Multivariate survival analysis using Cox's regression model

Abstract

In recent years, the Cox regression model has been used increasingly for analysis of censored survival data. With this model, the pattern of association (covariation) of many predictor variables with survival is analyzed to identify the combination of variables which best predicts survival. The results can be presented as a "pocket chart," by which a prognostic index for a new subject can easily be obtained. By a simple graph, the prognostic index can be translated to estimates of the probability of surviving a given time or the median survival time predicted for the subject. In controlled clinical trials, the Cox model can be used to adjust for imbalance in variables influencing prognosis and to identify variables being associated with the treatment effect (therapeutic variables). This paper describes in rather simple and practical terms some of the concepts behind the model, how to perform the analyses and how to interpret and utilize the results.