Family Practice, 2016, Vol. 33, No.

5, 449–452
doi:10.1093/fampra/cmw051
Advance Access publication 25 June 2016

Epidemiology

Association between unexplained

hypoalbuminaemia and new cancer diagnoses
in UK primary care patients

Downloaded from https://academic.oup.com/fampra/article/33/5/449/2197629 by guest on 20 December 2021

Samuel W D Merriela,*, Robert Carrolla, Fergus Hamiltonb and
William Hamiltonc
a
Centre for Academic Primary Care, School of Social and Community Medicine, University of Bristol, Bristol, UK,
b
Dunedin Public Hospital, Dunedin, New Zealand and cCollege House, University of Exeter Medical School, Exeter, UK

*Correspondence to Samuel W D Merriel, Centre for Academic Primary Care, School of Social and Community Medicine,
University of Bristol, Canynge Hall, 39 Whatley Road, BS8 2PS Bristol, UK; E-mail: sam.merriel@bristol.ac.uk

Abstract
Background.  The association between hypoalbuminaemia and a new diagnosis of cancer is as yet
unknown.
Objective.  This study aimed to assess whether unexplained hypoalbuminaemia was associated
with an increased risk of subsequent new cancer diagnosis within the next 12 months.
Methods.  A cohort study was performed using a large UK database of adult primary care patients.
Patients with a serum albumin test, either low or normal, were followed for 12 months for a new
diagnosis of non-skin cancer. Logistic regression was used to assess for relationships between
hypoalbuminaemia and cancer diagnoses.
Results.  A total of 100 122 participants had at least one albumin test result. Of these, 5753 (5.75%)
had a result <35 g/l, of whom 1634 developed cancer within 12  months. Of the 94 116 patients
with normal albumin values, 13 906 developed cancer. Hypoalbuminaemia was associated with
an increased risk of subsequent cancer diagnosis within 12  months (odds ratio [OR]: 2.29; 95%
confidence interval [CI]: 2.15–2.43). This association was smaller in magnitude after controlling for
other conditions known to affect albumin levels (OR: 1.29; 95% CI: 1.12–1.49).
Conclusions.  Low albumin levels were associated with an increased risk of cancer. This finding
needs to be confirmed in other primary care populations.

Key words: Hypoalbuminaemia, neoplasm, primary health care.

Introduction One of the possible reasons for poorer cancer survival rates in the
UK is due to diagnostic delay, leading to fewer cancers being diag-
The number of new cancer diagnoses in the UK in 2011 was over
nosed at an early stage. Hansen et al. (3) proposed a model to cat-
331 000, with a European age-standardized rate of 396.2 new
egorize diagnostic delay into subgroups: patient delay, doctor delay
cases per 100 000 people. The International Cancer Benchmarking
and system delay. A significant minority (18%) of patients consult
Partnership (ICBP) study showed that 1-year and 5-year cancer sur-
their GP three or more times before being referred to secondary care
vival was lower in each of the UK nations relative to other com-
for further assessment (4), although this could be for a number of
parable high-income countries (1). Recently published data show
reasons (5). It has also been highlighted that GPs in the UK have
England making faster improvements in 1-year survival rates for
less access to certain diagnostic tests compared to counterparts in
certain cancer types compared to the other ICBP countries (2), but
comparable high-income countries and appear to use existing tests
still lag behind in most outcomes.
less (6). Maximizing the use of existing investigations could help GPs

