Twenty-five-year Experience in Cardiovascular and All-causes Deaths

Transcription

1 Clinical Studies The Seven Countries Study in Japan Twenty-five-year Experience in Cardiovascular and All-causes Deaths Hironori TOSHIMA, M.D., Yoshinori KOGA, M.D.,* Allessandro MENOTTI, M.D.,** Ancel KEYS, M.D.,** Henry BLACKBURN, M.D.,** David R. JACOBS, Ph.D.,** and Fulvia SECCARECCIA M.D.,*** SUMMARY Within the Seven Countries Study two Japanese cohorts of men aged were enrolled, one from the farming village of Tanushimaru (n=508) and the other from the fishing village of Ushibuka (n=502), both located in Kyushu. Cardiovascular risk factors were measured at entry and 10 years later in the survivors. The 25-year death rate for all causes was 30% higher in Ushibuka than in Tanushimaru (p<0.001) and higher also from specific causes except for violence. Baseline differences in the levels of age, blood pressure, serum cholesterol, smoking habits, body mass index and heart rate explained only 19% of the difference in all-cause mortality between the two towns. Multivariate models from data of the pooled cohorts showed that age was a significant predictor of coronary heart disease, stroke, cancer, all other causes and all-cause mortality. Cigarette smoking predicted coronary heart disease, cancer and all-cause mortality. Systolic blood pressure predicted coronary heart disease, stroke and all-cause mortality. Serum cholesterol, body mass index and heart rate predicted none of these five causes of death. Changes in systolic blood pressure during the first 10 years of follow-up were directly related to deaths between 10 and 25 years of follow-up, significantly improving the predictive power of the multivariate model. Changes in the other risk factors did not contribute, however, to improved prediction. (Jpn Heart J 36: , 1995) Key words: Risk factors in Japan Twenty-five year prediction HE Seven Countries Study of cardiovascular diseases was started in the late 1950s and early 1960s to compare cardiovascular disease morbidity and From Kurume University Hospital Medical Center, Kurume, *3rd Department of Medicine, Kurume University School of Medicine, Kurume, Japan **Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, Minnesota, USA, and ***Laboratorio di Epidemiologia e Biostatistica, Istituto Superiore di Sanita', Rome, Italy. Address for correspondence: Hironori Toshima, M.D., Professor and Director, Kurume University Medical Center, Kokubumachi, Kurume 830, Japan. Received for publication November 14, Accepted November 30,

2 180 TOSHIMA ET AL Jpn Heart J March 1995 mortality in cultures contrasting in lifestyles and purported cardiovascular disease rates. After pilot studies, cohorts were enrolled and surveyed, 1 in the USA, 2 in Finland, 1 in the Netherlands, 3 in Italy, 2 in Croatia, (former Yugoslavia), 3 in Serbia, (former Yugoslavia), 2 in Greece and 2 in Japan, for a total of about 12,000 men, aged ) The two Japanese cohorts were selected to represent a contrast with European and North American populations because the Japanese were said to have low rates of coronary heart disease and relatively high rates of cerebrovascular diseases. Data derived from the Japanese cohort were reported for 5- and 10-year experience in major monographs of the study2,3) and in a number of other papers.4-7) In a recent review of the Seven Countries Study,8) 17 papers were cited dealing with the Japanese experience. These dated from the 1960s or 1970s when the follow-up was still relatively short. With completion of the 25-year follow-up for vital status and causes of death in all 16 cohorts of the Seven Countries Study, detailed analysis of the Japanese data is now possible. A major finding after 25 years is a 30% higher total mortality rate in one Japanese village (fishing village, Ushibuka) than in the other (farming village, Tanushimaru). The objectives of this paper are: 1) to describe the 25-year mortality experience in the two cohorts and attempt to explain their differences; 2) to provide multivariate models for prediction of causes of death; and 3) to establish whether early changes in risk factors from baseline to 10 years contribute to the mortality differences after 25 years. MATERIAL AND METHODS The two Japanese