Differential ferritin interpretation methods that adjust for inflammation yield discrepant iron deficiency prevalence

Transcription

1 bs_bs_banner Original Article Differential ferritin interpretation methods that adjust for inflammation yield discrepant iron deficiency prevalence DOI: /mcn Elsmari Nel*, Herculina S. Kruger*, Jeannine Baumgartner*, Mieke Faber and Cornelius M. Smuts* *Centre of Excellence for Nutrition, North-West University, Potchefstroom, South Africa, and Non-communicable Diseases Research Unit, Medical Research Council, Tygerberg, South Africa Abstract We reassessed the iron deficiency (ID) prevalence in a South African trial that formed part of the International Research on Infant Supplementation study by comparing four methods that account for the high prevalence of acute (28.6%) and chronic (41.8%) inflammation observed in the study. Serum ferritin (SF) was measured as marker of iron status in 192 apparently healthy, 4 13-month-old infants. Alpha-1 glycoprotein and C-reactive protein concentrations were determined to indicate chronic and acute inflammation, respectively. The ID prevalence was obtained by four methods that adjust for inflammation: (1) excluding infants with inflammation; (2) using a higher cut-off (SF < 30 μgl 1 ); (3) using different cut-offs for infants with vs. without inflammation (SF < 30 μg L 1 vs. SF < 12 μgl 1 ); and (4) adjusting SF concentrations with correction factors (CFs) were compared with a reference method (SF < 12 μgl 1 ) not accounting for inflammation. Using the higher SF cut-off method resulted in the highest ID prevalence (52.1%), followed by using two different cut-offs (31.8%), using CFs (21.9%) and excluding subjects with inflammation (17.6%).The CF method showed the best agreement with the reference method. Disregarding inflammation resulted in a significantly lower ID prevalence (17.2%). ID anaemia (IDA) prevalence ranged from 13.2% to 24.5%, with the lowest prevalence (12.0%) for the reference method. Our analysis highlights the challenge of assessing ID and IDA using only SF as marker of iron status in the presence of inflammation. We demonstrate the importance of measuring inflammation markers to account for their elevating effect on SF. Keywords: acute phase proteins, serum ferritin, anaemia, infants, infection, IRIS study, South Africa. Correspondence: Professor Herculina Salome Kruger, Centre of Excellence for Nutrition, North West University, Potchefstroom Campus, 11 Hoffman Street, Room 158, Building G16, 2520 Potchefstroom, South Africa. salome.kruger@nwu.ac.za Introduction Iron deficiency (ID) remains an important contributor to the prevalence of anaemia worldwide. The pooled International Research on Infant Supplementation (IRIS) study (Smuts et al. 2005a), which was conducted in four developing countries, reported the highest ID prevalence for Peru (38.6%) and Indonesia (34.5%), while in South Africa and Vietnam the ID prevalence was 18.3% and 12.4%, respectively. Iron deficiency anaemia (IDA) is the most severe form of ID and the most common micronutrient deficiency in infants aged 6 24 months (Stoltzfus & Dreyfuss 1998) with an estimated prevalence ranging between 30% and 45% in sub-saharan Africa (Lutter 2008). IDA has been implicated as a cause of stunting, developmental delay, reduced cognition and impaired immunity (WHO/UNICEF/UNU 2001). It is however important to confirm ID as the cause of anaemia This article forms part of a supplement sponsored by Sight and Life. [This statement was added after original online publication] 221

2 222 E. Nel et al. before initiating iron supplementation because ironreplete infants do not have an effective excretion pathway for the oversupply of iron, besides through feces (Chaparro 2008). Excess iron may have similar effects to IDA, which may affect linear growth, head circumference and weight gain. Free iron exerts a prooxidant in the body, affecting the genes that regulate growth (Dewey et al. 2002; Kelleher 2006). Furthermore, bacteria and parasites rely on free iron for multiplication and growth. Zimmermann et al. (2010) found that the poorly absorbable iron used for fortification favoured the growth of pathogenic gut microbiota in anaemic African children. The ratio of fecal enterobacteria to bifidobacteria and lactobacilli changed and the iron became available for the growth of pathogenic bacteria in the colon (Zimmermann et al. 2010). Many beneficial barrier bacteria, such as Lactobacilli, on the other hand, do not require iron for growth (Archibald 1983) and their growth was not enhanced by free iron (Zimmermann et al. 2010). Although serum ferritin (SF) is the recommended measure to describe the prevalence of ID if only one indicator