© The Author 2016. Published by Oxford University Press. All rights reserved.
449
For permissions, please e-mail: journals.permissions@oup.com.
450
with cancer diagnosis in the short- to medium-term. This requires
clinicians to recognize the possibility of cancer—this may follow
a patient describing symptoms of possible cancer or an abnormal
laboratory result. Anaemia and hypercalcaemia are examples of the
latter; this study examined for an association between serum albu-
min and new cancer diagnoses.
Albumin is the main protein of human blood plasma and has
many functions. Cancer, as well as a number of other diseases such
as trauma, sepsis, heart failure, liver failure and nephrotic syndrome,
may cause hypoalbuminaemia. Serum albumin levels can be very
low in advanced, metastatic cancer (7). Hypoalbuminaemia is also
a prognostic factor in late-stage cancer (8–12), heart failure (13),
circulatory disease (14) and acute medical admissions to hospital
(15). Cancer reduces circulating albumin due to metabolic and
vascular effects of the tumour (16), and the association of severe
hypoalbuminaemia and advanced cancer is established. Therefore,
albumin levels must start falling at some stage in the development of
a malignancy, and this process may start to occur before a diagnosis
of cancer has been made. There have been no studies to date investi-
gating the association between serum albumin levels and new cancer
diagnosis or any particular cancer types.
This exploratory study aimed to assess whether unexplained inci-
dent hypoalbuminaemia was associated with an increased risk of
subsequent cancer diagnosis within the next 12 months.
Methods
Participants in this cohort study were drawn from the control arm of
a separate large case–control study of UK primary care patients. The
original case–control study formed part of the DISCOVERY pro-
gramme (17), which is a multicentre research programme that aims
to improve cancer diagnosis. In the current cohort study, we included
previous control (cancer free) participants who had been age- and
sex-matched to case patients. Participant data came from the General
Practice Research Database (now known as the Clinical Practice
Research Datalink) and included information on basic patient demo-
graphics, investigation results and diagnostic Read codes.
Participants
Inclusion criteria were as follows: a serum albumin level recorded
between 1 January 2003 and 31 December 2007. Patients were clas-
sified as hypoalbuminaemia if their albumin result was below 35 g/l
and there was no previous record of hypoalbuminaemia (‘exposed’)
or normal if all albumin results up to 31 December 2007 were above
35 g/l (‘unexposed’), though they could have hypoalbuminaemia
after this date. Exclusion criteria were as follows: previous cancer
(other than non-melanoma skin cancer), a new diagnosis of non-
melanoma skin cancer and other relevant comorbidities that can
cause hypoalbuminaemia (liver dysfunction, sepsis/systemic immune
response syndrome, chest/abdominal trauma, heart failure, liver
failure or nephrotic syndrome). The data for these exclusions were
obtained by searching for any entry of a relevant clinical read code in
the participants’ medical record prior to the index date.
Dates and categorization of hypoalbuminaemia
Hypoalbuminaemia was defined as a result less than 35 g/l. Serum
albumin results were graded as ‘normal’ (albumin ≥ 35 g/l), ‘mild’
(30–34.99 g/l), ‘moderate’ (20–29.99 g/l) or ‘severe’ (<20 g/l) hypoal-
buminaemia. The date of the first test result showing hypoalbumi-
naemia was labelled the index date. Only the index date test result
was included, and any repeat albumin results within the 12-month
Family Practice, 2016, Vol. 33, No. 5
follow-up period were not considered. Participants with no recorded
low albumin level were set an index date based on their first normal
albumin test within 5  years. Individual socioeconomic status data
were not available for participants.
Outcomes
The primary outcome was the diagnosis of new cancer in the year
after the index date, excluding non-melanoma skin cancer. These
were identified using a validated list of 1079 cancer codes, which cat-
egorizes cancers into 21 sites, plus a miscellaneous group. Subsidiary
