How to correctly perform log-rank tests in a study population of 15 observations and less than 10 events?