Relation of neonatal iron status to individual variability in neonatal temperament

ABSTRACT: The relation between indices of neonatal iron status and individual differences in neonatal temperament were investigated in a sample of 148 low-income Peruvian women and their newborn infants. Using cord blood, at birth we obtained measures of neonatal ferritin, serum iron, and hemoglobin. While neonates were still in the hospital, their behavior during a structured anthropometry exami-nation was videotaped and subsequently coded on four temperament dimensions: activity level, negative emotionality, alertness, and soothability. The same dimen- sions were coded using a videotape obtained during a subsequent visit to the neonates' homes. Results indicated that lower levels of neonatal hemoglobin andserum iron were related to higher levels of negative emotionality and to lower levels of alertness and soothability. A similar pattern was found for ferritin, but only forfemales. For the most part, relations between neonatal iron measures and neonatal temperament were linear, operating across the full range of iron values. Our pattern of iron-temperament results could not be attributed to variation in family demo-graphics, low birth weight, gestational age, maternal dietary intake, or markers ofneonatal illness and maternal diabetes. Our findings are consistent with priorresearch with older infants relating iron deficiency to temperament. These resultssupport the importance of increased research on the early functional-behavioralconsequences of individual differences in iron status as well as on the mechanismsthat underlie such consequences. ß 2005 Wiley Periodicals, Inc. Dev Psychobiol46: 141-153, 2005.
Keywords: temperament; serum iron; neonatal ferritin; hemoglobin; emotionality Temperament is defined as: ''Biologically rooted indivi- gical problems. For example, the expression of neonatal dual differences in behavior tendencies that are present temperament may vary depending on the environmental early in life and are relatively stable across various kinds context within which the neonate is assessed (Ricciuti & of situations and over the course of time'' (Bates, 1989, Breitmeyer, 1988; Wachs, Pollitt, Cueto, & Jacoby, p. 4). Implicit in this definition is the hypothesis that we 2004); however, researchers have been able to identify should be able to identify individual differences in at least neonatal behavioral patterns fitting temperament dimen- some domains of temperament even in the neonatal period sions such as irritability, activity level, and alertness (Rothbart, Derryberry, & Posner, 1994). Obviously, iden- (Rothbart, Derryberry, & Hershey, 2000; Wachs et al., tifying stable individual behavioral characteristics in the 2004). These neonatal temperament dimensions can be neonatal period is an enterprise fraught with methodolo- reliably scored (Reise, 1983; Ricciuti & Breitmeyer,1988) and show short-term stability across the first fewmonths of life (Crockenberg & Smith, 1982; Korner,Hutchinson, Koperski, Kraemer, & Schneider, 1981; Received 24 December 2003; Accepted 5 October 2004 Worobey, 1986; Worobey & Lewis, 1989). Further, some Correspondence to: T. D. WachsContract grant sponsor: NSF studies also have demonstrated modest levels of pre- diction between indices of neonatal temperament and measures of temperament assessed at or after the first year (www.interscience.wiley.com). DOI 10.1002/dev.20049 of life (Korner et al., 1985; Newnham et al., 1997; Reise, Given that there is a biological basis for individual neonatal temperament. Thus, the focus in the present differences in temperament, and that individual differ- article is on relations between neonatal temperament and ences in neonatal temperament can be reliably measured and have predictive value, an important question is what Using a relatively large sample of pregnant Peruvian biological factors are associated with variability in early women and their newborn infants, we assessed maternal appearing individual differences in temperament? There anthropometry, diet, and iron status during the second and has been a substantial body of research showing genetic third trimesters of pregnancy as well as measures of fetal contributions to temperament (e.g., Braungert, Fulker, & growth, neonatal anthropometry, neonatal iron status, and Plomin, 1992; Plomin et al., 1993; Robinson, Kagan, neonatal temperament. Because our analyses indicated Reznick, & Corley, 1992; Saudino, Plomin, & DeFries, that variability in neonatal temperament was unrelated to 1996); however, what little evidence that is available also our measures of maternal and neonatal anthropometry, indicates that genetic influences upon temperament maternal diet, maternal iron status, or fetal growth, we appear to be far less during the neonatal period (Reise, report only our findings for neonatal iron. The lack of 1990). While other studies have related biomedical risk significant prediction for our measures of maternal an- factors to indices of neonatal temperament, reviews thropometry, diet, and iron status and fetal and neonatal indicate that evidence from these studies does not yield anthropometry was not unexpected given the nature of our a consistent pattern of results (Wachs & Bates, 2001).
sample. While the overwhelming majority of mothers Conceptually, intrauterine nutrition as well as the in our sample had low intakes of iron, calcium, folate, levels of specific nutrients assessed at birth (e.g., iron) also and zinc, this was not a sample characterized by severe could contribute to individual differences in temperament.
malnutrition, severe anemia, or fetal growth retardation Both infrahuman and human data have documented the (see descriptive data). In addition, there is evidence in- sensitivity of the fetal central nervous system (CNS) dicating that fetuses are at least partially buffered against to both general malnutrition and to specific nutritional maternal malnutrition, so that unless there is severe deficiencies (Morgane et al., 1993; Rao & Georgieff, maternal malnutrition or severe maternal micronutrient 2000). In addition, those aspects of CNS structure and deficiencies, there are not likely to be functional con- neurotransmitter metabolism that have been shown to be sequences to the fetus (Dobbing, 1990; Mahomed, 2003).
influenced by variability in nutritional intake are, in many For example, while the iron supply of the fetus is basically cases, the same CNS areas and metabolic processes that derived from maternal iron stores during pregnancy have been implicated in individual variability in tempera- (Allen, 1997; Michaelsen, Milman, & Samuelson, 1995; ment (Wachs, 2000). One specific nutrient that may be of O'Brien, Zavaleta, Abrams, & Caulfield, 2003), measures particular relevance for individual differences in tempera- of neonatal hemoglobin have generally not been found to ment is iron. Infrahuman research has shown that the CNS be related to maternal iron status during pregnancy, even areas affected by pre- and perinatal iron deficiency include when mothers are anemic (Allen, 2000; Halvorsen, 2000).
those involved in emotional processing (de Ungria et al., Given evidence for fetal buffering plus the fact that ours 2000). At the human level, Vaughn, Brown and Carter was not a population of severely malnourished mothers or (1986) reported higher levels of irritability in newborns a population characterized by severe intrauterine growth whose mothers were iron deficient. In earlier studies with retardation, it is not surprising that our maternal or fetal older infants, Lozoff and colleagues (Lozoff et al., 1998; Lozoff, Wolf, & Jimenez, 1996) reported higher levels However, there is evidence suggesting that early iron of negative affect and lower levels of attention to people status may be predictive, even in a population that is not and objects in 12- to 24-month-old infants with iron- severely iron deficient. Individual differences in maternal deficiency anemia, as compared to nonanemic infants. In iron status during pregnancy are related to indices of less a more recent article, Lozoff et al. (2003) reported that severe neonatal iron deficiency such as neonatal serum 6-month-old infants who received iron supplementa- iron (Agrawal, Tripathi, & Agarwal, 1983), neonatal tion were higher on positive affect, were more oriented ferritin (Hokama et al., 1996; Milman, Agger, & Nielsen, towards people in their environment, and were more 1994), and neonatal serum transferrin receptors (Choi, soothable at 12 months of age than were infants who did Kim, & Par, 2000). Further, even in the absence of not receive iron supplementation. While the previous neonatal iron-deficiency anemia, there is a greater risk evidence is supportive of a potential link between early of later iron deficiency or iron-deficiency anemia for iron status and early temperament, little is known about neonates with reduced levels of iron (Georgieff, Wewerka, the behavioral consequences of differences in neonatal Nelson, & Deregnier, 2002) whose mothers were iron de- iron status (Beard & Connor, 2003); in our review of the ficient during pregnancy (Colmer et al., 1990; Preziosi literature, we were unable to find any studies directly et al., 1997). In addition, some evidence suggests that neo- relating measures of neonatal iron status to measures of natal iron status may be predictive of later temperament, Neonatal Iron Status and Neonatal Temperament even in the absence of severe iron deficiency. Specifically, sample size of 148, power analysis indicated that we should be Tamura et al. (2002) reported that neonates whose ferritin able to detect a medium to small effect size for iron with power values fell in the lowest quartile of their sample showed greater than.80 and a medium to small covariate effect size with significantly lower alertness and tractability (an index of self-regulation) at 5 years of age, as rated by their mothers.
