Association of Fetuin-A with Incident Diabetes in Community-Living Older Adults: The Cardiovascular Health Study

Abstract

Background

The liver secreted protein fetuin-A induces peripheral insulin resistance in vitro. In a pilot study, we observed that higher fetuin-A levels were associated with diabetes in older persons. However, this finding has not been confirmed in large cohorts. We sought to confirm the association of fetuin-A with incident diabetes in older persons, and to determine if the association differs by age, sex, and race, and among persons with CVD.

Methods and Results

Among 3,710 community-living individuals aged ≥ 65 years without diabetes at baseline, serum fetuin-A was measured in serum collected in 1992–93. Participants were followed for 10.6 years (median) for incident diabetes. Cox regression models evaluated the association of fetuin-A with incident diabetes. Interaction terms evaluated heterogeneity by age, sex, race, and CVD. Mean age was 75years, 60% were female, 15% were black, and 16% had CVD. Mean fetuin-A concentrations were 0.47 ± 0.10 g/L. During follow-up, 305 incident diabetes cases occurred. Each 0.10g/L (SD) greater fetuin-A was associated with 19% higher risk of diabetes (HR 1.19; 95%CI 1.06–1.33) after adjusting for demographics, life-style factors, albumin, kidney function, and CVD. Further adjustment for potential mediators (body mass index, waist circumference, hypertension, lipids, and CRP) moderately attenuated the association (HR 1.13; 95% CI 1.00–1.28). Results were similar by sex, race, and CVD status, but were stronger in persons <75 years old (P-interaction 0.01).

Conclusions

Higher fetuin-A is associated with incident diabetes in older persons, irrespective of sex, race, or prevalent CVD status. The association may be attenuated in those aged ≥ 75 years.

Introduction

Secular trends in diet and lifestyle have resulted in an epidemic of diabetes.^1,2 Diabetes incidence and prevalence are both high in older persons;³ an ominous trend as our population ages, as this age group already bears the greatest burden of cardiovascular disease (CVD). Yet aging is associated with a redistribution of fat centrally and with sarcopenia,^4,5 so standard clinical risk factors such as body mass index (BMI) may become less reliable markers of diabetes risk in older age.⁶

Recent studies have identified a family of proteins secreted from adipose tissue, collectively termed adipokines, that regulate glucose metabolism. Their study has provided new insights to the biology of glucose regulation in younger and older populations alike. In contrast to adipokines which come from fat, fetuin-A (also known as α Heremans-Schmid glycoprotein [Ahsg]) is secreted from the liver.⁷ In vitro, fetuin-A reversibly binds the insulin receptor tyrosine kinase in peripheral tissues thereby inhibiting the insulin-induced intracellular signal cascade, producing peripheral insulin resistance.^8–10 Consistent with this function, fetuin-A knock-out mice are insulin sensitive,^11,12 whereas wild-type mice treated with exogenous fetuin-A acutely develop insulin resistance.¹³

To date, two studies have evaluated the association of serum fetuin-A levels with incident diabetes in human populations. Stefan and colleagues demonstrated that higher fetuin-A levels were associated with incident diabetes in 2,867 middle aged and predominantly Caucasian participants in Europe.¹⁴ In a pilot study, we made similar observations among 519 well functioning 70–79 year olds who participated in the Health Aging and Body Composition Study.¹⁵ Younger age, female gender, and Caucasian race were all strongly associated with higher fetuin-A levels in our prior study. Moreover, fetuin-A are known to inhibit arterial calcification,¹⁶ and lower levels have been associated with subclinical cardiovascular disease (CVD) in some,^17–20 but not all studies.^21,22 Whether or not the association of fetuin-A with diabetes differs by age, sex, race, or CVD status remains unknown. The modest sample size of our pilot study did not provide sufficient statistical power to evaluate heterogeneity across these subgroups.

In this study, we measured serum fetuin-A concentrations among a large sample of community-living individuals who participated in the Cardiovascular Health Study (CHS) and were followed for 10.6 years (median) for incident diabetes. We aimed to confirm the association of fetuin-A with incident diabetes in a larger sample of older individuals, and secondarily to evaluate for heterogeneity in this association by age, sex, race, and prevalent CVD status.

