Association between Inflammation and Biological Variation in Hemoglobin A1c in U.S. Nondiabetic Adults

Abstract

Context:

Inflammation is associated with higher glycated hemoglobin (HbA1c) levels. Whether the relationship is independent of blood glucose concentration remains unclear.

Objective:

The hemoglobin glycation index (HGI) was used to test the hypothesis that interindividual variation in HbA1c is associated with inflammation.

Participants:

This study used nondiabetic adults from the National Health and Nutrition Examination Survey (1999–2008).

Main Outcome Measures:

A subsample of participants was used to estimate the linear regression relationship between HbA1c and fasting plasma glucose (FPG). Predicted HbA1c were calculated for 7323 nondiabetic participants by inserting FPG into the equation, HbA1c = 0.017× FPG (mg/dL) + 3.7. HGI was calculated as the difference between the observed and predicted HbA1c and the population was divided into low, moderate, and high HGI subgroups. Polymorphonuclear leukocytes (PMNL), monocytes, and C-reactive protein (CRP) were used as biomarkers of inflammation.

Results:

Mean HbA1c, CRP, monocyte, and PMNL levels, but not FPG, progressively increased in the low, moderate, and high HGI subgroups. There were disproportionately more Blacks than whites in the high HGI subgroup. CRP (ß, 0.009; 95% confidence interval [CI], 0.0001–0.017), PMNL (ß, 0.036; 95% CI, 0.010–0.062), and monocyte count (ß, 0.072; 95% CI, 0.041–0.104) were each independent predictors of HGI after adjustment for age, sex, race, triglycerides, hemoglobin level, mean corpuscular volume, red cell distribution width, and obesity status.

Conclusions:

HGI reflects the effects of inflammation on HbA1c in a nondiabetic population of U.S. adults and may be a marker of risk associated with inflammation independent of FPG, race, and obesity.

Glycated hemoglobin (HbA1c) is measured clinically to estimate mean blood glucose (MBG) concentration during the previous 2 to 3 months. Numerous studies have shown, however, that variation in HbA1c in human populations cannot be fully explained by interindividual variation in MBG (1–6). In vivo and in vitro studies suggest that HbA1c levels are influenced by erythrocyte turnover, pH, and other factors besides blood glucose concentration (1, 2, 5, 7–11). Moreover, clinical studies have convincingly demonstrated that some individuals and racial groups have persistently lower or higher than expected HbA1c levels compared with others with similar blood glucose levels (2–6, 11–13). Interindividual variation in HbA1c due to factors other than blood glucose concentration complicates the use of HbA1c for the diagnosis and management of diabetes (4–6, 11).

The underlying biochemical mechanisms responsible for the observed disparity between HbA1c and blood glucose remain underexplored. However, several studies suggest a role for inflammation and oxidative stress. For example, higher total and differential white blood cell (WBC) counts have been associated with higher HbA1c levels in both diabetic and nondiabetic populations (14–19). Analysis of data from NHANES III suggested that C-reactive protein (CRP), a principal downstream mediator of the acute phase inflammatory response, was significantly positively associated with HbA1c in a population of nondiabetic U.S. adults (20). Gustavsson et al (21) reported that CRP, WBC, and fibrinogen were all positively correlated with HbA1c in nondiabetic patients with coronary artery disease.

There is strong evidence that cardiovascular disease (CVD) is a multifactorial disease with correlated risk factors such as obesity, dyslipidemia, diabetes, hypertension, and chronic low grade systemic inflammation, collectively known as a risk factor cluster. Evidence of a relationship between HbA1c and inflammation was previously reported in studies that used CRP and WBC count to assess inflammation (15, 17, 20, 21). Yasunari et al (16) reported that measures of oxidative stress in polymorphonuclear leukocyte (PMNL) and mononuclear cells were associated with higher HbA1c regardless of diabetes status. Recently, Jiang et al (17) demonstrated that WBC count was positively associated with HbA1c, but not FPG, in a large representative sample of the general Chinese population independent of body mass index (BMI). Farah et al (22) also reported that biomarkers of inflammation (PMNL and WBC counts) were associated with HbA1c but not FPG in type 2 diabetes patients. Collectively, these reports suggest that variation in inflammation influences variation in HbA1c more than variation in FPG.