cohorts of the Seven Countries Study consisted of men aged enrolled in "chunk" samples in the two villages, Tanushimaru and Ushibuka, located in Kyushu. Tanushimaru is a farming community whereas Ushibuka is characterized by a mix of farming and fishing activities. The entry examinations were held in 1958 in Tanushimaru (n=508) and in 1960 in Ushibuka (n=502) with nearly complete participation (100% in Tanushimaru and 99.6% in Ushibuka). Entry examination included numerous measurements reported elsewhere.1) For the present analysis, a limited set of risk measurements is considered: Age in years, rounded off to the nearest birthday; height and weight (following rules published in WHO Cardiovascular Survey Methods9) and used to compute the body mass index, weight in kg divided by height in m squared); Blood pressure measured in the right arm, in the supine position, at the end of a physical examination, by trained physicians using a mercury sphygmomanometer.9) (Mea-

3 Vol 36 No 2 THE SEVEN COUNTRIES STUDY IN JAPAN 181 surements were made to the nearest 2mmHg. Two measurements were taken 1 minute apart; systolic and diastolic 4th and 5th phase were recorded. For this analysis the average of the two systolic measurements was employed). Serum cholesterol was measured on a casual blood sample using the method of Abell and Kendall as modified by Anderson and Keys10) and was expressed in mg/dl. Cigarette consumption was self-reported on a questionnaire where average daily consumption was estimated; heart rate, in beats per minute, was taken as the average of rates measured in lead 1 and lead V6 of a standard 12-lead resting electrocardiogram. A similar examination was repeated 10 years after the first one. Complete follow-up was made for vital status and causes of death with complete data available for 25 years, based on documents at local Administrative Offices, supplemented by abstracts of hospital records, interviews with the local physicians and families of the dead persons and other witnesses of the fatal event. This information was later coded independently following the rules of the Seven Countries Study (2) and using the 8th Revision of the WHO-ICD.11) In case of multiple causes of death, a standard preference was given to the underlying cause of death, adopted with the following decreasing hierarchy: violence, cancer in advanced stages, coronary heart disease, stroke, and other causes as listed. The analysis involved computation of the mean values of risk factors measured at entry examination and after 10 years, and the 25-year, age-adjusted death rates in the two cohorts. The proportional hazards model12) was solved for the two cohorts lumped, due to the small numbers of deaths for coronary heart disease, strokes, all cancers, all other causes and all-causes mortality. The 6 risk factors were independent variables, with the addition of an extra covariate, a dummy variable, identifying the two cohorts (Tanushimaru as reference). In separate models of 10- to 25-year all-causes mortality, 10-year changes in the risk factors were entered as covariates, following a step-wise procedure. Subjects with missing values were excluded. The Lee method,13) based on the multiple logistic function, was used for estimation of the proportion of the difference in the 25-year total mortality between the two cohorts attributable to differences in risk factor levels. In particular, we report the ratio of the adjusted difference in rates between cohorts to the unadjusted difference. RESULTS The baseline characteristics of the two population samples are given in Table I. All the factors exhibit higher mean levels in Ushibuka than in Tanushimaru, except age, which was slightly and nonsignificantly higher in

4 182 TOSHIMA ET AL Jpn Heart J March 1995 Table I. The Seven Countries Study in Japan: Mean Levels of Cardiovascular Risk Factors in Tanushimaru and Ushibuka at Entry Examination (1958, 1960) Table II. The Seven Countries Study in Japan: Twenty-five Year Age-Adjusted Death Rates per 1000 from Different Causes. Tanushimaru and Ushibuka Table III. The Seven Countries Study in Japan: Solutions of the Proportional Hazards Model in the Lumped Cohorts with Five Causes of Death and Six Risk Factors as Covariates (n=851) Tanushimaru, and serum cholesterol, which was significantly higher in Tanushimaru by more than 5mg/dl. The higher levels in Ushibuka were statistically significant for body mass index, for systolic and diastolic blood pressure and heart rate. The 25-year death rates are given in Table II for single causes and for all-

