New-Onset Diabetes and Risk of All-Cause and Cardiovascular Mortality: The Cardiovascular Health Study

The CHS is a community-based, prospective cohort study of risk factors for cardiovascular disease in the elderly. Participants were recruited during 1989–1990 from four U.S. communities (Washington County, MD; Pittsburgh [Allegheny County], PA; Forsyth County, NC; and Sacramento County, CA) based on a randomly generated sampling frame from Health Care Financing Administration files (7,8).

The study protocol consisted of a baseline clinic visit followed by semiannual contacts that alternated between phone contacts and clinic visits through June 1999. After June 1999, semiannual contacts were conducted by phone only.

Extensive data collection was performed annually at each clinic visit and included information on medical histories, clinical measures, and medication use. Questionnaires were administered, which collected information on health behaviors and medical conditions including smoking status, history of coronary disease, history of cerebrovascular disease, and diabetes. Blood pressure was measured annually. Phlebotomy was performed on fasting participants in study years 1989–1990, 1992–1993, and 1996–1997 and was performed on nonfasting participants in years 1993–1994, 1994–1995, and 1995–1996. Medication inventories were conducted in person at each clinic visit through June 1999 and were collected annually by phone thereafter (9). Medications used within the past 2 weeks were recorded by study personnel using medicine containers that were brought to the clinic by the participants or by having the label read to them when the inventory was conducted by phone. For these analyses, we used data from 12 study encounters (10 clinical visits and 2 follow-up telephone contacts) that extended through 30 April 2002.

The cohort consisted of 5,201 community-dwelling adults aged ≥65 years who participated in the baseline clinic visit in 1989–1990 plus an additional 687 African-American adults aged ≥65 years who were recruited for a baseline clinic visit in 1992–1993 to augment minority representation. All study participants gave informed consent.

For these analyses of new-onset diabetes, subjects were excluded at baseline if they had an 8-h fasting plasma glucose >125 mg/dl (n = 424), if they were being treated with insulin or an oral hypoglycemic agent (n = 516), or if either assessment was missing or the participant was not fasting (n = 85). These exclusions produced a cohort of 4,863 nondiabetic subjects at baseline: 4,375 from the original cohort and 488 from the African-American cohort.

New-onset diabetes was defined as initiating antidiabetes medication according to annual medication inventories or by a single, in-study fasting plasma glucose >125 mg/dl. Once an individual was defined as having diabetes by either criterion, they were considered to have diabetes throughout follow-up.

All-cause and cardiovascular mortality were the outcomes of interest and were obtained through active surveillance of CHS participant contacts and through obituary searches at the four clinic sites (10). Cause-specific mortality was determined by the CHS Events Committee using information from several sources including autopsies, death certificates, hospitalization records, and physician notes. Cardiovascular mortality was defined as all atherosclerotic coronary heart disease (fatal myocardial infarction and define and possible fatal coronary heart disease), cerebrovascular disease (fatal ischemic and hemorrhagic stroke), and other atherosclerotic and cardiovascular disease deaths (such as a ruptured abdominal aortic aneurysm, valvular heart disease, and pulmonary embolism). This article includes mortality follow-up through 30 June 2002.

Demographics included sex and self-reported race (African American versus other). Other covariates were measured at the time of the clinic visit (or phone contact) immediately following diabetes identification and included age, BMI, impaired fasting glucose (<100 vs. ≥100 and ≤125 mg/dl), treated hypertension status (self-report of hypertension and current use of antihypertensive medication), systolic blood pressure, diastolic blood pressure, total cholesterol (measured in fasting and nonfasting phlebotomy years), smoking status (current versus not-current smoker), and exercise intensity (no exercise or low, moderate, and high intensity). Documentation for baseline and follow-up cardiovascular events that occurred in the study were reviewed and adjudicated by the CHS Event Committee (10,11). Study participants were classified as having prevalent coronary artery disease at the time of diabetes diagnosis if there was a confirmed history of angina, myocardial infarction, or coronary revascularization. Prevalent stroke was defined as a confirmed history of stroke. For all covariates, when data were missing for a visit, values were carried forward from previous visits.

