EPIDEMIOLOGIC STUDIES OF CORONARY HEART DISEASE AND STROKE ... · Stroke, Grant No. HL14783 from the National Heart Casualty Commission. and Lung Institute, the National Heart and

AMERICAN JOURNAL OF EPIDEMIOLOGY Vol. 102, No. 6

Copyright © 1975 by The Johns Hopkins University School of Hygiene and Public Health Printed in U.S.A.

EPIDEMIOLOGIC STUDIES OF CORONARY HEART DISEASE ANDSTROKE IN JAPANESE MEN LIVING IN JAPAN, HAWAII AND

CALIFORNIA: PREVALENCE OF CORONARY AND HYPERTENSIVEHEART DISEASE AND ASSOCIATED RISK FACTORS1

M. G. MARMOT,2 S. L. SYME,2 A. KAGAN,3 H. KATO," J. B. COHEN,2 AND J. BELSKY6

Marmot. M. G. (School of Public Health. U. of California, Berkeley. CA94720). S. L. Syme, A. Kagan, H. Kato. J. B. Cohen and J. Belsky. Epi-demiologic studies of coronary heart disease and stroke in Japanese menliving in Japan, Hawaii and California: Prevalence of coronary and hyper-tensive heart disease and associated risk factors. Am J Epidemiol 102:514-525. 1975.

A study of coronary heart disease (CHD) among Japanese migrantscompared with Japanese living in Japan provided the opportunity to studyfactors possibly responsible for the high rates of CHD in America as com-pared with Japan. Comparable methods were employed in examining 11,900men of Japanese ancestry aged 45—69 living in Japan, Hawaii and Cali-fornia. The age-adjusted prevalence rates for definite CHD as determinedby ECG were: Japan 5.3, Hawaii 5.2 and California 10.8/1000. For definiteplus possible CHD the rates were 25.4, 34.7 and 44.6. The prevalence ofangina pectoris and pain of possible myocardial infarction, determined byquestionnaire, showed a similar gradient. Elevated serum cholesterol showeda Japan-Hawaii-California gradient, but the prevalence of hypertension inJapan was intermediate between the prevalence in Hawaii and the higherprevalence in California. The three geographic locations were compared asto prevalence of CHD at comparable levels of blood pressure and cholesterol.At each blood pressure level and at each cholesterol level, the greater preva-lence of CHD in California persisted. These facts, plus the near universalityof smoking in Japan, suggest that conventional risk factors only partly ex-plain the observed gradient in CHD.

Received for publication September 9, 1974, and in Institutes of Health Special Research Resourcesfinal form June 12, 1975. Grant RR-3.

Abbreviation: CHD, coronary heart disease. "Program in Epidemiology, School of Public1 From the Epidemiology Program, School of Pub- Health, University of California at Berkeley,

lie Health, University of California at Berkeley; 3 Honolulu Heart Study, Honolulu, Hawaii.Honolulu Heart Study, Honolulu, Hawaii; and the 'Department of Epidemiology and Statistics,Atomic Bomb Casualty Commission, Hiroshima and Atomic Bomb Casualty Commission, Hiroshima, andNagasaki, Japan. Hiroshima Branch, Japanese National Institute of

Supported by: Grant No. 5 PO1 NB06818 from the Health, Hiroshima, Japan.National Institute of Neurological Diseases and "Formerly Department of Medicine, Atomic BombStroke, Grant No. HL14783 from the National Heart Casualty Commission.and Lung Institute, the National Heart and Lung The combined articles in this issue on "Epidemi-Institute Intramural Program, and by funds of the ologic Studies of Coronary Heart Disease and StrokeNational Heart and Lung Institute made available to in Japanese Men Living in Japan, Hawaii and Cali-the Atomic Bomb Casualty Commission through the fornia," are available on request from: EpidemiologyAtomic Energy Commission. Research Unit, Dept. of Biomedical and Environmen-

Computing assistance was obtained from the tal Health Sciences, School of Public Health, Univer-Health Sciences Computing Facility, University of sity of California, Berkeley, CA 94720.California at Los Angeles, sponsored by National

514

CHD PREVALENCE IN JAPANESE IN JAPAN AND AMERICA 515

blood pressure; coronary heart disease; epidemiology; glucose; lipids;migrants; mortality

INTRODUCTION

Worth et al. (1) confirmed Gordon'sfinding of a gradient in coronary heartdisease mortality among men of Japaneseancestry, increasing from Japan to Hawaiito California (2, 3). The three cohorts ofJapanese men in Japan, Honolulu and theSan Francisco Bay Area were examined tosee if disease prevalence followed the samegradient.

The present paper reports the prevalenceof coronary and hypertensive heart diseaseas well as the prevalence of high bloodpressure and elevated serum cholesterol,triglyceride and glucose found on examina-tion of the three study groups. Other pa-pers from this study report details of theresearch plan (4), methodology for compar-ison of diet (5), the relationship betweendiet and biochemical measurements (6),summary statistics of the characteristics ofthe three populations (7) and, in thisseries, detailed analyses of blood pressure(8) and biochemical measurements (9).