The hemoglobin glycation index was developed to quantify interindividual variation in HbA1c due to factors other than blood glucose concentration (1, 6, 11, 13, 23, 24). HGI was strongly predictive of microvascular disease in type 1 diabetes patients in the Diabetes Control and Complications Trial (2, 24). It was also reported to be higher in Black than white children with type 1 diabetes (1, 11). Racial disparity in HGI was also observed in adults with type 2 diabetes participating in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial (25). Analysis of HGI in ACCORD suggested that HGI is a CVD risk factor which in turn suggests that HGI may be associated with other CVD risk factors such as biomarkers of inflammation. The present study thus used HGI and data from the 1999–2008 cycles of the National Health and Nutrition Examination Survey (NHANES) to test the hypothesis that interindividual variation in HbA1c is associated with inflammation in a nondiabetic population of U.S. adults independent of fasting blood glucose concentration.

Research design and method

NHANES is an ongoing national survey directed by the Centers for Disease Control that uses a stratified multistage probability sampling design to represent the noninstitutionalized U.S. civilian population (26). Details of the NHANES sampling strategy and data collection procedures are described elsewhere in detail (26–28). Briefly, the survey consists of an in-home interview and a subsequent visit to a mobile examination center (MEC) where physical examinations, medical testing, and interviews were performed with selected participants. The five cycles included in the present analysis (1999–2000, 2001–2001, 2003–2004, 2005–2006, and 2007–2008) included 9965, 11 039, 10 122, 10 348, and 10 149 participants, respectively. There were 2772, 3162, 2915, 2859, and 2792 participants remaining in each corresponding cycle after omitting subjects who did not have blood collected, were not fasting, or whose fasting plasma glucose (FPG) or HbA1c was missing. Of these 14 500 subjects, 1365 were classified as having diabetes based on a self-reported history of diabetes or of taking antidiabetes medications (including insulin), or if they had an FPG of at least 126 mg/dL. After excluding subjects with diabetes, anemia (hemoglobin < 13 g/dL for men and < 12 g/dL for women; n = 658) or whose age was less than 20 years (n = 4167), a total of 8310 NHANES adults without diabetes were included in the present analysis.

Data collection

The National Centers for Health Statistics Ethics Review Board approved the NHANES study protocol and each participant provided written informed consent. Demographic information and medical histories were collected using standardized questionnaires. Physical and laboratory examinations were performed at MEC visits. A blood specimen was drawn from each participant's antecubital vein by trained phlebotomists according to a standard protocol (29). The blood was processed, stored, and shipped to various laboratories for analysis. Complete blood counts with differential were performed during MEC visits. PMNL count was calculated as total WBC count minus the sum of the lymphocyte count and the monocyte count. Serum CRP was measured using latex-enhanced nephelometry. For some analyses we used conventional clinical cut points to subdivide participants into low (< 0.8 k/μl) and high (≥ 0.8 k/μl) monocyte (30) subgroups, or low (< 0.3 mg/dL), moderate (0.3 to < 1.0 mg/dL), and high (≥ 1.0 mg/dL) CRP (31) subgroups. Obesity status was based on body mass index (BMI) defined as underweight (BMI < 18.5 kg/m²), normal weight (18.5 ≤ BMI < 25 kg/m²), overweight (25 ≤ BMI < 30 kg/m²), and obese (BMI ≥ 30 kg/m²). Fasting plasma glucose was measured enzymatically using a hexokinase method. All HbA1c assays used during the five NHANES cycles were certified by the National Glycohemoglobin Standardization Program.

Calculating the hemoglobin glycation index

A 10% subsample of 831 subjects was randomly extracted from the 8310 participants who met the inclusion criteria. Data from the subsample were used to estimate the linear relationship between FPG and HbA1c in the study population: HbA1c = 0.017 × FPG (mg/dL) + 3.7. A predicted HbA1c was then calculated for the remaining 7479 participants by inserting FPG into the subsample linear regression equation. HGI was calculated by subtracting the predicted HbA1c from the observed HbA1c. Each participant was then assigned to low, moderate, or high HGI subgroups based on weighted HGI tertile (33.3%) cut points (low HGI < −0.133; n = 2178; moderate HGI, −0.133–0.130; n = 2381; high HGI > 0.130; n = 2920). Further analyses were performed on 7323 participants after excluding individuals with incomplete laboratory data (n = 154) or acute infection (n = 2, defined as WBC > 11 (10⁹/L) or CRP > 10 mg/dL).

