To evaluate the long-term safety and efficacy of the combination of hydroxychloroquine, hydroxyurea and didanosine. We recruited antiretroviral-naive patients with viral loads less than 100 000 HIV-1 RNA copies/mL and CD4 counts greater than 150 cells/microL. All patients received hydroxychloroquine (200 mg), hydroxyurea (500 mg) and didanosine (125-200 mg) twice daily. Clinical and laboratory safety assessments and measurements of viral load and CD4 count were made at regular intervals, and genotypic resistance testing was performed on samples with detectable viral load at 48, 96 and 144 weeks. Fourteen of the 17 patients who commenced therapy remained on treatment at 144 weeks. Treatment was well tolerated but caused neutropenia, usually mild and transient, in 12 patients (71%). Mean viral load was reduced by 1.6 log(10) copies/mL below baseline (P<0.001), eight patients (47%) had undetectable viral load (<400 copies/mL), and two patients (12%) had detectable viral load but no detectable resistance mutations at week 144. Four patients (24%) had detectable viral load together with major resistance mutations (three with both 74 V and 184 V, and one with both 62 V and 65R) at week 144, but still had viral load suppression below baseline. Mean CD4 count was increased by 106 cells/microL above baseline (P=0.07) at week 144. This novel and well-tolerated combination controls viral replication during long-term follow up, with development of few resistance mutations. With careful monitoring it may be a useful strategy for delaying highly active antiretroviral therapy (HAART) and associated toxicity in selected patients with low initial viral loads.