Evaluation of Different Tests for the Serodiagnosis of Tuberculosis and the Use of Likelihood Ratios in Serology

Abstract

A serologic test for the diagnosis of tuberculosis was evaluated in 91 newly diagnosed tuberculosis (TB) patients, of whom 15 were HIV positive, in 17 TB patients during treatment, and in 220 control subjects (including individuals from endemic areas and patients with sarcoidosis or Crohn's disease). Purified proteins of M. tuberculosis with molecular weights of 10,000, 16,000, 24,000, 30,000, 38,000, and 70,000 were tested by ELISA. In addition, monoclonal antibody TB72 was tested by competition ELISA. The cutoff values were set at the mean plus three standard deviations of the values obtained in 100 healthy Dutch army recruits. Only the ELISA with the 10,000, 16,000, and 24,000 antigens and the TB72 assay discriminated between patients with TB who were not HIV positive and control subjects. Specificity varied from 95 to 98% and sensitivity from 29 to 51% with the different antigens. Combination of the test results of the ELISA with the 16,000 antigen and the TB72 assay had a sensitivity of 65% (95% confidence interval, 53 to 75%) and a specificity of 96% (95% confidence interval, 92 to 98%). The assay was useful for the diagnosis of both pulmonary and extrapulmonary TB. Optimal use of the serologic assay could be obtained when likelihood ratios for each test value are calculated instead of using the test dichotomized (positive or negative). High posttest probabilities indicate the presence of TB; low posttest probabilities do not exclude the disease and should lead to additional investigations.