Research

JAMA Internal Medicine | Original Investigation

Comparing Effectiveness and Safety of SGLT2 Inhibitors

vs DPP-4 Inhibitors in Patients With Type 2 Diabetes
and Varying Baseline HbA1c Levels
Elvira D’Andrea, MD, PhD; Deborah J. Wexler, MD, MSc; Seoyoung C. Kim, MD, ScD; Julie M. Paik, MD, ScD, MPH;
Ethan Alt, PhD; Elisabetta Patorno, MD, DrPH

Supplemental content
IMPORTANCE Sodium-glucose cotransporter 2 inhibitor (SGLT2i) therapy has been associated
with cardiovascular benefits and a few adverse events; however, whether the comparative
effectiveness and safety profiles vary with differences in baseline hemoglobin A1c (HbA1c)
levels is unknown.

OBJECTIVE To compare cardiovascular effectiveness and safety of treatment with SGLT2i vs

dipeptidyl peptidase 4 inhibitor (DPP-4i) in adults with type 2 diabetes (T2D) (1) overall and
(2) at varying baseline HbA1c levels.

DESIGN, SETTING, AND PARTICIPANTS A new-user comparative effectiveness and safety

research study was conducted among 144 614 commercially insured adults, initiating
treatment with SGLT2i or DPP-4i and with a recorded T2D diagnosis at baseline and
at least 1 HbA1c laboratory result recorded within 3 months before treatment initiation.

INTERVENTIONS The intervention consisted of the initiation of treatment with SGLT2i

or DPP-4i.

MAIN OUTCOMES AND MEASURES Primary outcomes were a composite of myocardial

infarction, stroke, or all-cause death (modified major adverse cardiovascular events [MACE])
and hospitalization for heart failure (HHF). Safety outcomes were hypovolemia, fractures,
falls, genital infections, diabetic ketoacidosis (DKA), acute kidney injury (AKI), and lower-limb
amputation. Incidence rate (IR) per 1000 person-years, hazard ratios (HR) and rate
differences (RD) with their 95% CIs were estimated controlling for 128 covariates.

RESULTS A total of 144 614 eligible adults (mean [SD] age, 62 [12.4] years; 54% male
participants) with T2D initiating treatment with a SGLT2i (n = 60 523) or a DPP-4i
(n = 84 091) were identified; 44 099 had an HbA1c baseline value of less than 7.5%, 52 986
between 7.5% and 9%, and 47 529 greater than 9%. Overall, 87 274 eligible patients were 1:1
propensity score–matched: 24 052 with HbA1c less than 7.5%; 32 290 with HbA1c between
7.5% and 9%; and 30 932 with HbA1c greater than 9% (to convert percentage of total
hemoglobin to proportion of total hemoglobin, multiply by 0.01). The initiation of SGLT2i Author Affiliations: Division of
vs DPP-4i was associated with a reduction in the risk of modified MACE (IR per 1000 Pharmacoepidemiology and
Pharmacoeconomics; Brigham and
person-years 17.13 vs 20.18, respectively; HR, 0.85; 95% CI, 0.75-0.95; RD, −3.02; 95% CI, Women’s Hospital and Harvard
−5.23 to –0.80) and HHF (IR per 1000 person-years 3.68 vs 8.08, respectively; HR, 0.46; Medical School, Boston,
95% CI, 0.35 to 0.57; RD −4.37; 95% CI, −5.62 to −3.12) over a mean follow-up of 8 months, Massachusetts (D’Andrea, Kim, Paik,
Alt, Patorno); Diabetes Center,
with no evidence of treatment effect heterogeneity across the HbA1c levels. Treatment with
Massachusetts General Hospital,
SGLT2i showed an increased risk of genital infections and DKA and a reduced AKI risk Harvard Medical School, Boston
compared with DPP-4i. Findings were consistent by HbA1c levels, except for a more (Wexler); Division of Kidney
pronounced risk of genital infections associated with SGLT2i for HbA1c levels of 7.5% to 9% Medicine, Brigham and Women’s
Hospital, Boston, Massachusetts
(IR per 1000 person-years 68.5 vs 22.8, respectively; HR, 3.10; 95% CI, 2.68-3.58; (Paik); New England Geriatric
RD, 46.22; 95% CI, 40.54-51.90). Research Education and Clinical
Center, VA Boston Healthcare
CONCLUSIONS AND RELEVANCE In this comparative effectiveness and safety research study System, Boston, Massachusetts
among adults with T2D, SGLT2i vs DPP-4i treatment initiators had a reduced risk of modified (Paik).
MACE and HHF, an increased risk of genital infections and DKA, and a lower risk of AKI, Corresponding Author: Elvira
regardless of baseline HbA1c. D’Andrea, MD, PhD, Division of
Pharmacoepidemiology and
Pharmacoeconomics,
Brigham and Women’s Hospital
and Harvard Medical School,
JAMA Intern Med. 2023;183(3):242-254. doi:10.1001/jamainternmed.2022.6664 1620 Tremont St, Boston, MA, 02120
Published online February 6, 2023. (elvira.dandrea@mail.harvard.edu).