Besides these data, there is an almost complete absence of studies investigating relations between neonatal iron Each trimester, trained dieticians assessed the energy and status and neonatal behavioral patterns. Given the almost nutrient intakes of the mothers using standardized 24-hr total absence of information, we believe it is important to dietary-recall assessments taken on 2 nonconsecutive days.
present our findings on neonatal iron and temperament, if Intakes were compared with current dietary recommended only to suggest the potential importance of additional intakes (e.g., Institute of Medicine, 2000). In terms of specific dietary deficiencies, those nutrients with the greatest probabilityof inadequacy for women at 10 to 24 weeks and 28 to 30 weeks ofpregnancy, respectively, were folate, calcium, iron, and zinc, with most of the population not covering 25% of the recom-mended intake for these nutrients across each of the trimesters.
The levels of maternal dietary intake found in our sample aresimilar to those reported by Sacco, Caulfield, Zavaleta, and The Canto Grande Maternal Child Health Center (MCHC) was Retamozo (2003) for pregnant women living in another peri- the site of the study. The MCHC is a state health facility that urban area of Lima. Alcohol use during pregnancy was relatively serves low-income families within Canto Grande. Canto Grande rare, with over 90% of women in our sample reporting no intake is a low-income, semi-urbanized district located northeast of of alcohol during pregnancy and less then 1% reporting an intake Lima, Peru, at an altitude of 200 m above sea level. The MCHC is of more then one drink per week. Over 75% of the women in our the largest public health facility in the area where mothers sample reported no caffeine intake over the course of pregnancy.
living in this district could deliver their babies. The rates of At the end of the first trimester, 8 ml of fasting venous blood unemployment and underemployment for the local population were drawn from the mothers during their regularly scheduled were considered high according to national census criteria. The prenatal visit to the MCHC; 6 ml of fasting venous blood were mean number of people per household was five. While 97% of drawn at the end of the second trimester of pregnancy. Measures the houses in our sample had electricity and 83% had a public of maternal hemoglobin, serum iron, total iron binding capacity, water supply, only 65% of houses had access to sewage services.
blood glucose level, and C-reactive protein were obtained from Pregnant women who lived in the MCHC-Canto Grande maternal blood samples. Measures of maternal iron status district were identified in the first trimester of pregnancy. Criteria declined significantly across trimesters. While the average level for inclusion of pregnant women were: less than 14 weeks/ of maternal Hb was greater than 11.0 g/dl (SD ¼ 1.0) in the first gestation as determined from last menstrual period and con- trimester, by the end of the second trimester the average Hb level firmed later via ultrasound exam, no significant maternal physi- had significantly dropped to below 10.5 g/dl (M ¼ 10.4, cal or mental health problems, mother greater than 14 years old, SD ¼ 1.00); this latter level is recommended as the cutoff point and mother willing to attend the two subsequent prenatal checks defining maternal anemia in the later stages of pregnancy programmed for the study. A total of 249 mothers met these (Milman, Byg, & Agger, 2000). Ultrasound measures were taken inclusion criteria and formed our initial sample; however, in the second and third trimesters to assess fetal growth rates. Our 99 mothers of newborns in the initial sample chose to have their ultrasound measurements indicated that fetuses showed ade- baby at a hospital other then MCHC, which was our primary site.
quate growth during the course of pregnancy, with significant While we were able to make arrangements with these other gains in estimated fetal weight occurring between Trimesters 1 hospitals to assess temperament on these 99 neonates while they were still in the hospital, we were unable to arrange for neonatalblood collection at birth. In addition, for 2 infants born at MCHC,problems in videotaping led to loss of temperament assessments.
Thus, our final sample consisted of 148 newborns for whom we After the infant was born, using a standardized examination had complete data on pregnancy measures during the second and procedure, trained Institute for Nutritional Investigation (IIN) third trimesters and iron status and temperament at birth.
Comparison of infants for whom we had complete data versusthose for whom we had no data on neonatal iron indicated 1Because the primary focus in this article is on neonatal iron, we only higher levels of alertness (t ¼ 4.59, p <.01) and activity level briefly summarize the measures and procedures used to assess maternal (t ¼ À3.62, p <.01), as assessed in our structured laboratory and fetal measures taken during pregnancy. Readers wishing a detailed procedure, for neonates in the no-iron measures group; however, description of how we assessed and coded our measures of maternal there were no other group differences in the other six tem- anthropometry, maternal diet, maternal biochemistry, and fetal growthduring pregnancy or our measure of neonatal anthropometry can obtain perament measures or in measures of family SES, prenatal status this information by writing to the first author. Readers wishing a detailed measures, (e.g., fetal weight, maternal glucose) or measures of description of our nonsignificant findings for these variables also can newborn status (e.g., gestational age, birth weight). With a obtain this information by writing to the first author.
field workers assessed the infant's body weight, length, and 15 min of videotape when the infant was awake. If the infant fell skinfold thickness. During the examination, field workers also asleep before the 15 min were up, the examiner returned no later examined each child for the presence of minor physical than the following day and continued videotaping until we had at anomalies. To assess levels of neonatal hemoglobin, hematocrit, serum ferritin, serum iron, total iron binding capacity, and C- All videotapes were sent to Purdue University, where they reactive protein, 4 ml of cord blood taken at birth were used. To were coded by a trained graduate-student observer or by the obtain cord blood, the cord was cut and clamped. The umbilical Purdue University co-principal investigator. In all cases, coders cord from the placenta to the clamp was held by surgical were unaware of the neonates' iron status. Coding was done tweezers, and a needle with a syringe was inserted into the vein.
using 15-s observational cycles. The coder observed the video The blood sample was removed, put into a test tube, and until the sound of a recycling timer indicated that 15 s had centrifuged within 30 min to separate out the serum prior to elapsed. At this time, the coder paused the videotape and storage. All blood determinations were performed at the recorded the ratings. When the recycling timer indicated that laboratory of the IIN, and a standardized set of procedures were another 15-s cycle was starting, the coder began the videotape implemented to ensure that cord blood drawn would reach the and observed for another 15-s cycle. Both coders independently IIN laboratory without contamination or deterioration. The coded a subset of tapes to assess intercoder reliability. Intraclass analytic quality of data from the IIN laboratory has been correlations were used to assess level of intercoder reliability evaluated and has received high rankings from a number of (Bartko, 1976). Across the five dimensions coded (discussed standardization agencies including the National Institute of next), the average intraclass r ¼.95, with a range from r ¼.91 to Standards and Technology (Gaithersburg, MD, USA) and the Swedish National Food Administration Quality Control Pro- Based on the neonatal temperament coding criteria developed by Riese (1987) and by Ricciuti and Breitmeyer (1988), in both Neonatal temperament outcomes were assessed on two assessment contexts we coded neonatal state plus four dimen- separate occasions. The first assessment occurred within 2 days sions of neonatal temperament: alertness, activity level, negative after the infant was born and took place during the structured emotionality (distress), and soothability. Initial state at the start laboratory procedure assessing neonatal anthropometry and of the temperament assessment procedure was coded using a 6- minor physical anomalies. The second assessment occurred after point scale ranging from asleep to crying, with midpoint codes the infant was discharged from the hospital and was an being regarded as optimal state. Alertness was coded using a 4- unstructured observation at the infant's home between the Days point scale ranging from no visual orienting to high alertness.
