Across five University of California medical centers, 70,982 propensity-matched adults with type 2 diabetes starting a GLP-1 receptor agonist, sulfonylurea, SGLT2 inhibitor or DPP4 inhibitor were followed for up to five years across 181 outcomes. Of 118 robust associations, 65 favored GLP-1RAs: acute myocardial infarction (HR 0.53), myocarditis/cardiomyopathy (0.38 vs SGLT2i), heart failure (0.71 vs DPP4i), nephritis/nephrosis (0.73), thromboembolism (~0.70) and tobacco-related disorders (0.60). Fifty-three were harms, led by osteoarthritis (1.48), stress fractures (1.68), sleep-wake disorders and diverticular disease.
The cardiovascular and renal signals match the big randomized outcome trials, which is reassuring for the method. The tobacco-related finding fits emerging work on GLP-1 effects on addictive behavior. The harms are the newer story. Musculoskeletal risk rising 22-fold in count of associations plausibly reflects lean-mass loss alongside rapid weight loss, and stress fractures hint at bone loading problems, though osteoarthritis may also be detected more often in people who become more active. Diverticular disease and sleep disturbance deserve scrutiny too.
Caveats matter. This is observational electronic-record data, so residual confounding and surveillance bias persist despite 11,000-covariate matching and E-values. Screening 181 outcomes invites false positives, hence the strict robustness filters, yet comparator choice shapes every hazard ratio. Findings come from one academic health system and cannot prove causation. Editorially, this is confirmatory on benefits and hypothesis-generating on harms: patients on GLP-1RAs should pair treatment with resistance training and adequate protein, and discuss bone and joint symptoms with their clinician.