Risk of diabetes mellitus among users of immune checkpoint inhibitors: A population-based cohort study

Immune checkpoint inhibitors (ICIs) are increasingly established cancer therapeutics, but they are associated with new-onset diabetes mellitus (DM). Such risks have not been adequately quantified, and between-class and -sex differences remain unexplored. This was a prospective cohort study of cancer patients receiving any ICI in Hong Kong between 2013 and 2021. Patients with known DM were excluded. Due to few patients using other ICIs, only programmed cell death 1 inhibitors (PD-1i) and programmed death ligand 1 inhibitors (PD-L1i) were compared, alongside between-sex comparison. When comparing PD-1i against PD-L1i, patients with the use of other ICIs or both PD-1i and PD-L1 were further excluded. Inverse probability treatment weighting (IPTW) was used to minimize between-group covariate imbalances. Altogether, 3375 patients were analyzed (65.2% males, median age 62.2 [interquartile range 53.8-69.5] years old). Over a median follow-up of 1.0 [0.4-2.4] years, new-onset DM occurred in 457 patients (13.5%), with a 3-year risk of 14.5% [95% confidence interval 13.3%, 15.8%]. IPTW achieve acceptable covariate balance between sexes, and between PD-1i (N = 622) and PD-L1i (N = 2426) users. Males had significantly higher risk of new-onset DM (hazard ratio 1.35 [1.09, 1.67], p = 0.006), while PD-1i and PD-L1i users did not have significantly different risks (hazard ratio vs PD-L1i 0.81 [0.59, 1.11], p = 0.182). These were consistent in those with at least 1 year of follow-up, and on competing risk regression. Users of ICI may have a substantial risk of new-onset DM, which may be higher in males but did not differ between PD-1i and PD-L1i.