Sudden Death, Impaired Glucose Tolerance, and Diabetes in Japanese American Men
Background Diabetes and glucose intolerance have been shown to increase the risk of cardiovascular mortality in a number of different populations. Most studies have been based on short follow-up periods, and few have had sufficient numbers to allow researchers to look at sudden death as an outcome.
Methods and Results The relation of sudden death, defined as unexpected death occurring within either 1 or 24 hours of first symptoms, to glucose intolerance measured by a nonfasting 1-hour postload measurement made in 1965 or history of diabetes was examined by use of 23 years of follow-up on the 8006 participants enrolled in the Honolulu Heart Program. After adjustment for baseline body mass index, hypertension, cholesterol, triglycerides, smoking, alcohol consumption, and left ventricular hypertrophy or strain, the relative risks for sudden death within 24 hours in individuals with high-normal (151 to 224 mg/dL), asymptomatic high glucose values (≥225 mg/dL), and diabetes compared with those with lower glucose values (<151 mg/dL) were 1.59, 2.22, and 2.76, respectively. All these relative risks were statistically significant (P≤.05). Trends for sudden death in 1 hour were similar. Among men with sudden death <1 hour after onset of symptoms, the strength of the association between diabetes and sudden death was stronger among those classified as having died of unknown causes who thus were more likely to have died of an arrhythmia than among those classified as having died of coronary heart disease.
Conclusions The relations seen in these analyses indicate that individuals with glucose intolerance or diagnosed diabetes are at increased risk for sudden death.
Diabetes and impaired glucose tolerance have been shown to increase the risk of cardiovascular mortality in a number of different populations.1 2 3 4 5 6 However, most studies have been based on relatively short follow-up periods, most are in white populations, and few have had sufficient numbers of subjects or events to allow researchers to look at sudden death as an outcome. The Framingham Study is one of the few sources of long-term follow-up data on this relation. In that study, diabetes was associated with a doubling of atherosclerotic cardiovascular mortality. Diabetes had a greater influence on this relation in women than in men; its effect diminished with age and was strongly influenced by the presence of other risk factors.1 Diabetes was strongly associated with the 26-year incidence of sudden death in women but not in men.7 Data from the Honolulu Heart Program, a long-term prospective study of coronary heart disease (CHD) and stroke in Hawaii, provide an opportunity to examine these relations in a nonwhite population. Thus, this investigation was undertaken to examine the relation between glucose intolerance and diabetes and 23-year incidence of sudden death in Japanese American men in Hawaii.
Between 1965 and 1968, 8006 men 45 to 68 years of age who were of Japanese ancestry but were residents of the United States living on the island of Oahu in 1965 were examined at the baseline examination of the Honolulu Heart Program. In this report, up to 23 years’ worth of follow-up data for each subject is used to examine the relation between diabetes and impaired glucose tolerance and sudden death.
All men were followed through a comprehensive hospital and death record surveillance system since entry into the study. Follow-up for mortality is virtually complete. At the last examination of the cohort during 1991 through 1993, vital status could be confirmed on 8002 of the original 8006 men in the cohort. Although there were three subsequent examinations of the entire cohort during the 23 years of follow-up, no further blood glucose analyses were performed. Thus, “control” of diabetes cannot be ascertained. History of diabetes and history of use of insulin or other drugs for diabetes were obtained 2 and 6 years after the baseline examination.
The underlying cause of death listed on the death certificate is not assumed to be the cause of death in the Honolulu Heart Program. Abstracted death, hospital and autopsy records, and reports from the family, attending physician, or medical examiner when warranted are reviewed by a panel of physicians to determine cause of death. These records are also used to determine the period between onset of symptoms and death for classification of sudden death. The time period entered on the death certificate is also not assumed to be correct. Solid evidence is required to classify an individual as having died suddenly and to establish time of death.
Deaths are normally classified as CHD-related if there is evidence of chest pain or ECG or laboratory evidence of acute coronary disease. This committee classifies the cause of death as unknown when the individual dies or is found dead within a short period of time after being seen alive but with no immediately obvious evidence or symptoms of the disease process that resulted in the death. Deaths that require assumptions or speculation as to possible cause are generally coded unknown. This results in significant variation of the panel’s classifications from underlying cause of death derived from death certificates. It is likely that many of those deaths classified as unknown represent arrhythmic deaths, especially when the individual is known to have been asymptomatic less than 1 hour before death. The committee codes cases of documented acute ventricular arrhythmias without prior symptoms that result in death in individuals who were monitored for cardiac activity before death as unknown causes of death if there is no evidence that myocardial ischemia or another process was responsible for that condition. For unobserved deaths, time of death is estimated conservatively on the basis of information from medical records and family members regarding the last time an individual was seen alive and well. Thus, some cases of unobserved deaths due to CHD or unknown causes may well have occurred less than 1 hour after the onset of symptoms but are conservatively classified as sudden death <24 hours after the onset of symptoms, because of a lack of confirmation of the time of death.
