Polygenic Risk Score Improves the Accuracy of a Clinical Risk Score for Coronary Artery Disease

9 10

The value of polygenic risk scores (PRS) towards improving guideline-recommended clinical 11 risk models for coronary artery disease (CAD) prediction is controversial. Here we examine whether an 12 integrated polygenic risk score improves prediction of CAD beyond pooled cohort equations. 13

An observation study of 291,305 unrelated White British UK Biobank participants enrolled 14 from 2006 to 2010 was conducted. A case-control sample of 9,499 prevalent CAD cases and an equal 15 number of randomly selected controls was used for tuning and integrating of the polygenic risk scores. A 16 separate cohort of 272,307 individuals (with follow-up to 2020) was used to examine the risk prediction 17 performance of pooled cohort equations, integrated polygenic risk score, and PRS-enhanced pooled co-18 hort equation for incident CAD cases. Performance of each model was analyzed by discrimination and 19 risk reclassification using a 7.5% threshold. 20

In the cohort of 272,307 individuals (mean age, 56.7 years) used to analyze predictive accuracy, 21 there were 7,036 incident CAD cases over a 12-year follow-up period. Model discrimination was tested 22 for integrated polygenic risk score, pooled cohort equation, and PRS-enhanced pooled cohort equation 23 with reported C-statistics of 0.640 (95% CI, 0.634-0.646), 0.718 (95% CI, 0.713-0.723), and 0.753 (95% CI, 24 0.748-0.758), respectively. Risk reclassification for the addition of the integrated polygenic risk score to 25 the pooled cohort equation at a 7.5% risk threshold resulted in a net reclassification improvement of 26 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. 2 0.117 (95% CI, 0.102 to 0.129) for cases and -0.023 (95% CI, -0.025 to -0.022) for noncases [overall: 27 0.093 (95% CI, 0.08 to 0.104)]. For incident CAD cases, this represented 14.2% correctly reclassified to 28 the higher-risk category and 2.6% incorrectly reclassified to the lower-risk category. 29

and Relevance: Addition of the integrated polygenic risk score for CAD to the pooled cohort 30 questions improves the predictive accuracy for incident CAD and clinical risk classification in the White 31 British from the UK biobank. These findings suggest that an integrated polygenic risk score may enhance 32 CAD risk prediction and screening in the White British population. 33 34 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 3

35 Cardiovascular disease (CVD) is a major cause of death worldwide.1 Risk estimates for CVD have 36 become particularly important for disease prevention and clinical practice.2,3,4,5 Current guidelines from 37 the American College of Cardiology and American Heart Association suggest lipid-lowering treatments 38 for individuals with greater than a 7.5% 10-year absolute risk of developing CVD based on pooled cohort 39 equations (PCE).6 Because of the central role of accurate risk estimates in CVD prevention, improving ac-40 curacy beyond those already used in clinical practice like PCE, could save lives by better identifying high 41 risk individuals. 42 Substantial advancements have been made over the past decades in identifying genetic variants 43 associated with coronary artery disease (CAD).7,8,9,10 Recent advances in polygenic risk scores (PRSs) have 44 sparked a great interest in enhancing disease risk prediction by using the information on millions of vari-45 ants across the genome.11,12,13,14 However, population health utility of PRSs in CAD risk prediction is con-46 troversial. Several studies have shown that PRSs can improve risk prediction accuracy for incident and 47 prevalent CAD cases compared with individual conventional risk factors15,16 and combing risk prediction 48 models (like PCE) with PRS improves the performance in terms of net reclassification improvement.17 On 49 the other hand, several studies18,19 integrating PRSs into PCE to assess possible clinical utility have con-50 cluded that the current benefits of incorporating PRSs were minimal (although statistically significant) 51 and were not considered clinically significant to warrant their use over current clinical used prediction 52 models. In this manuscript, we investigate why different studies have reached different and controver-53 sial conclusions. Specifically, we analyzed UK Biobank data to test the hypothesis that integrated PRSs 54 leveraging multiple newly developed PRS methods, and several genome-wide association study (GWAS) 55 datasets, can improve risk prediction for CAD over the widely used PCE and thus provide improved clini-56 cal utility in European populations.9,20,21,22,23,24,25 Furthermore, in secondary analysis, we extended our 57 integrated method to analyze its predictive performance in non-European populations. 58 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 4

59 Study Populations 60 Our study utilized the UK Biobank which includes 502,536 participants ranging in age from 40 to 61 69 at baseline recruitment.26 Biomarker data were collected from stored serum and red blood cells, de-62 tails of which are described elsewhere.27 Ethical approval for the UK Biobank study was obtained from 63 the National Health Service’s National Research Ethics Service North West (11/NW/0382). The current 64 research project (application number 48240) was approved by UK Biobank. Our study design is outlined 65 in Figure 1. 66 The primary endpoint for our study was CAD, for which several large GWAS results are availa-67 ble.8,9,25 We limited our primary investigation to unrelated White British individuals (as defined by UK 68 Biobank data-field 22006) to reduce the influence of population heterogeneity and related samples; un-69 related individuals were obtained by only keeping individuals with no relative 3rd degree or closer.28 We 70 further excluded outliers for heterozygosity or genotype missing rates (0.2> missing rate). Participants 71 with inconsistent reported and genotypic inferred sex and withdrawn consent were likewise removed. 