Risk of non-infectious uveitis or myasthenia gravis in patients on checkpoint inhibitors in a large healthcare claims database.

Abstract

To determine if checkpoint inhibitors (CPIs) confer an increased risk of non-infectious uveitis or myasthenia gravis (MG) compared to patients on non-checkpoint inhibitor (N-CPI) chemotherapy. A retrospective cohort study was performed comparing patients in a large commercial and Medicare advantage database exposed to CPI compared to N-CPI. All patients who initiated a CPI (ipilimumab, pembrolizumab, nivolumab, atezolizumab, avelumab, cemiplimab and durvalumab) were eligible. Date of earliest CPI in the exposure group and N-CPI chemotherapy in the comparator group was considered the index date. Exclusion occurred in both cohorts for any history of uveitis or MG diagnosis and having <1 year in the insurance plan prior to the index date, and <6 months in plan following the index date. Every exposed patient was matched up to 1:10 based on demographics and index year to patients on N-CPI chemotherapy. Multivariate Cox proportional hazards regression modelling was performed. For evaluation of incidence of non-infectious uveitis, 26 (0.3%) of 8678 patients on CPI and 123 (0.2%) of 76 153 N-CPI comparators were found to have non-infectious uveitis. After multivariate analysis, CPIs showed an increased hazard for uveitis compared to N-CPI (HR=2.09; 95% CI 1.36 to 3.22, p=0.001). For the MG analysis, 11 (0.1%) of 9210 patients developed MG in the CPI group and 36 (0.04%) of 80 620 comparators. The CPI cohort had a higher hazard of developing MG (HR=2.60; 95% CI 1.34 to 5.07, p=0.005) compared to controls in multivariate analysis. Exposure to CPI confers a higher risk for non-infectious uveitis and MG compared to N-CPI chemotherapy.