Setting: Karonga District, northern Malawi, 1980-1989.
Objective: To measure and interpret incidence and prevalence of tuberculin sensitivity.
Design: Tuberculin testing carried out within two total population surveys. Tuberculin 'positivity' and 'conversion' were defined using criteria recommended by the Tuberculosis Surveillance Research Unit and the American Thoracic Society, respectively.
Results: Data on 64,225 tests were available for analysis, including paired results on 6991 individuals tested in both surveys. Frequency distributions of induration varied by age, sex, BCG scar status and zone within the district. The prevalence of 'positivity' was similar in males and females until age 15, then higher among males, and was consistently higher among individuals with than among those without a BCG scar. Tuberculin 'conversion' rates estimated from cross-sectional data ranged from 0.34 to 1.15 per cent per annum. Conversion rates derived from longitudinal data were found to increase linearly with age, and the reversion rates declined rapidly with age among younger individuals. Such trends, which have been reported in other populations, are shown here to arise as an artefact of test instability. Prospective follow-up of observed converters showed greatly increased risks of tuberculosis, in particular during the two years following the second ('converted') test (relative risk > 10).
Conclusion: Estimation of a convincing 'true' annual risk of infection from tuberculin survey data is not possible from either cross-sectional or longitudinal data, due to misclassifications and the instability of delayed type hypersensitivity over time. An apparent increase in infection risk with age can arise as an artefact of test instability. BCG-induced tuberculin sensitivity declines rapidly in this population in most individuals. It is necessary to consider tuberculin reversions, whether real or apparent, when interpreting tuberculin data on individuals or populations.