The hypothesis that healthy individuals with latent (sub-clinical) infection with M tuberculosis show the same distribution as patients with manifest tuberculosis (previous slide) cannot usually be tested because of the disturbing influence of cross-reactions resulting from infection with environmental mycobacteria. To test the hypothesis, an environment with (virtual) absence of environmental myucobacteria must be chosen which is a difficult undertaking given their ubiquitous presence in our environment. Seemingly, in the dry and hot climate of Djibouti, such an environment has been identified.

A tuberculin skin test was given to patients with sputum smear-positive tuberculosis and all patients with a tuberculin skin test induration of sero millimeters were excluded to obtain the denominator. Similarly, healthy children were given the same dose of tuberculin (2 TU of PPD RT23) and all children with zero millimeters induration were excluded to obtain the denominator. The fraction of reactors within each 1-mm group were then plotted and are shown to be virtually identical for both patients with tuberculosis and children with assumed latent infection with M tuberculosis. The distributions are symmetrical with a mean of about 19 millimeters and the same standard deviation, and the distribution resembles those belonging to the normal family.

This observation would give credibility to the technique called "mirror" image that uses as an estimate for the prevalence of infection with M tuberculosis the number with an induration greater than (a clearly discernible maximum) mode, multiplies it by two and adds the number reacting at the mode to obtain the numerator. This approximation would be quite appropriate to correct for the sensitivity at the mode, but is sensitive to the precise location of the mode. Furthermore, it neglects the possible influence of the lack of specificity at the mode as the distribution of reactions attributable to infection with environmental mycobacteria might be log-normally distributed with a long tail.