Proper care of persons infected with isoniazid-resistant tubercle bacilli is controversial because there are few data on the risks and benefits of the preventive treatment alternatives. Decision analysis was used to facilitate a comparison of outcomes using different strategies and the Delphi technique to obtain estimates of relevant probabilities. For all probabilities of isoniazid resistance, the observation and no-drug alternative is unsatisfactory because it sould result in a twofold to sevenfold greater number of tuberculosis cases than any of the drug regimens. With a low probability of isoniazid resistance, isoniazid is the preventive treatment of least cost and proved efficacy. As the probability of isoniazid resistance increases, more cases are prevented by rifampin-containing regimens, but at added cost. These findings can be used to formulate appropriate preventive treatment recommendations.