All-cause mortality stratified by history of diabetes melitus (DM) and coronary heart disease (CHD) among 91 285 men.
Coronary heart disease (CHD) mortality stratified by history of diabetes mellitus (DM) and CHD among 91 285 men.
Lotufo PA, Gaziano JM, Chae CU, Ajani UA, Moreno-John G, Buring JE, Manson JE. Diabetes and All-Cause and Coronary Heart Disease Mortality Among US Male Physicians. Arch Intern Med. 2001;161(2):242–247. doi:10.1001/archinte.161.2.242
While diabetes has long been associated with increased risk of coronary heart disease (CHD), the magnitude of risk of diabetes-related CHD is uncertain.
To evaluate the impact of diabetes and prior CHD on all-cause and CHD mortality.
In a prospective cohort study of 91 285 US male physicians aged 40 to 84 years, participants were divided into 4 groups: (1) a reference group of 82 247 men free of both diabetes and CHD (previous myocardial infarction and/or angina) at baseline, (2) 2317 men with a history of diabetes but not CHD, (3) 5906 men with a history of CHD but not diabetes, and (4) 815 men with a history of both diabetes and CHD. Rates of all-cause and CHD mortality were compared in these groups.
Over 5 years (49 7952 person-years of follow-up), 3627 deaths from all causes were documented, including 1242 deaths from CHD. Compared with men with no diabetes or CHD, the age-adjusted relative risk of death from any cause was 2.3 (95% confidence interval [CI], 2.0-2.6) among men with diabetes and without CHD, 2.2 (95% CI, 2.0-2.4) among men with CHD and without diabetes, and 4.7 (95% CI, 4.0-5.4) among men with both diabetes and CHD. The relative risk of CHD death was 3.3 (95% CI, 2.6-4.1) among men with diabetes and without CHD, 5.6 (95% CI, 4.9-6.3) among men with CHD and without diabetes, and 12.0 (95% CI, 9.9-14.6) among men with both diabetes and CHD. Multivariate adjustment for body mass index, smoking status, alcohol intake, and physical activity as well as stratification by these variables did not materially alter these associations.
These prospective data indicate that diabetes is associated with a substantial increase in all-cause and CHD mortality. For all-cause mortality, the magnitude of excess risk conferred by diabetes is similar to that conferred by a history of CHD; for mortality from CHD, a history of CHD is a more potent predictor of death. The presence of both diabetes and CHD, however, identifies a particularly high-risk group.
DIABETES MELLITUS has long been linked with an excess risk of coronary heart disease (CHD).1 Several epidemiologic studies indicate that men with diabetes mellitus are 2 to 3 times more likely to die from CHD than those without diabetes,2- 5 and women with diabetes face an even higher risk.6,7 Diabetes increases the case-fatality rate of myocardial infarction (MI) events,8- 10 and prehospital mortality from acute coronary events is higher among people with diabetes.11 Patients with diabetes experienced a smaller decline in heart disease mortality in the period between 1982 and 1984 vs the period between 1971 and 1975 compared with nondiabetic individuals in the same periods.12 The magnitude of diabetes-related CHD risk and whether it rivals the excess risk associated with prior MI, however, are matters of some controversy. The National Cholesterol Education Program recommends more aggressive lipid-lowering therapy for people with a history of MI than for those without known CHD but with risk factors such as diabetes.13 However, a recent analysis of middle-aged Finnish men and women found that subjects who had diabetes but were free of cardiovascular disease had the same level of risk of CHD mortality as those subjects with a history of MI but without diabetes.14
The US Physicians' Health Study (PHS) enrollment cohort provided a unique opportunity to address the relationships of diabetes and a history of CHD to all-cause and CHD mortality. In this large, prospective cohort of 91 285 US men aged 40 to 84 years at baseline, we compared the risk of CHD mortality and all-cause mortality among 4 subsets of participants: (1) men who were free of diabetes and CHD (past MI and/or angina pectoris) at baseline, (2) men diagnosed with diabetes but not CHD, (3) men diagnosed with CHD but not diabetes, and (4) men diagnosed with both diabetes and CHD.
