Soya, maize and sorghum ready-to-use therapeutic foods are more effective in correcting anaemia and iron deficiency than the standard ready-to-use therapeutic food: randomized controlled trial

Background The prevalence of anaemia and iron deficiency (ID) among children with severe acute malnutrition (SAM) and their correction during nutritional rehabilitation are not well documented. This study assessed anaemia and ID prevalence and their predictors at start of SAM treatment, and the efficacy of their treatment and effect on gut health of two novel Ready-To-Use Therapeutic foods (RUTF) prepared from soybean, maize and sorghum (SMS) with (MSMS-RUTF) or without added milk (FSMS-RUTF) compared to those of the standard formulation prepared from peanut and milk (PM-RUTF). Methods This was a 3-arms parallel groups, simple randomised, controlled non-inferiority trial in 6–59 months old Central Malawian children with SAM. Anaemia was defined using altitude- and ethnicity-adjusted haemoglobin. Iron status was defined using soluble transferrin receptor (sTfR) and body iron stores (BIS). We used Pearson’s chi-square test, t-test for paired or unpaired data, Kruskal-Wallis test for between-arm differences as appropriate and logistic regression to identify independent predictors of anaemia or iron deficiency anaemia (IDA). Results The sample size was 389. At admission, the prevalence [%(95%CI)] of anaemia was 48.9(41.4–56.5)% while that of ID and IDA were 55.7(48.6–62.5)% and 34.3(28.2–41.0)% when using sTfR criterion and 29.1(24.4–34.4)% and 28.9(23.7–34.9)% when using BIS criterion, respectively. At discharge, nutrition rehabilitation with SMS-RUTF was associated with the lowest prevalence of anaemia [12.0(6.9–20.3)% for FSMS-RUTF, 18.2(11.9–26.8)% for MSMS-RUTF and 24.5(15.8–35.9)% for PM-RUTF; p = 0.023] and IDA [7.9(3.4–17.3)% for FSMS-RUTF, 10.9(4.8–22.6)% for MSMS-RUTF and 20.5(10.7–35.5)% for PM-RUTF; p = 0.028]. SMS-RUTF was also associated with the highest increase in BIS [Change in BIS (95%CI)] among the iron deplete at admission [6.2 (3.7; 8.6), 3.2 (0.8; 5.6), 2.2 (0.2; 4.3) for the same study arms; Anova p = 0.045]. Compared to P-RUTF, FSMS-RUTF had the highest adjusted recovery rate [OR (95%CI = 0.3 (0.2–0.5) with p < 0.001 for FSMS-RUTF and 0.6 (0.3–1.0) with p = 0.068 for MSMS-RUTF]. No effect of iron content on risk of iron overload or gut inflammation was observed. Conclusions Anaemia and ID are common among children with SAM. FSMS-RUTF is more efficacious in treating anaemia and correcting BIS among this group than PM-RUTF. Trial registration This study was registered on 15 April 2015 (PACTR201505001101224). Electronic supplementary material The online version of this article (10.1186/s12889-019-7170-x) contains supplementary material, which is available to authorized users.


