Longitudinal Evaluation of Externalizing and Internalizing
Behavior Problems Following Iron Deficiency in Infancy
Feyza Corapci,1,* PHD, Agustin Calatroni,1MS, Niko Kaciroti,1PHD, Elias Jimenez,3MD, and
1Center for Human Growth and Development,2Department of Pediatrics and Communicable Diseases,
University of Michigan, and3Hospital Nacional de Nin ˜os
of both iron status in infancy and infant characteristics.
children who either had chronic, severe iron deficiency or good iron status in infancy were followed for 19 years.
Results Mother ratings of externalizing and internalizing problems from age 5 to 11–14 years were higher for
the chronic iron deficiency group compared with those with the good iron status. Iron deficiency in infancy
predicted persisting externalizing problems over this time period, especially for those with low physical activity
in infancy. Beyond adolescence, youth in the chronic iron deficiency group did not report more problems than
those in the good iron group. ConclusionsThese findings underscore the importance of considering infant
iron status along with early behavioral characteristics to better identify those children at greatest risk for
persisting long-term behavior problems.
This study examined externalizing and internalizing behavior problem trajectories as a function
MethodsA sample of 185 healthy Costa Rican
Key words behavior problems; longitudinal design; infant characteristics; iron deficiency.
Iron deficiency anemia (IDA) is the most common single
nutrient disorder, affecting 20–25% of the children world-
wide, with a higher proportion of having ID without
anemia (Stoltzfus, 2001). Recent epidemiological data
also suggest that 12% of the Hispanic children and 6%
of the white and 6% of the African–American children in
the USA have ID (Brotanek, Gosz, Weitzman, & Flores,
2007). Infants with ID or IDA generally test lower in
mental and motor development and show affective differ-
ences (i.e. wariness, hesitance, less positive affect, and
less social interaction). Most studies continue to show
behavioral and affective differences after iron therapy
(Grantham-McGregor & Ani, 2001; Lozoff et al., 2006).
Several studies have found that infants with ID, com-
pared with those with good iron status, are more likely to
show poorer mental and motor performance at preschool
or school age (see Lozoff et al., 2006; Thomas, Grant, &
Aubuchon-Endsley, 2009 for reviews). Our longitudinal
study in Costa Rica followed participants from infancy to
the transition to adulthood (19 years) and documented
adverse persisting effects of early, chronic, and severe ID
on cognitive performance years later (Lozoff, Jimenez, &
Smith, 2006). However, less emphasis has been given to
the long-term social–emotional effects of ID. This oversight
is noteworthy because executive cognitive deficits are
widely recognized as a major risk factor for adjustment
problems (Moffitt & Caspi, 2001). The few available stu-
dies showed that children with poor iron status at birth
were significantly less alert and compliant with rules at
5 years of age (Tamura et al., 2002). Children in the
Costa Rica study described above, who had chronic,
severe ID in infancy, displayed lower levels of physical
activity, positive affect, and verbalization compared with
children with good iron status in infancy during a struc-
tured task at school entry (Corapci, Radan, & Lozoff,
2006). Differences in mother-interaction quality (i.e. poor
beyond infancy were also observed (Corapci et al.,
2006). The 11- to 14-year assessment of the Costa Rica
study—the only follow-up study into adolescence—found
that children with early, chronic, and severe ID were at
increased risk for externalizing (i.e. aggression, defiance)
*Present address: Feyza Corapci, Psychology Department, Bogazici University, Istanbul, Turkey; Agustin Calatroni, Rho Inc.,
Chapel Hill, North Carolina; Elias Jimenez, Hospital CIMA-San Jose, Costa Rica.
All correspondence concerning this article should be addressed to Betsy Lozoff, MD, Center for Human Growth and
Development, Department of Pediatrics and Communicable Diseases, University of Michigan, 300N. Ingalls, Ann Arbor, MI
48109-0406, USA. E-mail: firstname.lastname@example.org
Journal of Pediatric Psychology 35(3) pp. 296–305, 2010
Advance Access publication September 7, 2009
Journal of Pediatric Psychology vol. 35 no. 3 ? The Author 2009. Published by Oxford University Press on behalf of the Society of Pediatric Psychology.
All rights reserved. For permissions, please e-mail: email@example.com
and internalizing (i.e. anxiety, depression) problems, as
rated by both teachers and parents (Lozoff, Jimenez,
Hagen, Mollen, & Wolf, 2000). In several studies, the
iron status effect remained statistically significant after
control for child and family background factors.
emotional adjustment, especially in countries where the
prevalence of ID is high, may help understand how early
ID restricts children from reaching their developmental
potential (Walker et al., 2007). The concept of ‘‘functional
isolation’’ in the nutrition field has been used to help
explain long-lasting effects of chronic ID. According
to the functional isolation hypothesis (Levitsky, 1979),
nutritional deficiencies contribute to changes in infant
affect and activity, which in turn compromise infants’
ability to seek and/or receive stimulating and responsive
interactions from their caregivers, thereby contributing to
their isolation from the environment and poorer develop-
mental outcomes over time. Lozoff and colleagues
(Lozoff et al., 1998) have offered an integrated biological
and environmental view of brain–behavior relations. They
postulated that chronic, severe ID has a direct impact
on neurobiological changes in the developing brain
(e.g. myelination, neurotransmitter systems, and neural
metabolism). Given that dopamine and serotonin metabo-
lism and functioning are involved in emotion processing,
attention, and behavioral activation/inhibition (Beard,
2003), alterations in these neurotransmitter systems are
of particular relevance to the social–emotional domain
and may explain the observed pattern of wariness/
hesitance, low activity, and low positive affect observed
among ID infants. The integrated model also includes
the contribution of limitations in caregiving (e.g. lack
of responsivity and stimulation) to maintain or even
accentuate these child characteristics, especially in a
disadvantaged environment. Such dynamic, reciprocal
influences among the neural, social, and ecological factors
combine over time to undermine child self-regulatory com-
petence, which refers to the ability to modulate affective,
pathways to psychopathology (Posner & Rothbart, 2000).