The PHS was a randomized, double-blind, placebo-controlled trial testing the benefits and risks of aspirin and beta carotene in the prevention of cardiovascular disease and cancer.15,16 Beginning in 1982, letters of invitation, informed consent forms, and baseline questionnaires were sent to 261 248 male physicians. By December 31, 1983, 104 388 physicians had answered the baseline enrollment questionnaire. Men with a history of cancer, stroke, liver disease, or renal disease were excluded because their preexisting conditions might have increased the likelihood of detecting diabetes and CHD, leaving a total of 91 285 participants. Our analysis includes a reference group of 82 247 men without a history of either diabetes or CHD; 2317 men who, at baseline, reported a history of diabetes but not CHD; 5906 men who reported a history of CHD (previous MI and/or angina pectoris) but not diabetes; and 815 men who reported a diagnosis of both diabetes and CHD. A subset of this cohort (approximately 25%) was subsequently randomized into the trial.
On the baseline questionnaire, respondents were asked to report their ages and previously diagnosed medical conditions, including history of MI, angina pectoris, and diabetes mellitus. They were also asked to record data on coronary risk factors, including systolic and diastolic blood pressure, history of drug treatment for hypertension, cholesterol level, history of drug treatment for high cholesterol, cigarette smoking (never, past, or current, and number of cigarettes smoked daily for current smokers), frequency of vigorous exercise (rarely/never, 1-3 times per month, 1-4 times per week, or 5 or more times per week), and frequency of alcohol intake (never, rarely, monthly, weekly, or daily). Body mass index was calculated from self-reported weight and height.
Information was not collected about type and duration of diabetes. Based on the age distribution of the participants, it is likely that the majority had type 2 diabetes.
Self-reports of the presence or absence of diabetes and/or CHD at baseline among the enrollees in the PHS were not validated by medical record review. However, during the randomized phase of the PHS, 95% of reports of postrandomization angina and/or coronary revascularization were confirmed by a review of medical records in a sample of 100 randomly selected participants who reported this diagnosis,17 and previous studies of health professionals show the reporting of diabetes and CHD to be reliable.18
Deaths were identified through systematic searches of the National Death Index for the entire enrollment cohort, and death certificates were obtained from state agencies for all deaths that occurred before February 1, 1988. The deaths were classified by trained nosologists according to the International Classification of Diseases, Ninth Revision (ICD-9). The "Automated Classification of Medical Entities Decision Tables" was used to select the underlying cause for deaths that occurred during a mean follow-up period of 5 years. We chose as end points all deaths and deaths caused by CHD (ICD-9 codes 410-414). The reliability of the National Death Index for epidemiologic purposes among health professionals has been validated.19 To detect a possible shift of the death certification from CHD to other cardiovascular diseases, all cardiovascular deaths (except stroke) were also analyzed as an end point, but the results were not materially different from those obtained when considering CHD deaths separately.
We calculated means or proportions of baseline risk factors for the reference group (subjects without diabetes and without CHD) and the other 3 categories (men with a history of diabetes without CHD, men with a history of CHD without diabetes, and men with a history of both diabetes and CHD). We used the Cox proportional hazards model to calculate age-adjusted and multivariate-adjusted relative risks for each category compared with the reference group, primarily because this model takes into account the time to event (in this case, time to death).20 The multivariate analysis was adjusted for age, body mass index, smoking, exercise, and alcohol consumption. We did not adjust for hypertension and high cholesterol in the primary models because diabetes increases the risk of these conditions, and thus they are potential intermediates in the causal pathway. We included hypertension and high cholesterol in secondary models to assess whether they mediated the effect of diabetes, but the results were similar to those of the primary multivariate model.
During a mean follow-up period of 5 years, 3627 (4%) of the physicians in the study died. Cardiovascular diseases (excluding stroke) accounted for 1676 (46%) of the deaths; 1246 of these were classified as CHD deaths.