METHODS

Subjects studied. The data in this reportare for men of Japanese ancestry aged45-69 at time of examination: 2141 menwere studied by the Atomic Bomb Casu-alty Commission in Hiroshima and Naga-saki, Japan; 8006 men were studied inHonolulu, Hawaii; 1844 men were studiedin the San Francisco Bay Area, California.

Additional details regarding the develop-ment of the study samples, age distribu-tions, and response rates are reported inthe Introduction to this series (10).

Examination methods. Standardized in-terviews were used in the three centers. Allsubjects completed a self-administeredquestionnaire detailing various demo-graphic and socioeconomic characteristics.Following this, a medical history question-

naire was completed giving full details ofpast and present health. The medical his-tory questionnaire was administered by anurse-interviewer in Japan and Hawaii andwas self-administered in California. In Ha-waii and California, blood pressure withthe subject seated was taken by a nurse; inJapan it was taken by a physician. Dupli-cate measurements were taken; systolicand phase V diastolic blood pressures wererecorded, and the first measurement used.Details of examination methods are re-ported elsewhere (4).

A 50 gm glucose load was given early inthe clinic procedure and blood drawn onehour later. Subjects were non-fasting. Cali-fornia subjects who reported a history ofdiabetes were not given a glucose load.Serum cholesterol, uric acid and glucosevalues were determined in San Franciscofor the California and Hawaii cohorts andin Hiroshima for the Japanese cohort.Serum triglycerides for all three cohortswere determined in San Francisco. Proce-dures were devised to control variationwithin and between laboratories, and anal-ysis indicated that the methods usedyielded comparable data from the twolaboratories (9).

Conventional 12 lead electrocardiogramswere performed using a Sanborn elec-trocardiograph at 25 mm per second. Cop-ies of all ECG tracings were sent to theLaboratory of Physiological Hygiene inMinnesota for objective ratings accordingto the revised Minnesota code (11). Trac-ings sent to Minnesota were not identifia-ble by area by those reading the ECG's.

Diagnostic criteria. To avoid the possiblenon-comparability of clinical judgmentsamong the three study areas, only stan-dardized procedures were used to assessprevalence.

Angina pectoris and pain of possiblemyocardial infarction were determined by

516 MARMOT, SYME, KAGAN, KATO, COHEN AND BELSKY

means of the London School of Hygiene-Cardiovascular Questionnaire (11). A diag-nosis of angina pectoris was made if thesubject reported pain in the chest in theappropriate location while walking uphillor hurrying and if the pain went away in 10minutes or less when the subject stopped orslowed down. Possible history of myocar-dial infarction was said to exist if thesubject had experienced a severe painacross the front of the chest which lastedfor half an hour or more.

Prevalence of CHD was restricted to thefinding of abnormal Q and QS patterns onECG. Minnesota codes 1-1-1 through 1-1-7constituted definite CHD and 1-2-1through 1-3-6 constituted possible CHD.

World Health Organization criteria wereused to classify hypertension: i.e., hyper-tension represents a systolic blood pressure> 160 mm Hg and/or a diastolic bloodpressure > 95 mm Hg; normotension im-plies both systolic blood pressure < 140mm Hg and the diastolic pressure < 90 mmHg; borderline hypertension represents theresidual category, i.e., the systolic bloodpressure < 160 mm Hg and the diastolicblood pressure < 95 mm Hg, but pressuresare not simultaneously below both 140 mmHg systolic and 90 mm Hg diastolic.

Left ventricular hypertrophy was de-fined solely on the basis of the ECG, andwas diagnosed in the presence of a highvoltage ORS complex and ST-T wave ab-normalities (Minnesota codes 3-1 plus 4-1,4-2 or 4-3 plus 5-1, 5-2 or 5-3). Hyperten-sive heart disease was considered to bepresent if the subject had hypertension asdefined above plus ECG evidence of leftventricular hypertrophy.

The decision as to the levels at which abiochemical or physiological variableshould be considered a "risk" factor in agiven population is an arbitrary one andshould ultimately be determined on thebasis of long-term outcome. Available datasuggest that for cholesterol and blood pres-sure, and perhaps for the other risk factorsas well, the risk of subsequent development

of CHD increases at each increasing levelof the parameter (12). For ease of presenta-tion, however, we have chosen commonlyused arbitrary cutpoints.

RESULTS

Prevalence of coronary heart disease.The prevalence of coronary heart disease,as defined by ECG, and the prevalence oftwo classes of chest pain, as determined bystandard questionnaire, are presented intable 1. With the low prevalence ratesobserved, especially in Japan, the confi-dence intervals around the age-specificrates are wide relative to the magnitude ofthe rate. We have therefore summarizedthese as age-adjusted prevalence rates.Age-adjustments were performed by theindirect method using the three popula-tions pooled as the standard from whichthe standard age-specific rates were de-rived. There is no difference in prevalence

TABLE 1

Prevalence of coronary heart disease as determined byECG and standard questionnaire for Japanese males

by geographic location

Observationalbase and diagnosis

ECGDefinite CHD*Definite and

possibleCHDf

QuestionnaireAngina pec-

torisrj:Possible infarc-

tion:):

(No. of men) §

Japan

5.325.4

11.2

7.3

(2141)

Age-adjustedprevalence/1000

Hawaii

5.234.7

14.3

13.2

(8003)

Cali-fornia

10.844.6

25.3

31.4

(1834)

* Major Q/QS abnormalities: Minnesota Codes1-1-1 through 1-1-7.

t Definite and possible CHD = major and minorQ/QS abnormalities: Minnesota Codes 1-1-1 through1-3-6.

