Head-to-head evaluation of host blood biomarkers for point-of-care TB triage: diagnostic accuracy study

ABSTRACT Global tuberculosis (TB) incidence is slowly decreasing, partly because undiagnosed TB continues to fuel transmission. Point-of-care TB triage tests (TTTs) performed near a patient or at the site of patient care may help bridge this gap, but head-to-head comparisons of different host blood biomarkers within the same cohort are limited. In ambulatory adults in Uganda aged ≥15 years with any cough, fever, weight loss, or night sweats, we evaluated the diagnostic accuracy of C-reactive protein (CRP), monocyte-to-lymphocyte ratio (MLR), interferon-gamma-induced protein 10 (IP-10), and hemoglobin (anemia) head-to-head using a reference standard comprised of culture and/or Xpert MTB/RIF Ultra. Cut-offs of CRP and anemia were from World Health Organization (WHO) recommendations, MLR from the literature, and IP-10 from Youden’s index. We included 623 participants (193 people living with HIV [PLWHIV]), of whom 111 (18%) were reference standard-positive. Sensitivity and specificity (95% CI) were 91% (84%–96%) and 78% (74%–81%) for CRP (≥10 mg/L); 72% (63%–78%) and 80% (77%–84%) for MLR (≥0.378); 86% (79%–92%) and 73% (69%–77%) for IP-10 (≥28.75); 69% (60%–78%) and 76% (72%–80%) for anemia (<12 g/dL [female], <13 g/dL [male]), respectively. At 90% sensitivity, specificity was 76% (58%–83%) for CRP; 62% (47%–74%) for MLR; 64% (47%–75%) for IP-10, and 28% (17%–51%) for anemia. There was no significant difference for any of the four TTTs in diagnostic odds ratio between participants with and without a history of previous TB, between smokers and non-smokers, and between males and females. However, MLR performed significantly ( P = 0.001) worse in PLWHIV. Combining two index tests in parallel testing improved sensitivity up to 96% at the expense of reduced specificity. The applicable WHO minimum target product profile of 90% sensitivity and 60% specificity was met for CRP irrespective of HIV status, and for IP-10 and MLR when cut-offs were adjusted. Sex, prior history of smoking, and TB did not impact diagnostic performance. Parallel testing may be suitable in high TB prevalence settings, where minimizing missed cases and ensuring timely referral for confirmatory testing are priorities. IMPORTANCE This study provides strong evidence that host blood protein biomarkers (C-reactive protein [CRP], interferon gamma-induced protein 10 [IP-10], and monocyte-to-lymphocyte ratio [MLR]) can serve as effective point-of-care tools for tuberculosis (TB) triage. CRP reliably meets the World Health Organization (WHO) high-sensitivity target product profile criteria across people living with HIV and HIV-negative populations, while optimized cut-offs enable IP-10 and MLR to achieve comparable performance, expanding feasible options where CRP is unavailable or unaffordable. The minimal influence of sex, smoking history, or prior TB supports broad, unstratified implementation. Evidence that parallel testing reduces missed TB cases is particularly relevant in high-burden settings. Clinically, these findings support integrating biomarker-based triage into routine screening to improve early case detection, although it may require further external validation to ensure cut-offs used are robust and not overly data-driven. For policymakers, the results justify incorporating biomarker-based point-of-care tests into national TB screening algorithms to accelerate diagnosis, improve referral efficiency, and reduce TB transmission.