5 Vol 36 THE SEVEN COUNTRIES STUDY IN JAPAN 183 No 2 causes and were higher in Ushibuka than Tanushimaru for most causes. Coronary heart disease, congestive heart failure and all heart disease death were all more frequent in Ushibuka than in Tanushimaru. While lung and liver cancers were more frequent in Ushibuka, stomach cancers were more frequent in Tanushimaru. Violent death was more frequent in Tanishumaru. Probably owing to small numbers, only "other causes" and "all causes" of death showed statistically significantly higher rates in Ushibuka than in Tanushimaru. In the category "other causes" there was, in Ushibuka, an excess of infectious disease deaths, with a rate more than double that of Tanushimaru (29.6 and 12.9 per 1,000 respectively). All-causes mortality was 30% higher in Ushibuka than in Tanushimaru. The Cox model, predicting 25-year mortality for coronary heart disease, stroke, cancer, "other causes" and all-cause mortality was computed as a function of 7 risk factors. This is summarized in Table III for both cohorts combined. Age was a significant predictor for each of the several causes of death. Cigarette smoking was significant only for coronary heart disease, cancer and for all causes, but not for strokes or for "other causes". Systolic blood pressure had a significant coefficient for coronary heart disease, stroke and all-causes mortality, whereas serum cholesterol and heart rate were never significantly predictive of any cause of death. Body mass index was significantly inversely associated with coronary heart disease and with all-causes mortality. The dummy variable identifying the two cohorts indicated a significant unexplained excess risk in Ushibuka for allcause deaths. Because in some models the coefficients for serum cholesterol and for body mass index exhibited a negative sign, all models were re-computed by adding a quadratic term for those two variables. In no case did the magnitude of the coefficients, or their t values, suggest a parabolic relation between cholesterol, body mass index and the selected mortality end-points. Comparison of risk factors measured at year 0 and year 10 in the pool of the two cohorts is made in Table IV. The Seven Countries Study in Japan: Mean Levels of Risk Factors in the Lumped Cohorts at Examinations in Year 0 and Year 10

6 184 TOSHIMA ET AL Jpn Heart J March 1995 Table V. The Seven Countries Study in Japan: Solutions of the Proportional Hazards Model in the Lumped Cohorts with All-cause Mortality Occurring between Year 10 and 25 of Follow-up, with and without the Addition of Changes in Risk Factors in the First 10 Years (Step-Wise) *Improvement in Chi-square between the two models due to change in risk factors is 7.51, p= Table VI. The Seven Countries Study in Japan: "Explanation" of Differences in 25-Year All-Cause Mortality by Risk Factors (Lee Method) Table IV. A nonsignificant 10-year decline in mean levels occurred for cigarette smoking, body mass index and heart rate, whereas a large and significant rise occurred in blood pressure. The significant decline seen for serum choresterol level (>13mg/dl) was likely an artifact; remeasurement of serum cholesterol in Tanushimaru 3 years after baseline yielded a mean of 150 and standard deviation 41mg/dl. The multivariate model computed for all-cause mortality occurring from 10 to 25 years after the entry examination is given in Table V, with and without the addition of risk factor changes between baseline and 10 years. The solution reiterates the findings of Table III. That is, there is a significant predictive role of age, cigarette smoking and systolic blood pressure. In the step-wise solution including 10-year risk factor change, only blood pressure change entered the model with a significant coefficient. The baseline factors retained their original predictive power. None of the other 10-year changes in risk factors reached the threshold for entering the model.