Methods

Participants

The CHS is a community-based study of older adults, designed to evaluate risk factors for development and progression of CVD. The study design and protocols have been described previously.^23,24 In brief, eligibility required age ≥ 65 years, expectation to remain in the area for 3 years after recruitment, no active cancer treatment, and the ability to provide consent. Between 1989 and 1990, 5201 participants were recruited from 4 communities using Medicare eligibility lists in each area (Sacramento, CA; Forsyth County, NC; Washington County, MD; and Allegheny County, PA). An additional 687 predominantly African-American participants were recruited in 1992–93. We considered the 1992–93 visit as the baseline visit for this analysis. Among 5,265 individuals who participated at this visit, we excluded individuals with missing or insufficient blood specimens for fetuin-A measurement (n=551), those with prevalent diabetes at baseline (n=708), those with missing information on diabetes status at baseline (n=81), individuals with no follow-up information regarding diabetes status after the baseline visit (n=124), and those missing covariate data (n=91), resulting in a final analytic sample of 3,710 participants for this analysis. All participants provided written informed consent, and the study was approved by the investigational review boards of the 4 clinical sites, the Data Coordinating Center at the University of Washington, and the lead investigator’s institution at the University of California San Diego.

Measurements

Fetuin-A

Serum was collected at the 1992–93 study visit and stored at −70° Celsius until 2010 when it was thawed and measured for fetuin-A using an enzyme linked immunosorbent assay (ELISA) kit (Epitope Diagnostics, San Diego, CA). The assay uses a 2-site “sandwich” technique with polyclonal antibodies that bind different epitopes of human fetuin-A. Serum samples were measured twice in each participant, and results were averaged. We observed coefficients of variation between 3.3 and 9.1%.

Incident Diabetes

Glucose was measured on fasting blood samples obtained during the annual clinic examinations in 1992–93, 1996–97, 1998–99, and 2005–06 and non-fasting blood samples in 1994–95. Medication use was assessed at baseline and annually thereafter by medication inventory through 2007.²⁵ We classified participants as having diabetes if fasting glucose was ≥ 126 mg/dL, casual glucose was ≥ 200 mg/dL, or individuals used insulin or oral hypoglycemic agents. Participants with diabetes at baseline were excluded from further analysis, and those who did not have diabetes by this definition but developed it during follow-up were classified as incident diabetes cases.

Other Measurements

Age, sex, race, smoking status, physical activity, and alcohol consumption were based on self-report. Cystatin C concentrations were measured using a BN II nephelometer (Seimans; www.siemans.com) as described previously.²⁶ The interassay and intraassay coefficients of variation were < 2.9% and < 3.2%, respectively. Estimated glomerular filtration rate (eGFR) was calculated using the equation eGFR=76.7 * cystatin C (mg/L) ⁻¹.¹⁹.²⁷ Participants with prior history of myocardial infarction, revascularization (percutaneous coronary intervention or coronary artery bypass graft surgery), stroke, or transient ischemic attach were classified as having prevalent CVD. Anthropometric measurements were made by trained personnel using standardized protocols. Participants wore light-weight examination suits and no shoes. Standing height was measured using a stadiometer calibrated in centimeters. Body weight was measured using a balance beam scale calibrated in kilograms. Body mass index (BMI) was calculated (kilograms body weight/height in meters squared). Waist circumference was measured at the level of the umbilicus. After 5 minutes at rest, seated blood pressure was determined in three times using standard mercury sphygmomanometers (Hawksley & Sons Ltd., Sussex, United Kingdom). The second and third measurements were averaged. Hypertension was defined as systolic blood pressure > 140 mmHg, diastolic blood pressure > 90 mmHg, or anti-hypertensive medication use and a physician diagnosis of hypertension. The Olympus Demand System (Olympus, Lake Success, New York) was used to measure serum triglyceride and high density lipoprotein (HDL) cholesterol concentrations. C-reactive protein (CRP) was determined by ultrasensitive ELISA as described elsewhere.²⁸ Insulin was measured with a competitive radioimmunoassay,²⁹ and was combined with fasting glucose measurements to calculate homeostasis model assessment of insulin resistance (HOMA-IR).³⁰

