Serum Ferritin and Risk of the Metabolic Syndrome in U.S. Adults

The Third National Health and Nutrition Examination Survey (NHANES III) was conducted by the National Center for Health Statistics between 1988 and 1994. The survey was a multistage nationwide probability sample of the civilian, noninstitutionalized U.S. population. The study design included a home interview and a detailed clinical evaluation (13).

The present analysis was restricted to NHANES III participants ≥20 years of age, who were not pregnant or nursing at the time of the examination and who were randomly assigned to receive a morning examination (n = 7,907). We excluded individuals who had fasted <9 h at the time of the examination (n = 600), who had received treatment for anemia within the last 3 months (n = 186), with known current hepatitis C infection (n = 184), who had donated blood within the last 4 months (n = 195), or who had probable hemochromatosis based on abnormal values of multiple iron parameters (men: serum iron >190 μg/dl, serum ferritin >300 μg/l, and transferrin saturation >60%; women: serum iron >175 μg/dl, serum ferritin >200 μg/l, and transferrin saturation >60%) (n = 24) (14). Because some individuals met more than one exclusion criterion, the total number eligible for the study was 6,646. Individuals missing measurements for serum ferritin, fasting insulin, or any of the components of the metabolic syndrome were further excluded in the analysis (n = 602). Therefore, the final analytical sample consisted of 6,044 individuals.

Data from NHANES III were collected by household interview followed by a detailed physical examination conducted at a mobile examination center or at the participant’s home. Information on age, race/ethnicity, education, smoking history, alcohol intake, and use of antihypertensive and diabetes medications was collected during the interview. Menopausal status was defined as self-reported cessation of menstruation or hysterectomy. The average of all available blood pressure readings taken during the home interview and physical examination (up to six) was used in the analysis. Waist circumference was measured by a trained technician to the nearest 0.1 cm in a horizontal plane at the level of the high point of the iliac crest at minimal respiration. BMI was calculated as weight in kilograms divided by the square of height in meters. Alcohol intake was categorized as never, zero to two alcoholic drinks per day, and more than two alcoholic drinks per day based on self-reported average over the past 12 months. Smoking was categorized as current, former, and never.

NHANES laboratory procedures and quality control have been published previously (15). Serum ferritin was measured on fasting blood samples using a single-incubation two-site immunoradiometric assay (Bio-Rad Laboratories, Hercules, CA). The interassay coefficients of variation ranged from 4 to 6%. Plasma glucose was measured using a modified hexokinase enzymatic reference method (COBAS MIRA; Roche Diagnostics Laboratory Systems, Indianapolis, IN). Serum insulin was measured using the Pharmacia Insulin radioimmunoassay kit (Pharmacia Diagnostics, Uppsala, Sweden). Serum triglycerides and HDL cholesterol were measured enzymatically (Boehringer Mannheim Diagnostics, Indianapolis, IN). Alanine aminotranferase (ALT) and aspartate aminotransferase (AST) were assayed using the α-ketoglutarate reaction. Serum C-reactive protein (CRP) concentration was determined using latex-enhanced nephelometry with a limit of detection of 0.21 mg/dl.

The metabolic syndrome was defined as the presence of three or more of the following criteria set forth by the National Cholesterol Education Program Adult Treatment Panel III guidelines (16): 1) elevated blood pressure (average systolic blood pressure ≥130 mmHg or diastolic blood pressure ≥85 mmHg) or current blood pressure medication use, 2) low HDL cholesterol (<1.036 mmol/l in men and <1.295 mmol/l in women), 3) elevated serum triglycerides (≥1.695 mmol/l), 4) elevated fasting plasma glucose (≥6.105 mmol/l) or current antidiabetes medication use, and 5) abdominal obesity (waist circumference ≥102 cm in men and ≥88 cm in women). Diabetes was defined as a fasting plasma glucose value >7.0 mmol/l or current use of antidiabetes medication.

Insulin resistance was determined using the homeostasis model assessment (HOMA) estimate of insulin resistance (HOMA-IR = fasting insulin [μU/ml] × fasting glucose [mmol/l]/22.5) (17), fasting insulin level, and the ratio of triglyceride to HDL cholesterol.

