Background
Using data from autopsied young people aged 15 to 34 years, the Pathobiological Determinants of Atherosclerosis in Youth (PDAY) study developed a risk score based on age, sex, smoking status, high-density lipoprotein and non–high-density lipoprotein cholesterol levels, and the presence of obesity, hyperglycemia, and hypertension to predict advanced coronary artery atherosclerosis.
Methods
The Coronary Artery Risk Development in Young Adults (CARDIA) study assessed coronary artery calcium (CAC) by computed tomography in young adults participating in the 15-year examination. The PDAY risk score was calculated from risk factors measured at the CARDIA examinations at years 0, 5, 10, and 15.
Results
Odds ratios for amount of CAC (6 ordinal categories) for a 1-point increase in risk score computed from the modifiable risk factors ranged from 1.10 to 1.16 (all statistically significant). Odds ratios for presence of any amount of CAC ranged from 1.09 to 1.15 (all statistically significant), with the highest odds ratio for the risk score at year 0. An increase in risk score between years 0 and 15 increased the odds of CAC, and a decrease in risk score decreased the odds of CAC. A positive family history of cardiovascular disease increased the odds of CAC. The c statistics ranged from 0.752 to 0.770, with the highest discrimination based on the year 0 revised PDAY risk score that included family history and increased the points for the sex differential.
Conclusion
The PDAY risk score predicts CAC up to 15 years before its assessment, and risk score change during 15 years affects the risk of CAC.
Cardiovascular events in adulthood are linked to atherosclerosis, a process that begins in youth and is accelerated by the presence of risk factors, including dyslipidemia, hypertension, obesity, smoking, and diabetes mellitus.1-8 Preventive efforts directed at young individuals range from encouraging healthy lifestyle behaviors that prevent cardiovascular risk factors (primordial prevention) to more aggressive behavioral and pharmacologic treatment of high-risk individuals (primary prevention).9 Recent developments in imaging allow the noninvasive assessment of advanced atherosclerosis before the occurrence of clinical events.10-13 These techniques have usually been applied to older individuals to enhance risk prediction, and their findings may influence clinical management in those with intermediate risk.14-17 The techniques have also been used to detect atherosclerosis in high-risk young individuals,10 to determine the relationship of risk factors in youth to the progression of atherosclerosis in young adulthood,12,13 and to evaluate early pharmacologic treatment of familial hypercholesterolemia.18
In persons aged 15 to 34 years who died of external causes (accidents, homicides, and suicides), the Pathobiological Determinants of Atherosclerosis in Youth (PDAY) study identified the association of atherosclerosis with cardiovascular risk factors measured postmortem.2-4,6,7 A PDAY risk score, based on age, sex, and the cardiovascular risk factors, was developed to estimate the likelihood of advanced atherosclerosis in the coronary arteries of an individual.19
The Coronary Artery Risk Development in Young Adults (CARDIA) study performed computed tomographic (CT) imaging of coronary artery calcium (CAC) in its year 15 examination (when participants were aged 33-45 years). We used risk factor data from multiple prior examinations to determine how well the PDAY risk score computed from risk factors measured in youth predicted CAC measured years later in CARDIA subjects, and we estimated the effect of change in the PDAY risk score over time on its prediction of future CAC. We also revised the risk score by incorporating family history (information unavailable in the PDAY study) and compared the PDAY risk score prediction with that of the Framingham risk score.8
The PDAY risk score was derived from the associations of cardiovascular risk factors measured postmortem with target atherosclerotic lesions in the coronary arteries of 1117 cases.19 The target lesions were American Heart Association grade IV or V lesions in the left anterior descending coronary artery,6,20 greater than 9% of the intimal surface area of the right coronary artery involved with gross raised lesions, or both. Risk factors included high-density lipoprotein (HDL) and non-HDL cholesterol levels and thiocyanate concentration (as a marker of smoking) in postmortem serum,3 body mass index (calculated as weight in kilograms divided by height in meters squared) at autopsy of at least 30 to define obesity,7 hyperglycemia assessed by a red blood cell glycohemoglobin of at least 8%,2 and hypertension assessed by the intimal thickness of the small renal arteries.4 Risk factors and their point values are given in Table 1. Race/ethnicity was not a significant predictor of risk. The risk score was normalized so that a 1-point increase was equivalent to a multiplicative change in the odds (additive change in the logarithm of the odds) of a target lesion due to a 1-year increase in age.
