Cumulative impact of axial, structural, and repolarization ECG findings on long-term cardiovascular mortality among healthy individuals in Japan: National Integrated Project for Prospective Observation of Non-Communicable Disease and its Trends in the Aged, 1980 and 1990

Abstract

Aims

Various cohort studies have shown a close association between long-term cardiovascular disease (CVD) outcomes and individual electrocardiographic (ECG) abnormalities such as axial, structural, and repolarization changes. The combined effect of these ECG abnormalities, each assumed to be benign, has not been thoroughly investigated.

Methods and Results

Community-dwelling Japanese residents from the National Integrated Project for Perspective Observation of Non-Communicable Disease and its Trends in the Aged, 1980–2004 and 1990–2005 (NIPPON DATA80 and 90), were included in this study. Baseline ECG findings were classified using the Minnesota Code and categorized into axial (left axis deviation, clockwise rotation), structural (left ventricular hypertrophy, atrial enlargement), and repolarization (minor and major ST–T changes) abnormalities. The hazard ratios of the cumulative impacts of ECG findings on long-term CVD death were estimated by stratified Cox proportional hazard models, including adjustments for cohort strata. In all, 16,816 participants were evaluated. The average age was 51.2 ± 13.5 years; 42.7% participants were male. The duration of follow up was 300,924 person-years (mean 17.9 ± 5.8 years); there were 1218 CVD deaths during that time. Overall, 4203 participants (25.0%) had one or more categorical ECG abnormalities: 3648 (21.7%) had a single abnormality, and 555 (3.3%) had two or more. The risk of CVD mortality increased as the number of abnormalities accumulated (single abnormality HR 1.29, 95% CI 1.13–1.48; ≥2 abnormalities HR 2.10, 95% CI 1.73–2.53).

Conclusions

Individual ECG abnormalities had an additive effect in predicting CVD outcome risk in our large-scale cohort study.

Introduction

The use of screening electrocardiography (ECG) in healthy individuals is controversial. The US Preventive Services Task Force (USPSTF) does not recommend the use of screening ECGs because of insufficient evidence of any benefit.¹ Similarly, the American Heart Association guidelines consider screening ECGs for cardiovascular disease (CVD) risk assessments in asymptomatic individuals reasonable only in those with risk factors, such as hypertension and diabetes.² In Japan, various recommendations coexist. Screening ECGs are recommended as part of the worksite annual health check up, but in community-based medical examinations, they are only recommended for individuals with a high-risk profile.³ The reason for this inconsistency is partly due to a lack of studies that have assessed the impact of frequently seen, but nonspecific, ECG findings.

The performance of screening ECGs have been studied previously in various situations, but their obvious prognostic implications remains limited to rarely encountered, significant ECG findings in healthy individuals. These finding include ST elevations/depressions, abnormal Q-waves, high-degree atrioventricular blocks, or arrhythmias such as atrial fibrillation, atrial flutter, and Wolff-Parkinson-White (WPW) syndrome. Large-scale cohort studies, including ours, have demonstrated statistically significant but clinically negligible associations between long-term CVD outcomes and nonspecific ECG findings, including major and minor ST–T segment abnormalities, left ventricular hypertrophy, left axis deviations, and clockwise rotations.^4–15 Despite this statistical association, the additional benefit that screening ECGs provide to asymptomatic individuals and whether they reclassify a patient’s CVD risk remain controversial.

We hypothesized that the presence of these nonspecific, but highly frequent, ECG findings has an additive and clinically meaningful impact on long-term CVD outcomes. To assess the cumulative impact of nonspecific ECG abnormalities, we analysed data from the National Integrated Project for Prospective Observation of Non-Communicable Disease and its Trends in the Aged, 1980–2004 and 1990–2005 (NIPPON DATA80 and 90). These were cohort studies, with large sample sizes, of healthy Japanese individuals. Assessing the cumulative risk of screening ECG abnormalities was thought to possibly help identify the individuals who are at higher risk of CVD events; this would help establish the potential benefit of screening ECGs in asymptomatic individuals.

Methods

Participants

NIPPON DATA80 and 90 were studies conducted by the National Survey on Circulatory Disorders, Japan. The two cohort studies performed baseline surveys in 1980 and 1990, respectively. The details of these cohort studies have been previously reported.^13,16–27 Approval for the present study was obtained from the institutional review board of Shiga University of Medical Science (no. 12-18, 2000).

