Changes in cardiovascular disease risk factors in three Japanese national surveys 1971-1990.

Trends of cardiovascular disease risk factors were evaluated through comparison of three national surveys in Japan, 1971-1990. Data from three Japanese national surveys on circulatory disorders, conducted in 1971, 1980, and 1990, were analyzed. Variables common to the three national surveys were selected for analysis. Serum total cholesterol was also but was examined only in 1980 and 1990. Age- and sex-specific mean values, standard deviations, median values, and proportions in extreme categories were determined for all continuous variables and proportions in categories interest for all discrete variables. Trends from 1971-1990 and 1980-1990 were estimated by linear regression analysis for continuous variables, and logistic regression analysis for binary variables. Systolic blood pressure was decreasing constantly during three surveys in both men and women (beta = -0.22 in men and -0.34 in women, p < 0.01 in both). Total cholesterol was increasing rapidly in both men and women between 1980 and 1990 (beta = 1.27 in men and 1.41 in women, p < 0.01 in both). For other risk factors, the results were less consistent among age- or sex-specific groups. Continued monitoring of risk factor trends in Japan will be important for predicting and explaining future trends in the occurrence of coronary heart disease and stroke in this population.

Atherosclerotic and hypertensive cardiovascular diseases are a major cause of death in both developed and developing coun-tries1,2). A great deal of interest in their prevention has been expressed 3,4) ; nevertheless, preventive efforts remain insufficient since cardiovascular risk factors, such as high blood pressure, adverse lipid profiles, cigarette smoking, and obesity are highly prevalent in many populations2).
In Japan, cerebrovascular mortality, though falling, is still relatively high, while that for coronary heart disease is the lowest for any industrialized country 1,2) and has declined since 19705,6). These changes have been thought to result primarily from high blood pressure control, including both primary and secondary preventive measures7), and from dietary changes reflected, for example, by reduced frequency of extremely low serum total cholesterol concentrations, which have been associated with the risk of cerebral hemorrhages8). On the other hand, the role of serum total cholesterol as a major risk factor of coronary heart disease requires that changes in its distribution be carefully monitored9). Okayama et al. reported previ-ously, on increasing serum cholesterol concentrations in Japan10). Several studies have also reported increases in the prevalence of diabetes mellitus in Japan, a trend suggesting increased, and not decreased, cardiovascular disease risks [11][12][13] The Ni-Hon-San Study showed that Japanese suffer from coronary heart disease at a higher rate when exposed to western culture [14][15][16][17] and although the incidence is still not as high as that of Americans in general18), Japanese are clearly susceptible to the effects of adverse risk factor changes 16).
Understanding the trends of cardiovascular disease risk factors in Japan is important if potentially epidemic coronary heart disease is to be prevented. Previously, Ueshima et al6.7). showed the trends of coronary risk factors based sex and age specific mean values from the National Nutrition Survey and other national surveys between 1956 and 1980. They found that decline in blood pressure levels and in prevalence of hypertension and increasing treatment rate for cardiovascular disease might contribute to declining mortality from ischemic heart disease. Okayama et al.10) showed changes in total cho-lesterol and other risk factors based on the National Survey on Circulatory Disorders in 1980 and the National Nutrition Survey between 1980 and 1989. Their main results were increase in age-and sex-specific serum total cholesterol and decline in systolic blood pressure. Shimamoto et al8). also showed trends of total cholesterol, total serum protein, systolic blood pressure, and diastolic blood pressure based on the data in a rural community between 1963 and 1983. Like these other surveys, they reported increase in serum total cholesterol and serum total protein concentrations, and decline in systolic blood pressure levels in every age and sex group.
These analyses were all based on either sex-and age-specific mean values for grouped data or local community cohort data. It is preferable to analyze randomly sampled individual data from the whole nation. For the purpose of this study, to determine the trends of risk factors of cardiovascular diseases in Japan as precisely as possible, we have therefore utilized individual data from three national surveys on circulatory disorders which were carried out in 1971, 1980, and 1990.