Background
Globally, severe acute malnutrition (SAM) affects about 16.4 million children aged 6-59 months, out of whom 23% currently access treatment using ready to use therapeutic food (RUTF) [1,2]. Moreover, in Communitybased Management of Acute Malnutrition (CMAM), RUTF is the recommended diet for children in this age group being treated for SAM [3]. RUTF is a fortified energy dense food intended to cover all nutritional needs during recovery from SAM [3][4][5]. The most common formulation of RUTF is based on peanut and milk. However, research has been ongoing to develop alternative formulations that are equally effective and are cheaper and more acceptable in settings where peanut is not popular.
SAM is known to be a pluri-deficiency disease combining macronutrients and multiple micronutrients [6]. Thus, diets for SAM treatment should aim at correcting micronutrients deficiencies in addition to fulfilling requirements for catch up growth [6][7][8][9]. Anaemia is one of the most widespread nutritional deficiencies affecting over 1.6 billion people worldwide [10]. The burden varies from one study to another but the most recent representative estimates indicate that overall, 47.4% children below 5 years are affected worldwide with the highest prevalence of 64.6% found among African children [10][11][12][13].
In children, anaemia (defined as a Haemoglobin below 110 g/L) is a clinical condition that results in weakness, difficulty in concentrating, susceptibility to infections, delayed cognitive and motor development and increased mortality [14,15]. Anaemia can be classified into nutritional and non-nutritional anaemia, and is caused by several different factors including nutritional deficiencies, infections and inherited red blood cell disorders [16,17]. Studies indicate that at least 50% of anaemia in children below 5 years of age from low and middle income countries is due to Iron deficiency (ID) [13,14]. This has led to various initiatives including supplementation and food fortification to increase intake of iron [18,19].
Studies conducted in Africa and Asia have reported prevalence of over 80% of anaemia among SAM children admitted in tertiary care hospitals [20][21][22][23]. To our knowledge, no study has reported the prevalence among children admitted into Community-based Management of Acute Malnutrition (CMAM) programmes, a treatment approach that allows early initiation of nutrition rehabilitation [24,25]. The contribution of ID in anaemia during SAM is not well established, and available data suggest that the prevalence of ID may be context specific and depends on the co-existence of other factors contributing to the severity of SAM [26][27][28][29]. Despite the dearth of data on prevalence of IDA in SAM, the high prevalence of anaemia in children aged below 5 years and its serious clinical impacts underscore the importance of correcting ID during treatment of SAM.
Moreover, the high iron demand during catch-up growth makes recovering SAM cases particularly susceptible to ID. The presence of ID means that these children are at low risk of iron supplementation side effects even in malaria endemic areas as most side effects are observed in iron replete individuals [30,31]. Indeed, the World Health Organisation still recommends routine iron supplementation for prevention and treatment of IDA in malaria endemic areas where prevalence of IDA is also usually high [31].
Absorption of iron is influenced by both dietary and physiological factors, with the former being determined mainly by the form of iron and by the presence of inhibitors and enhancers in a particular meal [32]. Among other factors, phytic acid, milk proteins and calcium, have been identified as nutritional factors inhibiting iron absorption, while ascorbic acid enhances its absorption [32][33][34]. These factors are present in varying amounts in formulations based on soybean, maize and sorghum [34], and their effects on anaemia and iron status are not well understood.
The currently recommended iron content of RUTF of 10-14 mg/100 g tries to balance the need to correct ID and provide iron to support the rapid tissue synthesis required for catch-up growth with the need to minimise the production of free radicals and the risk of enhancing bacterial growth and virulence [3,35,36]. However, a study has shown that RUTF providing 22 mg per day of iron is less effective in maintaining or improving haemoglobin level than the therapeutic milk F100 providing 60 mg/day of iron [37]. More recently, we published results indicating that an RUTF containing 43.8 mg/100 g iron is more effective in treating anaemia than an RUTF containing 10-14 mg/100 g, however, the study was underpowered to demonstrate statistical difference in haemoglobin change in unadjusted analysis [38].
Concerns remain regarding the level of iron in therapeutic or supplementary food despite some studies showing the positive effect of iron supplementation on growth. These concerns are related to the potential side effects that some studies have attributed to iron fortified food of inducing an increase in the population of pathogenic bacteria in the gut that may induce intestinal inflammation [36]. Despite that a study conducted in South Africa showed that the iron form used in this study was not associated with increased bacterial growth, data comparing food with different iron contents are scarce making this a priority research area [36,39].
This study aimed to determine the burden of anaemia among children with uncomplicated SAM and assess the efficacy of two novel RUTF formulations in correcting anaemia and ID during SAM treatment. The specific objectives were: 1. To determine the prevalence and risk factors of anaemia and ID among children with uncomplicated SAM, 2. To analyse the impact of the novel recipes with reduced milk content on the efficacy of treating anaemia and ID, and 3. To examine the impact of increased iron intake on intestinal inflammation.

