Prevalence and Identification ofAbnormal Lipoprotein Levels in a Biracial

PopuIat~on Aged 23 to 35 Years(The CARDIA Study)

Samuel S. Gidding, MD, Kiang Liu, PhD, Diane E. Bild, MD, John Flack, MD,Julius Gardin, MD, Karen J. Ruth, MS, and Albert Oberman, MD

This study examines the prevalence of abnormal low-density lipoprotein (LDL) and high-density lipoprotein(HDL) cholesterol levels in young adults to determine theability of National Cholesterol Education Program AdultTreatment Panel (ATP) guidelines to identify persanswith elevated LDL cholesterol, to campare ather algo-rithms with those af the ATP, and to determine the con-tributions of race, gender, and ather caronary artery dis-ease risk factors to identifying patients with elevated LDLand low HDL cholesterol. The cohort was population-based, aged 23 to 35 years, and included relativelyequal numbers of blacks and whites, and men andwomen. The prevalence of LDL cholesterol =160 mg/dl(>4.1 mmol/L) was 5% in black women, 4% in whitewomen, 10Yo in black men, and 9’XO in white men. ATPidentified most participants with elevated LDL cholesterol(range: 58.8% of white men to 70.7% of black women).

dentification of a high-risk cohort of young adultswith dyslipidemia must be accomplished using an

efficient screening algorithm.1,2The Coronary ArteryRisk Development in Young Adults Study (CAR-DIA), a longitudinal study of cardiovascular riskfactors in black and white men and women aged 23to 35 years at the time of this study, provides aunique opportunity to assess the efficiency of differ-ent algorithms for identifying abnormal lipoproteinlevels in a community-based sample of youngadults.3The primary aims of this investigation were:(1) to determine the prevalence of abnormal lipo-protein levels and other coronary risk factors as de-fined by the National Cholesterol Education ProgramAdult Treatment Panel (ATP), (2) to assess theability of the ATP algorithm to identify elevatedlow-density lipoprotein cholesterol (LDL) and lowhigh-density lipoprotein cholesterol (HDL), (3) toassess the usefulness of algorithms that use nonlipidrisk factors including obesity as screening criteria,and (4) to determine the influence of race and genderon the performance of the algorithms.

From theChildren’s Plaza, Chicago, Illinois This studywas supportedby Contracts NO1-HC-48047, N01-HC-48048, NOl-HC-48049 and NO1-HC-48050 from the National Heart,Lurw,andBlood Institute,Bethesda, Mar+and. Manuscript receivedNo~ember6, 1995; revised manuscript received and accepted March 4,1996.

Address for reprints: Samuel S. Gidding, MD, 2300 Children’sPlaza/Mail Code 21, Chicaga, Illinois 60614.

Lipoprotein panels would have been required in 6% to7% of womenand to 15%to 18%of men. Aigarithmsthat used nonlipid risk factors required more lipoproteinpanels and identified fewer additional pcuficipants atrisk. The prevalence of HDL cholesterol <35 mg/dl (0.9mmo[/L) was 3% in women, 7y0 in black men, and 13Y0in white men. Algorithms *at used nonlipid risk factorsbefore measuring HDL cholesterol would require HDLcholesterol measurements in 35°A of whites and 56’%0ofblacks, but reduced sensitivity far identi&ing low HDLcholesterol (range: 58% in white men to 93~0 in blackwomen). In young adults, algorithms based on nonlipidrisk factors and family history have lower sensitivi~,and increase rather than decrease the number of fastinglipoprotein panels required when compared with ATPlevels.

(Am J Cardiol 1996;78;304-308)

METHODSCARDIA is a longitudinal study of the evolution

of coronary artery disease risk factors in 5,115 youn%adults aged 18 to 30 years at initial examination.

Four clinical centers conducted baseline examina-tions between March 1985and June 1986,and 3 sub-sequent examinations have since been completed.For the Birmingham, Alabama, Chicago, Illinois,and Minneapolis, Minnesota centers, the participantswere randomly recruited from the total communityor from selected census tracts in the community. Forthe Oakland, California center, the participants wererandomly recruited from the Kaiser-Permanentehealth plan membership’.Detailed descriptions of thesampling methods have been published.3The CAR-DIA cohort has relatively equal-sized groups of menand women, blacks and whites, and participants withless than and more than a high school education. Forthis study, measurements made at the third ( 1990 to1991) examination, when participants were aged 23to 35 years, were used. The study was approved bythe institutional review boards at all participatingcenters.

