A novel test to compare two treatments based on endpoints involving both nonfatal and fatal events.

TitleA novel test to compare two treatments based on endpoints involving both nonfatal and fatal events.
Publication TypeJournal Article
Year of Publication2015
AuthorsPotthoff, Richard F., and Susan Halabi
JournalPharm Stat
Volume14
Issue4
Pagination273-83
Date Published2015 Jul-Aug
ISSN1539-1612
KeywordsCarcinoma, Renal Cell, Clinical Trials, Phase III as Topic, Computer Simulation, Data Interpretation, Statistical, Endpoint Determination, Humans, Kidney Neoplasms, Models, Statistical, Randomized Controlled Trials as Topic, Research Design, Statistics, Nonparametric, Time Factors, Treatment Outcome
Abstract

In a clinical trial comparing two treatment groups, one commonly-used endpoint is time to death. Another is time until the first nonfatal event (if there is one) or until death (if not). Both endpoints have drawbacks. The wrong choice may adversely affect the value of the study by impairing power if deaths are too few (with the first endpoint) or by lessening the role of mortality if not (with the second endpoint). We propose a compromise that provides a simple test based on the time to death if the patient has died or time since randomization augmented by an increment otherwise. The test applies the ordinary two-sample Wilcoxon statistic to these values. The formula for the increment (the same for experimental and control patients) must be specified before the trial starts. In the simplest (and perhaps most useful) case, the increment assumes only two values, according to whether or not the (surviving) patient had a nonfatal event. More generally, the increment depends on the time of the first nonfatal event, if any, and the time since randomization. The test has correct Type I error even though it does not handle censoring in a customary way. For conditions where investigators would face no easy (advance) choice between the two older tests, simulation results favor the new test. An example using a renal-cancer trial is presented.

DOI10.1002/pst.1683
Alternate JournalPharm Stat
Original PublicationA novel test to compare two treatments based on endpoints involving both nonfatal and fatal events.
PubMed ID25894200
PubMed Central IDPMC4503508
Grant ListU01 CA157703 / CA / NCI NIH HHS / United States
P01 CA142538 / CA / NCI NIH HHS / United States
R01 CA155296 / CA / NCI NIH HHS / United States
CA142538 / CA / NCI NIH HHS / United States
CA155296 / CA / NCI NIH HHS / United States