2016 Volume 80 Issue 11 Pages 2343-2348
Background: The aim of this study was to examine whether the burden of diabetes on cardiovascular disease (CVD) in Japan has increased in recent years.
Methods and Results: Three cohorts were established, consisting of Japanese residents aged 40–69 years, in 1992–1995 (n=8,744), 1996–1999 (n=7,996), and 2000–2003 (n=7,273). All participants had follow-up for a median of 10 years. Diabetes was defined according to the following criteria: (1) fasting serum glucose ≥7.0 mmol/L; (2) non-fasting serum glucose ≥11.1 mmol/L; or (3) anti-diabetic treatment at baseline. During follow-up, the number of CVD incidents was 277 in the first, 214 in the second, and 190 in the third cohorts. The prevalence of diabetes increased slightly over time. Adjusting for traditional cardiovascular risk factors, multivariable HR (95% CI) for diabetes as a cardiovascular risk factor were 1.40 (0.91–2.14) in the first, 1.93 (1.25–3.00) in the second, and 2.59 (1.77–3.81) in the third cohorts. The population attributable fraction of CVD due to diabetes was 2.8%, 5.6%, and 12.4%, respectively.
Conclusions: This is the first study in middle-aged Japanese people to clarify an increased burden of CVD due to diabetes since the early 1990 s. Further efforts are needed to prevent and control diabetes through lifestyle modification and treatment. (Circ J 2016; 80: 2343–2348)
Nine percent of adults aged ≥18 years had diabetes in 2014, and worldwide 347 million people are estimated to have diabetes. According to the 2012 World Health Organization (WHO) statistics, diabetes causes 1.5 million deaths and is one of the leading causes of non-communicable disease. The prevalence of diabetes has been increasing gradually since the 1980 s and this rise is largely driven by physical inactivity and obesity.1 However, the relatively larger burden of diabetes in Asia than in Western countries was not fully explained by obesity.2 There are differences in the prevalence of diabetic complications and their related factors between Asian and European populations.3
Editorial p 2293
Diabetes is a strong risk factor for coronary artery disease (CAD) and ischemic stroke,4,5 as well as heart failure.6 A few studies have examined the population attributable fraction (PAF) of cardiovascular disease (CVD) due to diabetes and reported it to be at most 10%,4,7 and the impact might be smaller in comparison with that of hypertension, at approximately 50%.7 In the present study, we investigated the trends in the prevalence of diabetes and the PAF of CVD using three 10-year cohorts from the 1990 s to the 2000 s.
The present analysis included residents aged 40–69 years in four communities of the Circulatory Risk in Communities Study (CIRCS):8 (1) Ikawa town, Akita Prefecture; (2) Minami-Takayasu district, Yao City, Osaka Prefecture; (3) Noichi town, Kochi Prefecture; and (4) Kyowa town, Ibaraki Prefecture. We defined the first cohort as 1992–1995, the second cohort as 1996–1999, and the third cohort as 2000–2003. The follow-up was terminated at the end of 2002 in the first cohort, 2007 (2005 in Kochi) in the second cohort, and 2012 (2005 in Kochi and 2010 in Ibaraki) in the third cohort. The number of participants was 8,973 (3,325 men, 5,648 women) in the first, 8,145 (2,982 men, 5,163 women) in the second, and 7,398 (2,660 men, 4,738 women) in the third cohort. After the exclusion of participants with a history of CAD or stroke at baseline, we analyzed the data from 8,744 individuals (3,178 men, 5,566 women), 7,996 individuals (2,889 men, 5,107 women), and 7,273 individuals (2,575 men, 4,698 women), respectively.
The study protocol conformed to the Declaration of Helsinki and received ethics approval by the institutional review boards of the Osaka Center for Cancer and Cardiovascular Disease Prevention and of Osaka University.