of iron status can be measured (WHO 2011a), it is well known that inflammation has an elevating effect on SF concentrations (Finch 1994; Skinner et al. 2010; Righetti et al. 2013). Therefore, SF measurements are usually coupled with the analysis of acute phase proteins (APPs) to detect whether inflammation is present, and to adjust for its elevating effect on SF when present (WHO 2011a). In areas without safe water and sanitation, chronic inflammation may be caused by intestinal parasites, gastrointestinal, respiratory or other infections. Alternatively, being overweight or obese even in children as young as 3 years of age may contribute to inflammation by secreting a variety of pro-inflammatory cytokines (Cook et al. 2000; Visser et al. 2001; Tam et al. 2010). Skinner et al. (2010) showed that even very obese 1 2-year-olds already had elevated inflammatory markers. We revisited the baseline ID prevalence (based on SF) of the South African leg of the IRIS multi-centre study reported by Smuts et al. (2005b).The IRIS study was done in a rural site in the KwaZulu-Natal province of South Africa and only considered the influence of acute inflammation on SF. Recent literature, however, strongly demonstrates that the prevalence of chronic inflammation cannot be ignored, especially in an area like KwaZulu-Natal where the prevalence of HIV/AIDS is widespread (Department of Health 2002). We therefore recalculated the ID prevalence using four different methods that account for the effect of both acute and chronic inflammation on SF concentrations (Thurnham et al. 2010; WHO 2011a). These prevalence values were compared with a reference method that did not take inflammation into account (WHO 2011a). We postulated that the derived ID prevalence would vary depending on the method chosen, and that it would be higher when accounting for both acute and chronic inflammation. Methods Study population and participants A secondary, cross-sectional data analysis was performed on the baseline data of the South African trial Key messages Interpreting serum ferritin concentration as marker for iron status is challenging in populations with high prevalence of acute and/or chronic inflammation. Prevalence of iron deficiency varies depending on the method that was used to account for the elevating effect of acute and chronic inflammation on serum ferritin concentration. Prevalence of iron deficiency was highest (52.1%) when using a higher serum ferritin cut-off value (30 μgl 1 ); followed by using two different cut-offs (31.8%), i.e. for subjects with inflammation (30 μgl 1 ) vs. no inflammation (12 μgl 1 ); using correction factors according to the stage of inflammation (21.9%); and excluding subjects with inflammation (17.6%). Prevalence of iron deficiency was lowest (17.2%) when not accounting for inflammation.

3 Differential ferritin interpretation methods 223 of the IRIS study. This double-blind, placebocontrolled intervention examined the efficacy of multi-micronutrient supplementation on infants from a rural South African population (Smuts et al. 2005b). This study was conducted in the Valley of a Thousand Hills, a rural area situated 40 km northwest of Durban, KwaZulu-Natal, South Africa. Around predominantly Zulu-speaking people live scattered over the mountainous area which is somewhat better off in terms of welfare than many other rural areas in South Africa (Smuts et al. 2005b). The study sample consisted of 192 infants, aged 4 13 months, who were randomly selected to take part in the IRIS study, which lasted 6 months (April to September 2000), until they were around months old. Exclusion criteria included premature birth or low birthweight (<2.5 kg), congenital defects, chronic illness, severe wasting, fever (>39 C) or severe anaemia (Hb < 8gL 1 ).The Ethics Committee of the South African Medical Research Council approved the protocol. The local community leaders gave permission to do the study in the area. The caregivers of all participating infants gave written informed consent (Smuts et al. 2005b). Anthropometric measurements and analysis The anthropometric measurements taken in the original study were described in detail (Smuts et al. 2005b). The IRIS researchers (Smuts et al. 2005b) used the United States National Centre for Health Statistics median as reference to yield three measures of nutritional status. We recalculated the nutritional status of the infants by using the WHO (2006) growth standards to express their length-for-age (HAZ) and weight-for-length (WHZ) z-scores. Stunting, or chronic under nutrition, was defined as a HAZ below 2 standard deviations (SD) of the reference median. Overweight was defined as a WHZ above +2 SD, and obesity as a z-score for the