In this study, we analysed integrated data from NIPPON DATA80 and 90; a total of 18,929 healthy participants (10,546 from NIPPON DATA80 and 8383 from NIPPON DATA90) were included. There were 8143 men and 10,786 women, all aged 30 years or more, from 300 randomly selected districts throughout Japan. Participants were followed from 1980 through 2004 in NIPPON DATA80, and from 1990 through 2005 in NIPPON DATA90. The data consisted of patient histories, physical examination results, blood test results, standard 12-lead ECG recordings, and self-administered lifestyle questionnaires.

We excluded 2113 of the 18,929 participants for the following reasons: absence of a permanent address that was needed to link to vital statistical records (n = 1388; 1104 from NIPPON DATA80 and 284 from NIPPON DATA90); missing information in the baseline survey (n = 118; 2 from NIPPON DATA80 and 116 from NIPPON DATA90); history of known myocardial infarction or stroke (n = 392; 153 from NIPPON DATA80 and 239 from NIPPON DATA90); and specific ECG findings, including a moderate or severe Q-wave abnormality (Minnesota Code, MC, 1-1, 1-2), complete atrioventricular block (MC 6-1), WPW syndrome (MC 6-4), or atrial fibrillation or flutter (MC 8-3-1 or 8-3-2) (n = 215; 122 from NIPPON DATA80 and 93 from NIPPON DATA90). The final sample comprised 16,816 participants (Figure 1).

Baseline examination

At the time of the baseline survey (1980 for NIPPON DATA80 and 1990 for NIPPON DATA90), a standard 12-lead ECG was recorded, with each patient in the supine position. A Working Group of ECG Coding for the National Survey on Circulatory Disorders evaluated the electrocardiograms; two independently trained coders in accordance with the MC guidelines.^{13,16,17,20,25} In cases of discordant results, the final judgment was made by a panel of study epidemiologists and cardiologists.

After exclusion of the above-mentioned, significant ECG findings that required medical attention, the remaining ECG findings were classified into three categorical abnormalities based on their electrophysiological perspectives: axial (left axis deviation or clockwise rotation; MC 2-1 or 9-4-2), structural (left ventricular hypertrophy or atrial enlargement; MC 3-1, 3-3, 9-3-1, or 9-3-2), and repolarization (minor and major ST–T changes; MC 4-1, 4-2, 4-3, 4-4, 5-1, 5-2, 5-3, or 5-4). Patients were classified into three groups: no ECG abnormality, a single categorical abnormality, and ≥2 categorical abnormalities.

Baseline blood pressures were measured in the right arm of seated participants in both cohorts by a trained public health nurse, using a standard mercury sphygmomanometer. Hypertension was defined as a systolic blood pressure of ≥140 mmHg, a diastolic blood pressure ≥90 mmHg, use of antihypertensive agents, or any combination of these. Nonfasting blood samples were drawn, centrifuged within 60 minutes of collection, and stored at –70℃ until analysis. Serum total cholesterol was measured by the Liebermann-Burchard direct method in NIPPON DATA80²³ and enzymically in NIPPON DATA90.²⁷ Serum glucose was measured by the neocuproine method in NIPPON DATA80.²⁴ Because the current standard method of blood glucose levels is the hexokinase method, the neocuproine serum glucose levels were adjusted by the following formula: 0.047 × (glucose concentration in mg/dl) – 0.541.24 Plasma glucose was measured enzymically in NIPPON DATA90.¹⁸ Diabetes mellitus was defined as a plasma or serum glucose level of ≥11.1 mmol/l, the use of medications for diabetes mellitus, or both. Serum creatinine was measured using the alkaline picric acid method (Jaffe’s method) in both cohorts. Body mass indexes were calculated as the patient’s weight in kg divided by the square of his/her height in metres. Public health nurses obtained medical histories, including each patient’s smoking and alcohol consumption status.

Follow-up survey

The National Vital Statistics Database of Japan was used to obtain the underlying causes of death for patients who died during the study period, with permission from the Management and Coordination Agency, Government of Japan. The causes of death were coded in accordance with the International Classification of Diseases ninth revision (ICD-9) through to the end of 1994 and the ICD tenth revision (ICD-10) from the beginning of 1995. The details of the cause-of-death classification in the present study are described elsewhere.^13,16–27

Statistical analysis

We compared the hazard ratios among the three groups (no ECG abnormality, a single categorical abnormality, and ≥2 categorical abnormalities). To compare baseline characteristics, the chi-squared test or Fisher’s exact test was used for categorical variables and one-way analysis of variance (ANOVA) was used for continuous variables. Event-free survival was estimated by the Kaplan–Meier method and statistical differences were evaluated using a Log-rank test. Cox proportional hazards models were used to examine the cumulative effects of ECG categorical abnormalities on CVD outcomes. We analysed the integrated data of the NIPPON DATA80 and 90 cohorts to ensure sufficient numbers of events. Baseline hazards between the study cohorts were allowed to be different by including an adjustment for cohort strata in the Cox models.²⁸

