https://orcid.org/0000-0002-9998-1640
Figures
Abstract
Aim
To develop a model for cardiovascular disease (CVD) risk, combining polygenic risk score (PRS) with traditional risk factors while assessing the added value of PRS in two cohorts of biobank participants.
Methods
Data of 128 209 participants from the Estonian Biobank recruited between 2002–2017 and 2018–2022 without prevalent cardiovascular disease, was included. Hazard ratios (HR) for polygenic risk versus conventional risk factors were estimated with Cox proportional hazards models, cumulative incidence was assessed with Aalen-Johansen curves. Predictive performance was tested using a split-sample approach and competing risk modelling. Age at CVD event served as the outcome, and the impact of the PRS was evaluated by age group (25–59 vs. 60+), sex, and recruitment period, using HRs, Harrell’s C-index, and net reclassification indices (NRI).
Results
The estimated HR per one standard deviation (SD) of PRS ranged from 1.1, 95% CI 1.06–1.15 (age 60 + , earlier cohort) to 1.36, 95% CI 1.24–1.49 (men 25–59, later cohort). Adding PRS to the conventional risk factors in the age group 25–59 increased the C-statistic by 0.028 (p < 0.0001) for men. In the age group 60 + , the increase was 0.016 (p = 0.0002) across all. In the independent validation set, the continuous NRI was 19.1% (95% CI 13.3%–24.9%) in the 25–59 group and 13.9% (95% CI 8.1%–19.6%) in the 60 + group.
Citation: Puusepp T, Põld A, Milani L, Elken A, Estonian Biobank Research Team, Jürisson M, et al. (2025) Development and validation of a risk prediction algorithm for high-risk populations combining genetic and conventional risk factors of cardiovascular disease. PLoS One 20(10): e0335064. https://doi.org/10.1371/journal.pone.0335064
Editor: Giuseppe Novelli, Universita degli Studi di Roma Tor Vergata, ITALY
Received: May 26, 2025; Accepted: October 6, 2025; Published: October 21, 2025
Copyright: © 2025 Puusepp et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Individual level data from the Estonian Biobank cannot be shared publicly due to legal and ethical restrictions. According to the Human Genes Research Act, which regulates the operations of the Estonian Biobank, the data include potentially identifiable genetic and health information and are therefore subject to restricted access. Access to these data is only possible directly through the Estonian Biobank upon ethics approval from the Estonian Committee on Bioethics and Human Research. Researchers may request access following the procedure described here: https://genomics.ut.ee/en/content/estonian-biobank#dataaccess.
Funding: This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101060011. The research was conducted using the Estonian Center of Genomics/Roadmap II funded by the Estonian Research Council (project number TT17). The work of TP and KF was supported by Estonian Research Council (grant PRG1197). The work of AE was supported by Estonian Research Council (grant PRG2078). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: We have no conflicts of interest to disclose.
Introduction
Atherosclerotic cardiovascular diseases (CVD), including coronary artery disease and cerebrovascular disease, are the leading cause of death in numerous European countries [1]. Polygenic risk scores (PRSs) have demonstrated their effectiveness as a valuable and innovative method for assessing genetic risk related to CVD and improving the accuracy of disease risk prediction [2]. Several studies indicate that incorporating genetic risk assessment into existing risk stratification algorithms could significantly enhance their efficiency [3–5].
Polygenic risk scores have been validated in several studies and have been found to enhance CVD risk prediction independently of many traditional factors such as smoking, hypercholesterolemia, hypertension, obesity, and family history of CVD [2,6–8]. Typically, a PRS combines the effect of a large number (from hundreds to millions) of single nucleotide polymorphisms (SNPs) as a weighted sum of allele counts [9]. It has been shown that elevated polygenic scores contribute to a significantly higher percentage of early-onset myocardial infarction cases than monogenic variants for familial hypercholesterolemia [10,11]. This implies that integrating genetic predisposition complements CVD risk prediction and, when combined with traditional factors, can significantly improve disease risk prediction and facilitate decision making in primary prevention of CVD [12]. Studies assessing CVD risk combining PRS with clinical and lifestyle data show promising results, yet more rigorous validation and comparisons between existing models are necessary to justify their clinical utility [12–16].