7 Vol 36 No 2 THE SEVEN COUNTRIES STUDY IN JAPAN 185 Another model was solved by adding the mean of the baseline and 10-year levels instead of their differences. In this analysis the mean level entered the model but the baseline measurement of blood pressure completely lost predictive power. The linear correlation coefficient of the 0- to 10-year systolic blood pressure change vs. the baseline value was -0.30, whereas that of 0- and 10-year mean values vs. the baseline values was The correlation of the entry pressure levels to 10-year levels of systolic blood pressure was Table VI presents estimates of the observed differences in all-cause mortality between Tanushimaru and Ushibuka that could not be accounted for by the available risk factors. Blood pressure alone explained 13% of the differential mortality, blood pressure plus smoking habits explained 19%. No more was explained when cholesterol, body mass index and heart rate were added. Thus, 81% of the differential mortality is not accounted for by the risk factors measured. DISCUSSION The all-cause 25-year mortality rates in these Japanese cohorts were relatively high, mainly in the fishing village of Ushibuka, compared to the other Seven Countries cohorts.7) As in each country where two rural cohorts were studied, high and low mortality cohorts were seen. That was the case for the two rural Finnish cohorts, the two Italian rural cohorts, the two Croatian rural cohorts, the two rural Serbian cohorts and the two rural Greek cohorts,7) although the differences were not significant in Italy. Overall in the Seven Countries cohorts the variance in coronary death rates within countries was smaller than the variance among countries, but it was the reverse for all-cause mortality. Altogether, as previously reported after 10 years of follow-up,14) each of the several main causes of deaths was higher in the fishing village of Ushibuka than in the farming village of Tanushimaru, except for violent deaths. Although mean levels of each of the risk factors available for these analyses were systematically higher in Ushibuka than in Tanushimaru, except serum cholesterol, the greater all-cause mortality rate in Ushibuka was poorly explained (19%) by risk factor differences. Alcohol intake, particularly the Japanese beverage shochu which was preferred in Ushibuka may partially explain the difference in death rates between the two communities, as suggested in previous papers.15,16) Alcohol intake was not computerized and it was therefore not possible to consider it in this paper. On the other hand, 25-year coronary mortality rates in these two Japanese villages were the lowest of the SCS overall, with the exception of the cohort in

8 186 TOSHIMA ET AL Jpn Heart J March 1995 the Greek island of Crete which scored between the two Japanese areas (46 per 1,000 in Crete, 45 per 1,000 in Tanushimaru and 63 pr 1,000 in Ushibuka), Speculation about the reasons for the low 25-year death rates from coronary heart disease in these Japanese cohorts has been made elsewhere7) and, in recent papers not yet published (Kromhout, personal communication, and Verschuren, personal communication). Low dietary fat consumption, particularly of saturated fatty acids, and low levels of serum cholesterol (around 160mg/dl on the average) were partial explanations for the difference, mainly in coronary disease, as reported in papers dealing with shorter follow-up periods.2-4,17,18)the observed decline in serum cholesterol levels is believed to be related to technical problems in cholesterol measurement at the baseline examination. Other reasons for the low level of coronary deaths might be the low levels of hemostatic factors and high levels in dietary omega-3 fatty acids described in Japanese populations compared to Western ones.19,20) Multivariate prediction of the various causes of death as a function of 6 risk