaffini R, Cherubini V, Maffeis C, Scaramuzza A.
Has COVID-19 delayed the diagnosis and worsened the
presentation of type 1 diabetes in children? Diabetes Care 2020;
43: 2870–72.
75 Ruan Y, Mercuri L, Papadimitriou D, et al. Increase in
hypoglycaemia and hyperglycaemia in people with diabetes
admitted to hospital during COVID-19 pandemic.
J Diabetes Complications 2023; 37: 108474.
76 Abdou M, Hassan MM, Hassanein SA, Elsebaie EH, Shamma RA.
Presentations, complications, and challenges encountered during
management of type 1 diabetes in Egyptian children during
COVID-19 pandemic: a single-center experience. Front Endocrinol
2022; 13: 814991.
77 Al Agha AE, Alharbi RS, Almohammadi OA, Yousef SY,
Sulimani AE, Alaama RA. Impact of COVID-19 lockdown on
glycemic control in children and adolescents. Saudi Med J 2021;
42: 44–48.
78 Alaqeel A, Aljuraibah F, Alsuhaibani M, et al. The impact of
COVID-19 pandemic lockdown on the incidence of new-onset type 1
diabetes and ketoacidosis among Saudi children. Front Endocrinol
2021; 12: 669302.
79 Ansar A, Livett T, Beaton W, Carrel AL, Bekx MT. Sharp rise in
new-onset pediatric diabetes during the COVID-19 pandemic. WMJ
2022; 121: 177–80.
146
80 Baechle C, Eckert A, Kamrath C, et al. Incidence and presentation of
new-onset type 1 diabetes in children and adolescents from Germany
during the COVID-19 pandemic 2020 and 2021: current data from
the DPV registry. Diabetes Res Clin Pract 2023; 197: 110559.
81 Basatemur E, Jones A, Peters M, Ramnarayan P. Paediatric critical
care referrals of children with diabetic ketoacidosis during the
COVID-19 pandemic. Arch Dis Child 2021; 106: e21.
82 Birkebaek NH, Kamrath C, Grimsmann JM, et al. Impact of the
COVID-19 pandemic on long-term trends in the prevalence of
diabetic ketoacidosis at diagnosis of paediatric type 1 diabetes:
an international multicentre study based on data from 13 national
diabetes registries. Lancet Diabetes Endocrinol 2022; 10: 786–94.
83 Boboc AA, Novac CN, Ilie MT, et al. The impact of SARS-CoV-2
pandemic on the new cases of T1DM in children. A single-centre
cohort study. J Pers Med 2021; 11: 551.
84 Bogale KT, Urban V, Schaefer E, Bangalore Krishna K. The impact of
COVID-19 pandemic on prevalence of diabetic ketoacidosis at
diagnosis of type 1 diabetes: a single-centre study in central
Pennsylvania. Endocrinol Diabetes Metab 2021; 4: e00235.
85 Botelho TA, Santos JMN, Pinho CMS, et al. Ketoacidosis in new-
onset type 1 diabetes: did the severity increase during the COVID-19
pandemic? J Pediatr Endocrinol Metab 2022; 35: 73–77.
86 Chambers MA, Mecham C, Arreola EV, Sinha M. Increase in the
number of pediatric new-onset diabetes and diabetic ketoacidosis
cases during the COVID-19 pandemic. Endocr Pract 2022;
28: 479–85.
87 Cherubini V, Marino M, Scaramuzza AE, et al. The silent epidemic
of diabetic ketoacidosis at diagnosis of type 1 diabetes in children
and adolescents in Italy during the COVID-19 pandemic in 2020.
Front Endocrinol 2022; 13: 878634.
88 Danne T, Lanzinger S, de Bock M, et al. A worldwide perspective on
COVID-19 and diabetes management in 22,820 children from the
SWEET project: diabetic ketoacidosis rates increase and glycemic
control is maintained. Diabetes Technol Ther 2021; 23: 632–41.
89 d’Annunzio G, Bassi M, De Rose EL, et al. Increased frequency of
diabetic ketoacidosis: the link with COVID-19 pandemic.
Front Clin Diabetes Healthc 2022; 3: 846827.
90 Dovc K, Osredkar SR, Schweiger DS, Battelino T, Bratina N.
Nationwide digital/virtual diabetes care of children, adolescents and
young adults with type 1 diabetes during a COVID-19 pandemic in
Slovenia. Slov Med J 2020; 89: 626–33.
91 Dżygało K, Nowaczyk J, Szwilling A, Kowalska A. Increased
frequency of severe diabetic ketoacidosis at type 1 diabetes onset
among children during COVID-19 pandemic lockdown:
an observational cohort study. Pediatr Endocrinol Diabetes Metab 2020;
26: 167–75.
92 Fathi A, Levine GK, Hicks R, Morphew T, Babbitt CJ. Has variable
access to health care during the COVID-19 pandemic impacted the
severity of paediatric diabetic ketoacidosis? Pract Diabetes 2022;
39: 17–22.
93 Han MJ, Heo JH. Increased incidence of pediatric diabetic
ketoacidosis after COVID-19: a two-center retrospective study in
Korea. Diabetes Metab Syndr Obes 2021; 14: 783–90.
94 Hawkes CP, Willi SM. A trend towards an early increase in
ketoacidosis at presentation of paediatric type 1 diabetes during the
coronavirus-2019 pandemic. Diabet Med 2021; 38: e14461.
95 Ho J, Rosolowsky E, Pacaud D, et al. Diabetic ketoacidosis at type 1
diabetes diagnosis in children during the COVID-19 pandemic.
Pediatr Diabetes 2021; 22: 552–57.
96 Jalilova A, Ata A, Demir G, et al. The effect of the SARS-CoV-2
pandemic on presentation with diabetic ketoacidosis in children
with new onset type 1 diabetes mellitus. J Clin Res Pediatr Endocrinol
2023; 15: 264–67 .
97 Kaya G, Cimbek EA, Yesilbas O, Bostan YE, Karaguzel G. A long-
term comparison of presenting characteristics of children with
newly diagnosed type 1 diabetes before and during the COVID-19
pandemic. J Clin Res Pediatr Endocrinol 2022; 14: 267–74.
98 Knip M, Parviainen A, Turtinen M, et al. SARS-CoV-2 and type 1
diabetes in children in Finland: an observational study.
Lancet Diabetes Endocrinol 2023; 11: 251–60.
99 Lawrence C, Seckold R, Smart C, et al. Increased paediatric
presentations of severe diabetic ketoacidosis in an Australian tertiary
centre during the COVID-19 pandemic. Diabet Med 2021; 38: e14417.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

