A Statistical Framework for Decision Making in Confirmatory Multipopulation Tailoring Clinical Trials
- PDF / 345,034 Bytes
- 10 Pages / 602.986 x 782.986 pts Page_size
- 46 Downloads / 175 Views
A Statistical Framework for Decision Making in Confirmatory Multipopulation Tailoring Clinical Trials
Drug Information Journal 46(6) 647-656 ª The Author(s) 2012 Reprints and permission: sagepub.com/journalsPermissions.nav DOI: 10.1177/0092861512454116 http://dij.sagepub.com
Brian A. Millen, PhD1, Alex Dmitrienko, PhD2, Stephen Ruberg, PhD1, and Lei Shen, PhD1
Abstract This article focuses on statistical analysis of clinical trials pursuing tailored therapy objectives, wherein evaluation of treatment effect occurs in the overall population as well as in a predefined subpopulation(s). The design and analysis principles presented provide a framework for decision making based on these novel multipopulation tailoring trial designs, considering the particular case of confirmatory trials. These principles include traditional multiple testing considerations, as well as 2 new analysis principles. Keywords subgroup analysis, tailored therapy, type I error rate, influence condition, interaction condition
1. Introduction The promise of tailored therapeutics and personalized medicine has resulted in increased attention on evaluation of treatment effects in focused subpopulations in clinical trials. The subpopulations of interest may be defined by demographics, clinical markers, genetic markers, or a combination of these. Although subgroup analyses that explore treatment effects in subpopulations (defined by demographics and other characteristics) has been a standard part of clinical trial analysis plans for decades, the intent of those analyses historically had not been to test a priori hypotheses regarding treatment effect in the subpopulation(s). Instead, these analyses were exploratory and hypothesis generating in nature; the inference set for these trials was the overall patient population only. As interest in focused subpopulations advanced, single population tailoring trials became more common. In these trials, evaluation of treatment effect in a targeted subpopulation was the primary objective, and so-called enrichment designs1 were employed. The Herceptin program2 provides an early and prominent example. More recent examples include the Xalkori registration trials,3 the Zelboraf registration trials,4 and the Alimta program for nonsquamous non–small cell lung cancer.5 At times, such trials were undertaken only after negative results were obtained in overall population trials and the subpopulation was hypothesized based on exploratory subgroup analyses from the negative trial(s). In other cases, overall population trials were omitted altogether in favor of target subpopulation
trials. In these cases, treatment effect in the remainder of the population is unknown. Today, clinical trials with more complex objectives, such as evaluating treatment effects in focused subpopulations, as well as in the broader overall population, are conducted to realize the promise of tailored therapeutics. These multipopulation tailoring trials offer several advantages over the single population trials. As these trials provide inference
Data Loading...