72 In secondary analysis, we focused on African ancestry participants in the UK Biobank. Following 73 others,29,30 we used imputed data released by the UK Biobank to determine continental ancestry (African 74 (AFR), East Asian (EAS), European (EUR), South Asian (SAS)) and projected participants onto genetic prin-75 cipal components calculated in the 1000 Genome Project (N= 2000: AFR = 504; EAS = 504; EUR = 503; 76 SAS = 489). We excluded populations identified as African Caribbeans in Barbados (ACB) and Americans 77 of African Ancestry in SW (ASW) from the AFR population and all individuals of American ancestry (AMR) 78 due to their complex admixture patterns. Participants were assigned to ancestries based on likelihoods 79 calculated from their first 5 principal components. Samples were assigned via random forest to an an-80 cestry when their likelihood for a given ancestry was > 0.3. If two ancestries exceeded 0.3, we assigned 81 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 5 ancestries as: AFR over EUR, SAS over EUR, and EUR over EAS. Participants were excluded if no likeli-82 hood was > 0.3 or if 3 ancestry groups were > 0.3 (n = 8). The same quality control used in primary analy-83 sis was then applied to the resulting AFR ancestry population. 84 The study population was divided into (1) a case-control study (tuning dataset) established from 85 prevalent CAD cases (see Cardiovascular Outcome Definitions Subsection for details) and randomly se-86 lected controls and (2) an independent prospective cohort study (testing dataset) of participants with no 87 history of CAD at baseline recruitment. The tuning dataset was used for building risk prediction models 88 and the testing dataset was used for unbiasedly evaluating their performance. Of note, there were no 89 overlapping participants between these two datasets, ensuring the testing results were valid. 90 Definition of Risk Score Variables 91 The updated pooled cohort equation (PCE) model, a clinically used risk prediction model, was 92 used as our baseline. We matched variables available in the cohort to the predictors of the updated 93 PCE,3 including information on age, sex, race and ethnicity, smoking status, total and HDL cholesterol, 94 systolic blood pressure, diabetes, and the use of lipid lowering and blood pressure lowering medica-95 tions. Definitions for Type 1 and Type 2 diabetes, blood pressure lowering and lipid lowering medica-96 tions use as well as categorization of smoking status were defined based on UK-recommended QRISK3 97 scores.31,32 Details of variable definitions and protocol for handling missing values are relegated to the 98 eMethods section of Supplementary. 99 Figure 1. Study Design and Flowchart for Coronary Artery Disease (CAD) 100 A. Selection of PRS in case-control study 101 102 B. Cohort Study 103 104 105 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 6 Cardiovascular Outcome Definitions 106 The UK Biobank data have been linked to Hospital Episode Statistics (HES) and national death 107 and cancer registries. HES records diagnosis information via International Classification of Diseases 108 (ICD)-9th and 10th Revisions and codes operative procedures via OPCS-4. Death registries include death 109 date and both primary and secondary causes of death coded in ICD-10. We defined CAD by combining 110 HES, death registries, operation codes,31,32 as well as related self-reported diagnoses and previous proce-111 dure codes (Supplementary Tables 1 and 2). Following others,18 CAD was defined as myocardial infarc-112 tion, including related sequelae. 113 The date of event was determined via recorded episode date, admission date, or operation date 114 indicated in the hospital statistics. For participants with multiple CAD event dates, the earliest recorded 115 date were used as the dates of event. Age of event was determined by self-reported age and calculated 116 age based on the date of event; when both ages were available, the smaller value was used.15 Prevalent 117 cases at baseline were defined as individuals with age of event earlier than age at UK Biobank enroll-118 ment time. Follow-up time was calculated as the number of days from assessment date until the event 119 of interest (CAD event), a competing cause of death, or censorship date (2020/12/31) occurred. 120 Polygenic Risk Scores (PRS) 121 Information on genotyping and imputation has been described in detail elsewhere.27,33 Standard 122 quality-control procedures were applied to the imputed UK Biobank genotype data. Briefly, we re-123 stricted our analyses to autosomal genetic variants, kept variants with imputation information score 124 (INFO) score > 0.3, minor allele frequency > 1%, Hardy-Weinberg equilibrium P > 10-10, and genotype 125 missing rate < 10%. We further removed variants with ambiguous strands (A/T or C/G). 126 PRS for CAD were derived as weighted sums of risk alleles using 3 CAD GWAS datasets (CARDIo-127 GRAMplusC4D, FinnGen Biobank, Japan Biobank) that had no overlap with the present UK Biobank study 128 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 7 (Figure 1).8,9,25 The 3 GWAS datasets were filtered to only include SNPs present in the imputed UK Bi-129 obank data. For all datasets, we aligned β and allele frequencies to the hg19 alternate allele. First, we 130 performed a fixed-effect meta-analysis focused on GWAS datasets with subjects of European ancestry, 131 specifically the CARDIoGRAMplusC4D and FinnGen datasets, using METAL.34 Second, the PRSs were cal-132 culated by using either Japan Biobank data or combined European data and their corresponding popula-133 tion-specific 1000 Genome Project constructed LD reference panels. 134 Tuning of the PRS was implemented using seven methods: (1) clumping and thresholding using 135 PRSice-2 software (version 2.3.3),35 (2) LDpred,20 (3) lassosum,36 (4) PRS-CS ,21 (5) sBayesR,22 (6) LDpred-136 funct,23 and (7) DBSLMM.24 Detailed information on each PRS method and their associated parameters 137 are described in the eMethods section of Supplementary. All methods utilized were adjusted for geno-138 type measurement batch and the first five genetic principal components calculated by the UK Biobank. 139 Since different PRS methods and datasets may capture different information, we constructed the inte-140 grated PRS by ∑ 𝛽#! 𝑃𝑅𝑆! " !