100 Lee Y, Kim M, Oh K, et al. Comparison of initial presentation of
pediatric diabetes before and during the coronavirus disease 2019
pandemic era. J Korean Med Sci 2022; 37: e176.
101 Leiva-Gea I, Fernandez CA, Cardona-Hernandez R, et al. Increased
presentation of diabetic ketoacidosis and changes in age and
month of type 1 diabetes at onset during the COVID-19 pandemic
in Spain. J Clin Med 2022; 11: 4338.
102 Mangus CW, Parker SJ, DeLaroche AM, et al. Impact of COVID-19
on the associated complications of high-risk conditions in a
statewide pediatric emergency network.
J Am Coll Emerg Physicians Open 2022; 3: e12865.
103 McCluskey CK, Zee-Cheng JE, Klein MJ, et al. The temporal
relationship between local school closure and increased
incidence of pediatric diabetic ketoacidosis. Front Pediatr 2022;
10: 812265.
104 Nagl K, Waldhor T, Hofer SE, et al. Alarming increase of
ketoacidosis prevalence at type 1 diabetes-onset in Austria—results
from a nationwide registry. Front Pediatr 2022; 10: 820156.
105 Ordooei M, Karimi M, Akbarian E, Rasoulizadeh Z. Diabetic
ketoacidosis in children before and during COVID-19 pandemic:
a cross-sectional study. Int J Endocrinol Metab 2023; 21: e132809.
106 Pelletier JH, Rakkar J, Au AK, Fuhrman D, Clark RSB, Horvat CM.
Trends in US pediatric hospital admissions in 2020 compared with
the decade before the COVID-19 pandemic. JAMA Netw Open 2021;
4: e2037227.
107 Pillai SS, Cao C, Drees CJ, Chu TC, Mason K, Topor LS. Delays in
presentation of new onset diabetes at the start of the COVID-19
pandemic. R I Med 2022; 105: 46–50.
108 Rivero-Martin MJ, Rivas-Mercado CM, Cenal-Gonzalez-Fierro MJ,
et al. Severity of new-onset type 1 diabetes in children and
adolescents during the COVID-19 pandemic.
Endocrinol Diabetes Nutr 2022; 69: 810–15 (in Spanish).
109 Rusak E, Seget S, Macherski M, Furgal N, Dys P,
Jarosz-Chobot P. Has the COVID-19 pandemic affected the
prevalence of diabetic ketoacidosis in Polish children with newly
diagnosed type 1 diabetes? An example of the largest Polish
pediatric diabetes center (Upper Silesia-Katowice, Poland).
Healthcare 2022; 10: 348.
110 Sellers EAC, Pacaud D. Diabetic ketoacidosis at presentation of
type 1 diabetes in children in Canada during the COVID-19
pandemic. Paediatr Child Health (Oxford) 2021; 26: 208–09.
111 Vorgucin I, Savin M, Stankovic D, et al. Incidence of type 1 diabetes
mellitus and characteristics of diabetic ketoacidosis in children and
adolescents during the first two years of the COVID-19 pandemic
in Vojvodina. Medicina 2022; 58: 1013.
112 Zubkiewicz-Kucharska A, Seifert M, Stepkowski M, Noczynska A.
Diagnosis of type 1 diabetes during the SARS-CoV-2 pandemic:
does lockdown affect the incidence and clinical status of patients?
Adv Clin Exp Med 2021; 30: 127–34.
113 Ayoub AA, Addas MJ, Alghamdi AA, et al. The effect of COVID-19
lockdown on cerebrovascular accidents, acute coronary syndrome,
and diabetic ketoacidosis visits the emergency department:
a retrospective study. Cureus 2022; 14: e33154.
114 Elbarbary NS, Dos Santos TJ, de Beaufort C, Agwu JC, Calliari LE,
Scaramuzza AE. COVID-19 outbreak and pediatric diabetes:
perceptions of health care professionals worldwide. Pediatr Diabetes
2020; 21: 1083–92.
115 Godsey C, Gabor R, Oelstrom M, et al. Changes in pediatric
intensive care admissions in Wisconsin during the 2020 COVID-19
pandemic. WMJ 2022; 121: 194–200.
116 Lavik AR, Ebekozien O, Noor N, et al. Trends in type 1 diabetic
ketoacidosis during COVID-19 surges at 7 US centers: highest
burden on non-Hispanic Black patients. J Clin Endocrinol Metab
2022; 107: 1948–55.
117 Lockhart M, Green D, Smith D. The impact of COVID-19 lockdown
on glycaemic control in young adults with type 1 diabetes mellitus.
Ir J Med Sci 2022; 192: 671–73.
118 Misra S, Khozoee B, Huang J, et al. Comparison of diabetic
ketoacidosis in adults during the SARS-CoV-2 outbreak and over
the same time period for the preceding 3 years. Diabetes Care 2021;
44: e29–31.
119 Raucci U, Musolino AM, Di Lallo D, et al. Impact of the COVID-19
pandemic on the emergency department of a tertiary children’s
hospital. Ital J Pediatr 2021; 47: 21.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Review
120 Wallett L, Kempegowda P, Melson E, et al. Differences in
presentation, severity and management of DKA in type 1 and type 2
diabetes during the COVID-19 pandemic. Clin Med 2021;
21 (suppl 2): 1–2.
121 Valabhji J, Barron E, Vamos EP, et al. Temporal trends in
lower-limb major and minor amputation and revascularization
procedures in people with diabetes in England during the
COVID-19 pandemic. Diabetes Care 2021; 44: e133–35.
122 Bregovskiy V, Karpova I. Analysis of specialized care for patients
with diabetic foot syndrome in St Petersburg for 2010–2021.
Diabetes Mellit 2022; 25: 477–84.
123 Casciato DJ, Yancovitz S, Thompson J, et al. Diabetes-related major
and minor amputation risk increased during the COVID-19
pandemic. J Am Podiatr Med Assoc 2023; 113: 20–224.
124 de Mestral C, Gomez D, Wilton AS, et al. A population-based
analysis of diabetes-related care measures, foot complications, and
amputation during the COVID-19 pandemic in Ontario, Canada.
JAMA Netw Open 2022; 5: e2142354.
125 Ergisi Y, Ozdemir E, Altun O, Tikman M, Korkmazer S, Yalcin MN.
Indirect impact of the COVID-19 pandemic on diabetes-related
lower extremity amputations: a regional study. Jt Dis Relat Surg
2022; 33: 203–07.
126 Kendirci M, Sahiner IT, Sezikli I, Akin M, Yasti AC. Effects of the
COVID-19 pandemic on the management of diabetic foot ulcers:
experiences from a dedicated diabetic foot care center. Wounds
2022; 34: 146–50.
127 Mariet AS, Benzenine E, Bouillet B, Verges B, Quantin C, Petit JM.
Impact of the COVID-19 epidemic on hospitalization for diabetic
foot ulcers during lockdown: a French nationwide population-
based study. Diabet Med 2021; 38: e14577.
128 Mehanathan PB, Edwards AA, Robinson T. Experience of a surgeon
at the emergency department during COVID-19 pandemic.
Ann Med Surg 2020; 60: 245–48.
129 Radellini S, Vigneri E, Smeraldi L, et al. Evidence of greater severity
of diabetic foot ulcers during COVID-19 pandemic: a real-life
single-centre cohort study. Diabetes Metab Res Rev 2023; 39: e3626.
130 Rastogi A, Hiteshi P, Bhansali AA, Jude EB. Virtual triage and
outcomes of diabetic foot complications during COVID-19
pandemic: a retro-prospective, observational cohort study.
PLoS One 2021; 16: e0251143.
131 Rubin G, Feldman G, Dimri I, Shapiro A, Rozen N. Effects of the
COVID-19 pandemic on the outcome and mortality of patients with
diabetic foot ulcer. Int Wound J 2023; 20: 63–68.
132 Schmidt BM, Munson ME, Rothenberg GM, Holmes CM,
Pop-Busui R. Strategies to reduce severe diabetic foot infections
and complications during epidemics (STRIDE).
J Diabetes Complications 2020; 34: 107691.
133 Schmidt BM, Shin L. Tackling diabetic foot: limb salvage during
the COVID-19 pandemic. Ther Adv Endocrinol Metab 2023;
14: 20420188231157203.
134 Viswanathan V, Nachimuthu S. Major lower-limb amputation
during the COVID pandemic in South India.
Int J Low Extrem Wounds 2023; 22: 475–79.
135 Yunir E, Tarigan TJE, Iswati E, et al. Characteristics of diabetic foot
ulcer patients pre- and during COVID-19 pandemic: lessons learnt
from a national referral hospital in Indonesia.
J Prim Care Community Health 2022; 13: 21501319221089767.
136 Carro GV, Carlucci EM, Torterola I, Breppe P, Ticona Ortiz MÁ,
Palomino Pallarez JE. Diabetic foot and COVID-19. Medical
consultation and severity of lesions compared to 2019.
Medicina (B Aires) 2020; 80 (suppl 6): 30–34.
137 Lipscomb D, Smith A, Adamson S, Rezazadeh E. Diabetic foot
ulceration in COVID-19 lockdown: cause for concern or unexpected
benefit? Diabet Med 2020; 37: 1409.
138 Liu C, You J, Zhu W, et al. The COVID-19 outbreak negatively
affects the delivery of care for patients with diabetic foot ulcers.
Diabetes Care 2020; 43: e125.
139 Urbančič-Rovan V. Diabetic foot care before and during the
COVID-19 epidemic: what really matters? Diabetes Care 2021;
44: e27–28.
140 Im JHB, Jin YP, Chow R, Dharia RS, Yan P. Delayed anti-VEGF
injections during the COVID-19 pandemic and changes in visual
acuity in patients with three common retinal diseases: a systematic
review and meta-analysis. Surv Ophthalmol 2022; 67: 1593–602.
147
 Review
141 Al-Dwairi R, Rwashdeh H, Otoom M. The influence of COVID-19
lockdown in Jordan on patients with diabetic retinopathy:
a case-control study. Ther Clin Risk Manag 2021; 17: 1011–22.
142 Bulut MN, Sonmez HS, Gokce G, et al. The impact of delayed anti-
vascular endothelial growth factor treatment for retinal diseases
during the COVID-19 lockdown. Photodiagnosis Photodyn Ther 2021;
35: 102449.
143 Choi HG, Kim SY, Baek SU. Changes in mean and variance of
ophthalmic disease incidences during COVID-19 pandemic in
Korea. Sci Rep 2022; 12: 20364.
144 Das AV, Narayanan R, Rani PK. Effect of COVID-19 pandemic on
presentation of patients with diabetic retinopathy in a multitier
ophthalmology network in India. Cureus 2021; 13: e19148.
145 Gomel N, Shor R, Lippin N, et al. COVID-19 pandemic lockdowns’
impact on visual acuity of diabetic macular edema: a large cohort.
Ophthalmologica 2023; 246: 1–8.
146 Prajapati V, Shah K, Shah D, Wanjari MB, Singhal D. Effect of the
COVID-19 pandemic lockdown on the management of diabetic
retinopathy: a cross-sectional study. Cureus 2022; 14: e27623.
147 Bambra C, Riordan R, Ford J, Matthews F. The COVID-19
pandemic and health inequalities. J Epidemiol Community Health
2020; 74: 964–68.
148 Kretchy IA, Asiedu-Danso M, Kretchy J-P. Medication management
and adherence during the COVID-19 pandemic: perspectives and
experiences from low- and middle-income countries.
Res Social Adm Pharm 2021; 17: 2023–26.
149 Lippi G, Henry BM, Sanchis-Gomar F. Physical inactivity and
cardiovascular disease at the time of coronavirus disease 2019
(COVID-19). Eur J Prevent Cardiol 2020; 27: 906–08.
148
150 Li LZ, Wang S. Prevalence and predictors of general psychiatric
disorders and loneliness during COVID-19 in the United Kingdom.
Psychiatry Res 2020; 291: 113267.
151 Banerjee M, Chakraborty S, Pal R. Diabetes self-management amid
COVID-19 pandemic. Diabetes Metab Syndr 2020; 14: 351–54.
152 Eberle C, Stichling S. Impact of COVID-19 lockdown on glycemic
control in patients with type 1 and type 2 diabetes mellitus:
a systematic review. Diabetol Metab Syndr 2021; 13: 1–8.
153 Sim R, Chong CW, Loganadan NK, Hussein Z, Adam NL,
Lee SWH. Impact of COVID-19 lockdown on glycemic, weight,
blood pressure control and medication adherence in patients with
type 2 diabetes. Patient Prefer Adherence 2023; 17: 2109–17.
154 Ramírez Manent JI, Altisench Jané B, Sanchís Cortés P, et al.
Impact of COVID-19 lockdown on anthropometric variables, blood
pressure, and glucose and lipid profile in healthy adults: a before
and after pandemic lockdown longitudinal study. Nutrients 2022;
14: 1237.
155 Khunti K, Aroda VR, Aschner P, et al. The impact of the COVID-19
pandemic on diabetes services: planning for a global recovery.
Lancet Diabetes Endocrinol 2022; 10: 890–900.
156 Hartmann-Boyce J, Morris E, Goyder C, et al. Diabetes and
COVID-19: risks, management, and learnings from other national
disasters. Diabetes Care 2020; 43: 1695–703.
157 Hacker KA, Briss PA, Richardson L, Wright J, Petersen R.
COVID-19 and chronic disease: the impact now and in the future.
Prev Chronic Dis 2021; 18: E62.
Copyright © 2024 Elsevier Ltd. All rights reserved.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
 Articles

Analysis of type 2 diabetes heterogeneity with a tree-like

representation: insights from the prospective German
Diabetes Study and the LURIC cohort
Martin Schön, Katsiaryna Prystupa, Tim Mori, Oana P Zaharia, Kálmán Bódis, Maria Bombrich, Clara Möser, Iryna Yurchenko, Yuliya Kupriyanova,
Klaus Strassburger, Pavel Bobrov, Anand T N Nair, Gidon J Bönhof, Alexander Strom, Graciela E Delgado, Sema Kaya, Rainer Guthoff,
Norbert Stefan, Andreas L Birkenfeld, Hans Hauner, Jochen Seissler, Andreas Pfeiffer, Matthias Blüher, Stefan Bornstein, Julia Szendroedi,
Svenja Meyhöfer, Sandra Trenkamp, Volker Burkart, Vera B Schrauwen-Hinderling, Marcus E Kleber, Alexander Niessner, Christian Herder,
Oliver Kuss, Winfried März, Ewan R Pearson, Michael Roden, Robert Wagner, for the German Diabetes Study Group