242 (Reprinted) jamainternalmedicine.com

Downloaded From: https://jamanetwork.com/ on 03/20/2023

 Effectiveness and Safety of SGLT2 Inhibitors vs DPP-4 Inhibitors in Patients With Type 2 Diabetes
T
ype 2 diabetes (T2D) affects more than 11% of the US
population and is associated with increased morbidity
and mortality from cardiovascular and kidney disease.1,2
Mitigating the risk of these complications is a priority in the
management of diabetes. In large-scale postmarketing ran-
domized cardiovascular outcome trials (CVOTs), sodium-
glucose cotransporter 2 inhibitors (SGLT2i) have demon-
strated a cardiorenal protective effect in patients with T2D and
established cardiovascular or kidney conditions.3-6 These ben-
efits have been reproduced in real-world evidence studies,
which encompass a patient population with a broader spec-
trum of cardiovascular risk as seen in clinical practice.7,8 How-
ever, it is still unclear whether patients with different levels
of hyperglycemia can similarly benefit from the use of SGLT2i.
Cardiovascular outcome trials of SGLT2i have explored po-
tential for treatment effect heterogeneity by hyperglycemia,
defined as baseline assessment of glycated hemoglobin (HbA1c),
identifying some potential variation in the cardiovascular ef-
fects of SGLT2i by HbA1c level.3-6,9 However, subgroup analy-
ses within CVOTs are generally underpowered to detect mean-
ingful differences,9 and patients with uncontrolled diabetes
were often underrepresented due to strict inclusion criteria.3,4,6
Further, since SGLT2i induce a glycosuric response by reduc-
ing kidney tubular glucose reabsorption, those medications
can have a more pronounced effect on hyperglycemia in pa-
tients with poor glycemic control due to the increased amount
of filtered glucose.10 The accompanying diuretic and natri-
uretic effect of SGLT2 inhibition may lead to a more marked
improvement in volume status in patients with elevated vs
controlled glycemia resulting in a lower risk for heart failure.
Conversely, the higher concentration of glucose in the urine
in patients with severe hyperglycemia could lead to an in-
creased risk of adverse effects such as mycotic infections, vol-
ume depletion due to the osmotic diuresis induced by glycos-
uria, consequent increased risk of falls and fractures, and
diabetic ketoacidosis (DKA).11
In this large comparative effectiveness and safety re-
search study of patients with T2D, we evaluated cardiovascu-
lar and safety events associated with the initiation of SGLT2i
treatment compared with dipeptidyl peptidase 4 inhibitor
(DPP-4i), which clearly lack glycosuric adverse effects and are
generally considered safe (1) in the overall population and (2)
across subgroups of patients with controlled, above-target, or
elevated HbA1c levels at baseline.
Methods
Study Design and Data Source
We performed a new-user comparative effectiveness and
safety research study using a US health insurance data set (dei-
dentified Optum Clinformatics Data Mart Database) with na-
tionwide commercial coverage including Medicare Advan-
tage plans. Through linkage with national laboratory test
provider chains, results for outpatient laboratory tests are avail-
able for a subset of approximately 45% of beneficiaries (in-
cluding laboratory test results of HbA1c), representative of the
full insured population (see eMethods and eTable 1 in Supple-
jamainternalmedicine.com
Downloaded From: https://jamanetwork.com/ on 03/20/2023
Original Investigation Research
Key Points
Question Does the effectiveness and safety of sodium-glucose
cotransporter 2 inhibitors (SGLT2i) differ from that of dipeptidyl
peptidase 4 inhibitors (DPP-4i) in patients with type 2 diabetes
(T2D) overall and at varying baseline hemoglobin A1c (HbA1c)
levels?
Findings In this large new-user comparative effectiveness and
safety research study including 87 274 propensity-scored matched
adults with T2D, SGLT2i treatment initiators had a reduced risk of
major cardiovascular events, heart failure, and acute kidney injury
and an increased risk of genital infections and diabetic
ketoacidosis compared with DPP-4i treatment initiators,
regardless of their baseline HbA1c level.
Meaning Despite concern that use of SGLT2i at higher HbA1c
levels would cause excess risk, the findings of this study suggest
that patients with T2D can benefit from the use of SGLT2i
regardless of glycemic control, with the expected adverse effect
profile when compared with DPP-4i, with no additional risk of
adverse effects in patients with elevated HbA1c levels.
ment 1). The Mass General Brigham institutional review board
provided ethics approval. Informed consent was waived be-
cause the study used deidentified secondary data.
Study Population
The study population included patients 18 years and older who
initiated treatment with a SGLT2i (canagliflozin, dapagli-
flozin, empagliflozin, or ertugliflozin) or a DPP-4i (alogliptin,
saxagliptin, linagliptin, or sitagliptin) between April 1, 2013 (con-
sistent with the US Food and Drug Administration [FDA] ap-
proval of the first SGLT2i), and June 30, 2021. Treatment with
DPP-4i was selected as the comparator because these medica-
tions are also frequently used as second-line therapy for T2D,
have similar out-of-pocket costs as SGLT2i but a different mecha-
nism of action, which does not involve inhibition of kidney glu-
cose reabsorption and osmotic diuresis, and have shown no as-
sociation with atherosclerotic cardiovascular outcomes. Cohort
entry was the day of the first filled prescription of either SGLT2i
or DPP-4i, with no use in the previous 6 months. Study eligi-
bility was limited to patients with at least 6 months of continu-
ous health plan enrollment, a recorded T2D diagnosis before co-
hort entry, and at least 1 HbA1c laboratory result recorded within
3 months before cohort entry. We excluded patients with rec-
ords of type 1, secondary, or gestational diabetes; malignant neo-
plasms; end-stage kidney disease; kidney replacement therapy;
no laboratory results for creatinine; or nursing home resi-
dence within 6 months preceding cohort entry (eFigure 1 and
eTable 2 in Supplement 1). Based on the most recent HbA1c base-
line value, we identified 3 different subcohorts which com-
prised patients with controlled (HbA1c <7.5%), above-target
(HbA1c 7.5%-9%), or elevated (HbA1c >9%) glycemia, respec-
tively (to convert percentage of total hemoglobin to propor-
tion of total hemoglobin, multiply by 0.01). The cutoffs for HbA1c
stratification were chosen by both inspecting terciles of the
HbA1c distribution among SGLT2i treatment initiators and con-
sidering the thresholds currently recommended to define
controlled vs uncontrolled hyperglycemia.12,13
(Reprinted) JAMA Internal Medicine March 2023 Volume 183, Number 3 243
 Research Original Investigation Effectiveness and Safety of SGLT2 Inhibitors vs DPP-4 Inhibitors in Patients With Type 2 Diabetes
Outcomes and Follow-up
The primary effectiveness outcomes were (1) modified major
adverse cardiovascular events (MACE), a composite cardio-
vascular end point of myocardial infarction, ischemic or hem-
orrhagic stroke, and all-cause death, and (2) hospitalization
for heart failure (HHF). Secondary effectiveness outcomes were
myocardial infarction, ischemic or hemorrhagic stroke, and
all-cause mortality. In prior studies, the positive predictive val-
ues of claims-based algorithms were at least 87% for myocar-
dial infarction and stroke,14-16 and 84% to 100% for HHF.17
Safety outcomes included hypovolemia, nonvertebral frac-
tures, falls, genital infections, DKA, acute kidney injury (AKI),
and lower-limb amputations. Definitions were either vali-
dated against medical records18-21 or used in prior pharmaco-
epidemiologic studies assessing SGLT2i 22-24 (eTable 3 in
Supplement 1).
Adopting an as-treated approach, the follow-up started the
day after cohort entry and continued until treatment discon-
tinuation (allowing a 30-day grace period after termination of
the last prescription’s supply), switch to or augmentation with
a drug in the comparator class, occurrence of study outcome,
death, end of continuous health plan enrollment, or end of
available data, whichever came first.
Baseline Patient Characteristics
Patient characteristics were selected a priori as potential con-
founders and measured at treatment initiation (demograph-
ics), as last recorded value within 3 months before cohort en-
try (HbA1c), or within 6 months before cohort entry (all other
patient characteristics). Covariates included demographics,
cardiovascular and other comorbidities, general health state
indexes, such as combined comorbidity score and claims-
based frailty index,25,26 HbA1c laboratory results, diabetes-
specific complications, use of glucose-lowering and other medi-
cations, indicators of health care utilization as proxy for disease
state, surveillance, and intensity of care. Based on baseline
creatinine laboratory results, we calculated the estimated glo-
merular filtration rate (eGFRCr) using a version of the creatinine-
based Chronic Kidney Disease–Epidemiology Collaboration
(CKD-EPI) equation without the race term as correction
factor.27,28 Other laboratory test results were also measured at
baseline but were available for only a subset of the study popu-
lation (see eTables 4-6 in Supplement 1 for a complete list of
the baseline covariates).
Statistical Analysis
Propensity score (PS) matching was used to control for con-
founding. The PS for initiating SGLT2i vs DPP-4i therapy was
calculated within each HbA1c subcohort separately through a
logistic regression model with 128 prespecified covariates.
Laboratory data, except for HbA1c and eGFRCr, were not in-
cluded in the model because of the substantial proportion of
missing information. Initiators of SGLT2i therapy were 1:1
matched to initiators of DPP-4i therapy on their estimated PS
within each HbA1c subcohort using the nearest neighbor ap-
proach with a caliper width of 0.01 on the PS scale. Covariate
balance was assessed with standardized differences, with
meaningful imbalances set at values higher than 10%.29,30
244 JAMA Internal Medicine March 2023 Volume 183, Number 3 (Reprinted)
Downloaded From: https://jamanetwork.com/ on 03/20/2023
We also reviewed the balance in laboratory test results not in-
cluded in the PS model, to evaluate potential residual con-
founding after PS matching.
We tabulated numbers of events, incidence rates (IRs), and
rate differences (RDs) per 1000 person-years. Hazard ratios
(HRs) and 95% CIs were estimated by Cox proportional haz-
ard models. We used Kaplan-Meier methods to plot cumula-
tive incidence of primary outcomes and log-rank tests to com-
pare hazard rates between drug classes. Two-sided P values
for homogeneity were obtained by performing Wald tests and
values <.05 were considered indicative of treatment hetero-
geneity.
We inspected the robustness of the main findings through
sensitivity analyses (see eMethods in Supplement 1), address-
ing potential informative censoring, time-lag bias,31 unmea-
sured confounding for high risk for recurrence, and DPP-4i ef-
fects on HHF (since saxagliptin and alogliptin showed an
increased HHF rate in CVOTs,32,33 which resulted in an FDA
warning,34 we conducted a sensitivity analysis for the HHF out-
come redefining the comparator group as sitagliptin only).
All analyses were implemented using Aetion Evidence
Platform (Aetion Inc) and Stata statistical software, version 15.1
(StataCorp LLC).
Results
Study Population and Baseline Characteristics
A total of 144 614 eligible adults (mean [SD] age, 62 [12.4] years;
54% male participants) with T2D initiating treatment with a
SGLT2i (n = 60 523) or a DPP-4i (n = 84 091) were identified;
44 099 had an HbA1c baseline value of less than 7.5%, 52 986
between 7.5% and 9%, and 47 529 greater than 9%. Overall,
patients newly prescribed SGLT2i vs DPP-4i were younger,
more likely to have obesity, a higher eGFRCr, and to be treated
with more than 1 glucose-lowering medication (particularly
glucagon-like peptide-1 receptor agonists and insulin), and less
likely to have a diagnosis of diabetic nephropathy or CKD
(eTables 4-6 in Supplement 1).
After PS matching, 87 274 patients were retained: 24 052
with glycemia at target (HbA1c <7.5% mean, 6.8%), 32 290 with
glycemia above target (HbA1c 7.5%-9% mean, 8.2%), and 30 932
with elevated glycemia (HbA1c >9% mean, 10.6%) (Figure 1);
all baseline characteristics were well balanced (Table), includ-
ing the laboratory test results not included in the PS model
(eTables 4-6 in Supplement 1), and the PS distributions over-
lapped completely (eFigure 2 in Supplement 1).
In the overall population, 6.7% had moderate to ad-
vanced CKD, 8.7% had a history of mycotic infection, 0.7% had
a history of fractures, 2.1% had a history of falls, and 14.0% were
prescribed insulin on the day of cohort entry. Patients with
HbA1c levels of more than 9% treated with SGLT2i were younger
than those with HbA1c levels between 7.5% and 9% and those
with HbA1c levels less than 7.5% (57.7 vs 62.0 vs 62.5 years, re-
spectively), mostly male participants (57.2% vs 54.5% vs
50.2%), less frequently White (46.2% vs 51.8% vs 53.7%), and
more likely to receive insulin (20.1% vs 14.0% vs 7.8%). They
had higher eGFRCr (83.2 vs 77.4 vs 74.9) and lower burden of
jamainternalmedicine.com
 Effectiveness and Safety of SGLT2 Inhibitors vs DPP-4 Inhibitors in Patients With Type 2 Diabetes Original Investigation Research

Figure 1. Study Flowchart

903 430 Patients initiating SGLT2 or DPP-4

inhibitors between April 2013 and
June 2021 and with ≥365 d of
continuous plan enrollment prior
to cohort entry date

758 816 Patients excluded

640 732 Use of index drugs within 365 d before
cohort entry
808 Use of both index drugs on cohort entry
17 Aged <18 y at cohort entry
49 Age or sex missing
82 583 No HbA1c laboratory result recorded
within 90 d before or on cohort entry
847 No type 2 diabetes diagnosis
400 End-stage kidney disease, dialysis,
or kidney transplant
5540 Cancer (except nonmelanoma)
1429 Type 1 diabetes
1309 Secondary or gestational diabetes
709 Nursing home stay before cohort entry
23 850 No creatinine value recorded at baseline
543 Did not begin follow-up

144 614 Eligible adult patients with

type 2 diabetes initiating SGLT2
(n = 60 523) or DPP-4
(n = 84 091) inhibitors

44 099 Eligible patients with HbA1c 52 986 Eligible patients with HbA1c 47 529 Eligible patients with
value <7.5% value between 7.5% and 9% HbA1c value >9%

Eligible population included in the

24 052 1:1 PS-matched patients
with HbA1c <7.5% initiating
SGLT2 or DPP-4 inhibitors
87 274 1:1 PS-matched patients with
32 290 1:1 PS-matched patients
with HbA1c between 7.5%
and 9% initiating SGLT2 or
DPP-4 inhibitors
type 2 diabetes initiating
SGLT2 (n = 43 637) or
DPP-4 (n = 43 637) inhibitors
30 932 1:1 PS-matched patients
with HbA1c >9% initiating
SGLT2 or DPP-4 inhibitors
study cohort initiating SGLT2 or
DPP-4 inhibitors between April 2013
and June 2021 before and after
matching, overall and stratified by
HbA1c baseline level. DPP-4 indicates
dipeptidyl peptidase-4;
HbA1c, hemoglobin A1c;
PS, propensity score;
SGLT2, sodium-glucose
cotransporter 2.

