Utility of a three-gene transcriptomic signature in the diagnosis of tuberculosis in a low-endemic hospital setting

Host transcriptomic blood signatures have demonstrated diagnostic potential for tuberculosis (TB), requiring further validation across different geographical settings. Discriminating TB from other diseases with similar clinical manifestations is crucial for the development of an accurate immunodiagnostic tool. In this exploratory cohort study, we evaluated the performance of potential blood-based transcriptomic signatures in distinguishing TB disease from non-TB lower respiratory tract infections in hospitalised patients in a TB low-endemic country. Quantitative real-time polymerase chain reaction qPCR) was used to evaluate 26 previously published genes in blood from 31 patients (14 TB and 17 lower respiratory tract infection cases) admitted to Oslo University Hospital in Norway. The diagnostic accuracies of differentially expressed genes were determined by receiver operating characteristic curves. A significant difference (<i>p</i> &lt; .01) in the age distribution was observed between patients with TB (mean age, 40 ± 15 years) and lower respiratory tract infection (mean age 59 ± 12 years). Following adjustment for age, <i>ETV7, GBP1, GBP5, P2RY14</i> and <i>BLK</i> were significantly differentially expressed between patients with TB and those with LRI. A general discriminant analysis generated a three-gene signature (<i>BAFT2, ETV7</i> and <i>CD1C</i>), which diagnosed TB with an area under the receiver operating characteristic curve (AUC) of 0.86 (95% CI, 0.69 − 1.00), sensitivity of 69.23% (95% CI, 38.57%−90.91%) and specificity of 94.12% (95% CI, 71.31%−99.85%). The three-genes signature may have potential to improve diagnosis of TB in a hospitalised low-burden setting. However, the influence of confounding variables or covariates such as age requires further evaluation in larger studies.