At each of the 11 follow-up encounters, study participants with newly diagnosed diabetes were matched on age (by year), sex, and race (African American versus all others) to three participants without diabetes to create an inception cohort for these analyses. This approach permits adjustments for risk factors at the time of new-onset diabetes or control sampling. Sampling participants independently at each study encounter allowed nondiabetic participants to be sampled more than once but excludes those with new-onset diabetes from being sampled after identification.

Kaplan-Meier survival curves were plotted for all-cause mortality among those with and without new-onset diabetes, and separation of curves was tested with a log-rank test. All-cause and cardiovascular disease mortality rates per 100 person-years were calculated for the analytic cohort and for sex- and race-specific strata. Relative risks are provided and 95% CIs were calculated using exact methods. Proportional hazards regression models using diabetic and sampled nondiabetic participants estimated hazard ratios (HRs) while adjusting for age, sex, and race and the potential confounding factors of BMI, impaired fasting glucose, treated hypertension, systolic blood pressure, total cholesterol, current smoking status, exercise intensity, and prevalent coronary artery and cerebrovascular disease at the time of diabetes onset or at the time when they were sampled. Subjects entered the analysis when they were identified as having diabetes or when they were selected as a control subject. All were followed until death or the end of follow-up, at which time they were censored. For cardiovascular mortality analyses, noncardiovascular deaths were censored.

Risk differences for new-onset diabetes according to sex and race where tested with cross-product terms in models. To examine differences in mortality risk associated with new-onset diabetes over follow-up, HRs were calculated for four follow-up intervals: 0–2, 3–4, 5–6, and >6 years (maximum was 11 years, average 6.8 years). A cross-product interaction term was created using new-onset diabetes and a four-category ordinal measure of interval to test for changes in risk over time in multivariate models.

In sensitivity analyses, we restrict outcomes to those with coronary and cerebrovascular deaths. We also conducted sensitivity analyses where we restricted analyses to those recruited from 1989 to 1990 when a 2-h glucose challenge test was performed at study entry. For these analyses, we exclude participants who had a diabetes-level 2-h glucose at baseline and adjust for impaired glucose tolerance in survival analyses.

During the 11 years of follow-up, 282 participants were diagnosed with new-onset diabetes. Of those, 95 were identified by antidiabetes medication use alone, 162 by a single measure of fasting plasma glucose, and 25 by both criteria. Of the 162 identified by fasting plasma glucose alone, 71 (44%) were eventually treated with antidiabetes medication, and of the remaining 91 (54%), 60 had a subsequent fasting glucose measure and 36 (60%) of these had glucose levels >125 mg/dl. These participants with diabetes were matched to 837 participants without diabetes (a total of 9 participants without diabetes could not be found to match the age/sex/race criteria of those with new-onset diabetes). Table 1 describes participant characteristics at the time of entry into the analytic cohort. Matching on age, sex, and race was successful, and there were no differences between the two groups on these characteristics. Those with new-onset diabetes were likely to have treated hypertension, higher fasting plasma glucose levels and BMI, and more prevalent coronary artery disease.

There was a median of 5.9 years of follow-up (interquartile range 4.2–9.2) after entry into the analytic cohort. During follow-up, there were 352 deaths, of which 143 (41%) were classified as cardiovascular (95 atherosclerotic coronary disease deaths, 27 cerebrovascular disease deaths, and 21 other cardiovascular deaths). Figure 1 depicts the Kaplan-Meier survival curves for the first 8 years of follow-up. Overall survival was worse among those with new-onset diabetes compared with those without (P < 0.001).

