Soluble transferrin receptor level, inflammation markers, malaria, alpha‐thalassemia and selenium status are the major predictors of hemoglobin in children 6–23 months in Malawi

Abstract In sub‐Saharan Africa, nearly three‐fourths of children 6–23 months are anemic. Yet, the underlying causes had not been sufficiently explored. This study, based on data (n = 348) extracted from the Malawi Micronutrient Survey–2015/2016 dataset, evaluated the contribution of multiple factors to the hemoglobin status of children 6–23 months. The association between hemoglobin and 19 predictors was assessed using multiple linear regression analysis, and the relative contribution of the covariates was determined based on delta‐R 2 value. The study found that 43.9% of children were anemic and 76.9% had elevated soluble transferrin receptor (sTfR) levels. Unit changes in serum ferritin (µg/L) and sTfR (mg/L) were associated with 0.01 g/dl rise (p = .041) and 0.05 g/dl decline (p < .001) in hemoglobin, respectively. Each 1 ng/ml increase in plasma selenium was met with 0.007 g/dl (p = .02) rise in hemoglobin. Hemoglobin showed negative relationships with α‐1‐acid glycoprotein (AGP) (β = −.339, p = .007) and C‐reactive protein (CRP) (β = −.014, p = .004) and positive association with child's age in months (β = .038, p = .003) and altitude in meters (β = .001, p = .015). Children affected by α‐thalassemia (β = −.75, p < .001), malaria (β = −.43, p = .029), and fever (β = −.39, p = .008) had significantly lower hemoglobin levels. On the contrary, nine variables including serum zinc and retinol binding protein were not significant predictors of hemoglobin. sTfR had the highest delta‐R 2 contribution (9.1%) to hemoglobin variations, followed by inflammation (5.2%), α‐thalassemia (2.5%), age (2.1%), fever (1.9%), and malaria (1.5%). The analysis suggested iron status, inflammation, and malaria were the major predictors of hemoglobin among Malawian infants and young children.


