Joint model applied in a cohort study on HIV carriers coinfected with HBV and HCV

Abstract

Abstract The framework that permeates joint modeling has great appeal for biological interpretation, in addition to allowing the quantification of the contribution of the longitudinal biomarker in survival. This effect on the risk function is incorporated by the original variability of the observations made over time. In addition, this methodology provides more efficient estimates for the effect of the groups under a longitudinal aspect and with less bias in the survival model. A joint model is proposed in this research, adopting a mixed multivariate normal model for the longitudinal response and the relative risk model for survival. Our sample is composed of HIV carriers exposed to cART, characterized in groups: control (HIV), and co-infections with HBV and HCV. The joint distribution of the biomarker count, the square root of CD4, was modeled in relation to the start time of cART until the CD4:CD8 ratio = 0.9 was reached. The model selected, by testing the likelihood ratio, showed greater severity in HBV co-infections. The association was significant, indicating that the biomarker and risk of CD4:CD8 ratio = 0.9 should be analyzed together. This result corroborates with clinical evidence that points to this possible relationship and is supported by the residual analysis, which describes the good adequacy of the model.