Effect of iron supplementation on incidence of infectious illness in children: systematic reviewBMJ 2002; 325 doi: http://dx.doi.org/10.1136/bmj.325.7373.1142 (Published 16 November 2002) Cite this as: BMJ 2002;325:1142
- Tarun Gera, senior resident,
- H P S Sachdev, professor. ()
- Division of Clinical Epidemiology, Department of Paediatrics, Maulana Azad Medical College, New Delhi 110002, India
- Correspondence to: H P S Sachdev E-6/12, Vasant Vihar, New Dehli 110057, India
- Accepted 28 June 2002
Objective:To evaluate the effect of iron supplementation on the incidence of infections in children.
Design: Systematic review of randomised controlled trials.
Data sources:28 randomised controlled trials (six unpublished and 22 published) on 7892 children.
Interventions: Oral or parenteral iron supplementation or fortified formula milk or cereals.
Outcomes:Incidence of all recorded infectious illnesses, and individual illnesses, including respiratory tract infection, diarrhoea, malaria, other infections, and prevalence of positive smear results for malaria.
Results: The pooled estimate (random effects model) of the incidence rate ratio (iron v placebo) was 1.02 (95% confidence interval 0.96 to 1.08, P=0.54; P<0.0001 for heterogeneity). The incidence rate difference (iron minus placebo) for all recorded illnesses was 0.06 episodes/child year (−0.06 to 0.18, P=0.34; P<0.0001 for heterogeneity). However, there was an increase in the risk of developing diarrhoea (incidence rate ratio 1.11, 1.01 to 1.23, P=0.04), but this would not have an overall important on public health (incidence rate difference 0.05 episodes/child year, −0.03 to 0.13; P=0.21). The occurrence of other illnesses and positive results on malaria smears (adjusted for positive smears at baseline) were not significantly affected by iron administration. On meta-regression, the statistical heterogeneity could not be explained by the variables studied.
Conclusion:Iron supplementation has no apparent harmful effect on the overall incidence of infectious illnesses in children, though it slightly increases the risk of developing diarrhoea.
What is already known on this topic
Iron supplementation is recommended to prevent iron deficiency, which is a major health problem, especially in the developing countries
Conflicting data exist regarding the possibility of an increase in the incidence of infections with iron supplementation, resulting in concern about the safety of this intervention
What this study adds
Iron supplementation has no apparent harmful effect on the overall incidence of infectious illnesses in children
Iron administration increases the risk of developing diarrhoea
Fortification of foods may be the safest and most beneficial mode of supplementation in relation to infectious illnesses
Anaemia caused by iron deficiency is a major public health problem, affecting 46% of school children globally.1 Iron deficiency has adverse effects on psychomotor development2 and on the capacity to work. The reversible consequences in childhood have prompted recommendations for early intervention. The proposed interventions rely primarily on enhancing iron intake either through supplementation or fortification of food. 3 4
Because of these proposed interventions their safety needs to be unequivocally established. The role of iron in resistance to disease remains controversial. Iron deficiency may be an important defence mechanism, and the term “nutritional immunity” was coined to highlight the importance of hypoferraemia in preventing bacterial growth.5 Conversely, data suggest that iron deficiency is associated with impairment of cell mediated immunity and the bactericidal activity of neutrophils, thus increasing susceptibility to infection. 6 7 Iron supplementation may also cause damage to cells mediated through free radicals.8 Objective safety data from longitudinal studies of iron supplementation are conflicting; trials have shown either beneficial effects,9 no effect,10 or an increase in infectious illnesses. 11 12 Children, particularly infants and those living in developing countries, are vulnerable to infectious diseases. It is thus important to establish the safety of iron supplementation in children on a public health scale. We conducted a systematic review to determine the effect of iron supplementation on infectious illnesses.
To be included trials had to be randomised placebo controlled trials—except for those in which iron was given parenterally, in which case trials could be non-placebo controlled because it would be difficult to administer a similar placebo; had to investigate iron supplementation through the oral or the parenteral route or as formula milk or cereals fortified with iron; and evaluate one or more infectious illnesses as an outcome measure. We also included studies in which other micronutrients and drugs were simultaneously administered if the only difference between the study and the control groups was iron supplementation.
We searched computerised bibliographic medical databases, including Medline, Cochrane controlled trials register, Embase, IBIDS, and Healthstar. We also reviewed reference lists of identified articles and hand searched reviews, bibliographies of books, and abstracts and proceedings of international conferences or meetings. Donor agencies, “experts,” and authors of recent iron supplementation trials were contacted to identify any additional or ongoing trials. The title and abstract of the studies identified in the computerised search were scanned to exclude studies that were obviously irrelevant. We retrieved the full text of the remaining studies and identified studies that fulfilled the inclusion criteria. To avoid publication bias we included published and unpublished trials.