Statistical Analysis

We categorized participants into quartiles based on the distribution of fetuin-A concentrations in the study sample, and evaluated the distribution of demographics and diabetes risk factors across fetuin-A categories. Tests for trend across ordered categories were performed using linear regression, or a nonparametric test for trend,³¹ as appropriate. Differences in the distribution of categories of alcohol consumption and smoking status across quartiles of fetuin-A were tested using chi-square tests. Next, Kaplan Meier curves and the log rank test were used to evaluate the unadjusted association of fetuin-A categories with time to incident diabetes. To allow adjustment for potential confounders, we fit multivariable models using Cox regression, allowing the lowest fetuin-A quartile to serve as the reference category. As the functional form of the association of fetuin-A with incident diabetes was fairly linear across quartiles, we also evaluated fetuin-A as a continuous predictor variable (per standard deviation [SD] greater) to maximize statistical power. Sequential models were developed. An initial model adjusted for age, sex, race, and field center site. A subsequent model additionally adjusted for lifestyle factors and variables that we considered potential confounders of the association (physical activity, smoking [current, former, never], and alcohol use [0, 0–7, > 7 drinks per week], serum albumin, eGFR, and prevalent CVD). Fetuin-A has been associated with obesity, hyperlipidemia, and inflammation in prior studies,^13,32,33 but the directions of these associations remain uncertain. Thus, we constructed a final model that additionally adjusted for variables considered either potential mediators or confounders of the association of interest (BMI, waist circumference, hypertension, triglycerides, HDL, and CRP). We used Schoenfeld residuals to evaluate proportional hazards assumptions and found no appreciable evidence of violations. Last, we created multiplicative interaction terms for fetuin-A* age (≥75 years vs. younger), sex, race, and prevalent CVD status. Statistical significance of interaction terms were evaluated in the final adjusted model. Analysis was conducted using Stata, version 11.1 (StataCorp LP, College Station, Texas) and P-values < 0.05 were considered statistically significant for all analyses including tests of interactions.

Results

The mean age of the 3,710 person study sample was 74.8 ± 5.2 years, 60% (n=2,239) were female, and 14.5% (n=537) were African American. The mean fetuin-A concentration was 0.47 ± 0.10 g/L, and the distribution was approximately normal within the study sample. Table 1 shows the distribution of demographics and diabetes risk factors by fetuin-A quartiles. Compared to participants in the lowest quartile, those with higher fetuin-A concentrations were younger, were more frequently female and Caucasian, less likely to drink alcohol, had lower eGFR, and greater BMI, waist circumference, triglycerides, HDL cholesterol, CRP concentrations, serum albumin, and HOMA-IR.

Table 1.

Baseline Characteristics by Fetuin-A Quartiles in the Cardiovascular Health Study

	Fetuin-A Quartiles				P-value*
	1	2	3	4
N	939	919	931	921
Range (g/L)	≤0.41	>0.41–0.47	>0.47–0.53	>0.53
Age (yrs) ± SD	75.3 ± 5.7	75.0 ± 5.2	74.6 ± 5.0	74.3 ± 4.8	<0.001
Female n (%)	514 (54.7)	523 (56.9)	593 (63.7)	609 (66.1)	<0.001
African-American Race n (%)	220 (23.4)	144 (15.7)	114 (12.2)	59 (6.4)	<0.001
Physical Activity (kcal/week), median (IQR)	878 (293–1856)	870 (270–2085)	911 (323–1955)	960 (375–1995)	0.19
Smoking n (%)					0.36
Current	100 (10.7)	98 (10.7)	93 (10.0)	84 (9.1)
Former	423 (45.1)	410 (44.6)	381 (40.9)	406 (44.1)
Alcoholic drinks per week n (%)					<0.001
0	434 (46.2)	472 (51.4)	492 (52.9)	543 (59.0)
1–7	336 (35.8)	322 (35.0)	309 (33.2)	301 (32.7)
> 7	169 (18.0)	125 (13.6)	130 (14.0)	77 (8.4)
eGFR (ml/min/1.73m2) ± SD	74.7 ± 18.8	73.7 ± 17.7	73.4 ± 17.9	70.4 ± 18.2	<0.001
Prevalent Cardiovascular Disease n (%)	173 (18.4)	141 (15.3)	138 (14.8)	159 (17.3)	0.45
Body mass index (kg/m2) ± SD	25.8 ± 4.8	26.2 ± 4.4	26.8 ± 4.6	27.0 ± 4.3	<0.001
Waist circumference (cm) ± SD	94.5 ± 13.5	95.6 ± 12.2	97.5 ± 12.7	97.3 ± 12.5	<0.001
Hypertension n (%)	506 (53.9)	490 (53.3)	522 (56.1)	502 (54.5)	0.530
Triglycerides (mg/dL), Median (IQR)	102 (78–137)	114 (85–155)	128 (94–173)	146 (105–203)	<0.001
HDL cholesterol (mg/dL) ± SD	56.4 ± 15.2	54.1 ± 14.0	53.9 ± 14.4	53.5 ± 14.3	<0.001
Albumin (g/L) ± SD	3.92 ± 0.26	3.96 ± 0.26	3.98 ± 0.27	4.03 ± 0.27	<0.001
C reactive protein (mg/L), Median (IQR)	2.27 (1.08–5.45)	2.26 (1.09–5.46)	2.32 (1.09–5.41)	2.68 (1.29–5.69)	0.09
HOMA-IR Score, Median (IQR)	1.93 (1.42–2.69)	2.11 (1.51–3.07)	2.42 (1.69–3.35)	2.61 (1.86–3.85)	<0.001

