Statistical methods for comparing survival rates at a fixed time point / 中华流行病学杂志

ABSTRACT

Comparison of survival curves between two groups is an important part of disease prognosis study.Log-rank test is commonly used,but when the two curves' later intersecting opening is too large,the proportion of assumptions is not established,thus the Log-rank test is ineffective.We introduces five statistical tests to compare two survival curves at a fixed time points:classic method,logarithmic transformation,cloglog transformation,arcsine transformation and logit transformation.Through the study we found that if the overall survival curves are difficult to compare between groups tested with Log-rank test method or Two-stage test method,the fixed time point test can effectively determine whether there was significant difference in survival rate at a fixed time point.Among the five fixed time point tests,cloglog transformation could give more precise result.