Modelling the Survival Time of HIV Positive Pediatrics using Cox PH and Accelerated Failure Time Models

Abstract

The introduction of antiretroviral therapy in 1996 improved the longevity and wellbeing of peoples living with HIV in the industrialized world including children. This survival benefit of antiretroviral therapy (ART) in reducing HIV related deaths has been well studied in the developed world. In resource-poor settings, where such treatment was started recently, there is inadequate information about impact of ART on the survival of patients especially in children. Hence, this study aimed to modelling the overall survival (OS) after HAART in HIV/AIDS pediatrics by Applying Accelerated Failure Time (AFT), Cox Proportional Hazard (PH) and Cox with time varying Coefficient models. Methods: Institution based retrospective follow up study was carried out among HIV positive pediatrics from September 2005 to September 2013 at JUSH. Out of a population of HIV positive pediatrics who were taking antiretroviral therapy in the hospital during that period, data on 218 patients are included in this study. The study subjects were pediatrics in the age less than 15 years. Non-parametric methods, Kaplan-Meier method and log-rank tests, were employed to compare the survival between the different categories of the explanatory variables. Semi-parametric methods, Cox PH models with time-dependent covariates and parametric methods, parametric PH model and AFT model was applied with the objective of identifying potential predictors of mortality. Results: After initiation of the antiretroviral treatment, HIV positive pediatrics lived for an average of 40.14 months; the median survival time was found to be 38 months. The baseline functional status, CD4 counts, WHO clinical stage, TB/HIV co-infected and opportunistic infections significantly influence the survival of the patients. Conclusion: Although the proportional hazard assumption not holds for Cox PH model, the weibull parametric model fitted the data well and can be taken as an alternative for Cox PH model.