The Association of Hemoglobin A1c With Incident Heart Failure Among People Without Diabetes: The Atherosclerosis Risk in Communities Study

  1. Elizabeth Selvin1
  1. 1Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland;
  2. 2Division of General Internal Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland;
  3. 3Medical School of Ribeirao Preto, University of Sao Paulo, Ribeirao Preto, Brazil;
  4. 4Department of Epidemiology and Prevention, Wake Forest University Health Sciences, Winston-Salem, North Carolina;
  5. 5Department of Medicine, Division of Cardiology, University of North Carolina, Chapel Hill, North Carolina.
  1. Corresponding author: Kunihiro Matsushita, kmatsush{at}jhsph.edu.

Abstract

OBJECTIVE This study sought to investigate an association of HbA1c (A1C) with incident heart failure among individuals without diabetes and compare it to fasting glucose.

RESEARCH DESIGN AND METHODS We studied 11,057 participants of the Atherosclerosis Risk in Communities (ARIC) Study without heart failure or diabetes at baseline and estimated hazard ratios of incident heart failure by categories of A1C (<5.0, 5.0–5.4 [reference], 5.5–5.9, and 6.0–6.4%) and fasting glucose (<90, 90–99 [reference], 100–109, and 110–125 mg/dl) using Cox proportional hazards models.

RESULTS A total of 841 cases of incident heart failure hospitalization or deaths (International Classification of Disease, 9th/10th Revision, 428/I50) occurred during a median follow-up of 14.1 years (incidence rate 5.7 per 1,000 person-years). After the adjustment for covariates including fasting glucose, the hazard ratio of incident heart failure was higher in individuals with A1C 6.0–6.4% (1.40 [95% CI, 1.09–1.79]) and 5.5–6.0% (1.16 [0.98–1.37]) as compared with the reference group. Similar results were observed when adjusting for insulin level or limiting to heart failure cases without preceding coronary events or developed diabetes during follow-up. In contrast, elevated fasting glucose was not associated with heart failure after adjustment for covariates and A1C. Similar findings were observed when the top quartile (A1C, 5.7–6.4%, and fasting glucose, 108–125 mg/dl) was compared with the lowest quartile (<5.2% and <95 mg/dl, respectively).

CONCLUSIONS Elevated A1C (≥5.5–6.0%) was associated with incident heart failure in a middle-aged population without diabetes, suggesting that chronic hyperglycemia prior to the development of diabetes contributes to development of heart failure.

Diabetes is one of the most important risk factors for heart failure (1). Among people with diabetes, a dose relationship between glycemia measured by HbA1c (A1C) and heart failure risk has been reported in observational studies (16). The risk of hospitalization for heart failure increases 8–32% per 1% unit increase in A1C (36).

In contrast, to our knowledge, no previous study has investigated the association between A1C and the risk of heart failure in a population without diabetes. In this population, fasting glucose is only marginally or not associated with heart failure risk (7,8). This may be partially attributable to relatively high variability in glucose measurements (9) and to the fact that fasting glucose levels do not necessarily reflect postprandial hyperglycemia, a condition potentially contributing to development of cardiovascular disease (10,11).

In January 2010, the American Diabetes Association published new clinical guidelines recommending the use of A1C as a diagnostic test for diabetes (12), with cut-points based largely on the documented association of A1C with microvascular disease. Little is known, however, regarding the risk relationship of A1C levels with heart failure incidence in nondiabetic adults. The objective of this study was to investigate a possible relationship between A1C and the incidence of heart failure in a community-based study of people without diabetes. We also compared the associations of A1C and fasting glucose levels with risk of incident heart failure in this middle-aged nondiabetic population.

RESEARCH DESIGN AND METHODS

The Atherosclerosis Risk in Communities (ARIC) Study is a community-based cohort study of 15,792 people aged 45–64 years at baseline sampled from four U.S. communities: Forsyth County, North Carolina; suburban Minneapolis, Minnesota; Washington County, Maryland; and Jackson, Mississippi (13). The first examination was conducted during 1987–1989, with three triennial follow-up visits (visit 2 (1990–1992), visit 3 (1993–1995), and visit 4 (1996–1998)). Visit 2 was the only visit at which A1C was measured and was the baseline for the present study. There were 14,348 participants (91.4%) who attended visit 2. Of these, we excluded participants reporting race other than Caucasian or African American (n = 42); missing values of A1C (n = 278); with prevalent heart failure defined as self-reported treatment for heart failure, hospitalization for heart failure between visit 1 and 2, or the Gothenburg stage 3, a status with dyspnea due to cardiac causes and under treatment with digitalis or loop diuretics (n = 455) (14,15); or missing information about incident heart failure during follow-up (n = 245). We also excluded participants with diabetes defined by a fasting glucose of ≥7.0 mmol/l (≥126 mg/dl), nonfasting glucose of ≥11.1 mmol/l (≥200 mg/dl), A1C ≥6.5% (12), self-reported physician diagnosis of diabetes, or use of glucose-lowering medication (n = 2,174) or missing information for diabetes (n = 97) at either of visit 1 or visit 2, for a final study population size of 11,057 participants. The study was approved by the Institutional Review Boards of all participating institutions, and all participants gave informed consent.