At baseline, blood was drawn from seated participants into a plain, siliconized glass tube and the serum was separated. Serum glucose was measured using enzymatic methods with the SMAC automatic analyzer (Technicon, Tarrytown, NY, USA) in 1992, Hitachi 7250 (Hitachi Medical, Ibaraki, Japan) in 1993–2000 and AU2700 (Olympus, Tokyo, Japan) in 2001–2003, under consistent and comparable quality control at the laboratory of the Osaka Medical Center for Health Science and Promotion. Fasting ≥8 h was not required prior to blood drawing. Serum glucose was classified into three categories: diabetes, pre-diabetes, and normal. Diabetes was defined as fasting glucose ≥7.0 mmol/L, non-fasting glucose ≥11.1 mmol/L, and/or anti-diabetic medication. Pre-diabetes was defined as fasting glucose 6.1–6.9 mmol/L, or non-fasting glucose 7.8–11.0 mmol/L. All others were classified as normal.
In the current study, CVD was defined as CAD (myocardial infarction [MI], effort angina, or sudden cardiac death) and stroke. Incidents of CVD were ascertained from at least one of the following: death certificate; national health insurance claim; report from local physicians, public health nurses and health volunteers; annual cardiovascular risk survey; or household questionnaire. To confirm the diagnosis of CVD, we called, visited or invited the subjects or their families to obtain information of the symptoms and time course at onset. We then reviewed the medical records at local clinics and hospitals. Several physician epidemiologists independently determined whether each case was a definite or suspect case of CAD or stroke by reviewing available information, blinded to the data of the annual health check-up.
Diagnosis of MI was based on modified WHO criteria for CAD.9 Individuals were diagnosed with definite MI if they met the two following criteria: (1) typical severe chest pain (lasting >30 min); and (2) new abnormal Q or QS waves on electrocardiography, or consistent changes in cardiac enzyme levels. Probable MI was defined as meeting criterion (1) but not (2). Effort angina was defined as repeated episodes of chest pain during effort, especially when walking, usually disappearing rapidly after the cessation of effort or with sublingual nitroglycerin. Other than MI or effort angina, deaths that occurred within 1 h of onset were regarded as sudden cardiac death. CAD was defined as including definite or probable MI, effort angina, or sudden cardiac death. For sensitivity analysis, we included percutaneous coronary intervention (PCI) and coronary artery bypass grafting (CABG) accompanied by atypical chest pain as CAD. Stroke was diagnosed if individuals had focal neurological symptoms with rapid onset, persisting for at least 24 h or until death. Stroke was classified into ischemic or hemorrhagic stroke, primarily based on computed tomography or magnetic resonance imaging, available in approximately 90% of cases.
Serum total, high-density lipoprotein cholesterol (HDL-C) and triglycerides were measured using standardized methods10,11 at the laboratory of the Osaka Medical Center for Health Science and Promotion, an international member of the Cholesterol Reference Method Laboratory Network.
Blood pressure was measured by trained physicians using standard mercury sphygmomanometers and standardized epidemiological methods.8 Hypertension was defined as either systolic blood pressure (SBP) ≥140 mmHg, diastolic blood pressure ≥90 mmHg or use of antihypertensive medication. Height was measured with the subjects in stocking feet, and weight was measured while wearing light clothing. Body mass index (BMI) was calculated as weight (kg) divided by the square of height (m2). Overweight was defined as BMI ≥25 kg/m2. Obesity was defined as BMI ≥30 kg/m2.
Interview was conducted to ascertain smoking status, usual alcohol intake per week and the use of medications for diabetes, hypertension and/or hypercholesterolemia.
The prevalence of diabetes was calculated on analysis of covariance (ANCOVA) adjusting for age and sex, and was tested for time trends across cohorts using logistic regression. To compare baseline characteristics across the glucose categories, ANCOVA was used for the means, and chi-squared test for proportions. Given that serum triglyceride was affected by fasting status, the levels were adjusted for the time since the last meal as well as for age and sex. The time trend of the baseline characteristics across cohorts was evaluated in the same glucose category of each cohort using linear regression with time (0, 1, and 2 for three cohorts).
