HHS Public Access

Author manuscript
Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
Author Manuscript

Published in final edited form as:

Circ Cardiovasc Qual Outcomes. 2010 November ; 3(6): 661–667. doi:10.1161/CIRCOUTCOMES.
110.957936.

Low Hemoglobin A1c and Risk of All-Cause Mortality Among US

Adults Without Diabetes
April P. Carson, PhD, Caroline S. Fox, MD, MPH, Darren K. McGuire, MD, MHSc, Emily B.
Levitan, ScD, Martin Laclaustra, MD, PhD, Devin M. Mann, MD, MS, and Paul Muntner, PhD
Department of Epidemiology (A.P.C., E.B.L., P.M.), University of Alabama at Birmingham,
Birmingham, Ala; the National Heart, Lung, and Blood Institute’s Framingham Heart Study
Author Manuscript

(C.S.F.), Framingham, Mass; the Department of Endocrinology and Metabolism (C.S.F.), Brigham
and Women’s Hospital and Harvard Medical School, Boston, Mass; the Department of Internal
Medicine (D.K.M.), University of Texas Southwestern Medical Center at Dallas, Dallas, Tex; the
Department of Cardiovascular Epidemiology and Population Genetics (M.L.), National Center for
Cardiovascular Research (CNIC), Madrid, Spain; the Division of General Internal Medicine
(D.M.M.), Mount Sinai School of Medicine, New York, NY; and the Department of Medicine
(P.M.), University of Alabama at Birmingham, Birmingham, Ala

Abstract
Background—Among individuals without diabetes, elevated hemoglobin A1c (HbA1c) has
been associated with increased morbidity and mortality, but the literature is sparse regarding the
prognostic importance of low HbA1c.
Author Manuscript

Methods and Results—National Health and Nutrition Examination Survey III (NHANES III)
participants, 20 years and older, were followed up to 12 years (median follow-up, 8.8 years) for
all-cause mortality. Cox proportional hazards regression was used to calculate hazard ratios (HR)
and 95% confidence intervals (CI) for the association between HbA1c levels and all-cause
mortality for 14 099 participants without diabetes. There were 1825 deaths during the follow-up
period. Participants with a low HbA1c (<4.0%) had the highest levels of mean red blood cell
volume, ferritin, and liver enzymes and the lowest levels of mean total cholesterol and diastolic
blood pressure compared with their counterparts with HbA1c levels between 4.0% and 6.4%. An
HbA1c <4.0% versus 5.0% to 5.4% was associated with an increased risk of all-cause mortality
(HR, 3.73; 95% CI, 1.45 to 9.63) after adjustment for age, race-ethnicity, and sex. This association
was attenuated but remained statistically significant after further multivariable adjustment for
lifestyle, cardiovascular factors, metabolic factors, red blood cell indices, iron storage indices, and
Author Manuscript

liver function indices (HR, 2.90; 95% CI, 1.25 to 6.76).

Conclusions—In this nationally representative cohort, low HbA1c was associated with
increased all-cause mortality among US adults without diabetes. Additional research is needed to
confirm these results and identify potential mechanisms that may be underlying this association.

Correspondence to April P. Carson, PhD, Department of Epidemiology, University of Alabama at Birmingham, 1530 3rd Ave South,
RPHB 230N, Birmingham, AL 35294-0022. apcarson@uab.edu.
Disclosures
Dr McGuire received consultancy fees <$10 000 from Tethys Bioscience, Biosite, Inc, F. Hoffmann La Roche, and Daiichi Sankyo.
 Carson et al. Page 2
Author Manuscript

Keywords
hemoglobin A1c; epidemiology; mortality

Elevated hemoglobin A1c (HbA1c) is associated with an increased risk of cardiovascular

events among individuals without diabetes.1,2 Additionally, a recent study identified HbA1c
as a better predictor of cardiovascular events compared with fasting plasma glucose.3
Recently, the American Diabetes Association updated its clinical practice recommendations
to include HbA1c as a diagnostic test for diabetes,4 which may lead to its widespread use.
Along with the broad use of this assay, low HbA1c values, including very low values
<4.0%, may be frequently detected. HbA1c ≥6.5% has been shown to be reasonable for
diagnosing diabetes,5 but it is currently uncertain how to interpret low HbA1c values.

Very low HbA1c values among persons without diabetes may reflect underlying biological
Author Manuscript

processes. Certain health conditions that decrease erythrocyte life span (eg, iron-deficiency
anemia) are known to alter HbA1c values and make them unreliable.6 However, little is
known about other biological factors that result in low HbA1c values among individuals
without diabetes. Low HbA1c may not reflect metabolic control among individuals without
diabetes but may be reflecting other biological factors, such as red blood cell markers,
inflammation, or decreased liver function. The purpose of the present study is to describe the
health characteristics of individuals with low HbA1c levels, evaluate the association
between low HbA1c and all-cause mortality, and assess the underlying biological processes
that explain the association in a nationally, representative sample of US adults without
diabetes.

Methods
Author Manuscript

Study Population
The National Health and Nutrition Examination Survey (NHANES III) is a stratified,
multistage probability survey designed to select a representative sample of the civilian
noninstitutionalized US population. Overall, 18 825 adults 20 years of age and older
completed the NHANES III interview and examination between 1988 and 1994. Through
linkage with the National Death Index, participants were followed for mortality through
December 31, 2000. We excluded 3022 participants without plasma glucose or HbA1c
measurements and 19 participants who did not have follow-up information. We also
excluded 1685 participants with diabetes, defined as a fasting plasma glucose ≥126 mg/dL, a
nonfasting plasma glucose ≥200 mg/dL, an HbA1c ≥6.5%, and/or a self-reported history of
diabetes with concurrent use of antidiabetes medication. After these exclusions, 14 099
Author Manuscript

NHANES III participants without diabetes were included in the current analyses. The
protocol for NHANES III was approved by the National Center for Health Statistics of the
Centers for Disease Control and Prevention Institutional Review Board. All participants
gave informed consent.

