|Home | About | Journals | Submit | Contact Us | Français|
Iron-deficiency anaemia is common during childhood. Iron administration has been claimed to increase the risk of malaria.
To evaluate the effects and safety of iron supplementation, with or without folic acid, in children living in areas with hyperendemic or holoendemic malaria transmission.
We searched the Cochrane Infectious Diseases Group Specialized Register; the Cochrane Central Register of Controlled Trials (CENTRAL), published in the Cochrane Library, MEDLINE (up to August 2015) and LILACS (up to February 2015). We also checked the metaRegister of Controlled Trials (mRCT) and World Health Organization International Clinical Trials Registry Platform (WHO ICTRP) up to February 2015. We contacted the primary investigators of all included trials, ongoing trials, and those awaiting assessment to ask for unpublished data and further trials. We scanned references of included trials, pertinent reviews, and previous meta-analyses for additional references.
We included individually randomized controlled trials (RCTs) and cluster RCTs conducted in hyperendemic and holoendemic malaria regions or that reported on any malaria-related outcomes that included children younger than 18 years of age. We included trials that compared orally administered iron, iron with folic acid, and iron with antimalarial treatment versus placebo or no treatment. We included trials of iron supplementation or fortification interventions if they provided at least 80% of the Recommended Dietary Allowance (RDA) for prevention of anaemia by age. Antihelminthics could be administered to either group, and micronutrients had to be administered equally to both groups.
The primary outcomes were clinical malaria, severe malaria, and death from any cause. We assessed the risk of bias in included trials with domain-based evaluation and assessed the quality of the evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach. We performed a fixed-effect meta-analysis for all outcomes and random-effects meta-analysis for hematological outcomes, and adjusted analyses for cluster RCTs. We based the subgroup analyses for anaemia at baseline, age, and malaria prevention or management services on trial-level data.
Thirty-five trials (31,955 children) met the inclusion criteria. Overall, iron does not cause an excess of clinical malaria (risk ratio (RR) 0.93, 95% confidence intervals (CI) 0.87 to 1.00; 14 trials, 7168 children, high quality evidence). Iron probably does not cause an excess of clinical malaria in both populations where anaemia is common and those in which anaemia is uncommon. In areas where there are prevention and management services for malaria, iron (with or without folic acid) may reduce clinical malaria (RR 0.91, 95% CI 0.84 to 0.97; seven trials, 5586 participants, low quality evidence), while in areas where such services are unavailable, iron (with or without folic acid) may increase the incidence of malaria, although the lower CIs indicate no difference (RR 1.16, 95% CI 1.02 to 1.31; nine trials, 19,086 participants, low quality evidence). Iron supplementation does not cause an excess of severe malaria (RR 0.90, 95% CI 0.81 to 0.98; 6 trials, 3421 children, high quality evidence). We did not observe any differences for deaths (control event rate 1%, low quality evidence). Iron and antimalarial treatment reduced clinical malaria (RR 0.54, 95% CI 0.43 to 0.67; three trials, 728 children, high quality evidence). Overall, iron resulted in fewer anaemic children at follow up, and the end average change in haemoglobin from base line was higher with iron.
Iron treatment does not increase the risk of clinical malaria when regular malaria prevention or management services are provided. Where resources are limited, iron can be administered without screening for anaemia or for iron deficiency, as long as malaria prevention or management services are provided efficiently.
Why the review is important
Children living in malarial areas commonly develop anaemia. Long-term anaemia is thought to delay a child's development and make children more likely to get infections. In areas where anaemia is common, health providers may give iron to prevent anaemia, but there is a concern amongst researchers that this may increase the risk of malaria. It is thought that the iron tablets will increase iron levels in the blood, and this will promote the growth of the Plasmodium parasite that causes malaria. We aimed to assess the effects of oral iron supplementation in children living in countries where malaria is common.
Main findings of the review
Cochrane researchers searched the available evidence up to 30 August 2015 and included 35 trials (31,955 children). Iron did not increase the risk of malaria, indicated by fever and the presence of parasites in the blood (high quality evidence). There was no increased risk of death among children treated with iron, although the quality of the evidence for this was low. Among children treated with iron, there was no increased risk of severe malaria (high quality evidence). Although it is hypothesized that iron supplementation might harm children who do not have anaemia living in malarial areas, there is probably no increased risk for malaria in these children (moderate quality evidence). In areas where health services are sufficient to help prevent and treat malaria, giving iron supplements (with or without folic acid) may reduce clinical malaria. In areas where these services are not available, iron supplementation (with or without folic acid) may increase the number of children with clinical malaria (low quality evidence). Overall, iron resulted in fewer anaemic children at follow up, and the end average change in haemoglobin from base line was higher with iron.
Our conclusions are that iron supplementation does not adversely affect children living in malaria-endemic areas. Based on our review, routine iron supplementation should not be withheld from children living in countries where malaria is prevalent and malaria management services are available.
Childhood anaemia is a major, widespread public health problem in sub-Saharan Africa and other low-income areas (WHO 2008a; Kassebaum 2014). The highest prevalence of anaemia is found among children younger than five years of age who are living in low-income countries (Kassebaum 2014). Causes of anaemia in developing countries are numerous and often multifactorial, and include iron deficiency, infectious diseases (such as malaria, intestinal helminths, and schistosomiasis), haemoglobinopathies, and chronic kidney disease (WHO 2011a; Kassebaum 2014).
Iron is an important mineral needed to produce haemoglobin. It is also a component of many enzymes that are essential for proper cell development and cell growth of the brain, muscle, and the immune system (Beard 2001). It is a component of the peroxidase and nitrous oxide-generating enzymes that participate in the immune response to infections and is probably involved in regulating the production and action of cytokines (mediators of immune function released during early stages of infection). Since free iron is toxic to cells, it is stored as ferritin, an intracellular protein.
A relatively large amount of iron is required to produce red blood cells (erythropoiesis) in the first few months after birth. This is usually derived from the iron stored by the foetus in the last months of pregnancy. However, by the time a child is four to six months old, these stores become marginal or depleted. A child whose diet does not provide enough iron risks development of iron-deficiency anaemia. Infants with low total body iron at birth are particularly prone to iron deficiency; this is often exacerbated by the early introduction of cereal-based weaning food from which iron absorption can be as low as 5% (FAO/WHO 2005). Iron deficiency may be worsened by chronic blood loss from the intestines that results from intestinal parasitic infections (Stoltzfus 1997).
Iron deficiency is common and affects approximately two billion people worldwide, which results in over 500 million cases of anaemia (WHO 2004). In most areas, and specifically in all low- and middle-income regions, the most significant contributor to the onset of anaemia is iron deficiency (WHO 2008a; Kassebaum 2014). In sub-Saharan Africa, the prevalence of iron-deficiency anaemia is estimated to be around 60% overall (WHO 2004), with 40% to 50% of all children under five years in developing countries being iron-deficient (UNICEF 1998).
Based on estimates of iron-deficiency anaemia as a risk factor for death, iron deficiency has been estimated to cause 726,000 deaths in the perinatal and childhood periods globally, with the greatest toll in Southeast Asia and in Africa (WHO 2004; FAO/WHO 2005). Experimental and observational studies have linked iron deficiency to adverse effects on child development, including impairments of cognitive, emotional, and motor development (Pollitt 1993; Grantham-McGregor 2001; Gewa 2009), growth (Lawless 1994), immune function, and increased risk of infection (Berger 2000; Beard 2001). The relative risk for mental retardation associated with a 1 g/dL increase in population mean haemoglobin level has been estimated at 0.78 (95% confidence interval (CI) 0.70 to 0.86) (WHO 2004). However, these studies have been criticized for their inability to fully adjust for confounders and to establish causality (Oppenheimer 2001). Systematic reviews of randomized controlled trials (RCTs) on iron supplementation's effect on mental development, intelligence scores, motor development, and growth reported conflicting results (Bhandari 2001; Ramakrishnan 2004; Sachdev 2005; Iannotti 2006; Sachdev 2006; Low 2013; Thompson 2013; Wang 2013). Notably the time frame of many RCTs may not have allowed for a full evaluation of developmental outcomes.