Design of the surveys
The Ministry of Health and Welfare in Japan conducted National Surveys on Circulatory Disorders in 1971 and 197219), 198020), and 199021). The 1972 survey was a follow-up of participants in the study of 1971 and is therefore excluded from the present analysis.
Two-hundred districts were selected randomly from the total of 420 National Health Survey districts, which were selected randomly from 10,787 census districts, for the survey in 197119. These districts represented all Japan ; however, three districts could not contribute to the survey. All 19,128 individuals who resided in these districts and were 30 years of age or older as of September 30 were regarded as eligible subjects for the study. The survey was conducted on October 14 and October 20. The questionnaires were collected from 18,401 individuals, 96.2% (18,401/19,128) of all subjects. The main reasons for non-response were reports of being "busy" (55.3% of non-respondents) or "sick at home" (7.7%). Being "busy" was more frequent in the younger age groups, and being "sick at home" was more frequent in the older age groups. Due to incomplete answer and inadequate data collection, the number of data suitable for analysis was 12,964, 67.8% of total eligible subjects (12,964/19,128). The exact frequencies of available data by sex and age were not obtainable, but the estimated response rates in the age groups 50-59 and 60-69 were three to four percent higher and those in the age groups 30-39 and 70+ were two to three percent lower than the average 19). The survey consisted of a questionnaire and health examination. The questionnaire included the medical history of the examinees and family members, subjective symptoms, smoking status, drinking status, and other variables. The health examination included height, weight, skin fold thickness, blood pressure, urinary protein, urinary glucose, electrocardiogram (ECG), an examination of the fundus of the eye, and a general examination by a physician.
Data were collected mainly by medical and paramedical staff of health centers, such as physicians, public health nurses, and laboratory technicians, all trained in advance of the survey. The questionnaire was conducted by the trained public health workers. The blood pressure was measured by a mercury sphygmomanometer on the right upper arm in a seated position. The diastolic blood pressure.was recorded at phase V of the Korotkoff sounds. Urinary protein and glucose were determined by paper colorimetry using respective standard color charts. Slight color changes between positive and negative findings were classified as false positive.
The survey in 1980 was conducted on November 1 and 30 in three hundred survey districts selected randomly from the health statistics survey districts of 1980, which were selected from whole census districts20). If the residents of the districts were 30 years or older as of 1 November 1980, then they were regarded as subjects for the study. The total number of eligible residents by the same age criterion as in 1971 was 13,771. The survey methods were generally the same as in 1971, except that a blood examination, which included serum total cholesterol, total protein, albumin, uric acid, creatinine, and blood glucose, was introduced. The main reasons for non-response in 1980 were reports of being "busy" (41.2% of non-respondents) or "being out for a long period" (13.5%). The participation rates were 79.1% (10,897/13,771) as a whole, 73.5% for men and 84.2% for women. The rates by age were 76.8%, 79.1%, 82.6%, 84.1%, and 74.0% for ages 30-39, 40-49, 50-59, 60-69, and 70+, respectively. The regional participation rate was the highest in Hokuriku-Tokai (87.5%) and the lowest in Minami Kanto (67.3%).
For the survey in 1990, 300 districts were drawn at random from 1,040 districts designated for the 1990 National Livelihood Survey of households and house members, which were randomly selected from census districts in all parts of Japan 21). The total number of eligible residents by the same criterion as in 1971 and 1980 was 10,956. The number of individuals who were examined was 8,926 and the participation rate was 81.5% (8,926/10,956). The main reasons for nonresponse were, as in 1971, reports of being "busy" (59.0% of non-respondents), or being "sick at home" (6.5%). The participation rates by sex were 76.1% for men and 86.0% for women. The rates by age were 77.7%, 79.0%, 82.9%, 86.7%, and 82.9% for ages from 30-39 through 70+ as in 1980, respectively.
