PURPOSE: Estimating the rate of adverse events (AEs) caused by a treatment in clinical trials typically involves comparing the proportions of patients experiencing AEs in intervention and control groups. However, potentially important information, including duration, recurrence, and intensity of events, is lost. In this study, we illustrate how the additional information can be obtained and incorporated into analyses of AEs.
METHODS: Data on psychiatric AEs were extracted from clinical study reports (CSRs) provided by the manufacturer of oseltamivir in 4 prophylaxis randomised trials in adults and adolescents. We analysed the incidence, recurrence, duration, and intensity of events, using logistic regression models where the outcome compared was proportion of days suffering from an event, and developed novel presentation techniques.
RESULTS: Psychiatric AEs were generally more frequent, longer, and more intense in the treatment than placebo arms. Logistic regression models confirm the apparent association overall (odds ratio [OR] 3.46, 95% confidence interval [CI] 1.28 to 9.32), particularly for events classified as severe (OR 34.5, 95% CI 3.66 to 325). However, the absolute difference in proportion of days suffering from severe psychiatric AEs between groups was small.
CONCLUSIONS: This example analysis shows evidence of a causal effect of oseltamivir on psychiatric AEs, not apparent in the published versions of the same trials and a Cochrane review which showed a nonsignificant 81% increased odds of experiencing a psychiatric event. This unique and important finding was dependent on obtaining previously unavailable data from clinical study reports and using novel analyses and presentation methods.