ischemic heart disease, atherosclerosis, and...

Ischemic Heart Disease, Atherosclerosis,and Longevity

By JOHN W. GOFMAN, M.D., PH.D., WEI YOUNG, PH.D., AND ROBERT TANDY

IN THE LATE 1940's Duff and McMillan' 2made the remarkable observation that al-

loxan-diabetic rabbits developed equal orhigher blood cholesterol levels, but markedlyless atherosclerosis, than did nondiabetic rab-bits fed an equivalent high-cholesterol diet.From the excessive visible lipemia character-istic of the alloxan-diabetic rabbits in thisexperiment, Duff and McMillan surmised thatthe physical state of blood cholesterol mightbe altered in the alloxanized animals and thatthis alteration might account for their lowerdegree of atherosclerosis.During those same years, the difficulties at-

tending the study of serum macromolecules inthe ultracentrifuge were understood and re-solved.3-5 It was shown that in humans and inseveral other mammalian species a spectrum oflipoproteins, bearing cholesterol, triglycerides,and phospholipids, existed spontaneously orcould be induced by a variety of experimentalmeans. In the rabbit it was demonstratedearly6' 7that aortic atherogenesis, at least, wasstrongly related to the serum level of lipopro-teins of a rather narrow size-range, the levelof lipoproteins larger or smaller in size beingunrelated to the rate of development of theatherosclerotic lesion. The Duff observationswere explained when it was demonstrated thatthe alloxanized, cholesterol-fed rabbit trans-ported its massive concentrations of cholesterolin lipoproteins of Sf values greater than 80,with almost no transport in the form of athero-sclerosis-associated Sf 12-40 lipoproteins.7

From the University of California, Bio-Medical Re-search Division, Lawrence Radiation Laboratory,Livermore, Califomia.

This communication was presented as the LymanDuff Memorial Lecture, entitled "Atherosclerosis,1965," at the meeting of the Council on Arteriosclero-sis, American Heart Association, Bal Harbour, Florida,October 13, 1965.Circulation, Volume XXXIV, October 1966

The obvious and exciting question presenteditself, "Could a similar phenomenon be opera-tive in the human population developingatherosclerosis of such critical arterial sites asthe coronary or cerebral beds?" It was evidentthat such a question had to be studied in thehuman, for valid concern existed that rabbitatherosclerosis might bear little or no rela-tionship to the stenotic arterial disease re-sponsible for ischemic heart or brain diseasein the human.Two potential avenues of investigation were

evident: (1) a study of the relationship ofvarious serum lipoproteins to the occurrenceof overt clinical ischemic disease, such asischemic coronary heart disease; and (2) astudy of the relationship of various serumlipoproteins to the degree of intimal arterialnarrowing (a key feature of human athero-sclerosis). Both avenues had much to com-mend them; both faced formidable obstaclesand extensive requirements on the path tosuccessful execution.To a large extent, studies of both types have,

by now, been completed. Certain apparentparadoxes seem to attend the results. Our pur-pose herein is to present all the evidencetogether with a framework which allows forat least a first-order, paradox-free understand-ing of how the various facets of evidence fittogether.

The Relationship of Serum Lipidsand Lipoproteins to Overt Clinical

Ischemic Heart DiseaseThe early premise in this approach was that

persons manifesting overt clinical ischemicheart disease at a given point in life must havebeen developing stenotic coronary artery dis-ease at a greater rate than cohorts not mani-festing clinical disease at that same point inlife. Ideally, such information must be soughtin a prospective study, that is, one in which

679

by guest on April 29, 2018

http://circ.ahajournals.org/D

ownloaded from

http://circ.ahajournals.org/

GOFMAN ET AL.

the "cases" of clinically overt heart diseasearise out of an ostensibly healthy substratesample of the population. Only in this waycan it be assured that the "cohorts" truly aresuch. A study of this type that included mea-surement of serum cholesterol had been initi-ated by the United States Public Health Ser-vice at Framingham, Massachusetts. In 1950,through the participation of the DonnerLaboratory, studies of the low-density lipo-proteins in serum were added to the Framing-ham program. In 1953, a similar study,including serum cholesterol, low-density lipo-proteins, and high-density lipoproteins inindustrial employees of the Lawrence Radia-tion Laboratory at Livermore, California, wasinitiated. The results of 12 years of follow-upin the Framingham study and 10 years in theLivermore study will be presented here.

Retrospective Studies of ClinicalIschemic Heart Disease

Pending the anticipated long period forcompletion of the prospective studies, it wasfelt that preliminary insight might be obtainedthrough the study of blood lipoproteins incases of already established clinical ischemicheart disease, such as survivors of myocardialinfarction. The prime advantage was that suchcases were immediately available and inreasonably large numbers. But major disad-vantages were also present: (1) The possibilitythat blood lipoprotein status might be non-representative in the postmyocardial infarctionperiod was not readily to be ruled out in anunequivocal manner. (2) Even more distress-ing, was the inordinate difficulty in assuringa proper choice of cohorts without overtischemic heart disease. The cases of clinicallyovert ischemic heart disease came from hos-pital sources, from physicians' practices, orfrom the affected individuals still at work inindustrial populations under study. Establish-ment of true "cohorts," including such vari-ables as weight status, lipid status, and bloodpressure status, was painful at best, and neverreally successful. Nevertheless, highly sug-gestive information did emerge from thoseearly studies, but always with the reservationthat the prospective study might alter at least

some quantitative aspects of the conclusions.For example, the mean levels of four serumlipoprotein classes, Sf0 0-12, Sf0 12-20, Sfo20-100, and Sf0 100-400, and of the serumcholesterol were significantly elevated in theclinical ischemic heart disease cases contrastedwith "ostensible cohorts," matched by age andsex. -10A major objective of all such studies was to

determine whether, for the human coronaryarteries, any particular segment of the entirelipoprotein spectrum was the atherogenic seg-ment or was more atherogenic than borderingsegments, as had already been clearly demon-strated for aortic atherogenesis in the rabbit.It became clear that the weighting factors,expressing the relative importance to athero-genesis of individual segments of the lipopro-tein spectrum, were extraordinarily sensitiveboth to the source of clinical material and to"ostensible cohort" selection. As a result, amore limited objective was attempted, namely,to assign a relative weighting factor to twobroad bands in the lipoprotein spectrum, theSfo 0-12 and the Sf0 12-400 classes. To providesome working basis for ongoing studies, pend-ing completion of the prospective studies, afunction was defined as follows:

Atherogenic index, or A.I. =0.1 (Sfo 0-12) + 0.175 (Sfo 12-400).

The choice of 0.175 was made by the lineardiscriminant analysis of the then existingclinical cases9 plus its "ostensible cohort popu-lation." It has been repeatedly stressed11 12that this relative importance factor was basedupon the study of survivors of myocardialinfarction, that the prospective study mightvery well alter it, and further that some partof the spectrum from Sf0 0 to Sf0 400 mightdrop out of consideration with respect toprovision of an independent contribution todevelopment of disease.13

Still another difficulty arose early in thestudy of already established clinical ischemicheart disease. Both for the serum cholesterolmeasure and for the low-density lipoproteinmeasures, the difference between overt diseaseand "ostensible cohorts" decreasd markedly

Circulation, Volume XXXIV, October 1966

680



ownloaded from


ATHEROSCLEROSIS

with increasing age of the subjects.", 14 Thisextremely important feature was itself sorelybedeviled by the "appropriate cohort" prob-lem. Here again, only the prospective studiescould be expected to provide resolution andunderstanding. There were some who inferredfrom the age-dependent difference that severaldifferent diseases were involved, one or severalfor young subjects, another, or several others,for older subjects.15 Others suggested that thewrong lipid parameters were being measured,and that the proper parameter would not showsuch an age-dependent relation.'6 Neither ofthese suggestions has ever appeared reason-able or attractive to the present authors. In-deed, the considerations below will lead to asingle unifying concept to explain this age-related phenomenon, the existence of whichhas been amply confirmed in the prospectivestudies.Prospective Studies of iClinicalIschemic Heart Disease

Sufficient numbers of de novo cases ofischemic heart disease have now arisen bothIn the Framingham and the Livermore studiesto warrant examination of the results. How-ever, even from some 357 new events inthe two studies combined, certain questionsare still difficult to answer with high con-fidence.The first set of results is that for the Fram-

ingollam study, including the serum lipoproteinand cholesterol data determined at DonnerLaboratory (records of which have been onfile at the National Heart Institute for over 10years). Through the generosity of Dr. ThomasR. Dawber, the director of the Framinghamstudy, a listing of de novo cases arising during12 follow-up years has been made availablefor this evaluation. Included in this group arean appreciable number of subjects for whomserum lipoprotein and cholesterol measure-ments are available for two examinations, 1 to

