Introduction

Numerous studies have established a relationship between iron deficiency anemia (IDA) and developmental delays in infants and children (Grantham-McGregor & Ani, 2001). Randomized clinical trials have demonstrated that iron deficiency anemia seems to be related to concurrent and future risk of poor development, however, the research has been less conclusive in its ability to attribute cause and effect to this association. Reasons include differences in the populations studied in terms of their ages as well as other nutritional, health and psychosocial problems, all of which are potential confounders clouding the nature of the IDA–behavior relationship. Furthermore, the degree and chronicity of IDA varied as did the behavioral and cognitive outcomes studied. Finally, there has been substantial variability in the experimental designs (Pollitt et al, 1983, 1986; Seshadri & Gopaldas, 1989; Lozoff et al, 2000).

One of the difficulties in attempting to understand the effects of IDA on developmental outcomes has been that much of the research has focused on broad indicators of development (ie BSID, developmental quotients); these may not adequately measure the putative effects of IDA (Pollitt, 2001). As has been noted (Pollitt et al, 1986), these tests tap different abilities at different ages; thus, it is impossible to understand the precise nature of the effect of IDA on cognition. Furthermore, many significant findings have suggested that more specific cognitive processes, such as attention, which can influence speed of mental processing, response times, and learning under certain conditions, are affected by IDA. The effect of such processes may not translate directly to performance on global developmental tests. In addition, while much of the research has focused on infants (under age 2 y) there is some evidence that the effects of IDA may be of particular importance during the preschool period, particularly because preschool cognitive abilities have been shown to significantly affect school achievement (Pollitt et al, 1993); thus, anything that adversely affects cognitive abilities during this time has the potential for adversely affecting subsequent school achievement.

To date only four treatment trials have examined the effects of IDA on development among preschoolers. In three of these trials the samples were drawn from populations in which undernutrition is common (Pollitt et al, 1986; Soewondo et al, 1989; Seshadri & Gopaldas, 1989) and in one the sample was drawn from a generally well-nourished population (Pollit et al, 1983). One study of male preschoolers in India reported that anemic preschoolers treated with iron but not placebo exhibited significant improvements in verbal (+9.8 points) and performance IQ scores (+19.4) (Seshadri & Gopaldas, 1989). Three other studies focused on specific cognitive processes including tests of discrimination learning, oddity learning, and short-term memory. All three studies showed that children with IDA performed significantly worse on one or more discrimination learning tasks at baseline (Pollitt et al, 1983, 1986; Soewondo et al, 1989). After supplementation with iron however, differences between iron replete and iron-deficient anemic–iron-supplemented groups were no longer significant due to improvements among those IDA treated with iron (Pollitt et al, 1983, 1986; Soewondo et al, 1989). Performance on the discrimination learning tasks is hypothesized to reflect both the child's ability to attend to the relevant features of the task and learning rate (Pollitt et al, 1983). Because of the specific forms of the discrimination learning tasks that were affected, the authors concluded that effects on this test reflected a deficit in the IDA preschooler's ability to attend to relevant information in a problem–solving situation, not a learning rate deficit (Pollitt et al, 1983, 1986). In one of the studies, the authors suggested that it was visual attention specifically that was affected (Soewondo et al, 1989).

These studies also examined effects of IDA on the Oddity Learning tests; performance on these tests are thought to reflect conceptual learning (Pollitt et al, 1986). Conflicting results were found, both in terms of baseline differences between anemic and iron-replete children and the effect of iron treatment. In one of the studies (Cambridge, Massachusetts) no differences in performance on the oddity learning tasks were found between IDA and replete preschoolers (Pollitt et al, 1983), while in the others (Guatemala and Indonesia), pre–schoolers with IDA did significantly worse on one or more of these tasks (Pollitt et al, 1986; Soewondo et al, 1989). Iron treatment in IDA children was associated with improvements in one or more of the oddity learning tasks in Indonesian preschoolers (Soewondo et al, 1989) but not in the Guatemalan preschoolers (Pollitt et al, 1986).

These four studies provide support for the hypothesis that iron treatment of iron–deficient anemic children can im-prove cognitive function during the preschool period. There were consistent effects on the discrimination learning tasks which were interpreted as reflecting an effect of IDA on attention but it is unclear whether the speed of information processing, or ability to discriminate is mediating this effect. It is also unclear whether conceptual learning is affected by iron deficiency given the conflicting findings evidenced by these studies.

The present study was undertaken to assess whether a low-dose iron supplement would affect vigilance (ie sustained attention), attention and conceptual learning of mildly IDA, but otherwise well-nourished healthy preschoolers in Greece. It differs from previous studies on preschool-aged children in several ways. The sample was drawn from a generally well-nourished population, there were stringent exclusion criteria (ie based on birthweight, IQ, growth, and blood lead) there was random assignment to iron or placebo, and iron was administered in a physiological rather than a pharmacological dosage. Moreover, this study focuses on specific measures of attention, speed of information processing and conceptual learning, rather than on global measures of development. It is the first study that uses cognitive tests in preschoolers which can simultaneously do the following: separate effects of iron on the speed (reaction time) with which an individual processes information when completing elementary cognitive tasks from the ability to discriminate selective stimuli and examine sensitivity, specificity, and accuracy of responses. Finally, this study controls for the potential confounding effects of changes in vitamin status as the result of supplementation because all subjects were started on a children's multivitamin (MV) prior to randomization.

