Obesity increases the risk for iron deficiency, but the underlying mechanism is unclear. It is possible that overweight individuals may have lower dietary iron intake and/or bioavailability. Alternatively, obesity-related inflammation may increase hepcidin concentrations and reduce iron availability. Circulating hepcidin levels have not been compared in normal weight vs overweight individuals.
The objective of this study was to compare iron status, dietary iron intake and bioavailability, as well as circulating levels of hepcidin, leptin and interleukin-6 (IL-6), in overweight vs normal weight children.
In 6–14-year-old normal and overweight children (n=121), we measured dietary iron intake, estimated iron bioavailability and determined body mass index s.d. scores (BMI-SDS). In all children (n=121), we measured fasting serum ferritin, soluble transferrin receptor (sTfR), C-reactive protein (CRP) and leptin; in a subsample, we measured IL-6 (n=68) and serum hepcidin (n=30).
There were no significant differences in dietary iron intake or bioavailability comparing normal and overweight children. The prevalence of iron-deficient erythropoiesis (an increased sTfR concentration) was significantly higher in the overweight than in the normal weight children (20 vs 6%, P=0.022, with sTfR concentrations of 4.40±0.77 and 3.94±0.88 mg l−1, respectively, P=0.010). Serum hepcidin levels were significantly higher in the overweight children (P=0.001). BMI-SDS significantly correlated with sTfR (P=0.009), serum hepcidin (P=0.005) and the three measures of subclinical inflammation, namely CRP (P<0.001), IL-6 (P<0.001) and leptin (P<0.001). In a multiple regression model, serum hepcidin was correlated with BMI-SDS (P=0.020) and body iron (P=0.029), but not with the inflammatory markers.
Our findings indicate that there is reduced iron availability for erythropoiesis in overweight children and that this is unlikely due to low dietary iron supply but rather due to hepcidin-mediated reduced iron absorption and/or increased iron sequestration.
In industrialized countries, rates of iron deficiency are higher in overweight children and adults compared with their normal weight counterparts.1, 2, 3, 4, 5, 6, 7, 8, 9 The reason for this is unclear. Hepcidin is a 25 amino-acid peptide hormone that has a key role in body iron regulation. It is produced mainly by the liver but also by adipose tissue.10, 11 Increased circulating hepcidin levels restrict intestinal iron absorption and macrophage iron release. Hepatic hepcidin expression is modulated in response to body iron stores, hypoxia and inflammation, and low-grade inflammation is a characteristic of obesity. Proinflammatory cytokines, such as interleukin-6 (IL-6), effect hepcidin gene transcription through JAK (Janus kinase)–STAT3 (signal transducer and activator of transcription 3) interactions.12, 13 In addition, the adipokine leptin upregulates hepatic hepcidin expression through the JAK2–STAT3 signaling pathway.14 Thus, increased inflammation and/or leptin levels in obese individuals could reduce iron availability. In humans, higher body mass index (BMI) predicts decreased iron absorption measured using stable isotopes, and also an impaired response to iron fortification.9, 14, 15, 16
It has also been suggested that reduced iron intake due to poor dietary choices by overweight individuals may contribute to poor iron status, but there is little data to support this.1, 9, 15, 16 In adults, dietary iron intake and its estimated bioavailability are similar in overweight and normal weight adults,17 but no data are available for children. Even if iron intakes and estimated dietary bioavailability are similar, absorption of dietary iron could be reduced in overweight individuals if the circulating hepcidin level is increased. To gain insight into the links between obesity and iron status, our study aim was to compare iron status, dietary iron intake and bioavailability, as well as circulating levels of hepcidin, leptin and IL-6 in overweight vs normal weight children.
Subjects and methods
The subjects included in this study were 6–14-year-old children (n=121), living in the German-speaking part of Switzerland. The children were a convenience sample recruited in two steps (first step n=68, second step n=53) mainly through letters to primary schools, but as recruitment of overweight children proved to be difficult, also to pediatricians. Data collection started in the spring of 2005 and was completed in the fall of 2007. At primary schools, all children interested in participating were invited, and the children recruited through pediatricians were otherwise healthy overweight and obese children not yet enrolled in a weight-loss program. Data on relationships between diet and metabolic factors from a subgroup of these children (n=79) have been published previously.18, 19, 20, 21 Informed written consent was obtained from the parents and informed oral assent from the children. Ethical approval for the study was obtained from the ethics committee of the Swiss Federal Institute of Technology in Zürich.