3 years apart, before the occurrence of de novo

ischemic heart disease. For such a group,errors due to technical and biological variationare minimized, and hence the results for thisgroup are presented separately.The second set of results is for 1,961 male


subjects employed at Livermore and studiedinitially between 1954 and 1957. The follow-upperiod is approximately 10 years. The recordsof the cases of de novo ischemic heart diseasehave been made available for this studythrough the cooperation of Dr. Max Biggs,Medical Director at Lawrence RadiationLaboratory, Livermore, and his staff, andthrough the diligent efforts of Mrs. MargaretSoderberg of the Bio-Medical Research Divi-sion there. This second, independent studyallows examination of the possibility, alwaysconsidered remote, that some special or non-representative feature might be present in theFramingham population sample. It also allowsexamination of the high-density lipoproteinsin relation to the prospective occurrence ofischemic heart disease.*

Test of the Hypothesis That Levels of Certain SerumLipoproteins and Serum Cholesterol Are Elevatedin Advance of de novo Ischemic Heart DiseaseThe Framingham Study. All subjects for

whom lipoprotein measurements are availableare included in this analysis with the followingexceptions: (1) subjects with definite heartdisease at entry to the study, and (2) dia-betics known to be such at entry. New case3of heart disease arising in Framingham areprovided (in nonduplicative manner) in fivecategories of ischemic heart disease: (1) myo-cardial infarction, (2) coronary heart disease(death, but not sudden), (3) coronary in-sufficiency, (4) sudden death, and (5) anginapectoris. The basis for this reporting systemhas been described in publications of thebroader aspects of the Framingham study.'7-19While for certain purposes it might be im-portant to use only the group with theapparently strongest manifestation of ischemicheart disease, there is a distinct overridingrationale for combining all cases into onegroup, labeled "de novo ischemic heartdisease" (I.H.D.). The rationale is that thetrue problem to which we address ourselvesis the stated "epidemic of our time-ischem-ic coronary heart disease," rather than any

*In all the tables the data are for the Framinghampopulation sample except where specifically labeledas Livermore.

681



ownloaded from


GOFMAN ET AL.

subsegment thereof. No significant differencesexist either for serum lipoproteins or serumcholesterol between the strongest diagnosticcategory of de novo ischemic heart disease,that is, myocardial infarction, and the otherfour categories combined.The findings in 221 cases of de novo is-

chemic heart disease in men and in 98 casesof such disease in women are presented intables 1 and 2, respectively. The mean agesnoted in those tables are the means for thegroup of de novo cases at the time of entryinto this study. In both tabulations the denovo cases are compared with the substratepopulation out of which they arose, with allparameters adjusted to match the de novocases upon the variable of age. Age-matchingwas achieved in the following manner. Sincethe de novo cases include cases from the30-39, 40-49, and 50-69 year age groups, themean parameter values for the substrate agegroups were weighted in the same proportion

Table 1MENAgeGroup 30-69yrs. DONovo Ischemic Heart Disease (221 Cases)Man Age *49.2 yrs. VS.12yr. Followup Base Population (2022 Subjects)

DeNovo Base * SignificanceMeasure LKD. Population Difference Test

Sf 0-12 386.3 359.0±82.1 27.3 p 0.00lSf 12-20 76.4 67.0+27.8 9.4 p-0.001Sf 20100 120.9 108.2±48.8 2.7 p 0.00ISt 100-400 94.0 76.7 74.1 7.3 pa 0.001

A.I. 89.2 80.0±24.9 9.2 pQ0.001Cholesterol 249.0 230.7±40.8 8.3 p 0.001Systolic B. P 143.8 138.3 ±20.8 5.5 p-0.001Diastdic B.P 91.8 88.5±12.5 3.3 p-0.001Reltive Wt 16.3 1132± 15.1 3.1 p- 0.001

mMEAN I STANDARD DEVIATION OF DISTRIBUTION

Table 2WOMEN

Age Group 4069yrs. DeNovo Ischemic Heart Disease (98 Cases)Men Age 54.2 yrs. VS.12yr Fotlowup Base Population (1689 Subjects)

DeNuoro Base * SignificanceMeasure LHD. Population Difference Test

Sf 0-12 383.4 364.7±76.2 8.7 p- 0.02Sf 12-20 94.4 90.2±45.7 4.2 N. S.Sfo 20-100 109.8 98.1±47.8 1.7 p= 0.02Sf 100400 57.7 50.3±43.2 7.4 N. S.

A. I. 84.2 78.2±23.7 6.0 p=0.0Cholesterol 267.2 253.5±45.4 3.7 pa 0.005Systolic B.P 161.6 149.0- 26.8 2.6 pc 0.001Diastolic B.B 94.8 90.5± 13.3 4.3 p= 0.001Relative Wt. 126.0 121.0±21.4 5.0 pa 0.02* MEAN ± STANDARD DEVIATION OF DISTRIBUTION

as the distribution of de novo cases. A similarprocedure was adopted to provide a weightedstandard deviation in the base populationvalues. Failure to make such an age-matchingcan lead to deceptive results.For men, it can be stated, at or beyond

the P = 0.001 level of confidence, that Sfo0-12, Sfo 12-20, Sf0 20-100, Sfo 100-400 lipo-proteins, the atherogenic-index (A.I.) func-tion, the serum cholesterol level, the systolicand diastolic blood pressures, and the relativeweight are all elevated in those destined todevelop ischemic heart disease during a 12-year period following initial study. However,as a result of correlation, imperfect butdefinite, between several of the measuredvariables,'1 it is not possible to translatethe positive findings into any statement ofindependent contribution of each variable tothe predictive association with ischemic heartdisease. Even this series of 221 cases of denovo ischemic heart disease suffers fromsufficiently large standard errors of the differ-ence in means (I.H.D. vs. base population)for several measures to preclude a reliableassessment of weighting factors, and, in turn,strength of independent contribution. Theanalogous data for women show similar trends,but for all the measures of lipid and lipo-protein, the differences between I.H.D. andbase population are much smaller than theyare for men. Also the overall group ofwomen with de novo ischemic heart diseaseis 5 years older at entry than the correspond-ing overall group of men.Age Trend in Segregation of de novo Ischemic HeartDiseaseThe data for men are subdivided into

four categories in tables 3 to 6, utilizing agegroups 30-39 years, 40-49 years, 50-55 years,and 56-69 years. For all lipid and lipoproteinparameters the differences (I.H.D. vs. basepopulation) are much greater for the 30-39year group than for those above 40 years. Asfor the overall group, the standard errors ofdifferences in means for I.H.D. versus basepopulation do not allow accurate assessmentof independent weighting factors for the agesubcategories. Overall, it appears clear that


682



ownloaded from


ATHEROSCLEROSIS

Table 3MEN-AgeGroup 30-39yrs. DeNoo Ischemic Heart Disease (32 Cases)n Age *35.8 yrs. VS.

12yc Followup Base Population (687 Subjects)

DeNovo Base * SignificanceMeasure L H.D. Population Difference Test

Sf 0-12 390.2 341.9±83.0 48.3 p=0.001Sf 12-20 75.4 62.1 ± 27.2 13.3 p =0.01Sf 20-100 139.7 102.3±50.2 37.4 p 0.00IS 100-400 145.8 77.3±83.6 68.5 p 0.00I

A. I 102.1 76.6±27.3 25.5 p.0.00ICholesterol 267.7 222.6±44.2 45.1 p =0.001Systolic B.R 1362 130.3±14.8 5.9 p -0.05Diastolic B.P 91.6 84.6± 10.9 7.0 p =0.001Relative Wt. 119.3 111.9± 15.3 7.4 p <0.0I* MEAN ± STANDARD DEVIATION OF DISTRIBUTION

Table 4mmAge Group40-49y. DeAfvo Ischemic Heart Disease (76 Cases)Mean Age-45.3 y. VS.12 yr. Followup Base Population C 678 Subjects)

ANvo Base SignificanceMeasure I.H.D. Population * Difference Test

S; 0-1 2 385.8 358.8±83.5 27.0 p O0.0IS; 12-20 77.3 67.6±26.7 9.7 p 0.00IS;20-100 119.0 110.2±52.2 8.8 N.S.Sf 100-400 96.1 86.8±97.1 9.3 N.S.

A.. 89.8 82.3±28.2 7.5 p 0.0ICholesterol 249.5 232.8±43.5 16.7 p <0.00ISystolic B. P 139.0 135.1 ±18.1 3.9 0.01 cp c0.05Diastolic B.P 90.4 87.7±1 1.5 2.7 O.01cp<0.05RelativeWt 117.2 I 2.8± 14.8 4.4 p=0.001* MEAN ± STANDARD DEVIATION OF DISTRIBUTION

Table 5MENAgeGGraup505y DeNovo Ischemic Heart Disease (51 Cases)Mean Ago- 52.7 ym. VS.12yc Followup Base Population (353 Subjects)

DeNove Base * SignificanceMeasure LIH.D. Population Difference TTes

S; 0-12 398.2 364.6±86.1 33.6 p 'o.0S; 12-20 84.0 69.6±30.0 14.4 p <O.00IS; 20-100 1 22.1 109.8±46.0 1 2.3 0.01-pcO.05Sf 100-400 85.5 73.9 ±70.5 1 1.6 N. S.