Statistical analysis

Data are presented as means (± SE), medians (interquartile range [IQR] from the 25th to the 75th weighted percentile), or counts (weighted percentages). Descriptive statistics were compared between low, moderate, and high HGI subgroups. Group comparisons used ANOVA or Kruskal-Wallis tests for normally distributed and nonnormally distributed continuous variables, respectively. Rao-Scott F adjusted χ² tests were used to analyze categorical variables. Because the distributions of inflammatory biomarkers were skewed, especially CRP, Kendall's tau rank correlation coefficients were used to assess relationships between metrics of metabolic control (ie, HbA1c, FPG, and HGI) and biomarkers of inflammation using jackknife variances to account for the complex sampling design of continuous NHANES data. Crude means and adjusted least squares means for each metabolic parameter were compared across CRP or monocyte subgroups using multiple linear regression and Tukey's post-hoc tests.

Relationships between HGI as a continuous variable and biomarkers of inflammation were examined by adjusting for potential confounders that were associated with HbA1c level including age, sex, race, hemoglobin level, mean corpuscular volume (MCV), red cell distribution width (RDW), and triglyceride (basic adjustment, Model 1). Inspection of descriptive statistics and scatter plots showed that the distributions of CRP and triglyceride were highly skewed in which case the natural log transformation of each variable was used in the model. Additional adjustment for obesity status or waist circumference was also performed (full adjustment, Model 2). Regression coefficients (ß) and 95% confidence intervals (CI) were estimated for each model. Standardized regression coefficients were also computed by dividing the estimated ß by the ratio of the sample SD of the dependent variable to the sample SD of the regressor. Sensitivity analysis was performed using waist circumference instead of BMI to estimate obesity status. The study population was also evaluated with and without pregnant women (n = 407) included in the analyses. Tests of statistical significance were based on a 2-tailed type 1 error at P < .05. All analyses were performed using SAS version 9.4 (SAS Institute, Inc., Cary, NC), STATA version 13.0 (StataCorp LLP, College Station, TX), and R version 3.1.1 with techniques appropriate for complex sampling design survey data.

Results

Characteristics of the NHANES participants included in this study are presented in Table 1. Comparison of results between HGI subgroups shows that unlike FPG, HbA1c progressively increased in the low, moderate, and high HGI subgroups as did all three inflammatory biomarkers (CRP, PMNL, and monocytes). Red cell distribution width also progressively increased in the low, moderate, and high HGI subgroups whereas hemoglobin level and MCV progressively decreased. High HGI subjects were older and had higher waist circumference than low and moderate HGI subjects. Evaluation of the population using race and obesity as categorical variables (Table 1) showed that there was a disproportionate number of Black participants and obese subjects in the high HGI subgroup.

Table 1.

Clinical and Demographic Characteristics of Adult Nondiabetic NHANES Participants by HGI Subgroup

Variable (n = 7323) WTP = 154 195 876	Total	HGI subgroup
		Low	Moderate	High	Pa
Hemoglobin Glycation Index, %	−0.011 ± 0.009	−0.359 ± 0.007	−0.008 ± 0.002	0.338 ± 0.005	<.0001
HbA1c, %	5.3 ± 0.01	5.0 ± 0.01	5.3 ± 0.01	5.6 ± 0.01	<.0001
Fasting plasma glucose, mg/dL	96.5 ± 0.2	97.1 ± 0.4	95.9 ± 0.3	96.5 ± 0.3	.006
Age, y	44.6 ± 0.3	39.3 ± 0.5	43.8 ± 0.5	50.6 ± 0.4	<.0001
Sex, No. (%)					.09
Female	3687	992 (30.0)	1239 (35.3)	1456 (34.6)
Male	3636	1152 (37.2)	1092 (31.3)	1392 (31.5)
Race or ethnic group, No. (%)					<.0001
White	3883	1267 (35.6)	1323 (34.4)	1293 (30.0)
Black	1177	214 (19.0)	270 (24.6)	693 (56.4)
Other	2263	663 (32.5)	738 (33.2)	862 (34.2)
Obesity status, %					<.0001
Underweight (BMI < 18.5)	126	47 (41.1)	43 (38.4)	36 (20.5)
Normal weight (18.5 ≤ BMI < 25)	2329	810 (39.6)	764 (32.8)	755 (27.6)
Overweight (25 ≤ BMI < 30)	2664	778 (33.1)	865 (34.1)	1021 (32.8)
Obese (BMI ≥ 30)	2204	509 (26.3)	659 (32.7)	1036 (41.0)
Waist Circumference, cm	95.4 ± 0.2	93.1 ± 0.4	95.1 ± 0.3	98.1 ± 0.4	<.0001
Median C-reactive protein, mg/dL	0.19	0.14	0.18	0.23	<.0001
(IQR)	(0.08–0.42)	(0.06–0.34)	(0.08–0.41)	(0.10–0.51)
PMNL, k/μl	4.26 ± 0.03	4.14 ± 0.05	4.29 ± 0.04	4.34 ± 0.04	.003
Monocytes, k/μl	0.54 ± 0.003	0.52 ± 0.01	0.53 ± 0.004	0.55 ± 0.004	<.0001
Hemoglobin, g/dL	14.7 ± 0.04	14.9 ± 0.05	14.7 ± 0.04	14.6 ± 0.05	<.0001
Mean corpuscular volume, fl	90.4 ± 0.1	91.1 ± 0.1	90.4 ± 0.1	89.7 ± 0.2	<.0001
Red cell distribution width, %	12.5 ± 0.02	12.3 ± 0.02	12.5 ± 0.03	12.8 ± 0.02	<.0001
Median Triglycerides, mg/dL	103	96	106	106	<.0001
(IQR)	(71–152)	(67–145)	(72–153)	(76–155)

