Abstract
The diagnosis of mycobacterial infections has been dramatically improved by the introduction of molecular methods aimed to reduce the time to diagnosis as compared with culture. The broad range pan-mycobacterial PCR can detect all the mycobacterial species directly from clinical specimens. We aimed to evaluate its usefulness and its clinical added value for the diagnosis of nontuberculous mycobacterial (NTM) infections. We performed a retrospective study (2003-2013) including 952 samples taken from 639 patients with clinical suspicion of NTM infection. The performance of smear microscopy, PCR and culture was established using clinical data to investigate discrepant results. We also compared the time to microbial diagnosis between the direct PCR and culture. The sensitivity, specificity, positive and negative predictive values of the PCR were 61.6% (53.5-69.1), 99.1% (98.2-99.6), 92.8% (85.8-96.5) and 93.4% (91.6-94.9), respectively, when considering all specimens. When considering smear-positive specimens and smear-negative specimens, the sensitivity was 81.6% and 40%, respectively. The sensitivity for pulmonary and extra-pulmonary smear-positive specimens was 85.2% versus 72.7%. The median time to identification at species level was 35 days (SD, 17.67) for culture and 6 days (SD, 2.67) for the PCR (when positive), which represents a 29-day shorter time to results (p < 0.0001). The 16S rRNA gene pan-mycobacterial PCR displays a substantial benefit in terms of time to diagnose NTM infections when compared with culture. Despite an excellent specificity, its sensitivity is yet limited in particular for smear-negative specimens, which might be improved by relying onto real-time PCRs.