Several large cohort and country-specific prognostic CVD risk models have been developed based on traditional risk factors such as the European SCORE2, the American pooled cohort equations (PCE), and the UK-specific QRISK3 algorithm, yet their generalisability across different populations remains limited [17–21]. A comparable challenge arises when considering the questionable effectiveness of PRS-based risk assessment algorithms that utilize UK Biobank data, given that the United Kingdom has a low prevalence of cardiovascular disease compared to high-risk populations in Middle and Eastern Europe [22–24].
This study describes the development and validation of the novel risk assessment model combining traditional cardiovascular risk factors with PRSs.
Methods
Ethics
The activities of the EstBB are regulated by the Human Genes Research Act, which was adopted in 2000 specifically for the operations of the EstBB. Individual level data analysis in the EstBB was carried out under ethical approval 1.1–12/624 from the Estonian Committee on Bioethics and Human Research (Estonian Ministry of Social Affairs), using data according to release application 6–7/GI/24836 from the Estonian Biobank. All participants have signed a written informed consent.
Sources of data
Data from the Estonian Biobank (EstBB) was used to compose the study cohort [25]. EstBB is a volunteer-based biobank that includes genotype and clinical events data of more than 210 000 participants. The health records are regularly updated using national registries, hospital databases and the database of the national health insurance fund which covers data from both primary and secondary care. Additionally, the cholesterol data was quantified using nuclear magnetic resonance (NMR) spectroscopy. Participants have been recruited during two distinct phases. The active recruitment periods for the first wave were in 2002–2004 and 2007–2010, during which over 51 000 participants were recruited. An additional approximately 1000 volunteers joined in 2011–2017, resulting in a cohort of about 52 000 participants [25]. The second wave recruited over 150 000 participants during its active national campaign in 2018–2019, followed by about 9000 additional participants recruited in 2020–2024, resulting in a cohort of almost 160 000 participants. The recruitment process used the national digital ID system, allowing electronic consent and blood sampling at nearby healthcare providers or pharmacies, making the process simpler, faster and more accessible for potential participants, compared to the first wave. The total number of participants across both recruitment waves (over 210 000) accounts for roughly 20% of the Estonian adult population. The data were accessed for research purposes on October 26, 2024. The authors did not have access to information that could identify individual participants during or after data collection.
The PRS used in this study is the multi-ancestry PRS for coronary artery disease developed by Patel, et al. and obtained from the PGS catalog [26,27]. This PRS was selected from a pool of 151 candidate CAD PRSs for its highest z-score in predicting prevalent CVD via logistic regression adjusted for age at recruitment and sex, utilising EstBB data comprising 15 095 CVD cases and 119 694 controls at baseline.
Participants
Sample size.
All 185 760 EstBB participants aged at least 25 years at recruitment with genotyping data available were considered for the analysis. After applying the inclusion and exclusion criteria (see below), the total number of individuals included in the study was n = 128 209, with 32 554 from the first recruitment wave (2002–2017) and 95 655 from the second wave (2018–2022). The cohorts from the two recruitment waves were analysed separately to avoid potential selection biases arising from differences in recruitment processes and calendar time effects.
Exclusion criteria.
- -. Prevalent CVD cases, i.e., individuals diagnosed with non-fatal CVD (ICD-10 codes I20, I21–I25, I60–I69 excluding I60, I62, I67.1, I67.5, I68.2) before recruitment (n = 25 894).
- -. Individuals with diabetes mellitus (E10–E14) at baseline (n=11 555)
- -. Individuals with familial hypercholesterolemia (FH) (n=76)
- -. Individuals with missing lipid values (Total cholesterol (Total-C), HDL cholesterol (HDL-C)) (n = 3 569)
- -. Individuals with missing systolic blood pressure (SBP) or with SBP<50 mmHg or SBP>300 mmHg (n=27 212)
- -. Individuals with missing body mass index (BMI) or with BMI less than 15 kg/m2 or more than 50 kg/m2 (n=752)
- -. Individuals with missing smoking data (n=1 591)
Outcome
The outcome of interest was incident non-fatal or fatal CVD event. Incident CVD events were identified using the atherosclerotic CVD definition provided by the SCORE2 working group [17]. A detailed list of the ICD-codes is in S1 Table. As the outcome data is based on Electronic Health Records (EHR) linkage, we assumed there is no missing outcome data. The data from EHR was available up to December 31st, 2023. The outcome event was observed in 6 893 individuals and deaths from non-CVD causes (n = 2 124) were treated as competing events.
Predictors
The model combined conventional predictors of CVD and pre-calculated PRS for CAD. Predictor management has been described in the S1 Text.