The covariates included in the model were age, gender, body mass index, smoking status (never or past versus current smoker), diabetes mellitus, systolic blood pressure, total cholesterol, and serum creatinine. These covariates were clinically related to CVD events and significantly associated with CVD deaths in our univariate analysis (p < 0.05). All p-values were 2-sided, and the significance level was set at p = 0.05 for all analyses. All statistical analyses were performed using Statistical Package for the Social Sciences version 20 (SPSS, Chicago, IL, USA).

Results

A total of 16,816 participants were analysed. The age of the participants was (mean ± SD) 51.2 ± 13.5 years, and 42.7% were male. The total duration of follow up was 300,924 person-years (mean 17.9 ± 5.8 years). There were 1218 CVD deaths during the follow-up period: 248 due to coronary artery disease, 239 due to heart failure, and 548 due to stroke events. Overall, 4203 participants (25.0%) had one or more categorical ECG abnormalities: 3648 (21.7%) had a single categorical abnormality (1127 axis abnormalities, 1808 structural abnormalities, 713 repolarization abnormalities) and 555 (3.3%) had two or more categorical abnormalities (Figure 1).

Table 1 shows the baseline characteristics of the participants, according to the number of categorical ECG abnormalities. The comparison of the baseline characteristics of NIPPON DATA80 and 90 are described in Supplementary Table 1 (available online). The number of abnormalities increased with increasing age, systolic and diastolic blood pressure values, fasting blood glucose levels, and creatinine levels. Furthermore, the number of abnormalities was also correlated with the prevalence of hypertension, diabetes mellitus, and current smoking.

Figure 2 shows Kaplan–Meier survival curves for CVD deaths, according to the numbers of categorical ECG abnormalities. As indicated in the curve, participants with single or more categorical ECG abnormalities at baseline had a significantly lower survival rate during the follow-up period, and individual ECG categorical abnormalities had an additive effect on CVD mortality (Log-rank test; p < 0.001).

Table 2 shows the adjusted hazard ratios for each outcome. After adjusting for confounding factors, the risk of CVD mortality increased as the number of categorical ECG abnormalities accumulated (HR 1.29, 95% CI 1.13–1.48 in patients with a single abnormality; HR 2.10, 95% CI 1.73–2.53 in patients with ≥2 abnormalities). The prognostic impact of the accumulated categorical ECG abnormalities was significant in all types of CVD deaths. Each study cohort (NIPPON DATA80 versus 90) demonstrated the same tendencies seen in the overall analysis (Supplementary Table 2), although there were insufficient event numbers to draw conclusive results from the individual cohorts.

Figure 3 shows the adjusted hazard ratios for CVD mortality in each group. The presence of axial, structural, and repolarization abnormalities was independently associated with adverse CVD events (HR 1.30, 95% CI 1.05–1.61 for axial abnormalities; HR 1.25, 95% CI 1.04–1.50 for structural abnormalities; HR 1.37, 95% CI 1.11–1.70 for repolarization abnormalities). A cumulative effect of ECG abnormalities on increasing CVD mortality was also identified (HR 2.10, 95% CI 1.73–2.53 for ≥2 categorical abnormalities).

Discussion

Individual, nonspecific ECG abnormalities were shown to have a cumulative impact on predicting the risk of CVD outcomes in a Japanese population without a history of myocardial infarction or stroke. This impact was applicable to all types of CVD mortalities, including coronary heart disease, heart failure, and stroke. The cumulative effect of individual ECG abnormalities may be useful for risk stratification in healthy individuals.

Whether screening ECGs provide benefits to asymptomatic individuals, or whether they aid in reclassifying their CVD risk, remains controversial.^1–3 CVD is the leading cause of death throughout the world²⁹ and is often asymptomatic before the first cardiac event, which may consist of sudden cardiac death, myocardial infarction, or heart failure.³⁰ In Japan, regular health checkups are widely conducted, particularly among the working population. Screening ECGs are mandated for all employees, under legal regulations on industrial safety and health.³ In contrast, no regulations or guidelines mandate routine ECG screenings for individuals who are self-employed, homemakers, or retired. In 2004, the USPSTF recommended against screening with resting ECGs in asymptomatic individuals.¹ This recommendation was based on the lack of clinical trials evaluating outcomes after ECG screenings and on the concept that abnormalities in resting ECGs are unlikely to change management decisions for persons who are already classified on the basis of traditional risk factor assessments.³¹ The results of our study, that a single ECG abnormality has a minimal impact on CVD outcomes, would seem to agree with the USPSTF recommendation. However, we also observed that cumulative ECG abnormalities have a strong impact on CVD outcomes; this may indicate a potential benefit for screening ECGs.

