When would you use a log-rank test?

• A. To compare means of continuous outcomes at a fixed time point
• B. To compare survival curves between groups in the presence of censorship
• C. To evaluate risk factors in cross-sectional data
• D. To test equivalence of two proportions in a cohort

Answer: (B)

The main concept is comparing survival curves when some data are censored. The log-rank test is designed for time-to-event data, where not everyone experiences the event during the study and some observations are censored. It evaluates whether two or more groups have the same survival experience over the entire follow-up period by looking at, at each event time, how many events occur versus how many would be expected if the groups shared the same survival curve. These observed versus expected counts are summed across all event times to form a test statistic that follows a chi-square distribution under the null hypothesis of identical survival curves. This test handles censoring properly and does not rely on a specific distribution for survival, making it the appropriate tool for comparing survival curves.

Why the other options don’t fit: comparing means of a continuous outcome at a fixed time point ignores the timing of events and censoring; cross-sectional data lack time-to-event information altogether; and testing equivalence of two proportions doesn't incorporate time-to-event data or censoring.