QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Oppenheimer, S. J. Search for Related Content PubMed PubMed Citation Articles by Oppenheimer, S. J.

---

Supplement

Iron and Its Relation to Immunity and Infectious Disease^{1 ,2}

Green College, Oxford, UK

	ABSTRACT

TOP ABSTRACT INTRODUCTION Criteria for inclusion of... In vitro studies of... Observational studies on iron... Iron intervention does not... Long-term prospective controlled... Long-term studies in malaria... Causal relationships: do they... Is there a connection... SUMMARY DISCUSSION REFERENCES

 
The continuing unresolved debate over the interaction of iron and infection indicates a need for quantitative review of clinical morbidity outcomes. Iron deficiency is associated with reversible abnormalities of immune function, but it is difficult to demonstrate the severity and relevance of these in observational studies. Iron treatment has been associated with acute exacerbations of infection, in particular, malaria. Oral iron has been associated with increased rates of clinical malaria (5 of 9 studies) and increased morbidity from other infectious disease (4 of 8 studies). In most instances, therapeutic doses of oral iron were used. No studies in malarial regions showed benefits. Knowledge of local prevalence of causes of anemia including iron deficiency, seasonal malarial endemicity, protective hemoglobinopathies and age-specific immunity is essential in planning interventions. A balance must be struck in dose of oral iron and the timing of intervention with respect to age and malaria transmission. Antimalarial intervention is important. No studies of oral iron supplementation clearly show deleterious effects in nonmalarious areas. Milk fortification reduced morbidity due to respiratory disease in two very early studies in nonmalarious regions, but this was not confirmed in three later fortification studies, and better morbidity rates could be achieved by breast-feeding alone. One study in a nonmalarious area of Indonesia showed reduced infectious outcome after oral iron supplementation of anemic schoolchildren. No systematic studies report oral iron supplementation and infectious morbidity in breast-fed infants in nonmalarious regions.

KEY WORDS: • iron • infection • malaria • morbidity • clinical trial

	INTRODUCTION

TOP ABSTRACT INTRODUCTION Criteria for inclusion of... In vitro studies of... Observational studies on iron... Iron intervention does not... Long-term prospective controlled... Long-term studies in malaria... Causal relationships: do they... Is there a connection... SUMMARY DISCUSSION REFERENCES

 
Iron deficiency is the most common micronutrient deficiency in the world, especially in the tropics. Prevention of iron deficiency is perceived by health workers as a desirable worldwide goal, preferably provision of iron by the oral route, including fortification of milk and other foodstuffs, although parenteral administration is still used in certain circumstances (Oppenheimer and Hendrickse 1983 ). However, this dream of universal freedom from an easily preventable disorder has been stalled somewhat since the 1980s.

Although this failure may be mainly one of implementation, one often-mentioned contribution to the problem is the unresolved concern as to the interaction among iron status, iron supplementation and susceptibility to infection (Dhur et al. 1989 , Farthing 1989 , Hershko et al. 1988 , Hershko 1993 , Oppenheimer and Hendrickse 1983 , Oppenheimer 1994 and 1998 , Scrimshaw and San Giovanni 1997 ). This difficult subject has been polarized by partisan claims either that iron deficiency always helps (the so-called nutritional immunity hypothesis [Kochan 1973 , Weinberg 1978 ]) or always hinders defenses against infection.

The early prospective intervention studies conducted in deprived populations of temperate, developed countries tended to support the value of iron supplements in reducing rates of respiratory infections in infants (Andelman and Sered 1966 , Cantwell 1972 , MacKay 1928 ). This rosy picture was not sustained with intervention studies published from the late 1970s on. Side effects of treatment, particularly with parenteral iron, were one problem, and later reports from the tropics seemed to indicate a deleterious effect on susceptibility to both malaria and respiratory infections, thus emphasizing the fact, already known from animal studies, that different organisms interact differently with iron in their hosts.

The malaria issue has dominated the picture since the 1970s, and few significant studies evaluating systematic iron supplementation and infectious morbidity in nonmalarious areas were published in the 1980s. The controversy has been compounded by the lack of adequate control in earlier published prospective clinical intervention studies and the practical impossibility of conducting adequately controlled nonintervention (i.e., observational) studies in iron-deficient humans. Even laboratory measures of iron deficiency are grossly confounded by the immediate presence of infection (Oppenheimer and Hendrickse 1983 ).

The result has been that although experimental iron deficiency in animals and in vitro functional effects on immunity in humans have been usefully studied for decades, numerous clinical reviews still seem unable to make clear quantitative statements about how important iron deficiency is to human infectious morbidity. Although more controlled studies must be done, there is still useful information to be gleaned from the literature as long as the older intervention studies are not all rejected out of hand because of design faults.

	Criteria for inclusion of studies and evaluation of a relationship

TOP ABSTRACT INTRODUCTION Criteria for inclusion of... In vitro studies of... Observational studies on iron... Iron intervention does not... Long-term prospective controlled... Long-term studies in malaria... Causal relationships: do they... Is there a connection... SUMMARY DISCUSSION REFERENCES

 
This review is directed primarily at the clinical and epidemiological evidence for a causal relationship between iron deficiency and infectious morbidity as shown by controlled intervention studies of iron supplementation. As has been stated, clinical studies in countries with endemic malaria tend to have produced effects opposite to those in nonmalarious climes; thus, results are discussed and summarized in tabular form separately for these two situations.