Summary
Background Heterogeneity in type 2 diabetes can be represented by a tree-like graph structure by use of reversed Lancet Diabetes Endocrinol
graph-embedded dimensionality reduction. We aimed to examine whether this approach can be used to stratify key 2024; 12: 119–31

pathophysiological components and diabetes-related complications during longitudinal follow-up of individuals with Published Online
December 21, 2023
recent-onset type 2 diabetes.
https://doi.org/10.1016/
S2213-8587(23)00329-7
Methods For this cohort analysis, 927 participants aged 18–69 years from the German Diabetes Study (GDS) with See Comment page 87
recent-onset type 2 diabetes were mapped onto a previously developed two-dimensional tree based on nine simple *Members of the German
clinical and laboratory variables, residualised for age and sex. Insulin sensitivity was assessed by a hyperinsulinaemic- Diabetes Study Group are listed
euglycaemic clamp, insulin secretion was assessed by intravenous glucose tolerance test, hepatic lipid content was at the end of the Article
assessed by 1H magnetic resonance spectroscopy, serum interleukin (IL)-6 and IL-18 were assessed by ELISA, and Institute for Clinical
peripheral and autonomic neuropathy were assessed by functional and clinical measures. Participants were followed Diabetology, German Diabetes
Center, Leibniz Center for
up for up to 16 years. We also investigated heart failure and all-cause mortality in 794 individuals with type 2 diabetes Diabetes Research at Heinrich
undergoing invasive coronary diagnostics from the Ludwigshafen Risk and Cardiovascular Health (LURIC) cohort. Heine University, Düsseldorf,
Germany (M Schön PhD,
Findings There were gradients of clamp-measured insulin sensitivity (both dimensions: p<0·0001) and insulin K Prystupa MD,
O P Zaharia MD, K Bódis MD,
secretion (pdim1<0·0001, pdim2=0·00097) across the tree. Individuals in the region with the lowest insulin sensitivity had M Bombrich MD, C Möser MD,
the highest hepatic lipid content (n=205, pdim1<0·0001, pdim2=0·037), pro-inflammatory biomarkers (IL-6: n=348, I Yurchenko MD,
pdim1<0·0001, pdim2=0·013; IL-18: n=350, pdim1<0·0001, pdim2=0·38), and elevated cardiovascular risk (nevents=143, pdim1=0·14, Y Kupriyanova PhD,
pdim2<0·00081), whereas individuals positioned in the branch with the lowest insulin secretion were more prone to G J Bönhof MD, A Strom PhD,
S Trenkamp PhD, V Burkart PhD,
require insulin therapy (nevents=85, pdim1=0·032, pdim2=0·12) and had the highest risk of diabetic sensorimotor Prof V B Schrauwen-Hinderling
polyneuropathy (nevents=184, pdim1=0·012, pdim2=0·044) and cardiac autonomic neuropathy (nevents=118, pdim1=0·0094, PhD, Prof C Herder PhD,
pdim2=0·06). In the LURIC cohort, all-cause mortality was highest in the tree branch showing insulin resistance Prof M Roden MD,
(nevents=488, pdim1=0·12, pdim2=0·0032). Significant gradients differentiated individuals having heart failure with Prof R Wagner MD); German
Center for Diabetes Research,
preserved ejection fraction from those who had heart failure with reduced ejection fraction. München-Neuherberg,
Germany (M Schön, K Prystupa,
Interpretation These data define the pathophysiological underpinnings of the tree structure, which has the potential T Mori MSc, O P Zaharia, K Bódis,
M Bombrich, C Möser,
to stratify diabetes-related complications on the basis of routinely available variables and thereby expand the toolbox
I Yurchenko, Y Kupriyanova,
of precision diabetes diagnosis. K Strassburger PhD,
P Bobrov PhD, G J Bönhof,
Funding German Diabetes Center, German Federal Ministry of Health, Ministry of Culture and Science of the state of A Strom, S Meyhöfer MD,
S Trenkamp, V Burkart,
North Rhine-Westphalia, German Federal Ministry of Education and Research, German Diabetes Association, German
Prof V B Schrauwen-Hinderling,
Center for Diabetes Research, European Community, German Research Foundation, and Schmutzler Stiftung. Prof C Herder, Prof O Kuss PhD,
Prof M Roden, Prof R Wagner);
Copyright © 2024 Elsevier Ltd. All rights reserved. Institute for Biometrics and
Epidemiology, German
Diabetes Center, Leibniz Center
Introduction heterogeneity.4 Recent studies have shown that prediabetes5 for Diabetes Research at
Multiple biological pathways are involved in the and diabetes6–8 can be stratified into distinct endotypes Heinrich Heine University,
development of type 2 diabetes.1 However, the present delineated by key pathobiology-related variables, which Düsseldorf, Germany
(T Mori, K Strassburger,
classification endorsed by WHO2 fails to explain the differentiate the risk of complications.6,7,9
P Bobrov, Prof O Kuss); Division
pathophysiological variability underlying type 2 diabetes.3 Nevertheless, some data suggest that models of Endocrinology and
Networks of specific pathophysiological alterations could recognising type 2 diabetes as a phenotypic continuum Diabetology, Medical Faculty
underlie each case of diabetes and influence disease outperform discrete endotype-based stratification in and University Hospital
Düsseldorf, Heinrich Heine
course. Therefore, the current one-size-fits-all classification predicting complications.10 Moreover, endotypes might University Düsseldorf,
of type 2 diabetes does not account for disease not necessarily align with the underlying molecular

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 119

 Articles
Düsseldorf, Germany
(M Schön, O P Zaharia, K Bódis,
C Möser, G J Bönhof,
Prof C Herder, Prof M Roden,
Prof R Wagner); Division of
Population Health and
Genomics, School of Medicine,
University of Dundee, Dundee,
UK (A T N Nair PhD,
Prof E R Pearson PhD);
5th Department of Medicine
(Nephrology, Hypertensiology,
Rheumatology, Endocrinology,
Diabetology), Medical Faculty
Mannheim, University of
Heidelberg, Mannheim,
Germany (G E Delgado MSc,
M E Kleber PhD,
Prof W März MD); Center for
Preventive Medicine and
Digital Health Baden-
Württemberg, Medical Faculty
Mannheim, Heidelberg
University, Mannheim,
Germany(G E Delgado);
Department of
Ophthalmology, Medical
Faculty and University Hospital
Düsseldorf, Heinrich Heine
University Düsseldorf,
Düsseldorf, Germany
(S Kaya MD, Prof R Guthoff MD);
Institute for Diabetes Research
and Metabolic Diseases,
University of Tübingen,
Tübingen, Germany
(Prof N Stefan MD,
Prof A L Birkenfeld MD);
Institute of Nutritional
Medicine, School of Medicine,
Technical University of Munich,
München, Germany
(Prof H Hauner MD); Diabetes
Research Group, Medical
Department 4, Ludwig-
Maximilians University
Munich, München, Germany
(Prof J Seissler MD); German
Institute of Human Nutrition
Potsdam-Rehbrücke, Nuthetal,
Germany (Prof A Pfeiffer MD);
Department of Medicine,
Endocrinology and
Nephrology, University of
Leipzig, Leipzig, Germany
(Prof M Blüher MD); Helmholtz
Institute for Metabolic, Obesity
and Vascular Research of the
Helmholtz Zentrum München
at the University of Leipzig and
University Hospital Leipzig,
Leipzig, Germany
(Prof M Blüher); Department of
Internal Medicine III, Dresden
University of Technology,
Dresden, Germany
(Prof S Bornstein MD);
Department of Medicine I and
Clinical Chemistry, University
Hospital of Heidelberg,
Heidelberg, Germany
120
Research in context
Evidence before this study
We searched PubMed for articles published until
April 20, 2023, using the search terms “type 2 diabetes” AND
“subclassification” OR “precision medicine” OR “phenotype”
without language restrictions. A recent study applied reverse
graph embedding based on nine routinely available variables to
a Scottish cohort to visualise diabetes heterogeneity on
a tree-like structure as a disease continuum, in a manner that
was different from categorising diabetes into discrete
endotypes. Phenotypic variation at the onset of type 2 diabetes
was shown to affect disease progression and drug response.
Added value of this study
Using gold-standard methods of phenotyping, in this cohort
analysis we uncovered the unique variability in metabolic
phenotypes for insulin sensitivity, insulin secretion, adipose
tissue distribution and pro-inflammatory profile based on
simple clinical variables. Stratification of phenotypic variation
at recent-onset type 2 diabetes identified individuals who were
disease architecture. Instead, they represent a snapshot
of transient metabolic states, with endotype assignments
potentially changing over time, as we recently
demonstrated.7
To address these issues, a recent study11 applied data
dimensionality reduction on routinely collected clinical
data in individuals with newly diagnosed type 2 diabetes.
The data were reduced into a two-dimensional space in
the form of a minimal spanning tree structure by use of
reverse graph embedding.11 The central graphical result of
these data is a main tree trunk with branch-like structures.
Each resulting branch comprises individuals with a
specific phenotype pattern extending to the tip.
Individuals with mixed characteristics are located
centrally in the principal tree trunk, whereas distinct
phenotypes are closer to the tips.9 This method provided
insights into how phenotypic variation at the time of
type 2 diabetes diagnosis relates to diabetes outcomes.11
The application of this method was made available
as an online tool to visualise the phenotypes of
people with recently diagnosed type 2 diabetes along with
the potential to predict the risk of various complications.
In this cohort analysis, we aimed to replicate the
continuous representation of type 2 diabetes hetero­­
geneity using a tree-like graph structure in the German
Diabetes Study (GDS) and the Ludwigshafen Risk and
Cardiovascular Health (LURIC) cohorts; to investigate how
the tree structure discriminates key pathophys­ iological
hallmarks and glycaemic traits as well as their changes
during disease progression; to investigate how specific
metabolic phenotypes predict mortality and diabetes-
related complications across the tree-like representation;
and to ascertain whether continuous representation of
heterogeneity overlaps with diabetes endotypes.
susceptible to decreased insulin secretion or relevant arterial
hypertension or dyslipidaemia, and translated into variations in
the risks of diabetes outcomes, including those incompletely
addressed in previous studies, such as mortality, steatotic liver
disease, neuropathy, diabetic foot syndrome, and depression.
Implications of all the available evidence
The algorithm presented here and available as an online tool
stratifies individuals with diabetes on the basis of simple
variables, which are affordable and easy to collect in routine
clinical practice; this algorithm could help to identify those at
increased risk for specific comorbidities. These findings
advocate for precision medicine in diabetology and support the
need for further studies to apply targeted prevention and
treatment measures on individuals with type 2 diabetes
identified by this algorithm who are susceptible to
development or worsening of specific diabetes-related
outcomes.
Methods
German Diabetes Study population
For this cohort analysis, we recruited 927 individuals of
European descent aged 18–69 years and diagnosed with
type 2 diabetes within the past 12 months, based on the
criteria of the American Diabetes Association,12 from
the prospective observational GDS.13 Exclusion criteria
were diabetes from other causes; pregnancy; acute or
severe chronic heart, liver, renal, or psychiatric diseases;
and immunosuppressive therapy.13 Follow-up visits in
the GDS take place in 5-year intervals; standardised
telephone interviews are conducted annually for up to
16 years to collect data on medication, self-reported
outcomes, disease course, and compliance. The GDS is
an ongoing study, so most of the participants with
missing data at 5-year, 10-year or 15-year follow-up visits
were not lost to follow-up, but their follow-up
visits were not yet due. Selection bias has been
minimised in the GDS through the application of
identical inclusion and exclusion criteria throughout
the recruitment period.
The GDS was approved by the ethics committee at the
Faculty of Medicine of Heinrich Heine University
Düsseldorf (reference number 4508) and other study
sites. The study was performed according to the
Declaration of Helsinki, is registered on ClinicalTrials.
gov (NCT01055093), and written informed consent was
acquired from all participants before inclusion.
German Diabetes Study procedures
Participants were asked to discontinue glucose-lowering
medication, to refrain from strenuous physical activity
and alcohol intake, and to adhere to a balanced isocaloric
diet for 3 days before study visits. All examinations
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