same indicator above +3 SD of the median of the reference population (De Onis et al. 2006). Methods used to reassess ID and IDA prevalence The method by which blood was collected and analysed in the laboratory was described previously (Smuts et al. 2005b). C-reactive protein (CRP) and alpha-1 glycoprotein (AGP) were measured by means of a sandwich enzyme-linked immunosorbent assay (ELISA), and SF was assessed with an ELISA test (Smuts et al. 2005b). An Hb cut-off of < 110gL 1 identified anaemia (WHO 2011b). CRP and AGP measured the effects of short-term and long-term inflammatory status, respectively. The IRIS researchers excluded all cases with elevated CRP (>12 mg L 1 ) at baseline from the statistical analysis for SF (Smuts et al. 2005b). We however defined elevated APP concentrations as CRP > 5mgL 1 or AGP > 1gL 1 (Thurnham et al. 2010). We opted to apply the commonly used cut-off values for the APPs to facilitate comparison with other studies, and to add to the limited understanding of the relationship between markers of inflammation, and SF as an indicator of iron status. We then interpreted SF concentrations by applying four methods that adjust for the effect of inflammation on SF and compared the results to the ID prevalence obtained from a reference method not taking inflammation into account. Our reassessment was based upon the WHO working group s recommendations (WHO 2011a) together with the recently published approach of Thurnham et al. (2010). In the first method, we raised the SF cut-off concentration that defines ID to 30 μgl 1 due to widespread inflammation in our sample (WHO 2011a). The second method excluded the SF values of all individuals with elevated concentrations of CRP and/or AGP from the determination of ID prevalence (WHO 2011a). Thirdly, we proposed a different interpretation of thewho working group s recommendation to apply the higher cut-off of 30 μgl 1 only to individuals with elevated CRP and/or elevated AGP, and the normal cut-off of 12 μgl 1 to individuals without inflammation. Our fourth method adjusted individuals SF concentrations by means of correction factors (CFs) specific to each subject s inflammatory status (Thurnham et al. 2010). For this method, individuals were categorised into four groups based on their CRP and/or AGP concentrations: (1) an apparently healthy reference group (CRP 5mgL 1 and AGP 1gL 1 ); (2) an incubation group (CRP > 5mgL 1 and AGP 1gL 1 ); (3) an early convalescence group (CRP > 5mgL 1 and

4 224 E. Nel et al. AGP > 1gL 1 ); and (4) a late convalescence group (CRP 5mgL 1 and AGP > 1gL 1 ). Individual SF concentrations were then adjusted by using the relevant, group-specific CF as a multiplier and repeating the calculation to determine the prevalence of ID after correction. For the incubation group a CF of 0.77 was used, and for the early and late convalescence groups CFs of 0.53 and 0.75 were used, respectively. Different CFs are needed to account for inflammation during the various stages of inflammation, because the increase in SF after infection follows a different pattern from that of either CRP or AGP. SF concentrations rise significantly within a few hours of the onset of inflammation, and concentrations remain high even after CRP concentrations have subsided, and while AGP concentrations remain elevated (Thurnham et al. 2010). We assessed the agreement between each of the methods that adjust for the effect of inflammation on SF, respectively, and the reference method. We did not assess the agreement with the second method (exclusion of SF values of all individuals with elevated concentrations of CRP and/or AGP), because more than 50% of our participants had to be excluded with this method, resulting in two samples with different sample size. Statistical analysis All statistical analyses were conducted using SPSS (IBM Statistical Package for the Social Sciences, version 21, Chicago, IL, USA). Significance was defined as P < Descriptive statistics were done for all variables. Continuous variables were visually examined for adherence to the normal distribution using Q_Q plots and histograms. Non-normally distributed variables were transformed for statistical analyses, or non-parametric methods were used for analyses. Two-way frequency tables described the prevalence of ID, IDA and anaemia not associated with ID for the four methods that account for inflammation, as well as for the reference method. Chi-square analysed differences in the ID prevalence when comparing the four methods that adjusted for inflammation to the reference method, while McNemar tests indicated whether the differences were significant. We assessed the kappa statistic of agreement between three of the methods that adjust