PURPOSE: Recent trials suggest serious toxicity in HIV-associated non-Hodgkin's lymphoma (NHL) with rituximab (R) and chemotherapy (CT), offsetting the benefit of rituximab. METHOD: We retrospectively reviewed experience with CHOP-R vs. CT in 40 patients with HIV-associated diffuse large B-cell lymphoma (DLBCL) diagnosed between December 1992 and February 2006, all of whom were treated with curative intent. RESULTS: In a univariate analysis, International Prognostic Index (IPI) score, prior AIDS, HAART, and rituximab were significant for overall survival (OS). In a multivariate analysis, IPI 0-1 (p < .02), no prior AIDS (p < .0002), and receiving CHOP-R (p < .01) were significant for improved OS, and HAART use (p < .09) retained a trend for improved OS. The hazard ratio (HR) for patients with high IPI receiving CHOP-R was 0.3 (95% CI 0.1-0.8). Patients without prior AIDS receiving CHOP-R had an HR of 0.5 (95% CI 0.1-1.7). The OS at 30 months in patients not receiving HAART was 0%. With HAART, OS was 33% for CT and 86% for CHOP-R; HR for CHOP-R was 0.4 (95% CI 0.1-1.2). Toxic deaths were 3 (33%) for CHOP-R and 6 (25%) for CT (p = ns); all toxic deaths with CHOP-R were in patients not receiving HAART. Rituximab-treated patients had a lower death rate from lymphoma (CHOP-R, 2 [16%] vs. CT, 15 [63%]; p < .04), and overall mortality (CHOP-R, 5 [42%] vs. CT, 21 [88%]; p < .01). CONCLUSION: These retrospective data suggest that fatal toxicity of rituximab in HIV-NHL is not increased provided HAART is used, that the addition of rituximab to CT improved outcome, and that further prospective trials investigating this issue are warranted.