Dietary assessment was made using two 24-h recalls and a 1-day weighed food record in the first group of children (n=68) and, to further improve data quality, two 24-h recalls combined with one 3-day weighed food record in the second group of children (n=53). All interviews were conducted by well-trained female interviewers in the family home. Each family was visited twice by the same interviewer within 3 weeks. The 24-h recalls were performed at each visit and at the first visit the interviewer gave instructions and guidelines for the weighed food record. At the second visit, the weighed food record was carefully reviewed with the child and caretaker. Volumes and portion sizes for the 24-h recalls were estimated using photographs of different portions of various foods and meals. A combination of 24-h recalls and a food record was used to provide a good overview of the children's habitual diet; this approach has been validated in children as young as 8 years of age.22 An appointment was scheduled at the hospital clinic, at the parents’ convenience, usually within 1–2 weeks of the dietary assessment.
The children presented to the hospital clinic in the morning after a 12 h overnight fast. A volume of 12 ml of blood was collected by venipuncture. Height was measured to the nearest 0.5 cm and weight to the nearest 100 g using a digital balance (Beurer BF 18, Beurer, Ulm, Germany). Pubertal staging was performed by presenting drawings of the different Tanner stages to the child and the parents.
Dietary data obtained from the three to five records were thoroughly checked and entered into a nutrition software system (EBISpro for Windows 6.0, Dr J Erhardt, University of Hohenheim, Germany). This system translates the amount of food eaten into individual nutrients and classifies consumed foods into 22 food groups. The program is based on the German Food and Nutrition Data Base BLS 2.3 (Federal Health Department, Berlin, Germany), and, for foods specific to Switzerland, incorporates values from the Swiss Food Composition Database.23 Energy and nutrient data were averaged across the 3 or 5 days required to obtain a mean daily energy and nutrient intake for each child.
Total dietary iron intake was calculated using the program EBISpro, but for the determination of heme and non-heme iron all iron-containing products were separately divided into animal tissues and other components. Percentages of heme iron in cooked animal products were estimated to be 60% for beef, 40% for pork, 30% for chicken and 25% for fish.24 Heme iron bioavailability was assumed to be 23%.25 Bioavailability of non-heme iron was calculated according to the equation of Monsen and Balintfy,26 which takes into account two enhancers of iron absorption, namely vitamin C and animal tissue. Enhancing factors (EFs) were calculated as the sum of milligrams of ascorbic acid plus grams of cooked animal tissue. The further calculation of % (percentage) iron bioavailability is meal based, but as the meal-to-meal variability in iron intake is modest,27 it was decided to use the mean daily non-heme iron and EF intake of each child and to divide the value by 3 to achieve an approximate per meal intake value. The following equations were then used to calculate non-heme iron bioavailability:26
If Σ EF <75: % bioavailability=3+8.93 × ln (EF+100/100)
If Σ EF 75: % bioavailability=8
Body mass index of the children was calculated as weight (kg) divided by height (m2). To scale the data for comparison across ages and sex, BMI s.d. scores (BMI-SDS: individual BMI value−reference mean BMI value divided by s.d.) were calculated using the program Epi Info (Version 3.4.1, Centers for Disease Control and prevention (CDC), Atlanta, GA, USA) and the reference values proposed by the US CDC (US Centers for Disease Control and Prevention) in 2000. Age- and gender-specific criteria also collected from CDC28 were used to classify children as normal weight or overweight (above the 85th percentile). These criteria have been previously validated in Swiss children between the age group of 6 and 12 years.29
Serum was stored at −20 °C until analysis: serum ferritin (SF) and C-reactive protein (CRP) were measured using chemiluminescent immunometry (Serum Ferritin/high sensitivity CRP, IMMULITE Bühlmann Laboratories AG, Allschwil, Switzerland), soluble transferrin receptor (sTfR) by immunturbidimetry (sTfR, Cobas Integra, Roche, Basel, Switzerland), IL-6 and leptin using ELISA (enzyme-linked immunosorbent assay) test kits (high sensitivity ELISA, Quantikine HS Human IL-6 Immunoassay, R&D Systems, Minneapolis, MN, USA; Leptin (human) ELISA Kit, BioVender, Alexis Biochemicals, Lausen, Switzerland). For a subsample of 30 subjects (9 normal weight, 6 overweight and 15 obese), in which adequately stored serum was still available, serum hepcidin concentrations