Median Survival, Hazard Ratios and Bounding Drug Effectiveness
21 Pages Posted: 2 Jan 2014
Date Written: December 31, 2013
Abstract
This paper shows that differences in median survival rates are not be informative of a drug's effectiveness. The existence or lack of existence of a difference does not inform public policy or patient decision making. Moreover, other commonly used summary measures such as Kaplan-Meier plots and hazard ratios are generally not informative either. Published results from the EMILIA trial show that median survival difference was 5.8 months and the hazard ratio was statistically significantly less than 1 for T-DM1 against the benchmark. The FDA approved T-DM1 for metastatic breast cancer treatment based upon these results. However, without strong and non-credible assumptions we can only say that at least 7% of the sampled population would live longer on T-DM1 relative to the benchmark and some positive proportion of patients who would live 6 months longer on T-DM1 relative to the benchmark. The presented results do not allow us to rule out the possibility that most of the sampled population would live longer on the benchmark than on T-DM1.
Keywords: treatment effect, median survival, bounds
JEL Classification: C34, C90, I12, I18
Suggested Citation: Suggested Citation