Abstract
Background
Genetic risk for coronary artery disease (CAD) can be estimated using polygenic risk scores (PRS), but greater clarity is needed on how PRS might inform clinical risk assessment and prevention. We assessed the risk for CAD jointly conferred by LDL-C and CAD PRS relative to established treatment thresholds.
Methods
This study followed 257,158 UK Biobank (UKBB) participants and 67,668 from the All of Us Research Program (AoU), a longitudinal U.S. cohort, without prior CAD, stroke, or diabetes. In UKBB, Cox proportional hazards models estimated CAD hazard ratios for each LDL-C × CAD PRS stratum relative to individuals with LDL-C < 100 mg/dL. The severe hypercholesterolemia (LDL-C ≥ 190 mg/dL) group, the guideline-designated threshold for statin initiation, served as a benchmark against which groups were compared. Parallel analyses were performed in AoU cross-sectionally.
Results
Over a median (IQR) follow-up of 13.5 (12.8–14.2) years, 13,886 (5.4%) participants developed CAD in UKBB. Higher CAD PRS was associated with CAD risk that was comparable to, or in some strata exceeding, that associated with LDL-C ≥ 190 mg/dL by pairwise Wald tests at progressively lower LDL-C concentrations in both cohorts.
Conclusions
Individuals with high genetic risk and moderate LDL-C elevations, representing 14.8% of the study population, have CAD risk comparable to or greater than that of SH. Consideration of dynamic LDL-C thresholds among those with high CAD PRS may identify individuals with risk sufficiently high to warrant preventive therapy.
Keywords: Polygenic risk score, Low-density lipoprotein cholesterol, Coronary artery disease
1. Introduction
Coronary artery disease (CAD) remains the leading cause of premature death among adults globally [1]. The current paradigm of primary prevention of CAD centers on forecasting patients’ 10-year risk of developing atherosclerotic cardiovascular disease (ASCVD) using validated clinical risk scores to identify high-risk individuals for targeted intervention [[2], [3], [4]]. While recent versions of these calculators have expanded to account for an increasing number of clinical risk factors [5,6], genetic predisposition remains unaccounted for in either these risk models or clinical management guidelines to date [7].
The development of polygenic risk scores (PRS), which aggregate many risk-conferring alleles into a single metric, has now made it possible to meaningfully quantify genetic risk for CAD [8,9]. Recent CAD PRS have demonstrated improved risk prediction relative to validated clinical risk prediction scores alone [[10], [11], [12], [13]]. While PRS are increasingly available for clinical use [[14], [15], [16]], how they should be incorporated into clinical decision-making remains unclear [15,17,18], underscoring the need for better characterization of PRS interactions with other risk factors in determining CAD risk.
In this study, we assessed the joint risk conferred by low-density lipoprotein cholesterol (LDL-C) and CAD PRS compared to current LDL-C–based treatment thresholds. Prior work has shown that the effect of LDL-C on CAD risk is amplified in those with higher PRS [13,19]. We hypothesized that individuals with high PRS and moderately elevated LDL-C may have CAD risk comparable to those with LDL-C ≥ 190 mg/dL, the threshold for lipid-lowering therapy initiation by LDL-C alone per guidelines [2,20,21]. Thus, we sought to establish an LDL-C–by–CAD PRS nomogram to better risk-stratify individuals with non-severe LDL-C elevations who might warrant comparable intervention.
2. Methods
2.1. Study population
The UK Biobank (UKBB) is a population-based cohort of approximately 500,000 adults recruited between 2006 and 2010 and followed via national health records [22]. Secondary analyses of UKBB data under Application 7089 were approved by the Massachusetts General Hospital Institutional Review Board. From 501,940 UKBB participants, we excluded individuals with prior CAD or stroke (n = 30,876), lipid-lowering therapy use (n = 62,374), missing genetic data or inclusion in the GPSMult training set (n = 106,147), missing LDL-C (n = 14,830), type 2 diabetes (n = 10,478), or missing covariates (n = 20,077), including age, sex, the first 10 principal components of ancestry, smoking status, systolic blood pressure, and BMI. The resulting cohort included 257,158 participants, with subgroup analyses performed in 16,722 individuals of genetically inferred non-European ancestry and 240,436 individuals of genetically inferred European ancestry. Cohort construction details are provided in Supplementary Fig. 1.
For external replication, we used the NIH All of Us Research Program (AoU), a longitudinal U.S. cohort designed to include populations historically under-represented in biomedical research. Detailed study protocols, including eligibility and genomic data curation, have been described previously [23,24]. The Institutional Review Boards of the All of Us Research Program approved all study procedures and informed consent was obtained from all participants. From the Curated Data Repository version 7 (CDR v7), we filtered the AoU cohort from 413,456 to 67,668 individuals by excluding those with missing LDL-C values (n = 272,962), missing polygenic risk scores (n = 38,976), type 2 diabetes (n = 28,102), or missing covariates including smoking status, systolic blood pressure, body mass index, or the first 10 genetic principal components (n = 5,748). Details of this cohort construction are visualized in Supplementary Fig. 2.
2.2. Low-density lipoprotein cholesterol (LDL-C)
LDL-C was measured at baseline in UKBB and defined in AoU as the highest recorded clinical value. For the analysis including individuals taking lipid-lowering medications, LDL-C was divided by 0.7 in both cohorts to approximate average imputed effect of lipid-lowering medications [25]. Lipid-lowering medications included all ATC code C10 compounds except omega-3 triglycerides.
2.3. Genotyping and quality control
Participants in UKBB were genotyped using either the UK BiLEVE Axiom array or the UK Biobank Axiom array [22]. Genotypes were imputed centrally by UKBB using phased haplotypes and reference panels from the Haplotype Reference Consortium (HRC) [26,27] and a merged UK10K + 1000 Genomes Phase 3 panel [24]. Quality control (QC) measures included filtering out variants with poor call rates, deviations from Hardy-Weinberg equilibrium, or significant genotype frequency differences between sexes.
Participants in AoU underwent high-coverage whole genome sequencing [26]. Variant calling was performed jointly across samples using the Genome Analysis Toolkit and aligned to the GRCh38 reference genome. QC measures included filtering out samples with sex discordance, a cross-individual contamination rate > 3%, and a call rate < 98%.
2.4. Genetically inferred ancestry
Genetic ancestry was inferred for UKBB participants by us and for AoU participants by All of Us Genome Centers and Data and Research Center. A random forest classifier, trained on reference samples from 1000 Genomes Project [26] (UKBB) and Human Genome Diversity Project [28] (AoU), assigned each sample to one of the ancestries using the first 10 (UKBB) or 16 (AoU) genetic principal components as features.
2.5. Polygenic risk score
We utilized the genome-wide polygenic score for multiple ancestries (GPSMult) developed by Patel et al., [12] which was constructed using genome-wide association study data spanning five major ancestral populations (African, East Asian, European, Hispanic, and South Asian) and ten CAD risk factors. GPSMult has demonstrated enhanced predictive performance across all ancestries, outperforming all previously published CAD PRS. Multiple newer CAD PRS are largely comparable [29,30].