for the effect of inflammation on SF, respectively, and the reference method (Posner et al. 1990). Results Characteristics of the study participants are shown in Table 1. The mean HAZ was below the WHO (2006) growth standard s population median, while the mean WHZ was above +1. The prevalence of elevated APPs indicating inflammation was 52.6%. We determined that 11.5% of the subjects with inflammation were in the incubation, 17.2% in the early convalescent, and 24% in the late convalescent stages. Figure 1 presents the differences in the prevalence of ID, IDA and proportion of anaemia not associated with ID for the five different methods used to interpret SF concentrations. Compared with the ID prevalence obtained from the reference method (17.2%), the ID prevalence obtained from all the other methods was significantly higher (P < 0.05). The ID prevalence from the exclusion method (17.6%) could not be compared with the other methods due to differences in sample size. The higher cut-off method classified the most infants as being ID (52.1%), and the reference method the least. The reference method classified 34.9% fewer infants as ID when compared with using the higher cut-off method, 14.6% less when Table 1. Characteristics of study participants Characteristic N* Distribution/prevalence Age (months) ± 2.02 Male : female ratio (%) :52.6 Length-for-age (z-score) ± 1.12 Weight-for-length (z-score) ± 1.00 CRP (mg L 1 ) (0.54; 6.29) AGP (g L 1 ) (0.75; 1.19) Inflammation (% with AGP > 1gL 1 or CRP > 5mgL 1 ) Haemoglobin (g L 1 ) ± 1.18 Anaemia (%, Hb < 110gL 1 ) Unadjusted SF (μg L 1 ) (15.1; 52.76) Values are mean ± SD for variables with a normal distribution; median (IQR) for variables not normally distributed and % for frequency. *Number of participants varies due to missing values.

5 Differential ferritin interpretation methods Fig. 1. Differences in the prevalence of ID, IDA and anaemia not associated with ID for the five methods employed to interpret SF concentrations. 1 Higher cut-off method; 2 AGP 1gL 1 and/or CRP > 5mgL 1 ; 3 Different cut-offs for SF for subjects with inflammation (30 μgl 1 ) vs.no inflammation (12 μgl 1 ). *ID prevalence differed significantly from the reference method (P < 0.005), McNemar test Ref*: unadjusted SF < 12 µg L 1 SF < 30 µg L 1 1 Excluding inflammation 2 Two different cut-offs 3 Adjusted SF with CF % ID % IDA % Anaemia not associated with ID compared with using two different cut-offs, and 4.7% less when compared with using the CF method. Accordingly, IDA prevalence varied from 12% to 24.5% for the four methods and the reference method. The contribution of IDA to anaemia varied from 29.9% to 61% for the different methods. The highest level of agreement with the reference method was found for the CF method (κ = 0.85, P < ), followed by the method using two different cut-offs (κ = 0.62, P < ), and the lowest agreement with the higher cut-off method (κ = 0.32, P < ). Discussion We reassessed the baseline ID prevalence of the South African trial of the IRIS multi-centre study using recently proposed methods to interpret SF concentrations in the presence of both acute and chronic inflammation. Findings of this study showed that the different methods used to account for the effect of inflammation on SF concentration yielded different estimates of the prevalence of ID, and the proportion of anaemia associated with ID. The methods that adjusted for inflammation, namely using a higher cutoff (52.1%), the exclusion method (17.6%), using two different cut-offs (31.8%), and the CF method (21.9%) yielded higher ID prevalence than the reference method (17.2%) that did not account for inflammation and used a SF cut-off of 12 μgl 1. We found a significantly higher ID prevalence when using the higher SF cut-off (<30 μgl 1 ) compared with the ID prevalence obtained from the other methods. This result is in agreement with the findings of Engle-Stone et al. (2013), who observed that by using the higher cut-off the prevalence of ID was underestimated in infants with inflammation and overestimated in those without. Subjects with infection can be categorised into different phases of the inflammatory process, and the extent of elevation in SF concentrations varies between these phases (Thurnham et al. 2010). It is therefore imprecise to use only one specific SF cut-off concentration as an international standard (Kung u et al. 2009), especially because the increase in SF concentration after infection follows a different pattern from that of either CRP or AGP (Thurnham et al. 2010). We further agree that the higher cut-off (<30 μgl 1 ) method should be used only to estimate the prevalence of ID if data have already been collected in a population known