^*p for trend, except smoking and drinking, for which a chi-square p-value is provided

Participants were followed for a median of 10.6 (interquartile range 5.9, 14.0) years, during which time 305 cases of incident diabetes were ascertained. In unadjusted analysis, we observed a graded increase in the rate of diabetes with higher fetuin-A quartiles. (Figure 1; log-rank P-value = 0.02). Table 2 shows the association of fetuin-A with incident diabetes in a series of adjusted models. When adjusted for age, sex, race, and field center site, compared to the lowest quartile, subjects in the highest quartile were at 69% higher risk of incident diabetes, and each SD greater fetuin-A level was associated with a 23% greater risk. This association was modestly attenuated with additional adjustment for lifestyle factors, eGFR, and prevalent CVD (model 2). Next, we adjusted for additional factors where it was less certain if they function as mediators or confounders in the pathway between fetuin-A and incident diabetes (BMI, waist circumference, triglycerides, HDL cholesterol, serum albumin, and CRP). In this model, each SD higher fetuin-A level remained associated with a 13% higher risk of diabetes (P=0.05).

Figure 1.

Kaplan Meier Curve Showing the Association of Fetuin-A Quartiles with Incident Diabetes in Community-living Older Persons: The Cardiovascular Health Study

Table 2.

Association of Fetuin-A with Incident Diabetes in the Cardiovascular Health Study

Range (g/L)	Fetuin-A Quartiles				Continuous
	I	II	III	IV	Per SD (0.10 g/L) greater
	≤0.41	>0.41–0.47	>0.47–0.53	>0.53		P-value
Events/ Person Years at Risk	56/8811	71/8913	93/8989	85/8670
Model 1*; HR (95% CI)	1.00 (Ref)	1.31 (0.92–1.87)	1.75 (1.25–2.45)	1.69 (1.20–2.40)	1.23 (1.09–1.38)	0.001
Model 2**; HR (95% CI)	1.00 (Ref)	1.27 (0.89–1.80)	1.72 (1.23–2.41)	1.56 (1.10–2.21)	1.19 (1.06–1.34)	0.003
Model 3†; HR (95% CI)	1.00 (Ref)	1.16 (0.81–1.65)	1.52 (1.08–2.13)	1.37 (0.95–1.96)	1.13 (1.00–1.28)	0.05

Abbreviations: HR=Hazard Ratio; CI=confidence interval; Ref=reference category

^*Model 1 is adjusted for age, sex, race, and field center site

^**Model 2 is adjusted for model 1 variables plus physical activity, smoking, alcohol use, eGFR, and prevalent CVD

^† Model 3 is adjusted for model 2 variables plus body mass index, waist circumference, hypertension, triglycerides, HDL, serum albumin, and CRP.

We tested for heterogeneity by age, sex, race, and prevalent CVD status. The association of fetuin-A with incident diabetes was similar in men and women, black and white individuals, and those with and without prevalent CVD (p-interactions all > 0.10). However, the association was stronger in individuals <75 years old compared to those who were older (P-interaction = 0.01). In participants < 75 years old, compared to the lowest fetuin-A quartile, those in the highest quartile were at 84% greater risk, and each SD greater fetuin-A was associated with a 23% greater risk of diabetes in the final adjusted model, whereas no association was observed in those ≥ 75 years of age (Table 3). Results were similar when models were further adjusted for self-reported health status (data not shown).

