Using serial observations to identify predictors of progression to aids in the Toronto Sexual Contact Study

The Toronto Sexual Contact Study comprises a cohort of 249 male sexual contacts of men with HIV disease which has been followed every 3 months for almost 5 years. On enrolment 143 were seropositive and 16 seroconverted during the follow-up period. By 31 December 1989,41 of the 159 seropositive cohort members had developed AIDS. Using Cox relative risk regression models, we investigated the association of a number of laboratory and clinical variables and progression to AIDS. Fixed covariate models examined laboratory variables from the enrolment visit of cohort members, with time calculated from this date. In models assessing time dependent covariates, time was calculated from the estimated date of HIV infection. In the univariate models of either fixed or time dependent covariates, many variables were significantly associated with risk of progression to AIDS (T4 cell count, T4/T8 ratio, blastogenic responses to phytohemagglutinin, concanavalin A, and pokeweed mitogen, serum IgA, appearance of p24 antigen, and the development of oral hairy leukoplakia, thrush, or herpes zoster). Appearance of persistent generalized lymphadenopathy was not associated with increased risk of progression. In the multivariate model which evaluated fixed laboratory covariates, T4/T8 ratio, IgA level, and PHA response at enrolment were significantly associated with elevated risk. In order to evaluate the impact of the serial observations of immune function variables, these variables were lagged one year in a model with other time dependent covariates; the following point estimates of relative risk of progression to AIDS were observed: for a unit decline in T4/T8 ratio, RR = 60.8 ( p < 0.0001); for a 10,000 cpm decline in PHA response, RR = 1.09, ( p = 0.07); for a 100 μg/l increase in IgA, RR = 1.33, ( p = 0.02); for the development of oral hairy leukoplakia, thrush, or herpes zoster, RR = 2.69, ( p = 0.02); and treatment with zidovudine, RR = 0.06, ( p = 0.01).