Author + information
- Received July 29, 2003
- Revision received October 15, 2003
- Accepted October 20, 2003
- Published online June 16, 2004.
- Joshua I. Barzilay, MD*,* (, )
- Richard A. Kronmal, PhD†,
- John S. Gottdiener, MD§,
- Nicholas L. Smith, PhD, MPH‡,
- Gregory L. Burke, MD, MS∥,
- Russell Tracy, PhD¶,
- Peter J. Savage, MD# and
- Michelle Carlson, PhD**
- ↵*Reprint requests and correspondence:
Dr. Joshua I. Barzilay, Kaiser Permanente of Georgia, 200 Crescent Center Parkway, Tucker, Georgia 30084.
Objectives The purpose of this study was to determine if fasting glucose levels are an independent risk factor for congestive heart failure (CHF) in elderly individuals with diabetes mellitus (DM) with or without coronary heart disease (CHD).
Background Diabetes mellitus and CHF frequently coexist in the elderly. It is not clear whether fasting glucose levels in the setting of DM are a risk factor for incident CHF in the elderly.
Methods A cohort of 829 diabetic participants, age ≥65 years, without prevalent CHF, was followed for five to eight years. The Cox proportional hazards modeling was used to determine the risk of CHF by fasting glucose levels. The cohort was categorized by the presence or absence of prevalent CHD.
Results For a 1 standard deviation (60.6 mg/dl) increase in fasting glucose, the adjusted hazard ratios for incident CHF among participants without CHD at baseline, with or without an incident myocardial infarction (MI) or CHD event on follow-up, was 1.41 (95% confidence interval 1.24 to 1.61; p < 0.0001). Among those with prevalent CHD at baseline, with or without another incident MI or CHD event on follow-up, the corresponding adjusted hazard ratio was 1.27 (95% confidence interval 1.02 to 1.58; p < 0.05).
Conclusions Among older adults with DM, elevated fasting glucose levels are a risk factor for incident CHF. The relationship of fasting glucose to CHF differs somewhat by the presence or absence of prevalent CHD.
Congestive heart failure (CHF) and diabetes mellitus (DM) are age-related disorders that frequently coexist (1,2). Despite their close association, there is uncertainty whether elevated glucose levels are a risk factor for CHF. Cohort (3–7)and clinic (8–11)studies of people with DM have reached conflicting results. In a recent analysis of the diabetic and nondiabetic cohorts of the Cardiovascular Health Study (CHS)—a longitudinal study of elderly adults whose purpose is to identify factors related to the onset and course of cardiovascular disease and stroke—we found that a history of DM predicted CHF, but not a glucose level >125 mg/dl (12). The uncertainty regarding the association of elevated glucose levels and incident CHF may stem from the fact that CHF is a heterogeneous disorder. It may occur as a result of myocardial damage secondary to coronary heart disease (CHD), but may also occur in the absence of CHD due to myocardial dysfunction. If elevated glucose levels have a differing degree of association with these conditions, then the relationship of elevated glucose levels to CHF risk would vary as well.
In the present study, we examine the diabetic cohort of CHS to determine the role of fasting glucose levels for incident CHF in those with and without CHD at baseline or on follow-up.
Recruitment methods for CHS have been published (13). In brief, a random sample of individuals, ≥65 years of age, derived from Medicare eligibility lists, and other household members ≥65 years, were invited to participate in the study. Potential participants were excluded if they had illnesses that were expected to lead to early death. A total of 5,201 participants were recruited in 1989 to 1990 (called the original cohort), and 687 were recruited in 1992 to 1993 to provide additional representation of African Americans (called the new cohort). For these analyses, only participants with DM at entry were studied (n = 910). All participants signed informed consent.
Upon entry into the study, participants were invited for a baseline interview. Information on prescription medications used in the preceding two weeks (including insulin and oral hypoglycemic agents) was collected (14). During a subsequent clinic visit, venipuncture was done after an overnight fast. Plasma and serum were frozen at −70°C, and shipped to the CHS Central Laboratory (University of Vermont, Burlington, Vermont). Fasting serum chemistry analyses were performed (15). Glucose levels were measured on a Kodak Ektachem 700 Analyzer (Ektachem Test Methodologies, Eastman Kodak, Rochester, New York, March 1985) and assayed within 30 days. Average monthly coefficient of variation was 0.93%. One year after the baseline examination, a subset of CHS participants had a hemoglobin A1c test performed as part of a substudy.