All analyses were performed separately for men, premenopausal women, and postmenopausal women, including calculation of specific quartiles of serum ferritin. Serum ferritin values were right skewed and were log transformed. Odds ratios for the prevalence of the metabolic syndrome and its components were calculated for quartiles of serum ferritin using logistic regression. Tests for trend across quartiles were computed by including a variable with the median value for each quartile as a continuous variable in the logistic regression models. All models were adjusted for age, race/ethnicity, BMI, CRP, smoking, and alcohol intake. Statistical analyses were performed using STATA (version 8.1; College Station, TX) “svy” commands to account for complex survey design and included sampling weights to provide nationally representative prevalence estimates.

Mean levels of serum ferritin were significantly lower in premenopausal women 33.6 μg/l (95% CI 31.5–35.8) compared with postmenopausal women 93.4 μg/l (87.9–99.2) and lower in postmenopausal women compared with men 139.9 μg/l (133.6–146.6). The prevalence of the metabolic syndrome was 10.2% in premenopausal women, 27.8% in postmenopausal women, and 17.5% in men.

In all three groups, the highest prevalence of the metabolic syndrome occurred in the highest quartile of serum ferritin (Table 1). The prevalence of elevated blood pressure, elevated plasma glucose, elevated triglycerides, and abdominal adiposity all increased significantly with increasing serum ferritin in men and postmenopausal women. In premenopausal women, the prevalence of elevated triglycerides and abdominal adiposity also increased with increasing serum ferritin. For all three groups, the greater the number of metabolic syndrome components present, the greater the serum ferritin level (Fig. 1).

After adjusting for age, race/ethnicity, BMI, smoking status, alcohol consumption, and CRP, serum ferritin remained positively associated with prevalence of the metabolic syndrome in all three groups (Table 2). Further adjustment for triceps skinfold thickness, as an additional marker of adiposity, did not change the results. The metabolic syndrome was more common in the highest compared with the lowest quartile of serum ferritin in premenopausal women (OR 2.4 [95% CI 1.1–5.2]), in postmenopausal women (2.7 [1.7–4.1]), and in men (1.6 [0.9–2.7]) (Table 2). Ferritin was also positively associated with elevated triglycerides and elevated glucose levels in postmenopausal women and in men. These results remained unchanged if individuals taking antihypertensives or hypoglycemic medications are excluded from the analysis. The association between serum ferritin and abdominal adiposity observed in univariate analyses was attenuated when BMI was included in multivariate models.

Because serum ferritin is an acute- phase reactant (18), we repeated our analyses after excluding individuals with a white blood cell count above or below the reference range (>11.7 or <3.0 × 10⁹ cells/l) (n = 136), CRP >1.0 mg/dl (n = 474), or elevation of one or more of the following liver enzymes: ALT greater than two times the upper limit of normal (>80 units/l for men and >62 units/l for women) or AST greater than two times the upper limit of normal (>74 U/l for men or >62 units/l for women) (n = 32) (15). Similar results were observed after these individuals were excluded. For example, the adjusted odds ratios of metabolic syndrome comparing the highest with lowest quartile of serum ferritin for premenopausal women, postmenopausal women, and men were 1.5 (95% CI 0.9–2.6), 2.3 (0.8–7.0), and 2.0 (1.2–3.4), respectively.

Mean values of estimated insulin resistance were calculated by quartile of serum ferritin (Table 3). This analysis was restricted to individuals without diabetes. We did not observe any consistent association between HOMA-IR or fasting insulin with serum ferritin in premenopausal women. We did observe a trend of increasing insulin resistance (HOMA-IR, fasting insulin, and triglyceride-to-HDL ratio) with increasing serum ferritin in both postmenopausal women and men. In Table 4, we present the multivariate-adjusted ratio of insulin resistance measures for the second, third, and fourth quartiles of serum ferritin compared with the first. Similar to the unadjusted analyses, the ratio of insulin resistance values increased with increasing quartiles of serum ferritin in men and postmenopausal women. For example, levels of the triglyceride-to-HDL ratio were 21% higher in the highest quartile of serum ferritin compared with the lowest quartile of serum ferritin for men. In postmenopausal women, the observed association between insulin resistance and ferritin appeared to follow a j-shaped pattern in which values were slightly elevated in the first quartile, lower in the second quartile, and then increased in the third and fourth quartiles.

