I am conducting a meta-analysis of survival time. Pooling in survival
rates at t=1 is straightforward using the metan command for relative risk
of survival. However, what about pooling median survival times?
I have thought of creating a table of ordered failure times, treating
individual studies as if they were patients in a trial, and then comparing
them using the log rank test. Or, calculating the mean and confidence
interval of the median survival times for each trial, and simply reporting
those without a statistical test. In my data I have each treatment group in
each trial:
