Abstract 16467: A Novel CT-derived Radiotranscriptomic Signature of Perivascular Adipose Tissue Stratifies COVID-19 Vascular Cytokine Burst and Predicts in Hospital Outcomes

• Published in: Circulation (New York, N.Y.) N.Y.), 2020-11, Vol.142 (Suppl_3 Suppl 3), p.A16467-A16467
• Main Authors: Kotanidis, Christos P, Xie, Cheng, Kotronias, Rafail, Siddique, Muhammad, Thomas, Sheena, Schottlander, David, Channon, Keith, Neubauer, Stefan, Deanfield, John, Antoniades, Charalambos
• Format: Article
• Language: English
• ISSN: 0009-7322; 1524-4539
• DOI: 10.1161/circ.142.suppl_3.16467

IntroductionCOVID-19 is characterised by severe vascular inflammation. Perivascular adipose tissue (PVAT) has the ability to change its texture in response to vascular inflammation. HypothesisComputed Tomography Angiography (CTA)-based radiotranscriptomic phenotyping of PVAT may quantify COVID-19-induced vascular inflammation, predicting clinical outcomes. MethodsIn Study 1, RNA sequencing of 60 internal mammary artery (IMA) biopsies from patients undergoing coronary bypass surgery was performed to build a transcriptomic fingerprint similar to that observed in COVID-19. This fingerprint was used to train an extreme gradient boosting algorithm, C19-RS, using CTA-derived radiomic features of PVAT around the IMA and descending thoracic aorta. In Study 2, C19-RS was validated in pulmonary artery CTAs from an independent cohort of 201 patients for COVID-19 detection and test its prognostic value in COVID-19. ResultsUnsupervised hierarchical clustering of RNASeq data in Study 1 identified 2 clusters of vascular inflammation (A). Machine learning was used to train C19-RS to detect vascular inflammation based on 31 radiomic features. In study 2, 22 deaths and 32 ICU admissions were recorded. Patients with high C19-RS had an OR=3.11[95%CI:1.06-9.85] for COVID-19 adjusted for age, sex, risk factors, hsCRP, WBCC, COPD and CT tube voltage. C19-RS significantly improved the discrimination of a baseline model containing the above variables, for COVID-19 detection (delta[AUC]=0.03, p=0.008, B). C19-RS was significantly associated with in-hospital death (C), and a composite endpoint of in-hospital death and ICU admission, with adjusted HR4.29 (95% CI1.48-13.52, p=0.009). ConclusionCOVID-19-induced vascular inflammation can be quantified by a radiotranscriptomic signature (C19-RS) derived from CT analysis of PVAT. C19-RS stratifies vascular inflammatory burden in COVID-19, and has striking prognostic value for in-hospital outcomes.