The CARDIA study is a longitudinal study of cardiovascular risk factors begun in 1985-1986 that included 5115 men and women from 4 centers (Birmingham, Ala; Chicago, Ill; Minneapolis, Minn; and Oakland, Calif).20 The sample was population based and was approximately equally divided by sex, race/ethnicity (African American vs white), and education level (completed vs did not complete high school). At its year 15 examination, CT scanning was performed on 3043 participants. The present study used risk scores calculated from the examinations at years 0, 5, 10, and 15. Participants were excluded if they (1) did not attend the examination, (2) were pregnant at the examination, or (3) were missing data needed for completion of the PDAY risk score or the Framingham risk score.8 At the year 0 examination, the CARDIA cohort included participants younger than 20 years, for whom the Framingham risk score could not be computed. Therefore, the sample sizes were 2967 for the PDAY risk score and 2719 for the Framingham risk score at year 0, and they were 2769, 2732, and 2975 at years 5, 10, and 15, respectively, for the PDAY and Framingham risk scores. The CARDIA study was approved by the institutional review boards of all participating centers.
Age, sex, race/ethnicity, and smoking status were determined by self-report. Body mass index was calculated from weight and height measured in light clothing and without shoes. Blood pressure was measured 3 times by random zero sphygmomanometer, and the mean of the second and third measurements was used. Hypertension was defined as having a systolic blood pressure of at least 130 mm Hg or a diastolic blood pressure of at least 85 mm Hg or taking antihypertensive medications. Total cholesterol and HDL cholesterol levels were measured by enzymatic techniques, and non-HDL cholesterol levels were calculated by subtraction.21 Diabetes mellitus was defined as having a fasting glucose level of at least 126 mg/dL (≥3.26 mmol/L); taking diabetic medication at the examinations at years 0, 10, and 15 (fasting glucose levels were unavailable at year 5); or both. Family history of cardiovascular disease was defined as having a parent with a cardiovascular event. Because family history was not updated at year 15, the year 10 family history was used.
Participants underwent CT scanning with electron beam CT (Imatron C-150) or multidetector CT (GE Lightspeed or Siemens VZ), depending on clinic location. All participants had 2 scans performed 1 to 2 minutes apart. For each scan, 40 consecutive 3-mm transverse images from the root of the aorta to the apex of the heart were obtained at 80% of the electrocardiographic R-R interval. Each image was examined by a radiologic technologist, who removed bony structures from the images and identified a region of interest around each potential focus of CAC. A focus was defined as a region of 4 adjacent pixels or more (approximately 1.87 mm2) with a CT number higher than 130 Hounsfield units. Quantification of lesions of this size is reproducible.22 A total CAC score was computed as the sum of the Agatston score of calcified lesions within each artery (left main, left circumflex, left anterior descending, and right) and across all arteries. Coronary artery calcium was defined as present if each score was greater than 0; the mean of the 2 scans was the final score. Each scan set with at least 1 nonzero score was reviewed by an expert investigator blinded to the scan scores to verify the presence of CAC. The CAC scores were adjusted for between-center differences using a standard calcium phantom scanned underneath each participant. The amount of CAC was categorized into 6 categories (0, >0 to <10, 10 to <20, 20 to <100, 100 to <400, or ≥400 units).23
The association of CAC with risk score was analyzed using binary logistic regression, with the presence of any amount of CAC as the outcome and the predictor variables of sex, age at CAC assessment, and PDAY risk score computed from the modifiable risk factors. The amount of CAC was analyzed by ordinal logistic regression.24,25 The association between CAC category and quartile of PDAY risk score at year 0 was assessed using the Goodman and Kruskal γ statistic.25
Tracking was measured by the extent to which individuals in the lowest or highest quartiles for the PDAY risk score at year 0 remained in the same quartiles at year 15. The year 0 and year 15 PDAY risk scores computed from the modifiable risk factors were divided into examination-specific approximate quartiles for each sex group and race/ethnicity. The examination-specific quartiles were cross-tabulated to determine the percentage of those in a given quartile at year 0 who were in each quartile at year 15. If no tracking occurred, those in the highest (or lowest) quartile at year 0 were expected to be evenly distributed across quartiles at year 15. Quartiles are only approximately 25% of the sample. Heavy ties among PDAY risk scores precluded quartiles from having equal sample sizes.
