Associations of mortality and sepsis after GLP-1 receptor agonist initiation versus SGLT2 inhibitor or metformin initiation in people with HIV, overweight or obesity, and type 2 diabetes: A target trial emulation

Abstract

People living with HIV (PLWH) with overweight or obesity and type 2 diabetes (T2D) carry elevated cardiometabolic and infectious morbidity. Comparative effectiveness data for glucagon-like peptide-1 receptor agonists (GLP-1 RAs)against active comparators are scarce in this population. To estimate associations between GLP-1 RA initiation andclinical outcomes compared withsodium-glucose cotransporter-2 inhibitors (SGLT2i) or metformin. Target trial emulation of new-user, active-comparator cohorts in the TriNetX US Collaborative Network (2017 to 2025), with pre-specified replication in the Global Collaborative Network. Adults with HIV, overweight or obesity (ICD-10-CM E66 or body mass index 27 kg/m2 or higher), and T2D newly starting a GLP-1 RA, SGLT2i, or metformin after a 6-month washout were eligible. The primary endpoint was all-cause mortality within 3 years of the index date; secondary endpoints were A41.9-coded sepsis, candidiasis, major adverse cardiovascular events, major adverse kidney events, other or unspecified encephalopathy, and three nested exploratory code-based neurocognitive definitions (definitions 1, 1 to 2, and 1 to 3). We built 1:1 propensity score matched cohorts and fitted Cox models with Benjamini-Hochberg correction within each comparator stratum. Analyses included 2,720 matched pairs vs SGLT2i and 4,604 vs metformin. GLP-1 RA initiation was associated with a lower estimated hazard of all-cause mortality during up to 3 years of follow-up vs SGLT2i (hazard ratio [HR] 0.63, 95% confidence interval [CI] 0.49–0.81) and vs metformin (HR 0.65, 0.53–0.81). Kaplan-Meier estimated 3-year mortality risk was 6.7% versus 9.4% against SGLT2i and 5.4% versus 7.4% against metformin. A41.9-coded sepsis showed a lower hazard vs SGLT2i (HR 0.71, 0.53–0.94) and metformin (HR 0.75, 0.60–0.94). Cardiovascular and kidney composites differed less clearly, with renal outcomes directionally favoring SGLT2i. Code-based neurocognitive endpoints were lower vs metformin, but were exploratory and limited by administrative coding; reported significance reflects adjusted P values. GLP-1 RA initiation was associated with a lower observed hazard of all-cause mortality and of A41.9-coded sepsis than SGLT2i or metformin in PLWH with overweight or obesity and T2D. These associations remainvulnerable to residual confounding and require prospective confirmation. Emulation is not equivalent to randomization.