| BACKG ROU N D
Anemia is a complex disorder characterized by suboptimal hemoglobin or red blood cell mass that results in diminished cellular perfusion (Kraemer & Zimmermann, 2007). Though it is commonly asserted that at least half of the global burden of anemia is attributable to iron deficiency, other micronutrient deficiencies including folate, vitamin B12, vitamin A, and zinc and genetic disorders like sickle cell disease and thalassemia, chronic and infectious diseases, hemolysis or blood loss, bone marrow disorders lead to anemia. Furthermore, the epidemiological significance of the specific causes is likely to vary across populations (de Benoist, McLean, Egli, & Cogswell, 2008;Kraemer & Zimmermann, 2007).
Globally, more than 1.6 billion people, equivalent to a quarter of the world population, are affected by anemia (de Benoist et al., 2008); and preschool children (43%) and pregnant women (38%) take the highest burden (WHO, 2015). While anemia has global public health significance, sub-Saharan Africa (SSA) and South-East Asia are disproportionally affected, and in these regions, nearly half of women and two-thirds of children are anemic (de Benoist et al., 2008). Developing countries also contribute as high as 90% for the global anemia-related disability (Kassebaum, 2016). Though the prevalence of anemia significantly dropped in the last three decades, the decline had been steady and the Global Nutrition Target for reducing anemia by 50% between 1990 and 2025 is unlikely to be achieved (Kassebaum et al., 2014; WHO, 2017).
The complementary feeding period, which typically extends between 6 and 23 months of age, is a vulnerable period for growth faltering, micronutrient deficiencies, and common childhood illnesses (UNICEF, 2012). This is especially true in low-income countries where access to fortified and hygienic complementary foods is limited. As infants gradually transit from exclusive breastfeeding to family food, they frequently get exposed to unhygienic, monotonous, and nutrient-poor complementary foods (Gebremedhin, 2019). Analysis of multiple demographic and health surveys (DHS) indicated that the prevalence of anemia among children 6-23 months of age in SSA is as high as 75% and the consumption pattern of iron-rich or iron-fortified is very low (Gebremedhin, 2019;Prieto-Patron, Van der Horst, Hutton, & Detzel, 2018).
During early childhood period, anemia and iron deficiency anemia cause multiple consequences including death due to severe anemia, impaired physical and psychomotor developments, weakened cell-mediated immunity, and increased susceptibility to infectious diseases (Cherayil, 2010;Larson, Phiri, & Pasricha, 2017;UNICEF & WHO, 2011). A study based on data from multiple African countries estimated that each 1 g/dl increase in the population hemoglobin concentration is associated with a 24% decline in the risk of infant mortality (Scott, Chen-Edinboro, Caulfield, & Murray-Kolb, 2014).
World Health Organization (WHO) recommends that children 6-23 months living in areas where the prevalence of anemia is above 40% should daily receive 10-12.5 mg elemental iron supplementation for 3 consecutive months per year (WHO, 2016).
Previous studies conducted in Malawi indicated that anemia was extremely common among preschool children (Calis et al., 2008;National Statistical Office, 2017). The recent national DHS found that in 2015, 62% of preschool children were anemic and the prevalence exceeded 80% among children younger than 18 months of age (National Statistical Office, 2017). Underlying risk factors of anemia among Malawian children include the following: iron, vitamin B-12 and vitamin A deficiencies, infectious diseases including malaria, HIV infection and hookworm infestation, and genetic glucose-6-phosphate dehydrogenase deficiency (Calis et al., 2008).
Rural place of residence, maternal illiteracy, and low household socioeconomic status were also associated with anemia (National Statistical Office, 2017).
Despite the exceptionally high burden of anemia among infants and young children (6-23 months of age) in SSA (Gebremedhin, 2019;Prieto-Patron et al., 2018), the existing literature is primarily focused on the entire under five children and most studies only explored the sociodemographic determinants of anemia. Accordingly, the underlying causes of anemia and their epidemiological significance in children 6-23 months in SSA have not been adequately elucidated.
This study, based on the secondary data of the Malawi Micronutrient Survey (MMS)-2015/2016, evaluated the contribution of multiple risk factors including micronutrient deficiencies, inflammation markers, genetic disorders, and infant feeding practices, to the hemoglobin level of children 6-23 months of age.

| Design and study subjects
This cross-sectional study was conducted based on the secondary data of MMS, which was carried out between December 2015 and February 2016 on a subsample of households selected for the 2015-2016 Malawi DHS. The MMS collected diverse information relevant to anemia in multiple age groups including preschool children (6-59 months). For this specific study, the data of infant and young children 6-23 months were analyzed. The exclusion criteria were as follows: missing hemoglobin or child's age data, not listed in the original DHS dataset and, having age mismatch between the MMS and DHS datasets. Ultimately the data of 348 eligible infants and young children were analyzed ( Figure 1).

| Sample size
As the study was conducted based on secondary data, priory sample size estimation was not made. However, the adequacy of the available sample size for identifying predictors of hemoglobin was evaluated by G*power program (Faul, Erdfelder, Buchner, & Lang, 2009) using an approach suitable for multiple linear regression analysis. The specifications made during the computation were as follows: two-tailed test, 80% power, 95% confidence level, medium effect size (Cohen's f 2 = 0.15), and 19 predictors in the regression model. In the post hoc sample size estimation, medium effect size was assumed because small effect size may not be clinically relevant and large effect size inflates type II error. Data were collected from the mothers of the index children.

| Sampling procedure and data collection approach of the primary survey
The MMN survey collected diverse information relevant to anemia.
Comprehensive sociodemographic data were collected using the standard DHS questionnaire. Occurrences of common childhood illnesses (cough, fever, diarrhea) in the preceding 2 weeks were evaluated based on the reports of the caregivers. Infant and young child feeding practices were assessed using an approach compatible with the existing WHO standard (WHO, 2008).
Furthermore, information pertaining to iron biomarkers (serum ferritin and soluble transferrin receptor [sTfR]), serum zinc, retinol binding protein (RBP), plasma selenium, inflammation markers (C-reactive protein [CRP], and α-1 acid glycoprotein [AGP]), genetic blood disorders pertinent to anemia (sickle cell, alpha-thalassemia, glucose-6-phosphate dehydrogenase [G6PD] deficiency), malaria, and hematuria status, was collected. Hemoglobin was determined in the field using Hemocue 301 ® and was adjusted for altitude (CDC, 1998). RBP, inflammation markers, ferritin, and sTfR were measured using ELISA. Inherited blood disorders were determined from dried blood spot using PCR and serum zinc was measured via atomic emission spectrometry. Malaria and hematuria diagnosed in the filed using rapid tests (National Statistical Office et al., 2017).