analyses examined cancer diagnoses in the second year after the
index date, and specific cancer sites.
Statistical analysis
Downloaded from https://academic.oup.com/fampra/article/33/5/449/2197629 by guest on 20 December 2021
Data were extracted and then analysed using Stata version 13.1
(Statacorp, College Station, TX). Chi-square and Student’s t-tests
were utilized to analyse differences between study subgroups.
Logistic regression techniques, controlling for known confounding
factors (age, gender and history of chest/abdominal trauma, sepsis,
liver failure, heart failure or nephrotic syndrome), were employed to
assess the risk of a subsequent diagnosis of cancer 12 months after
a participant was found to have hypoalbuminaemia. All models
included age at index date and sex. Sample size calculations for two
groups with a dichotomous endpoint, assuming a 0.5% difference
in cancer incidence with a power of 80% and an alpha of 0.05,
suggested that a minimum sample size of 9346 would be required.
Results
229 903 participants were eligible. Of these, 100 122 participants
had at least one albumin test recorded during the study period.
There was no difference in gender between included and excluded
participants, although the included participants were significantly
older (mean age 75.58 versus 73.61  years, P  <  0.001). Participant
demographics are shown in Table 1. The most common cancer types
diagnosed were urinary tract (9.6%), lymphoma (7.8%) and breast
(7.7%).
The mean albumin result for included participants was 41.24 g/l
(SD 4.47). 6% (5753) of participants had hypoalbuminaemia, and
over a quarter (1634) of these patients were subsequently diagnosed
with cancer within the following 12 months, giving a positive predic-
tive value of 28.4% (95% confidence interval [CI]: 26.77–30.03).
Figure 1 shows the relationship between specific albumin levels and
cancer outcomes, with the highest proportion of cancers seen with
albumin levels between 20 and 30 g/l.
Unadjusted regression analysis showed that the odds ratio (OR)
of a cancer diagnosis within 12  months for exposed participants
compared to participants with a normal albumin level was 2.29
Table 1.  Baseline demographic characteristics of included partici-
pants, divided by albumin level
Hypoalbumi- Normal albumin
naemia patients patients
Number 5753 94 116
Age, median (IQR) 84 (76–90) 76 (67–84)
Male, n (%) 2761 (47.99%) 45 882 (48.75%)
Cancer within 12 months, n (%) 1634 (28.40%) 13 906 (14.78%)
Albumin, median (IQR) 32 (30–34) 42 (40–44)
IQR, interquartile range.
Association between hypoalbuminaemia and new cancer diagnosis
Figure 1.  Proportion of patients diagnosed with cancer within 12 months by index albumin level. Black bars indicate normal albumin and white bars indicate
hypoalbuminaemia
(95% CI: 2.15–2.43). After adjustment for potential confounding
comorbidities, the OR of cancer diagnosis between the subgroups in
the subsequent 12 months was 1.29 (95% CI: 1.12–1.49), and there
was no difference in the risk after 24  months. There was no clear
trend towards increasing odds of a cancer diagnosis with increasing
severity of hypoalbuminaemia grading (Table 2).
Discussion
This study has shown in a large UK primary care population that
patients with unexplained hypoalbuminaemia are more likely to
develop cancer within the next 12 months. The risk of cancer is greater
over shorter time periods, and whilst the effect was partially attenuated
after controlling for relevant comorbidities, the increased risk remained.
In our crude analysis, there was no clear evidence of an association
between risk of cancer and grade of hypoalbuminaemia. Following
adjustment for potential confounders, the odds of cancer appeared to
be increased in patients with increasing severity of hypoalbuminaemia.
However, the statistical evidence backing this dose–response was weak.
This is the first study to explore the relationship between serum
albumin levels and new cancer diagnosis in primary care. Almost all
studies of albumin and cancer have been conducted in secondary
and tertiary care settings and focus on the utility of albumin levels
in predicting surgical mortality and disease-related prognosis (8–12).