were determined at the Department of Clinical Chemistry of the Radboud University Nijmegen, The Netherlands (http://www.hepcidinanalysis.com), using a combination of weak cation exchange chromatography and time-of-flight mass spectrometry.30 The subsample did not differ from the complete sample with regard to age, BMI, iron status, dietary iron intake or inflammatory markers. Normal reference ranges are sTfR concentration <5.0 mg l−1 (as indicated by the manufacturer) and SF concentrations of >15 μg l−1.31 As sTfR concentrations measured using the above-mentioned method are on average 30% lower as compared with the values obtained using the ELISA assay (Ramco Laboratories, Stafford, VA, USA,32 the values were corrected before the calculation of body iron stores as proposed by Cook et al.33 For the correction of sTfR concentrations, the following formula was used: sTfR corr (Ramco Laboratories)=(sTfR (Ramco Laboratories)−0.299)/0.631.32
After correction, body iron stores were calculated as follows:
Body iron (mg kg−1)=−[log(sTfR corr/SF)−2.8229]/0.1207.33
Statistical analysis was carried out using the statistical package SPSS 13.0 for Windows (SPSS, Chicago, IL, USA). Non-normally distributed variables were log-transformed for comparisons. Normally distributed data were presented as means±s.d. and non-normally distributed data as medians (range). The data of three children were excluded before analysis: one child reported having had influenza and had an increased CRP concentration (3.5 mg l−1); in two other children (one in the normal weight and one in the overweight group) hepcidin concentrations were extreme outliers (>3 s.d. above the mean). An independent samples t-test was used to compare between the two weight categories, and multiple regression models were used to study the effect of iron intake and bioavailability, as well as the effects of obesity and inflammatory markers on iron status, and the interactions between hepcidin, obesity, inflammation and iron status. All equations were checked for confounding factors such as adiposity, age, puberty and gender, and they were added as covariates if necessary. Fisher's exact test was used to compare proportions of iron-deficient and non-iron-deficient subjects between normal weight and overweight subjects.
Anthropometric characteristics, iron status parameters, markers of subclinical inflammation and iron intake, all by weight classification, are shown in Table 1. There was no significant difference in age between the two groups. Iron status, SF and body iron stores were not different between the groups, whereas sTfR was significantly higher in the overweight group (P=0.01). The prevalence rates of iron deficiency (based on the sTfR concentration) in the overweight and normal weight groups were 20 and 6% respectively (P<0.022). The three markers of subclinical inflammation (CRP, IL-6 and leptin) increased with increasing adiposity (P<0.01). Hepcidin concentrations were significantly higher in the overweight children compared with the normal weight children (P=0.002).
The dietary variables are shown in Table 2. Total iron intakes did not differ significantly between groups. Heme iron intake and consumption of meat products were significantly higher in the overweight children compared with the normal weight children (P=0.006 and P=0.002, respectively). However, % bioavailability of non-heme iron and total bioavailable iron did not differ significantly between the groups. There was a nonsignificant trend toward higher amounts of bioavailable iron in the overweight group (P=0.077).
In a continuous, univariate analysis, BMI-SDS was significantly correlated with (1) sTfR (r=0.240, P=0.009, see Figure 1); (2) serum hepcidin concentrations (r=0.527, P=0.005, see Figure 2); and (3) the three measures of subclinical inflammation, CRP (r=0.488, P<0.001), IL-6 (r=0.434, P<0.001) and leptin (r=0.809, P<0.001). Moreover, BMI-SDS was correlated with heme iron intake (r=0.272, P=0.003) and the intake of meat products (r=0.269, P=0.003), as well as total bioavailable iron (r=0.192, P=0.037) but not with total iron intake. After controlling for BMI-SDS, SF was significantly correlated with CRP (r=0.311, P=0.001) but not with IL-6 or leptin. Serum hepcidin was correlated with body iron (r=0.467, P=0.014, see Figure 2) but not with any of the inflammatory markers.
In a multiple regression model with sTfR as the dependent variable, only BMI-SDS was the significant predictor (P=0.027, β=0.238); iron intake, iron bioavailability, age, gender and pubertal stage were not significant predictors. In a multiple regression model with serum hepcidin as the dependent variable, both BMI-SDS (P=0.020, β=0.680) and body iron (P=0.039, β=0.364) were significant predictors, but the inflammatory markers CRP and leptin (n=28, adjusted r2=0.417) were not.