comorbidities and frailty. Compared with SGLT2i treatment
initiators with HbA1c levels of less than 7.5% and greater than
9%, those with HbA1c levels between 7.5% and 9% had higher
prevalence of diabetes-related complications such as dia-
betic nephropathy (15.3% vs 15.1% vs 13.2%) and retinopathy
(7.8% vs 5.6% vs 7.3%) and lower prevalence of untreated dia-
betes at baseline (Table).
Duration of follow-up on treatment varied slightly based
on the outcome. In the overall population, the mean
follow-up was 240 days for modified MACE and 241 days for
HHF. Most patients were censored due to treatment discon-
tinuation (approximately 60%). Details on follow-up and
censoring reasons are reported in eTable 7 in Supplement 1.
Primary Effectiveness Outcomes Analyses
After PS matching, the IRs per 1000 person-years for modi-
fied MACE were overall 17.13 vs 20.18 in SGLT2i vs DPP-4i
initiators, respectively, showing among new users of SGLT2i
vs DPP-4i a 15% decreased risk (HR, 0.85; 95% CI, 0.75-0.95),
or 3 fewer events in 1000 person-years (RD –3.02; 95% CI,
–5.23 to −0.80). The results across subgroups were consis-
tent with the overall findings with no evidence of effect
heterogeneity (HbA1c <7.5% HR, 0.84; 95% CI, 0.66-1.07;
HbA1c 7.5%-9% HR, 0.88; 95% CI, 0.72-1.07; and HbA1c >9%
HR, 0.83; 95% CI, 0.68-1.00; P for homogeneity = .91),
although the degree of uncertainty was higher and the point
estimates less precise due to the reduced statistical power
(Figure 2).
Overall, 3.68 vs 8.08 HHF events per 1000 person-years
were estimated in SGLT2i vs DPP-4i treatment initiators, re-
spectively. The initiation of SGLT2i vs DPP-4i was associated
with a 54% decreased risk of HHF (HR, 0.46; 95% CI, 0.35-
0.57), corresponding to approximately 4 fewer cases per 1000
person-years (RD −4.37; 95% CI, −5.62 to −3.12). This was
consistent across subgroups with no evidence of effect
heterogeneity (HbA1c <7.5% HR, 0.48; 95% CI, 0.33-0.72; HbA1c
7.5%-9% HR, 0.44; 95% CI, 0.30-0.64; HbA1c >9% HR, 0.47;
95% CI, 0.31-0.71; P = .95 for homogeneity) (Figure 2).
Kaplan-Meier curves comparing the cumulative inci-
dence of modified MACE and HHF between initiators of SGLT2
vs DPP-4i therapy were consistent with these results and across
subgroups (Figure 3 and eFigure 3 in Supplement 1). Clinical
benefits were observed within the first 3 months of follow-up
(Figure 3).

jamainternalmedicine.com (Reprinted) JAMA Internal Medicine March 2023 Volume 183, Number 3 245

Downloaded From: https://jamanetwork.com/ on 03/20/2023

 246
Table. Selected Baseline Characteristics of Unmatched and 1:1 Propensity Score–Matched Patients Initiating SGLT2 Inhibitor vs DPP-4 Inhibitor Therapy Overall and Stratified by HbA1c Levels
Characteristics before Subgroup HbA1c <7.5% Subgroup HbA1c 7.5%-9% Subgroup HbA1c >9% Overall population
PS-matching
SGLT2 inhibitor DPP-4 inhibitor St. Diff SGLT2 inhibitor DPP-4 inhibitor St. Diff SGLT2 inhibitor DPP-4 inhibitor St. Diff SGLT2 inhibitor DPP-4 inhibitor St. Diff
(n = 16 316) (n = 27 783) (n = 22 312) (n = 30 674) (n = 21 895) (n = 25 634) (n = 60 523) (n = 84 091)
Age, mean (SD) 61.1 (11.5) 66.7 (11.6) −0.5 60.8 (11.4) 64.9 (11.9) −0.4 56.7 (11.9) 60.0 (13.0) −0.3 59.4 (11.6) 64.0 (12.1) −0.4
Male, No. (%) 8466 (51.9) 12 942 (46.6) 0.1 12 554 (56.3) 15 567 (50.7) 0.1 12 782 (58.4) 14 047 (54.8) 0.1 33 802 (55.8) 42 556 (50.6) 0.1
Female, No. (%) 7850 (48.1) 14 841 (53.4) −0.1 9758 (43.7) 15 107 (49.3) −0.1 9113 (41.6) 11 587 (45.2) −0.1 26 721 (44.2) 41 535 (49.4) −0.1
Race, No. (%)
Research Original Investigation

Asian 1098 (6.7) 2679 (9.6) −0.1 1329 (6.0) 2548 (8.3) −0.1 991 (4.5) 1545 (6.0) −0.1 3418 (5.6) 6772 (8.1) −0.1
Black 1902 (11.7) 3762(13.5) −0.1 2495 (11.2) 3851 (12.6) −0.04 2946 (13.5) 3859 (15.1) −0.1 7343 (12.1) 11 472 (13.6) −0.04
Hispanic 3252 (19.9) 6244 (22.5) −0.1 4970 (22.3) 7838 (25.6) −0.1 5932 (27.1) 7749 (30.2) −0.1 14 154 (23.4) 21 831 (26.0) −0.1
White 9144 (56.0) 13 584 (48.9) 0.1 12 186 (54.6) 14 686 (47.9) 0.1 10 712 (48.9) 11 060 (43.1) 0.1 32 042 (52.9) 39 330 (46.8) 0.1

Downloaded From: https://jamanetwork.com/ on 03/20/2023

Othera or unknown 20 (5.6) 1514 (5.4) 0.01 1332 (6.0) 1751 (5.7) 0.01 1314 (6.0) 1421 (5.5) 0.02 3566 (5.9) 4686 (5.6) 0.01
Obesity, No. (%) 6254 (38.3) 6889 (24.8) 0.3 8107 (36.3) 8178 (26.7) 0.2 7952 (36.3) 6932 (27.0) 0.2 22 313 (36.9) 21 999 (26.2) 0.2
Laboratory results,
mean (SD)
HbA1c value, %b 6.8 (0.6) 6.8 (0.6) 0.1 8.2 (0.6) 8.2 (0.6) 0.1 10.6 (1.4) 10.6 (1.5) −0.01 8.7 (1.0) 8.5 (1.0) 0.3
mmol/mol 51 (NA) 50 (NA) NA 66 (NA) 66 (NA) NA 91 (NA) 93 (NA) NA 72 (NA) 69 (NA) NA
eGFRCrc 76.2 (23.9) 68.0 (24.4) 0.3 78.7 (23.4) 73.4 (24.2) 0.2 84.0 (24.5) 80.1 (25.3) 0.2 80.0 (23.9) 73.7 (24.6) 0.3

JAMA Internal Medicine March 2023 Volume 183, Number 3 (Reprinted)

Burden of comorbidities,
mean (SD)
Combined comorbidity 1.1 (2.0) 1.4 (2.3) −0.2 1.0 (1.8) 1.2 (2.0) −0.1 1.0 (1.7) 1.0 (1.9) −0.03 1.0 (1.8) 1.2 (2.1) −0.1
scored
Frailty scoree 0.2 (0.04) 0.2 (0.1) <0.01 0.1 (0.04) 0.2 (0.04) −0.3 0.1 (0.04) 0.1 (0.04) <0.01 0.1 (0.04) 0.2 (0.04) −0.3
Diabetes-related
comorbidities, No. (%)
Diabetic nephropathy 2353 (14.4) 5948 (21.4) −0.2 3350 (15.0) 5884 (19.2) −0.11 2830 (12.9) 4019 (15.7) −0.1 8533 (14.1) 15 851 (18.8) −0.13
Diabetic retinopathy 964 (5.9) 1803 (6.5) −0.02 1842 (8.3) 2473 (8.1) 0.01 1718 (7.8) 1986 (7.7) <0.01 4524 (7.5) 6262 (7.4) <0.01
Lower-limb amputations 66 (0.4) 120 (0.4) <0.01 104 (0.5) 146 (0.5) <0.01 145 (0.7) 169 (0.7) <0.01 315 (0.5) 435 (0.5) <0.01
Diabetic ketoacidosis 22 (0.1) 35 (0.1) <0.01 32 (0.1) 42 (0.1) <0.01 66 (0.3) 78 (0.3) <0.01 120 (0.2) 155 (0.2) <0.01
Other comorbidities, No. (%)
History of cardiovascular 3873 (23.7) 7194 (25.9) −0.1 4783 (21.4) 7110 (23.2) −0.04 3953 (18.1) 4636 (18.1) <0.01 12 609 (20.8) 18 940 (22.5) −0.04
disease
Heart failure 1284 (7.9) 2378 (8.6) −0.03 1325 (5.9) 2093 (6.8) −0.04 1402 (5.5) 1633 (5.6) <0.01 4011 (6.6) 6104 (7.3) −0.03
Chronic kidney disease, 1156 (7.1) 4930 (17.7) −0.3 1333 (6.0) 3868 (12.6) −0.2 941 (4.3) 2243 (8.8) −0.2 3430 (5.7) 11 041 (13.1) −0.3
stage <3
Acute kidney injury 368 (2.3) 1235 (4.4) −0.1 368 (1.6) 841 (2.7) −0.1 328 (1.5) 596 (2.3) −0.1 1064 (1.8) 2672 (3.2) −0.1
Mycotic infections 1389 (8.5) 2638 (9.5) −0.03 1857 (8.3) 2912 (9.5) −0.04 1986 (9.1) 2440 (9.5) −0.01 5232 (8.6) 7990 (9.5) −0.03
Fractures 138 (0.8) 288 (1.0) −0.02 144 (0.6) 258 (0.8) −0.02 144 (0.7) 191 (0.7) <0.01 426 (0.7) 737 (0.9) −0.02
Falls 353 (2.2) 802 (2.9) −0.04 441 (2.0) 736 (2.4) −0.03 437 (2.0) 540 (2.1) −0.01 1231 (2.0) 2078 (2.5) −0.03

(continued)

jamainternalmedicine.com
Effectiveness and Safety of SGLT2 Inhibitors vs DPP-4 Inhibitors in Patients With Type 2 Diabetes
 Table. Selected Baseline Characteristics of Unmatched and 1:1 Propensity Score–Matched Patients Initiating SGLT2 Inhibitor vs DPP-4 Inhibitor Therapy Overall and Stratified by HbA1c Levels (continued)
Diabetes treatment
No use of any 2610 (16.0) 6053 (21.8) −0.2 1898 (8.5) 3843 (12.5) −0.1 2584 (11.8) 4630 (18.1) −0.2 7092 (11.7) 14 526 (17.3) −0.2
glucose-lowering drugs
at baseline; No. (%)
No. glucose-lowering 1.1 (0.9) 1.0 (0.8) 0.2 1.4 (0.9) 1.2 (0.8) 0.2 1.4 (0.9) 1.3 (0.8) 0.1 1.3 (0.9) 1.2 (0.8) <0.01
drugs, mean (SD)f

jamainternalmedicine.com
Metformin, No. (%)f 10 120 (62.0) 16 545 (59.6) 0.1 14 915 (66.8) 20 337 (66.3) 0.01 14 364 (65.6) 17 562 (68.5) −0.1 39 399 (65.1) 54 444 (64.7) 0.01
Sulfonylureas (second 2863 (17.5) 6334 (22.8) −0.1 6756 (30.3) 11 174 (36.4) −0.1 6263 (28.6) 9159 (35.7) −0.2 15 882 (26.2) 26 667 (31.7) −0.1
generation), No. (%)f
GLP-1 receptor agonists, 2464 (15.1) 580 (2.1) 0.5 3637 (16.3) 802 (2.6) 0.5 3510 (16.0) 806 (3.1) 0.5 9611 (15.9) 2188 (2.6) 0.5
No. (%)f
Insulin, No. (%)f 1750 (10.7) 1536 (5.5) 0.2 4226 (18.9) 3405 (11.1) 0.2 5433 (24.8) 4476 (17.5) 0.2 11 409 (18.9) 9417 (11.2) 0.2
Characteristics after Overall Population

