Estimation of time-dependent rate ratios in case-control studies: comparison of two approaches for exposure assessment

Publisher Website
Google Scholar
Abstract
Purpose In pharmacoepidemiology, it is well recognized that the rate of adverse events may vary as a function of the cumulative duration of the drug exposure and/or the time since the end of the exposure. In case‐control studies, two different approaches have been used to estimate temporal effects of drug exposure: the time‐windows (T‐Ws) approach and the duration‐specific (D‐S) approach. We decided to conduct a simulation study to compare the two approaches when the rate ratios (RRs) vary as a function of the cumulative duration of exposure and/or the time since the end of exposure. Methods We generated three cohorts of 500 000 individuals in which the rate of the event was varying as a function of the cumulative duration of exposure and the time since the end of exposure. For each cohort, a nested case‐control analysis was performed using both the D‐S and the T‐Ws approaches. In the T‐Ws approach, a RR is estimated within specific periods of time prior to the outcome, while a RR is estimated within periods of cumulative duration of exposure and time since the end of exposure in the D‐S approach. Results We found that the RRs obtained from the D‐S approach exactly corresponded to the RRs obtained from the cohort analyses, while the RRs obtained from the T‐Ws approach generally not. RRs obtained from the T‐Ws approach were difficult to interpret in terms of the effect of the duration and timing of the exposure. Conclusion The D‐S approach should be used to investigate the duration‐related effects of exposure in case‐control studies. Copyright © 2005 John Wiley & Sons, Ltd.