A double-blind placebo-controlled trial of antituberculosis chemoprophylaxis was undertaken in silicotic subjects in Hong Kong where there is a high prevalence of both silicosis and tuberculosis. During 1981 to 1987, 679 Chinese men with silicosis, with no history of previous antituberculosis chemotherapy and no evidence of active tuberculosis, were admitted to the trial and have been studied for between 2 and 5 yr. They were allocated at random to four series—rifampin for 12 wk (R3), isoniazid and rifampin for 12 wk (HR3), isoniazid alone for 24 wk (H6), or placebo (PI)—in a double-blind design with matching placebos for isoniazid and rifampin as appropriate. Active pulmonary tuberculosis developed more frequently during the 5 yr in the placebo series than in the three chemoprophylaxis series (p < 0.01, log-rank test), but there were no significant differences between the chemoprophylaxis series. The estimated proportions of patients with active pulmonary disease in the placebo series were 9% at 2 yr, 15% at 3 yr, 20% at 4 yr, and 27% at 5 yr. In contrast, in the three chemoprophylaxis series combined they were 5, 8, 10, and 13%, respectively. Thus, although chemoprophylaxis halved the proportion of patients in whom tuberculosis developed, this proportion was still substantial. There was no evidence that chemoprophylaxis led to the selection of drug-resistant strains of bacilli. Adverse effects were reported with a similar frequency in all four series, suggesting that few were drug related. During the first 12 wk, hepatic toxicity was reported in 8 (1%) patients (3 HR3, 3 H6, and 2 PI), but only 1 (H6) had symptomatic hepatitis. The serum alanine aminotransferase concentrations during chemoprophylaxis were higher in the HR3 and H6 series than in the R3 series (p < 0.001); there was no significant difference between the R3 and PI series. In conclusion, more effective antituberculosis chemoprophylaxis regimens for silicotic subjects are needed; rifampin on its own probably carries a very low risk of hepatic toxicity.