Various definitions have been used for sudden death in published reports. For the purposes of these analyses, sudden death is defined as being classified as such by the Honolulu Heart Program physician panel, using criteria of nontraumatic death occurring suddenly or unexpectedly <1 or <24 hours after the onset of the terminal episode and resulting from CHD or unknown causes. Rapid demise resulting from other known causes is not included in these analyses.
At the baseline examination, subjects were interviewed regarding their medical histories and had brief medical examinations, 12-lead ECGs, three sitting blood pressure measurements, nonfasting 1-hour post–50 g glucose load serum glucose tests, and serum measurements of cholesterol, among other measurements. Although the nonfasting 1-hour postload glucose test used to screen this population in 1965 to 1968 undoubtedly reflects glucose tolerance status, it is not clear how individual levels compare with the standard 2-hour post–75 g load administered in the fasting individual, which is the standard today. Thus, for the purposes of this investigation, all eligible subjects were somewhat arbitrarily classified into four groups according to participant-reported history of diabetes and/or post–50 g load glucose values at the baseline examination (Table 1⇓). This classification scheme puts primary emphasis on the actual test of glucose intolerance and secondary emphasis on what was felt to be the less precise measure of participant recall of history of diabetes. The first group comprised those individuals with or without a history of diabetes who had what will be called a “low-normal” postload glucose, defined as a value of <151 mg/dL. The second group was called “high-normal,” defined as a postload glucose of between 151 and 224 mg/dL. The third, “asymptomatic-high,” was made up of individuals with no history of being diagnosed as having diabetes or taking diabetes medication but who had a 1-hour postload glucose of >225 mg/dL, a value that reflects approximately the 90th percentile in this population. The last group, the “diabetics,” was composed of those with a history of diabetes who were not taking diabetes medication but had a glucose level > 225 mg/dL and those taking diabetes medication regardless of glucose status on the postload measurement.
The Research and Institutional Review Committee of the Kuakini Medical Center approved the Honolulu Heart Program, and all subjects gave informed consent. The procedures followed were in accordance with institutional guidelines.
There were 417 men with evidence of prior heart disease or stroke. Glucose was not measured in 40 subjects at the initial examination. For most analyses, men in these two groups were eliminated. This left an at-risk cohort of 7549 men. Age-adjusted mean levels of cardiovascular risk factors were compared across the four glucose tolerance categories by use of a general linear models procedure from SAS8 9 with low-normal as the reference group. Age-adjusted incidence rates of sudden death per 1000 men were calculated with the entire study population as the standard. To determine whether diabetes or glucose intolerance was independently associated with sudden death over the next 23 years, the Cox proportional-hazards model was used, with the dependent variable being the time from the glucose tolerance test at baseline to death. Relative risks of sudden death for the high-normal, asymptomatic-high, and diabetes groups were calculated with the low-normal group as the reference group. These were adjusted for potential confounders, including age, hypertension, body mass index (BMI), serum cholesterol, serum triglycerides, smoking, alcohol, and left ventricular hypertrophy or strain. Analyses were also conducted in which individuals with prevalent cardiovascular disease were not excluded and in which relaxed definitions of sudden death based on time since onset of symptoms only and not cause of death were used.
After individuals with prevalent disease at baseline were eliminated, in follow-up averaging 23 years, 122 individuals suffered a nontraumatic death occurring suddenly or unexpectedly <1 hour after the onset of the terminal episode; 115 of these were attributed to CHD (43%) or unknown causes (53%) and 5 to stroke or aneurysm (4%) by the Honolulu Heart Program physician panel. By the less conservative World Health Organization criteria of death <24 hours after the onset of the terminal episode, 298 participants had sudden death; 244 of these were thought to be due to CHD (41%) or unknown causes (43%) and 46 (16%) to stroke, other cerebrovascular disease, or aneurysm.
Table 2⇓ shows the age-adjusted mean levels of selected variables by glucose tolerance status at baseline. The mean level of prevalence of smoking, hypertension, and alcohol intake is highest in the asymptomatic-high category for glucose level. The mean level of cholesterol, triglycerides, and BMI increases with worsening glucose tolerance status, with the highest mean levels found in known diabetics. As expected, when that mean was compared with the low-normal reference group, there were significant differences in most instances.