Based on their PDAY risk score computed from modifiable risk factors, participants were classified into low-risk, intermediate-risk, and high-risk groups. The cutoff points were selected so that the low-risk and high-risk groups would have at least the lowest and highest 25% of the year 0 PDAY risk scores, respectively. Within each of these 3 groups, participants were categorized into approximate thirds of the 15-year change (year 15 minus year 0) in the PDAY risk score to examine the effect of changing the modifiable risk factors. A revised PDAY risk score included points for family history of cardiovascular disease and increased the points for the sex differential. Points for family history were determined by dividing the logistic regression coefficient for family history by the coefficient for the PDAY risk score and by rounding to the nearest integer. The points for sex were determined in a similar manner.
To validate the allocation of risk score points among sex, age at the time of risk factor measurement, modifiable risk factors, and family history, we computed c statistics24 (equivalent to area under the receiver operating characteristic curve) to assess discrimination (ie, the ability to separate individuals with and without CAC); statistical significance among receiver operating characteristic curves at each visit was assessed.26 At years 5, 10, and 15, the CARDIA study included participants older than 34 years. For these examinations, the age at the time of risk factor measurement was included as a predictor in the logistic regression model along with the PDAY risk score computed from the other risk factors.
Characteristics of the cardia cohort
Table 2 gives descriptive statistics for the CARDIA cohort at each examination. For comparison between PDAY and Framingham risk scores, the age term was excluded. Over the 4 examinations, smoking frequency decreased, body mass index increased, non-HDL cholesterol levels trended upward, HDL cholesterol levels declined between years 5 and 10, prevalence of hypertension increased, prevalence of diabetes mellitus increased at year 10, and prevalence of positive family history increased through year 10. Year 15 CAC scores greater than 0 occurred in 201 (14.8%) of 1358 men and in 80 (5.0%) of 1609 women; the mean, median, minimum, and maximum CAC scores greater than 0 were 81.8, 21.8, 0.8, and 3520.5, respectively.
Relative risk due to modifiable risk factors
Table 3 gives the odds ratios (ORs) for the presence of any amount of CAC at year 15 and for the amount of CAC for a 1-point increase in PDAY risk scores computed from the modifiable risk factors at each of the 4 examinations, adjusted for sex and age at measurement of CAC. All ORs were significantly different from 1.00. Accounting for the amount of CAC increased the ORs at each examination by 0.01 compared with the ORs for the presence of CAC. Table 4 gives the relationship between the amount of CAC and the PDAY risk score computed from modifiable risk factors; this relationship was positive and significant (γ = 0.51 [95% confidence interval, 0.43-0.58]). Of those with CAC scores greater than 100, 33 (71.7%) of 46 had PDAY risk scores of at least 3.
As summarized in Table 5, the PDAY risk score tracked well. About 50% (minimum 37% and maximum 57%) of young people in the first or fourth quartile of risk at year 0 remained within that quartile at year 15.
Effect of changing the modifiable risk factors
Figure 1 shows the effect of risk score change on the odds of CAC in subjects at 3 levels of risk score computed from the modifiable risk factors at year 0 (lowest, −1 or 0; intermediate, 1-2; and highest, ≥3). The reference category (OR, 1.00) is the lowest year 0 risk and lowest change group. In the lowest risk group at year 0, for those whose risk score increased the most, the likelihood of CAC increased, but the OR was not statistically significant. Subjects with intermediate risk at year 0 that changed little or increased were significantly more likely to develop CAC than the low-risk and low-change group; the subjects who decreased their score were not significantly different from the reference group. Subjects with the highest risk at year 0 were significantly more likely to develop CAC than the low-risk and low-change group regardless of their risk score change. The intermediate and highest subjects at year 0 who decreased their risk score experienced a decreased OR relative to the subjects who remained the same or increased their risk score slightly.
Contribution of family history
Table 6 gives the ORs for a positive family history of cardiovascular disease (all ORs were significantly different from 1.00). The ORs for the PDAY risk score computed from the modifiable risk factors were only slightly different from the ORs estimated in a model without family history (Table 3).
A revised PDAY risk score was constructed using −8 points for female sex and 4 points for positive family history of cardiovascular disease. The PDAY risk score points (Table 1) were used for the modifiable risk factors.
Prediction of cac using the pday risk score
The foregoing analyses showed that the PDAY risk score computed from the modifiable risk factors identified a risk of CAC years in advance of CAC assessment. Table 7 gives c statistics for the risk score, including age and sex at the 4 examinations. The c statistics ranged from 0.728 to 0.746 and represented acceptable discrimination.24
Comparison of pday, revised pday, and framingham risk scores
The revised PDAY risk score performed better than the original PDAY risk score, with c statistics ranging from 0.752 to 0.770 (Table 7). The highest c statistics were obtained for risk scores computed from the year 0 and year 5 examinations. The Framingham risk score performed less well, with c statistics below 0.7. At each examination, the c statistics among the 3 risk scores were significantly different. Figure 2 shows the receiver operating characteristic curves for predicting CAC at year 15 using the PDAY risk score, the revised PDAY risk score, and the Framingham risk score at year 0.
