Mice that had been inoculated intravenously with 6.30 log10 Mycobacterium tuberculosis H37Rv 14 days earlier were administered one of three combinations of drugs, i.e., isoniazid (INH)-rifampin (RMP)-pyrazinamide (PZA), INH-RMP, and RMP-PZA, during an initial 2-month period to mimic the initial phase of chemotherapy for human tuberculosis and during a later 4-month period to mimic the continuation phase of chemotherapy. At the end of the initial phase, all three combined regimens were found to have been highly effective in terms of the number of CFUs in the spleens of infected mice. The bactericidal activities of INH-RMP-PZA and INH-RMP were similar, whereas that of RMP-PZA was significantly greater. The spleens of all of the mice that had been treated initially with INH-RMP-PZA were culture negative by the end of 6 months of treatment, regardless of the regimen employed during the continuation phase. However, after an additional period of 6 months without treatment, the proportion of spleen culture positivity, or relapse rate, was significantly smaller in the subgroup treated with RMP-PZA during the continuation phase than in the subgroups treated with INH-RMP-PZA or INH-RMP; the relapse rate did not differ significantly between the latter two subgroups. These results suggest that antagonism occurs between INH and the combination RMP-PZA during both the initial and continuation phases of chemotherapy, compromising the benefit conferred by the addition of PZA to the combined regimen.The preliminary pharmacokinetic analysis suggested that the pharmacological interaction between INH and RMP was very likely to be involved in the mechanism of antagonism, as concomitant treatment with INH had significantly reduced the peak serum level and the area under the serum concentration-time curve of RMP in mice.