One other study has looked at the relationship between serum cal-
cium levels and cancer diagnosis in primary care using this cohort
of patients, finding an increased risk of subsequent cancer diagnosis
with rising calcium levels (18).
This study has been performed using a large, widely used, repre-
sentative primary care database of patients from across the UK (19).
The study findings may be generalizable to the elderly adult UK pop-
ulation and could be relevant for comparable high-income countries.
The longitudinal nature of the data allowed temporal relationships
to be assessed and risks calculated at different time points. Whilst
this study is addressing an as-yet unexplored research question, the
findings are consistent with what might be expected given the known
relationship between advanced cancer and hypoalbuminaemia (7),
raising the biological plausibility of the association.
This study is observational in nature, so causality cannot be
assumed. There may be an element of confounding by indication in
the main findings, given sicker patients undergo more investigations
and are probably at higher risk of cancer. The proportion of patients
451
Downloaded from https://academic.oup.com/fampra/article/33/5/449/2197629 by guest on 20 December 2021
Table  2.  Logistic regression analysis of hypoalbuminaemia grad-
ing and risk of cancer diagnosis
Odds ratio 95% Confidence interval
Unadjusted analysis
  Mild hypoalbuminaemia 2.22 2.08–2.38
  Moderate hypoalbuminaemia 2.7 2.39–3.05
  Severe hypoalbuminaemia 0.96 0.58–0.161
Adjusted analysisa
  Mild hypoalbuminaemia 1.3 1.11–1.53
  Moderate hypoalbuminaemia 1.21 0.91–1.59
  Severe hypoalbuminaemia 1.99 0.81–4.91
a
Adjusted for age, gender and history of chest/abdominal trauma, sepsis,
liver failure, heart failure or nephrotic syndrome.
in our group with normal albumin levels who subsequently devel-
oped cancer is higher than the national incidence rate, exemplifying
that. However, as we believe this risk will be similar across exposure
groups (those with/without hypoalbuminaemia), this should not
have biased our findings. Individual socioeconomic status data were
not available for participants of this study, and their possible effect
on the measured associations could not be included in the analysis
(20). Clinical Practice Research Datalink data relies on accurate cod-
ing by primary care practitioners, though the entry of laboratory
results are very reliable, and cancer diagnoses are also well recorded
(21). Albumin levels in the dataset were standardized to grams per
litre, but standardization of the testing process was not possible. The
nondifferential measurement error introduced through different test-
ing methods may have led to an underestimation of the true associa-
tion between hypoalbuminaemia and cancer risk. As a result of the
DISCOVERY participants being older than the UK general popu-
lation, the cohort also had a higher prevalence of cancer. Included
participants were significantly older than those who were not. As
increasing age is associated with increased risk of cancer, this may
have affected the generalizability of the results.
Conclusions
This is the first study to demonstrate a possible association between
hypoalbuminaemia and subsequent cancer diagnosis in primary care.
The findings of this study, and the temporal relationship between
452
albumin and cancer, need to be confirmed in subsequent epidemio-
logical studies in different primary care populations before any clear
conclusions can be drawn. Diagnostic studies would also need to be
performed to assess for the utility of unexplained hypoalbuminae-
mia in identifying undiagnosed cancers and GPs’ clinical reasoning
for using the test as a potential diagnostic tool. Improving the detec-
tion of cancer at an earlier stage may help to improve outcomes for
many patients suffering from this increasingly common disease.
Declaration
Funding: SWDM is undertaking an NIHR academic clinical fellowship in gen-
eral practice.
Ethical approval: Independent Scientific Advisory Committee—protocol
09-110.
Conflict of interest: WH is an associate editor of Family Practice. The remain-
ing authors have no conflicts of interest to declare.
