Statistical methods for the analysis of HIV-1 core polypeptide antigen data in clinical studies

Levels of HIV-1 core polypeptide were assessed in serum and in lymphocyte cultures obtained from ARC and AIDS patients enrolled in a prospectively randomized, placebo-controlled study of zidovudine (AZT). Because these data have special features uncharacteristic of most laboratory data, a comprehensive account of statistical methods appropriate for their analysis is contained in this paper. Standard methods are described for the analysis of HIV-1 antigen in serum collected repeatedly over time in the same individual. For the analysis of lymphocyte culture data, more sophisticated statistical techniques based on nonparametric survival analysis methods are proposed. Using microcomputer software available upon request and developed to implement this statistical procedure, a significant decline in lymphocyte HIV-1 virus expression was noted between pretreatment and 3 months after the initiation of therapy among AZT-treated patients (p = 0.0017) that was not seen in placebo-treated patients (p = 0.25). Statistically significant between-group differences were also noted in the change from baseline at 3 months in HIV-1 antigen serum data (p = 0.040). We conclude that HIV-1 core polypeptide is an important measure of the antiretroviral activity of AZT and that the demonstrated clinical efficacy of AZT relative to placebo parallels its antiretroviral effect.