List of the included predictors:
- -. Age (years)
- -. Sex (M/F)
- -. Current smoking (y/n)
- -. SBP (mmHg)
- -. Total cholesterol (mmol/L)
- -. HDL cholesterol (mmol/L)
- -. BMI (kg/m2)
- -. PRS for CAD
Statistical methods
- a). Impact of PRS on the outcome
Aalen-Johansen curves, accounting for competing causes of death and using age as time scale, were estimated to assess how PRS differences impact cumulative CVD event incidence in men and women aged 25–70 at recruitment across three PRS groups (bottom 10%, 10%–90%, top 10%). Crude hazard ratios with 95% CIs were calculated using the Cox proportional hazards models, separately for the earlier and later cohort.
- b). PRS effect and discrimination compared with that of conventional risk factors
Separate models were fitted for earlier and later cohort and two age groups (25–59 and 60 + years) to estimate the effect of PRS and compare its discrimination with that of conventional risk factors (current smoking, SBP, total cholesterol, HDL cholesterol, BMI). In the younger group, models were sex-specific due to significant interactions, while in the older group, sex was a stratification variable. Cox models used age at CVD event as the time scale to avoid bias due to left truncation. Harrel’s C-index was used to assess discrimination for single-risk-factor models and models with all covariates, with and without PRS.
- c). Model’s predictive ability in an independent sample
An independent dataset was used to assess the model’s predictive ability using a split-sample approach, dividing the cohort into a training set (one-third) and a validation set (two-thirds of the cohort). In each group (defined by age, sex, and recruitment period), two Cox proportional hazard models were fitted in the training datasets with time from recruitment to the incident CVD event as the main outcome of interest. The models accounted for competing risks using the multi-state modelling principle, where other causes of death were considered as competing events [28].
The event times for participants with no events or with an event occurring later than 5 or 10 years after recruitment were censored at 5 or 10 years for the later and the earlier cohort, respectively. Models were first fitted using traditional CVD risk factors, then with these factors plus PRS.
Model calibration was assessed by comparing the number of observed CVD events within quintiles of predicted risk with those predicted from the models. Continuous net reclassification index (NRI) and categorical NRI were computed to compare the 5-year predicted risk with observed event rates. To convert the linear predictors from the survival models adjusted for traditional risk factors and the PRS into absolute 5- or 10-year CVD risk predictions, the absolute risk formula of Benichou and Gail was used (implemented in the riskRegression package version 2023.9.20 of the R software) [29,30].
All analyses were done using R version 4.2.2 [31].
Results
Cohort description and outcomes
A total of 128 209 Estonian Biobank participants were included in the analysis. Baseline characteristics of the earlier cohort (n = 32 554, recruited in 2002–2017) and later cohort (n = 95 655, recruited in 2018–2022) are presented in Table 1. The mean age at recruitment in both cohorts is similar, approximately 44 years (44.4 in the earlier cohort and 43.7 in the later) and the sex distributions are comparable (about two thirds of the cohorts being women). The median follow-up period is 14.9 years for the earlier cohort and 5.1 years for the later cohort. Regarding risk factors, there is a large difference in smoking prevalence: 30% of the earlier cohort were current smokers at recruitment, compared to 19% in the later cohort. We can also see somewhat higher average blood pressure and LDL cholesterol levels in the earlier cohort among individuals aged 60 and older. The 5-year cumulative CVD incidence differs being 5.7% in the earlier and 0.6% the later cohort. These differences reflect the longer follow-up and higher prevalence of traditional risk factors in the earlier cohort, as well as improvements in population health and differences in recruitment procedures between the two cohorts.
Cumulative incidence in different PRS percentiles
The cumulative incidence curves on age scale (Fig 1) show higher CVD incidence in men than women across all PRS groups and cohorts. By age 70, about 40% of men in the earlier cohort had experienced CVD, compared to about 25% in the later cohort. For women, the rates were 24% and 13%, respectively. Men in the later cohort had a similar CVD risk to women in the earlier cohort.
The effect of PRS is consistent across sexes and cohorts. Individuals in the highest PRS decile have significantly higher CVD risk compared to those in the middle or lowest percentiles. For example, in the earlier cohort, men in the highest PRS decile have over double the cumulative incidence of CVD by age 70 compared to those in the lowest decile. Men in the top PRS decile reach a 20% cumulative incidence of CVD six years earlier than average, while those in the lowest decile reach it five years later, creating a decade-long difference between extreme deciles. Similar patterns are observed in women.