Downloaded From: https://jamanetwork.com/ on 03/20/2023

Subgroup HbA1c <7.5% Subgroup HbA1c 7.5%-9% Subgroup HbA1c >9%
PS-matching
SGLT2 inhibitor DPP-4 inhibitor St. Diff SGLT2 inhibitor DPP-4 inhibitor St. Diff SGLT2 inhibitor DPP-4 inhibitor St. Diff SGLT2 inhibitor DPP-4 inhibitor St. Diff
(n = 12 026) (n = 12 026) (n = 16 145) (n = 16 145) (n = 15 466) (n = 15 466) (n = 43 637) (n = 43 637)
Age, mean (SD) 62.5 (11.4) 62.3 (11.3) 0.01 62.0 (11.5) 62.0 (11.4) 0.01 57.6 (12.2) 57.7 (12.1) −0.01 60.6 (11.7) 60.6 (11.6) <0.01
Male, No. (%) 6034 (50.2) 6047 (50.3) <0.01 8794 (54.5) 8784 (54.4) <0.01 8846 (57.2) 8859 (57.3) <0.01 23 674 (54.3) 23 690 (54.3) <0.01
Female, No. (%) 5992 (49.8) 5979 (49.7) <0.01 7351 (45.5) 7361 (45.6) <0.01 6620 (42.8) 6607 (42.7) <0.01 19 963 (45.7) 19 947 (45.7) <0.01
Race, No. (%)
Asian 914 (7.6) 906 (7.5) <0.01 1092 (6.8) 1113 (6.9) <0.01 795 (5.1) 785 (5.1) <0.01 2801 (6.4) 2804 (6.4) <0.01
Black 1483 (12.3) 1471 (12.2) <0.01 1930 (12.0) 1890 (11.7) 0.01 2193 (14.2) 2184 (14.1) <0.01 5606 (12.8) 5545 (12.7) <0.01
Hispanic 2482 (20.6) 2513 (20.9) −0.01 3807 (23.6) 3768 (23.3) 0.01 4418 (28.6) 4439 (28.7) <0.01 10 707 (24.5) 10 720 (24.6) <0.01
White 6463 (53.7) 6436 (53.5) <0.01 8358 (51.8) 8380 (51.9) <0.01 7138 (46.2) 7141 (46.2) <0.01 21 959 (50.3) 21 957 (50.3) <0.01
Othera or unknown 684 (5.7) 700 (5.8) <0.01 958 (5.9) 994 (6.2) −0.01 922 (6.0) 917 (5.9) <0.01 2564 (5.9) 2611 (6.0) <0.01
Effectiveness and Safety of SGLT2 Inhibitors vs DPP-4 Inhibitors in Patients With Type 2 Diabetes

Obesity, No. (%) 3995 (33.2) 3959 (32.9) 0.01 5198 (32.2) 5209 (32.3) <0.01 4941 (31.9) 4922 (31.8) <0.01 14 134 (32.4) 14 090 (32.3) <0.01
Laboratory results,
mean (SD)
HbA1c value %b 6.8 (0.6) 6.8 (0.6) 0.02 8.2 (0.6) 8.2 (0.6) <0.01 10.6 (1.4) 10.6 (1.4) <0.01 8.7 (1.0) 8.7 (1.0) <0.01
mmol/mol 51 (NA) 51 (NA) NA 66 (NA) 66 (NA) NA 93 (NA) 93 (NA) NA 71 (NA) 71 (NA) NA
eGFRCrc 74.9 (23.6) 75.1 (23.6) −0.01 77.4 (23.3) 77.8 (23.5) −0.02 83.2 (24.4) 83.2 (24.5) <0.01 78.8 (23.8) 79.0 (23.9) −0.01
Burden of comorbidities,
mean (SD)
Combined comorbidity 1.1 (2.0) 1.1 (2.0) <0.01 1.0 (1.9) 1.0 (1.8) 0.01 0.9 (1.7) 0.9 (1.7) 0.01 1.0 (1.9) 1.0 (1.8) 0.01
scored
e
Frailty index 0.2 (0.04) 0.2 (0.04) <0.01 0.1 (0.04) 0.1 (0.04) <0.01 0.1 (0.04) 0.1 (0.04) <0.01 0.14 (0.04) 0.14 (0.04) <0.01
Diabetes-related
comorbidities, No. (%)
Diabetic nephropathy 1818 (15.1) 1788 (14.9) 0.01 2477 (15.3) 2460 (15.2) <0.01 2042 (13.2) 2051 (13.3) <0.01 6337 (14.5) 6299 (14.4) <0.01
Diabetic retinopathy 672 (5.6) 677 (5.6) <0.01 1267 (7.8) 1261 (7.8) <0.01 1134 (7.3) 1139 (7.4) <0.01 3073 (7.0) 3077 (7.1) <0.01
Lower-limb amputations 44 (0.4) 40 (0.3) 0.02 73 (0.5) 76 (0.5) <0.01 90 (0.6) 94 (0.6) <0.01 207 (0.5) 210 (0.5) <0.01
Diabetic ketoacidosis 14 (0.1) 17 (0.1) <0.01 21 (0.1) 19 (0.1) <0.01 43 (0.3) 49 (0.3) <0.01 78 (0.2) 85 (0.2) <0.01

(continued)

(Reprinted) JAMA Internal Medicine March 2023 Volume 183, Number 3

Original Investigation Research

247
 248
Table. Selected Baseline Characteristics of Unmatched and 1:1 Propensity Score–Matched Patients Initiating SGLT2 Inhibitor vs DPP-4 Inhibitor Therapy Overall and Stratified by HbA1c Levels (continued)
Other comorbidities, No. (%)
History of cardiovascular 2806 (23.3) 2739 (22.8) 0.01 3462 (21.4) 3374 (20.9) 0.01 2625 (17.0) 2663 (17.2) −0.01 8893 (20.4) 8776 (20.1) 0.01
disease
Heart failure 874 (7.3) 861 (7.2) <0.01 951 (5.9) 929 (5.8) <0.01 767 (5.0) 731 (4.7) 0.01 2592 (5.9) 2521 (5.8) <0.01
Chronic kidney disease, 1009 (8.4) 974 (8.1) 0.01 1158 (7.2) 1121 (6.9) 0.01 817 (5.3) 777 (5.0) 0.01 2984 (6.8) 2872 (6.6) 0.01
stage <3
Acute kidney injury 286 (2.4) 302 (2.5) −0.01 289 (1.8) 288 (1.8) <0.01 254 (1.6) 250 (1.6) <0.01 829 (1.9) 840 (1.9) <0.01
Research Original Investigation

Mycotic infections 1013 (8.4) 975 (8.1) 0.01 1365 (8.5) 1383 (8.6) <0.01 1418 (9.2) 1421 (9.2) <0.01 3796 (8.7) 3779 (8.7) <0.01
Fractures 105 (0.9) 99 (0.8) 0.01 111 (0.7) 103 (0.6) 0.01 108 (0.7) 102 (0.7) <0.01 324 (0.7) 304 (0.7) <0.01
Falls 272 (2.3) 260 (2.2) 0.01 341 (2.1) 353 (2.2) −0.01 306 (2.0) 296 (1.9) 0.01 919 (2.1) 909 (2.1) <0.01
Diabetes treatment

Downloaded From: https://jamanetwork.com/ on 03/20/2023

No use of any 2266 (18.8) 2276 (18.9) <0.01 1699 (10.5) 1687 (10.4) <0.01 2308 (14.9) 2290 (14.8) <0.01 6273 (14.4) 6253 (14.3) <0.01
glucose-lowering drugs
at baseline; No. (%)
No. glucose-lowering 1.0 (0.8) 1.0 (0.8) <0.01 1.3 (0.9) 1.3 (0.8) 0.01 1.3 (0.9) 1.3 (0.8) −0.01 1.2 (0.9) 1.2 (0.8) <0.01
drugs, mean (SD)f
Metformin, No. (%)f 7400 (61.5) 7573 (63.0) −0.03 10 788 (66.8) 11 062 (68.5) −0.04 10 360 (67.0) 10 584 (68.4) −0.03 28 548 (65.4) 29 219 (67.0) −0.03
Sulfonylureas (second 2280 (19.0) 2311 (19.2) −0.01 5269 (32.6) 5300 (32.8) <0.01 4937 (31.9) 4937 (31.9) <0.01 12 486 (28.6) 12 548 (28.8) <0.01
generation), No. (%)f
GLP-1 receptor agonists, 754 (6.3) 537 (4.5) 0.1 1024 (6.3) 760 (4.7) 0.1 976 (6.3) 772 (5.0) 0.1 2754 (6.3) 2069 (4.7) 0.1

JAMA Internal Medicine March 2023 Volume 183, Number 3 (Reprinted)

No. (%)f
Insulin, No. (%)f 939 (7.8) 838 (7.0) 0.03 2268 (14.0) 2199 (13.6) 0.01 3109 (20.1) 3086 (20.0) <0.01 6316 (14.5) 6123 (14.0) 0.01
Abbreviations: DPP-4i, dipeptidyl peptidase 4 inhibitors; eGFRCr, creatinine-based estimated glomerular (CKD-EPI) equation with no inclusion of race as correction factor.
filtration rate; GLP-1, glucagon-like peptide-1; HbA1c, hemoglobin A1c; NA, not applicable; PS, propensity score; d
A higher combined comorbidity score is associated with a greater risk of mortality during the follow-up.
SGLT2, sodium-glucose cotransporter-2; St. Diff, standardized differences, ie, the difference in means or e
Individuals are considered prefrail when the frailty index is between 0.15 and 0.24 and frail when the frailty index
proportions divided by the pooled.
is at least 0.25.
a
American Indian, Alaska Native, Native Hawaiian, or other Pacific Islander. f
Treatment prescriptions overlapping the date of initiation of the study drugs (ie, concurrent use).
b
To convert to proportion of total hemoglobin, multiply by 0.01.
c
eGFRCr has been estimated applying the creatinine-based Chronic Kidney Disease–Epidemiology Collaboration

jamainternalmedicine.com
Effectiveness and Safety of SGLT2 Inhibitors vs DPP-4 Inhibitors in Patients With Type 2 Diabetes
 Effectiveness and Safety of SGLT2 Inhibitors vs DPP-4 Inhibitors in Patients With Type 2 Diabetes Original Investigation Research

