Coronary Risk Stratification, Discrimination, and Reclassification Improvement Based on Quantification of Subclinical Coronary Atherosclerosis

Transcription

1 Journal of the American College of Cardiology Vol. 56, No. 17, by the American College of Cardiology Foundation ISSN /$36.00 Published by Elsevier Inc. doi: /j.jacc Coronary Risk Stratification, Discrimination, and Reclassification Improvement Based on Quantification of Subclinical Coronary Atherosclerosis The Heinz Nixdorf Recall Study Imaging of Coronary Calcium Raimund Erbel, MD,* Stefan Möhlenkamp, MD,* Susanne Moebus, PHD, Axel Schmermund, MD, Nils Lehmann, PHD, Andreas Stang, MD, Nico Dragano, PHD, Dietrich Grönemeyer, MD, Rainer Seibel, MD,# Hagen Kälsch, MD,* Martina Bröcker-Preuss, PHD,** Klaus Mann, MD,** Johannes Siegrist, MD, Karl-Heinz Jöckel, PHD, for the Heinz Nixdorf Recall Study Investigative Group Essen, Frankfurt am Main, Halle, Düsseldorf, Witten, and Mülheim, Germany Objectives Background Methods Results Conclusions The purpose of this study was to determine net reclassification improvement (NRI) and improved risk prediction based on coronary artery calcification (CAC) scoring in comparison with traditional risk factors. CAC as a sign of subclinical coronary atherosclerosis can noninvasively be detected by CT and has been suggested to predict coronary events. In 4,129 subjects from the HNR (Heinz Nixdorf Recall) study (age 45 to 75 years, 53% female) without overt coronary artery disease at baseline, traditional risk factors and CAC scores were measured. Their risk was categorized into low, intermediate, and high according to the Framingham Risk Score (FRS) and National Cholesterol Education Panel Adult Treatment Panel (ATP) III guidelines, and the reclassification rate based on CAC results was calculated. After 5 years of follow-up, 93 coronary deaths and nonfatal myocardial infarctions occurred (cumulative risk 2.3%; 95% confidence interval: 1.8% to 2.8%). Reclassifying intermediate (defined as 10% to 20% and 6% to 20%) risk subjects with CAC 100 to the low-risk category and with CAC 400 to the high-risk category yielded an NRI of 21.7% (p ) and 30.6% (p ) for the FRS, respectively. Integrated discrimination improvement using FRS variables and CAC was 1.52% (p ). Adding CAC scores to the FRS and National Cholesterol Education Panel ATP III categories improved the area under the curve from to (p 0.003) and from to (p ), respectively. CAC scoring results in a high reclassification rate in the intermediate-risk cohort, demonstrating the benefit of imaging of subclinical coronary atherosclerosis. Our study supports its application, especially in carefully selected individuals with intermediate risk. (J Am Coll Cardiol 2010;56: ) 2010 by the American College of Cardiology Foundation Coronary artery disease (CAD) risk prediction in asymptomatic individuals represents a major challenge. Currently, risk stratification algorithms are applied, allowing global risk assessment rather than focusing on single risk factors (1,2). The widely used Framingham algorithm is based on traditional CAD risk factors, including age, sex, smoking, total cholesterol or low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, blood pressure, and diabetes. The lifetime prognostic power of these risk See page 1415 From the *Department of Cardiology, West-German Heart Center Essen, University Duisburg-Essen, Essen, Germany; Institute of Medical Informatics, Biometry, and Epidemiology, University Duisburg-Essen, Essen, Germany; Cardioangiological Center Bethanien, Frankfurt am Main, Germany; Institute of Clinical Epidemiology, Medical Faculty, Martin-Luther-University of Halle-Wittenberg, Halle, Germany; Institute of Medical Sociology, University Düsseldorf, Düsseldorf, Germany; Institute of Diagnostic and Interventional Radiology, University Witten/Herdecke, Witten, Germany; #Diagnosticum, Mülheim, Germany; and the **Department of Endocrinology and Department of Clinical Chemistry and Laboratory Medicine, University Duisburg-Essen, Essen, Germany. This study is supported by the German Ministry of Education and Science (BMBF), and the German AeroCenter (Deutsches Zentrum für Luft- und Raumfahrt [DLR]), Bonn, Germany. Assessment of psychosocial factors and neighborhood level information is funded by the German Research Council (DFG; Project SI 236/8-1 and SI 236/9-1). The authors acknowledge the support of the Sarstedt AG & Co. (Nümbrecht, Germany) concerning laboratory equipment. The authors have reported that they have no relationships to disclose. Presented in part at the Late Breaking Clinical Trial, 58th Annual Scientific Session of the American College of Cardiology, 2009, and American Heart Association Scientific Sessions Manuscript received February 17, 2010; revised manuscript received June 9, 2010, accepted June 15, 2010.