were conducted after a 10 h overnight fast.13 Routine
laboratory parameters were analysed in the central
laboratory (German Diabetes Center, Düsseldorf,
Germany).13
To assess whole-body insulin sensitivity and insulin
secretion, a modified Botnia clamp comprising an
intravenous glucose tolerance test and a hyperinsulinaemic-
euglycaemic clamp was performed.13 The intravenous
glucose tolerance test started with a 30% glucose bolus,
and insulin secretion was calculated from the incremental
area under the curve (AUC) of C-peptide concentra­
tions. The hyperinsulinaemic-euglycaemic clamp started
with a priming dose of insulin over 10 min followed by
a constant infusion (Insuman Rapid; Sanofi, Frankfurt am
Main, Germany) for 3 h. Whole-body insulin sensitivity
was derived from mean glucose infusion rates during the
steady state. Fasting adipose-tissue insulin resistance was
ascertained from fasting free fatty acid (mmol/L) and
insulin (pmol/L) concentrations.14
Insulin secretion was also ascertained from a mixed
meal tolerance test based on standardised liquid meals
(237 mL; 250 kcal, 6 g fat, 28 g carbohydrates, 20 g protein;
Boost Nestlé Health Care Nutrition, Lausanne,
Switzerland) as described and validated previously.15
Insulin secretion based on the mixed meal tolerance test
was calculated as the AUC from 6-point C-peptide
concentrations during the first 3 h after the ingestion.
To assess body composition and fat distribution, MRI
and magnetic resonance spectroscopy (¹H-MRS) were
performed in the 3-T MR scanner (Achieva X-series,
Philips Healthcare, Best, the Netherlands). Total, sub­
cutaneous, and visceral adipose tissue volumes were
assessed from whole-body images with T1-weighted axial
fast spin echo. Hepatic lipid content was quantified by
¹H-MRS with the stimulated echo acquisition mode
(STEAM), using an echo time of 10 ms in a voxel
positioned within a homogeneous part of liver tissue to
avoid major vessels and the gallbladder.16 A bioimpedance
analysis was done to assess fat mass and fat-free mass
(BioElectrical Impedance Analyzer System, RJL Systems,
Detroit, MI, USA).
Whole-body energy metabolism was evaluated as
follows: resting energy expenditure was estimated by
indirect calorimetry by use of Vmax Encore 29n
(CareFusion, Höchberg, Germany);13 peak aerobic
capacity (VO₂max) was measured on a cycle ergometer
by an incremental exhaustive exercise test;13 self-reported
habitual physical activity during the preceding 12 months
was assessed by a modified Baecke questionnaire;17
and serum levels of interleukin (IL)-6 and IL-18 and
total adiponectin were assessed by enzyme-linked
immunosorbent assays.18
Insulin requirement was recorded during annual
phone interviews and was defined as the first timepoint
at which insulin had been prescribed.
Complications were monitored in various ways.
Cardiovascular diseases (transient ischaemic attack, stroke,
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Articles
heart attack, angina, heart failure, peripheral arterial (Prof J Szendroedi MD); Institute
disease, or hospital admission for heart failure) were for Endocrinology & Diabetes,
University of Lübeck, Lübeck,
documented annually through phone interviews. Urinary Germany (S Meyhöfer);
albumin concentrations were measured to assess Department of Internal
microalbuminuria (20–200 mg/L) and macroalbuminuria Medicine 1, Endocrinology &
(>200 mg/L). The estimated glomerular filtration rate Diabetes, University of Lübeck,
Lübeck, Germany (S Meyhöfer);
(eGFR) was ascertained from cystatin C and creatinine.19 SYNLAB MVZ für
Peripheral nerve function was assessed by electrophys­ Humangenetik Mannheim
iological testing, quantitative sensory testing, and clinical GmbH, Mannheim, Germany
neuropathy score surveys. Motor nerve conduction velocity (M E Kleber); Division of
Cardiology, Department of
in the peroneal nerve, sensory nerve conduction Internal Medicine II, Medical
velocity, sensory nerve action potentials in the sural University of Vienna, Austria
nerve, and vibration and thermal detection thresholds (Prof A Niessner PhD); Centre for
were measured.20 Diabetic sensorimotor polyneuropathy Health and Society, Medical
Faculty and University Hospital
(DSPN) was assessed with the Neuropathy Disability Score Düsseldorf, Heinrich Heine
based on modified Toronto Consensus criteria.21 Diagnostic University Düsseldorf,
criteria for DSPN (three of ten points) were reached in 28% Düsseldorf, Germany
(Prof O Kuss); SYNLAB
(184 of 652) participants at the last available measurement.
Academy, SYNLAB Holding
Autonomic nerve function was measured with a battery of Deutschland GmbH, Augsburg
seven cardiovascular tests by use of a VariaCardio TF5 and Mannheim, Munich,
system (MIE Medical Research, Leeds, UK).22 Cardiac Germany (Prof W März)
autonomic neuropathy was defined as the presence of at Correspondence to:
least three abnormal results. Self-reported clinical signs or Prof Robert Wagner, Institute for
Clinical Diabetology, German
history of diabetic foot syndrome and depression were Diabetes Center, Leibniz Center
recorded. for Diabetes Research at Heinrich
Assignment to diabetes endotypes was done with Heine University Düsseldorf,
a mapping algorithm described previously.7 40225 Düsseldorf, Germany
robert.wagner@ddz.de
For the online tool see
LURIC study population https://atn-uod2018.shinyapps.
The prospective LURIC study recruited individuals of io/Prediction_diabetes_
European descent with clinically suspected silent or outcome_18082021
symptomatic coronary heart disease or acute coronary
syndrome (acute myocardial infarction and unstable
angina) who underwent coronary angiography between
July 22, 1997, and January 25, 2000.23 In the current study,
we analysed data from 794 individuals diagnosed with
type 2 diabetes within the past 5 years and who had not
yet received insulin treatment.
LURIC procedures
Heart failure with preserved ejection fraction or reduced
ejection fraction was diagnosed according to current
recommendations,24 as reported previously.25 Data about
deceased participants were acquired from local
population registers.
Statistical analysis
We included participants with type 2 diabetes without
glutamic acid decarboxylase (GAD) antibodies (measured
by a radioligand assay, cutoff 1·0 U/mL) and with complete
baseline datasets (age, BMI, total and HDL-cholesterol,
triglycerides, HbA1c, creatinine, alanin-aminotransferase
[ALT], and systolic and diastolic blood pressure) in the
analysis. For assigning each individual to a position on
the reference tree, we used a mapping function based
on the nine variables residualised for age and sex.11 We
applied rank-based inverse normal transformation for
121
 Articles

continuous variables. Outliers that were more than

5 SD from the mean were excluded. After each participant
was assigned to a coordinate of the tree by the mapping
algorithm, we tested the separation of variables across the
tree by fitting linear regressions to the two dimensions.
Normal distribution of the residuals was tested also from
Person 1 Person 2 Person 3
HbA1c ↓ • ↑ the QQ-plots. The homoscedascity of the residuals was
BMI ↓ • ↑ also tested by assessing residuals-by-fitted plots.
Total cholesterol • ↑ • To allow non-linear fits along each of the two coor­
HDL-cholesterol ↑ • • dinates, natural splines with three degrees of freedom
Triglycerides • • ↑ were applied. The reported p values for the gradients are
Alanine aminotransferase • • ↑ based on the F-tests of the sequential analysis of variance
Serum creatinine • • • tables for the linear models, in which the respective
Systolic blood pressure • ↑ •
outcome was modelled by the two tree dimensions. The
Diastolic blood pressure • ↑ •
corresponding null hypothesis was that the association of
Figure 1: Visual representation of phenotypic characteristics on a tree-like graph structure as a result of data the coefficients with the spline terms of the respective
dimensionality reduction of nine clinical variables tree dimension is equal to zero.
The DDRTree algorithm was used to reduce nine clinical variables residualised for age and sex into a non-linear tree
To ascertain the natural change in these variables
structure and map individuals from the GDS cohort into the pre-existing tree derived from the Scottish cohort.11
Yellow, blue, and red dots represent examples of key characteristics of individuals positioned in the respective during the first 5 years of a diabetes diagnosis, the
branches of the tree. DDRTree=Data dimensionality reduction Tree. GDS=German Diabetes Study. gradient of individual slopes (deltas) across the two tree
dimensions was investigated. Deltas with a gradient
Baseline (n=927) 5-year follow- 10-year follow- 15-year greater than 1·0 were considered as a relevant change in
up (n=447) up (n=147) follow-up the position across the tree.
(n=26)
The gradients of hazard ratios (HRs) across the tree
Lost to follow-up ·· 190 (43%) 44 (30%) 6 (23%) during the disease course were assessed by Cox propor­
Sex tional hazards regression, using the tree coordinates as
Female 318 (34%) 126 (28%) 45 (31%) 6 (23%) exposures with adjustments for age and sex. Individuals
Male 609 (66%) 321 (72%) 102 (69%) 20 (77%) either had an event during the follow-up period or were
Age, years 53·4 (10·0) 58·7 (10·0) 63·5 (10·3) 64·5 (8·2) censored at their last recorded observation if no events
BMI, kg/m² 31·6 (6·1) 31·5 (5·9) 31·0 (5·7) 30·7 (5·1) occurred. Gradients of HRs across the tree are shown for
Waist-to-hip ratio 0·97 (0·08)* 0·98 (0·08) 0·97 (0·07) 1·00 (0·09) one representative male individual aged 50 years, unless
Fasting blood glucose, mg/dL 123 (35·2)† 145 (46·5) 171 (58·6) 188 (50·6) indicated otherwise. Schoenfeld residuals, Concordance
HbA1c (%) 6·4% (0·9) 6·9% (1·1) 7·2% (1·2) 7·4% (1·2) indices, and Bayesian Information Criterion of both the
HbA1c (mmol/mol) 46·6 (9·7) 49·2 (17·9) 53·4 (16·9) 54·4 (20·2) tree coordinate-based models and the models with
Whole-body insulin sensitivity, 6·1 (2·6)‡ 5·5 (2·4) 5·2 (2·3) 4·2 (1·9) the nine variables underlying the tree are summarised
mg/kg*min in the appendix (p 3).
Insulin secretion (total AUC of 121·2 (76·4)§ 94·7 (71·2) 67·8 (68·2) ·· Binary variables (use of selected medications and
C-peptide) cardiac autonomic neuropathy) and their predicted
hsCRP, mg/dL 0·41 (0·63)¶ 0·28 (0·39) 0·54 (2·52) 0·17 (0·11) probabilities were mapped to the tree coordinates with
eGFR, mL/min per 1·73 m² 87·5 (15·8)‖ 84·0 (16·9) 78·5 (19·1) 75·5 (24·9) the odds derived from the logit function of logistic
Creatinine, mg/dL 0·88 (0·18) 0·89 (0·19) 0·91 (0·23) 0·98 (0·30) regression models using the tree coordinates. Logistic
Total cholesterol, mg/dL 199 (42·3) 196 (44·5) 187 (42·4) 163 (46·5) regression was used to analyse cardiac autonomic
LDL-cholesterol, mg/dL 127 (36·3) 124 (38·5) 118 (39·0) 96 (45·9) neuropathy because 67% (79 of 118) of cases had already
HDL-cholesterol, mg/dL 46·8 (13·0) 47·4 (13·1) 48·9 (13·8) 46·7 (13·8) occurred at baseline. An increased risk of cardiac
Triglycerides, mg/dL 163 (134) 187 (176) 189 (169) 206 (241) autonomic neuropathy early after diabetes diagnosis has
ALT, U/L 34·0 (18·7) 33·9 (21·8) 36·4 (49·2) 31·8 (14·9) also been described previously.26
Systolic blood pressure, mm Hg 139 (17·4) 141 (17·0) 143 (16·9) 149 (18·4) To map the position of diabetes endotypes onto the tree,
Diastolic blood pressure, mm Hg 85·0 (10·7) 84·7 (10·5) 82·2 (10·2) 83·3 (9·1) we estimated the distribution from the endotype
Data are n (%) or mean (SD). ALT=alanine aminotransferase. AUC=area under the curve. eGFR=estimated glomerular assignment of participants with multinomial regression,
filtration rate. hsCRP=high-sensitivity C-reactive protein. Data on insulin secretion at 15-year follow-up are not using the most prevalent7 mild age-related diabetes as
presented due to the small sample size. *Data available for 923 individuals. †Data available for 883 individuals. ‡Data a reference endotype. Exposure for groups of medications
available for 652 individuals. §Data available for 827 individuals. ¶Data available for 901 individuals. ‖Data available
for 811 individuals. was calculated for each participant as a percentage of time
during the total duration of the study. To ascertain the
Table: Characteristics of the German Diabetes Study cohort at baseline and at 5-year follow-up visits stability of the gradient in prediction models, we applied
bootstrapping for variable predictions from the regression
models. A power calculation for the primary research
question (whole-body insulin sensitivity predicted by the

