Construction And Utility Evaluation Of Comprehensive Risk Prediction Models For Coronary Artery Disease Based On Environmental And Genetic Risks In Chinese Adults

Background and objectiveCoronary artery disease(CAD),with high mortality and morbidity,is the primary factor of mortality and disease burden globally.Both external(environmental factors)and internal(genetic variants)factors influence the development of CAD.Risk assessment and identification of high-risk individuals for preventive intervention using these risk factors is effective to prevent and control CAD.Currently,guidelines at home and abroad recommended constructing atherosclerotic cardiovascular disease(ASCVD)(including CAD and stroke)risk prediction models with environmental factors(age,blood pressure,lipids and glucose,etc)to assess 10-year and lifetime(for example,till age of 80 years)clinical risks,such as the Prediction for Atherosclerotic Cardiovascular Disease Risk in China(China-PAR)10-year risk prediction model in Chinese and the pooled cohort equation(PCE)in American.Compared with traditional environmental factors,genetic variants remain stable during the life course.Through integrating common genetic variants associated with CAD or cardio-metabolic risk factors to construct polygenic risk scores(PRSs)provided the opportunity to evaluate the genetic risk.For example,based on 540 genetic variants of CAD and related traits,a recent study from China constructed a CAD PRS and validated it in cohorts comprising over 40,000 individuals.They found that the PRS had good performance in risk prediction,demonstrating power in early prevention.Moreover,studies from European and north American countries have reported modest improvements in predictive accuracy by adding CAD PRSs to ASCVD clinical risk models,while whether PRSs could refine clinical risk stratification based on contemporary guidelines and inform clinical decision-making for prevention still needed to be clarified.How to assess the CAD risk using existing ASCVD prediction models?Would integrate environmental and genetic risks improve predictive accuracy and identify potential at-risk individuals beyond guideline-based risk assessment and guide clinical decision-making?To test these hypotheses,firstly,using the China-PAR 10-year clinical risk prediction model,the 10-year and lifetime clinical risk prediction models of CAD based on environmental factors were constructed and evaluated in large cohorts.Risk prediction and stratification of the CAD PRS for Chinese was further investigated.Finally,we integrated clinical and genetic risks to construct comprehensive risk prediction models and evaluate the value in precision prevention of CAD.Subjects and methodsIndividuals were from three subcohorts in the China-PAR project:China Multi Center Collaborative Study of Cardiovascular Epidemiology(ChinaMUCA)1998,International Collaborative Study of Cardiovascular Disease in Asia(InterASIA)and Community Intervention of Metabolic Syndrome in China and Chinese Family Health Study(CIMIC).Baseline surveys of the ChinaMUCA 1998 and InterASIA were carried out in 1998 and 2000-2001 respectively,and the first follow-up surveys of these two subcohorts were constructed in 2007-2008,along with the baseline survey of the CIMIC.Subsequently,all three subcohorts were followed up for two times in 2012-2015 and in 2018-2020.41,271 individuals aged 30-74 years with genotyping information and without CAD at baseline were included in this study.Standardized questionnaires were utilized to collect the geographical demographic characteristics,lifestyles,personal and family disease history,medication information and perform physical examination.We collected fasting peripheral blood samples for laboratory tests and genotyping.Six risk factors of CAD were defined:current smoker,overweight,hypertension,diabetes,hypercholesterolemia and family disease history of CAD.In the present study,CAD included the initial occurrence of unstable angina pectoris,nonfatal acute myocardial infarction or CAD deaths.Deaths from other diseases rather than CAD were defined as competing events.Construction and evaluation of the clinical risk prediction models of CAD:we used Cox model and Fine&Gray’s model to recalibrate the China-PAR 10-year clinical risk equation of ASCVD,and to construct the 10-year and lifetime clinical risk prediction models of CAD.We drew calibration curves to compare the predicted risk with the Kaplen-Meier adjusted observed risk,calculated C-statistics and calibration χ2 to evaluate the predictive accuracy of the models.We determined the thresholds for intermediate,high and very high risk of CAD