Methods

Study design and study site

The original study design was a double-blind randomized controlled trial of iron supplementation. The term 'supplementation' rather than 'treatment' was used because the latter term would not be appropriate for the group with good iron status included in this study. It consisted of 2 months of baseline and 2 months of intervention (see Figure 1). Multivitamin supplementation was initiated at the beginning of the baseline period (T0) at all Day Care Centers (DCCs). At baseline and at the end of the intervention (T2) (MV+iron or MV alone), computerized tests of attention and learning were administered to all subjects. At the end of the baseline period (T1) subjects were randomly assigned to receive either 15 mg iron per day (multivitamins + iron) or placebo (multivitamins alone). Randomization was conducted within iron class and day care center. Day care providers and researchers were blinded as to the subjects' iron status and treatment assignment. Randomization was carried out within each day care center to minimize potential for confounding by subject variability within each DCC.

Figure 1.

Time points of study. T0, screening (hematology and enrollment criteria), and enrollment into study; T1; randomization; T2; post-treatment hematologic and cognitive testing.

Full figure and legend (7K)

Two criteria were used to choose the study site: (a) the widespread use of non-iron-fortified infant formulas at the time of the study and (b) a relatively high prevalence of iron deficiency anemia among infants and children (Kattamis, 1982). The sample was drawn from preschoolers attending nine public DCCs in Athens, Greece. At the time of this study provision of day care for working mothers was subsidized in Greece and attendance was independent of family income. To ensure compliance, supplements were given to the children at the DCCs (ie 5 days per week).

The DCCs were open from 0700 to 1600 and they provided two meals and two snacks per day. Children usually began attending these centers when they were 2.5–3 y old. This study was approved by the University of California Davis Human Subject's Review Board and the Committee of Medical Ethics of the Institute of Child Health in Athens.

Sample

Children, 3–4-y of age (N=150) were screened and those meeting the following criteria (n=124) were considered eligible to be included in the trial: birth weight 2500 g, currently healthy, benign past medical history, IQ (based on the Griffith's Mental Development Scales) 1 s.d. below the age-adjusted mean (Kattamis, 1982), blood Pb 200 ppb and height, weight and head circumference for age 10th percentile (US National Center for Health Statistics, 1976). At time of this study the aforementioned lead value corresponded to the Center for Disease Control's cut-off of 20 g/dl. The authors acknowledge that this has now changed to 10 g/dl (CDC, 2000) (All, except one child, who was excluded, had lead values under 53 ppb).

The fact that the IDA in this population was of mild nature made certain classification of iron status difficult for a large proportion of the children; thus, the results presented here were not based on the stringent RCT that was originally designed and implemented because it was necessary to use response to treatment to reclassify subjects in order to minimize misclassification. To prevent heterogeneity within iron class, the final composition of the iron groups followed two steps, one before and one after the intervention. Just prior to randomization, a preliminary classification of iron status was used as follows: (a) suspected IDA (N=19)=Hgb 115 g/l and at least one of the following: transferrin saturation (TS) 16% or serum ferritin <12 g/l, (b) suspected good iron status (N=49)=Hgb 120 g/l and at least one of the following: TS 20% or ferritin 12 g/l, (c) Iron status uncertain (N=55)=those not falling into either of the above classes. One 'suspected IDA' and three 'suspected good iron status' children dropped out of the study prior to randomization; thus sample sizes were 18 and 46, respectively.

After establishing this classification, the subjects, within iron class, were randomly assigned to the treatment (iron+multivitamins) or placebo (multivitamins). After the intervention, response to treatment among those receiving iron was utilized to further validate our original classification and subjects were reclassified as follows. For Hgb this cutoff was based on the 5th %tile for the population of 3–4-y olds according to NHANES II (Pilch & Senti, 1984). Iron-deficient anemic was defined as: (1) baseline Hgb <112 g/l and TS <16% and Ferritin <12 g/l or (2) Hgb rise of >10 g/l (T2–T0) with iron supplementation. Good iron status was defined as baseline levels of Hgb >120 g/l and either TS >20% or serum ferritin >12 g/l. Among good iron status, iron-treated subjects, only those who did not respond to treatment were retained for the final analysis. Nonresponse was defined as a change in Hgb of 7.5 g/l (1 s.d. of the change among iron-treated subjects). Table 1 indicates the origin of the final sample as well as those who could not be classified. The final sample consisted of 21 anemic (14 iron-treated and seven placebo) and 28 good iron status (18 iron-treated and 10 placebo) subjects. Hereafter, all reported statistics refer to these two final groups. Thus, two alternate techniques were used to classify subjects' as having IDA: either a strict IDA criterion or evidence of a response to treatment. This approach improved the confidence with which classification of iron status could be made. There were 71 children whose iron status remained uncertain and thus could not be classified as either anemic or as having good iron status (Table 1). Thus they were dropped from sub-sequent analyses.

Table 1 - Origin of final sample and those not classified.

