Random variation in tuberculin sensitivity in schoolchildren. Serial skin testing before and after preventive treatment for tuberculosis.

Schoolchildren were Mantoux-tested with 2 TU freeze-dried PPD RT23, and the strong reactors with indurations of 14.0 mm or more were selected for treatment with one of three different fixed drug combinations containing isoniazid or with placebo for 2 to 6 months. The initial tuberculin test was repeated after 8, 14, and 27 months. Of the 8,934 black schoolchildren initially tested, 5,165 did not react to the skin test, 2,898 had indurations up to 14.0 mm, and 871 reacted strongly. Of these strong reactors, 808 were allocated to four preventive treatment groups. On completion of treatment, the mean tuberculin reaction for all groups was significantly decreased. Because the placebo group showed changes similar to those seen in the other treatment groups, the tuberculin skin test is probably not suitable for monitoring the success of preventive therapy. Differences between skin test results before and after treatment when retesting only strong reactors are caused by a combination of effects that are difficult to distinguish. Assuming random variation in tuberculin sensitivity, the decrease can be explained as a combined effect of regression to the mean and some boosting. The increased reaction sizes in the subsequent Mantoux tests are explained by the booster phenomenon and possibly by reinfection. When using a cutting point for deriving a positive reactor, the chance of being selected for preventive treatment may depend primarily on the moment in time when the test is done. Thus, all reactors with no recent BCG vaccination should equally be considered for treatment.