% Cardiovascular questionnaire (Rose and Black-burn, 1968).

§The numbers of men in each of the tables varydue to differences in the numbers of missing values.


of definite CHD between Japan and Ha-waii, but the California Japanese havemore than twice the rate of the other twocohorts. When the criteria for CHD arebroadened to include lesser Q/QS abnor-malities, i.e., definite and possible CHD, aJapan-Hawaii-California gradient of in-creasing prevalence is seen.

Angina pectoris and possible myocardialinfarction are diagnosed by questionnaire.Both of these show the same Japan-Hawaii-California gradient in prevalenceand, as with CHD by electrocardiogram,the differences between California and Ha-waii are greater than the Japan-Hawaiidifferences.

Doubt may be raised about the use, inlinguistically different populations, of aquestionnaire which was developed in Eng-lish and only subsequently translated intoJapanese. It has been shown that thecultural background of the respondentsand the mode of administration of thequestionnaire can affect prevalence esti-mates (3). However, the fact that theprevalence estimates, based on question-naire responses, of angina and, to a lesserextent, of possible myocardial infarctionare parallel to prevalence numbers of CHDdetermined by electrocardiogram, suggeststhat the differences in prevalence are realand not simply a function of differentialresponse to the questionnaire.

Prevalence of hypertension and hyper-tensive heart disease. Given the previouslyreported inverse relationship betweenstroke and cardiovascular disease mortal-ity among the three Japanese populations,the prevalence of hypertension in the threegroups is of considerable interest (table 2).California Japanese have a higher preva-lence of definite hypertension and highermean blood pressures than the other co-horts and, with the exception of the oldestand the youngest age groups, the Japancohort has a higher prevalence of hyperten-sion than is seen in Hawaii. The detailedblood pressure distributions are reportedseparately by Winkelstein et al. (8).

The prevalence of hypertensive heartdisease and of left ventricular hypertrophy(LVH) is presented in table 3. Hyperten-sive heart disease is defined as a combina-tion of hypertension and left ventricularhypertrophy on ECG. Despite the lowerprevalence of hypertension in Japan, ahigher prevalence of hypertensive heartdisease is observed in Japan than in Cali-fornia. This suggests that the occurrence ofLVH as defined by ECG occurs morefrequently in Japan than in the other twocohorts, regardless of blood pressure level.This is confirmed in table 3.

Prevalence of biochemical abnormali-ties. The distributions of the biochemicalvariables in the three populations by ageare presented in a companion paper (9).Here attention is confined to the meansand the proportions of individuals whosevalues for these variables lie above arbi-trarily defined cutpoints.

The prevalence of hypercholesterolemia(serum cholesterol > 260 mg/100 ml) andmean serum cholesterols are shown in table4. There is a Japan-Hawaii-California gra-dient of increasing prevalence of elevatedserum cholesterol and of increasing meanserum cholesterols. The lower cholesterollevels in the oldest group might in part bethe result of the type of sample usuallyrepresented in a cross-sectional examina-tion; the oldest, sickest group being under-represented.

The comparison of serum triglyceridelevels is complicated by the fact that thesubjects in the study were non-fasting.Results must therefore be interpreted withcaution and the use of conventional cut-points becomes questionable. For this rea-son, the 80th percentile for the pooled datafrom all three sites was selected as thecriterion of elevation, and the proportion ofindividuals in each age-area group fallingabove that arbitrary cutpoint (280 mg/100ml) is compared (table 5). There is no cleardifference between Hawaii and California,but both the American cohorts show a verymuch higher prevalence of elevated serum


TABLE 2

Prevalence of hypertension* and mean blood pressure for Japanese males by age and geographic location

Age

45-49

50-54

55-59

60-64

65-69

Age-adjusted(No. of men)

Mean blood pressure and preva-lence of hypertension

(per 1000)

Hypertensives/1000Mean systolic (mm Hg)Mean diastolic(No. of men)

Hypertensives/1000Mean systolicMean diastolic(No. of men)




rate of hypertensives/1000

Japan

139.5125.780.6

(294)

194.6130.482.4

(442)

255.0136.084.8

(451)

280.6140.083.1

(506)

318.0143.483.0

(434)

223.7(2127)

Hawaii

142.5128.381.8

(1832)

183.3131.882.2

(2788)

199.1134.282.6

(1592)

247.9138.482.0

(1335)

352.8141.281.1

(451)

194.4(7998)

California

234.8132.587.5

(707)

286.8136.388.6

(509)

263.6141.789.4

(264-)

384.2143.489.0

(164)

423.8147.389.3

(151)

315.8(1795)

1 Hypertension = systolic blood pressure >160 mm Hg and/or diastolic blood pressure >95 mm Hg.

TABLE 3

Prevalence of hypertensive heart disease and leftventricular hypertrophy for Japanese males by

geographic location

Diagnosis

Age-adjustedprevalence/1000

Japan Hawaii Cali-fornia

Hypertensive heartdisease*

Left ventricular hy-pertrophyf

(No. of men)

9.3

16.4

(2127)

1.4

5.7

(7998)

4.6

6.1

(1795)

* Definite hypertension (systolic > 160 mm Hg or diastolic>95 mm Hg) plus left ventricular hypertrophy on ECG.

t Left ventricular hypertrophy = Minnesota codes 3-1 plus4-1, 4-2 or 4-3 plus 5-1. 5-2 or 5-3.

triglyceride than the Japanese cohort. Themean serum triglycerides, in general, showa Japan-Hawaii-California gradient.

