Using likelihood ratios to estimate diagnostic accuracy of a novel multiplex nested PCR in extra-pulmonary tuberculosis

Abstract

A tertiary care centre in Mumbai with a referral bias towards treatment failures. To standardise and evaluate a novel single tube multiplex nested polymerase chain reaction (PCR) targeting insertion sequence (IS) 6110, mpb64, rrs and rpoB genes for rapid diagnosis of extra-pulmonary tuberculosis (EPTB). The PCR assay was evaluated among 489 consecutive consenting patients, and results were compared against a composite reference standard comprising smear microscopy, culture, clinical symptoms, radiological scan and histology. PCR assay reported a pooled sensitivity of 94.5% (242/256, 95%CI 91-97): 91.9% (125/136, 95%CI 86-96) for smear-negative composite reference standard (CRS) positive cases and 97.5% (117/120, 95%CI 93-99) for smear-positive CRS-positive cases. The PCR positivity rate increased from 91.7% (235/256, 95%CI 88-95) when presence of IS6110 was considered alone for reporting a test as positive to 94.5% (242/256, 95%CI 91-97) when used in combination with other three gene targets, with a specificity of 96.4% (212/220, 95%CI 93-98). A positive likelihood ratio of 26 (95%CI 13-51) and a negative likelihood ratio of 0.06 (95%CI 0.03-0.09) makes the test useful for ruling out and ruling in the disease. Culture should not be replaced by PCR as a gold standard; however, PCR can be used as a rapid, accurate tool in the diagnosis of EPTB.