Compared to individuals in the 10th–90th PRS percentile range, the hazard ratio (HR) for those in the highest PRS decile is 1.7 (95% CI 1.5–1.9) for men and 1.5 (95% CI 1.3–1.7) for women in the earlier cohort, 1.9 (95% CI 1.6–2.4) for men and 1.6 (95% CI 1.3–2.0) for women in the later cohort. When the highest PRS decile is compared to the lowest, HRs range from 1.9 for women in the earlier cohort to 2.9 for men in the later cohort.
Effect of PRS in a model with conventional risk factors
HRs corresponding to PRS are shown in Fig 2. These were obtained from Cox models with age as time scale, adjusted for current smoking, SBP, total cholesterol, HDL cholesterol and BMI. Models were fitted separately for the two cohorts, sexes, and age groups (in individuals aged 60 + , sex-stratified models were fitted).
The effect of PRS is strongest in younger men, with similar HRs across cohorts. Younger women show slightly lower HRs, with a stronger effect in the later cohort. In individuals aged 60 + , the effect is more pronounced in the later cohort, with no significant interaction between PRS and sex in this age group. Detailed parameter estimates are shown in S2 Table.
Discriminatory power of PRS and conventional risk factors: comparison of C-indices
Cox models using age as time scale and stratified by cohort (and additionally by sex in the age group 60+) were fitted with each traditional risk factor and PRS as a single covariate to assess their relative importance. Harrell’s C-index estimates remained below 0.6 (Fig 3), reflecting the model’s discriminative ability for individuals of the same baseline age, as age was used as time scale.
In younger men, PRS is the strongest single predictor, with a C-index of 0.60 (95% CI 0.58–0.61), nearly equivalent to that of the model with traditional risk factors. Combining traditional risk factors and PRS leads to a C-index of 0.63 (95% CI 0.61–0.64), representing an improvement of 0.028 (p < 0.0001) over the model without PRS.
In younger women, the effect of PRS alone (C-index 0.56, 95% CI 0.54–0.57) is not stronger than that of SBP or BMI, but it still leads to a slight improvement in the combined risk factor model (C-index 0.61, 95% CI 0.60–0.63, increase 0.004, p = 0.085).
In individuals aged 60 + , PRS is the strongest single predictor (C-index 0.55, 95% CI 0.53–0.56). Adding PRS to the traditional risk factor model increases the C-index to 0.57 (95% CI 0.55–0.58, increase 0.016, p = 0.0002).
In a separate analysis of the two cohorts, there were only minor differences in C-index values across cohorts, with the exception of 60 + age category, where the improvement was clearly higher in the new cohort (adding PRS increased the C-index by 0.008 in the 2002–2017 cohort and by 0.03 in the 2018–2022 cohort; details in S3 Table).
The performance of model-based predictions in an independent sample
To assess the performance of the predictive algorithm on independent data, models including all traditional risk factors, with and without PRS, were refitted using the training set of 42 827 individuals. The parameter estimates were used to calculate the linear predictor values for 85 382 individuals in the validation set. As seen in S1 Fig, calibration of the model with PRS is adequate in the validation set when the predicted risk of CVD in groups defined by quintiles of predicted risk is compared to the number of CVD events in these groups.
Net reclassification analysis comparing the conventional model to the conventional model with PRS were done separately for the two age groups (25–59 and 60+) but jointly for the two cohorts (Table 2). NRI was calculated according to predicted 5-year risk of CVD. For the categorical NRI analyses, risk categories (low, intermediate, and high) were defined as <1.25%, 1.25% to 5%, and >5% within 5 years for the younger age group (25–59), and as <5%, 5% to 10%, and >10% within 5 years for the older age group (60+) [32]. A detailed description of the NRI analysis is in the S2 Text.
There is a significant improvement in reclassification for both events and non-events in both age groups. In the age group 25–59, the overall NRI was 19.1% (95% CI 13.3%–24.9%). The categorical NRI in this group showed a modest but significant improvement of 3.0% (95% CI 1.2%–4.8%). In the age group 60 + , the overall NRI was slightly lower but still significant at 13.9% (95% CI 8.1%–19.6%), and the categorical NRI in this age group was slightly higher at 3.1% (95% CI 1.1%–5.0%). Detailed results of categorical NRI analysis are in S4 Table.
Fig 4 displays results of the categorical NRI analysis, focusing on the reclassification of individuals initially classified in the intermediate risk group. In the age group 25–59, the initial intermediate risk category included 19 871 individuals, with a 5-year CVD incidence of 2.6%. After PRS adjustment, 8.1% of these individuals were reclassified to low risk, where the incidence was 0.9%, and 3.7% were moved to high risk, where the incidence increased to 5.8%. In the age group 60 + , the initial intermediate risk category included 2 931 individuals, with a 5-year CVD incidence of 8.8%. After PRS incorporation, 15.1% were reclassified to low risk, where the incidence was 5.4%, and 11.4% to high risk, where the incidence was 12.9%.
Arrows indicate the movement of individuals between categories, with corresponding percentages representing the proportion of individuals reclassified.
Potential for using PRS-based risk prediction in practice: an illustration
The practical use of the proposed risk prediction algorithm involves communication of the effect of unmodifiable genetic component (PRS) alongside potentially modifiable risk factors. We illustrate the message that could be delivered to men and women of age 50. Fig 5 shows the predicted 10-year risk for non-smoking individuals with average PRS, as well as predictions for individuals whose PRS exceeds the mean by two SDs and/or who are current smokers. The values of all other risk factors were fixed at their mean values for men or women aged 25–59 in the 2002–2017 cohort.
The predictions are calculated from the Cox model fitted using the full data of the age group 25–59 recruited in 2002–2017 and using time on study as time scale. The high and average PRS correspond to PRS values of 2 (mean + 2SD) and 0 (mean/median), respectively.
The plot indicates that the 10-year CVD risk for a non-smoking man with a high PRS (20.2%, 95% CI 17.0%–23.5%) exceeds that of a smoker with an average PRS (15.6%, 95% CI 14.1%–17.3%). Among women, a current smoker with an average PRS faces a comparable risk (8.4%, 95% CI 7.4%–9.5%) to a non-smoker with a high PRS (8.8%, 95% CI 7.6%–10.2%).
Communicating this information to healthcare professionals enhances their understanding of the importance of PRS and supports patient discussions, encouraging high-risk individuals to adjust their behaviour while considering the individual differences in genetic predisposition.
Discussion
To the best of our knowledge this is the first tailored model for a high CVD-risk population combining polygenic and traditional risk factors for estimating cardiovascular disease risk. This study found a significant improvement in risk discrimination for both men and women when incorporating the CAD PRS to the prediction model.
Our model used Estonian Biobank data and demonstrated a significant rise in the risk of CVD events and/or mortality with higher CAD PRS. Moreover, individuals with high PRS may experience the onset of the disease up to a decade earlier compared to those with medium or low PRS. This suggests that a high PRS (top 10%) constitutes a substantial risk factor, comparable to that of smoking or high cholesterol, with elevated risks apparent as early as in one’s 30s. Our study demonstrates that the PRS-inclusive risk model performed best in younger cohorts, which is crucial given the pressing need to develop high-quality risk assessment tools for younger populations, especially since existing models like SCORE-2 can only be used starting at the age of 40 [17]. Similar results have also been shown in previous PRS population-based models created with Finnish and UK Biobank data [6,22,33].
We constructed models utilizing two distinct cohorts from different time periods, yet the significance and consistency of the PRS effect remained consistent in both, underscoring its importance in clinical risk assessment, as also reported by Patel, et al [27]. The Estonian population is categorised as high risk for cardiovascular disease based on standardised cardiovascular disease mortality rates, thus serving as a proxy to multiple high-risk populations in Eastern Europe [17]. The baseline risk differs between the two Biobank cohorts, with the later cohort having lower levels of conventional risk factors resulting in lower CVD incidence rates compared to the earlier cohort (reflecting contemporary declining trends in CVD incidence in developed countries). Regardless of cohort differences, the inclusion of PRS in the model demonstrates its consistent effect on overall risk across both cohorts, highlighting the model’s robustness. As public health advancements and behavioural changes reduce the impact of traditional risk factors on disease risk, the relative importance of PRS grows [34]. Consequently, models like ours can help identify high-risk individuals within the population and guide targeted interventions.
