Background: We aimed to investigate the association between aortic arch and coronary artery calcification (CAC). We postulated that low‐ and high‐risk CAC scores could be predicted with the evaluation of standard chest radiography for aortic arch calcification (AAC). Patients and methods: Consecutive patients who were referred for a multidetector computerized tomography (MDCT) examination were enrolled prospectively. All patients were scanned using a commercially available 64‐slice MDCT scanner for the evaluation of CAC score. A four‐point grading scale (0, 1, 2 and 3) was used to evaluate AAC on the standard posterior‐anterior chest radiography images. Results: The study group consisted of 248 patients. Median age of the study group was 52 (IQR: 10) years, and 165 (67 %) were male. AAC grades (r = 0.676, p < 0.0001) and age (r = 0.518, p < 0.0001) were significantly and positively correlated with CAC score. Presence of AAC was independently associated with the presence of CAC (OR: 11.20, 95 % CI 4.25 to 29.52). An AAC grade of ≥ 2 was the strongest independent predictor of a high‐risk CAC score (OR: 27.42, 95 % CI 6.09 to 123.52). Receiver operating characteristics curve analysis yielded a strong predictive ability of AAC grades for a CAC score of ≥ 100 (AUC = 0.892, P < 0.0001), and ≥ 400 (AUC = 0.894, P < 0.0001). Absence of AAC had a sensitivity, specificity and accuracy of 90 %, 84 % and 89 %, respectively, for a CAC score of < 100. An AAC grade of ≥ 2 predicted a CAC score of ≥400 with a sensitivity, specificity and accuracy of 68 %, 98 % and 95 %, respectively. Conclusions: AAC is a strong and independent predictor of CAC. The discriminative performance of AAC is high in detecting patients with low‐ and high‐risk CAC scores.
Application of machine learning algorithms to predict coronary artery calcification with a sibship-based design