Full table

Although this reclassification is not consistent with an 'intent to treat' analysis, it was necessary to be fully certain that those classified as deficient were truly deficient and those classified as having good iron status truly had good iron status. Given that the gold standard for the determination of iron deficiency anemia is response to iron treatment, the only way to ensure that classification was accurate was by assessing response to treatment when treatment was adequate; otherwise, a lack of change in hemoglobin would simply reflect an inadequate dosage.

Intervention

The iron intervention was 15 mg Fe in the form of ferrous fumarate plus multivitamins (MV); the placebo (MV) consisted of MV alone. Assignment was randomized within each DCC. The providers distributed the supplements (intervention or placebo) at the DCCs. Attendance was monitored during the supplementation period. The composition of the supplements is shown in Table 2. These supplements were kindly provided by the Pharmavite Corporation. Their contribution to this project is gratefully acknowledged.

Table 2 - Composition of iron supplement (multivitamins+iron) and placebo (multivitamins).

Full table

Variables and measurements

Hematology

Venous blood samples at baseline (T0) and postsupplementation (T2) were collected in the morning. A complete blood panel (hemoglobin, hematocrit, mean corpuscular volume, and red and white blood cell counts) was carried out using the automated Coulter 80S. Serum iron and iron binding capacity were assessed using kits from Boehringer Mannheim BmbH. Serum ferritin was measured using an enzyme immunoassay (Ferrizyme: Abbott Laboratories). Whole blood lead was determined by directly assaying the samples using the furnace attachment to the Perkin-Elmer AA.

Cognitive function

The assessment of cognitive function included a computerized test battery of five different tests: a simple reaction time test, a continuous performance task (CPT), and three oddity learning (OL) tasks. Simple reaction time measures the speed at which simple information is processed. For the simple reaction time test, the child is not asked to discriminate between any of the stimuli but to simply respond to all of them as quickly as he or she can. The CPT measures the speed of information processing when having to selectively respond to a specific stimulus; thus, it includes the speed of discrimination. It also measures the accuracy with which a child can discriminate and respond to selective stimuli. The CPT was selected because of its track record in detecting alterations in attention and its ability to assess information processing speed, including discrimination. Oddity learning is a test of the rate of conceptual learning, and was selected because previous research has suggested differences on these tasks.

Procedure

These tests were pretested among preschoolers in the United States and Greece to establish the suitability of the test stimuli, test format, and speed of presentation for this age. The tests were administered by either the first author or a trained tester. The trained tester had been educated as a primary school teacher in Greece. Both were blind to the child's iron status and treatment assignment. The elapsed time between the T0 hematological assessment and the baseline cognitive assessment ranged from 2 to 4 weeks for the CPT and simple reaction time tests (3–6 weeks prior to randomization) and 6–8 weeks for the oddity learning task (0–2 weeks prior to randomization).

Simple reaction time test

The simple reaction time test assesses the speed of response. The importance of simple reaction time is that it does not involve discrimination or thought. One figure flashes repeatedly on the screen in intervals of 400 to 800 ms (a total of 18 times or trials per block). The subject is instructed to respond as quickly as possible by pressing the space bar each time the figure appears. Test results provide a simple reaction time (RT) score in milliseconds. After two training blocks (not included in any analyses), two blocks (of 18 trials each) were performed. At baseline, there was no significant difference in scores between the first and second blocks for any subgroup. The mean simple reaction time of the two blocks at each time point (T1 and T2) was used in the final analyses.

Continuous performance task (CPT)

The CPT measures the speed of information processing when having to discriminate between probes to selectively respond to a specific stimulus. It also measures the accuracy with which a child discriminates. Therefore, CPT reaction time includes cognitive processing and decision-making as compared to simple RT. This task consists of a variety of familiar figures (eg, a fish, a cat, a happy face) flashing randomly on the screen (a total of 18 trials per block) in intervals that randomly range from 400 to 800 ms. They remain on the screen for either 400 or 500 ms. The test consisted of three blocks. The child is instructed to press the space bar each time a probe (critical stimulus (CS) eg a cat) is flashed on the screen and to refrain from responding upon presentation of other stimuli (noncritical stimuli (NCS)(eg a happy face). The probe appears randomly 6 times per block (ie six CS and 12 NCS). Test results include errors of commission (ie responding to a NCS), errors of omission (ie not responding to a CS), and choice reaction time (ie speed of response when respond to a CS). In addition, the following combined measures of performance were calculated: (a) Specificity (# NCS to which subject refrained from responding/total NCS), (b) Sensitivity (# CS to which subject responded/total # CS), and (c) Accuracy (total # correct/total trials). Note that accuracy is a combination of specificity and sensitivity. Both the mean of the three blocks and the test results from each block were separately used as outcomes. Unless the results differed by block, the results reported are the mean of the three blocks.

Two additional indicators (efficiency and impulsivity) that combine accuracy and speed (Salkind & Wright, 1977) were also calculated. For this purpose, total errors and reaction time were transformed into z-scores based on the sample distribution. Efficiency was defined as total errors (z-score)+RT (z-score) and impulsivity as total errors (z-score) – RT (z-score). A lower score on efficiency is indicative of faster and more accurate performance. A higher score on impulsivity represents a faster response at the cost of accuracy.

In addition, since the difference between the speed of response in the choice vs simple reaction times is theoretically the time needed to discriminate the probe from the other figures, choice reaction time was regressed on simple reaction time. The residual, thought to reflect the amount of time the child takes to discriminate, was also used as an outcome variable.