2.6. Clinical endpoints
Incident CAD in UKBB was defined using ICD-10 and OPCS-4 hospital records plus doctor-diagnosed and self-reported MI/revascularization codes, as detailed in Supplementary Table 1. In survival analyses, follow-up was censored at 15 years.
In AoU, CAD was defined as the presence of at least two diagnosis codes or at least one procedural code of the ICD-9/10, CPT-4, HCPCS, DRG, and SNOMED terminology codes listed in Supplementary Table 2, as previously described [31].
2.7. External replication analysis
To assess external validity, we performed parallel analyses in AoU, with adaptations made to account for differences in dataset parameters. Given the relative recency of AoU, when excluding individuals with prior CAD, the median follow-up length was only 1.95 (IQR, 0.67–2.74) years (versus UKBB’s median of 13.5 years). To address this, we used a logistic regression, rather than a Cox proportional hazards, model including both incident and prevalent CAD cases with the same covariates used in the primary analysis. Additionally, to maintain sample size and thus optimize power, we adjusted for, rather than excluding, lipid-lowering medication usage.
2.8. Statistical analysis
Individuals were stratified by combinations of LDL-C and CAD PRS ranges. LDL-C was categorized as < 100, 100–130, 130–160, 160–190, and ≥ 190 mg/dL. PRS was stratified into < 50th, 50th–70th, 70th–80th, 80th–90th, and 90th–100th percentiles. Cox proportional hazards models estimated HRs for incident CAD for each LDL-C × PRS stratum compared to a reference group containing all individuals with LDL-C < 100 mg/dL (0–100th percentile PRS). Models were adjusted for age, sex, first 10 genetic principal components of ancestry, smoking status, systolic blood pressure, and BMI. The proportional hazards assumption was graphically inspected by log-minus-log plot.
To provide absolute risk estimates, incidence rates per 1,000 person-years and 95% confidence intervals were estimated using exact Poisson methods within each LDL-C × PRS stratum.
To assess potential incremental clinical utility of adding CAD PRS beyond clinical covariates, we compared a clinical Cox model versus a model additionally including CAD PRS using discrimination (Harrell’s C-index) and a 10-year decision curve analysis in UKBB evaluating net benefit across clinically relevant risk thresholds; for the 10-year analysis, we evaluated predicted 10-year risk among participants with ≥ 10 years of follow-up or a CAD event within 10 years.
To formally compare CAD risk across LDL-C × CAD PRS strata relative to the severe hypercholesterolemia (SH) benchmark, we conducted post hoc pairwise contrasts of model-based estimates using two-sided Wald tests (on log HRs in UKBB and log odds ratios [ORs] in AoU) using the ‘emmeans’ package in R [32].
All analyses were conducted using R, version 4.3 (R Project for Statistical Computing). This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.
3. Results
3.1. Findings in UKBB
3.1.1. Baseline characteristics
In the primary analysis cohort, UKBB, 257,158 individuals were followed, of whom 13,886 developed CAD over a median (IQR) follow-up period of 13.5 (12.8–14.2) years. At enrollment, mean (standard deviation [SD]) age was 55.2 (8.1) years, 150,066 (58.4%) of participants were female, mean (SD) BMI was 26.8 (4.4) kg/m², 24,767 (9.6%) were current smokers, and 84,401 (32.8%) were previous smokers. Mean LDL-C was 143.5 mg/dL with an SD of 31.4 mg/dL. The reference group in LDL-C–by–CAD PRS analyses, comprising all individuals with LDL-C < 100 mg/dL (n = 18,020; 7.0%), had a mean CAD PRS corresponding to the 40.3rd percentile with an SD of 28.7. In contrast, individuals with LDL-C ≥ 190 mg/dL (n = 19,418; 7.6%) had a mean CAD PRS corresponding to the 55.2nd percentile (SD = 27.9). Baseline demographic and clinical characteristics of participants who developed incident CAD versus those who did not develop incident CAD in UKBB are shown in Table 1.
Table 1.
Baseline characteristics by Incident coronary artery disease status in the UK Biobank.
| Characteristic | Without incident CAD (n = 243,272) | Developed incident CAD (n = 13,886) | P value |
|---|---|---|---|
| Age, y, mean (SD) | 55.0 (8.1) | 59.3 (7.2) | <0.001 |
| Sex, No. (%) | |||
| Female | 144,783 (59.5) | 5,283 (38.0) | <0.001 |
| Male | 98,489 (40.5) | 8,603 (62.0) | |
| Genetically Inferred Ancestry,* No. (%) | |||
| Admixed American | 2,705 (1.1) | 153 (1.1) | <0.001 |
| African | 6,200 (2.5) | 184 (1.3) | |
| East Asian | 1,813 (0.7) | 43 (0.3) | |
| European | 227,310 (93.4) | 13,126 (94.5) | |
| South Asian | 5,244 (2.2) | 380 (2.7) | |
| Smoking Status, No. (%) | |||
| Current | 22,735 (9.3) | 2,032 (14.6) | <0.001 |
| Previous | 79,115 (32.5) | 5,286 (38.1) | |
| Never | 141,422 (58.1) | 6,568 (47.3) | |
| Body Mass Index, mean (SD) | 26.7 (4.4) | 27.7 (4.4) | <0.001 |
| Systolic Blood Pressure, mmHg, mean (SD) | 137.8 (19.4) | 145.8 (19.9) | <0.001 |
| Diastolic Blood Pressure, mmHg, mean (SD) | 81.9 (10.7) | 84.7 (10.8) | <0.001 |
| Total Cholesterol, mg/dL, mean (SD) | 227.7 (40.8) | 236.1 (41.8) | <0.001 |
| LDL-C, mg/dL, mean (SD) | 143.0 (31.3) | 151.9 (31.9) | <0.001 |
| HDL-C, mg/dL, mean (SD) | 57.9 (14.7) | 53.4 (13.8) | <0.001 |
| Triglycerides, mg/dL, median (IQR) | 123.9 (88.1–178.7) | 149.3 (106.6–213.7) | <0.001 |
| CAD PRS percentile, mean (SD) | 47.1 (28.6) | 57.3 (28.3) | <0.001 |
Baseline demographic and clinical characteristics of participants who developed incident CAD versus those who did not develop incident CAD within the filtered UK Biobank cohort. Continuous variables were compared between groups using t-test for normally distributed continuous variables, Wilcoxon rank sum test for skewed continuous variables, and χ2 test for categorical variables. Total cholesterol measurements were available for 243,223 individuals who did not develop CAD and 13,882 individuals who developed CAD. HDL-C measurements were available for 222,078 individuals who did not develop CAD and 12,701 individuals who developed CAD. Triglyceride measurements were available for 243,120 individuals who did not develop CAD and 13,880 individuals who developed CAD. CAD indicates coronary artery disease; HDL-C, high-density lipoprotein cholesterol; IQR, interquartile range; LDL-C, low-density lipoprotein cholesterol; PRS, polygenic risk score; and SD, standard deviation.
Genetically inferred ancestry was assigned using genetic data and reflects broad ancestry groupings; therefore, these categories may not correspond to participants’ self-identified race or ethnicity.
3.1.2. Individual effects of LDL-C and CAD PRS on incident CAD risk
Relative to individuals with LDL-C < 100 mg/dL, those in the following LDL-C ranges (mg/dL) had the following HRs (95% confidence intervals [CIs]): 100–130, HR of 0.95 (95% CI, 0.87–1.04); 130–160, HR of 1.09 (95% CI, 1.00–1.19); 160–190, HR of 1.20 (95% CI, 1.10–1.32); and ≥ 190, HR of 1.48 (95% CI, 1.34–1.63). CAD PRS was associated with incident CAD with an HR of 1.49 (95% CI, 1.47–1.52) per SD. The mean (SD) CAD PRS percentile was 57.3 (28.3) among participants who developed incident CAD and 47.1 (28.6) among those who did not. The C-index for incident CAD from a Cox model including age, sex, and clinical risk factors (LDL-C, systolic blood pressure, body mass index, smoking status, and genetic principal components) was 0.723 (95% CI, 0.719–0.727). With the addition of CAD PRS, the C-index increased to 0.744 (95% CI, 0.740–0.748). In 10-year decision curve analysis, adding CAD PRS yielded a small increase in net benefit compared with the clinical model at thresholds of 5% and 7.5% 10-year risk (Δ net benefit ≈0.0016 at both thresholds).
3.1.3. Primary analyses for the risk of incident CAD by combined LDL-C and PRS strata
After stratifying participants by LDL-C and CAD PRS ranges, we obtained incident CAD HR estimates shown in Fig. 1A and Table 3A. Absolute incidence rates per 1,000 person-years (95% Poisson confidence intervals) for each LDL-C × PRS stratum are reported alongside HRs in Table 3A. A statistically significant but numerically very modest interaction effect was observed between continuous LDL-C and CAD PRS (HRinteraction, 1.0013 [95% CI, 1.0007–1.0018] per 1 mg/dL LDL-C change × per 1 percentile CAD PRS change; P for interaction = 4 × 10−6). Using LDL-C < 100 mg/dL as the reference group, the LDL-C ≥ 190 mg/dL stratum (SH) had an HR of 1.81 (95% CI, 1.65–2.00). Several LDL-C × CAD PRS strata demonstrated CAD risk that was not significantly lower than that of the SH group, including individuals with LDL-C 160–190 mg/dL and PRS in the 70th–80th percentile (HR, 1.76 [95% CI, 1.55–1.99]; P for the difference compared with SH [PvsSH], 0.57; 2.2% of cohort) and 80th–90th percentile (HR, 1.97 [95% CI, 1.75–2.22]; PvsSH, 0.10; 2.2% of cohort), as well as individuals with LDL-C 130–160 mg/dL and PRS in the 80th–90th percentile (HR, 1.80 [95% CI, 1.61–2.02]; PvsSH, 0.89; 3.4% of cohort), and those with LDL-C 100–130 and 90th–100th percentile PRS (HR, 1.86 [95% CI, 1.61–2.14]; PvsSH, 0.74; 1.9% of cohort).
Fig. 1A.
Hazard Ratios for Incident CAD by LDL-C and CAD PRS Strata in the UK Biobank.
Hazard ratios for incident coronary artery disease (CAD) stratified by low-density lipoprotein cholesterol (LDL-C) ranges and CAD polygenic risk score (PRS) strata in the UK Biobank cohort. Hazard ratios are quantified relative to individuals with LDL-C < 100 mg/dL and any CAD PRS. Error bars represent 95% confidence intervals.
Table 3A.
Hazard ratios and incidence rates for CAD by LDL-C and CAD PRS strata.
| PRS Percentile | LDL-C (mg/dL) |
|||
|---|---|---|---|---|
| 100–130 | 130–160 | 160–190 | ≥ 190 | |
| 90th–100th |
1.86 [1.61–2.14] 4,928 / 299 4.66 [4.15–5.22] |
2.55 [2.29–2.84] 7,819 / 776 7.75 [7.22–8.32] |
2.84 [2.53–3.18] 5,427 / 624 9.07 [8.37–9.81] |
1.81 [1.65–2.00] 19,418 / 1,639 6.57 [6.26–6.90] |
| 80th–90th |
1.35 [1.16–1.56] 5,672 / 269 3.61 [3.19–4.07] |
1.80 [1.61–2.02] 8,714 / 657 5.84 [5.40–6.30] |
1.97 [1.75–2.22] 5,704 / 502 6.83 [6.25–7.46] |
|
| 70th–80th |
1.34 [1.17–1.54] 6,041 / 300 3.79 [3.37–4.24] |
1.57 [1.40–1.76] 9,091 / 622 5.27 [4.86–5.70] |
1.76 [1.55–1.99] 5,578 / 451 6.28 [5.71–6.89] |
|
| 50th–70th |
1.11 [0.99–1.25] 13,297 / 568 3.25 [2.99–3.53] |
1.31 [1.18–1.45] 19,112 / 1,148 4.61 [4.34–4.88] |
1.46 [1.31–1.62] 11,133 / 776 5.39 [5.02–5.79] |
|
| 0–50th |
0.81 [0.74–0.90] 41,706 / 1,384 2.51 [2.38–2.65] |
0.86 [0.79–0.95] 50,132 / 2,080 3.16 [3.03–3.30] |
0.95 [0.86–1.05] 25,366 / 1,222 3.68 [3.48–3.89] |
|
Format for each cell: hazard ratio (95% confidence interval), number of participants / CAD events, and incidence rate per 1,000 person-years (95% confidence interval). Hazard ratios are estimated relative to participants with LDL-C < 100 mg/dL. Red cells indicate strata whose hazard was not significantly lower than, or exceeded, that observed in participants with LDL-C ≥ 190 mg/dL, based on formal pairwise comparisons. CAD indicates coronary artery disease; LDL-C, low-density lipoprotein cholesterol; and PRS, polygenic risk score.
At higher PRS levels, CAD risk significantly exceeded that of the SH benchmark. Individuals in the 90th–100th percentile at LDL-C 130–160 mg/dL (HR, 2.55 [95% CI, 2.29–2.84]; PvsSH, 6.9 × 10−15; 3.0% of cohort) and 160–190 mg/dL (HR, 2.84 [95% CI, 2.53–3.18]; PvsSH, 2.6 × 10−21; 2.1% of cohort) exhibited significantly higher risk than the SH benchmark. Collectively, these LDL-C × CAD PRS strata comprised approximately 14.8% of the cohort and exhibited CAD risk that was not significantly lower than, and in some cases significantly higher than, that associated with SH.
3.1.4. Subgroup analyses
To assess generalizability across ancestries, we conducted subgroup analyses in non-European ancestry (non-EUR; n = 16,722) and European ancestry (EUR; n = 240,436) participants. In the non-EUR cohort, using LDL-C < 100 mg/dL as the reference group, the LDL-C ≥ 190 mg/dL stratum (SH) had a HR of 2.50 (95% CI, 1.69–3.71). Despite attenuated PRS performance, all but one (LDL-C 100–130 mg/dL with PRS in the 90th–100th percentile) of the LDL-C × CAD PRS strata identified in the primary analysis continued to show risk not statistically different from the SH benchmark. One additional stratum, LDL-C 130–160 mg/dL with PRS in the 70th–80th percentile, was not statistically different in risk from SH. Across most strata, confidence intervals were substantially wider than in the primary analysis, consistent with smaller sample size and reduced PRS performance in non-EUR populations (Supplementary Table 3 and Supplementary Fig. 3). In contrast, the EUR-only analysis reproduced the risk-equivalence patterns observed in the primary analysis (Supplementary Table 4 and Supplementary Fig. 4).
Given known interactions between age and the effects of both LDL-C [33,34] and PRS [35,36] on CAD, we also conducted age-stratified sub-analyses in UKBB (Supplementary Table 5 and Supplementary Fig. 5). For each age group, we evaluated LDL-C × PRS strata relative to individuals with LDL-C < 100 mg/dL within that same age range and compared LDL-C × PRS strata against those with LDL-C ≥ 190 mg/dL within the same age range. Among participants aged < 50 years, CAD risk was not significantly lower than SH among individuals with LDL-C 100–190 mg/dL and PRS ≥ 90th percentile, as well as among those with LDL-C 160–190 mg/dL and PRS in the 80th–90th percentile. In the 50–60-year group, a broader set of strata exhibited CAD risk not significantly lower than SH, including individuals with LDL-C 100–130 mg/dL and PRS ≥ 90th percentile, those with LDL-C 130–160 mg/dL and PRS ≥ 80th percentile, and those with LDL-C 160–190 mg/dL and PRS ≥ 70th percentile, with CAD risk exceeding the SH benchmark among those with LDL-C 160–190 mg/dL and PRS ≥ 90th percentile. Among participants aged ≥ 60 years, CAD risk not significantly lower than SH extended further across LDL-C strata, encompassing LDL-C 100–130 mg/dL with PRS ≥ 70th percentile, LDL-C 130–160 mg/dL with PRS ≥ 70th percentile, and LDL-C 160–190 mg/dL with PRS ≥ 50th percentile, while CAD risk exceeded that of SH among those with LDL-C 130–190 mg/dL and PRS ≥ 90th percentile. Collectively, these findings demonstrate a progressive age-dependent expansion of LDL-C × PRS combinations associated with SH-comparable or greater CAD risk.
3.1.5. Sensitivity analyses
We performed three more sensitivity analyses within the UKBB: 1) adjusting for, rather than excluding by, lipid-lowering therapy; 2) applying a logistic regression model to evaluate both incident and prevalent CAD in the same model; and 3) redefining the benchmark group to participants with LDL-C ≥ 190 mg/dL and CAD PRS in the 40th–60th (rather than 0–100th) percentile. In all three sensitivity analyses, the same high-risk LDL-C × PRS strata observed in the primary analysis were identified (Supplementary Tables 6–8 and Supplementary Figs. 6–8). Sensitivity analysis (3) yielded additional strata whose HRs were not significantly lower than SH, reflecting the lower comparison risk that results from restricting the benchmark group. That is, when the benchmark is restricted to participants with LDL-C ≥ 190 mg/dL and PRS in the 40th–60th percentile, the comparison hazard ratio (HR = 1.57) is lower than in the primary analysis, where the benchmark included all PRS percentiles (HR = 1.81). This difference reflects the higher mean CAD PRS among participants with LDL-C ≥ 190 mg/dL compared with those with LDL-C < 190 mg/dL (55.2nd percentile [SD, 27.9] vs 40.7th percentile [SD, 28.7]). As a result, the primary benchmark reflects very high LDL in a group modestly enriched for higher PRS, whereas restricting the benchmark to the 40th–60th percentile attenuates this enrichment, lowers the comparison HR, and allows additional LDL-C × PRS strata with HRs not lower than SH.
3.2. External replication in AoU
Replication analysis was conducted on 67,668 individuals enrolled in AoU, of whom 5,344 experienced incident or prevalent CAD. At enrollment, mean (SD) age was 55.0 (16.2) years, 63.9% (n = 43,221) were female, 31.4% (21,262) were of non-European ancestry, mean (SD) BMI was 29.2 (7.1), mean (SD) adjusted LDL-C was 147.3 (59.0) mg/dL, mean (SD) systolic blood pressure was 118.2 (8.6) mmHg, and 38.2% (n = 25,876) of individuals had ever smoked. The reference group, containing all individuals with adjusted LDL-C < 100 mg/dL (n = 13,958; 20.6% of total cohort), had a mean (SD) CAD PRS percentile of 43.7 (28.7). The benchmark group, containing all individuals with adjusted LDL-C ≥ 190 mg/dL (n = 14,514; 21.4% of cohort), had a mean (SD) CAD PRS percentile of 47.5 (27.6). The mean (SD) CAD PRS percentile of those who experienced CAD was 57.4 (27.8) and of those who did not experience CAD was 49.4 (28.9). Baseline demographic and clinical characteristics of AoU participants with CAD (incident or prevalent) versus those without CAD are shown in Table 2.
Table 2.
Baseline characteristics by all coronary artery disease status in the All of Us research program.
| Characteristic | Without CAD (n = 62,324) |
With CAD (incident or prevalent) (n = 5344) |
P value |
|---|---|---|---|
| Age, y, mean (SD) | 54.1 (16.1) | 66.2 (12.6) | <0.001 |
| Sex, No. (%) | |||
| Male | 21,297 (34.2) | 3,150 (58.9) | <0.001 |
| Female | 41,027 (65.8) | 2,194 (41.1) | |
| Genetically Inferred Ancestry,* No. (%) | |||
| Admixed American | 7,903 (12.7) | 427 (8.0) | <0.001 |
| African | 9,522 (15.3) | 833 (15.6) | |
| East Asian | 1,372 (2.2) | 54 (1.0) | |
| European | 42,417 (68.1) | 3,989 (74.6) | |
| Other | 1,110 (1.8) | 41 (0.8) | |
| Ever Smoker, No. (%) | 23,046 (37.0) | 2,830 (53.0) | <0.001 |
| Body Mass Index, mean (SD) | 29.1 (7.1) | 29.3 (6.5) | 0.042 |
| Systolic Blood Pressure, mmHg, mean (SD) | 118.2 (8.6) | 118.5 (9.0) | 0.010 |
| Diastolic Blood Pressure, mmHg, mean (SD) | 77.2 (10.4) | 76.5 (11.6) | <0.001 |
| Total Cholesterol, mg/dL, mean (SD) | 208.9 (45.9) | 204.9 (52.8) | <0.001 |
| LDL-C, mg/dL, mean (SD) | 145.5 (58.0) | 169.0 (65.6) | <0.001 |
| HDL-C, mg/dL, mean (SD) | 63.8 (19.5) | 59.4 (19.1) | <0.001 |
| Triglycerides, mg/dL, median (IQR) | 126.0 (88.0–185.0) | 152.0 (106.0–219.0) | <0.001 |
| CAD PRS percentile, mean (SD) | 49.4 (28.9) | 57.4 (27.8) | <0.001 |
Baseline demographic and clinical characteristics of participants with CAD (incident or prevalent) versus those without CAD within the filtered All of Us Research Program cohort. Continuous variables were compared between groups using t-test for normally distributed continuous variables, Wilcoxon rank sum test for skewed continuous variable, and χ2 test for categorical variables. Diastolic blood pressure measurements were available for 61,311 individuals without CAD and 5271 individuals with CAD. Total cholesterol measurements were available for 55,961 individuals without CAD and 4826 individuals with CAD. HDL-C measurements were available for 57,874 individuals without CAD and 4881 individuals with CAD. Triglyceride measurements were available for 58,685 individuals without CAD and 5082 individuals with CAD. CAD indicates coronary artery disease; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; IQR, interquartile range; PRS, polygenic risk score; and SD, standard deviation.
