Original Article | Published:

Overweight impairs efficacy of iron supplementation in iron-deficient South African children: a randomized controlled intervention

International Journal of Obesity volume 37, pages 2430 (2013) | Download Citation



Many countries in the nutrition transition have high rates of iron deficiency (ID) and overweight (OW). ID is more common in OW children; this may be due to adiposity-related inflammation reducing iron absorption.


We investigated whether weight status predicts response to oral iron supplementation in ID South African children.


A placebo-controlled trial of oral iron supplementation (50 mg, 4 × weeks for 8.5 months) was done in ID 6- to 11-year-old children (n=321); 28% were OW or obese. BMI-for-age z-scores (BAZ), hepcidin (in a sub-sample), hemoglobin, serum ferritin (SF), transferrin receptor (TfR), zinc protoporphyrin (ZnPP) and C-reactive protein (CRP) were measured; body iron was calculated from the SF to TfR ratio.


At baseline, BAZ correlated with CRP (r=0.201, P<0.001) and CRP correlated with hepcidin (r=0.384, P<0.001). Normal weight children supplemented with iron had significantly lower TfR concentrations at endpoint than the OW children supplemented with iron and the children receiving placebo. Higher BAZ predicted higher TfR (β=0.232, P<0.001) and lower body iron (β=−0.090, P=0.016) at endpoint, and increased the odds ratio (OR) for remaining ID at endpoint in both the iron and placebo groups (iron: OR 2.31, 95% CI: 1.13, 4.73; placebo: OR 1.78, 95% CI: 1.09, 2.91). In the children supplemented with iron, baseline hepcidin and BAZ were significant predictors of endpoint TfR, with a trend towards a hepcidin × BAZ interaction (P=0.058).


South African children with high BAZ have a two-fold higher risk of remaining ID after iron supplementation. This may be due to their higher hepcidin concentrations reducing iron absorption. Thus, the current surge in OW in rapidly developing countries may undercut efforts to control anemia in vulnerable groups. The trial is registered at clinicaltrials.gov as NCT01092377.


Countries in the ‘nutrition transition’ are undergoing rapid dietary and lifestyle changes that produce a double burden of malnutrition: their populations suffer from increasing overconsumption (for example, obesity, diabetes) but continue to have high rates of micronutrient deficiencies (for example, iron deficiency anemia (IDA)).1, 2, 3 Not only do overweight (OW) and iron deficiency (ID) often coexist in these populations, it appears they also interact, with adverse consequences. Studies in high- and low-income countries have consistently found OW individuals at all ages have higher rates of ID than normal weight (NW) counterparts.4, 5, 6, 7, 8

There is increasing evidence that inflammatory cytokines produced by excess adipose tissue in obesity stimulate secretion of hepcidin by the liver.4, 9, 10, 11, 12 High circulating hepcidin concentrations reduce iron efflux from enterocytes and thereby lower dietary iron absorption; they also reduce iron release by macrophages and thereby lower iron cycling from senescent erythrocytes. These mechanisms may at least partially explain the increased risk of ID in OW individuals.13 Compared with NW children, OW children have higher C-reactive protein (CRP), interleukin (IL)-6, leptin and hepcidin concentrations that predict poor iron status.4 Adiposity in young women is associated with lower iron absorption measured by stable iron isotopes.10 Weight loss in adults and children is associated with lower hepcidin concentrations and improved iron status.12, 14 Overall, the data suggest that the link between OW and ID is likely to be due to an inflammation-induced increase in circulating hepcidin. However, studies on this topic have been cross-sectional or uncontrolled, making conclusions on causality uncertain.

South Africa is undergoing the nutrition transition and the current surge in OW in its population may impair efforts to control iron deficiency anemia (IDA) in vulnerable groups, such as young women and children.15, 16 Concurrent IDA and obesity in children may be particularly detrimental as both can reduce activity levels and impair cognition.17, 18, 19 Therefore, our study aim was to determine if body weight status modifies the efficacy of iron supplementation in South African children with poor iron status. Our hypothesis was that greater adiposity would reduce iron absorption and blunt the response to iron supplementation.

