On the Estimation of Expected Survival Time of AIDS Patients Undergoing Antiretroviral Therapy Using Censored Generalized Poisson Regression Model.

Objective: The progression of HIV infection depends not only on the combination of drugs and adherence to the Anti Retroviral therapy (ART), but also on various other factors. The survival of an HIV patient depends on interaction of these factors with the therapy. Material and Methods: The parametric approach with and without covariates have been used to analyze the survival data of HIV/AIDS patients. When covariates are not considered, survival distributions are fitted and best one chosen. The inclusion of covariates is analyzed using Censored Generalized Poisson Regression Model (CGPR). Results: For a retrospective right censored data of 1689 HIV/AIDS patients undergoing ART at the Ram Manohar Lohia Hospital, New Delhi, India, appropriate survival distributions in the absence of covariates are fitted. Using the Akaike Information Criterion, Gamma distribution is found to be the distribution of best fit. Assuming Gamma survival distribution with right censored survival time data, the mean survival time of AIDS patients using the method of maximum likelihood is found to be 11.14 years. Also, CGPR model for estimating the survival time while accounting for the impact of cofactors is used for the first time on the AIDS dataset. Conclusion: The advantage of using a CGPR model over the gamma right censored distribution is that the mean survival time of the AIDS patient undergoing ART can be estimated in the presence of significant prognostic factors. Using this model, age, gender, smoking status, alcoholism, WHO staging, improvement in CD4 count, opportunistic infections, number of visits and weight at initiation of ART are identified as significant prognostic factors effecting the survival time. On averaging the predicted survival times based on CGPR model, the mean survival time of AIDS patients on ART is found to be 12.12 years. [ABSTRACT FROM AUTHOR]