The current European Society of Cardiology (ESC) Cardiovascular Disease Prevention Guidelines do not advocate for the routine collection of genetic data in primary prevention [32]. However, a recent clinical consensus highlights the critical need to quantify the potential benefits of PRS in clinical practice [35]. In European countries, polygenic risk scores are not yet integrated into routinely collected administrative health data for risk prediction at either the population or individual level. This practice should be reassessed as the global focus on prevention and health promotion increases, highlighting the growing importance of genetics and polygenic risk in everyday healthcare. Clinicians often rely on risk prediction models like SCORE2 and QRISK3 for cardiovascular risk management, but the potential of genetic testing in primary prevention requires more rigorous evaluation. Before integrating PRS into clinical practice, the effectiveness of PRS-based primary CVD prevention strategies, such as lipid-lowering treatments, must be validated through randomized clinical trials. As our study has revealed that individuals with very high CAD PRS face a significantly elevated risk of CVD, we argue that such individuals should be directed toward more intensive primary preventive measures at an early age [36].
Strengths and limitations
A key strength of our study is the use of the Estonian Biobank database, which uniquely integrates clinical and genetic data—unlike most biobanks where these datasets are typically separate [37]. The availability of CAD PRS for nearly all biobank participants allowed us to leverage a large and comprehensive population for model development. This analysis was further enhanced by the high quality of the genetic data utilized to compute the polygenic scores as all SNPs were measured consistently using the same genotyping arrays.
Despite the EstBB cohort being a relatively large sample, a key limitation is the varied follow-up time for some individuals (7.3 years being the average follow-up time) and potential selection effects, due to volunteer-based sampling scheme. In addition, not all predictors had been uniformly measured for every individual, and the demographic composition, including ethnic and racial representation, reflects the Estonian population rather than Europe as a whole. To assure that the best possible model based on conventional predictors is used in both cohorts, accounting for possible differences from a random population-based sample, we did not rely on standardized risk-prediction algorithms such as SCORE2 but instead developed the models that provided the best fit for our data. To maintain simplicity, we chose not to include data on cardiovascular disease (CVD) treatment nor calculate the Charlson Comorbidity Index in the model [38].
Conclusions
The results of this study emphasize the importance of using polygenic risk scores in combination with traditional risk factors for identifying individuals at high risk for atherosclerotic cardiovascular disease. From a primary prevention perspective, polygenic risk scores allow for the early assessment of risk, enabling the implementation of proactive prevention strategies aimed at reducing the burden of cardiovascular disease, particularly in a younger age.
Supporting information
S2 Table. Hazard ratios and p-values from CVD risk models with and without the PRS.
The models are derived in the full data of the earlier and later cohort using sex-specific and sex-stratified analysis for age groups 25–59 and 60 + , respectively.
https://doi.org/10.1371/journal.pone.0335064.s003
(PDF)
S3 Table. Model discrimination in age and recruitment groups.
The models are derived and the C-indices are calculated using the full data from the earlier and later cohort, with sex-specific and sex-stratified analyses for age groups 25–59 and 60 + , respectively.
https://doi.org/10.1371/journal.pone.0335064.s004
(PDF)
S4 Table. Categorical reclassification of 5-year CVD risk.
The conventional model with and without PRS are compared separately in 25–59 and 60 + age group.
https://doi.org/10.1371/journal.pone.0335064.s006
(PDF)
S1 Fig. Calibration of absolute CVD risk in validation set.
The calibration is assessed by sex, age group, and recruitment period. 10- and 5-year CVD risks are used for the earlier and the later cohort, respectively.
https://doi.org/10.1371/journal.pone.0335064.s007
(PDF)
Acknowledgments
The Estonian Biobank Research Team, including Andres Metspalu, Lili Milani, Tõnu Esko, Reedik Mägi, Mait Metspalu, Mari Nelis, and Georgi Hudjashov, contributed to data collection, genotyping, QC, and imputation, while Priit Palta, Nele Taba, Erik Abner, Jaanika Kronberg, and Urmo Võsa contributed to the generation, development, and QC of the NMR data.