WHAT IS KNOWN

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
 Carson et al. Page 3

• The widespread use of hemoglobin A1c as a diagnostic test for diabetes may
Author Manuscript

lead to the frequent detection of low hemoglobin A1c values.

• The prognostic importance of low hemoglobin A1c values is unknown, but the
natural assumption of clinicians may be that low HbA1c is beneficial.

WHAT THE STUDY ADDS

• Hemoglobin values as low as 2.8% were detected in this nationally

representative study.

• Low hemoglobin A1c (<4.0%) was associated with an increased risk of all-
cause mortality.

• Other biological processes, such as inflammation and liver function, may

underlie the association between low hemoglobin A1c (<4.0%) and all-cause
Author Manuscript

mortality.

Baseline Data Collection

NHANES III baseline data were collected during an in-home interview and a subsequent
visit to a mobile examination center when a physician conducted a medical evaluation.
During the in-home interview, researchers collected sociodemographic, behavioral, and
health-related information including prior diagnoses and current medications using a
standardized questionnaire. The medical evaluation included a physical examination,
anthropometric and blood pressure measurements, and blood and spot-urine collection.

HbA1c was analyzed at the University of Missouri at Columbia using standardized

laboratory protocols and quality control procedures. HbA1c was measured using a Bio-Rad
Author Manuscript

Diamat ion exchange high-performance liquid chromatography system. If hemoglobin

variants (hemoglobin C, D, F, and S) were present, an alternative standardized affinity
chromatographic method was used. HbA1c results <4% or >11% were repeated for
verification. Two quality control approaches were used for HbA1c: (1) batch quality control
specimens placed in each run, and (2) 5% sample specimens randomly selected from each
batch which were reanalyzed in another batch. The coefficients of variation ranged from
2.0% to 3.1% for low HbA1c quality control pools and 1.1% to 2.3% for high HbA1c
quality control pools.7

Mortality Follow-Up
The current analyses use participant follow-up for mortality through December 31, 2000.
The method of probabilistic matching was used to link NHANES III participants with the
Author Manuscript

National Death Index to ascertain vital status. Identical matching methodology applied to the
NHANES I Epidemiological Follow-up Study for validation purposes found that 96.1% of
deceased participants and 99.4% of living participants were correctly classified.8

Statistical Methods
Characteristics of participants were calculated by a priori–defined groups of HbA1c levels
(<4.0%, 4.0% to 4.4%, 4.5% to 4.9%, 5.0% to 5.4%, 5.5% to 5.9%, and 6.0% to 6.4%) as

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
 Carson et al. Page 4

means for continuous variables and proportions for categorical variables. All-cause mortality
Author Manuscript

rates, standardized to the age distribution of US adults without diabetes, were calculated by
HbA1c levels. Follow-up for each study participant was calculated as the time between their
NHANES III examination and the date of death or December 31, 2000, for those still alive
at the end of follow-up.

Cox proportional hazards regression was used to obtain the hazard ratios (HR) and 95%
confidence intervals (CI) for all-cause mortality associated with HbA1c levels. The variables
for the models were selected and grouped into 7 categories, based on possible associations
with HbA1c and all-cause mortality: (1) demographic factors including age, race-ethnicity,
and sex; (2) lifestyle factors including education, income, current smoking, alcohol
consumption, physical inactivity, body mass index, and aspirin use; (3) cardiovascular
factors including systolic blood pressure, antihypertensive medication use, total and HDL
cholesterol, log-transformed triglycerides, C-reactive protein, and history of cardiovascular
Author Manuscript

disease (CVD); (4) metabolic factors including prior diagnosis of thyroid disease, thyroid-
stimulating hormone, estimated glomerular filtration rate, and albuminuria; (5) red blood
cell indices including hemoglobin, red blood cell distribution width, mean cell volume, and
serum folate; (6) iron storage indices including serum albumin, ferritin, and transferrin
saturation; and (7) liver function indices including hepatitis C seropositivity, aspartate
aminotransferase (AST), and alanine aminotransferase (ALT). The initial model adjusted for
demographic factors. The next set of models adjusted for demographic factors plus each of
the other variable groups listed above, separately. The final model adjusted for demographic
factors plus all of the other variable groups, simultaneously. To further explore the HR for
all-cause mortality across the full range of HbA1c levels, we used restricted quadratic
splines with knots at the 2.5, 10, 50, 90, and 97.5 percentiles (4.3%, 4.7%, 5.3%, 5.9%, and
6.2%, respectively) in a full multivariable adjusted Cox regression model.
Author Manuscript

Sensitivity analyses were conducted after excluding participants with cancer, CVD, anemia,
and hepatitis C in the full multivariable adjusted model. The proportional hazards
assumption was assessed using Schoenfeld residuals, and no violations were noted. Data
were analyzed using SUDAAN (version 9.0; Research Triangle Institute, Research Triangle
Park, NC) and R (version 2.6.1; R Foundation for Statistical Computing, Vienna, Austria) to
account for the complex NHANES sampling design, including unequal probabilities of
selection, oversampling, and nonresponse.