Methods
This study was embedded within a larger 3-arms parallel group, simple randomised, controlled trial assessing the efficacy in treating SAM of two novel formulations of RUTF, using the standard peanut-milk RUTF as control.
The methodological aspects were reported in detail previously in the paper presenting on the main findings of the randomised clinical trial in recovery rates from SAM [40]. Below we only include the aspects of the methods required to understand the findings of this paper.

Study design
This was a randomised clinical trial assessing the efficacy of 3 different formulation of RUTF in correcting anaemia and ID during the treatment of SAM. A subsample from the study population of the larger study, was randomly selected. The two principal outcomes of interest for this embedded study were: 1) prevalence and predictors of anaemia and ID at admission and at discharge; and 2) between arms difference in prevalence of anaemia and ID at discharge.

Study setting and participants
The study was conducted at 21 day-care feeding centres scattered in three districts in Central Malawi (Lilongwe, Dedza and Mchinji). The altitude of this area ranges from 578 m to 1300 m above sea level. The incidences of malaria, intestinal parasitic infestation, viral and bacterial infections are high during childhood [41,42]. A subsample of 389 children aged 6-59 months diagnosed with SAM (mid upper arm circumference [MUAC] < 115 mm or bilateral pitting oedema of any degree) who also had good appetite and no medical complication, were selected from the larger study sample of 1347 children randomly allocated in the 3 treatments group by the simple randomization (ratio 1:1:1) based on a computer generated random list and using a closed envelope method as already described [40]. The subsample was selected using a predetermined schedule, where the phlebotomist nurse visited 3 to 4 sites per day and collected samples for baseline data in all those who were admitted 48 h before the visit. The collection of samples for baseline data was conducted across all the sites and was stopped when the required sample was reached. These same subjects were followed for collection of further samples at the subsequent sampling time-points.

RUTF formulations
The two new RUTF formulations were based on extruded soya, maize and sorghum, one containing no milk (FSMS-RUTF) and the other 9% milk (MSMS-RUTF). The control RUTF (PM-RUTF) was based on peanuts and dried skim milk and contained 28% milk. As shown in Table 1, all the RUTFs were designed to meet the World Health Organisation (WHO) recommendations for RUTF mineral and vitamin levels [1] with the exception of iron and zinc. Iron, zinc and vitamin C levels in FSMS-RUTF and MSMS-RUTF were increased to attain a phytic acid / iron molar ratio, ascorbic acid / iron weight ratio and zinc / iron weight ratio of < 2.5, 3.0-16.0 and 0.8-3.5, respectively to enhance iron and zinc absorption [43,44]. Iron was in the form of ferrous sulphate monohydrate. All samples were identical in packaging except a colour code which was not understood by the participants and field implementers.

Treatment protocol and Iron intake
The nutrition and medical management of children in all study groups were similar and followed the Malawi national guidelines for the management of acute malnutrition with the exception of the day-care approach and the fact that children were allowed to eat the RUTF ad libitum. Study assistant nurses fed the children with the support of the caregivers and closely monitored intake and occurrence of symptoms and clinical signs such as appetite, diarrhoea, vomiting, abdominal pain, flatulence, fever, skin eruption, cough and respiratory distress. No iron was added to the treatment in case of mild or moderate anaemia. As per the national protocol, a blood transfusion was needed only if the haemoglobin was < 5 g/dl.

