A Nineteen-Year Cohort Study on the Relationship of Electrocardiographic Findings to All Cause Mortality Among Subjects in The National Survey on Circulatory Disorders, NIPPON DATA80.

BACKGROUND: Electrocardiogram (ECG) is one of the most popular tools for daily clinics and health checkup, and has been used for the National Survey on Circulatory Disorders to assess the health status in Japanese people. The meaningfulness to predict mortality from all causes among people with ECG abnormality is to be clarified using national samples. METHODS: ECG findings recorded among 9,638 subjects for National Survey on Circulatory Disorders 1980 in Japan, were classified using the Minnesota Codes (mc). Their relationships to all cause mortality over 19 years were examined using Cox proportional hazard models adjusting for sex, age, systolic blood pressure, blood glucose, and smoking habits. RESULTS: Subjects with abnormal Q-QS findings showed significantly high hazard ratios ( 3.71(mc1-1) and 1.57 (mc1-3)) for mortality to the subjects who were free from any major ECG findings. Hazard ratios were 1.37 (mc2-1) to 4.16 (mc2-5) for axis deviation, and 1.34 (mc3-1) to 1.35 (mc3-3) for left high R waves. Those were 1.63 (mc4-3) to 2.59 (mc4-1) for ST depression, and 1.54 (mc5-3) to 2.33 (mc5-1) for T abnormality. The lower the second number of the Minnesota Codes was, the higher hazard ratio was observed in the Q-QS, ST, T codes. The hazard ratios of junction-type ST depression (mc4-4), and low T waves (mc5-4, 5-5) were not significant. CONCLUSIONS: ECG findings defined by the Minnesota Codes were useful to predict the risk for mortality from all causes even after adjusting for the other major risk factors, and the results supported a usefulness of the ECG for health check-ups.

The number of subjects with ECG findings and deaths of all causes are showed in Table 1, along with the death percentages in 19 years. The number of subjects in the reference group was 5,535 and the deaths were 836 (15.1%) during the follow-up period. The number of subjects with any major ECG code was 4,103, and the number of deaths was 1,174 deaths (28.6%). (Table 2) With analysis using Cox proportional hazard models, the death

ECG and Mortality in NIPPON DATA80
For an epidemiologic purpose, the Minnesota Code system was developed by Henry Blackburn and others in 1960, 2 and revised by Ronald Prineas and others in 1982. 3 Recently the revised code system was used worldwide to confirm the objectivity of ECG findings.
The Japanese Government performed a series of National Surveys on the Circulatory Disorders almost every ten years. 1 A research group had been organized to follow-up those participants to clarify the relation of risk factors, such as obesity, 4 smoking, 5 serum uric acid, 6 total cholesterol, 7 blood pressure, 8 resting heart rate, 9 and others. 10,11 These participants were from a national stratified random sample throughout the country, not merely nationwide pooled samples by relevant researchers. It was an epoch making achievement to follow these sampled participants and to evaluate their ECG findings objectively with the by Minnesota Code system. The attempt could be achieved by support of Ministry of Health and Welfare (presently, new Ministry of Health, Labour, and Welfare) and extensive cooperation of nearly 300 local public health centers along with collaborators listed in Appendixes I and II. Similar type national samples were available in the United States for National Health and Nutrition Surveys 1971Surveys -75, 1976Surveys -80, and 1988, and many valuable reports presenting year-by-year, [12][13][14][15][16][17] but not yet for extensive ECG findings.
The purpose of this study was to investigate the relationship of ECG findings coded by the Minnesota Code system to deaths of all causes by following-up 19 years of national samples participated in the National Survey on Circulatory Disorders conducted in 1980.
The Ministry of Health and Welfare of the Japanese government carried out the National Survey on Circulatory Disorders in order to clarify the recent status of Japanese people by examining all residents living in the 300 stratified random sample areas selected throughout the country in 1980. A Working Committee charged local public health centers on health check-up in the assigned areas under the guidance for the National Survey (the details described in an Official Report by the Ministry 1 ). In 1980, 10,546 people participated in the Survey (Response rate was 79.1%), and the Cohort Study Group on Activity in Daily Living and Prevention of Degradation in Quality of Life among the Elderly followed them up to 1999.
The Cohort Study Group identified 9,638 (4,103 males and 5,535 females) alive or dead people with ECG record successfully by 15 September, 1999. Among them, 118 were lost after the date of moving, but they were definitely alive till the time of moving out according to their official records, "residence registration record" kept in any one of local administrative offices throughout the country. The number of deaths was 1,327 (13.8%), in which 711 deaths were males (16.8%) and 616 deaths were females (11.4%). The detail causes of deaths among this cohort appeared in the previous papers 6  ard ratio for code 3-2 was significantly high only in males. The ratio for code 3-2 tended to be higher than those for code 3-1 and code 3-3 in males. (Table 4) 1) ST depression codes: Among the hazard ratios for code 4-1, 4-2, 4-3, and 4-4 (ST depression), those for code 4-1 and code 4-2 were significantly high in the sex combined, in males, and in females. The hazard ratios of subjects with ST depression were ordered inversely by the second code numbers in the sex combined, in males and in females.