The present study included social–emotional evalua-
tions in infancy and four subsequent follow-ups (5, 11–14,
15–18, and 19 years) from an ongoing study in Costa Rica
to address two novel research questions. First, rather than
examining the effect of ID on social–emotional functioning
at a given point at time, we investigated the pattern of
change in externalizing and internalizing problems over
time between participants with chronic, severe ID and
those with good iron status in infancy. Second, we exam-
ined whether early infant characteristics (e.g. low positive
affect and activity level) might operate as vulnerability fac-
tors in the context of chronic and severe ID. Finally, the
present study also extended the current knowledge by
investigating the role of early chronic ID on behavior
problems beyond early adolescence.
Based on the functional isolation hypothesis and past
research, we hypothesized that children with chronic ID in
infancy would show persisting levels of externalizing and
internalizing problems from early childhood to adolescence
and beyond, compared with children with good iron status
in infancy. We also predicted that early chronic ID status
would interact with infant characteristics consistent with
functional isolation (i.e. higher levels of wary/distressed
emotionality, lower activity) to result in greater behavior
problems over time.
The original study was conducted between 1983 and 1985
with 191 infants from an urban community near San Jose,
the capital of Costa Rica. Enrollment entailed door-to-door
screening of 12- to 23-month-old infants in the entire
community, and refusal was 11.6%. Recruited infants
had a birth weight of ?2.5kg and were free of acute
or chronic medical problems with no evidence of growth
failure or other nutrient deficiencies. At the time of enroll-
ment, mothers averaged 26.9 years of age (SD¼6.3)
and had 9.6 years of schooling (SD¼3.3). Families were
generally lower middle to working class. Details of the
original study have been published previously (Lozoff
et al., 1987).
Only 15 participants were completely lost after the
5-year follow-up. Fifty-one per cent of the sample had
complete social/emotional assessment data in all the four
follow-up periods (5, 11–14, 15–18, and 19 years). The
rest of the participants provided data from a varying
number of time points after infancy. Overall, 74% of
the original sample participated in assessments at both
5 years and early adolescence, and 73% of the original
sample participated in assessments at both early and
mid-adolescence. Lack of participation in our study was
primarily due to difficulty in locating a family. The
15 participants who were lost after the 5-year follow-up
did not differ from the remaining participants with respect
to iron status, gender, family socioeconomic status (SES),
and behavior problem scores at the age of 5 years (all
p-values >.05), suggesting that the sample was not
biased toward those families who were less stressed or
toward those children who were more psychologically
adjusted. Parental signed informed consent for all phases
Behavior Problems Following Iron Deficiency
of the study was obtained by the project pediatrician.
Research protocols were approved by the Institutional
Review Boards of the collaborative institutions.
Of the 191 infants in the initial study, six children
were excluded due to lack of information about their
iron status after treatment. Fifty-three infants were classi-
fied as having chronic ID and 132 were classified as having
good iron status as described further. At study entry, the
chronic ID group had a greater proportion of male infants
[73% vs. 46%, Pearson w2(1, N¼142)¼8.06, p<.01]
and came from families with significantly lower SES
as measuredby the Hollingshead
Index [M¼27.2, SD¼10.8 vs. M¼31, SD¼12.6,
t(183)¼?1.9, p?.05]. Of the 151 children (mean
age¼60.2 months, range¼59–63 months) assessed at
the 5-year follow-up, 41 of these were in the chronic ID
group in infancy and 110 were in the good iron status
group. Of the 160 children re-evaluated at 11–14 years
them were in the chronic ID group. A mid-adolescence
follow-up provided data for 133 participants (mean
age¼16.5 years, range¼15.0–17.8); 34 of them belonged
to the chronic ID group. An assessment at 19 years
provided data for 119 participants (mean age¼19 years,
range¼18–20); 34 of them were in the chronic ID group
Measures and Procedure
Iron Status in Infancy
Admission to the study in infancy was based on a venous
blood sample. Iron status was determined by concen-
trations of hemoglobin (Hb), transferrin saturation,
erythrocyte protoporphyrin, and serum ferritin, and
response to 3 months of iron treatment. Iron status
ranged from iron sufficiency to ID with moderate
anemia. Iron sufficiency was defined as hemoglobin
?120g/l and no abnormal iron measures. Mild anemia
was defined as hemoglobin content 101–105g/l, and mod-
erate anemia was defined as hemoglobin content being
?100g/l. ID was defined as serum ferritin ?12?g/l, and
either erythrocyte protoporphyrin ?100?g/dl packed red
blood cells or transferrin saturation <10%. Infants with
hemoglobin <120g/l and ID were given either intramus-
cular iron or 3 months of oral iron therapy in two doses
daily. Infants with hemoglobin ?120g/l who were either
ID or iron depleted (i.e. serum ferritin <12mg/l) received
oral iron treatment; those with normal values on all iron
status measures received placebo. Hematologic response to
iron therapy in infancy was excellent (Lozoff et al., 1987).