A preexisting database of reports (Oppenheimer 1994 and 1998 ) was supplemented with the use of the following: 1) MEDLINE search, using combinations of the keyword "iron" crossed separately with "supplement," "infection," "trial" and "immunity"; the results were hand-searched; 2) personal communications with workers involved in existing trials; and 3) search of the Cochrane Controlled Trials Register. All studies mentioning infectious outcome were reviewed, but only controlled studies with quantified clinical infectious morbidity are discussed here. The only further exclusions from the latter were studies in which another micronutrient apart from iron was combined with iron as an intervention but not given to the placebo group.

Design, participant selection and multiple outcomes of these studies were so qualitatively heterogeneous that meta-analysis may have limited meaning, and this review concentrates on the possible ecological reasons for the different outcomes. To obtain some consistency in tabular and graphic presentation of outcomes, however, odds ratios (OR)³ [with a fixed-effects model and 95% confidence intervals (CI)—RevMan 4.0.4 (Cochrane Collaboration)] based on the dichotomous contingency "no morbid event vs. one or more events" per individual, over a stated time period—are given, where published data allow. In some studies, this contingency is not available because total morbid events rather than individuals were given as the numerator. If OR could not be calculated from available information, the actual published rates, relative risks or both are quoted in tables only and omitted from OR plots. In several studies, follow-up wastage was high, thus introducing potential systematic bias in rates. For these studies, therefore, realistic average surveillance denominators are estimated from available information. Only clinically relevant morbidity outcomes are cited. Functional laboratory immunological outcomes and malarial parasite prevalence rates have been dealt with elsewhere (Dhur et al. 1989 , Farthing 1989 , Hershko 1993 , Oppenheimer and Hendrickse 1983 , Oppenheimer 1994 and 1998 , Scrimshaw and San Giovanni 1997 , Shankar et al. 2000 ).

During the review of studies, subsidiary questions were addressed, such as whether potential causal effects are graded and what, if any, are the laboratory measurements that predict the morbidity outcome most closely. In the absence of clear answers to the primary questions in humans, such issues can only be flagged, not answered.

Before the discussion of interventions, an overview will be made of the in vitro evidence for functional immunological effects of iron deficiency and supplementation along with evidence for how this interfaces with specific infections. This is supplemented by a review of nonintervention iron deficiency and morbidity reports.

	In vitro studies of iron deficiency and functional immunity

TOP ABSTRACT INTRODUCTION Criteria for inclusion of... In vitro studies of... Observational studies on iron... Iron intervention does not... Long-term prospective controlled... Long-term studies in malaria... Causal relationships: do they... Is there a connection... SUMMARY DISCUSSION REFERENCES

 
Two component systems of active immunity—humoral and cell-mediated immunity—have been studied extensively, mainly in vitro, in relation to iron deficiency in both humans and animals. Little evidence exists for systematic major humoral deficiencies in iron-deficient humans, and although specific defects in cell-mediated immunity have been well described and reviewed, even in latent iron deficiency, note that such minor functional changes cannot be compared with the devastating effects of the well-defined immunodeficiency syndromes (Dhur et al. 1989 , Farthing 1989 , Hershko 1993 , Oppenheimer and Hendrickse 1983 , Scrimshaw and San Giovanni 1997 ). Little information is available on whether effects are graded by degree of iron deficiency; they may even be present before hemoglobin is lowered. Intensively studied deleterious effects of iron deficiency on cellular defenses that are reversible with iron therapy include the following: 1) Reduced polymorph neutrophil function with decreased myeloperoxidase activity and nitro blue toluene reduction reversed by iron administration. Intracellular bacteriocidal activity has been reported as impaired, but not all studies give consistent results (Dhur et al. 1989 , Farthing 1989 , Hershko 1993 , Oppenheimer and Hendrickse 1983 , Scrimshaw and San Giovanni 1997 ). 2) Depression of T-lymphocyte numbers with thymic atrophy, and most but not all studies concur (Dhur et al. 1989 , Farthing 1989 , Hershko 1993 , Oppenheimer and Hendrickse 1983 , Scrimshaw and San Giovanni 1997 ). 3) Defective T lymphocyte–induced proliferative response, with slightly more reports showing an effect than not showing an effect (Farthing 1989 ). 4) Impaired natural killer cell activity (Dhur et al. 1989 ). 5) Impaired interleukin-2 production by lymphocytes (Galan et al.1992 ). 6) Reduced production of macrophage migration inhibition factor (Dhur et al. 1989 , Farthing 1989 ). 7) Reversible impairment of delayed cutaneous hypersensitivity, including tuberculin reactivity; in general there is agreement for this finding (Farthing 1989 , Moraes-de-Souza et al. 1984 ).

Proteins in iron metabolism.

The third set of defense systems that are uniquely associated with iron metabolism are cellular and extracellular iron-binding proteins (transferrins and lactoferrin). These inhibit bacterial growth by withdrawing iron.

Almost 30 years ago, Bullen et al. (1972) showed that the bacteriostatic action of human milk is abolished by in vitro addition of iron. In this context, Murray et al. (1980) showed an increase in Entamoeba histolytica infection in nomads who drank cow’s milk during supplementation with oral iron. This was not seen in the nonrandomized control group or in those receiving parenteral iron. They related the effect to saturation of the cow’s milk lactoferrin by oral iron.

