Seasonality in Adverse Drug Events: Time-Series Analysis of JADER Using ARIMA/SARIMA and Prophet

The Japanese Adverse Drug Event Report (JADER) database is a spontaneous reporting system that compiles real-world data on adverse events (AEs) associated with drug use in Japan, including date of occurrence, allowing investigation of temporal trends. This study aimed to identify AEs that exhibit seasonal variation by applying time-series models—the autoregressive integrated moving average (ARIMA), seasonal ARIMA (SARIMA), and Prophet—to monthly AE reports from January 2005 to December 2019. Models were optimized using grid search and time-series cross-validation, and model performance was evaluated using mean absolute error, root mean squared error, and mean absolute percentage error. AEs whose seasonal models outperformed non-seasonal ones were considered likely seasonal. Sixteen AEs, including influenza, Guillain-Barré syndrome, encephalitis, and adverse reactions related to antineoplastic agents or immune checkpoint inhibitors, were identified as exhibiting potential seasonality. Some AEs showed clear peaks in specific months, while others demonstrated discrepancies between observed data and modeled seasonal components. These results indicate that certain AEs may be influenced by seasonal factors such as infectious diseases, climate, or patient behaviors. Identifying these seasonal trends can support proactive risk management, inform medication safety strategies, and assist clinicians, regulators, and researchers in strengthening pharmacovigilance and improving patient outcomes.