The methods of examination were described in the report of the survey21). Blood tests were conducted for 11 items : hemoglobin, GOT, GPT, *-GTP, total cholesterol, HDL-cholesterol, total protein, blood glucose, fructosamine, creatinine, and uric acid. For this survey, cooperation was obtained from the Department of Epidemiology and Mass Examination for Cardiovascular Diseases, Osaka Medical Center for Cancer and Cardiovascular Diseases, which has 16 years of experience in the U.S. CDC-NHLBI Lipid Standardization Program of the Centers for Disease Control 22,23). This quality control program showed that the lipid determinations in the 1980 and 1990 surveys were comparable.

Statistical Methods
Variables to these three national surveys were selected for analysis : smoking rate, height, weight, systolic and diastolic blood pressure, urinary protein and glucose, and certain broad classes of ECG abnormalities. Serum total cholesterol was added for 1980 and 1990. Information on alcohol consumption was excluded from the analysis because the data were not comparable across surveys. Analysis was performed separately by sex and age or was age-adjusted in multivariate analysis.
First, sex-and age-specific mean, standard deviation, and median values by survey year were determined for each continuous variables : systolic blood pressure (mm Hg), diastolic blood pressure (mm Hg), serum total cholesterol concentration (mg/dl), and body mass index (kg/m2). The prevalence ratios for high systolic blood pressure (140mm Hg and over and 160mm Hg and over), high diastolic blood pressure (90mm Hg and over and 100mm Hg and over), low serum total cholesterol concentration (lower than 150mg/dl), high cholesterol (220mg/dl and over and 260mg/dl and over), and high body mass index (25kg/m2 and over and 30kg/m2 and over) were also calculated. Second, age-specific prevalence ratios were estimated for each categorical variable : smoking rate, ECG findings, urinary protein, and urinary glucose. Cigarette smoking status was analyzed in three categories : non-smokers., smokers, and ex-smokers. Urinary protein and glucose findings were classified as negative, false positive, and positive according to the results of paper colorimetry. ECG abnormalities were classified as none, minor, and major, in accordance with the Minnesota Code (Appendix).
The trends of continuous variables were analyzed by multiple linear regression models to control age effects, by use of the SAS REG procedure 24). Year was coded as 0, 9, and 19 for 1971, 1980, and 1990, respectively, to calculate the trends over the three surveys ; it was coded 0 and 10 for 1980 and 1990 to calculate the trends between 1980 and 1990. In order to estimate the trends for smoking, ECG findings, urinary protein, and urinary glucose, these were treated as binary variables: smokers vs. others ; major abnormal ECG findings vs. others; positive urinary protein vs.' others ; and positive urinary glucose vs. others. The binary variables were analyzed by multiple logistic regression models by use of the SAS LOGISTIC procedure 25). The coding of year for the logistic regressior models was performed as for the multiple linear regressior models.

Sex-and Age-Specific Changes in Risk Factors
The changes in the distributions of risk factors by sex and age groups are summarized in Table 1 through Table 8.
Mean and median values of systolic blood pressure by age and the prevalence of high systolic blood pressure, at 140 mm Hg and over and 160 mm Hg and over, are shown in Table 1. In every age group in both men and women, mean systolic blood pressure declined over the survey period, especially at older ages, and in most age-sex groups the decrease in mean values was greater than in median values. The prevalence of high systolic blood pressure at 140 mm Hg and over and especially at 160 mm Hg and over also declined remarkably, in every age-sex group.
The mean and median values of diastolic blood pressure by age and the prevalence of high diastolic blood pressure, at 90 mm Hg and over and 100 mm Hg and over, are shown in Table 2. These results are contrary to those for systolic blood pressure : mean values of diastolic blood pressure did not decrease consistently among age-sex groups and increased in some ; prevalence ratios for high diastolic pressure did generally decrease, but not for men at>=90 mm Hg.
Mean and median values of serum total cholesterol concentration by age and prevalence ratios for values lower than 150mg/dl, 220mg/dl and over, and 260mg/dl and over are shown in Table 3. In every age group in both men and women, marked increases were observed, except for the decreasing prevalence of values below 150mg/dl in every agesex group.