The diagnosis of iron deficiency and iron deficiency anaemia relies mainly on the measurement of a person's haemoglobin, iron, and ferritin (Pasricha 2013). The measurement of haemoglobin alone is not sufficiently sensitive (due to overlapping values in iron-replete and iron-deficient individuals) and is not specific because of the numerous causes of anaemia in developing countries. Ferritin is the most commonly accepted measure of iron status (Mei 2005). However, there is a complex interaction between infection, inflammation (even when subclinical), and ferritin. Infection and inflammation increase ferritin, which is an acute phase reactant. The increase is proportional to the baseline ferritin levels and available iron stores (Thurnham 2010). It decreases only slowly after the resolution of infection and remains elevated in the convalescent phases of infection. Thus, in developing countries it is difficult to interpret ferritin levels and their use as a biomarker of iron deficiency may underestimate the true prevalence of iron deficiency (Nyakeriga 2004; Zimmermann 2005). Other biomarkers or combinations of biomarkers have been suggested for the assessment of iron deficiency in locations with a high prevalence of infection. These include the serum transferrin receptor, zinc protoporphyrin, transferrin saturation, and the ratio of serum transferrin receptor to serum ferritin (Lynch 2011), as well as the adjustment of ferritin to C-reactive protein or alpha1-acid glycoprotein levels, or both (Mburu 2008; Thurnham 2010). The World Health Organization (WHO) and the Centers for Disease Control and Prevention (CDC) recommend the use of concurrent measurements of haemoglobin, ferritin, and transferrin receptor to assess the iron status of a group (WHO/CDC 2004; WHO 2011b). The concurrent measurement of the inflammatory markers C-reactive protein and alpha1-acid glycoprotein facilitates the interpretation of ferritin levels. However, the exclusion of children with elevated markers of inflammation from iron deficiency assessment is not reasonable, since up to 69% of children in malaria-endemic areas may have elevated markers of inflammation (Darboe 2007).
Malaria is a leading cause of morbidity and mortality in children in sub-Saharan Africa (Breman 2001; WHO 2008b). Most infections are caused by the most virulent parasite species, Plasmodium falciparum ( WHO 2008b), which is transmitted to humans by the bite of an infected female Anopheles mosquito. Trends and general patterns of malaria transmission vary greatly geographically. Children are vulnerable to malaria from the age of approximately three months, when immunity acquired from the mother wanes. Malaria is an important contributor to anaemia in endemic regions through the destruction of parasitized red blood cells (haemolysis), increased clearance of infected and uninfected red blood cells by the spleen, cytokine-induced dyserythropoiesis (abnormal production of red blood cells), and probably also decreased absorption of dietary iron (Menendez 2000; Ekvall 2003; Glinz 2015).
There is an ongoing debate on whether iron deficiency offers protection from malaria and whether an excess of iron increases the risk of malaria or severe malaria (Oppenheimer 2001; Stoltzfus 2010; Suchdev 2010; Oppenheimer 2012). Iron is required by many pathogens for their survival and pathogenicity (killing ability) (Beard 2001). Removal of free circulating iron seems to be an important part of the host (human) response to infection. The theory that iron deficiency may be an important defence mechanism has been termed "nutritional immunity" (Kochan 1973). The erythrocytic form of the Plasmodium parasite requires free iron (which is lacking in an iron-deficient person). In one observational study iron deficiency was associated with a small, albeit significant, degree of protection from episodes of clinical malaria in a cohort of young children living on the Kenyan coast (Nyakeriga 2004).
In areas where the prevalence of anaemia is 40% or more in young children, guidelines generally recommend that children of normal birthweight receive oral iron (2 mg/kg/day of elemental iron, daily, for three months) between the ages of six months and two years, and that children with a low birthweight receive the same amount of iron starting at two months (Stoltzfus 1998; INACG 1999).
Several meta-analyses have previously examined the benefits and risks of iron supplementation in children (INACG 1999; Oppenheimer 2001; Gera 2002; Iannotti 2006; Gera 2007). These have shown that iron treatment increases haemoglobin and prevents anaemia. The absolute effects on haemoglobin were larger among children who were anaemic at baseline and smaller in malarial hyperendemic regions compared with non-endemic regions, and with iron-fortified food compared with oral medicinal iron (Gera 2007). An increased risk of malaria has been highlighted by several meta-analyses. Most studies reported parasitaemia (being slide-positive for P. falciparum at the end of supplementation) rather than clinical malaria (INACG 1999; Oppenheimer 2001; Gera 2002; Iannotti 2006). The effects on parasitaemia were associated with baseline rates of parasitaemia (Gera 2002). In Gera 2002 other infections were also assessed. Overall there was no difference in the incidence rate ratio for all recorded infections. Diarrhoea was more frequent in the iron-supplemented group.
In 2006, the results of a large RCT that evaluated the effect of iron and folate supplements in a malaria-endemic area of Zanzibar (Pemba Island) were published (Sazawal 2006 (C)a). The study was terminated prematurely on the recommendation of the study data safety and management board following the higher proportion of hospitalization or death among participants randomized to treatment with iron and folic acid. A subgroup analysis revealed that the risk was limited to children who were iron-replete when iron supplementation was started. This trial heightened global concern about the routine, non-selective iron supplementation policy in areas where malaria is highly prevalent. Before this trial, the WHO guidelines for children living in malaria-endemic areas were no different than the general recommendations (WHO 2003). In 2007 a consultation convened to consider the recommendations for children living in malaria-endemic areas (WHO 2007). The trial's subgroup analysis suggested that it might be necessary to screen for iron deficiency and treat only iron-deficient children. However, such a recommendation is difficult or impossible to implement. There is no consensus on the most appropriate biomarker to assess iron deficiency or monitor iron status during supplementation in regions with a high prevalence of infection (Mburu 2008; Thurnham 2010; Lynch 2011; Pasricha 2013). Furthermore, as a public health intervention, screening of all children before iron supplementation is impractical in most malaria-endemic areas. Thus, it has become critical to examine the safety and effects of iron supplementation in malaria-endemic areas considering all the available evidence. In 2013 another large RCT that evaluated the effect of iron added to micronutrient powder was conducted in Ghana. Insecticide-treated bed nets were provided to all participants and antimalarial treatment was systematically administered when indicated (Zlotkin 2013 (C)). The use of iron in this trial did not result in an increased incidence of malaria among participants.
In view of the newly available data we set to examine the complete evidence in all RCTs that assessed iron supplementation for children in malaria-endemic areas. In a previous version of the review, we did not observe an increased risk of malaria with iron supplementation overall nor was iron harmful in the treatment of malaria (Ojukwu 2007; Ojukwu 2009; Okebe 2011). Since any theoretical harm of iron administration would be expected to occur in the setting of intense malaria transmission, in the current review version we limited our analysis to areas with hyperendemic or holoendemic transmission of malaria or to trials conducted in other areas, but reported malaria-related outcomes. We specifically searched for outcomes related to malaria and data for all-cause mortality, which ultimately combines benefit and harm. Due to the conflicting results of the studies conducted in Pemba Island, Sazawal 2006 (C)a, and in Ghana, Zlotkin 2013 (C), we compared the effect of iron administration on the incidence of clinical malaria in studies in which prevention or management of malaria were offered as an integral part of the study design, with studies in which neither malaria prevention nor malaria management were systematically administered.
To evaluate the effects and safety of iron supplementation, with or without folic acid, in children living in areas with hyperendemic or holoendemic malaria transmission.