2 1398 Erbel et al. JACC Vol. 56, No. 17, 2010 Reclassification Based on Coronary Calcium Testing October 19, 2010: Abbreviations and Acronyms ATP Adult Treatment Panel AUC area under the curve CAC coronary artery calcification CAD coronary artery disease FRS Framingham Risk Score HDL high-density lipoprotein IDI integrated discrimination improvement LDL low-density lipoprotein NRI net reclassification improvement ROC receiver-operator characteristic RR relative risk factors is well established. However, their prognostic effects depend on the duration of exposure and on the magnitude of the deviation of the risk factors from normal, which may vary over time (3). Atherosclerotic plaque formation can be regarded as the intermediate between risk factor exposure and clinical events and can be imaged at a pre-clinical stage. The quantity of coronary artery calcification (CAC), a specific marker of coronary atherosclerosis, is correlated with coronary atherosclerotic plaque burden and seems to reflect the cumulative effects of risk factors and vascular aging (4). Clinical studies have demonstrated its ability to predict cardiovascular risk (5), but methodological shortcomings include selective recruitment, lack of sex-specific analysis, use of soft end points (i.e., revascularization or angina), and nonblinded assessment of results and outcome (6). In order to overcome these shortcomings, the HNR (Heinz Nixdorf Recall) study was initiated (7). Meanwhile, the MESA (Multi-Ethnic Study of Atherosclerosis) could demonstrate an incremental prognostic value of CAC over traditional risk factors in a representative U.S.-American population (8,9). Similarly, biomarkers and other parameters are used for improvements in risk prediction. The clinical value is expressed as their ability to reclassify individual risk (10,11). We hypothesized that CAC testing can be used for reclassification and improved risk prediction of hard events (i.e., nonfatal myocardial infarction and coronary death) beyond traditional risk factors. Methods Study design and population. The Heinz Nixdorf Recall Study is a population-based cohort study with subjects randomly selected from mandatory lists of residence. Study methods have been described in detail (7,12,13). Briefly, 4,814 participants 45 to 75 years of age were enrolled between 2000 and 2003 in the metropolitan Ruhr area in Germany. Of these, 327 (6.8%) reported a history of CAD at baseline and were excluded from this analysis (Fig. 1). The CAC scores were not reported to participants. All participants gave their written informed consent, and the study was approved by the institutional local ethical committees. Cardiovascular risk factors. The risk factor assessment has previously been described (Online Appendix) (13). Risk stratification algorithms. We assessed 2 different algorithms for coronary risk stratification: 1) the Framingham Risk Score (FRS) was defined as published using LDL cholesterol charts (1); and 2) in order to estimate the effect of the high-risk variables symptomatic carotid stenosis, stroke, peripheral artery disease, or diabetes, we allocated these subjects to the high-risk category following the National Cholesterol Education Program Adult Treatment Panel (ATP) III algorithm (14). In both algorithms, participants were categorized into low ( 10% or 6% in 10 years), intermediate (10% to 20% or 6% to 20%) and high ( 20%) risk categories (9 11,15). Electron-beam computed tomography. Nonenhanced electron-beam computed tomography scans were performed with a C-100 or C-150 scanner (GE Imatron, San Francisco, California) in 2 radiology institutions (D.G. and R.S.), as outlined in the Online Appendix. Follow-up. Annual postal questionnaires and a second medical examination assessed the morbidity status during follow-up (i.e., hospital admissions, outpatient diagnoses of cardiovascular disease). Participants were followed for a median of 5 years (mean years). Exclusion of subjects is shown in Figure 1, leaving 4,129 participants (53% women) for the present analysis. Study end points and verification of study end points. Primary end points for this study were based on unequivocally documented incident coronary events that met predefined study criteria (7). We considered a myocardial infarction event based on symptoms, signs of electrocardiography, and enzymes (levels of creatine kinase), as well as troponin T or I, and necropsy as nonfatal acute myocardial infarction and coronary death (16). For all primary study end points, hospital and nursing home records, including electrocardiograms, laboratory Figure 1 Flow Chart of the HNR Study Cohort Study cohort includes 4,487 participants 45 to 75 years of age without coronary artery disease (CAD). MI myocardial infarction.