2) T abnormality codes:
Among the hazard ratios of subjects with T abnormality, that for code 5-1 was the highest in females and in the sex combined. The ratio of code 5-2 was the highest in males. The hazard ratios for code 5-4 and for code 5-5 were not statistically significant. (Table 4) 1) Atrio-ventricular codes: Among the death hazard ratios of subjects with incomplete atrio-ventricular block, that of code 6-2 was the highest in males and in females. The hazard ratio for code 6-5 (short PQ interval) was also significantly high in males or females. The ratio for code 6-3 (PQ prolongation) was significantly high only in males. 2) Ventricular codes: Among the death hazard ratios of subjects with intraventricular block code (code 7-1, 7-2, 7-3, 7-4, and 7-5), those for code 7-1 and 7-2 were significantly high. Although the ratio for code 7-4 was also high but not significant in males and in the sexes combined. The hazard ratio of subjects with complete right bundle branch block (code 7-2) was significant only in females and in the sexes combined. (Table 4) 1) Ectopic beats: The hazard ratio of subjects with frequent ectopic beats (code 8-1) was over double but not significant in males and close to double in the sex combined. The hazard ratio with less frequent ectopic beats (code 8-9-1, an optional code for hazard ratios of subjects with any one of major ECG findings were formulated in Table 2, adjusted for sex, age, systolic blood pressure, blood glucose, and smoking habits. The hazard ratio of females to males was 0.65. By separate analysis by sex, the hazard ratio of subjects with abnormal ECG among males was 1.34, and tended to be higher than those (1.19) in females. The hazard ratios tended to be higher in females than those in males for smoking habits (1.29 vs 1.18). (Table 3) The Cox hazard ratios of mortality from all causes were essentially similar for the subjects with abnormal ECG finding before (1.28) and after excluding the deaths within 5 years after the examination (1.21). Those of sex, systolic blood pressure, blood glucose, or smoking did not change appreciably.

Effect by exclusion of the deaths close to the date of examination
Based on these results, the further analysis by Cox proportional hazard model was performed with all deaths throughout the follow-up period. (Table 4) 1) Q-QS codes: The hazard ratios of subjects with each Minnesota Code adjusted for age, systolic blood pressure, blood glucose, smoking habits, and sex if applicable, were shown in Table 4. The ratios for Q-QS code 1-1 were significantly high in both sexes.The ratio for minor Q-QS code 1-3 also showed a significant elevation.

2) Axis deviation codes:
The hazard ratio for axis code 2-1 (left axis deviation) was 1.37 for male and female combined. The ratio was not significant in males (1.19), contrary to that in females (1.81).
The ratios of subjects with axis code 2-3 (minor right axis deviation) were significantly high among males and females combined and among males. The hazard ratio for axis code 2-4 (extreme axis deviation, 2.85) and 2-5 (4.16) was significantly high among the sex combined and among males, and tended to be higher than those codes for 2-1, 2-2, or 2-3.

3) High R codes:
The hazard ratios for codes 3-1 and 3-3 (left high R codes) were significantly high at the same levels.   Minnesota codes (mc) in parencesis * : Secondary mention of cerebrovascular disease. **: Secondary mention of cardiovascular disease.
subjects (Table 5). This result corresponds to the fact that a subject with history of myocardial infarction has the higher risk of recurrence.
Although prevalence of the extreme axis deviation (code 2-4, n=5) or indeterminate axis (code 2-5, n=6) were relatively rare the observed high hazard ratios of early death might come from any disposition of the heart or abnormal propagation of the excitation due to any serious heart disease. The genesis of indeterminate axis was suggested as a posterior, rightward and superior orientation of terminal QRS forces, which might result from number of causes, by a quantitative vectorcardiographic analysis by Goldberger. 21 The high Cox hazard ratios of subjects with left high R (code 3-1 or code 3-3), due to hypertension, cardiomyopathy, valvular disease, or sports heart as well as merely thin chest, show the significance of myocardial hypertrophy for prevention of early deaths. The Cox hazard ratio of subjects with right high R (code 3-2) was the higher than those of code 3-1 and code 3-3, though significance was lower maybe due to fewer cases with 3-2 code. The highest but not significant hazard ratio was observed for subjects with bi-ventricular high R (code 3-4) which was added in the revised Minnesota Codes 1982. 3 The hazard ratio of subjects with ST depression was higher than that of those with T abnormality. Because ST depression code should be with T abnormality by definition in Minnesota Codes, 2 the hazard ratio of subjects with ST depression may show a combination effect of ST and T abnormalities. Besides coronary atherosclerosis, ST depression with T abnormality might be developed in hypertension. ST depression might be assumed as a sign of arterio-sclerosis and/or hypertensive stress leading to the relevant complication such as not only heart disease but also cerebrovascular disease, finally to death, as shown in Table 5.
One of the striking facts was that the T abnormality was always more frequent in females than males in almost all of Japanese communities. However, the hazard ratios of subjects with T abnormality (code 5-1 to code 5-3) were high in males and in females. The hazard ratio of subjects with T abnormality code 5-4 was high but not significant in this analysis. The code 5-5, minimally low T, was added according to Japanese scientists to Minnesota codes. 18 However, this study could not support the significance at the moment.
The hazard ratio of subjects with complete left bundle branch block was higher than that of complete right bundle branch block in males. However, the ratio of complete right bundle branch block in female was significantly high and larger than that of complete left bundle branch block.
It was reasonable that the hazard ratio of subjects with frequent ectopic beats was higher than that of less frequent ectopic beats (code 8-9-1). The frequent ectopic beats might lead to atrial fibrillation in case of supraventricular beats, and ventricular fibrillation in case of ventricular beats. The hazard ratio of atrial fibrillation was comparable with major high-risk codes such as Q-QS, ST depression, and T abnormality. the National Survey) was also over double and significant in males.