The cohort for the present analyses comprisedpersons (n = 4,244) with measurement of lipopro-teins. Exclusion criteria were fasting <12 hours (n= 524), pregnant or possibly pregnant (n = 113),and taking lipid-lowering medication (n = 10). Anadditional group had absent information on riskstatus (n = 111). Some persons had multiple exclu-sion criteria. This left a cohort of 3,486: 920 black

304 01996 by Excerpta Medica, IncAll rights reserved.

0002-9149/96/$15.00Pll S0002-9149(96]00282-2

women, 959 white women, 688 black men, and 919white men. Results were similar when analyzed withand without those without information on risk status.

Measurements: Total cholesterol, triglycerides,and HDL cholesterol were determined using enzy-matic procedures at the Northwest Lipid ResearchClinic Laboratory. The laboratory participates in thelipid standardization program of the Center for Dis-ease Control. HDL cholesterol was measured afterdextran sulfate-magnesium precipitation of apoli-poprotein B–containing particles. LDL cholesterolwas estimated using the Friedewald equation4 (LDLcholesterol = total cholesterol – HDL cholesterol –triglycerides/5). Subjects with triglycerides 2400mg/dl (4.5 mmol/L) did not have calculable LDLcholesterol and were analyzed as a separate group.For construction of algorithms based on ATP, thesingle lipid and lipoprotein value obtained on eachparticipant was considered the “true” value. This isdifferent from the ATP algorithm, which suggeststhat final classification be based on the average of=2 calculated LDL cholesterol values 1.

Smoking status, parental history of myocardialinfarction, or coronary-related death (including ageat the time of the event), presence of diabetes mel-litus, and use of antihypertensive medications wereobtained by questionnaire. After a 5-minute rest in aquiet room, blood, pressure was measured 3 timeswith a random-zero sphygmomanometer. The last 2readings were then averaged. Body weight was mea-sured to the nearest 0.2 pound using a calibratedscale with the participant in light clothing withoutshoes. Height was measured to the nearest 0.5 cmusing a vertical ruler. Body mass index was calcu-lated as weight (kg)/height(m)2.

Assignment of risk factors: Risk factors were as-signed according to ATP criteria.l Elevated or high-risk total cholesterol was 2240 mg/dl (6.2 mmol/L), and borderline high was a value from 200 to 239mg/dl (5.2 to 6.2 mmol/L). Elevated or high-riskLDL cholesterol was =160 mg/dl (4.1 mmol/L),and borderline high risk was a value between 130and 159 mg/dl (3.4 to 4.1 rnmol/L). Subjects withborderline high LDL cholesterol and =2 nonlipidrisk factors (as defined below) were also consideredhigh risk. The extremes of the HDL cholesterol dis-tribution were defined as low, <35 mg/dl (0.9mmol/L) or protective, =60 mg/dl ( 1.6 mmol/L).Triglycerides were considered borderline high if be-tween 200 and 400 mg/dl (2.3 to 4.5 mmol/L) andelevated if =400 mg/dl (4.5 mrnol/L).

Nonlipid risk factors were defined as in the ATPalgorithm: current smoker, diabetes mellitus, hyper-tension (taking antihypertensive medication, havingsystolicblood pressure >140 mm Hg, and/or havingdiastolic blood pressure =90 mm Hg), and positivefamily history (either a mother with myocardial in-farction or coronary death before age 65 years or afather with a similar history before age 55 years).Those who were unaware of their family history forcoronary artery disease were considered to have anegative family history. As all participants were aged

<45 years, age was not considered in the algorithms.History of coronary artery disease was present in 2.Obesity was defined as body mass index =30 kg/m2(analogous to 130%ideal weight, determined by cal-culating body mass index for 1309Z0ideal weight onMetropolitan Life tables for both men and women,verified in this cohort, data not shown).