factors was computed only on the cohorts combined, owing to small numbers. The unexplained mortality differences between the two cohorts were estimated by entering a dummy variable to identify the cohorts. This variable always suggested an excess mortality in Ushibuka not explained by risk factors. Only in one of the models, however, did the coefficient reach statistically significant levels, again possibly due to small numbers. One of the main features of the multivariate analysis in the Japanese cohorts was the absence of any significant relationship between serum cholesterol level and coronary heart disease. This was seen in only a few other groups of the Seven Countries Study, such as in Croatia and in the Italian Railroad samples (A. Menotti, unpublished). In all other cohorts of the Seven Countries Study serum cholesterol exhibited a strong direct relationship with future coronary rates.21-23) The absence of a positive relation in these populations may be attributable to the overall low mean cholesterol levels observed at the entry examination, and to the subsequent major decline in levels observed in the next 10 years. This might have resulted in the coronary events falling in a range where the slope of the cholesterol-coronary heart disease relationship is not steep. The small numbers involved probably played a major role. In these Japanese cohorts combined (n=1010) only 57 fatal coronary events were recorded in 25 years (46 in the subset of 851 suitable for multivariate analysis.) The literature shows contrasting evidence on the issue of blood cholesterol and risk. In a report from the Japan- Hawaii study, and in an independent Japanese study,24,25) a positive relationship between individual serum cholesterol levels and coronary events was found both in Japanese men living in Japan and those living in Hawaii. Other reports from the same Japan-Hawaii study did not, however, reach the same conclusion.26) On

9 Vol 36 THE SEVEN COUNTRIES STUDY IN JAPAN 187 No 2 the other hand, a Chinese population with very low levels overall of serum cholesterol, and a huge sample size, showed the usually found direct relationship.27) The roles of systolic blood pressure and cigarette smoking are similar to that reported in other areas of the Seven Countries Study for different lengths of follow-up2-5,21-23) and in other studies conducted in Japanese men.8,25-28,29) The stroke death rates in these Japanese cohorts were among the highest in the Seven Countries Study, as reported in the 20-year analysis.6) As expected, blood pressure was a major predictor of stroke but no contribution was found from the other factors, except age. In a previous report of the 20-year experience on stroke mortality in 6 countries of the Seven Countries Study,6) in none were serum cholesterol, physical activity, body mass index and cigarette smoking significant predictors of stroke mortality (with the exception of Yugoslavia for smoking habits). Reports from other studies with contrasting results on this issue could be better understood if a clear separation were possible between haemorrhagic and thrombotic stroke.30-39) Unfortunately this is not possible for Seven Countries Study data. In the overall mortality experience of the two Japanese cohorts combined, coronary heart disease represents only 12% of total mortality and stroke 21%; the two major cardiovascular causes combined are responsible for one-third of the total morality. It is therefore no surprise that in the models predicting allcause morality, the impact is only from age, blood pressure and cigarette smoking, considering the positive relationship of smoking habits to cancer morality rates. Cancer accounts for more than one third of all deaths in both areas. The overwhelming predictive role of systolic blood pressure for mortality, and the systematic increase in blood pressure observed during the