This study links a cardiovascular risk score based on anatomically demonstrated atherosclerosis in autopsied subjects19 and the noninvasive assessment of CAC by CT scanning in living subjects from the CARDIA study. Young people with modifiable risk factors were at increased risk of developing CAC, and this increased risk was detected up to 15 years before the measurement of CAC. The risk score computed from risk factors measured 10 to 15 years before the assessment of CAC performed better than the risk score computed from risk factors at the time of CAC assessment. If the risk score increased between year 0 and year 15, the likelihood of developing CAC increased; if the risk score decreased, the likelihood of developing CAC decreased. The PDAY risk score and a revised PDAY risk score that included family history of cardiovascular disease and a greater differential between the sexes performed better than the Framingham risk score.
Comparison with other studies
Our findings are similar to those of other studies showing the association of risk factors measured in youth with CAC27 and carotid intimal thickness12,13 measured later in life. Our results also are consistent with the tracking of individual risk factors.28,29
For years 0 and 5, the ORs for a 1-point increase in the PDAY risk score (Table 3) were similar to the OR of 1.18 obtained in the PDAY sample used for risk score development.19 Discrimination of the PDAY risk score (c statistic range, 0.728-0.746; Table 7) validates the PDAY risk score model.
Incorporation of family history of cardiovascular disease and a larger differential for sex than in the original PDAY risk score improved prediction. The highest discrimination and ORs were obtained for the revised PDAY risk score at the year 0 examination (c statistic, 0.770; Table 7), although this discrimination was not appreciably better than the discrimination at other visits. Coronary artery calcium may vary by race/ethnicity,30 but we did not adjust for this factor because race/ethnicity was not a significant predictor of advanced atherosclerosis in the PDAY sample.19 Additional refinements using risk factors not measured in the PDAY study may improve prediction; for example, a recent CARDIA publication reported an inverse association between CAC and education level.23
Discrimination of the PDAY risk score was higher than discrimination of the Framingham risk score in predicting CAC in young adults. The Framingham risk score was developed from predominantly middle-aged and older individuals (different from the CARDIA cohort) to predict cardiovascular events that result from atherosclerosis and thrombosis.8 The Framingham risk score is useful but imperfect for identifying middle-aged and older adults at risk for cardiovascular events; the PDAY risk score is useful but imperfect for identifying young persons likely to have precocious atherosclerosis.14-17
Early control of risk factors
The objective of beginning cardiovascular risk prevention in youth is to retard the early progression of atherosclerosis. This study shows the effect of risk acquired early in life and the importance of risk reduction in young adulthood to prevent atherosclerosis. For most individuals, behavioral intervention to maintain cardiovascular health includes not smoking, maintaining normal weight for height, engaging in regular exercise, and consuming a proper diet.9 Among subjects in the Atherosclerosis Risk in Communities study,31 individuals with a sequence variation in PCSK9 had 28% (among African American subjects) and 15% (among white subjects) reductions in low-density lipoprotein cholesterol levels, with 88% (among African American subjects) and 47% (among white subjects) reductions in risk of CHD. This result supports the benefit of even modest reduction in risk factors during a long period.
The PDAY risk score will be helpful in identifying subjects who should undergo imaging studies and those most likely to have rapid progression of atherosclerosis. Imaging can stratify individuals who are at high risk based on the risk factors into those with and without premature atherosclerosis and can assess the rate of progression of atherosclerosis over time with and without intervention.18,32
Because event rates at young ages are low, it is difficult to conduct clinical trials to show that risk reduction reduces event rates.17 Noninvasive imaging of atherosclerosis will facilitate studies of behavioral and pharmacologic risk factor intervention in adolescents and young adults.32 Results from such studies will provide the gold standard of evidence regarding the value of cardiovascular risk factor control in youth.
The following findings support that primordial and primary prevention prevents or delays CHD: (1) Risk scores computed from the CHD risk factors in young people are strongly associated with CAC 10 to 15 years later. (2) Risk scores track. (3) An increase in risk score over time increases the likelihood of CAC. (4) A decrease in risk score over time decreases the risk of CAC.
