PCN4: COMPARING MEAN VERSUS MEDIAN SURVIVAL AS A PRELUDE TO COST-EFFECTIVENESS (C/E) ANALYSES

Abstract

BACKGROUND: In a multinational trial designed to determine the efficacy of epoetin alfa in chemotherapy-induced anemia, patients receiving non-platinum chemotherapy having a hemoglobin 10.5 g/dL or less, or a decline in hemoglobin of 1.5 g/dL or greater were randomized (2:1) to epoetin alfa or placebo. A total of 375 patients (251 epoetin alfa, 124 placebo) were assessed for survival status twelve months after completing the protocol, but prior to unblinding. A log-rank test showed a trend in survival favoring epoetin alfa (median of 17 vs. 11 months, p = 0.128). OBJECTIVES: The primary efficacy endpoint for many cancer clinical trials is median survival. In preparing for an economic analysis, however, we analyzed mean survival, the appropriate survival endpoint for a C/E analysis. METHODS: Sampling with replacement, we conducted a post hoc analysis that examined the difference in mean survival by drawing 10,000 samples in a bootstrapping simulation. Within each sample the survival curves were truncated to maintain identical follow-up periods between treatment groups. The difference in mean survival was computed for each sample. The probability of superior efficacy was obtained by sorting the results from the samples. RESULTS: The average mean survival difference, across the 10,000 samples, showed a 0.212-year survival benefit for epoetin alfa. The probability that the difference in mean survival favors epoetin alfa was 0.965. CONCLUSIONS: Comparing differences in median and mean survival may lead to different conclusions about the value of a therapy. Given that mean survival is the appropriate effectiveness endpoint for survival-based C/E analyses, a non-significant difference in median survival does not preclude full C/E analyses. Specifically, the mean survival results from this trial warrant a full C/E analysis of epoetin alfa in treating anemia for patients receiving non-platinum chemotherapy.