Cow’s milk formulas contain no active lactoferrin, and therefore iron fortification in theory should not compromise artificially fed infants. However, there has been debate about the advisability of giving oral iron to breast-fed infants. One thing that is clear is that formula milk, with or without iron, can carry a higher risk of infectious morbidity than breast-feeding in deprived communities (Heresi et al. 1995 ). Lonnerdal et al. (1980) pointed out that because the bulk of the iron in breast milk is not attached to lactoferrin and yet is highly bioavailable for the infant, an alternative method of oral iron supplementation for infants might be to supplement lactating mothers.

    "Nutritional immunity." The growth of a variety of bacteria and fungi are inhibited in vitro by transferrin and lactoferrin (Kochan 1973 , Weinberg 1978 ). The thesis of nutritional immunity elaborates this well-known defense mechanism by arguing that "further lowering of the saturation of iron in transferrin or lactoferrin further enhances immunity." In practice, however, this thesis appears to have been overstated for extracellular microbial pathogens (Oppenheimer and Hendrickse 1983 ). Virulent invasive pathogens usually have their own powerful siderophores that are quite capable of removing iron from transferrin, whereas less virulent opportunistic infections such as Escherichia coli are equally inhibited from growing in plasma over a wide range of transferrin saturations, corresponding to different physiological states of iron balance (Oppenheimer and Hendrickse 1983 ). In other words, iron-deficient individuals may not be especially protected from such opportunistic infections. Equally, lowered transferrin saturation does not affect the virulence of disseminating organisms with high iron avidity (Oppenheimer and Hendrickse 1983 ).

One group of microorganisms with an iron requirement in a special situation in relation to iron availability comprises the erythrocytic forms of malaria. Although inside cells with the highest iron content in the vertebrate body, plasmodia are apparently unable to use this source effectively nor do they seem able to extract transferrin-bound iron from the plasma surrounding the red cells. Much work over the past 15 years has centered around the observation that iron chelators can inhibit malarial growth in vitro as well as in vivo (Hershko et al. 1988 ). A recent in vitro study appears to confirm the longstanding theory that a small labile pool of iron in red cells (which may be crucially smaller in people with iron deficiency) provides the iron the parasite requires (Loyevsky et al. 1999 ). This may provide one mechanism that underlies clinical observations suggesting that iron deficiency may protect from malaria.

The qualitative differences between iron-pathogen interactions in the extracellular compartment and those that are intracellular have recently been refocused by study of the natural resistance associated macrophage proteins (NRAMP 1, 2) and the role of NRAMP 1 in resistance to infection by actively removing intracellular iron. NRAMP 1 targets the membrane of microbe-containing phagosomes in macrophages and monocytes. Allelic variants of NRAMP 1 have recently been found to be associated with susceptibility to tuberculosis and leprosy in humans (Canonne-Hergaux et al. 1999 ). This finding may be relevant to the differential susceptibility of people to infection by intracellular pathogens according to their iron status.

To summarize in vitro evidence, iron deficiency depresses certain aspects of cell-mediated immunity, including lymphocyte, neutrophil and macrophage function; humoral immunity is unaffected and the significance of hypoferremia (as opposed to normal transferrin saturation) on growth of microorganisms is uncertain. In contrast, one group of intracellular organisms, Plasmodia, may have a specific disadvantage in iron deficiency. Because there are conflicting effects of deficiency and treatment on defense systems, it becomes more important to study the situation in vivo.

	Observational studies on iron deficiency and infectious morbidity

TOP ABSTRACT INTRODUCTION Criteria for inclusion of... In vitro studies of... Observational studies on iron... Iron intervention does not... Long-term prospective controlled... Long-term studies in malaria... Causal relationships: do they... Is there a connection... SUMMARY DISCUSSION REFERENCES

 
Iron deficiency and its effect on infection are difficult to study in humans using observational or noninterventional means. This is because iron deficiency is part of a cluster of nutrient and social deprivations, ultimately resulting from poverty, that are inevitably interrelated. There is also an ethical problem with prolonged study of people known to be iron deficient while withholding treatment. Therefore, we will start with a brief overview of the effects of iron deficiency (and changes after supplementation) on infectious morbidity in animals. These results, although unable to answer the main questions, may give pointers as to what to look for in clinical and epidemiological studies.

In vivo studies in iron-deficient animals.

In view of the potential problems of confounding when relating iron deficiency to immune status in humans, it is worthwhile looking at the more controlled observations available from animal studies. Experimental studies in laboratory animals uniformly show reversible deleterious effects of iron administration on tests of functional immunity. These may occur even in mild deficiency. Reports of graded effects are contradictory (Dhur et al. 1989 , Scrimshaw and San Giovanni 1997 ). The picture is not clear, however, with experimental studies of effects on morbidity. Experimental studies of infectious challenge and subsequent morbidity in iron-deficient animals have produced conflicting results (Dhur et al. 1989 ). Hart et al. (1982) using Proteus mirabilis–induced pyelonephritis in rats showed differential effects, with severe iron deficiency protecting less than mild deficiency. Baggs and Miller (1973) claimed that severe deficiency enhanced defenses in rats against invasive Salmonella, whereas mild deficiency impaired them. Preweaning iron deficiency produced permanent immune defects. Puschmann and Ganzoni (1977) showed increased resistance of iron-deficient mice to invasive Salmonella typhimurium infection, whereas Chu et al. (1976) showed an increased mortality of severely iron-deficient rats infected with Streptococcus pneumoniae compared with controls. Harvey et al. (1985) showed reduced parasitemias and reduced mortalities in iron-deficient mice infected with Plasmodium chabaudi. Perhaps the only partially unifying message to be gained from these studies is that effects are microorganism specific and that iron deficiency may be more likely to protect against intracellular than extracellular pathogens.