Data collection.

ARIC study participants provided information on demographic and behavioral variables and medical history to a trained interviewer at each visit. In this study, we used information obtained at visit 2, unless otherwise noted. Smoking status and alcohol intake were determined by self-report. Participants were asked to bring all medications, which were coded by trained personnel. Information about completed years of education was obtained at visit 1. Certified technicians measured three systolic and diastolic blood pressures with participants in the sitting position after 5 min of rest using a random-zero sphygmomanometer. The average of the second and third readings was recorded. A1C was measured using a high-performance liquid chromatography instrument (Tosoh 2.2 Plus in 2003–2004 and the Tosoh G7 in 2007–2008, Tosoh Corporation, Tokyo, Japan) on all participants with available stored whole blood (16). We have previously demonstrated the reliability of measurements from these stored samples (17). Fasting serum glucose was measured by the optimized DART GLUCOSE reagent method and cholesterol, triglycerides, and HDL cholesterol were determined using enzymatic methods. LDL cholesterol was calculated using the Friedewald equation (18). Insulin was measured by radioimmunoassay (125Insulin kit; Cambridge Medical Diagnosis, Bilerica, MA) at visit 1 (19). Serum creatinine concentration was measured using a modified kinetic Jaffe method. Estimated glomerular filtration rate was computed by the Modification of Diet in Renal Disease Study equation (20). Evidence of atherosclerosis of the common carotid arteries (shadowing/plaque on either side or none) was determined by ultrasound examination (13,21).

Outcome.

ARIC investigators conduct continuous, comprehensive surveillance for all cardiovascular disease-related hospitalizations and deaths in the four communities. Incident heart failure was defined as death from heart failure in any position on the death certificate or as the first heart failure hospitalization with the International Classification of Diseases Code, Ninth Revision (ICD-9) 428 or Tenth Revision (ICD-10) I50 in any position of the hospital discharge list (6). Incident heart failure from visit 2 to January 1, 2006, was analyzed in the present study.

Statistical analyses.

We categorized A1C using the following cut-points: <5.0, 5.0–5.4, 5.5–5.9, and 6.0–6.4%. Baseline characteristics of the population were compared across these A1C categories. We evaluated the continuous association between A1C and the incidence rates of heart failure using a Poisson regression model incorporating linear spline terms for A1C (knots at 5.0, 5.5, and 6.0%) with adjustment for age, sex, and race. Cox proportional hazards models were used to quantify the association between the above categories of A1C and incident heart failure. We tested for interactions using the likelihood-ratio test. For 10,866 participants (98.3%) who provided fasting (≥8 h) blood samples, we also evaluated the association of fasting glucose levels and incident heart failure by using clinical categories of glucose concentration as follows: <5.0, 5.0–5.5, 5.6–6.0, 6.1–6.9 mmol/l (<90, 90–99, 100–109, and 110–125 mg/dl). We used the most prevalent category within a normal range as a reference group for both A1C (5.0–5.4%) and fasting glucose (5.0–5.5 mmol/l) levels.

We implemented three models for the adjustment for covariates. Model 1 was adjusted for age, sex, and race. Model 2 was further adjusted for level of education, carotid atherosclerosis, systolic blood pressure, antihypertensive medication, smoking, alcohol intake, BMI, LDL cholesterol, HDL cholesterol, a self-reported history of coronary heart disease (CHD) by visit 2, clinical examination, or hospital records and estimated glomerular filtration rate. Model 3 was adjusted for all variables in model 2 plus either baseline fasting glucose or A1C, as appropriate. We also investigated the association of quartiles of A1C or fasting glucose with heart failure risk.