Dataanalysis:All analyses were stratified by raceand gender. Prevalence of lipid and nonlipid riskfactors were calculated. Test characteristics (sensi-tivity, specificity,positive predictive value, negativepredictive value) were calculated for the ATP algo-rithm. A positive test was a test that identified a par-ticipant with a lipoprotein abnormality. Sensitivitywas the percentage of those with the abnormalityidentified by the screening test ([n with a positivetest and the abnormality/n with the abnormality] X100). Specificity was the percentage of those withnormal screening tests who did not have a lipopro-tein abnormality ([n with a negative test and’no ab-normality/n with no abnormality] X 100). Positivepredictive value was the percentage of positive teststhat were truly positive (the lower the value, themore “false positive” tests). Negative predictivevalue was the percentage of negative tests that weretruly negative (the lower the number, the more‘‘false negative’ tests). Lipid and lipoprotein levelsof those with and without specific nonlipid risk fac-tors were compared using unpaired t tests. Risk fac-tor sums were calculated as the sum of the nonlipidrisk factors present plus the risk assigned to HDLcholesterol (low 1, intermediate O, protective –1;possible range of the sum, 1 to 5). These sums wereused to identify participants with cardiovascular riskfactors in analyses of the ATP and other algorithms.

RESULTSThe prevalence of elevated total cholesterol, ele-

vated LDL cholesterol, elevated triglycerides, lowHDL cholesterol, and protective HDL cholesterol areshown in Tables I and II. Elevated total and LDLcholesterol levels were twice as likely to occur inmen than in women. Low HDL cholesterol was pres-ent in about 3Y0of women, but was present in 7.09i0of black men and 12.89toof white men. Conversely,protective HDL cholesterol ( 260 mg/dl [1.6 mmol/L]) was relatively common in women and much lesscommon in men, particularly whites. The highestprevalence of elevated triglycerides was in whitemen.

The prevalence of nonlipid risk factors variedwidely by race and gender (Table II). Smoking wasby far the most common risk factor. Blacks weremore likely to smoke, have hypertension, or have apositive family history. Women were more likely tobe diabetic. The prevalence of =2 nonlipid risk fac-tors is about 79Z0in all race/gender groups, exceptfor white women in whom multiple risk is much lesscommon, 2.6$Z0.

Algorithms for identifying elevated low-density lipo-protein cholesterol: Table III lists the test characteris-tics of the ATP algorithms for classifying risk by

PREVENTIVE CARDIOLOGY/LIPOPROTEINS IN YOUNG ADULTS 305

ITABLE I Prevalence of Elevated IDL Chalesteral, Elevated Tatal Cholesterol, and Elevated Triglyceridesas Defined by the Adult Treatment Panel

Black Women White Wamen Black Men White Men(n= 920) (n = 959) [n = 688] (n =919)

LDL 44 (4.8) 41 (4,3) 68 (9,9] 80 (8.9]LDL cholesterol 130-159 mg/dl 154 (1 6.8) 122 [12.7] 131 [19.1] 186 (20.6)Total cholesterol =240 mg/dl 28 (3.0) 35 [3,6] 51 [7.4] 57 (6.2)Total cholesterol 200-239 mg/dl 185 (20.1 ) 143 (14.9) 145 [21 ,1]Triglycerides =400 mg/dl

196 (21 .3)1 (o. 1) 2 (0.2) 3 (0.4) 16 [1 .7)

Triglycerides 200-399 mg/dl 3 [0.3) 11(1.1) 19(2.8) 66(7.2)Values are expressed os number (%).For low-density lipoprotein cholesterol prevalence,persons with triglycerides 2400 mg/dl were excluded.

LDL= Iow-densiiy lipoprotein,

TABLE II Prevalence af Risk Factors and Protective Factor as Defined by the Adult Treatment Panel

Black Women White Women Black Men White Men(n= 919) (n= 957) (n= 685) (n = 903)

Positive risk factors

Current smokers 264 (28,7] 207 (21 .6] 247 [36,1] 212 (23.5)Family history of CAD 75 [8,2] 36 [3.8] 33 [4,8] 18 (2.0)Diabetes mellitus 32 (3,5] 20 (2,1] 6 [0.9] 6 (0.7)Systemic hypertension 57 (6.2] 11 (1,1] 44 (6.4) 38 (4.2)HD1 cholesterol <35 mg/dl 27 [2.9] 28 (2.9) 48 (7.0) 116 [1 2.8)

Protective factorHD1 cholesterol =60 mg/dl 331 [36,0) 390 (40.8] 174 (25.4) 180 [1 2,0]=2 Risk factors 67 (7.3) 25 (7.6) 42 (6.1) 63 [7.0]

Values are expressed os number (%).