However, some infants still had biochemical alterations
even after 3 months of iron treatment (e.g. erythrocyte
protoporphyrin >100?g/dl packed red blood cells or
transferrin saturation ?10%).
Analyses reported further compared chronic ID and
good iron status groups based on the approach used in
previous studies of this cohort, beginning with the results
of the 5-year follow-up (Lozoff, Jimenez, & Wolf, 1991).
At 5 years, children who had moderate IDA as infants
continued to test lower in mental and motor functioning,
as did children with higher hemoglobin levels who still had
some biochemical evidence of ID after iron therapy in
infancy. The same was true in early adolescence; therefore,
both groups were combined. They were considered to have
severe and chronic ID in infancy, evidenced by lower initial
hemoglobin levels and higher initial erythrocyte protopor-
phyrin values. Since anemia is a late manifestation of ID
and hemoglobin level reflects chronicity and severity once
anemia develops, infants with moderate IDA had to have
chronic, severe ID. Furthermore, enrollment occurred at
12–23 months, making it likely that ID had been present
for some time. The good iron status group consisted of
infants who were iron sufficient at study entry and those
with lower hemoglobin and any degree of ID who became
iron sufficient after iron treatment. Iron status of the
participants was excellent at the 5- and 11- to 14-year
follow-ups (Lozoff et al., 1991; Lozoff et al., 2000).
At the 19-year follow-up, ID was present in <5%, and
no one had IDA except for four females, two of whom
were pregnant (Lozoff, Jimenez et al., 2006).
Social/emotional Assessment in Infancy
Multiple methods and reporters were used to collect socio-
emotional data in infancy. First, trained examiners used
the Infant Behavior Record (IBR) of the Bayley Scales of
Infant Development (Bayley, 1969) with summary scores
based on prior research (Matheny, 1980; Wolf & Lozoff,
1985) to rate infants’ affect and behavior. Second, obser-
vers completed quantitative behavior coding of play and
developmental testing from videotape to rate the physical
activity and proximity of the infant to caregiver. All exam-
iners were blind to the infant’s iron status or treatment.
Inter-rater reliability, assessed by intraclass correlation
coefficient, averaged from 0.88 to 0.93 (Lozoff et al.,
1998). Multiple spot observations of infant behavior and
affect were conducted in the home (Rogoff, 1978). Each
approach distinguished between the chronic ID and good
iron status groups (Lozoff et al., 1998) providing support
for the measures’ validity.
Socioemotional Adjustment from Early Childhood
through Early Adulthood
The Spanish version of the Parent (CBCL), Youth (YSR)
and the Young Adult (YASR) forms of the Child Behavior
Corapci et al.
Checklist (CBC; Achenbach & Edelbrock, 1983, 1991)
were used to assess behavior problems. The CBC is a
valid and reliable measure that has been used in various
countries (Crijnen, Achenbach, & Verhulst, 1997). We
computed the externalizing and internalizing scale scores
based on Achenbach’s empirically derived two-factor,
second-order structure, which cluster similarly across
cultures (De Groot, Koot, & Verhulst, 1994, 1996).
Maternal ratings on the aggressive and delinquent behavior
subscales were used for externalizing scale scores at 5- and
11- to 14-year follow-ups. Maternal ratings on the social
withdrawal, somatic complaints, and anxiety/depression
subscales were used for internalizing scale scores. As
children get older, their parents’ evaluations may not
reflect the full range of their children’s behavior problems
(Leve, Kim, & Pears, 2005). Therefore, at the 15- to
18-year and 19-year follow-ups, youth ratings were
obtained. Maternal ratings of the CBCL were not available
at these later follow-ups. Cronbach-a for internalizing and
externalizing scales ranged from .86 to .91 in our sample.
All analyses were conducted using t-scores relative to age-
normed scores based on the U.S. standardization samples
at each time point.
Family SES in infancy was assessed using the Hollingshead
Four Factor Index (Hollingshead, 1975). Pubertal status
was assessed with the standard Tanner staging system
(Tanner, 1962) during a pediatric examination at the
11- to 14-year follow-up. Based on our sample distribu-
tion, three levels of Tanner pubertal stages were used in
data analysis, i.e. early (Stage 1), middle (Stages 2 and 3),
and late (Stages 4 and 5). As part of the 19-year follow-up,
participants filled out the Child Health and Illness Profile-
Adolescent Edition (CHIP-AE; Starfield et al., 1995) to
report on their school and work performance.
Infancy Data Reduction
A total of 25 variables pertaining to infant’s affect and
activity, proximity to mother, and engagement with objects
came from examiner ratings on the IBR and from direct
observational coding during play and development testing
in the clinic and from home observations. An exploratory
factor analysis was conducted with principle component
analysis as the extraction method to reduce data and
empirically isolate major infancy variables. Orthogonal
(varimax) and oblique (promax) rotations were conducted.