Figure 2. Primary and Secondary Effectiveness Outcomes in 1:1 Propensity Score–Matched Patients
Initiating SGLT2 Inhibitor vs DPP-4 Inhibitor Therapy Stratified by HbA1c Levels

SGLT2 DPP-4 SGLT2 inhibitor vs DPP-4

inhibitor inhibitor inhibitor, relative scale
events, No. events, No. RD/1000 PY Favors Favors P value
(IR/1000 PY) (IR/1000 PY) (95% Cl) HR (95% Cl) SGLT2 DPP-4 for
Primary effectiveness outcomes homogeneity
Modified MACEa
Subgroup HbA1c <7.5% (n = 24 052) 119 (14.31) 139 (17.04) -2.73 (-6.56 to 1.10) 0.84 (0.66 to 1.07) .91
Subgroup HbA1c 7.5%-9% (n = 32 290) 182 (16.3) 205 (18.6) -2.33 (-5.81 to 1.15) 0.88 (0.72 to 1.07)
Subgroup HbA1c >9% (n = 30 932) 199 (20.5) 224 (24.95) -4.45 (-8.79 to -0.12) 0.83 (0.68 to 1.00)
Overall population (n = 87 274) 500 (17.13) 568 (20.18) -3.02 (-5.23 to -0.80) 0.85 (0.75 to 0.95)
Hospitalization for HF
Subgroup HbA1c <7.5% (n = 24 052) 37 (4.44) 75 (9.18) -4.74 (-7.27 to -2.22) 0.48 (0.33 to 0.72)
Subgroup HbA1c 7.5%-9% (n = 32 290) 37 (3.3) 84 (7.61) -4.31 (-6.25 to -2.37) 0.44 (0.30 to 0.64) .95
Subgroup HbA1c >9% (n = 30 932) 34 (3.48) 69 (7.65) -4.17 (-6.32 to -2.02) 0.47 (0.31 to 0.71)
Overall population (n = 87 274) 108 (3.68) 228 (8.08) -4.37 (-5.62 to -3.12) 0.46 (0.35 to 0.57)
Secondary effectiveness outcomes
Myocardial infarction
Subgroup HbA1c <7.5% (n = 24 052) 54 (6.49) 56 (6.86) -0.37 (-2.86 to 2.13) 0.94 (0.65 to 1.37)
Subgroup HbA1c 7.5%-9% (n = 32 290) 70 (6.26) 81 (7.34) -1.08 (-3.25 to 1.09) 0.85 (0.62 to 1.18) .84
Subgroup HbA1c >9% (n = 30 932) 86 (8.84) 83 (9.22) -0.37 (-3.10 to 2.35) 0.97 (0.72 to 1.31)
Overall population (n = 87 274) 210 (7.18) 220 (7.80) -0.67 ( -2.07 to 0.74) 0.92 (0.74 to 1.09)
Stroke
Subgroup HbA1c <7.5% (n = 24 052) 27 (3.24) 29 (3.54) -0.31 (-2.08 to 1.47) 0.91 (0.54 to 1.55)
Subgroup HbA1c 7.5%-9% (n = 32 290) 45 (4.02) 48 (4.35) -0.33 (-2.03 to 1.37) 0.94 (0.62 to 1.41) .84
Subgroup HbA1c >9% (n = 30 932) 54 (5.53) 63 (6.99) -1.45 (-3.72 to 0.82) 0.80 (0.56 to 1.16)
Overall population (n = 87 274) 126 (4.30) 140 (4.96) -0.57 (-1.66 to 0.50) 0.86 (0.65 to 1.08)
All-cause mortality
Subgroup HbA1c <7.5% (n = 24 052) 43 (5.15) 64 (7.8) -2.65 (-5.11 to -0.20) 0.66 (0.45 to 0.97)
Subgroup HbA1c 7.5%-9% (n = 32 290) 77 (6.86) 88 (7.95) -1.09 (-3.35 to 1.17) 0.86 (0.63 to 1.17) .55
Subgroup HbA1c >9% (n = 30 932) 70 (7.16) 90 (9.95) -2. 79 (-5.44 to -0.14) 0.71 (0.52 to 0.98)
Overall population (n = 87 274) 190 (6.47) 242 (8.54) -2.08 (-3.49 to -0.68) 0.74 (0.59 to 0.88)

0.2 1 2.0
HR (95% CI)

Number of events and incidence rates (IR) by treatment group and the point estimates of the effect sizes are shown overall and for HbA1c subcohorts.
Hazard ratios are indicated by squares; 95% CIs, by horizontal lines. DPP-4 indicates dipeptidyl peptidase 4; HbA1c, hemoglobin A1c; HR, hazard ratio;
MACE, major adverse cardiovascular events; PY, person-years; RD, rate difference; SGLT2, sodium-glucose cotransporter 2.
a
Modified MACE is composite outcome including myocardial infarction and stroke events and all-cause deaths. A detailed definition is reported in Supplement 1.

Secondary Effectiveness Outcomes Analyses
No overall differences between SGLT2i and DPP-4i treat-
ments in the risk of myocardial infarction (HR, 0.92; 95% CI,
0.74-1.09) or stroke (HR, 0.86; 95% CI, 0.65-1.08) were found.
In the overall population, the initiation of treatment with
SGLT2i vs DPP-4i was associated with a 26% reduced risk of
all-cause mortality (HR, 0.74; 95% CI, 0.59-0.88), correspond-
ing to 2 fewer deaths per 1000 person-years. No evidence of
effect heterogeneity on either the relative or the absolute
scale was found between subgroups for any of the secondary
outcomes (Figure 2).
Safety Outcomes Analyses
The risks of hypovolemia, nonvertebral fractures, falls and
lower-limb amputations were similar among patients initiat-
ing treatment with SGLT2i vs DPP-4i (Figure 4). In the overall
population, SGLT2i vs DPP-4i initiators had a 2.17-fold in-
creased risk of genital infections (HR, 2.17; 95% CI, 1.98-
2.36), corresponding to approximately 38 additional events per
1000 person-years, with evidence of treatment effect hetero-
geneity across subgroups on both the relative and absolute
scales (P < .01 for homogeneity). Patients with HbA1c levels of
7.5% to 9% had a 3.1-fold increased risk for yeast infections (IR
per 1000 person-years 68.5 vs 22.8, respectively; HR, 3.10;
95% CI, 2.68-3.58) vs an approximately 2-fold increased risk
in patients with HbA1c levels of greater than 7.5% (HR, 2.41;
95% CI, 2.04-2.85) and greater than 9% (HR, 1.82; 95% CI, 1.60-
2.08), corresponding to RD of 46.22 (95% CI, 40.54-51.90) for
HbA1c 7.5%-9%, vs 33.96 (95% CI, 27.69-40.23) for HbA1c <7.5%,
and 29.66 (95% CI, 23.00-36.32) for HbA1c >9% additional cases
per 1000 person-years. Overall, a 1.7-fold increased risk of DKA
was found in SGLT2i treatment initiators (HR, 1.73; 95% CI, 1.06-
2.43). Although the estimates are less precise and the uncer-
tainty is higher due to the low number of DKA events within
each HbA1c subgroup, the stratified results appear consistent
with the overall finding (Figure 4). In the overall cohort, a 27%
decreased risk of AKI was associated with the initiation of
SGLT2i vs DPP-4i (HR, 0.73; 95% CI, 0.66-0.81), correspond-
ing to approximately 8 fewer cases per 1000 patient-years.
Similar results were obtained in subgroup analyses by HbA1c

jamainternalmedicine.com (Reprinted) JAMA Internal Medicine March 2023 Volume 183, Number 3 249

Downloaded From: https://jamanetwork.com/ on 03/20/2023

 Research Original Investigation Effectiveness and Safety of SGLT2 Inhibitors vs DPP-4 Inhibitors in Patients With Type 2 Diabetes

Figure 3. Cumulative Incidence of Modified MACE and HHF Comparing 1:1 Propensity Score–Matched Patients
Initiating SGLT2 Inhibitor vs DPP-4 Inhibitor Therapy

A Modified MACE

5
Log-rank P <.01

Cumulative incidence, %
4
DPP−4 inhibitors
3

2
SGLT2 inhibitors
1

0
0 3 6 9 12 15 18 21 24 27 30
Time, mo
No. at risk
DPP−4 inhibitors 43 637 29 568 17 256 11 306 7917 5800 4297 3311 2617 2042 1616
SGLT2 inhibitors 43 637 28 946 17 422 11 816 8555 6292 4749 3719 2928 2294 1775

B HHF
5
Log-rank P <.01
Cumulative incidence, %

2
DPP−4 inhibitors

SGLT2 inhibitors
0
0 3 6 9 12 15 18 21 24 27 30 DPP-4 indicates dipeptidyl
Time, mo peptidase 4; HHF, hospitalization
No. at risk for heart failure; MACE, major
DPP−4 inhibitors 43 637 29 582 17 280 11 326 7941 5820 4314 3328 2628 2055 1624 adverse cardiovascular events;
SGLT2 inhibitors 43 637 28 973 17 457 11 847 8583 6325 4777 3740 2947 2303 1786 SGLT2, sodium-glucose
cotransporter 2.

with no evidence of treatment effect heterogeneity by HbA1c istics differed among subgroups (for example patients with el-
(Figure 4). evated HbA1c were younger, more likely to receive insulin, had
a higher eGFRCr and fewer comorbidities than others), ben-
Sensitivity Analyses efits and adverse effects of SGLT2i vs DPP-4i were similar across
Findings remained consistent when an intention-to-treat HbA1c subcohorts.
approach was adopted and when the internal validity of Overall, patients receiving a SGLT2i had a 15% lower risk
the effectiveness and safety outcome analyses was tested of the composite of myocardial infarction, stroke, or death from
(eTables 8-9 in Supplement 1) with some fluctuations in point all causes (approximately 3 fewer cases per 1000 person-
estimates driven by the small number of events in the sub- years) and a 54% lower risk of HHF (approximately 4 fewer
group analyses by HbA1c. No difference in treatment effect was cases per 1000 person-years) than those receiving a DPP-4i.
found in patients with vs without cardiovascular diseases. These findings with respect to the modified MACE outcome
parallel those of the placebo-controlled CANVAS trial for
canagliflozin, 4 and the placebo-controlled EMPA-REG
OUTCOME trial for empagliflozin,3 and of large cohort stud-
Discussion ies comparing SGLT2i vs DPP-4i.7,8,35-37 Similarly, our HHF re-
In this large comparative effectiveness and safety research sults are in line with those from CVOTs,3-6 and large compara-
study of 87 274 adults with T2D, including 24 052 with con- tive effectiveness studies. 37,38 The effect estimates were
trolled HbA1c levels, 32 290 with above-target HbA1c levels, and consistent across HbA1c subgroups for both modified MACE
30 932 with elevated baseline HbA1c levels, we found that (1) (HR range, 0.83-0.88), in line with exploratory analyses from
initiating treatment with SGLT2i was associated with a re- a network meta-analysis39 and HHF (HR range, 0.46-0.48).
duced risk of modified MACE, HHF, and AKI and a higher risk The safety analyses showed that overall patients initiat-
of genital infections and DKA compared with DPP-4i; and that ing a SGLT2i had a higher risk of genital infections and
(2) the results did not vary based on preexposure HbA1c levels DKA, both of which are known adverse effects of these
for most outcomes evaluated. Although individual character- medications,21,22,40 and a lower risk of AKI, a previously