3 and 7 of life. In the structured observation, the neonate was Activity was the summed score across two codes: head and undressed and trained examiners weighed the neonate, assessed number of limbs moved plus vigor of movement (coded on a 4- length using a Franklin plane, and assessed skinfold thickness point scale ranging from no movement to very large move- using calipers. To assess minor physical anomalies, examiners ments). Negative emotionality was coded using a 4-point scale assessed distance between the eyes, whether ears were set low on ranging from no negative affect to intense distress. When the head, 5th-finger curvature, palmer crease, and asymmetry in neonates were coded as showing moderate or intense distress, toe size using the procedure developed by Waldrop, Bell, soothability was then coded. The soothability code was based on McLaughlin, & Halverson (1978). Because both the anthro- the number of different soothing manipulations it took pometry and the physical anomalies exam required manipulating (range ¼ 0-6) before the neonate returned to a nondistressed the neonate in specific ways, this procedure can be regarded as state (Detailed coding criteria for each dimension of tempera- the more stressful of the two examination situations. If the ment are available from the principal author upon request.) neonate became distressed during the examination, except forholding the child between procedures, no attempts at soothing The unstructured observation was carried out at home at a A culturally relevant SES inventory developed at IIN was time between feedings when the neonate was awake. The administered upon enrollment of the mothers. This instrument neonate was placed on his or her back on a bed or a couch in the included data on education of the parents, construction material home, and mothers were instructed not to touch or talk directly to of their house, services at home (water, electricity, sewage), their child unless instructed to do so by the examiner. If the number of siblings, occupation of the parents, and household neonate became distressed, maternal intervention was requested possessions. Measures of maternal dietary intake and maternal as part of a standardized soothing procedure. In contrast to the caffeine and alcohol use during pregnancy were obtained via structured procedure, no demands were placed on the neonate dietary recall measures obtained during the second and third during the unstructured home observation, and neonatal distress trimesters of pregnancy, as described earlier. Measures of was responded to relatively quickly.
maternal C-reactive protein and maternal glucose levels during Both structured and unstructured assessments were video- pregnancy and neonatal C-reactive protein at birth were obtained taped. Because of the more intrusive nature of the structured from blood biochemistry assessments, as described previously.
assessments, neonates did not fall asleep during these proce- Gestational age was assessed from ultrasounds taken during dures; therefore, all structured assessments were videotaped in a pregnancy, as also described previously. Measures of birth single session that took between 15 to 20 min to complete. For weight, type of delivery, and neonatal Apgar scores were taken the unstructured procedure, our goal was to obtain a minimum of Neonatal Iron Status and Neonatal Temperament correlation between the eight measures of neonatal temperamentand the three measures of neonatal iron status. A test of the null In assessing the possible contributions of neonatal biochemistry, hypothesis that all correlations between multiple predictor and we utilized the most direct measures of early iron status: neonatal multiple criterion variables are zero order is equivalent to a serum iron, serum ferritin, and hemoglobin levels. Both visual nonsignificant canonical correlation between a set of predictor analysis of the distribution and descriptive statistics indicated and a set of criterion variables (Johnson & Wichern, 1982); that hemoglobin (skew ¼ À.300, kurtosis ¼ À.147) and serum however, if the overall canonical correlation between our indices iron (skew ¼.136, kurtosis ¼.177) approximated a normal of neonatal temperament and our indices of neonatal iron status distribution. While the distribution for ferritin was shifted to is significant, this means that at least some dimensions of the left (skew ¼ 1.325, kurtosis ¼ 3.173), this distribution was neonatal iron status are significantly related to neonatal tem- normalized utilizing a square-root transformation (skew ¼.342, perament over and above chance levels. To determine which kurtosis ¼.748); however, given that our results were essentially specific nutritional and temperament variables were producing identical when we used the nontransformed or the square-root the overall significant canonical correlation in the second stage transformed ferritin values, for ease of comparison we report the of data analysis, we computed a series of eight multiple re- findings for the nontransformed ferritin values.
gressions, regressing our neonatal iron measures onto each In contrast to results for older children and adults, far less is measure of neonatal temperament. By use of this two-stage known about the functional consequences of deficits in measures strategy, we both minimize the likelihood of capitalizing on of neonatal iron status, particularly with regard to the question of chance (initial canonical correlation) while maintaining our whether a continuous or a discrete (threshold) model is more ability to identify specific iron-temperament relations (Stage 2 appropriate. Given the lack of definitive evidence on an appropriate cutoff point, we treated each of our neonatal iron In two sets of additional analyses, we used the mediation test measures as a continuous variable. When significant relations model described by Baron and Kenny (1986). According to were found between neonatal iron and neonatal temperament Baron and Kenny, if a mediating variable is responsible for the measures using a linear model, we then reanalyzed the data to see relation between predictor and outcome, then statistically if a nonlinear model provided a better fit. When a nonlinear controlling for the influence of the mediator will reduce the model provided a better fit to the data, we inspected the scatter predictor-outcome relation to nonsignificance. The standard plot to determine what range of neonatal iron values was most mediation test model described by Baron and Kenny requires that both the predictor and alternative mediating variables Given the potential importance of considering context when be significantly related to both each other and to the outcome looking at indices of neonatal temperament, we chose to variable. Given the stringency of this criteria and the exploratory separately analyze the four neonatal temperament dimensions nature of our research, it was decided to utilize a less stringent set for each testing situation. Because of deviations from a normal of criteria in our second set of mediation analyses, designed to distribution, activity level in the structured situation and determine if non-iron variables were responsible for observ- alertness in the unstructured observation were transformed using ed relations between our measures of neonatal iron and neonatal log transformation (thus reversing directionality). For distress, temperament. Specifically, variables were chosen for further we used a percentage score, reflecting the percent of observa- study in our second set of mediation analysis if they were tional blocks the child was scored as displaying either moderate significantly related to any one of our three iron predictors or if or high levels of distress. The %score in turn was transformed they were significantly related to more then one of our eight using a square-root transformation. For soothability, using cluster analysis procedures children were classified into one ofthree groups reflecting either low soothability, moderatesoothability, or high soothability (Details of group-compositioncriteria may be obtained from the principal author.) Because of the paucity of previous data on the question of nutritional contributions to variability in neonatal temperament, we felt that greater than usual consideration needed to be given tominimizing Type II errors, in addition to the usual emphasis on Only 2% of our sample of newborns had birth weights less avoiding Type I errors. Our analytic goal was to find an approach than 2.5 kg, 66.3% had birth weights between 2.5 and which minimized the likelihood of capitalizing on chance by 3.49 kg, and the remaining 31.5% had birth weights using too many comparisons, yet would allow us to look at greater than 3.5 kg. Only 4 neonates in our sample had relatively specific predictor-outcome relations. Because of the 1-min Apgar scores below 7, and only 1 neonate had a many unknowns with regard to relations among our different 5-min Apgar score below 7. There were only 2 preterm measures of neonatal iron status and our measures of neonatal infants in our sample (The range of gestational ages at temperament, we did not feel that structural equation modelingwould be appropriate given that this approach requires having birth was 32-39 weeks.) Data for the preterm infants were explicit and plausible models of the links among the different predictors and different outcomes (Schumacker & Lomax, With regard to our measures of neonatal iron, the mean 1996). Given our analytic requirements, we chose a two-stage value for serum iron was 148.79 ug/dl (SD ¼ 44.99 ug/dl).
