Current evidence on diagnostic accuracy of commercially based nucleic acid amplification tests for the diagnosis of pulmonary tuberculosis

Abstract

Even though commercial nucleic acid amplification tests (NAATs) have become the most frequently used molecular tests for laboratory diagnosis of pulmonary tuberculosis (TB), published studies report variable estimates of their diagnostic accuracy. We analysed the accuracy of commercial NAATs for the diagnosis of pulmonary TB in smear positive and smear negative respiratory samples using culture as a reference standard. English language studies reporting data sufficient for calculating sensitivity and specificity of commercial NAATs on smear positive and/or smear negative respiratory samples were included. Meta-regression was used to analyse associations with reference test quality, the prevalence of TB, sample and test type. Predictive values for different levels of pre-test probability were quantified using Bayes' approach. Sixty three journal articles published between 1995 and 2004 met the inclusion criteria. Pooled sensitivity and specificity were 0.96 and 0.85 among smear positive samples and 0.66 and 0.98 among smear negative samples. The number of culture media used as reference test, the inclusion of bronchial samples, and the TB prevalence were found to influence the reported accuracy. The test type had no effect on the diagnostic odds ratio but seemed to be correlated with sensitivity or specificity, probably via a threshold effect. Commercial NAATs can be confidently used to exclude TB in patients with smear positive samples in which environmental mycobacteria infection is suspected and to confirm TB in a proportion of smear negative cases. The methodological characteristics of primary studies have a considerable effect on the reported diagnostic accuracy.