Combining survival endpoints within a single statistical analysis
Suppose that two time-to-event responses are observed on each individual in a comparative clinical trial. For example we may record time to first cardiac event and time to death from any cause; time to loss of vision in the left eye and time to loss of vision in the right eye; or time to the first and time to the second epileptic seizure. This project concerns methods for combining the analyses of the two responses. This leads to a need to estimate the correlation between two test statistics, and in particular between two logrank statistics. In fact, versions of the logrank statistic suitable for interval censored survival data are investigated as a tractable alternative. The new approach is compared with the method of Wei, Lin and Weissfeld, both in terms of its construction and in terms of its performance in the analysis of real and simulated data.