3 JACC Vol. 56, No. 17, 2010 October 19, 2010: Erbel et al. Reclassification Based on Coronary Calcium Testing 1399 Baseline (NonfatalCharacteristics Myocardial Baseline Infarction Characteristics of the HNR and Coronary Study of thepopulation HNR Deaths) StudyDuring With Population and 5-Year Without Follow-Up and HardWithout Events Hard Events Table 1 (Nonfatal Myocardial Infarction and Coronary Deaths) During 5-Year Follow-Up With Events (n 93) Without Events (n 4,036) p Value Age, yrs Male sex 65 (70) 1,887 (47) BMI, kg/m Overweight (BMI kg/m 2 ) 43 (47) 1,851 (46) 0.10 Obese (BMI 30.0 kg/m 2 ) 30 (33) 1,071 (27) Waist circumference, cm Lipids Total cholesterol, mg/dl LDL cholesterol, mg/dl HDL cholesterol, mg/dl Triglycerides, mg/dl Total cholesterol 240 mg/dl 45 (48) 1,572 (39) 0.07 Intake of lipid-lowering drugs 10 (11) 364 (9) 0.56 Total cholesterol 240 mg/dl or lipid-lowering drugs 52 (56) 1,832 (45) 0.04 Blood pressure Systolic blood pressure, mm Hg Diastolic blood pressure, mm Hg Prevalence of hypertension JNC 7 stage 1 25 (27) 1,078 (27) JNC 7 stage 2 20 (22) 456 (11) Prevalence of antihypertensive drugs 43 (46) 1,284 (32) JNC 7 stage 1 antihypertensive drugs 61 (66) 2,198 (55) 0.03 Smoking status Never smoked 35 (38) 1,780 (44) Former smoker 33 (35) 1,341 (33) 0.20 Current smoker 25 (27) 915 (23) Diabetes mellitus 16 (17) 289 (7) Glucose, mg/dl History of stroke 3 (3) 89 (2) 0.51 History of peripheral artery disease 6 (7) 48 (1) Median serum C-reactive protein, mg/dl 0.21 (0.10/0.50) 0.14 (0.07/0.31) Serum creatinine, mg/dl Risk score algorithms Framingham risk score, % Median, % 14 (9/22) 9 (6/14) Low risk, 10% 26 (28) 2,204 (54) Low risk, 6% 7 (8) 933 (23) Intermediate risk, 10% 20% 38 (41) 1,324 (33) Intermediate risk, 6% 20% 57 (61) 2,595 (64) High risk, 20% 29 (31) 508 (13) ATP III score Low risk, 10% 22 (23) 2,106 (52) Low risk, 6% 5 (5) 903 (22) Intermediate risk, 10% 20% 33 (36) 1,157 (29) Intermediate risk, 6% 20% 50 (54) 2,360 (59) High risk, 20% 38 (41) 773 (19) CAC score 746 1, Median 183 (33/1,005) 11 (0/108) CAC score categories 0 11 (12) 1,311 (32) (26) 1,666 (41) (25) 671 (17) (37) 388 (10) Data are expressed as n (%), mean SD, or median (1st/3rd quartile). ATP III National Cholesterol Education Program Adult Treatment Panel III; BMI body mass index; CAC coronary artery calcification; HDL high-density lipoprotein; JNC 7 Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure; LDL low-density lipoprotein.

4 1400 Erbel et al. JACC Vol. 56, No. 17, 2010 Reclassification Based on Coronary Calcium Testing October 19, 2010: values, and pathology reports, were collected. For deceased subjects, death certificates were collected and interviews with general practitioners, relatives, and eyewitnesses were undertaken if possible. Medical records were obtained in 100% of all reported end points. An external end point committee blinded for risk factor status and CAC scores reviewed all documents and classified the end points. Statistical analysis. Based on previous studies, we hypothesized a relative risk (RR) 3.0 of primary end points in the highest versus lowest CAC score quartile (7). However, for sample size calculation, we used a more conservative estimate of a 2.5-fold increase in the RR of myocardial infarction or coronary death associated with the 4th quartile compared with the 1st quartile of CAC distribution. In order to detect an RR 2.5, a sample size of 4,200 eligible subjects at study entry was considered to be necessary for the primary end point analysis at a power of 90% after 5 years. Here, we assumed 12% of subjects with overt coronary heart disease at baseline (to be excluded) and a 5-year drop-out rate of 15%. The significance level (alpha error) was set at 5% for the 1-sided test (7). Demographic data and risk factors are expressed as mean (SD) or median (25th to 75th percentile), and frequencies are given as counts (%). Quartiles of CAC distribution were calculated by computing the common and the sex-specific 25th and 75th CAC percentiles within the present cohort of 4,129 participants. In addition, pre-defined CAC score categories 0, 1 to 99, 100 to 399, and 400 were used (13). Differences in proportions were statistically evaluated using chi-square or Fisher exact test and trends in proportions using the Cochran-Armitage trend test; location measures of continuous quantities were compared using Student t test or Mann-Whitney U test. Net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were derived from logistic regression models (17) based on Framingham risk factors (age, sex, diabetes, systolic blood pressure, LDL and HDL cholesterol, present smoking status) with and without CAC. Here, we utilized the logarithmic transformation of the CAC score, log(cac 1). For calculating the NRI, rescaled individual predicted risks from models with and without log(cac 1) were compared with established Framingham risk thresholds. The rescaling factor was derived by dividing the average 10-year Framingham risk by the observed 5-year event rate. To evaluate model calibration, we calculated the Hosmer-Lemeshow chisquare, which is a measure of deviation between observed and predicted outcomes in deciles of predicted risk. We also give the reclassification accounting for CAC, starting from FRS or ATP III categories, when performed using pre-defined Agatston score categories (0 to 99, 100 to 399, 400) (13), with CAC 100 leading to downclassification and CAC 400 leading to up-classification. Original risk categories and the resulting new classification were compared by computing the NRI (17). To assess further the ability of risk categories (or FRS values), log(cac 1), and both combined to distinguish individuals who will develop an event, we also estimated the receiver-operator characteristic (ROC) curves and areas under ROC curves (AUC or c-statistic with 95% confidence interval [CI]) in corresponding logistic models (18). This was also carried out for the model with Framingham risk factors, with and without log(cac 1). C-statistics from time-to-event analyses (Harrell c), calculated as sensitivity analysis, were numerically almost identical to c-statistics from logistic regression. Observed 5-year cumulative risks are given with their 95% CIs. We used univariate and multivariate relative risk regression (Poisson regression with a log link) (19) to calculate unadjusted and adjusted RRs and corresponding 95% CIs for the occurrence of primary end points. RR estimates for CAC scores were adjusted for conventional risk categories. All calculations were performed with SAS (version 9.2, SAS Institute, Cary, North Carolina). Results Baseline characteristics. Table 1 shows the distribution of cardiovascular risk factors at baseline for participants with and without events. The distribution for men and women is shown in Online Table 1. Number ATP HNR Intermediate 6% toiii Study 20% Risk of Coronary During Number Categories Risk Thresholds 5 Hard Years of Coronary and Events ofrelative Follow-Up: 10% Hard intothe Risks 20% Events 3 FRS and in in the and the 3 FRS and ATP III Risk Categories and Relative Risks in the Table 2 HNR Study During 5 Years of Follow-Up: Intermediate Risk Thresholds 10% to 20% and 6% to 20% Risk Category 10-Year Event Rate No. With Events No. in Group RR 95% CI FRS 10% 20% Low risk ( 10%) 26 2, Reference Intermediate risk (10% 20%) 38 1, High risk ( 20%) % 20% Low risk ( 6%) Reference Intermediate risk (6% 20%) 57 2, High risk ( 20%) ATP III risk score 10% 20% Low risk ( 10%) 22 2, Reference Intermediate risk (10% 20%) 33 1, High risk ( 20%) % 20% Low risk ( 6%) Reference Intermediate risk (6% 20%) 50 2, High risk ( 20%) Data are expressed as n in each group and category, as well as RR with 95% CIs. ATP Adult Treatment Panel; CI confidence interval; FRS Framingham Risk Score; RR relative risk.