Abbreviation: WTP, weighted to the total population.

Unless otherwise noted values are means ± se for continuous variables or number (weighted %) for categorical variables.

^aOverall differences between HGI groups using either ANOVA, Kruskal-Wallis tests or Rao-Scott F adjusted χ² tests.

Univariate Kendall's tau rank correlation was used to evaluate simple relationships between metrics of metabolic control (HbA1c, FPG, and HGI) and biomarkers of inflammation expressed as continuous variables (CRP, PMNL, and monocytes). The results (Table 2) showed that all three metrics of metabolic control were positively correlated with all three biomarkers of inflammation in the study population. Moreover, correlation coefficients were higher for HbA1c than either FPG or HGI. Multiple linear regression analyses (Table 3) showed that CRP, PMNL, and monocyte count were each independent predictors of HGI after basic adjustment (Model 1). The observed relationships between HGI and inflammatory biomarkers persisted after further adjustment for obesity (Model 2). Monocyte count had the strongest relationship with HGI, with one SD increase in monocyte count associated with a 0.063 increase in HGI.

Table 2.

Correlation Coefficient between Inflammation Biomarkers and Glycemic Profiles

	HbA1c	FPG	HGI
	C-reactive protein, mg/dL	0.137 (0.128–0.147)	0.076 (0.066–0.085)	0.104 (0.097–0.112)
PMNL, k/μL	0.062 (0.055–0.070)	0.046 (0.038–0.055)	0.043 (0.037–0.450)
Monocytes, k/μL	0.077 (0.067–0.086)	0.069 (0.062–0.076)	0.048 (0.040–0.056)

Values are Kendall's tau rank correlation coefficients and 95% CI. All P < .0001.

Table 3.

Adjusted Regression Coefficients on HGI

Predictors	Model 1b			Model 2c
	ß	Standardized ß	P	ß	Standardized ß	P
Age	0.006 (0.005–0.007)	0.288	<.0001	0.006 (0.005–0.007)	0.291	<.0001
Female vs male	0.023 (−0.001–0.047)	0.035	.059	0.021 (−0.004–0.046)	0.031	.099
Black vs white	0.180 (0.154–0.207)	0.154	<.0001	0.177 (0.151–0.023)	0.151	<.0001
Other race vs white	0.065 (0.035–0.095)	0.074	<.0001	0.066 (0.036–0.096)	0.075	<.0001
Triglyceridesa, mg/dL	0.010 (−0.006–0.027)	0.018	.205	0.007 (−0.009–0.024)	0.013	.362
Hemoglobin, g/dL	−0.012 (−0.023–−0.002)	−0.047	.024	−0.012 (−0.023–−0.002)	−0.049	.020
Mean corpuscular volume, fl	−0.009 (−0.012–−0.007)	−0.130	<.0001	−0.009 (−0.012–−0.007)	−0.126	<.0001
Red cell distribution width, %	0.039 (0.026–0.053)	0.099	<.0001	0.038 (0.025–0.052)	0.097	<.0001
CRPa, mg/dL	0.013 (0.004–0.021)	0.046	.004	0.010 (0.001–0.018)	0.035	.036
PMNL, k/μl	0.013 (0.006–0.019)	0.050	.003	0.013 (0.006–0.019)	0.050	.003
Monocytes, k/μl	0.126 (0.065–0.187)	0.063	<.0001	0.126 (0.065–0.186)	0.063	<.0001
Underweight vs normal weight				0.016 (−0.039–0.071)	0.007	.566
Overweight vs normal weight				−0.006 (−0.029–0.017)	−0.009	.590
Obese vs normal weight				0.029 (0.004–0.054)	0.039	.002

