Greenan-Barrett, James;
Mendelsohn, Simon C;
Scriba, Thomas J;
Noursadeghi, Mahdad;
Gupta, Rishi K;
(2025)
Single-gene transcripts for subclinical tuberculosis: an individual participant data meta-analysis.
The Lancet Microbe
, Article 101186. 10.1016/j.lanmic.2025.101186.
(In press).
Preview |
Text
PIIS2666524725001144.pdf - Published Version Download (7MB) | Preview |
Abstract
Background: Translation of blood RNA signatures might be accelerated by identifying biomarkers composed of the minimum number of gene transcripts. We aimed to test the hypothesis that single-gene transcripts provide similar accuracy for detection of subclinical tuberculosis to multi-gene signatures and benchmark their accuracy and clinical utility against interferon-γ release assays (IGRAs). // Methods: For this individual participant data meta-analysis, we searched PubMed from database inception to June 10, 2024, using terms for “tuberculosis”, “subclinical”, and “RNA” to identify studies in which participants underwent whole-blood RNA sampling with at least 12 months of follow-up for development of clinical tuberculosis. We performed a one-stage individual participant data meta-analysis to compare the accuracy of multi-gene signatures against single-gene transcripts to discriminate individuals with subclinical tuberculosis—defined as asymptomatic prevalent or incident tuberculosis (diagnosed ≥21 days from enrolment, irrespective of symptoms) over a 12-month interval—from individuals who remained disease free. We performed decision curve analysis to evaluate the net benefit of using single-gene transcripts and IGRAs, alone or in combination, to stratify preventive treatment compared with strategies of treating all or no individuals. // Findings: 276 articles were identified in the search; of these, seven met the eligibility criteria and all had IPD available. We evaluated 80 single-genes and eight multi-gene signatures in a pooled analysis of four RNA sequencing and three quantitative PCR datasets, comprising 6544 total samples and including 283 samples from 214 individuals with subclinical tuberculosis. Distributions of transcript and signature Z scores after standardisation were similar and there was little heterogeneity between datasets. Five single-gene transcripts (BATF2, FCGR1A/B, ANKRD22, GBP2, and SERPING1) had equivalent areas under the receiver operating characteristic curves (0·75 [95% CI 0·71–0·79] to 0·77 [0·73–0·81]) to the best-performing multi-gene signature over 12 months, but none met the WHO minimum target product profile (TPP) for a tuberculosis progression test. IGRAs approximated the TPP in low-burden settings but showed much lower specificity in high-burden settings (74% [95% CI 72–76] vs 32% [30–35]). By contrast, sensitivity (67% [47–82] in high-burden settings vs 78% [67–86] in low-burden settings) and specificity (72% [70–74] vs 67% [64–69]) of the best-performing single-gene transcript was similar across settings. Decision curve analysis showed that in high-burden settings, stratifying preventive treatment using single-gene transcripts had greater net benefit than using IGRAs, which offered little net benefit over treating all individuals. In low-burden settings, IGRAs offered greater net benefit than single-gene transcripts to stratify treatment, but combining both tests provided the highest net benefit for tuberculosis programmes aiming to treat fewer than 50 people to prevent a single case. // Interpretation: Single-gene transcripts are equivalent to multi-gene signatures for detection of subclinical tuberculosis, with consistent performance across settings. Single-gene transcripts show potential clinical utility to stratify preventive treatment, particularly when used in combination with IGRAs in low-burden settings. // Funding: National Institute for Health Research, Wellcome Trust.
Archive Staff Only
 |
View Item |