The Figure⇓ shows age-adjusted rates of sudden death in <1 and <24 hours over an average of 23 years of follow-up by the glucose tolerance states outlined above. As can be seen, the rates increase stepwise with worsening baseline glucose tolerance status from 21.0 per 1000 men in the low-normal group to 35.3 per 1000 in the high-normal, 54.4 per 1000 in the asymptomatic-high, and 69.4 per 1000 in known diabetics for the group in which sudden death occurred in <24 hours. For sudden death in 1 hour, a subset of the group in which sudden death occurred in <24 hours, the rates go from 11.8 per 1000 men in the low-normal group to 15.0 per 1000 in the high-normal, 28.5 per 1000 in the asymptomatic-high, and 31.6 per 1000 in known diabetics.
As Table 3⇓ shows, after adjustment, the relative risk of sudden death in < 24 hours for individuals with high-normal postload glucose was 1.59 compared with the low-normal group. For the asymptomatic-high postload glucose group, the relative risk was 2.22. For diabetics, the relative risk was 2.76. The 95% CIs did not include one for any of these relative risks. Relative risks for sudden death in 1 hour were 1.19, 2.05, and 2.24 for the high-normal, asymptomatic-high, and diabetic groups, respectively. Except for the high-normal group, the 95% CIs did not include one. As Table 3⇓ shows, the relative risks for sudden death in <1 hour among individuals with sudden death classified as resulting from unknown causes and thus more likely to have resulted from arrythmia were substantially greater than among those deaths classified as sudden death caused by CHD in <1 hour. For sudden death in <24 hours, the relative risks between those deaths caused by CHD and those of unknown causes were more comparable.
Three fifths of coronary fatalities have been said to occur outside the hospital, with the majority occurring suddenly.10 Thus, reduction of risk for sudden death is an important component of reduction of CHD mortality. The data presented here suggest that prevention of diabetes and impaired glucose tolerance may have important implications for the prevention of sudden death. The relations seen in these analyses indicate that individuals with impaired glucose tolerance or diagnosed diabetes are at increased risk for sudden death. This risk appears to be increased with increasing degree of abnormality. This graded effect lessens the likelihood that the associations seen here are artifactual.
The relations seen are apparent for sudden death defined as death within 24 hours of the onset of symptoms as well as within 1 hour. These relations appear to be independent of other risk factor abnormalities. Although the measure of glucose tolerance used in 1965 is not consistent with current standards, it probably reflects risk associated with impairment of glucose tolerance in this population. Associations between stroke and this measure of glucose tolerance are consistent with this assumption.11 It was not possible in this data set to assess the effect of control of diabetes or changes in glucose tolerance status on the outcome of sudden death. However, if there was an effect, the bias introduced by such temporal changes would be likely to weaken the apparent associations between impaired glucose tolerance and diabetes and sudden death. Thus, the associations reported in this paper might be expected to underestimate the real effects.
To investigate the effect of adding newly diagnosed diabetics to the diabetic group, an analysis was carried out that compared the predictive value of diabetes by history at baseline for sudden death in <1 hour to the predictive value of diabetes at the third examination, which then included incident new diabetes over 6 years. The relative risk of sudden death through 1988 for diabetics compared with nondiabetics at baseline after multivariate adjustment for multiple variables was 1.6. The relative risk from the third examination through 1988 was 1.5. Thus, in these analyses, although the number of cases is reduced and the CIs are wide, newly diagnosed diabetes appeared to have little impact on the relative risk of sudden death in addition to that attributable to prevalent diabetes.
