Introduction: Thoracic aortic disease is an important cause of morbidity and mortality in the US, and aortic diameter is a heritable contributor to risk. Question: Could a polygenic prediction of ascending aortic diameter improve detection of aortic aneurysm? Methods: We used deep learning to measure ascending thoracic aortic diameter in 56,556 UK Biobank participants and conducted a genome-wide association study (GWAS) in 39,524. We built a 1.1-million SNP polygenic score using PRScs . In 4,896 non-GWAS participants (“training set”), we built linear models from clinical variables and the polygenic score (“AORTA Gene”) and simpler comparators. The models were assessed in the remaining 4,962 unrelated UK Biobank participants with imaging (“test set”). The models were externally validated in All Of Us participants over the age of 40. Results: In the test set, the AORTA Gene model explained 39.9% (95% CI 37.8-42.0%) of the variance in thoracic aortic diameter compared to 29.2% for the clinical model (95% CI 27.1-31.4%). For diameter ≥ 4cm, the area under the receiver operating characteristic curve (AUROC) was 0.834 for AORTA Gene, superior to the clinical model (0.765, P=7.3E-10). Thresholding at the top 10% of the score, AORTA Gene had a sensitivity of 52% and a specificity of 91.2% (F1: 0.197) for detecting the 113 participants with diameter ≥ 4. For the clinical score, the respective values were 36.3%, 90.9%, and 0.137. In 610 All Of Us participants with genetic data and aortic measurements, the AORTA Gene model had a correlation of 0.591 (95% CI 0.537-0.640) with aortic diameter vs 0.538 (95% CI 0.479-0.592) for the clinical model (P=2.2E-13 against a null hypothesis of the genetics being uninformative beyond clinical factors). It had an AUROC of 0.827 for identifying diameter ≥4cm vs 0.791 for the clinical model (P=7.8E-03). AORTA Gene also had a higher AUROC for prevalent thoracic aneurysm diagnoses (0.760 vs 0.739 for the clinical score, N=1,904 cases, P=2.4E-16) and incident diagnoses (0.748 vs 0.729, N=1,632 events, P=9.5E-10). Conclusions: Genetic information improved prediction of thoracic aortic diameter and aneurysm when added to clinical risk factors. Larger and more diverse samples will be needed to develop more powerful scores.
Read full abstract