Development and evaluation of Z-score based aortic diameter thresholds for early detection of thoracic aortic dissection and aneurysm: Analysis in the UK Biobank
Authors
Affiliations (1)
Affiliations (1)
- University College London
Abstract
Background & PurposeClinical guidelines recommend using an absolute ascending aortic diameter (AAD) cutoff of 4.5 cm for monitoring and 5.5 cm for surgery in people at risk of thoracic aortic dissection. However, AAD varies with age, sex, height and body surface area (BSA). Using imaging data from the UK Biobank, we investigated different AAD cutoffs in the prediction of aortic dissection or aneurysm based on normal population variation in AAD, with and without adjustment for age, sex, height and BSA. MethodsUsing cardiac magnetic resonance (CMR) images of UK Biobank participants, we applied a pre-trained neural network to segment and measure AAD. We evaluated four approaches to predict a subsequent clinical diagnosis of thoracic aortic dissection or presence of a thoracic aneurysm. We evaluated (i) the performance of the guideline endorsed AAD cutoff of 4.5 cm, and (ii) cutoffs based on the standardized deviation from the population mean aortic diameter (Z-score): a) without accounting for age, sex or anthropometric measures (ZAAD); b) accounting for age, height, and sex (ZH) and c) accounting for age, body surface area (BSA) and sex (ZBSA). We assessed performance using the detection rate (DR) and false positive rate (FPR). ResultsWe measured AAD from CMR images of 77,527 participants from UK Biobank, among whom there were 72 subsequent diagnoses of thoracic dissection or aneurysm (incidence rate [~]0.1%, approximately 1 in 1100). The AAD was 4.2 cm (SD 0.6) in affected individuals and 3.3 cm (SD 0.4) in unaffected individuals (p<0.001). The guideline endorsed 4.5 cm absolute AAD cutoff had a DR of 25% at a 0.3% FPR. A cutoff of ZAAD>2 detected 57% of affected individuals at a 3% FPR, while ZH>2 and ZBSA>2 had DRs of 53% and 47% respectively, with FPRs of 4% and 2%, respectively. The area under the ROC curve (AUC) was 0.90 for ZAAD, 0.84 for ZH and 0.83 for ZBSA. ConclusionThe current guideline-recommended absolute AAD cutoff of 4.5 cm has a low FPR but a low DR, missing nearly three-quarters of individuals who later developed thoracic aneurysm or dissection. Z-score-based approaches, particularly the ZAAD, improve the DR at the expense of a higher FPR. These findings support the consideration of population-derived Z-scores in clinical monitoring and risk stratification for thoracic aortas at risk of dissection.