according to the distribution of 10-year clinical risk of ASCVD.As a comparison,we recalibrated the PCE models and compared the performance with the 10year risk prediction model of CAD.Cox model with adjustment for age,sex and the first four genetic principal components was used to evaluate the association between PRS and CAD and the corresponding hazard ratio(HR)and its 95%confidence interval(CI).C-statistics for the PRS and clinical risk factors were calculated.Genetic risk groups were defined by quintiles of the PRS:low(<20%),intermediate(20%-80%)and high(≥80%).We constructed 10-year and lifetime comprehensive risk prediction models integrating clinical and genetic risks and evaluated the utility of CAD PRS in terms of improvement in predictive accuracy and clinical risk stratification over the clinical risk model:(1)C-statistic and net reclassification index(NRI)were calculated when adding PRS to the 10-year risk prediction model of CAD.(2)the 10year and lifetime(till age of 80 years)cumulative comprehensive risks according to the joint categories of genetic and clinical risks of CAD were calculated.The first four genetic principal components were additionally adjusted in both models.ResultsThe age(standard deviation,SD)of 41,271 subjects in the present study was 52.3± 10.6 years,including 42.5%male subjects.We certified 1,303 CAD cases during follow-up(a median of 12.0 years).The predicted events of CAD by the 10-year clinical risk model were close to the Kaplan-Meier adjusted observed events(total population:949.9 versus 899.2,male:463.4 versus 449.1,female:486.5 versus 450.1).In all individuals,male and female,the Cstatistic(95%CI)was 0.756(0.743 to 0.769),0.743(0.724 to 0.761)and 0.763(0.744 to 0.781),and the calibration χ2 was 12.9,10.6 and 10.3(all P>0.05).We defined the thresholds for intermediate,high,and very high clinical risk of CAD as 2.5%,4.5%,and 6.0%,respectively.Compared with the recalibrated China-PAR model,the C-statistics for PCE models were slightly lower.For the CAD lifetime clinical risk prediction model,the C-statistic(95%CI)was 0.728(0.713 to 0.742),0.704(0.683 to 0.725)and 0.732(0.711 to 0.753)in all individuals,male and female,respectively.Compared with any of six conventional factors(current smoker,overweight,hypertension,diabetes,hypercholesterolemia and family disease history of CAD),PRS had the highest C-statistic(C=0.728,95%CI:0.714 to 0.743).For per SD increase in PRS,the relative risk of CAD increased evidently,with the HR(95%CI)of 1.44(1.36 to 1.52)and the effect of PRS was more pronounced in male and hypertensive individuals,with the HR(95%CI)of 1.56(1.44to 1.68)and 1.51(1.40 to 1.63),respectively.Compared to those with low genetic risk,those with high genetic risk conferred a three-fold risk of developing CAD,with the HR(95%CI)of 2.92(2.43 to 3.51).Finally,we integrated clinical and genetic risks to construct 10-year and lifetime comprehensive risk prediction models,and evaluated the improvements in predictive accuracy and clinical risk stratification beyond the clinical risk model.The addition of PRS to the 10-year clinical risk prediction model of CAD modestly improved the predictive accuracy,with the increment in C statistic by 1%(P<0.001),and the NRI(95%CI)of 3.5%(1.2%to 6.0%).Based on the current guidelines for CAD primary prevention,the PRS could also refine clinical risk stratification.Within any given strata of clinical risk,marked gradients in the absolute risk across PRS categories were found.Of note,among intermediate clinical risk individuals with uncertain decision for intervention in primary prevention,the 10-year risk for those with high genetic risk(4.6%,95%CI:4.5%to 4.6%)would meet the threshold of high clinical risk(4.5%).In these individuals,the 10-year risk(4.6%versus 4.8%)and lifetime risk(17.9%versus 16.6%)reached the level of those with high clinical risk and 20%-80%of PRS,and these findings were also observed in all the age and sex-specific groups.ConclusionIn the present study,we recalibrated the guideline-based 10-year ASCVD risk prediction model to construct and evaluate the 10-year and lifetime clinical risk prediction models of CAD for Chinese.PRS was independently associated with incident CAD,displaying good performance in risk prediction and stratification.The clinical and genetic comprehensive risk prediction models could further refine clinical risk stratification and guide clinical decision-making for prevention.Therefore,integrating environmental and genetic risks for early risk screening and primary prevention could open up a new dimension for precision prevention as well as improve the health condition of citizens.