Person-years for each individual were calculated as the duration from the date of enrollment in the cohort until the incidence date of CVD, the date of death, or the date of failure to follow-up, whichever came first. The incidence of first-ever CVD was calculated by the person-year method, and was standardized to the distribution of age and sex in the Japanese population in 1995 by the direct method. Incident curves were calculated to estimate the cumulative incidence of CVD by serum glucose category for each cohort. Hazard ratios (HR) and 95% CI of CVD were calculated for diabetes and pre-diabetes using Cox proportional hazards regression compared with normal individuals. In model 1, we adjusted for age (years) and sex; and in model 2 (multivariable-adjusted model), we further adjusted for BMI category (quartiles in each cohort), serum total cholesterol category (quartiles in each cohort), serum triglycerides (mmol/L), SBP (mmHg), anti-hypertensive medication use (no or yes), cigarette smoking (never, former or current smoker), alcohol intake (never, former, or current drinker [<46, 46–68 or ≥69 g ethanol/day]), time since the last meal (0–<1, 1–<2, 2–<3, 3–<4, 4–<8 or ≥8 h), and community.
We calculated the PAF of CVD for diabetes and pre-diabetes, which is the proportion of CVD events in the population that would be attributable to diabetes or pre-diabetes. We used a category-specific attributable fraction with the formula pdi (1–1/RRi), which produces internally valid estimates when confounding exists,12 where pdi is the proportion of total cases in the population arising from the ith exposure category and RRi is the multivariable-adjusted HR for the ith exposure category relative to the unexposed group. We also calculated approximate estimates of 95% CI for the PAF.13 The analysis was repeated, stratified by treated status for diabetes.
All statistical analysis was performed with SAS (version 9.4; SAS Institute, Cary, NC, USA). All P-values for statistical tests were 2-tailed, and P<0.05 was regarded as statistically significant.
The age- and sex-adjusted prevalence of diabetes at baseline was 4.4% in the first, 4.8% in the second, and 5.6% in the third cohorts (P for trend <0.001). The corresponding prevalence of pre-diabetes was 8.9%, 8.4% and 8.6% (P for trend=0.40).
Table 1 lists the baseline characteristics for all three cohorts. In all cohorts, compared with the normal glucose category, pre-diabetes and diabetes were positively associated with age, BMI, SBP, serum triglycerides, and the prevalence of overweight, obesity, hypertension and anti-hypertensive medication use. Serum total cholesterol and the prevalence of current smokers were higher for diabetes in the first cohort. Serum HDL-C was lower and the prevalence of lipid-lowing medication use was higher for diabetes in the second and the third cohorts.
Data adjusted for age and sex. Serum triglycerides were also adjusted for time since the last meal. BMI, body mass index; DBP, diastolic blood pressure; HDL-C, high-density lipoprotein cholesterol; SBP, systolic blood pressure.
In each glucose category, mean age, serum total cholesterol and HDL-C, and the prevalence of lipid-lowering medication use increased from the first to the third cohorts (Table 1). Mean BMI did not differ among the three cohorts both in the normal and diabetic categories, but it increased in the pre-diabetic category over time. In the diabetic category, mean SBP decreased from the first to the third cohorts, whereas the prevalence of anti-hypertensive medication use and hypertension did not change substantially over time. In each category, the prevalence of current smokers did not change over time.
The median follow-up period for each cohort was 10.1 years. Only 5.8% of the participants in the first cohort, 5.2% in the second cohort, and 3.4% in the third cohort were lost to follow-up on censored death or moving out of the communities.