analytic strategy. In the first stage, we computed the canonical The normal range of serum iron in full-term neonates is 125 to 225 ug/dl (Siner & Newman, 2002). A total of significant, correlations (St. James-Roberts & Wolke, 30.8% of our sample had serum iron levels below 125 ug/ 1988). There was a trend for neonates to be in a more alert, dl; 3.8% had values above 225 ug/dl. The mean ferritin active state at the start of the unstructured observation level in our sample was 153.19 ug/L (SD ¼ 78.67 ug/L).
session. As would be expected, the correlation of initial The normal range of ferritin in full-term neonates is neonatal state across testing contexts was nonsignificant between 25 to 200 ug/L (Siner & Newman, 1997). Only 1 neonate had a ferritin value below 25 ug/l whereas22.6% had values exceeding 200 ug/l. The mean hemo- Neonatal Temperament and Neonatal Iron Status. The globin level in our sample was 15.6 g/dl (SD ¼ 1.4 g/dl).
overall canonical correlation between our three measures The normal range of hemoglobin values in full-term of neonatal iron status (Hb, serum ferritin, and serum iron) neonates is 14.0 to 20.0 g/dl (Siner & Newman, 2002).
and our eight neonatal temperament scores (soothability, Fifteen percent of our sample had hemoglobin levels alertness, activity, and distress assessed in the structured below 14 g/dl, and none were above 18.9 g/dl. While and unstructured settings) was statistically significant neonatal serum iron and ferritin levels were significant- (Rc ¼.46, Bartlett's w2 ¼ 55.87, df ¼ 24, n ¼ 148, p <.01).
ly correlated with each other in the expected direc- As discussed earlier, significance of the total canonical tion (r ¼.21, p <.01), the correlation between neonatal correlation gives us overall protection against a Type 1 ferritin and hemoglobin, while significant, was negative error when we break down analyses to determine what (r ¼ À.22, p <.01); serum iron and hemoglobin levels aspects of neonatal iron status and neonatal temperament were uncorrelated (r ¼ À.01, n.s.).
were driving the significant canonical correlation.
The means and SDs (in parentheses) for neonatal To break down the significant canonical correlation, we temperament scores are shown in Table 1. As shown, ran eight multiple regressions using our three neonatal analysis of neonatal temperament in the structured and iron status measures as predictors and individual neonatal unstructured contexts revealed that while there were no temperament dimensions assessed in either the structur- differences in activity level across the two contexts, there ed or the unstructured setting as outcome variables. Five were significantly higher levels of neonatal alertness in the of the eight regressions were significant (Table 2). For unstructured testing situation, along with significantly temperament assessed in the structured laboratory situa- higher levels of distress and higher numbers of less tion, neonates with higher hemoglobin levels were rated soothable neonates in the structured testing situation.
as more alert. Also shown in Table 2, neonates with higher Given the differing nature of the testing situations, these levels of serum iron were rated as showing less distress significant differences would be expected. As shown in and as being less active (activity log transformed, so Table 1, analyses also indicated low, but significant, cross- directionality reversed and higher activity scores mean contextual correlations between each of our temperament lower activity). In the unstructured assessment, results domains. These results are consistent with results from again indicated significantly greater alertness for neonates other studies examining the stability of different measures with higher hemoglobin levels (alertness log transform- of neonatal temperament assessed in different contexts in ed, so directionality reversed). In addition, neonates with the first week of life which also reported modest, but higher levels of serum iron were more likely to be rated as % time blocs showing high or moderate distress tp <.06.
aTrue directionality of relation reversed due to log transformation of activity in the structured laboratory assessment and alertness in the unstructured Neonatal Iron Status and Neonatal Temperament Significant Regression Analyses Relating Neonatal range. Increased activity level was found for neonates with serum iron levels below 85 ug/dl, and decreases inactivity level were found for neonates with serum iron Given evidence cited earlier on the pattern of relations between iron and temperament, our finding that lower serum iron level is related to higher laboratory-assessed activity level seems counterintuitive; however, our find- ings do suggest an alternative interpretation. As reported earlier, higher serum iron is related to both lower activity level and lower distress in our laboratory assessment. Our results also indicate that higher distress in the laboratory situation is related to higher activity level (r ¼ À.52, p <.01; activity level log transformed, hence the negative correlation). This pattern suggests the possibility that neonatal distress may be mediating the relation between serum iron and activity level. Mediation occurs when significant relations between a specific predictor (serum iron) and outcome variable (activity level) occur because of the influence of a third (mediating) variable common to both predictor and outcome (distress). To test this mediat- ing hypothesis, using the Baron and Kenny procedure described earlier, we reran the regression between our neonatal iron measures and lab activity, initially entering neonatal lab distress. Under these conditions, the previ- ously significant relation between serum iron and labo- ratory activity level dropped to nonsignificance in theregression (t ¼ 1.187, b ¼.091, n.s.). What this result indicates is that neonates with low serum iron levels are more likely to be distressed, and that one of the con- aTrue directionality of relation reversed due to log transformation of sequences of greater distress is higher activity level.
activity in the structured laboratory assessment (low score means highactivity) and alertness in the unstructured home assessment (low score Additional Tests for Mediation Effects. A final set of means high alertness) or to reverse coding of soothability group (low analyses was utilized to assess whether observed relations between neonatal temperament and measures of neonataliron status could be attributed to an alternative mediating easier to soothe following distress (Negative beta reflects variable other than iron, per se. As described earlier, variables were chosen for further mediation analysis if We reanalyzed each of the five regressions shown in they were significantly related to any one of our three iron Table 2 to determine if a better fit to the data could be predictors or if they were significantly related to more than obtained using a nonlinear model (log, quadratic, cubic) one of our eight temperament outcome measures.
rather than our original linear model. For all analyses In the initial step of our mediation analysis, based on involving hemoglobin and for those analyses involving our review of the literature and reviewers' comments, serum iron, distress, and soothability, no improvement we identified 10 potential alternative mediating variables was found when a nonlinear analysis was used, suggesting that were available in our database and which previous that relations are occurring across the full range of iron research had related to either neonatal iron or neonatal values. In only one analysis involving serum iron and lab temperament. Four of the mediating variables chosen had activity (change in R2 ¼.047, p <.01) was a better fit ob- been identified in previous research as potentially relevant tained using a nonlinear cubic model.2 For serum iron and to individual differences in neonatal temperament.
lab activity, a cubic model suggested little relation be- These variables included neonatal state (Sameroff, tween activity level and serum iron in the 85- to 185-ug/dl Krafchuck, & Bakow, 1978), birth weight (Garcia-Coll,Halpern, Vohr, Seifer, & Oh, 1992; Riese, 1994; Sajaniemi, Salokorpi, & vonWendt, 1998), maternal Readers wishing to have a copy of our results for the nonlinear regressions can obtain these by writing to the first author.
medication during delivery (Lester, Als, & Brazelton, 1982; because we did not have adequate record data of use intake, gestational age, and Apgar. While there was of anesthetic or analgesics during delivery, we assessed modest variation in both betas and significance levels, whether the baby was born by C-section as a proxy), and with one exception all significant iron-temperament maternal caffeine intake during pregnancy (Engle et al., relations reported in Table 2 remained significant after 1999). An additional three variables chosen had been accounting for the influence of state, maternal glucose identified in previous research as potentially relevant to during pregnancy, maternal alcohol intake during preg- individual differences in neonatal iron status. These vari- nancy, gestational age, and 1-min Apgar.3 The one ex- ables included family SES measures (parental education ception involved the relation between serum iron and level, home quality, and family possessions: Grantham- laboratory activity level, which dropped below the tradi- McGregor, Fernald, & Sethuraman, 1999), neonatal tional criterion for statistical significance (b ¼.137, biomedical status (Allen, 2000; Lee & Nieman, 1996: t ¼ 1.68, p <.10) after variance associated with the neo- assessed via 1- and 5-min Apgar scores and neonatal nate's initial state at the start of the laboratory session was C-reactive protein), and maternal diabetes (Georgieff, Schmidt, Mills, Radner, & Widness, 1992; Nelson et al., As discussed previously, we also investigated the ques- 2000; Petry et al., 1992; measures of maternal blood tion of whether gender moderated our pattern of findings.
