Analysis of over 2,400 modern phase I cancer trials: Composition, outcomes, and use of surrogate endpoints

作者：Thomas G. Roberts, Bernardo H. L. Goulart, Sarah Stallings, Lee S. Squitieri, Bruce Allan Chabner, Stan N. Finkelstein, Jeffrey William Clark · 发表于：Journal of Clinical Oncology · 年份：2004 · DOI：10.1200/jco.2004.22.14_suppl.2084 · 被引用次数：12 · 研究领域：Statistical Methods in Clinical Trials、Cancer Genomics and Diagnostics、Health Systems, Economic Evaluations, Quality of Life

2084 Background: There has been no comprehensive analysis of phase I cancer trials since 1991, despite the transition from cytotoxic to targeted drugs. Trends in response rates (RR) and toxic death rates (TDR) for modern phase I trials are therefore unknown. We set out to perform the most extensive outcomes analysis to date of phase I cancer trials. Methods: We analyzed the composition and outcomes of all phase I cancer trials submitted to ASCO for years 1991 through 2002. For trials testing unapproved, single-agents in solid tumors, we reviewed published reports and performed patient-level analysis. The major outcomes variables included overall response rate (RR) and toxic death rate (TDR). Because of changes in phase I trial design, we also analyzed the use of biomarkers (BM) and surrogate endpoints (SE). Results: The overall data set included 2,439 trials. Composition and outcomes varied by year of submission to ASCO (Table). Over the period of analysis, phase I trials have become significantly more international, more complex in design, and more likely to identify a commercial sponsor. Drugs under investigation have become more likely to be given by the oral route, less likely to be produced by recombinant technology, and less likely to be cytotoxic. The strongest predictor for inclusion of a BM or SE was NIH sponsorship (odds ratio: 2.9). The overall RR and TDR for single-agent studies published in final form was 3.96% and 0.6%, respectively. Both RR and TDR fell signifi...