Figure shows the cumulative incidence of CVD according to glucose category over the decade of follow-up in the three cohorts. Diabetic patients had the highest incidence of total CVD compared with the pre-diabetic and normal glucose categories in each cohort. Table 2 lists the incidence rate and HR of CVD according to serum glucose category in the three cohorts. The age- and sex-adjusted incidence rates of total CVD in normal individuals decreased by 32%, whereas in diabetic subjects the total CVD incidence rates increased by 33% from the first to the third cohorts. Among diabetic subjects, the CAD incidence rate was 1.8-fold higher in the third cohort than in the first cohort.
Cumulative incidence of cardiovascular disease (CVD) according to serum glucose category (normal, pre-diabetic and diabetic) during 10 years of follow-up in (A) the first cohort (1992–1995); (B) the second cohort (1996–1999); and (C) the third cohort (2000–2003).
CAD, coronary artery disease; CVD, cardiovascular disease.
Compared with normal glucose category, the age- and sex-adjusted risk of total CVD for diabetes was 2.0-fold higher in the first and second cohorts and 3.0-fold higher in the third cohort. Although further adjustment for traditional cardiovascular risk factors attenuated the excess risk of total CVD with diabetes, the association remained statistically significant in the second and third cohorts. The multivariable HR (95% CI) of total CVD for diabetes were 1.40 (0.91–2.14) in the first cohort, 1.93 (1.25–3.00) in the second cohort, and 2.59 (1.77–3.81) in the third cohort. When PCI and CABG were included as outcomes, the corresponding HR (95% CI) of total CVD were 1.40 (0.92–2.12), 2.28 (1.52–3.42) and 2.83 (1.97–4.05), respectively. The multivariable HR of CAD and ischemic stroke for diabetes showed similar upward trends. There was no significant excess risk of hemorrhagic stroke for diabetes in three cohorts.
The PAF of total CVD for diabetes was 2.8% in the first, 5.6% in the second, and 12.4% in the third cohort (Table 3). The PAF for diabetes was 16.2% for CAD and 11.1% for total stroke, and, notably, 15.9% for ischemic stroke in the third cohort. The PAF of total CVD and other outcomes for pre-diabetes did not change consistently from the first to the third cohorts.
PAF, population attributable fraction. Other abbreviations as in Table 2.
The PAF of total CVD for both untreated and treated diabetes showed upward shifts, with the PAF for untreated diabetes at –0.4% (95% CI, –3.3 to 2.4) in the first cohort, 2.2% (95% CI, –1.2 to 5.6) in the second cohort, and 2.8% (95% CI, –1.1 to 6.5) in the third cohort, and the corresponding PAF for treated diabetes at 3.0% (95% CI, 0.3–5.7), 3.4% (95% CI, 0.2–6.5) and 9.6% (95% CI, 4.4–14.5).
In middle-aged Japanese people, the population burden of diabetes for total CVD increased significantly from the late 1990 s through the early 2000 s, and one-ninth of total CVD incidents were attributable to diabetes in the early 2000 s. There was a slight increase in the prevalence of diabetes, and a substantial increase in relative risk of total CVD for diabetes over time.
This is the first study in Asia to investigate time trends for the excess risk of CVD events associated with diabetes. From the first to third cohorts, diabetic individuals had decreased mean glucose and increased prevalence of anti-diabetic medication use. Between the second and the third cohorts, diabetic individuals had decreased mean SBP and triglycerides. Mean serum total cholesterol and prevalence of lipid-lowering medication, however, increased over time. Mean glucose was not different between the cohorts, despite the increased prevalence of diabetes, because the proportion of treated diabetes increased over time. Therefore, a net effect of these risk factor changes on total CVD is uncertain. It is possible that the increased risk of CVD among the diabetes was attributable to the elongation of the duration of diabetes, although not examined in the present cohorts. There is also the increasing chance of detection of CVD in the treated diabetes subjects.