5 JACC Vol. 56, No. 17, 2010 October 19, 2010: Erbel et al. Reclassification Based on Coronary Calcium Testing 1401 Cumulative events. During 5-year follow-up, 93 (2.25%) primary end points occurred among 4,129 participants, including 29 coronary deaths (31%) and 64 nonfatal myocardial infarctions (69%) (Fig. 1). The number of coronary events and the relative risks in each Framingham risk category for the total cohort are listed in Table 2 (distribution for men and women is shown in Online Table 2). Reclassification of risk for all risk categories. CAC categories were used for reclassification (Figs. 2A and 2B). For example, in the intermediate FRS category, subjects with low CAC score had an event rate of only 1.4% (95% CI: 0.7% to 2.5%), which was not significantly different from that in the low FRS category of 1.2% (95% CI: 0.8% to 1.7%) (p 0.58). The intermediate group with high CAC score, however, had a risk of 8.7% (95% CI: 5.2% to 13.6%), which was similar to that of the total high-risk FRS cohort of 5.4% (95% CI: 3.7% to 7.7%) (p 0.10), with similar results using the 6% to 20% threshold for intermediate risk. Using ATP III instead of FRS influenced the results only slightly (Online Figs. 1A and 1B). Figure 2 Observed Cumulative Rates With 95% CIs of Hard Events (Nonfatal Myocardial Infarction and Coronary Deaths) in the HNR Study Stratified by the Framingham Risk Score (FRS) in the risk categories of 10%, 10% to 20%, and 20% at 10 years (A) and 6%, 6% to 20%, and 20% (B). In addition, the event rates in each category of coronary artery calcification (CAC) scores within each FRS category are given. Bars represent 95% confidence intervals (CIs).

6 1402 Erbel et al. JACC Vol. 56, No. 17, 2010 Reclassification Based on Coronary Calcium Testing October 19, 2010: Reclassification 10% or Follow-Up Without to 20% Based Hard Reclassification andof Coronary 6% Intermediate CAC to 20%) Results Events of Risk Intermediate (Defined During Participants 5-Year as (Defined With as 10% to 20% and 6% to 20%) Risk Participants With Table 3 or Without Hard Coronary Events During 5-Year Follow-Up Based on CAC Results Reclassification Accounting for CAC Scores Total Classification According to the FRS Low Intermediate High No. 10% 20% Participants with events Low, 10% in 10 yrs Intermediate, 10% 20% in 10 yrs High, 20% in 10 yrs Total number with events Participants without events Low, 10% in 10 yrs 2, ,204 Intermediate, 10% 20% in 10 yrs ,324 High, 20% in 10 yrs Total number without events 3, ,036 Net reclassification improvement 21.7% (p ) (estimated) 6% 20% Participants with events Low, 6% in 10 yrs Intermediate, 6% 20% in 10 yrs High, 20% in 10 yrs Total number with events Participants without events Low, 6% in 10 yrs Intermediate, 6% 20% in 10 yrs 1, ,595 High, 20% in 10 yrs Total number without events 2, ,036 Net reclassification improvement 30.6% (p ) (estimated) Data are expressed in the intermediate-risk category for those with and without events, but grouping all individuals in the low and high category together in one group for estimating net reclassification improvement based on coronary artery calcification scores. CAC coronary artery calcification; FRS Framingham Risk Score. Reclassification of intermediate-risk subjects with CAC 100 to the low-risk category and with CAC 400 to the high-risk category (Table 3) resulted in correct upclassification of 17 (18%) and 18 (19%) and incorrect down-classification of 12 (13%) and 27 (29%) of 93 subjects with events for the thresholds 10% to 20% and 6% to 20%, respectively. Correct down-classification occurred in 836 (21%) and 1,870 (46%) and incorrect up-classification occurred in 178 (4.4%) and 246 (6%) of 4,036 subjects without events, which yielded an NRI of 21.7% (p ) and 30.6% (p ), respectively. The same high rates were also achieved using ATP III risk categories (Online Table 3), yielding an NRI of 18.8% (p ) and 29.0% (p ), respectively. In the total cohort (Table 4), reclassification based on rescaled individual predicted event probabilities from logistic regression models with and without log(cac 1) resulted in an NRI of 22.4% (p ) and 19.6% (p 0.004) using intermediate-risk thresholds of 10% to 20% and 6% to 20%, respectively. The rescaling factor was 5.07, equaling (average Framingham 10-year risk) divided by 2.25 (observed 5-year event rate). Crude and adjusted cumulative risks in CAC score categories. The RR of coronary events (Table 5) for the upper versus the lower CAC quartile was 6.40 (95% CI: 3.37 to 12.16; p ). Crude cumulative risks among subjects with a CAC score of 100 to 399 and 400 were nearly 5 and 10 times, respectively, the cumulative risk of subjects without