Statistical Evaluation of Responder Analysis in Stem Cell Clinical Trials

In evaluating treatment efficacy, there is an ongoing discussion about which endpoint is more efficient to represent the treatment effect. Absolute change (AC) is the difference between before and after treatment, while relative change (RC) is the AC relative to the baseline value. Principal investigators sometimes support the credibility of relative change, but the FDA is more likely to support absolute change. Therefore, whether these two endpoints can be translated or combined is worth investigating in order to satisfy both parties.

In this article, a motivating example is presented to show that the choice of endpoint will result in different conclusions. The compared relationship of AC and RC is discussed in terms of required sample size, power, and precision. A new type of responder endpoint that combines the concepts of AC and RC is proposed. The comparative relationship regarding sample size, power, and precision of the proposed responder endpoint and the original two endpoints are also investigated.

As a result, the performance of AC and RC is highly dependent on the choice of threshold that is often informed based on minimum clinically important difference or other clinical experience. Therefore, an absolute translation between them is hard to achieve. Inspired by the concept of responder analysis, three types of responder endpoints are proposed and discussed. The pattern of the third type of responder endpoint is having higher power, higher precision, and less required sample size in estimating the treatment effect compared to AC and RC within a range of thresholds. This advantage becomes more obvious when applying higher AC and RC thresholds and lower [Formula: see text] threshold.

The proposed endpoint incorporates the information from the AC and RC endpoints and could be another wise choice when designing clinical trials especially when there is no absolute preference between AC and RC.