Genetically inferred ancestry was assigned using genetic data and reflects broad ancestry groupings; therefore, these categories may not correspond to participants’ self-identified race or ethnicity.
Relative to individuals with LDL-C < 100 mg/dL, those in the following LDL-C ranges (mg/dL) exhibited the respective ORs for lifetime CAD development: 100–130, OR of 0.73 (95% CI, 0.65–0.81); 130–160, OR of 0.82 (95% CI, 0.74–0.91); 160–190, OR of 1.08 (95% CI, 0.97–1.21); and ≥ 190, OR of 1.26 (95% CI, 1.15–1.39). CAD PRS was associated with an OR of 1.58 (95% CI, 1.52–1.63) per SD. The AUC for a logistic regression model including age, sex, and clinical risk factors was 0.765 (95% CI, 0.759–0.771). With the addition of CAD PRS, the AUC increased to 0.778 (95% CI, 0.772–0.784).
When participants were jointly stratified by LDL-C and CAD PRS, model-based OR estimates across LDL-C × PRS strata demonstrated patterns directionally consistent with those observed in UKBB, with progressively higher CAD risk at higher PRS levels within each LDL-C range. Formal post-hoc contrasts comparing each LDL-C × PRS stratum directly against the SH benchmark identified all of the same strata as in the primary analysis as carrying risk not significantly lower than that of the SH benchmark group (OR = 1.64), with the addition of individuals with LDL-C 100–130 mg/dL and CAD PRS ≥ 80th percentile (Fig. 1B and Table 3B).
Fig. 1B.
Odds Ratios for All CAD by LDL-C and CAD PRS Strata in the All of Us Research Program.
Odds ratios for coronary artery disease (CAD) stratified by low-density lipoprotein cholesterol (LDL-C) ranges and CAD polygenic risk score (PRS) strata in the All of Us Research Program. Odds ratios are quantified relative to individuals with LDL-C < 100 mg/dL and any CAD PRS. Error bars represent 95% confidence intervals.
Table 3B.
Odds ratios for CAD and counts of participants and events by LDL-C and CAD PRS strata in the All of Us research program.
| PRS Percentile | LDL-C (mg/dL) |
|||
|---|---|---|---|---|
| 100–130 | 130–160 | 160–190 | ≥ 190 | |
| 90th–100th | 1.46 [1.14–1.86] 1,150 / 89 |
1.65 [1.29–2.09] 939 / 93 |
1.88 [1.43–2.47] 535 / 71 |
1.36 [1.24–1.49] 14,514 / 1,937 |
| 80th–90th | 1.08 [0.84–1.38] 1,285 / 82 |
1.54 [1.23–1.92] 1,103 / 106 |
2.14 [1.68–2.72] 653 / 95 |
|
| 70th–80th | 0.87 [0.68–1.12] 1,404 / 75 |
1.04 [0.82–1.31] 1,268 / 91 |
1.68 [1.31–2.14] 707 / 89 |
|
| 50th–70th | 1.00 [0.85–1.18] 3,161 / 197 |
1.15 [0.98–1.35] 2,878 / 226 |
1.09 [0.90–1.31] 1,573 / 150 |
|
| 0–50th | 0.54 [0.47–0.61] 9,954 / 406 |
0.59 [0.52–0.67] 8,327 / 444 |
0.91 [0.80–1.04] 4,259 / 397 |
|
Format for each cell: odds ratio (95% confidence interval), number of participants / CAD events. Odds ratios are estimated relative to participants with LDL-C < 100 mg/dL. Red cells indicate strata whose odds ratio was not significantly lower than, or exceeded, that observed in participants with LDL-C ≥ 190 mg/dL, based on formal pairwise comparisons. This analysis included 67,668 individuals, with 5,344 total cases of CAD (incident or prevalent). CAD indicates coronary artery disease; LDL-C, low-density lipoprotein; and PRS, polygenic risk score.
4. Discussion
In this study, we found that individuals with higher CAD PRS reached CAD risk that was comparable to that of individuals with SH (LDL-C ≥ 190 mg/dL) at progressively lower LDL-C concentrations. Specifically, CAD risk was comparable to the SH benchmark among individuals with LDL-C 100–130 mg/dL and CAD PRS ≥ 90th percentile, those with LDL-C 130–160 mg/dL and CAD PRS ≥ 80th percentile, and those with LDL-C 160–190 mg/dL and CAD PRS ≥ 70th percentile. CAD significantly exceeded the SH benchmark only among individuals with both LDL-C ≥ 130 mg/dL and CAD PRS ≥ 90th percentile. These patterns were consistent across multiple sensitivity analyses, across ancestry subgroups, and in external validation in AoU, supporting robustness of our findings.
These results align with prior work examining how CAD risk is jointly predicted by LDL-C levels and polygenic risk. Other studies have quantified this risk gradient across LDL-C and polygenic risk strata, including analyses in UKBB [19] and, more recently, another in a multi-ethnic cohort of 47,576 participants reported by Iribarren et al., [37] each demonstrating a similar pattern of risk. While both studies effectively demonstrated the joint influence of LDL-C and polygenic risk, their PRS stratification used broader risk categories (e.g., low, intermediate, or high risk). By contrast, our study stratified into finer PRS strata, enabling more granular risk assessment and identification of specific joint cutoffs into a nomogram that could inform clinical decision-making.
Importantly, our study establishes that these LDL-C × PRS strata are at risk comparable to those with SH but does not demonstrate that lipid-lowering therapy would achieve equivalent risk reductions in these groups. Whether individuals with moderately elevated LDL-C and high PRS derive benefit comparable to those with SH requires explicit quantification in future studies. Nonetheless, existing trial data suggest that individuals with high CAD PRS may derive disproportionately greater absolute risk reduction from lipid-lowering therapies. In the WOSCOPS primary prevention trial, participants in the highest PRS quintile experienced greater absolute risk reductions from statin therapy (7.9%) compared to all other individuals (2.7%) over a 13-year follow-up [38]. However, both groups had nearly identical baseline LDL-C levels (∼192 mg/dL), which limits the applicability of this finding to individuals with more moderate LDL-C. A similar pattern of heightened therapeutic benefit in high-PRS individuals was observed in a trial of alirocumab after acute coronary syndrome, where the highest PRS quintile experienced a 6.0% absolute risk reduction in MACEs versus 1.5% in all other individuals [39]. Notably, median baseline LDL-C levels were much lower in this study: 88.4 mg/dL and 86.1 mg/dL in the high and low PRS groups respectively, suggesting that heightened responsiveness to lipid lowering holds even at lower baseline LDL-C ranges. Together, these studies suggest that individuals with high PRS derive disproportionately greater benefit from lipid-lowering, across both high and low LDL-C ranges. However, neither directly addresses whether individuals with moderately elevated LDL-C and high PRS might derive benefit comparable to those with SH and average PRS.