Data collection and procedures
Nutritional and morbidity parameters were monitored and recorded daily on specially designed forms. Blood samples for measurement of anaemia (haemoglobin, haematocrit and full blood count), iron status (ferritin, soluble transferrin receptor), retinol-binding protein and inflammation (Creactive protein, and alpha-1-acid glycoprotein) parameters were measured on admission and at discharge. Trained paediatric phlebotomists collected blood by antecubital or metacarpal venipuncture into appropriate tubes provided by the Lilongwe University of North Carolina research laboratory. All the blood specimens were immediately stored in CubeCooler™ and kept at 4°C [45] and delivered to the Lilongwe research laboratory within 6 h. Samples for the measurement of haemoglobin, haematocrit and full blood count were analysed at the Lilongwe UNC laboratory immediately after delivery. For other parameters, samples were later shipped to Germany for the measurement of the 5 plasma proteins [ferritin, soluble transferrin receptor (sTfR), retinol-binding protein (RBP), C-reactive protein, and alpha-1-acid glycoprotein (AGP)]. At the Lilongwee laboratory, cells and plasma were separated within 24 h and the plasma was subsequently stored at − 80°C until shipping to Germany. Before shipping to Germany, the samples were unfrozen to allow for the transfer of 0.2 ml of plasma into the VitMIN laboratory special pre-labelled storage tubes (Willstaett, Germany). All the samples were refrozen after the transfer. During shipping to Germany, the sample cold-chain was maintained using cooler boxes with dry ice.
To assess the impact of the different study RUTFs on gut inflammation, stools samples were collected from the subset of subjects from whom fresh stools could be obtained during the day of the visit by the samples collectors in plastic pots at baseline, at 3 weeks and at discharge for the measurement of calprotectin [47,48]. A quantitative point-of-care chromatographic immunoassay kit (Quantum Blue®, Alpha laboratories, Hampshire, UK) was used for the measurement [49,50]. As per the manufacturer's instructions, two drops of homogenised stool were applied inside a plastic cassette and the cassette inserted into a portable electronic reader that displayed numerical value representing calprotectin concentration. The cassette reader was calibrated for each day of tests.

Sample size
Sample size was calculated to demonstrate the noninferiority of the new RUTFs in terms of haemoglobin change [40]. A total of 192 children studied at both admission and discharge were required to be 90% sure that the lower limit of a one-sided 95% confidence interval would be above the non-inferiority limit of − 0.5 g dl-1 indicating no difference between the arms. We increased this sample to a minimum of 300 children at admission and 225 (75 per arm) at discharge, to account for loss to follow up during nutrition rehabilitation due to defaulting, death, withdrawal of consent for blood collection, difficulties in blood collection or the requirement for a transfusion during treatment. This sample size was also sufficient for the determination of the prevalence of anaemia and ID at admission (calculated sample size = 292 children assuming prevalence of 50 ± 5%) and discharge (calculated sample size of 233 children assuming prevalence of 25 ± 5%), the detection of a trend in the prevalence of anaemia at discharge equivalent to a 9% difference between groups and the existence of a difference in intestinal inflammation between the study arms.

Variable transformation, adjustments and definitions
Individual haemoglobin values were adjusted for altitude and ethnicity according to the methods by WHO [51] and Sullivan et al. [52]. The adjusted values were then used to define anaemia using the recommended cut-off for children aged 6-59 months of 110 g/L [51,52].
As β-thalassemia trait is common among Malawian children, we calculated the Mentzer index to serve as a proxy of presence of β-thalassemia trait [16,53,54]. The index is calculated as the quotient of the mean corpuscular volume (MCV, in fL) divided by the red blood cell count (RBC, in Millions per microliter). β-thalassemia trait was assumed if the index was < 13 [53,54]. Plasma ferritin, soluble transferrin receptor (sTfR) and body iron stores (BIS) were used to describe iron status [55,56]. As these biomarkers are impacted by inflammation, they were adjusted using the inflammation biomarkers C-reactive protein and alpha-1-acid glycoprotein (AGP), according to the method proposed by Thurnham et al. [57]. We report iron sufficiency or deficiency using BIS (ratio of inflammation-adjusted Plasma Ferritin to inflammation-adjusted sTfR [58] as a continuous variable with a BIS < 0 mg.kg − 1 indicating ID, a BIS between 0 and 2.9 mg.kg − 1 low iron status and a BIS ≥3 mg.kg − 1 indicating normal iron status. With adjusted sTfR, a value > 8.3 mg/L indicated ID while a value below this cut-off indicated absence of ID. We used Plasma Ferritin to define excess body iron stores or iron overload [59], taking a cutoff > 100 μ/L to define excess body iron stores [59].

