Overweight is increasing in transition countries, while iron deficiency remains common. In industrialized countries, greater adiposity increases risk of iron deficiency. Higher hepcidin levels in obesity may reduce dietary iron absorption. Therefore, we investigated the association between body mass index (BMI) and iron absorption, iron status and the response to iron fortification in populations from three transition countries (Thailand, Morocco and India).
In Thai women (n=92), we examined the relationship between BMI and iron absorption from a reference meal containing ∼4 mg of isotopically labeled fortification iron. We analyzed data from baseline (n=1688) and intervention (n=727) studies in children in Morocco and India to look for associations between BMI Z-scores and baseline hemoglobin, serum ferritin and transferrin receptor, whole blood zinc protoporphyrin and body iron stores, and changes in these measures after provision of iron.
In the Thai women, 20% were iron deficient and 22% were overweight. Independent of iron status, a higher BMI Z-score was associated with decreased iron absorption (P=0.030). In the Indian and Moroccan children, 42% were iron deficient and 6.3% were overweight. A higher BMI Z-score predicted poorer iron status at baseline (P<0.001) and less improvement in iron status during the interventions (P<0.001).
Adiposity in young women predicts lower iron absorption, and pediatric adiposity predicts iron deficiency and a reduced response to iron fortification. These data suggest the current surge in overweight in transition countries may impair efforts to control iron deficiency in these target groups. Interactions of the ‘double burden’ of malnutrition during the nutrition transition may have adverse consequences.
Transition countries are undergoing rapid dietary and lifestyle changes that produce a ‘double burden’ of malnutrition and overweight.1, 2 Three transition countries in which this is occurring are Thailand, Morocco and India.3, 4, 5 In Bangkok, Thailand, it is estimated nearly one-third of women are overweight and 24% are anemic.6, 7 In Morocco, 24% of children less than 5 years are stunted and 30% of school-age children are anemic,8 at the same time, 9% of school-age children are overweight.9, 10 Similarly, in middle class Indian school children, anemia is present in 19–88%, while overweight affects 9–29%.11 In many lower income countries, the prevalence of overweight is increasing at 2–4 times the rate of the industrialized world.12
Studies in industrialized countries have consistently found higher rates of iron deficiency in overweight children13, 14, 15, 16, 17, 18 and adults.19, 20, 21, 22, 23 Although the mechanism is unclear, this may be due to lower iron intakes and/or increased iron requirements in overweight individuals.14, 23 In addition, the chronic inflammation and increased leptin production characteristic of obesity increase hepcidin secretion from the liver,24 which, along with hepcidin produced by adipose tissue,25 could reduce dietary iron absorption.26
The major adverse effects of iron deficiency are impaired cognitive development in children27 and poorer pregnancy outcome in women.28 A potential interaction between the ‘double burden’ of adiposity and iron deficiency has not been examined in transition countries. Thus, the aim of this study was to investigate the association between weight status, iron deficiency, iron absorption and the response to iron fortification in children and women from transition countries.
Subjects and methods
Iron absorption studies in women
A total of 92 apparently healthy premenopausal Thai women (18–50 years of age) with a maximum body weight of 70 kg were recruited at Mahidol University in Bangkok. They all received the same iron isotope-labeled reference meal. Sixty-seven participated in studies of iron-fortified fish sauce29 and the effect of spices on iron absorption,30 and 25 were healthy controls in a study of the effects of hemoglobinopathies on iron absorption (M Zimmermann, unpublished data). The first two studies have been reported previously.29, 30 Exclusion criteria included pregnancy, lactation, gastrointestinal disorders or metabolic diseases. The absence of thalassemia was verified by hemoglobin typing using a commercial HPLC-based autoanalyzer (Variant, Bio-Rad, Hercules, CA, USA). The study protocols were approved by the Ethical Committees of Mahidol University, Thailand, and ETH Zurich, Switzerland. Written informed consent was obtained from the participants.