Clarifying this question will require explicit quantification, ideally through randomized trials or well-designed observational studies capable of assessing treatment-effect heterogeneity across LDL-C × PRS strata. Even in the absence of definitive evidence of equivalent treatment benefit, identifying individuals whose LDL-C × PRS profiles confer risk not significantly lower than SH has meaningful implications for prevention and risk communication.
In age-stratified analyses, we observed a progressive expansion of LDL-C × PRS strata associated with CAD risk comparable to or exceeding that of SH. This pattern is consistent with known age-dependent effects of both LDL-C and CAD-PRS on CAD risk and supports the interaction between LDL-C and CAD-PRS. The broader set of SH-comparable strata in older individuals may reflect cumulative exposure to LDL-C and other age-related risk factors, which could amplify the penetrance of polygenic susceptibility. Conversely, the restriction of SH-level risk to the highest-PRS strata in younger adults suggests that PRS may be most clinically useful in early life as a precipitant for earlier management, despite low short-term risk [36,40]. Notably, widely used risk prediction models already incorporate age × lipid interaction terms [41], yet do not presently account for PRS-related interactions. Our findings suggest that incorporating age × PRS and LDL-C × PRS interactions may further improve risk stratification across the lifespan.
This study has several limitations. First, although time-to-event Cox models were used in UKBB, the relatively short follow-up in AoU necessitated cross-sectional logistic regression including both incident and prevalent CAD, limiting comparability and precluding assessment of time-dependent risk in the replication cohort. Second, LDL-C measurement differed between cohorts—baseline study measurement in UKBB versus the highest recorded clinical value in AoU—and adjustment for lipid-lowering therapy by dividing LDL-C by 0.7 provides only an approximation of true pre-treatment LDL-C. Third, both cohorts are observational and subject to residual confounding and selection bias.
An additional limitation relates to ancestry and generalizability. Despite use of GPSMult, which has improved multi-ancestry performance, predictive performance remains attenuated in non-European populations. While risk-equivalence relationships differed slightly in the non-EUR cohort, the overall risk patterns were consistent, with minor differences likely reflecting statistical variability due to decreased sample sizes and attenuated PRS performance rather than meaningful deviations in risk. Additionally, limited non-European sample sizes in both cohorts precluded robust estimates within finer ancestry subdivisions. Further efforts in larger multi-ancestry cohorts and further refinement of PRS in underrepresented populations will be essential to ensure that these risk relationships are broadly generalizable.
Finally, our analyses focused specifically on LDL-C and CAD PRS, whereas clinical decision-making takes into account many more factors. This work addresses only one guideline-based pathway for lipid-lowering therapy—the LDL-C–specific threshold—and does not evaluate other major indications, including diabetes (diabetes with LDL-C ≥ 70 mg/dL), elevated 10-year ASCVD risk (≥ 7.5%), or secondary prevention, each of which represents a distinct clinical context in which PRS may provide additional nuance and could be quantified. In this context, the present study provides one component of a broader evidence base needed to contextualize PRS in preventive decision-making: empirical identification of LDL-C × PRS strata that reach or exceed a recognized LDL-C–based treatment threshold. Future work should evaluate how CAD PRS interacts with other major risk indication pathways, with a potential long-term goal of integration into multivariable risk calculators, to further refine precision prevention strategies.
5. Conclusions
Individuals with moderate LDL-C elevations and high CAD PRS, comprising approximately 14.8% of the study population, carry comparable CAD risk to that of individuals with SH (LDL-C ≥ 190 mg/dL). These findings support consideration—pending further evaluation—of the following CAD PRS-adjusted LDL-C thresholds for primary prevention measures: 160 mg/dL for individuals with CAD PRS above the 70th percentile, 130 mg/dL for those with CAD PRS above the 80th percentile, and 100 mg/dL for those with CAD PRS above the 90th percentile.
Central illustration.
Declaration of generative AI and AI-assisted technologies in the manuscript preparation process
During the preparation of this work the authors used ChatGPT in order to generate codes for analyses and proofreading the manuscript. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.
Sources of funding
A.P.P., S.C., S.K., G.M.P., and T.N. are supported by grants from the National Institute of Health (NIH)/National Heart, Lung, and Blood Institute (NHLBI) (K08HL168238, K99HL177340, K99HL169733, R01HL127564, and K99HL165024, respectively). Research reported in this publication was supported by NIH for the project “Polygenic Risk Methods in Diverse Populations (PRIMED) Consortium”, with grant funding for Study Site FFAIR-PRS (U01HG011719) to P.N., and the Coordinating Center (U01HG011697). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Ethical statement
Secondary analyses of the data in the UK Biobank under Application 7089 were approved by the Massachusetts General Hospital Institutional Review Board. The Institutional Review Boards of the All of Us Research Program approved all study procedures and informed consent was obtained from all participants.
CRediT authorship contribution statement
Colin Harper: Writing – original draft, Visualization, Formal analysis, Data curation, Conceptualization. Anika Misra: Data curation. Aniruddh P. Patel: Data curation. So Mi Cho: Methodology. Satoshi Koyama: Data curation. Gina M. Peloso: Writing – review & editing. Whitney Hornsby: Project administration. Tetsushi Nakao: Writing – review & editing, Visualization, Validation, Supervision, Formal analysis, Data curation. Pradeep Natarajan: Writing – review & editing, Visualization, Funding acquisition, Conceptualization.