Oddity learning (OL)

OL is a test of conceptual learning. More specifically it reflects how quickly (ie how many trials) a child forms categories of information in order to facilitate problem-solving. A set of three figures, two of which were identical and one different, appears on the screen. The task (or problem) consists of identifying which figure is associated with a computerized beep (positive feedback). The subject has to learn that this is always the odd figure. In the case of an erroneous response, the subject is told which figure would have been correct. There are 3 OL tasks, with 30 possible sets (or trials) of figures in each. The order of presentation (ie right, middle, and left) within each set and the order of the sets are random. In the first task, new figures are presented for each set; next the figures are repeated once (eg, a bee, a bee and a hive, might become a hive, bee and hive). In the third task, figures are repeated twice. The criterion for learning is seven consecutive correct responses.

Trials to the learning criterion and proportion correct were recorded for each task. In order to include the results of subjects not reaching the learning criterion for a particular task, the maximum value for trials to criterion was set at 37 (the sum of the maximum score (30) plus 7 (the learning criterion)). Because many subjects did not learn how to choose the correct figure during the first task (particularly at baseline) and were still trying to ascertain this in tasks 2 and 3, the total number of trials to reach criterion was also used as an outcome measure. This was the sum of the trials across all tasks.

To consolidate the results from the OL tasks, principal components analysis on the three tasks was conducted for the entire sample at baseline and postsupplementation. One factor with an eigenvalue >1 emerged at both baseline (56.2% of variance explained) and T2 (73.6% of variance explained). At baseline this factor loads high on trials to criterion in OL2 and OL3, but lower on trials to criterion in the first OL task, which suggests that it may not reflect initial learning but rather the transferal of learning to a more complex problem. At T2 however, this factor loaded similarly high on all three OL tasks.

To summarize, the outcome variables from the OL test were: (a) trials to criterion, (b) proportion correct for each task separately, and (c) two summary variables: a factor score, and the sum of all trials to criterion for the three tasks.

Statistical analysis

Because randomization occurred within iron class, the hematologic results for those with good iron status and anemic subjects were analyzed separately, as follows: (1) repeated measures analysis of variance of T0 and T2 hematology, (2) regression analysis (analysis of covariance format) of T2 values controlling for T0, and (3) analysis of covariance of the change scores (T2–T0).

For the cognitive outcomes, two sets of regression analyses were conducted. The first set examined the effect of supplementation separately for anemic and good iron status subjects. Age and baseline cognitive test scores were used as covariates. Inclusion of baseline test scores in the model partials out between-group differences at baseline, and allows for the best assessment of the effect of the intervention. The second set of analyses examined the main effects of both iron status and treatment and their interaction in the same regression model. This was done to compare the response to treatment in the anemic and iron-replete subjects, but can also describe the effect of treatment in each group. Because the results were similar, only the combined analyses will be reported here.

A two-tailed test for differences was used. For all main effects, the significance level used was P<0.05; results for the interactive term were reported if the P-value was <0.10 (Neter et al, 1985).

Results

The mean (s.d.) ages (y) of the anemic subjects and those with good iron status at baseline were significantly different (P<0.05); consequently, all subsequent analyses on cognitive outcomes were age-adjusted. There was no significant difference in mean age of the iron-treated and placebo groups; however, good iron status/Fe: 3.71, good iron status/placebo 3.72, anemic/Fe: 3.47, anemic/Pl: 3.61 y. The mean ages of mothers and fathers were 31.9 and 37.0 y, respectively. On average, both mothers and fathers had completed the 12th grade (the equivalent of high school in Greece). There were no significant differences by iron status or treatment assignment for any of the following potential confounders (Table 3): age of mother and father, years that the mother and father had lived in Athens, maternal or paternal educational level (y). There were also no significant differences in mean birth weight or number of weeks of breastfeeding between those with good iron status and anemic subjects (results not shown).

Table 3 - Parental socio-demographic characteristics by child's iron status and treatment assignment (mean (s.d.)).

Full table

Hematologic values

Iron status pre- and postsupplementation (means s.d.s), categorized by both iron status and treatment assignment, are shown in Table 4, in order to assess any differences between iron-treated and placebo groups. Baseline means of the anemic-placebo group for serum iron, TS, and ferritin were significantly lower than those for the anemic–iron-treated group (P<0.05), since the IDA placebo group had to meet the more stringent criteria for classification in that group, as response to treatment could not be used. There was no difference at T0 on hemoglobin. Groups classified as having good iron status were similar in all iron indicators at baseline (Table 4), however at T2, those treated with iron had significantly higher (P<0.05) mean serum iron and transferrin saturation than those given placebo.

Table 4 - Iron-deficient anemic and those with good iron status: means (s.d.) of hematological indicators of iron status pre (T0) and post (T2) supplementation.

Full table

The results described below on the repeated measures analysis of variance are not shown in a table and address the question of how treatment affected hematological changes over time for anemic and good iron status subjects:

Hemoglobin: among anemic subjects, there was a significant time by treatment interaction (F-value=34.53, P<0.0001) showing a significantly greater increase among anemic children treated with iron (mean increase=16 g/l) than among children in the placebo group (mean increase=0.3 g/l, P<0.0001 for difference). There was no time or time-by-treatment interaction for the subjects with good iron status.