2) Atrial fibrillation or flutter:
Among the hazard ratios of subjects with arrhythmia, code 8-3 (atrial fibrillation or flutter) was the highest in females and in the sexes combined. The hazard ratios for code 8-7(sinus tachycardia) and code 8-8 were significantly high only in males, but not in females at all.

3) Miscellaneous codes:
The death hazard ratio of subjects with low voltage (code 9-1) was significantly high only in the sex combined. The hazard ratio of subjects with ST elevation (code 9-2) was significantly high in the sexes combined and in males, but not in females. The hazard ratio of subjects with counter clockwise rotation (code 9-4-1) was significantly high only in males, and that with clockwise rotation (code 9-4-2) was significantly high in each sex and in both sexes combined. The ratio with high T wave (code 9-5) was significantly high in males and in the sexes combined. (Table 5) The proportion of heart disease death in the abnormal ECG group in females tended to be higher than that in males (21.9 vs 17.2 %). The proportions of death from heart diseases tended to be higher in males with Q-QS abnormality, atrio-ventricular conduction defect, and ventricular codes than in females. Although the number of cases was limited, it was true of the deaths from ischemic heart disease.

Deaths from heart disease among the subjects with ECG codes
However, the percentage proportions of death from heart diseases tended to be higher in females with the other ECG code groups, especially left high R, ST depression, T abnormality, ectopic beats, atrial fibrillation or flutter, low voltage, or even counter clock or clock-wise rotation codes. The death proportion of deaths from cerebral infarction tended to be high in females with atrial fibrillation or flutter, but not so high in males (25.9% vs 12.5 %).
One of major contributions of this study comes from the extensive national samples residing in exactly 300 stratified randomly sampled areas which were sub-samples defined by the Statistical Information Bureau, Ministry of Health and Welfare (Ministry of Health, Labour, and Welfare after reorganization in 2001) for several national surveys in 1962, 1972, 1980, 1990, and 2000. The over-all response rate was over 79% in the relevant National Survey on Circulatory Disorders in 1980, 1 but the response rates in municipal area were lower than those in rural areas.
The high hazard ratio of the subjects with abnormal Q-QS findings showing the possible past history of myocardial infarction was observed even considering for sex, age, systolic blood pressure, blood glucose, and smoking habits. In the subjects with Q-S finding, proportion of death from heart disease was two third of all deaths and that was almost double to those in the reference national sample to give clear data the relationship of ECG findings objectively diagnosed to deaths from all causes. These data would be the base to prevent early deaths by intervening as a national project, such as the Health Japan 21 or Healthy people 2010 in the United State. Although we could not discuss much about the subjects group of small size, there would be a starting point for further studies.
The morbidity hazard ratio of specific disease, such as cerebral stroke and myocardial infarction, would be much more interesting along with mortality, but it was very hard to get such morbidity data of the subjects throughout country at the moment.
In this analysis, each specific ECG finding was evaluated using the Cox proportional hazard model for deaths of all cause, and any combination of ECG findings would show a much higher significant hazard ratio. Some of the further analysis on combination of ECG findings will come in our later papers.
In order to get an official permission to use the national database of deaths from the Japanese government, Masumi Minowa, M.D. (National Institute of Public Health) worked hard and deserved the authors' special appreciation. The authors would like to express many thanks to the contribution by many listed and unlisted collaborators with the principal investigators and associates listed in Appendixes II and I.