first 10 years of follow-up, explain why pressure changes were the only risk change adding power to the prediction of 10 to 25-year deaths. Finally, the large difference in total mortality between the two Japanese areas remains largely unaccounted for. The few available risk factors measured in the late 1950s and early 1960s allow no better explanation. REFERENCES 1. Keys A, Aravanis C, Blackburn H, Van Buchem FS, Buzina R, Djordjevic BD, Dontas AS, Fidanza F, Karvonen MJ, Kimura N, Lecos D, Monti M, Puddu V, Taylor HL: Epidemiological studies related to coronary heart disease; characteristics of men aged in Seven Countries. Acta Med Scand 180: 1, Keys A (ed): Coronary heart disease in Seven Countries. Circulation 41 (Suppl 1): 1, Keys A (ed): Seven Countries. A multivariate analysis of death and coronary heart disease. Harvard University Press, Harvard, Mass, p 1, Keys A, Menotti A, Aravanis C, Blackburn H, Djordevic BS, Buzina R, Dontas AS, Fidanza F,

10 188 TOSHIMA ET AL Jpn Heart J March 1995 Karvonen MJ, Kimura N: The Seven Countries Study deaths in 15 years. Prev Med 13: 141, Menotti A, Keys A, Aravanis C, Blackburn H, Dontas A, Fidanza F, Karvonen MJ, Kromhout D, Nedeljkovic S, Nissinen A: Seven Countries Study; first 20 year mortality data in 12 cohorts of six countries. Ann Med 21: 175, Menotti A, Keys A, Blackburn H, Aravanis C, Dontas A, Fidanza F, Giampaoli S, Karvonen M, Kromhout D, Nedeljkovic S: Twenty-year stroke mortality and prediction in twelve cohorts of the Seven Countries Study. Int J Epidemiol 19: 309, Menotti A, Keys A, Kromhout D, Blackburn H, Aravanis C, Bloemberg B, Buzina R, Dontas A, Fidanza F, Giampaoli S, Karvonen M, Lanti M, Mohacek I, Nedeljkovic S, Nissinen A, Pekkanen J, Punsar S, Seccareccia F, Toshima H: Inter-cohort differences in coronary heart disease mortality in the 25-year follow-up of the Seven Countries Study. Eur J Epidemiol 9: 527, The Seven Countries Study. A scientific adventure in cardiovascular disease epidemiology. (Kromhout D, Menotti A, Blackburn H eds) RIVM, Bilthoven NL, Brouwer Offset bv, Uthrecht, NL, p 1-219, Rose G, Blackburn H: Cardiovascular Survey Methods. World Health Organization, Geneva, Anderson JT, Keys A: Cholesterol in serum and lipoprotein fractions: its measurement and stability. Clin Chem 2: 145, WHO: International Classification of Diseases, Eight Revision, WHO, Geneva, Cox DR: Regression models and life tables. J Roy Stat Soc B43: 187, Lee J: Covariance adjustment of rates based on the multiple logistic regression model. J Chron Dis 34: 415, Kimura N, Toshima H, Nakayama Y, Mizuguchi T, Takayama K, Yoshinaga M, Fukami T, Tashiro H, Katayama F, Abe K, Arima T, Yokota Y, Minagawa E, Tanaka R, Akiyoshi T, Soejima K, Yamada K, Mizunoe A, Nakamura K, Oshima F, Tanaka K, Akasu K, Niizaki T, Ikeda H, Nakamichi E, Ageta M, Miike Y, Inoue T, Nakagawa T, Nanbu S, Tanioka T, Shimada S, Fukumoto T: Population survey on cerebrovascular and cardiovascular disease. The ten years experience in the farming village of Tanushimaru and the fishing village of Ushibuka. Jpn Heart J 13: 118, Shibata A, Hirohata T, Toshima H, Tashiro H: The role of drinking and cigarette smoking in the excess deaths from liver cancer. Jpn J Cancer Res 77: 287, Fukumoto I: Prospective population survey on dietary habits and natural span of life; a 15 year followup in the fishing village of Ushibuka, rural southern Japan. J Kurume Med Assoc 46: 43, 1983 (in Japanese) 17. Keys A, Aravanis C, Vam Buchem F: The diet and all causes death rate in the Seven Countries Study. Lancet 2: 58, Keys A, Menotti A, Karvonen MJ: The diet and 15-year death rate in the Seven Countries Study. Am J Epidemiol 124: 903, Iso H, Folsom AR, Wu KK, Finch A, Sato S, Munger RG, Shimamoto T, Terao A, Komachi Y: Hemostatic variables in Japanese and Caucasian men; tissue plasminogen activator, antithrombin III, and protein C and