CAD = coronary artery disease; HDL = highdensity lipoprotein

have required a lipoprotein profile (total cholesterol>240 mg/dl [6.2 mmol/L], HDL cholesterol <35mg/dl [0.9 mmol/L], or total cholesterol 200 to 239mg/dl [5.2 to 6.2 mmol/L] and =2 risk factors)varies by race/gender group: 6.5~ofor white women,7.0% for black women, 14.6% for black men, and18.2% for white men. For all race/gender groups,sensitivity for identifying those with elevated LDLcholesterol ranged from a low of 58.8?Z0for whitemen to a high of 70.7~0for white women. Negativepredictive values (the accuracy of a negative test) of98% to 99% were achieved for women and 96% formen. Positive predictive value (the accuracy of apositive test) and specificity (the percentage of neg-ative tests truly negative) were reduced by the largenumber of profiles required because of low HDLcholesterol, particularly for white men who had thehighest prevalence of low HDL cholesterol.

The ATP algorithm also identified 8 of 12 blackwomen (67?ZO),4 of 4 white women ( 100?ZO),11 of13black men (85%), and 16 of 17 white men (94%)with borderline high LDL cholesterol and =2 asso-ciated risk factors who would also be candidates foraggressive dietary intervention.

Table III also lists the test characteristics of astrategy that recommends lipoprotein panels for allpersons with -1 established risk factor. This strat-egy is of limited usefulness because the number whorequire fasting lipoprotein profiles is much higher(two- to sixfold depending on the race/gender

elevated LDL cholesterol. The number that would group) than when using the ATP algorithm, and thenumber tested varies widely by gender independentof the prevalence of elevated LD-Lcholesterol (e.g.,black men have a lower likelihood of having ele-vated LDL cholesterol but require more lipoproteinpanels than white men). Table III also shows thatthe percentage of subjects identified as having highLDL (the sensitivity) is only slightly greater thanthe percentage tested for black women and whitemen; for black men the untested population actuallyhas a higher percentage of patients with elevatedLDL cholesterol (comparison of percentage testedand sensitivity in Table III). This is because of apoor association of elevated LDL with nonlipid riskfactors including family history (data not shown).

The addition of obesity (body mass index =30kg/m2) to the list of nonlipid risk factors in the ATPalgorithm increased the number of persons needinglipoprotein panels but did not significantly increasesensitivity. An additional 18 black women (2.090),5 white women (0.5%), 5 black men (0.7%), and 4white men (0.490) would have lipoprotein panelsperformed identifying 2 additional black women, 1white woman, and 1 black man with elevated LDLcholesterol.

Algorithms for identifying low high-density lipoproteincholesterol: The current algorithm of universalscreening for HDL cholesterol identifies all personswith low HDL cholesterol. Because the prevalenceof HDL cholesterol <35 mg/dl (0.9 mmol/L) islow, particularly for women, alternative strategies

306 THE AMERICAN JOURNAL OF CARDIO1OGY” VOL 78 AUGUST 1, 1996

TABLE Ill Test Characteristics

Black Women White Women Black Men White Men

Test Characteristics of the Adult Treatment Panel Algorithm for Identifying LD1 Cholesterol = 160 mg/dl [%)

% having a lipoprotein panel 7,0 6.5 14,6 10.2Sensitivity 65.9 70.7 69.1Specificity

58.896.0 96.4 91 .A 85.s

Pos. Pred. value 45.3 46.13 47,0 26.7Neg. pred, value 98.2 98.7 96,4 95.5

Test Characteristics of cm Algorithm thot Meosures lipoproteins on All Participants With z 1 Adult Treatment Panel Ncmlipid Risk Factor

to Identify those With IDL Cholesterol >160 mg/dl (%)

% having a lipoprotein panel 39.6 26.8 42,2Sensitivity

28.054.5 26.8 30.9 A5.O

Specificity 61.1 73.3 56.6 73.6Pos. pred. value 6.6 4.3 7.3 14.2Neg. Pred. value 96.4 95.7 88.1 93.2

LDL= low-densitylipoprotein; Neg. pred. = negative predictive; Pos. pred. = positive predictivevalue.