Serum ferritin: among anemic, iron-treated subjects, serum ferritin increased (mean change=+8 g/l), but the change was significant only at P<0.10. Good iron status, iron-treated subjects exhibited a significant increase (mean change=+12 g/l; P<0.01) whereas the mean change among placebo subjects was not significant (+2 g/l).

Transferrin saturation: among anemic subjects there was no significant change over time in either treatment group. TS significantly declined between T0 and T2 in the good iron status–placebo subgroup (P<0.01), but not in the iron-treated group.

Cognitive tests

At baseline (T0) there were no significant differences between the anemic and good iron status subgroups' mean information processing (simple RT and CPT) scores, mean trials to criterion, or in the proportion reaching criterion in any of the OL tasks (Table 5). Unexpectedly, baseline proportion correct on the first OL task was higher for anemic in comparison to subjects with good iron status (0.81 vs 0.66, P<0.01). There were no significant baseline (T0) differences between iron-treated and placebo groups either within the good iron status or anemic group on any of the cognitive measures (results not shown). Table 6 depicts the R² and adjusted means for all the cognitive outcomes for which there was a statistically significant interaction between iron status and treatment. The question of primary interest is what the effect of iron supplementation was on the IDA subjects and how this compared to the good iron status group. The statistical model included baseline score, age, iron class, treatment assignment, and an interactive term between iron class and treatment assignment. Inclusion of the interaction term allows us to compare the performance of the following four groups: good iron status–iron treated, good iron status–placebo, IDA–iron treated, and IDA–placebo. Thus, the effect of the treatment within each iron class as well as a comparison between iron class and treatment can be assessed.

Table 5 - Mean (s.e.) baseline scores on information processing and oddity learning tests for anemic and good iron status subjects.

Full table

Table 6 - Adjusted means for significant iron status^*treatment interactions.

Full table

Simple reaction time (RT)

There was no treatment effect on simple reaction time, either in the good iron status or IDA group, nor was there a significant interaction between iron status and treatment. The statistical model included iron status, treatment assignment, baseline score, age, and the interaction term (iron status by treatment assignment). The statistical model did not explain a significant portion of the variance in simple RT (data not shown).

Continuous performance task (CPT)

The models explained a significant portion of the variance (Table 6) in choice (CPT) reaction time, reaction time residual, CPT specificity (P<0.05), accuracy, and efficiency (P<0.01) but not in mean CPT sensitivity (data not shown). After iron treatment, the anemic subjects made significantly fewer errors of commission (14% higher specificity, P<0.05) and exhibited 8% higher accuracy (P<0.05) compared with those given placebo (Table 6 for adjusted means) although sensitivity was similar between treatment groups. Thus, they were better at refraining from responding to irrelevant stimuli than those anemic children given placebo. Among those who had good iron status, no effect of treatment was seen on sensitivity (data not shown), specificity, or accuracy (Table 6). For choice reaction time, although the interaction was significant (P<0.05), the differences by treatment group within each iron class were only significant at P<0.10. Nevertheless, effects of treatment on choice reaction time were in opposite directions for those with good iron status and IDA. Iron supplementation among those with IDA was associated with a faster response as compared to the placebo group (647 vs 685 ms respectively, P<0.10), whereas iron supplementation among those who had good iron status was associated with a slower response (666 vs 636 ms, respectively, P<0.10). Finally, those anemic children given iron were significantly more efficient than those given placebo (mean efficiency score=-0.55 vs 0.54, P<0.05). Thus, the anemic-iron-treated subjects were not only faster but made a fewer number of errors than anemic subjects given placebo. Again no effect of the treatment was seen on CPT efficiency among subjects with good iron status.

As expected, age was inversely associated with efficiency and reaction time residual, and positively associated with accuracy (P<0.05; data not shown) indicating that older children were faster and more accurate then younger children. Baseline performance was positively associated with performance in the postsupplementation period for CPT sensitivity and accuracy (P<0.05).

Oddity learning (OL) task

None of the models explain a significant portion of the variance in the OL outcomes (ie trials to criterion or proportion correct for OL1, OL2, or OL3). There were also no statistically significant interactions between iron status and treatment for any of the OL outcomes. There was however, what could be considered a trend; for the OL1, anemic iron treated children tended to have a higher proportion correct than those given placebo but this was significant only at P<0.10 (Table 6).

Analysis by intent to treat

These data were also reanalyzed using intent to treat among those initially "suspected anemic" and those with 'suspected good iron status'. Improved performance after iron supplementation was observed among those classified as "suspected anemic" on choice reaction time (P<0.05), reaction time residual (ie speed of discrimination; P<0.05)) and CPT efficiency (P<0.10). Conversely, there was no effect found on sensitivity, specificity, or accuracy of response on the CPT or on the OL outcomes.