Both rotations yielded comparable results (r’s between the
corresponding factors produced by varimax and promax
rotations were >.95), and the promax rotation indicated
that the correlation between the factors was quite negligi-
ble. Therefore, we chose to report the findings with the
orthogonal rotation to have uncorrelated and more
meaningful factors. The first two factors in the rotated
matrix pertained to infant affect and activity level and
were retained because they were the most theoretically
meaningful and interpretable with regard to the functional
isolation hypothesis. These factors accounted for 25% of
the variance. An affect and inactivity factor score were
computed using the regression method. A factor loading
of .40 or higher was used. High scores on the affect factor
characterized infants who were more likely to be wary,
upset, tense, unhappy, and easily fatigued during mental
and motor testing and play session. Higher scores on the
inactivity factor characterized infants who were less likely
to be walking or playing with objects interactively and
more likely to be asleep or doing nothing during home
visits. These infants were also less likely to explore new
areas in play, change their proximal relations to their
mothers as well as vocalize in play, and more likely to
be in a playpen.
Longitudinal Data Analyses
(Proc Mixed in SAS) to examine behavior problem trajec-
tories depending on iron status in infancy. This approach
accounts for the within subject correlation due to having
repeated measures (RMs) on the same subjects over time
for externalizing and internalizing problems. It also uses
all cases, whether or not they have an equal number of
available data across all follow-up periods, and accommo-
dates data that are missing at random using a restricted
maximum likelihood estimation method (Fitzmaurice,
Laird, & Ware, 2004). A single growth curve analysis
from age 5 to 19 years was not possible given the simul-
taneous change both in the period of assessment (early to
mid-adolescence) and informant (mother to youth).
Therefore, two sets of analyses were conducted to model
the difference between two time points in behavior pro-
blems as a function of ID status. The first set included
CBCL ratings at age 5 and 11–14 years per mother
report. The second set of analyses covered the period
from age 15–18 to 19 years per youth self-report.
To account for the within subject correlation of the
CBCL ratings between two observation points, we consid-
ered both a random coefficient (intercept) model and RMs
approach using an autoregressive covariance structure.
The RM approach had a better fit than the model with
random intercept based on the Bayesian Information
Criteria (lower BIC value) and was used in the analysis.
All analyses included iron status in infancy, time and their
Behavior Problems Following Iron Deficiency
interaction as independent variables, and controlling for
gender and SES in infancy. Pubertal status was also
included as a covariate when analyzing change in behavior
problems from 5 to 11- to 14 years. The model examining
change in behavior problems from age 15–18 to 19 years
included mother ratings of CBCL from age 11–14 years as
a covariate. Entering this variable allowed the examination
of the stability of behavior problems over the adolescence
years (Leve et al., 2005; Sorlie, Hagen, & Ogden, 2008)
and also created a conservative test of the unique predic-
tive role of ID in infancy, over and above the contribution
of behavior problems at early adolescence.
Group Differences and Change in Mother Ratings
of Behavior Problems from 5 to 11–14 Years
The two-way interaction term between the iron status
group and time was not significant in the LMM analyses
of externalizing and internalizing scores. However, there
were significant main effects of iron group and time. As
predicted, mother ratings of externalizing and internalizing
problems were higher for children in the chronic ID group
compared with the good iron group, F(1, 174)¼4.9,
p<.05, and F(1, 174)¼9, p<.01, respectively (Table I).
The direction of change (averaged across iron group) was
toward a decrease in externalizing problems and an increase
in internalizing problems over time, F(1, 174)¼15.6,
p<.001, and F(1, 174)¼4.2, p<.05, respectively. The
same pattern of significant findings remained after con-
trolling for gender, SES, and pubertal status. As shown in
Table I, Cohen’s effect sizes for iron status ranged from .35
SD to .54 SD, indicating moderate-sized effects.
Change in Behavior Problems from 5 to 11–14
Years as a Function of Infant Characteristics
To examine the interactive effect of infant activity level and
iron status group on externalizing problems over time, the
LMM analysis included a three-way interaction term
between iron group, time, and infant activity level, all
lower order interaction terms, and significant covariates.
After covariate control, the three-way interaction term
was significant, F(1, 171)¼6.05, p<.05. To understand
the three-way interaction, the effects of ‘‘high’’ and ‘‘low’’
infant inactivity (?1 SD from the mean) on externalizing
problem scores were computed separately for the two-way
interaction between iron group and time (Aiken & West,
1991). As seen in Fig. 1, when infant activity level was
high, children in the chronic ID group improved from
the subclinical to the normal range (6-point decrease;
from a t-score of 64.5 to 58.5), compared with those in
the good iron status, who showed little change (1-point
decrease; from 58.3 to 57.2). On the other hand, as seen in
Fig. 2, when infant activity was low, children in the chronic
ID group remained in the subclinical range of externalizing
problems over time as expected (2-point decrease;
Table I. Mother CBCL Ratings of Externalizing and Internalizing Problems at Age 5 and 11–14 years by Iron Status Group in Infancy
Chronic ID (mean?SD)
Good iron (mean?SD)
Effect size (95% CI)Iron groupTime
11–14 11–145 11–14FF
*p<.05; **p<.01; ***p<.001.
aAdjusted values control for gender, SES in infancy, and pubertal status at early adolescence. Adjusted SD was calculated based on SEMˇN.
