OBJECTIVE AND METHODS: In 1998 we reported on a randomized comparison between stavudine plus didanosine plus placebo versus stavudine plus didanosine plus hydroxyurea (HU), in patients with a CD4 count of 200-500 x 10(6) cells/l. After 3 months, the HU group had a higher proportion of patients with viral load < 200 x 10 cells/l. At the end of the 3 months blinded period, patients in the placebo group had the option to add HU if their viral load remained > 200 x 10(6) cells/l. We report results after 24 months. RESULTS: Seventy-two patients were randomized to the HU arm, and a further 30 elected to add HU after 12 weeks. Twenty-four months after the start of the trial, only 25% of the 72 patients originally randomized to HU, and 20% of the 30 who added HU after week 12, were still taking it. The reasons for stopping HU were: lack of efficacy (45%), adverse events (37%) and patient or physician preference (18%). Side effects were more frequent in the didanosine/stavudine/HU group than in the didanosine/stavudine group: neuropathy (35 versus 15%, P< 0.02), fatigue (22 versus 7%, P< 0.01), and nausea or vomiting (26 versus 9%, P< 0.01). Of those who had discontinued HU, 73% were taking three drugs including a protease inhibitor. Patients who had started HU were compared with similar patients who had started protease inhibitors in the Swiss cohort. The probability of stopping HU was higher than the probability of stopping nelfinavir or indinavir, and similar to the probability of stopping ritonavir. CONCLUSION: HU increased the antiviral effect of stavudine plus didanosine. However, side effects were more frequent, and after 24 months the majority of patients had switched to protease inhibitor regimens.