New Insights into the Limitations of Host Transcriptional Biomarkers of Tuberculosis

Blood RNA biomarkers for tuberculosis screening in people living with HIV before antiretroviral therapy initiation: a diagnostic accuracy study

BACKGROUND

Undiagnosed tuberculosis remains a major threat for people living with HIV. Multiple blood transcriptomic biomarkers have shown promise for tuberculosis diagnosis. We sought to evaluate their diagnostic accuracy and clinical utility for systematic pre-antiretroviral therapy (ART) tuberculosis screening.

METHODS

We enrolled consecutive adults (age ≥18 years) referred to start ART at a community health centre in Cape Town, South Africa, irrespective of symptoms. Sputa were obtained (using induction if required) for two liquid cultures. Whole-blood RNA samples underwent transcriptional profiling using a custom Nanostring gene panel. We measured the diagnostic accuracy of seven candidate RNA signatures (one single gene biomarker [BATF2] and six multigene biomarkers) for the reference standard of Mycobacterium tuberculosis culture status, using area under the receiver-operating characteristic curve (AUROC) analysis, and sensitivity and specificity at prespecified thresholds (two standard scores above the mean of healthy controls; Z2). Clinical utility was assessed by calculating net benefit in decision curve analysis. We compared performance with C-reactive protein (CRP; threshold ≥5 mg/L), WHO four-symptom screen (W4SS), and the WHO target product profile for tuberculosis triage tests.

FINDINGS

A total of 707 people living with HIV (407 [58%] female and 300 [42%] male) were included, with median CD4 count 306 cells per mm3 (IQR 184-486). Of 676 participants with available sputum culture results, 89 (13%) had culture-confirmed tuberculosis. The seven RNA signatures were moderately to highly correlated (Spearman rank coefficients 0·42-0·93) and discriminated tuberculosis culture positivity with similar AUROCs (0·73-0·80), but none statistically better than CRP (AUROC 0·78, 95% CI 0·72-0·83). Diagnostic accuracy was similar across CD4 count strata, but lower among participants with negative W4SS (AUROCs 0·56-0·65) compared with positive (AUROCs 0·75-0·84). The RNA biomarker with the highest AUROC point estimate was a four-gene signature (Suliman4; AUROC 0·80, 95% CI 0·75-0·86), with sensitivity 83% (95% CI 74-90) and specificity 59% (55-63) at the Z2 threshold. In decision curve analysis, Suliman4 and CRP had similar clinical utility to guide confirmatory tuberculosis testing, but both had higher net benefit than W4SS. In exploratory analyses, an approach combining CRP (≥5 mg/L) and Suliman4 (≥Z2) had sensitivity of 80% (70-87), specificity of 70% (66-74), and higher net benefit than either biomarker alone.

INTERPRETATION

RNA biomarkers showed better clinical utility to guide confirmatory tuberculosis testing for people living with HIV before ART initiation than symptom-based screening, but their performance did not exceed that of CRP and fell short of WHO recommended targets. Interferon-independent approaches might be required to improve accuracy of host-response biomarkers to support tuberculosis screening before ART initiation.

FUNDING

South African Medical Research Council, European and Developing Countries Clinical Trials Partnership 2, National Institutes of Health National Institute of Allergy and Infectious Diseases, The Wellcome Trust, National Institute for Health and Care Research, Royal College of Physicians London.

Toward a molecular microbial blood test for tuberculosis infection

The World Health Organization's aim to end the global tuberculosis (TB) epidemic by 2050 cannot be achieved without taking measures to identify people with asymptomatic Mycobacterium tuberculosis (Mtb) infection and offer them an intervention to reduce the risk of disease progression, such as preventive antimicrobial therapy. Implementation of this strategy is limited by the fact that existing tests for Mtb infection, which use immunosensitization to Mtb-specific antigens as a proxy for infection, have low positive predictive value for progression to TB. A blood test that detects Mtb deoxyribonucleic acid (DNA) could allow preventive therapy to be targeted at individuals with microbiological evidence of persistent infection. In this review, we summarize recent advances in the development of molecular microbial blood tests for Mtb infection and discuss potential explanations for discordance between their results and those of immunodiagnostic tests in adults with recent exposure to an infectious index case. We also present a roadmap for further development of molecular microbial blood tests for Mtb infection, and highlight the potential for research in this area to provide novel insights into the biology of Mtb infection and yield new tools to support efforts to control the global TB epidemic.

Poor performance of paired tests of latent tuberculosis in highly immune-compromised individuals exposed to multidrug-resistant tuberculosis: time for new diagnostic markers

Guideline-based recommendations for diagnosis of latent TB in highly immune suppressed populations are difficult to interpret and poorly characterised. More accurate biomarkers independent of T-cell functions are urgently required. https://bit.ly/41P8vTa.

Blood RNA biomarkers for tuberculosis screening in people living with HIV prior to anti-retroviral therapy initiation: A diagnostic accuracy study

Background

Undiagnosed tuberculosis (TB) remains a major threat for people living with HIV (PLHIV). Multiple blood transcriptomic biomarkers have shown promise for TB diagnosis. We sought to evaluate their diagnostic accuracy and clinical utility for systematic pre-antiretroviral therapy (ART) TB screening.