Table 1 presents the baseline characteristics of the participants. Men who had both diabetes and CHD were generally older, more likely to be smokers, and less physically active; drank less alcohol; and had a higher frequency of hypertension and higher cholesterol compared with men in the other 3 categories. Those with a history of CHD but not diabetes had a higher frequency of high cholesterol compared with subjects without CHD. Subjects with diabetes, regardless of their CHD status, reported drinking less alcohol than subjects without diabetes. Those without CHD, regardless of their diabetes status, were more likely to be never smokers.
Among the 6721 participants who reported having CHD at baseline, 2509 (37.3%) reported having had at least one MI without concomitant angina pectoris; 2233 (33.2%) reported having only angina; and 1979 (29.4%) reported having both conditions.
The age-adjusted relative risk of all-cause mortality was similar for subjects with a history of diabetes but not CHD and subjects with a history of CHD but not diabetes (Table 2). As expected, the risk of all-cause mortality was much higher among those with both diabetes and CHD. Compared with men with no diabetes or CHD, the age-adjusted relative risk of CHD mortality was 3-fold higher for those with diabetes but not CHD, 5-fold higher for those with CHD but not diabetes, and 12-fold higher for those with both diabetes and CHD. For both all-cause and CHD mortality, the results were not materially altered after multivariate adjustment for body mass index, smoking habits, physical activity, and alcohol intake. These results are also shown by Kaplan-Meier curves in Figure 1 and Figure 2.
After categorizing participants into 3 age strata (40-54, 55-69, and 70-84 years), excess mortality risks associated with diabetes and/or CHD were observed in all age groups. The magnitudes of the associations were greatest for men in the youngest group (Table 3).
Table 4 shows the modifying effect of covariates other than age in our models. Diabetes and CHD were each associated with elevated mortality risk in all strata of body mass index, in the presence or absence of hypertension, in the presence or absence of high cholesterol, with past or current smoking status, with vigorous exercise, and with alcohol intake.
These prospective data demonstrate that self-reported diabetes is associated with a substantial increase in mortality from CHD and from all causes. The total risk of death associated with diabetes was essentially equal to that associated with prior CHD, each conferring a 2-fold increased risk of death. The observed risk of all-cause mortality was similar for subjects with a history of diabetes but not CHD and subjects with a history of CHD but not diabetes, which is likely attributable to the increased burden of noncardiovascular death among people with diabetes.21 In terms of CHD mortality, diabetes alone more than tripled the risk of CHD death. While a history of CHD carried a greater risk for CHD death than diabetes alone, the combined presence of diabetes mellitus and CHD conferred a dramatic 12-fold increase in risk of CHD death compared with those men with neither disease. These increased risks underscore the impact of diabetes on cardiovascular health. Diabetes is associated with increases in traditional risk factors for cardiovascular disease, including hypertension and dyslipidemia, increased oxidative stress, and increased tissue glycosylation, as well as platelet and fibrinolytic disorders that lead to a procoagulant state.22
Our data differ from the results of Haffner et al,14 whose 7-year follow-up of 2432 Finnish men and women aged 45 to 64 years found similar high risks of CHD mortality among diabetic patients without a prior MI and nondiabetic patients with a prior MI.14 Because diabetes is a weaker risk factor for CHD in elderly individuals than in middle-aged individuals, the different mean ages of the Finnish and PHS populations may account for the divergent results from that study and ours. However, we were able to verify in the PHS enrollment cohort that a significant difference in risk persisted even in the older age groups between subjects with a history of diabetes without CHD and those with a history of CHD without diabetes. Another important difference between the 2 studies is that all participants in the PHS were male, while the Finnish study enrolled almost equal numbers of men and women. Numerous prospective cohort studies have demonstrated that diabetes is a stronger risk factor for CHD in women than in men, with age-adjusted CHD mortality rates 3 to 7 times higher in diabetic women than in nondiabetic women6,7 and 2 to 3 times higher in diabetic men than in nondiabetic men.2- 5