122 www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

 Articles

tree coordinates) showed an ƒ² of 0·26, indicating a fairly pattern (figure 2D). The stability of the predicted
large effect size. The statistical power was 1·0 at an α-level gradients for these variables was similar across the
of 0·05. For all data management, statistical analysis, and bootstrap samples (appendix pp 12–13). In a subset of
generation of graphs, R (version 4.2.0) was used. individuals with both tests available (n=294), we
compared the distribution of insulin secretion derived See Online for appendix
Role of the funding source
The funding sources had no influence on the design and
A Whole-body insulin sensitivity
conduct of this study, data collection, data analysis or Measured n=652 Predicted pdim1<0·0001 pdim2 <0·0001
data interpretation; or on the preparation, review, or 1·0 Quartiles
approval of this manuscript for publication. 1
0·5 2
Results 3

Dimension 2
The first examination in the GDS study started in 4
0
Sept 7, 2005, and the last baseline visit was in Nov 11, 2020.
0·5
The last follow-up interview used in the proportional
–0·5 0
hazard models was from Sept 8, 2022. The overall –0·5
dropout rate of the cohort was 26% (241 of 927 participants). –1·0
–1·0
Participants in the GDS were mapped to the prototype
tree by HbA1c, BMI, total cholesterol, HDL-cholesterol,
triglycerides, ALT, creatinine, systolic blood pressure, B ADIPO-IR
Measured n=848 Predicted pdim1<0·0001 pdim2<0·0001
and diastolic blood pressure at baseline by use of
1·0
a mapping algorithm11 (figure 1, table, appendix p 10). Quartiles
1
Among the underlying variables of the tree, triglycerides
0·5 2
had the largest impact on the tree coordinates, followed
3
by HDL-cholesterol, systolic blood pressure, diastolic
Dimension 2

4
blood pressure, and BMI, with the numerically smallest 0

effect size of creatinine (appendix p 10). Characteristics 0·5

of the LURIC cohort are presented in the appendix (p 4), –0·5
0
with the mapped positions of the participants shown in –0·5
the appendix (p 11). –1·0
–1·0
Whole-body insulin sensitivity showed gradients in
both dimensions across the tree, with the highest levels C Insulin secretion
in the upper-left branch and the lowest in the lower-right Measured n=827 Predicted pdim1<0·0001 pdim2=0·00097

branch (figure 2A). Adipose-tissue insulin resistance 1·0

Quartiles
showed an inverse gradient pattern (figure 2B). Insulin 1
secretion (assessed via intravenous glucose tolerance 0·5 2
test) showed different gradients across both dimensions 3
Dimension 2

of the tree compared with insulin sensitivity, with the 4

0
highest levels in the lower branches and the lowest levels 0·6
in the upper-right branch (figure 2C). Adjustment for –0·5 0·3
whole-body insulin sensitivity did not change this 0
–0·3
–1·0

D Insulin secretion (residualised)

Measured n=626 Predicted pdim1 <0·0001 pdim2 =0·0096
1·0
Figure 2: Visualisation of phenotypic heterogeneity in insulin sensitivity Quartiles
(A, B) and insulin secretion (C, D) at baseline 1
A mapping function was used to position individuals of the GDS cohort into the 0·5 2
reference tree derived from the Scottish cohort.11 We investigated the gradient 3
Dimension 2

of whole-body insulin sensitivity (hyperinsulinaemic-euglycaemic clamp, n=652,

4
panel A), fasting adipose tissue insulin resistance (n=848, panel B), insulin 0
secretion (intravenous glucose tolerance test, n=827, panel C), and
insulin secretion residualised for whole-body insulin sensitivity (n=626, panel D) 0·5
–0·5
across the tree. In the first panel, each dot represents the position of one
0
individual from the GDS cohort. The second panel represents the computed
gradient of each value in the tree with linear regression models. The yellow –1·0 –0·5
colour indicates a high value, and the dark blue colour indicates a low value of
1·0 0·5 0 –0·5 –1·0 1·0 0·5 0 –0·5 –1·0
the phenotype. ADIPO-IR=fasting adipose tissue insulin resistance. GDS=German
Diabetes Study. Dimension 1 Dimension 1

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 123

 Articles

A Hepatocellular lipids B AST/ALT ratio

Measured n=205 Predicted pdim1<0·0001 pdim2=0·037 Measured n=923 Predicted pdim1<0·0001 pdim2<0·0001
1·0 Quartiles Quartiles
1 1
0·5 2 2
3 3
Dimension 2

4 4
0
0·5
0 0·4
–0·5
–0·5 0
–1·0 –0·4
–1·0

C FIB-4 D Interleukin-6
Measured n=664 Predicted pdim1<0·0001 pdim2=0·052 Measured n=348 Predicted pdim1<0·0001, pdim2=0·013
1·0

0·5
Dimension 2

0
0·8
0·8
–0·5 0·3
0·4
0 0
–0·3 –0·4
–1·0

E Interleukin-18 F Adiponectin
Measured n=350 Predicted pdim1<0·0001 pdim2=0·38 Measured n=351 Predicted pdim1=0·00042 pdim2<0·0001
1·0

0·5
Dimension 2

0
0·50 0·5
0·25
0
–0·5 0
–0·25 –0·5
–0·50 –1·0
–1·0
–0·75 –1·5
1·0 0·5 0 0·5 –1·0 1·0 0·5 0 0·5 –1·0 1·0 0·5 0 0·5 –1·0 1·0 0·5 0 0·5 –1·0
Dimension 1 Dimension 1 Dimension 1 Dimension 1

Figure 3: Visualisation of phenotypic heterogeneity in hepatic lipid content, liver fibrosis, pro-inflammatory cytokines, and adiponectin at baseline
A mapping function was used to position individuals of the GDS cohort into the reference tree derived from the Scottish cohort11 for hepatocellular lipid content (assessed by magnetic resonance
spectroscopy, n=205, panel A), surrogate for steatotic liver disease (AST/ALT ratio, n=923, panel B), surrogate for liver fibrosis (FIB-4, n=664, panel C), interleukin-6 (n=348, panel D), interleukin-18
(n=350, panel E), and adiponectin (n=351, panel F). In the first panel, each dot represents the position of one individual from the GDS cohort. The second panel represents the computed gradient of
each value on the tree. The yellow colour indicates a high value, and the dark blue colour indicates a low value of the phenotype. ALT=alanine aminotransferase. AST=aspartate aminotransferase.
FIB-4=index for liver fibrosis. GDS=German Diabetes Study.

from the intravenous glucose tolerance test and from the
mixed meal tolerance test (appendix p 16). The highest
levels of insulin secretion remained in the lower parts of
the tree (dimension two) in both tests, but with opposing
gradients along the x-axis.
Total adipose tissue volume assessed by MRI in
a subset of 166 individuals showed a gradient in the first
dimension, with individuals on the left branches having
lower adipose tissue volume than those on the right
(appendix p 17). Bioimpedance measurements of whole-
body adipose tissue volume also showed a gradient in the
second dimension (appendix p 17). Subcutaneous and
visceral fat content (appendix p 17) were also differentially
distributed in the first dimension. Fat-free mass and
waist-to-hip ratio (appendix p 17) showed gradients across
both dimensions, with higher values in the right and
lower parts.
Hepatic lipid content showed gradients across both
dimensions in 205 individuals, and the distribu­
tion pattern closely mirrored that of whole-body
insulin sensitivity and adipose-tissue insulin resistance
(figure 3A). The gradient for hepatic lipid content was
recapitulated by the aspartate aminotransferase/alanine
aminotransferase (AST/ALT) ratio in a larger sample of
923 individuals (figure 3B). Participants with the highest
risk of liver fibrosis based on the biochemical index
were positioned in the right part of the tree (figure 3C).
We observed a higher variance in the distribution
of body fat gradient when assessed by MRI
(appendix pp 12–13), especially in the lower branches of

124 www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

 Articles

A Cardiovascular endpoints B Antihypertensive medication

Events=143 Predicted, median HR=1·1 Events n=435 pdim1<0·0001 pdim2=0·066
(95% CI 0·8–1·3)
pdim1=0·14 pdim2=0·00081
Event Event
1·0
occurred occurred
0·5 No No
Dimension 2

Yes Yes
0 2·0 2·0
1·5 1·8
–0·5 1·0 1·6
0·5 1·4
–1·0

C All-cause mortality D Heart failure Heart failure with Heart failure with Heart failure Event occured
Events n=488 Predicted, median HR 1·7 Events n=738 preserved ejection preserved ejection with preserved Yes
(95% CI 1·5–2·0) pdim1=0·010, fraction fraction ejection fraction No
pdim1=0·12 pdim2<0·0032 pdim2=0·00096 Heart failure with
1·3
Event 1·2 preserved ejection
1·0
occurred 1·1 fraction
0·5 No Heart failure with
Dimension 2

Heart failure reduced ejection

Yes
0 with reduced fraction
2·4 ejection fraction
–0·5
1·8 1·6
–1·0 1·4
1·2 1·2
–1·0 –0·5 0 0·5 1·0
E Urinary albumin concentrations F eGFR Dimension 1
Measured n=824 Predicted pdim1<0·000 Measured n=811 Predicted pdim1<0·0001
pdim2=0·97 pdim2=0·22
Quartiles Quartiles
1·0 1 1
2 2
0·5
Dimension 2

