Evolution of statistical methodology and design of phase II/III clinical trials in non-small cell lung cancer (NSCLC)

Abstract

7108 Background: A significant heterogeneity exists in the design and reporting of phase II and III therapeutic clinical trials in NSCLC. This has led to difficulty in interpretation of these trials leading to over- or underestimation of therapeutic efficacy. We set out to investigate the statistical methodology and design reporting of chemotherapeutic trials in NSCLC published in the Journal of Clinical Oncology (JCO) over 20 years. Methods: We identified all phase II and III NSCLC chemotherapy trials published in the JCO from January 1983 to August 2005. All manuscripts were reviewed to evaluate components of statistical design that were reported, including: sample size calculation, power, type I error, single or multiple drug trials, relative response sought in phase II trials and improvement in survival time or response rate sought in phase III trials. Results: One hundred forty eight trials were identified. 52% of studies were phase III and 48% were phase II. The majority (78%) were conducted in advanced stage NSCLC. Sample size calculations were reported for only 58% of phase III studies and 31% of phase II studies. Power was reported in 66% of phase III studies and 13% of phase II trials. Type I error was reported in 47% of phase III studies and 17% in phase II studies. 60% of phase III trials defined endpoints (percentage improvement in survival time, improvement in survival time in months or increase in response rate). 41% of phase II trails defined the target response rate, ranging from response rates of 15% to 70%. The frequency of adequate reporting of statistical design was shown to increase from 31% in 1990–1995 to 64% in 2000–2005 ( table ). Conclusions: Significant heterogeneity exists in trial design and reporting of phase II and III trials in NSCLC. This impacts the ability to adequately interpret these studies. More widespread application of statistical methods in planning and reporting of lung cancer clinical trials are necessary to increase reliability of data. [Table: see text] No significant financial relationships to disclose.