|
Abstract:
|
In many multicenter, randomized, controlled clinical trials in the device industry, a primary or secondary endpoint, occurring during the course of the clinical trial, may be best analyzed with time to event modeling techniques. Further there may be an interest in long term follow-up as confirmatory evidence of the success for a particular measure. Often the analysis can be enhanced by consideration of covariates that may influence results. Correctly adding in covariates to the analysis as well as correctly assessing analysis assumptions is critical for proper interpretation of treatment effect in this setting. This investigation considers different model building strategies using the Cox regression model which include assessing fitted Proportional Hazards assumptions, methods of selecting model covariates, and utility of time-dependent covariates for this analysis.
|