Data management and analysis
Data were double entered by two enumerators into a customized Epidata database prepared for this study [60]. Quality of data entry was monitored by the supervisors who cross-checked a random selection of 10% of the records. Cleaned data were exported to stata-13 for analysis [61].
Some data for key biological markers of anaemia and iron status at discharge data collection point could not be obtained. An assessment of the reasons identified the following: discontinuation of collection of specimens for this sub-study when the number that the budget allowed to be included was reached; child dropout from the study; child death; and no return after being transferred for management of complication. The budget shortage was due to the unexpected rise in prices of the tests obliging us to reduce the sample of participants to be surveyed at discharge. The retrospective follow up of defaulters revealed that the defaults were attributed to relocation, death or mothers related reasons (fatigue due to advanced pregnancy, delivery of new baby, taking care of a hospitalised household member, divorce and farming obligations). As all these reasons are responsible of missing values that can be classified as "missing completely at random" we opted for complete case analysis approach that drops the missing values in the analysis [62,63].
Means and standard deviations (SD), medians and interquartile ranges (IQR) or proportions and 95% Confidence intervals (95%CI) were used to describe the admission and discharge parameters, as appropriate. All 95% CIs used clustered robust estimates of the variance to account for clustering at the level of the day care centre. Means were compared using t-test for paired or unpaired data, as appropriate, medians were compared using the Kruskal-Wallis test and proportions compared using the Pearson's chi-square test or the chi-square test for trend as appropriate. We used logistic regression to identify predictors of anaemia and IDA at admission and discharge. The predetermined candidate predictors were study arm, age, sex, MUAC of admission, presence of oedema at admission, adjusted haemoglobin of admission, length of stay in programme, daily RUTF intake and presence of β-thalassemia trait. The final model was obtained using the approach recommended by Greenland for the selection of variables [56]. We started with the full model with all the potential predictors listed above and thereafter manually removed nonsignificant variables one by one using the p-value and the change-in-estimate method [64]. Multivariable linear regression was used to obtain adjusted prevalence at discharge of anaemia, ID and IDA. The adjustment variables were gender, mother education level, anaemia at admission, iron status of admission as measured by BIS, and inflammation category of admission.

Results
Out of the 1303 children screened and accepted into the larger efficacy study, 389 were randomly selected for this embedded study of whom haemoglobin and iron status at admission was determined in 386 and 343, respectively. The other 46 were excluded due to difficulties of obtaining adequate blood samples (n = 43) and withdrawal of consent (n = 3). Recruitment and follow up for this study was conducted between September 2015 and June 2016. Table 2 presents the characteristics at admission of the enrolled children. The sample was balanced in terms of sex and breastfeeding status. There were more children below 24 months and the great majority had both parents alive with almost all the heads of the households being subsistence farmers. The three groups were different with regard to maternal education status, with PM-RUTF group having the lowest proportion of mothers with no education. Overall, less than 20% of the mothers reached the secondary level of education. There were marginally insignificant differences in gender and inflammation among the three groups. Access to safe drinking water was high but utilisation of insecticide treated net (ITN) was reported by less than 50%. The two clinical forms of SAM were equally represented in the sample. The majority had some form of inflammation with elevated values of C-reactive protein or AGP. The Mentzer index classified only few of the children as probably having βthalassemia trait. The between arms comparison showed that there was no difference in the other parameters evaluated.

Prevalence of anaemia, iron deficiency and iron deficiency anaemia at admission
Out of the 386 children for whom haemoglobin was measured on admission (3 children withdrew consent after randomisation), anaemia was diagnosed in [% (95%CI)] 48.9(41.4-56.5)%. Table 3 presents the distribution of anaemia per selected parameters. Children with β-thalassemia trait had low prevalence of anaemia when compared to those without the trait while those with inflammation at incubation and early convalescence had increased risk of anaemia when compared to those with no inflammation. Logistic regression analysis (detailed table provided as Additional file 1) confirmed the protective effect for anaemia of having β-thalassemia trait  . Among children below 24 months of age, those still breastfed had a significantly higher prevalence of ID (p = 0.004). A third of all enrolled children had IDA and > 60% of the anaemia was associated with ID (Table 3). ID was also observed in non-anaemic children ( Table 3). The independent predictors of IDA were largely similar to those for anaemia (Additional files 1 and 2); they included the absence of β-thalassemia trait Regarding ID and IDA, Table 4 also shows that FSMS-RUTF arm had higher cure rates and lower frequency of new episodes for both ID and IDA than the two other arms. The linear trend was also statistically significant. Figure 1 shows the distribution of individuals' values of BIS at admission and discharge according to iron status at admission. Overall, there was an increase in BIS but with variations according to type of RUTF used for nutrition rehabilitation and according to BIS category at admission (Additional file 5). Some children with normal BIS at admission became iron deficient during the course of nutrition recovery. For children with normal iron stores, the BIS decreased in all study arms by an average (Δ(95%CI) of − 0.9 (− 1.9; 0.1) for the FSMS-RUTF arm (p = 0.074), − 0.3 (− 1.9; 1.3) for the MSMS-