their relations to coronary risk factors. Am J Epidemiol 132: 41, Iso H, Sato S, Folsom AR, Shimamoto T, Terao A, Munger RG, Kitamura A, Konishi M, Iida M, Komachi Y: Serum fatty acids and fish intake in rural Japanese, urban Japanese, Japanese American and Caucasian American men. Int J Epidemiol 18: 374, Menotti A, Keys A, Kromhout D, Nissinen A, Blackburn H, Fidanza F, Giampaoli S, Karvonen MJ, Pekkanen J, Punsar S: Twenty-five year mortality from coronary heart disease and its prediction in five cohorts of middle aged men in Finland, the Netherlands and Italy. Prev Med 19: 270, Dontas AS, Menotti A, Aravanis C, Corcondilas A, Lekos D, Seccareccia F: Long-term prediction of coronary heart disease mortality in two rural Greek populations. Eur Heart J 14: 1153, Menotti A, Mariotti S, Seccareccia, Giampaoli S: The 25 year estimated probability of death from some specific causes as a function of twelve risk factors in middle aged men. Eur J Epidemiol 4: 60, Robertson TL, Kato H, Gordon T, Kagan A, Rhoads GG, Land CE, Worth RM, Belsky JL, Dock, DS, Miyanishi M, Kawamoto S: Epidemiologic studies of coronary heart disease and stroke in Japanese men living in Japan, Hawaii and California; coronary heart disease risk factors in Japan and Hawaii. Am J Cardiol 39: 244, 1977

11 Vol 36 No 2 THE SEVEN COUNTRIES STUDY IN JAPAN Kodama K, Sasaki H, Shimizu Y: Trend of coronary heart disease and its relationship to risk factors in the Japanese population: a 26-year follow-up, Hiroshima-Nagasaki Study. Jpn Circ J 54: 475, Yano K, MacLean C J, Reed DM, Shimizu Y, Sasaki, H, Kodama K, Kato H, Kagan H: A comparison of the 12-year mortality and predictive factors of coronary heart disease among Japanese men in Japan and Hawaii. Am J Epidemiol 127: 476, Chen Z, Peto R, Collins R, MacMahon S, Lu J, Li W: Serum cholesterol concentration and coronary heart disease in population with low cholesterol concentration. Br Ned J 303: 993, Kiyohara Y, Ueda K, Fujishima M: Smoking and cardiovascular disease in the general population in Japan. J Hypertens 8 (Suppl 5): S9, Fujishima M, Kiyohara Y, Ueda K, Hasuo Y, Karo I, lwamoto H: Smoking as cardiovascular risk factor in low cholesterol population: the Hisayama Study. Clin Exper Hypertenion 14: 99, Shimamoto T, Komachi Y, Inada H, Doi M, Iso H, Sato S, Kitamura A, Iida M, Konishi M, Nakanishi N: Trends of coronary heart disease and stroke and their risk factors in Japan. Circulation 79: 503, Dawber TR. The Framingham Study: The epidemiology of atherosclerotic disease. Harvard University Press, Cambridge, MA, USA, p 1, Aronow W J, Gutstein H, Lee NH, Edwards M: Three year follow-up of risk factors correlated with a new atherothrombotic brain infarction in 708 elderly patients. Angiol 39: 563, Reed D, Yano K, Kagan A: Lipids and lipoproteins as predictors of coronary heart disease, stroke and cancer in the Honolulu Heart Program. Am J Med 80: 871, Welin L, Svardsudd K, Wilhelmsen L, Larsson B, Tibblin G: Analysis of risk factors for stroke in a cohort of men born in N Engl J Med 317: 521, Abbott RD, Yin Y, Reed DM, Yano K: Risk of stroke in male cigarette smokers. N Engl J Med 315: 171, Jacobs D: The relationship between cholesterol and stroke. in International Conference of Stroke Prevention and Epidemiology. Health Reports Statistics Canada 6: 87, Wolf PA, D'Agostino RB, Kannel WB, Bonita R, Belanger AJ: Cigarette smoking as a risk factor for stroke: the Framingham Study. JAMA 259: 1025, Menotti A, Lanti P, Seccareccia F, Giampaoli S, Dima F: Multivariate prediction of the first cerebrovascular event in an Italian sample of middle aged men followed-up for 25 years. Stroke 24: 42, Benfante R, Yano K, Hwang L J, Curb D, Kaga A, Ross W: Elevated serum cholesterol is a risk factor for both coronary heart disease and thromboembolic stroke in Hawaiian Japanese men; implications of shared risk. Stroke 25: 814, 1994