References
- 1. Global Burden of Cardiovascular Diseases and Risks 2023 Collaborators. Global, Regional, and National Burden of Cardiovascular Diseases and Risk Factors in 204 Countries and Territories, 1990-2023. J Am Coll Cardiol. 2025. pmid:40990886
- 2. Weale ME, Riveros-Mckay F, Selzam S, Seth P, Moore R, Tarran WA, et al. Validation of an Integrated Risk Tool, Including Polygenic Risk Score, for Atherosclerotic Cardiovascular Disease in Multiple Ethnicities and Ancestries. Am J Cardiol. 2021;148:157–64. pmid:33675770
- 3. Aragam KG, Dobbyn A, Judy R, Chaffin M, Chaudhary K, Hindy G, et al. Limitations of Contemporary Guidelines for Managing Patients at High Genetic Risk of Coronary Artery Disease. J Am Coll Cardiol. 2020;75(22):2769–80. pmid:32498804
- 4. Rotter JI, Lin HJ. An Outbreak of Polygenic Scores for Coronary Artery Disease. J Am Coll Cardiol. 2020;75(22):2781–4. pmid:32498805
- 5. Li L, Pang S, Starnecker F, Mueller-Myhsok B, Schunkert H. Integration of a polygenic score into guideline-recommended prediction of cardiovascular disease. Eur Heart J. 2024;45(20):1843–52. pmid:38551411
- 6. Elliott J, Bodinier B, Bond TA, Chadeau-Hyam M, Evangelou E, Moons KGM, et al. Predictive Accuracy of a Polygenic Risk Score-Enhanced Prediction Model vs a Clinical Risk Score for Coronary Artery Disease. JAMA. 2020;323(7):636–45. pmid:32068818
- 7. Tada H, Melander O, Louie JZ, Catanese JJ, Rowland CM, Devlin JJ, et al. Risk prediction by genetic risk scores for coronary heart disease is independent of self-reported family history. Eur Heart J. 2016;37(6):561–7. pmid:26392438
- 8. Ding K, Bailey KR, Kullo IJ. Genotype-informed estimation of risk of coronary heart disease based on genome-wide association data linked to the electronic medical record. BMC Cardiovasc Disord. 2011;11:66. pmid:22151179
- 9. St-Pierre J, Zhang X, Lu T, Jiang L, Loffree X, Wang L, et al. Considering strategies for SNP selection in genetic and polygenic risk scores. Front Genet. 2022;13:900595. pmid:36819922
- 10. Khera AV, Chaffin M, Aragam KG, Haas ME, Roselli C, Choi SH, et al. Genome-wide polygenic scores for common diseases identify individuals with risk equivalent to monogenic mutations. Nat Genet. 2018;50(9):1219–24. pmid:30104762
- 11. Thériault S, Lali R, Chong M, Velianou JL, Natarajan MK, Paré G. Polygenic Contribution in Individuals With Early-Onset Coronary Artery Disease. Circ Genom Precis Med. 2018;11(1):e001849. pmid:29874178
- 12. Isgut M, Sun J, Quyyumi AA, Gibson G. Highly elevated polygenic risk scores are better predictors of myocardial infarction risk early in life than later. Genome Med. 2021;13(1):13. pmid:33509272
- 13. Klarin D, Natarajan P. Clinical utility of polygenic risk scores for coronary artery disease. Nat Rev Cardiol. 2022;19(5):291–301. pmid:34811547
- 14. O’Sullivan JW, Raghavan S, Marquez-Luna C, Luzum JA, Damrauer SM, Ashley EA, et al. Polygenic Risk Scores for Cardiovascular Disease: A Scientific Statement From the American Heart Association. Circulation. 2022;146(8):e93–118. pmid:35862132
- 15. Viigimaa M, Jürisson M, Pisarev H, Kalda R, Alavere H, Irs A, et al. Effectiveness and feasibility of cardiovascular disease personalized prevention on high polygenic risk score subjects: a randomized controlled pilot study. Eur Heart J Open. 2022;2(6):oeac079. pmid:36600884
- 16. Kumuthini J, Zick B, Balasopoulou A, Chalikiopoulou C, Dandara C, El-Kamah G, et al. The clinical utility of polygenic risk scores in genomic medicine practices: a systematic review. Hum Genet. 2022;141(11):1697–704. pmid:35488921
- 17. SCORE2 working group and ESC Cardiovascular risk collaboration. SCORE2 risk prediction algorithms: new models to estimate 10-year risk of cardiovascular disease in Europe. Eur Heart J. 2021;42(25):2439–54. pmid:34120177
- 18. Goff DC Jr, Lloyd-Jones DM, Bennett G, Coady S, D’Agostino RB Sr, Gibbons R, 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. 2014;63(25 Pt B):2935–59. pmid:24239921
- 19. Hippisley-Cox J, Coupland C, Brindle P. Development and validation of QRISK3 risk prediction algorithms to estimate future risk of cardiovascular disease: prospective cohort study. BMJ. 2017;357:j2099. pmid:28536104
- 20. Sofogianni A, Stalikas N, Antza C, Tziomalos K. Cardiovascular Risk Prediction Models and Scores in the Era of Personalized Medicine. J Pers Med. 2022;12(7):1180. pmid:35887677
- 21. Tillmann T, Läll K, Dukes O, Veronesi G, Pikhart H, Peasey A, et al. Development and validation of two SCORE-based cardiovascular risk prediction models for Eastern Europe: a multicohort study. Eur Heart J. 2020;41(35):3325–33. pmid:33011775
- 22. Sun L, Pennells L, Kaptoge S, Nelson CP, Ritchie SC, Abraham G, et al. Polygenic risk scores in cardiovascular risk prediction: A cohort study and modelling analyses. PLoS Med. 2021;18(1):e1003498. pmid:33444330
- 23. Riveros-Mckay F, Weale ME, Moore R, Selzam S, Krapohl E, Sivley RM, et al. Integrated Polygenic Tool Substantially Enhances Coronary Artery Disease Prediction. Circ Genom Precis Med. 2021;14(2):e003304. pmid:33651632
- 24.