Randomized controlled trials (RCTs) that randomized individuals or clusters, and were conducted in countries defined as hyperendemic or holoendemic for malaria (Hay 2004; Table 1) or reported on any malaria-related outcome. We excluded studies if the publication specifically stated, or we obtained information from the study authors, that the study was conducted in an area or period without malaria activity. We considered cluster RCTs eligible only if they included at least two units per trial arm. We excluded trials conducted in hypoendemic and mesoendemic areas, unless they reported malaria-reported outcomes. In addition, we excluded studies that did not report at least one of the review-defined primary or secondary outcomes.
Children (less than 18 years of age), with or without anaemia, and with or without malaria or parasitaemia at baseline. We excluded pregnant women.
We only included trials that allocated antiparasitics or other micronutrients (for example, zinc, vitamin A, vitamin C) if both trial arms received the same dose and schedule. Iron could be administered orally in any form of tablet, elixir, supplementation, or fortification (including fortification of food, drink, sprinkles, or other modes of iron administration as long as it provided at least 80% of the Recommended Dietary Allowance (RDA) recommended by the World Health Organization (WHO) for prevention of anaemia by age (Table 2; Stoltzfus 1998). Eighty per cent of the RDA would approximate the Estimated Average Requirement (EAR), which is the daily intake value of a nutrient that is estimated to meet the nutrient requirement of half the healthy population, by age (Institute of Medicine 1998). Iron could be administered for any duration or interval of administration.
We constructed the following comparisons:
We excluded studies that did not report at least one of the review-defined primary or secondary outcomes.
We attempted to identify all relevant trials regardless of language or publication status (published, unpublished, in press, and in progress).
The Information Specialist of the Cochrane Infectious Diseases Group (CIDG) editorial base, Vittoria Lutje, searched the following databases and used the search terms and strategy described in Table 3: the CIDG Specialized Register; the Cochrane Central Register of Controlled Trials (CENTRAL), published in the Cochrane Library (February 2015); MEDLINE (1966 to August 2015); EMBASE (1980 to February 2015); and LILACS (1982 to February 2015). We also searched the metaRegister of Controlled Trials (mRCT) and World Health Organization International Clinical Trials Registry Platform (WHO ICTRP) using 'iron' and 'malaria' as search terms.
We contacted the primary investigators of all included trials, ongoing trials, and those awaiting assessment to ask for unpublished data and further trials.
We scanned the bibliographies of all included trials, pertinent reviews, and previous meta-analyses for additional references.
Several review authors (Juliana Ojukwu, Dafna Yahav (DY), Joseph Okebe (JO), and Mical Paul (MP) for first version of this Cochrane review (Ojukwu 2009); Rana Shbita (RS), DY, JO, and MP for the second version (Okebe 2011); and Ami Neuberger (AN), JO, and MP for this review update) independently inspected the abstract of each identified reference and obtained the full text of relevant articles. Two review authors independently reviewed the articles and applied the inclusion criteria. If needed, we contacted the study authors to clarify study eligibility. We resolved any areas of disagreement by discussion with a third review author. Each trial was scrutinized to identify multiple publications from the same data set. We documented the justification for exclusion of studies from the review. We named studies by the first author and year of publication (with the addition of a, and b, for different studies from the same author and year of publication). The addition of (C) to the trial's identification denotes that the trial was cluster randomized.
Two review authors independently extracted data into a prepiloted data-extraction spreadsheet which detailed relevant epidemiologic and clinical data. We resolved any differences in the data extracted by discussion. One review author entered data into Review Manager (RevMan) (RevMan 2014).
For individually RCTs, we recorded the number of participants that experienced the event and the number of participants analysed in each treatment group or the effect estimate reported (for example, risk ratio (RR)) for dichotomous outcome measures. For count data, we recorded the number of events and the number of child-months of follow-up in each group. Whenever a trial did not report the number of child-months, we used the product of the duration of follow-up and the number of children evaluated to estimate this figure. For continuous data, we extracted means (arithmetic or geometric) and a measure of variance (standard deviation (SD), standard error (SE), or confidence interval (CI)) and the numbers analysed in each group. We calculated haemoglobin values in g/dL by multiplying hematocrit or packed cell volume values by 0.34, when a trial did not report haemoglobin values.
In cluster RCTs, we recorded the unit of randomization (for example, household, compound, sector, or village), the number of clusters in the trial, and the average cluster size. We documented the statistical methods used to analyse the trial alongside details that described whether these methods adjusted for clustering or other covariates. We extracted estimates of the intracluster correlation coefficient (ICC) for each outcome whenever possible. Where results had been adjusted for clustering, we extracted the treatment effect estimate and the SD or CI. If we did not adjust the results for clustering, we extracted the data reported (see adjustment below).
Two review authors independently assessed the risk of bias. For all included trials we assessed the following.
We graded the generation of randomization sequence and allocation concealment as at either low, high, or unclear risk of bias, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).
For cluster RCTs we also assessed the following.
For dichotomous data we calculated RRs and for continuous data absolute mean differences, with 95% CIs. We computed SDs from SEs or 95% CIs, and assumed a normal distribution of the values. We calculated risk difference for the outcome of all-cause mortality to allow the inclusion of the large number of trials with no deaths in both trial arms in the analysis. For count data, we calculated the rate ratio and SE for each study. We replaced zero events by 0.5. When the original included trials reported covariate-adjusted incidence rate ratios, we used these data with SEs. We analysed infectious episodes, hospitalizations, and clinic visits as count data, and reported rate ratios per child-months. We calculated standardized mean differences for the outcomes of weight and height, since we combined absolute values with weight/height for age Z scores.
We contacted the trial authors if the available data were unclear, missing, or reported in a format that was different from the one required.
We aimed to perform an intention-to-treat analysis, where the trial authors accounted for all randomized participants; otherwise we performed a complete case analysis.
We assessed heterogeneity in the included trials by visual examination of the forest plot to detect non-overlapping CIs, using the Chi² test of heterogeneity (P < 0.1 indicating statistical significance) and the I² statistic of inconsistency (with a value = 50% denoting moderate levels of heterogeneity). When statistical heterogeneity was present, we investigated the reasons for it using subgroup analysis.
We constructed a funnel plot to assess the effect of small trials on the main outcome (when we included more than 10 trials).
We conducted analyses using RevMan (RevMan 2014). We included cluster RCTs in the main analysis after adjustment for clustering (see above). We performed the meta-analysis using the Mantel Haenszel (M-H) fixed-effect model or the generic inverse variance method (when adjustment for clustering was performed by adjusting SEs). Regarding the outcomes of haemoglobin and anaemia, we used a random-effects model where we expected a priori heterogeneity to be displayed due to different mean baseline haemoglobin values and definitions of anaemia in different studies.
We assessed the quality of the evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach and constructed 'Summary of findings' tables using GRADEpro Guideline Development Tool (GDT) (http://www.gradepro.org). We presented 'Summary of findings' tables for the primary outcomes and hospital admissions.
When we detected heterogeneity, we attempted the following subgroup analyses. We defined subgroups by trial (or trial arm) level and not at the level of individual participants, since most trials targeted the participant subgroup of interest as the main study (for example, trials were conducted on anaemic or non-anaemic children, and recruited children within a narrow age range). Moreover, trials most commonly did not present all outcomes for children subgroups.
We primarily stratified analyses by the presence of anaemia at baseline or malaria endemicity (selected by relevance to the outcome assessed), regardless of the presence or absence of heterogeneity to address clinically relevant populations. Similarly, for the outcomes related to malaria we assessed the effects of age. We performed comparisons between subgroups with RevMan (RevMan 2014).
