The effects of maraviroc versus efavirenz in combination with zidovudine/abacavir on the CD4/CD8 ratio in treatment-naïve HIV-infected individuals.

For the principal objective, longitudinal changes in CD4 and CD8 counts and in the CD4/CD8 ratio will be assessed using linear mixed models with random intercepts. Age, gender, risk factors for HIV acquisition, year of ART initiation, baseline CD4+ T-cell count and CD8+ T-cell will be included in multivariate analyses as fixed-effects. Interaction terms were created to assess whether these changes over time differed significantly between treatment arms.


For the secondary objective, the outcome variable will be the time to the clinical event. The primary correlate of interest in the analysis of time to clinical event will be the time-updated binary indicator of CD4:CD8 normalization. Kaplan Meier methods for time-dependent covariates will be used to examine the associations of CD4:CD8 normalization with the incidences and rates of survival and selected clinical event. Proportional hazard models will be used to estimate the association of CD4:CD8 normalization with incidences and rates of survival and selected clinical events, after adjusting for other variables hypothesized to be associated with non-AIDS-related events, based on the background literature. When covariates are considered to be collinear (such as IDU and hepatitis C), the variable with the strongest association will be included in the multivariable model. The final model will be adjusted at least by the following variables: age, gender, risk factors for HIV acquisition, year of ART initiation, treatment arm, baseline CD4+ T-cell count and CD8+ T-cell count and time-updated HIV RNA.

For both objectives, sensitivity analyses will be conducted to examine the effects of 1) defining normalization of the CD4:CD8 ratio as >0.5, >0.8, >1 or >1.5, 2) excluding late presenters, defined as having an AIDS-defining events within 3 months after ART initiation and 3) including only patients reaching >500 CD4/uL during follow-up. Missing data will be handled using multiple imputation.