Clinical data on 7840 HIV-positive patients, representing 43 745 patient-years of follow-up, has been collected. All patients with ARL since 1986 (n = 150) were assessed at presentation for prognostic factors and outcomes recorded. Comparisons are made between cases in the pre-HAART era (1988-1995), and the HAART era (1996-1999). Statistical models are used to calculate the incidence of ARL and factors predicting its development. The incidence of ARL has not changed over time (3 to 7 of 1000 patients per year, P = .933), but contributes to a greater percentage of first AIDS-defining illnesses (ADI) in the HAART era (P < or = .0001). Older age, nadir CD4 count, and no prior HAART use, predict the development of ARL. There has been no change in stage at presentation, presence of B symptoms, performance status, or marrow involvement between the 2 time cohorts or between patients with or without prior HAART exposure. Similarly, there is no difference in survival duration between the pre-HAART and HAART era (log rank P = .15) or specifically in patients treated with HAART before ARL diagnosis (log rank P = .12). The use of HAART has not yet been shown to influence the incidence or survival of ARL. However, because nadir CD4 count and use of HAART are independent predictors of ARL development, this may translate into a future fall in new cases. (Blood. 2000;96:2730-2734)