Subjects and methods

Study site

The study was conducted between November 2009 and November 2010 at four primary schools serving low-income rural villages in the Province of KwaZulu-Natal, in eastern South Africa. This study was part of a larger study that investigated the effects of combined supplementation with iron and n-3 fatty acids on cognition. When n-3 fatty acid status was included in the regression models as a covariate, it did not significantly affect the relationships between weight and iron status, and the fatty acid data are not discussed here. Before the start of the study, the parents of all children in the first five grades at the schools were invited to attend a meeting at which the study purpose and procedures were explained. The subjects were then asked to join the study and written informed consent from the parents or guardians and verbal assent from the children were obtained. The ethical committees of the North-West University, Potchefstroom, South Africa and the ETH Zürich, Switzerland approved the study protocol.


In total, 926 children participated in the baseline screening. Inclusion criteria for the intervention study were: (1) age 6–11 years; (2) hemoglobin (Hb) >8 g dl−1; (3) ID, defined as either serum ferritin (SF) <20 μg l−1, zinc protoporphyrin (ZnPP) >70 μmol per mol heme in washed red blood cells or serum transferrin receptor (TfR) >8.3 mg l−1; (4) apparently healthy, with no chronic illness; (5) no consumption of iron-containing dietary supplements. Children found to have an Hb <8 g dl−1 were referred for medical treatment. In the screening, 349 children met the inclusion criteria and were asked to join the intervention study. Sample size calculation for this study was done based on the correlation between BMI z-score (BAZ) and changes in TfR observed in earlier iron supplementation trials.10 Using an r2 of 0.025, an alpha of 0.05 and a power of 80%, sample size calculation indicated a sample size of 244 subjects would be sufficient; to cover potential drop-outs the sample size was increased by 20% to a final sample size of 300.

Randomization and blinding

To maintain anonymity, all subjects participating in the baseline screening received a personal code, which was used throughout the study. The subjects meeting the inclusion criteria were stratified first by school and by grade. Randomization was performed by means of a computer-generated list, blocked by school, and the enrolled subjects were assigned to treatment codes and respective group colors, which were used on supplement containers and subject name tags throughout the trial. Participants, investigators, staff and the sponsors were blinded to treatment assignment. The group codes were held by a member of an independent Safety Monitoring Board until data analysis was completed.

Study design

The study was a placebo-controlled, double-blind intervention trial. All participating children were dewormed at baseline and at midpoint (4 months) with an oral dose of 400 mg mebendazole. The iron group received one oral tablet containing 50 mg iron as iron sulfate (Lomapharm, Paul Lohmann GmbH, Emmertal, Germany) 4 days per week during school days. This iron dose was chosen based on the results from previous supplementation trials that found beneficial effects on cognition, which was the primary outcome of the main study. The control group received the same regimen of placebo tablets (Lomapharm, Paul Lohmann GmbH) identical in appearance to the iron tablets. The tablets were administered in the morning, directly after the children had arrived at school (before 0800 hours). Tablets were swallowed with a 200 ml fruit-flavored beverage that contained ≈10 mg vitamin C per serving. Trained fieldworkers directly supervised tablet consumption and recorded compliance.

The study was supervised by an independent Safety Monitoring Board. Two board members performed two field visits during the study and data collection. The intervention period was for 8.5 months, with a total of 105 supplementation days. All children remaining ID at the end of the study received additional iron supplementation.

