Diagnostic accuracy study of a novel blood-based assay^# for identification of TB in people living with HIV

Abstract

A non-sputum triage test to rule out TB disease is a WHO high-priority diagnostic and a combinatory score based on a 3-gene host-signature has shown promise in discriminating TB from other illnesses. We evaluated the accuracy of an early-prototype cartridge-assay (Xpert MTB Host Response, or Xpert-MTB-HR-Prototype) of this 3-gene signature on bio-banked blood-samples from PLHIV against a comprehensive microbiological reference standard (CMRS) and against Xpert MTB/RIF on first sputum alone. We depict results based on performance targets set by WHO in comparison with a laboratory-based CRP assay. Of 201 patients included, 67 were culture-positive for Mycobacterium tuberculosis. AUC for the Xpert-MTB-HR-Prototype was 0.89 (CI 0.83-0.94) against the CMRS and 0.94 (CI 0.89-0.98) against Xpert MTB/RIF. Considering Xpert-MTB-HR-Prototype as a triage test (at nearest upper value of sensitivity to 90%), specificity was 55.8% (CI 47.2-64.1) compared to the CMRS and 85.9% (CI 79.3-90.7) compared to Xpert MTB/RIF as confirmatory tests. Considering Xpert-MTB-HR-Prototype as a stand-alone diagnostic test, at a specificity near 95%, the test achieved a sensitivity of 65.7% (CI 53.7-75.9) while CRP achieved a sensitivity of only 13.6% (CI 7.3-23.4). In this first accuracy study of a prototype blood-based host-marker assay, we show the possible value of the assay for triage and diagnosis in PLHIV.

Blood-based host biomarker diagnostics in active case finding for pulmonary tuberculosis: a diagnostic case-control study

“…Our results corroborate a previous study conducted by Södersten and colleagues in a very different population and clinical setting-screening HIVinfected individuals presenting with TB symptoms to clinics in South Africa and Peru-which demonstrated feasibility of measuring the 3-gene signature on this platform. 11 Consistent with this and earlier studies on the 3-gene signature, we found that TB cases had lower Xpert-MTB-HR scores, indicating higher expression levels (up-regulation) of GBP5 and DUSP3 compared with KLF2, compared with controls. [9][10][11]20 The AUC in this study (0.84) was lower than that reported by Södersten (0.94 against sputum Xpert MTB/RIF), 11 but consistent with a previous smaller evaluation of the 3-gene signature by laboratory-based qRT-PCR in this low HIVprevalence, actively screened population (0.87).…”

“…11 Consistent with this and earlier studies on the 3-gene signature, we found that TB cases had lower Xpert-MTB-HR scores, indicating higher expression levels (up-regulation) of GBP5 and DUSP3 compared with KLF2, compared with controls. [9][10][11]20 The AUC in this study (0.84) was lower than that reported by Södersten (0.94 against sputum Xpert MTB/RIF), 11 but consistent with a previous smaller evaluation of the 3-gene signature by laboratory-based qRT-PCR in this low HIVprevalence, actively screened population (0.87). 10 Earlier studies have also investigated CRP testing for distinguishing individuals with active TB from those without disease.…”

“…A recent study demonstrated its potential in triage and diagnosis of active TB among people living with HIV and presenting to healthcare facilities with TB symptoms. 11 Whether this assay would perform well for active case finding, irrespective of symptoms, in a low HIV prevalence setting is unknown. To address this knowledge gap, we performed a diagnostic evaluation study nested within a prospective active TB case finding study in prisons, evaluating the accuracy of a blood based 3-gene signature assay as potential in triage and diagnosis of active TB.…”

Multi-cohort analysis of host immune response identifies conserved protective and detrimental modules associated with severity irrespective of virus

“…However, such a conserved gene signature, identified using a large amount of heterogeneous data across multiple cohorts, can be further analyzed to identify a parsimonious, clinically useful, point-of-care test that is generalizable across patient populations. For example, we have previously described a 3-gene host response-based signature for diagnosis of tuberculosis that has been shown to be generalizable across a large number of patient populations, and has been translated into a proof-of-concept point-of-care cartridge with high accuracy (Sodersten et al, 2020). In addition, given the high pairwise correlation between genes within each module, only a small subset of genes within each module would provide the same discriminatory power, further allowing the selection of a parsimonious gene signature.…”

“…For example, a 3-gene host response-based signature has been repeatedly demonstrated to predict progression from latent Mycobacterium tuberculosis (Mtb) infection to active tuberculosis 6 months prior (Gupta et al, 2020; Turner et al, 2020; Warsinske et al, 2018). Finally, and most importantly, these generalizable robust host response-based signatures can be translated onto a targeted platform as a point-of-care test with rapid turnaround time (Sodersten et al, 2020).…”