3 3
0 4 4
0·6
–0·5 0·4 0·25
0·2 0
–1·0 0 –0·25
–0·50

G Change in urinary albumin concentrations H Change in eGFR

Measured n=344 Predicted pdim1=0·74 Measured n=326 Predicted pdim1=0·041,
pdim2=0·46 pdim2=0·91
1·0
Quartiles Quartiles
1 1
0·5
Dimension 2

2 2
0 3 3
4 4
–0·5 0 –0·25
–0·50
–2·5 –0·75
–1·0 –1·00
–5·0 –1·25
–1·0 –0·5 0 0·5 1·0 –1·0 –0·5 0 0·5 1·0 –1·0 –0·5 0 0·5 1·0 –1·0 –0·5 0 0·5 1·0
Dimension 1 Dimension 1 Dimension 1 Dimension 1

Figure 4: Visualisation of type 2 diabetes heterogeneity for cardiovascular complications, all-cause mortality, and chronic kidney disease
A mapping function was used to position individuals of the GDS cohort into the reference tree derived from the Scottish cohort11 to estimate the probability of cardiovascular outcomes (n=143,
panel A) and use of antihypertensive medication at baseline (n=435, panel B) and from the LURIC cohort for all-cause mortality (n=488, panel C) and heart failure (n=738, panel D). A mapping
function was used to position individuals in the GDS cohort into the reference tree derived from the Scottish cohort11 for urinary albumin concentrations (n=824, panel E) and eGFR (n=811, panel F)
at baseline, and for the change in urinary albumin concentrations (n=344, panel G) and change in eGFR (n=326, panel H) in the first 5 years after diagnosis of type 2 diabetes. The yellow colour
indicates a high HR, and the dark blue colour indicates a low HR for the event. In the first panel, each dot represents the position of one individual from the GDS or LURIC cohort. The second panel
represents the computed gradient of each value into the tree. eGFR=estimated glomerular filtration rate. GDS=German Diabetes Study. HR=hazard ratio. OR=odds ratio.

the tree, possibly due to the low number of participants
in these areas. However, the stability of measurements
performed in the larger study population was high
(appendix pp 12–13).
Resting energy expenditure was strongly associated
with the first dimension, but did not reach statistical
significance in the second dimension (appendix p 18).
Whereas VO₂max did not vary across the tree in any
dimension, participants with increased habitual physical
activity were positioned in the upper branches of the tree
(appendix p 18).
The pro-inflammatory biomarkers IL-6 (figure 3D) and
IL-18 (figure 3E) were both correlated with dimension 1
and IL-6 also correlated with dimension 2, with the
lowest levels in the upper-left branch and the highest
levels in the lower-right branch (similar to whole-body
and adipose tissue insulin resistance, hepatic lipid
content, and resting energy expenditure). By contrast,
adiponectin concentrations showed gradients in
the opposite direction (figure 3F). The stability of the
predicted gradients for IL-6, IL-18, and adiponectin was
high (appendix p 14).

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 125

 Articles
5 years after a diabetes diagnosis, 90 (20%) of
447 participants changed their position across the tree
(appendix p 19). When investigating the changes
in selected variables 5 years after the diagnosis in
447 participants, we observed that whereas changes
in HbA1c did not reveal gradients in either dimension
(appendix pp 20–21), changes in BMI showed variation
in the first dimension (appendix pp 20–21), with
individuals in the left branches (with lower BMI at
baseline) being more prone to bodyweight gain.
Changes in whole-body insulin sensitivity showed
minimal variation (appendix pp 20–21), but insulin
secretion decreased in the left tree branches
(appendix pp 20–21). Changes in systolic blood pressure
and diastolic blood pressure (appendix pp 20–21)
showed variation across both dimensions, with
individuals in the lower-left part being the most prone
to increased blood pressure. However, similarly to
changes in BMI and insulin secretion, the upper-right
part of the tree already had higher blood pressure at
baseline. Adjusting for the migration distance of each
individual after 5 years on the tree did not abolish the
significance of the gradients of any of the variables
(appendix pp 5–6). We observed relatively low variance
in changes to selected variables after 5 years
(appendix p 14).
The HR of cardiovascular outcomes was increased in
the lower parts of the tree (HRmax 2·0, median HR 1·1
[95% CI 0·8–1·3]; figure 4A), positioning individuals
with low whole-body insulin sensitivity and high adipose-
tissue insulin resistance (figure 2A, B), whose blood
pressure as well as LDL-cholesterol increased 5 years after
diagnosis (appendix pp 20–21). Notably, overall
cardiovascular risk was not increased for individuals with
the highest blood pressure at baseline, positioned in the
upper-right branch of the tree. Participants in this branch
were more commonly prescribed with antihypertensive
medication than those in the left and lower parts of the
tree at baseline (n=435, Pdim1=0·001, Pdim2=0·066; figure
4B). Individuals positioned in the middle and left lower
branches of the tree were not commonly receiving
antihypertensive medication at baseline (figure 4B),
which could contribute to the increased HR for
cardiovascular events during follow-up, despite baseline
blood pressure being within a normal range (appendix
p 11). Interestingly, participants positioned in the lower
branches of the tree with the highest cardio­vascular risk
in the GDS cohort also had the highest all-cause mortality
(figure 4C) along with heart failure with reduced ejection
fraction (figure 4D) in the LURIC cohort.
Chronic kidney disease, proxied by urinary albumin
concentrations and eGFR, was worse at baseline in the
right branches of the tree (figure 4E, F). Although we
did not observe a change in urinary albumin
concentrations during the first 5 years (figure 4G),
eGFR decreased more rapidly in the right-sided
branches (figure 4H). We observed a higher DSPN score
126
(n=652, Pdim1=0·012, Pdim2=0·044; figure 5A) and higher
probability of cardiac autonomic neuropathy (n=118,
Pdim1=0·009, Pdim2=0·060; figure 5B) in the upper right
part of the tree, corresponding to the regions with lower
baseline insulin secretion. The risk of diabetic foot
syndrome (HRmax 2·5, median HR 1·2 [95% CI 0·8–1·9];
figure 5C) and depression (HRmax 1·5, median HR 1·0
[0·8–1·5]; figure 5D) was the highest in the lower
branches of the tree, similar to cardiovascular outcomes.
Individuals in the upper right part of the tree with the
lowest insulin secretion at baseline and increased DSPN
score (figure 5A) and increased risk of cardiac autonomic
neuropathy (figure 5B) were more likely to require
insulin treatment (HRmax 2·0, median HR 1·1 [95% CI
0·9–1·2]; figure 5E). Adjusting for the migration
distance of each individual after 5 years on the tree did
not abolish the significance of the gradients of the
complica­tions. However, the significance of the gradient
of insulin requirement was lost after this adjustment
(appendix pp 5–6). The variability of prediction
outcomes was slightly higher in the lower branches of
the tree for cardiovascular outcomes, diabetic foot
syndrome, and depression (appendix p 15), possibly due
to a lower number of participants with more extreme
characteristics in the more distal branches of the tree.
Severe insulin-resistant diabetes was positioned in the
lower-right part of the tree, with the lowest whole-body
insulin sensitivity, the highest BMI, adipose-tissue
insulin resistance, pro-inflammatory biomarkers, and
hepatic lipid content at baseline, and with the highest HR
for cardiovascular disease, diabetic foot syndrome, and
depression (appendix p 22). Mild obesity-related diabetes
was positioned in the central parts, whereas mild
age-related diabetes was positioned in the upper left and
right branches. A small sample size of participants with
severe insulin-deficient diabetes precluded mapping of
this endotype.
At baseline, participants in the lower branches of the
tree were more likely to use metformin, whereas those
in the left branches were more often prescribed with
insulin (appendix pp 23–24). We did not observe any
gradient across the tree for the use of DPP-4 inhibitors
(appendix pp 23–24) and only a few participants were
using sulfonylureaz (n=20), SGLT2 inhibitors (n=1), or
GLP-1 receptor agonists (n=1). Individuals positioned
in the right branches of the tree were also more
commonly taking antihypertensive therapy, including
β-blockers and calcium channel blockers or renin–
angiotensin–aldosterone system (RAAS) inhibitors, or
both (appendix pp 23–24). The gradient of calcium
channel blockers was significant in both dimensions.
Its use was highest in the upper right branch,
corresponding to increased blood pressure. There was
no distribution gradient for the use of statins at baseline
(appendix pp 23–24).
When looking at the exposure to these groups of
medications during the observation period of the study,
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
 Articles

A Diabetic sensorimotor polyneuropathy B Cardiac autonomic neuropathy

Measured n=652 Predicted pdim1=0·012, pdim2=0·044 Measured n=118 Predicted pdim1=0·0094 pdim2=0·060
1·0 1·0 Event
Quartiles occurred
1 No
0·5 2 0·5 Yes
Dimension 2

Dimension 2
3
OR
0 4 0
1·30
3 1·25
–0·5 –0·5 1·20
2
1·15
1 1·10
–1·0 –1·0
0

C Diabetic foot syndrome D Depression

Events n=51 Predicted, medium HR 1·2 (95% CI 0·8–1·9) Events n=173 Predicted, median HR 1·0 (95% CI 0·8–1·5)
pdim1=0·81 pdim2=0·017 pdim1=0·58 pdim2=0·001
1·0 Event 1·0 Event
occurred occurred
No No
0·5 0·5
Yes Yes
Dimension 2

Dimension 2
HR HR
0 0
2·5
2·0 1·5
–0·5 1·5 –0·5
1·0 1·0
0·5
–1·0 –1·0
1·0 0·5 0 0·5 –1·0 1·0 0·5 0 0·5 –1·0
Dimension 1 Dimension 1
E Requirement for insulin
Events n=85 Predicted, medium HR 1·1 (95% CI 0·9–1·2)
pdim1=0·032, pdim2=0·12
1·0 Event
occurred
No
0·5 Yes
Dimension 2

HR
0
2·0
–0·5 1·5
1·0
–1·0
1·0 0·5 0 0·5 –1·0 1·0 0·5 0 0·5 –1·0
Dimension 1 Dimension 1

Figure 5: Visualisation of type 2 diabetes heterogeneity for diabetes-associated neuropathy, diabetic foot syndrome, depression, and insulin requirement
A mapping function was used to position individuals from the GDS cohort into the reference tree derived from the Scottish cohort11 to estimate the probability of diabetic sensorimotor
polyneuropathy (n=652, panel A) and cardiac autonomic neuropathy (n=118, panel B), diabetic foot syndrome (n=51, panel C), depression (n=173, panel D), and insulin requirement (n=85, panel E).
The diagnostic criteria based on the Neuropathy Disability Score were met in 184 of 652 individuals in whom diabetic sensorimotor polyneuropathy was assessed. The yellow colour indicates a
high HR, and the dark blue colour indicates a low HR for the event. In the first panel, each dot represents the position of one individual from the GDS cohort. The second panel represents the computed
gradient of each value into the tree. HR=hazard ratio. OR=odds ratio.