Table 2 depicts the number of deaths, person-years of exposure, and death rates by diabetes status for the new-onset diabetic cohort and for the cohort stratified by sex and race. Age-, sex-, and race-adjusted relative rates (RRs) and 95% CIs are also presented. The risk of all-cause mortality was 50% greater (RR 1.5 [95% CI 1.2–1.9]), and the risk of cardiovascular mortality was twofold greater (2.1 [1.5–3.0]) in those with new-onset diabetes compared with those without diabetes in age-, sex-, and race-adjusted comparisons. Additional adjustments for BMI, impaired fasting glucose, treated hypertension, exercise intensity, and prevalent coronary artery disease in multivariate survival analyses had little effect on risk estimates: all-cause mortality (HR 1.9 [95% CI 1.4–2.5]) and cardiovascular mortality (2.2 [1.4–3.4]). Further adjustment for systolic blood pressure, smoking, and stroke did not change risk estimates. Risk estimates were similar for men and women and were larger in African Americans than in non–African Americans. There was no evidence for effect modification by sex or race (P > 0.45 for all sex-diabetes and race-diabetes interactions). Among the 120 participants with new-onset diabetes who were initially classified by antidiabetes medication use, all-cause mortality risk (HR 2.1 [95% CI 1.5–3.1]) and cardiovascular mortality risk (2.2 [1.3–3.8]) did not differ from risks among the 187 participants with new-onset diabetes who were initially classified by fasting glucose levels: all-cause mortality risk (1.7 [1.2–2.3]) and cardiovascular mortality risk (2.3 [1.4–3.8]). Among those classified by antidiabetes medication, 71% were started on a second-generation sulfonylurea, and all-cause morality risk was similar across other oral hypoglycemic classes and insulin.

Table 3 provides all-cause and cardiovascular mortality risk associated with new-onset diabetes stratified by years of follow-up. For all-cause mortality in the first 2 years of follow-up, there was a twofold increase in risk of death among those with new-onset diabetes compared with those without diabetes (HR 2.3 [95% CI 1.2–4.2]) and risks in subsequent periods remained elevated. There was no suggestion of an increasing trend in the diabetes-mortality association across periods (trend test had a nonsignificant negative slope). For cardiovascular mortality in the first 2 years of follow-up, there was a multifold increase in risk of death among those with new-onset diabetes compared with those without diabetes (4.3 [1.7–10.8]). In post hoc analyses, cardiovascular mortality in the first 2 years after diagnosis was elevated (1.5 [0.7–3.3]) compared with the new-onset diabetes and cardiovascular mortality risk in subsequent periods.

When cardiovascular deaths were restricted to atherosclerotic coronary disease deaths and cerebrovascular disease, the risk of cardiovascular mortality was essentially unchanged (HR 2.6 [95% CI 1.7–4.1]). Among deaths that were noncardiovascular, new-onset diabetes was not associated with cancer deaths (0.8 [0.4–1.4]). Other death categories were small and were not investigated. When analyses were restricted to those with a 2-h glucose measure that did not exceed 200 mg/dl (151 participants with new-onset diabetes and 447 control subjects), mortality risk was similar for all-cause (1.6 [1.1–2.3]) and cardiovascular mortality (1.8 [1.0–3.1]). Cardiovascular mortality risk remained elevated in the first 2 years, and there was no suggestion of increasing all-cause or cardiovascular mortality risk over time, although CIs were wide.

New-onset diabetes, defined by the initiation of antidiabetes medication or by a fasting plasma glucose >125 mg/dl, was associated with a 90% increase in risk of all-cause mortality and a 120% increase in risk of cardiovascular mortality compared with study participants without diabetes. Contrary to our hypothesis, there was no increase in all-cause or cardiovascular morality risk over time. Instead, we found a significant, large increase in cardiovascular mortality in the first 2 years of follow-up that diminished over time, whereas all-cause mortality risk was similar and elevated throughout follow-up.