Echocardiograms were obtained during the baseline examination for 99.1% of the original cohort in 1989 to 1990. Echocardiograms for the African-American participants were obtained in 1994 to 1995, one year after enrollment in the study. Therefore, these latter echocardiograms are not used in this analysis because some incident events may have occurred before performing the echocardiogram. Moreover, laboratory values, medication use, and anthropometric measurements would have differed from the baseline examination to the time that the echocardiogram was performed. For those with incident CHF, left ventricular (LV) systolic function data was obtained in many (but not all) cases from echocardiographic reports at the point of care during index hospitalization or office visit at which the clinical diagnosis of CHF was made. The reports were reviewed by CHS investigators.
Diabetes mellitus was defined by a fasting glucose ≥126 mg/dl or the use of antidiabetic agents. Self-report of DM was not a defining criterion. Subjects who were not fasting at blood draw or whose diabetic status could not be determined from medication lists were excluded (n = 64).
Baseline and incident CHD was defined as a history of myocardial infarction (MI) or a non-MI event, such as angina pectoris, or a revascularization procedure (coronary artery bypass grafting or percutaneous transluminal coronary angioplasty).
Global LV function was qualitatively assessed from the baseline two-dimensional echocardiogram as normal, borderline, or abnormal ejection fraction. Inter-reader agreement was 94%, and intra-reader agreement was 98% of paired studies (16). Qualitatively assessed normal LV function on the echocardiogram corresponded to an ejection fraction ≥55%, borderline was 45% to 54%, abnormal was <45%.
The diagnosis of incident CHF was based on data from an index hospitalization, outpatient visits for CHF, or self-report of a physician diagnosis of CHF (17). These reports were confirmed by documentation in the participant's medical records of a constellation of symptoms (dyspnea, fatigue, orthopnea, paroxysmal nocturnal dyspnea) and physical signs (edema, pulmonary rales, gallop rhythm, displaced LV apical impulse), and by supporting clinical findings such as chest X-ray and medications (digoxin, hydralazine, or loop diuretics). Supporting clinical findings were adjudicated by the CHS events committee, which classified all cardiovascular events (17).
Analyses for this study are based on events through June 1998. Follow-up was >95% complete for the original and the African-American cohorts based on telephone contact and review of Medicare data tapes.
Baseline characteristics of the cohort, categorized by development of CHF, were compared using the chi-square test for discrete values and the ttest for continuous data. The Cox proportional hazards models were used to examine the effect of fasting glucose levels on incident CHF. A fasting glucose of 200 mg/dl was approximately the 80th percentile of this cohort. Models were generated in stages. Demographic, laboratory, and clinical measures were taken first. Once significant predictors were identified, time-dependent variables (new MI or non-MI CHD event) were added to each model to evaluate the risk of these events on CHF. To determine the risk associated with fasting glucose in those without baseline CHD or an incident MI/other CHD event before the occurrence of incident CHF, significant interactions of risk factors and time-dependent variables were included in the model. At each stage, stepwise selection was used with a value of p < 0.05. All analyses were done using SPSS Version 11.0 (SPSS Inc., Chicago, Illinois).
There were 910 participants in CHS with DM at baseline (15.5% of the two cohorts combined). Of these, 81 (8.9%) had prevalent CHF and were not considered in this analysis. Of the remaining 829 participants, there were 626 without clinical CHD and 203 with clinical CHD at baseline (Fig. 1).
Incident CHF in those without baseline CHD
There were 123 incident CHF events among the 626 individuals without CHD at baseline—55 after an incident CHD/MI event, and 68 without an incident CHD/MI event. Baseline characteristics of these 626 participants, categorized by development or absence of incident CHF, are shown in Table 1. Those who developed CHF were older, more centrally obese, and had higher systolic blood pressure, fasting glucose, uric acid, and C-reactive protein levels than those who did not develop CHF. They did not differ significantly with respect to LV function at baseline or hypoglycemic medication use. Systolic LV function at baseline was usually normal. At the time of hospitalization for the incident CHF, LV function was preserved in more than one-half of cases (among those with no MI or CHD event before CHF, 14 of 32 [43.8%] had abnormal ejection fraction; among those with MI or CHD event before CHF, 12 of 30 [40.0%] had abnormal ejection fraction).