Often, because of problems with the classification of the cause of death, such as having only death certificates available, definitions of sudden death use only the timing of onset of symptoms until death and do not take into account the presumed cause. The Honolulu Heart Program physician panel examines all available records, including autopsy reports, information from the family, hospital records, and death certificates, to determine cause and time of death. As noted previously, that panel determined that 96% of the sudden deaths that occurred <1 hour after onset of symptoms were CHD-related or of unknown causes. On the other hand, 16% of the sudden deaths occurring in <24 hours were thought to be due to causes other than CHD or unknown causes. Thus, the cause of death was somewhat more heterogeneous for sudden death in deaths occurring in <24 hours. Therefore, for the purposes of these analyses, we used only those cases of sudden death classified by the Honolulu Heart Program physician panel as due to CHD or unknown causes. Previous analyses of the Honolulu Heart Program data also indicated that the sudden death–risk factor relations among the total group of sudden deaths, defined by time since onset of symptoms, may be somewhat different for men who die in <1 hour and those who die in <24 hours.12 However, when comparable analyses in which less rigid definitions of sudden death were used and when prevalent cases of CHD and stroke were not excluded at baseline, the relations found were similar to those seen in the primary analyses for sudden death in both <24 hours and <1 hour. Individuals with diabetes are prone to earlier development of both macrovascular disease (atherosclerotic coronary, cerebral, and peripheral blood vessels) and microvascular disease (kidney, retina, and nerve). Both types of pathogeneses may contribute to an increased risk of sudden death. In a study of 1085 autopsies of Japanese American men in Hawaii, those classified as having had sudden death were associated with severe coronary atherosclerosis in at least one of the three main coronary arteries in 71% of cases compared with only 34% of nonsudden deaths.13 Davies and Thomas14 found that among individuals who died of ischemic heart disease in <6 hours, 74% were associated with coronary thrombosis. Among 48 of the 74% with thrombosis, the site was an area of a preexisting high-grade atherosclerotic lesion. However, others have found that <10% of autopsied sudden deaths had coronary thrombi.15 16 Diabetic individuals with prolonged QT intervals on ECG, a condition that may reflect autonomic neuropathy, do seem prone to sudden death.17 18
Variation in myocardial sensitivity to the development of ventricular arrhythmias is probably an important factor in sudden death, especially in death occurring in <1 hour, whether or not thrombosis is involved. Clinical diabetes has been associated with sudden deaths attributed to both arrhythmia and circulatory failure.19 The primary cause of sudden death in 142 closely observed men was the sudden development of a cardiac arrhythmia. This was the case in 91% of deaths in <1 hour and 85% of deaths in <24 hours.20 It is probable, however, that the pathogenesis of sudden death is a combination of atherosclerosis, thrombosis, and neural factors in many individuals. Thus, the propensity for sudden death in individuals with coronary disease may be enhanced by diabetes-associated autonomic neuropathy or microvascular disease affecting the conduction system of the heart. In these analyses, the fact that, among men with sudden death in <1 hour after onset of symptoms, the strength of the association between diabetes and sudden death is stronger for those participants classified as dying of unknown causes and thus more likely to have had an arrhythmic death than among those classified as dying of CHD would support this hypothesis. In addition, a stronger, more consistent relation was found when those participants with preexisting atherosclerotic diseases are excluded from the analyses; this would also seem to indicate that there may be a different balance of these factors in individuals with diabetes.
Thus, the data presented here indicate that existent diabetes and impaired glucose tolerance in middle-aged men are important predictors for long-term risk of sudden death. It seems likely that this is related not only to an increased propensity for fatal myocardial infarction but also to an increased risk for arrhythmic death in susceptible individuals.
This work was supported by contract NO1-HC-05102 from the NHLBI, Bethesda, Md.
- Received November 7, 1994.
- Accepted December 12, 1994.
- Copyright © 1995 by American Heart Association
Kannel WB, McGee DL. Diabetes and glucose tolerance as risk factors for cardiovascular disease: the Framingham Study. Diabetes Care. 1979;2:120-126.
Butler WJ, Ostrander LD, Carman WJ, Lamphiear DE. Mortality from coronary heart disease in the Tecumseh Study: long-term effect of diabetes mellitus, glucose intolerance and other risk factors. Am J Epidemiol. 1985;121:541-547.
Kannel WB, McGee DL. Epidemiology of sudden death: insights from the Framingham Study. In: Sudden Cardiac Death. Josephson ME, ed. Philadelphia, Pa: Cardiovascular Clinics, FA Davis Co; 1985:93-104.
SAS Institute, Inc. SAS/STAT User’s Guide, Version 6. 4th ed. Cary, NC: SAS Institute, Inc; 1990;2:891-996.
Hinkle LE, Whitney LH, Lehman EW, Dunn J, Benjiman B, King R, Plakun A, Flehinger B. Occupation, education and coronary heart disease. Science. 1968;160:238-246.
Burchfiel CM, Curb JD, Rodriguez BL, Abbott RD, Chiu D, Yano K. Glucose intolerance and 22-year stroke incidence: the Honolulu Heart Program. Stroke. 1994;25:951-957.
Kagan A, Yano K, Reed DM, MacLean CJ. Predictors of sudden cardiac death among Hawaiian-Japanese men. Am J Epidemiol. 1989;130:268-277.
Hinkle LE, Thaler HT. The clinical classification of cardiac deaths. Circulation. 1982;65:457-464.
Hinkle LE. Short-term risk factors for sudden death. <0kn;-1>Ann N Y Acad Sci. 1982;382:22-38.