A. L. 88.9 80.9±24.6 8.0 p = 0.01Cholesterol 249.5 233.8±38.9 5.7 p 0.0ISystolic B.P 145.4 138.1 ±21.6 7.3 p <0.0Diastolic B. P 92.7 188.6± 12.4 4.1 O.0l1p-0.05Relative WtI 16.7 114.3 ± 14.1 2.4 N. S.* MEAN ± STANDARD DEVIATION OF DISTRIBUTION

Table 6MEN

Age Group 56-62yrs. DeNovo Ischemic Heart Disease ( 62 Cases)Mean Age * 58.2 yrs. VS.12 yr. Followup Base Population (304 Subjects)

De *Base SignIficanceMeosure I. H. D. Population Difference Test

S, 0-12 375.0 360.8±82.9 14.2 N.S.S, 12-20 69.7 67.6±28.5 2.1 NS.S, 20-100 112.6 107.0±49.3 5.6 N.S.S,f100-400 71.7 64.8±53.3 6.9 N.S.

AI. v 82.0 78.0±22.3 4.0 N.S.Cholesterol 238.2 230.2±40.8 8.0 N.S.Systolic B.P 152.5 1426±24.7 9.9 p 0.001Diastolic RP 92.7 88.0±13.9 4.7 p n0.01Relative Wt 1 13.4 112.0±15.8 1.4 N.S.* MEAN ± STANDARD DEVIATION OF DISTRIBUTION

Table 7WOMEN

Age Group 30-49yrs. DeNo Ischemic Heart Disease ( 19 Cases)Mean Age 44.4yra. VS.12yr. Followup Base Population (1635 Subjects)

DeNo"o Base * SignificanceMeasure IHD. Populatlon Difference Test

S. 0-12 367.1 336.4±80.1 30.7 o.0llp<0.05sf 12-20 69.6 72.1±33.6 -2.5 N. S.S; 20-100 70.4 80.4±38.6 -10.0 N.S.s; 100-400 28.9 3a 1±35.0 -9.2 N.S.

A. I. 66.3 67.1±20.9 -0.8 N.S.Cholesterol 260.4 230.9± 41.8 29.5 p<.oSystolic B.P 141.2 134.8± 17.9 6.4 N.S.Diastolic B.P 86.3 84.9± 11.0 1.4 N.S.Relative Wt. 119.7 115.6±32.4 4.1 N.S.* MEAN ± STANDARD DEVIATION OF DISTRIBUTION

Table 8WOMENAge 6roup 50-55yr5 DeNovo Ischemic Heart Disease (39 Cases)Mean Ag**52.7yrm V.&12yr.Followup Base Population (468 Subjects)

DeN~o Bass * SignIficanceMeasure L.KD. Populatlon DIfference Test

ST 0-12 3867 366.1±79.7 20.6 0.01<p<0.05ST 12-20 10&5 91.6±47.7 14.9 0.0l<p<0.05S; 20-100 119.2 98.3±48.4 20.9 p<0.0S; 100-400 54.0 49.3±42.5 4.7 N.S.

A.I. 87.7 78.6±24.5 9.1 0.0l<p<0.05Cholesterol 274.0 256.3±47.0 17.7 0.01<p<0.05Systolic B.P 163.4 148.0±28.0 15.4 p<-000Diastolic B.P 97.5 90.4±13.4 7.1 p- 0.001Relative Wt. 129.3 1 21.9±21.0 7.4 0.01<p<0.05* MEAN ± STANDARD DEVIATION OF DISTRIBUTION

these prospective studies do not support a

weighting factor of 1.75 for the Sf° 12-400lipoprotein classes in comparison with 1.00for the Sf0 0-12 lipoprotein class.The similar breakdown for women is avail-

able in tables 7 to 9. In both men andwomen the segregation of de novo ischemicheart disease from the base population byany of the lipid or lipoprotein parameters


is weak indeed in the older age groups. Itwill be recalled that the early retrospectivestudies of the 1950's suggested this age-relatedphenomenon; the prospective studies prove

it beyond doubt. Kagan and associates18 havealready called attention to the loss in segre-gation power of the serum cholesterol mea-

sure in women over 50 years of age.

It should be asked whether some feature

683



ownloaded from


GOFMAN ET AL.

Table 9WOMENAg Group 56-69yrs. DeNovo Ischemic Heart Disease (43 Cases)Mean Ag* 58.7 yrs. VS.12 yr. Followup Base Population (384 Subjects)

DNOwvo Bas * SignificanceMeasure I.KD. Population Difference Test

Sf 0-12 381.8 374.1 ±75.8 7.7 N.S.Sf 12-20 90.9 97.8±48.9 -6.9 N.S.Sf20-100 114.7 107.3±52.1 7.4 N.S.S'f100-400 7 1.5 56.1 ±45.9 5.4 0.0 lp<0.05

A I. 86.6 83.2 ±24.2 3.4 N.S.Cholesterol 259.1 261.4±45.4 -2.3 N. S.Systolic B. P 166.4 155.2 t27.5 1.2 p <0.0l1DiasiolicaP 94.9 91.8t 13.9 3.1 N.SRelative Wt 124.6 122.2+21.9 2.4 N.S.* MEAN ± STi NDARD DEVIATION OF DISTRIBtU ION

Table 10MENAgeGroup 20-66yrs. DeNovo Ischemic Heart Disease (38 Cases)MeanAge 43.7yrs. VS.

LIVERMORE Base Population (1961 Subjects)

DeNovo Base * SignificanceMeasure INHD. Population Difference Tast

Sf 0-12 429.3 373.6±81.1 55.7 p'0.00ISf 12-20 71.4 54.7±23.3 16.7 p <0.00ISf 20-100 127.5 101.3±60.0 26.2 p 0.01Sf 100-400 69.5 59.5±77.1 10.0 N.S.

A. I. 90.0 74.8 ±26.7 15.2 p O.001Cholesterol 266.7 235.5 ±43.7 31.2 p 0.001Systolic B.P 129.6 123.4t 13.5 6.2 p<0.0lDiastolic B. P 75.6 73.2± 9.0 2.4 N. S.Relative Wt. 108.8 106.3± 14.1 2.5 N.S.* MEAN ± STANDARD DEVIATION OF DISTRIBUTION

of this study itself accounts for the decreasingpower of segregation of ischemic heart diseasethrough blood lipid parameters with increas-ing age. The possibility that a 12-year follow-up period might have washed out segregationwas checked by dividing the entire studyinto two 6-year follow-up periods. No appre-ciable difference in segregation upon any

lipid or lipoprotein parameter was apparentfor the first 6 years versus the second 6years. The possibility that combined biologi-cal plus technical error might result in lossof segregation power was evaluated by con-

sideration of individuals who had been studiedtwice prior to the occurrence of de novo

ischemic heart disease. The mean value ofeach parameter was used in this evaluation.While the base population available for thisevaluation was much smaller than the overallgroup, it is clear that the multiply sampledgroup does not demonstrate appreciably im-proved segregation of de novo ischemicheart disease from the base population.

Lastly, could the loss of segregation power

of blood lipid parameters with increasing agebe accounted for by some characteristic ( s)peculiar to the Framingham population sam-

ple? Here the follow-up results of the Liver-more study are pertinent. The comparison ofde novo ischemic heart disease and basepopulation is presented in table 10. The over-

all lipid differences, in men, for I.H.D. versus

base population appear larger for the Liver-more sample than for the Framingham group.

However, the mean age at entry to studyis 43.7 years for the I.H.D. group at Livermore

versus 49.2 years at Framingham, so that thelarger differences in the Livermore group areconsistent with the general trends observedwithin the Framingham study. Thus, theLivermore results are entirely consistent withthe Framingham results. This similarity speaksagainst any peculiarity of either sample beingresponsible for the age trend in segregationby blood lipid parameters. Presented in tables11 to 13 are the strengths of the lipid param-eter segregation as a function of age forthree variables, Sf° 0-12, A.I., and serumcholesterol for the Framingham and Livermorestudies. The similarities are striking, bothwith respect to absolute values and changesin relation to age. Beyond age 55 years, noneof the data demonstrate any significant segre-gation power for blood lipid parameters.

Sf 20-100, Sf 100-400, High-Density Lipoproteins,Pre-f-Lipoprotein, and Triglycerides in Relation toSegregation of Ischemic Heart Disease

As mentioned earlier, a major objective ofall such endeavors is to determine for thehuman lipoprotein spectrum the lower andupper limits with respect to association withischemic heart disease. Ultimately, it is notenough to prove association of a particularlipoprotein class with the disease; rather, itis necessary to know whether an independentcontribution to the association is made bythat particular lipoprotein class. In the Fram-ingham data, the Sf° 100-400 class is signifi-cantly elevated in the 30-39 year old groupwith de novo ischemic heart disease versusthe base population. In no other age categorystudied could this class be shown to be


684



ownloaded from


ATHEROSCLEROSIS

MEN

Table 11

AGE vs. Lipid Parameter Segregation OfDeNovo Ischemic Heart Disease.. ..