The incidence of diabetes increases dramatically for adults aged >40 years. Findings from Rochester County, MN, and the Netherlands suggest that diabetes incidence continues to increase through the 8th decade of life in men and women but may level off or possibly decrease for adults aged ≥80 years (2–5). Numerous epidemiologic studies have demonstrated that type 2 diabetes increases the risk of cardiovascular morbidity and mortality (12–14). Prolonged exposure to advanced glycation end products that arise from a hyperglycemic environment may contribute to adverse vascular outcomes (15–18). It is known that glucotoxins that result from mild hyperglycemia interfere with normal endothelial function and promote inflammation, thrombosis, and atherosclerosis (19,20). Epidemiologic data demonstrate that exposure to impaired levels of fasting and 2-h glucose are associated with a modest increase in mortality risk (HRs 1.1 and 1.4, respectively) and that diabetic levels of either measure is associated with greater risk (HRs 1.7 and 1.9, respectively) (6).

Data from our study suggest no increase in mortality risk with time and that there may be short-term adverse mortality effects that follow the diagnosis of new-onset diabetes in older adults. We speculated that the long-term damaging effects of hyperglycemia on atherosclerosis and on the vasculature would contribute substantially to mortality and that these effects would not manifest early. Our data suggest otherwise, although statistical comparisons between follow-up categories lacked statistical power. The strong evidence of short-term mortality risk is compelling and suggests that limited exposure to a hyperglycemic environment, where fasting plasma glucose exceeds 125 mg/dl, has detrimental effects, possibly through nonatherosclerotic pathways. This observation held true when new-onset diabetes was identified either by hyperglycemia or by medication initiation. It also held true when we excluded participants with 2-h glucose measures >200 mg/dl at baseline.

This investigation was conducted using standardly collected longitudinal data from a well-defined cohort that has virtually complete mortality follow-up. Several limitations need consideration nonetheless. Fasting plasma glucose was not assessed annually, and we relied on a single measurement to classify new-onset diabetes, which can lead to misclassification. Our data indicate, however, that a large proportion of those classified by a single fasting measure were eventually treated pharmaceutically or had a subsequent elevated fasting glucose. New-onset diabetes identification dates were assigned to the date of the clinic visit or phone contact, but the actual onset of diabetes certainly preceded the assigned date. This is especially true for diabetes defined by impaired fasting glucose, which was assessed only three times during the study. It is possible that some new-onset diagnoses were missed if the participant died before the subsequent study contact. Measurement of 2-h postchallenge fasting glucose, which has been shown in these and other data to be more strongly related to morbidity and mortality (6,18,21), was only available at baseline and at 7 years’ follow-up; thus, it was not feasible to use this measure as a criterion for diabetes diagnosis. Type 2 diabetes is often underdiagnosed in the community setting, and the extent of underdiagnosis, though mitigated by repeated fasting glucose measures, is unknown (2).

We found that mortality risks were similar if diabetes was identified by hyperglycemia or by initiating antidiabetes medication. This observation parallels clinical trial findings where the success of treatment with oral hypoglycemic agents or insulin to prevent macrovasculature damage has been modest in terms of long-term benefits (15,22). Clinically, there has been more success reducing mortality differentials by controlling cardiovascular risk factors such as hypertension in adults with diabetes (23). We have previously demonstrated, however, that the treatment and control of cardiovascular risk factors in this population-based cohort of older adults with diabetes has been suboptimal (24). The effect of adequate control of glycemia and other cardiovascular risk factors to reduce mortality in older adults with new-onset diabetes remains uncertain.

In summary, we found no increase in all-cause or cardiovascular mortality over time associated with new-onset diabetes in older adults. An early differential in mortality risk was detected, and these findings suggest that long-term macrovascular damage from atherosclerosis may not be primarily responsible for increased risk.

The research reported in this article was supported by contracts N01-HC-85079 through N01-HC-85086, N01-HC-35129, N01-HC-15103, N01-HC-55222, U01 H1080295, and R01-HL-73410 from the National Heart, Lung, and Blood Institute and grant R01 AG09556 from the National Institute on Aging, with additional contributions from the National Institute of Neurological Disorders and Stroke.

A complete list of participating CHS investigators and institutions can be found at http://www.chs-nhlbi.org.