#$ , where 𝛽#! is the estimated coefficient of 𝑃𝑅𝑆! in the logistic regression using 141 the tuning dataset and 𝑃𝑅𝑆! is the jth PRS.37 Selection of PRS methods for the integrated model was de-142 termined based on are under the curve (AUC) results from the tuning dataset. Methods with the largest 143 AUC improvement over the PCE model were selected and analyzed in the testing dataset until the inclu-144 sion of additional PRS methods failed to improve the predictive performance of the integrated model. 145 We assessed the performance of the integrated model against the individual PRS methods in the testing 146 dataset as well as models combining the European meta-analysis data and Japan Biobank data. 147 Statistical Analysis 148 Participants were excluded from the study for multiple factors, including missing genetic data, 149 mismatches in reported and genotypic sex, withdrawal of informed consent, and missing predictor val-150 ues. Using previously published baseline coefficients for each predictor variable and baseline hazard, we 151 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 8 calculated the updated pooled cohort equation scores (PCE).3 We examined several models as defined in 152 previous studies18,19: (1) PCE; (2) (integrated) PRS for CAD; and (3) PCE and (integrated) PRS. We per-153 formed Cox proportional hazard regression using follow-up time as the time variable in the testing data. 154 As a sensitivity analysis, all models were reexamined after removing participants that reported taking 155 lipid-lowering medications at baseline of the UK Biobank study. 156 We examined the discrimination of each model via Harrell’s C-statistic and its 95% confidence 157 interval. 38,39,40 In brief, the C-statistic is a measure of the discriminatory power of a risk prediction 158 model, with values ranging from 0.5 (no discrimination) to 1.0 (perfect discrimination). Calibration and 159 recalibration of the baseline models were graphically assessed by comparing observed probabilities via 160 Kaplan-Meier estimates to the mean predicted probability within tenths of the predicted probabilities. 161 During recalibration, the baseline survival function was estimated in the testing cohort and combined 162 with predicted hazard ratios from the validation dataset in a Cox model to obtain recalibrated predicted 163 probabilities.3,18 We assessed the recalibration results via the calibration slope and Greenwood-Nam-164 D’Agostino test.41 165 We evaluated risk prediction accuracy using the net reclassification improvement (NRI)42 at a 166 threshold of 7.5% (clinically used in the United States), continuous NRI, and associated integrated dis-167 crimination improvement (IDI).43 These metrics quantify how well a new model (PCE plus PRS) reclassi-168 fies individuals compared to an old model (PCE); a brief explanation of these metrics can be found in the 169 eMethods section of the Supplementary. 170 Statistical analyses were conducted in R software, version 4.0.0 (R Project for Statistical Compu-171 ting).44 Anaconda, version 3.8.3, was also used for PRS methods that utilized Python programming lan-172 guage.45 173

174 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 9 Following removal of participants with missing data and selecting for only unrelated white Brit-175 ish participants, the UK Biobank dataset contained 291,305 participants which were subsequently di-176 vided into case-control and cohort study datasets (Figure 1). The case-control study contained 9,499 177 prevalent CAD cases and an equal number of controls used for tuning of the PRS methods. The inde-178 pendent cohort study was comprised of 272,307 individuals (mean age: 56.7) with 7,036 incident cases. 179 Participants with CAD at baseline were not included in the cohort study population. The cohort study 180 had a median follow-up time of 12.33 years (interquartile range, 1.42), while incident CAD cases had a 181 median follow-up time of 5.02 years (interquartile range, 4.07). Baseline characteristics (such as age, 182 smoking status, cholesterol, and systolic blood pressure) were similar for participants included in the co-183 hort analysis and excluded due to missing covariates (Supplementary Tables 3A-3C). 184 For the case-control study, each PRS method for CAD was performed across multiple parameter 185 settings to determine optimal values that would be combined for the cohort study. We classified the 186 “optimal” parameter values as those achieving the highest AUC values for that individual method. Spe-187 cific details on each method’s tuning parameters and individual AUC values were provided in Supple-188 mentary Tables 4A and 4B for the European meta-analysis (EUR) and Japan Biobank (Japan) datasets. 189 We combined the PRS for CAD based on the combination of the three GWAS datasets. As expected, be-190 cause the combined EUR + Japan methods fully utilized all three GWAS datasets and several comple-191 mentary PRS methods, it achieved the highest AUC [0.641 (95% CI, 0.635-0.648)] and thus we focused 192 on this PRS (denoted by integrated CAD PRS or simply PRS) for the remaining analysis. The maximal inte-193 grated CAD PRS model for this study was determined to include the EUR and Japan derived clumping 194 and thresholding, LDpred, lassosum, PRS-CS, and LDpred-funct methods. During this step, we evaluated 195 the PRS methods for collinearity concerns and determined the different methods tended to not be 196 highly correlated (Supplementary Figure 1). 197 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 10 In the cohort analysis, following selection of white British participants, as well as excluding indi-198 viduals with missing data, and selecting the case-control subjects, 272,307 participants were used. The 199 discrimination of the integrated CAD PRS remained similar as that in tuning case-control study; the C-200 statistic for the integrated CAD PRS was 0.640 (95% CI, 0.634 -0.646) (Table 1). The discrimination of the 201 PCE (C-statistics, 0.718 [95% CI, 0.713-0.723]) was higher than the integrated CAD PRS. Addition of indi-202 vidual PRSs to the PCE resulted in improved discrimination of the model with PRS-CS applied to the Eu-203 ropean meta-analysis showing the highest discrimination (C-statistics, 0.749 [95% CI, 0.744-0.754]) (Sup-204 plementary Table 5A-5C). We observed the most significant improvement in discrimination when the 205 integrated CAD PRS were added to the PCE, showing a C-statistic increase to 0.753 (95% CI, 0.748-206 0.758), an associated change over the PCE alone of 0.035 (95% CI, 0.03 – 0.04) (Table 1 and Figure 2). 