A.IL Cholesterol 0-120 A H U S A H U 0 A a

Age Group byrsd 30-3940-49 50-55 55-69 30-39 40-49 50-55 5-69 30-39 40-4 50-55 55-69

Mean Age (yrs) 35.7 45.3 52.7 58.2 35.7 45.3 52.7 58.2 35.7 45.3 52.7 58.2

Ischemic Heart Disease 102.1 89.8 88.9 82.0 267.7 249.5 249.5 238.2 390.2 385.8 3982 375.0

Base Population 76.6 82.3 80.9 78.0 22.6 232.8 233.8 2302 341.9 358.8 364.6 360.8

I. H. D.- Base 25.5 7.5 8.0 4.0 45.1 16.7 15.7 8.0 48.3 27.0 33.6 14.2

BASED UPON:DeNovo I. H. D.

* = 32 CASES ARISING IN 687 SUBJECTS* = 76 CASES ARISING IN 678 SUBJECTSA SI CASES ARISING IN 353 SUBJECTSi = 62 CASES ARISING IN 304 SUBJECTS

Table 12

4WOMEN AGE vs. Lipid Parameter Segregation OfDeNovo Ischemic Heart Disease.. ..

A. I. Cholesterol S 0-120A s 0 A 0 ~~~~0 A

Age Group (yrs) 30-49 50-55 56-69 30-49 50-55 56-69 30-49 50-55 56-69Mean Age (yrs) 44.4 52.7 58.7 44.4 52.7 58.7 44.4 52.7 58.7

Ischemic Heart Disease 66.3 87.7 86.6 260.4 274.0 259.1 367.1 386.7 381.8

Base Population 67.1 78.6 83.2 230.9 256.3 261.4 336.4 366.1 374.1

LK D.- Base -0.8 9.1 3.4 29.5 17.7 -2.3 30.7 20.6 7.7

BASED UPON:DeNovo I.H.D.

* a 19 CASES ARISING OUT OF 1635 SUBJECTS* a 39 CASES ARISING OUT OF 468 SUBJECTSA - 43 CASES ARISING OUT OF 384 SUBJECTS

significantly elevated. The data for the 30-39year old Framingham group are not goodenough to resolve the issue of independentcontribution from Sf° 100-400 lipoproteins.Circulation, Volume XXXIV, October 1966

The observed elevation could be due inpart, or even in whole, to interclass correla-tions. The weakness of segregation power forSf° 100-400 in all other groups suggests that

685



ownloaded from


GOFMAN ET AL.

Table 13

MEN(Livermore)lOyr. Folowup

Ischemic Heart Disease

Base Population

I.H. Il-Base

AGE vs. Lipid Parameter Segregation OfDeNovo Ischemic Heart Disease..

A.I. Cholesterol S 0-12

21 CASES LESS THAN 45 YRS. (MEAN AGE= 38.0 YRS.)17 CASES 45 YRS. OR OLDER (MEAN AGE= 50.8 YRS.)

the contribution must be lower than that forthe other classes. Precise establishment of theupper limit of the lipoprotein spectrum forassociation with ischemic heart disease is notpossible. While segregation power appearsto drop above Sf° 100, some importance can-not be ruled out for lipoproteins of Sf° > 400.

On the other side of the lipoprotein spec-trum are the high-density lipoproteins HDL2and HDL3 and a class known as HDL1,lying between the low- and high-density lipo-proteins. No measurements of these classeswere available for the Framingham populationsample. In the Livermore study, all three

Table 14MENAge Group 20-66yrs.Mean Age=43.7 yrs.10 yr. FollowupLIVERMORE

DeNovo Ischemic Heart Disease (38 Cases)vs.

Base Population (1961 Subjects)(HIGH DENSITY LIPOPROTEINS)

DeNovo BaseLH.D. Population * Difference

SignificonceTest

H. D. LI 25.6 25.0 ±16.8 0.6 N. S.

H. D. L2 25.8 38.0 ±29.1 -12.2 p= 0.0I

H. D. L3 204.8 223.6 ±47.8 -18.8 pWO.02

* MEAN ± STANDARD DEVIATION OF DISTRIBUTIONCirculation, Volume XXXIV, October 1966

LS THAN 4 LESS THAN rESSTHAN 43 +YRS4453 +YRS 43 S. 43+ YRS. 43 YR&S 1_5+YS

96.7 81.6_ 280.5 249.6 451.0 402.573.0 76.8 231.7 239.4 365.0 381.7-23.7 4.8 48.8 10.2 86.0 20.8

Measure

686



ownloaded from


ATHEROSCLEROSIS

classes, HDLj, HDL2, and HDL3 were mea-

sured for all subjects. The results for de novo

ischemic heart disease and base populationsample are presented in table 14. It is evidentthat no segregation of ischemic heart diseasecan be proved for HDL, lipoproteins. BothHDL2 and HDL3 classes are significantlylower in the cases of de novo ischemic heartdisease than in the base population. It isknown that levels of both HDL2 and HDL3are inversely correlated with levels of thelow-density lipoproteins, Sf0 0-400.11' 20 Fromthese data it is not possible to concludewhether or not the observed lowerings ofHDL2 and HDL3 in ischemic heart diseaseare in excess of those anticipated from theinverse correlations. This, again, would ulti-mately be desirable information, since if thereis any lowering beyond that expected frominterclass correlations, the possibility of a pro-

tective role of high-density lipoproteins wouldrequire consideration. The lowering of HDL2and HDL3 lipoprotein classes is consistentwith reports from Barr's group2l' 22 concerningchemically measured a-lipoproteins in retro-spective studies of myocardial infarction.

Recently Albrink and associates'6 publisheddata on serum triglyceride in a small seriesof 50-59 year old male survivors of myo-

cardial infarction. They concluded that serum

triglyceride level shows powerful segrega-

tion of ischemic heart disease vs. base popu-

lation for the age group above 50 years andsuggest further that poor segregation in thisage group is the result of measurement ofless consequential blood lipid parameters.Their study of myocardial infarction survivorsis fraught with all the difficulties describedabove for the choice of an appropriate basepopulation for comparison with the infarc-tion survivors. They speculated that the tri-glyceride elevation is probably due to eleva-tion in the very low-density, glyceride-richSf0 20-400 lipoproteins. Our data are clearlyat variance with the conclusions of Albrinkand associates, since the segregation power

beyond 50 years of age, while poor in generalfor all blood lipid parameters, seems especial-ly poor for the triglyceride-rich Sf'0 20-400


lipoproteins. Carlson,23 who also studied myo-cardial infarction survivors, concluded thatserum triglyceride segregation is poor in pa-tients beyond 50 years of age. Fredrickson andLees,24 quoting Besterman's data25 on infarc-tion survivors, remarked on the powerful seg-regation of myocardial infarction survivorsmeasured electrophoretically by pre-,-lipo-proteins, which are known to be essentiallyequivalent to Sf0 20-400 lipoproteins.The only prospective study of this question

is that reported herein for Sf0 20-400 lipo-proteins in the Framingham and Livermorepopulations. Since segregation power for suchlipoproteins is poor for persons beyond 50years of age, it is likely that, when serumtriglyceride and pre-pt-lipoprotein measuresare subjected to the careful scrutiny of aprospective study, they also will prove equallypoor for such age groups.Estimation of Ischemic Heart Disease Risk fromBlood Lipid Parameters

The broad outlines of what may be expectedin predictive power of lipoproteins of anyclass from Sf0 0 to Sf0 400, for A.I., or forserum cholesterol is evident from the behaviorof the differences in means (I.H.D. vs. basepopulation), for the several age groups underconsideration. Predictive power for de novoischemic heart disease should be appreciablein the young age group and should deterioratewith increasing age. Tables 15 and 16, whichprovide incidence rate of ischemic heart dis-ease for the population ranked in deciles,demonstrate precisely this trend for A.I. andserum cholesterol in men in the Framinghamstudy. It is evident that above 55 years thereis essentially no predictive power. The agetrend is not clear cut for women within thesedata, but predictive power for women more

than 55 years of age does not exist. Sinceit appears unlikely that a variation in techniqueof measurement of blood lipoproteins or lipidswill materially alter this situation, it is ap-propriate to consider certain major implica-tions of these findings.