The incident rate of CHF in the cohort was 31.1 cases per 1,000 person-years. The rate in nondiabetic CHS participants without baseline CHD was 12.7 cases per 1,000 person-years. The incidence rates of CHF events per 1,000 patient-years of follow-up by fasting glucose quartiles are shown in Table 2. There was a near doubling of rate in the 2nd and 3rd quartiles as compared with the 1st quartile (31.8 and 28.8 cases per 1,000 patient-years, respectively, vs. 17.7). There was a near tripling in rate in the 4th quartile (47.7 cases per 1,000 patient-years) relative to the 1st quartile.
Results of the Cox proportional hazards models examining the effect of fasting glucose on incident CHF risk are shown in Table 3. In unadjusted analysis, a 1 standard deviation increase in fasting glucose level (60.6 mg/dl) was associated with a 37% increased risk of CHF (model I, p < 0.0001). When adjustment was made for other baseline risk factors, a further increase in association of fasting glucose with incident CHF was found (model II, p < 0.0001). Addition of time-dependent variables (incident MI or other CHD event) did not change this association (model III, p < 0.0001). A test for interaction between fasting glucose levels and incident MI before incident CHF was not significant, suggesting the effect of glucose on CHF did not differ by the presence or absence of incident MI. Several other significant interactions between incident MI and baseline variables had little effect on the risk associated with fasting glucose levels and CHF (data not shown). When analysis was repeated with fasting glucose levels as quartiles, there was an increase in risk from the first to the 2nd and 3rd quartiles. There was a further increase in the 4th quartile. With adjustment for baseline factors (model II) and incident MI and CHD events (model III), the association of fasting glucose with CHF increased, especially for those in the 4th quartile.
Incident CHF in those with baseline CHD
There were 203 participants at baseline without CHF who had clinical CHD. Among them, 83 incident CHF events occurred for a rate of 78.3 cases per 1,000 person-years. The rate in nondiabetic CHS participants with baseline CHD was 30.9 cases per 1,000 person-years. Among those for whom echocardiographic data was available at the time of the CHF event, 12 of 21 (57.1%) had an abnormal ejection fraction without an additional incident MI or CHD event, and 7 of 15 (46.7%) had an abnormal ejection fraction with an incident MI or CHD event.
In unadjusted analysis, a 1 standard deviation increase in fasting glucose level (60.6 mg/dl) was associated with a 16% increased risk of CHF (model I, p = 0.15) (Table 4). When adjustment was made for other baseline risk factors, an increase in association was found (model II, p < 0.05). Addition of time-dependent variables (incident MI or other CHD event) did not significantly change this association (model III, p < 0.05). When the analysis was repeated with fasting glucose levels as quartiles, there was an increase in risk for CHF in the 4th quartile relative to the other quartiles, especially in models II and III.
Of the 829 individuals in this study, 136 had a hemoglobin A1c test done one year after the baseline examination. The correlation coefficient between the baseline fasting glucose levels and percent hemoglobin A1c was 0.59 (p < 0.000001).
This study shows that fasting glucose levels are related to CHF risk in older adults with DM. Among those without prevalent CHD, our models show an ∼40% increase in risk. Among those with prevalent CHD, the risk of CHF associated with elevated fasting glucose level is also present but not as strong. The relationship, however, is not well estimated because of the small number of participants in the group.
Raised glucose levels may lead to CHF in the absence of CHD by several mechanisms. Raised levels may reflect poorer compliance with medications or poorer medical care. In this study, however, the use of hypoglycemic agents was the same in those who did or did not develop CHF. Second, hyperglycemia impairs endothelial function and vasodilation, impeding the ability of coronary arteries to increase myocardial blood flow (18). Finally, elevated glucose levels may have a deleterious effect on the myocardium leading to fibrosis and stiffness (19–23)with diminished diastolic filling. Diastolic dysfunction has been described as a cause of CHF in adults with DM (24). In CHS, most participants with CHF had normal, or borderline, impairment of systolic function, presumably reflecting isolated diastolic heart failure (25,26).
We have previously shown that incident CHD among individuals with glucose disorders is related to elevated fasting glucose levels (27). In that analysis, 20 mg/dl increments of fasting glucose levels increased CHD risk by 6% (95% confidence interval 1.00 to 1.12). Thus, it is not surprising that in those in whom CHF occurred in association with prevalent CHD that the risk of CHF was associated with the fasting glucose level.
