Transcriptomic signatures for diagnosing tuberculosis in clinical practice: a prospective, multicentre cohort study

Area of Science:

• Genomics
• Infectious Diseases
• Diagnostic Technologies

Background:

• Blood transcriptomic signatures have shown promise for tuberculosis (TB) diagnosis in case-control studies.
• However, prospective validation in diverse clinical presentations, including extrapulmonary TB and mimics, is lacking.

Purpose of the Study:

• To evaluate the diagnostic accuracy of transcriptomic signatures in adults with clinically suspected tuberculosis in a routine care setting.
• To assess novel and previously published signatures against established performance benchmarks.

Main Methods:

• Prospective, multicentre cohort study (VANTDET) in England involving 628 participants with suspected TB.
• Whole blood microarray analysis to measure transcript abundance, followed by 6-12 months of diagnostic follow-up.
• Assessment of six novel and three published transcriptomic signatures using receiver-operating characteristic curves (AUC-ROC), sensitivity, and specificity.

Main Results:

• The best novel signature achieved an AUC-ROC of 0.87 for all active TB, with 77% sensitivity and 84% specificity.
• The best published signature had an AUC-ROC of 0.83, 78% sensitivity, and 76% specificity.
• No signatures met the World Health Organization's targets for paired sensitivity and specificity for non-sputum-based TB diagnostics.

Conclusions:

• Transcriptomic signatures demonstrated inadequate diagnostic accuracy for tuberculosis in a clinically representative cohort in routine practice.
• These findings suggest limited current clinical utility for transcriptomic signatures in the diagnostic evaluation of suspected TB.