TABLE IV Test Characteristics of Three Algorithms to Identifylow HD1 Cholesterol [<35 mg/dl)*

NI.RFAll (%) (%] NLRF + BMI (%]

Black Women

% having HDL 100 39.6 56.7measured

Sensitivity 100 81.5 92.6

Specificity — 61.7 44.4

Pos. pred. value 2.9 6.0 4,8

Neg. pred. value — 99.1 99,5

White Women

% having HDL 100 26.8 34.8

measuredSensitivity 100 60.7 78.6

Specificity — 74.3 66.5

Pas, pred. value 2.9 6.6 6.6

Neg. pred. value — 98,4 99.0

Black Men

% having HDL 100 42.2 54,3

measuredSensitivity 100 37.5 68.8

Specificity — 57.5 46.8

Pas. pred. value 7.0 6.2 8.9

Neg. pred. value — 92.4 95.2

White Men

% having HDL 100 28.0 35,4

measuredSensitively 100 43.1 57.8

Specificity — 74.2 67.9

Pos. pred. value 12.8 19.8 20.9

Neg. pred. value — 89,8 91.6

● Testing Everyone (All], Testing Those with = 1 National Cholesterol Educa.tion Program Adult Treatment Panel Nonlipid Risk Factor (NLRF), and TestingThose with a NLRF or with body mass index =30 kg/m2 [NLRF + BMI).

BMI = body mass index; HDL = highdensity lipoprotein; NLRF = nonlipid riskfactor.

short of universal screening were considered. In Table IV, the usefulness of measuring HDL cholesterolonlv in those with 1 ATP risk factor and in thosewith severe obesity (body mass index =30 kg/m2)is shown. HDL cholesterol is generally decreased in

Persons with ATP risk factors (data not shown). In~hemodel that includes obesity; over half of biacksand roughly 1/3 of whites would have HI)L choles-terol measured. The percentage identified rangesfrom a low of 58% of white men to 93% of blackwomen. This is because white men have a high prev-alence of low HDL cholesterol and a low prevalenceof nonlipid risk factors including severe obesity;black women have a high prevalence of obesity.

Identification of elevated triglycerides: A total of 22participants (16 white men and 6 others) had incal-culable LDL cholesterol on the basis of z 1value fortriglycerides =400 mg/dl (4.5 mmol/L). Of these,19 of 22 would have had fasting triglycerides mea-sured in the ATP algorithm either because of ele-vated total cholesterol, low HDL cholesterol, or bor-derline high total cholesterol and >2 risk factors.

DISCUSSIONThis study has quantified the sensitivity and

specificity of the ATP algorithm in identifying per-sons with elevated LDL cholesterol and low HDLcholesterol in young adults, and compared this strat-egy with others that could be considered. Prescreen-ing by universal measurement of nonfasting total andHDL cholesterol would limit the number of fastinglipoprotein determinations necessary to 7% ofwomen and to 15% to 18% of men. Roughly 2/3 ofthose with elevated LDL cholesterol and all of thosewith low HDL cholesterol would be identified. Noother LDL prescreening strategy tested achieved theefficiency of this algorithm; all others required morelipoprotein panels with comparable or lower sensi-tivity. Another study,by Sempos et al,shas estimatedthe outcome of applying ATP to young adults usingthe National Health and Nutrition Survey III con-ducted between 1988 and 1991. An estimated 23%of men and 18% of women aged 20 to 34 yearswould require fasting lipoprotein measurements.These values are slightly higher for men and consid-erably higher for women because of a higher prev-alence of coronary artery disease, defined as a pos-itive response to the Rose questionnaire for angina

PREVENTIVE CARDIOLOGY/LIPOPROTEINS IN YOUNG ADULTS 307

and a 2% higher prevalence of elevated LDL cho-lesterol.

Recently, the value of universal screening ofyoung adults to identify abnormal lipoproteins hasbeen challenged.2,6Young adults,particularlywomen,

have a relatively low incidence of coronary artery dis-ease, and the benefitsof primary preventionbeginningat a young age are potentially diminished by the pos-sibilityof unintendedharmful effects of medical ther-apy. In addition, this approach is unlikely to have afavorable cost/benefit ratio.’ It has been suggestedthat cholesterol testing be restricted to those withknown coronary artery disease, several other risk fac-tors, or with documented family history of familialhypercholesterolemia.2 The present study suggeststhat selective screening strategies based on nonlipidrisk factors, including family history or presenceof coronary artery disease, would not help identifythose with abnormal lipoproteins.Selective screeningwould underidentify those young adults who couldbenefit from nonpharmacologiclowering of their riskstatus, particularly black men. The ATP algorithm isthe most efficient,limiting the measurementof fastinglipoproteinsto a minority of young adults and testingmore of those most likely to have a future event—men with elevated total cholesterol or other risk fac-tors.8’9However, it does rely on a calculated LDLcholesterol that may be less reliable in persons withhigh triglycerides.

