Clinical trial design in small cell lung cancer: Surrogate endpoints and statistical evolution.

Abstract

7087 Background: Small cell lung cancer (SCLC) is a devastating disease for which little recent therapeutic advance has been achieved. SCLC trial design and reporting may have an impact on the interpretation of studies conducted. Furthermore the use of surrogate endpoints in SCLC has not been explored. Methods: Through examining all phase II and III SCLC trials published in the Journal of Clinical Oncology (8,471 patients from 66 trials between 1983-2010), we examined how SCLC trial reporting and design has evolved, determining if the type I error, power, and sample size calculations were provided. We assessed primary endpoints for all trials and sought to discover if response rate (RR) or progression-free survival (PFS) correlated with overall survival (OS). We analyzed response and survival outcomes for associations using Pearson correlation coefficient and differences between data groups by ANOVA followed by Tukey’s multiple comparison procedure. Results: There were increased reporting trends of statistical design in power (16.7% in 1986-1996 to 77.8% of trials in 2006-2010 [p=0.001]) and type I error (22.2% in 1986-1996 to 72.2% in 2006-2010 [p=0.005]). 72.2% of trials published in 1986-1996 failed to report a primary endpoint whereas only 5.56% of trials in 2006-2010 failed to do so (p=0.004). Of phase II trials, primary endpoint was identified as RR in 65%, OS in 25%, and PFS in 10% (p<0.0001). There is a strong correlation between RR and both PFS (p=0.013) and OS (p=0.012) in extensive-stage disease (ED) but not limited-stage disease (LD) (p=0.978 for PFS, p=0.193 for OS). This correlation is for partial response rate (pRR) but not complete response rate. RR (p=0.029) exhibits a negative trend over time with a dramatic and significant decrease in RR across all studies starting in 2005. A strong positive correlation exists between PFS and OS for LD (p=0.036) and ED (p=0.058). We found no change in median overall survival (p=0.383). Conclusions: Over time there has been improvement in reporting the essential statistical data for proper interpretation of SCLC trials. No change in survival was seen over the past 28 years. RR should be evaluated as a surrogate endpoint for OS in ED but not LD. pRR correlates with OS and PFS in both LD and ED.