Quality of methods
We assessed the quality of trials using recommended criteria. 13 14 Concealment of allocation was classed as adequate, unclear, inadequate, or not used. To assess completeness of follow up we classified studies by percentage of participants excluded (<3%, 3-9.9%, 10-19.9%, and 20%). Blinding was classified as double blinding, single blinding, no blinding, and unclear. TG Abstracted all data.
We used preformed questionnaires to abstract data. The data included in this review were derived from the published papers or were provided by the authors. Illnesses and the outcomes included were as defined by the authors. Whenever possible we contacted the authors for clarifications.
The presence of bias in the extracted data was evaluated by funnel plots.15 We used the metabias command in Stata software to perform the statistical tests for funnel plot asymmetry.16 The pooled estimates of incidence rate ratio and incidence rate difference were calculated by StatsDirect statistical software (version 1.9.5; StatsDirect, Cambridge) with fixed effects and random effects model assumptions.17 This program also computes the formal test of heterogeneity (Q statistic). We primarily report random effects estimates because most of the pooled results obtained were statistically heterogeneous. We chose incidence rate summary to account for the differences in duration of follow up in the various extracted studies. The data were recorded in the form of the total number of episodes of illness and the person time exposed (in child years). For trials in which the results were available in this format we recorded the figures directly from the publication, and this category of studies was labelled as the “actual” group. In the “computed” group of trials, the person time of follow up was not provided, and we calculated estimates from the product of the duration of follow up and the sample sizes available at the beginning and the end of the study. In some trials data were obtained by quantitative analysis of published graphs.
Some studies had reported only on the prevalence of malaria parasitaemia confirmed from smears at the beginning and the end of the supplementation period. Pooled estimates of the odds ratio of positive smears at the end of the supplementation period were computed by the “meta” command in Stata software.16 We also performed a meta-regression (restricted maximum likelihood iteration) through the “metareg” command in Stata software to determine the pooled log odds ratio of developing malaria in the group with iron supplementation compared with the placebo group. The covariate in the meta-regression equation was the log odds ratio at the beginning of the trial to adjust for the baseline differences in the prevalence of malaria.
We carried out stratified analyses for quality of methods; case detection (active field based or passive facility based); specificity of case definition; route of iron administration (parenteral, oral supplement, or fortified food); dose—this was initially planned but could not be performed as it could not be extracted for each study; duration of supplementation; type of illness (gastrointestinal, respiratory, malaria, non-diarrhoeal, or others); and baseline haemoglobin concentration in the supplemented group. The contribution of these variables to heterogeneity was also explored by meta-regression. 16 17
We identified 47 randomised controlled trials that were potentially eligible. Of these, 38 trials were published in medical journals or were theses9–52 and 9 were unpublished (box B1). Nineteen studies were ineligible (table 1). We therefore evaluated 28 studies (22 published 1011 31–44 47–52 two theses, 45 46 and six unpublished) in this systematic review.
Papers presented at International Nutritional Anaemia Consultative Group (INACG) Symposium, Hanoi, Vietnam, 2001
Allen LH, Lopez P, Galvaz IA, Garcia DP, Isoard F, Rosado JL. Does multiple micronutrient supplementation increase haemoglobin and iron status more than iron alone?
Lonnerdal B, Domellof M, Dewey KG, Cohen R, Rivera LL, Hernell O. Effects of iron supplementation of breastfed infants in Honduras and Sweden from 4-9 or 6-9 months of age.
Ninh NX, Berger J, Tolvanen M, Trung NQ, Nhien NV, Lien DK, et al. Control of iron deficiency anaemia in Vietnamese infants by efficacy of iron and zinc supplementation to reduce anaemia and growth faltering in Vietnamese infants.
Zimmermann M, Hess S, Adou P, Torresani T, Cook J, Hurrell R. Treatment of iron deficiency in goitrous children improves the efficacy of iodized salt.
Quyen DT, Berger J, Ninh NX, Khan NC, Khoi HH. Control of iron deficiency anaemia in Vietnamese infants by weekly and daily iron supplementation: efficacy and effectiveness.