metformin had the highest exposure, followed by RAAS
inhibitors, statins, β-blockers, and calcium channel
blockers. The mean exposure of other analysed medications
was lower than 10% (appendix p 7).
After adjustment for medication in the Cox models, the
dimensions of the tree were still significantly associated
with cardiovascular risk, diabetic foot syndrome, and
depression. When adjusting the risk of diabetes-related
comorbidities to exposure to selected medications,
cardiovascular risk was positively associated with the use
of β-blockers and statins, which is likely to have been
caused by a prescription bias. Interestingly, the use of
insulin was negatively associated with prevalent
depression, but the association with diabetic foot
syndrome was not significant (appendix pp 8–9). The
impact of SGLT2 inhibitors and GLP-1 receptor agonists
could not be taken into account because of the low
number of participants exposed to these medications
(appendix pp 8–9).
Discussion
In the present study, an algorithm based on simple
routinely available variables detected essential phy­
siological attributes representing diabetes heterogeneity
in insulin sensitivity, insulin secretion, adipose tissue
distribution, and pro-inflammatory profiles in individ­
uals with recent-onset type 2 diabetes. Stratifying
type 2 diabetes on the basis of pathophysiological
differences translated into identification of individuals
prone to decreasing insulin secretion or worsening
arterial hypertension or dys­lipidaemia, or both, within
the first 5 years after diagnosis and also identified
variations in mortality risk and the risk of certain diabetes
outcomes collected up to 16 years.

www.thelancet.com/diabetes-endocrinology Vol 12 February 2024 127

 Articles
The two-dimensional tree-like structure formed by the
algorithm represents a practical tool for early detection of
phenotypic variations in type 2 diabetes, facilitating
differential predictions of disease course, mortality, and
complication risks. For example, individuals in the lower-
right branch, characterised by the lowest insulin
sensitivity, highest levels of visceral adipose tissue, and
pro-inflammatory markers, had an increased risk of
cardiovascular events, diabetic foot syndrome, and
depression. Conversely, those in the upper-right branch
had moderate insulin resistance but markedly decreased
insulin secretion early after diabetes diagnosis despite
negative GAD antibodies, which translated into an
increased risk of insulin therapy as well as diabetes-
associated neuropathy.
The association of the tree dimensions with the
underlying variables showed similar patterns to previously
reported population-based cohorts.11 Notably, the variance
of HbA1c was lower in the GDS cohort, possibly due to the
study design and recruitment strategy. Furthermore, the
mean age of participants was approximately 10 years lower
in the GDS than in the Scottish cohort (63·2 years).11 Because
of the exclusion criterion for people with HbA1c values
greater than 9% at baseline, the GDS does not capture
individuals with glycaemic decompensation, who were
mostly positioned in the right branches of the tree as
demonstrated in population-based cohorts.11 Nonetheless,
state-of-the-art techniques of deep metabolic phenotyping
in the GDS allowed us to detect different patterns of
insulin sensitivity and insulin secretion, which would
have been challenging in large-scale population-based
studies. When insulin sensitivity and secretion are
measured with surrogates based on identical parameters
such as homeostatic model assessment, they reflect
mirrored results.11 Whereas the pattern for insulin
sensitivity showed similarities to the pattern observed in
the ADOPT trial,11 insulin secretion showed an opposite
pattern (the highest in the lower left part and the lowest in
the upper right part), even when adjusted for the degree of
insulin sensitivity.
The lower-right part of the branch was characterised
with the highest degree of insulin resistance early after
diabetes onset, which was mirrored by the distribution
gradient for metformin use. This part of the tree
showed elevated cardiovascular risk, which aligns with
the findings of Nair and colleagues11 and partially covers
the complication profile of the severe insulin-resistant
diabetes endotype,6,27 which is characterised by insulin
resistance,6,7 a pro-inflammatory phenotype,28 and
steatotic liver disease.7 However, whereas the lowest
probability of major adverse cardiovascular events in the
Scottish and UKBB cohort was in the upper left branch,11
in the GDS cohort it was located in the upper right
branch. One explanation for this discrepancy could be
the use of early and effective antihypertensive medication
in individuals belonging to these branches. Our data
corroborate the view that liver lipid content, alongside
128
whole-body and adipose tissue insulin resistance, is
associated with increased cardiovascular risk. Notably,
we confirmed this cardiovascular risk pattern in an
independent cohort with all-cause mortality risk closely
reflecting the risk of major adverse cardiovascular events
in the Scottish and UKBB cohorts.11 Furthermore, our
data show that participants with this metabolic
phenotype, visualised in the lower right branch, have an
increased risk of diabetic foot syndrome and depression.
These diabetes complications were often not addressed
in previous studies on type 2 diabetes heterogeneity,29
despite their increasing prevalence in people with
type 2 diabetes.30,31 Interestingly, we observed different
patterns for heart failure with preserved ejection
fraction compared with heart failure with reduced
ejection fraction. Whereas heart failure with preserved
ejection fraction was more probable in the tree area
associated with hypertension, heart failure with reduced
ejection fraction was linked to the regions characterised
by insulin resistance, inflammation, and obesity.
The tree-mapping effectively discriminated differences
in insulin secretion, which differed from the pattern of
insulin sensitivity, identifying individuals in the upper-
right branch with decreased insulin secretion at baseline,
accelerated insulin initiation, and elevated risk of
peripheral and autonomic neuropathy. The association
between low insulin secretion and the risk of neuropathy
was previously reported by our group7 and corresponded
to the severe insulin-deficient diabetes endotype.
Notably, the algorithm in the present study captured this
association without any direct measure of insulin
secretion, in contrast to the assignment to diabetes
endotypes, which involves a biochemical surrogate
for insulin secretion. This convergence of results
from different approaches of diabetes subphenotyping
is important since reverse graph embedding and
subphenotyping diabetes into discrete endotypes differ
not only in the analytical method but also with regard to
the clinical variables used for stratification. Insulin
treatment mirrored the deterioration of insulin secretion
in the early course of the disease. Participants more
prone to require insulin treatment were positioned in
the upper right areas of the tree, in contrast to the
finding from the Scottish cohort, which related the
bottom right areas to higher levels of HbA1c,11 often
serving as an indication for starting insulin in clinical
settings. Our observations from the GDS cohort link
insulin requirement to low insulin secretion at baseline
as assessed by the intra­venous glucose tolerance test. By
contrast, tree branches with low insulin secretion
assessed with a mixed meal tolerance test (upper left
branch) were not prone to insulin initiation during
follow-up.
Individuals in the upper right branch received
antihypertensive treatment early after diabetes onset,
probably resulting in a decrease in blood pressure
5 years after diagnosis. Importantly, individuals prone
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