207 We further stratified the population by age group (younger and older than 55 years of age) and sex (men 208 and women) separately and observed higher discrimination in women than men and higher discrimina-209 tion in the younger age group than in the older age group (Table 1). Participants that were not receiving 210 lipid-lowering medication at baseline were also examined and demonstrated similar discrimination per-211 formance (Table 1). 212 Table 1. C-Statistics for Coronary Artery Disease for Full Population and Stratified by Sex and Age 213 Group (Younger and Older than 55 Years of Age) A,B 214 215 216 Figure 2. Receiver Operator Characteristic Curves and C-Statistics for Different Models in Cohort Analyses 217 of White British and African Ancestry Populations 218 219 220 When evaluating model performance, we compared observed and predicted cumulative inci-221 dences of CAD events across each tenth of predicted risk and determined the addition of our integrated 222 PRS method to the baseline model overestimated risk. Following others,17,18 we recalibrated the model 223 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 11 by fitting predicted log-HRs as covariates in the model, resulting in considerable improvement in model 224 calibration (Supplementary Figure 2). 225 We investigated the potential of the PRS-enhanced PCE model in the risk assessment of CAD. 226 We found that an individual’s integrated CAD PRS were generally uncorrelated (Pearson correlation co-227 efficient r, 0.01) with their PCE, which partially explains why adding integrated CAD PRS to PCE model 228 (denoted by PRS-enhanced PCE) improves the discrimination power. We evaluated the hazard ratios HR 229 via a Cox regression. The PCE model had an adjusted HR of 1.653 (95% CI: 1.628-1.679) per standard de-230 viation increase (P < .001) while the PRS-enhanced PCE model reported an adjusted HR of 1.77 (95% CI: 231 1.745 - 1.796) per standard deviation increase of PRS (P < .001). The PRS-enhanced PCE model further 232 improves the discrimination power of PCE model (Figure 3). For example, in the PRS-enhanced PCE 233 model, there was a 7.77-fold (95% CI: 7.61- to 7.92-fold) risk of CAD for individuals in the top quintile 234 compared to those in the bottom quintile. The PCE model, in comparison, reported a 5.29-fold (95% CI: 235 5.21- to 5.39-fold) risk of CAD between the top and bottom quintiles. 236 Figure 3. Cumulative Absolute Risk of Developing CAD 237 238 After adding PRS for CAD to the PCE model, predicted risk changed by greater than 1% for 35.5% 239 of participants while changing by 5% or greater for 1.9% of participants (Figure 4A). There were 7,005 240 incident CAD cases and 256,072 noncases at the 10-year follow-up; 9,230 individuals were censored due 241 to lack of disease or follow-up at 10 years. At the suggested 7.5% risk threshold, 992 of 7,005 cases 242 (14.2%) were correctly reclassified to the higher-risk category and 182 of 7,005 cases (2.6%) were incor-243 rectly moved to the lower-risk category. For non-case participants, 3,443 of 256,072 (1.3%) were cor-244 rectly moved down to the lower-risk category while 9,331 of 256,072 (3.6%) were incorrectly moved up 245 to the high-risk category (Figure 4B). 246 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 12 When comparing integrated PRS for CAD model to the PCE model, the NRI for cases was 11.7% 247 (95% CI, 10.2% to 12.9%) and –2.3% (95% CI, -2.5% to –2.2%) for noncases (Figure 4C). Following the ad-248 dition of the integrated CAD PRS to PCE, the IDI metric indicated an increase in risk difference between 249 cases and noncases of 0.056 (95% CI, 0.053 to 0.059) (Figure 4C). Stratification by sex indicated higher 250 NRI improvement in men over women; stratification by age group saw similar overall NRI improvement 251 (Supplementary Table 6). 252 Figure 4. Change in Predicted Probabilities and Risk Reclassification 253 A. Difference between 10-y risk by PCE and PRS-enhanced PCE 254 B. PCE + PRS 10-year risk reclassification 255 256 C. Net reclassification improvement and integrated discrimination improvement results 257 Secondary Analyses 258 There were 6,971 participants in the AFR ancestry population that were divided into case-con-259 trol and cohort datasets. The case-control dataset consisted of 109 prevalent CAD cases and an equal 260 number of controls. The cohort population was composed of 6,753 participants (median follow-up: 261 12.75, interquartile range: 1.25) in which 88 incident CAD cases were observed (median follow-up: 5.97, 262 interquartile range: 3.3). Baseline characteristics are presented in Supplementary Tables 3D-3F. 263 In case-control analysis, the optimized integrated CAD PRS model that achieved the highest AUC 264 (0.717 [95% CI, 0.644-0.769]) was determined to include the EUR clumping and thresholding, LDpred, 265 PRS-CS, and LDpred-funct methods as well as the Japan LDpred, PRS-CS, and sBayesR methods. In cohort 266 analysis, the integrated CAD PRS C-statistic was 0.542 (95% CI, 0.485-0.6) (Figure 1). Discrimination of 267 the PCE model (0.714 [95% CI, 0.659-0.769]) outperformed the integrated CAD PRS. In contrast to the 268 White British population, the incremental value of the addition of the integrated CAD PRS to the PCE 269 model was minimal (increase in C-statistic, .002 [95% CI, 0.006 to -0.001]) (Table 1). We further stratified 270 by gender and age and observed higher discrimination in women and in the older age group, however, 271 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 13 we noticed slightly greater improvement in discrimination with the addition of our integrated CAD PRS 272 in both men and the younger age group. Participants not on lipid-lowering medication at baseline saw 273 slightly higher, but still minimal discrimination improvement than the full population (Table 1). NRI and 274 IDI metrics were likewise minimal and incrementally smaller than in the White British population (Sup-275 plementary Table 7). 276 277

