Please enable JavaScript to view this site.

This guide is for an old version of Prism. Browse the latest version or update Prism

In many clinical and animal studies, the outcome is survival time. The goal of the study is to determine whether a treatment changes survival. Prism creates survival curves, using the product limit method of Kaplan and Meier, and compares survival curves using both the logrank test and the Gehan-Wilcoxon test.

Censored data

Creating a survival curve is not quite as easy as it sounds. The difficulty is that you rarely know the survival time for each subject.

Some subjects may still be alive at the end of the study. You know how long they have survived so far, but don't know how long they will survive in the future.

Others drop out of the study -- perhaps they moved to a different city or wanted to take a medication disallowed on the protocol. You know they survived a certain length of time on the protocol, but don't know how long they survived after that (or do know, but can't use the information because they weren't following the experimental protocol). In both cases, information about these patients is said to be censored.

You definitely don't want to eliminate these censored observations from your analyses -- you just need to account for them properly.The term “censored” seems to imply that the subject did something inappropriate. But that isn't the case. The term “censored” simply means that you don't know, or can't use, survival beyond a certain point. Prism automatically accounts for censored data when it creates and compares survival curves.

Not just for survival

The term survival curve is a bit restrictive as the outcome can be any well-defined end point that can only happen once per subject. Instead of death, the endpoint could be occlusion of a vascular graft, first metastasis of a tumor, or rejection of a transplanted kidney. The event does not have to be dire. The event could be restoration of renal function, discharge from a hospital, or graduation.

Analyzing other kinds of survival data

Some kinds of survival data are better analyzed with nonlinear regression. For example, don't use the methods described in this section to analyze cell survival curves plotting percent survival (Y) as a function of various doses of radiation (X). The survival methods described in this chapter are only useful if X is time, and you know the survival time for each subject.

Proportional hazards regression

The analyses built in to Prism can compare the survival curves of two or more groups. But these methods (logrank test, Gehan-Breslow-Wilcoxon test) cannot handle data where subjects in the groups are matched, or when you also want to adjust for age or gender or other variables. For this kind of analysis, you need to use proportional hazards regression, which Prism does not do.

© 1995-2019 GraphPad Software, LLC. All rights reserved.