If, for any blood lipid parameter, the differ-ence in value between I.H.D. and the base

687



ownloaded from


Table 15

A.I. vs. DeNAovo Ischemic Heart Disease Incidence Rate

MEN

DECILE F46-o- -44-21 1/68 2/67 3/35 4/302 0/68 3/67 4/35 3/303 1/68 7/67 2/35 10/304 3/68 7/67 2/35 9/305 1/68 6/67 3/35 4/306 2/68 5/67 8/35 5/307 5/68 13/67 9/35 6/308 3/68 10/67 4/35 2/309 4/68 15/67 7/35 11/3010 12/75 8/75 9/38 8/34

WOMENYEAS

DECILE 3 040 --~~0 - -400/79 3/83 5/46 2/38

2 0/79 0/83 1/46 6/383 0/79 1/83 3/46 2/384 1/79 0/83 3/46 4/385 1/79 3/83 2/46 3/386 0/79 0/83 2/46 4/387 0/79 2/83 4/46 8/388 I1/79 2/83 7/46 2/389 0/79 4/83 5/46 6/3810 0/82 1/90 7/54 6/42

Table 16

Serum Cholesterol vs. DeNovo Ischemic Heart Disease Incidence Rate

MENYEARS

DECIL 30 40- 50456-62I 0/68 4/67 1/35 3/302 0/60 5/67 2/35 8/303 1/68 6/67 6/35 7/304 2/68 5/67 2/35 7/305 1/68 7/67 5/35 5/306 2/68 10/67 6/35 4/307 3/68 8/67 6/35 6/308 8/68 9/67 5/35 8/309 5/68 10/67 7/35 5/3010 10/75 12/75 11/38 9/34

WOMENYEARtS

DECILE l YE30404 56-S1 0/79 2/83 5/46 7/382 0/79 2/83 2/46 4/383 0/79 0/83 2/46 3/384 1/79 0/83 2/46 6/385 0/79 0/83 6/46 3/386 2/79 2/83 3/46 2/387 0/79 1/83 5/46 1/388 0/79 1/83 2/46 5/389 0/79 2/83 5/46 5/38Io 0/79 6/90 7/54 7/42

population approaches zero, and if the distri-bution of values of that parameter about thecentral value is similar in the de novo I.H.D.group and the base population, then it followsthat the risk of future ischemic heart diseasein such a population is independent of theblood lipid parameter, even for widely differing

values of the parameter. It follows, further,that under such circumstances, there wouldexist no rationale for the expectation thatdietary or pharmacological alteration of suchblood lipid parameters might alter the riskof de novo ischemic heart disease in such apopulation.


GOFMAN ET AL.688



ownloaded from


ATHEROSCLEROSIS

Relationship of Serum Lipids andLipoproteins to Degree of

Arterial NarrowingThe study of atherosclerosis in autopsy

material from subjects whose blood lipid andlipoprotein parameters had been determinedduring life represents the second major avenueof approach. Such a prospective study hasbeen carried out by Paterson and his co-workers.26 27 Ideally, such studies should coverthe same age ranges (30-69 years) as do theprospective studies of the development of denovo clinical ischemic heart disease. Thedifficulties attendant upon such a study, espe-cially for the younger age groups, are self-evident. Hence, the youngest group availablein the Paterson study is the 60-69 year agecategory. In spite of the limitations in agerange, the findings of that study are instructive.Multiple determinations of low-density lipo-proteins, A.I. values, and serum cholesterolwere performed in a population sample ofinstitutionalized psychotic individuals. Thechoice of a psychotic population sample forsuch a study may be questioned. No knownfactors about psychotic subjects would leadone to expect a distortion of relationshipbetween blood lipid parameters and athero-sclerosis. Nevertheless, generalization of re-sults to nonpsychotic population samples issubject to reservations on this issue. Forthose for which autopsy was available, thelipid parameters during life were comparedwith degree of atherosclerosis of coronaryarteries measured at autopsy. Since blood lipidparameters were measured on a regular basisduring life, the values are, for many of thecases, much more reliable as indices of bloodlipid status than a single determination wouldbe. The grading system for degree of athero-sclerosis was certainly not of the quantitativetype reported by Young and associates.28However, it is doubtful that this materiallyperturbed the central findings that emergedfrom Paterson's study. Essentially it wasdemonstrated by Paterson and associates, thatthe degree of coronary artery atherosclerosiswas unrelated, or at best very weakly related,to the blood levels during life of serum choles-Circulation, Volume XXXIV, October 1966

terol, A.I., or individual lipoprotein classesfor the 60-69 year group of subjects. Thereis no reason to question the validity of theobservations. There is excellent reason toquestion their interpretation of these findings.

Paterson and associates interpreted the ab-sence of relationship between blood lipid pa-rameters and degree of coronary sclerosis assounding the death knell for the blood lipidtheory of atherogenesis. If these were theonly existing data relevant to the problem,it would be fair to agree that, at least for60-69 year old men (assuming his populationsample to be representative), blood lipidparameters are unrelated to degree of athero-sclerosis in the coronary arteries. Since nostudies are available for men in the youngerage categories, the Paterson data allow forno conclusions concerning these age groups.The Paterson group has stated27 that thispoint is not relevant, as follows: "We do notconcede, of course, that a man in his sixties isvery old." From considerations to be elaborat-ed herewith, it may well be concluded that60-69 years of age is indeed very old forcoronary atherogenesis.

It is of prime importance at this point toconsider how the findings of the Patersongroup concerning the degree of coronaryatherosclerosis at autopsy compare with theconclusions reached from the study of de novoischemic heart disease in the Framingham andLivermore population samples. The Patersondata are restricted to the 60-69 year agecategory, so the appropriate comparative evi-dence concerning clinical ischemic heart dis-ease is for this same age group.The basic premise (vide stupra) is that blood

lipid parameters derive predictive ability forde novo ischemic heart disease indirectly viarelationship of blood lipid parameters withcoronary atherosclerosis. Since no significantdifference could be proved to exist betweende novo I.H.D. and base population for sub-jects over 55 years of age, the conclusion wouldbe drawn that blood lipid parameters areunrelated to coronary atherosclerosis in sub-jects above 55 years of age. Thus, the Paterson

689



ownloaded from


GOFMAN ET AL.

autopsy findings are precisely those to be ex-

pected from the study of de novo ischemicheart disease in the Framingham and Liver-more groups. Unfortunately, but for excellentreasons, the Paterson studies provide no evi-dence concerning the relationship of bloodlipid parameters to coronary atherosclerosis inthe 30-39 year age group-a group in whichrisk of de novo ischemic heart disease isstrongly related to blood lipid parameters.

Interpretation of the Studies ofde novo Ischemic Heart Disease and

Coronary AtherosclerosisNone of the observational material presented

so far gives any conflicting evidence or realparadoxes. Some integration of all the findingsis, however, highly desirable. The central ques-

tion is, "Why should blood lipid parametersbe related to risk of ischemic heart disease(and presumably with coronary atherosclero-sis) at approximately 35 years of age, andlose such relationship beyond 55 years of age?"At the outset it can be stated that there isno a priori reason to expect that a relationshipbeween blood lipid parameters and coronary

atherosclerosis should be equally strong at allages or that it should change in any pre-

scribed manner. An understanding of theage-dependence of such relationship dependsupon an insight into the process of athero-genesis. There are ancillary items of evidenceconcerning atherosclerosis that are relevant tosuch insight. Several years ago Young andassociates2831 conducted an extensive quanti-tative study of human coronary and cerebralintimal sclerosis. In those studies histologicalcross-sections of arteries were quantitativelydescribed with respect to degree of sclerosis.The area (in cross-section) of tissue internalto the internal elastic lamella was determinedby planimetry and designated as I. The totalarterial cross-section area was designated as

E. Intimal sclerosis was defined as I/E, thearea of intimal tissue per unit cross-sectionof artery. No differentiation was made be-tween lipids and any other space-occupyingmaterial in the intima. While some observersmay debate assignment of all intimal materialas sclerotic material, there is a reasonable

Table 17Correlations of Degree of Atherosclerosis Withinthe Cerebral Artery Bed

Correlationcoefficient

(Pearson r)

Middle cerebral artery vs.average of all cerebral branches

Posterior cerebral artery vs.average of all cerebral branches

Carotid and basilar arteries vs.average of all cerebral branches

Anterior cerebral artery vs.average of all cerebral branches

+ 0.96

+ 0.95

+ 0.92

+ 0.89

basis for regarding the quantity of intimalmaterial in a "normal" artery to be negligible.In any event, this particular issue can bebypassed, since the operational definition ofI/E is clear.

Several important generalizations emergedfrom those studies: (1) There was an ex-tremely high degree of correlation of I/Evalues for any particular cerebral artery withI/E values for all other cerebral arteries(table 17). This evidence suggests stronglythat a general factor is implicated in thepathogenesis of the human cerebral arteriallesion. Blood lipid level could be such ageneral factor, since the same blood bathesall regions of the cerebral artery bed. (2)There was a moderately high correlation ofI/E values between certain parts of the coro-nary arterial bed and a relatively poor correla-tion of I/E values between other parts (table18). No general factor, such as blood lipid

Table 18Correlations of Degree of Atherosclerosis Withinthe Coronary Artery Bed

Correlationcoefficient

(Pearson r)

Left circumflex branch vs.average of all coronary branches + 0.82

Anterior descending branch vs.average of all coronary branches + 0.80

Right coronary artery vs.average of all coronary branches + 0.74

Left anterior descending branch vs.average of all coronary branches + 0.60

Left main coronary branch vs.average of all coronary branches + 0.44


690



ownloaded from


ATHEROSCLEROSIS

level, can explain this finding. Therefore,these latter data bespeak the operation of astrong focal factor in determination of degreeof sclerosis, varying appreciably in differentregions of the coronary bed of a single indi-vidual. (3) Of especial consequence in thepresent considerations was the demonstration,both in the coronary and cerebral arterialbeds, that I/E values are strongly related toarterial radius (see tables 19 and 20). Inthe cerebral arteries, the I/E value wasproportional to arterial radius over the rangeof radii existing in that bed. In the coronaryarteries, where the average degree of sclerosisin I/E units is higher than that for the cere-bral arteries at any particular radius, the linearrelationship between I/E and arterial radiusexisted for radii less than 2.0 mm. Above thisradius value, the I/E value did not appearto change with radius at least up to 2.5 mm.Certainly no general factor, such as bloodlipid level, can explain this radius effect.