One hour post-load serum glucose,

rather than blood glucose, was used toassess the prevalence of hyperglycemia.The glucose levels, shown in table 6, aretherefore 30-35 mg/100 ml higher than thecorresponding blood glucose figures wouldhave been (14). Various biochemical crite-ria have been used in other studies toassess diabetes and these vary greatly.Whichever criteria are used, comparativelyhigh cutoff points will result in few falsepositives and many false negatives, whilethe reverse will be true if the cutoff pointsare comparatively low. Here, a cutoff pointof 200 mg/100 ml was used and the result-ing prevalence figures are referred to ashyperglycemia rather than diabetes. Itmay be seen that the frequency of hyper-glycemia is approximately equal in Hawaiiand California, but both the American


TABLE 4

Prevalence of hypercholesteroiemia* and mean serum cholesterol for Japanese males by age and geographiclocation

Age

45-49

50-54

55-59

60-64

65-69

Serum cholesterol

>260 mg/100 ml (rate/1000)Mean serum cholesterol (mg/100 ml)(No. of men)





Age-adjusted prevalence/1000, serum cholesterol >260mg/100 ml

(No. of men)

Japan

27.9179.8(287)

33.9182.5(442)

30.6181.5(457)

27.1182.2(517)

27.6180.9(435)

31.6

(2138)

Hawaii

131.4219.4(1819)

135.8219.4(2783)

132.7218.7(1582)

127.1216.7(1330)

73.8211.1(447)

124.0

(7961)

California

138.9223.4(720)

198.4228.2(514)

161.0226.8'(267)

160.5223.6(162)

189.5224.0(153)

162.5

(1816)

* Serum cholesterol >260 mg/100 ml.

cohorts show a greater frequency thanJapan.

The California figures probably repre-sent an underestimate of the true preva-lence of hyperglycemia, as subjects whoreported that they had ever been told by adoctor that they had diabetes were notgiven an oral glucose load. Hence, a com-parison of data based on the presence ofeither an elevated serum glucose and/or apositive history of diabetes is presented intable 6. This raises the prevalence esti-mates in all three geographic locations, butproduces little change in the differencesbetween them.

The relationship of blood pressure andserum cholesterol to CHD. A higher preva-lence of electrocardiographic and sympto-matic evidence of CHD has been observedin California. In general, there is a Japan< Hawaii < California gradient, withrelatively smaller differences betweenJapan and Hawaii than between Hawaii

and California. It is of interest to note thatthe frequency of elevated cholesterol levelsin the three areas does not precisely paral-lel the CHD prevalence. For cholesterolelevation there are substantial differencesbetween Japan and the two American co-horts, and smaller differences between Ha-waii and California.

It is also surprising that the distributionof hypertension did not parallel the CHDdistribution. For hypertension, Japan hasprevalence levels intermediate between thelower levels in Hawaii and the higher levelsin California.

A closer study of these relationships ispresented in figures 1 and 2. For any givenlevel of blood pressure (figure 1), the Ja-pan-Hawaii-California gradient in CHDprevalence persists. Given the differencesin blood pressure distributions between thecohorts, it is possible that these broadblood pressure groupings do not satisfacto-rily "control" for differences in blood pres-


TABLE O

Prevalence of hypertriglyceridemia* and mean serum triglyceride for Japanese males by age and geographiclocation

Age

45-49

50-54

55-59

60-64

65-69

Serum triglyceride

>28O mg/100 ml (rate/1000)Mean serum triglyceride (mg/100 ml)(No. of men)

>28O mg/100 ml (rate/1000)Mean serum triglyceride (mg/100 ml)(No. of men)

>280 mg/100 ml (rate/1000)Mean serum triglyceride (mg/100 ml)(No. of men)



Age-adjusted prevalence/1000, serum triglyceride >280mg/100 ml

(No. of men)

Japan

79.5149.1(264)

64.5142.7(372)

62.0140.7(371)

38.4125.8(443)

41.1123.6(365)

61.4

(1815)

Hawaii

272.1235.8(1742)

278.8247.7(2679)

242.9225.6(1523)

180.5200.3(1280)

193.5198.6(434)

245.5

(7658)

California

278.4241.7(722)

269.4250.2(516)

251.9237.3(270)

262.2231.3(164)

187.5220.7(153)

251.9

(1825)

* Serum triglyceride >280 mg/100 ml.

TABLE 6

Prevalence of hyperglycemia* and history of diabetes for Japanese males by age and geographic location

Age

45-4950-5455-5960-6465-69

Age-adjusted rate(No. of men)

Prevalence of hyperglycemia/1000

Japanf

72.6133.091.7

153.3146.7

113.1(1195)

Hawaii

141.4179.8220.7259.8319.2

201.6(7977)

California

164.8160.8217.2186.3264.9

192.2(1805)

Hyperglycemia and/or history ofdiabetes (prevalence/1000)

Japan

153.2183.5154.2230.0223.3

175.0(1195)

Hawaii!

