Development and Evaluation of a Comprehensive Prediction Model for Incident Coronary Heart Disease Using Genetic, Social, and Lifestyle-Psychological Factors: A Prospective Analysis of the UK Biobank

Abstract

Background: Clinical risk calculators for coronary heart disease (CHD) do not include genetic, social, and lifestyle-psychological risk factors.

Objective: To improve CHD risk prediction by developing and evaluating a prediction model that incorporated a polygenic risk score (PRS) and a polysocial score (PSS), the latter including social determinants of health and lifestyle-psychological factors.

Design: Cohort study.

Setting: United Kingdom.

Participants: UK Biobank participants recruited between 2006 and 2010.

Measurements: Incident CHD (myocardial infarction and/or coronary revascularization); 10-year clinical risk based on pooled cohort equations (PCE), Predicting Risk of cardiovascular disease EVENTs (PREVENT), and QRISK3; PRS (Polygenic Score Catalog identification: PGS000018) for CHD (PRS_CHD); and PSS_CHD from 100 related covariates. Machine-learning and time-to-event analyses and model performance indices.

Results: In 388 224 participants (age, 55.5 [SD, 8.1] years; 42.5% men; 94.9% White), the hazard ratio for 1 SD increase in PSS_CHD for incident CHD was 1.43 (95% CI, 1.38 to 1.49; P < 0.001) and for 1 SD increase in PRS_CHD was 1.59 (CI, 1.53 to 1.66, P < 0.001). Non-White persons had higher PSS_CHD than White persons. The effects of PSS_CHD and PRS_CHD on CHD were independent and additive. At a 10-year CHD risk threshold of 7.5%, adding PSS_CHD and PRS_CHD to PCE reclassified 12% of participants, with 1.86 times higher CHD risk in the up- versus down-reclassified persons and showed superior performance compared with PCE as reflected by improved net benefit while maintaining good calibration relative to the clinical risk calculators. Similar results were seen when incorporating PSS_CHD and PRS_CHD into PREVENT and QRISK3.

Limitation: A predominantly White cohort; possible healthy participant effect and ecological fallacy.

Conclusion: A PSS_CHD was associated with incident CHD and its joint modeling with PRS_CHD improved the performance of clinical risk calculators.

Primary funding source: National Human Genome Research Institute.

Figure 1.. Study design.

CHD = coronary heart disease; PCE = pooled cohort equations; PREVENT = Predicting Risk of cardiovascular disease EVENTs; PRS_{CHD =} polygenic risk score for CHD; PSS_{CHD =} polysocial score for CHD; UKB = UK Biobank. * The numbers for each exclusion item may overlap.

Figure 2.. Coefficients of components of PSS_CHD for CHD risk prediction.

In the elastic net regression, age, sex, self-reported White ethnicity, and components of PSS_CHD were incorporated, and the coefficients obtained from the training cohort and used to calculate PSS_CHD in the test and final cohorts are presented. To enhance clarity, we present the 50 most important variables with the highest coefficients in the training cohort, excluding age (coefficient = 0.54), sex (coefficient = 0.52), and self-reported White ethnicity (coefficient = 0.01) from the figure. AS = advanced subsidiary; CHD = coronary heart disease; DIY= do-it-yourself; GBP = British pound sterling; HNC = higher national certificate; HND = higher national diploma; NVQ = national vocational qualification; PSS_CHD = polysocial score for CHD.

Figure 3.. Comparison of CHD risk calculated by PCE versus CRS_PCE considering 7.5% as the actionable threshold.

The figure shows that participants up-reclassified had 86% higher risk for CHD compared with those down-reclassified (hazard ratio, 1.86 [95% CI, 1.32 to 2.62]; P < 0.001). A sensitivity analysis considering 10% as the actionable threshold is presented in Supplement Figure 15 (available at Annals.org). Similar comparisons of CHD risk using PREVENT versus CRS_PREVENT and QRISK3 versus CRS_QRISK3 are presented in Supplement Figures 16 to 19 (available at Annals.org). CHD = coronary heart disease; CRS_PCE = comprehensive risk score based on PCE; CRS_PREVENT = comprehensive risk score based on PREVENT; CRS_QRISK3 = comprehensive risk score based on QRISK3; PCE = pooled cohort equations; PREVENT = Predicting Risk of cardiovascular disease EVENTs.

Figure 4.. Decision curves of a comprehensive prediction model compared with the PCE, across 2 actionable thresholds.

This figure shows the net benefit of using PCE versus CRS_PCE for prevention of adverse events across the 2 actionable thresholds. The vertical dashed lines mark the key decision thresholds for 10-year CHD risk at 7.5% and 10%. The black line represents the strategy of preventing no adverse event, resulting in a net benefit of 0. This approach avoids any preventive interventions but fails to address persons at high risk, leading to no net benefit. The dotted-and-dashed line depicts the strategy of preventing adverse events in everyone, which maximizes net benefit at very low thresholds due to the high number of events identified. However, as the threshold increases, the net benefit rapidly declines to 0 due to unnecessary preventive interventions and associated harms or costs in low-risk persons. The dashed line shows the net benefit of using the PCE to guide preventive strategies, which consistently show positive net benefit across actionable thresholds. The dotted line represents CRS_PCE, providing higher net benefit than the PCE at each threshold, suggesting superior clinical utility in guiding prevention strategies. CHD = coronary heart disease; CRS_PC_E = comprehensive risk score based on PCE; PCE = pooled cohort equations.