Results
Characteristics of Participants Without Diabetes by HbA1c Level
Author Manuscript

The range of HbA1c values was 2.8% to 6.4%, and 97 participants had an HbA1c <4.0%.
Participants with an HbA1c between 4.0% and 4.4% were the youngest; those with an
HbA1c of 6.0 to 6.4% were the oldest (Table 1). A greater proportion of participants with
HbA1c <5.0% were women. J-shaped or U-shaped associations were present for age, non-
Hispanic black race-ethnicity, less than a high school education, and annual household
income <$20 000. Mean cell volume, ferritin, transferrin saturation, and liver function
measures were elevated in the HbA1c <4.0% group, whereas mean systolic blood pressure,
diastolic blood pressure, total cholesterol, and median triglyceride levels were lowest among

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
 Carson et al. Page 5

those with an HbA1c <4.0%. Also, the proportion of participants with a history of CVD
Author Manuscript

disease was lowest among those with an HbA1c <4.0%, whereas the proportion of
participants with hepatitis C was the highest in this group.

HbA1c and All-Cause Mortality Among Participants Without Diabetes

Over a median follow-up of 8.8 years, 1825 deaths occurred among participants without
diabetes. The age-standardized mortality rates by HbA1c level for participants without
diabetes and the HRs (95% CIs) for all-cause mortality associated with HbA1c levels are
presented in Table 2. Using the group with HbA1c of 5.0% to 5.4% as the reference, HbA1c
levels <4.0% were associated with an increased risk of all-cause mortality after adjustment
for age, race-ethnicity, and sex (HR, 3.73; 95% CI, 1.45 to 9.63). The HRs increased after
separate adjustment for lifestyle factors, cardiovascular factors, or iron storage. In contrast,
the HRs were attenuated after separate adjustment for metabolic factors (HR, 3.47; 95% CI,
Author Manuscript

1.36 to 8.82), red blood cell indices (HR, 2.68; 95% CI, 1.08 to 6.66), or liver function
measures (HR, 3.09; 95% CI, 1.20 to 7.92). In the multivariable adjusted model including all
variable categories, the HR for all-cause mortality among those with an HbA1c <4.0%
versus 5.0% to 5.4% remained almost 3 times as high and statistically significant (HR, 2.90;
95% CI, 1.25 to 6.76).

Elevated HbA1c among participants without diabetes was also associated with an increased
risk of all-cause mortality. An HbA1c of 6.0% to 6.4%, compared with an HbA1c of 5.0% to
5.4%, was associated with an increased risk of all-cause mortality after adjustment for age,
race-ethnicity, and sex (HR, 1.28; 95% CI, 1.06 to 1.54). This association persisted after
separate adjustment for the different variable groups and in the full multivariable adjusted
model (HR, 1.30; 95% CI, 1.06 to 1.60).
Author Manuscript

Using a multivariable adjusted quadratic spline with HbA1c=5.2% as the cut-point, low and
high HbA1c values were associated with an increased risk of all-cause mortality among
individuals without diabetes (Figure).

Sensitivity Analyses
Sensitivity analyses excluding participants with cancer, CVD, anemia, or hepatitis C are
presented in Table 3. Among participants with an HbA1c <4.0%, excluding those with
anemia increased the effect estimate (HR, 3.56; 95% CI, 1.38 to 9.17), whereas excluding
those with hepatitis C attenuated the association (HR, 2.26; 95% CI, 0.96 to 5.42). After
excluding participants with cancer, CVD, anemia, or hepatitis C, the association between
low HbA1c and all-cause mortality persisted (HR, 2.81; 95% CI, 1.03 to 7.67). For
participants with an HbA1c of 6.0% to 6.4%, compared with participants with an HbA1c of
Author Manuscript

5.0% to 5.4%, the association with all-cause mortality also persisted (HR, 1.41; 95% CI,
1.05 to 1.89) after excluding participants with cancer, CVD, anemia, or hepatitis C.

Discussion
In the present study of US adults, very low HbA1c was associated with an increased risk of
all-cause mortality among persons without diabetes. Specifically, an HbA1c<4.0% was
associated with a large, statistically significant association with all-cause mortality that

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
 Carson et al. Page 6

persisted after multivariable adjustment for demographic, lifestyle, cardiovascular factors,

Author Manuscript

metabolic factors, red blood cell indices, iron storage, and liver function indices. Although
participants with HbA1c <4.0% had a generally favorable cardiovascular profile at their
baseline NHANES III examination, red blood cell indices, iron storage, and liver function
measures were generally adverse compared with participants with higher HbA1c levels.
However, participants with an HbA1c <4.0% still had an increased risk of all-cause
mortality after adjusting for these differences.

Low HbA1c values are likely to be detected among persons without diabetes as the use of
HbA1c as a diagnostic test for diabetes becomes part of routine clinical practice, as recently
recommended by the American Diabetes Association.4 However, the clinical relevance of
low HbA1C values remains unclear, with little known about the biological underpinning of
such low values or the diagnostic and prognostic utility of such observations. Although
several studies have investigated the association of HbA1c levels with morbidity and
Author Manuscript

mortality among persons without diabetes,9–14 they primarily have focused on the
association between higher levels of HbA1c and health outcomes. Elevated HbA1c has been
associated with an increased risk of cardiovascular morbidity1,10,12,15 and mortality9,14,16–18
among persons without diabetes, although some of these associations were attenuated after
adjustment for cardiovascular factors. In contrast, a couple of studies have also reported that
elevated HbA1c is not associated with cardiovascular morbidity13 and mortality11 among
persons without diabetes. In the present study, participants with an HbA1c of 5.5% to 5.9%
or 6.0% to 6.4% had an increased risk of all-cause mortality that remained after adjusting for
various biological factors.