172.1220.7270.5316.8361.6

244.7(7911)

California

206.7225.5280.9285.7351.0

256.5(1805)

* Serum glucose >200 mg/100 ml one hour after 50 gm glucose load. No glucose load was given to diabetics inCalifornia, but they had a serum glucose measured and included in the tabulations.

fin Japan, subjects who were examined in the evening did not have a post-load glucose determination.% For 66 subjects in Hawaii, the history of diabetes was unknown.

sure. To assess this possibility the meanblood pressures within each hypertensioncategory were compared between cohorts.The differences in means were small forboth systolic and diastolic blood pressures,suggesting that differences in blood pres-

sure alone do not account for the differ-ences in CHD.

Figure 2 shows that the Japan-Hawaii-California CHD gradient persists at eachlevel of serum cholesterol. There are toofew subjects in Japan with elevated serum


PrevalenceCHD/lOOO

(Age odjus

Btooo

NO 0! «,«.

7 0

60

50

40-of

30t o )

20-

10-

| | JAPAN 620

^^M HAWAII W

BSCAUF —

Sm

iS

Normal

724 4,578 722

SI 5

SBS

11Borderline

316 1,873 503 3

676Illllllll

1H,,n

30 1,527 523

FIGURE 1: Prevalence of "all CHD" (age-adjustedrate/1000) at categorical levels of blood pressure forJapanese males by geographic location. CHD =Minnesota code 1:1 on ECG or angina or myocardialinfarction determined by questionnaire. Normal, bor-derline, high—WHO criteria.

70-

60-

50-

40-Prtvolence ofCHD/lOOO 3Q.(Age adjusted)

20-

10-

Choleiterol (Mg%)

No of men

1 1 JAPAN1 ' 696

^ ^ | HAWAI1 SPI III

^ ^ SB

r^i-nii<220

1,163 4,279 81 1

s sS 48 3 S

5 Is

l! II220-259 >26O

132 2,662 661. (42) 996 290

FIGURE 2: Prevalence of "all CHD" (age-adjustedrate/1000) at categorical levels of serum cholesterol forJapanese males by geographic location. CHD =Minnesota code 1:1 on ECG or angina or myocardialinfarction determined by questionnaire.

cholesterol to compute CHD prevalencenumbers. Once again the possibility wasassessed that differences between the co-horts in the underlying distributions ofserum cholesterol are inadequately takencare of by using these broad serum choles-terol groupings. There was a 20 mg/100 mldifference in mean serum cholesterol be-tween Japan and the other two cohorts inthe < 220 mg/100 ml category. This isconsistent with the possibility that thelower CHD prevalence in Japan in this lowcholesterol category is a result of lowerserum cholesterol levels in Japan. How-ever, the mean serum cholesterols by loca-tion, within each of the two higher choles-

terol groupings were very close. This sug-gests that the cholesterol groupings"220-259" and ">260 mg/100 ml" do con-trol for serum cholesterol differences. Thefact that within these categories the Cali-fornian Japanese have a higher CHD prev-alence indicates that differences in serumcholesterol do not completely account forthe Japan-Hawaii-California differences inCHD prevalence.

The Mantel-Haenszel procedure (36)was used to examine these CHD prevalencedata, controlling simultaneously for age,blood pressure and serum cholesterol. Con-trolling for these three variables, and tak-ing the Japan CHD rate as 1, the relativerisk (strictly, relative prevalence) in Ha-waii is 1.6 and in California is 2.1.

Smoking and CHD. Figure 3 presents theprevalence of CHD, controlling for currentsmoking pattern. Within each smokingcategory, the Japan-Hawaii-CaliforniaCHD gradient persists. For Japan andHawaii, the high prevalence in the ex-smoker category is likely the result ofindividuals changing their smoking habitsconsequent upon developing the disease. Itshould be pointed out that although smok-ing is very common in Japan, fewer smok-ers in Japan are "heavy" smokers as com-pared with Hawaii and California (7).Hence the apparent difference in CHDprevalence between California and Japanamong smokers may be overstated.

Prevalence olCHD/lOOO

70

60

50-

40-

30.

20

10-

| |jAMN

^ ^ H HAWAII

^ • j CALIF *A!

sS

nil

707

62 0 S

s s2 Ss s

IE •!496 I4fi 2O93 663 991 3503 f

FIGURE 3: Prevalence of CHD in Japan, Hawaii,and California controlling for smoking. CHD = Min-nesota code 1:1 on ECG or angina or myocardialinfarction determined by questionnaire.


It will be noted that for these last analy-ses a combined category of "all CHD" wasused to increase the number of cases. Thereare many suggestions that angina pectorisand other manifestations of coronary heartdisease may be epidemiologically distinct.However, since the patterns of variation inangina and in prevalence of CHD (definedby ECG) between geographic sites arequite similar in the present study, it seemsreasonable to combine them. In California,where the prevalence of CHD was highest,analyses were done separately on thesymptom categories and ECG categories.These separate categories showed essen-tially the same relationships to serum cho-lesterol, blood pressure and smoking as didthe combined category of "all CHD".

