Performance of 2 Finger-Stick Blood Tests to Triage Adults With Symptoms of Pulmonary Tuberculosis: A Prospective Multisite Diagnostic Accuracy Study

Abstract

Background: Non-sputum-based, point-of-care triage tests for pulmonary tuberculosis could enhance tuberculosis diagnostic programs. We assessed the diagnostic accuracy of 2 finger-stick blood tests: the Cepheid 3 gene host-response cartridge (Xpert-HR), which measures 3 host messenger RNA transcripts, and the 3-host protein multibiomarker test (MBT).

Methods: We performed a prospective diagnostic accuracy study of consecutive participants with symptoms compatible with pulmonary tuberculosis in The Gambia, South Africa, Uganda, and Vietnam. A composite reference standard for active pulmonary tuberculosis incorporated chest radiography, symptom resolution, and sputum microbiological test results. A training-test set approach was used to evaluate test cutoff specificities at 90% sensitivity.

Results: Between 1 November 2020 and 1 May 2023, we screened 1262 participants aged 12-70 years with cough lasting >2 weeks and another symptom suggestive of tuberculosis. Of those who were classifiable by reference tests, 1154 participants had evaluable Xpert-HR results and 961 had evaluable MBT results. Xpert-HR had an area under the receiver operating characteristic (AUROC) curve of 0.92 at a cutoff of -1.275 or below, with a sensitivity of 92.8%, specificity of 62.5%, positive predictive value of 47.9%, and negative predictive value of 95.9%. The MBT had an AUROC of 0.91 at a cutoff of ≥0.42, with a sensitivity of 91.4%, specificity of 73.2%, positive predictive value of 52.0%, and negative predictive value of 96.4%.

Conclusions: Our results show that both Xpert-HR and the MBT are promising non-sputum-based point-of-care tests. The MBT met the World Health Organization target product profile for a triage test, which suggests it should be further developed.

Keywords: Biomarkers; diagnosis; point-of-care; triage; tuberculosis.

Graphical Abstract

A graphic depicting the study population, the diagnostic tests being assessed, the reference standard, and the main results. This graphical abstract is also available at Tidbit: https://tidbitapp.io/institutional-portal/clinical-infectious-diseases/tidbits/performance-of-two-fingerstick-blood-tests-to-triage-adults-with-symptoms-of-pulmonary-tb-a-prospective-multisite-diagnostic-accuracy-study/update

Figure 1.

STARD diagram of participants in study. Details of the composite reference standard are provided in Table 1. Abbreviations: CRS, composite reference standard; MBT, multibiomarker test; Xpert-HR, Xpert MTB Host Response assay (Cepheid).

Figure 2.

Performance of the Xpert MTB Host Response assay (Xpert-HR). A, Area under the receiver operating characteristic (ROC) curve (AUROC) curve for the Xpert-HR TB score (using TBP in place of KLF2) against the composite reference standard (CRS). B, ROC curves for the Xpert-HR TB score at different study sites. C, D, Comparative median TB scores in participants classified as “tuberculosis positive” (C) or “tuberculosis negative” (D) by the CRS, at different study sites. Results of independent median tests were significant for comparisons of Uganda versus Vietnam (P < .001), Uganda versus South Africa (P = .03), and The Gambia versus Vietnam (P = .004) in the tuberculosis-negative group and for Uganda versus The Gambia (P < .001) in the tuberculosis-positive group. Dashed red lines represent the optimal cutoff between tuberculosis-positive and tuberculosis-negative classifications in this analysis. Abbreviation: CI, confidence interval.

Figure 3.

Forest plots comparing the measures of diagnostic accuracy for the Xpert MTB Host Response assay (Xpert-HR) between subgroups and sites. A, Sensitivity. B, Specificity. C, Positive predictive value (PPV). D, Negative predictive value (NPV). Vertical lines represent World Health Organization target product profile cutoffs of ≥90% sensitivity and ≥70% specificity. Abbreviations: CI, confidence interval; CRS, composite reference standard; HIV, human immunodeficiency virus.

Figure 4.

Performance of the multibiomarker test (MBT). A, Receiver operating characteristic (ROC) curve for the MBT against the composite reference standard. B, ROC curves for the MBT at different study sites. C, Effect of bacteria detectable with nasopharyngeal swab sample polymerase chain reaction (PCR) on the MBT (P = .004 for the tuberculosis-negative group). D, Effect of a virus detectable with nasopharyngeal swab sample PCR on the MBT. Dashed red lines represent the optimal cutoff between “tuberculosis positive” and “tuberculosis negative” in this analysis. Abbreviations: AUROC, area under the ROC; CI, confidence interval.

Figure 5.

Forest plots comparing the measures of diagnostic accuracy of the multibiomarker test (MBT) between subgroups and sites. A, Sensitivity. B, Specificity. C, Positive predictive value (PPV). D, Negative predictive value (NPV). Vertical lines represent World Health Organization target product profile cutoffs of ≥90% sensitivity and ≥70% specificity. Abbreviations: CI, confidence interval; CRS, composite reference standard; HIV, human immunodeficiency virus.