Laboratory analysis

Venous blood samples (10 ml) were drawn into EDTA-coated and into trace element-free Vacutainer tubes (Becton Dickinson, Woodmead, South Africa) at baseline and endpoint. Hb concentration was measured on site on an aliquot of whole blood by direct cyanmethemoglobin method (Ames Mini-Pak Hb test pack and Ames Minilab, Bio Rad Laboratories (PTY) Ltd, Johannesburg, South Africa), using Drabkins solution and a standard mini-photometer. The remaining blood samples were centrifuged at 500 g for 15 min at room temperature, and plasma and serum were aliquoted and stored at −20 °C for the duration of the fieldwork (4 days). Red blood cells were washed twice with 0.15 mol l−1 NaCl, and centrifuged at 500 g for 10 min to remove the buffy coat. ZnPP was measured on washed red blood cells using a hematofluorometer (Aviv Biomedical, Lakewood, NJ, USA) and three-level control material provided by the manufacturer within the same day of blood sampling. After completion of the fieldwork, samples were transported on dry ice and stored at −80 °C until analysis. SF and CRP were measured using an automated chemiluminescent immunoassay system (IMMULITE, DPC Bühlmann GmbH, Aschwil, Switzerland). Serum TfR was measured using an in vitro enzyme immunoassay (Ramco Laboratories, Inc., Stafford, TX, USA). Body iron was calculated from the ratio of TfR:SF according to the equation of Cook et al.20 For reporting of prevalence and statistical analyses, ID was defined as either ZnPP >70 μmol per mol heme,21 TfR >8.3 mg l−1 (test kit reference value), or SF <15 μg l−1.22 Anemia was defined as Hb <11.5 g dl−1.23 Inflammation was defined as CRP >5 mg l−1. SF values of subjects with CRP >5 mg l−1 were excluded from the analysis, due to the confounding effects of inflammation on SF. Plasma hepcidin concentrations (nmol l−1) were determined in a sub-sample of the study children (n=99, all children ranked by BAZ and every third child selected to represent the entire BAZ range) by a combination of weak-cation-exchange chromatography and time-of-flight mass spectrometry (WCX-TOF MS). More specifically, an internal standard (synthetic hepcidin-24, Peptide International Inc., Louisville, KY, USA)24 was added to 50 μl plasma before total hepcidin was enriched from the sample with Macro-Prep CM Support beads (Bio-Rad Laboratories, Hercules, CA, USA) at neutral pH. Next, hepcidin-enriched samples were applied to a MSP 96-polished steel MALDI target plate followed by the addition of cyano-4-hydroxy-cinnamic acid as energy-absorbing matrix, all in nitrogen atmosphere. Peptide spectra were generated on a Microflex LT matrix-enhanced laser desorption/ionization TOF MS platform (Bruker Daltonics GmbH, Bremen, Germany). The lower limit of detection of this method was 0.5 nmol l−1; ranges for the coefficients of variation were 2.2–3.7% (intra-run) and 3.9–9.1% (inter-run).25


Body weight was measured without shoes and outerwear, in light clothing, to the nearest 0.01 kg on a load-cell-operated digital scale (Masskot, UC-300 Precision Health Scale; A&D Co, Stanger, South Africa) calibrated using fixed weights. Height was measured (without shoes) to the nearest 0.1 cm using a rigid stadiometer, which was calibrated using a steel tape. Age- and sex-specific height-for-age (HAZ), weight-for-age (WAZ) and BMI-for-age z-scores (BAZ) were calculated using the 2007 World Health Organization (WHO) growth standards for children at age of 5–19 years with the software WHO Anthro Plus for personal computers, version 1.0.3, 2010 (WHO, 2009). Children with a BAZ >1 and <2 were classified as OW and those with BAZ 2 were classified as obese.

Statistical analysis

Statistical analysis was done by using IBM SPSS Statistics Version 19 (IBM Company, Armonk, NY, USA). All data were checked for normal distribution and for the presence of outliers (± 3 s.d. from the mean). Non-normally distributed data were log-transformed before data analysis. No suitable transformation could be found for CRP, therefore non-parametric tests were applied for this variable. Drop-outs and missing data were treated using multiple imputations, under the assumption that values were missing at random. The lower detection limit of the hepcidin assay was 0.5 nmol l−1, and values below this limit were imputed using multiple imputations specifying the maximum value as 0.5 nmol l−1. Baseline characteristics in NW and OW (including obese) children were analyzed using independent t-tests for normally distributed continuous variables, Mann–Whitney U-test for CRP and χ2 tests for categorical variables. Differences between treatment groups and respective weight sub-groups at baseline were investigated using ANOVA. Differences between treatment groups at endpoint were analyzed using analysis of covariance (ANCOVA), including gender, age and school as individual covariates. Bivariate Pearson’s correlations (except for CRP, where Spearman correlations were done) and multiple linear regression models were used to study associations between continuous variables at baseline. For the multiple linear regression analyses, diagnostics were performed on the final models to ensure that all statistical assumptions were met (linearity, normal distribution of residuals, multicollinearity). Effects of treatment and weight group (NW vs OW), as well as potential interactions on endpoint iron status measurements were investigated using two-way ANCOVA controlling for gender, age, school and respective baseline iron status measurements. Multiple linear regression analyses were done on the endpoint iron status measurements, including age, gender, school and the baseline measurement as individual level covariate. Furthermore, we added a treatment × BAZ interaction term (orthogonalized) to the model to test whether the effect of BAZ on iron status was in response to iron treatment. Moreover, we examined the odds ratios (OR) for being ID at endpoint with increasing BAZ using binary logistic regression analyses, adjusting for age, gender, school and respective baseline ID prevalence, again using the treatment × BAZ interaction term (orthogonalized). Multivariate regression analyses including endpoint iron status measurements as dependent and baseline hepcidin, baseline BAZ and a hepcidin × BAZ interaction term (orthogonalized) as independent variables were used to examine the influence of hepcidin on the relationship between BAZ and iron status at endpoint. P-values <0.05 were considered significant.