DISCUSSION

The findings indicate that Japanese liv-ing in California have a higher prevalenceof all manifestations of CHD than doJapanese living in Hawaii or Japan andthat Japanese in Hawaii, for three of thefour measures used, have higher rates thanJapanese in Japan. Before these findingscan be accepted, the possibility that theymay be due to biases inherent in cross-sec-tional examination or to artifacts of diag-nostic methods should be examined.

Clearly, if a disorder has equal incidencein two populations, the prevalence ratesshould be proportional to the survivalrates, other things being equal. It is un-likely that lower survival rates are respon-sible for the lower prevalence rates inJapan and Hawaii since case-fatality ratesfor Japanese under study in Hiroshima andNagasaki are reported to be quite low (15),and since this proposed high case-fatalityrate is not consistent with the gradient inCHD mortality reported by Gordon (2, 3)and largely confirmed by our own mortalitysurveillance (1). It is possible that non-response due to factors other than deathmay have biased the prevalence estimates.For this factor to have produced the gradi-ent of prevalence rates observed, non-

respondents in California would have tohave been at much lower risk of CHD thannon-respondents in Hawaii and Japan.Based on our mortality surveillance stud-ies, there is no evidence that this is thecase. In addition it may be argued that thelower response rates to examination inHawaii and California as compared withJapan may have biased the prevalenceestimates; but as non-respondents may ingeneral have poorer health than respond-ents (37), the bias would be in the directionof lowering the rates in Hawaii and Cali-fornia and could not account for the ob-served prevalence differences.

The possibility remains that these differ-ences in prevalence represent variation indiagnostic methods. This is especiallylikely with a questionnaire used for inter-national comparisons, with the difficultiesencountered in translation and differentmodes of administration (13). However, allECG's were centrally coded in a uniformfashion. The fact that the prevalence esti-mates from the questionnaire generallyshow geographic variations similar to thetrends in prevalence determined by ECG,and that both these prevalence estimatesshow a gradient similar to that predictedby Gordon's mortality report, is support forthe validity of the observed gradient inCHD.

It should be further borne in mind thatwhen ECG items are compared this is notthe same as comparing the degree ofatheroma in the coronary arteries and it ispossible that the relationship between vari-ous ECG items and "true" coronary heartdisease may differ slightly in these Japa-nese populations from the relationship thatis assumed to exist in Caucasian popula-tions. This is currently under investigationin the present study.

Hypertensive heart disease is far morecommon in Japan than in the Americancohorts, a finding which would not havebeen predicted from the blood pressuredistributions. This high prevalence of leftventricular hypertrophy, which is more a


function of high voltage of the QRS com-plex than of ST-T wave changes, has beenfound in other studies in Japan where thepossibility was raised that the high voltageseen on ECG was a function of thin chestwall in the Japanese rather than of in-creased myocardial bulk (16, 17). How-ever, high voltage cannot be considered tobe totally physiological as Japanese menand women who display this finding have asubstantially increased risk of subsequentCHD mortality (15). This increased risk ofCHD mortality among subjects with leftventricular hypertrophy is in accordancewith findings from white men and womenin Framingham (18). The apparent dis-cordance between prevalence of hyperten-sive heart disease and prevalence of hyper-tension between Japan and California mayalso be partly a function of the circum-stances surrounding the measurement ofblood pressure. The study sample fromJapan has been under observation throughthree cycles of examinations whereas thedata from the Hawaii and California sam-ples are from a first cycle of clinical exami-nations. The experience in Hiroshima andNagasaki, as well as in Framingham (19),has been that blood pressure recordingswere higher in the first cycle than insubsequent cycles and it is possible there-fore that the blood pressures of the Ameri-can cohorts are spuriously high relative tothose in Japan. This possibility could ex-plain only part of the differences in bloodpressure distributions observed and cer-tainly does not account for the Hawaii-California differences.

Diabetes is relatively common in Japan(20, 21). In the present study the preva-lence of hyperglycemia is high, especiallyin Hawaii and California. Comparisonswith other US populations are hindered bydifferences in methods used. However,when allowances are made for these differ-ences and the problems inherent in com-parisons of serum glucose with blood glu-cose, then it would appear that the preva-lence of hyperglycemia in the (predomi-

nantly Caucasian) population samplestudied in the US Health ExaminationSurvey is intermediate between the preva-lence in Japan and the prevalence in thetwo Japanese-American cohorts (22).Given this high prevalence of hyper-glycemia, it will be of considerable interestto directly examine the relationship be-tween diagnosed diabetes and CHD inJapanese-Americans, especially as it hasbeen reported that CHD is a comparativelyrare complication of diabetes in Japan(21).

The levels of serum cholesterol in Hawaiiand California, although higher than inJapan, are lower than those reported in theUS Health Examination Survey's popula-tion sample (23). The very low levels ofserum cholesterol in Japan, where theprevalence of CHD is very low, are consist-ent with the established role of hyper-cholesterolemia as a risk factor. However,at equivalent levels of serum cholesterol,California Japanese still have higher CHDprevalence. Similarly, at equivalent levelsof blood pressure, the high California prev-alence persists. To make definite state-ments about relationship of risk factors todisease from prevalence figures would beinappropriate, but the cautious interpreta-tion from these simple prevalence compari-sons is that other factors interact withblood pressure and cholesterol in the eti-ology of CHD.