We conducted sensitivity analyses by methods of allocation concealment to assess the effect of risk of bias on primary outcomes. We restricted the analysis of malaria-related outcomes to P. falciparum. When we assessed all malaria species together, we included in this analysis trials where over 85% of malaria spp. diagnosed were P. falciparum. We excluded trials that counted multiple episodes of the outcome in individuals and trials whose outcome assessment occurred at a different point in time from that used in other included trials.
After the filtration of publications that were irrelevant or clearly incompatible with the inclusion criteria, we initially considered in full 120 studies that were conducted in hyperendemic or holoendemic malaria areas or that reported on malaria. Of these we excluded 83 publications for the reasons we have detailed in the 'Characteristics of excluded studies' tables. Overall, we included 52 publications, which represent 35 individual RCTs. Eleven trials were multi-armed, which led to 51 comparisons included in the review. We added five new trials in this current review update (Hop 2005; Giovannini 2006; Thi 2006; Esan 2013; Zlotkin 2013 (C)). We excluded 17 trials that assessed iron for the prevention or treatment of anaemia, which were included in the previous review version (Okebe 2011) from the current review update since they were conducted in hypoendemic or mesoendemic areas for malaria or provided an insufficient iron dose. Similarly, we excluded four trials that assessed iron as part of the treatment of malaria, as we dropped this analysis from the current version of this review.
Where the full publication did not provide enough information, we attempted to contact the authors of included and potentially relevant trials. We requested data primarily on malaria and all-cause mortality. We established correspondence with 26 trial authors, of whom 21 supplied further information.
We have provided a description of the included trials in the 'Characteristics of included studies' tables. The trials were published between the years 1973 and 2013. Overall the included trials recruited 31,955 children: 7953 in 26 individually RCTs and 24,002 (73%) in nine cluster RCTs. The largest cluster RCT included two separate, independent cohorts: the main trial, Sazawal 2006 (C)a, and an independent substudy, Sazawal 2006 (C)b. We only included two arms of this trial in the review (iron, folic acid, and vitamin A versus vitamin A alone), totaling 15,956 children in the main study and 1619 children in the substudy (analysed as separate trials in the review). We included unpublished data supplied by the trial authors on the outcomes of malaria and death from the substudy. Our attempts to obtain data on the two trial arms that compared iron and folic acid and vitamin A and zinc versus zinc and vitamin A (Sazawal 2007) at the time the iron trial arm had been stopped (August 2014) were unsuccessful. This analysis could have been a major contribution to the evidence. Twenty trials reported adherence, and the average overall adherence to all trial drugs was good (89%).
All trials assessed the administration of iron or iron plus folic acid for the prevention or treatment of anaemia among children without an acute illness. The mean iron supplementation dose was 2 mg/kg/day, and the mean duration of treatment was 4.5 months (one to 12 months). Twenty-seven trials added the antimalarial treatment to the iron arm or both trial arms, 12 trials added anthelminthics to both trial arms, and eight trials added micronutrients to both trial arms. Twenty-three trials reported one or more of the review-defined malaria-related outcomes (65.7% of trials included in the review. We have described the types of outcomes and their definitions in Table 4. Severe malaria, as defined per protocol, was reported in a single trial and its substudy that reported on cerebral malaria (Sazawal 2006 (C)a; Sazawal 2006 (C)b). Five trials reported clinical malaria with high-grade parasitaemia (Smith 1989 (C); Adam 1997 (C); Massaga 2003; Ayoya 2009; Zlotkin 2013 (C)). Twelve trials reported only or mostly (over 80%) on P. falciparum malaria (Table 4). Most trials that reported malaria-related outcomes performed regular surveillance for malaria using blood smears at baseline and during treatment (either at regular intervals or whenever children were febrile), and offered trial participants treatment when they were symptomatic (Table 4). Notably, no surveillance or treatment outside the hospital was offered in the main trial (Sazawal 2006 (C)a, unlike its substudy (Sazawal 2006 (C)b), where monitoring was performed and treatment was offered to children at their home. The baseline rate of malaria parasitaemia (reported in 11 of 19 trials) ranged from 0% to 70% of children (mean 45%). The mean baseline haemoglobin was lower than 10 g/dL in 17 trials (when iron was most commonly administered for the treatment of anaemia) and 10 g/dL or higher in 22 (when iron was administered for the prevention of anaemia). The trial population consisted of children aged less than two years of age in 12 trials, two to five years of age in 11 trials, and over five years of age in 16 trials. The respective number of trials that reported on malaria-related outcomes in the three age groups were eight, nine, and six trials.
We have detailed the specific reasons for exclusion in the 'Characteristics of excluded studies' tables. The major reasons for exclusion of studies were the following.
We judged 20 of the 35 trials (57.1%) as at low risk of bias related to allocation concealment. One trial was at a high risk of bias (Smith 1989 (C)). All the remaining trials either did not describe their methods clearly or did not provide a description of them. We judged the generation of randomization sequence to be at low risk of bias in 26/35 (74.3%) trials, at a high risk for bias in the one trial using alternation, and at unclear risk in all the others. Overall, we considered the allocation procedure (both allocation concealment and generation) as at low risk of bias in 18 (51.4%) trials.
Twenty-seven trials out of 35 trials (77.1%) described double-blinding or stated that the trial was double-blind, but gave no description of the blinding techniques; we considered all of these trials to be at low risk of bias (see the 'Risk of bias' tables). We considered six trials to be at high risk, and three trials at an unclear risk of bias.
The included trials explained the reasons for exclusion of participants from malaria-related and haemoglobin outcome reporting and related it to an inability to obtain blood samples from the participants. The reasons for the exclusion of randomized children from mortality assessments were unclear and we considered this to be a serious risk for bias since deaths could have occurred among the excluded children. We have provided the details of the number of participants randomized and evaluated in the 'Risk of bias' tables.
We did not have access to protocols to compare planned outcomes with those reported in the final publication.
One trial specified methods for assessment of malaria throughout the trial (without defining these as study outcomes), but did not report the results per trial arm (Olsen 2006). We could not contact the authors of this trial. We contacted authors of trials that did not report on malaria in their methods or result sections; the authors of one trial reported that data had been collected in the trial but were no longer available (Powers 1983), while the authors of seven trials replied that malaria-related outcomes were not collected in their trials (Greisen 1986 (C); Latham 1990; Dossa 2001a; Dossa 2001b; Hall 2002 (C); Hess 2002; Zlotkin 2003). Thus, selective reporting bias is unlikely with regard to the outcomes related to malaria.
Only one trial defined death as an outcome (Sazawal 2006 (C)a; Sazawal 2006 (C)b), although 16 trials reported these results and we obtained these results from authors of another 14 trials. Thus, it is not possible to discuss reporting bias in relation to the outcome of mortality.
Nine of the included RCTs were cluster randomized.
See: Summary of findings for the main comparison Oral iron versus placebo or no treatment for children in malaria-endemic areas; Summary of findings 2 Effects of oral iron with or without folic acid on malaria among children in malaria-endemic areas; Summary of findings 3 Oral iron with antimalarial prophylaxis versus placebo or no treatment for children in malaria-endemic areas
We included trials or trial arms that compared iron plus antimalarial versus antimalarial alone in this comparison. We did not separate the comparisons of iron versus placebo/no treatment and iron plus antimalarial versus antimalarial, unless there was significant heterogeneity in results explained by this factor. Four-armed trials including both comparisons appear twice in the analyses (Menendez 1997; Verhoef 2002; Massaga 2003), once for the comparison of iron versus placebo/no treatment and once for the comparison of iron plus antimalarial versus antimalarial (each arm included different children).