Of the 321 children enrolled, 294 completed the study, with a similar drop-out rate in the two treatment (iron vs placebo) groups of approximately 8%. For data analyses, the children were further divided by their baseline BAZ into OW or obese (BAZ >1) and NW (BAZ 1) sub-groups. Baseline anthropometric characteristics and iron status indices by these weight sub-groups are shown in Table 1. Nearly 30% of the children were OW or obese, and compared with NW, the OW or obese children had significantly higher CRP concentrations (P=0.007) indicating greater levels of inflammation. In bivariate correlations, CRP significantly correlated with SF (rs=0.297, P<0.001), ZnPP (rs=0.231, P<0.001) and body Fe (rs=0.252, P<0.001), but not with TfR or Hb. The baseline prevalence of ID was 26.0%, 11.4% and 62.7% based on SF, TfR and ZnPP, respectively. The prevalence of anemia was 21%, but only 9.0% of the children had IDA based on SF and 15.2% based on ZnPP. There were no significant baseline differences in Hb or iron status indices comparing NW with OW children. In bivariate and multivariate correlations (controlling for age, gender and school), BAZ was a weak significant predictor of ZnPP (r=0.142, P=0.013; β=0.139, P=0.015), but of no other iron index. In addition, BAZ correlated with CRP (rs=0.201, P<0.001; β=0.182, P<0.001). However, in the multivariate analyses, there were no significant BAZ × CRP interactions on the iron status indices. Thus, OW children began the intervention with higher levels of subclinical inflammation but comparable iron status to their NW counterparts.

Table 1: Baseline characteristics of the iron-deficient South African children enrolled in the iron intervention by weight status, before randomization into the two treatment groupsa

Iron status indices at baseline and after the 8.5-month intervention by weight status and treatment group are shown in Table 2. There were no significant differences in the iron indices among the four groups at baseline. In the two-way ANCOVA to assess the effects of treatment and weight status on endpoint iron status, there were significant iron treatment effects on all iron indices for improved iron status compared with the placebo group (all P<0.001). There were significant increases (within the normal reference range) in median TfR from baseline to endpoint within all groups, except within the NW group receiving iron (P=0.07). There was a significant effect of iron treatment for lower TfR concentrations at endpoint (P<0.001). The baseline prevalence of ID based on TfR was significantly higher in the children allocated to iron treatment compared with the children allocated to placebo (15.6% vs 7.5%, P<0.001). The prevalence of ID based on TfR decreased from 15.6% at baseline to 8.1% at endpoint (P<0.05) in the children receiving iron, whereas the ID prevalence increased from 7.5% to 23.6% in those receiving placebo (P<0.001). Furthermore, there was a significant effect of OW including obesity for higher endpoint TfR concentrations (P<0.001). The NW children treated with iron had significantly lower TfR concentrations at endpoint than the OW children treated with iron and the NW and OW children receiving placebo. Mean changes in TfR from baseline to endpoint by group are shown in Figure 1. There were no significant treatment × weight group interactions on any of the iron status indices. Thus, irrespective of treatment group, being an OW or obese child was a predictor of poorer iron status at endpoint.

Table 2: Effects of iron treatment and weight group on iron status indicators in South African children over 8.5 monthsa
Figure 1
Figure 1

Mean changes in TfR concentrations from baseline to endpoint in four groups of iron-deficient South African children: overweight including obese (OW, BAZ >1) and treated with iron (Iron-OW, n=46), NW (BAZ 1) treated with iron (Iron-NW, n=114), OW and treated with placebo (placebo-OW, n=45) and NW and treated with placebo (placebo-NW, n=116). Values are means±s.e.m. Labeled means without a common letter differ significantly from each other, P<0.05.

A similar pattern was apparent in the continuous analyses as shown in Table 3. Treatment group was a significant predictor of all endpoint iron status indices. In addition, controlling for treatment group, BAZ was a significant positive predictor of endpoint TfR (Figure 2) and a negative predictor of body iron, and showed a non-significant positive trend for endpoint ZnPP (P=0.078). These correlations were not significantly influenced by treatment group (BAZ × group interactions, P>0.05 all).

