The Isoniazid Debate-Reply

In Reply. —  The letters by Tsevat et al and Moulding reaffirm a central point in my article.¹ Mathematical modeling works, but the results are critically dependent on the hard data incorporated as probabilities for the occurrence of chance outcomes. When good hard data are not available, investigators have to make choices about which data will be used. Obviously, different choices will lead to different results. I do not intend to quibble over why I chose to use certain data in my analysis. Instead, I want to re-pose a question already raised in my article. The mathematical models have clearly identified the critical variable in the isoniazid chemoprophylaxis debate. Why has the appropriate federal agency not published unimpeachable hard data on the isoniazid hepatitis case—fatality rate for relevant subgroups, eg, those under 35 years of age, pregnant women, women, etc? Only with these data available can we determine for