Table 19Relationship Between Cerebral Atherosclerosisand Radius of Cerebral Arteries

No. ofsections

37151

518354

2

Range of radius(mm)

2.1-2.51.6-2.01.1-1.50.6-1.00.2-0.5

Mean radius(mm)

2.2461.7001.2140.8670.470

(I/E)b

58.549.032.019.818.0

Note: (I/E )b = atherosclerosis of the brain, whereI = intimal material area; E = total arterial cross-section area; and b = brain.

Table 20Relationship Between Coronary Atherosclerosisand the Radius of the Coronary Arteries

No. of Range of radius Mean radiussections (mm) (mm) (I/E)c

48 2.6-3.0 2.672 58.2152 2.1-2.5 2.206 58.5212 1.6-2.0 1.737 55.0228 1.0-1.5 1.253 43.8118 0.5-1.0 0.798 28.1

Note: (I/E ) c = coronary atherosclerosis, whereI = intima material area; E = total arterial cross-section area; and c = coronary artery.Circulation, Volume XXXIV, October 1966

At this juncture it is not necessary to under-stand how arterial radius comes to be related tointimal sclerosis expressed in I/E units. Butthe existence of the relationship in arteriesof 60-89 year old men is not disputable. Youngand associates29 suggested that the radiusrelation might operate to decelerate the rateof development of new atherosclerosis as theradius of any particular artery is reducedthrough accumulation of intimal material.Construction of a Model of AtherogenesisLet us now consider a simple model of the

development of cerebral and coronary intimalsclerosis utilizing the above findings. (1)Assume that atherosclerosis proceeds by con-centric reduction of arterial radius. (Actuallyeccentricity is more common in advanceddisease, but we may reserve this considerationas possibly a second-order correction to themodel.) (2) Assume that some blood lipidparameter operates as a general factor in allarteries developing sclerosis. (3) Assume thatfor each artery segment the focal factor canbe expressed as a single value.

Thus,let r = radius of lumen of artery at time t,let accumulation of sclerotic intimal ma-

terial be described in terms of radius reduc-tion, -dr,

let k = the focal factor in any particularartery (differing within and between indi-vidual subjects), and

let c =the value of the pertinent bloodlipid parameter (in this first approximationthe same blood lipid parameter will beassumed to operate for all arterial beds).

Now, explore first the simplest possible de-pendence of accumulation of sclerosis uponk, c, and r. Such a dependence is writtenfor any particular arterial segment as

-dr = (k) (c) (r)dt or - dr = (k) (c)dt. (1)r

Integrating,

ln r = - kct, assuming r = r. at some ar-ro bitrary time taken as 0 time

for the processor r = ree-t

691



ownloaded from


GOFMAN ET AL.

At time t, the degree of sclerosis in (I/E)units is expressed as follows:

(I/E)t= 1rr2 r r2

Substituting for r, we have

r,,2 -r2e-2kct(I/E)t 2

and

(I/E)t 1- e-2kctBut this result is clearly at variance with theobservation which shows I/E values to berelated to initial arterial radius. Therefore,this simple relationship (equation 1) cannotbe correct.

Let us next assume dependence upon r'rather than r, writing

dr (k)(c)(r2)dt, or- drr2 (k) (c)dt. (2)

Integrating,

r(,I_

1,r

r

-kct, with r ro at some arbitrarytime taken as 0 for this process.

+ kctrO

rOI + ro kct

( -I(r02 r2)(L/E)t ~ and substituting for r,

(I/E)t-

(I/E)t

rO2r22I- ±r0 kct)

(1+ ro ket) )

It is evident that in this relationship, (I/E)tis a function of initial radius r0, and thatfor particular values of k, c, and t, the degreeof sclerosis, (I/E )t increases with increasein r,. Also, the boundary conditions are satis-fied, since at t=0, (I/E)0r0 and as t be-comes very large, (I/E)t approaches 1, whichexpresses complete luminal obliteration by in-timal sclerosis. I/E can never become greaterthan 1.0.

Validity of the Model of Atherogenesis

We may now explore whether this simplerelationship fits: (1) the de novo ischemicheart disease findings of Framingham andLivermore, (2) the Paterson findings con-cerning blood lipids and coronary sclerosisin 60-69 year old men, and (3) the observa-tions of Young and associates relating arterialradius to I/E values.As a first approximation, this test was done

with the following simplifications:Let c = 5 represent the median value of

the blood lipid parameter and assume no agedependence for the blood lipid parameteritself.

Let the process start, arbitrarily, at age 25years. After preliminary trials of appropriatek values, a value of k 0.01 was selected,and I/E values were calculated as a functionof c, t, and ro, using the integrated formof equation 2. In figure 1 are presented thecalculations for I/E in an artery of r0 =2.0mm, as a function of blood lipid parameterand age. In figure 2 are presented the calcu-lations for I/E as a function of initial arterialradius and age, at the median value for theblood lipid parameter, c 5. Considerationof these figures allows exploration of wherethis simple model fits reality and where itdoes not.

1. From figure 1 it is noted that in earlyyears after inception of the process, the degreeof sclerosis differs appreciably with change invalue of c, the blood lipid parameter. Pro-gressively with age, the degree of sclerosis,

00

x

0

0

cc

w-j

(0

z

25 30 35 40 45 50 55 60 65 70 7'AGE (yrs.)

Figure 1


1r

692



ownloaded from


ATHEROSCLEROSIS

O AIMLtKUbtNLtbI1

ATH00E RtfOGENE dSIS 0m90x rr= kcrdt

U k . 0.01

9)0

Z1 20 -ter odius -0

Cco

25 30 35 40 45 50 55 60 65 70 75AGE (yrs.)

Figure 2

in I/E units, differs less and less over awide range of c values. At age 30 years,I/E=0.55 for c=5, I/E=0.75 for c=10.Thus for a twofold increase in blood lipidparameter, I/E increases by 36%. In contrast, atage 60 years, I/E = 0.94 for c = 5, I/E = 0.97for c = 10. Thus at age 60 years, I/E increasesby 3% for a twofold increase in blood lipidparameter. Thus, early in this idealized de-velopment of atherosclerosis, degree of athero-sclerosis is strongly related to blood lipidparameter. If clinical incidence of ischemicheart disease is related to degree of athero-sclerosis, then such incidence should showstrong dependence upon blood lipid parameter.This was observed in the Framingham andLivermore studies. Late in the idealizeddevelopment of atherosclerosis, degree ofatherosclerosis is only minimally related toblood lipid parameter. Hence, on the samegrounds as above, clinical incidence shouldshow very little relationship with blood lipidparameter, even over a very large rangeof blood lipid parameter values. This also wasobserved in the Framingham and Livermorestudies.

2. Late in the idealized development ofatherosclerosis, such as at age 65 years, thedegree of sclerosis is calculated to be triviallydifferent over a wide range of blood lipidparameter values. This is precisely what thePaterson group found-namely, that the degreeof coronary atherosclerosis was not significant-ly related to blood lipid parameters.

3. Lastly, this idealized development of

atherosclerosis leads to a prediction of in-creasing I/E values at any time with in-creasing initial radius of artery (see figure 2),as observed by Young and associates.A major disagreement with reality in figure

1 is that the predicted degree of coronarysclerosis (I/E) approaches values of 0.9 to1.00, which is at variance with the observa-tions of Young and associates. Lowering ofthe value chosen for k will improve thisaspect of the model, but will worsen the agree-ment of the calculated values both with thedata on incidence of ischemic heart diseaseand with the Paterson studies of degree ofsclerosis in relation to blood lipid parameters.However, if k itself is allowed to decreaseas appreciable atherosclerosis develops, theonly major disagreement with reality is re-moved. Examination of atherosclerosis in au-topsy material reveals in general that newsclerosis is not the rule over the "fibrous cap"commonly described in moderately advancedlesions. This observation would be consistentwith a decrease in k as atherosclerosis becomesappreciable, over and above the radius effectitself.A simple function expressing the decline in

k with appreciable development of atheroscle-rosis is the following:

k = ko e-m(I/E)2 (3)The choice of m = 6.93 allows k to become= 0.5 k. for (I/E )2 = 0.10, that is, where(I/E) = 0.32.