Table 3: Multivariate linear regressions between the BMI-for-age z-scores at baseline and the endpoint iron status parameters in South African children after iron or placebo supplementation for 8.5 months (n=321)a
Figure 2
Figure 2

Correlation between BMI-for-age z-scores at baseline and transferrin receptor (TfR) concentrations at endpoint after iron (n=160) or placebo (n=161) supplementation for 8.5 months in iron-deficient South African school children (n=321). Correlations (using log-transformed TfR values): for iron-supplemented children, r=0.290 and P=0.001; for placebo-supplemented children, r=0.263 and P=0.001.

Iron treatment significantly decreased the OR for being ID (defined by TfR) at endpoint (OR=0.229, 95% CI: 0.08, 0.66), and higher BAZ significantly increased the OR (OR=1.66, 95% CI: 1.06, 2.58), including age, gender, school and ID prevalence at baseline as covariates, with no treatment group × BAZ interaction (P=0.735). In both the iron and placebo groups, higher BAZ significantly increased the OR for endpoint ID with the effect being stronger in the iron group (OR=2.31, 95% CI: 1.13, 4.73) than in the placebo group (OR=1.78, 95% CI: 1.09, 2.91). Higher baseline BAZ also significantly increased the OR for being ID at endpoint based on ZnPP (OR=1.90, 95% CI: 1.01, 3.56) with no significant treatment × baseline BAZ interaction (P=0.083).

At baseline, hepcidin concentration was significantly correlated with CRP (r=0.384, P<0.001), age (r=0.287, P=0.014), SF (r=0.486, P<0.001), body iron (r=0.537, P<0.001), and TfR (r=−0.218, P=0.032). In multivariate regressions controlling for age, gender and school (and for CRP for iron indices), these associations remained significant (SF, β=0.480, P<0.001; body iron, β=0.512, P<0.001; and TfR, β=−0.346, P=0.006). In bivariate correlations, hepcidin was not significantly correlated with BAZ (r=0.047, P=0.650) and median hepcidin was not significantly different in NW (n=71) and OW children (n=28) at baseline (P=0.272 between groups) (Table 1).

However, to investigate the influence of baseline hepcidin on the relationship between BAZ and endpoint TfR concentrations, we used multivariate regression with TfR as the dependent variable and hepcidin, BAZ and a hepcidin × BAZ interaction term as independent variables, as well as age, gender, school and respective baseline TfR as covariates. As iron treatment was a significant predictor of endpoint TfR, the analysis was done in the children receiving iron and placebo separately. In the children receiving placebo, none of the predictors were significant. However, in the children supplemented with iron, baseline BAZ was significant predictor of endpoint TfR (P=0.042). Furthermore, baseline hepcidin was a significant predictor of endpoint TfR (P=0.030) and there was a trend for a hepcidin × BAZ interaction (P=0.058). This hepcidin × BAZ interaction is apparent in Figure 3, where the correlations between baseline hepcidin and endpoint TfR for the NW and OW groups supplemented with iron are shown separately. Higher baseline hepcidin tended to be associated with higher TfR (poorer iron status) at endpoint only in the OW group (r=0.553 and P=0.078).

Figure 3
Figure 3

Correlation between baseline hepcidin and endpoint transferrin receptor concentrations in overweight including obese (OW, n=11; BAZ >1) and NW (n=25; BAZ 1) iron-deficient South African children supplemented with iron. Correlations (using log-transformed hepcidin and TfR values): for OW children, r=0.553 and P=0.078; for NW children, r=0.157 and P=0.453.


This is the first prospective study demonstrating that OW ID children, compared with their NW ID counterparts, have a reduced response to iron supplementation. The strengths of the study are its randomized controlled design, the provision of a high dose of well-absorbed supplemental iron, its fairly large sample size in a low-income population with a high rate of OW and a study site in an African country undergoing the nutrition transition.

South Africa is struggling to reduce high rates of anemia in young women and children in the face of a sharp increase in OW in these target groups. According to a 2005 national survey, 14% and 52% of children (aged 1–9 years) and women, respectively, are OW or obese, whereas 28% and 29% are anemic.26 Not only do OW and ID often coexist, it appears they also interact. Cross-sectional studies in high-income countries have repeatedly shown that OW individuals at all ages are at increased risk of ID4, 5, 6, 7, 27, 28, 29 and a similar pattern is seen in transition countries.8 In a previous study that reanalyzed data from iron fortification studies, higher BMI z-score predicted poorer iron status at baseline and less improvement in iron status during interventions in Moroccan and Indian school children.10 Limitations of that study were its secondary analysis of data, the lack of hepcidin measurements in the children and a low prevalence of OW at 6.3%.

