Decision analyses should be helpful in selecting indications for isoniazid prophylaxis. In reality, however, they are not helpful because they use widely different assumptions that lead to vastly different conclusions. This is demonstrated by the contrasting analyses by Salpeter and colleagues [1], which concluded that isoniazid given to all tuberculin rectors would result in significant savings and large health benefits, and Tsevat and colleagues [2], which concluded that isoniazid reduces life expectancy.

Salpeter and associates assumed that the annual rate of the development of tuberculosis would be a constant 0.0006 cases per year, starting with the 15th year after the results of a tuberculin test were known. In contrast, Tsevat and coworkers assumed that the rates would decrease with the passage of time, an assumption reinforced by Chiba's study [3]. The latter showed a threefold decrease in rates between the 20th and 30th years.

Salpeter and coworkers assumed lifetime protection of 85% after 6 months of isoniazid therapy, whereas the study they cited showed only 65% protection with 6 months of isoniazid therapy for the 5 years that were studied [4]. Furthermore, the rates in the placebo group approached the rates in the isoniazid group in the fifth year, suggesting minimal protection thereafter.

The fatal isoniazid-related hepatitis rate of 0.00001 to 0.00002 deaths per isoniazid recipient assumed by Salpeter and colleagues is probably low because of serious underreporting of isoniazid-related deaths. Salpeter and coworkers used age-specific tuberculosis case-fatality rates that were higher and more realistic than the rates used by Tsevat and colleagues but did not mention the health concern factor. The latter probably reduces the tuberculosis case-fatality rate and cost of treating tuberculosis in persons concerned enough to take self-administered preventive treatment [5].

If decision analyses for isoniazid prophylaxis are to be useful, they must be repeated after the assumptions are refined by careful study and discussion among knowledgeable parties.