Few of the studies investigating outcomes by HbA1c level for individuals without diabetes
have reported the morbidity and mortality risk associated with low levels. The lowest
Author Manuscript

HbA1c quintile evaluated in the Women’s Health Study included values between 2.27% and
4.79%, with a median HbA1c of 4.7%.18 The association between HbA1c and mortality,
modeled using a cubic spline, indicated an increased risk for mortality at higher levels of
HbA1c, but no increased risk was observed at lower HbA1c levels. In contrast, in the
Atherosclerosis Risk in Communities Study, low HbA1c levels (<5.0%) were not associated
with coronary heart disease, ischemic stroke, or diabetes but were associated with an
increased risk of all-cause mortality after adjusting for cardiovascular risk factors and
baseline fasting glucose.3

It is not clear what biological processes may be underlying the association between low
HbA1c and all-cause mortality. In the present study, participants with low HbA1c values
had unfavorable profiles of red blood cell related factors, iron storage, and liver function.
Author Manuscript

Red blood cell distribution width has been associated with an increased risk of morbidity
and mortality19–22 and with inflammation.21 Younger red blood cells are typically larger and
have more size variability compared with older red blood cells,20 which could affect red
blood cell distribution width and HbA1c values.6,23 Also, iron stores were elevated among
those with low HbA1c in the present study. Iron-deficiency anemia has been associated with
increased HbA1c values among individuals without diabetes; however, patients with iron-
deficiency anemia who were treated with iron therapy had decreases in HbA1c values.24,25
Also, ferritin was elevated among those with low HbA1c in the present study. Elevated

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
 Carson et al. Page 7

ferritin has been associated with an increased risk of myocardial infarction26 and
Author Manuscript

atherosclerosis,27 but it is also an acute-phase reactant and could be reflecting an

inflammatory response. Finally, liver function enzymes were elevated and hepatitis C was
more prevalent among those with an HbA1c <4.0% in the present study. Elevated ALT is
frequently associated with fatty liver and an adverse cardiometabolic risk factor profile.
ALT was positively associated with HbA1c levels among participants with and without
diabetes in the British Women’s Heart and Health Study, with a stronger association noted
among women without diabetes.28 Certain combination therapies for hepatitis C have been
reported to temporarily lower HbA1c levels29; however, the low HbA1c-mortality
association observed in the present study was still evident after excluding participants with
hepatitis C. The overall biomarker profile of the participants without diabetes and an HbA1c
< 4.0% is suggestive of red blood cell markers, inflammation, and liver function as part of
the biological underpinning for the association noted in this study. However, the low HbA1c
Author Manuscript

and all-cause mortality association observed in this study persisted after adjustment for
lifestyle, cardiovascular, metabolic, red blood cell, iron storage, and liver function indices.
The association also persisted after excluding participants with cancer, CVD, anemia, or
hepatitis C in sensitivity analyses. Additional research is needed to explore the potential
health effects of aberrations in red blood cell markers, inflammation, and liver function
indices and associations with all-cause mortality among individuals with a low HbA1c.

The present study should be interpreted in the context of possible limitations. HbA1c results
may be affected by hemoglobin variants and conditions that affect erythrocyte life span (eg,
hemolytic anemia).6,30 The study protocol included an alternative testing method for
hemoglobin variants, but data were not available to assess decreased red blood cell survival.
When participants with anemia and other conditions were excluded from the analysis, low
HbA1c remained associated with all-cause mortality. Despite these limitations, this study
Author Manuscript

has multiple strengths including the large, ethnically diverse, and representative sample of
US adults without diabetes. Additionally, NHANES III included a broad collection of data
using a standardized protocol and core laboratory measurement of HbA1c with quality
control procedures that allowed for the assessment of extensive participant characteristics by
HbA1c level.

The present findings may have important clinical implications as the use of HbA1c as a
diagnostic test for diabetes becomes more common. Low HbA1c levels will be identified in
clinical practice and the natural assumption may be that low HbA1c is beneficial. However,
the findings from the present study suggest that there may be a threshold below which
HbA1c is associated with an increased all-cause mortality risk among persons without
diabetes. Additional research is needed to confirm these results and identify the putative
Author Manuscript

mechanisms involved in this association.

Acknowledgments
Sources of Funding

Dr Laclaustra is supported by CP08/00112 “Miguel Servet” Grant (Instituto de Salud Carlos III, Spain).

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
 Carson et al. Page 8

References
Author Manuscript

1. Selvin E, Coresh J, Golden SH, Brancati FL, Folsom AR, Steffes MW. Glycemic control and
coronary heart disease risk in persons with and without diabetes: the atherosclerosis risk in
communities study. Arch Intern Med. 2005; 165:1910–1916. [PubMed: 16157837]
2. Selvin E, Marinopoulos S, Berkenblit G, Rami T, Brancati FL, Powe NR, Golden SH. Meta-
analysis: glycosylated hemoglobin and cardiovascular disease in diabetes mellitus. Ann Intern Med.
2004; 141:421–431. [PubMed: 15381515]
3. Selvin E, Steffes MW, Zhu H, Matsushita K, Wagenknecht L, Pankow J, Coresh J, Brancati FL.
Glycated hemoglobin, diabetes, and cardiovascular risk in nondiabetic adults. N Engl J Med. 2010;
362:800–811. [PubMed: 20200384]
4. Standards of medical care in diabetes: 2010. Diabetes Care. 2010; 33(Suppl 1):S11–S61. [PubMed:
20042772]
5. Carson AP, Reynolds K, Fonseca VA, Muntner P. Comparison of A1C and fasting glucose criteria
to diagnose diabetes among US adults. Diabetes Care. 2010; 33:95–97. [PubMed: 19808920]
6. Little RR, Sacks DB. HbA1c: how do we measure it and what does it mean? Curr Opin Endocrinol
Author Manuscript