Figure 1. Interaction between iron status and time predicting mother
ratings of externalizing problems for infants with high activity.
Corapci et al.
from 63 to 61), whereas those in the good iron group
improved from the subclinical to the normal range
(7-point decrease; from 63.3 to 56.5).
In a separate LMM, we also examined the interactive
effect of infant affect with iron status over time. No such
three-way interaction was observed for infant affect.
Contrary to our expectations, LMM analyses of internaliz-
ing scores did not reveal statistically significant two- or
three-way interactions. Main effects for infant activity or
infant affect were also not significant.
Group Differences and Change in Behavior
Problems from 15–18 to 19 Years (Self-report)
The two-way interaction term between the iron group
status and time as well as the main effect of iron group
were not statistically significant (Table II). These findings
did not support the expected iron group differences in
behavior problems beyond adolescence. Only the time
effect was significant; a decrease in externalizing problems
from 15–18 to 19 years was detected, before and after
F(1, 143)¼6.8, p?.01, respectively. After covariate con-
trol, a significant decrease in internalizing problems was
also detected, F(1, 137)¼3.9, p<.05. There were no
gender differences in relation to change in either type of
behavior problems over time. Neither infant activity nor
affect had a significant interaction with iron status on
the trajectory of externalizing and internalizing problems
from 15–18 to 19 years. There were also no significant
main effects on infant activity or affect.
We examined the link between iron status and school and
work performance from the CHIP-EA. We found that
a greater proportion of the individuals in the chronic ID
group dropped out of school in the past 2 years compared
withthose in the good
w2(1, N¼114)¼4.96, p<.05, Crame ´r’s V¼.21. The
relation between iron status and work productivity was
statistically marginal, Pearson w2
p¼.10, Crame ´r’s V¼.26. Somewhat a larger proportion
of young adults in the chronic ID group described
themselves as inefficient at work compared with their
The goal of the present study was to examine the change in
externalizing and internalizing problems as a function of
both iron status in infancy and infant characteristics. Our
interpretations of both our significant and non-significant
findings are discussed in the following sections.
Table II. Youth Self-ratings of Externalizing and Internalizing Problems at Age 15–18 and 19 years by Iron Status Group in Infancy
Chronic ID (mean?SD)
Good iron (mean?SD)
Effect size (95% CI)Iron groupTime
19 19 15–1819FF
aAdjusted values control for gender, SES in infancy and behavior problems at early adolescence. Adjusted SD was calculated based on SEMˇN.
Figure 2. Interaction between iron status and time predicting mother
ratings of externalizing problems for infants with low activity.
Behavior Problems Following Iron Deficiency
Early, Chronic ID and Externalizing and
Internalizing Problems from 5 to 11–14 Years
Our first set of analyses covered the period from preschool
to early adolescence and included mother ratings of
externalizing and internalizing problems. Contrary to
expectations, the pattern of change in behavior problems
between the chronic ID and good iron groups did not differ
significantly. However, as expected, we found that children
in the chronic ID group, on average, had higher levels
of externalizing and internalizing problems from age 5 to
11–14 years compared with those with good iron status.
While we cannot eliminate all possible factor(s) closely
associated with ID, our results remained significant after
controlling for the effects of gender, SES, and pubertal
We also described differences in behavior problem
trajectories as a function of both iron status and infant
characteristics. Consistent with our predictions, we
found that chronic ID status, when accompanied by low
physical activity in infancy, predicted persisting externaliz-
ing problems from age 5 to 11–14 years, whereas less
active children in the good iron group showed decreases
in externalizing problems. Conversely, children in the
chronic ID group with high activity showed improvements
in externalizing problems. Finally, children in the good
iron group who were active infants remained in the
normal ranges at both ages.
The current findings make sense in the framework of
the functional isolation model (Levitsky, 1979) and the
integrated biological/environmental model (Lozoff et al.,
1998). In line with the predictions of these models, we
found that chronic ID in infancy contributed to long-last-
ing behavior problems, possibly through the interplay of
intermediary neural and environmental mechanisms.