250 JAMA Internal Medicine March 2023 Volume 183, Number 3 (Reprinted) jamainternalmedicine.com

Downloaded From: https://jamanetwork.com/ on 03/20/2023

 Effectiveness and Safety of SGLT2 Inhibitors vs DPP-4 Inhibitors in Patients With Type 2 Diabetes Original Investigation Research

Figure 4. Safety Outcomes in 1:1 Propensity Score–Matched Patients Initiating SGLT2 Inhibitor vs DPP-4 Inhibitor Therapy Stratified by HbA1c Levels

SGLT2 DPP-4 SGLT2 inhibitor vs DPP-4

inhibitor inhibitor inhibitor, relative scale
events, No. events, No. RD/1000 PY Favors Favors P value
(IR/1000 PY) (IR/1000 PY) (95% Cl) HR (95% Cl) SGLT2 DPP-4 for
Hypovolemia homogeneity
Subgroup HbA1c <7.5% (n = 24 052) 75 (9.03) 67 (8.21) 0.82 (-2.02 to 3.65) 1.10 (0.79 to 1.53)
Subgroup HbA1c 7.5%-9% (n = 32 290) 101 (9.04) 89 (8.07) 0.97 (-1.46 to 3.40) 1.13 (0.85 to 1.50) .11
Subgroup HbA1c >9% (n = 30 932) 81 (8.33) 101 (11.24) -2.92 (-5.76 to 0.07) 0.76 (0.57 to 1.02)
Overall population (n = 87 274) 257 (8.80) 258 (9.12) -0.23 (-1.77 to 1.32) 0.93 (0.76 to 1.09)
Nonvertebral fractures
Subgroup HbA1c <7.5% (n = 24 052) 38 (4.57) 38 (4.64) -0.08 (-2.15 to 1.99) 0.98 (0.63 to 1.54) .39
Subgroup HbA1c 7.5%-9% (n = 32 290) 35 (3.12) 42 (3.8) -0.68 (-2.23 to 0.87) 0.82 (0.52 to 1.28)
Subgroup HbA1c >9% (n = 30 932) 43 (4.41) 30 (3.33) 1.08 (-0.69 to 2.86) 1.35 (0.84 to 2.15)
Overall population (n = 87 274) 116 (3.96) 110 (3.89) 0.04 (-0.97 to 1.06) 0.96 (0.69 to 1.23)
Falls
Subgroup HbA1c <7.5% (n = 24 052) 49 (5.88) 44 (5.39) 0.49 (-1.80 to 2.79) 1.10 (0.73 to 1.65) .69
Subgroup HbA1c 7.5%-9% (n = 32 290) 49 (4.37) 56 (5.07) -0.70 (-2.50 to 1.11) 0.86 (0.59 to 1.26)
Subgroup HbA1c >9% (n = 30 932) 44 (4.51) 47 (5.21) -0.70 (-2.70 to 1.30) 0.88 (0.58 to 1.32)
Overall population (n = 87 274) 142 (4.85) 147 (5.21) -0.40 (-1.56 to 0.76) 0.92 (0.70 to 1.14)
Genital infections
Subgroup HbA1c <7.5% (n = 24 052) 469 (58.22) 196 (24.27) 33.96 (27.69 to 40.23) 2.41 (2.04 to 2.85) <.01
Subgroup HbA1c 7.5%-9% (n = 32 290) 738 (68.5) 243 (22.28) 46.22 (40.54 to 51.90) 3.10 (2.68 to 3.58)
Subgroup HbA1c >9% (n = 30 932) 640 (68.04) 341 (38.38) 29.66 (23.00 to 36.32) 1.82 (1.60 to 2.08)
Overall population (n = 87 274) 1847 (65.41) 780 (27.99) 37.53 (33.98 to 41.09) 2.17 (1.98 to 2.36)
Diabetic ketoacidosis
Subgroup HbA1c <7.5% (n = 24 052) 6 (0.72) 4 (0.49) 0.23 (-0.52 to 0.98) 1.49 (0.42 to 5.26) .62
Subgroup HbA1c 7.5%-9% (n = 32 290) 13 (1.16) 10 (0.9) 0.25 (-0.59 to 1.10) 1.27 (0.65 to 2.89)
Subgroup HbA1c >9% (n = 30 932) 47 (4.81) 22 (2.43) 2.38 (0.67 to 4.09) 2.06 (1.62 to 3.42)
Overall population (n = 87 274) 66 (2.25) 36 (1.27) 0.45 (-0.09 to 0.98) 1.73 (1.06 to 2.43)
Acute kidney injury
Subgroup HbA1c <7.5% (n = 24 052) 162 (19.55) 220 (27.22) -7.68 (-12.37 to -2.99) 0.72 (0.59 to 0.89)
Subgroup HbA1c 7.5%-9% (n = 32 290) 211 (18.92) 295 (27.03) -8.11 (-12.12 to -4.11) 0.70 (0.59 to 0.84) .63
Subgroup HbA1c >9% (n = 30 932) 225 (23.25) 268 (30.05) -6.80 (-11.51 to -2 .09) 0.79 (0.66 to 0.94)
Overall population (n = 87 274) 598 (20.54) 783 (28.05) -7.60 (-10.15 to -5.04) 0.73 (0.66 to 0.81)
Lower-limb amputation
Subgroup HbA1c <7.5% (n = 24 052) 9 (1.08) 13 (1.59) -0.51 (-1.62 to 0.60) 0.69 (0.29 to 1.61)
Subgroup HbA1c 7.5%-9% (n = 32 290) 32 (2.85) 22 (1.99) 0.86 (-0.43 to 2.15) 1.44 (0.84 to 2.48) .37
Subgroup HbA1c >9% (n = 30 932) 37 (3.78) 33 (3.62) 0.16 (-1.57 to 1.90) 1.05 (0.66 to 1.68)
Overall population (n = 87 274) 78 (2.56) 68 (2.46) 0.09 (-0.67 to 0.85) 1.02 (0.66 to 1.38)

0.2 1 6.0
HR (95% CI)

Number of events and incidence rates (IR) by treatment group and point estimates of the effect sizes are shown overall and for HbA1c subcohorts.
Hazard ratios are indicated by squares; 95% CIs, by horizontal lines. DPP-4 indicates dipeptidyl peptidase 4; HbA1c, hemoglobin A1c; HR, hazard ratio;
PY, person-years; RD, rate difference; SGLT2, sodium-glucose cotransporter 2.

observed benefit,22,41,42 than those receiving a DPP-4i. Rates
of hypovolemia, falls, bone fracture events, and lower-limb
amputations were similar in the SGLT2i and DPP-4i groups.
The analysis of the safety profile of SGLT2i across patients
with different HbA1c levels, which has not been investigated
in CVOTs, is a main strength of our study. Medications in the
SGLT2i class are responsible for pharmacologically induced
renal glycosuria by suppressing sodium and glucose reab-
sorption in the proximal tubule. Given that the urinary glu-
cose concentration and consequent osmotic diuresis is
higher in SGLT2i users with uncontrolled glycemia than in
those with better controlled glycemia, a common hypothesis
is that the risk of hypovolemia (and consequent falls and
fractures) from polyuria, genitourinary infections from glu-
cosuria, amputations from a reduced limb perfusion due to
hypotension and an increased risk of peripheral ischemia
due to hemoconcentration and hyperviscosity, and DKA
from decreased plasma glucose and insulin release might
have been further increased in patients with elevated HbA1c
compared with patients with lower HbA1c.43-45 The findings
of this study do not support this hypothesis. Results were
largely consistent across all HbA1c subcohorts, except for
some evidence of treatment effect heterogeneity for genital
infections, with the highest risk observed in patients with
baseline HbA1c levels between 7.5% and 9%. This subgroup
had the highest prevalence of both diabetes-related compli-
cations and prescriptions of glucose-lowering medications,
suggesting a more advanced stage of diabetes compared
with other subgroups. This may explain the increased risk of
yeast infections observed in these patients.

jamainternalmedicine.com (Reprinted) JAMA Internal Medicine March 2023 Volume 183, Number 3 251

Downloaded From: https://jamanetwork.com/ on 03/20/2023

 Research Original Investigation Effectiveness and Safety of SGLT2 Inhibitors vs DPP-4 Inhibitors in Patients With Type 2 Diabetes