to have high rates of inflammation, where CRP and/or AGP were not measured (Engle-Stone et al. 2013). It is important that researchers using this method, however, keep in mind that overestimations of the ID prevalence are very likely. We furthermore postulate that this cut-off needs to be adapted, because it appears to be too high for this particular study sample and showed the lowest agreement with the reference method. When we used different SF cut-offs for infants with inflammation (<30 μgl 1 ) and without inflammation (<12 μgl 1 ), we obtained a lower ID prevalence than when employing the higher cut-off to all infants, but a higher prevalence than when excluding all infants with inflammation. This supports our assumption that

6 226 E. Nel et al. the cut-off proposed by the WHO (2011a) for subjects with inflammation may be too high for our particular study population, leading to an overestimation of the ID prevalence. Excluding subjects with inflammation, on the other hand, could bias the results, as ID individuals might be more prone to infection (Thurnham et al. 2010) which may lead to an underestimation of the ID prevalence (WHO 2011a). Furthermore, excluding subjects with inflammation could substantially reduce the sample size in populations with a high prevalence of inflammation (Engle-Stone et al. 2013). Such was the case in our study where we had to exclude more than 50% of our subjects to apply the exclusion method. By using CFs to adjust SF concentrations, we derived an ID prevalence higher than what was previously reported for the same study infants (Smuts et al. 2005b), or obtained from the reference or exclusion methods, but lower than what was found using the higher cut-off or different cut-off methods. Using this method, researchers are able to assess the iron status of populations with high rates of inflammation while retaining the use of all data (Knowles et al. 2013). Furthermore, it enables researchers to follow the recommendations of the WHO to use SF as a marker of iron status, while incorporating the effects of APPs (Thurnham et al. 2010; Grant et al. 2012). This method provides, in our opinion, a more accurate estimation of ID prevalence and showed the best agreement with the reference method. The CFs published by Thurnham et al. (2010) can be applied to other subject populations because the authors found that the elevating effect on SF concentrations was proportionate to baseline SF measurements at each stage of the infection cycle, irrespective of age or gender. The single limitation of the CF method is that it can be used only when both APPs are measured, which is not always possible in population studies subject to budget constraints. Depending on the method applied to define ID, the prevalence of IDA in our study population ranged from 12.0% to 24.5% (Fig. 1), while the prevalence of anaemia not associated with ID ranged from 15.6% to 28.1%. The causes of anaemia are often multifactorial in settings with high rates of inflammation (Zetterström 2004), such as this study sample, and may include nutrient deficiencies, malaria, helminth infestation, HIV or certain congenital haemoglobinopathies (DeMaeyer et al. 1989; Gillespie & Johnston 1998), for which we had no data. The results of Sazawal et al. (2006) suggest that only children with IDA benefited from iron supplementation in terms of hospital admissions and mortality. Infants with anaemia not associated with ID may even show adverse effects to iron supplementation, such as an increased susceptibility to infection or poor growth (Sazawal et al. 2006). Therefore, misdiagnosis, or assuming that 50% of anaemia is due to ID (DeMaeyer & Adiels-Tegman 1985) when specific iron status indicators are not available, may lead to inappropriate and/or ineffective treatment (Engle- Stone et al. 2013; Knowles et al. 2013). It is important to measure both APPs when interpreting iron status based on SF concentrations (Grant et al. 2012), because the sensitivity of SF may be reduced if only one marker of inflammation is taken into account. Thurnham et al. (2010) showed that the prevalence of ID is underestimated in a population with high rates of inflammation, if SF concentrations are not corrected for inflammation. Righetti et al. (2013), who studied infants and children from the Ivory Coast, and Engle-Stone et al. (2013), who included Cameroonian households with a child aged months old into their study, also demonstrated a significant increase in the prevalence of ID when SF concentrations were corrected for inflammation. Our study contributes to an increasingly rich literature (Thurnham et al. 2010; Grant et al. 2012; Engle-Stone et al. 2013; Knowles et al. 2013) that documents the effect of inflammation on SF, and reinforces the pivotal role of APP measurements to account for the effect of inflammation, when using SF as a marker of iron status. We recognise that because we only used SF and Hb, together with the APPs, to interpret the iron status of these infants, we are unable to conclude which of the four methods determined true ID prevalence. We therefore propose additional research to investigate the effects of inflammation on different indicators of iron status in diverse populations. The results will enable health professionals to advise on the best measure of iron status in the presence of inflammation. Furthermore,