Declaration of competing interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:
Tetsushi Nakao reports financial support was provided by National Heart Lung and Blood Institute. Pradeep Natarajan reports financial support was provided by National Human Genome Research Institute. Aniruddh P. Patel reports financial support was provided by National Heart Lung and Blood Institute. So Mi Cho reports financial support was provided by National Heart Lung and Blood Institute. Satoshi Koyama reports financial support was provided by National Heart Lung and Blood Institute. Gina M. Peloso reports financial support was provided by National Heart Lung and Blood Institute. Tetsushi Nakao reports a relationship with Kowa Company Ltd that includes: speaking and lecture fees. Pradeep Natarajan reports a relationship with Allelica that includes: consulting or advisory and funding grants. Pradeep Natarajan reports a relationship with Amgen Inc that includes: funding grants. Pradeep Natarajan reports a relationship with Apple Inc that includes: consulting or advisory and funding grants. Pradeep Natarajan reports a relationship with Boston Scientific Corporation that includes: funding grants. Pradeep Natarajan reports a relationship with Cleerly Inc that includes: funding grants. Pradeep Natarajan reports a relationship with Genentech Inc that includes: consulting or advisory and funding grants. Pradeep Natarajan reports a relationship with Roche that includes: consulting or advisory and funding grants. Pradeep Natarajan reports a relationship with Ionis Pharmaceuticals Inc that includes: funding grants. Pradeep Natarajan reports a relationship with Novartis that includes: consulting or advisory and funding grants. Pradeep Natarajan reports a relationship with Silence Therapeutics that includes: funding grants. Pradeep Natarajan reports a relationship with AIRNA that includes: consulting or advisory. Pradeep Natarajan reports a relationship with AstraZeneca that includes: consulting or advisory. Pradeep Natarajan reports a relationship with Brain Capital that includes: consulting or advisory. Pradeep Natarajan reports a relationship with Blackstone Life Sciences that includes: consulting or advisory. Pradeep Natarajan reports a relationship with Bristol Myers Squibb Co that includes: consulting or advisory. Pradeep Natarajan reports a relationship with Creative Educational Concepts Inc that includes: consulting or advisory. Pradeep Natarajan reports a relationship with CRISPR Therapeutics Inc that includes: consulting or advisory. Pradeep Natarajan reports a relationship with Eli Lilly and Company that includes: consulting or advisory. Pradeep Natarajan reports a relationship with Esperion Therapeutics Inc that includes: consulting or advisory. Pradeep Natarajan reports a relationship with Foresite Capital that includes: consulting or advisory. Pradeep Natarajan reports a relationship with Foresite Labs that includes: consulting or advisory. Pradeep Natarajan reports a relationship with GV that includes: consulting or advisory. Pradeep Natarajan reports a relationship with HeartFlow Inc that includes: consulting or advisory. Pradeep Natarajan reports a relationship with Magnet Biomedicine, Inc. that includes: consulting or advisory. Pradeep Natarajan reports a relationship with Merck & Co Inc that includes: consulting or advisory. Pradeep Natarajan reports a relationship with Novo Nordisk Inc that includes: consulting or advisory. Pradeep Natarajan reports a relationship with TenSixteen Bio that includes: consulting or advisory and equity or stocks. Pradeep Natarajan reports a relationship with Tourmaline Bio, Inc. that includes: consulting or advisory. Pradeep Natarajan reports a relationship with Bolt that includes: equity or stocks. Pradeep Natarajan reports a relationship with Candela that includes: equity or stocks. Pradeep Natarajan reports a relationship with Mercury that includes: equity or stocks. Pradeep Natarajan reports a relationship with MyOme that includes: equity or stocks. Pradeep Natarajan reports a relationship with Parameter Health that includes: equity or stocks. Pradeep Natarajan reports a relationship with Preciseli that includes: equity or stocks. Pradeep Natarajan has patent licensed to Recora. P.N. declares spousal employment at Vertex Pharmaceuticals. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
We thank the participants, contributors, and researchers of the UK Biobank and All of Us Research Program.
Footnotes
Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.ajpc.2026.101487.
Contributor Information
Colin Harper, Email: colinharper@hms.harvard.edu.
Anika Misra, Email: misraani@broadinstitute.org.
Aniruddh P. Patel, Email: aniruddh@broadinstitute.org.
So Mi Cho, Email: somi@broadinstitute.org.
Satoshi Koyama, Email: skoyama2@mgh.harvard.edu.
Gina M. Peloso, Email: gpeloso@bu.edu.
Whitney Hornsby, Email: whornsby@mgb.org.
Tetsushi Nakao, Email: tnakao@broadinstitute.org.
Pradeep Natarajan, Email: pnatarajan@mgh.harvard.edu.
Appendix. Supplementary materials
References
- 1.Collaborators GBDCoD Global burden of 288 causes of death and life expectancy decomposition in 204 countries and territories and 811 subnational locations, 1990-2021: a systematic analysis for the Global Burden of Disease Study 2021. Lancet. May 18, 2024;403(10440):2100–2132. doi: 10.1016/S0140-6736(24)00367-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Arnett D.K., Blumenthal R.S., Albert M.A., et al. 2019 ACC/AHA Guideline on the primary prevention of Cardiovascular Disease: executive Summary: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. J Am Coll Cardiol. Sep 10 2019;74(10):1376–1414. doi: 10.1016/j.jacc.2019.03.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Lloyd-Jones D.M., Braun L.T., Ndumele C.E., Smith S.C., Jr., Sperling L.S., Virani S.S., Blumenthal R.S. Use of risk assessment tools to guide decision-making in the primary prevention of atherosclerotic cardiovascular disease: a special report from the American Heart Association and American College of Cardiology. Circulation. Jun 18 2019;139(25):e1162–e1177. doi: 10.1161/CIR.0000000000000638. [DOI] [PubMed] [Google Scholar]
- 4.Wong N.D., Budoff M.J., Ferdinand K., et al. Atherosclerotic cardiovascular disease risk assessment: an American Society for Preventive Cardiology clinical practice statement. Am J Prev Cardiol. Jun 2022;10 doi: 10.1016/j.ajpc.2022.100335. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.group Sw, collaboration ESCCr SCORE2 risk prediction algorithms: new models to estimate 10-year risk of cardiovascular disease in Europe. Eur Heart J. Jul 1 2021;42(25):2439–2454. doi: 10.1093/eurheartj/ehab309. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Khan S.S., Coresh J., Pencina M.J., et al. Novel prediction equations for absolute risk assessment of total cardiovascular disease incorporating cardiovascular-kidney-metabolic health: a scientific statement from the American Heart Association. Circulation. Dec 12 2023;148(24):1982–2004. doi: 10.1161/CIR.0000000000001191. [DOI] [PubMed] [Google Scholar]