In the present study, nearly 30% of children were OW or obese. Because ID was an inclusion criterion, we did not expect weight status to be strongly correlated with iron status at baseline, and only one of the iron indices, ZnPP, was weakly correlated with BAZ at baseline. This was desirable in that it allowed us to better test our study hypothesis of an impaired response to iron in OW and obesity, because the OW or obese children began the intervention with comparable iron status to their NW counterparts. Our data suggest weight status is a significant predictor of iron status. Higher BAZ predicted poorer iron status at the end of the intervention and significantly increased the OR for ID at endpoint. These effects occurred not only in children receiving iron treatment but also in children receiving placebo. This may have been due to the mitigating effects of OW and obesity on iron absorption from the native diet (maize flour in South Africa is iron fortified) during the study period, including factors in the placebo group that may have increased basal iron intake (sensitization of the local populace to the prevalence and adverse effects of ID, seasonal changes in diet) and/or increased basal iron absorption (provision of extra vitamin C in the drink given with the tablets could have improved iron absorption from foods consumed at a late breakfast). Another potential explanation for the increased risk of ID in OW children receiving placebo, might be that the iron requirements in growing OW children are higher than in their NW counterparts. Thus, at a similar level of iron intake the iron status of OW would deteriorate to a greater extent than in NW children during periods of growth. Nonetheless, the effect of OW or obesity on endpoint iron status was more pronounced in the iron treatment group and the risk of ID at endpoint with higher BAZ was greater in the iron group (OR=2.31) than in the placebo group (OR=1.78).

One challenge in studies linking OW, inflammation and ID is defining iron status. Indicators of iron stores, such as SF and serum iron are affected by inflammation independent of iron status;30 thus, they may be confounded by adiposity-related inflammation. In NW individuals with ID, SF is decreased and directly related to transferrin saturation.30 In contrast, in OW people with ID, SF tends to be higher than in NW individuals and inversely related to transferrin saturation.31, 32, 33 Thus, in obesity, SF reflects both obesity-related inflammation and iron status. This effect may explain why in the present study there were associations between BAZ and TfR and ZnPP, but not with SF at endpoint.

Why is iron supplementation less effective in OW children? There are several possible explanations. OW individuals may have higher iron requirements because of an increased blood volume34 or increased integumentary iron losses due to greater body surface area. However, increasing evidence suggests absorption and/or utilization of dietary iron may be reduced in OW individuals because adipose tissue-related inflammation may increase circulating hepcidin. Hepatic hepcidin expression is modulated by both body iron stores and inflammation.9 Increased IL-6 and leptin levels are characteristic of the inflammation of OW, and leptin, a proinflammatory adipokine, as well as IL-6, activate hepcidin gene transcription through JAK (Janus kinase)–STAT3 (signal transducer and activator of transcription interactions).35, 36, 37 Previous studies have reported associations between adiposity-related inflammation and circulating hepcidin.4, 14, 38, 39 Our data support this hypothesis: (a) there was a significant association between BAZ and low-grade inflammation (as measured by CRP) at baseline; (b) there was a significant correlation between inflammation and hepcidin, independent of iron status; and (c) baseline hepcidin and BAZ were significant predictors of endpoint TfR in children receiving iron supplements with a trend towards a hepcidin × BAZ interaction: Higher baseline hepcidin was associated with higher TfR (poorer iron status) at endpoint only in the OW group.

In the children receiving placebo, neither BAZ nor hepcidin was a significant predictor of endpoint TfR. Thus, increased circulating hepcidin concentrations in OW children may have caused reduced iron absorption from iron supplements during the study period, whereas in children that received placebo the negative effect of higher BAZ on endpoint TfR might be explained by other factors than reduced absorption, such as increased requirements or losses. However, the main limitation of the current study is that the hepcidin analysis was performed only at baseline and in a sub-sample, which may have resulted in a lack of statistical power to detect further significant associations.

In summary, our findings suggest that the current surge in OW in transition countries may impair efforts to control ID in vulnerable target groups. Moreover, both ID19 and obesity17, 18 have been associated with decreased exercise capacity and impaired cognitive function. Thus, interactions of the ‘double burden’ of malnutrition during the nutrition transition may have adverse consequences.