Discussion

One important advantage to using measures of information processing, such as reaction time is that they are less sensitive to educational and cultural differences than most other tests used in the past (Pollitt, 1993). In this study, we examined the effects of iron supplementation of those with good iron status and IDA preschoolers on speed of information processing, accuracy of discrimination (CPT), and conceptual (oddity) learning. Those anemic preschoolers supplemented with iron responded more accurately in the CPT than those anemic preschoolers given placebo. They responded significantly more accurately because they were better at refraining from responding to irrelevant stimuli (ie better specificity) than the anemic placebo preschoolers, not because they responded to a greater proportion of the critical stimuli (ie sensitivity). Thus, they appeared to be better able to discriminate between the critical stimulus to which they were instructed to respond and better able to refrain from responding to irrelevant stimuli. Similarly, when examining accuracy and speed of response simultaneously (ie efficiency) anemic preschoolers given iron were more efficient (ie faster and more accurate) than those given placebo. They also performed similarly to the good iron status groups on the CPT. Conversely, significant effects of the iron supplement were not observed on those who had good iron status on any of the cognitive tests used in this study.

In contrast, after the intervention, the anemic subjects who received placebo performed more poorly on these tasks than did the other three groups. Whereas the anemic–iron-treated group was as fast and accurate as both of the good iron status groups, the anemic placebo group was slower to respond and made a greater number of errors. From the analysis of specificity and sensitivity, it appears that the errors most affected by supplementation were errors of commission (ie specificity). In other words, the anemic placebo preschoolers responded more frequently to stimuli to which they should not have responded. This suggests that they were indiscriminate in their responses.

The results observed in this study are in agreement with previous research that has shown an improvement in cognitive measures in infants (Oski & Honig, 1978; Walter et al, 1983; Driva et al, 1985; Idjradinata & Pollitt, 1993; Harahap et al, 2000; Lozoff et al, 2003), preschoolers (Pollitt et al, 1983, 1986; Seshadri & Gopaldas, 1989; Soewondo et al, 1989;) and school age children (Seshadri & Gopaldas, 1989; Bruner et al, 1996; Lynn & Harland, 1998) after treatment with iron.

This study's findings advances our understanding in a number of ways particularly in delineating the type of attentional processes affected by IDA. It builds upon previous studies that have used other measures of attention in infants (Lozoff et al, 1998, 2003) and in school-aged children (Bruner et al, 1996; Lozoff et al, 2000). IDA infants have previously been found to be 'less attentive' than iron-replete infants, although there was no effect of iron treatment on this outcome. Similarly, no effect of iron therapy was found on the 'Brief Test of Attention' in high school children, although iron treatment was found to affect verbal learning (Bruner et al, 1996). Finally, school-aged children who had chronic iron deficiency anemia during infancy but were treated with iron performed more poorly on tests of attention (specifically selective attention) than their nonanemic peers 10 y later (Lozoff et al, 2000). In this aforementioned study, Lozoff et al (2000) noted that those who had formerly been chronically iron deficient showed a delay in developing the ability to attend selectively and inhibit attention to the irrelevant (Lozoff et al, 2000). Moreover, a recent prevention trial on a large sample of infants found that those not given iron (compared to those given iron) had a longer looking time on the Fagan test but there was no difference in novelty preference (Lozoff et al, 2003). The authors drew the inference that those who did not receive iron had less efficient information processing (Lozoff et al, 2003). The present study's findings are consistent with these previously reported findings since the anemic–placebo group committed a great number of errors in which they responded to irrelevant stimuli, while the anemic–iron-treated group did better both by being faster and by refraining from responding to irrelevant stimuli, thus being more selective in their response. However, it takes a step further than simply supporting the overall importance of attention because it points specifically to selective attention as being affected by IDA.

Information processing tasks involve a number of processes including speed, and discrimination. Based on our data reported here, one hypothesis is that the ability to discriminate may be sensitive to iron deficiency anemia. That is, effects were seen on the CPT (ie reaction time and accuracy) whose performance is dependent on the child's ability to discriminate one probe from the rest but not on simple reaction time which does not require discrimination. The ability of a child to discriminate has implications for the basic school readiness skills at this age, such as letter recognition and learning to read (Nurss, 1987). Alternatively, as some have suggested, it may be that it is the ability of preschoolers to inhibit their attending to and potentially responding to the irrelevant which has been affected by IDA (Lozoff et al, 2000). Other studies have used the OL task in preschoolers; however, mean number correct rather than mean trials to criterion has typically been used (Pollitt et al, 1983; Soewondo et al, 1989). One study showed no effects of iron treatment on performance on the OL task (Pollitt et al, 1983), while the second showed an effect of iron treatment on the third OL task (that is when the figures were repeated twice) (Soewondo et al, 1989).

It is paradoxical that there were no significant differences in cognitive outcomes at baseline between anemic and good iron status subjects. Because of the pattern of the findings, one explanation is that the normal improvement over time that occurs with age and previous exposure to the task is detrimentally affected by iron deficiency anemia. Studies in experimental animals have demonstrated a similar phenomenon: previous exposure to a confusing learning task resulted in significantly better performance among nonanemic animals but did not benefit iron-deficient animals (Massaro, 1982). The present study is consistent with these experimental studies on animals in suggesting that iron treatment prevented a decline in cognitive test performance among the anemic children. Evidence for this stems from the fact that adjusted CPT efficiency scores declined in the anemic placebo group (0.41 z-score decline) over time whereas the other groups either improved or remained the same. Similarly, for the OL1 all subgroups improved through time except for the anemic placebo group. This is also consistent with the research that has shown that among malnourished infants and children from disadvantaged populations, there are declines in relative cognitive performance with age (Saco-Pollitt et al, 1985).