Observational studies in iron-deficient humans.

Because of problems of control and confounding, few observational clinical studies in iron-deficient humans convincingly relate such deficiency to substantial morbidity due to infections. Some studies have noted anemia in children admitted to hospital for various infections, but these data could not establish causality (Kaplan and Oski 1980 , Lovric 1970 , Oppenheimer 1980 ). Higgs and Wells (1973) noted that of 31 patients with chronic mucocutaneous candidiasis, 23 were iron deficient and 9 of 11 improved with oral and parenteral iron therapy alone, with a regression of oral lesions and development of delayed hypersensitivity to Candida. There was no control group. In another report, 16 patients with recurrent staphylococcal furunculosis also had nonanemic iron deficiency. Furunculosis resolved after 3–4 wk of iron therapy in all but one patient (Weijmer et al. 1990 ).

In one prospective study, postoperative complications, in particular, infections after abdominal surgery, were reported to be significantly more common in 228 patients with low preoperative serum ferritin compared with 220 patients with normal ferritin; confounders including hemoglobin levels were taken into account in the analysis (Harju 1988 ). Reports from the tropics are difficult to evaluate because of limited ascertainment of the causes of anemia. In an often misquoted study, Masawe et al. (1974) reported fewer bacterial infections in patients admitted with simple iron deficiency anemia than in a control inpatient group with a variety of other causes of anemia (megaloblastic and refractory), but they also reported more frequent malaria in the iron-deficient patients (8 of 16 cases after initiation of therapy). Unfortunately, they did not give details of which patients were receiving oral or parenteral therapy or whether patients had received therapy before admission.

In their report of the effects of iron treatment in Somali nomads, Murray et al. (1978) also noted that nomads entering a feeding camp had no infections if they were iron deficient (n = 26) in contrast to a high rate of infection in those with normal iron status (19 of 64). This unblinded study is difficult to assess. In contrast, Snow and colleagues (1991) attempted to determine whether measurement of iron status in 1- to 9-y-old Gambian children before the start of the malaria transmission season could predict malarial experience and morbidity during that season. They did not find a significant correlation between any hematological measures of iron status and subsequent malarial experience in this older age group.

Possible evidence for a protective effect of low iron stores at birth on subsequent malarial morbidity was obtained in a prospective study conducted by the author in Papua New Guinea (described in more detail below) (Oppenheimer et al. 1986a and 1986b ). Infants with lower hemoglobin values at birth were less likely to have malaria at field follow-up and less likely to be admitted to hospital during y 1 of life. Because birth hemoglobin is the main iron source during the first year, this association may mean that iron deficiency protects from malaria and other infections. This association may, alternatively, have resulted from the associated protective effect of homozygous single-deletion -thalassemia, which is present in >50% of that population. Indeed the high prevalence of single-deletion -thalassemia in many tropical areas may have a confounding effect in many studies of iron, anemia and morbidity because the mutation both causes anemia and protects against malaria and other infections (Oppenheimer et al. 1987 ). Allen and colleagues (1997) showed in a subsequent case-control study in the same part of New Guinea that people with homozygous +-thalassemia were much less likely than controls to develop severe malaria (OR 0.40; 95% CI: 0.22–0.74) and to be admitted to hospital for nonmalarial infections (OR 0.36; 95% CI: 0.22–0.60).

In the study of Oppenheimer et al. (1986a) , the predictive value of hemoglobin measured at birth was an exception to the generally low predictive value of hematological measures for subsequent morbidity. Measures of iron status such as hemoglobin, transferrin saturation and ferritin taken at birth and at 2 and 6 mo were of no predictive value for subsequent infectious outcome. Changes in both ferritin and transferrin saturation acted more as acute-phase responses to acute infection. Use of hemoglobin as a surrogate for iron status in infectious morbidity studies is thus clearly fraught with potential for confounding. This was clearly demonstrated in the large iron intervention study of Gebreselassie (1996) in Ethiopia (see also below). Pretrial cross-sectional assessment of the cohort showed that anemia was significantly associated with malaria (OR 1.31; 95%CI: 1.03–1.66), acute respiratory infection (OR 1.33; 95% CI: 1.05–1.69) and diarrhea (OR 1.74; 95% CI: 1.28–2.40). However, no relationship was detected between the same morbidity indicators and the more reliable iron measures of mean cell volume and serum ferritin taken at the same time. Whether the lower hemoglobin of the sick children resulted from iron deficiency or was secondary to their infections is therefore not clear.

Confounders in observational studies on iron deficiency.

Clearly there are many problems and differences in design between these observational reports on iron deficiency in the tropics that make it difficult to draw useful conclusions. Another major confounding difference is the age of the study respondents, i.e., younger immune-naive infants appear to show an advantage of low iron status, whereas older semi-immune children do not. Other confounders may interact with immediate effects of iron treatment and will be considered next.

	Iron intervention does not simply reverse iron deficiency

TOP ABSTRACT INTRODUCTION Criteria for inclusion of... In vitro studies of... Observational studies on iron... Iron intervention does not... Long-term prospective controlled... Long-term studies in malaria... Causal relationships: do they... Is there a connection... SUMMARY DISCUSSION REFERENCES

 
Much of the confusion and many of the conflicting results of intervention studies stem from confounding factors that may affect immune and iron status of the populations under study. These include age, past immune experience, diet and cooking practice, and common inherited disorders of globin genes, which may, depending on type and zygosity, protect from malaria, lose their protective efficacy during iron supplementation or predispose to iron overload. Temporary overload may also result from treatment, in particular with parenteral iron.

