EVIDENCE IN FAVOUR OF A LOW DOSE FOR THE TUBERCULIN TEST.

In an attempt to solve the problem of whether a high or a low dose should be used for tuberculin testing a new method of statistical analysis is presented. It consists in tabulating the frequency of tuberculin "positive" children in households where the adults are tuberculin "positive" and tuberculin "negative" respectively; these frequencies are compiled for a wide range of arbitrarily chosen limits between "positive" and "negative" reactions. Values for these statistics have been obtained from data collected in Kenya; they strongly suggest that the reactors to the low-dose test are those infected with tubercle bacilli and the non-reactors to this test are the uninfected. However, a more detailed analysis is possible and this paper demonstrates that statistics of the type described may be used for testing the plausibility of a specific hypothesis about the patterns of tuberculin reactions among infected and uninfected persons. By testing a number of different hypotheses, the most plausible was found to be that the tuberculin sensitivities of infected and uninfected persons correspond to those of the reactors and the non-reactors to the low-dose test respectively. On the basis of the data presented it thus appears that the low-dose test is better suited than the high-dose test for distinguishing between infected and uninfected persons.