Evaluation of the Xpert MTB Host Response assay for the triage of patients with presumed pulmonary tuberculosis: a prospective diagnostic accuracy study in Viet Nam, India, the Philippines, Uganda, and South Africa

Abstract

Background: Non-sputum-based triage tests for tuberculosis are a priority for ending tuberculosis. We aimed to evaluate the diagnostic accuracy of the late-prototype Xpert MTB Host Response (Xpert HR) blood-based assay.

Methods: We conducted a prospective diagnostic accuracy study among outpatients with presumed tuberculosis in outpatient clinics in Viet Nam, India, the Philippines, Uganda, and South Africa. Eligible participants were aged 18 years or older and reported cough lasting at least 2 weeks. We excluded those receiving tuberculosis treatment in the preceding 12 months and those who were unwilling to consent. Xpert HR was performed on capillary or venous blood. Reference standard testing included sputum Xpert MTB/RIF Ultra and mycobacterial culture. We performed receiver operating characteristic (ROC) analysis to identify the optimal cutoff value for the Xpert HR to achieve the target sensitivity of 90% or more while maximising specificity, then calculated diagnostic accuracy using this cutoff value. This study was prospectively registered with ClinicalTrials.gov, NCT04923958.

Findings: Between July 13, 2021, and Aug 15, 2022, 2046 adults with at least 2 weeks of cough were identified, of whom 1499 adults (686 [45·8%] females and 813 [54·2%] males) had valid Xpert HR and reference standard results. 329 (21·9%) had microbiologically confirmed tuberculosis. Xpert HR had an area under the ROC curve of 0·89 (95% CI 0·86-0·91). The optimal cutoff value was less than or equal to -1·25, giving a sensitivity of 90·3% (95% CI 86·5-93·3; 297 of 329) and a specificity of 62·6% (95% CI 59·7-65·3; 732 of 1170). Sensitivity was similar across countries, by sex, and by subgroups, although specificity was lower in people living with HIV (45·1%, 95% CI 37·8-52·6) than in those not living with HIV (65·9%, 62·8-68·8; difference of 20·8%, 95% CI 13·0-28·6; p<0·0001). Xpert HR had high negative predictive value (95·8%, 95% CI 94·1-97·1), but positive predictive value was only 40·1% (95% CI 36·8-44·1). Using the Xpert HR as a triage test would have reduced confirmatory sputum testing by 57·3% (95% CI 54·2-60·4).

Interpretation: Xpert HR did not meet WHO minimum specificity targets for a non-sputum-based triage test for pulmonary tuberculosis. Despite promise as a rule-out test that could reduce confirmatory sputum testing, further cost-effectiveness modelling and data on acceptability and usability are needed to inform policy recommendations.

Funding: National Institute of Allergy and Infectious Diseases of the US National Institutes of Health.

Translations: For the Vietnamese and Tagalog translations of the abstract see Supplementary Materials section.

Figure 1:. Participant flow chart

Reasons for indeterminate reference standard results are detailed in appendix 3 (p 10). MGIT=mycobacterium growth indicator tube. Xpert HR=Xpert MTB Host Response.

Figure 2:. Xpert HR TB Score ROC curve analysis against microbiological reference standard

(A) The overall ROC curve for TB Score and by country. (B) ROC curves by blood collection method. (C) ROC curves by HIV status. (D) ROC curves by diabetes status. The shaded region in charts A–D shows cutoff values that meet both the WHO target product profile minimal target sensitivity (≥90%) or specificity (≥70%), the dashed grey lines show the WHO target product profile minimal targets. AUCs and 95% CIs overall, and by country and subgroup, are in appendix 3 (p 14). (E) Xpert HR TB Score by microbiological reference standard. Median TB Score was −2·85 (IQR −3·28 to −2·18) for participants positive by microbiological reference standard, and −1·05 (IQR −1·50 to −0·70) for participants negative by microbiological reference standard. AUC=area under the ROC. ROC=receiver operating characteristic. Xpert HR=Xpert MTB Host Response.

Figure 3:. Diagnostic accuracy of the Xpert HR compared with the microbiological reference standard

Sensitivity, specificity, positive predictive value, and negative predictive value of the Xpert HR are compared with the microbiological reference standard overall and by subgroup. Vertical dashed line shows WHO target product profile minimum values for sensitivity (≥90%) and specificity (≥70%). Xpert HR=Xpert MTB Host Response.

Figure 4:. Decision curve analysis

Decision curve analysis shown as net reduction in sputum-based tuberculosis tests per 100 people with presumptive tuberculosis tested with Xpert HR weighted by a range of threshold probabilities from 5 to 30%. At a 5% threshold probability, triage testing using Xpert HR would lead to a net reduction of approximately eight tuberculosis tests per 100 people. At a 20% threshold probability, triage testing using Xpert HR would lead to a net reduction of approximately 40 tuberculosis tests per 100 patients. Dashed lines indicate the 5% and 20% threshold probabilities. Xpert HR=Xpert MTB Host Response.