Unlike findings in other studies, our findings suggest that adding up the relatively benign ECG findings of axial, structural, and repolarization abnormalities to obtain a cumulative number may be more beneficial than focusing on the individual ECG findings. Many previous studies addressed the concept of ‘major’ and ‘minor’ abnormalities in screening ECGs, and the impact of these abnormalities on CVD outcomes;^{4,11,14,32–34} however, these studies had several limitations. First, the definitions of ‘major’ and ‘minor’ abnormalities varied.³¹ Second, these studies did not address some ECG abnormalities, such as clockwise rotations, which have recently been reported as important. Third, the same categorical abnormalities, such as ST–T changes or T-wave inversions, were included in both ‘major’ and ‘minor’ abnormality classifications. Therefore, the additive effect of ECG abnormalities was not assessed. As some changes were assigned to both the ‘major’ and ‘minor’ categories, adding these categories would have led to duplication and misrepresentation of the additive effect. Our new concept of using cumulative ECG abnormalities may help physicians to understand the clinical significance of these abnormalities and to stratify their potential hazards.

Limitations

Our study has several limitations. Although the current study involved only Japanese participants, previous studies have shown that the prognostic value of several ECG findings, proven in the USA and Europe, was also applicable to Japanese populations.^16,20,25 Second, we used a single, baseline ECG result for analysis. Single biological measurements are known to be subject to variability and the observed ECG abnormalities may have changed over time. Third, although we had independently trained ECG readers to code each tracing, automated measurement systems for ECGs were not available at the time of the baseline evaluations (1980 and 1990). Therefore, a panel of experts, including epidemiologists and cardiologists, confirmed the findings. Fourth, we did not analyse a single cohort but the integrated data from two cohorts. However, as discussed above, the two study populations showed the same trends in individual analyses as in the combined analysis, and the ECG findings are objective and, therefore, insensitive to differences between cohorts. Fifth, only a single underlying cause of death is described in the National Vital Statistics Database of Japan, which was used to obtain the causes of death in this study. For this reason, the mortality statistics for coronary heart disease may have been underestimated because deaths coded as ‘heart failure’ may have hidden some coronary events.³⁵ However, we believe this does not affect our results since we also assessed ‘heart failure’ as an endpoint, and the present study focused mainly on all cardiovascular mortality. Finally, the present study did not show the actual reclassification based on the results of screening ECGs, according to the previously established risk stratifications (e.g. Framingham risk scores). However, using Framingham risk scores is probably not helpful, because they have been proven to not work well in Japanese populations, in which the age-adjusted mortality due to coronary artery disease is the lowest amongst all developed countries.³⁶ The new guidelines of the Japan Atherosclerosis Society (JAS) do not use Framingham risk scores for absolute risk assessment;³⁷ instead, they use the NIPPON DATA risk chart.³⁸ An assessment of how accurately the model containing nonspecific ECG findings classifies subjects into risk categories per the JAS guidelines may be feasible; however, such an analysis was not within the scope of the present study.

Conclusions

We found that individual nonspecific ECG abnormalities had an additive effect on predicting the risk of CVD outcomes in this large-scale Japanese cohort. This cumulative impact was notable for all types of CVD mortality, indicating the potential benefit of screening ECGs in healthy individuals.

Declaration

Dr Inohara had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Acknowledgements

The authors wish to extend our appreciation to the members of the NIPPON DATA80/90 Research Group, who are listed in a supplementary file (available online).

Funding

This study was supported by a grant-in-aid by from the Ministry of Health, Labor, and Welfare under the auspices of the Japanese Association for Cerebro-Cardiovascular Disease Control, a research grant for cardiovascular diseases (7A-2) from the Ministry of Health, Labor, and Welfare, and research grants from Health and Labor Sciences (Comprehensive Research on Aging and Health H11-Chouju-046, H14-Chouju-003, H17-Chouju-012, H19-Chouju-Ippan-014; Comprehensive Research on Life Style-Related Diseases Including Cardiovascular Diseases and Diabetes Mellitus H22-Jyunkankitou-Seisyu-Sitei-017 and H25-Jyunkankitou-Seisyu-Sitei-022).

Conflict of interest

The authors declare that there is no conflict of interest.

References