- 7.O'Sullivan J.W., Raghavan S., Marquez-Luna C., et al. Polygenic risk Scores for cardiovascular disease: a scientific statement from the American Heart Association. Circulation. Aug 23 2022;146(8):e93–e118. doi: 10.1161/CIR.0000000000001077. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Aragam K.G., Natarajan P. Polygenic scores to assess atherosclerotic cardiovascular disease risk: clinical perspectives and basic implications. Circ Res. Apr 24 2020;126(9):1159–1177. doi: 10.1161/CIRCRESAHA.120.315928. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Khera A.V., Emdin C.A., Drake I., et al. Genetic risk, adherence to a healthy lifestyle, and coronary disease. N Engl J Med. Dec 15 2016;375(24):2349–2358. doi: 10.1056/NEJMoa1605086. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Aragam K.G., Dobbyn A., Judy R., et al. Limitations of contemporary guidelines for managing patients at high genetic risk of coronary artery disease. J Am Coll Cardiol. Jun 9 2020;75(22):2769–2780. doi: 10.1016/j.jacc.2020.04.027. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Naderian M., Norland K., Schaid D.J., Kullo I.J. 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. Ann Intern Med. Jan 2025;178(1):1–10. doi: 10.7326/ANNALS-24-00716. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Patel A.P., Wang M., Ruan Y., et al. A multi-ancestry polygenic risk score improves risk prediction for coronary artery disease. Nat Med. Jul 2023;29(7):1793–1803. doi: 10.1038/s41591-023-02429-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Truong B., Ruan Y., Haidermota S., et al. Modification of coronary artery disease clinical risk factors by coronary artery disease polygenic risk score. Med. May 10, 2024;5(5):459–468.e3. doi: 10.1016/j.medj.2024.02.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Hao L., Kraft P., Berriz G.F., et al. Development of a clinical polygenic risk score assay and reporting workflow. Nat Med. May 2022;28(5):1006–1013. doi: 10.1038/s41591-022-01767-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Lennon N.J., Kottyan L.C., Kachulis C., et al. Selection, optimization and validation of ten chronic disease polygenic risk scores for clinical implementation in diverse US populations. Nat Med. Feb 2024;30(2):480–487. doi: 10.1038/s41591-024-02796-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Lewis A.C.F., Perez E.F., Prince A.E.R., et al. Patient and provider perspectives on polygenic risk scores: implications for clinical reporting and utilization. Genome Med. Oct 7 2022;14(1):114. doi: 10.1186/s13073-022-01117-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Kullo I.J., Lewis C.M., Inouye M., Martin A.R., Ripatti S., Chatterjee N. Polygenic scores in biomedical research. Nat Rev Genet. Sep 2022;23(9):524–532. doi: 10.1038/s41576-022-00470-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Purvis R., Forrest L.E., Young M.A., Limb S., James P., Taylor N. Defining next steps in the clinical implementation of polygenic scores: a landscape analysis of professional groups' perspectives. Genet Med. Jun 2025;27(6) doi: 10.1016/j.gim.2025.101414. [DOI] [PubMed] [Google Scholar]
- 19.Bolli A., Di Domenico P., Pastorino R., Busby G.B., Botta G. Risk of coronary artery disease conferred by low-density lipoprotein cholesterol depends on polygenic background. Circulation. Apr 6 2021;143(14):1452–1454. doi: 10.1161/CIRCULATIONAHA.120.051843. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Force U.S.P.S.T., Mangione C.M., Barry M.J., et al. Statin use for the primary prevention of cardiovascular disease in adults: US Preventive Services Task Force Recommendation Statement. JAMA. Aug 23 2022;328(8):746–753. doi: 10.1001/jama.2022.13044. [DOI] [PubMed] [Google Scholar]
- 21.Mach F., Baigent C., Catapano A.L., et al. 2019 ESC/EAS Guidelines for the management of dyslipidaemias: lipid modification to reduce cardiovascular risk. Eur Heart J. Jan 1 2020;41(1):111–188. doi: 10.1093/eurheartj/ehz455. [DOI] [PubMed] [Google Scholar]
- 22.Bycroft C., Freeman C., Petkova D., et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. Oct 2018;562(7726):203–209. doi: 10.1038/s41586-018-0579-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.The “all of us” research program. N Engl J Med. 2019;381(7):668–676. doi: 10.1056/NEJMsr1809937. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Genomic data in the All of Us Research Program. Nature. Mar 2024;627(8003):340–346. doi: 10.1038/s41586-023-06957-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Graham S.E., Clarke S.L., Wu K.H., et al. The power of genetic diversity in genome-wide association studies of lipids. Nature. Dec 2021;600(7890):675–679. doi: 10.1038/s41586-021-04064-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Genomes Project C. Auton A., Brooks L.D., et al. A global reference for human genetic variation. Nature. Oct 1 2015;526(7571):68–74. doi: 10.1038/nature15393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Huang J., Howie B., McCarthy S., et al. Improved imputation of low-frequency and rare variants using the UK10K haplotype reference panel. Nat Commun. Sep 14 2015;6:8111. doi: 10.1038/ncomms9111. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Rosenberg N.A., Pritchard J.K., Weber J.L., Cann H.M., Kidd K.K., Zhivotovsky L.A., Feldman M.W. Genetic structure of human populations. Science. Dec 20 2002;298(5602):2381–2385. doi: 10.1126/science.1078311. [DOI] [PubMed] [Google Scholar]
- 29.Misra A., Truong B., Urbut S.M., et al. Instability of high polygenic risk classification and mitigation by integrative scoring. Nat Commun. Feb 12 2025;16(1):1584. doi: 10.1038/s41467-025-56945-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Truong B., Hull L.E., Ruan Y., et al. Integrative polygenic risk score improves the prediction accuracy of complex traits and diseases. Cell Genom. Apr 10 2024;4(4) doi: 10.1016/j.xgen.2024.100523. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Nakao T., Koyama S., Truong B., et al. Genomic, phenomic, and geographic associations of leukocyte telomere length in the United States. MedRxiv. Nov 4 2024:2024.11.02.24316529. doi:10.1101/2024.11.02.24316529. [DOI] [PMC free article] [PubMed]
- 32.emmeans: estimated Marginal means, aka Least-Squares means. 2025. https://cran.r-project.org/package=emmeans.
- 33.Prospective Studies C. Lewington S., Whitlock G., et al. Blood cholesterol and vascular mortality by age, sex, and blood pressure: a meta-analysis of individual data from 61 prospective studies with 55,000 vascular deaths. Lancet. Dec 1 2007;370(9602):1829–1839. doi: 10.1016/S0140-6736(07)61778-4. [DOI] [PubMed] [Google Scholar]
- 34.Xiao J., Wei H., Gao Z., Chen L., Ye W., Huang W. Differential age-specific associations of LDL cholesterol and body mass index with coronary heart disease. Atherosclerosis. Jun 2024;393 doi: 10.1016/j.atherosclerosis.2024.117542. [DOI] [PubMed] [Google Scholar]
- 35.Marston N.A., Pirruccello J.P., Melloni G.E.M., et al. Predictive utility of a coronary artery disease polygenic risk score in primary prevention. JAMA Cardiol. Feb 1 2023;8(2):130–137. doi: 10.1001/jamacardio.2022.4466. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Urbut S.M., Cho S.M.J., Paruchuri K., et al. Dynamic importance of genomic and clinical risk for coronary artery disease over the life course. Circ Genom Precis Med. Feb 2025;18(1) doi: 10.1161/CIRCGEN.124.004681. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Iribarren C., Lu M., Elosua R., Gulati M., Wong N.D., Rana J.S. Joint consideration of LDL-C and polygenic risk for coronary heart disease risk assessment. JACC Adv. Nov 2025;4(11 Pt 1) doi: 10.1016/j.jacadv.2025.102228. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Natarajan P., Young R., Stitziel N.O., et al. Polygenic risk score identifies subgroup with higher burden of atherosclerosis and greater relative benefit from Statin therapy in the primary prevention setting. Circulation. May 30, 2017;135(22):2091–2101. doi: 10.1161/CIRCULATIONAHA.116.024436. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Damask A., Steg P.G., Schwartz G.G., et al. Patients with high genome-wide polygenic risk scores for coronary artery disease may receive greater clinical benefit from Alirocumab treatment in the ODYSSEY OUTCOMES trial. Circulation. Feb 25 2020;141(8):624–636. doi: 10.1161/CIRCULATIONAHA.119.044434. [DOI] [PubMed] [Google Scholar]
- 40.Urbut S.M., Yeung M.W., Khurshid S., et al. MSGene: a multistate model using genetic risk and the electronic health record applied to lifetime risk of coronary artery disease. Nat Commun. 2024;15(1):1–14. doi: 10.1038/s41467-024-49296-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Goff D.C., Jr., Lloyd-Jones D.M., Bennett G., et al. 2013 ACC/AHA guideline on the assessment of cardiovascular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol. Jul 1 2014;63(25 Pt B):2935–2959. doi: 10.1016/j.jacc.2013.11.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.