All trials defined clinical malaria as fever (usually greater than 37.5°C) and parasitaemia (any density). Overall, there was no significant difference in the risk ratio (RR) for clinical malaria between iron treatment and placebo or no treatment, with a trend in favour of iron treatment (RR 0.93, 95% CI 0.87 to 1.00; 14 trials, 7168 children). Specifically, there was no difference in the risk for clinical malaria between iron treatment versus placebo or no treatment in the subgroup of trials including non-anaemic children (RR 0.97, 95% CI 0.86 to 1.09; five trials, 2112 children) without heterogeneity, Analysis 1.1). Significant heterogeneity was present in the subgroup analysis with anaemia (P = 0.01, I² statistic = 56%) and in the overall analysis. We rated the quality of the evidence as high for the overall analysis and moderate for the subgroups of anaemic and non-anaemic children (Summary of findings for the main comparison).
The RR of malaria was lower for children younger than two years of age (RR 0.89, 95% CI 0.82 to 0.97), and this result was largely driven by the recent study conducted in Ghana (Zlotkin 2013 (C)). In children aged between two and five years, and among children older than five years, we did not observe any effect of iron treatment on the relative risk of clinical malaria (Analysis 1.2).
The results favoured iron treatment when we only included trials that described malaria caused solely or primarily by P. falciparum (RR 0.91, 95% CI 0.84 to 0.99; nine trials, 5503 children, Analysis 1.3). Similarly, in six trials that reported malaria with high-grade parasitaemia as an outcome, the RR was lower among children treated with iron when compared to children receiving placebo or no treatment (RR 0.90, 95% CI 0.81 to 0.98; Analysis 1.6) and we rated the quality of this evidence as high (Summary of findings for the main comparison).
Ten of the 17 trials included in the analysis were at low risk of bias with respect to allocation concealment, and all but two of the trials were double blinded. Sensitivity analysis restricted to trials at low risk of bias for allocation concealment did not affect results. The funnel plot was asymmetrical, which indicated that small studies that favour iron could be missing (Figure 3). Three trials reported either on episodes of malaria (Richard 2006; Leenstra 2009) or clinic visits for malaria (Smith 1989 (C)) rather than participants with their first or only episode. The exclusion of these trials did not affect the pooled RR for this comparison. One trial reported on children with clinical malaria only at end of follow-up, which was six months after completion of iron supplementation (Menendez 1997). Its exclusion did not affect the results.
15/17 trials reported mortality, and in most trials no deaths occurred among the evaluable children (control event rate 1%). Overall, there was no difference between the iron and placebo/no treatment groups, without heterogeneity (Analysis 1.5). We assessed the quality of the evidence for this outcome as low due to imprecision (very low event rate) and incomplete outcome assessment in most of the trials (a drop-out range of 2% to 62% of participants), as deaths might have occurred among the people lost to follow up (Summary of findings for the main comparison).
There was no statistically significant difference in the prevalence of parasitaemia of any level, with a trend favouring placebo or no treatment (RR 1.11, 95% CI 1.0 to 1.23; nine trials, 3393 children, without significant heterogeneity;Analysis 1.4). We converted odds ratios to RRs to allow for the use of data on parasitaemia from one trial (Mebrahtu 2004 (C)). This outcome was not affected by age, anaemia at baseline, or Plasmodium species assessed (Analysis 1.4; Analysis 1.7; Analysis 1.8). Despite the lack of heterogeneity overall, there was a significant difference between trials that described adequate allocation concealment (RR 0.98, 95% CI 0.83 to 1.15, four trials, 1727 children, Analysis 1.9) and trials with unclear or inadequate methods that showed significantly higher rate of parasitaemia with iron treatment (RR 1.22, 95% CI 1.06 to 1.40, five trials, 1666 children; P = 0.04 for the difference between subgroups; Analysis 1.9). All of the trials included in the comparison of parasitaemia were double blinded. There was no statistically significant difference in the occurrence of high-grade asymptomatic parasitaemia, most commonly defined as = 5000 parasites/μL (RR 1.13, 95% CI 0.93 to 1.37; five trials, 2565 children Analysis 1.10). In trials that continued follow-up after the cessation of iron administration, there was a higher prevalence of any level of parasitaemia at the end of follow-up among participants treated with iron (RR 1.23, 95% CI 1.09 to 1.40; five trials, 1150 children; Analysis 1.11).
It was difficult to establish whether the trials reported on children with parasitaemia or on parasitaemia episodes. Gebreselassie 1996 clearly reported on cumulative incidence and Leenstra 2009 included repeated episodes. Leenstra 2009 reported incidence rate ratios with 95% CIs adjusted for age, baseline parasitaemia, and school. We used these in our analysis as relative risks. The exclusion of these trials did not affect the results.
The trials reported parasite density differently, with differences that referred both to the unit of measurement and the denominator (Table 7). A meta-analysis was therefore not possible; we have shown the results in Table 7 for each trial. Qualitatively, parasite density was higher in the iron supplemented group in four trials, lower in one, and similar in one of the six trials that reported on parasite density at the end of treatment.
Six trials reported hospitalizations or clinic visits. Overall, there was no difference between iron and placebo or no treatment (Analysis 1.12); these results were not affected by the administration of antimalarial medications. We assessed the quality of the evidence as very low due to inconsistency and indirectness of this outcome ('Summary of findings' table 1).
Analyses for haemoglobin were highly heterogeneous, since the absolute magnitude of treatment effect in trials differed, and the 95% CIs were narrow. However, the heterogeneity stemmed from different magnitudes of increase in haemoglobin with iron supplementation and not in the direction of the result. Overall, at the end of treatment there was a mean difference of haemoglobin level of 0.75 g/dL (95% CI 0.48 to 1.01; 16 trials, 5261 children; I² statistic = 93%; Analysis 1.13). In trials with anaemic children at baseline, the children gained 0.95 g/dL haemoglobin (95% CI 0.38 to 1.51; seven trials, 2481 participants) with iron supplementation, while in trials including mostly children without anaemia the end haemoglobin was higher than control by 0.61 g/dL (95% CI 0.38 to 0.85; nine trials, 2780 participants; P = 0.28 for subgroup difference). The mean change of haemoglobin from the baseline at end of treatment was 0.67 g/dL (95% CI 0.42 to 0.92; 12 trials, 2462 children; I² statistic = 82%; Analysis 1.19).
The RR for anaemia at the end of treatment, as defined in the trial, was 0.63 (95% CI 0.49 to 0.82; 15 trials, 3784 children; Analysis 1.15), with similar substantial heterogeneity mainly in the magnitude of benefit.
We did not observe any differences when we analysed the effect of iron on haemoglobin by age groups, or by the addition of antimalarial medications or multinutrients to both study arms (analyses not shown). Heterogeneity was maintained in all these subgroup analyses.
Six studies provided data on respiratory infections. There was no difference between iron and placebo overall (rate ratio 0.99, 95% CI 0.85 to 1.15; six trials, 21,767 child-months; I² statistic = 0%; Analysis 1.20). Trials usually reported diarrhoea as 'infectious diarrhoea', although we could not clearly differentiate the symptoms from diarrhoea related to iron or iron/zinc supplementation. We stratified this analysis by zinc co-administration (Analysis 1.17). Overall, treatment was associated with an increased risk of diarrhoea (rate ratio 1.15, 95% CI 1.06 to 1.26; eight trials, 23,912 child-months; I² statistic = 40%). This association was driven by the effect of the iron-zinc combination (rate ratio 1.29, 95% CI 1.15 to 1.44; three trials, 6346 children), and not by iron treatment alone (rate ratio 0.99, 95% CI 0.87 to 1.13; seven trials, 17566 children). Six trials reported the number of febrile episodes; there was no difference between iron treatment and control arm (RR 1.03, 95% CI 0.93 to 1.14; 15531 children). One trial reported more days with fever among iron-treated participants, and one trial reported more infectious episodes among participants randomized to the iron treatment arm. Definitions and reporting methods were highly variable; results are shown per outcome (Analysis 1.18).
