The AORTA Gene score for detection and risk stratification of ascending aortic dilation

Abstract

Background and aims: This study assessed whether a model incorporating clinical features and a polygenic score for ascending aortic diameter would improve diameter estimation and prediction of adverse thoracic aortic events over clinical features alone.

Methods: Aortic diameter estimation models were built with a 1.1 million-variant polygenic score (AORTA Gene) and without it. Models were validated internally in 4394 UK Biobank participants and externally in 5469 individuals from Mass General Brigham (MGB) Biobank, 1298 from the Framingham Heart Study (FHS), and 610 from All of Us. Model fit for adverse thoracic aortic events was compared in 401 453 UK Biobank and 164 789 All of Us participants.

Results: AORTA Gene explained more of the variance in thoracic aortic diameter compared to clinical factors alone: 39.5% (95% confidence interval 37.3%-41.8%) vs. 29.3% (27.0%-31.5%) in UK Biobank, 36.5% (34.4%-38.5%) vs. 32.5% (30.4%-34.5%) in MGB, 41.8% (37.7%-45.9%) vs. 33.0% (28.9%-37.2%) in FHS, and 34.9% (28.8%-41.0%) vs. 28.9% (22.9%-35.0%) in All of Us. AORTA Gene had a greater area under the receiver operating characteristic curve for identifying diameter ≥ 4 cm: 0.836 vs. 0.776 (P < .0001) in UK Biobank, 0.808 vs. 0.767 in MGB (P < .0001), 0.856 vs. 0.818 in FHS (P < .0001), and 0.827 vs. 0.791 (P = .0078) in All of Us. AORTA Gene was more informative for adverse thoracic aortic events in UK Biobank (P = .0042) and All of Us (P = .049).

Conclusions: A comprehensive model incorporating polygenic information and clinical risk factors explained 34.9%-41.8% of the variation in ascending aortic diameter, improving the identification of ascending aortic dilation and adverse thoracic aortic events compared to clinical risk factors.

Keywords: All of Us; Aortic dissection; Ascending aorta; Framingham Heart Study; Polygenic score; UK Biobank.

Structured Graphical Abstract

A clinical model to estimate ascending aortic diameter was developed in 44 420 UK Biobank participants with cardiovascular magnetic resonance imaging, with the arrows depicting interaction terms between the clinical risk factors. A polygenic score derived from a genome-wide association study of 39 524 participants was incorporated based on data from 4896 participants to produce the AORTA Gene model. The AORTA Gene model was used to estimate ascending aortic diameter in UK Biobank, All of Us, Framingham Heart Study, and Mass General Brigham and to predict adverse thoracic aortic outcomes, including thoracic aortic dissection, in UK Biobank and All of Us. The aortic drawings are derived from Servier Medical Art, CC BY 4.0. Height image by Fengquan Li, aging image by Adrien Coquet, blood pressure cuff image by Luis Prado, scale image by Gacem Tachfin, and DNA image by TkBt, from thenounproject.com, CC BY 3.0. FHS, Framingham Heart Study; GWAS, genome-wide association study; MGB, Mass General Brigham.

Figure 1

Score derivation and variance explained. (A) The AORTA Gene model was derived from clinical measurements and a polygenic score derived from a genome-wide association study of samples that did not overlap with the internal UK Biobank validation set. The graph connecting different risk factors represents the ability for the model to use statistical interactions to estimate aortic diameter. (B) For each cohort, the variance in ascending aortic diameter explained by the model derived from age and sex (left), the clinical model (AORTA Score, middle), and the clinical model incorporating the polygenic score (AORTA Gene, right) is depicted for the internal validation set and the three external validation sets. FHS, Framingham Heart Study; MGB, Mass General Brigham. Pictograms are copyright © Noun Project

Figure 2

Study cohorts. The sample diagram depicts subsets of people. UK Biobank participants with magnetic resonance imaging data were split by random ID into a genome-wide association study group (blue, N = 39 524) and a residual group. Individuals in the residual group related within three degrees of kinship were removed (N = 557), and the residual group was then split into a reweighting group (purple, N = 4896) and an internal validation group (yellow, N = 4394 after excluding 568 participants with missing data). As depicted in the legend, the genome-wide association study and PRS development was conducted in the genome-wide association study group (blue, N = 39 524). Clinical model development was conducted in the genome-wide association study and reweighting groups (blue and purple, N = 44 420). The PRS was incorporated into the clinical score using a linear model to produce the AORTA Gene model in the reweighting group (purple, N = 4896). All models were validated in the internal validation set (yellow, N = 4962) and the external validation sets (orange): Mass General Brigham , Framingham Heart Study, and All of Us. Additionally, models were validated for electronic health record-based diagnoses of thoracic aortic aneurysm and adverse thoracic aortic events in UK Biobank (N = 401 453) and All of Us (164 668; orange). FHS, Framingham Heart Study; GWAS, genome-wide association study; MGB, Mass General Brigham; MRI, magnetic resonance imaging

Figure 3

Impact of leaving variables out of the model. A simplified linear model for aortic diameter having the AORTA Gene main terms (but no interaction terms, and no body mass index term to allow weight and height to be independent) was fitted in the training set while leaving out one predictor and assessed in the UK Biobank validation set. This was repeated for each predictor in the AORTA Gene model. The point estimate and 95% confidence interval for the relative change in variance explained (R²) as a percentage of the variance explained by the full model (y axis) are shown after the removal of one predictor (x axis). The right-sided y axis shows absolute change in variance explained on a 0–1 scale compared to the full model. The horizontal black line at y = 0 represents an identical variance explained compared to the full model

Figure 4

Score calibration curves in the UK Biobank validation set. Counterclockwise from top left: AORTA Gene, followed by the clinical AORTA Score, the age and sex model, and the age/sex/genetics model. The x axis represents the estimated ascending aortic diameter; the y axis represents the measured ascending aortic diameter; both are truncated at the same points so that the x and y axes have the same span. Each point represents one of the 4962 UK Biobank internal validation set participants; orange points represent women while navy points represent men. The blue line shows the smoothed average value and only extends along the x axis to the limits of the observed data. The black line shows the line of ideal calibration, where a 1 cm greater score would be met by a 1 cm greater aortic diameter. For all plotted scores, the slopes were statistically indistinguishable from one and the intercepts were statistically indistinguishable from zero as detailed in the Results

Figure 5

Receiver operating characteristic curves. Receiver operating characteristic curves for the AORTA Gene model (red) and the clinical AORTA Score model (blue) in the validation cohorts. The dashed diagonal line represents the no-information baseline. Counterclockwise from top left: UK Biobank internal validation, Mass General Brigham, Framingham Heart Study, and All of Us. The area under the receiver operating characteristic curve for detecting aortic diameter ≥ 4 cm for the AORTA Gene model in these cohorts was, respectively, 0.836, 0.808, 0.856, and 0.827. The respective area under the receiver operating characteristic curve for the clinical AORTA Score was 0.776, 0.767, 0.818, and 0.791. FHS, Framingham Heart Study; MGB, Mass General Brigham