Not always a good night's sleep: A global pharmacovigilance risk profiling of drug-induced sleep disorders.

OBJECTIVES

Sleep disorders and disturbances affect more than one-third of the global population. Although certain drugs have been implicated, the real-world breadth and strength of these signals across therapeutic classes remain poorly characterised. This study aimed to systematically profile drug-induced sleep disorders and disturbances using large-scale pharmacovigilance data and to identify drugs with disproportionate reporting, including those not previously well recognised.

METHODS

We analyzed 21 years of post-marketing surveillance data using four disproportionality algorithms (Reporting Odds Ratio [ROR], Proportional Reporting Ratio [PRR] with χ2, Bayesian Confidence Propagation Neural Network [BCPNN], and Multi-item Gamma Poisson Shrinker [MGPS]), leveraging their complementary strengths in sensitivity, bias reduction, stability, and rare-event detection. Positive signals were defined as meeting pre-specified thresholds for all four algorithms simultaneously.

RESULTS

Among over 22 million reports, 411,736 involved sleep disorders (58.9% female). Overall, 131 drugs showed disproportionate signals; 24 (18.3%) lacked FDA-labelled warnings. Varenicline accounted for the most reports (n = 16,200), whereas montelukast showed the strongest signal (ROR 11.78, 95% CI 11.40-12.18). Median time to onset was 36 days (IQR: 6-260), with an early failure-type profile (Weibull β = 0.449, 95% CI 0.447-0.452). Among therapeutic classes, antineoplastic and immunomodulating agents (ATC L01) showed the highest disproportionality (ROR 9.19, 95% CI 8.86-9.53).

CONCLUSIONS

Multiple drug classes show disproportionate reporting for sleep disorders and disturbances with an early-onset risk profile, supporting monitoring at treatment initiation. Approximately one in five of these drugs lack US labelling warnings, highlighting the need for clinical awareness and mechanistic studies.