The prevention of relapse in quiescent Crohn's disease remains a major therapeutic challenge. The present study is a double blind placebo (P)-controlled, randomized, multicentre cooperative trial designed to test the effectiveness of levamisole (L) in the prophylaxis against flare up in patients with quiescent Crohn's disease. The trial included 2 successive phases: a) phase I:167 patients with inactive disease (but who had not had previous resection of all diseased tissue) were randomly and double blindly assigned to receive either L (150 mg orally once weekly) or P; patients were randomized in 2 strata: those having experienced a recent flare up (within the 3 months preceding their entry into the trial: red strata) versus others (blue strata). Patients were followed up at 3 monthly intervals during 2 years. Initially there was no significant difference between L and P groups as regard to age, sex ratio, duration of disease at the time of randomization, incidence of prior intestinal resection, Crohn's disease topography, clinical activity; biological activity was slightly but significantly higher (P less than 0.05) in the P group. Twelve patients were withdrawn from analysis (lost to follow-up: 2; inadequate respect of the protocol: 10), leaving 155 patients (78 L, 77 P) who completed the study. L did not significantly influence any of the following parameters: incidence of attacks (L: 37 p. 100; P: 35 p. 100), lag time between the entry into the trial and the occurrence of the attack (L: 32.7 +/- 5.2; P: 41.8 +/- 5.8; m +/- SEM; weeks), curves of maintenance in remission (Kaplan-Meier method), outcome rank, severity of attacks. Attempts to analyse separately certain subgroups--subjects with purely colonic (+/- anal) disease, subjects with small bowel localization (+/- anus), patients of the red or blue strata--did not show any statistical difference between L and P. Thirteen patients left the trial for minor intolerance (10 L, 3 P). b) Phase II lasted one further year and involved the patients still in remission and in the trial at the end of phase I (n = 57). Those who had received L during phase I were randomized between continuance of L (L leads to L) vs. a change to P (L leads to P); those who had been on P during phase I were randomized between continuance of P vs. a switch to L.(ABSTRACT TRUNCATED AT 400 WORDS)