We thank Dr. Moulding for his comments. We concluded that isoniazid prophylaxis modestly extends life expectancy for treated patients and would prevent 35 176 deaths if the entire population of low-risk tuberculin reactors older than 35 years of age received prophylaxis. Tsevat and colleagues' analysis [1], performed a decade ago, concluded that the use of isoniazid was a "close call" for an individual patient and decreased life expectancy by a few days per treated patient. One important difference between the two analyses is that we used estimates of the fatality rate with monitored isoniazid prophylaxis that were based on large pooled studies that were not available previously; in contrast, Tsevat and colleagues used data on unmonitored prophylaxis. Current evidence suggests that monitored prophylaxis decreases the mortality rate from isoniazid to within a range of 0.001% to 0.002%, a rate substantially lower than that used in previous analyses. Underreporting of prophylaxis-related deaths is a concern, as Dr. Moulding suggests, but our sensitivity analyses showed that the probability of isoniazid-induced death had to be 35 times higher than our base-case estimate for the no-prophylaxis strategy to be preferred.

Another difference between our analysis and those of other investigators is that we formally evaluated the effect of secondary transmission of tuberculosis on the benefits and costs of prophylaxis [2]. Our estimate of tuberculosis activation rates after 15 years of known tuberculin reactivity was based on the pooled results of many studies. We recently published a detailed analysis of the activation rates as part of a population model for tuberculosis transmission [2].

Studies of 12 months of isoniazid prophylaxis indicate a 93% reduction in the activation of tuberculosis in subgroups with good compliance. The study mentioned by Dr. Moulding indicates that a 6-month course of isoniazid is equivalent to a 12-month course for patients who have minimal changes on chest radiography [3], a characteristic of the patients in our analyses. Therefore, we used a conservative estimate of 85% efficacy for a completed 6-month course and an overall efficacy of 71% that accounts for incomplete compliance with prophylaxis.

To our knowledge, the "health concern factor" has never been studied for tuberculosis. In fact, the high mortality rate from tuberculosis in patients older than 65 years of age is usually seen in patients without other illnesses, such as alcoholism, injection drug abuse, or HIV infection, and who presented for treatment as soon as they began feeling ill [4]. We believe that these factors tip the balance in favor of isoniazid prophylaxis.