Effect of RUTF used for nutrition rehabilitation on gut inflammation
Gut inflammation as measured by concentration of calprotectin in stools was high in all study arms at admission and remained high after 3 weeks of treatment and at discharge, with no statistically significant difference across the study arms.

Discussion
In this study, we have shown that prevalence of anaemia, ID and IDA are high in Malawian children with uncomplicated SAM admitted into a CMAM programme, above the WHO criterion of 40% defining a severe public health problem and much higher than the 29% reported by the most recent community-based study among Malawian children below 5 years [16,65]. Similar high prevalence of anaemia among children with SAM have been reported in other countries and continents [66,67], indicating that SAM children should be considered as an "at risk group" for anaemia. Emerging evidence suggests that anaemia in low income countries is multifactorial with dietary factors an important cause [20,28,29,67]. Our results indicate that a high proportion of Malawian children who develop SAM during hunger seasons, which occur around November to March when family food reserves are largely depleted, have IDA and a pre-existing iron deficiency before developing SAM indicating that the usual factors causing iron deficiency are also in play for children with SAM [26,[68][69][70]. This finding contradicts the historical view that children with SAM usually have elevated body iron at diagnosis and supports the current practise of presumptive treatment of IDA during treatment of SAM.
This study also demonstrates that the content of cow's milk in RUTF is inversely related to the product's efficacy in treating anaemia and replenishing Body Iron Stores (BIS). This is more demonstrated by the difference in efficacy between the two SMS-based formulations which had similar phytic acid/iron and ascorbic acid / iron molar ratios and ascorbic acid / iron weight ratios. It is well known that casein, whey protein and  [15,33,71]. In this study, the milk-free FSMS-RUTF product containing 35.1 mg Fe / 100 g performed significantly better than MSMS-RUTF containing 9.3% cow's milk and comparable iron content of 31.6 mg / 100 g. In turn, the MSMS-RUTF performed significantly better than the standard PM-RUTF product containing 28.2% cow's milk and 10.5 mg Fe /100 g. These results are similar to those of a study conducted in Burkina Faso that also reported a better absorption of zinc when provided as tablet than when given concurrently with a low dose lipid-based nutrient supplement (LNS) containing cow's milk powder [72]. In order to achieve optimal mineral delivery through ready to use foods, these findings suggest that further research is needed to determine the optimal balance between levels of mineral fortification and the composition of the food including the content of milk. The superior performance of the pulse/grain-based formulations compared to the milk-based formulation demonstrates that the high content of phytates in cereals and pulses is not necessarily a barrier to using these ingredients in the manufacture of RUTFs. To address this problem, the iron and vitamin C levels in the two new  RUTF formulations were increased to achieve phytic acid: iron molar ratio and vitamin C: iron weight ratios that are optimum for iron absorption [43,44,73]. A potential problem with increasing the level of iron is that unabsorbed iron could change the gut microbiome and promote the growth of pathogens [47,[74][75][76]. This is particularly relevant as this study was conducted in a malaria-endemic setting where some studies have suggested that iron supplementation may enhance morbidity in iron replete children [77] and controversy exists over whether foods in such settings should be fortified with iron. However as reported in our previous study, no differences in morbidity was observed in our studies between the high iron and low iron RUTF arms, indicating that the iron content of the foods was not associated with an increased frequency of episodes of fever [40]. This finding mirrors that of a previous study using an LNS with a much higher level of iron for a long period of 6 months [78].
The absence of any differences in stools calprotectin concentration, a marker of intestinal inflammation between the study arms at any of the three points of measurement also indicates that the higher levels of iron were not associated with increased gut mucosal inflammation [47,79]. These findings again reflect those of a study conducted among South African children that concluded that supplementation with ferrous sulphate had no detrimental effect on microbiota [39]. However, it is interesting to note that high faecal calprotectin levels persisted in all three study arms, a finding reported elsewhere in children recovering from SAM [80] and suggestive of persistent intestinal inflammation that is not completely resolved by the treatment. Although several studies have shown that nutrition rehabilitation with RUTF is associated with improvement of the microbiota, the improvement is incomplete and transient [81,82]. In children with cystic fibrosis who have similar gut mucosal abnormalities to those seen in SAM, the use of probiotics has been associated with improvements in mucosal structure [83,84] suggesting that adding probiotics into RUTF may help to achieve sustained normalisation of intestinal mucosa. Interestingly, 8 years ago we published a pioneering study that assessed the effect of adding probiotic in RUTF and yielded promising results suggesting positive effect on mortality [85].
Increasing the iron was also not associated with iron overload and in this study changes in BIS were inversely proportional to BIS at the start of treatment, demonstrating that in the presence of tissue ID there is an upregulation of intestinal absorption and vice versa. This indicates that tissue iron status in children with SAM still plays the central role in regulating (upregulating and downregulating) iron intestinal absorption and mobilisation from body stores [86][87][88][89]. As a result, during the usual duration of CMAM treatment biological regulation mechanisms appear able to keep tissue iron concentration below the excess cut-off and the risk of excess body iron stores is limited. This finding contradicts observations gathered before the advent of CMAM that reported a deregulation of iron metabolism at the time of admission into therapeutic feeding programmes [24,86,90]. In practice, it means that the fear of an increased absorption of iron during SAM is not a reason to prevent upward revision of the recommended iron density for RUTF used for uncomplicated SAM.