Utilizing this variation in k, with k. = 0.005,calculations similar to those described abovewere carried through with the same integratedform of equation 2 as before for variationin I/E with age, blood lipid parameter, andinitial arterial radius. These are presented infigures 3 and 4. By choosing a particular age,the variation of I/E with initial arterial radiuscan be replotted from the calculations offigure 4. This was done for early, intermediate,and late periods in figure 5, in which thevariation in I/E as a function of initialarterial radius alone is presented at early,intermediate, and late periods. Superimposedon this last plot are the observational data


693



ownloaded from


GOFMAN ET AL.

0-

0a:

o

Jco.Ni

Figure 3

0

0

0wo80

J 60

E 40

z 20

0

0

hitoa

o loo

800

a60a 402-

Z. 201

0'U -- ---25 30 35 40 45 50 55 60 65AGE (yr..)

Figure 4

70 75

I I I I I-

VARIATION OF INTIMAL SCLEROSIS withARTERIAL RADIUS

Predicted by Model of AtherogenesisExperimentol Observations of Young, etal(Coronary Arteries)

LATE DISEASE

jTEMMM%TEIN,EARLY IN DEVELOPMENT

Ivl l l la-O.s 1.0 1.5 2-0 2-5 3.0 3.5 4.0

ARTERIAL RADIUS,mmFigure 5

of Young and associates for I/E in relationto arterial radius in coronary arteries of 60-89year old men.

The calculational data presented in figures3 to 5 are in harmony with (1) the Framing-ham and Livermore data concerning the de-creasing association of blood lipid parameter

with risk of de novo ischemic heart diseaseas a function of age; (2) the lack of relation-ship between degree of coronary atheroscle-rosis and blood lipid parameter in the 60-69year age bracket-as already observed byPaterson; (3) the approximate extent of thecoronary (I/E) values to be expected in the60-89 year age bracket observed by Youngand associates, and (4) the positive relation-ship between I/E and arterial radius ob-served by Young and co-workers, includingthe nearly linear relationship with radius formoderate disease, and a flattening out at radiiabove 2.0 mm with advanced disease.

Implications

A first-approximation description of the de-velopment of atherosclerosis has been presentedin relationship to a general parameter (bloodlipid level), a focal parameter (the value ofk), arterial radius, and age. One may legiti-mately ask what mechanism of atherogenesisunderlies the very simple equations employedand what assurance there is that the equationsand the description are correct. The mechanismof atherogenesis remains unknown, so it can

be unequivocally stated that the descriptionand equations do no violence to well-estab-lished mechanism concepts. The correctnessof a concept such as we have presented isbest tested against reality in the clinical sense.

Several major relationships having reality inthe clinical sense have been shown herein tobe consistent with the concepts and the equa-

tions. The crucial additional tests lie in theprediction of additional phenomena concern-

ing atherogenesis and in a comparison of pre-

diction with reality. We may now considerwhat some of the additional phenomena are

that would be predicted. For some, data are

already available. For others, as data develop,the validity or lack of validity of the pre-

dictions will become evident.Prediction 1. Blood lipid parameters lose

predictive value for de novo ischemic heartdisease in the neighborhood of 55 years ofage, not because of the improper choice ofparameters, but rather because of the natureof the evolution of the atherosclerotic process.


I I I V I I I I

PREDICTED DEVELOPMENT OF INTIMAL SCLEROSIS v.

AGE AND ARTERIAL RADIUS (Bosed Upon Model OfAtherogenoses)

-drr2 =kcdtk = ko e-m(1/E)2

RodiRa iu =3 ( "' mm\=R~~~~~~~~~~.Atrotf Rcdius

694



ownloaded from


ATHEROSCLEROSIS

The loss of predictive value for those param-eters already measured is evident from thedata presented herein. The prodiction is thatsuch parameters as serum triglyceride, pre-,8-lipoproteins, or other lipid parameters willdo no better if subjected to the rigorousscrutiny of a valid prospective study.

Prediction 2. Blood lipids are intimately re-lated to the pathogenesis of atherosclerosis,but on the basis of the nature of the evolutionof the process we can predict: (a) no ap-preciable relationship between degree of scle-rosis and blood lipids when the process isadvanced, and (b) an appreciable relationshipwhen the process is early or moderate.The lack of relationship observed by the

Paterson group between coronary atherosclero-sis and blood lipids was discussed in detail.It may be noted in observational materialof Young and associates that cerebral athero-sclerosis is much less advanced than coronaryatherosclerosis, even in old subjects (60-89years of age). Paterson and associates, intheir most recent report,27 reported that eventhough no relationship existed between coro-nary atherosclerosis and blood lipids in 60-69year old men, a weak but significant relation-ship did exist between blood lipids andcerebral atherosclerosis.

Further evidence on this prediction mustbe sought in (a) a study of younger subjects,which is difficult, or (b) animals. If an ap-propriate experiment were done in animalsdeveloping appreciable coronary atherosclero-sis, it would be predicted that early in thecourse of the disease blood lipid levels wouldbe related to degree of coronary sclerosis, butwith advancing disease the relationship wouldweaken progressively.

Prediction 3. Blood lipid parameters are notlikely to provide any prognostic informationin already established ischemic heart diseasein persons above 50 years of age or probablyeven in younger groups.Roe and associates32 have already tested

this and found it to be the case for subjectswho have experienced myocardial infarction.

Prediction 4. Moderate alteration of bloodlipid parameters by dietary or pharmacologicalCirculation, Volume XXXIV, October 1966

means in subjects beyond 50 years of age canhardly be expected to alter the risk of denovo development of ischemic heart disease.Over a wide range of blood lipid parameters,

the data and concepts presented herein indicateno appreciable difference in risk for subjectswho are over 50 years of age. Thus, it is hardlylikely that alterations of blood lipid parameterswithin this range can produce measurablealteration in risk.

It is essential to qualify this prediction lestit be misunderstood. A dietary regimen or apharmaceutical agent may influence the dis-ease by other means than via effects uponblood lipid parameters. Prediction 4 is neces-sarily silent about effects other than thoserelated to blood lipid parameters.

Prediction 5. If dietary and pharmacologi-cal alteration of blood lipid parameters cannotbe expected to alter the risk in ostensiblyhealthy subjects over 50 years old, there isno reason to expect that such measures willalter the risk for persons with establishedischemic heart disease.Two recent controlled studies, one with

low fat diets33 and the other with corn oilsupplements,34 indicate no suggestion thatsuch measures alter prognosis in establishedischemic heart disease.

Prediction 6. If moderate reduction in bloodlipid parameters does prove to alter the riskof de novo ischemic heart disease in ostensiblyhealthy subjects of young age groups (wellbelow 55 years of age), the effect will belost when the subjects reach 55 years of age,even if the regimen prescribed is adhered tofaithfully. In essence this is a statement thatthe life expectancy at 55 years of age andbeyond will not be materially altered, so faras coronary disease is concerned, by moderatereductions in blood lipid levels within therange of values characteristic of the U.S. popu-lation at present.

Prediction 7. If longevity as limited bycoronary atherosclerosis is to be alteredthrough effects upon blood lipid parameters,a drastic rather than moderate reduction insuch parameters is required. (Note effect ofa value of 1.0 in figure 5.)

695



ownloaded from


GOFMAN ET AL.

At present, we do not know how to provokesuch drastic alteration in blood lipids, nor dowe know whether such drastic lowerings areotherwise consistent with health. A small pro-portion of the U.S. population is spontaneous-ly characterized by extremely low blood lipidlevels and appears to enjoy good health. Suchindividuals deserve intensive study if clinicaluse is ever to be made of drastic loweringof serum lipoproteins or cholesterol.

Prediction 8. The k value, expressing thefocal determinant of development of athero-sclerosis deserves intensive study. At presentwe have essentially no information concern-ing factors that may reduce the value of kor how it comes to have the value it doeshave. It is clear that k differs within segmentsof a single artery, between arteries in a singleindividual, and between individuals in thepopulation for a particular artery. It is notinconceivable that a concerted attack uponunderstanding k may be more rewarding thanefforts to alter blood lipid parameters.

Prediction 9. All the findings discussed here-in pertain to blood lipid parameters as theyexist in free-living population samples studiedin the United States. Specifically, they referto two samples, at Framingham and at Liver-more. It is conceivable that alteration incomposition of blood lipoproteins, as suggestedby Geer and McGill35 might lead to alterationsin development of atherosclerosis even thoughno alteration in absolute lipoprotein levelsoccurred. A priori, a particular alteration maybe possible for better, for worse, or for nochange with respect to atherogenesis. Cer-tainly the burden of proof that any alterationin composition is beneficial rests upon thoseproposing such alteration.

Prediction 10. In the absence of significantprogress in beneficial alteration of k valuesor of ability to produce safe widespreaddrastic lowering of serum lipids, it wouldappear highly desirable to devote major at-tention to the acute determinants of myocar-dial infarction or other clinical manifestationsof coronary heart disease rather than to coro-nary atherogenesis itself.