glucose level obtained during pregnancy were used as a While there were no gender differences in our iron mea- marker variable for risk of maternal diabetes). Three sures, males were more alert and less distressed than additional measures also were selected based on prior females in the laboratory setting while females were more research relating these variables to both neonatal tempera- active in the home assessment. To test for moderation, as ment and neonatal iron status. These three additional mea- recommended by Baron and Kenny (1986), Gender  Iron sures were gestational age (Luchtman-Jones, Schwartz, & interaction terms were entered as a second step in those Wilson, 2002; Ricciuti & Breitmeyer, 1988), maternal regressions previously identified as significant in our intake during pregnancy of calcium and zinc (Guiang & initial analyses. Two significant Gender  Iron interac- Georgieff, 1998; Merialdi, Caulfield, Zavaleta, Figueroa, tions were identified. The first was a Gender  Ferritin & DiPietro, 1998), and maternal alcohol intake during interaction for alertness assessed in the laboratory pregnancy (Miller, Roskams, & Connor, 1995; Weinberg, situation (t ¼ À2.29, b ¼ À.469, p <.05); the second 1997). In addition to these 10 potential mediators, we also also was a Gender  Ferritin interaction for soothability looked at gender as a possible moderator of our findings assessed at home (t ¼ À2.10, b ¼ À.431, p <.05). Break- relating neonatal temperament to neonatal iron measures, down of the first interaction indicated that the relation given evidence suggesting the possibility of gender dif- between ferritin and alertness assessed in the laboratory ferences in both iron metabolism at birth (Choi et al., setting was significant for females (r ¼.29, p <.01), but 2000; Tamura et al., 1999) and infant temperament (Eaton not for males (r ¼ À.08, n.s.), with the difference between the two correlations statistically significant by r-z trans- In Step 2 of our mediational analysis, we assessed form test (z ¼ 2.34, p <.05). Similarly, the correlation whether the 10 variables selected met our criteria of being between ferritin and soothability assessed in the home was related to any one of our neonatal iron measures or to significant for females (r ¼ À.26, p <.05), but not for more than one of our neonatal temperament measures.
males (r ¼ À.18, n.s.), with the difference between the two In this analysis, our measures of family SES, maternal correlations again being statistically significant (z ¼ 2.70, dietary intake of calcium and zinc during pregnancy, maternal caffeine intake during pregnancy, C-sectionbirth, neonatal C-reactive protein, and birth weight weredropped from further consideration since they did not meet either of these criteria. Five variables met one of theaforementioned criteria, namely neonatal state (met the Until now, the largest body of research on the biological Baron & Kenny, 1986, criteria), maternal glucose levels roots of temperament has focused on the genetics of during pregnancy (related to more than one tempera- individual differences in temperament. In terms of other ment outcome), maternal alcohol intake during pregnancy potential biological influences, at least for our sample, (related to more than one temperament outcome), neo- results indicate little relation between prenatal nutritional natal gestational age (related to more than one tempera- factors such as maternal nutritional intake or fetal growth.
ment outcome), and 1-min Apgar score (related to Our nonsignificant findings for these predictors does not neonatal hemoglobin). These variables were each ini-tially entered into our iron-temperament regressions to 3Readers wishing to have a copy of our results for these mediational determine if our pattern of findings would change after analyses or our moderation analysis can obtain these by writing to the accounting for the influence of state, glucose, alcohol Neonatal Iron Status and Neonatal Temperament necessarily mean that maternal nutrition during preg- neonatal temperament. To test for alternative mechanisms nancy is unrelated to individual differences in early besides iron, we first identified a set of 10 potential temperament. There remains the question of whether such mediators based on our review of the literature and prediction might occur with a sample that is more severely suggestions from reviewers of this article. We eliminated a malnourished than the women in our sample.
number of these potential mediators from consideration What our results do suggest is the potential relevance (family SES, maternal caffeine, calcium and zinc intake of neonatal iron status to variability in early tempera- during pregnancy, C-section birth, 5-min Apgar, neonatal ment. Specifically, our findings indicate that lower levels C-reactive protein, and birth weight) based on the fact of neonatal hemoglobin are related to lower levels of that they were unrelated to any of our three neonatal iron neonatal alertness, assessed in both a structured labora- measures and to less than two of our eight neonatal tory-examination procedure and in an unstructured temperament measures. We then recomputed our analyses observation done in the neonate's home. The relation to determine if our pattern of findings would change after between hemoglobin level and neonatal alertness appears partialling out the variance associated with the remaining to be linear, operating across the full range of hemoglobin mediators that did correlate with any measure of neonatal values, with no evidence for a threshold above or below iron or more than one of our temperament measures: which hemoglobin is unrelated to alertness. Our results neonatal state, maternal glucose level during pregnancy, also indicate that lower levels of neonatal serum iron are maternal alcohol use during pregnancy, 1-min Apgar, related to increased neonatal negative emotionality, as and gestational age. With the exception of the relation seen in greater distress and distress-mediated activity between serum iron and neonatal activity, all initial level assessed during the laboratory testing and reduced significant findings on the relation of hemoglobin and soothability assessed during the unstructured home ob- serum iron to neonatal temperament remained significant, servation. What appears to be a setting difference (Serum even after controlling for the influence of these alternative iron predicts distress in the laboratory assessment and soothability in the home assessment.) may well reflect Obviously, elimination of the aforementioned vari- procedural factors, given that examiners did not attempt to ables as potential mediators of our findings in and of soothe neonates during the laboratory assessment while itself does not indicate that we can attribute variability in distress was responded to quickly in the home-observation neonatal temperament primarily to variability in neonatal setting. Rather than setting differences, what our findings iron status, in part because of the limited amount of unique suggest are differential relations, with hemoglobin related variance associated with neonatal iron and because this to alertness, which may be an early manifestation of list of mediating factors is not exhaustive. The impact later developing self-regulation processes (Derryberry & of other potential mediating variables such as maternal Rothbart, 1997) while individual differences in serum cigarette use during pregnancy, specific drugs taken iron level are related to indices of negative emotional during pregnancy, or when the umbilical cord is clamped reactivity. While ferritin was unrelated to any of our following delivery could not be assessed because this temperament measures when assessed as a main information was not in our database. While we cannot effect, our results do suggest the possibility of gender conclusively eliminate all possible mediators as an alter- differences, with higher levels of ferritin being related to native explanation for our findings, we can conclude that higher alertness and soothability for females, but not for our findings remained robust even after taking into account some of the most likely alternative explanations.
