Citation

Duan JZ. J. Clin. Pharmacol. 2009; 49(5): 540-552.

Copyright

(Copyright © 2009, American College of Clinical Pharmacology, Publisher SAGE Publishing)

DOI

10.1177/0091270009333485

PMID

unavailable

Abstract

The purpose of this study is to compare and evaluate logistic regression and time-to-event analysis, 2 commonly used methods for exposure—adverse event (AE) analyses. An AE data set selected from clinical trials is analyzed by both methods and the results are compared. The parameter estimates, odds ratios for logistic regression, and hazard ratios for time-to-event analysis for each AE are compared and further analyzed. In a data set involving 822 patients, 25 AEs are analyzed. A linear relationship is demonstrated between the parameter estimates from the 2 methods and between the odds ratios and hazard ratios. The small differences between the 2 analyses are related to the lower rate of the events and the weaker effects of the risk factors. Although the 2 methods can make predications for the risks, the severity, duration, and recurrence are not well defined. AE time profiles showing the onset, duration, and offset of AEs are important for risk assessment and management. Both analyses can provide information about exposure—AE relationship, and the results from the 2 analyses are consistent in most cases. One should not use the logistic model when length of follow-up varies because of biased estimates. The application of the 2 methods should be combined with AE profiling.