Twenty-five patients with good-prognosis polyarteritis nodosa or Churg-Strauss syndrome entered a prospective, randomized, multicentre study comparing two treatments: either oral corticosteroids and oral cyclophosphamide (CY; 2 mg/kg/day) for 1 yr (group A), or oral corticosteroids and monthly i.v. CY pulses (0.6 g/m2) (group B) for 1 yr. The objective was to determine the optimal CY regimen. Judgement criteria were the efficacy of the treatment in controlling the disease and the development of side-effects. Among the 25 patients who could be analysed, complete recovery was achieved with the experimental treatment in 9/12 patients in group A and 10/13 patients in group B. Two patients in each group relapsed after the end of therapy and were well controlled by corticosteroids or other drugs. One failure occurred in each group. The mean follow-up was 60.8 +/- 14.5 months after the beginning of the treatment. Side-effects associated with the administration of CY and steroids were noted 27 times in group A vs 14 times in group B (not significant). The oldest patient in these series (group B) died of pneumonia. No superiority in terms of efficacy could be established between the two regimens; however, the number of patients included was too small to conclude definitively. Toxic side-effects were significantly more frequent in women (P < 0.02). The high number of adverse effects leads us to recommend pulse over oral CY and an overall lowering of the doses of immunosuppression.