Estimation of the effects of iron treatment on body iron compartments depends on the dose and route of administration and time elapsed since the start of treatment. There is a massive but short-term hyperferremia after parenteral administration of iron that lasts up to 3 wk after intramuscular iron dextran (Will 1968 ) or 2–3 d after intravenous iron dextran (Kanakakorn et al. 1973 ). Circulating iron dextran complex may be a source of iron for bacterial growth immediately after injection, and serum bacteriostatic action is lost during this period. This size of effect is not seen with oral iron supplementation in normal doses (Gross 1968 ), although gut intraluminal iron may be high (Murray et al. 1980 ). The hyperferremic effect of parenteral iron is a particular danger during the neonatal period, when such therapy is contraindicated in all circumstances (Becroft et al. 1977 ).

In the 1970s, several poorly controlled studies incriminated iron therapy in acute exacerbations of preexisting or latent infections. In most of these reports, parenteral iron was used. The report of Masawe et al. (1974) , which included infections after iron therapy, was already mentioned. Byles and D’Sa (1970) , in a poorly controlled study, reported 11 cases of clinical malaria in 917 pregnant women immediately after parenteral iron therapy.

In New Zealand, serious E. coli sepsis was reported in the mid-1970s by Barry and Reeve (1977) in 2% of Polynesian neonates who received 250 mg iron dextran at birth (in five daily doses) (Table 1 ). The increase in infections was confined to the week after the last injections and was detected retrospectively with the attendant problem of validity. Neonatal sepsis fell to 0.2% after discontinuation of this practice. A similar finding, also in New Zealand, was made by Farmer and Becroft (1976) . In this case it was neonatal E. coli meningitis that was specifically higher in those receiving parenteral iron (Table 1 ; Fig. 1 ). In confirmation of the postulated mechanism, Becroft et al. (1977) showed a marked reduction in the bacteriostatic action of serum of these neonates on E. coli in vitro. Despite the obvious methodological problems of these retrospective studies, the evidence for increased E. coli sepsis in neonates is strong.

View larger version (11K): [in this window] [in a new window]   Figure 1. Iron trials in nonmalarious regions. Outcome: serious infections (all types). Method of administration: neonatal parenteral iron. Method of morbidity quantification: retrospective passive case detection using historical controls (1: Barry and Reeve 1977 , Farmer and Becroft 1976 ); prospective passive case detection of admissions and in-patients (2: Cantwell 1972 ). Abbreviations: OR, odds ratio; CI, confidence interval (all figures).

Although most the above immediate adverse effects were associated with parenteral therapy, an important and often-quoted study in the 1970s used oral iron. A family team (Murray et al. 1978 ) conducted a prospective placebo-controlled randomized trial of 30 d oral iron supplementation in 137 adult Somali nomads with iron-deficiency anemia. Iron treatment increased hemoglobin and transferrin saturation during the study. Although no malaria was noted at the start of the study in either group, 13 clinical attacks of malaria occurred in the iron group (n = 71) and only 1 in the control group (n = 67) by the end of the trial; these attacks were presumed by the authors to be reactivations (Table 3 ). Fevers were significantly more common in the iron group and eight nomads in the iron group started excreting Schistosoma ova in the urine compared with no new cases in the placebo group. The study was single blind, and no follow-up was made after the 30 d of treatment, although the treatment group still had high reticulocyte counts and higher transferrin saturations than the control groups at the end of this period. Documentation of infections was limited to laboratory identification of a pathogen.

	Long-term prospective controlled iron intervention studies

TOP ABSTRACT INTRODUCTION Criteria for inclusion of... In vitro studies of... Observational studies on iron... Iron intervention does not... Long-term prospective controlled... Long-term studies in malaria... Causal relationships: do they... Is there a connection... SUMMARY DISCUSSION REFERENCES

Prospective studies in nonmalarious areas.

The earliest longitudinal inquiry into the effects of iron supplementation on infection rates was that of Helen MacKay (1928) who reported that infants in London aged 3 wk to 18 mo given dietary supplements of iron had 50% fewer respiratory and gastrointestinal infections than infants not supplemented. Odds ratios estimated from her data showed a significantly reduced risk of respiratory infections associated with iron supplementation. Both method of administration (which included fortified milk) and dosage of iron (50–100 mg/d) varied among infants; unfortunately, observations on study and control groups were not blinded and, although covering 1 y in each group, were not made in the same year (Table 1 ; Fig. 2 ).

View larger version (26K): [in this window] [in a new window]   Figure 2. Iron trials in nonmalarious regions. Outcome: respiratory infections. Methods of administration: food fortification: infants (1–3: Andelman and Sered 1966 , Javaid et al. 1991 , MacKay 1928 ); oral supplementation: preschoolers (4: Angeles et al. 1993 ); parenteral: newborns (5: Cantwell 1972 , James and Combes 1960 ). Methods of morbidity quantification: unblinded oral recall (1 and 2); field clinical assessment (3 and 4); prospective hospital based clinical assessment of serious infections (5); active case detection (1–4); prospective passive case detection (5), historical controls (1). LRTI, lower respiratory tract infection; URTI, upper respiratory tract infection; OPD, outpatient detection.