The investigation of such mediating mechanisms was
beyond the scope of the present study, but we can specu-
late that the brain effects of ID have contributed to
dopamine-related behaviors such as infant inactivity
However, a child’s activity level partly depends
on parental behavior and thus may not be the result
of ID per se. In fact, previously published reports of
the infancy cohort and the 5-year follow-up sample of
the Costa Rica study provide evidence for the lack
of warmth and responsivity in the mother–child interac-
tions of children with ID in infancy (Corapci et al.,
2006; Lozoff et al., 1998). These observations point
to the highly interactive and transactional influences of
childand caregiving characteristics(Sameroff&
MacKenzie, 2003; Wright, 2008). Based on the con-
temporary theories of child development and empirical
evidence, there is an increasing recognition that nutri-
tional deficiencies are translated into child behavior
problems through such developmentally inhibiting trans-
actional processes, which in turn appear to undermine
the development of children’s self-regulatory skills such
as the ability to modulate affect, attention, and behavior
(Wachs, 2009; Wachs, Pollitt, Cuerto, Jacoby, & Creed-
The importance of ‘‘child effects’’ as a starting point in
the transactional processes in the case of ID in infancy has
been documented in preventive trials, animal studies and
developmental testing observations. Large randomized,
controlled preventive trials (Black et al., 2004; Lozoff,
De Andraca, Castillo, Smith, Walter, & Pino, 2003)
provided evidence that a greater percentage of the infants
who did not receive iron supplementation showed less
positive affect and poorer social interaction compared
with iron-supplemented infants. These results suggest
that the lack of iron in infancy causes alterations in
infant affect and interaction, independent of child or
family differences. Animal research has shown that
monkey infants on prenatally iron-deprived diets showed
reduced activity (Golub, Hogrefe, Germann, Capitano, &
Lozoff, 2006). Furthermore, in the infancy phase of the
Costa Rica study, the tester, who was uninformed about
the infants’ hematologic status, offered fewer demonstra-
tions and encouragements to IDA and comparison group
infants during developmental testing, indicating that
adults other than the mother behaved differently with
Our findings also suggest that in the context of lower
environmental risk for the good iron group, caregivers
might have been more likely to perceive their low-activity
children as ‘‘easy’’ babies and promote the development of
self-regulation in their children with adequate monitoring,
effective disciplining, and modeling of emotion and
behavior regulation, thereby diminishing the likelihood of
externalizing problems. In contrast, being an active infant
emerged as a buffering factor for those infants in the
chronic ID group and led to decreases in externalizing
problems from age 5 to 11–14 years. It is possible that
these active children were more likely to elicit parental
attention and involvement. It is also possible that parents,
who were informed about their children’s ID status at
the end of the infancy study, might have perceived their
highly active and chronic ID children at risk for later
problems and thus might have been more sensitive to
their children’s cues.
Corapci et al.
Early, Chronic, Severe ID, and Externalizing and
Internalizing Problems from 15–18 to 19 Years
Our second set of analyses covered the period from
mid-adolescence to early adulthood and included youth
self-ratings of externalizing and internalizing problems.
The pattern of change in behavior problems from age
15–18 to 19 years was the same for each iron group
status. Contrary to our expectations, youth in the chronic
ID group reported problems in the normal range, as did
those in the good iron group. Since behavioral ratings were
not obtained from other informants after early adolescence,
we do not know whether the association between early,
chronic ID and behavioral adjustment detected early
on dissipated by late adolescence or whether the lack of
association was due to change in informant.
In contrast to the lack of differences in behavior
problems in late adolescence, recent findings from this
sample revealed differences in cognitive test scores from
infancy to 19 years. The gap between the good iron status
and ID group widened, especially for those children
coming from low-SES families and reached a magnitude
of 25–28 points of difference (Lozoff, Jimenez et al.,
2006). It seems likely that a cumulative cognitive deficit
of the observed magnitude would impair self-regulation,
verbal reasoning and executive functioning, and thus
would contribute to impulsive responding, inattention as
well as poor planning, thus increasing the risk for psycho-
pathology (Moffitt & Caspi, 2001; Pennington & Ozonoff,
1996). It is possible that cognitive limitation in the chronic
ID group compromised the reliability and quality of their
self-report. In our exploration of the real-life impacts
in early adulthood of ID in infancy, we found a higher
prevalence of school dropout and work unproductivity at
19 years in the chronic ID group. This finding makes
a fading effect of ID less likely and again suggests
that informant bias may have contributed to the lack of
difference in behavior problems by self-report.
A discrepancy between parent and youth reported
externalizing and internalizing problems has been reported
for children with low birth weight (Saigall, Pinelli,
Hoult, Kim, & Boyle, 2003). Significant differences were
detected between the groups based on parent report, but
not based on youth self-report. Differences in parent and
child reports of behavior problems have also been docu-
mented in previous clinical psychology research. Youth,
compared with their parents, under-report problems,
perhaps due to failure to acknowledge their problems
(Achenbach, Krukowski, Dumenci, & Ivanova, 2005;
Jensen et al., 1999; Martin, Ford, Dyer-Friedman, Tang,
& Huffman, 2004).
Limitations, Conclusions, and Future Research
The lack of behavioral ratings from other informants after
early adolescence is a limitation of our study. Thus, a single
growth curve analysis was not possible to use to examine
the developmental trajectory of behavior problems because
we could not tell whether differences from early to mid-
adolescence related to actual changes in behavior problems
or to their measurement (i.e. mother- vs. self-report).
Although long-term effects of early, severe and chronic
ID on behavior problems beyond early adolescence were
not detected per youth self-report, this finding should be
considered tentative in light of cognitive impairments and
functional outcomes and be replicated in future research.
Future studies with multiple reporters and multiple
measures of impairment (both objective and subjective)
are warranted to differentiate a potential waning influence
of early iron status from self-report biases.
A causal relationship between chronic, severe ID in
infancy and behavioral problems cannot be proved
although iron group differences from age 5 to 11–14
years remained significant after control for child and
family background variables. Some uncontrolled or unmea-
sured factor(s) closely associated with ID might account
for the group differences. Another limitation is that there is
no study comparable with our multi-method approach for
social–emotional data in infancy, and hence there is no
possibility of direct comparisons regarding our data reduc-
tion. It is thus unclear whether our findings are general-
izable to other samples, including children who are not
healthy or full term. Furthermore, we know relatively
little about how early ID impacts the immediate social
context. For example, mother–child interactions are
central to the transactional framework of the functional
isolation hypothesis and suggest not only an important
mediating mechanism but also avenues for intervention.