Results for height and weight were inconsistently reported as end values or as the change from baseline and absolute values or z-scores matched for age, height, or weight. The analyses shown are based on absolute weight in kg and height in cm at the end of treatment or weight/height for age z-scores. For end of treatment weight, there was no statistically significant difference between iron and placebo/no treatment study arms (Analysis 1.14), while the change from baseline favoured iron (Analysis 1.16). The latter analysis was heterogenous and the heterogeneity was not explained by the review-defined subgroups. There were no significant differences in end values or change from baseline of height, with similar heterogeneity in the change from baseline analysis (Analysis 1.21; Analysis 1.22).
The only trial that reported on malaria-related outcomes was the Pemba study (Sazawal 2006 (C)a; Sazawal 2006 (C)b). Malaria-related outcomes reported were admissions for malaria (Analysis 2.1) and cerebral malaria (Analysis 2.2). The results for the main study and the substudy were significantly different, and the main study showed a higher risk for severe malaria with iron plus folic acid and the substudy showed a lower risk. Therefore we did not pool the results. Children in the substudy were older than children in the main trial (mean age of 22.5 versus 18.3 months) and the baseline haemoglobin for the substudy cohort was probably higher than that of the main trial, because children with severe anaemia (haemoglobin < 7 g/dL) were excluded only from the substudy. However the main difference, as described by the trial authors, was that children in the substudy were monitored and offered treatment for malaria at home throughout the trial period. In the main trial, Sazawal 2006 (C)a, there was no organized infrastructure for the diagnosis and treatment of uncomplicated malaria.
The pooled risk difference for mortality was 0.00 per 1000 children (95% CI −0.00 to 0.01; five trials, 18,034 children, Analysis 2.3), and Sazawal 2006 (C)a contributed 88.4% of the weight of this analysis.
Hospital admissions were reported only in Sazawal 2006 (C)a (RR 1.08, 95% CI 0.96 to 1.22; 22,959 children). Haemoglobin at end of treatment was similarly reported in a single study, Giovannini 2006, and was higher with iron and folate, mean difference 0.90 g/dL (95% CI 0.51 to 1.29; 124 children Analysis 2.5). The RR for anaemia at end of treatment was 0.49 (0.25 to 0.99; three trials, 633 children, Analysis 2.6). No consistent data were reported for respiratory infections, other febrile episodes, and diarrhoea. There were no significant differences in the absolute end values of weight (Analysis 2.7) and height (Analysis 2.8).
We did not create any 'Summary of findings' tables for this comparison, as malaria-related outcome were based on a single trial. The risk of bias for the trial was low, except for the fact that it was discontinued for harm (Sazawal 2006 (C)a; Sazawal 2006 (C)b).
We analysed a single outcome for this comparison to enable the compilation of malaria in all trials that administered iron versus placebo or no treatment. We subgrouped the analysis by the presence of malaria prevention and management services in the trial setting and showed a significant lower risk of malaria with iron with or without folic acid when services were present (RR 0.91, 95% CI 0.84 to 0.97; seven trials, 5586 children), and a significantly higher risk of malaria when such services were absent (RR 1.16, 95% CI 1.02 to 1.31; nine trials, 19,086 children, analysis 3.1, P < 0.001 for subgroup difference). We assessed the quality of the evidence as low for both subgroups, due to inconsistency and suspected publication bias when malaria prevention and management services were present and indirectness in the analysis without malaria prevention and management services, as the latter was dominated was the Sazawal 2006 (C)a study that assessed only admissions due to malaria rather than all events of malaria ('Summary of findings' table 2).
Three trials reported on clinical malaria and all were individually RCTs. The trials uniformly showed that the intervention was protective for clinical malaria (pooled RR 0.54, 95% CI 0.43 to 0.67; three trials, 728 children, I² statistic = 0%; Analysis 4.1), high quality evidence ('Summary of findings' table 3). These trials did not assess severe malaria.
There was no difference in the risk of death for the three trials combined (RR 1.05, 95% CI 0.52 to 2.11; three trials, 728 participants, Analysis 4.2). The quality of the evidence was low due to imprecision ('Summary of findings' table 3).
Both the number of hospitalizations and the number of clinic visits were significantly reduced in two trials (Analysis 4.3). Iron plus antimalarial significantly improved haemoglobin in one trial and decreased the prevalence of anaemia in two trials (Analysis 4.4; Analysis 4.5). Respiratory infections, diarrhoea, and other infections were not reported in these studies.
The trials included in this comparison were four-armed trials that assessed iron, placebo, iron with an antimalarial, and an antimalarial alone. The comparison of antimalarial treatment alone versus placebo showed identical results to the comparison of iron plus antimalarial versus placebo, except for the outcomes of haemoglobin/anaemia where the addition of iron conferred a higher benefit (analysis not shown). This observation strengthens the lack of effect of iron on malaria or other adverse outcome.
Oral iron supplementation alone did not increase the risk for clinical malaria, and the upper level of the 95% CI indicated no harm (risk ratio (RR) 0.93, 95% CI 0.87 to 1.00). The risk was probably not increased in trials that recruited children who were not anaemic at baseline (RR 0.97, 95% CI 0.86 to 1.09). There was no increased risk of P. falciparum malaria or malaria with high-grade parasitaemia, and pooled results favoured iron treatment. The combination of iron and folate resulted in a higher rate of admissions for malaria and cerebral malaria in one large trial, Sazawal 2006 (C)a, but a substudy pointed at the opposite direction. The difference in malaria surveillance and management strategies in the two parts of this trial led to an analysis of all iron supplementation trials by malaria management infrastructure. Overall, iron treatment, with or without folate, was safe and may be protective in settings where malaria prevention or management programmes were implemented (RR 0.91, 95% CI 0.84 to 0.97). Conversley, it might be associated with an increased risk of clinical malaria when neither prevention nor antimalarial treatment were available (RR 1.16, 95% CI 1.02 to 1.31; P < 0.001 for the difference between subgroups). This finding agrees with the analysis of iron with an antimalarial drug as the intervention, which significantly decreased the occurrence of clinical malaria compared to placebo or no treatment. In all analyses, mortality was low in general and not different between iron-treated participants, and those that received placebo or no treatment.
There were no effects of iron administration alone on infections, hospitalizations and clinic visits, and parasitaemia at the end of treatment. Iron treatment alone resulted in higher rates of parasitaemia at the end of follow-up in trials with unclear allocation concealment methods. The combination of iron and antimalarial medication resulted in fewer clinic visits and hospitalizations, which probably reflected the effect of the antimalarial drug. The combination of zinc and iron was associated with more episodes of diarrhoea, while iron monotherapy was not. In all comparisons, iron supplementation increased haemoglobin and decreased anaemia. All analyses were highly heterogenous, which precluded a precise pooled estimate of effect, but the increase in haemoglobin was substantial in most individual trials and the prevalence of anaemia was reduced by 40% to 50% in most comparisons. We did not observe a clear beneficial or adverse effect of iron supplementation on weight or height.
In this current review update we applied more stringent inclusion criteria to focus on the question of iron's safety and beneficial effect in areas where these are doubtful. Firstly, since the debate that concerned iron administration centred around its effect in areas with significant burden of malaria, we chose to include only trials conducted in hyperendemic or holoendemic areas for malaria transmission, or trials that reported malaria outcomes. Secondly, we included trials of iron fortification and for all included trials we excluded trials that administered low-dose iron that is not expected to result in iron repletion. Of the five trials that we added to the current review update, one large trial had a large impact on the pooled malaria outcome results (Zlotkin 2013 (C)). Thirdly, we omitted from this update an analysis of iron administration during proven malaria episodes. We included four trials in the 2011 version of this review, Okebe 2011, that assessed this intervention and no new trials have been published in recent years.