Diabetes Obes. 2009; 16:113–118. [PubMed: 19300091]

7. Centers for Disease Control and Prevention (CDC). National Center for Health Statistics (NCHS).
Laboratory Procedures Used for Third National Health and Nutrition Examination Survey
(NHANES III), 1988–1994. US Department of Health and Human Services, Centers for Disease
Control and Prevention; National Health and Nutrition Examination Survey. Available at: http://
www.cdc.gov/nchs/data/nhanes/nhanes3/cdrom/nchs/manuals/labman.pdf [Accessed March 2010]
8. Centers for Disease Control and Prevention (CDC). National Center for Health Statistics (NCHS).
NHANES I Epidemiologic Follow-Up Survey: Calibration sample for NDI matching methodology.
US Department of Health and Human Services, Centers for Disease Control and Prevention;
National Health and Nutrition Examination Survey. Available at: http://www.cdc.gov/nchs/data/
datalinkage/mort_calibration_study.pdf [Accessed March 2010]
9. Barr EL, Boyko EJ, Zimmet PZ, Wolfe R, Tonkin AM, Shaw JE. Continuous relationships between
non-diabetic hyperglycaemia and both cardiovascular disease and all-cause mortality: the Australian
Diabetes, Obesity, and Lifestyle (AusDiab) study. Diabetologia. 2009; 52:415–424. [PubMed:
19130039]
Author Manuscript

10. Blake GJ, Pradhan AD, Manson JE, Williams GR, Buring J, Ridker PM, Glynn RJ. Hemoglobin
A1c level and future cardiovascular events among women. Arch Intern Med. 2004; 164:757–761.
[PubMed: 15078645]
11. Chonchol M, Katz R, Fried LF, Sarnak MJ, Siscovick DS, Newman AB, Strotmeyer ES, Bertoni
A, Shlipak MG. Glycosylated hemoglobin and the risk of death and cardiovascular mortality in the
elderly. Nutr Metab Cardiovasc Dis. 2010; 20:15–21. [PubMed: 19364638]
12. Meigs JB, Nathan DM, D’Agostino RB Sr, Wilson PW. Fasting and postchallenge glycemia and
cardiovascular disease risk: the Framingham Offspring Study. Diabetes Care. 2002; 25:1845–
1850. [PubMed: 12351489]
13. Pradhan AD, Rifai N, Buring JE, Ridker PM. Hemoglobin A1c predicts diabetes but not
cardiovascular disease in nondiabetic women. Am J Med. 2007; 120:720–727. [PubMed:
17679132]
14. Qiao Q, Dekker JM, de Vegt F, Nijpels G, Nissinen A, Stehouwer CD, Bouter LM, Heine RJ,
Tuomilehto J. Two prospective studies found that elevated 2-hr glucose predicted male mortality
Author Manuscript

independent of fasting glucose and HbA1c. J Clin Epidemiol. 2004; 57:590–596. [PubMed:
15246127]
15. Khaw KT, Wareham N, Bingham S, Luben R, Welch A, Day N. Association of hemoglobin A1c
with cardiovascular disease and mortality in adults: the European prospective investigation into
cancer in Norfolk. Ann Intern Med. 2004; 141:413–420. [PubMed: 15381514]
16. de Vegt F, Dekker JM, Ruhe HG, Stehouwer CD, Nijpels G, Bouter LM, Heine RJ.
Hyperglycaemia is associated with all-cause and cardiovascular mortality in the Hoorn population:
the Hoorn Study. Diabetologia. 1999; 42:926–931. [PubMed: 10491751]

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
 Carson et al. Page 9

17. Khaw KT, Wareham N, Luben R, Bingham S, Oakes S, Welch A, Day N. Glycated haemoglobin,
diabetes, and mortality in men in Norfolk cohort of European prospective investigation of cancer
Author Manuscript

and nutrition (EPIC-Norfolk). BMJ. 2001; 322:15–18. [PubMed: 11141143]

18. Levitan EB, Liu S, Stampfer MJ, Cook NR, Rexrode KM, Ridker PM, Buring JE, Manson JE.
HbA1c measured in stored erythrocytes and mortality rate among middle-aged and older women.
Diabetologia. 2008; 51:267–275. [PubMed: 18043905]
19. Felker GM, Allen LA, Pocock SJ, Shaw LK, McMurray JJ, Pfeffer MA, Swedberg K, Wang D,
Yusuf S, Michelson EL, Granger CB. Red cell distribution width as a novel prognostic marker in
heart failure: data from the CHARM Program and the Duke Databank. J Am Coll Cardiol. 2007;
50:40–47. [PubMed: 17601544]
20. Patel KV, Ferrucci L, Ershler WB, Longo DL, Guralnik JM. Red blood cell distribution width and
the risk of death in middle-aged and older adults. Arch Intern Med. 2009; 169:515–523. [PubMed:
19273783]
21. Perlstein TS, Weuve J, Pfeffer MA, Beckman JA. Red blood cell distribution width and mortality
risk in a community-based prospective cohort. Arch Intern Med. 2009; 169:588–594. [PubMed:
19307522]
Author Manuscript