| Variables of the study
The dependent variable of the analysis was hemoglobin level adjusted for altitude. The independent variables include the following: age, sex, and breastfeeding status of the child; child's

| Data analysis
The data analysis was performed using STATA/SE 14.0 program. The data of 348 children 6-23 months analyzed errors. Mixed-effects model was not considered because the number of observations per cluster was too few for many of the clusters.
All the 19 independent variables were entered into one multivariable model using the "enter" approach. The number of covariates in the model was not reduced because the sample size per covariate ratio was within the acceptable range (15-20 observations per variable; Austin & Steyerberg, 2015). The presence of multicollinearity among the covariates was assessed using variance inflation factor (VIF) and found to be within the acceptable range (VIF < 10). The VIF ranged from 1.10 for wasting to 1.82 to CRP. Partial plots were used for evaluating the normality and homoscedasticity of error terms. The goodness of fit was measured by adjusted-R 2 value and F-statistic.
The relative contribution of each of the predictor to the model was evaluated using delta-R 2 value, as determined by nested regression approach. Two independent variables (hematuria and sickle cell disease) were dropped from the regression analysis because they had

| Inflammation, illness, and genetic abnormalities
Inflammation appeared to be very common among infant and young children in Malawi. The median (IQR) of CRP concentrations was 1.8 (0.6-6.0) mg/L, and 28.3% had elevated levels (>5 mg/L). Similarly, the median AGP was 1.0 (0.0-2.0) g/L and 53.9% had elevated AGP  Consumption pattern of animal sources foods including flesh foods (33.8%), eggs (7.4%), and dairy products excluding breastmilk (4.4%) was relatively lower. had selenium deficiency (plasma selenium <70 ng/ml) (Combs, 2015) and 57.1% were zinc-deficient based on cutoff that has taken into consideration time of blood collection and fasting status of the subjects (King et al., 2016). On the basis of inflammation-adjusted ferritin levels (Suchdev et al., 2016), the prevalence of iron deficiency (42.0%) was also very high. Conversely, vitamin A deficiency was rare (3.2%) ( Table 2).

| D ISCUSS I ON
In SSA, infants and young children 6-23 months of age take the highest burden of anemia. However, the underlying causes had not