Certainly the role of blood pressure inthe etiology of stroke and heart disease isfar from simple. It is known in the UnitedStates that blacks have higher blood pres-sures than whites (24-26) but have lowerprevalence and incidence of heart diseasethan do white Americans and higher preva-lence and incidence of stroke (27, 28). Inthe present study, the Hawaii Japanese,whose reported rates of stroke mortalityand CHD mortality are intermediate be-tween Japan and California, have the low-est prevalence of hypertension. The preva-lence of hypertension is high in both Japanand California, but is comparable to previ-


ous studies on Japanese men (17, 29, 30).The reason why high blood pressure shouldcontribute to CHD in the California Japa-nese and to stroke in the Japanese in thehome country is not clear. It may be, as hasbeen suggested, that high blood pressurepredisposes to CHD only in the presence ofa particular set of factors, including ele-vated serum cholesterol, and that highblood pressure in the presence of otherfactors predisposes to stroke (31). Theelucidation of these relationships mustawait the accumulation of precise inci-dence data.

As with any cross-sectional study, theseprevalence data only relate current riskfactor levels to existence of CHD. This maybe a problem with smoking, for example, ascurrent smoking habits do not reflect life-time exposure. However, as the higherCHD prevalence in California was seen ineach smoking category, it seems unlikelythat these CHD differences are all due todifferences in smoking.

If incidence data confirm that choles-terol, blood pressure and smoking differ-ences do not completely explain these gra-dients in cardiovascular disease, then wemust look elsewhere for explanations.These gradients may be a result of geneticfactors, i.e., the migrant Japanese may begenetically different from those resident inJapan in ways that make them more sus-ceptible to heart disease and less suscepti-ble to stroke. This seems unlikely sincesuch genetic differences would have toexplain differences between Japanese inHawaii and California as well as betweenmigrants and home populations. It is morelikely, then, that environmental factorsplay a role in these and other chronicdiseases. Other migrant studies of CHD,chronic respiratory disease (32, 33) andmany cancers (34, 35) have suggested thatmigrants commonly experience rates ofdisease intermediate between the levels intheir home country and the levels in theadopted country, and the results of the

present study are in accordance with thatpattern.

The Japanese in America lead verydifferent lives from the Japanese in thehome country. Among other things, theyeat different diets, they have differentpatterns of occupation and they live in adifferent social and cultural milieu. Therelationship between the changes in thesefactors and CHD, both within and betweenstudy cohorts, will be the subject of laterreports.

REFERENCES

1. Worth RM, Rhoads G, Kagan A, et al: Epidemio-logic studies of coronary heart disease and strokein Japanese men living in Japan, Hawaii, andCalifornia: Mortality. Am J Epidemiol102:481-490, 1975

2. Gordon T: Mortality experience among the Japa-nese in the United States, Hawaii and Japan.Public Health Rep 72:543-553, 1957

3. Gordon T: Further mortality experience amongJapanese Americans. Public Health Rep 82:973-984, 1967

4. Belsky JL, Kagan A, Syme SI.: EpidemiologicStudies of Coronary Heart Disease and Stroke inJapanese Men Living in Japan, Hawaii andCalifornia. Research Plan. Atomic Bomb CasualtyCommission Technical Report 12-71, 1971. Mi-crofiched and stored at: Bay Microfilm, Inc, 737Loma Verde Ave, Palo Alto, CA

5. Tillotson JL, Kato H, Nichaman MZ, et al:Epidemiology of coronary heart disease andstroke in Japanese men living in Japan, Hawaii,and California: Methodology for comparison ofdiet. Am J Clin Nutr 26:177-184, 1973

6. Kato H, Tillotson J, Nichaman M, et al: Epide-miologic studies of coronary heart disease andstroke in Japanese men living in Japan, Hawaiiand California: Serum lipids and diet. Am JEpidemiol 97:372-385, 1973

7. Kagan A, Harris BR, Winkelstein W, et al:Epidemiologic studies of coronary heart diseaseand stroke in Japanese men living in Japan,Hawaii and California: Demographic, physical,dietary, and biochemical characteristics. JChronic Dis 27:345-364, 1974

8. Winkelstein W, Kagan A, Kato H, et al: Epidemi-ologic studies of coronary heart disease and strokein Japanese men living in Japan, Hawaii andCalifornia: Blood pressure distributions. Am JEpidemiol 102:502-513, 1975

9. Nichaman MZ, Hamilton HB, Kagan A, et al:Epidemiologic studies of coronary heart diseaseand stroke in Japanese men living in Japan,Hawaii and California: Distribution of biochemi-cal risk factors. Am J Epidemiol 102:491-501,1975

525

10. Syme SL, Marmot MG, Kagan A, et al: Epidemi-ologic studies of coronary heart disease and strokein Japanese men living in Japan, Hawaii andCalifornia: Introduction. Am J Epidemiol102:477-480, 1975

11. Rose GA, Blackburn H: Cardiovascular SurveyMethods. Monogr Ser No 56, Geneva, WHO, 1968

12. Report of Inter-Society Commission for HeartDisease Resources (J. Stamler, A. Lilienfeld,chairmen). Primary prevention of the atheroscle-rotic diseases. Circulation 42:A55-A95, 1970