In the 2011 version of this review, Okebe 2011, we deemed that iron treatment not harmful regarding clinical malaria. In the current analysis, we found a trend towards a favourable effect of iron treatment on clinical malaria, with a statistically significant benefit in several subgroup analyses: children younger than two years of age, children infected primarily or exclusively by P. falciparum, and children included in trials in which malaria prevention or management strategies were implemented as an integral part of the trial's design. These results strengthen our conclusions from the previous versions of this Cochrane review (Ojukwu 2009; Okebe 2011); iron administration for the prevention or treatment of anaemia is safe in malaria-endemic areas if malaria is adequately prevented, diagnosed, and treated.
We have provided the results and the quality of evidence assessments in the 'Summary of findings' tables. There were several main reasons that led to downgrading of the evidence. We downgraded the quality of the evidence for imprecision whenever the 95% confidence intervals (CIs) showed possible harm with iron. Some of the important patient-relevant outcomes assessed, including hospital admissions, clinical visits, and deaths, can obviously be triggered by causes other than malaria and the difference between arms is probably not only explained by malaria. Thus, we downgraded these outcomes for indirectness. For deaths, a serious concern of missing data with or without publication bias led to downgrading of the evidence; mortality data were reported in 26/39 of the trials included in the analyses and most of the trials that reported on mortality referred only to children available for analysis at the end of treatment or follow-up. Deaths should be assessed among all children randomized, mainly those lost to follow-up. Inconsistency was apparent in the outcome definitions, mainly with the large Pemba trial that reported only on malaria requiring hospital admission, while all other trials reported on fever with parasitaemia (Sazawal 2006 (C)a; Sazawal 2006 (C)b). Some degree of imprecision might exist in the latter outcome, as asymptomatic parasitaemia is common in malaria-endemic regions and fever with parasitaemia may be caused by infections other than malaria. Funnel plot asymmetry was present in the analyses of iron versus placebo or no treatment, with an excess of small trials that favoured the control arm. However, we did not consider this as a reflection of publication bias since malaria was not the primary outcome assessed in most trials (specifically, it was not the primary outcome in all small trials) and the direction of the small trials' effect was against the safety of iron.
For the intervention of iron with folic acid, the large Pemba trial reported a significantly increased risk for death or hospital admission (Sazawal 2006 (C)a; Sazawal 2006 (C)b), but the relative contributions of the addition of folic acid and the poor infrastructure for diagnosis and surveillance of malaria to the adverse effects of iron in this trial are unclear. Thus, data on the safety of iron and folate supplementation are unclear.
There is an interest in the assessment of the effect of iron supplementation on malaria by iron status and iron-deficiency anaemia at baseline. We could not conduct a subgroup analysis by individual children's iron status or haemoglobin at baseline due to the lack of subgroup data reported in the included trials for the outcomes of deaths and malaria. Our analyses stratified by anaemia are based on the study groups' mean haemoglobin level. Most of the trials recruited a uniform population of anaemic or non-anaemic children, and thus enabled this stratification in the meta-analysis. Although this analysis could mask an adverse effect in individual iron-replete, non-anaemic children compensated by benefit in iron-deplete, anaemic children, the lack of heterogeneity in the analyses for malaria and deaths makes this possibility unlikely. This possibility is not supported by a trial-level stratified analysis restricted to trials that recruited only anaemic or non-anaemic children (data not shown). However, we could not conduct an analysis by true iron status or iron-deficiency anaemia at baseline. Since all trials of non-anaemic children had malaria prevention and management services, we could not assess iron supplementation in non-anaemic populations where malaria prevention and management services were not present.
There was heterogeneity regarding the management of children identified as anaemic during the trial and after its conclusion. Some trials supplied iron to all children identified as anaemic below a certain threshold during the trial. After treatment, during a follow-up period, children who remained anaemic were all given iron per protocol, were offered the possibility of treatment, or the issue was not addressed specifically. We could not assess the effects of this variable due to the large heterogeneity in trial protocols and poor reporting. This factor could underlie some of the unexplained heterogeneity observed in our analyses for haemoglobin and anaemia.
Much of the evidence relies on cluster RCTs. Naturally, these were the largest trials and thus carried a large weight in the meta-analysis. However, the major outcomes assessed in these trials are probably correlated within clusters, including anaemia, iron status, malaria, and other infectious complications. In classes or schools, the correlation between individuals may be smaller than among families, but the large cluster size increases the cluster effect. Ideally, we would want these trials to be planned and analyzed accordingly. The trial should report on the unit of randomization, the average cluster size (number of children in household or class), the number of clusters and individuals randomized, the intracluster correlation coefficient (ICC) value for each outcome (denoting the degree of similarity between individuals in the same cluster) and an effect estimate adjusted for clustering. Unadjusted effect estimates, calculated as if the trial was individually randomized, may result in an exaggerated precision of the effect estimate, thus inflating the weight of the trial in the meta-analysis. Out of the nine cluster RCTs included in our review, only Sazawal 2006 (C)a, Sazawal 2006 (C)b, and Zlotkin 2013 (C) reported adjusted analyses for the primary outcomes. In our analyses we used estimated ICCs to adjust the weight of the cluster RCTs in the meta-analysis. We cannot be sure that the contribution of these trials to the compiled analysis is correct.
Other analyses or outcomes lacking from our Cochrane review include an assessment of the effect of children's nutritional status at baseline on the results; analyses stratified by the schedule of iron supplementation (daily versus weekly); psychomotor and cognitive outcomes assessed in another Cochrane review (Wang 2013); tuberculosis, and age/weight/height-adjusted Z scores for growth. Finally, we do not have the data from Sazawal 2006 (C)a on malaria-related events, hospital admissions, and deaths for the two trial arms: iron plus folic acid plus vitamin A plus zinc versus zinc plus vitamin A. These data could add 16,196 more children to the analyses of iron plus folic acid versus placebo treatment. Similarly, we cannot exclude the existence of more unpublished RCTs that could contribute to the evidence on iron supplementation and malaria, such as those identified in PhD theses (Gebreselassie 1996; Adam 1997 (C)) or others (Roschnik 2003 (C)).
Sazawal 2006 (C)a is by far the largest trial to date, and probably the only trial to date, powered to assess the effect of iron supplementation on severe malaria. It showed a significantly increased risk for the composite outcome of death or hospitalization among children, most of whom who were younger than two years of age. The risk of death was increased, but without statistical significance. In a stratified analysis of an independent substudy, we observed adverse effects of iron supplementation among children who were iron-replete and non-anaemic at baseline, while among children who were iron-deficient and anaemic there were fewer adverse events. At the time there was no malaria control programme in Zanzibar, and the main trial's protocol did not offer children special malaria prevention or management services. In the substudy, Sazawal 2006 (C)b, surveillance for parasitaemia was performed and children received treatment according to the study protocol in their homes. The main trial showed that iron is harmful, while the substudy showed that iron supplementation is protective for severe malaria. Our analysis, which included all trials that assessed iron supplementation, concords with this conclusion (Analysis 3.1).
In 2007, based on Sazawal 2006 (C)a, the World Health Organization (WHO) released a statement that recommended that iron supplementation should be prescribed only after screening for iron deficiency (WHO 2007). Universal screening for anaemia and iron status places a significant burden on healthcare systems in low-income countries. This Cochrane review was originally published in 2009, Ojukwu 2009, and added further debate into the conundrum of iron supplementation for children in malaria-endemic areas (Roth 2010; Stoltzfus 2010; Suchdev 2010; Oppenheimer 2012). The previous versions of this systematic review, Ojukwu 2009 and Okebe 2011, challenged the recommendations for universal screening, and placed the Sazawal 2006 (C)a trial in the context of the complete evidence. In 2011 the WHO amended its recommendation to state that “In malaria-endemic areas, the provision of iron should be implemented in conjunction with measures to prevent, diagnose and treat malaria" (WHO 2011c). In 2013 the results of the Zlotkin 2013 (C) trial became available, with no harm observed for iron-treated children. Children included in the latter trial underwent initial screening for malaria, received insecticide-treated bed nets, and were provided with first-line antimalarial treatment if ill. The results of the Zlotkin 2013 (C) trial and this current analysis further shift the emphasis of decision making prior to iron supplementation on malaria prevention and management rather than on the assessment of iron status.
