In an attempt to find a vaccine that gives greater and more consistent protection against leprosy than BCG vaccine, we compared BCG with and without killed Mycobacterium leprae in Venezuela. Close contacts of prevalent leprosy cases were selected as the trial population since they are at greatest risk of leprosy. Since 1983, 29,113 contacts have been randomly allocated vaccination with BCG alone or BCG plus 6 x 10(8) irradiated, autoclaved M leprae purified from the tissues of infected armadillos. We excluded contacts with signs of leprosy at screening and a proportion of those whose skin-test responses to M leprae soluble antigen (MLSA) were 10 mm or more (positive reactions). By July, 1991, 59 postvaccination cases of leprosy had been confirmed in 150,026 person-years of follow-up through annual clinical examinations of the trial population (31 BCG, 28 BCG/M leprae). In the subgroup for which we thought an effect of vaccination was most likely (onset more than a year after vaccination, negative MLSA skin-test response before vaccination), leprosy developed in 11 BCG recipients and 9 BCG/M leprae recipients; there were 18% fewer cases (upper 95% confidence limit [CL] 70%) in the BCG/M leprae than in the BCG alone group. For all cases with onset more than a year after vaccination irrespective of MLSA reaction the relative efficacy was 0% (upper 95% CL 54%; 15 cases in each vaccine group). Retrospective analysis of data on the number of BCG scars found on each contact screened suggested that BCG alone confers substantial protection against leprosy (vaccine efficacy 56%, 95% CL 27-74%) and there was a suggestion that several doses of BCG offered additional protection. There is no evidence in the first 5 years of follow-up of this trial that BCG plus M leprae offers substantially better protection against leprosy than does BCG alone, but the confidence interval on the relative efficacy estimate is wide.