GLP-1 Drugs Beat SGLT2 Inhibitors for Weight Loss — But Only in Women and Non-Diabetics
A real-world study of nearly 4,000 patients reveals GLP-1 receptor agonists outperform SGLT2 inhibitors for weight loss, with striking sex and diabetes-status differences.
Summary
Researchers used the NIH All of Us database to compare 12-month weight loss in 2,932 adults starting GLP-1 drugs (such as semaglutide) and 941 starting SGLT2 inhibitors (such as empagliflozin). After statistical balancing for age, BMI, heart failure, kidney disease and other factors, GLP-1 users lost about 4.1% of body weight versus 2.9% for SGLT2 users, a 1.2 percentage-point gap. More GLP-1 users lost at least 10% of body weight (about 19% vs 12%). The advantage showed up in people without type 2 diabetes and in women, but not in people with diabetes or in men. The authors stress these subgroup findings were exploratory and need confirmation in prospective trials. Blood pressure did not differ, and HbA1c fell slightly more with GLP-1 drugs.
Detailed Summary
Excess weight drives diabetes, cardiovascular disease and shortened healthspan, and two drug classes are widely used in adults with overweight or obesity: GLP-1 receptor agonists (GLP-1RAs) such as semaglutide, and SGLT2 inhibitors (SGLT2is) such as empagliflozin. Weight loss is an approved indication only for the former, yet SGLT2is are often prescribed to patients with excess weight. Trials comparing them are limited and enroll narrow populations, so real-world evidence on who benefits most is valuable.
This study used the NIH All of Us Research Program, a deliberately diverse U.S. cohort, in an active comparator new-user design. It included adults aged 18-80 with BMI of at least 27 who started either drug class between 2018 and 2023: 2,932 GLP-1RA initiators and 941 SGLT2i initiators, each with no prior exposure to the other class. Inverse probability of treatment weighting, including heart failure and chronic kidney disease, balanced all 28 diagnostics (maximum standardized difference 0.077).
At 12 months, weighted mean weight change was -4.13% with GLP-1RAs and -2.94% with SGLT2is, a difference of -1.20 percentage points (95% CI -1.91 to -0.48; P = .001). Observed rates of at least 10% weight loss were 19.7% vs 10.2%. After weighting, the odds ratio was 1.75 (95% CI 1.40-2.19), the marginal risk difference was +7.04 points, and the number needed to treat was 14. Bootstrap resampling gave a similar interval.
The gap was not uniform. In people with obesity but no type 2 diabetes, GLP-1RA users lost 1.96 points more (P < .001), while in those with type 2 diabetes the difference was essentially zero (+0.08; interaction P = .006). Women lost 1.71 points more on GLP-1RAs (P = .001), whereas men showed no difference (+0.12; interaction P = .017). By race/ethnicity, the effect reached significance in Non-Hispanic White participants (-1.84 points) but was imprecise in Black (-0.81) and Hispanic/Latino (-0.20) participants, with all estimates favoring GLP-1RAs.
Secondary outcomes were modest. HbA1c fell 0.261 points more with GLP-1RAs (P = .032; n = 1,191 with data), while systolic blood pressure did not differ (-0.64 mmHg; P = .588). Excluding the 58 tirzepatide users barely changed the result (-1.09 points), and adding baseline weight to the model left the estimate at -1.19.
The 1.2-point difference is far smaller than GLP-1RA versus placebo in trials, because both arms received active drugs, adherence and dosing varied, and follow-up was only about a year. Key caveats are that the diabetes and sex interaction tests were post-hoc and hypothesis-generating, unmeasured confounding (such as prescriber specialty or adherence) is possible, and the SGLT2i group was small. Confirmation in prospective, sex-stratified studies is needed.
Key Findings
- GLP-1RA initiators lost 4.13% of body weight vs 2.94% for SGLT2i initiators at 12 months: a 1.20 percentage-point difference (95% CI -1.91 to -0.48; P = .001)
- At least 10% weight loss was reached by 19.7% of GLP-1RA vs 10.2% of SGLT2i users; weighted OR 1.75, RR 1.61, risk difference +7.04 points, NNT 14
- Advantage concentrated in obesity without type 2 diabetes (-1.96 points; P < .001) and absent in type 2 diabetes (+0.08; 95% CI -1.04 to +1.19; interaction P = .006)
- Women lost 1.71 points more on GLP-1RAs (P = .001); men showed no difference (+0.12; P = .821; interaction P = .017)
- HbA1c fell 0.261 points more with GLP-1RAs (95% CI -0.500 to -0.022; P = .032), a modest difference; systolic blood pressure did not differ (-0.64 mmHg; P = .588)
- Excluding tirzepatide users (n = 58) left results unchanged (-1.09 points; OR 1.69), so the finding is not driven by the most potent agent
- By race/ethnicity: Non-Hispanic White -1.84 points (P = .0006); Black -0.81 and Hispanic/Latino -0.20, both with CIs crossing zero and limited power
Methodology
Retrospective active comparator new-user cohort study using electronic health records from the NIH All of Us Research Program (2018-2023), with 2,932 GLP-1RA and 941 SGLT2i initiators aged 18-80 with BMI of at least 27 and available weights. Inverse probability of treatment weighting (17-covariate propensity model, stabilized weights truncated at the 99th percentile) was used, with weighted linear and logistic regression and HC3 robust standard errors. The primary outcome was percent weight change at 12 months (6-18-month window), analyzed by intention-to-treat. Sensitivity analyses included bootstrapping, crude regression, excluding tirzepatide, and adding baseline weight; the diabetes and sex interaction tests were post-hoc.
Study Limitations
This is an observational study, so unmeasured confounding from prescriber preference, adherence, dosing and prescribing context cannot be excluded, and intention-to-treat follow-up without censoring at switching may dilute differences. The diabetes and sex interactions were post-hoc, the SGLT2i group was small (limiting subgroup precision, and the Asian subgroup was suppressed), and baseline HbA1c was missing for 58.9% of participants. The supplied text does not report conflicts of interest.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe: