Statistical considerations in clinical trial design with event-free survival as the primary efficacy endpoint

In late-phase confirmatory clinical trials in the oncology field, time-to-event (TTE) endpoints are commonly used as primary endpoints for establishing the efficacy of investigational therapies. Among these TTE endpoints, overall survival (OS) is always considered as the gold standard. However, OS data can take years to mature, and its use for measurement of efficacy can be confounded by the use of post-treatment rescue therapies or supportive care. Therefore, to accelerate the development process and better characterize the treatment effect of new investigational therapies, other TTE endpoints such as progression-free survival and event-free survival (EFS) are applied as primary efficacy endpoints in some confirmatory trials, either as a surrogate for OS or as a direct measure of clinical benefits. For evaluating novel treatments for acute myeloid leukemia, EFS has been gradually recognized as a direct measure of clinical benefits. However, the application of an EFS endpoint is still controversial mainly due to the debate surrounding definition of treatment failure (TF) events. In this article, we investigate the EFS endpoint with the most conservative definition for the timing of TF, which is Day 1 since randomization. Specifically, the corresponding non-proportional hazard pattern of the EFS endpoint is investigated with both analytical and numerical approaches.

Automated summary

In late-phase confirmatory clinical trials in oncology, time-to-event endpoints such as progression-free survival and event-free survival are increasingly used as primary efficacy endpoints alongside overall survival. Event-free survival is being recognized as a direct measure of clinical benefits in evaluating novel treatments for acute myeloid leukemia, despite controversies surrounding its definition of treatment failure.