Many participants could not provide accuratefamily history either of a coronary event or of theage at which the event occurred. This is in contrastto other nonlipid risk factors that are more easilyverified by an examining physician. In this study,family history was not particularly good at identi-fying participants with dyslipidemia. In the BogalusaHeart Study, family history did identify trends to-ward adverse risk profiles in adolescents; however,there was a substantial inaccuracy rate in the assess-ment of family history.’”

Obesity is not considered a risk factor in ATP,whereas it was included in the original ATP guide-lines.ll Obesity is associated with low HDL choles-terol; therefore, it is helpful as a criteria in selectivescreening strategies to identify those with low HDL

cholesterol but unnecessary in universal screeningstrategies.

In summary, cholesterol screening strategies suchas ATP, which include the initial measurement oftotal cholesterol and HDL cholesterol, are more ef-ficient than those that depend on nonlipid risk fac-tors. Young adult men are more likely to benefit froma universal screening strategy than young women be-cause they have the highest prevalence of elevatedLDL cholesterol and low HDL cholesterol. Use ofnonlipid prescreening criteria in algorithms overev-aluates black women (high prevalence of nonlipidrisk factors and low prevalence of elevated LDL cho-lesterol), and has very poor test characteristics inblack men (high prevalence of nonlipid risk factors,poor relation of these to elevated LDL cholesterol).

1. Expert Panel on Detection, Evaluation, and Treatment of High Blood Cho-lesterol in Adults. Summiwy of the second report of rfre National CholesterolEducrttion Program (NCEP) expert panel cm detection, evalurttion, and treat-ment of high blood cholesterol in adults (Adult Treatment Panel II). JAMA1993;269:3015-3023.2. Hulley SB, Newman TB, Grady D, Garber AM, Baron RB, Browner WS.Should we be measuring blood cholesterol in young adults? JAMA1993;269:1416-1419.3. Friedman GD, Cutter G, Donahue RP, Hughes GH, Hulley SE, Jacobs DRJr, I.iu K, Savage PJ. CARDIA: stndy design, recruitment, and some charac-teristics of examined subjects. J C[in Epidemiol 1988;44:1105-1116.4. Friedewald WT. Levy RI, Fredrickson DS. Estimation of the concentrationof low density lipoprotein cholesterol in plasma, without the use of the prepar-ative centrifuge. Clin Chem1972;18:499–501.5. Sempm CT, Cleeman JI, Carroll MD, Johnson CL, Bachorik PS, GordonDJ, Burt VL, Bricfel RR, Braun CD, Lippel K, Rifkind BM. Prevalence of highblood cholesterol among US adults: an update based on guidelines from thesecond repofi of the National Cholesterol Education program Adult TreatmentPanel. JAMA 1993;269:3009–3014.6. Krahn M, Naylor DC, Basinski AS, Detsky AS. Comparison of an aggressive(U. S.) and a less aggressive (Canadian) .policy for cholesterol screening andtreatment. Ann Intern Med 1991;115:248–255.7. Goldman L, Gordon DJ, Rifkind BM, Hulley SB, Detsky AS, Goodman DS,Kinasian B, Weinstein MC. Cost and health implications of cholesterol low-ering. Circuladcm 1992;85:1939–1941.8. Klag MJ, Ford DE, Mead LA, He J, Whelton PK, Liang KY, Levine DM.Sernm cholesterol in young men and subsequent cardiovascular disease. NEnglJMed 1993;328:313-318.9. Starnler J, Stamler R, Brown WV, Gotto AM, Greenland P, Grrmdy S,Hegsted M, Luepker RV, Neaten JD, Steinberg D, Stone N, Van Horn L, Wiss-ler RW. Serum cholesterol: doing the right thing. Circulation 1993;88:1954-1960.10. Bao W, Srinivasan SR, Wattigney WA, Berenson GS. The relationship ofparental cardiovascular disease to risk factors in children and young adults: theBogalusa Heat Study. Circukzdon 1995;91:365-371.11.Report of the National Cholesterol Education Program Expert Panel ondetection, evaluation, and treatment of high blood cholesterol in adults. ArchInter-nMed 1988;148:36-69.

308 THE AMERICAN JOURNAL OF Cardiology@ VOL. 78 AUGUST 1, 1996