European Association of Preventive Cardiology. European Risk Regions. Based on SCORE2 and SCORE2-OP risk regions. https://www.heartscore.org/en_GB/heartscore-europe-risk-regions. Accessed 2023 September 9.
- 25. Leitsalu L, Haller T, Esko T, Tammesoo M-L, Alavere H, Snieder H, et al. Cohort Profile: Estonian Biobank of the Estonian Genome Center, University of Tartu. Int J Epidemiol. 2015;44(4):1137–47. pmid:24518929
- 26. Lambert SA, Gil L, Jupp S, Ritchie SC, Xu Y, Buniello A, et al. The Polygenic Score Catalog as an open database for reproducibility and systematic evaluation. Nat Genet. 2021;53(4):420–5. pmid:33692568
- 27. Patel AP, Wang M, Ruan Y, Koyama S, Clarke SL, Yang X, et al. A multi-ancestry polygenic risk score improves risk prediction for coronary artery disease. Nat Med. 2023;29(7):1793–803. pmid:37414900
- 28. Putter H, Fiocco M, Geskus RB. Tutorial in biostatistics: competing risks and multi-state models. Stat Med. 2007;26(11):2389–430. pmid:17031868
- 29.
Gerds T, Ohlendorff J, Ozenne B. Risk regression models and prediction scores for survival analysis with competing risks. 2023. https://CRAN.R-project.org/package=riskRegression
- 30. Benichou J, Gail MH. Estimates of Absolute Cause-Specific Risk in Cohort Studies. Biometrics. 1990;46(3):813.
- 31.
R Core Team, R. [Online]. Available: https://www.r-project.org/
- 32. Visseren FLJ, Mach F, Smulders YM, Carballo D, Koskinas KC, Bäck M, et al. 2021 ESC Guidelines on cardiovascular disease prevention in clinical practice. Eur Heart J. 2021;42(34):3227–337. pmid:34458905
- 33. Tikkanen E, Havulinna AS, Palotie A, Salomaa V, Ripatti S. Genetic risk prediction and a 2-stage risk screening strategy for coronary heart disease. Arterioscler Thromb Vasc Biol. 2013;33(9):2261–6. pmid:23599444
- 34. Yanes T, McInerney-Leo AM, Law MH, Cummings S. The emerging field of polygenic risk scores and perspective for use in clinical care. Hum Mol Genet. 2020;29(R2):R165–76. pmid:32620971
- 35. Schunkert H, Di Angelantonio E, Inouye M, Patel RS, Ripatti S, Widen E, et al. Clinical utility and implementation of polygenic risk scores for predicting cardiovascular disease: A clinical consensus statement of the ESC Council on Cardiovascular Genomics, the ESC Cardiovascular Risk Collaboration, and the European Association of Preventive Cardiology. Eur Heart J. 2025;46(15):1372–83. pmid:39906985
- 36. Marston NA, Pirruccello JP, Melloni GEM, Koyama S, Kamanu FK, Weng L-C, et al. Predictive Utility of a Coronary Artery Disease Polygenic Risk Score in Primary Prevention. JAMA Cardiol. 2023;8(2):130–7. pmid:36576811
- 37.
Ehealth - for continuity of care: proceedings of mie2014. In Studies in health technology and informatics, no. v. 205. Washington, DC: IOS Press, 2014.
- 38. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373–83. pmid:3558716