Atukorala S, de Silva A, Ahluwalia N. Evaluation of iron status of children in the presence of infections: effect of iron supplementation on iron status, infection and morbidity.*
Other unpublished papers
Rice AL, Stoltzfus RJ, Tielsch JM, Savioli L, Montresor A, Albonico M, et al. Iron supplementation and mebendazole treatment do not affect respiratory or diarrhoeal morbidity incidence rates in Tanzanian preschoolers. 1999.*
Agarwal D, Sachdev HPS, Mallika V, Singh T. Iron supplementation in breast fed, full term, low birth weight infants. 1999.*
Nagpal J, Sachdev HPS, Mallika V, Singh T. Iron supplementation with complementary feeding in predominantly breastfed infants. 2000.*
*Included in the review
Baseline characteristics of the studies
Table 4 depicts the baseline characteristics of the included trials. Thirteen trials were in children aged <1 year, 10 studies included preschool children ( 5 years), and five trials included children aged >5 years. Eleven trials were from Africa, eight from Asia, five from the Americas, two from Europe, and two from Australia and New Zealand. The eligibility and exclusion criteria varied. Most of the studies used oral iron supplementation (20/28; 71%). Three trials used parenteral administration, and five studies used iron fortified foods.
Differences in the mode of administration may have implications for bioavailability of iron and its possible effect on the immune function. The supplementation dose used could influence the degree to which illness was affected. As a crude generalisation, the fortified formulas had the lowest dosage and the parenteral route had the highest. The duration of supplementation and follow up for oral intake varied from 2 months to 30 months.
The specificity of the definition used for illness was variable. Specificity of diagnosis has the potential to bias the observed effect of supplementation on illness. For example, low specificity definitions could underestimate the effect of iron supplementation on malaria due to a high rate of misclassification of non-malarial fevers as malaria. In some studies, fever was recorded as an additional infectious illness because fever in children is mostly attributed to infectious diseases. 4151 Inclusion of fever as a separate infection may lead to duplication of data because fever may accompany malaria, respiratory tract infection, and diarrhoea. However, we have included it on the assumption that an equal distribution of fever in both groups would eliminate any bias and also prevent non-inclusion of any observed infection.
The methods of surveillance varied: 15 were clinic based whereas 13 were field trials with active surveillance for cases. If iron supplementation has selective effects on mild rather than more severe episodes of illness then differences in methods of case detection may influence the observed effects of iron supplementation.
Bias detection for included studies
The funnel plot (fig 1) seems symmetrical, and we found no evidence of bias using the Egger (weighted regression) method (P=0.663 for bias) or the Begg (rank correlation) method (continuity corrected P=0.488).
Pooled and stratified estimates
We collected data on 7892 children followed up for 5650 child years—4027 children and 2802 child years in the iron supplemented group and 3865 children and 2848 child years in the placebo group (table 3). The pooled estimate of the incidence rate ratio (iron versus placebo) for all the recorded morbidities was 1.02 (95% confidence interval 0.96 to 1.08; P=0.54; test for heterogeneity Q=78.29, P<0.0001, fig 2). Calculations of incidence rate ratio based on “actual” data (when available) and computations from sample size at the end of the study (1.03, 0.97 to 1.08, P=0.21; test for heterogeneity Q=72.19, P<0.0001) were virtually identical with computations based on sample sizes at the beginning of the study. Besides the incidence rate ratio, from the public health perspective the incidence rate difference is considered to be more informative. The incidence rate difference (iron minus placebo) for all the recorded illnesses was 0.06 episodes per child year (−0.06 to 0.18, P=0.34; test for heterogeneity Q= 80.01, P<0.0001).
Stratified analysis for the effect on individual infectious illnesses showed that children in the iron supplementation group had an 11% (1% to 23%) higher risk (incidence rate ratio) of developing diarrhoea (P=0.04; test for heterogeneity Q= 30.24, P= 0.04, table 4). The effect on other individual illnesses was not significant. However, the incidence rate difference (public health impact) for diarrhoea was 0.05 episodes per child year (−0.03 to 0.13, P=0.21; test for heterogeneity Q=42.03, P=0.001). Further stratification showed that the significantly increased risk of diarrhoea associated with iron supplementation was restricted to oral supplementation (nine studies; incidence rate ratio 1.15, 1.01 to 1.32, P=0.04; incidence rate difference 0.18 episodes per child year, −0.01 to 0.37; P=0.07). The individual studies had not determined the cause of the diarrhoea, though dysentery indicates severe infectious diarrhoea. Only two studies provided information on dysentery; they showed no difference in the incidence between the two groups. Meta-regression showed that the route of iron administration (oral versus other) was not significantly associated with incidence rate ratio for diarrhoea (risk ratio 1.06, 0.85 to 1.32, P=0.59).
From the available data we found no increased risk of severe illness associated with iron supplementation (analysis possible only for lower respiratory tract infection and dysentery).