7 Differential ferritin interpretation methods 227 studies should aim to validate peripheral blood iron markers against bone marrow iron smears (the golden standard for assessing iron status), similar to a study in Malawian children aged 6 59 months (Phiri et al. 2009). However, the collection of bone marrow iron smears is very invasive and not suitable for screening purposes (Phiri et al. 2009). Finally, our results affirm that SF should rather not be used on its own as proxy indicator for ID in infant populations with widespread inflammation because the ID prevalence will be underestimated. We conclude that, when APPs cannot be measured in the presence of inflammation, and SF is the singular ID measure, the SF cut-off of < 30 μgl 1 is recommended. This cut-off should be redefined for different populations and age categories. When only SF and both APPs are available, we recommend adjusting SF using CF. Acknowledgement The study is a secondary data analysis of the South African leg of the IRIS study. Source of funding None. Conflicts of interest The authors declare that they have no conflicts of interest. Contributions EN is responsible for the writing and finalisation of the article. HSK provided input into the content and writing of the article. JB provided input into the statistical analysis, content and writing of the article. MF provided input into the content and writing of the article. CMS is the principal investigator of the original IRIS study, provided input into the content and writing of the article. References Archibald F. (1983) Lactobacillus platarum, an organism not requiring iron. FEMS Microbiology Letters 19, Chaparro C.M. (2008) Setting the stage for child health and development: prevention of iron deficiency in early infancy. Journal of Nutrition 138, Cook D.G., Mendall M.A., Whincup P.H., Carey I.M., Ballam L., Morris J.E. et al. (2000) C-reactive protein concentration in children: relationship to adiposity and other cardiovascular risk factors. Atherosclerosis 149, De Onis M., Onyango A.W., Borghi E., Garza C., Yang H. & WHO Multicentre Growth Reference Study Group (2006) Comparison of the World Health Organization (WHO) child growth standards and the National Center for Health Statistics/WHO international growth reference: implications for child health programmes. Public Health Nutrition 9 (7), DeMaeyer E.M. & Adiels-Tegman M. (1985) The prevalence of anaemia in the world. World Health Statistics Quarterly 38, DeMaeyer E.M., Dallman P., Gurney J.M., Hallberg L., Sood S.K. & Srikantia S.G. (1989) Preventing and Controlling Iron Deficiency Anaemia through Primary Health Care: A Guide for Health Administrators and Programme Managers, pp World Health Organization: Geneva. Department of Health (2002) Summary report: national HIV and syphilis antenatal sero-prevalence survey in South Africa. Pretoria: South African Department of Health, pp Dewey K.G., Domellöf M., Cohen R.J., Rivera L.L., Hernell O. & Lönnerdal B. (2002) Iron supplementation affects growth and morbidity of breast-fed infants: results of a randomized trial in Sweden and Honduras. Journal of Nutrition 132, Engle-Stone R., Nankap M., Ndjebayi A.O., Erhardt J.G. & Brown K.H. (2013) Plasma ferritin and soluble transferrin receptor concentrations and body iron stores identify similar risk factors for iron deficiency but result in different estimates of the National prevalence of iron deficiency and iron-deficiency anaemia among women and children in Cameroon. Journal of Nutrition 143, Finch C. (1994) Regulators of iron balance in humans. Blood 84 (6), Gillespie S. & Johnston J.L. (1998) Expert Consultation on Anemia Determinants and Interventions, pp The Micronutrient Initiative: Canada. Grant F.K.E., Suchdev P.S., Flores-Ayala R., Cole C.R., Ramakrishnan U., Ruth L.J., et al. (2012) Correcting for inflammation changes estimates of iron deficiency among rural Kenyan preschool children. Journal of Nutrition 142, Kelleher S. (2006) Effects of age and mineral intake on the regulation of iron absorption in infants. Journal of Paediatrics 149 (5S), S69 S73.