An important finding from this study is that none of the formulations used, even the higher iron non-milk formulation, adequately corrected ID and IDA by the time of reaching CMAM anthropometric discharge criteria in a substantial proportion of children. This is a serious deficiency. The improved performance of the plant-based RUTFs containing higher quantities of iron and the absence of negative side effects of increasing the iron content [38,40] strongly suggest that the current recommendation on iron density of RUTF needs to be revised upwards [3].
We also found that the iron status of some children deteriorated during treatment, particularly with the WHO recommended lower iron formula, indicating that the recommended level of iron in RUTF is less effective in covering the increased need for iron during rapid catch up [91]. This finding is in line with that of an earlier study conducted in Senegal that showed that the currently recommended iron density in RUTFs is not sufficient to cover the needs of children with SAM during the period of rapid catch up growth, particularly in those children with higher weight gain velocities [37].
This study also provided evidence that current anthropometric discharge criteria are not a good proxy for attainment of homeostasis for iron, with anthropometric discharge criteria being met before a significant proportion of children attains replenished BIS, especially those who were depleted at the start of treatment. Similar issues have been shown in a recent study on immunity in children recovering from SAM [92], and taken together indicate the need for further research to determine the optimum criteria of discharging children that takes into account the fact that RUTF intake should not be prolonged unnecessarily given its high content of fat and other nutrients, and the need to minimize risk of relapse due to excessive morbidity. In parallel to the determination of new discharge criteria, there is a need to accelerate research aiming at determining which follow up intervention may ensure post-discharge recovery of iron homeostasis. Several studies have recently published results of interventions that can be combined with CMAM to ensure such post-discharge recovery [72,93,94]. With the increasing interest in ensuring long term benefits of SAM treatment, such interventions may contribute to the reduction of excessive morbidity observed during the first 3 months after discharge. Indeed, the high prevalence of ID and IDA among the breastfed children with SAM in this study points toward insufficient iron intake from the readily available complementary foods used by children below 5 years of age in communities where children with SAM come from [95,96].
This study demonstrates that RUTFs should always have an appropriate density of bioavailable iron to enable reversal of ID and IDA. The FSMS-RUTF iron profile can serve as a starting point for further research and improvements that take into account the effect of cow's milk and other ingredients and factors such as phytate in the food. As an important proportion of children recovering from SAM meet anthropometric criteria for cure before reversal of their IDA, there is a need to develop an appropriate post-discharge follow up package for continuation of ID correction such as proper administration of well-formulated micronutrient powders or lipid-based nutrients supplements, anti-helminth and anti-malaria treatments, and improvements of environmental hygiene.
Our results have to be interpreted taking into account some methodological limitations. The first is that we used indicators of iron status that are influenced by inflammation and a high proportion of children enrolled into the study had inflammation at admission. To address this, we adjusted these parameters for inflammation [58,97] and used the sTfr and the sTfr/ferritin index to define the iron status [98]. Studies have shown that these are less affected by inflammation and are good indicators of iron status in Malawian children [42,56,58,99] and we therefore believe that our estimates are the most precise that can practically be obtained in the absence of bone marrow or liver biopsy analysis. Although serum ferritin was also measured, we did not use this biomarker for defining iron status in this study because of the clear absence of correlation of this biomarker and anaemia in our study population. This finding was in accordance with findings from other teams showing that serum ferritin is less accurate in assessing ID in Malawian children even after adjustment for inflammation as malaria infection may lead to long lasting ferritin elevation [42,100]. The second limitation is the absence of screening for malaria and any inherited disorders known to influence the markers of iron status or the biological regulation of iron homeostasis [16,[101][102][103]. This may have resulted in overestimating the prevalence of ID and IDA. However, as this study was embedded in an RCT with good randomisation at baseline, it is unlikely that this factor introduced a difference between the arms. Finally, because of financial constraints we were not able to include any analysis of the microbiota. This could have permitted a better understanding of the bacterial population of the gut at the onset and during the course of treatment and enabled us to ascertain the safety of iron densities used with regard to pathogenic gut bacterial growth and selection. Moreover, having data on microbiota changes could have allowed us to examine the influence of the microbiota on the observed difference in treatment effect on haemoglobin and iron status.