Mean and median values of body mass index by age and the prevalence of body mass index of 25kg/m2 and over or 30kg/m2 and over are shown in Table 4. In men, mean body mass index increased in every age group ; the prevalence of 25 kg/m2 and over also increased in all but the youngest men over the period of the three surveys, but the prevalence ratio for 30 kg/m2 and over did not. In women, changes in body mass index varied with age : in the age groups younger than 50 years mean and median values decreased and increased in elder women, but all changes were small. For women under age 60, percentages with high body mass index decreased.
The prevalence of smokers and ex-smokers by sex and age group is shown in Table 5. The prevalence of smoking in men decreased consistently in all age groups. In women, smoking rates increased in the age groups younger than 50 years, and decreased in the age groups 50 years and older. At all ages, smoking was much less frequently reported by women than by men. The prevalence of abnormal ECG findings by sex and age is shown in Table 6. The prevalence of major abnormal findings decreased in all but the oldest men and in every age group of women. In both men and women, the proportions with normal ECGs were greatest in 1990, in all age groups.
The prevalence of positive tests for urinary protein by sex and age group over the survey years is shown in Table 7. In men, the percent positive increased slightly in all but the youngest age group. In women, decreases in prevalence 96 K. Sakata and D.R. Labarthe Table 1. Changes in systolic blood pressure (mm Hg) by sex and age occurred for all but the 40-49 year age group.
The prevalence of positive tests for urinary glucose by sex and age group over survey years is shown in Table 8.
Increases were observed except in men aged 30-39 and women aged 40-49. Trends of Risk Factors over the Survey Periods Table 9 shows the trends of systolic blood pressure, diastolic blood pressure, and body mass index by multiple regression analysis with control for age over the three surveys. Systolic blood pressure decreased significantly in both men and women over the period ; however, diastolic blood pressure increased for men and decreased for women. Body mass index increased Table 2. Changes in diastolic blood pressure (mm Hg) by sex and age in men but did not change in women. For diastolic blood pressure in men, no significant change was found between 1980 and 1990. Total cholesterol increased significantly in both men and women from 1980 to 1990. Body mass index again showed no significant trend in women, as in the three-survey analysis, above. Trends of smoking rate, the prevalence of major abnormal findings on ECG, the prevalence of positive tests for urinary Table 3. Changes in total cholesterol (mg/dl) by sex and age protein, and the prevalence of positive tests for urinary glucose over the three surveys are presented in Table 11. The trends were estimated using multiple logistic regression models to control age effects. The smoking rate declined in men, but not in women, over the three surveys. The prevalence of major abnormal findings on ECG decreased in both men and women. The trend of the prevalence of the positive tests for urinary protein in women was significantly negative. The prevalence of positive tests for urinary glucose increased in both men and women.
The decrease in the smoking rate in women between 1980 and 1990 became significant in contrast to the change from 1971 (Table 12). The trend of the prevalence of major abnormal findings on ECG was no longer significant in men between 1980 and 1990. The urinary glucose positivity rate in men between 1980 and 1990 was no longer significant.

DISCUSSION
This study, based on individual data from three national surveys in 1971, 1980, and 1990, shows that cardiovascular disease risk factors in Japan are changing in a complex manner.
The observed decrease in systolic blood pressure is consistent with other studies 7.10,26,27). Treatment of high blood pressure has become popular in Japan since the 1960s, because the Health Insurance Law became effective in 1961, and every Japanese has been covered under the insurance. The main reason for the decline seems to be the treatment of high blood pressure 7). It should be noted, however, that for systolic blood pressure the median values decreased and this measure, unlike the mean, would not be expected to reflect treatment effects at the upper extreme of the distribution. Changes in salt intake may partly explain the changes in systolic pressure, yet per Table 4. Changes in body mass index (kg/m2) by sex and age capita salt intake has been rising since 198728).
