Sources of variation in studies of the early bactericidal activity of antituberculosis drugs

The early bactericidal activity (EBA) of antituberculosis drugs can be measured as the daily fall in cfu counts of Mycobacterium tuberculosis in sputum, usually during the first 2 days of treatment. In studies of low potency drugs, it is necessary to compare the treated group of patients with a group who receives no chemotherapy (Nil group). Over the past 10 years, five Nil groups of between five and 13 patients have been studied in Cape Town and two Nil groups in Hong Kong. There was a suggestion of an increase in variation within the Cape Town groups, as shown by a regression of variance size against study date (P = 0.06), which could not be attributed to technical causes. It might indicate increasing host resistance in the Western Cape epidemic of tuberculosis. Since the weighted mean of all Nil groups at Cape Town was 0.00036, very close to zero, it would seem safe to test means of treated groups against zero thus increasing precision and avoiding ethical problems in delaying treatment. In contrast to Nil groups, the variation found in five groups who received 300 mg isoniazid daily (INH 300) was uniform and homoscedastic, possibly because the additional variation was caused mainly by individual differences in plasma isoniazid concentrations and patient body weights. The mean EBA in the INH 300 series was 0.575 with 95% confidence limits of 0.515 and 0.636.