In the USA, the Framingham Study reported that the age- and sex-adjusted HR (95% CI) of CVD for diabetes did not change substantially: 3.0 (2.3–3.9) in 1952–1974 and 2.5 (1.9–3.2) in 1975–1998.14 They showed that the impact of diabetes as a CVD risk factor was not declining and there had been a significant increase in the prevalence of diabetes: 8.1% in 1952–1974 and 14.6% in 1975–1998. The PAF of CVD for diabetes then increased from 5.4% to 8.7%, respectively. In Finland, the PAF of CVD for diabetes increased in men from 11.4% in 1992 to 13.8% in 2002 along with the increased prevalence of diabetes from 2.3% to 4.1%, while the PAF in women declined from 20.1% to 16.9%, despite the increase in the prevalence of diabetes from 2.5% to 3.2%.15 They used the data from the Finish national registers and did not include untreated diabetes. In the present study the HR of total CVD for diabetes increased steadily from 1992–1995 to 2000–2003, and the prevalence of diabetes slightly increased. As a result, the PAF of CVD for diabetes increased from 2.8% to 12.4%. The rising trend for the prevalence of diabetes in the general population was also reported from the Hisayama Study: the age-adjusted prevalence of type 2 diabetes increased from 14.6% in 1988 to 20.8% in 2002 in men, and from 9.1% to 11.2% in women.16
The strengths of the present study include the large population-based sample of middle-aged Japanese subjects, and the use of standardized methods for the measurement of serum glucose and cardiovascular risk factors. Bias for the exposure variables can be minimized due to the cohort design. The CVD events were routinely ascertained and a high percentage of stroke events were confirmed using imaging.
This study has several potential limitations. First, the criteria for the diagnosis of diabetes at general clinical cites were updated in accord with the guideline revision by the Japan Diabetes Society in 1999.17 The criteria change may have affected the increase in the prevalence of treated diabetes in particular between the second and third cohorts because the cut-off for the definition of diabetes shifted from 7.8 mmol/L to 7.0 mmol/L. In the present study, however, mean glucose level in treated diabetes subjects did not differ substantially between the two cohorts: mean glucose at fasting was 8.8 mmol/L in the second cohort to 8.9 mmol/L in the third cohort (P for difference=0.82), and that at non-fasting was 10.8 mmol/L to 10.1 mmol/L, respectively (P for difference=0.29). Thus, overtreatment due to criteria change was unlikely to occur. Second, we relied on a single glucose measurement because glycated hemoglobin was not measured in the first cohort, or oral glucose tolerance test in any of the three cohorts. We also relied on the WHO diagnostic criteria using impaired fasting and non-fasting glucose levels for pre-diabetes and diabetes.18 The criteria are liable to misclassification, which may dilute the associations of pre-diabetes and diabetes with risk of CVD. Quality control, however, was maintained for the laboratory data across the three automatic analyzers, and the definition of serum glucose category was consistent between the cohorts. Third, we did not require the presence of symptoms for the definition of diabetes because systematic and unbiased collection of symptoms would be difficult in epidemiological studies. Thus, the present classification system is more likely to identify false positives of diabetes, and would weaken the association between diabetes and risk of CVD. The extent of misclassification of the categories, however, would be similar across the cohorts, and therefore such misclassification would not have substantially altered the conclusions. Fourth, there may be an increasing chance of detection with the increased proportion of treated diabetes because of their management, leading to early detection of CVD. Last, this study used data for middle-aged Japanese men and women, and it is not clear whether the results are generalizable to other ethnic groups.
The Western Pacific countries with increased burden of CVD19 have more people with diabetes than any other region in the world, and the prevalence of diabetes was projected to increase by 71% in South-East Asia and 46% in the Western Pacific by 2035.20 The increasing PAF of CVD due to diabetes noted in the present study could serve to predict the increasing burden in other Asian countries in the future.
In conclusion, the present population-based cohort study has shown that the proportion of CVD attributable to diabetes has increased during three survey periods between 1992 and 2003. Further efforts are needed to prevent and control diabetes through lifestyle modification and treatment in Japan.