Survival Analysis Using Primary and Surrogate Endpoints

GAIL E. TUDOR. Survival Analysis Using Primary and Surrogate Endpoints. (Under the direction of TIMOTHY M. MORGAN) Current survival techniques do not provide a good method for handling clinical trials with a large percent of censored observations. Some methods that have been proposed to increase the power of analyses· include covariate adjustment, multivariate models, use of surrogate endpoints and the use of time dependent covariates. This research proposes using time dependent surrogates of survival as outcome variables, in. conjunction with observed survival time to improve the precision in comparing the relative effects of two treatments on the distribution of survival time, or in estimating the parameters in the survival model, where the outcome measure is time to a specified event. This is in contrast to the standard method which uses the marginal density of survival time, T, only, or the marginal density of the surrogate, X, only, therefore, ignoring some available information. The surrogate measure, X, may be a fixed value or a time dependent variable, X(t). X is a summary measure of some of the covariates measured throughout the trial that provide additional information on a subject's survival time. It is possible to model these time dependent covariate values and relate the parameters in the model to the parameters in the distribution of T given X. The efficiency of the standard method compared to the proposed method is presented for several different cases which use a fixed or time dependent variable. The last example applies data from a study on the effects of lipids on coronary heart disease. The proposed method is always as efficient or more efficient than the standard method. When the treatment effects survival by having a multiplicative effect on the conditional hazard the proposed method is most efficient when the values of X are very heterogeneous and there is little censoring. When treatment affects the distribution of the surrogate measure, X, only, the proposed method is most efficient when the values of X are homogeneous and there is a large amount of censoring. When both are affected the largest gain is achieved when X is very heterogeneous and there is little censoring.