Our data demonstrate a striking 5-fold increased risk of all-cause mortality and a 12-fold increased risk of CHD mortality in men with both diabetes and CHD, highlighting the importance of aggressive secondary prevention in patients with coexisting diabetes and CHD. This is further supported by secondary analyses from the Scandinavian Simvastatin Survival Study,23 the Gruppo Italiano per lo Studio della Sopravvivenza nell'Infarto miocardico 3 (GISSI-3) trial,24 and the Heart Outcomes Prevention Evaluation trial,25 in which post-MI treatment with simvastatin, lisinopril, and ramipril, respectively, was associated with significantly greater reductions in mortality in diabetic patients than in nondiabetic patients. Greater efforts are needed to target this high-risk population, however, as demonstrated in a recent study of survivors of first MI in the United Kingdom, in which people with diabetes were less likely to be treated with lipid-lowering agents, angiotensin-converting enzyme inhibitors, and aspirin than were those without diabetes.26
In terms of primary prevention of CHD in people with diabetes, expert panels currently recommend a more radical reduction in blood pressure among people with diabetes and coexisting hypertension,27 the use of low-dose aspirin for people with diabetes and other CHD risk factors,28 and the use of lipid-lowering drugs for people with diabetes who have elevated lipoprotein levels.29 However, these recommendations are aimed at diabetic persons who have concomitant CHD risk factors. Our data suggest that diabetes is associated with an increased risk of CHD mortality regardless of the presence or absence of other risk factors (Table 4) and that having diabetes alone warrants a more intensive application of primary prevention measures, as recently recommended by the American Heart Association.30
The strengths of our study are its size, its prospective design, and the relatively homogeneous nature of the cohort, which minimize confounding by several variables, including early symptom awareness, access to medical care, educational attainment, and socioeconomic status. At the same time, the study has several potential limitations. The data were based on self-reports, which can lead to misclassification. Studies of health professionals, however, have found self-reporting to be reliable for cardiovascular risk factors,18 with a 95% rate of confirmation of CHD events in the randomized phase of the PHS.17 Other limitations include the use of medical death certificates to define CHD end points,31 the lack of data about the duration or type of diabetes, and the lack of data about glycemic control. However, tight control of blood glucose levels has not been conclusively documented to be associated with cardiovascular end points in either type 1 or type 2 diabetes.32,33
Our results may not be generalizable to women because diabetes is a stronger risk factor for cardiovascular disease in women than in men; so far, no observational studies among women using the same design as our study have been published. Furthermore, because more than 90% of our participants were white, these results may not be generalizable to men of other racial groups, in whom the prevalence of diabetes34 and the risk of CHD35 may differ.
The increased risk of all-cause and CHD mortality that we observed among individuals with preexisting diabetes or CHD may, in fact, represent a best-case scenario, given the relatively high access to medical insurance and medical care in the PHS population. The relative risks may be even higher in a general population with poorer access to health care, and thus possibly less prompt and less aggressive diagnosis and care of diabetes and CHD.
In conclusion, our data demonstrate that diabetes is a strong risk factor for mortality from all causes and from CHD and that individuals with coexisting diabetes and CHD are part of a particularly high-risk group. These findings support the need for aggressive primary and secondary prevention measures in individuals with diabetes. Clinical trials addressing primary prevention in people with diabetes free of coronary disease, such as the proposed Prevention of Cardiovascular Disease in Diabetes Mellitus Study and the subgroup of diabetic participants in the ongoing Antihypertensive and Lipid Lowering Treatment to Prevent Heart Attack Trial (ALLHAT), and data from observational studies are needed to develop new approaches to reduce the burden of cardiovascular disease among people with diabetes.
Accepted for publication July 10, 2000.
This study was supported by grant HL-42441 from the National Institutes of Health, Bethesda, Md. Dr Lotufo is the recipient of fellowship 97-02424-8 from Fundação de Amparo a Pesquisa do Estado de São Paulo, Sao Paulo, Brazil.
The authors thank Rimma Dushkes, MS, and P. J. Skerrett, MS, for their expert assistance.
Corresponding author and reprints: JoAnn E. Manson, MD, DrPH, Division of Preventive Medicine, Brigham and Women's Hospital, 900 Commonwealth Ave E, Boston, MA 02215-1204 (e-mail: firstname.lastname@example.org).