Progression to AIDS and predictors of AIDS in seroprevalent and seroincident cohorts of homosexual men

Abstract

As part of an ongoing prospective study of seropositive homosexual men in Vancouver, Canada, a seroprevalent cohort of 246 subjects (i.e. duration of infection unknown) and a seroincident cohort of 102 subjects (i.e. duration of infection known) were followed a median of 63 and 45 months, respectively. Follow-up with validation utilizing record linkage with the Canadian Federal Centre for AIDS registry revealed 58 and nine cases of AIDS in the seroprevalent and seroincident cohorts, respectively, through July 1988. These data yield product limit estimates of the cumulative progression rates to AIDS at 60 months of 23.0% for the seroprevalent cohort, 13.0% for the seroincident cohort, and 21.0% for the combined groups. Univariate analyses revealed the following to be statistically and clinically significant predictors of AIDS progression: low CD4 counts, low CD4/CD8 ratios, elevated immune complexes, elevated immunoglobulin G (IgG) and immunoglobulin A (IgA) levels, and low platelet counts. Cox regression revealed that elevated IgA levels, low CD4 counts, elevated immune complexes, two or more symptoms, and more than 20 male sexual partners in high-risk areas in the 5 years prior to enrolment were independent predictors of progression to AIDS over the subsequent 5 years. A multivariate risk function based on the latter five variables delineated low-, medium- and high-risk groups whose 5-year progression rates to AIDS were 6.7, 15.6 and 64.4%, respectively. The high-risk group contained 75% of all subjects who progressed to AIDS. Only 6% of the high-risk group would have qualified for zidovudine therapy under current guidelines at the beginning of the observation period. A simplified, more widely applicable function based only on IgA, immune complexes, lymphocyte counts, and symptoms delineated groups whose 5-year progression rates to AIDS were 6.9, 18.8 and 56.4%, respectively. Our data suggest that such models may have much greater predictive value than simplistic rules based on CD4 count alone. Further development and testing of such prediction rules should be encouraged so that we may better delineate those infected persons most likely to benefit from interventions at an early stage in their infection.