Given that obesity is associated with subclinical inflammation, and that SF is an acute-phase protein, sTfR is likely to be the best clinical measure of iron status in overweight individuals.16 In our data, adiposity was a predictor of poor iron status as measured by sTfR, but not SF. In iron deficiency, sTfR is increased because cell expression of the transferrin receptor is upregulated to increase the uptake of circulating iron, primarily into marrow red cell precursors. sTfR concentrations are not significantly affected by inflammation, and are therefore useful in differentiating iron deficiency from inflammatory hypoferremia.34 Thus, in iron-deficient overweight children, the sTfR and SF may be discrepant because of the confounding effect of obesity-associated inflammation on SF. In our children, all three inflammatory markers (CRP, IL-6 and leptin) significantly increased with increasing adiposity, and CRP was a significant predictor of SF, but not TfR, independent of adiposity. Thus, our data indicate poorer iron status in overweight compared with normal weight children, consistent with previous studies1, 4, 5, 6, 9 and emphasize the limitations of SF as an iron status indicator in overweight individuals.
It has been suggested that obesity may be associated with a low-quality diet containing little iron.9 However, comparisons of dietary iron intake of normal and overweight adults9, 17 and children35 have found no significant differences in total iron intake. However, in iron nutrition, total dietary iron is less important than the amount of bioavailable iron. In humans, dietary iron is consumed in two major forms, namely heme and non-heme iron. As the bioavailability of these forms of dietary iron varies considerably, total dietary iron intake is not necessarily a good predictor of absorbed iron.36 Although in western diets the intake of heme iron accounts for only ∼10% of total dietary iron, it usually supplies most of the absorbed iron as it is relatively well absorbed (25–35%).26, 37 On the other hand, non-heme iron accounts for ∼90% of total dietary iron, but its absorption is strongly influenced by dietary inhibitors and enhancers and is usually only 3–8%, depending on iron status.26 Although total dietary iron intake in this study did not differ between weight groups, the intake of heme iron was significantly higher in the obese children because of their higher meat intake. Our findings agree with a recent adult study, in which total iron consumption and iron bioavailability did not differ between obese and non-obese subjects, whereas heme iron intake and the consumption of animal protein were higher in obese adults.17
Calculating dietary iron bioavailability is challenging. Several algorithms are available for estimating the bioavailability of non-heme dietary iron, based on the intake of inhibiting factors and EFs.25, 26, 38, 39, 40, 41 In this study, we applied the algorithm by Monsen and Balintfy,26 which includes two EFs, vitamin C and animal tissue, but no inhibiting factors. The major inhibitor of non-heme iron absorption, phytic acid, is included in several other algorithms,25, 38, 39, 40 but because reliable data on the phytic acid content of many foods consumed in Switzerland are not available, we were unable to use these algorithms. Despite this caveat, our data suggest that decreased dietary iron intake and/or bioavailability are an unlikely explanation for poorer iron status in overweight children.
In our study children, estimates of dietary iron intake and bioavailability did not vary with adiposity. However, the bioavailability of iron may be modulated not only by dietary factors but also by physiological factors. Iron status, mediated by circulating hepcidin levels, is a primary determinant of dietary iron absorption. However, hepatic hepcidin expression is modulated not only by body iron stores but also by hypoxia and inflammation. Proinflammatory cytokines, such as IL-6 and leptin, activate hepcidin gene transcription through a STAT3 binding motif at position −64/−72 of the promoter.12
The major new finding in this study is that circulating hepcidin concentrations are increased in overweight children, independent of iron status. Higher levels of hepcidin in overweight children may be due to obesity-related inflammation, and three inflammatory markers, CRP, IL-6 and leptin, were increased in the overweight children. However, we found no association of hepcidin with any of these inflammatory markers. However, hepcidin is not only expressed in the liver but also at messenger RNA and protein levels in adipose tissue, and hepcidin messenger mRNA expression is increased in adipose tissue of obese patients.10 Thus, it is possible that the increased hepcidin levels in our overweight children originated at least partially from adipose tissue rather than from inflammatory STAT3 activation of liver hepcidin expression.
Our findings confirm that there is reduced iron availability for erythropoiesis in overweight children and that this is unlikely due to low dietary iron supply but rather due to hepcidin-mediated reduced iron absorption and/or increased iron sequestration. However, this hypothesis needs confirmation by intervention studies in overweight children that show that weight loss reduces circulating hepcidin levels, increases iron absorption and improves iron status.
Conflict of interest
The authors declare no conflict of interest.
We would like to thank all the children and families who participated in the study. We also thank Isabella Sciaroni, Nicole Beljean, Alexandra Uster and Evelyne Pflugi (ETH Zürich, Switzerland) who assisted with the study. Special thanks also go to Harold Tjalsma from http://www.hepcidinanalysis.com in Nijmegen for his assistance with the hepcidin measurements.
Each of the authors contributed to the study design and the statistical analysis of the data as well as the writing and editing of the paper. Data collection and laboratory analysis was done by IA.
This study was funded by the ETH Zurich.