TA B L E 3 (Continued)
been adequately explored. This analysis identified multiple risk factors of anemia, with their relative epidemiologic significance, among Malawian children 6-23 months.
In this study, two iron biomarkers (sTfR and ferritin) found to be significant predictors of hemoglobin. While ferritin displayed marginally significant relationship, sTfR emerged as the strongest predictor with the highest delta-R 2 contribution to the model. The weak association observed between ferritin and hemoglobin can be due to the high prevalence of inflammation and the acute phase reactant nature of ferritin. Though ferritin is a sensitive and valuable index of iron nutrition, its utility in settings with high prevalence of inflammation is limited (Dignass, Farrag, & Stein, 2018;Nemeth & Ganz, 2014).
Convincing evidence exists that sTfR is more resistant to inflammation and it is an important biomarker of iron-deficient erythropoiesis in hospitalized patients and populations with high prevalence of inflammation, including preschool children (Beguin, 2003;Ragab, Ibrahim, Eid, Kotb, & Konsowa, 2002;Rohner et al., 2017). The strong association between hemoglobin and sTfR levels, the high prevalence of iron-deficient erythropoiesis (77%), and the low consumption of flesh foods (34%) suggested that iron deficiency is the foremost important cause of anemia in young children in Malawi.
The regression analysis might even have underestimated the significance of iron due to the statistical adjustment of sTfR for ferritin.
Nevertheless, both were kept in the equation with the intension of maximizing the goodness of fit of the model.
A negative association was also observed between hemoglobin and inflammation, and in combination, the two inflammation markers had the second highest delta-R 2 contribution (about 5%) to the model. This is consistent to understand that anemia of inflammation (AI) is the second most common cause of anemia worldwide, next to iron deficiency anemia (Ganz & Nemeth, 2009;Nemeth & Ganz, 2014). It has been hypothesized that, during inflammation, the pro-inflammatory cytokine IL-6 increases hepcidin to cause iron sequestration that in turn leads to iron-restricted erythropoiesis.
Selenium level was found to be an independent predictor of hemoglobin. The association between selenium and hemoglobin has not been extensively investigated in humans before. Yet, the available cross-sectional studies among school-aged children (Houghton, Parnell, Thomson, Green, & Gibson, 2016;Nhien et al., 2008), adults The analysis showed that hemoglobin level tends to increase with rising altitude. Conclusive evidence exists that hemoglobin rises at high altitude as a physiological response to elevation-induced hypoxia (CDC, 1998;WHO, 2011). Yet more surprising is the association was sustained despite the use of altitude adjusted hemoglobin values recommended by CDC (1998 The study found that malaria was very common among infants and young children (24%) in Malawi and children tested positive for malaria on average had 0.4-0.7 g/dl lower hemoglobin concentrations. This is consistent to the understanding that malaria causes anemia through multiple pathways including hemolysis of erythrocytes, suppressing of erythropoiesis, and maldistribution of iron (Helleberg et al., 2005;White, 2018). Further, malaria infection increases hepcidin that reduces absorption of iron in the blood (Prentice, Ghattas, Doherty, & Cox, 2007). In this analysis, we might have even underestimated the contribution of malaria to hemoglobin due to the statistical adjustment of malaria to multiple inflammation markers, fever, and other symptoms of the disease.
In this study, child's age showed an independent negative association with hemoglobin status. This is consistent to the findings of studies conducted in other low-income countries (Huang et al., 2018;Mohammed, Habtewold, & Esmaillzadeh, 2019). Prolonged exclusive breastfeeding, delayed initiation of complementary foods, and initiating complementary feeding with poorly diversified diet could contribute for the lower hemoglobin levels in infants. Further, a study that analyzed hepcidin pattern among young Kenyan children concluded that hepcidin-a negative regulator of iron absorption-declines throughout infancy and reaches the lowest level between 1 and 2 years of age, suggesting iron absorption is higher in the second year of age than during infancy (Atkinson et al., 2015). .004 *

1.9
Alpha-thalassemia bias and may underestimate the strength of association with hemoglobin. Further, adjustment of child feeding indicators (breastfeeding status and dietary diversity) for micronutrient status can underestimate their significance for anemia because improving micronutrient status is among the major pathways through which optimal child feeding prevents anemia. Finally, despite the fact that many covariates had been measured and adjusted in the study, some relevant predictors (e.g., B-12 and folate status, dysentery) were still missing. This might have lowered the goodness of fit of the model and induced residual confounding.

| CON CLUS ION
Iron deficiency and inflammation showed major contributions to the hemoglobin status of infants and young children 6-23 months in Malawi, possibly suggesting iron deficiency and anemia of infection are the major underlying causes of anemia in the country.
Other factors that emerged significant predictors of hemoglobin include alpha-thalassemia, malaria, other febrile illnesses, and serum selenium level. The assortation between plasma selenium and hemoglobin status should be investigated through randomized controlled trials.

ACK N OWLED G M ENTS
The author acknowledges Measure DHS and National Statistical Office of Malawi for allowing access to this dataset.

CO N FLI C T O F I NTE R E S T
The author declares that he does not have any conflict of interest.

E TH I C A L A PPROVA L
This original study was approved by the Institutional Review Board of Malawi National Health Sciences Research Committee.

I N FO R M E D CO N S E NT
In the original study, written informed consent was obtained from all study participants.