gms | German Medical Science

From 1984 to 2002 more than 900 patients have received a cardiac transplantation at Cardio Thoracic Surgery in Vienna. Physicians are especially interested in the impact of graft arteriosclerosis on survival after heart transplantation. Other factors such as recipient age at surgery, donor age, CMV infection, perioperative infection, any other infection, rejection, gender, diabetes mellitus and induction therapy have to be considered as well. It is important to have an extended follow-up because graft arteriosclerosis will develop over a long period of time.

The most popular model used for survival analysis is the proportional hazards regression model proposed by Cox [Ref. 1]. Nevertheless the fundamental assumption of the Cox model is the proportionality of the hazards. For many applications this assumption is doubtful. In the analysis proportional hazards were explored by the residual score test and graphically using a log(-log(survival)) plot and Schoenfeld residuals [Ref. 2].

To overcome the violation of proportional hazards, 2 strategies are used: the incorporation of time dependent covariates and the fit of an alternative model, the accelerated failure time model. Accelerated failure time models (AFT) provide an alternative framework to fit the data [Ref. 3]. A further advantage of the AFT model is, that a direct effect of the explanatory variables on the survival time is measured and not a conditional probability as in the Cox model.

The results of the AFT model and the Cox proportional hazards model with time dependent covariates are compared according to their performance.