We conducted several sensitivity analyses to assess the robustness of our results. First, we examined heart failure occurring in the absence of clinical CHD. To accomplish this, we conducted our analysis censoring incident CHD cases that occurred prior to the date of heart failure (n = 1,088). Second, we repeated our analyses after excluding participants who had incident diabetes in the first 6 years of follow-up (between visit 2 and visit 4) defined by a fasting glucose of ≥7.0 mmol/l, nonfasting glucose of ≥11.1 mmol/l, self-reported physician diagnosis of diabetes, or use of glucose-lowering medication at visit 3 or visit 4 (n = 600). After visit 4, the ARIC Study obtained self-reported information on diabetes diagnosis and medication use by annual telephone calls, for a maximum of 15 years of follow-up. Using this information, we investigated heart failure occurring in the absence of diagnosed diabetes by censoring incident diabetes cases occurring before the heart failure event (n = 1,497).

We also examined the association of 1% unit increase in A1C with heart failure risk. To evaluate whether this association was consistent in groups with low-/high-risk profile, we evaluated this association in the subgroup of participants according to the absence/presence of cardiovascular risk factors. All analyses were conducted using Stata 10.1 software (Stata Corp, College Station, TX) and a P value of <0.05 was considered statistically significant.

RESULTS

Demographic characteristics of participants by categories of A1C are shown in Table 1. Participants with A1C ≥5.5% were more likely to be older, African American, and smokers but less likely to be current drinkers as compared with the reference group (A1C 5.0–5.4%). Individuals with A1C 6.0–6.4% at baseline were also more likely to have higher BMI, higher blood pressure, adverse lipid profile, and higher prevalence of CHD and carotid atherosclerosis. A1C and fasting glucose were weakly, but significantly, correlated (r = 0.32, P < 0.001).

TABLE 1

Characteristics of participants according to categories of A1C

During a median follow-up time of 14.1 years, there were 841cases of incident heart failure. The overall incidence rate of heart failure was 5.7 per 1,000 person-years. The continuous associations of A1C and fasting glucose levels with incidence rate of heart failure with the adjustment for age, sex, and race are shown in Fig. 1. The incidence rate of heart failure increased linearly above A1C 5.0% (tests for differences in slopes above A1C 5.0% were not statistically significant; data not shown) and was ∼2-fold or higher in a range of A1C ≥6.0% as compared with that of A1C 5.0%. Although the incidence rate of heart failure increased at a fasting glucose level of 5.6 mmol/l, the slope was much shallower than that for A1C and was flat at the range of 5.0–5.5 mmol/l. Increased risk of heart failure was observed at the low ranges of A1C (<5.0%) and fasting glucose (<5.0 mmol/l), although 95% CIs were wide, reflecting imprecision of the estimate.

FIG. 1.

Incident rates of heart failure (HF) according to A1C and fasting glucose. The graph shows incidence rates (per 1,000 person-years) and 95% CIs (shaded area) of heart failure with spline terms of A1C (knots at 5.0, 5.5, and 6.0%) (A) and fasting glucose (knots at 5.0, 5.6, and 6.1 mmol/l [90, 100, and 110 mg/dl]) (B) adjusted for age, sex, and race. The histograms represent the frequency distribution of A1C (4.5–6.5%) and fasting glucose (4.2–6.9 mmol/l [75–125 mg/dl]) in the study sample.

We estimated the hazard ratios and corresponding 95% CIs for incident heart failure by categories of A1C using Cox proportional hazards models adjusting for multiple covariates (Table 2). As compared with participants with A1C 5.0–5.4, the hazard ratios of heart failure rose progressively from 1.44 (95% CI, 1.24–1.68) to 2.04 (95% CI, 1.63–2.54) across categories of A1C ≥5.5% in the model adjusted for age, sex, and race (model 1). The association among individuals with A1C 6.0–6.4% remained significant even after adjustment for all traditional cardiovascular risk factors (model 2, hazard ratio 1.38 [95% CI, 1.09–1.75]), although the association among participants with A1C 5.5–5.9% was attenuated to borderline significance (hazard ratio 1.16 [0.98–1.36], P = 0.08). The adjustment for fasting glucose did not alter the results (model 3). These associations did not change appreciably after further adjustment for use of antihypertensive drugs (i.e., β-blockers, ACE inhibitors, or diuretics), which might potentially affect both glucose metabolism and risk of heart failure (data not shown).

TABLE 2

Adjusted hazard ratios (HRs; 95% CI) for incident heart failure (HF) according to A1C categories

There was no evidence of effect modification by a history of CHD at baseline (P for interaction = 0.83), and similar, but slightly attenuated, associations were observed when we censored participants without prevalent CHD at baseline who developed CHD prior to heart failure (hazard ratio 1.27 [95% CI, 0.95–1.70] for A1C 6.0–6.4% and trend P = 0.095). We obtained similar results even after further adjusting for insulin levels. The exclusion of participants who developed diabetes during the first 6 years or censoring participants who self-reported diagnosed diabetes before heart failure during follow-up did not alter the results (data not shown).