References1. DUFF, G. L., AND MCMILLAN, G. C.: Inhibition

of experimental cholesterol atherosclerosis byalloxan diabetes in the rabbit. (Abstr.) AmerHeart J 36: 469, 1948.

2. DUFF, G. L., AND MCMILLAN, G. C.: Effect ofalloxan diabetes on experimental cholesterolatherosclerosis in the rabbit. J Exp Med 89:611, 1949.

3. PEDERSEN, K. O.: Ultracentrifugal Studies onSerum and Serum Fractions. Upsala, Almquist& Wiksells, A.B., 1945.

4. GOFMAN, J., LINDGREN, F., AND ELLIOTT, H.:Ultracentrifugal studies of lipoproteins of hu-man serum. J Biol Chem 179: 973, 1949.

5. LiNDREN, F. T., ELLIOTT, H. A., GOFMAN, J. W.,AND STIsowER, B.: Ultracentrifugal composi-tion of normal rabbit serum. J Biol Chem 182:1, 1950.

6. GOFMAN, J. W., ET AL.: The role of lipids andlipoprotein in atherosclerosis. Science 111: 166,1950.

7. PIERCE, F. T., JR.: Relationship of serum lipopro-teins to atherosclerosis in the cholesterol-fedalloxanized rabbit. Circulation 5: 401, 1952.

8. JONES, H., GOFMAN, J., LINDGREN, F., LYON, T.,GRAHAM, D., AND STRISOWER, B.: Lipoproteinsin atherosclerosis. Amer J Med 11: 358, 1951.

9. GOFMAN, J., STRISOWER, B., DE LALLA, O.,TAMPLIN, A., JONEs, H., AND LINDGREN, F.:Index of coronary artery atherogenesis. ModMed 119: June 15, 1953.

10. TAMPLIN, A. R., STRISOWER, B., DE LALLA, O.,GOFMAN, J., AND GLAZIER, F.: Lipoproteins.aging, and coronary artery disease. J Geront9: 403, 1954.

11. GOFMAN, J., ET AL.: Serum lipoprotein transportsystem in health, metabolic disorders, athero-sclerosis, and coronary heart disease. Plasma 2:414, 1954.

12. Technical Group of the Committee on Lipo-proteins and Atherosclerosis: Evaluation ofserum lipoprotein and cholesterol measurementsas predictors of clinical complications ofatherosclerosis, Appendix A. Circulation 14:725, 1954.

13. GOFMAN, J., AND YOUNG, W.: Filtration -conceptof atherosclerosis and serum lipids in the di-agnosis of atherosclerosis. In Atherosclerosisand Its Origin, edited by Maurice Sandler andGeoffrey H. Bourne. New York, AcademicPress, 1963, p. 197.

14. GOFMAN, J. W.: Coronary Heart Disease. Spring-field, Illinois, Charles C Thomas, 1959, p. 50.

15. OLIVER, M. F.: Coronary disease, hypercholes-terolaemia and its treatment. Scot Med J 3:225, 1958.

16. ALBRINK, M. J., MEIGS, J. W., AND MAN, E. G.:

Circulation. Voluine XXXIV. October 1966

696



ownloaded from


ATHEROSCLEROSIS

Serum lipids, hypertension, and coronary

artery disease. Amer J Med 31: 4, 1961.17. DAWBER, T. R., MooRE, F. E., AND MANN, G. V.:

Coronary heart disease in the Framinghamstudy. Amer J Public Health 47: 4, 1957.

18. KAGAN, A., DAWBER, T. R., KANNEL, W. B., AND

REVOTSKIE, N.: The Framingham study: Aprospective study of coronary heart disease.Fed Proc 21: 52, 1962.

19. KAGAN, A., KANNEL, W. B., DAWBER, T. R., AND

REVOTSKIE, N.: The coronary profile. AnnNY Acad Sci 97: 833, 1963.

20. DE LALLA, O., AND GOFMAN, J. W.: On thepopulation distributions of both the low-densitylipoproteins and the high-density lipoproteinsand the serum proteins. In Lipide und Lipo-proteide un Blutplasma, Berlin, Springer-Ver-lag, 1961, p. 138.

21. BARR, D. P., Russ, E. M., AND EDER, H. A.:Influence of estrogens on lipoproteins in athero-sclerosis. Trans Ass Amer Physicians 65: 102,1952.

22. BARR, D. P., Russ, E. M., AND EDER, H. A.:Protein-lipid relationships in human plasma:II. In atherosclerosis and related conditions.Amer J Med 11: 480, 1951.

23. CARLSON, L. A.: Serum lipids in men with myo-

cardial infarction. Acta Med Scand (Stock-holm) 167: 399, 1960.

24. FREDRICKSON, D. S., AND LEES, R. S.: Diseasescharacterized by evidence of abnormal lipidmetabolism: Familial hyperlipoproteinemia. InMetabolic Bases of Inherited Disease, ed. 2,edited by G. B. Stanbury, J. B. Wyngaarden,and D. S. Fredrickson. New York, McGraw-Hill, 1960, p. 429.

25. BESTERMAN, E. M. M.: Lipoproteins in coronaryartery disease. Brit Heart J 19: 503, 1957.

26. PATERSON, J. C., CORNISH, B. R., AND ARM-STRONG, E. C.: Serum lipids in human athero-sclerosis. Circulation 13: 224, 1956.

27. PATERSON, J. C., ARMSTRONG, R., AND ARM-STRONG, E. C.: Serum lipid levels and theseverity of coronary and cerebral atherosclero-sis in adequately nourished men, 60 to 69years of age. Circulation 27: 229, 1963.

28. YOUNG, W., GOFMAN, J. W., MALAMUD, N.,SIMON, A., AND WATERS, E. S. G.: Interrela-tionship between cerebral and coronary athero-sclerosis. Geriatrics 11: 413, 1956.

29. YOUNG, W., GOFMAN, J. W., TANDY, R., MALA-MUD, N., AND WATERS, E. S. G.: Quantitationof atherosclerosis: I. Relationship to arterysize. Amer J Cardiol 6: 288, 1960.

30. YOUNG, W., GOFMAN, J. W., TANDY, R., MALA-MUD, N., and WATERS, E. S. G.: Quantitationof atherosclerosis: II. Quantitative aspects ofthe relationship of blood pressure and athero-sclerosis. Amer J Cardiol 6: 294, 1960.

31. YOUNG, W., GOFMAN, J. W., TANDY, R., MALA-MUD, N., and WATERS, E. S. G.: Quantitationof atherosclerosis: III. Extent of correlation ofdegrees of atherosclerosis within and betweenthe coronary and cerebral vascular beds. AmerJ Cardiol 6: 300, 1960.

32. ROE, R. D., LITTLE, A., AND SHANOFF, H.: Prog-nosis and serum lipids in survivors of myocardi-al infarction. Circulation 22: 800, 1960.

33. Research Committee of the Edgware General,the West Middlesex, and St. George's Hos-pitals, London: Low fat diet in myocardialinfarction: A controlled trial. Lancet 2: 501,1965.

34. RosE, G. A., THOMSON, W. G., AND WILLIAMS,R. T.: Corn oil in treatment of ischemic heartdisease. Brit Med J 1: 1531, 1965.

35. McGILL, H. C., JR.: Comparison of experimental-ly induced animal atherosclerosis with spon-

taneous human atherosclerosis. In ComparativeAtherosclerosis, edited by James C. Roberts,Jr., and Reuben Straus. New York, Hoeber,1965, p. 354.


697



ownloaded from


JOHN W. GOFMAN, WEI YOUNG and ROBERT TANDYIschemic Heart Disease, Atherosclerosis, and Longevity

Print ISSN: 0009-7322. Online ISSN: 1524-4539 Copyright © 1966 American Heart Association, Inc. All rights reserved.

is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231Circulation doi: 10.1161/01.CIR.34.4.679

1966;34:679-697Circulation.

http://circ.ahajournals.org/content/34/4/679.citationlocated on the World Wide Web at:

The online version of this article, along with updated information and services, is

http://circ.ahajournals.org//subscriptions/

is online at: Circulation Information about subscribing to Subscriptions:

http://www.lww.com/reprints Information about reprints can be found online at: Reprints:

document. and Rights Question and Answer

Permissionsthe Web page under Services. Further information about this process is available in thewhich permission is being requested is located, click Request Permissions in the middle column ofClearance Center, not the Editorial Office. Once the online version of the published article for

can be obtained via RightsLink, a service of the CopyrightCirculationoriginally published in Requests for permissions to reproduce figures, tables, or portions of articlesPermissions:



ownloaded from

http://circ.ahajournals.org/content/34/4/679.citation

http://www.ahajournals.org/site/rights/

http://www.ahajournals.org/site/rights/

http://www.lww.com/reprints

http://circ.ahajournals.org//subscriptions/


ischemic heart disease, atherosclerosis, and...

Documents