Regression coefficient (95% CI).

^aNatural log transformed.

^bModel 1: Adjusting for survey years, age, sex, race, or ethnic group, hemoglobin level, mean corpuscular volume, red cell distribution width, and triglyceride.

^cModel 2: Model 1 plus obesity status (underweight, normal weight, overweight, obesity).

Figure 1 shows adjusted mean values for HbA1c, FPG, and HGI by CRP or monocyte count after controlling for survey years, age, sex, race, hemoglobin level, MCV, red cell distribution width, triglyceride level, and obesity. The results show that HbA1c and HGI but not FPG were significantly higher in the higher CRP and high monocyte subgroups. Sensitivity analyses showed that the estimated ß coefficients remained stable when pregnant women were excluded from the analyses. Waist circumference was not significantly associated with HGI when included in the model instead of obesity.

Figure 1.

Effect of inflammation on metrics of metabolic control after full adjustment including obesity. y-axis, from left to right: FPG (mg/dL), HbA1c (%), HGI (%). x-axis, upper panel: C-reactive protein (mg/dL); lower panel: monocyte count (k/μl). Groups: Upper panel, Group 1: blank bar (CRP < 0.3 mg/dL); Group 2: shaded bar (0.3 ≤ CRP < 1.0 mg/dL); Group 3: black bar (CRP ≥ 1.0 mg/dL). Lower panel: Group 1: blank bar (monocyte count < 0.8 k/μl); Group 2: black bar (monocyte count ≥ 0.8 k/μl). ^a,b,c, means ± SE with different characters are significantly different within CRP or monocyte count subgroups; P < .05.

Discussion

Sources of HbA1c variation in human populations can be broadly divided into two categories, 1) interindividual variation in blood glucose concentration, and 2) idiosyncratic differences in other factors that influence HbA1c levels. Because of how it is calculated, HGI reflects interindividual variation in HbA1c due to these other factors (1–4, 11–13, 32). That variation in HbA1c is attributable to the combined influences of variation in both blood glucose concentration and HGI is supported by the observation that correlation coefficients between biomarkers of inflammation and HbA1c were greater than those for either FPG or HGI (Table 2).

The present study used HGI to test the hypothesis that interindividual variation in HbA1c is associated with inflammation in a nondiabetic population of U.S. adults independent of the effect of blood glucose concentration. This hypothesis is supported by the observation that mean CRP, monocyte, and PMNL levels were progressively higher in the low, moderate, and high HGI subgroups (Table 1) while no trend was observed in FPG levels. Furthermore, HbA1c and HGI were both significantly higher in the high monocyte group and higher CRP groups (Figure 1) while FPG remained unchanged. Multiple linear regression (Table 3) showed that all three inflammation biomarkers were independent predictors of HGI after adjustment for covariates using either Model 1 (adjusted for survey years, age, sex, race, triglycerides, hemoglobin level, MCV, and red cell distribution width) or Model 2 (adjusted for Model 1 covariates plus obesity status).

When assessing the clinical relevance of these results it is important to note that although the apparent effects of inflammation on HbA1c and HGI were relatively small, it is possible that the same degree of inflammation could have a more profound influence on vascular diseases mediated by protein glycation. Consider, for example, that protein glycation is a cumulative process where chemically modified protein amino groups accumulate over the life of each protein. Red blood cells (and thus hemoglobin molecules) have an average biological life span of approximately 120 days. In contrast, collagen molecules can exist in the extracellular matrix surrounding blood vessels for many years. Thus although chronic inflammation in nondiabetic NHANES participants seems to have a relatively small effect on glycation of a short-lived protein such as hemoglobin, the same degree of inflammation over longer periods of time could have much greater effect on the glycation and function of long-lived proteins such as collagen that have a direct role in the pathophysiology of vascular disease.