In this study, we observed a positive association between elevated iron stores, measured by serum ferritin levels, and the prevalence of the metabolic syndrome. Ferritin levels also correlated with individual components of the metabolic syndrome, particularly serum triglycerides and plasma glucose, as well as markers of insulin resistance.

Several cross-sectional studies have previously examined the association between iron stores and individual metabolic cardiovascular risk factors, including hypertension (3), dyslipidemia (4,5), elevated fasting insulin and blood glucose (6), and central adiposity (7). Our results are consistent with the previous literature and extend these observations to include the metabolic syndrome. The explanation for the observed j-shaped association between ferritin levels and markers of insulin resistance in postmenopausal women is unknown. It is possible that a carbohydrate-rich, protein-poor diet, which may lead to iron deficiency, may also contribute to insulin resistance (19).

Several lines of evidence support the association between elevated iron stores and insulin resistance. Type 2 diabetes is commonly observed in patients with hemochromatosis, a group with demonstrated iron overload (20). Second, serum ferritin levels correlated with various measures of insulin sensitivity in several different populations, including individuals without diabetes (6,21), individuals with diabetes (22), women with gestational diabetes (23), and patients with thalassemia (24). Third, a small intervention study of phlebotomy in patients with type 2 diabetes resulted in improved insulin sensitivity (22). Finally, two prospective studies have shown an association between elevated iron stores and incident diabetes (1,2). These studies were conducted in Finnish men (1) and U.S. nurses (2), and both provide evidence that elevated ferritin levels precede the development of diabetes.

It is unclear whether elevated iron stores may simply be another marker of insulin resistance or whether elevated iron stores may contribute to the pathogenesis of altered metabolic states. Iron is a transition metal capable of causing oxidative tissue damage by catalyzing the formation of free radicals (12). Iron overload may contribute to insulin resistance through mechanisms related to both reduced extraction of insulin and impaired insulin secretion (25). Results of oral glucose tolerance tests in patients with hemochromatosis have suggested that hepatic iron overload results in impaired insulin extraction (10). Others have suggested that iron deposition in pancreatic β-cells may also impair insulin secretion in more advanced states of iron overload (25).

Serum ferritin is a widely used marker of iron status in epidemiological studies (20) and accurately reflects differences in body iron stores by age and sex (26). However, serum ferritin is an acute-phase reactant and may be artificially elevated in the presence of inflammation (18). We attempted to minimize this potential source of confounding by adjusting for CRP and by excluding those individuals with suspected inflammation, infection, and liver disease in sensitivity analyses; however, we cannot rule out residual confounding by other inflammatory conditions. Similarly, we cannot rule out residual confounding by unmeasured factors that may be related both to the prevalence of the metabolic syndrome and to ferritin levels. A further limitation of this study is the cross-sectional design, which prevents us from making inferences about the directionality of the associations. Although there is increasing evidence to suggest that iron may influence glucose metabolism, it is possible that glucose metabolism may also influence body iron stores. Insulin, possibly through the activation of hypoxia-inducible factor-1α, may stimulate the production of erythropoietin, thus increasing the efficiency of iron absorption (27).

In summary, moderately elevated iron levels were associated with an increased prevalence of the metabolic syndrome and markers of insulin resistance. These associations were evident at moderately elevated iron levels, below levels associated with hemochromatosis. Given the high prevalence of elevated iron stores, especially in older ages, prospective studies are needed to determine whether moderately elevated iron stores precede the development of insulin resistance and contribute to the increased risk associated with it.

M.J. was supported by National Heart, Lung, and Blood Institute Training Grant 5T32-HL-007024-27.