We did not find an increased risk of clinical malaria and parasitaemia, all-cause mortality, or other infectious complications with iron supplementation alone for children living in areas with intense malaria transmission. Subgroup analyses did not point at an increased risk for these outcomes in children that were non-anaemic at baseline and in children younger than two years of age. Overall, iron supplementation may be associated with an increased risk of malaria in settings with no access to malaria prevention or management services, but is safe when such services are available. In such circumstances, administration of iron with an antimalarial drug confers significant protection from malaria, and probably reflects the effect of the antimalarial drugs.
Iron supplementation significantly improves haemoglobin levels and reduces the prevalence of anaemia in highly malaria-endemic areas. Universal screening for iron deficiency and anaemia can select the population most likely to benefit from iron administration, but such screening programmes are not currently feasible in most areas with intense malaria transmission.
Based on our Cochrane review, iron supplementation should not be withheld from children that live in malaria-endemic countries. The new data added to this meta-analysis since 2011 significantly strengthen this recommendation. Malaria prevention and management should be offered to children regardless of iron supplementation, since these interventions reduce malaria, mortality, and anaemia (Meremikwu 2008). Improvements in prevention and management of malaria have occurred in the last decade in sub-Saharan Africa, allowing for iron supplementation in safer settings than ever before (WHO GMP 2010; MDG 2011; Murray 2014).
There are not enough data to draw conclusions on the intervention of iron with folic acid. Folic acid may interfere with the efficacy of sulphadoxine-pyrimethamine, an antimalarial drug used for intermittent preventive treatment or treatment of clinical episodes of malaria (Mulenga 2006; Metz 2007). Furthermore, there is no evidence of folate deficiency among children under two years old in malaria-endemic areas (Metz 2007).
There is perhaps a remaining uncertainty whether iron supplementation alone results in an increased risk of malaria in a subset of iron-replete, non-anaemic children living in highly malaria-endemic areas. To address this question, an individual patient data (IPD) meta-analysis of existing trials might be of value. Such an analysis will allow a better assessment of the covariates of interest than the trial-level analysis presented herein. These include participants' iron and haemoglobin status at baseline (most trials reported baseline assessment of one or measures of iron status and haemoglobin), and more precise age stratification to address the main group of interest, that of children between six months and two years of age.
Well-conducted observational studies that assess the effects of iron supplementation and fortification (Stoltzfus 2011) are important since RCTs measure only a limited duration of iron supplementation and may not represent the child population in need of iron supplementation. Growth and developmental outcomes would probably be better assessed in such long-term studies.
Further studies should establish optimal malaria prevention and management programmes to prevent harm during iron administration.
The editorial base for the Cochrane Infectious Diseases Group (CIDG) is funded by the UK Department for International Development for the benefit of developing countries. Dafna Yahav (DY) received a grant from the Research Programme Consortium funded by the UK Department for International Development to complete the 2014 update. Ami Neuberger (AN) received funding from the Department of Nutrition for Health and Development, World Health Organization for the 2016 update.
The academic editor for this review is Professor Paul Garner.
We acknowledge the assistance and contribution to the review of Juliana U Ojukwu (JUO) (conceived the idea for the review, wrote the protocol, and contributed to the first edition of this review); Sarah Donegan (who advised and assisted with data analysis); Harriet G. McLehose (who assisted in the writing and final drafting of the protocol and first edition of the review); Paul Garner (who assisted with the study design, analysis, and co-ordination) and Leonard Leibovici (who assisted with data analysis and interpretation). Juliana U Ojukwu was awarded a Reviews for Africa Programme Fellowship (www.mrc.ac.za/cochrane/rap.htm), funded by a grant from the Nuffield Commonwealth Programme, through The Nuffield Foundation during the preparation of the first edition of this review.
We thank the CIDG for their support of this review, without which we could not have performed this work. Special thanks to Vittoria Lutje, Information Specialist of the CIDG, who designed and performed the searches. We thank Professor Jimmy Volmink, Dr Taryn Young, and Karen Essex of the South African Cochrane Centre for inviting us to meet and work on this review.
We thank Professor Sunil Sazawal for supplying unpublished data for the Sazawal 2006 (C)b substudy and Professors HP Sachdev and Tarun Gera for their help obtaining unpublished data for the trials included in this Cochrane review. We thank all the trial authors who responded to our requests for further data and provided the data where available (see the 'Characteristics of included studies' section). We thank the reviewers of our manuscript who raised important issues related to the analyses presented. We urge all the authors of the included studies to correct the data used in their studies, and if necessary, add data, if available, mainly for the primary outcomes of malaria and mortality, and to point out any other inaccuracies in our analyses.
Based on other trials included in this Cochrane review we assumed an average cluster size of 1.5 for households and 32 for classes, when the average cluster size or number of clusters and individuals were not reported (see 'Characteristics of included studies' tables, Table 5 and Table 6 for reported and assumed cluster sizes). The design effects (DEs) or intracluster correlation coefficients (ICCs) used for the different outcomes were the following.
JUO conceived the idea for the review, wrote the protocol, identified studies for inclusion and exclusion, extracted the data, entered the data in RevMan (RevMan 2014), participated in the data analysis, and reviewed all the drafts of the first edition of this review.
Joseph Okebe (JO) wrote the protocol, identified studies for inclusion and exclusion, extracted the data, entered data in RevMan (RevMan 2014), participated in the data analysis, and reviewed all the drafts and the final review.
DY extracted the data from all included studies, entered data in RevMan (RevMan 2014), participated in the data analysis, and reviewed all drafts and the final review.
Mical Paul (MP) planned the data extraction, extracted the data, entered data in RevMan (RevMan 2014), participated in the data analysis, and wrote the review.
All review authors approved the final publication.
MP updated the RevMan file (RevMan 2014), reorganized previous data, carried out the subgroup analyses, performed the GRADE classifications, wrote the final version of the update, and revised the final manuscript.
Rana Shbita (RS) performed the search for the 2011 update and identified studies for inclusion and exclusion, extracted the data for the new studies, participated in the data analysis, and reviewed all the drafts and the final review. This work was performed in partial fulfilment of RS Master in Epidemiology,
DY performed the search for the 2011 update, extracted the data for the new studies, performed the GRADE classifications, and revised the final manuscript.
AM identified studies for inclusion and exclusion, extracted the data for the new studies, entered data in RevMan (RevMan 2014), performed data analysis, and wrote the final version of the update.
MP updated the RevMan file (RevMan 2014), reorganized previous data, carried out the subgroup analyses, performed the GRADE classifications, co-wrote the final version of the update, and revised the final manuscript.
DY and JO performed the search for the 2015 update, and revised the final manuscript.
All review authors listed approved the revised publication.
AM has no known conflicts of interest.JO has no known conflicts of interest.DY has no known conflicts of interest.MP has no known conflicts of interest.
*Endemic Diseases; Anemia, Iron-Deficiency [etiology;*prevention & control]; Antimalarials [administration & dosage]; Dietary Supplements [adverse effects]; Folic Acid [adverse effects]; Iron [*administration & dosage; adverse effects];Malaria [chemically induced;*complications]; Parasitemia [chemically induced; complications]; Randomized Controlled Trials as Topic
Adolescent; Child; Child, Preschool; Humans
References to studies included in this review