Considerations for biomarker-targeted intervention strategies for tuberculosis disease prevention

Abstract

Current diagnostic tests for Mycobacterium tuberculosis (MTB) infection have low prognostic specificity for identifying individuals who will develop tuberculosis (TB) disease, making mass preventive therapy strategies targeting all MTB-infected individuals impractical in high-burden TB countries. Here we discuss general considerations for a risk-targeted test-and-treat strategy based on a highly specific transcriptomic biomarker that can identify individuals who are most likely to progress to active TB disease as well as individuals with TB disease who have not yet presented for medical care. Such risk-targeted strategies may offer a rapid, ethical and cost-effective path towards decreasing the burden of TB disease and interrupting transmission and would also be critical to achieving TB elimination in countries nearing elimination. We also discuss design considerations for a Correlate of Risk Targeted Intervention Study (CORTIS), which could provide proof-of-concept for the strategy. One such study in South Africa is currently enrolling 1500 high-risk and 1700 low-risk individuals, as defined by biomarker status, and is randomizing high-risk participants to TB preventive therapy or standard of care treatment. All participants are monitored for progression to active TB with primary objectives to assess efficacy of the treatment and performance of the biomarker.

Keywords: Biomarker; Correlate of risk; Study design; Tuberculosis; mRNA.

Fig. 1

Tuberculosis biomarker comparison. Infection with Mycobacterium tuberculosis (MTB) is associated with a broad clinical spectrum of pathogenesis that includes asymptomatic persistent MTB infection; incipient TB disease; asymptomatic, but microbiologically detectable, sub-clinical TB disease; and active symptomatic TB disease. In CORTIS-01 the TB disease endpoint is based on detection of MTB nucleic acid (Xpert MTB/RIF) and/or sputum culture (Mycobacteria Growth Indicator Tube assays, MGIT), a standard tool for TB diagnosis. The interferon gamma release assay (IGRA) and the tuberculin skin test (TST) detect MTB-specific T cells in the blood and are currently used to identify MTB-infected individuals. Most IGRA/TST+ individuals will not develop active disease. The COR is an mRNA expression signature that detects the type I/II interferon response and is associated with active disease (i.e. diagnostic) and with individuals who are likely to develop disease (i.e. prognostic). A test-and-treat strategy would treat only COR+ individuals to prevent active disease. Individuals with latent MTB infection that have a lower risk of developing active TB disease may be IGRA/TST+ and COR-, thus sparing them unnecessary treatment in a mass test-and-treat campaign.

Fig. 2

“Hybrid” treatment selection design for CORTIS-01. (A) Healthy, HIV-uninfected adults are recruited and screened using the COR blood-based biomarker. Within 28 days of screening, COR⁺ participants are enrolled and randomized to preventive therapy (Treatment/Rx) or standard of care (SOC). A fraction (F) of COR⁻ participants are enrolled in a SOC arm. All enrolled participants are followed for 15 months for TB disease. COR⁺ and COR⁻ participants are enrolled concurrently and therefore rate of enrolment directly depends on the prevalence of COR-positivity (π₀). (B) The rate of incident TB in the COR⁺(Rx) and COR⁺(SOC) groups can be compared to evaluate treatment efficacy, while comparison of the COR⁺(SOC) and COR⁻(SOC) groups yields information about biomarker performance. Strategy efficacy is assessed using all three groups.

Fig. 3

Performance characteristics of COR biomarker. The COR was measured at baseline in participants of the Adolescent Cohort Study (ACS). (A) Relative risk of developing TB disease for COR⁺ vs. COR⁻ participants was computed longitudinally using a cumulative incidence-based approach (blue shaded 95% confidence interval from bootstrap sampling). Data was approximated well by exponential decay from RR = 15 to RR = 1 with decay time constant of 12 months (dashed line). (B) Sensitivity (red) and specificity (blue) of COR in the ACS training and test set samples up to 1 year preceding TB disease. Classification performance depends on the threshold applied to the continuous readout. Using a 60% threshold, the COR had 71% sensitivity and 84% specificity.

Fig. 4

Simulated TB endpoints. CORTIS-01 was simulated stochastically 10 K times using parameters estimated from previous observation studies, including overall TB incidence of 1.1 cases per year, treatment efficacy of 80% and an initial COR relative-risk of 15, decreasing exponentially to 1 with decay constant of 12 months (see Supplement). (A) Kaplan-Meier survival curves and the associated 2.5th and 97.5th percentiles across simulations for the COR⁺(Rx) group (red), the COR⁺(SOC) group (blue) and the COR⁻(SOC) group (green) within each trial. (B) Power to detect statistically significant treatment efficacy (TE ≥ 20%; two-sided α = 0.1; blue line), relative-risk (RR_COR ≥ 2; two-sided α = 0.05; red line) and strategy efficacy (SE ≥ 0%; two-sided α = 0.1; black line) was computed as the fraction of simulated trials in which a significant effect was detected.