CAC. For CAC scores 400, RR was nearly identical in men and women (Online Table 4) and remained highly significant when they were adjusted for the ATP III algorithm. This was not different when data were adjusted for FRS categories (data not shown). c-statistics, calibration, and IDI. AUCs were similar among FRS values (0.681) and the ATP III categories (0.653; p 0.14) (Figs. 3A and 3B). The AUC for log(cac 1) was Adding log(cac 1) to the 2 algorithms significantly increased the AUCs to and 0.755, respectively (data for men and women are shown in Online Table 5). The AUC derived from a risk prediction model based on Framingham risk variables (age, sex, dia- Reclassification Models Hard Based 10% Coronary Using tocac 20% Reclassification FRS Events With and forvariables the 6% Definition During to Total 20% With for 5-Year Cohort ofthe Intermediate ortotal Follow-Up Without BasedCohort onrisk Based on Models Using FRS Variables With or Without Table 4 Hard Coronary Events During 5-Year Follow-Up Based on CAC With Definition of Intermediate Risk as 10% to 20% and 6% to 20% Classification Using a Model Based on FRS Variables 10% 20% Participants with events Reclassification Using Models Based on FRS Variables With and Without CAC Low Intermediate High Total No. Low, 10% in 10 yrs Intermediate, 10% 20% in 10 yrs High, 20% in 10 yrs Total number with events Participants without events Low, 10% in 10 yrs 2, ,386 Intermediate, 10% 20% in 10 yrs ,096 High, 20% in 10 yrs Total number without events 2, ,036 Net reclassification improvement 22.4% (p ) (estimated) 6% 20% Participants with events Low, 6% in 10 yrs Intermediate, 6% 20% in 10 yrs High, 20% in 10 yrs Total number with events Participants without events Low, 6% in 10 yrs 1, ,404 Intermediate, 6% 20% in 10 yrs 735 1, ,078 High, 20% in 10 yrs Total number without events 1,980 1, ,036 Net reclassification improvement 19.6% (p 0.004) (estimated) Abbreviations as in Table 3.

7 JACC Vol. 56, No. 17, 2010 October 19, 2010: Erbel et al. Reclassification Based on Coronary Calcium Testing 1403 Cumulative Table 5 and Cumulative Relativeand Risks Relative of HardRisks Events of Hard Associated EventsWith Associated Increasing WithCAC Increasing Scores CAC in thescores HNR Study in the HNR Study No. With Events No. in Group Crude RR 95% CI Adjusted RR* 95% CI CAC score categories , Reference 1.00 Reference , , Log 2 (CAC 1) 93 4, Quartiles of CAC scores 1st (0) 11 1, Reference 1.00 Reference 2nd ( ) rd ( ) 23 1, th ( 114.5) 55 1, Data are expressed as n in each group and category of CAC and quartiles, as well as crude and adjusted RR with the 95% CIs. *Adjusted for ATP III risk categories. Abbreviations as in Table 2. betes, LDL and HDL cholesterol, systolic blood pressure, and smoking) yielded an AUC of (95% CI: to 0.760). Adding log(cac 1) to this model increased the AUC to (95% CI: to 0.812; p 0.004). The Hosmer-Lemeshow test yielded a chi-square of 15.5 (p 0.05) for the model with Framingham risk factors and decreased to 9.1 (p 0.33) when log(cac 1) was entered into the model. The IDI was estimated as 1.52% (p ) comparing the models with Framingham risk factors with and without log(cac 1). Discussion This large population-based study with nearly 5,000 participants and 5-year follow-up period demonstrates the following: 1) CAC testing results in high net reclassification rates of 21.7% and even 30.6% in participants at intermediate risk defined as 10% to 20% and 6% to 20%, respectively; 2) the use of the ATP III categories in comparison with the FRS yields similar results; 3) the adjusted RR increases with the CAC scores; 4) a CAC score of 0 indicates an excellent prognosis with an event rate of only 0.16%/year; and 5) the CAC scoring shows significantly better results for c-statistics than traditional risk factors for men as well as women. Our findings compare well with those recently reported by MESA (15). In that study, NRI was 25% (95% CI: 16% to 34%). An additional 23% of subjects with events were reclassified to the high-risk category, and 13% additional subjects without events were reclassified to the low-risk category, which favorably compares with the 24% and 19% of subjects, respectively, in our study, considering the corresponding models based on Framingham variables and a Figure 3 ROC Curve for the Prediction of Risk of Hard Events Within 5 Years of Inclusion in the HNR Study Using (A) the Framingham risk score (FRS) or (B) the National Cholesterol Education Program Adult Treatment Panel III (ATP III) score in comparison with the coronary artery calcification (CAC) score expressed as log(cac 1) and the combination of the CAC and FRS or ATP III. ROC receiver-operator characteristic.