Participants with fasting glucose levels of 6.1–6.9 mmol/l but not 5.6–6.0 mmol/l had an increased risk of heart failure as compared with those with glucose levels of 5.0–5.5 mmol/l in model 1 (Table 3). However, the association was greatly attenuated after adjustment for multiple covariates (model 2) and no longer significant when A1C was included in the model (model 3, hazard ratio 1.11 [95% CI, 0.90–1.35]). By contrast, fasting glucose <5.0 mmol/l was associated with higher risk of heart failure as compared with the reference group, even after adjusted for multiple covariates (model 2, hazard ratio 1.51 [1.14–2.00]). This association remained significant even after adjusting for A1C (model 3) or restricting the sample to participants who contribute more than 5 years follow-up time (hazard ratio 1.55 [1.13–2.13]). We repeated the analyses using an average fasting glucose level at visit 1 and visit 2 and obtained similar results (data not shown).

TABLE 3

Adjusted HRs (95% CI) for incident HF according to fasting glucose categories

When we compared the quartiles of A1C and fasting glucose in model 2, the highest quartile of A1C (5.7–6.4%) but not fasting glucose (6.0–6.9 mmol/l [108–125 mg/dl]) was associated with heart failure risk as compared with the lowest quartile of A1C (<5.2%) and fasting glucose (<5.3 mmol/l [<95 mg/dl]) (hazard ratio 1.42 [1.13–1.78] and 1.03 [0.84–1.27], respectively). Similar results were observed when we used the second quartile of fasting glucose (5.3–5.6 mmol/l [95–100 mg/dl]) as the reference group (data not shown).

We also examined the joint association of A1C and fasting glucose with heart failure risk (Table 4). A1C 6.0–6.4% compared with 5.0–5.4% was significantly associated with increased risk for heart failure at fasting glucose levels of 5.0–5.5 mmol/l with similar association at other fasting glucose levels. In contrast, the association of elevated fasting glucose with heart failure was not significant at A1C 5.0–5.4%. Similarly, there was no consistent increase in heart failure risk associated with higher fasting glucose at other A1C categories. Although the relative risk associated with higher A1C tended to be larger among participants with low/normal fasting glucose levels as compared with those with elevated fasting glucose levels, the interaction of A1C and fasting glucose categories on heart failure risk was not significant (P = 0.257).

TABLE 4

Adjusted* HRs (95% CI) for incident HF according to the combination of A1C and fasting glucose categories

Finally, we modeled the association of heart failure risk per 1% unit increase in A1C and examined this association in different subgroups (Fig. 2). Overall, each 1% unit increase in A1C was associated with 39% (95% CI, 13–70%) increased risk of heart failure after adjusted for multiple covariates. These results were largely consistent across the different subpopulations (all Ps for interaction >0.05).

FIG. 2.

Hazard ratios (HRs) of heart failure per 1% unit increase in A1C. Hazard ratios overall and within subgroups adjusted for the same covariates as model 2 in Table 2 are shown. Error bars represent 95% CIs. eGFR = estimated glomerular filtration rate.

DISCUSSION

In this community-based population, we found that elevated A1C even in the range under 6.5% was associated with heart failure risk independently of traditional cardiovascular risk factors in middle-aged individuals during a median of 14 years of follow-up. The risk conferred by 1% unit increase in A1C in our study was 39% and was slightly higher than what has been reported in populations of people with diabetes (36). This association remained significant even after censoring participants who developed diabetes during median follow-up of 14 years before incident heart failure, suggesting that impaired glucose metabolism even before the development of diabetes is an independent risk factor of heart failure.

Given that the association between A1C and heart failure risk was somewhat attenuated by limiting to heart failure cases without preceding CHD, elevated A1C apparently confers heart failure risk partially through its association with increased CHD risk (22). Several other mechanisms might explain the positive association between A1C and risk of heart failure. People with glucose intolerance may have other comorbidities like hypertension or obesity predisposing to the development of heart failure (23). Hyperinsulinemia may also play a role by stimulating sodium retention and/or activating the sympathetic nervous system (8). Insulin is a known growth factor and may contribute to myocardial dysfunction via increases in cardiac mass (8). However, the association of A1C with risk of heart failure in our study was independent of hypertension, obesity, other traditional cardiovascular risk factors, and insulin concentration, suggesting direct effects of hyperglycemia on the development of heart failure. Indeed, hyperglycemia may harm the heart via oxidative stress (7,24). Increased oxidative stress is associated with cell injury or apoptosis, resulting in decreased cardiac contractile (24). Glucose may also interact with collagen and stimulate the production of advanced glycation end products (25). Advanced glycation end products are hypothesized to induce fibrosis in the heart, leading to myocardial stiffness and diastolic dysfunction (25,26).

