A survival analysis of a cohort of HIV-infected persons with hemophilia was undertaken to identify the prognostic significance of several hemophilia- and virus-related factors. Subjects were participants in the National Cancer Institute Multicenter Hemophilia Cohort Study. The study sample was restricted to subjects with hemophilia type A or B, of at least 10 years of age and consisted of 644 subjects with a median follow-up of 3.7 years Age at entry and age at HIV infection were correlated and five-year cumulative mortality risks were observed to be linearly associated with age. Several findings from this study have not been previously reported. Among the hemophilia-related variables examined, only antibodies to factor VIII were found to be prognostic (RR = 1.68, 95% CI = 0.96-2.94, adjusted for age and CD4%), among the elderly the age-inhibitor interaction was on an additive scale, and inhibitors were unrelated to hemorrhagic deaths. An indeterminate antibody reactivity to hepatitis C virus recombinant antigens was associated with a three-fold higher mortality rate (RR = 2.98, 95% = 1.77-5.02, adjusted for age and CD4%) was associated with low CD4% values, and was also associated with hepatic mortality among older subjects Results from this study also determined that moderate immune deficiency, defined by CD4 values between 14% and 29%, did not have prognostic relevance for all-cause mortality whereas lower immune states were significantly prognostic and time-dependent (RR = 6.62, 95% CI = 3.52-12.42 for low immune states prior to entry and RR = 4.76, 95% CI = 2.99-7.57 for low immune states occurring after entry). Strategies for modeling immune states were compared. A summary, single time-dependent covariate estimate resulted in an averaging of effect that overestimated or underestimated risk for subjects with shorter and longer durations of low immune states, respectively. A time-stratified approach resulted in more informative estimates Duration of HIV infection did not have prognostic significance and 16% of the cohort have survived at least 10 years with HIV infection Findings from a limited evaluation of antiretroviral treatment effects suggested that there was no survival advantage over three years from start of antiretroviral therapy when initiated within one year compared to initiation after longer durations of incident moderate or low immune states