8 1404 Erbel et al. JACC Vol. 56, No. 17, 2010 Reclassification Based on Coronary Calcium Testing October 19, 2010: NRI Table Using 6 Other NRI Using Variables Other of Variables Extended of Risk Extended Stratification Risk Stratification Risk Marker/Factor NRI (%) Intermediate 10-Yr Risk Definition (%) p Value First Author (Ref. #) Multiple biomarker score (troponin I, NT-proBNP, cystatin C, CRP) Zethelius et al. (10) Multiple biomarker score (MR-proADM, NT-proBNP) NS Melander et al. (11) HDL cholesterol (Framingham) Pencina et al. (17) HDL cholesterol (SCORE data) and Cooney et al. (23) Heart rate (SCORE data) and 5 10 NS Cooney et al. (24) hscrp (women) Cook et al. (25) hscrp (men) Wilson et al. (26) hscrp (men and women) Ridker et al. (27) HbA1c (men) and HbA1c (women) Simmons et al. (21) CAC (men and women, including revascularization) Polonsky et al. (15) CAC (model based on FRS-variables with and without CAC, all subjects) Erbel et al. (current study) CAC (reclassification based on the FRS only of intermediate-risk subjects) Erbel et al. (current study) Modified from Cooney et al. (22). The different publications on net reclassification are related to biomarkers, HDL cholesterol, heart rate, and signs of diabetes. In addition, results of our study are listed related to CAC. CRP C-reactive protein; HbA1c hemoglobin A1c; hscrp high-sensitive C-reactive protein; MR-proADM mid-regional pro-adrenomedullin; NRI net reclassification improvement; NT-proBNP N-terminal pro B-type natriuretic peptide; other abbreviations as in Table 1. 6% to 20% definition of intermediate risk. Thus 2 independent observational and prospective studies have now shown a high NRI using a CAC-based approach. Using a multiple biomarker model of troponin I, N-terminal pro B-type natriuretic peptide, cystatin C, and C-reactive protein, only 1 study could report a similar high NRI (26%), but in a cohort of older adults with known CAD (10). But these results could not be confirmed using biomarkers like midregional pro-adrenomedullin and N-terminal pro B-type natriuretic peptide for risk assessment (11). For other variables (Table 6) such as heart rate, C-reactive protein, or hemoglobin A1c, lower NRI values were reported (20 26). When intermediate risk was defined as 5% to 20%, the NRI was 16.7% extrapolated to a 10-year ATP III risk for the Reynolds score in asymptomatic men (27). On the one hand, despite these very promising results, physicians have to be aware that the proposed risk stratification algorithms are not perfect, as events do rarely occur in those who are classified as low risk even when CAC scores are low. On the other hand, differing from previous reports (28), a high CAC in low Framingham risk subjects was associated with an event rate comparable to that of the high-risk group. This event rate in the low-risk subgroup could be further reduced when subjects with risk equivalents were recognized as being per se at high risk, demonstrated in the ATP III analysis compared with the FRS results. The event rate was also smaller when the lower intermediate risk threshold was defined at 6% instead of 10% risk in 10 years. The annual event rate in the remaining subjects was then only 0.12%. Extending CAC scoring to these low-risk individuals would also mean a substantial increase in costs and unnecessary radiation exposure (29). This interpretation may change taking into account the 10-year or lifetime risk for coronary events (30), but based on the observed 5-year event rates, CAC scoring seems to not be indicated in routine clinical practice for low-risk individuals. In line with previous studies (31,32), subjects at high risk but low CAC scores appear to have a favorable prognosis, with an event rate within the range of the expected annual 0.25% to 0.5% events estimated from the Duke treadmill score (33). In these individuals, risk factor modification by lifestyle changes may be an alternative first-line therapy. However, there is lack of confirmation from prospective risk modification trials that this is safe (28,34). In addition, we found that in the high FRS and ATP III risk categories with low CAC levels, more subjects were already on risk factor modifying therapy than in the low-risk group, which might have resulted in lower coronary event rates (data not shown). Finally, other atherosclerotic end-organ manifestations in high-risk individuals cannot be neglected. Therefore, despite the graded event rate in high-risk subjects depending on coronary atherosclerosis burden, CAC scoring in high-risk individuals has little effect on clinical decision making. In summary, our data support the highest clinical benefit in