Diagnostic accuracy of the Xpert® MTB/RIF assay for extra-pulmonary tuberculosis: a meta-analysis

Abstract

Background: Xpert(®) MTB/RIF is a commercially available nucleic acid amplification test developed for the diagnosis of pulmonary tuberculosis (PTB).

Objective: To determine the diagnostic accuracy of Xpert for the detection of extra-pulmonary tuberculosis (EPTB).

Methods: We searched MEDLINE, EMBASE and Global Health databases from January 2010 to 15 August 2014 for studies of diagnostic performance in which Xpert was examined against culture for patients with clinically suspected EPTB. Bivariate random effects models were used to provide pooled estimates of diagnostic accuracy.

Results: Thirty-six studies were identified, with a pooled sensitivity and specificity of respectively 77% (95%CI 66-85) and 97% (95%CI 94-98). Substantial variations existed between study estimates of sensitivity (I(2) = 99%) and specificity (I(2) = 96%). Among site-specific estimates for lymph, pleural fluid, cerebrospinal fluid, gastro-intestinal and urinary samples, the pooled sensitivity was lower in pleural fluid (37%, 95%CI 26-50, meta-regression P < 0.001) and higher in lymph node samples (87%, 95%CI 75-95, meta-regression P = 0.03).

Conclusion: Xpert has high specificity but limited sensitivity for the detection of EPTB. Although positive Xpert test results may be useful in rapidly identifying EPTB cases, negative test results provide less certainty for ruling out disease.