An alternative explanation for the differences postsupplementation but not at baseline is motivation. The first time the children were exposed to these tests, they had never before seen a computer. The experience was novel and exciting. By the second wave of computerized tests, the novelty had worn off and the children may not have been as interested or excited about 'playing with the computer'. Previous research has reported that anemic children are more fearful, unreactive to usual stimuli, more solemn and less involved than their nonanemic counterparts (Grantham-McGregor & Ani, 2001). Although observations of subjects during testing was not a component of this study, it is possible that poorer motivation may be a manifestation of what others have observed. Other possibilities include differences in fatigue, frustration, or other aspects of affect during testing.

Many mechanisms have been proposed by which iron deficiency anemia can lead to altered cognition (Grantham-McGregor & Ani, 2001). These include changes in the structure and function of the CNS; this has been primarily supported by research in animals not children. Two frequently discussed mechanisms are effects on the dopaminergic system and effects on myelination (Grantham-McGregor & Ani, 2001). A comprehensive review of the biological role of iron in neuronal functioning noted that the dopaminergic system has been consistently sensitive to experimental changes in iron status (Beard, 2001). Linking iron's role in the brain and behavioral effects observed has been challenging. However, it has been noted that the process of attention to environmental information is dependent on rates of dopamine clearance from the interstitial space, and that this suggests that iron status may affect behavior through dopamine metabolism. Some have speculated that effects on selective attention are consistent with current understanding of iron's role in pre–frontal–striatal and hippocampal systems (Lozoff et al, 2000); dopamine has been noted to play an important role in the first of the aforementioned systems. Thus, one hypothesis is that the effects on information processing observed in this study may be due to effects of iron on the dopaminergic system. If this is the case, then it is the ability to inhibit response to the irrelevant, rather than ability to discriminate, which is being affected by iron deficiency. Others who have provided evidence for effects of iron deficiency anemia on auditory brain stem responses (ie prolonged central conduction time among anemic infants) and longer latencies on visually evoked potentials (Algarín et al, 2003) in infants and preschool children have suggested these effects provide support for the hypothesis that IDA has effects on myelination. However, they acknowledged the limitations of their data in pointing to myelination because they did not actually conduct anatomic neuroimaging (Algarín et al, 2003). In this study it is unlikely that changes in mylenation can explain the effects of iron on attention because, as noted, potential remediation in hypomyelination is likely a slow process (Grantham-McGregor & Ani, 2001).

One limitation to this study is that the randomized controlled trial design was not completely maintained because the primary analytic strategy was not intent to treat. The justification for this lies in the fact that response to treatment was the only method to minimize misclassification of iron status given that iron deficiency anemia in the sample was relatively mild. Adjusting for baseline cognitive performance in assessing effects of treatment on post-treatment performance was used in this study however; this is an accepted method used to reduce effects of confounding (Anderson et al, 1980). The fact that the results were similar on several of the outcomes when using intent to treat suggests that the results reported in this paper can be attributed to iron not to confounding. A related limitation is the possibility that the lack of response in the placebo group in terms of improvement in neurocognitive scores may be in part because the iron status of this group was poorer at the outset, indicating that they likely had more chronic iron deficiency anemia.

Another limitation is the small sample size which increases the risk of a Type II error (ie accepting the null hypothesis that there is no effect of iron supplementation when there is). This was not a problem for the results of the CPT test, since significant effects of iron supplementation were found. However, it is troublesome that in the case of the first OL task, the difference between treatment groups for the iron-deficient anemic subjects was not statistically significant (ie IDA-iron treated group took 5.74 fewer trials than IDA-placebo to reach criterion, P<0.10). Statistical power calculations indicate that to detect this difference at a P level of 0.05 with the statistical power of 0.90, the sample size would have had to be 12 subjects per anemic subgroup. On the other hand, if a one-tailed test of significance were carried out, this difference of 5.74 trials to criterion would have been significant at P<0.05. This points to the need for additional studies examining these outcomes in relation to iron deficiency anemia in preschool children.

In conclusion, this study demonstrated that iron supplementation of iron-deficient anemic preschool children improves a number of skills related to information processing. Further, these skills in which children with iron deficiency anemia showed delay could potentially be related to later learning. This is an important link to understanding mechanisms by which IDA has been associated with poor achievement.