The largest longitudinal study of iron supplementation was that of Andelman and Sered (1966) in Chicago, in which 1048 infants were randomly assigned to receive formula milk with or without iron fortification (10 mg elemental Fe/L). Follow-up was for 18 mo. Anemia [hemoglobin (Hb)<100 g/L] occurred in 76% of the control and 9% of the study group. These cases were then unfortunately removed from analysis. No details of methods of morbidity recording were given, and mother’s recall apparently was used. Odds ratios calculated from the presented data using estimated denominators show a significantly reduced risk of respiratory infections associated with iron supplementation at each 3-mo follow-up to 68 wk (Table 1 ; Fig. 2 ).

Burman (1972) conducted a similar placebo-controlled study using oral colloidal ferric hydroxide (10 mg Fe/d) in Bristol. Again, methodology of morbidity recording was not elaborated and no figures were actually given. No differences in illnesses were noted between the control and study groups, but the small difference in Hb values between the two groups reached significance only in the 2nd y, suggesting a low rate of iron deficiency in any case in the placebo group.

Salmi et al. (1963) reported twice the incidence of infections in the control over the study group in a prospective trial of parenteral chelated iron medication administered to premature infants in Finland. No details were given in the letter.

In a controlled 2-y prospective study in the same population and same part of New Zealand as the study of Barry and Reeve (1977) (see above), Cantwell (1972) studied Polynesian neonates who had received 250 mg iron dextran at birth (in five daily doses) and apparently found an opposite effect (Table 1 ; Fig. 1 ), namely, 42 and 32% hospitalization rates for infections in control and iron dextran treatment groups, respectively. These effects are apparent in the recalculated OR, although they achieve significance only for the subset of respiratory infections (Table 1 ; Figs. 1 2 3 ). The obvious difference between these contrasting reports is in the time scale studied, i.e., the adverse effects reported in the neonates were immediate and, although serious, affected a relatively much smaller proportion of infants than Cantwell’s finding of beneficial effects over a longer period. This contrast can be seen graphically in Figure 1 .

View larger version (18K): [in this window] [in a new window]   Figure 3. Iron trials in nonmalarious regions. Outcome: diarrheal infections (all types). method of administration: parenteral iron: infants (1: Cantwell 1972 , James and Combes 1960 ); food fortification: infants (2: MacKay 1928 and 3: Javaid et al. 1991 ); oral supplementation: preschoolers (4: Angeles et al. 1993 ). Methods of morbidity quantification: hospital based clinical assessment of serious infections (1); unblinded oral recall with historical controls (2); field clinical assessment (3 and 4); prospective passive case detection (1); prospective active case detection (2–4). OPD, outpatient detection.

In another study undertaken in Hungary by Hemminki and co-workers (1995) , a birth cohort received iron-fortified milk (6.5 mg/L) and were followed for 1 y. Both passive and active follow-up by various levels of health care worker were combined. This intervention failed to find any differences in morbidity; however, the only morbidity recorded was upper respiratory and there were no hematological differences between intervention groups. For these reasons, this study is not included in tables.

Heresi and colleagues (1995) conducted a controlled study using iron fortification of full-fat powdered cow’s milk (fortified with 15 mg Fe/100 g compared with unfortified milk) of infants in Santiago, Chile. Follow-up was from birth to 15 mo of age. By 9 mo of age, 15% of mothers had changed their children to breast milk alone. This latter group of children was then observed as a statistically segregated group. Partial improvement of hematological status in the test group suggested that iron deficiency anemia in the control group was inadequately reversed by the level of supplementation. The authors interpreted the complex subdivided outcomes as showing a trend of lower infectious morbidity in the supplemented group and also in the noniron-deficient children from both groups. However, the former differences were not significant. The group that clearly did best with respect to lower infection rates was the group that had changed to breast milk alone; this group had significantly lower morbidity rates than did either the iron intervention or control group (Table 1) .

Milk cereals with and without iron (mean daily intake 4.1–5.1 mg Fe) were given to groups of weaning infants (between 122 and 365 d of life) in a careful but complex 10-cell study conducted by Javaid and colleagues (1991) in Pakistan. There was a neighboring community control group with no nutritional supplementation. Although the iron was associated with a measurable effect on hematological values and infectious morbidity was reduced in all cereal-supplemented groups compared with the control group, there was no evidence of the iron intervention per se on such morbidity. It was concluded that the reduction in morbidity was a result of macronutrient supplementation (Table 1 ; Figs. 2 , 3 ).

Three nonfood oral iron supplementation studies have given conflicting results in nonmalarious tropical regions. Chwang and colleagues (1988) gave oral ferrous sulfate [10 mg/(kg · d)] to Indonesian school children in a placebo-controlled, double-blind study for 12 wk. Infectious morbidity was scored on a composite scale. Children were assessed as anemic or nonanemic at the start of the study by hemoglobin and transferrin saturation, thus giving four cells. Those who were anemic were smaller and had a significantly higher morbidity score at the start. They were also the stratum that benefited from iron intervention in terms of significantly reduced infectious morbidity and improved growth and hematological status as assessed from the analysis of covariance. The presentation of their results does not allow calculation of OR; thus the study is not tabulated, but the ratio of morbidity scores by iron intervention in the anemic subgroup was an impressive 4.0. This is one of the few studies in the literature to show an effect of iron intervention on length velocity.