It is obvious from our regressions that neonatal iron Given that our results relating variability in neonatal measures offer only a partial explanation of variability temperament to measures of neonatal iron status cannot be in neonatal temperament. Depending upon the specific easily attributed to alternative non-iron explanations, a iron and temperament measures assessed, neonatal iron critical question is the mechanism that underlies our accounts for between 5 to 10% of unique variance in findings. Any iron-related explanatory mechanisms offer- neonatal temperament. It is possible that the relatively ed must be considered in relation to the many issues small amount of predictive variance associated with neo- involved in conceptualizing the nature and consequences natal iron would have been greater had there been more of measures of individual differences in neonatal iron neonates with severe levels of iron deficiency in our status (Beard & Connor, 2003). The most obvious ex- population (discussed later); however, at least within this planatory mechanism would be iron deficiency, given study, our findings with regard to neonatal iron do appear what is known about the effects of iron deficiency on CNS to be robust. As part of our analytic strategy, we con- development (Beard & Connor, 2003; Rao & Georgieff, sidered the possibility that our results were due to the 2000) and the relation of CNS function to temperament actions of an alternative mediating variable that covaries (Wachs, 2000). Support for iron deficiency as the mech- with neonatal iron status and with individual variability in anism underlying our findings can be found in the consistency between our pattern of associations relating deficiency remains a critical question that must be dealt neonatal iron measures to increased neonatal distress and with before we can conclude that our findings represent reduced neonatal alertness and soothability with results a downward extension of the findings of Lozoff and from infrahuman studies and studies with older iron- colleagues (Lozoff et al., 1996, 1998, 2003), or that deficient infants. At the infrahuman level, our findings are relations between iron deficiency and individual varia- consistent with evidence indicating reduced attention to bility in temperament may be occurring earlier in life then environmental cues in iron-deficient as opposed to iron- sufficient young rats (Beard, Erikson, & Jones, 2002). At In terms of other potential mechanisms of iron meta- the human level, our results indicating reduced attention bolism besides iron deficiency, exposure to excess iron and increased negative emotionality are consistent with during the neonatal period has been linked to impaired the findings by Lozoff and colleagues (2003; Lozoff et al., neurological function during the first year of life, per- 1998; Lozoff et al., 1996) based on studies with iron- haps as the result of oxidation due to free radical toxicity (Buonocore et al., 2003). However, given inadequate However, before attributing our pattern of findings to maternal dietary iron during the prenatal period, a rel- iron deficiency, it is essential to consider the nature of our atively low frequency of mothers taking iron supplemen- population. While there was an increased risk of iron tation during pregnancy and the lack of markers indicating deficiency in our sample, the majority of neonates in our iron overload in neonates in our sample (e.g., excessively sample were not iron deficient. As illustrated in our high ferritin levels: Guiang & Georgieff, 1998) as well as descriptive data, while 30% of our sample was below the linear rather than nonlinear threshold patterns relating threshold for normal levels of serum iron and 15% was iron to temperament, iron excess does not seem to be a below threshold for hemoglobin, the mean sample values likely alternative mechanism. A related mechanism, sug- for serum iron and hemoglobin were above threshold.
gested by one reviewers of this article, was that the Further, only 1 neonate was below the normal threshold underlying mechanism could involve iron homeostasis for ferritin. In addition, our curve fitting analyses indi- rather then iron deficiency, per se. The importance of cated that in almost all cases a linear model fit the data, developing adequate iron regulation in neonates is well indicating that iron-temperament relations were operat- established (Guiang & Georgieff, 1998). It is possible ing across the full range of iron values rather than being that the negative correlation found between ferritin and restricted to neonates with the lowest iron values.
hemoglobin levels in our sample may reflect neonatal Given the nature of our sample, targeting iron defi- attempts to regulate iron homeostasis. Based on evidence ciency as an explanatory mechanism would require an from both infrahuman and human samples, Georgieff and additional assumption, namely that neonates are highly colleagues suggested that such a negative relation may be sensitive to even low levels of iron deficiency. The biologically plausible, reflecting a reallocation of storage possibility that early brain development may be highly iron into the red-cell mass to meet increased neonatal vulnerable to iron deficiency has been raised by re- iron demands (Amarnath, Ophoven, Mills, Murphy, & searchers studying infrahuman populations (Youdim, Georgieff, 1989; Georgieff et al., 1990; Georgieff, Ben-Shachar, & Yehuda, 1989). Congruent with the Widness, Mills, & Stonestreet, 1989; Guiang, Georgieff, conclusion of Youdim and colleagues (1989), Felt and Lambert, Schmidt, & Widness, 1997); however, negative Lozoff (1996) reported reduced orientation and activity correlations between storage iron and hemoglobin are levels in 8-day-old rat pups with normal hemoglobin more likely to occur in preterm neonates undergoing that had been exposed to an iron-deficient intrauterine specific stresses that require iron mobilization, such as environment during their early gestational period. Unfor- hypoxemia (Rao & Georgieff, 2002). These conditions do tunately, little human evidence is available testing the not characterize our sample. Further, in contrast to iron assumption of high susceptibility to low levels of early deficiency and iron excess where there are known biol- iron deficiency, particularly for behavioral-developmental ogical consequences that could act to influence neonatal outcomes. As noted earlier, Tamura et al. (2002) showed temperament, the underlying mechanism through which long-term developmental consequences for neonates with poorly regulated iron homeostasis could translate into reduced levels of cord serum ferritin; however, Tamura variability in neonatal temperament is unclear. Results et al.'s (2002) significant findings were restricted to from infrahuman studies have suggested that when stor- neonates in the lowest quartile of the ferritin distribu- age iron is depleted in the fetal or neonatal period, tion (<76 ug/l). Our results with a different age group physiological regulation processes prioritize remaining and outcome measures indicated a ferritin -temperament iron towards red blood cell production at the expense of link only for females, with no evidence supporting the organs such as the brain (Guiang et al., 1997). While little superiority of a nonlinear model. Thus, whether human human evidence is available on the developmental con- neonates are particularly susceptible to low levels of iron sequences of such iron-regulation processes, redirection Neonatal Iron Status and Neonatal Temperament of iron away from the brain could have potential in- Bates, J. (1989). Applications of temperament concepts. In fluences on neonatal behavioral patterns.
G. A. Kohnstamm, J. E. Bates, & M. K. Rothbart (Eds.), While much is known from infrahuman studies about Temperament in childhood (pp. 321-355). Chichester, the physiological consequences of iron deficiency in the fetal and neonatal period, there is remarkably little Beard, J., & Connor, J. (2003). Iron status and neural functioning. Annual Review of Nutrition, 23, 41-58.
evidence on the behavioral consequences, particularly in Beard, J., Erikson, K., & Jones, B. (2002). Neurobehavioral human populations (Beard & Connor, 2003). While our analysis of iron deficiency in rats. Behavioral Brain findings support a link between individual variability in neonatal iron status and individual variability in neonatal Braungart, J., Fulker, D., & Plomin, R. (1992). Genetic media- temperament, the nature of the process underlying these tion of the home environment during infancy. Developmental findings remains a question for future research. The results presented here support the importance of increas- Buonocore, G., Perrone, S., Longini, M., Paffetti, P., Vezzosi, ed research on the early functional-behavioral conse- P., Gatti, M., & Bracci, R. (2003). Nonprotein bound iron as quences of individual differences in iron status as well as early predictive marker of neonatal brain damage. Brain, on the mechanisms that underlie such consequences, particularly in populations where there is a high incidence Choi, J., Kim, C., & Par, S. (2000). Erythropoietic activity and soluble transferrin receptor level in neonate and maternal of more severe early iron deficiency.
blood. Acta Paediatrica, 89, 675-689.
Colmer, J., Colomer, C., Gutierrez, D., Jubert, A., Nolasco, A., Donat, J., Fernandez-Delgado, R., Donat, F., & Alvarez- Dardet, C. (1990). Anaemia during pregnancy as a risk factorfor infant iron deficiency: Report from the Valencia InfantAnaemia Cohort (VIAC) study. Paediatric and Perinatal The authors acknowledge the immense help of Patricia Ba'rrig for the supervision of psychological procedures, Lizette Ganoza Crockenberg, S., & Smith, P. (1982). Antecedents of mother- and Giovanna Rios for supervision of the nutritional field work,Feyza Corapci for scoring videotapes, and Fernando Andrade, infant interaction and infant irritability in the first three Mark Grudberg, and Pinar Gurkas for the statistical analysis. We months of life. Infant Behavior and Development, 5, 105- also acknowledge the help of Dr. Mario Merialdi for training in the ultrasound evaluation and Dr. Michael Georgieff for his Derryberry, D., & Rothbart, M. (1997). Reactive and effortful incisive comments on preliminary drafts of this article. Special control processes in the organization of temperament.
thanks are due to the Canto Grande Maternal Child Health Development and Psychopathology, 9, 633-652.