Because obesity is a major source of inflammation (19, 33, 34) it might also be a confounder or mediator of the relationship between inflammation and HGI. The results of multiple regression analyses showed that obesity was associated with HGI, with a 0.029% higher mean HGI in obese subjects compared with normal weight subjects. The standardized ß regression coefficients describing the relationships between HGI and either monocyte count or PMNL count remained relatively constant after further adjusting for obesity status. In contrast, the ß for CRP decreased, which is consistent with the results of clinical studies that have reported attenuation of the association between CRP and HbA1c after adjusting for BMI (14, 15, 20, 21, 35, 36). This observation suggests that the relationship between HGI and CRP might be partly mediated by obesity.

In the present study, RDW was highest in the high HGI group (Table 1) which suggests that RDW is associated with HbA1c independent of variation in FPG. High RDW is indicative of greater variation in red blood cell size and has been associated with iron deficiency anemia (37). Although several studies have clearly shown that iron deficiency anemia is associated with higher HbA1c levels (38) this cannot explain high RDW in high HGI subjects because NHANES participants with anemia were excluded from the analyses. Engström et al (39) reported that RDW was positively correlated with HbA1c in diabetic subjects. Our findings agree more closely with those of Veranna et al (40) who reported that RDW was positively associated with HbA1c independent of glucose in nondiabetic subjects. Several studies also link high RDW with higher CRP and increased CVD risk (41–43), which could mean that HGI, RDW, and inflammatory biomarkers are part of the same CVD risk factor cluster.

As previously reported in subjects with diabetes (1, 11, 32), nondiabetic Black subjects were disproportionately represented in the NHANES high HGI subgroup (Table 1). Furthermore, nondiabetic Black participants had a 0.177% higher mean HGI than nondiabetic whites with other covariates held constant after full adjustment (Table 3). This suggests that the Black-white difference in HGI is independent of variation in inflammation, obesity, and other covariates associated with HbA1c. These results support previous observations in patients with diabetes showing that Blacks have higher HGI than whites and extend this conclusion to individuals without diabetes. Because our study is based on 10 years of data from a large nationally representative sample, the results likely have better external validity compared with studies where sampling is limited to local populations or clinic settings.

This study has some limitations. For example, this is the first study to use HGI to assess biological variation in HbA1c in nondiabetic subjects and only the second study to use FPG to calculate HGI. Most previous studies of HGI in human populations used mean blood glucose to calculate HGI. Although HGI calculated using all glucose data downloaded from patient meters were highly correlated with HGI calculated using only prebreakfast glucose data (11), potential issues related to the use of FPG rather than MBG for calculating should be further explored in longitudinal studies. Also, due to the cross-sectional design of the study, causality of the observed associations remains speculative. Another potential limitation is the fact that NHANES demographic data are self reported, including whether participants have a history of diabetes. To minimize any confounding due diabetes misclassification, the definition of diabetes was conservatively expanded to also exclude participants who reported using antidiabetic medications (including insulin) and anyone with an FPG at least 126 mg/dL.

It is conceivable that interassay variation between study cycles could contribute to population variation in HbA1c because NHANES used different methods for analyzing HbA1c over the five cycles covered by these analyses, including the Primus CLC330 (1999–2004), the Tosoh A1C 2.2 Plus Glycohemoglobin Analyzer (2005–2006), and the Tosoh A1C G7 Automated HPLC Analyzer (2007–2008). However, laboratory method cross-over studies were conducted at the time of each change to a different HbA1c method and a comprehensive evaluation of HbA1c results across the 1999–2010 cycles failed to identify analytical issues as a significant source of intercycle differences in HbA1c (37). Finally, abnormal hemoglobin variants can affect the accuracy of some HbA1c assays but hemoglobin phenotypes were not assessed as part of the NHANES protocol and thus were not accounted for in the present study.

In conclusion, using HGI to analyze NHANES data confirmed that biological variation in HbA1c is positively associated with inflammation independent of blood glucose concentration and obesity in nondiabetic subjects. These observations concur with experimental observations from population-based studies that suggest that interindividual variation in HbA1c is related to low-grade inflammation in both diabetic and nondiabetic human populations.