Meta-analysis of progression-free survival as a predictor of overall survival in locally advanced or metastatic non-small cell lung cancer trials.

Abstract

7604 Background: Access to new oncology treatments for non-small-cell lung cancer (NSCLC) could be expedited if progression free survival (PFS) was accepted by payers as a valid endpoint predictive for overall survival (OS). We investigated the relationship between PFS and OS in advanced NSCLC, which has previously been limited by available data. Methods: A systematic review of the published literature was conducted (1988 to August 2011; Medline/Embase/Cochrane). Identified studies were assessed against the following criteria for inclusion in the analysis: randomised design; NSCLC; advanced disease (Stage IIIb or IV); pharmacological treatment in 2 or more groups; reported outcomes included median OS and median PFS. For each pair of treatment groups compared, the median OS difference and the median PFS difference were calculated. The data were analysed with locally weighted least squares regression (LOWESS) to validate the linear relationship. Unweighted and weighted linear regression was used to test the significance of the relationship between OS difference and PFS difference. Results: A total of 124 clinical trials, with 40,568 patients, were included in the analysis. The studies ranged from middle aged cohorts to the elderly. Median performance status (ECOG/WHO) was 1. 70% of studies were first line and therapy type was predominantly chemotherapy (57%). 19 studies (15%) reported use of crossover or rescue therapy. Weighted linear regression demonstrated a highly significant linear relationship between OS difference and PFS difference (Table). The results suggest that a 1 month difference in PFS is associated with approximately 1.3 months difference in OS. Conclusions: Analysis of NSCLC trials showed the treatment effect on PFS was predictive of the treatment effect on OS. With increased use of multiple lines of treatment and crossover/rescue therapy in new NSCLC trials OS differences are frequently confounded, making PFS an important outcome for health care decision making. [Table: see text]