Table 3.

Age Stratified Association of Fetuin-A with Incident Diabetes in the Cardiovascular Health Study^*

Range (g/L)	Fetuin-A Quartiles				Continuous
	I	II	III	IV	Per SD (0.10 g/L) greater
	≤0.41	>0.41–0.47	>0.47–0.53	>0.53		P-value
Age < 75 years
Events/ Person Years at Risk	29/5170	44/5367	62/5587	65/5619
Adjusted** HR (95% CI)	1.00 (Ref)	1.36 (0.84–2.19)	1.74 (1.10–2.74)	1.84 (1.15–2.95)	1.23 (1.06–1.43)	0.006
Age ≥ 75 years
Events/ Person Years at Risk	27/3641	27/3546	31/3402	20/3052
Adjusted** HR (95% CI)	1.00 (Ref)	0.95 (0.55–1.64)	1.26 (0.74–2.15)	0.85 (0.46–1.57)	0.96 (0.78–1.19)	0.73

Abbreviations: HR=Hazard Ratio; CI=confidence interval; Ref=reference category

^*P-value for interaction = 0.01

^**Adjusted for age, sex, race, field center site, physical activity, smoking, alcohol use, eGFR, and prevalent CVD, body mass index, waist circumference, hypertension, triglycerides, HDL, serum albumin, and CRP.

In a secondary analysis, we investigated whether this finding might reflect different relationships of fetuin-A with insulin resistance among the two age groups, by analyzing the cross-sectional relationship of fetuin-A with HOMA-IR at the baseline study visit. In the fully adjusted model, fetuin-A was associated with higher HOMA-IR in both age groups, but the association was again stronger in persons < 75 years of age (β per SD higher fetuin-A 0.27 [95% CI 0.20–0.34] in persons < 75 years, and 0.16 [95% CI 0.07–0.24] in persons ≥ 75 years; p-interaction = 0.02).

Discussion

In a large biracial and geographically diverse cohort of community-living older persons and with a median follow-up time over 10 years, we demonstrate that higher serum fetuin-A levels are associated with an increased risk of diabetes. This study confirms findings in a middle aged Caucasian population, and our findings in a pilot study of older persons.^14,15 In secondary analysis we demonstrate that fetuin-A remains associated with incident diabetes in men, blacks, and individuals with prevalent CVD, although the association was primarily limited to individuals < 75 years of age.

Recent studies have elucidated mechanisms through which fat derived adipokines influence insulin resistance. Fetuin-A differs as it is the only factor to our knowledge that is produced and secreted by the liver that directly influences peripheral insulin resistance.⁷ Recent work has demonstrated that fetuin-A binds the beta subunit of the insulin receptor; thus it does not compete with insulin binding.³⁴ To date, only two proteins are known to bind the ectodomain of the insulin receptor – insulin and fetuin-A – the former stimulates signal transduction while the later inhibits it.³⁴ In addition to binding and inhibiting the insulin receptor, recent in-vitro and in vivo studies in mice suggest that fetuin-A also stimulates production of inflammatory cytokines and suppresses production of the adipokine adiponectin.¹³ Other studies demonstrate that free fatty acids promote generation of fetuin-A in hepatocytes; a process that is more efficient in the presences of inflammatory stress.³⁵ Collectively, these studies suggest that fetuin-A may provide a mechanism for the liver to regulate glucose metabolism in other organs. Thus, the findings reported here provide evidence that one or more of these pathways may be important to the development of incident diabetes not only in middle aged individuals, but also older persons, at least until age 75.

Identification of mechanisms leading to incident diabetes in older persons is important to public health. Rates of incident diabetes are high in older persons, and incident diabetes is strongly associated with all-cause and CVD mortality in this age group.³⁶ Current population trends show marked increases in the prevalence of obesity and simultaneously, our population is aging.¹ These factors suggest that rates of diabetes in older persons are likely to increase in future years, which might fuel higher rates of CVD, kidney disease, visual impairment, and other established consequences of diabetes. Yet, because body composition and glucose regulation change with age, biological mechanisms and risk factors for incident diabetes may differ in older individuals. Stefan and colleagues reported that fetuin-A is associated with incident diabetes in a middle aged European cohort (mean age 49 years).¹⁴ Whether or not this association differed by age was not reported. In our pilot study, fetuin-A was associated with incident diabetes in a separate sample of well functioning persons aged 70–79 years,¹⁵ but the limited number of participants and narrow age range in this study limited the ability to test for heterogeneity by age. Here, we observed that the association of fetuin-A with diabetes was observed only in those < 75 years. Moreover, the association of fetuin-A with HOMA-IR was also stronger in the < 75 year old age group.

