|
Abstract:
|
Although many statistical tests compare two event time distributions exist in statistical literature, our recent systematic review demonstrated that 97% papers reporting phase III cancer clinical trial results used logrank test. The reason of this result would be because logrank test is the most powerful for proportional hazards (PH) alternatives. A natural question arises from here is how likely we have a PH case empirically. If PH is true in almost all trials, this analytic practice would be reasonable. However, if not, the analytic approach could be improved, especially in a way summarizing treatment effect. Currently, hazard ratio (HR) is routinely used for it, although issues and concerns of HR have been extensively discussed for the past couple of years. Such domination of using HR would be connected with the dominated use of logrank test in almost all trials because of their coherency. Motivated by this, we conduct a further systematic review and identify the proportion of PH cases in cancer clinical trials. This investigation will reveal if the current test/estimation approach is reasonable. It can be breakthrough and enhance utilization of alternatives to HR.
|
