Somatosensory and peripheral neuropathic outcomes after semaglutide or tirzepatide initiation compared with other anti-diabetic medications

Abstract

Glucagon-like peptide-1 (GLP-1) receptor agonists, including semaglutide and the dual glucose-dependent insulinotropic polypeptide and GLP-1 receptor agonist tirzepatide, are widely used for type 2 diabetes and obesity, and trials and pharmacovigilance reports have raised concern about altered sensation and peripheral neuropathy. Using de-identified electronic health records from a U.S. federated data network, we ascertained 14 new-onset somatosensory and peripheral neuropathic phenotypes over 12 months using diagnosis codes and physician-validated, AI-curated clinical-note concepts (positive predictive value 88.8%) among adult new users of semaglutide or tirzepatide without these phenotypes at baseline. Of 169,657 at-risk users, 40,142 (23.7%) had a newly documented phenotype, most often hypoesthesia (13.0%) and paresthesia (12.3%). Incidence rose with baseline glycated hemoglobin (HbA1c), from 19.2% below 5.7% to 33.7% at 12% or higher (Cochran–Armitage p < 0.001), and showed no consistent relation to achieved weight loss. In 35,187 propensity-score-matched pairs, GLP-1 users had a lower 12-month incidence of any phenotype than initiators of metformin, dipeptidyl-peptidase-4 inhibitors or sodium-glucose cotransporter-2 inhibitors (22.1% vs. 26.3%; relative risk [RR] 0.84, 95% CI 0.82–0.86), most markedly for polyneuropathy (RR 0.69, 0.64–0.74). Seven of nine negative-control outcomes differed between the arms, in both directions; after empirical calibration to these controls, no estimate was statistically significant (any phenotype, calibrated RR 0.85, 0.58–1.25). In this observational study we found no evidence of a population-level excess of documented somatosensory or peripheral neuropathic phenotypes after semaglutide or tirzepatide initiation; the modestly lower incidence was within the range of systematic error indicated by the negative controls.