For all subjects, the test meals were identical reference meals of steamed white rice (50 g dry weight) served with a soup prepared from local vegetables (50 g white cabbage, 50 g Chinese cabbage, 30 g Thai mushrooms and 20 g steamed carrots). Vegetables and rice were purchased in bulk and individual portions were prepared using standardized procedures. Each meal contained ∼4 mg of isotopically labeled fortification iron, as [57Fe/58Fe]-ferrous sulfate. All test meals were consumed in the morning between 0700 and 0900 after an overnight fast. Body weight was measured using a digital scale to the nearest 100 g and height was measured using a rigid stadiometer to the nearest 0.5 cm.31 Before the meal, a venous blood sample was drawn to determine hemoglobin (Hb), serum ferritin (SF) and C-reactive protein (CRP); CRP was measured in 59 out of the 92 subjects. Hb was measured using an ADVIA 120 Hematology System (Bayer AG, Leverkusen, Germany). Anemia was defined as Hb concentration <12 g l−1.32 SF and CRP were analyzed by chemiluminescent immunometric assay (Immulite/Ferritin and Immulite/High Sensitivity CRP, Bühlmann Laboratories AG, Allschwil, Switzerland) using low- and high-level quality control materials supplied by the company. Normal ranges: for SF, 5–300 μg l−1; for CRP; <5 mg l−1 in absence of infection. A second blood sample was drawn 14 days later. Blood samples were analyzed for isotopic composition by multicollector negative thermal ionization mass spectrometry to calculate fractional iron absorption.33
Baseline and intervention studies in children
The four pediatric studies analyzed in this paper were baseline surveys followed by controlled efficacy trials of iron fortification, and have been reported elsewhere.34, 35, 36 From the baseline surveys, we included all children in this analysis; from the intervention trials, because we wanted to examine the effect of weight status on response to the iron treatment, we included only children from the iron treatment groups. Body weight was measured using a digital scale to the nearest 100 g and height was measured using a rigid stadiometer or a nonstretchable pull-down measuring tape to the nearest 0.5 cm.31 The Moroccan studies34, 35 were done in semi-rural villages in the northern Rif mountains. The first study tested the efficacy of a salt containing 1 mg Fe per gram (as encapsulated ferrous sulfate) provided to households for 9 months, in 6- to 15-year-old children.34 The second study tested the efficacy of a salt containing 2 mg Fe per gram (as micronized ferric pyrophosphate) provided for 9 months, in 6- to 15-year-old children.35 The two Indian studies were done in Bangalore, Karnataka State.36 In the first, urban, 6- to 13-year-old children were given a rice-based lunch meal fortified with 20 mg Fe as ferric pyrophosphate for 7 months.36 The second study was in a semi-rural area south of Bangalore, where 5- to 16-year-old children were given fortified salt containing 2 mg Fe per gram salt as ferric pyrophosphate or encapsulated ferrous fumarate for 9 months (Personal communication; Maria Andersson, 2008). The study protocols were approved by the respective ethical committees in Morocco and India, as well as at the ETH Zurich, Switzerland. Written and/or oral informed consent was obtained from the parents of the children.
Whole blood was collected by venipuncture and transported on ice to the laboratory for determination of Hb, SF, whole blood zinc protoporphyrin (ZnPP), serum transferrin receptor (TfR). Hb was measured in whole blood using automated coulter counters: in India with a Cell-Dyn 1700 (Abbot Labs, Abbott Park, IL, USA), in Morocco with a AcT8 Counter (Beckman Coulter, Krefeld, Germany), with 3 level control materials provided by the company. Anemia was defined as a Hb <12 gl−1 in children aged >12 years and <11.5 gl−1 in children aged 5–11 years.32 ZnPP was measured on washed red blood cells by using hematofluorometry (Aviv Biomedical, Lakewood, NJ, USA) with control materials provided by the company. TfR was measured using an enzyme-linked immunosorbent assay (Ramco, Houston, TX, USA). SF was measured using an automated chemiluminescent immunoassay system (Immulite, Diagnostic Products Corporation) or an enzyme-linked immunosorbent assay (Ramco). Normal reference values: ZnPP, <40 μmol/mol heme;37 SF >15 μg l−132 and TfR <8.2 mg l−1.38 Iron deficiency was defined as SF <15 μg l−1 or TfR >8.2 mg l−1 and ZnPP >40 μmol/mole heme. Body iron stores (BFe) were calculated from the TfR:SF ratio.39
Statistical analyses were performed by using Microsoft EXCEL (2002; Microsoft Corporation, Redmond, WA, USA) and SPSS for Windows statistical software (version 13.0; SPSS Inc., Chicago, IL, USA). Data from the absorption studies, baseline studies and intervention trials were separately pooled. Three subjects in the iron absorption studies were excluded as outliers because their fractional iron absorptions were >3.0 s.d. above the geometric mean. To account for differences in iron status and their effect on dietary iron absorption, iron absorption in each subject was corrected to a value corresponding to a SF of 40 μg l−1.40 The body mass index (BMI; in kg/m−2) of the adults and children was calculated. For the Thai women, a BMI of >23 was used to classify subjects as overweight.6, 41 For the children, age and sex-specific criteria from the World Health Organization42 were used to calculate BMI Z-scores and classify children as overweight (Z-score >1) or obese (Z-score >2). BMI Z-scores were calculated as the difference between the obtained score and the age and sex-specific mean, divided by the s.d. Data were expressed as means±s.d., medians (25th percentile, 75th percentile) and/or ranges. CRP was not normally distributed and values were log transformed for comparisons. To analyze for associations, Pearson's correlations and multivariate regressions were done.