Table 5 shows the data extracted on malarial parasitaemia. The pooled odds ratio for positive smear tests for malaria at the end of the supplementation period (random effects model) was 1.43 (1.08 to 1.91, P=0.014; test for heterogeneity Q=11.611, P=0.114, fig 3. Meta-regression analysis of trials with relevant data (excluding the study by Oppenheimer et al34) indicated that this treatment effect was significantly associated with the baseline positivity of smear tests (for a unit increase in log odds ratio of baseline positivity, the treatment effect increased by 2.89; 1.37 to 6.10; P=0.005) but not iron supplementation (1.24; 0.98 to 1.57; P=0.076).
Meta-regression analyses to explore heterogeneity
Stratified estimates indicated that iron supplementation did not significantly (P>0.05) increase the incidence of infections (incidence rate ratio and incidence rate difference), irrespective of the quality of methods, methods of surveillance, route of iron supplementation, duration of supplementation, geographic location of the study population, or the basal haemoglobin concentration of the iron supplemented group (data not presented). Meta-regression analysis showed that the treatment effect (incidence rate ratio) was not significantly associated with any of these study characteristics (table 6).
The results from our analysis of these studies show that iron supplementation does not significantly increase the risk of overall infection. However, there was an increase in the risk of developing diarrhoea, but this would not have an important overall impact on public health. The occurrence of other illnesses and malarial parasitaemia (adjusted for positive smear results at baseline) was not significantly affected by iron administration (P>0.05).
Strengths and limitations of analysis
Despite wide clinical and methodological heterogeneity in the various trials, the main inference remained stable for the various sensitivity analyses that we performed. An important caveat is the lack of uniform definitions for the individual clinical morbidities. Uniform definitions and active surveillance would have provided greater weight to the conclusions. Furthermore, not all the included trials were of high quality. We could not explain the statistical heterogeneity by various study characteristics.
There are still some questions unanswered and some new issues raised. We could not determine whether the higher risk of diarrhoea was a result of increased gastrointestinal infections or a consequence of the irritant effect of iron on the gut motility, a known effect.53 Dysentery is invariably infective in origin, and the two trials that provided information found no evidence of an increase in dysentery in children receiving iron supplements.
We could not analyse the effect of dose on the incidence of infections. However, the near absence of any important adverse effects, particularly diarrhoea, in children receiving fortified foods (compared with medicinal iron) raises the possibility of a dose related effect. Interestingly, there was also a similar significant protective effect against the development of respiratory tract infections (four studies; incidence rate ratio=0.92; 0.86 to 0.98; P=0.02). However, our meta-regression analysis showed that the route of administration was not significantly associated with incidence rate ratio. Fortification with low doses of iron is closest to the physiological situation and could theoretically be considered the safest public health intervention. There is thus a case for concomitant evaluation of the possible beneficial effects of iron fortified foods on the haematological response and infections.
Meta-regression analysis suggested that the risk of acquiring infectious illnesses is inversely associated with the baseline haemoglobin concentration. Stratified analysis also suggested increased risk of infections in children who had a mean baseline concentration below 100 g/l. Iron supplementation promotes production of free radicals, and this may have a deleterious effect on the immunity of a child. Ironically, defences against free radicals are compromised the most in iron deficiency and malnutrition, 54 55 which are conditions likely to benefit the most from iron supplementation. Interestingly, all the studies included in this stratified subset were from regions of the African continent where malaria is endemic. Some data suggest indirectly that iron deficiency in such regions decreases the susceptibility to disease related to malaria, HIV, and tuberculosis.56 The safety of iron supplementation in people with anaemia, particularly in regions where malaria is endemic, may be difficult to determine because of the ethical problem of withholding treatment in a control group.
We thank Sunil Sinha for facilitating access to Embase and two PhD dissertations. We also thank L Satyanaryana for help in conducting the meta-regression analysis. Part of this paper was presented as a poster at the International Nutritional Anaemia Consultative Group symposium 2001, held at Hanoi, Vietnam.
Contributors: TG prepared the protocol, applied the search strategy, performed the retrieval of articles, and extracted the data from the included studies. HPSS developed the idea for the review, finalised the protocol and search strategy, and performed the statistical analysis. Both the authors contributed to the drafting of the final version of the paper. HPSS is guarantor.
Editorial by Tomkins
Competing interests International Life Science Institute (ILSI) sponsored TG for travel to Hanoi, Vietnam for the purpose of attending the International Nutritional Anaemia Consultative Group symposium and presenting part of the analysis as a poster.