This study augments the evidence provided by CVOTs of
SGLT2i, showing that patients with severe uncontrolled dia-
betes can benefit from the use of these medications in a fash-
ion similar to patients with better controlled glycemia, with
no further increase in the risk of adverse effects. The popula-
tion of patients identified in this study is 5 to 9 times larger
than the populations included in the CVOTs. Because of the
larger sample, we were able to identify 3 subgroups of pa-
tients with different ranges of HbA1c, and thus explore with
more granularity and reduced level of uncertainty the influ-
ence of increasing glycemic levels on the safety and effective-
ness of SGLT2i treatment. Another strength of this study is bet-
ter generalizability of the findings to routine care. Several
studies reported that a considerable number of patients with
T2D cared for in clinical practice do not have characteristics
similar to the patient populations included in CVOTs.46-48
A recent review showed that if the enrollment criteria of
CANVAS,4 EMPA-REG OUTCOME,3 and VERTIS-CV6 were ap-
plied to the real-world population, only 17% to 36% of pa-
tients with T2D would have been eligible, with only 49.5% of
real-world patients with T2D eligible for a CVOT with broader
inclusion criteria such as the DECLARE-TIMI-58 trial.48 Fur-
ther, while CVOTs restrict to patients with cardiorenal dis-
eases or multiple risk factors to achieve adequate statistical
power in the time frame of the trials, we examined the
comparative effectiveness of these drugs across the broader
spectrum of cardiovascular risk. Lastly, adopting an active-
comparator new-user design largely reduced the risk of
biased findings, increasing the study validity.49,50
Limitations
This study has limitations. First, residual confounding for un-
measured characteristics, such as duration of diabetes or body
mass index, cannot be entirely ruled out. However, we ob-
served that covariates not included in the PS model (labora-
tory test results available only for a subset of the analytic co-
hort) were balanced after adjustment. Additionally, compared
with other large cohort studies that compared SGLT2 vs
DPP-4i,7,8,22-24,38,40,41 we addressed potential confounding by
diabetes severity and kidney function by controlling for
eGFRCr. Second, the stratification by HbA1c levels reduced pre-
cision of some outcome estimates within subgroups, such as
for modified MACE and DKA. However, the direction and mag-
nitude of the effect for these outcomes were consistent with
the overall findings, supporting the lack of effect modifica-
tion by HbA1c. Third, as this study is based on routine care use
of SGLT2i or DPP-4i, the mean follow-up (ie, time on treat-
ment) was shorter compared with CVOTs, which introduce
substantial measures to improve treatment adherence. Con-
trary to randomized clinical trials that require long follow-up
to accumulate sufficient events for powered analyses, the size
of this study population allowed us to generate overall re-
sults with high precision despite a shorter length of follow-
up. Additionally, several trials showed that SGLT2i rapidly re-
duced the risk of cardiovascular death or HHF in patients with
T2D, with benefits that are sustained over time.51-54 Thus, as-
suming no time-varying hazards, these results should be gen-
eralizable to longer-term findings. Fourth, potential for out-
come misclassification cannot be entirely excluded; however,
the validated outcome definitions used in this study have high
positive predictive value and specificity and are not expected
to differ by treatment group. Last, we could not evaluate car-
diovascular death due to the lack of information on cause of
death in the data. Death for all causes may be limited by in-
complete death records in this data set, though we would not
expect this to differ by treatment groups.
Conclusions
In this large comparative effectiveness and safety study of
87 274 adults with T2D, patients who initiated SGLT2i therapy
had a reduced risk of MACE, HHF, and AKI, and an increased
risk of genital infections and DKA, compared with DPP-4i.
The cardiovascular effectiveness and safety of SGLT2i vs DPP-4i
did not vary based on baseline HbA1c levels. This study comple-
ments the evidence provided by CVOTs by showing that pa-
tients with T2D can benefit from the use of SGLT2i regardless
of glycemic control, with no additional increase in the risk of
adverse effects in patients with above-target or elevated HbA1c
levels, compared with DPP-4i initiators with similar glycemic
control.

ARTICLE INFORMATION
Accepted for Publication: December 7, 2022.
Published Online: February 6, 2023.
doi:10.1001/jamainternmed.2022.6664
Open Access: This is an open access article
distributed under the terms of the CC-BY License.
© 2023 D’Andrea E et al. JAMA Internal Medicine.
Author Contributions: Dr D’Andrea and Patorno
had full access to all of the data in the study and
take responsibility for the integrity of the data and
the accuracy of the data analysis.
Concept and design: D'Andrea, Wexler, Kim,
Patorno.
Acquisition, analysis, or interpretation of data:
All authors.
Drafting of the manuscript: D'Andrea.
Critical revision of the manuscript for important
intellectual content: All authors.
Statistical analysis: D'Andrea, Alt.
Obtained funding: Patorno.
Administrative, technical, or material support:
D'Andrea, Patorno.
Supervision: Wexler, Kim, Patorno.
Conflict of Interest Disclosures: Dr Wexler
reported participation in a data monitoring
committee for studies of oral and subcutnaeous
semaglutide from Novo Nordisk during the conduct
of the study. Dr Kim reported grants from Roche,
AbbVie, Pfizer, and Bristol Myers Squibb outside
the submitted work. Dr Patorno is an investigator of
a research grant to the Brigham and Women’s
Hospital from Boehringer Ingelheim outside the
submitted work. No other disclosures were
reported.
Funding/Support: This study was funded by a
grant (DB-2020C2-20326) from the Patient
Centered Outcomes Research Institute. Dr Patorno
was supported by a career development grant
K08AG055670 from the National Institute on
Aging and by a research grant from the Food and
Drug Administration (5U01FD007213).
Role of the Funder/Sponsor: The funders had no
role in the design and conduct of the study;
collection, management, analysis, and
interpretation of the data; preparation, review, or
approval of the manuscript; and decision to submit
the manuscript for publication.
Data Sharing Statement: See Supplement 2.
REFERENCES
1. Centers for Disease Control and Prevention.
National Diabetes Statistics Report. 2022. Accessed
April 28, 2022. https://www.cdc.gov/diabetes/
data/statistics-report/index.html