intermediate-risk subjects. Study limitations. The observed event rate of 2.25% extrapolates to 450 per 100,000/year and corresponds to the expected event rate of 300 to 500 per 100,000/year (7), and the RRs in our study correspond to the previous meta-analysis of CAC scoring (Fig. 4) (5). The cumulative risk was also in the expected range. However, the FRS overestimated risk more than 2-fold in this European cohort, in line with previous reports (35). Yet the FRS is the best validated risk stratification algorithm across various cohorts, and it did stratify our cohort into those with low, intermediate, and high cumulative risks, even though the expected thresholds were not reached. Furthermore, the FRS (36) has been developed to identify coro-

9 JACC Vol. 56, No. 17, 2010 October 19, 2010: Erbel et al. Reclassification Based on Coronary Calcium Testing 1405 Figure 4 Relative Risks and 95% CIs of CAC Categories 1 to 99, 100 to 399, 400 to 999, and >1,000 Versus CAC 0 During the 5-Year Follow-Up in the HNR Study For comparison, the results of a meta-analysis by Greenland et al. (5) are added. CAC coronary artery calcification; HNR Heinz Nixdorf Recall. nary events in men and women from the general population, which is in line with our study goals. We cannot exclude that other more recently proposed algorithms (2,37,38) would have predicted events better than the algorithms used in this analysis, given that other algorithms have in part used different end points or were developed based on selected populations. Finally, a cost-effectiveness analysis is warranted when evaluating a novel marker of cardiovascular risk (39). This requires a thorough investigation of medication prescription data, which is part of a separate ongoing health-economic project and will be the focus of a separate forthcoming report. Conclusions CAC scoring improves risk stratification, discrimination, and reclassification above and beyond traditional risk factor categories. Limiting CAC scoring to intermediate-risk subjects helps to correctly identify a high proportion of individuals at highest risk and can contribute to reducing the number of coronary events in the general population. However, physicians should be aware that these algorithms are not perfect, particularly in the low-risk category. Acknowledgments The authors thank the Heinz Nixdorf Foundation (Chairman: Martin Nixdorf, previous chairman: Dr. jur. Gerhard Schmidt), Germany, for their generous support of this study. The authors also thank Prof. Dr. K. Lauterbach (Department of Health Economy and Epidemiology, University of Cologne, Germany) for his valuable contributions in an early phase of the study. The authors are indebted to the all study participants and to the dedicated personnel of both the study centre of the Heinz Nixdorf Recall Study and the EBT-scanner facilities as well as to the investigative group, in particular to U. Roggenbuck, S. Slomiany, E.M. Beck, A. Öffner, S. Münkel, M. Bauer, S. Schrader, R. Peter, and H. Hirche. Reprint requests and correspondence: Dr. Raimund Erbel, Department of Cardiology, West-German Heart Center Essen, University Duisburg-Essen, Hufelandstrasse 55, D Essen, Germany. erbel@uk-essen.de. REFERENCES 1. Wilson PW, D Agostino RB, Levy D, Belanger AM, Silbershatz H, Kannel WB. Prediction of coronary heart disease using risk factor categories. Circulation 1998;97: Assmann G, Cullen P, Schulte H. Simple scoring scheme for calculating the risk of acute coronary events based on the 10-year follow-up of the prospective cardiovascular Münster (PROCAM) study. Circulation 2002;105: Sniderman AD, Furberg CD. Age as a modifiable risk factor for cardiovascular disease. Lancet 2008;371: Shaw L, Raggi P, Berman DS, Callister Q. CAC as a measure of biologic age. Atherosclerosis 2006;188: Greenland P, Bonow RO, Brundage BH, et al. ACCF/AHA 2007 clinical expert consensus document on CAC scoring by CT in global cardiovascular risk assessment and in evaluation of patients with chest pain: a report of the ACC Foundation Clinical Expert Consensus Task Force (ACCF/AHA Writing Committee to Update the 2000 Expert Consensus Document on Electron Beam Computed Tomography). J Am Coll Cardiol 2007;49: Redberg RF. Coronary artery calcium: should we rely on this surrogate marker? Circulation 2006;113: Schmermund A, Möhlenkamp S, Stang A, et al. Assessment of clinically silent atherosclerotic disease and established and novel risk factors for predicting myocardial infarction and cardiac death in healthy middle-aged subjects: rationale and design of the Heinz Nixdorf RECALL Study. Risk Factors, Evaluation of Coronary Calcium and Lifestyle. Am Heart J 2002;144:212 8.