Conclusion
Our study shows that anaemia including IDA is common among children with SAM and this justifies presumptive treatment against ID and IDA in this group. The study also indicates that the FSMS-RUTF with a higher level of iron and no cow's milk is more efficacious in correcting anaemia and BIS among children with SAM than the current standard milk-based formulation and that raising the levels of iron in RUTF was not associated with any adverse outcomes. Authors' information None.

Funding
The study was funded by Ajinomoto Co. Inc., the Japan International Cooperation Agency and Global Innovation Fund. Both funding organizations had no role in the study design, data collection, analysis and interpretation, or the decision to publish the findings.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Ethics approval and consent to participate Adherence to ethical principles has been extensively described elsewhere [32]. Briefly, the study was approved on 23rd March 2015 by the National Ethics Committee of the Malawi Ministry of Health and from Institutional Review Board of Ajinomoto Co., Inc., (Ref no. 2014-021) and was registered on 15th April 2015 in the Pan African Clinical Trial Registry (PACTR201505001101224) prior to starting recruitment. Participation was voluntary and the carer signing of the consent form confirmed the enrolment. The study was carried out in accordance with the declaration of Helsinki and adheres to CONSORT guidelines.

Consent for publication
Not applicable.

Competing interests
Valid Nutrition and Ajinomoto Co. Inc. designed and produced the SMS-RUTFs. PA is an employee of Valid Nutrition and HM is an employee of Ajinomoto Co. Inc. SC is the unpaid director of Valid Nutrition and a director of Valid International Ltd. Valid International Ltd. is the sister company of Valid Nutrition, and PB and KS are Valid International employees. All other authors had no conflict of interest. Valid Nutrition administered the study grant. Valid Nutrition and Valid International researchers participated in the study design, implementation, and interpretation of the results. Apart from contributing HM's expertise, Ajinomoto Co. Inc. had no role in the study design, data collection, analysis and interpretation, or the decision to publish the findings.