252 JAMA Internal Medicine March 2023 Volume 183, Number 3 (Reprinted) jamainternalmedicine.com

Downloaded From: https://jamanetwork.com/ on 03/20/2023

 Effectiveness and Safety of SGLT2 Inhibitors vs DPP-4 Inhibitors in Patients With Type 2 Diabetes
2. Rawshani A, Rawshani A, Franzén S, et al.
Mortality and cardiovascular disease in type 1 and
type 2 diabetes. N Engl J Med. 2017;376(15):
1407-1418. doi:10.1056/NEJMoa1608664
3. Zinman B, Wanner C, Lachin JM, et al;
EMPA-REG OUTCOME Investigators. Empagliflozin,
cardiovascular outcomes, and mortality in type 2
diabetes. N Engl J Med. 2015;373(22):2117-2128.
doi:10.1056/NEJMoa1504720
4. Neal B, Perkovic V, Mahaffey KW, et al; CANVAS
Program Collaborative Group. Canagliflozin and
cardiovascular and renal events in type 2 diabetes.
N Engl J Med. 2017;377(7):644-657. doi:10.1056/
NEJMoa1611925
5. Wiviott SD, Raz I, Bonaca MP, et al;
DECLARE–TIMI 58 Investigators. Dapagliflozin and
cardiovascular outcomes in type 2 diabetes. N Engl
J Med. 2019;380(4):347-357. doi:10.1056/
NEJMoa1812389
6. Cannon CP, Pratley R, Dagogo-Jack S, et al;
VERTIS CV Investigators. Cardiovascular outcomes
with ertugliflozin in type 2 diabetes. N Engl J Med.
2020;383(15):1425-1435. doi:10.1056/
NEJMoa2004967
7. Patorno E, Pawar A, Wexler DJ, et al.
Effectiveness and safety of empagliflozin in routine
care patients: results from the EMPagliflozin
compaRative effectIveness and SafEty (EMPRISE)
study. Diabetes Obes Metab. 2022;24(3):442-454.
doi:10.1111/dom.14593
8. Patorno E, Goldfine AB, Schneeweiss S, et al.
Cardiovascular outcomes associated with
canagliflozin versus other non-gliflozin antidiabetic
drugs: population based cohort study. BMJ. 2018;
360:k119. doi:10.1136/bmj.k119
9. D’Andrea E, Kesselheim AS, Franklin JM,
Jung EH, Hey SP, Patorno E. Heterogeneity of
antidiabetic treatment effect on the risk of major
adverse cardiovascular events in type 2 diabetes:
a systematic review and meta-analysis. Cardiovasc
Diabetol. 2020;19(1):154. doi:10.1186/s12933-
020-01133-1
10. Cherney DZ, Kanbay M, Lovshin JA. Renal
physiology of glucose handling and therapeutic
implications. Nephrol Dial Transplant. 2020;35
(suppl 1):i3-i12. doi:10.1093/ndt/gfz230
11. Hsia DS, Grove O, Cefalu WT. An update on
sodium-glucose co-transporter-2 inhibitors for the
treatment of diabetes mellitus. Curr Opin Endocrinol
Diabetes Obes. 2017;24(1):73-79.
12. Buse JB, Wexler DJ, Tsapas A, et al. 2019 update
to: management of hyperglycemia in type 2
diabetes, 2018: a consensus report by the American
Diabetes Association (ADA) and the European
Association for the Study of Diabetes (EASD).
Diabetes Care. 2020;43(2):487-493. doi:10.2337/
dci19-0066
13. Garber AJ, Handelsman Y, Grunberger G, et al.
Consensus statement by the American Association
Of Clinical Endocrinologists and American College
Of Endocrinology on the comprehensive type 2
diabetes management algorithm—2020 executive
summary. Endocr Pract. 2020;26(1):107-139. doi:10.
4158/CS-2019-0472
14. Kiyota Y, Schneeweiss S, Glynn RJ,
Cannuscio CC, Avorn J, Solomon DH. Accuracy of
Medicare claims-based diagnosis of acute
myocardial infarction: estimating positive
predictive value on the basis of review of hospital
jamainternalmedicine.com
Downloaded From: https://jamanetwork.com/ on 03/20/2023
records. Am Heart J. 2004;148(1):99-104. doi:10.
1016/j.ahj.2004.02.013
15. Wahl PM, Rodgers K, Schneeweiss S, et al.
Validation of claims-based diagnostic and
procedure codes for cardiovascular and
gastrointestinal serious adverse events in a
commercially-insured population.
Pharmacoepidemiol Drug Saf. 2010;19(6):596-603.
doi:10.1002/pds.1924
16. Tirschwell DL, Longstreth WT Jr. Validating
administrative data in stroke research. Stroke.
2002;33(10):2465-2470. doi:10.1161/01.STR.
0000032240.28636.BD
17. Saczynski JS, Andrade SE, Harrold LR, et al.
A systematic review of validated methods for
identifying heart failure using administrative data.
Pharmacoepidemiol Drug Saf. 2012;21(0 1)(suppl 1):
129-140. doi:10.1002/pds.2313
18. Bobo WV, Cooper WO, Epstein RA Jr,
Arbogast PG, Mounsey J, Ray WA. Positive
predictive value of automated database records for
diabetic ketoacidosis (DKA) in children and youth
exposed to antipsychotic drugs or control
medications: a Tennessee Medicaid Study. BMC
Med Res Methodol. 2011;11:157. doi:10.1186/1471-
2288-11-157
19. Newton KM, Wagner EH, Ramsey SD, et al.
The use of automated data to identify
complications and comorbidities of diabetes:
a validation study. J Clin Epidemiol. 1999;52(3):
199-207. doi:10.1016/S0895-4356(98)00161-9
20. Wright NC, Daigle SG, Melton ME, Delzell ES,
Balasubramanian A, Curtis JR. The design and
validation of a new algorithm to identify incident
fractures in administrative claims data. J Bone Miner
Res. 2019;34(10):1798-1807. doi:10.1002/jbmr.3807
21. Waikar SS, Wald R, Chertow GM, et al. Validity
of International Classification of Diseases, Ninth
Revision, Clinical Modification codes for acute renal
failure. J Am Soc Nephrol. 2006;17(6):1688-1694.
doi:10.1681/ASN.2006010073
22. Dave CV, Schneeweiss S, Patorno E.
Comparative risk of genital infections associated
with sodium-glucose co-transporter-2 inhibitors.
Diabetes Obes Metab. 2019;21(2):434-438. doi:10.
1111/dom.13531
23. Patorno E, Pawar A, Bessette LG, et al.
Comparative effectiveness and safety of
sodium-glucose cotransporter 2 inhibitors versus
glucagon-like peptide 1 receptor agonists in older
adults. Diabetes Care. 2021;44(3):826-835. doi:10.
2337/dc20-1464
24. Patorno E, Htoo PT, Glynn RJ, et al.
Sodium-glucose cotransporter-2 inhibitors versus
glucagon-like peptide-1 receptor agonists and the
risk for cardiovascular outcomes in routine care
patients with diabetes across categories of
cardiovascular disease. Ann Intern Med. 2021;174
(11):1528-1541. doi:10.7326/M21-0893
25. Gagne JJ, Glynn RJ, Avorn J, Levin R,
Schneeweiss S. A combined comorbidity score
predicted mortality in elderly patients better than
existing scores. J Clin Epidemiol. 2011;64(7):749-759.
doi:10.1016/j.jclinepi.2010.10.004
26. Kim DH, Schneeweiss S, Glynn RJ, Lipsitz LA,
Rockwood K, Avorn J. Measuring frailty in Medicare
data: development and validation of a claims-based
frailty index. J Gerontol A Biol Sci Med Sci. 2018;73
(7):980-987. doi:10.1093/gerona/glx229
(Reprinted) JAMA Internal Medicine March 2023 Volume 183, Number 3
Original Investigation Research
27. Inker LA, Eneanya ND, Coresh J, et al; Chronic
Kidney Disease Epidemiology Collaboration. New
creatinine- and cystatin c-based equations to
estimate GFR without race. N Engl J Med. 2021;385
(19):1737-1749. doi:10.1056/NEJMoa2102953
28. Paik JM, Patorno E, Zhuo M, et al. Accuracy of
identifying diagnosis of moderate to severe chronic
kidney disease in administrative claims data.
Pharmacoepidemiol Drug Saf. 2022;31(4):467-475.
doi:10.1002/pds.5398
29. Austin PC. Balance diagnostics for comparing
the distribution of baseline covariates between
treatment groups in propensity-score matched
samples. Stat Med. 2009;28(25):3083-3107.
doi:10.1002/sim.3697
30. Wang SV, Jin Y, Fireman B, et al. Relative
performance of propensity score matching
strategies for subgroup analyses. Am J Epidemiol.
2018;187(8):1799-1807. doi:10.1093/aje/kwy049
31. Suissa S. Lower risk of death with SGLT2
inhibitors in observational studies: real or bias?
Diabetes Care. 2018;41(1):6-10. doi:10.2337/dc17-1223
32. Scirica BM, Bhatt DL, Braunwald E, et al;
SAVOR-TIMI 53 Steering Committee and
Investigators. Saxagliptin and cardiovascular
outcomes in patients with type 2 diabetes mellitus.
N Engl J Med. 2013;369(14):1317-1326. doi:10.
1056/NEJMoa1307684
33. Zannad F, Cannon CP, Cushman WC, et al;
EXAMINE Investigators. Heart failure and mortality
outcomes in patients with type 2 diabetes taking
alogliptin versus placebo in EXAMINE:
a multicentre, randomised, double-blind trial. Lancet.
2015;385(9982):2067-2076. doi:10.1016/S0140-
6736(14)62225-X
34. US Food and Drug Administration. FDA drug
safety communication: FDA adds warnings about
heart failure risk to labels of type 2 diabetes
medicines containing saxagliptin and alogliptin.
April 5, 2016. Accessed June 26, 2022. https://
www.fda.gov/media/96895/download
35. Persson F, Nyström T, Jørgensen ME, et al.
Dapagliflozin is associated with lower risk of
cardiovascular events and all-cause mortality in
people with type 2 diabetes (CVD-REAL Nordic)
when compared with dipeptidyl peptidase-4
inhibitor therapy: a multinational observational
study. Diabetes Obes Metab. 2018;20(2):344-351.
doi:10.1111/dom.13077
36. Filion KB, Lix LM, Yu OH, et al; Canadian
Network for Observational Drug Effect Studies
(CNODES) Investigators. Sodium glucose
cotransporter 2 inhibitors and risk of major adverse
cardiovascular events: multi-database retrospective
cohort study. BMJ. 2020;370:m3342. doi:10.1136/
bmj.m3342
37. Kosiborod M, Cavender MA, Fu AZ, et al;
CVD-REAL Investigators and Study Group*. Lower
risk of heart failure and death in patients initiated
on sodium-glucose cotransporter-2 inhibitors
versus other glucose-lowering drugs: the
CVD-REAL study (comparative effectiveness of
cardiovascular outcomes in new users of
sodium-glucose cotransporter-2 inhibitors).
Circulation. 2017;136(3):249-259. doi:10.1161/
CIRCULATIONAHA.117.029190
38. Patorno E, Pawar A, Franklin JM, et al.
Empagliflozin and the risk of heart failure
hospitalization in routine clinical care. Circulation.
253
 Research Original Investigation
2019;139(25):2822-2830. doi:10.1161/
CIRCULATIONAHA.118.039177
39. Lin DS, Lee JK, Hung CS, Chen WJ. The efficacy
and safety of novel classes of glucose-lowering
drugs for cardiovascular outcomes: a network
meta-analysis of randomised clinical trials.
Diabetologia. 2021;64(12):2676-2686. doi:10.1007/
s00125-021-05529-w
40. Fralick M, Schneeweiss S, Patorno E. Risk of
diabetic ketoacidosis after initiation of an SGLT2
inhibitor. N Engl J Med. 2017;376(23):2300-2302.
doi:10.1056/NEJMc1701990
41. Zhuo M, Paik JM, Wexler DJ, Bonventre JV,
Kim SC, Patorno E. SGLT2 inhibitors and the risk of
acute kidney injury in older adults with type 2
diabetes. Am J Kidney Dis. 2022;79(6):858-867.e1.
doi:10.1053/j.ajkd.2021.09.015
42. Wanner Ch, Inzucchi SE, Zinman B.
Empagliflozin and progression of kidney disease in
type 2 diabetes. N Engl J Med. 2016;375(18):
1801-1802. doi:10.1056/NEJMoa1515920
43. Brady JA, Hallow KM. Model-based evaluation
of proximal sodium reabsorption through SGLT2 in
health and diabetes and the effect of inhibition with
canagliflozin. J Clin Pharmacol. 2018;58(3):377-385.
doi:10.1002/jcph.1030
44. Nakagaito M, Imamura T, Joho S, Ushijima R,
Nakamura M, Kinugawa K. Relationship between
HbA1c level and effectiveness of SGLT2 inhibitors in
decompensated heart failure patients with type 2
254 JAMA Internal Medicine March 2023 Volume 183, Number 3 (Reprinted)
Downloaded From: https://jamanetwork.com/ on 03/20/2023
Effectiveness and Safety of SGLT2 Inhibitors vs DPP-4 Inhibitors in Patients With Type 2 Diabetes
diabetes mellitus. Int Heart J. 2021;62(4):843-849.
doi:10.1536/ihj.20-764
45. Lytvyn Y, Bjornstad P, Udell JA, Lovshin JA,
Cherney DZI. Sodium glucose cotransporter-2
inhibition in heart failure: potential mechanisms,
clinical applications, and summary of clinical trials.
Circulation. 2017;136(17):1643-1658. doi:10.1161/
CIRCULATIONAHA.117.030012
46. Birkeland KI, Bodegard J, Norhammar A, et al.
How representative of a general type 2 diabetes
population are patients included in cardiovascular
outcome trials with SGLT2 inhibitors: a large
European observational study. Diabetes Obes Metab.
2019;21(4):968-974. doi:10.1111/dom.13612
47. Wittbrodt E, Chamberlain D, Arnold SV, Tang F,
Kosiborod M. Eligibility of patients with type 2
diabetes for sodium-glucose co-transporter-2
inhibitor cardiovascular outcomes trials: an
assessment using the Diabetes Collaborative
Registry. Diabetes Obes Metab. 2019;21(8):
1985-1989. doi:10.1111/dom.13738
48. Castellana M, Procino F, Sardone R, Trimboli P,
Giannelli G. Generalizability of sodium-glucose
co-transporter-2 inhibitors cardiovascular outcome
trials to the type 2 diabetes population:
a systematic review and meta-analysis. Cardiovasc
Diabetol. 2020;19(1):87. doi:10.1186/s12933-020-
01067-8
49. Bykov K, Patorno E, D’Andrea E, et al.
Prevalence of avoidable and bias-inflicting
methodological pitfalls in real-world studies of
medication safety and effectiveness. Clin
Pharmacol Ther. 2022;111(1):209-217. doi:10.1002/
cpt.2364
50. D’Andrea E, Vinals L, Patorno E, et al. How well
can we assess the validity of non-randomised
studies of medications: a systematic review of
assessment tools. BMJ Open. 2021;11(3):e043961.
doi:10.1136/bmjopen-2020-043961
51. Verma S, Leiter LA, Zinman B, et al. Time to
cardiovascular benefits of empagliflozin: a post hoc
observation from the EMPA-REG OUTCOME trial.
ESC Heart Fail. 2021;8(4):2603-2607. doi:10.1002/
ehf2.13374
52. Packer M, Anker SD, Butler J, et al. Effect of
empagliflozin on the clinical stability of patients
with heart failure and a reduced ejection fraction:
the EMPEROR-Reduced Trial. Circulation. 2021;143
(4):326-336. Published online October 21, 2020.
doi:10.1161/CIRCULATIONAHA.120.051783
53. Berg DD, Jhund PS, Docherty KF, et al. Time to
clinical benefit of dapagliflozin and significance of
prior heart failure hospitalization in patients with
heart failure with reduced ejection fraction. JAMA
Cardiol. 2021;6(5):499-507. doi:10.1001/
jamacardio.2020.7585
54. Vaduganathan M, Claggett BL, Jhund P, et al.
Time to clinical benefit of dapagliflozin in patients
with heart failure with mildly reduced or preserved
ejection fraction: a prespecified secondary analysis
of the DELIVER randomized clinical trial. JAMA
Cardiol. 2022;7(12):1259-1263. Published online
October 03, 2022. doi:10.1001/jamacardio.2022.
3750
jamainternalmedicine.com