278 In our analysis, the addition of genetic information to the PCE clinical risk score was associated 279 with a moderate improvement in predictive accuracy for CAD. Addition of PRS to the baseline PCE model 280 resulted in a 3.5% improvement in model concordance as well as a 9.3% net reclassification improve-281 ment (NRI) of incident CAD cases and noncases over the baseline PCE model at a 7.5% risk threshold. In 282 comparison, the integrated risk tool17 and Elliott et al.18 achieved 2.7% (in European population) and 283 4.0% (in all UK Biobank subjects) improvement in terms of NRI, respectively. While both studies improve 284 the performance by integrating PRS into PCE, they reached different conclusions regarding its clinical 285 utility, highlighting the importance of building a more powerful and accurate risk prediction model. 286 Our studies are innovative and are different from existing studies evaluating the clinical utility of 287 adding PRS over existing clinical risk models18,19,46,47 in the following aspects. While matching our defini-288 tion of CAD to that of a previous study performed with the UK Biobank,18 we were able to take ad-289 vantage of more recent incident CAD data. We also utilized a stricter definition for our target population 290 in the UK Biobank data as opposed to the entire UK Biobank data, which contain individuals of diverse 291 ancestry. Recent studies have shown population-specific bias and limited use of specific PRS methods 292 when used on non-European populations.48,49 We also used three distinct GWAS datasets to build the 293 PRS and integrated results from several advanced and more recent PRS methods,21,22,23,24 improving the 294 discrimination power of our integrated CAD PRS. 295 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 14 We found that integrating PRS to baseline PCE model resulted in significant continuous and cat-296 egorical NRI. Categorical NRI for incident cases was 11.7% and -2.3% for noncases. Our model greatly 297 improved reclassification for cases over previous studies,17,18,19 but resulted in more misclassification in 298 non-case individuals. This difference in performance for noncases may be due in part to model specifica-299 tions and cohort selection. In contrast to Moseley et al,19 in which the 2013 PCE model was used, we uti-300 lized the updated 2018 PCE as our baseline. The 2013 model was noted to overestimate risk across all 301 risk groups, prompting the development of the updated PCE model.3 We also used a younger cohort 302 compared to the two cohorts in Moseley et al (mean age 56.7 years compared to 62.9 and 61.8, respec-303 tively). As noted, we included only White British ancestry in our primary cohort. The inclusion of other 304 ethnicities in the cohort may significantly decrease the discrimination power of the PRS constructed. 305 This is shown in our secondary analysis of African ancestry, where the PRS results based on a European 306 ancestry GWAS dataset vastly underperformed compared to the White British population (C-statistics 307 0.715 vs 0.752, respectively) (Supplementary Table 5). 308 Our results suggest an association between predictive accuracy of PRS and incident CAD events 309 that varies based on age and sex. Men showed significantly higher C-statistic improvement than women 310 (0.051 vs 0.035) in the PRS-enhanced PCE model over the baseline PCE model. This is complemented by 311 an 11.6% overall categorical NRI improvement in men compared to 3.6% in women (Supplementary Ta-312 ble 6). Recent studies using PRS in the UK Biobank demonstrated comparable results with higher risks 313 for incident CAD in men than women.15,47,50 The improved performance in men may be attributed to 314 overrepresentation of male CAD cases in the case-control and cohort studies. The use of sex-specific 315 data may lead to improved prediction accuracy of PRS. 316 Our results also suggest a genetic component to early-onset cases of CAD and a possible applica-317 tion of PRS in identifying individuals at heightened risk of these cases, as the predictive accuracy of inci-318 dent CAD cases was higher in participants < 55 years of age. The observed C-statistic for the integrated 319 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 15 PRS-enhanced PCE model was 0.793 compared to 0.705 observed in the ≥ 55 age group. This observa-320 tion supports two recent studies that found high risk score predictions in genetic variants strongly asso-321 ciated with early-onset CAD (<40 years old) as well as improved risk classification of early-onset CAD to 322 higher-risk categories that were not classified as such by PCE.9,51 323 There are limitations in our study. First, our study was conducted in the UK Biobank and is, 324 therefore, limited by the characteristics of the cohort. The UK Biobank cohort is composed of primarily 325 European ancestries (further restricted to White British ancestry in this study) and limited to an age 326 range of 40 to 69 years, restricting its application to other ancestries and age groups. In addition, partici-327 pants in the UK Biobank assessment tend to be healthier and more well-off compared to the general UK 328 population,52 and thus population-level CAD risk may be underestimated in our study. In secondary anal-329 ysis, the limited genetic diversity of the UK Biobank cohort is apparent and resulted in significantly 330 smaller tuning and testing. The extent to which our results can be applied to larger non-European ances-331 tries, in particular African ancestry, warrants further investigation. These results also highlight the ur-332 gency of developing novel cross-ancestry PRS methods10,17,53,54,55 and using more diverse cohorts to con-333 struct PRSs.17 In addition, as the case-control and cohort analyses are derived from the same study, 334 more broad generalizability of the results requires further investigation. Second, this study included PRS 335 for low frequency and common genetic variants (MAF ≥ 1%) and did not examine the predictive accu-336 racy of rare variants known to affect CAD risk. Third, the algorithm for selection of CAD cases utilizes 337 self-report, death, and hospital inpatient data for the definition of prevalent and incident CAD cases. As 338 such, misclassification of cases is possible. Fourth, tuning of each PRS method in the case-control study 339 used prevalent CAD cases, which could introduce survival bias. However, simulation studies have 340 demonstrated a limited effect of survival bias on estimated genetic effects of event risks.56 Fifth, partici-341 pants with at least 1 missing predictor value were excluded from the study. Excluded participants were 342 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 16 not considerably different demographically from those included and thus the missing data are unlikely 343 to have a significant effect on the reported estimates. 344