References
1. Coleman MP, Forman D, Bryant H et  al. Cancer survival in Australia,
Canada, Denmark, Norway, Sweden, and the UK, 1995–2007 (the Inter-
national Cancer Benchmarking Partnership): an analysis of population-
based cancer registry data. Lancet 2011; 377: 127–38.
2. Walters S, Benitez-Majano S, Muller P et al. Is England closing the interna-
tional gap in cancer survival? Br J Cancer 2015; 113: 848–60.
3. Hansen R, Olesen F, Sørensen H, Sokolowski I, Søndergaard J. Socioeco-
nomic patient characteristics predict delay in cancer diagnosis: a Danish
cohort study. BMC Health Serv Res 2008; 8: 49.
4. Lyratzopoulos G, Abel GA, McPhail S, Neal RD, Rubin GP. Measures of
promptness of cancer diagnosis in primary care: secondary analysis of
national audit data on patients with 18 common and rarer cancers. Br J
Cancer 2013; 108: 686–90.
5. Lyratzopoulos G, Wardle J, Rubin G. Rethinking diagnostic delay in
cancer: how difficult is the diagnosis? BMJ 2014; 349: g7400.
6. Rose PW, Rubin G, Perera-Salazar R et  al. Explaining variation in
cancer survival between 11 jurisdictions in the International Cancer Bench-
marking Partnership: a primary care vignette survey. BMJ Open 2015; 5:
e007212.
7. Namendys-Silva SA, Gonzalez-Herrera MO, Texcocano-Becerra J, Her-
rera-Gomez A. Hypoalbuminemia in critically ill patients with cancer:
incidence and mortality. Am J Hosp Palliat Med 2011; 28: 253–7.
Family Practice, 2016, Vol. 33, No. 5
8. Al-Shaiba R, McMillan DC, Angerson WJ et  al. The relationship between
hypoalbuminaemia, tumour volume and the systemic inflammatory response
in patients with colorectal liver metastases. Br J Cancer 2004; 19: 205–7.
9. Brown DJF, Milroy R, Preston T, McMillan DC. The relationship between
an inflammation-based prognostic score (Glasgow Prognostic Score) and
changes in serum biochemical variables in patients with advanced lung
and gastrointestinal cancer. J Clin Pathol 2007; 60: 705–8.
10. Demir H, Abidin Akkus Z, Cebi A, Cakir T, Izmirli M. Catalase, carbonic
anhydrase and other biochemical parameters in esophageal cancers in
Turkey. Asian Pac J Cancer Prev 2010; 11: 1029–32.
11. Murri AMA, Bartlett JMS, Canney PA et al. Evaluation of an inflamma-
tion-based prognostic score (GPS) in patients with metastatic breast can-
cer. Br J Cancer 2006; 94: 227–30.
12. Seebacher V, Grimm C, Reinthaller A et al. The value of serum albumin
as a novel independent marker for prognosis in patients with endometrial
cancer. Eur J Obstet Gynecol Reprod Biol 2013; 171: 101–6.
Downloaded from https://academic.oup.com/fampra/article/33/5/449/2197629 by guest on 20 December 2021
13. Horwich TB, Kalantar-Zadeh K, MacLellan RW, Fonarow GC. Albumin
levels predict survival in patients with systolic heart failure. Am Heart J
2008; 155: 883–9.
14. Van Hemelrijck M, Eichholzer M, Faeh D, Rohrmann S. Ability of a bio-
marker-based score to predict death from circulatory disease and cancer in
NHANES III. BMC Public Health 2012; 12: 895.
15. Jellinge ME, Henriksen DP, Hallas P, Brabrand M. Hypoalbuminemia is a
strong predictor of 30-day all-cause mortality in acutely admitted medical
patients: a prospective, observational, cohort study. PLoS One 2014; 9:
e105983.
16. Fanali G, di Masi A, Trezza V et al. Molecular aspects of medicine. Mol
Aspects Med 2012; 33: 209–90.
17.
Hamilton W (ed.). The Discovery Programme. http://discovery-
programme.org.uk/index.php. (published on 1 January 2012; accessed on
20 March 2016).
18. Hamilton F, Carroll R, Hamilton W, Salisbury C. The risk of cancer in
primary care patients with hypercalcaemia: a cohort study using electronic
records. Br J Cancer 2014; 111: 1410–2.
19. Reeves D, Springate DA, Ashcroft DM et  al. Can analyses of electronic
patient records be independently and externally validated? The effect of
statins on the mortality of patients with ischaemic heart disease: a cohort
study with nest case-control analysis. BMJ Open 2014; 4: 1–11.
20. Cancer Research UK. Delay Kills. UK: Cancer Research UK, 2012,
pp. 1–12.
21. Boggon R, van Staa TP, Chapman M et al. Cancer recording and mortality
in the General Practice Research Database and linked cancer registries.
Pharmacoepidemiol Drug Saf 2013; 22: 168–75.