HIV-infected individuals are at high risk of
tuberculosis disease and those with prior
tuberculosis episodes are at even higher risk of disease recurrence. A non-sputum
biomarker that identifies individuals at highest
tuberculosis risk would allow targeted microbiological testing and appropriate treatment and also guide need for prolonged
therapy. We determined the utility of a previously developed whole blood transcriptomic correlate of risk (COR) signature for (1) predicting incident recurrent
tuberculosis, (2)
tuberculosis diagnosis and (3) its potential utility for
tuberculosis treatment monitoring in HIV-infected individuals. We retrieved cryopreserved blood specimens from three previously completed clinical studies and measured the COR signature by quantitative microfluidic real-time-PCR. The signature differentiated recurrent
tuberculosis progressors from non-progressors within 3 months of diagnosis with an area under the Receiver-operating characteristic (ROC) curve (AUC) of 0.72 (95% confidence interval (CI), 0.58-0.85) amongst HIV-infected individuals on antiretroviral
therapy (ART). Twenty-five of 43 progressors (58%) were asymptomatic at microbiological diagnosis and thus had subclinical disease. The signature showed excellent diagnostic discrimination between HIV-uninfected
tuberculosis cases and controls (AUC 0.97; 95%CI 0.94-1). Performance was lower in HIV-infected individuals (AUC 0.83; 95%CI 0.81-0.96) and signature scores were directly associated with HIV viral loads.
Tuberculosis treatment response in HIV-infected individuals on ART with a new recurrent
tuberculosis diagnosis was also assessed. Signature scores decreased significantly during treatment. However, pre-treatment scores could not differentiate between those who became sputum negative before and after 2 months. Direct application of the unmodified blood transcriptomic COR signature detected subclinical and active
tuberculosis by blind validation in HIV-infected individuals. However, prognostic performance for recurrent
tuberculosis, and performance as diagnostic and as treatment monitoring tool in HIV-infected persons was inferior to published results from HIV-negative cohorts. Our results suggest that performance of transcriptomic signatures comprising
interferon stimulated genes are negatively affected in HIV-infected individuals, especially in those with incompletely suppressed viral loads.