Whether lowering A1C via lifestyle modification or medication can reduce the risk of heart failure in nondiabetic populations is an important question. To our knowledge, no study has specifically investigated this issue. Some clinical trials (27,28), but not all (29), demonstrated that interventions with lifestyle modification or glucose-lowering medications may reduce cardiovascular risk in individuals with impaired glucose tolerance. The STOP-NIDDM Trial showed reduced risk of cardiovascular events by an α-glucosidase inhibitor, acarbose, but had few heart failure cases (27). Although an increased risk of heart failure with rosiglitazone treatment was reported in the DREAM Trial (29), the elevated risk of heart failure may be specific to this pharmacologic agent (30). Further studies are needed to evaluate whether early interventions to prevent glucose intolerance and lower A1C levels might reduce subsequent heart failure risk.

In contrast to the robust independent association of A1C with increased risk of heart failure, fasting glucose was only weakly associated and no longer significant after adjusting for A1C in our study population. This weak association between fasting glucose and incident heart failure in a nondiabetic population is consistent with the previous literature (7,8) and may result from relatively high variability in glucose measurements (9). Superiority of A1C to fasting glucose for disease prediction has also been observed for CHD, stroke, and total mortality (31). Nevertheless, our results suggest that A1C is a better risk marker of heart failure as compared with fasting glucose in a nondiabetic population.

Participants with a fasting glucose <5.0 mmol/l had an increased risk of heart failure compared with individuals with fasting glucose 5.0–5.5 mmol/l in our study. Such a J-shaped association has been observed for all-cause or cardiovascular mortality in some studies (3234) but not for heart failure (7,8,35). Given that excluding heart failure cases within 5 years of follow-up did not alter our findings, reverse causation is unlikely. There remains the possibility that comorbid conditions associated with both lower glucose concentration and risk of heart failure, e.g., liver dysfunction (36), might account for the observed higher heart failure risk at low serum glucose concentrations.

Several limitations of the present study should be mentioned. First, we evaluated the association of a single measurement of A1C and heart failure. A previous study showed that updated average A1C may predict future heart failure better than a single baseline A1C in patients with diabetes (4). However, in nondiabetic populations, a single A1C measurement is highly reliable (37,38). Second, as with any observational study, we cannot rule out the possibility of residual confounding despite adjustment for all major cardiovascular risk factors. Third, identification of heart failure cases relied entirely on ICD codes abstracted from hospital records and death certificates (6,39). Reliance on hospital discharge codes may underestimate heart failure incidence (40). Finally, because the ARIC Study consists of a middle-aged, bi-ethnic community-based population of the U.S., additional studies are needed in younger populations, the elderly, or other ethnicities.

In conclusion, elevated A1C (≥5.5–6.0%) was more strongly associated with increased risk of heart failure as compared with fasting glucose in a middle-aged bi-ethnic population without diabetes. Our findings suggest that chronic hyperglycemia even before the development of diabetes is an independent risk factor of heart failure and may contribute to the development of heart failure beyond its effect on CHD risk.

ACKNOWLEDGMENTS

The ARIC Study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts N01-HC-55015, N01-HC-55016, N01-HC-55018, N01-HC-55019, N01-HC-55020, N01-HC-55021, and N01-HC-55022. This research was supported by NIH/NIDDK Grant R21-DK-080294. S.B. was supported by NIH/NHLBI Grant 5T32-HL-007024. E.S. was also supported by NIH/NIDDK Grant K01-DK-076595. J.C. was also supported by NIH/NIDDK Grant R01-DK-076770. No potential conflicts of interest relevant to this article were reported.

K.M. analyzed the data and wrote the manuscript. S.B., A.P.-F., A.B., P.P.C., and J.C. contributed to discussion and reviewed/edited the manuscript. E.S. collected the data, contributed to discussion, and reviewed/edited the manuscript.

The authors thank the staff and participants of the ARIC Study for their important contributions.

Footnotes

  • The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

  • See accompanying commentary, p. 1868.

  • Received February 2, 2010.
  • Accepted May 7, 2010.

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. See http://creativecommons.org/licenses/by-nc-nd/3.0/ for details.

REFERENCES

| Table of Contents

This Article

  1. Diabetes vol. 59 no. 8 2020-2026
  1. All Versions of this Article:
    1. db10-0165v1
    2. 59/8/2020 most recent