Estimation of Survival of Liver Cirrhosis Patients, in the Presence of Prognostic Factors Using Accelerated Failure Time Model as an Alternative to Proportional Hazard Model

Accelerated Failure Time Model (AFTM) encompasses a wider range of survival time distributions as compared to Cox proportional hazard (PH) model. This article illustrates the use of accelerated failure time model as an alternative to the proportional hazard model in the analysis of time to event data. This kind of study is being done for the first time on Indian population wherein a retrospective data of 666 admitted patients suffering from liver cirrhosis has been obtained and analyzed by both Cox PH and AFT models to evaluate the effect of covariates on the survival of these patients. Model selection criteria include minimization of AIC and graphs showing approximation of cumulative Cox-Snell residuals to (-log) Kaplan-Meier estimates to select the best model. It was conclusively established through the selected model that patients with higher level of serum creatinine and presence of altered sensorium are the significant factors affecting the survival of these patients. In multivariate analysis, all AFT models were judged to be better than Cox regression; Log logistic AFT model was found to be the best fit among candidate models.