Another oral iron study (Fe 30 mg/d for 2 mo) in Indonesia was carried out with anemic malnourished preschoolers by Angeles and co-workers (1993) . Reductions in rates of fever, respiratory infections and diarrhea were associated with iron therapy, although there were small cell numbers and these differences were not significant (Table 1 ; Figs. 2 , 3 ). Conversely, Mitra and co-workers (1997) gave 125 mg of ferrous gluconate daily for 15 mo to half of a group of 349 Bangladeshi children aged 2–48 mo. No differences were noted in attack rates for diarrhea, dysentery or acute respiratory infections. The only exception to this was found on stratification of outcome by age. Children <1 y of age in the iron group had significantly more episodes of dysentery and more days of illness with this disease than their placebo counterparts. This last result could be important because it represents the only hint in the literature that oral iron may have any deleterious effects for infection at all in nonmalarious regions. Numbers of infants, however, were small (iron, n = 34; control, n = 44), and no overall effect was detected in the whole group (Table 1) .

Overview of iron intervention studies in nonmalarious areas.

In summary, iron intervention studies in nonmalarious regions can be divided into two types, i.e., parenteral and oral. Parenteral iron administration at birth carries significant risk of severe sepsis and meningitis (Barry and Reeve 1977 , Farmer and Becroft 1976 ). Delay of this risky practice to the postneonatal period dramatically reduces such risks. Paradoxically, when parenterally dosed infants were followed long-term, subsequent morbidity due to respiratory, gastrointestinal and other infections was, on balance, reduced (Table 1 ; Fig. 1 ). This overall direction of benefit is shared with the oral supplementation studies (Figs. 1 2 3) .

Oral iron supplementation was delivered by milk or cereal fortification in six of the studies reviewed. Two large studies carried out earlier in the past century both indicated that iron fortification carried an impressively lower risk of respiratory infections and a less clear reduction in diarrheal disease (Figs. 2 , 3) . This promise in these unreliable studies was not realized in four iron food-fortification studies performed in the 1990s that failed to show effects either way (Table 1) . Perhaps the real message to be drawn from the latter studies was that breast milk is best. Although, in principle, more careful work is required in nonmalarious regions, it is unlikely in practice that much more will now be learned with fortification studies. There is a strong case for iron fortification of formula milk and no evidence at all that this fortification per se increases infectious morbidity.

If any area is understudied in nonmalarious regions, it is the effect of oral iron supplementation on infectious morbidity in breast-fed infants. Three oral iron therapy interventions to mainly older children gave conflicting results. Two interventions showed a reduction in infectious morbidity and a third did not detect any change except an increase in dysentery in infants.

The least that can be said of the evidence from iron supplementation studies in the nonmalarious areas in the 1990s is that there is little or no evidence of harm. The study of Chwang et al. 1988 was the only one to show clear evidence of benefit. From the public health point of view, arguments for oral iron supplementation in nonmalarious areas should be determined more by its other known benefits.

	Long-term studies in malaria-endemic regions

TOP ABSTRACT INTRODUCTION Criteria for inclusion of... In vitro studies of... Observational studies on iron... Iron intervention does not... Long-term prospective controlled... Long-term studies in malaria... Causal relationships: do they... Is there a connection... SUMMARY DISCUSSION REFERENCES

 
Controlled studies of iron supplementation have been conducted in malarial countries since the late 1970s, with some comment about infectious outcome (Oppenheimer 1998 ). Many of these studies recorded important laboratory outcomes of the intervention, such as hemoglobin change and parasite prevalence. These surrogate outcomes have been dealt with elsewhere (Shankar et al. 2000 ) and are not discussed further here. Eleven studies had quantitative infectious morbidity outcome data in some form (Adam 1996 , Berger et al. 2000 , Gebreselassie 1996 , Harvey et al. 1989 , Menendez et al. 1997 , Murray et al. 1978 , Oppenheimer et al. 1986a and 1986b , Rice et al. unpublished data, 1998, Smith et al. 1989 , van den Hombergh et al. 1996 ).

The earliest of the prospective studies, the report of Murray et al. (1978) of infectious recrudescence during iron therapy, has already been mentioned under immediate treatment effects (Tables 2 and 3; Fig. 4 ). This study followed up only for 30 d, the final observations coinciding with the end of supplementation.

View larger version (18K): [in this window] [in a new window]   Figure 4. Iron trials in malarious regions. Outcome: clinical attacks of malaria. Method of administration: parenteral iron: infants (1: Oppenheimer et al. 1986a ); Oral supplementation: anemic preschoolers (3: Adam 1996b , Smith et al. 1989 ); oral supplementation: anemic school children (5: Gebreselassie 1996 , Harvey et al. 1989 ); oral supplementation: anemic adults (7: Adam 1996a , Murray et al. 1978 ). Method of morbidity quantification: prospective regular active case detection in field; field clinical + slide assessment (all).

A large prospective, randomized, double-blind, placebo-controlled trial of parenteral iron supplementation in infancy was carried out in Papua New Guinea, in the early 1980s by Oppenheimer and colleagues (1984a , 1984b , 1986a and 1986b) . It had been established that iron deficiency was prevalent among infants in the study population and that malarial transmission was intense. To avoid the known risks of iron therapy in the neonatal period (Barry and Reeve 1977 ), a single dose of iron dextran (150 mg elemental iron) was administered at 2 mo of age to the treatment group (n = 236); control infants (n = 250) received an injection of sterile pyrogen-free saline. Infants were reexamined and relevant blood samples were taken 1 wk after the injection and at 6 and 12 mo of age. All admissions to hospital were documented (Tables 2 and 3 ; Figs. 4 5 6 7 8 ).