10 1406 Erbel et al. JACC Vol. 56, No. 17, 2010 Reclassification Based on Coronary Calcium Testing October 19, 2010: Detrano R, Guerci AD, Carr JJ, et al. Coronary calcium as a predictor of coronary events in four racial or ethnic groups. N Engl J Med 2008;358: Budoff MJ, Nasir K, McClelland RL, et al. Coronary calcium predicts events better with absolute calcium scores than age-, sex-, race/ ethnicity percentiles: MESA (Multi-Ethnic Study of Etherosclerosis). J Am Coll Cardiol 2009;53: Zethelius B, Berglund L, Sundström J, et al. Use of multiple biomarkers to improve the prediction of death from cardiovascular causes. N Engl J Med 2008;358: Melander O, Newton-Cheh C, Almgren P, et al. Novel and conventional biomarkers for prediction of incident cardiovascular events in the community. JAMA 2009;302: Stang A, Moebus S, Dragano N, et al. Heinz Nixdorf Recall Study Investigation Group. Baseline recruitment and analyses of nonresponse of the Heinz Nixdorf Recall Study: identifiability of phone numbers as the major determinant of response. Eur J Epidemiol 2005;20: Erbel R, Möhlenkamp S, Lehmann N, et al., Heinz Nixdorf Recall Study Investigative Group. Sex related cardiovascular risk stratification based on quantification of atherosclerosis and inflammation. Atherosclerosis 2008;197: Third report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III) Final Report. Circulation 2002;106: Polonsky TS, McClelland RL, Jorgensen NW, et al. Coronary artery calcium score and risk classification for coronary heart disease prediction. JAMA 2010;303: Thygesen K, Alpert JS, White HD, et al., Joint ESC/ACCF/AHA/ WHF Task Force for the Redefinition of Myocardial Infarction. Universal definition of myocardial infarction. J Am Coll Cardiol 2007;50: Pencina MJ, D Agostino RB, D Agostino RB Jr., Vasan RS. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med 2008;27: DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated ROC-curves: a non-parametric approach. Biometrics 1988;443: Spiegelman D, Hertzmark E. Easy SAS calculations for risk or prevalence ratios and differences. Am J Epidemiol 2005;162: Shah T, Casas JP, Cooper JA, et al. Critical appraisal of CRP measurement for the prediction of coronary heart disease events: new data and systematic review of 31 prospective cohorts. Int J Epidemiol 2009;38: Simmons RK, Sharp S, Boekholdt SM, et al. Evaluation of the Framingham risk score in the European Prospective Investigation of Cancer-Norfolk cohort: does adding glycated hemoglobin improve the prediction of coronary heart disease events? Arch Intern Med 2008; 168: Cooney MT, Dudina AL, Graham IM. Value and limitations of existing scores for the assessment of cardiovascular risk: a review for clinicians. J Am Coll Cardiol 2009;54: Cooney MT, Dudina A, DeBacquer D, et al. How much does HDL cholesterol add to risk estimation? A report from the SCORE Investigators. Eur J Cardiovasc Prev Rehabil 2009;16: Cooney MT, Vartiainen E, Laatikainen T, Joulevi A, Dudina A, Graham I. Simplifying cardiovascular risk estimation using resting heart rate. Eur Heart J 2010;31: Cook NR, Buring JE, Ridker PM. The effect of including CRP in cardiovascular risk prediction models for women. Ann Intern Med 2006;145: Wilson PW, Pencina M, Jacques P, Selhub J, D Agostino R Sr., O Donnell CJ. C-reactive protein and reclassification of cardiovascular risk in the Framingham Heart Study. Circ Cardiovasc Qual Outcomes 2008;1: Ridker PM, Paynter NP, Rifai N, Gaziano JM, Cook NR. C-reactive protein and parental history improve global cardiovascular risk prediction: The Reynolds Risk Score for men. Circulation 2008;118; Arad Y, Goodman KJ, Roth M, Newstein D, Guerci AD. Coronary calcification, coronary disease risk factors, C-reactive protein, and atherosclerotic cardiovascular disease events: the St. Francis Heart Study. J Am Coll Cardiol 2005;46: Smith-Bindman R, Lipson J, Marcus R, et al. Radiation dose associated with common computed tomography examinations and the associated lifetime attributable risk of cancer. Arch Intern Med 2009;169: Berry JD, Liu K, Folsom AR, et al. Prevalence and progression of subclinical atherosclerosis in younger adults with low short-term but high lifetime estimated risk for CVD: The Coronary Artery Risk Development in Young Adults Study and Multi-Ethnic Study of Atherosclerosis. Circulation 2009;119; Sarwar A, Shaw LJ, Shapiro MD, et al. Diagnostic and prognostic value of absence of coronary artery calcification. J Am Coll Cardiol Img 2009;2: Raggi P, Shaw LJ, Berman DS, Callister TQ. Prognostic value of coronary artery calcium screening in subjects with and without diabetes. J Am Coll Cardiol 2004;43: Mark DB, Shaw L, Harrell FE Jr., et al. Prognostic value of a treadmill exercise score in outpatients with suspected coronary artery disease. N Engl J Med 1991;325: Greenland P, LaBree L, Azen SP, Doherty TM, Detrano RC. CAC score combined with Framingham score for risk prediction in asymptomatic individuals. JAMA 2004;291: Hense HW, Schulte H, Löwel H, Assmann G, Keil U. Framingham risk function overestimates risk of coronary heart disease in men and women from Germany: results from the MONICA Augsburg and the PROCAM cohorts. Eur Heart J 2003,24: Preis SR, Hwang SJ, Fox CS, et al. Eligibility of individuals with subclinical CAC and intermediate coronary heart disease risk for reclassification (from the Framingham Heart Study). Am J Cardiol 2009;103: Nicholls M. Preventive cardiology in Germany: HeartScore. Circulation 2006;114:f Hippisley-Cox J, Coupland C, Vnogradova Y, Robson J, May M, Brindle P. Derivation and validation of QRISK, a new cardiovascular disease risk score for the United Kingdom: prospective open cohort study. BMJ 2007;335: Hlatky MA, Greenland P, Arnett DK, et al., American Heart Association Expert Panel on Subclinical Atherosclerotic Diseases and Emerging Risk Factors and the Stroke Council. Criteria for Evaluation of Novel Markers of Cardiovascular Risk. A Scientific Statement From the American Heart Association. Circulation 2009;119: Key Words: cardiac death y coronary artery calcification y myocardial infarction y reclassification y risk factors y subclinical atherosclerosis. APPENDIX For an expanded Methods section, supplementary tables, and supplementary figures, please see the online version of this article.