Prospective multicentre head-to-head validation of host blood transcriptomic biomarkers for pulmonary tuberculosis by real-time PCR
Simon C. Mendelsohn, Stanley Kimbung Mbandi, Andrew Fioré-Gartland, Adam Penn‐Nicholson, Munyaradzi Musvosvi, Humphrey Mulenga, Michelle Fisher, Katie Hadley, Mzwandile Erasmus, Onke Nombida, Michèle Tameris, Gerhard Walzl, Kogieleum Naidoo, Gavin Churchyard, Mark Hatherill, Thomas J. Scriba
Abstract
Abstract Background Sensitive point-of-care screening tests are urgently needed to identify individuals at highest risk of tuberculosis. We prospectively tested performance of host-blood transcriptomic tuberculosis signatures. Methods Adults without suspicion of tuberculosis were recruited from five endemic South African communities. Eight parsimonious host-blood transcriptomic tuberculosis signatures were measured by microfluidic RT-qPCR at enrolment. Upper respiratory swab specimens were tested with a multiplex bacterial-viral RT-qPCR panel in a subset of participants. Diagnostic and prognostic performance for microbiologically confirmed prevalent and incident pulmonary tuberculosis was tested in all participants at baseline and during active surveillance through 15 months follow-up, respectively. Results Among 20,207 HIV-uninfected and 963 HIV-infected adults screened; 2923 and 861 were enroled. There were 61 HIV-uninfected (weighted prevalence 1.1%) and 10 HIV-infected (prevalence 1.2%) tuberculosis cases at baseline. Parsimonious signature diagnostic performance was superior among symptomatic (AUCs 0.85–0.98) as compared to asymptomatic (AUCs 0.61–0.78) HIV-uninfected participants. Thereafter, 24 HIV-uninfected and 9 HIV-infected participants progressed to incident tuberculosis (1.1 and 1.0 per 100 person-years, respectively). Among HIV-uninfected individuals, prognostic performance for incident tuberculosis occurring within 6–12 months was higher relative to 15 months. 1000 HIV-uninfected participants were tested for respiratory microorganisms and 413 HIV-infected for HIV plasma viral load; 7/8 signature scores were higher ( p < 0.05) in participants with viral respiratory infections or detectable HIV viraemia than those without. Conclusions Several parsimonious tuberculosis transcriptomic signatures met triage test targets among symptomatic participants, and incipient test targets within 6 months. However, the signatures were upregulated with viral infection and offered poor specificity for diagnosing sub-clinical tuberculosis.