View larger version (11K): [in this window] [in a new window]   Figure 5. Iron trials in malarious regions. Outcome: respiratory infections. Method of administration: parenteral iron to infants (1: Oppenheimer et al. 1986a ); oral supplementation anemic preschoolers (2: Rice et al. personal communication, and 3: Adam 1996 ). Method of morbidity quantification: prospective regular active case detection (all); full clinical assessment in field (1); oral recall (2 and 3).

View larger version (14K): [in this window] [in a new window]   Figure 6. Iron trials in malarious regions. Outcome: passively detected infectious morbidity in age-defined cohorts of infants. Method of administration: parenteral iron (1: Oppenheimer et al. 1986a ); oral supplementation (2: Menendez et al. 1997 ). Method of morbidity quantification: prospective passive case detection using clinic visits (OPD) and admissions for severe infection full clinical and laboratory assessment (1); oral recall (2). Abbreviations: AOM, acute otitis media; SLRI, severe lower respiratory infection; OPD, outpatient detection.

View larger version (14K): [in this window] [in a new window]   Figure 7. Iron trials in malarious regions. Outcome: diarrheal infections. Method of administration: parenteral iron to infants (1: Oppenheimer et al. 1986a ); oral supplementation to anemic preschoolers (2: Rice et al. personal communication, and 3: Adam 1996 ); oral supplementation to anemic school children (4: Harvey et al. 1989 ). Method of morbidity quantification: prospective regular active case detection (all); full clinical assessment in field (1 and 3); oral recall (2 and 4)

View larger version (11K): [in this window] [in a new window]   Figure 8. Iron trials in malarious regions. Outcome: other infectious disease. Method of administration: parenteral iron to infants (1: Oppenheimer et al. 1986a ); oral supplementation to anemic preschoolers (2: Adam 1996b ); oral supplementation to anemic women (3: Adam 1996a ). Method of morbidity quantification: regular prospective active case detection with full clinical assessment in field (1–3)

    Increased malaria associated with increased pneumonia morbidity rates. Although increased malaria rates both in hospital admissions and in the field were the most clearly demonstrable morbid effect of iron supplementation found in this study, clinical attacks of malaria were a relatively less common primary reason for admission than acute lower respiratory infections (16 vs. 63%) (Table 3 ; Fig. 6 ). All 12 deaths in the study were due primarily to pneumonia (5 iron group, 7 placebo).

Rates of admission for pneumonia and time spent in hospital with pneumonia were higher in the iron dextran group (Table 3 ; Fig. 6 ) (Oppenheimer et al. 1986b ). Because this conflicts with the results of longitudinal studies reported previously in infants from nonmalarious areas, it is worth examining the possibility that malaria might have had an effect on susceptibility to pneumonia. Indirect evidence is available for this, i.e., 89% of pneumonia admissions had evidence of malaria (blood slide positive, significant splenomegaly or both), a much higher rate than that observed among the study children in the field. Support for the idea of a promoting effect of malaria on other infectious morbidity, particularly pneumonia, through a presumed influence on immune susceptibility comes from a recent study in the same community (Allen et al. 1997 ). Studies with the permethrin-impregnated bed net have also shown a dramatic reduction in nonmalarial as well as malarial infectious morbidity, again supporting the protean nature of malaria (Alonso et al. 1991 ). If such a close interaction between malaria and pneumonia is substantiated, it will go a long way to explain the opposite effects of iron supplementation on respiratory disease in temperate and tropical countries. It also underlines the importance of measuring all infectious morbidity, not just malaria.

    Hyperferremic effects of parenteral iron are short-lived. It has been argued that any general interpretation of the results of the Oppenheimer study for iron intervention in malarious areas is limited because a parenteral iron preparation was used. This position may be overstated. At the current state of knowledge, the immediate hyperferraemic effects of parenteral iron do not last longer than several weeks (see above). Further, no significant morbidity difference was recorded at the 1-wk check after the injection, and clinically significant hyperferremia was not noted at the subsequent field follow-up visits up to 1 y of age. Differences in morbidity observed all relate to periods well after the intervention.

Oral iron intervention.

    Effect of age and malarial immunity. A more relevant issue is that the study group used by Oppenheimer was an age-defined cohort of infants aged 2 mo who may be regarded as less immune than other children in malaria-endemic areas. In contrast, in a report of a 16-wk study in the same area, Harvey et al. (1989) failed to show any adverse effects of oral iron supplementation to prepubescent schoolchildren, particularly in relation to malaria indices (Figs. 4 and 7). Morbidity assessment, however, was based on mother’s recall (Table 3) . Harvey and colleagues speculated that acquired malarial immunity in their schoolchildren may have masked the potential interaction between iron and malaria. Harvey’s report is now in a minority. Two subsequent intervention studies in schoolchildren that gave oral iron for anemia (Adam 1996 , Smith et al. 1989 ) and a further two in adults (Adam 1996 , Murray et al. 1978 ) have shown an increased risk of clinical malaria. One other showed a similar trend that did not achieve statistical significance (Table 3 , Fig. 4 ) (Gebreselassie 1996 ).

The potential effect of age on immunity to malaria is actually a strong argument in favor of using age-defined cohorts in randomized controlled trials. The Tanzanian trial of Menendez et al. (1997) was the only other childhood trial in the literature to use an ag