At 30 years into the HIV infection epidemic, the optimal antiretroviral (ARV) regimen for infected patients with cancer remains unknown. We therefore sought to retrospectively study different ARV regimens used in this population. Data from HIV-infected patients seen at The University of Texas MD Anderson Cancer Center in Houston, Texas, USA, from 2001 to 2012 were reviewed. Patients received nucleoside reverse transcriptase inhibitors (NRTIs) plus protease inhibitors (PIs), non-NRTIs (NNRTIs), integrase strand-transfer inhibitors (INSTIs), or combinations of these. A total of 154 patients were studied. Most patients were male (80%), white (51%) and had haematological malignancies (HMs) (58%). NRTIs were combined with PIs (37%), NNRTIs (32%), INSTIs (19%) or combinations of these (11%). INSTIs were the most commonly used in patients with HM and in those receiving high-dose steroids or topoisomerase inhibitors (p <0.05). Side-effects occurred in 35%, 14%, 3% and 6% of patients receiving PIs, NNRTIs, INSTIs and combinations, respectively (p 0.001). Grade 3-4 adverse events were uncommon. Multivariate logistic regression analysis demonstrated that INSTIs and NNRTIs were nine times (95% confidence interval (CI), 1.4-50.8) and 11 times (95% CI, 1.9-64.7) more likely to be effective at 6 months, respectively, than PIs. This is the largest reported analysis studying different ARV regimens in HIV-infected cancer patients. Combinations that included PIs were the least favourable. NNRTIs and INSTIs had comparable efficacy, but INSTIs appeared to be the better tolerated ARVs in patients with HM or those receiving various chemotherapeutic agents.