Several other points regarding our results should be noted. First, of the total CHF events in this diabetic cohort, only 33% (68 of 206) occurred in the absence of clinical CHD. This value is lower than that found in the Framingham Heart study (28). This difference, in all likelihood, reflects the higher rate of CHD in people with DM. Second, among those in whom echocardiographic reports were available at the time of CHF diagnosis, ∼45% to 60% had normal or borderline systolic ventricular function (depending on prevalent or incident MI or CHD events). This is in keeping with recent population-based data (29). Third, approximately one-half of the cohort had untreated DM. Most were newly diagnosed diabetic individuals. This is similar to prior studies of DM in which one-half of individuals with DM are unaware of having it (30).
The strength of this study is its methodology. It allowed for the evaluation of baseline risk factors as well as incident events. It was population-based and included women and minorities. Disadvantages of this study should also be noted. First, institutionalized individuals and those with short life expectancy were excluded. Thus, the sample is primarily of the relatively healthy elderly, and results cannot be extrapolated to those who were very ill. On the other hand, the relatively long follow-up of CHS should diminish this effect. Second, a fasting glucose level was used and not a measure of overall glucose control, such as glycosylated hemoglobin. Many epidemiological studies rely on fasting glucose levels. There was, however, a strong correlation between the baseline fasting glucose level and the hemoglobin A1c level in a subset of the study cohort done one year after baseline testing. This suggests that the fasting glucose level was related to overall glucose control. It also suggests that elevated glucose levels tracked over time and reflected continued elevated levels. Third, angiographic data are not available in CHS. The degree of underlying CHD in those without clinical CHD, therefore, cannot be assessed.
In conclusion, elevated fasting glucose levels in older adults with DM are associated with the risk of CHF in both those with and without MI/CHD. Ongoing prospective diabetes studies, such as the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial, will be able to determine if better glucose control results in a reduction of CHF risk.
☆ Supported by contracts NO1-HC-85079 through NO1-HC-85086, NO1-HC-35129, and NO1-HC-15103 from the National Heart, Lung, and Blood Institute.
- coronary heart disease
- congestive heart failure
- Cardiovascular Health Study
- diabetes mellitus
- left ventricle/ventricular
- myocardial infarction
- Received July 29, 2003.
- Revision received October 15, 2003.
- Accepted October 20, 2003.
- American College of Cardiology Foundation
- Solang L.,
- Malmberg K.,
- Ryden L.
- Iribarren C.,
- Karter A.J.,
- Go A.S.,
- et al.
- Nichols G.A.,
- Hillier T.A.,
- Erbey J.R.,
- Brown J.B.
- Chae CU, Glynn RJ, Manson JE, et al. Diabetes predicts congestive heart failure in the elderly. Circulation 1998 Suppl 1:I–658.
- Roblin D.W.,
- Barzilay J.I.,
- Gardner C.S.,
- Kawamura G.
- Lo S.S.,
- Leslie R.D.,
- Sutton M.S.
- Gottdiener J.S.,
- Arnold A.M.,
- Aurigemma G.P.,
- et al.
- Psaty B.M.,
- Lee M.,
- Savage P.J.,
- Rutan G.H.,
- German P.S.,
- Lyles M.
- Cushman M.,
- Cornell E.S.,
- Howard P.R.,
- Boville E.G.,
- Tracy R.P.
- Gardin J.M.,
- Siscovick D.,
- Anton-Culver H.,
- et al.
- Ives D.G.,
- Fitzpatrick A.L.,
- Bild D.E.,
- et al.
- Williams S.B.,
- Goldfine A.B.,
- Timimi F.K.,
- et al.
- Frustaci A.,
- Kajstura J.,
- Chimenti C.,
- et al.
- Taegtmeyer H.,
- McNulty P.,
- Young M.E.
- Poirier P.,
- Bogaty P.,
- Garneau C.,
- Marois L.,
- Dumesnil J.-G.
- Bell D.S.H.
- Kuller L.H.,
- Velentgas P.,
- Barzilay J.,
- Beauchamp N.J.,
- O'Leary D.H.,
- Savage P.J.
- Lloyd-Jones D.M.,
- Larson M.G.,
- Leip E.P.,
- et al.
- ↵Kenny SJ, Aubert RE, Geiss LS. Prevalence and incidence of non-insulin-dependent diabetes. In: Harris MI, Bennett PH, Boyko EJ, et al., editors. Diabetes in America. 2nd edition. 1995:47–67. NIH Publication No. 95-1468. Available at http://diabetes.niddk.nih.gov/dm/pubs/america/index.htm