Abstract
South Africa implemented Xpert MTB/RIF as the initial diagnostic test for pulmonary tuberculosis (TB). Xpert MTB/RIF's accuracy for diagnosing extrapulmonary tuberculosis (EPTB) was investigated. EPTB specimens (n = 7,916) from hospitalized patients received over a 6-month period at a high-throughput TB referral laboratory in Johannesburg were investigated. Large-volume specimens were centrifuged, tissue biopsy specimens homogenized, and all specimens checked for growth of contaminating bacteria on blood agar. Contaminated samples received NALC-NaOH (N-acetyl-l-cysteine-sodium hydroxide) decontamination prior to liquid culture. Residual specimens (volumes > 1 ml) after inoculation of culture (n = 1,175) were tested using the Xpert MTB/RIF sputum protocol. Using culture as the reference, Xpert MTB/RIF's overall sensitivity was 59% (95% confidence interval [95% CI], 53% to 65%) and specificity was 92% (CI, 90% to 94%), with the highest sensitivities of 91% (95% CI, 78% to 97%) for pus, 80% (95% CI, 56% to 94%) for lymph node aspirates, and 51% (95% CI, 44% to 58%) for fluids (ascitic, 59%; pleural, 47%). A difference in sensitivities was noticed between specimens classified as having a thick (87% [95% CI, 76% to 94%]) versus clear (watery) (48% [95% CI, 36% to 61%]) appearance. This was unchanged with traces of blood (52% [95% CI, 44% to 60%]) or precentrifugation (57% [95% CI, 28% to 82%]) among clear specimens. Xpert MTB/RIF generated an additional 124 specimen results that were contaminated by Mycobacterial Growth Indicator Tubes (MGIT; 10.5%) and diagnosed rifampin (RIF) resistance earlier (9.6% [25/260]). Xpert MTB/RIF's performance on EPTB specimens provides very promising results and should be considered for incorporation into national TB guidelines. Xpert MTB/RIF is less affected by contaminating bacteria and reduces laboratory labor and diagnostic delay compared to traditional methods.