The mechanisms responsible for these findings are uncertain. One possibility is that fetuin-A may become a less important regulator of peripheral insulin resistance in the oldest old. Another possibility is that participants who lived to ≥ 75 years and remained healthy enough to participate in this study differed systematically from the younger participants. Last, there were fewer cases of diabetes in the ≥ 75 year old age group (n=105 vs. 200 in the < 75 age group), and we tested multiple interactions. Although the p-value for interaction was statistically significant (p=0.01) and similar findings were confirmed evaluating HOMA-IR in cross-sectional analysis, it is possible that these were chance findings. Thus, we believe these findings require confirmation in other settings.

While multiple in vitro studies in different mammalian species,^7–9 studies in knock-out mice,^11,12 and studies in wild-type mice treated with exogenous fetuin-A¹³ all suggest that it may directly induce peripheral insulin resistance, there remains some controversy on this topic. Chen and colleagues transfected human fetuin-A into rat adipocytes, and tested whether this induced insulin resistance in the adipocytes in cell culture.³⁷ In contrast to fetuin-A’s effect on insulin resistance in other target tissues, these investigators failed to observe changes in insulin stimulated glucose transport. Thus, by the design of this experiment, transfected human fetuin-A would need to interact with the rat insulin receptor. To our knowledge, whether such interactions occur across species subtypes of fetuin-A and the insulin receptor is unknown, which may have contributed to the null findings. However, on the basis of these findings, others have speculated that high fetuin-A may simply reflect an overfed state or mark good hepatic synthetic function, similar to serum albumin.³⁸ Our study is the first to our knowledge to adjust for serum albumin when evaluating the association of fetuin-A with incident diabetes. Serum fetuin-A and albumin concentrations were only weekly correlated (r = 0.15), and adjusting for albumin had little influence on the association of fetuin-A with diabetes in our final models. While our epidemiologic study cannot prove causality, these data suggest that higher fetuin-A levels is not merely marking hepatic synthetic function, or serving as a surrogate of serum albumin. This distinction is important because if fetuin-A’s binding to the insulin receptor induces peripheral insulin resistance, then blockade of this interaction may ultimately serve as a novel target for therapy for insulin resistant states.¹⁵

Fetuin-A measurement may also ultimately have use to guide therapy with established medications for diabetes. For example, recent studies have demonstrated that the peroxisome proliferator-activated receptor (PPAR) gamma agonist pioglitazone lowers fetuin-A levels.³⁹ PPAR gamma agonists induce fluid retention, and have been associated with increased risk of heart failure and CVD events;^40,41 outcomes that are particularly common in older persons. Whether or not fetuin-A levels may inform the risk/benefit ratio of this class of medications in older persons is an important topic for future study. Short-term exercise interventions have also shown promise in lowering fetuin-A levels in young and middle aged persons in some,^42,43 but not all studies.⁴⁴ Future studies should investigate whether such interventions might lower fetuin-A in older persons, and if lowering fetuin-A translates into improvements in hard clinical end-points.

Strengths of this study include its large sample size, biracial and geographic diversity of the study participants, and long-term follow-up for incident diabetes. The study also has important limitations. Our study included individuals ≥ 65 years old, and within this age group, we observed heterogeneity in the association of fetuin-A with incident diabetes. Future studies are required to confirm this finding and evaluate heterogeneity across a wider age spectrum. Our study sample included only whites and blacks. Results may differ in other race/ethnicities. Fetuin-A was measured at a single time-point which may not reflect long-term exposure, and precludes evaluation of whether trajectories of fetuin-A may provide information on diabetes risk above and beyond a one-time measurement.

In conclusion, higher serum fetuin-A levels are associated with incident diabetes in community-living older persons. Fetuin-A remains associated with diabetes risk among men, blacks, and individuals with prevalent CVD, however the association of fetuin-A with diabetes may wane after age 75. Future studies are required to confirm heterogeneity by age, and to determine whether fetuin-A measurement may ultimately allow guided therapy to treat or prevent diabetes and associated complications.