Iron absorption studies
Table 1 shows the characteristics of the women (n=92) who participated in the iron absorption studies. The age range was 19–50 years. Four percent were anemic, 20% were iron deficient and 22% were overweight. After correcting for differences in iron status among subjects using SF,42 in a multivariate regression including age, Hb, CRP and BMI, fractional iron absorption was negatively correlated with CRP (standardized β=−0.422; P<0.001) and BMI (standardized β=−0.106; P<0.05). The relationship between BMI and CRP is shown in Figure 1, and the relationship between BMI and fractional iron absorption is shown in Figure 2.
Baseline and iron fortification intervention studies
The characteristics of the children in the baseline studies are shown in Table 2. Of the total number (n=1688), 991 were Moroccan and 697 were Indian. Mean weight and height were 25.3±7.8 kg and 128.0±13.5 cm, respectively. The prevalence of overweight, anemia and iron deficiency was 6, 33 and 42%, respectively. The number of subjects with a BMI Z-score <1 and <2 were 623 and 137, respectively. Table 3 shows the characteristics of children in the interventions and the mean changes in Hb and iron status during the interventions. Of the total number of children (n=727), 384 were Moroccan and 343 were Indian. The mean weight and height were 27.5±8.6 (kg) and 131.5±14.4 cm, respectively. The number of subjects with a BMI Z-score <1 and <2 were 306 and 86, respectively.
Figure 3 shows the inverse relationship between BMI Z-score and BFe in the baseline data. Table 4 shows the significant associations between BMI Z-score and baseline SF, TfR, ZPP and BFe, from the regression models including age, gender, study, Z-score for BMI. BMI Z-score was a highly significant negative predictor of BFe (P<0.0001) and a positive predictor of TfR and ZPP (P<0.0001); all three indicate greater BMI was associated with poorer iron status. BMI Z-score also positively predicted SF (P=0.021). This association was weaker than those between BMI and the other three indicators of iron status, and may reflect the effect of adipose-related inflammation on SF, an acute phase protein (see Discussion). In these baseline regressions, along with BMI Z-score, significant predictors (standardized β-coefficient, P-value) of SF were age (−0.151, <0.0001) and female gender (−0.073, 0.003); predictors of TfR were age (−0.055, 0.018) and female gender (0.067, 0.004); predictors of ZPP were age (−0.072, 0.003) and female gender (0.051, 0.034); a predictor of BFe was age (−0.071, 0.004). Significant predictors of Hb were age (standardized β=0.211; P<0.0001), but not gender, study or BMI Z-score.
Figure 4 shows the inverse relationship between BMI Z-score and change in BFe in the iron fortification intervention studies data. Table 5 shows the significant associations between BMI Z-score and change in SF, TfR, ZPP and BFe during the intervention, from the regression models including age, gender, study, Z-score for BMI and baseline value for the dependent variable. BMI Z-score was a significant negative predictor of change in SF and BFe, that is, higher BMI Z-score was associated with less of an increase in these variables during the intervention, and a significant positive predictor of change in TfR and ZPP, that is, higher BMI Z-score was associated with less of a decrease in these variables during the intervention (P<0.0001 for all). In these regressions, along with BMI Z-score, significant predictors (standardized β-coefficient, P-value) of SF were age (−0.112, <0.002) and baseline SF (−0.290, <0.0001); a predictor of TfR was baseline TfR (−0.427, <0.0001); predictors of ZPP were age (0.760, 0.026) and baseline ZPP (−0.386, <0.00001); predictors of BFe were age (−0.103, 0.002) and baseline BFe (−0.455, <0.0001). For all four indicators of iron status, baseline value was the strongest predictor, indicating children with more severe iron deficiency show a greater response to interventions. Significant predictors of change in Hb were Hb at baseline (standardized β=−0.330; P<0.0001) and age (standardized β=0.128; P=0.001), but not gender, study or BMI Z-score.