Future research should test models that include parent–
child interaction measures as time-varying covariates in
order to better understand how the level of behavior
problems following early, chronic ID changes in relation
to changes in caregiving quality.
In conclusion, our findings suggest that chronic ID in
infancy represents a risk factor for later behavior problems
and chronic ID along with infant inactivity identifies chil-
dren at greatest risk for long-lasting externalizing problems
from early childhood to early adolescence. These results
raise the question of whether the pediatric evaluation of
children with ID should include an assessment of child’s
behavioral and affective characteristics within the social
Behavior Problems Following Iron Deficiency
context in addition to hematologic measures. The iron
status effect sizes we observed may be meaningful for
policy and clinical implications, given the large number
of children in the world who suffer from chronic and
severe ID. Increasing caregivers’ sensitivity to their ID chil-
dren’s early activity level might be a useful intervention
target besides iron therapy to prevent behavior problems.
This research was supported by grants from the National
Institutes of Health (HD14122 and a MERIT Award to
Betsy Lozoff, R37 HD31606). We are grateful to all families
who participated in the study.
Received January 20, 2009; revisions received July 6, 2009;
accepted July 7, 2009
Achenbach, T. M., & Edelbrock, C. (1983). Manual for
the child behavior checklist and revised child behavior
profile. Burlington, VT: University of Vermont.
Achenbach, T. M., & Edelbrock, C. S. (1991). Manual
for the child behavior checklist/4-18 and 1991 profile.
Burlington, VT: University of Vermont, Department
Achenbach, T. M., Krukowski, R. A., Dumenci, L.,
& Ivanova, M. Y. (2005). Assessment of adult
psychopathology: meta-analysis and implications of
cross-informant correlations. Psychological Bulletin,
Aiken, L. S., & West, S. G. (1991). Multiple regression:
testing and interpreting interactions. Newbury Park:
Bayley, N. (1969). Bayley scales of infant development.
New York: Psychological Corporation.
Beard, J. (2003). Iron deficiency alters brain development
and functioning. Journal of Nutrition, 133,
Black, M., Baqui, A. H., Zaman, K., Ake Persson, L., El
Arifeen, S., Le, K., et al. (2004). Iron and zinc
supplementation promote motor development and
exploratory behavior among Bangladeshi infants.
American Journal of Clinical Nutrition, 80, 903–910.
Brotanek, J. M., Gosz, J., Weitzman, M., & Flores, G.
(2007). Iron deficiency in early childhood in the
United States: risk factors and racial/ethnic disparities.
Pediatrics, 120, 568–575.
Corapci, F., Radan, A. E., & Lozoff, B. (2006). Iron
deficiency in infancy and mother-child interaction at
5 years. Journal of Development and Behavioral
Pediatrics, 27, 371–378.
Crijnen, A. A., Achenbach, T. M., & Verhulst, F. C. (1997).
Comparisons of problems reported by parents of
children in 12 cultures: total problems, externalizing,
and internalizing. Journal of the American Academy of
Child and Adolescent Psychiatry, 36, 1269–1277.
De Groot, A., Koot, H. M., & Verhulst, F. C. (1994).
Cross-cultural generalizability of the child behavior
checklist form cross-informant syndromes.
Psychological Assessment, 6, 225–230.
De Groot, A., Koot, H. M., & Verhulst, F. C. (1996).
Cross-cultural generalizability of the youth self-
report and teacher’s report form cross-informant
syndromes. Journal of Abnormal Child Psychology, 24,
Fitzmaurice, G., Laird, N., & Ware, J. (2004). Applied
longitudinal analysis. Hoboken, NJ: Wiley-Interscience.
Golub, M. S., Hogrefe, C. E., Germann, S. L., Capitano, J.
P., & Lozoff, B. (2006). Behavioral consequences of
developmental iron deficiency in infant rhesus
monkeys. Neurotoxicology and Teratology, 28, 3–17.
Grantham-McGregor, S., & Ani, C. (2001). A review of
studies on the effect of iron deficiency on cognitive
development in children. Journal of Nutrition, 131,
Hollingshead, A. A. (1975). Unpublished manuscript.
Four-factor index of social status. New Haven, CT:
Yale University, Unpublished Manuscript.
Jensen, P., Rubio-Stipec, M., Canino, G., Bird, H.,
Dulcan, M., Schwab-Stone, M. E., et al. (1999).
Parent and child contributions to diagnosis of mental
disorder: are both informants always necessary?
Journal of the American Academy of Child and Adolescent
Psychiatry, 38, 1569–1579.
Leve, L. D., Kim, H. K., & Pears, K. C. (2005). Childhood
temperament and family environment as predictors
of internalizing and externalizing trajectories from ages
5 to 17. Journal of Abnormal Child Psychology, 33,
Levitsky, D. A. (1979). Malnutrition and the hunger
to learn. In D. A. Levitsky (Ed.), Malnutrition,
environment, and behaviour (pp. 161–179). Ithaca, NY:
Cornell University Press.
Lozoff, B., Beard, J., Connor, J., Felt, B., Georgieff, M. K.,
& Schallert, T. (2006). Long-lasting neural and
behavioral effects of iron deficiency in infancy.