  1. 1.

    WHO. 10 Facts on obesity. World Health Organisation: Geneva, 2010.

  2. 2.

    , , , , . Nutritional status of affluent Indian school children: what and how much do we know? Indian Pediatr 2007; 44: 204–213.

  3. 3.

    , , , , , et al. Comparison of the efficacy of wheat-based snacks fortified with ferrous sulfate, electrolytic iron, or hydrogen-reduced elemental iron: randomized, double-blind, controlled trial in Thai women. Am J Clin Nutr 2005; 82: 1276–1282.

  4. 4.

    , , . Overweight children have higher circulating hepcidin concentrations and lower iron status but have dietary iron intakes and bioavailability comparable to normal weight children. Int J Obesity 2009; 33: 1111–1117.

  5. 5.

    , , , . Iron deficiency in early childhood in the United States: risk factors and racial/ethnic disparities. Pediatrics 2007; 120: 568–575.

  6. 6.

    , , , , . Overweight children and adolescents: a risk group for iron deficiency. Pediatrics 2004; 114: 104–108.

  7. 7.

    , , , , , . Greater prevalence of iron deficiency in overweight and obese children and adolescents. Int J Obes 2003; 27: 416–418.

  8. 8.

    , , , , , et al. Sharply higher rates of iron deficiency in obese Mexican women and children are predicted by obesity-related inflammation rather than by differences in dietary iron intake. Am J Clin Nutr 2011; 93: 975–983.

  9. 9.

    , . Hepcidin, the iron watcher. Biochimie 2009; 91: 1223–1228.

  10. 10.

    , , , , , et al. Adiposity in women and children from transition countries predicts decreased iron absorption, iron deficiency and a reduced response to iron fortification. Int J Obesity 2008; 32: 1098–1104.

  11. 11.

    , , , , . Excess adiposity, inflammation, and iron-deficiency in female adolescents. J Am Diet Assoc 2009; 109: 297–302.

  12. 12.

    , , , , , et al. Decreased serum hepcidin and improved functional iron status 6 months after restrictive bariatric surgery. Obesity 2010; 18: 2010–2016.

  13. 13.

    , , . Does obesity increase risk for iron deficiency? A review of the literature and the potential mechanisms. Int J Vitam Nutr Res 2011; 80: 263–270.

  14. 14.

    , , , , , et al. Effect of body mass index reduction on serum hepcidin levels and iron status in obese children. Int J Obes 2010; 34: 1772–1774.

  15. 15.

    , , . Where does the black population of South Africa stand on the nutrition transition? Public Health Nutr 2002; 5: 157–162.

  16. 16.

    , , . Secular trends in the prevalence of stunting, overweight and obesity among South African children (1994-2004). Eur J Clin Nutr 2011; 65: 835–840.

  17. 17.

    , , , . Overweight is associated with decreased cognitive functioning among school-age children and adolescents. Obesity 2008; 16: 1809–1815.

  18. 18.

    , , , , . Lower cognitive function in the presence of obesity and hypertension: the Framingham heart study. Int J Obesity 2003; 27: 260–268.

  19. 19.

    , . An overview of evidence for a causal relation between iron deficiency during development and deficits in cognitive or behavioral function. Am J Clin Nutr 2007; 85: 931–945.

  20. 20.

    , , . The quantitative assessment of body iron. Blood 2003; 101: 3359–3364.

  21. 21.

    , , , , , et al. Soluble transferrin receptor and zinc protoporphyrin—competitors or efficient partners? Eur J Haematol 2005; 75: 309–317.

  22. 22.

    . Methods to assess iron and iodine status. Br J Nutr 2008; 99 (Suppl 3): S2–S9.

  23. 23.

    World Health Organisation, United Nations Children's Fund United Nations University. Iron deficiency anaemia. Assessment, prevention and control. World Health Organisation: Geneva, Switzerland, 2001.

  24. 24.

    , , , , , et al. Advances in quantitative hepcidin measurements by time-of-flight mass spectrometry. Plos One 2008; 3: 7.

  25. 25.

    , , , , , et al. Immunochemical and mass-spectrometry-based serum hepcidin assays for iron metabolism disorders. Clin Chem 2010; 56: 1570–1579.

  26. 26.