22. Tonelli M, Sacks F, Arnold M, Moye L, Davis B, Pfeffer M. Relation between red blood cell
distribution width and cardiovascular event rate in people with coronary disease. Circulation.
2008; 117:163–168. [PubMed: 18172029]
23. Cohen RM, Franco RS, Khera PK, Smith EP, Lindsell CJ, Ciraolo PJ, Palascak MB, Joiner CH.
Red cell life span heterogeneity in hematologically normal people is sufficient to alter HbA1c.
Blood. 2008; 112:4284–4291. [PubMed: 18694998]
24. Coban E, Ozdogan M, Timuragaoglu A. Effect of iron deficiency anemia on the levels of
hemoglobin A1c in nondiabetic patients. Acta Haematol. 2004; 112:126–128. [PubMed:
15345893]
25. Tarim O, Kucukerdogan A, Gunay U, Eralp O, Ercan I. Effects of iron deficiency anemia on
hemoglobin A1c in type 1 diabetes mellitus. Pediatr Int. 1999; 41:357–362. [PubMed: 10453183]
26. Salonen JT, Nyyssonen K, Korpela H, Tuomilehto J, Seppanen R, Salonen R. High stored iron
levels are associated with excess risk of myocardial infarction in eastern Finnish men. Circulation.
1992; 86:803–811. [PubMed: 1516192]
27. Kiechl S, Willeit J, Egger G, Poewe W, Oberhollenzer F. Body iron stores and the risk of carotid
Author Manuscript

atherosclerosis: prospective results from the Bruneck study. Circulation. 1997; 96:3300–3307.
[PubMed: 9396420]
28. Fraser A, Ebrahim S, Smith GD, Lawlor DA. A comparison of associations of alanine
aminotransferase and gamma-glutamyltransferase with fasting glucose, fasting insulin, and
glycated hemoglobin in women with and without diabetes. Hepatology. 2007; 46:158–165.
[PubMed: 17596883]
29. Greenberg PD, Rosman AS, Eldeiry LS, Naqvi Z, Brau N. Decline in haemoglobin A1c values in
diabetic patients receiving interferon-alpha and ribavirin for chronic hepatitis C. J Viral Hepat.
2006; 13:613–617. [PubMed: 16907848]
30. Sacks DB, Bruns DE, Goldstein DE, Maclaren NK, McDonald JM, Parrott M. Guidelines and
recommendations for laboratory analysis in the diagnosis and management of diabetes mellitus.
Clin Chem. 2002; 48:436–472. [PubMed: 11861436]
Author Manuscript

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
 Carson et al. Page 10
Author Manuscript
Author Manuscript

Figure.
Adjusted hazard ratios for the association between HbA1c and all-cause mortality among
participants without diabetes using a quadratic spline with knots at the 2.5, 10, 50, 90, and
97.5 percentiles. Adjusted for age, race-ethnicity, sex, lifestyle factors (education, income,
current smoking, alcohol consumption, physical activity, body mass index, and aspirin use),
cardiovascular factors (systolic blood pressure, antihypertensive medication use, total
cholesterol, HDL cholesterol, log triglycerides, elevated C-reactive protein, and history of
CVD), metabolic factors (prior diagnosis of thyroid disease, thyroid-stimulating hormone,
estimated glomerular filtration rate, and albuminuria), red blood cell indices (hemoglobin,
red blood cell distribution width, mean cell volume, and serum folate), iron storage indices
(serum albumin, ferritin, and transferrin saturation), and liver function indices (hepatitis C
seropositivity, AST, and ALT). Knots were placed at 4.3%, 4.7%, 5.3%, 5.9%, and 6.2%,
representing HbA1c levels at the 2.5, 10, 50, 90, and 97.5 percentiles; shaded area represents
Author Manuscript

95% CI.
Author Manuscript

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
 Author Manuscript Author Manuscript Author Manuscript Author Manuscript

Table 1

Characteristics of NHANES III Participants Without Diabetes (n=14 099) by HbA1c Levels, 1988 to 1994

HbA1c%
Carson et al.

<4.0 (n=97) 4.0 – 4.4 (n=469) 4.5– 4.9 (n=2675) 5.0 –5.4 (n=5691) 5.5–5.9 (n=4035) 6.0 – 6.4 (n=1132)
Demographics
Age, y 40.1 (2.3) 34.6 (0.7) 36.5 (0.4) 42.9 (0.5) 51.6 (0.7) 57.3 (0.9)
Women, % 57.1 60.2 59.8 50.4 46.1 50.4
Race-ethnicity
Non-Hispanic white, % 64.3 81.4 83.5 78.4 71.6 58.9
Non-Hispanic black, % 27.0 10.2 5.9 8.3 14.5 25.2
Mexican-American, % 2.8 3.0 4.4 5.3 5.6 4.7
Other, % 5.9 5.4 6.2 8.0 8.3 11.3
Lifestyle factors
Less than high school education, % 18.1 13.5 17.5 21.6 32.2 39.7
Income <$20 000, % 42.3 31.6 26.8 30.4 37.7 44.8
Current smoker, % 20.5 19.5 25.3 29.9 32.4 36.6
Consumes alcohol, % 52.7 65.0 62.1 56.8 47.8 38.0
Body mass index, kg/m2 25.4 (0.9) 24.7 (0.2) 25.1 (0.1) 26.1 (0.1) 27.5 (0.1) 28.7 (0.4)
Physically inactive, % 17.2 18.1 17.3 19.4 27.1 30.3
Aspirin use, % 21.5 36.8 36.2 37.9 39.4 39.5
Cardiovascular factors
Systolic blood pressure, mm Hg 115.0 (2.6) 115.0 (0.6) 115.8 (0.4) 120.8 (0.5) 127.0 (0.5) 132.3 (0.9)
Diastolic blood pressure, mm Hg 67.2 (1.6) 72.2 (0.5) 72.2 (0.3) 74.3 (0.2) 75.6 (0.2) 76.5 (0.5)
Antihypertensive medication use, % 3.3 5.3 6.6 8.9 14.6 28.0
Total cholesterol, mmol/L 4.42 (0.13) 4.81 (0.08) 4.98 (0.03) 5.25 (0.03) 5.54 (0.03) 5.73 (0.05)