Center and the families in Canto Grande who participated in our de Ungria, M., Rao, R., Wobken, J., Luciana, M., Nelson, M., & Georgieff, M. (2000). Perinatal iron deficiency decreasescytochrome c oxidase (CytOx) activity in selected regions ofneonatal rat brain. Pediatric Research, 48, 169-176.
Dobbing, J. (1990). Vulnerable periods in the developing brain.
In J. Dobbing (Ed.), Brain behavior and iron in the infant diet(pp. 1-17). London: Springer-Verlag.
Agrawal, R. M. D., Tripathi, A. M., & Agarwal, K. N. (1983).
Eaton, W., & Enns, L. (1986). Sex differences in human motor Cord blood haemoglobin, iron and ferritin status in maternal activity level. Psychological Bulletin, 100, 19-28.
anaemia. Acta Paediatrica Scandinavica, 72, 545-548.
Engle, P., VasDias, T., Howard, I., Romero-Abal, M., Quan de Allen, L. (1997). Pregnancy and iron deficiency: Unresolved Serrano, J., Bulux, J., Solomons, N., & Dewey, K. (1999).
issues. Nutrition Reviews, 55, 91-101.
Effects of discontinuing coffee intake on iron deficient Allen, L. (2000). Anaemia and iron deficiency: Effects on Guatemalan toddlers' cognitive development and sleep.
pregnancy outcome. American Journal of Clinical Nutrition, Early Human Development, 53, 251-269.
Felt, B., & Lozoff, B. (1996). Brain iron and behavior of rats Amarnath, U., Ophoven, J., Mills, M., Murphy, E., & Georgieff, are not normalized by treatment of iron deficiency anemia M. (1989). The relationship between decreased iron stores, during early development. Journal of Nutrition, 126, 693- serum iron and neonatal hypoglycemia in large-for-date newborn infants. Acta Paediatrica Scandinavica, 78, 538- Garcia-Coll, C., Halpern, L., Vohr, B., Seifer, R., & Oh, W.
(1992). Stability and correlates of change of early tempera- Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator ment in preterm and full term infants. Infant Behavior and distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Person- Georgieff, M., Landon, M., Mills, M., Hedlund, B., Faassen, A., ality and Social Psychology, 51, 1173-1182.
Schmidt, R., Ophoven, J., & Widness, J. (1990). Abnormal Bartko, J. C. (1976). On various intraclass correlation reliability iron distribution in infants of diabetic mothers: Spectrum and coefficients. Psychological Bulletin, 83, 762-765.
maternal antecedents. Journal of Pediatrics, 117, 455-461.
Georgieff, M., Schmidt, R., Mills, M., Radner, W., & Widness, Lozoff, B., Wolf, A., & Jimenez, E. (1996). Iron deficiency J. (1992). Fetal iron and cytochrome C status after in- anemia and infant development: Effects of extended oral iron trauterine hypoxemia and erythropoietin administration.
therapy. Journal of Pediatrics, 129, 382-389.
American Journal of Physiology: Regulatory, Integrative Luchtman-Jones, L., Schwartz, A., & Wilson, D. (2002).
and Comparative Physiology, 262, R485-R491.
Hematologic problems in the fetus and neonate. In A.
Georgieff, M., Wewerka, S., Nelson, C., & Deregnier, R.
Fanaroff & R. Martin (Eds.), Neonatal-perinatal medicine: (2002). Iron status at 9 months of infants with low iron stores Diseases of the fetus and infant (7th ed., pp. 1183-1238).
at birth. Journal of Pediatrics, 141, 405-409.
Georgieff, M., Widness, J., Mills, M., & Stonestreet, B. (1989).
Mahomed, K. (2003). Iron supplementation in pregnancy The effect of prolonged intrauterine hyperinsulinemia on iron (Cochrane Review). The Cochrane Library, Issue 3, U.K.
utilization in fetal sheep. Pediatric Research, 26, 467-469.
Merialdi, M., Caulfield, L., Zavaleta, N., Figueroa, A., & Grantham-McGregor, S., Fernald, L., & Sethuraman, K. (1999).
DiPietro, J. (1998). Adding zinc to prenatal iron and folate Effects of health and nutrition on cognitive and behavioural tablets improves fetal neurobehavioral development. FASEB development in children in the first 3 years of life: Part 2.
Infections and micronutrient deficiencies: Iodine, iron and Michaelsen, K., Milman, N., & Samuelson, G. (1995). A longi- zinc. Food and Nutrition Bulletin, 20, 76-99.
tudinal study of iron status in healthy Danish Infants. Acta Guiang, S., & Georgieff, M. (1998). Perinatal iron, trace minerals and vitamin metabolism. In R. Polin & W. Fox Miller, M., Roskams, A., & Connor, J. R. (1995). Iron (Eds.), Fetal and neonatal physiology, 2nd ed. (Vol. 1, regulation in the developing rat brain: Effect of in utero pp. 401-410). Philadelphia: Saunders.
ethanol exposure. Journal of Neurochemistry, 65, 373- Guiang, S., Georgieff, M., Lambert, D., Schmidt, R., & Widness, J. (1997). Intravenous iron supplementation effect Milman, N., Agger, A., & Nielsen, O. (1994). Iron status on tissue iron and hemoproteins in chronically phleboto- markers and serum erythropoietin in 120 mothers and mized lambs. American Journal of Physiology: Regulatory, newborn infants. Acta Obstetrica Gynecologica Scandinavia, Integrative and Comparative Physiology, 273, R2124- Milman, N., Byg, K., & Agger, A. (2000). Hemoglobin and Halvorsen, S. (2000). Iron balance between mother and infant erythrocyte indices during normal pregnancy and post- during pregnancy and breastfeeding. Acta Paediatrica, 89, partum in 206 women with and without iron supplementa- tion. Acta Obstetrica et Gynecologica Scandinavica, 79, Hokama, T., Takenaka, S., Hirayama, K., Yara, A., Yoshida, K., Itokazu, K., Kinhjo, R., & Yabu, E. (1996). Iron status of Morgane, P., Austin-LaFrance, R., Bronzino, J., Tonkiss, J., newborns born to iron deficient anaemic mothers. Journal of Diaz-Cintra, S., Cintra, L., Kemper, T., & Galler, J. (1993).
Prenatal malnutrition and development of the brain. Neuro- Institute of Medicine. (2000). Dietary reference intakes: Appli- science and Biobehavioral Reviews, 17, 91-128.
cations in dietary assessment. Washington, DC: National Nelson, C. A., Wewerka, S., Thomas, K. M., Tribby-Walbridge, S., de Regnier, R., & Georgieff, M. (2000). Neurocogni- Johnson, R., & Wichern, D. (1982). Applied multivariate tive sequelae of infants of diabetic mothers. Behavioural statistics (2nd ed.) Englewood Cliffs, NJ: Prentice Hall.
Korner, A. F., Hutchinson, C. A., Koperski, J. A., Kraemer, Newnham, C., McKenzie, B., Foddy, M., Prior, M., Ong, B., & H. C., & Schneider, P. A. (1981). Stability of individual Drew, J. (1997, April). The relationship between neonatal differences of neonatal motor and crying patterns. Child temperament-like behaviours and later temperament charac- teristics in preterm infants. Poster presented at the biennial Korner, A. F., Zeanah, C. H., Linden, J., Berkowitz, R. I., meeting of the Society for Research in Child Development, Kraemer, H. C., & Agras, W. S. (1985). The relation be- tween neonatal and later activity and temperament. Child O'Brien, K., Zavaleta, N., Abrams, S., & Caulfield, L. (2003).
