BACKGROUND Pancreatitis is a known adverse effect of the nucleoside reverse transcriptase inhibitors, particularly didanosine. Hydroxyurea has been used to potentiate the antiviral efficacy of didanosine, but recently there has been concern that severe and even fatal pancreatitis may be more likely to occur when hydroxyurea is used in combination with didanosine. We investigated the incidence of pancreatitis in patients using nucleoside analogues with or without hydroxyurea. METHODS Data were obtained from patients followed longitudinally on the Johns Hopkins HIV Clinic. Incidence rates of pancreatitis were calculated for each antiretroviral regimen that included zidovudine, stavudine, didanosine (+ hydroxyurea), and didanosine + stavudine (+ hydroxyurea). Poisson regression was used to compare the relative rate of pancreatitis for each regimen adjusting for other covariates. RESULTS A total of 2613 patients received at least one of the nucleoside reverse transcriptase inhibitor-containing regimens. There were 33 cases of pancreatitis. The crude incidence rate of pancreatitis ranged from 0.18 cases per 100 person-years on therapy for zidovudine to 6.25 cases per 100 person-years for didanosine + hydroxyurea. Compared to didanosine alone, and adjusting for CD4 cell count and other variables, the relative risk (RR) of pancreatitis was 8.56 [95% confidence interval) CI, 1.85-35.59] for didanosine + hydroxyurea, and 2.35 (95% CI, 0.46-11.89) for didanosine + stavudine + hydroxyurea. For any use of hydroxyurea, the RR = 4.01 (95% CI, 1.02-15.89). Other risk factors for pancreatitis included a CD4 cell count < 200 x 106 cells/l, female sex, and a history of pancreatitis. CONCLUSIONS Our data show that the risk of pancreatitis is four-fold higher when hydroxyurea is used. The use of hydroxyurea with didanosine should probably be discouraged if other treatment options are available.