The trends of diastolic blood pressure in men showed a different pattern from that of women. The reason is still unknown ; however, several factors may be considered. First, the difference of the treatment for hypertension between men and women may explain the difference in the trends between men and women. The treatment rate is higher in women than in men7). Second, body mass index is increasing in men, but it is not increasing in women. It is well known that obesity is related to hypertension M. Third, the increase in alcohol consumption may partly explain the discrepancy. Alcohol consumption as ethyl alcohol was 6870 ml per capita in 1970 in adults and 8290 ml in 199030). The rate of increase over the last 20 years was 20.7%. In Japan, 55.4% of men and 6.5% of women were habitual drinkers in 199021). Therefore, men are subject to the influence of alcohol more than women by the increase of alcohol consumption. However, it is uncertain whether these factors affected systolic blood pressure and diastolic blood pressure differently.
Serum total cholesterol concentrations increased remarkably in both men and women between 1980 and 1990. Since elevated total cholesterol concentration is a major risk factor for Although the prevalence of obesity in Japan (defined by body mass index equal or higher than 25 kg/m2) is still low compared with other countries2), the upward trend for men especially notable because of the concomitant increases in serum total cholesterol concentrations and the prevalence of diabetes mellitus11-13). A possible explanation of the discrepancy between men and women in trends of body mass index is cosmetic preference for leaner appearance in younger women.
The smoking rates in men, though decreasing, are still high in comparison with those in other countries2). In women, the smoking rates are low but increasing among those younger than 50 years. For both men and women, smoking prevention and cessation are needed to reduce risk of cardiovascular diseases, cancers, and respiratory diseases 36.37).Therefore, every effort should be made to eradicate the smoking habit. Major abnormal findings on ECG are important indicators to trends in cardiovascular diseases, whose decrease in men was no longer significant between 1980 and 1990. In this period, age-specific trends were slightly upward in men younger than 60 years, coronary heart disease or other heart diseases may be starting to increase in frequency in Japan. So far, only urinary glucose data are available as a crude index of trends of diabetes mellitus in Japan. The urinary glucose positive rate showed a substantial increase in both men Table  7. Changes in positive tests for urinary protein by sex and age and women, although the increase in men between 1980 and 1990 was not significant. The prevalence of diabetes mellitus based on the fructosamine and hemoglobin Alc of the same subjects in 1990 was estimated to be 9% to 18% in men and 7% to 15% in women39). This estimate also suggests an increase in prevalence of diabetes11-13). Based on the next national survey of circulatory disorders, improved estimates of trends in diabetes will become possible. Two advantage of this study over previous reports are the random selection of survey districts from the whole country and comparability of methods across surveys.
One possible bias of this study is the participation rate of the study : 67.8%,79.1%, 81.5% for 1971, 1980 , 1990, respectively. It is possible that those who did not visit the examination Table 8. Changes in positive tests for urinary glucose by sex and age affected the results of observation ; however, the effect would not be a serious bias for the trends of major risk factors because the participation rates by age showed a similar pattern over the three surveys, and the trends were mostly similar to other studies 7,8.10,40-42).
Finally, those who were inpatients were excluded from the subjects of the surveys. The prevalence of a rare risk factor could be affected considerably, although the inpatient rate in 1993 was 1.15%43). As a design feature common to the three surveys, this factor could not affect the estimation of trends in the non-institutionalized population.
Another possible limitation is the quality of the data.
Although most of the examiners were workers of public health centers and had experience for the survey, because of the large number of examiners, the problem of interobserver variability must be considered. To improve the quality of data, strict training is desirable44,45). In addition, the examination of blood pressure was performed only once. In order to reduce the intraindividual variation, it is desirable to measure twice or more 46). Nevertheless, since the survey methods were common over the three surveys, trends themselves would not be distorted.  Table 11. Trends of smokers, major abnormal findings on ECG, positive tests for urinary protein, positive tests for urinary glucose by sex after controlling for age over three surveys in 1971, 1980, and 1990 *Parameter estimates by logistic regression models . So far controlling systolic blood pressure seems to have been successful; however, Japanese people are about to face a new problem, hypercholesterolemia, and still have high prevalence of smokers in men. Careful monitoring of these factors is essential. Further studies including children and adolescents are needed to detect more precise changes of risk factors and to provide effective preventive measures against cardiovascular diseases. Appendix.
Classification of electrocardiographic results