345 Addition of the integrated CAD PRS to the PCE resulted in a statistically significant improvement 346 in predictive accuracy for incident CAD, especially in individuals under the age of 55 years old in White 347 British population. It was also associated with moderate improvement in risk reclassification across all 348 subgroups. However, the benefits of adding integrated CAD PRS to the PCE are minimal for African pop-349 ulation. In summary, the inclusion of genetic information to the pooled cohort equation can help im-350 prove clinical risk classification and demonstrates the potential for genetic screening in early life to im-351 prove clinical risk prediction in White British population. 352 353 354 Abbreviations 355 Cardiovascular disease (CVD); pooled cohort equations (PCE); coronary artery disease (CAD); polygenic 356 risk score (PRS); genome-wide association study (GWAS); African (AFR); East Asian (EAS); European 357 (EUR); South Asian (SAS); hospital episode statistics (HES); area under the curve (AUC); net reclassifica-358 tion improvement (NRI); integrated discrimination improvement (IDI); confidence interval (CI) 359

360 1. GBD 20019 Disease and Injuries Collaborators. Global burden of 369 diseases and injuries in 204 361 countries and territories, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. 362 Lancet. 2020; 396(10258):1204-1222. 363 2. Damen JA, Hooft L, Schuit E, et al. Prediction models for cardiovascular disease risk in the general 364 population: systematic review. BMJ. 2016;353:i2416. 365 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 17 3. Yadlowsky S, Hayward RA, Sussman JB, McClelland RL, Min YI, Basu S. Clinical implications of revised 366 pooled cohort equations for estimating atherosclerotic cardiovascular disease risk. Ann Intern Med. 367 2018;169(1):20-29. 368 4. Conroy R, Pyörälä K, Fitzgerald A, et al. Estimation of ten-year risk of fatal cardiovascular disease in 369 Europe: The SCORE project. European Heart Journal. 2003;24(11):987-1003. 370 5. D’Agostino R, Vasan R, Pencina M, et al. General Cardiovascular Risk Profile for Use in Primary Care. 371 Circulation. 2008;117(6):743-753. 372 6. Arnett DK, Blumenthal RS, Albert MA, et al. 2019 ACC/AHA guideline on the primary prevention of 373 cardiovascular disease: a report of the American College of Cardiology/American Heart Association Task 374 Force on Clinical Practice Guidelines. J Am Coll Cardiol. 2019;74(10):e177-e232. 375 7. Musunuru K, Kathiresan S. Genetics of common, complex coronary artery disease. Cell. 376 2019;177(1):132-145. 377 8. Nikpay M, Goel A, Won HH, et al. A comprehensive 1,000 Genomes-based genome-wide association 378 meta-analysis of coronary artery disease. Nat Genet. 2015;47(10):1121-1130. 379 9. Mars N, Koskela J, Ripatti P, et al. Polygenic and clinical risk scores and their impact on age at onset 380 and prediction of cardiometabolic diseases and common cancers. Nat Med. 2020;26(4).549-557. 381 10. Koyama S, Ito K, Terao C, et al. Population-specific and trans-ancestry genome-wide analyses identify 382 distinct and shared genetic risk loci for coronary artery disease. Nat Genet. 2020;52(11):1169-1177. 383 11. Knowles J, Ashley E. Cardiovascular disease: The rise of the genetic risk score. PLOS Medicine. 384 2018;15(3):e1002546. 385 12. Torkamani A, Wineinger N, Topol E. The personal and clinical utility of polygenic risk scores. Nature 386 Review Genetics. 2018;19(9):581-590. 387 13. Wise A, Manolio T, Mensah G, et al. Genomic medicine for undiagnosed diseases. Lancet. 388 2019;394(10197):533-540. 389 14. Claussnitzer M, Cho J, Collins R, et al. A brief history of human disease genetics. Nature. 390 2020;577(7789):179-189. 391 15. Inoyue M, Abraham G, Nelson C, et al. Genomic Risk Prediction of Coronary Artery Disease in 392 480,000 Adults: Implications for Primary Prevention. J Am Coll Cardiol. 2018;72(16):1883-1893. 393 16. Khera AV, Chaffin M, Aragam KG, et al. Genome-wide polygenic scores for common diseases identify 394 individuals with risk equivalent to monogenic mutations. Nat Genet. 2018;50(9):1219-1224. 395 17. Weale M, Riveros-McKay F, Selzam S, et al. Validation of an Integrated Risk Tool, Including Polygenic 396 Risk Score, for Atherosclerotic Cardiovascualr Disease in Multiple Ethnicities and Ancestries. Am J Cardiol. 397 2021;148:157-164. 398 18. Elliott J, Bodinier B, Bond T, et al. Predictive Accuracy of a Polygenic Risk Score-Enhanced Prediction 399 Model vs a Clinical Risk Score for Coronary Artery Disease. JAMA. 2020;323(7):636-645. 400 19. Mosley J, Gupta D, Tan J, et al. Predictive Accuracy of a Polygenic Risk Score Compared with a Clini-401 cal Risk Score for Incident Coronary Heart Disease. JAMA. 2020;323(7):627-635. 402 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 18 20. Vilhjálmsson B, Yang J, Finucane H, et al. Modeling Linkage Disequilibrium Increases Accuracy of Pol-403 ygenic Risk Scores. American Journal of Human Genetics. 2015;97(4):576-592. 404 21. Ge T, Chen C, Ni Y, et al. Polygenic prediction via Bayesian regression and continuous shrinkage pri-405 ors. Nature Communications. 2019;10(1). 406 22. Lloyd-Jones L, Zeng J, Sidorenko J, et al. Improved polygenic prediction by Bayesian multiple regres-407 sion on summary statistics. Nature Communications. 2019;10(1). 408 23. Márquez-Luna C, Gazal S, Loh P, et al. LDpred-funct: incorporating functional priors improves poly-409 genic prediction accuracy in UK Biobank and 23andMe data sets. bioRxiv. 2018;375337. 410 24. Yang S, Zhou X. Accurate and Scalable Construction of Polygenic Scores in Large Biobank Data Sets. 411 American Journal of Human Genetics. 2020;106(5):679-693. 