Correspondence: Samuel S. Gidding, MD, Alfred I. duPont Hospital for Children, 1600 Rockland Rd, Wilmington, DE 19899 (sgidding@nemours.org).
Accepted for Publication: August 15, 2006.
Author Contributions:Study concept and design: Gidding, McMahan, and Williams. Acquisition of data: Gidding and Schreiner. Analysis and interpretation of data: Gidding, McMahan, McGill, Colangelo, Williams, and Liu. Drafting of the manuscript: Gidding, McMahan, and McGill. Critical revision of the manuscript for important intellectual content: Gidding, McMahan, McGill, Colangelo, Schreiner, Williams, and Liu. Statistical analysis: Gidding, McMahan, Colangelo, Schreiner, Williams, and Liu. Obtained funding: Schreiner. Administrative, technical, and material support: Gidding. Study supervision: Gidding.
Financial Disclosure: None reported.
Funding/Support: The CARDIA study is supported by grants N01-HC-48048, N01-HC-48049, N01-HC-48050, and N01-HC-95095 from the National Heart, Lung, and Blood Institute. Dr Gidding was supported in part by grant 1 P20 RR020173-01 from the National Center for Research Resources.
1.Strong
JPMalcom
GTMcMahan
CA
et al. Prevalence and extent of atherosclerosis in adolescents and young adults: implications for prevention from the Pathobiological Determinants of Atherosclerosis in Youth Study.
JAMA 1999;281727- 735
PubMedGoogle ScholarCrossref 2.McGill
HC
JrMcMahan
CAMalcom
GTOalmann
MCStrong
JPPathobiological Determinants of Atherosclerosis in Youth (PDAY) Research Group, Relation of glycohemoglobin and adiposity to atherosclerosis in youth.
Arterioscler Thromb Vasc Biol 1995;15431- 440
PubMedGoogle ScholarCrossref 3.McGill
HC
JrMcMahan
CAMalcom
GTOalmann
MCStrong
JPPDAY Research Group, Effects of serum lipoproteins and smoking on atherosclerosis in young men and women.
Arterioscler Thromb Vasc Biol 1997;1795- 106
PubMedGoogle ScholarCrossref 4.McGill
HC
JrMcMahan
CATracy
RE
et al. Relation of a postmortem renal index of hypertension to atherosclerosis and coronary artery size in young men and women.
Arterioscler Thromb Vasc Biol 1998;181108- 1118
PubMedGoogle ScholarCrossref 5.Berenson
GSSrinivasan
SRBao
W
et al. Association between multiple cardiovascular risk factors and atherosclerosis in children and young adults: the Bogalusa Heart Study.
N Engl J Med 1998;3381650- 1656
PubMedGoogle ScholarCrossref 6.McGill
HC
JrMcMahan
CAZieske
AW
et al. Association of coronary heart disease risk factors with microscopic qualities of coronary atherosclerosis in youth.
Circulation 2000;102374- 379
PubMedGoogle ScholarCrossref 7.McGill
HC
JrMcMahan
CAHerderick
EE
et al. Obesity accelerates the progression of coronary atherosclerosis in young men.
Circulation 2002;1052712- 2718
PubMedGoogle ScholarCrossref 8.National Cholesterol Education Program, 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;1063143- 3421
PubMedGoogle Scholar 9.Williams
CLHayman
LLDaniels
SR
et al. Cardiovascular health in childhood: a statement for health professionals from the Committee on Atherosclerosis, Hypertension, and Obesity in the Young (AHOY) of the Council on Cardiovascular Disease in the Young, American Heart Association.
Circulation 2002;106143- 160[published correction appears in
Circulation. 2002;106:1178]
PubMedGoogle ScholarCrossref 10.Gidding
SSBookstein
LCChomka
EV Usefulness of electron beam tomography in adolescents and young adults with heterozygous familial hypercholesterolemia.
Circulation 1998;982580- 2583
PubMedGoogle ScholarCrossref 11.Fayad
ZAFuster
VNikolaou
KBecker
C Computed tomography and magnetic resonance imaging for noninvasive coronary angiography and plaque imaging: current and potential future concepts.
Circulation 2002;1062026- 2034
PubMedGoogle ScholarCrossref 12.Li
SChen
WSrinivasan
SR
et al. Childhood cardiovascular risk factors and carotid vascular changes in adulthood: the Bogalusa Heart Study.