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
HDL cholesterol, mmol/L 1.39 (0.05) 1.42 (0.03) 1.36 (0.01) 1.33 (0.01) 1.27 (0.01) 1.22 (0.02)
Triglycerides, mmol/L, median (interquartile range) 1.00 (0.73–1.26) 1.03 (0.75–1.54) 1.12 (0.79–1.68) 1.20 (0.86–1.80) 1.42 (0.98–2.08) 1.55 (1.07–2.37)
Elevated CRP, % 22.4 18.6 22.0 26.5 35.8 49.4
History of cardiovascular disease, % 0.0 1.0 1.6 3.2 6.8 9.5
Metabolic factors
Prior diagnosis of thyroid disease, % 1.9 4.3 2.7 4.1 5.5 4.9
Page 11
 Author Manuscript Author Manuscript Author Manuscript Author Manuscript

HbA1c%

<4.0 (n=97) 4.0 – 4.4 (n=469) 4.5– 4.9 (n=2675) 5.0 –5.4 (n=5691) 5.5–5.9 (n=4035) 6.0 – 6.4 (n=1132)
Thyroid-stimulating hormone, mU/L 1.9 (0.1) 1.7 (0.1) 1.8 (0.04) 2.0 (0.04) 2.4 (0.2) 2.6 (0.3)
Estimated GFR, mL/min/1.73 m2 104.6 (4.6) 100.6 (1.9) 98.6 (0.6) 93.5 (0.6) 87.8 (0.6) 83.1 (1.1)
Carson et al.

Albuminuria, % 13.9 9.3 5.3 7.7 10.7 17.5

Fasting plasma glucose, mmol/L* 4.93 (0.09) 4.98 (0.05) 5.10 (0.02) 5.23 (0.02) 5.49 (0.02) 5.74 (0.04)

Red blood cell indices

Hemoglobin, g/L 132 (2) 138 (1) 140 (1) 142 (1) 142 (1) 141 (1)
Red blood cell distribution width, % 13.54 (0.26) 12.65 (0.06) 12.71 (0.02) 12.88 (0.02) 13.22 (0.03) 13.52 (0.04)
Mean cell volume, fL 93.3 (1.0) 91.7 (0.4) 90.6 (0.1) 90.1 (0.2) 89.3 (0.2) 88.4 (0.3)
Serum folate, nmol/L 18.4 (2.7) 15.2 (0.9) 15.2 (0.5) 15.2 (0.5) 15.6 (0.7) 17.4 (0.9)
Iron storage indices
Serum albumin, g/L 39.7 (1.4) 42.7 (0.5) 42.4 (0.2) 42.1 (0.2) 41.3 (0.2) 40.5 (0.2)
Ferritin, μg/L 192.4 (38.7) 129.7 (11.1) 108.9 (3.3) 119.8 (2.5) 134.5 (3.4) 150.7 (4.5)
Transferrin saturation, % 36.3 (3.4) 28.9 (0.8) 27.8 (0.5) 26.4 (0.3) 25.0 (0.3) 24.1 (0.6)
Liver function indices
Hepatitis C seropositivity, % 11.1 2.3 1.4 2.4 2.3 3.1
AST, U/L 33.9 (5.9) 20.8 (0.8) 20.7 (0.3) 21.3 (0.3) 21.6 (0.2) 22.2 (0.6)
ALT, U/L 20.1 (3.3) 16.6 (1.0) 16.9 (0.4) 17.7 (0.5) 17.8 (0.5) 17.8 (0.8)

Values are means (SE) unless otherwise noted.

GFR indicates glomerular filtration rate; CRP, C-reactive protein; albuminuria, urinary albumin-to-creatinine ratio ≥0 mg/g; and elevated CRP, CRP ≥3 mg/L.

To convert SI units to conventional units, divide by the following conversion factors: cholesterol (0.02586); triglycerides (0.01129); fasting plasma glucose (0.05551); serum folate (2.266).
*
Limited to the fasting subsample of NHANES III participants. The sample size from lowest to highest HbA1c levels was n=45, 196, 1213, 2582, 1819, and 487.

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
Page 12
 Author Manuscript Author Manuscript Author Manuscript Author Manuscript

Table 2

Age-Standardized Mortality Rates and HRs (95% CI) for the Association Between HbA1c and All-Cause Mortality Among NHANES III Participants
Without Diabetes
Carson et al.

HbA1c, %

<4.0 (n=97) 4.0 – 4.4 (n=469) 4.5– 4.9 (n=2675) 5.0 –5.4 (n=5691) 5.5–5.9 (n=4035) 6.0 – 6.4 (n=1132)

