Evaluation of Amplicor MTB test as adjunct to smears and culture for direct detection of Mycobacterium tuberculosis in the French Caribbean

A total of 784 specimens collected from 370 individuals between January and August 1995 were analyzed by using the Amplicor Mycobacterium tuberculosis test (Roche Diagnostic Systems, Basel, Switzerland), a PCR-based test for the direct detection of organisms of the M. tuberculosis complex. The PCR results were compared with standard bacteriological data, including those obtained by acid-fast microscopy, culture, and biochemical identification as well as final clinical diagnosis for each patient, Several parallel controls were used: the kit DNA positive control, 10(3) CFU of M. tuberculosis, and three negative controls for each independent assay, No false-positive PCR results were obtained, and overall, M. tuberculosis was detected in 20 of 370 individuals screened. Five additional patients during the same time were found to be infected with mycobacteria other than tubercle bacilli; their specimens gave positive smear and/or culture test results, but Amplicor tests were always negative. The sensitivity, specificity, positive predictive value, and negative predictive value for the Amplicor MTB test compared with culture per specimen were 76.7, 97.7, 66.0, and 98.6%, respectively. For resolved cases, these values were, respectively, 69.4, 100, 100, and 96.8%; however, the sensitivity and negative predictive value increased to 90.9 and 99.2%, respectively, if PCR-negative nonrespiratory specimens (gastric washings) were not considered. When only specimens from proven tuberculosis patients were considered (n = 114) and the sum of PCR-positive and/or culture-positive samples from proven tuberculosis patients was considered the total number of positive samples, PCR had a sensitivity of 83.3% compared with 71.6% for culture. Results per patient (about three samples each) yielded 100% sensitivity and 100% specificity. We conclude that the Amplicor MTB test is highly specific and rapid for routine use in a clinical laboratory. However, in order to obtain a high degree of sensitivity, it should be run as an adjunct to smears and culture with at least three samples for each patient, and a single-sample PCR-negative result must be considered carefully because of potential false-negatives.