JAMA 2003;2902271- 2276
PubMedGoogle ScholarCrossref 13.Raitakari
OTJuonala
MKahonen
M
et al. Cardiovascular risk factors in childhood as predictors of carotid artery intima-media thickness in adulthood: the Cardiovascular Risk in Young Finns Study.
JAMA 2003;2902277- 2283
PubMedGoogle ScholarCrossref 14.Greenland
PLaBree
LAzen
SPDoherty
TMDetrano
RC Coronary artery calcium score combined with Framingham score for risk prediction in asymptomatic individuals.
JAMA 2004;291210- 215
PubMedGoogle ScholarCrossref 15.Mozaffarian
D Electron-beam computed tomography for coronary calcium: a useful test to screen for coronary heart disease?
JAMA 2005;2942897- 2901
PubMedGoogle ScholarCrossref 16.Arad
YGoodman
KJRoth
MNewstein
DGuerci
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;46158- 165
PubMedGoogle ScholarCrossref 17.Taylor
AJBindeman
JFeuerstein
ICao
FBrazaitis
MO’Malley
PG Coronary calcium independently predicts incident premature coronary heart disease over measured cardiovascular risk factors: mean three-year outcomes in the Prospective Army Coronary Calcium (PACC) project.
J Am Coll Cardiol 2005;46807- 814
PubMedGoogle ScholarCrossref 18.Wiegman
AHutten
BAde Groot
E
et al. Efficacy and safety of statin therapy in children with familial hypercholesterolemia: a randomized controlled trial.
JAMA 2004;292331- 337
PubMedGoogle ScholarCrossref 19.McMahan
CAGidding
SSFayad
ZA
et al. Risk scores predict atherosclerotic lesions in young people.
Arch Intern Med 2005;165883- 890
PubMedGoogle ScholarCrossref 20.Stary
HCChandler
ABDinsmore
RE
et al. A definition of advanced types of atherosclerotic lesions and a histological classification of atherosclerosis: a report from the Committee on Vascular Lesions of the Council on Arteriosclerosis, American Heart Association.
Arterioscler Thromb Vasc Biol 1995;151512- 1531
PubMedGoogle ScholarCrossref 21.Friedman
GDCutter
GRDonahue
RP
et al. CARDIA: study design, recruitment, and some characteristics of the examined subjects.
J Clin Epidemiol 1988;411105- 1116
PubMedGoogle ScholarCrossref 22.Agatston
ASJanowitz
WRHildner
FJZusmer
NRViamonte
M
JrDetrano
R Quantification of coronary artery calcium using ultrafast computed tomography.
J Am Coll Cardiol 1990;15827- 832
PubMedGoogle ScholarCrossref 23.Yan
LLLiu
KDaviglus
ML
et al. Education, 15-year risk factor progression, and coronary artery calcium in young adulthood and early middle age: the Coronary Artery Risk Development in Young Adults study.
JAMA 2006;2951793- 1800
PubMedGoogle ScholarCrossref 24.Hosmer
DWLemeshow
S Applied Logistic Regression. 2nd ed. New York, NY John Wiley & Sons Inc2000;
25.Agresti
A Categorical Data Analysis. 2nd ed. Hoboken, NJ John Wiley & Sons Inc2002;
26.DeLong
ERDeLong
DMClarke-Pearson
DL Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach.
Biometrics 1988;44837- 845
PubMedGoogle ScholarCrossref 27.Mahoney
LTBurns
TLStanford
W
et al. Coronary risk factors measured in childhood and young adult life are associated with coronary artery calcification in young adults: the Muscatine Study.
J Am Coll Cardiol 1996;27277- 284
PubMedGoogle ScholarCrossref 28.Clarke
WRSchrott
HGLeaverton
PEConnor
WELauer
RM Tracking of blood lipids and blood pressures in school age children: the Muscatine Study.
Circulation 1978;58626- 634
PubMedGoogle ScholarCrossref 29.Webber
LSCresanta
JLVoors
AWBerenson
GS Tracking of cardiovascular disease risk factor variables in school-age children.
J Chronic Dis 1983;36647- 660
PubMedGoogle ScholarCrossref 30.Bild
DEDetrano
RPeterson
D
et al. Ethnic differences in coronary calcification: the Multi-Ethnic Study of Atherosclerosis (MESA).
Circulation 2005;1111313- 1320
PubMedGoogle ScholarCrossref 31.Cohen
JCBoerwinkle
EMosley
THHobbs
HH Sequence variations in
PCSK9, low LDL, and protection against coronary heart disease.
N Engl J Med 2006;3541264- 1272
PubMedGoogle ScholarCrossref