Maternal iron status influences iron transfer to the fetus Lee, R., & Nieman, D. (1996). Nutritional assessment (2nd ed.) during the third trimester of pregnancy. American Journal of Lester, B., Als, H., & Brazelton, T. (1982). Maternal obstetric Petry, C. D., Eaton, M. A., Wobken, J. D., Mills, M. M., anesthesia and newborn behavior. Child Development, 53, Johnson, D. E., & Georgieff, M. K. (1992). Iron deficiency of liver, heart, and brain in newborn infants of diabetic mothers.
Lozoff, B., De Andraca, I., Castillo, M., Smith, J., Walter, T., Journal of Paediatrics, 121, 109-114.
& Pino, P. (2003). Behavioral and developmental effects of Plomin, R., Emde, R., Braungart, J., Campos, J., Corley, R., preventing iron deficiency anemia in healthy full-term Fulker, D., Kagan, J., Reznick, J., Robinson, J., Zahn-Waxler, infants. Pediatrics, 112, 846-854.
C., & DeFries, J. (1993). Genetic change in continuity from Lozoff, B., Klein, N., Nelson, E., McClish, D., Manuel, M., & 14 to 20 months. Child Development, 64, 1354-1376.
Chacon, M. (1998). Behavior of infants with iron deficiency Preziosi, P., Prual, A., Galan, P., Daooudaa, H., Boureima, H., anemia. Child Development, 69, 24-36.
& Hereberg, S. (1997). Effect of iron supplementation on the Neonatal Iron Status and Neonatal Temperament iron status of pregnant women: Consequences for newborns.
Saudino, K., Plomin, R., & DeFries, J. (1996). Tester rated American Journal of Nutrition, 66, 1178-1182.
temperament at 14, 20 and 24 months. British Journal of Rao, R., & Georgieff, M. (2000). Early nutrition and brain Developmental Psychology, 14, 129-144.
development. In C. Nelson (Ed.), The effects of early Schumacker, R., & Lomax, R. (1996). A beginner's guide to adversity on neurobehavioral development (pp. 1 - 38).
structural equation modelling. Mahwah, NJ: Erlbaum.
Siner, B., & Newman, N. (1997). Tables of normal values. In A.
Rao, R., & Georgieff, M. (2002). Perinatal aspects of iron Fanaroff & R. Martin (Eds.), Neonatal-perinatal medicine: metabolism. Acta Paediatrica Supplements, 438, 124-129.
Diseases of the fetus and infant, (6th ed., Vol. 2, pp. 1755- Ricciuti, H., & Breitmeyer, B. (1988). Observational asses- sment of infant temperament in the natural setting of a Siner, B., & Newman, N. (2002). Tables of normal values. In A.
newborn nursery. Merrill-Palmer Quarterly, 34, 281-299.
Fanaroff & R. Martin (Eds.), Neonatal-perinatal medicine: Riese, M. (1983). Assessment of behavioural pattern in Diseases of the fetus and infant (7th ed., Vol. 2, pp. 1645- neonates. Infant Behavior and Development, 6, 241-246.
Riese, M. (1987). Longitudinal assessment of temperament St. James-Roberts, I. & Wolke, D. (1988). Convergences and from birth to two years. Infant Behavior and Development, discrepancies among mothers' and professionals' assess- ments of difficult neonatal behaviour. Journal of Child Riese, M. L. (1990). Neonatal temperament in monozygotic Psychology and Psychiatry, 29, 21-42.
and dizgotic twin pairs. Child Development, 61, 1230-1237.
Tamura, T., Goldenberg, M., Hou, J., Johnston, K., Cliver, S., Riese, M. L. (1994). Discordant twin pairs: The relation Ramey, S., & Nelson, K. (2002). Cord serum ferritin between gestational age and neonatal temperament differ- concentrations and mental and psychomotor development ences in cotwins. Acta Geneticae Medicae et Gemellologiae: of children at five years of age. Journal of Pediatrics, 140, Riese, M. L. (1995). Mothers' ratings of infant temperament: Tamura, T., Hous, J., Goldenberg, R., Johnston, K., & Cliver, S.
Relation to neonatal latency to soothe by pacifier. Journal of (1999). Gender difference in cord serum ferritin concentra- tions. Biology of the Neonate, 75, 343-349.
Robinson, J., Kagan, J., Reznick, J., & Corley, R. (1992). The Vaughn, J., Brown, J., & Carter, J. (1986). The effects of heritability of inhibited and uninhibited behavior. Develop- maternal anemia on infant behaviour. Journal of the National Rothbart, M. (1989). Biological processes of temperament. In Wachs, T. D. (2000). Linking nutrition and temperament.
G. Kohnstamm, J. Bates, & M. Rothbart (Eds.), Tempera- In V. Molfese & D. Molfese (Eds.), Temperament and ment in childhood (pp. 77-110). New York: Wiley.
personality across the life-span (pp. 57-84). Mahwah, NJ: Rothbart, M., Derryberry, D., & Hershey, K. (2000). Stability of temperament in childhood: Laboratory infant assessment to Wachs, T. D., & Bates, J. (2001). Temperament. In G. Bremner parent report at seven years. In V. Molfese & D. Molfese & A. Fogel (Eds.), Handbook of infant development (pp.
(Eds.), Temperament and personality development across the 465-502). Oxford, England: Blackwell.
life span (pp. 85-120). Mahwah, NJ: Erlbaum.
Wachs, T. D., Pollitt, E., Cueto, S., & Jacoby, E. E. (2004).
Rothbart, M., Derryberry, D., & Posner, M. (1994). A Structure and cross-contextual stability of neonatal tempera- psychobiological approach to the development of tempera- ment. Infant Behavior and Development, 27, 382-396.
ment. In J. Bates & T. D. Wachs (Eds.), Temperament: Waldrop, M., Bell, R., McLaughlin, B., & Halverson, C. (1978, Individual differences at the interface of biology and behavior February). Newborn minor physical anomalies predict short (pp. 83-116). Washington, DC: American Psychological attention span, peer aggression and impulsivity at age 3.
Sacco, L. M., Caulfield, L. E., Zavaleta, N., & Retamozo, L.
Weinberg, N. (1997). Cognitive and behavioral deficits asso- (2003). Dietary pattern and usual nutrient intakes of Peruvian ciated with parental alcohol use. Journal of the American women during pregnancy. European Journal of Clinical Academy of Child & Adolescent Psychiatry, 36, 177-186.
Worobey, J. (1986). Convergence among assessments of tem- Sajaniemi, N., Salokorpi, T., & VonWendt, L. (1998). Tem- perament in the first month. Child Development, 57, 47-55.
perament profiles and their role in neurodevelopmental Worobey, J., & Lewis, M. (1989). Individual differences in the assessed pre-term children at two years of age. European reactivity of young infants. Developmental Psychology, 25, Child and Adolescent Psychiatry, 7, 145-152.
Sameroff, A., Krafchuck, E., & Bakow, H. (1978). Issues in Youdim, M., Ben-Shachar, D., & Yehuda, S. (1989). Putative grouping items from the Neonatal Behavioral Assessment biological mechanisms of iron deficiency on brain biochem- Scale. Monographs of the Society for Research in Child istry and behavior. American Journal of Clinical Nutrition, Development, 43, 46-59. Issues 5-6, serial no. 177.

Source: http://www.grade.org.pe/download/pubs/Neonatal%20iron%20status%20and%20development.pdf