Mortality rate* 39.3 13.5 9.9 8.8 11.3 12.5

Model 1 3.73 (1.45–9.63) 1.31 (0.68–2.53) 1.08 (0.80–1.46) 1 (ref) 1.23 (1.06–1.42) 1.28 (1.06–1.54)
Model 2a 3.97 (1.54–10.2) 1.37 (0.74–2.54) 1.13 (0.84–1.52) 1 (ref) 1.19 (1.02–1.38) 1.25 (1.04–1.51)
Model 2b 3.97 (1.54–10.2) 1.27 (0.66–2.44) 1.11 (0.81–1.50) 1 (ref) 1.22 (1.05–1.42) 1.25 (1.03–1.51)
Model 2c 3.47 (1.36–8.82) 1.36 (0.71–2.60) 1.16 (0.87–1.55) 1 (ref) 1.21 (1.04–1.42) 1.23 (1.02–1.47)
Model 2d 2.68 (1.08–6.66) 1.28 (0.69–2.35) 1.01 (0.76–1.34) 1 (ref) 1.27 (1.09–1.47) 1.41 (1.16–1.70)
Model 2e 3.87 (1.53–9.80) 1.30 (0.67–2.50) 1.12 (0.83–1.52) 1 (ref) 1.18 (1.00–1.38) 1.26 (1.05–1.51)
Model 2f 3.09 (1.20–7.92) 1.29 (0.67–2.48) 1.11 (0.82–1.50) 1 (ref) 1.22 (1.06–1.42) 1.31 (1.10–1.58)
Model 3 2.90 (1.25–6.76) 1.26 (0.69–2.28) 1.16 (0.86–1.55) 1 (ref) 1.26 (1.05–1.50) 1.30 (1.06–1.60)

*
Mortality rate standardized to age distribution of participants without diabetes, per 1000 person-years.

Model 1: Adjusted for age, race-ethnicity, and sex.

Model 2a: Adjusted for Model 1 variables plus education, income, current smoking, alcohol consumption, physical activity, body mass index, and aspirin use.

Model 2b: Adjusted for Model 1 variables plus cardiovascular factors (systolic blood pressure, antihypertensive medication use, total cholesterol, HDL cholesterol, log triglycerides, C-reactive protein, and
history of CVD).

Model 2c: Adjusted for Model 1 variables plus metabolic factors (prior diagnosis of thyroid disease, thyroid stimulating hormone, estimated glomerular filtration rate, and albuminuria).

Model 2d: Adjusted for Model 1 variables plus red blood cell indices (hemoglobin, red blood cell distribution width, mean cell volume, and serum folate).

Model 2e: Adjusted for Model 1 variables plus iron storage indices (serum albumin, ferritin, and transferrin saturation).

Model 2f: Adjusted for Model 1 variables plus liver function indices (hepatitis C seropositivity, AST, and ALT).

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
Model 3: Adjusted for age, race-ethnicity, sex, education, income, current smoking, alcohol consumption, physical activity, body mass index, aspirin use, cardiovascular factors (systolic blood pressure,
antihypertensive medication use, total cholesterol, HDL cholesterol, log triglycerides, C-reactive protein, and history of CVD), metabolic factors (prior diagnosis of thyroid disease, thyroid-stimulating
hormone, estimated glomerular filtration rate, and albuminuria), red blood cell indices (hemoglobin, red blood cell distribution width, mean cell volume, and serum folate), iron storage indices (serum
albumin, ferritin, and transferrin saturation), and liver function indices (hepatitis C seropositivity, AST, and ALT).
Page 13
 Author Manuscript Author Manuscript Author Manuscript Author Manuscript

Table 3

Sensitivity Analyses for the Association Between HbA1c and All-Cause Mortality Among NHANES III Participants Without Diabetes

HbA1c, %
Carson et al.

<4.0 (n=97) 4.0 – 4.4 (n=469) 4.5– 4.9 (n=2675) 5.0 –5.4 (n=5691) 5.5–5.9 (n=4035) 6.0 – 6.4 (n=1132)
Model* HR (95% CI)
Model 1 2.88 (1.17–7.09) 1.27 (0.70 –2.33) 1.21 (0.90 –1.64) 1 (ref) 1.33 (1.11–1.59) 1.40 (1.13–1.74)
Model 2 2.87 (1.25–6.62) 1.08 (0.56–2.05) 1.17 (0.85–1.60) 1 (ref) 1.31 (1.08–1.60) 1.34 (1.04–1.73)
Model 3 3.56 (1.38–9.17) 1.15 (0.64–2.05) 1.24 (0.88–1.75) 1 (ref) 1.34 (1.09–1.64) 1.29 (1.03–1.63)
Model 4 2.26 (0.96–5.42) 0.92 (0.50–1.62) 1.06 (0.78–1.42) 1 (ref) 1.23 (1.04–1.44) 1.27 (1.04–1.54)
Model 5 2.81 (1.03–7.67) 0.95 (0.54–1.94) 1.16 (0.79–1.87) 1 (ref) 1.40 (1.10–1.76) 1.41 (1.05–1.89)

Model 1 excludes n=530 with cancer.

Model 2 excludes n=814 with CVD.

Model 3 excludes n=1741 with anemia.

Model 4 excludes n=349 with hepatitis C.

Model 5 excludes n=3086 with cancer, CVD, anemia, or hepatitis C.

*
Adjusted for age, race-ethnicity, sex, education, income, current smoking, alcohol consumption, physical activity, body mass index, aspirin use, cardiovascular factors (systolic blood pressure,
antihypertensive medication use, total cholesterol, HDL cholesterol, log triglycerides, C-reactive protein, and history of CVD [except models 2 and 5]), metabolic abnormalities (estimated glomerular
filtration rate, albuminuria, prior diagnosis of thyroid disease, and thyroid-stimulating hormone), red blood cell indices (hemoglobin, red blood cell distribution width, mean cell volume, and serum folate),
iron storage indices (serum albumin, ferritin, and transferrin saturation), and liver function indices (hepatitis C seropositivity [except Models 4 and 5], AST, and ALT).

Circ Cardiovasc Qual Outcomes. Author manuscript; available in PMC 2016 February 01.
Page 14