8 228 E. Nel et al. Knowles J., Thurnham D.I., Phengdy B., Houamboun K., Philavong K., Keomoungkhone I. et al. (2013) Impact of inflammation on the biomarkers of iron status in a crosssectional survey of Lao women and children. British Journal of Nutrition 110, Kung u J.K., Wright V.J., Haji H.J., Ramsan M., Goodman D., Tielsch J.M. et al. (2009) Adjusting for the acute phase response is essential to interpret iron status indicators among young Zanzibari children prone to chronic malaria and helminth infections. Journal of Nutrition 139, Lutter C.K. (2008) Iron deficiency in young children in low-income countries and new approaches for its prevention. Journal of Nutrition 138, Phiri K.S., Calis J.C.J., Siyasiya A., Bates I. & Van Hensbroek M. (2009) New cut-off values for ferritin and soluble transferrin receptor for the assessment of iron deficiency in children in a high infection pressure area. Journal of Clinical Pathology 62, Posner K.L., Sampson P.D., Caplan R.A., Ward R.J. & Chenly F.W. (1990) Measuring inter-rater reliability among multiple raters: an example for nominal data. Statistics in Medicine 9, Righetti A.A., Wegmüller R., Glinz D., Quattara M., Adiossan L.G., N Goran E. et al. (2013) Effects of inflammation and Plasmodium falciparum infection on soluble transferrin receptor and plasma ferritin concentration in different age groups: a prospective longitudinal study in Côte dívoire. American Journal of Clinical Nutrition 97, Sazawal S., Black R.E., Ramsan M., Chwaya H.M., Stoltzfus R.J., Dutta A. et al. (2006) Effects of routine prophylactic supplementation with iron and folic acid on admission to hospital and mortality in preschool children in a high malaria transmission setting: communitybased, randomised, placebo-controlled trial. Lancet 367, Skinner A.C., Steiner M.J., Henderson F.W. & Perrin E.M. (2010) Multiple markers of inflammation and weight status: cross-sectional analyses throughout childhood. Pediatrics 125 (4), e801 e809. Smuts C.M., Lombard C.J., Benadé S., Dhansay M.A., Berger J., Hop L.T. et al. (2005a) Efficacy of a foodletbased multiple micronutrient supplement for preventing growth faltering, anemia, and micronutrient deficiency of infants: the four country IRIS trial pooled data analysis. Journal of Nutrition 135, 631S 638S. Smuts C.M., Dhansay M.A., Faber M., Van Stuijvenberg M.E., Swanevelder S., Gross R. et al. (2005b) Efficacy of multiple micronutrient supplementation for improving anaemia, micronutrient status, and growth in South African infants. Journal of Nutrition 135, 653S 659S. Stoltzfus R.J. & Dreyfuss M. (1998) Guidelines for the Use of Iron Supplements to Prevent and Treat Iron Deficiency Anemia, pp Washington ILSI Press: Washington DC, USA. Tam C.S., Clément K., Baur L.A. & Tordjman J. (2010) Obesity and low-grade inflammation: a paediatric perspective. Obesity Reviews 11, Thurnham D.I., McCabe L.D., Haldar S., Wieringa F.T., Northrop-Clewes C.A. & McCabe G.P. (2010) Adjusting plasma ferritin concentrations to remove the effects of subclinical inflammation in the assessment of iron deficiency: a meta-analysis. American Journal of Clinical Nutrition 92, Visser M., Bouter L.M., McQuillan G.M., Wener M.H. & Harris T.B. (2001) Low-grade systemic inflammation in overweight children. Pediatrics 107 (1), e13. WHO Multicentre Growth Reference Study Group (2006) WHO Child Growth Standards: Length/Height-For-Age, Weight-For-Age, Weight-For-Length, Weight-For-Height and Body Mass Index-For-Age: Methods and Development, p World Health Organization: Geneva. World Health Organization (WHO) (2011a) Serum ferritin concentrations for the assessment of iron status and iron deficiency in populations. Vitamin and Mineral Nutrition Information System, Geneva: WHO/NHD/MNM/ 11.1:1 6. World Health Organization (WHO) (2011b) Haemoglobin concentrations for the diagnosis of anaemia and assessment of severity. Vitamin and Mineral Nutrition Information System, Geneva: WHO/NMH/NHD/MNM/ 11.1:1 6. WHO/UNICEF/UNU. (2001) Iron deficiency anaemia assessment, prevention, and control: a guide for programme managers. Geneva: WHO/NHD/01.3: Zetterström R. (2004) Iron deficiency and iron deficiency anaemia during infancy and childhood. Acta Paediatrica 93, Zimmermann M.B., Chassard C., Rohner F., N Goran E.K., Nindjin C., Dostal A. et al. (2010) The effects of iron fortification on the gut microbiota in African children: a randomized controlled trial in Côte d Ivoire. American Journal of Clinical Nutrition 92,