Survival analysis: part 15 of a series on evaluation of scientific publications.

BACKGROUND: Survival times are often used to compare treatments. Survival data are a special type of data, and therefore have to be analyzed with special methods.

METHODS: We illustrate special techniques for analyzing survival times by applying them to a publication on the treatment of patients with brain tumors. The present article is based on textbooks of statistics, a selective review of the literature, and the authors' own experience.

RESULTS: Survival times are analyzed with the Kaplan-Meier method, which yields two measures of interest: survival rates and the median survival time. The log-rank test is used to compare survival times across treatment groups. Cox regression is used in multivariable models. The hazard ratio, a descriptive measure for differences in survival times, is explained.

CONCLUSION: If survival times are analyzed without the use of special techniques, or if the underlying assumptions are not taken into account, faulty interpretation may result. Readers of scientific publications should know these pitfalls and be able to judge for themselves whether the chosen analytical method is correct.