Nutrition Reviews, 64, S34–S43.
Lozoff, B., Brittenham, G. M., Wolf, A. W., McClish, D. K.,
Kuhnert, P. M., Jimenez, E., et al. (1987). Iron
deficiency anemia and iron therapy: effects on infant
Corapci et al.
developmental test performance. Pediatrics, 79,
Lozoff, B., De Andraca, I., Castillo, M., Smith, J.,
Walter, T., & Pino, P. (2003). Behavioral and
developmental effects of preventing iron-deficiency
anemia in healthy full-term infants. Pediatrics, 112,
Lozoff, B., Jimenez, E., Hagen, J., Mollen, E., & Wolf, A.
W. (2000). Poorer behavioral and developmental
outcome more than 10 years after treatment for
iron deficiency in infancy. Pediatrics, 105, E51.
Lozoff, B., Jimenez, E., & Smith, J. (2006). Widening
gap in cognitive test scores following iron deficiency
in infancy: a longitudinal analysis to 19 years.
Archives in Pediatrics Adolescent Medicine, 160,
Lozoff, B., Jimenez, E., & Wolf, A. W (1991). Long-
term developmental outcome of infants with iron
deficiency. New England Journal of Medicine, 325,
Lozoff, B., Klein, N. K., Nelson, E. C., McClish, D. K.,
Manuel, M., & Chacon, M. E. (1998). Behavior of
infants with iron deficiency anemia. Child
Development, 69, 24–36.
Martin, J. L., Ford, C. B., Dyer-Friedman, J., Tang, J.,
& Huffman, L. C. (2004). Patterns of agreement
between parent and child ratings of emotional and
behavioral problems in an outpatient clinical
setting: when children endorse more problems.
Journal of Developmental and Behavioral Pediatrics, 25,
Matheny, A. I. Jr. (1980). Bayley’s Infant Behavior Record:
behavioral components and twin analyses. Child
Development, 51, 1157–1167.
Moffitt, T., & Caspi, A. (2001). Childhood predictors
differentiate life-course persistent and adolescence-
limited antisocial pathways among males and females.
Development and Psychopathology, 13, 355–375.
Pennington, B. F., & Ozonoff, S. (1996). Executive
functions and developmental psychopathology.
Journal of Child Psychology and Psychiatry, 37, 51–87.
Posner, M. I., & Rothbart, M. K. (2000). Developing
mechanisms of self-regulation. Development and
Psychopathology, 12, 427–441.
Rogoff, B. (1978). A spot observation: an introduction and
examination. Quarterly Newsletter of the Institute for
Comparative Human Development, 2, 21–26.
Saigall, S., Pinelli, J., Hoult, L., Kim, M. M., & Boyle, M.
(2003). Psychopathology and social competencies of
adolescents who were extremely low birth weight.
Pediatrics, 111, 969–975.
Sameroff, A., & MacKenzie, M. J. (2003). Research
strategies for capturing transactional models of
development: the limits of the possible. Development
and Psychopathology, 15, 613–640.
Sorlie, M., Hagen, K. A., & Ogden, T. (2008). Social
competence and antisocial behavior: continuity and
distinctiveness across early adolescence. Journal of
Research on Adolescence, 18, 121–144.
Starfield, B., Riley, A. W., Green, B., Ensminger, M.,
Ryan, S. A., Kelleher, K., et al. (1995). The adolescent
child health and illness profile: a population-based
measure of health. Medical Care, 33, 553–566.
Stoltzfus, R. J. (2001). Defining iron-deficiency anemia in
public heath terms: a time for reflection. Journal of
Nutrition, 131, 565S–567S.
Tamura, T., Goldenberg, R. L., Hou, J., Johnston, K. E.,
Cliver, S. P., Ramey, S. L., et al. (2002). Cord serum
ferritin concentrations and mental and psychomotor
development of children at five years of age. Journal of
Pediatrics, 140, 165–170.
Tanner, J. M. (1962). Growth at adolescence (2nd ed.).
Oxford: Blackwell Scientific Publication.
Thomas, D. G., Grant, S. L., & Aubuchon-Endsley, N. L.
(2009). The role of iron in neurocognitive
development. Developmental Neuropsychology, 34,
Wachs, T. D. (2009). Models linking nutritional
deficiencies to maternal and child mental health.
American Journal of Clinical Nutrition, 89, 935S–939S.
Wachs, T. D., Pollitt, E., Cuerto, S., Jacoby, H., & Creed-
Kanashiro, H. (2005). Relation of neonatal iron status
to individual variability in neonatal temperament.
Developmental Psychobiology, 46, 141–153.
Walker, S. P., Wachs, T. D., Gardner, J. M., Lozoff, B.,
Wasserman, G. A., Pollitt, E., et al. (2007). Child
development: risk factors for adverse outcomes in
developing countries. Lancet, 369, 145–157.
Wolf, A. W., & Lozoff, B. (1985). A clinically interpretable
method for analyzing the Bayley Infant Behavior
Record. Journal of Pediatric Psychology, 10, 199–214.
Wright, R. O. (2008). Neurotoxicology: what can context
teach us? Journal of Pediatrics, 152, 155–157.
Behavior Problems Following Iron Deficiency