to development or worsening of arterial hypertension
were positioned in the three lower branches, but
only the lower-right branch received antihypertensive
treatment at baseline. The other untreated two lower
branches were associated with the highest increase in
both systolic and diastolic blood pressure, probably
contributing to the elevated cardiovascular risk. Deep
metabolic phenotyping revealed that individuals in
these two branches were sedentary at baseline, with
a high to moderate degree of insulin resistance initially
overcompensated with increased insulin secretion,
which subsequently decreased. Furthermore, these
two branches had the highest increase in total and LDL-
cholesterol. Therefore, these individuals might be
overlooked and could benefit from intensified treatment
concepts aiming to optimise cardiovascular risk factors
early after a diabetes diagnosis.
During the early stages of the disease course, we
did not observe any change in urinary albumin
concentrations. However, we found a steeper decline in
eGFR in the branches on the right, corresponding to
individuals with higher baseline blood pressure, and
consistent with the pattern of chronic kidney disease
from population-based cohorts.11
Interestingly, individuals with the highest degree of
insulin resistance at baseline along with high BMI
also had relatively elevated fat-free mass and resting
energy expenditure, possibly representing an early
compensatory mechanism to offset energy excess.
Variations in habitual physical activity did not translate
into variations in VO₂max. This might be due to
generally lower levels of VO₂max in cohorts with
diabetes,32 with compro­mised metabolic flexibility or
muscle mitochondrial respiration,33 or both, resulting
in an inability to increase VO₂max despite high levels
of physical activity. This phenomenon was suggested
to be present in 15–20% of individuals with
type 2 diabetes.34
Approximately 20% of participants changed their
position across the tree after 5 years, which might be due
to altered metabolism or treatment, or both, over time.
Notably, a similar proportion of participants changed
their endotype assignments over an identical interval.7 As
these two approaches differ not only in the method of
assessing heterogeneity but also in most of the
underlying variables, this similarity could indicate true
phenotypic changes over time in this proportion of the
population.
As evidenced by the absence of difference in the change
in HbA1c across the tree, these baseline features might
not differentiate glycaemic trajectories, although our
analysis revealed that they are effective in stratifying
complications.
The nine variables used to assign individuals to the
tree-like graph are based on standard measurements
and easy to determine, suggesting they could be an
attractive tool for precision diabetes diagnosis in
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Articles
settings with limited resources. Importantly, this
algorithm does not involve laboratory measurements
that are less standardised on an international level,
such as insulin or C-peptide.35,36 Finally, continuous
representation of diabetes heterogeneity by displaying
each individual on a spectrum of a disease continuum
could potentially provide more precise information
about an individual phenotype than discrete approaches.
Our study is not population-based, and the study design
excludes individuals with decompensated diabetes. As
a result, the diabetes cohort evaluated in this study
cannot represent the full spectrum of people with
diabetes under real-life circumstances. This limitation
might also affect the number and distribution of events
for specific diabetes-related outcomes at follow-up.
Notably, the predictive accuracy of our models was
limited, suggesting that inferences from the tree are
not deterministic, and so caution should be exercised
when interpreting these results in a clinical setting.
Self-reported outcomes might be influenced by
subjectivity and response or memory bias, or both.
Further limitations include the absence of comprehen­
sive genetic information, which could complement
findings related to comprehensive metabolic pheno­
typing and be linked to complications. Although
glucose-lowering medications were discontinued at
least 3 days before the metabolic tests and appropriate
statistical adjustments, we cannot rule out the impact
of glucose-lowering medications on the metabolic
phenotype. Last, it is uncertain whether the results are
generalisable to people of non-European descent.
In conclusion, this study clarifies the patho­
physiological underpinnings of the graph-derived tree
ascertained with nine ubiquitously available variables.
It extends the understanding of the phenotypic
heterogeneity of type 2 diabetes and establishes links
with diabetes outcomes. This algorithm turns complex
disease characteristics into a simple two-dimensional
model, enabling visualisation of the profile of a person
with type 2 diabetes compared with others with a similar
age and sex. When applied in clinical practice, the
position of an individual with diabetes will be displayed
on a tree-like structure followed by an estimation of
their risk for certain comorbidities, which will help to
better understand the person’s own phenotype. Overall,
this tool holds the potential to facilitate the interaction
between clini­cians and individuals with diabetes. When
further comple­mented with genomic information and
well-designed randomised controlled trials, this
algorithm can pave the way for novel precision treatment
strategies.
German Diabetes Study Group
Michael Roden (speaker), Hadi Al-Hasani, Bengt Frederik Belgardt,
Gidon J Bönhof, Gerd Geerling, Christian Herder, Andrea Icks,
Karin Jandeleit-Dahm, Jorg Kotzka, Oliver Kuß, Eckhard Lammert,
Wolfgang Rathmann, Sabrina Schlesinger, Vera Schrauwen-Hinderling,
Julia Szendroedi, Sandra Trenkamp, Robert Wagner.
129
 Articles
Contributors
MS analysed the data, researched the data, wrote the first draft, and
edited the manuscript. KP, OPZ, KB, MBo, CM, IY, YK, GJB, AS, GD,
RG, SK, and CH examined and researched the data. TM, KS, PB, ATN,
and OK contributed to data management and statistical analyses.
NS, AB, HH, JS, AP, MBl, SB, JS, SM, and ST contributed to data
acquisition and provided critical inputs to the revision of the manuscript.
VB, VSH, MK, AN, CH, OK, WM, and ERP contributed to the
discussion and reviewed the Article. MR contributed to the discussion,
and reviewed and edited the Article. RW conceived the idea for this
study, performed the statistical analyses, and reviewed and edited the
Article. RW is the guarantor of this work and, as such, had full access to
all the data in the study and takes responsibility for the integrity of the
data and the accuracy of the data analysis. All authors gave final approval
of this version of the manuscript for publication.
Declaration of interests
OPZ reports grants from the German Diabetes Association (DDG),
the European Foundation for the Study of Diabetes (EFSD), and
Profibildung 2020; and received honoraria from Sanofi. RG served
on an advisory board for Roche Pharma and Bayer Health Care.
AP received support from BMBF-DZD; grants from EFSD and
BMBF-FEI; honoraria from Novo Nordisk, Sanofi, Abbott,
Bundesverband der Pharmazeutischen Industrie (BDI), SCIARC
GmbH, and the University of Fulda; and consulting fees from Abbott.
MBl received honoraria as a consultant and speaker from Amgen,
AstraZeneca, Bayer, Boehringer-Ingelheim, Lilly, Novo Nordisk,
Novartis, and Sanofi; and served on an advisory board for Boehringer-
Ingelheim. ERP reports personal honoraria from Lilly and Novo
Nordisk. MR declares personal fees from AstraZeneca, Boehringer-
Ingelheim, Eli Lilly, Fishawack Group, Madrigal, Novo Nordisk,
Nutricia/Danone, Prosciento, Sanofi, and Target-RWE. RW reports
lecture fees from Novo Nordisk, Sanofi, and Eli Lilly; and served on
an advisory board for Akcea Therapeutics, Daiichi Sankyo, Sanofi,
Eli Lilly, and Novo Nordisk.
Data sharing
To ensure the data privacy of the participants, the generated datasets of
the ongoing GDS are not publicly available, especially since they are
subject to national data protection laws and restrictions by the ethics
committee. However, pseudonymised data can be requested through an
individual project agreement within the GDS with the principal
investigator of the study. Requests for pseudonymised data can be
submitted by email to beate.stodieck@ddz.de.
Acknowledgments
The GDS was initiated and financed by the German Diabetes Center
(DDZ), which is funded by the German Federal Ministry of Health
(Berlin, Germany), the Ministry of Culture and Science of the state North
Rhine-Westphalia (Düsseldorf, Germany), the German Federal Ministry
of Education and Research (BMBF) to the German Center for Diabetes
Research (DZD), the European Community (HORIZON-HLTH-2022-
STAYHLTH-02–01: Panel A) to the INTERCEPT-T2D consortium, the
German Research Foundation (GRK 2576), and the Schmutzler Stiftung;
and is receiving funding from the programme “Profilbildung 2020”, an
initiative of the Ministry of Culture and Science of the State of
Northrhine Westphalia. We thank all participants and staff involved in
conducting the GDS and LURIC studies for their invaluable
contributions. The sole responsibility for the content of this publication
lies with the authors.
References
1 Roden M, Shulman GI. The integrative biology of type 2 diabetes.
Nature 2019; 576: 51–60.
2 WHO. Classification of diabetes mellitus. April 21, 2019. https://
www.who.int/publications-detail-redirect/classification-of-diabetes-
mellitus (accessed Nov 27, 2023).
3 Pearson ER. Type 2 diabetes: a multifaceted disease. Diabetologia
2019; 62: 1107–12.
4 McCarthy MI. Painting a new picture of personalised medicine for
diabetes. Diabetologia 2017; 60: 793–99.
5 Wagner R, Heni M, Tabák AG, et al. Pathophysiology-based
subphenotyping of individuals at elevated risk for type 2 diabetes.
Nat Med 2021; 27: 49–57.
130
6 Ahlqvist E, Storm P, Käräjämäki A, et al. Novel subgroups of
adult-onset diabetes and their association with outcomes:
a data-driven cluster analysis of six variables.
Lancet Diabetes Endocrinol 2018; 6: 361–69.
7 Zaharia OP, Strassburger K, Strom A, et al. Risk of diabetes-
associated diseases in subgroups of patients with recent-onset
diabetes: a 5-year follow-up study. Lancet Diabetes Endocrinol 2019;
7: 684–94.
8 Zou X, Zhou X, Zhu Z, Ji L. Novel subgroups of patients with
adult-onset diabetes in Chinese and US populations.
Lancet Diabetes Endocrinol 2019; 7: 9–11.
9 Herder C, Roden M. A novel diabetes typology: towards precision
diabetology from pathogenesis to treatment. Diabetologia 2022;
65: 1770–81.
10 Dennis JM, Shields BM, Henley WE, Jones AG, Hattersley AT.
Disease progression and treatment response in data-driven
subgroups of type 2 diabetes compared with models based on
simple clinical features: an analysis using clinical trial data.
Lancet Diabetes Endocrinol 2019; 7: 442–51.
11 Nair ATN, Wesolowska-Andersen A, Brorsson C, et al.
Heterogeneity in phenotype, disease progression and drug response
in type 2 diabetes. Nat Med 2022; 28: 982–88.
12 American Diabetes Association Professional Practice C. 2.
Classification and diagnosis of diabetes: standards of medical care
in diabetes—2022. Diabetes Care 2022; 45 (suppl 1): S17–38.
13 Szendroedi J, Saxena A, Weber KS, et al. Cohort profile: the German
Diabetes Study (GDS). Cardiovasc Diabetol 2016; 15: 59.
14 Bódis K, Roden M. Energy metabolism of white adipose tissue
and insulin resistance in humans. Eur J Clin Invest 2018;
48: e13017.
15 Kössler T, Bobrov P, Strassburger K, et al. Impact of mixed meal
tolerance test composition on measures of beta-cell function in
type 2 diabetes. Nutr Metab 2021; 18: 47.
16 Kupriyanova Y, Zaharia OP, Bobrov P, et al. Early changes in
hepatic energy metabolism and lipid content in recent-onset type 1
and 2 diabetes mellitus. J Hepatol 2021; 74: 1028–37.
17 Pesta D, Jelenik T, Zaharia OP, et al. NDUFB6 polymorphism is
associated with physical activity-mediated metabolic changes in
type 2 diabetes. Front Endocrinol 2021; 12: 693683.
18 Herder C, Fürstos JF, Nowotny B, et al. Associations between
inflammation-related biomarkers and depressive symptoms in
individuals with recently diagnosed type 1 and type 2 diabetes.
Brain Behav Immun 2017; 61: 137–45.
19 Levey AS, Stevens LA, Schmid CH, et al. A new equation to
estimate glomerular filtration rate. Ann Intern Med 2009;
150: 604–12.
20 Bönhof GJ, Strom A, Püttgen S, et al. Patterns of cutaneous nerve
fibre loss and regeneration in type 2 diabetes with painful and
painless polyneuropathy. Diabetologia 2017; 60: 2495–503.
21 Young MJ, Boulton AJ, MacLeod AF, Williams DR, Sonksen PH.
A multicentre study of the prevalence of diabetic peripheral
neuropathy in the United Kingdom hospital clinic population.
Diabetologia 1993; 36: 150–54.
22 Ziegler D, Strom A, Bönhof G, et al. Differential associations of
lower cardiac vagal tone with insulin resistance and insulin
secretion in recently diagnosed type 1 and type 2 diabetes.
Metabolism 2018; 79: 1–9.
23 Winkelmann BR, März W, Boehm BO, et al. Rationale and design
of the LURIC study--a resource for functional genomics,
pharmacogenomics and long-term prognosis of cardiovascular
disease. Pharmacogenomics 2001; 2 (suppl 1): S1–73.
24 Paulus WJ, Tschöpe C, Sanderson JE, et al. How to diagnose
diastolic heart failure: a consensus statement on the diagnosis of
heart failure with normal left ventricular ejection fraction by the
Heart Failure and Echocardiography Associations of the European
Society of Cardiology. Eur Heart J 2007; 28: 2539–50.
25 Koller L, Kleber M, Goliasch G, et al. C-reactive protein predicts
mortality in patients referred for coronary angiography and
symptoms of heart failure with preserved ejection fraction.
Eur J Heart Fail 2014; 16: 758–66.
26 Zoppini G, Cacciatori V, Raimondo D, et al. Prevalence of
cardiovascular autonomic neuropathy in a cohort of patients with
newly diagnosed type 2 diabetes: the Verona Newly Diagnosed
Type 2 Diabetes Study (VNDS). Diabetes Care 2015; 38: 1487–93.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024

27 Pigeyre M, Hess S, Gomez MF, et al. Validation of the classification
for type 2 diabetes into five subgroups: a report from the ORIGIN
trial. Diabetologia 2022; 65: 206–15.
28 Herder C, Maalmi H, Strassburger K, et al. Differences in
biomarkers of inflammation between novel subgroups of
recent-onset diabetes. Diabetes 2021; 70: 1198–208.
29 Misra S, Wagner R, Ozkan B, et al. Precision subclassification of
type 2 diabetes: a systematic review. Commun Med 2023; 3: 138.
30 McDermott K, Fang M, Boulton AJM, Selvin E, Hicks CW. Etiology,
epidemiology, and disparities in the burden of diabetic foot ulcers.
Diabetes Care 2023; 46: 209–21.
31 Roy T, Lloyd CE. Epidemiology of depression and diabetes:
a systematic review. J Affect Disord 2012; 142 (suppl): S8–21.
32 Koliaki C, Roden M. Alterations of mitochondrial function and
insulin sensitivity in human obesity and diabetes mellitus.
Annu Rev Nutr 2016; 36: 337–67.
www.thelancet.com/diabetes-endocrinology Vol 12 February 2024
Articles
33 Szendroedi J, Phielix E, Roden M. The role of mitochondria in
insulin resistance and type 2 diabetes mellitus. Nat Rev Endocrinol
2011; 8: 92–103.
34 Stephens NA, Sparks LM. Resistance to the beneficial effects of
exercise in type 2 diabetes: are some individuals programmed
to fail? J Clin Endocrinol Metab 2015; 100: 43–52.
35 Rosli N, Kwon HJ, Lim J, Yoon YA, Jeong JS. Measurement
comparability of insulin assays using conventional immunoassay
kits. J Clin Lab Anal 2022; 36: e24521.
36 Hörber S, Orth M, Fritsche A, Peter A. Comparability of C-peptide
measurements - current status and clinical relevance.
Exp Clin Endocrinol Diabetes 2023; 131: 173–78.
131