References

1. Algarín C, Peirano P, Garrido M, Pizarro F & Lozoff B (2003): Iron deficiency anemia in infancy: long-lasting effects on auditory and visual system functioning. Ped. Res. 53, 217–223. | Article |
2. Anderson S, Auguier A, Hauck WW, Oakes D, Vandaele W & Weisberg H. (1980): Analyzing data from premeasure/postmeasure design. Statist. Methods Compar. Stud. 235–247 (Chapter 11).
3. Beard JL (2001): Iron biology in immune function, muscle metabolism, and neuronal functioning. J. Nutr. 131, 568S–580S. | PubMed | ISI | ChemPort |
4. Bruner A, Joffe A, Duggan A, Casella J & Brandt J (1996): Randomized study of cognitive effects of iron supplementation in non-anaemic iron-deficient adolescent girls. Lancet 348, 992–996. | Article | PubMed | ISI | ChemPort |
5. Centers for Disease Control and Prevention (CDC) (2000): Blood lead levels in young children—United States and Selected States, 1996–1999. MMWR 49, 1133–1137.
6. Driva A, Kafatos A & Solman M. (1985): Iron deficiency and the cognitive and psychomotor development of children: a pilot study with institutionalised children. Early Child Dev. Care 22, 73–82.
7. Grantham-McGregor S & Ani C. (2001): A review of studies on the effect of iron deficiency on cognitive development in children. J. Nutr. 131, 649S–668S. | PubMed | ISI | ChemPort |
8. Harahap H, Jahari AB, Husaini MA, Saco-Pollitt C & Pollitt E. (2000): Effects of an energy and micronutrient supplement on iron deficiency anemia, physical activity and motor and mental development in undernourished children in Indonesia. Eur. J. Clin. Nutr. 54 (Suppl), S114–S119. | Article | PubMed |
9. Idjradinata P & Pollitt E. (1993): Reversal of developmental delays in iron-deficient anaemic infants treated with iron. Lancet 341, 1–4. | Article | PubMed |
10. Kattamis C. (1982): Prospects, inabilities and achievements in epidemiologic studies on iron deficiency and iron deficiency anemia. Paediatrician 11, 121–135. | PubMed |
11. Lozoff B, Klein NK, Nelson EC, McClish DK, Manuel M & Chacon ME (1998): Behavior of infants with iron deficiency anemia. Child Dev. 69, 24–36. | PubMed |
12. Lozoff B, Jimenez E, Hagen J, Mollen E & Wolf AW (2000): Poorer behavioral and developmental outcome more than 10 years after treatment for iron deficiency in infancy. J. Pediatr. 105, 1–11.
13. Lozoff B, De Andraca I, Castillo M, Smith JB, Walter T & Pino P (2003): Behavioral and developmental effects of preventing iron-deficiency anemia in healthy full-term infants. Pediatrics 112, 846–854. | PubMed |
14. Lynn R & Harland EP (1998): A positive effect of iron supplementation on the IQS of iron deficient children. Person Ind. Diff. 24 (6), 883–885. | Article |
15. Massaro TF (1982): Effect of iron deficiency in learning. In Iron Deficiency: Brain Biochemistry and Behavior eds Pollitt E, Leibel RL New York, NY: Raven Press. pp 125–140.
16. Neter J, Waserman W & Kutter MH (1985): Applied linear statistical models, pp 700–702. Homerwood, IL: Richard D. Irwin Inc.
17. Nurss J (1987): Readiness for Kindergarten. ERIC Digest: ERIC Clearinghouse on Elementary and Early Childhood Education.
18. Oski FA & Honig AS (1978): The effects of therapy on the developmental scores of iron-deficient infants. J. Pediat. 92, 21–25. | PubMed |
19. Pilch SM & Senti FR eds. (1984): Assessment of the Iron Nutritional Status of the US Population based on data Collected in the Second National Health and Nutrition Examination Survey, 1976–1980. Report of the Life Sciences Research Office. Bethesda, MD: Federation of American Societies for Experimental Biology.
20. Pollitt E (1993): Iron Deficiency and Cognitive Function. Annu. Rev. of Nutr. 13, 521–537. | Article |
21. Pollitt E, Leibel RL & Greenfield DB (1983): Iron deficiency and cognitive test performance in preschool children. Nutr. Behav. 1, 137–146.
22. Pollitt E, Saco-Pollitt C, Leibel RL & Viteri FE (1986): Iron deficiency and behavioral development in infants and preschool children. Am. J. Clin. Nutr. 43, 555–565. | PubMed |
23. Pollitt E, Gorman KE, Martorell R & Ingel P (1993): Early supplementary feeding and cognition. Monogr. Soc. Res. Child Dev. 58, 1–12. | PubMed |
24. Pollitt E (2001): The Developmental and Probabilistic nature of the functional consequences of Iron-Deficiency Anemia in Children. J. Nutr. 131, 669S–675S. | PubMed |
25. Saco-Pollitt C, Pollitt E & Greenfield DB (1985): The cumulative deficit hypothesis in the light of cross-culture evidence. Int. J. Behav. Dev. 8, 75–97.
26. Salkind NJ & Wright JC (1977): The development of reflection–impulsivity and cognitive efficiency. An Integrated Model. Hum. Dev. 20, 377–387.
27. Seshadri S & Gopaldas T (1989): Impact of iron supplementation on cognitive functions in preschool and school-aged children: the Indian experience. Am. J. Clin. Nutr. 50 (Suppl), 675–684. | PubMed |
28. Soewondo W, Husaini M & Pollitt E (1989): Effects of iron deficiency on Attention and learning processes in preschool children: Bangdung Indonesia. Am. J. Clin. Nutr. 50 (Suppl), 667–673. | PubMed |
29. United States Center for Health Statistics: NCHS Growth Charts (1976): Monthly Vital Statistics Report, Vol 25 (3) Suppl (HRA) 76-1120 Health Resources Administration, Rockville MA, June.
30. Walter T, Kovalskys J & Stekel A (1983): Effect of mild iron deficiency on infant mental development scores. J. Pediat. 102, 519–522. | PubMed |