13. Marmot MG, Syme SL, Kagan A, et al: The useof a standard questionnaire to diagnose anginapectoris in international studies. Presented at thesixth annual meeting of Society for EpidemiologicResearch, Winnipeg, Canada, June 21-23, 1973

14. Committee on Statistics of the American Diabe-tes Association: Standardization of the oral glu-cose tolerance test. Diabetes 18:299-310, 1969

15. Johnson KG, Yano K, Kato H: Coronary heartdisease in Hiroshima. Report of a six-year periodof surveillance, 1958-1964. Am J Public Health58:1355-67, 1968

16. Takahashi N, Kato K, Suzuki K, et al: Epidemio-logic studies on hypertension and coronary heartdisease in a Japanese rural population. III. Elec-trocardiographic findings in Chiyoda. Jap Heart J5:37-48, 1964

17. Katsuki S, Hirota Y, Akazome T, et al: Epidemi-ological studies on cerebrovascular diseases inHisayama, Kyushu Island, Japan. Part I. Withparticular reference to cardiovascular status. JapHeart J 5:12-36," 1964

18. Kannel WB, Gordon T, Castelli WP, et al:Electrocardiographic left ventricular hypertrophyand risk of coronary heart disease. Ann InternMed 72:813-822, 1970

19. Kagan A, Gordon T, Kannel WB, et al: Bloodpressure and its relation to coronary heart diseasein the Framingham study. Hypertension. Vol 7:Drug Action, Epidemiology and Hemodynamics.Proceedings of the Council for High Blood Pres-sure Research, American Heart Association. NewYork, American Heart Association, 53-81,1959

20. Rudnick PA, Anderson PS: Diabetes mellitus inHiroshima, Japan. Diabetes 11:533-545, 1962

21. Blackard WG, Omori Y, Freedman LR: Theepidemiology of diabetes mellitus in Japan. JChronic Dis 18:415-427, 1965

22. US Department of Health, Education, and Wel-fare, National Center for Health Statistics: Glu-cose Tolerance of Adults. United States1960-1962. Vital and Health Statistics (PHSPublication 1000, Series 11, No 2), WashingtonDC, GPO, 1964

23. US Department of Health, Education, and Wel-fare, National Center for Health Statistics:Serum Cholesterol Levels of Adults. United

States 1960-1962. Vital and Health Statistics(PHS Publication 1000, Series 11, No 22), Wash-ington DC, GPO, 1967

24. McDonough JR, Garrison GE, Hames CG: Bloodpressure and hypertensive disease among Negroesand whites: A study in Evans County, Georgia.Ann Intern Med 61:208-228, 1964

25. Borhani NO, Borkman TS: Alameda CountyBlood Pressure Study. State of California, De-partment of Public Health, June 1968

26. US Department of Health, Education, and Wel-fare, National Center for Health Statistics: Hy-pertension and Hypertensive Heart Disease inAdults. United States 1960-1962. Vital andHealth Statistics (PHS Publication 1000, Series11, No 13), Washington DC, GPO, 1966

27. Cassel JC, Heyden S, Bartel AG, et al: Incidenceof coronary heart disease by ethnic group, socialclass, and sex. Arch Intern Med 128:901-906, 1971

28. Heyman A, Karp HR, Heyden S, et al: Cere-brovascular disease in the biracial population ofEvans County, Georgia. Arch Intern Med128:949-955, 1971

29. Suzuki K, Watanabe H, Takahashi N, et al:Epidemiologic studies on hypertension and coro-nary heart disease in a Japanese rural population.I. A study of blood pressure in Chiyoda. Jap HeartJ 3:544-554, 1962

30. Imai K, Takikawa M, Maeda Y, et al: National. survey of blood pressure and heart diseases in

Japan. Public Health Rep 80:321-329, 196531. Switzer S: Hypertension and ischemic heart dis-

ease in Hiroshima, Japan. Circulation28:368-380, 1963

32. Stamler J, Kjelsberg M, Hall Y, et al: Epidemio-logic studies on cardiovascular-renal diseases: III.Analysis of mortality by age-sex-nationality. JChronic Dis 12:464-475, 1960

33. Reid DD: Studies of disease among migrants andnative populations in Great Britain, Norway, andthe United States. I. Background and design. In:Epidemiological Study of Cancer and OtherChronic Diseases. Edited by W. Haenszel. NatlCancer Inst Monogr 19:287-299, 1966

34. Haenszel W: Cancer mortality among the foreign-born in the United States. J Natl Cancer Inst26:37-132, 1961

35. Haenszel W, Kurihara M: Studies of Japanesemigrants. I. Mortality from cancer and otherdiseases among Japanese in the United States. JNatl Cancer Inst 40:43-68, 1968

36. Mantel, Haenszel W: Statistical aspects of theanalysis of data from retrospective studies ofdisease. J Natl Cancer Inst 22:719-748, 1959

37. Doll R, Hill AB: Lung cancer and other causes ofdeath in relation to smoking. Br Med J 2:1070-1081, 1956

EPIDEMIOLOGIC STUDIES OF CORONARY HEART DISEASE AND STROKE ... · Stroke, Grant No. HL14783 from the National Heart Casualty Commission. and Lung Institute, the National Heart and

Documents