PA-347 Diagnostic accuracy of published host blood transcriptomic tuberculosis signatures in a low burden hospital setting

BackgroundSeveral host blood transcriptomic signatures have shown promise as tuberculosis (TB) diagnostic candidates, supporting the use of such biomarkers in prototype RNA-based point-of-care tests. Further validation of these signatures across different clinical and endemic settings in relevant patient cohorts is required to ascertain their global applicability. We aimed to evaluate the accuracies of published TB transcriptomic signatures in discriminating TB disease from lower respiratory tract infections (LRI) and latent tuberculosis infection (LTBI) in a low-burden hospital setting. MethodsWe evaluated the accuracy of 20 candidate blood transcriptomic TB signatures in diagnosing TB in 86 individuals (18 TB, 19 LRI, and 49 LTBI) recruited at Oslo University Hospital, Norway. Gene expression was measured using the fluidigm microfluidic qRT-PCR platform on Paxgene blood RNA samples collected at inclusion. The diagnostic performance of the signatures was assessed by area under the receiver operating characteristic curve (AUC-ROC).ResultsWhen a head-to-head comparison of the accuracies of the signatures was made between the study groups (TB vs LRI and TB vs LTBI) similar diagnostic performance was observed for the Sweeney 3-, Francisco 2-, Gjøen 7-, and Roe 3- gene signatures (AUC between 0.78 and 0.89). GBP1 and GBP5 were the most accurate individual genes (AUC ≥ 0.80), differentiating TB patients from individuals with LRI and LTBI. However, new 3- and 5-gene combinations, improved the accuracy in distinguishing TB from LRI (AUC=1.00, Sensitivity and specificity of 100%) and TB from LTBI (AUC=0.99, Sensitivity = 93.3%, Specificity = 97.9%), respectively. ConclusionHost blood transcriptomic TB signatures primarily identified in studies from high TB-burden settings showed equivalent diagnostic potential in a low-burden hospital setting. New combinations of transcripts with improved diagnostic accuracy, met the minimal WHO target product profile thresholds for non-sputum-based triage TB tests, and warrant further investigation in larger, prospective, and diverse patient cohorts.