412 25. Nagai A, Hirata M, Kamatani Y, et al. Overview of the BioBank Japan Project: Study design and pro-413 file. Journal of Epidemiology. 2017;27(3):S2-S8. 414 26. Sudlow C, Gallacher J, Allen N, et al. UK biobank: an open access resource for identifying the causes 415 of a wide range of complex diseases of middle and old age. PLoS Med. 2015;12(3):e1001779. 416 27. UK Biobank. Biomarker assay quality procedures: approaches used to minimize systematic and ran-417 dom errors (and the wider epidemiological implications): version 1.2. https://biobank.ctsu.ox.ac.uk/crys-418 tal/cyrstal/docs/biomarker_issues.pdf. Published April 2, 2019. Accessed August 10, 2021. 419 28. Bycroft C, Freeman C, Petkova D, et al. The UK Biobank resource with deep phenotyping and ge-420 nomic data. Nature. 2108;562(7726):203-209. 421 29. Wang Y, Guo J, Ni G, et al. Theoretical and empirical quantification of the accuracy of polygenic 422 scores in ancestry divergent populations. Nature Communications. 2020;11:3865. 423 30. Backman J, Li A, Marcketta A, et al. Exome sequencing and analysis of 454,787 UK Biobank partici-424 pants. Nature. 2021;599:628-634. 425 31. Hippisley-Cox J, Coupland C, Vinogradova Y, et al. Predicting cardiovascular risk in England and 426 Wales: prospective derivation and validation of QRISK2. BMJ. 2008;336(7659):1475-1482. 427 32. Hippisley-Cox J, Coupland C, Brindle P. Development and validation of QRISK3 risk prediction algo-428 rithms to estimate future risk of cardiovascular disease: prospective cohort study. BMJ. 2017;357:j2099. 429 33. UK Biobank. Genotype imputation and genetic association studies of UK Biobank: interim data re-430 lease. http://www.ukbiobank.ac.uk/wp-content/uploads/2014/04/imputation_documentation_May2015.pdf. 431 Published May 2015. Accessed January 10, 2019. 432 34. Willer C, Li Y, Abecasis G. METAL: Fast and efficient meta-analysis of genome wide association scans. 433 Bioinformatics. 2010;26(17):2190-2191. 434 35. Choi S, Mak T, O’Reilly P. Tutorial: a guide to performing polygenic risk score analyses. Nature Proto-435 cols. 2020;15(9):2759-2772. 436 36. Mak T, Porsch R, Choi S, et al. Polygenic scores via penalized regression on summary statistics. Ge-437 netic Epidemiology. 2017;41(6):469-480. 438 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 19 37. Wu C, Zhu J, King A, et al. Novel strategy for disease risk prediction incorporating predicted gene ex-439 pression and DNA methylation data: a multi-phased study of prostate cancer. Cancer Communications. 440 2021;41(12):1387-1397. 441 38. SOMERSD. Stata module to calculate Kendall’s tau-a, Somers’ D. and median differences [computer 442 program]. Version S336401: Boston College Department of Economics; 1998. 443 39. Harrell FE Jr, Califf RM, Pryor DB, Lee KL, Rosati RA. Evaluation the yield of medical tests. JAMA. 444 1982;247(18):2543-2546. 445 40. Newson R. Parameters behind “nonparametric” statistics: Kendall’s tau, Somers’ D and median dif-446 ferences. Stata J. 2002;2(1):45-64. 447 41. Demler O, Paynter N, Cook N. Tests of calibration and goodness-of-fit in the survival setting. Statis-448 tics in Medicine. 2015;34(10):1659-1980. 449 42. Leening M, Vedder M, Witteman J, et al. Net Reclassification Improvement: Computation, Interpre-450 tation, and Controversies A Literature Review and Clinician’s Guide. Ann Intern Med. 2014;160:122-131. 451 43. Pencina MJ, Steyerberg EW, D’Agostino RB Sr. Net reclassification index at event rate: properties 452 and relationships. Stat Med. 2017;36(28):4455-4467. 453 44. The R Project for Statistical Computing [computer Program]. Version 4.0.0, Vienna, Austria: 2013. 454 45. Anaconda Software Distribution [Internet]. Anaconda Documentation. Anaconda Inc.; 2020. Availa-455 ble from: https:/docs.anaconda.com/ 456 46. Aragam K, Dobbyn A, Judy R. et al. Limitations of Contemporary Guidelines for Managing Patients at 457 High Genetic Risk of Coronary Artery Disease. J Am Coll Cardiol. 2020;75(22):2769-2780. 458 47. Riveros-McKay F, Weale M, Moore R, et al. Integrated Polygenic Tool Substantially Enhances Coro-459 nary Artery Disease Prediction. Circulation: Genomic and Precision Medicine. 2021; 14(2):e003304. 460 48. Gola D, Erdmann J, Läll K, et al. Population Bias in Polygenic Risk Prediction Models for Coronary Ar-461 tery Disease. Circulation: Genomic and Precision Medicine. 2020;13(6):e002932. 462 49. Matsunaga H, Ito K, Akiyama M, et al. Transethnic Meta-Analysis of Genome-Wide Association Stud-463 ies Identifies Three New Loci and Characterizes Population-Specific Differences for Coronary Artery Dis-464 ease. Circulation: Genomic and Precision Medicine. 2020;13(3):e002670. 465 50. Manikpurage H, Eslami A, Perrot N, et al. Polygenic Risk Score for Coronary Artery Disease Improves 466 the Prediction of Early-Onset Myocardial Infarction and Mortality in Men. Circulation: Genomic and Preci-467 sion Medicine. 2021;14(6):e003452. 468 51. Thériault S, Lali R, Chong M, et al. Polygenic Contribution in Individuals With Early-Onset Coronary 469 Artery Disease. Circulation: Genomic and Precision Medicine. 2018;11(1). 470 52. Fry A, Littlejohns T, Sudlow C, et al. Comparison of Sociodemographic and Health-Related Character-471 istics of UK Biobank Participants with Those fo the General Population. American Journal of Epidemiology. 472 2017;186(9):1026-1034. 473 53. Fritsche L, Ma Y, Zhang D, et al. On cross-ancestry cancer polygenic risk scores. PLoS Genet. 474 2021;17(9):e1009670. 475 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 20 54. Chen C, Han J, Hunter D, Kraft P, Price A. Explicit Modeling of Ancestry Improves Polygenic Risk 476 Scores and BLUP Prediciton. Genetic Epidemiology. 2015;39(6):427-438. 477 55. Cai M, Xiao J, Zhang S, et al. A unified framework for cross-population trait prediction by leveraging 478 the genetic correlation of polygenic traits. AJHG. 2021;108(4):632-655. 479 56. Hu YJ, Schmidt AF, Dudbridge F, et al; The GENIUS-CHD Consortium. Impact of selection bias on esti-480 mation of subsequent event risk. Circ Cardiovasc Genet. 2017;10(5):e001616. 