In the baseline pediatric studies, higher BMI Z-score was a significant predictor of poorer iron status. This is consistent with previous studies in children in industrialized countries13, 14, 15, 16, 17, 18 reporting an inverse relationship between iron status and adiposity, For example, in the NHANES III sample of 9698 older US children, where 14% were at risk for overweight, 10% were overweight and 3% were iron deficient, those at risk for overweight and who were overweight were twice as likely to be iron deficient compared to those not overweight.17 Our data demonstrate this same relationship exists in children in transition countries with higher rates of iron deficiency and lower rates of adiposity. A limitation in previous studies was the use of iron status indicators (for example, SF and serum transferrin) that are acute-phase proteins28 and may be confounded by the adipose-related inflammation. We used multiple indicators of iron status, including TfR and ZPP, two measures less likely to be confounded by inflammation.37, 38
Several factors may explain why greater adiposity increases risk for iron deficiency. Overweight may be associated with poor quality or restricted diets low in iron, but when dietary iron intakes in overweight adults23 and children43 are estimated, they are not lower than in normal weight individuals. However, even if diets of overweight individuals are not lower in iron, the absorption of the iron may be reduced because increased circulating hepcidin in obesity may reduce iron absorption (see Discussion). Iron requirements in overweight individuals may be increased due to larger blood volume and higher basal iron losses with higher body weight,44 but this has not been directly measured. In addition, overweight girls tend to mature and begin their menses at an earlier age, increasing their iron requirements.45
Our findings in the Thai women demonstrate for the first time that greater adiposity is associated with lower fractional iron absorption in humans, independent of iron status (Figure 2). Hepcidin expression is increased in obesity25 and other chronic inflammatory states.46 Increased circulating hepcidin may reduce both gastrointestinal iron absorption26 and iron release from the reticuloendothelial system.47 Although we did not measure hepcidin concentrations, measurements of CRP indicated inflammation was clearly increased with greater adiposity (Figure 1). Both liver and adipose tissue produce hepcidin, and while liver hepcidin expression is positively associated with increased BFe, adipose tissue hepcidin expression is positively correlated with BMI and may be negatively associated with transferrin saturation.25 In addition, circulating leptin is higher in overweight subjects,48 and leptin increases hepcidin expression by liver cells in vitro via JAK/STAT signaling.24 Similarly, lipocalin-2 is an iron-binding protein that is upregulated by inflammation and may help sequester iron during infections.49 It is produced by adipose tissue and its expression is increased in db/db (leptin receptor-deficient) obese mice.50 Thus, low iron status in overweight individuals could result from a combination of nutritional (reduced absorption) and functional (increased sequestration) iron deficiency. However, this hypothesis is not supported by studies in genetically obese (ob/ob) mice.51, 52 When provided with an iron-sufficient diet, obese mice absorb twice as much iron as lean mice but have lower iron levels in the liver and small intestine; these studies predate the discovery of hepcidin and there is no data reported on hepcidin concentrations in these ob/ob mice.51, 52
In normal-weight individuals, SF concentrations are appropriately decreased and directly related to transferrin saturation if BFe are depleted.28 In contrast, in obese subjects, SF tends to be higher than that in normal weight individuals and inversely related to transferrin saturation.20, 21, 53 This suggests SF, an acute phase protein that can be elevated in inflammatory conditions even in the presence of iron deficiency, may be increased by adipose-mediated inflammation. Our data support this. In the Thai women, CRP was positively correlated with both BMI (P<0.0001; Figure 1) and SF (P=0.043), and in the pediatric studies, while a greater BMI Z-score was strongly correlated with higher TfR and ZPP concentrations and lower BFe (P<0.001, all indicating poorer iron status), it was also a modest predictor of a higher SF concentration (P=0.021; Table 4).
Our study has several limitations. It was cross-sectional in design, so no firm conclusions regarding cause and effect can be made. Although it is plausible that obesity decreases iron absorption by the mechanisms discussed above, it is also possible, but less likely, that iron deficiency could contribute to the development of obesity. Second, a number of variables that influence both iron and body weight status, including socioeconomic factors, physical activity and physical maturity were not included in our analyses and could be residual confounders. However, the strength of our associations makes this less likely. Third, we did not use a high-sensitivity CRP assay in the pediatric studies, or have measures of hepcidin or leptin in the adult studies, all of which may have provided important data on potential mechanisms. The strengths of the study include the large sample sizes, the varied ethnicity—Asian, Arabic, Indian—of the subjects, the use of BFe to define iron status in the children and the direct measurement of iron absorption from a reference meal using isotopes.
This is the first study to demonstrate the negative impact of adiposity on iron absorption and response to iron fortification. It suggests the rapid increase in overweight in transition countries could impair their efforts to control iron deficiency in women and children. These findings need to be confirmed in other populations and settings, but imply interactions of the ‘double burden’ of malnutrition in women and children during the nutrition transition may have adverse consequences.
We thank the participating women and children, as well as the local health care staff in Thailand, Morocco and India for their assistance on the project. We thank M Andersson (Zürich, Switzerland), D Moretti (Vlaardingen, The Netherlands), T Walcyzk (Singapore), P Thankachan and A Kurpad (Bangalore, India), S Tuntipopipat (Bangkok, Thailand), and A Saad (Rabat, Morocco). MZ, CZ, SM, PW, NC did the field work, MZ wrote the first draft of the manuscript, all authors contributed to the writing and editing of the manuscript. None of the authors has a conflict of interest relating to this paper. This study was supported by the Thrasher Foundation (Salt Lake City, USA) the Micronutrient Initiative (Ottawa, Canada); the Nestlé Foundation (Lausanne, Switzerland); the Foundation for Micronutrients in Medicine (Rapperswil, Switzerland) and the Swiss Federal Institute of Technology (Zürich, Switzerland).