A Composite Endpoint Approach to Evaluate Treatment Benefits Based on Patient Preferences
Regulatory approval of a medical product considers both benefits and harms that can be measured by multiple endpoints. The importance of these endpoints may vary among patients. Current approaches integrate multiple outcomes without reflecting heterogeneity of patient preferences. In this paper, we proposed a new composite desirability of outcome ranking (DOOR) to define a winning probability. Participants are stratified and randomized according to their baseline preference ranking. A Wilcoxon-Mann-Whitney U-statistic averages the pairwise DOOR between one treated and one controlled participant. The DOOR ranks outcomes with consideration of differences in participant’s ranking of outcome importance. In a pair of participants, the smallest common top ranking outcomes will be compared. If the treated participant has at least one better (or worse) outcome than the controlled participant and equivalent in the remaining outcomes, the treatment is more (or less) desirable. Otherwise, the pair has a tie and the next level of common top ranking outcomes will be explored. We estimate the winning probability of treatment over control arms and make statistical inferences about this estimand. Statistical properties of our method are examined and compared with other conventional approaches to demonstrate the differences. The approach is illustrated with three trial design examples.
This is a joint work with Drs. Qian Zhao, Shiyan Yan, Lu Tian, Xin Tu, Jie Chen, Lorene Nelson, and Ms. Jiying Zou.