    , , , , . Anthropometric status. In: Labadarios D, (ed). Consumption Survey, Fortification Baseline. Stellenbosch: South Africa, 2005, pp 121–160.

  27. 27.

    , , . Relation of Body Size and Composition to Clinical Biochemical and Hematologic Indexes in United-States Men and Women. Am J Clin Nutr 1989; 50: 1276–1281.

  28. 28.

    , , . Hypoferraemia in obese adolescents. Lancet 1962; 2: 327–328.

  29. 29.

    , , , , , et al. Inflammation and iron deficiency in the hypoferremia of obesity. Int J Obes 2007; 31: 1412–1419.

  30. 30.

    , . Nutritional iron deficiency. Lancet 2007; 370: 511–520.

  31. 31.

    , , , , , . Effect of hemochromatosis genotype and lifestyle factors on iron and red cell indices in a community population. Clin Chem 2001; 47: 202–208.

  32. 32.

    , , , , . Relative importance of female-specific and non-female-specific effects on variation in iron stores between women. Br J Haematol 2003; 120: 860–866.

  33. 33.

    , , . Iron metabolism in genetically obese (ob/ob) mice. J Nutr 1988; 118: 46–51.

  34. 34.

    Anonymous. Iron. In: Dietary Reference Intakes for Vitamin A, Vitamin K, Arsenic, Boron, Chromium, Copper, Iodine, Iron, Manganese, Molybdenum, Nickel, Silicon, Vanadium, and Zinc. National Academy Press: Washington, DC, 2001. pp 290–293.

  35. 35.

    , , , . Leptin increases the expression of the iron regulatory hormone hepcidin in HuH7 human hepatoma cells. J Nutr 2007; 137: 2366–2370.

  36. 36.

    , , , , , . STAT3 mediates hepatic hepcidin expression and its inflammatory stimulation. Blood 2007; 109: 353–358.

  37. 37.

    , . Interleukin-6 induces hepcidin expression through STAT3. Blood 2006; 108: 3204–3209.

  38. 38.

    , , , , , et al. Elevated systemic hepcidin and iron depletion in obese premenopausal females. Obesity 2009; 18: 1449–1456.

  39. 39.

    , , , , , et al. Hepcidin in obese children as a potential mediator of the association between obesity and iron deficiency. J Clin Endocrinol Metab 2009; 94: 5102–5107.

Download references


We thank all the fieldworkers, teachers and principals of the schools for their support of the study, the children and parents for their participation in the trial, MRC and NWU colleagues and students, especially J Greeff, for their assistance during field and laboratory work, Anja Fleisch for doing preliminary data analyses and interpretation within her Bachelor’s thesis, and PL Jooste, and JC Jerling for acting as the Safety Monitoring Board. We thank Paul Lohmann GmbH (Lomapharm, Emmertal, Germany) for supplying the iron tablets used in the trial, and ET Wiegerinck (Hepcidinanalysis.com, Nijmegen, The Netherlands) for technical assistance. Financial support for the study was provided by Unilever Research and Development (Vlaardingen, The Netherlands) and The Medicor Foundation (Vaduz, Principality of Liechtenstein).

Author contributions

MZ, CMS, JB and LM designed and conducted the study; JB and IA analyzed the data and performed the statistical analyses; HT conducted the hepcidin analyses; JB, IA and MZ wrote the first draft of the manuscript; and all authors read and edited the manuscript. All authors had full access to all the data and JB had the responsibility for the final content of the paper and the decision to submit for publication.

The trial is registered at clinicaltrials.gov as NCT01092377.

Author information


  1. Centre of Exellence for Nutrition (CEN), North-West University, Potchefstroom, South Africa

    • J Baumgartner
    • , C M Smuts
    •  & L Malan
  2. Laboratory of Human Nutrition, Institute of Food, Nutrition and Health, ETH Zürich, Switzerland

    • J Baumgartner
    • , I Aeberli
    •  & M B Zimmermann
  3. Laboratory of Genetic, Endocrine and Metabolic Diseases, Department of Laboratory Medicine, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands

    • H Tjalsma


  1. Search for J Baumgartner in:

  2. Search for C M Smuts in:

  3. Search for I Aeberli in:

  4. Search for L Malan in:

  5. Search for H Tjalsma in:

  6. Search for M B Zimmermann in:

Competing interests

The authors declare no conflict of interest.

Corresponding author

Correspondence to J Baumgartner.

About this article

Publication history







Further reading