Methods

We enrolled consecutive adults referred to start ART at a community health centre in Cape Town, South Africa, irrespective of symptoms. Sputa were obtained (using induction if required) for two liquid cultures. Whole-blood RNA samples underwent transcriptional profiling using a custom Nanostring gene-panel. We measured the diagnostic accuracy of seven candidate RNA biomarkers for the reference standard of Mycobacterium tuberculosis culture status, using area under the receiver-operating characteristic curve (AUROC) analysis, and sensitivity/specificity at pre-specified thresholds (two standard scores above the mean of healthy controls; Z2). Clinical utility was assessed using decision curve analysis. We compared performance to CRP (threshold ≥5mg/L), World Health Organisation (WHO) four-symptom screen (W4SS) and the WHO target product profile for TB triage tests.

Results

A total of 707 PLHIV were included, with median CD4 count 306 cells/mm3. Of 676 with available sputum culture results, 89 (13%) had culture-confirmed TB. The seven RNA biomarkers were moderately to highly correlated (Spearman rank coefficients 0.42-0.93) and discriminated TB culture-positivity with similar AUROCs (0.73-0.80), but none statistically better than CRP (AUROC 0.78; 95% CI 0.72-0.83). Diagnostic accuracy was similar across CD4 count strata, but lower among W4SS-negative (AUROCs 0.56-0.65) compared to W4SS-positive participants (AUROCs 0.75-0.84). The RNA biomarker with highest AUROC point estimate was a 4-gene signature (Suliman4; AUROC 0.80; 95% CI 0.75-0.86), with sensitivity 0.83 (0.74-0.90) and specificity 0.59 (0.55-0.63) at Z2 threshold. In decision curve analysis, Suliman4 and CRP had similar clinical utility to guide confirmatory TB testing, but both had higher net benefit than W4SS. In exploratory analyses, an approach combining CRP (≥5mg/L) and Suliman4 (≥Z2) had sensitivity of 0.80 (0.70-0.87), specificity of 0.70 (0.66-0.74) and higher net benefit than either biomarker alone.

Interpretation

RNA biomarkers showed better clinical utility to guide confirmatory TB testing for PLHIV prior to ART initiation than symptom-based screening, but their performance did not exceed that of CRP, and fell short of WHO recommended targets. Interferon-independent approaches may be required to improve accuracy of host-response biomarkers to support TB screening pre-ART initiation.

Funding

South African MRC, EDCTP2, NIH/NIAID, Wellcome Trust, NIHR, Royal College of Physicians London.

Development of diagnostic algorithm using machine learning for distinguishing between active tuberculosis and latent tuberculosis infection

BACKGROUND

The discrimination between active tuberculosis (ATB) and latent tuberculosis infection (LTBI) remains challenging. The present study aims to investigate the value of diagnostic models established by machine learning based on multiple laboratory data for distinguishing Mycobacterium tuberculosis (Mtb) infection status.

METHODS

T-SPOT, lymphocyte characteristic detection, and routine laboratory tests were performed on participants. Diagnostic models were built according to various algorithms.

RESULTS

A total of 892 participants (468 ATB and 424 LTBI) and another 263 participants (125 ATB and 138 LTBI), were respectively enrolled at Tongji Hospital (discovery cohort) and Sino-French New City Hospital (validation cohort). Receiver operating characteristic (ROC) curve analysis showed that the value of individual indicator for differentiating ATB from LTBI was limited (area under the ROC curve (AUC) < 0.8). A total of 28 models were successfully established using machine learning. Among them, the AUCs of 25 models were more than 0.9 in test set. It was found that conditional random forests (cforest) model, based on the implementation of the random forest and bagging ensemble algorithms utilizing conditional inference trees as base learners, presented best discriminative power in segregating ATB from LTBI. Specially, cforest model presented an AUC of 0.978, with the sensitivity of 93.39% and the specificity of 91.18%. Mtb-specific response represented by early secreted antigenic target 6 (ESAT-6) and culture filtrate protein 10 (CFP-10) spot-forming cell (SFC) in T-SPOT assay, as well as global adaptive immunity assessed by CD4 cell IFN-γ secretion, CD8 cell IFN-γ secretion, and CD4 cell number, were found to contribute greatly to the cforest model. Superior performance obtained in the discovery cohort was further confirmed in the validation cohort. The sensitivity and specificity of cforest model in validation set were 92.80% and 89.86%, respectively.

CONCLUSIONS

Cforest model developed upon machine learning could serve as a valuable and prospective tool for identifying Mtb infection status. The present study provided a novel and viable idea for realizing the clinical diagnostic application of the combination of machine learning and laboratory findings.

Toward a more generalizable blood RNA signature for bacterial and viral infections

Host-response profiles can discriminate different infections. A new 8-gene blood RNA signature to discriminate bacterial and viral infections extends our focus hitherto on the case mix from the US and Europe to include that of low- and middle-income countries.1 Challenges remain.