481 482 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 21 Figure 1. Study Design and Flowchart for Coronary Artery Disease (CAD) 483 A. Selection of PRS in case-control study 484 485 B. Cohort Study 486 487 488 489 To select the parameters for each method with the best discrimination based on area under the curve (AUC), clumping and thresholding, 490 LDpred, lassosum, PRS-CS, sBayesR, LDpred-funct, and DBSLMM were used to calculate polygenic risk scores (PRS) on the case-control set con-491 sisting of prevalent cases. For these calculations, summary data for three genome-wide association studies (GWAS) on CAD (CARDIoGRAM-492 plusC4D, Finngen Biobank, Japan Biobank) that excluded UK Biobank and data on linkage disequilibrium were used. The calculated PRS were 493 applied to a nonoverlapping set of participants from the UK Biobank with no preexisting CAD, aged 40 to 69 at baseline, and who were followed 494 up for incident CAD events. In this population, the pooled cohort equations (PCE) model was calculated and different models (PRS, PCE, PRS-495 enhanced PCE) were compared in terms of their predictive accuracy based on discrimination, calibration, and reclassification metrics. 496 497 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 22 Table 1. C-Statistics for Coronary Artery Disease for Full Population and Stratified by Sex and Age 498 Group (Younger and Older than 55 Years of Age) 1,2 499 C-Statistic (95% CI) All Men Women Participants Aged <55 y Participants Aged ≥ 55 y Participants Not Receiving Lipid-Lowering Treatment at Baseline A. White British Ancestry Participants, No. 272307 124155 148152 102330 169977 235172 Cases, No. 7036 5093 1943 1276 5760 5091 Polygenic risk score .64 (.634- .646) .643 (.636 -.651) .641 (.629 -.654) .69 (.626 -.705) .632 (.625 -.639) .646 (.638 -.653) Pooled cohort equation .718 (.713 -.723) .663 (.656 -.67) .706 (.695 -.717) .749 (.736 -.761) .665 (.658 -.671) .73 (.724 -.737) Polygenic risk score + pooled cohort equation .753 (.748 -.758) .714 (.708 -.721) .741 (.73 -.751) .793 (.781 -.806) .705 (.699 -.712) .766 (.76 -.772) B. African Ancestry Participants, No. 6753 2901 3852 4528 2225 5896 Cases, No. 88 46 42 42 46 63 Polygenic risk score .542 (.485 -.6) .574 (.494 -.654) .6 (.511 -.634) .548 (.46 -.637) .543 (.462 -.624) .534 (.464 -.604) Pooled cohort equation .714 (.659 -.769) .674 (.595 -.753) .734 (.653 -.815) .657 (.572 -.742) .721 (.656 -.787) .698 (.628 -.768) Polygenic risk score + pooled cohort equation .716 (.665 -.768) .695 (.622 -.768) .732 (.654 -.81) .679 (.597 -.761) .696 (.629 -.763) .707 (.64 -.774) 1. Cox proportional hazard models for CAD using recalibrated polygenic risk score, pooled cohort equations, and both combined 500 models. 2. C-Statistics shown for combined European meta-analysis + Japan Biobank PRS methods. Results are presented for 501 both White British and African ancestry populations. 502 503 504 505 506 507 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 23 Figure 2. Receiver Operator Characteristic Curves and C-Statistics for Different Models in Cohort Analyses 508 of White British and African Ancestry Populations 509 510 PCE indicates pooled cohort equation; PRS indicates integrated polygenic risk score. A) is the White British population of 272,307 individuals 511 over a mean 12 years of follow-up with 7036 incident CAD cases and B) is the African ancestry population of 6,753 individuals over a mean 13 512 years of follow-up with 88 incident CAD cases. 513 514 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 24 Figure 3. Cumulative Absolute Risk of Developing CAD 515 516 Cumulative absolute risk of developing CAD by quintiles of the overall polygenic score in A) the PCE model and B) the PRS-en-517 hanced PCE model. The shaded portions correspond to the 95% confidence interval. 518 519 520 521 522 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 25 Figure 4. Change in Predicted Probabilities and Risk Reclassification 523 A. Difference between 10-y risk by PCE and PRS-enhanced PCE 524 525 526 B. PCE + PRS 10-year risk reclassification 527 PCE + PRS Risk Threshold, % Threshold, % < .75 ≥ 7.5 % Reclassified Improved Classification Cases < 7.5 5098 992 14.2 + ≥ 7.5 182 733 2.6 - Noncases < 7.5 236916 9331 3.6 - ≥ 7.5 3443 6382 1.3 + 528 529 530 531 532 533 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint 26 C. Net reclassification improvement and integrated discrimination improvement results 534 No. of Participants Continuous Net Reclassification Improvement Categorical Net Reclassification Improvement Integrated Discrimination Improvement Cases 7036 0.215 (0.194 to 0.228) 0.117 (0.102 to 0.129) Noncases 256072 0.235 (0.224 to 0.25) -0.023 (-0.025 to - 0.022) Full Population 263108 0.45 (0.423 to 0.478) 0.093 (0.08 to 0.104) 0.0564 (0.0534 to 0.0594) Censored 9230 535 A, Change in the predicted probabilities of the recalibrated pooled cohort equations (PCE) model after the addition of polygenic risk scores 536 (PRS) for CAD. The x-axis shows the predicted probability from the baseline PCE model. The y-axis is the difference in 10-year risk probabilities 537 of a CAD event between the PRS-enhanced model and the baseline PCE model. The scatterplot has a random draw of 1% of the participants 538 shown. The histogram x- and y-axes are based on the full population. B, Reclassification table of predicted probabilities by PCE and PRS-en-539 hanced PCE models at 7.5% threshold. Rows indicating an improved classification with the PRS-enhanced PCE model are marked by a plus sign 540 while rows indicating a deteriorated classification are marked by a minus sign. C, Table of net reclassification improvement (NRI) and integrated 541 discrimination improvement (IDI). NRIa is defined in the continuous case as the sum of proportions of cases and noncases with improved com-542 bined score minus the sum of proportions with deteriorated combined score. In the categorical case, NRI is defined by changed at a 7.5% 543 threshold predicted probability. A positive NRI indicates a better combined score overall. IDIb measures the difference of average probabilities 544 of an event in cases and noncases. A larger IDI indicates more discrimination in the combined score. 545 a NRI = P(up|case) - P(down|case) - P(up|noncase) + P(down|noncase) b IDI = PPCE+PRS(case) - PPCE+PRS(noncase) - PPCE(case) + PPCE(noncase) 546 547 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 3, 2022. ; https://doi.org/10.1101/2022.06.02.22275933doi: medRxiv preprint