Effect of duration of breastfeeding on neuropsychological development at 10 to 12 years of age in a cohort of healthy children

Authors


Dr Francesco Gesualdo at Bambino Gesù Children’s Hospital, Epidemiology Unit, Piazza S. Onofrio, 4, 00165 Rome, Italy. E-mail: f.gesualdo@gmail.com

Abstract

Aim  The aim of this article was to explore the effect of duration of breastfeeding on neurocognitive development.

Method  The long-term effect of breastfeeding on neurodevelopment was examined through a battery of neuropsychological tests in 1403 children (693 females, 710 males; mean age 11y 9mo [SD 6mo], range: 10y 3mo–12y 8mo) who were originally recruited at 6 to 12 weeks of age for a clinical trial on acellular pertussis vaccines. An estimated IQ was obtained from scores of the vocabulary, similarities, block design, and coding tests. Breastfeeding data had been prospectively collected throughout the first year of life. Duration of exclusive breastfeeding was defined as the time during which children received breast milk without receiving any supplemental formula or food. Children were assessed at 10 to 12 years of age. We adjusted the analysis on test scores for multiple potential confounders.

Results  Multivariate analysis showed a significant association between exclusive breastfeeding duration and test scores in the vocabulary (odds ratio [OR] 0.05; confidence interval [CI] 0.00–0.10; p=0.04) and similarities (OR 0.06; CI 0.01–0.11; p=0.03) tests. These associations have a negligible effect size, however. Scores on one writing praxis test subcategory decreased with increasing duration of both exclusive breastfeeding (OR −0.06; CI −0.11 to −0.01; p=0.03) and breastfeeding irrespective of consumption of other foods (OR −0.06; CI −0.11 to −0.01; p=0.03). A negative association was also found between one subcategory of the California verbal learning test and breastfeeding duration longer than 6 months (OR −0.21; CI −0.42 to −0.01; p=0.04).

Interpretation  Breastfed healthy children may perform better on neuropsychological tests in the language domain at 10 to 12 years of age. However, the effect of breast milk on neuropsychological performance in healthy children may have a limited clinical relevance and is confounded by parental education.

What this paper adds

  •  There is a weak positive association between longer duration of exclusive breastfeeding and performance on neuropsychological language tests at age 10 to 12 years.
  •  Parental education may confound the effect of breastfeeding on neurodevelopment.

The World Health Organization1 recommends exclusive breastfeeding for the first 6 months of life in order to achieve optimal growth, development, and health; continuation of breastfeeding beyond this age is also strongly encouraged. Several studies have focused on the relationship between breastfeeding duration and cognitive development, with inconsistent results. A meta-analysis conducted by Anderson et al.2 in 1999 reported consistently higher IQs in breastfed than in formula-fed infants. Other studies published after this meta-analysis confirmed these results and conclusions.3

A recent large randomized study provided strong evidence that prolonged and exclusive breastfeeding improves children’s cognitive development.4

However, many authors claim that the apparent cognitive advantage given by prolonged breastfeeding is the effect of confounding factors, including maternal IQ.2,5 The cognitive advantage of breastfeeding seems particularly evident in preterm infants and those with a low birthweight.2 In 1992, Lucas et al.6 showed an 8.3-point advantage in IQ at 7 years 6 months to 8 years of age in preterm infants fed with mother’s breast milk versus standard formula, even after adjustment for mother’s education and social class.6 In a randomized controlled trial, the same authors showed an improvement, as measured by the Bayley Scales of Psychomotor Development, at 18 months in preterm infants who had been fed donor’s breast milk compared with those who had been fed term-formula.7 In the same group, no major advantage was seen in the Mental Development Index.7 Neither Motor Development Index scores nor the Psychomotor Development Index scores improved in preterm infants who had been fed donor’s breast milk compared with preterm infants fed with enriched formula.7

Another study provided evidence of slight developmental advantages in very-low-birthweight children who had received breast milk compared with children who had never received breast milk, although many of the observed effects were confounded by indicators of social advantage.8

On the other hand, in a study conducted by Der et al.,5 breastfeeding was found to have little or no effect on intelligence in children born at term and with adequate birthweight. This study shows that a 4-point increase in cognitive ability scores in breastfed versus non-breastfed infants disappears when adjustment is made for mother’s IQ.5 A recent study conducted in India, in a setting where prolonged breastfeeding is the rule, showed no beneficial effect of longer breastfeeding on later cognitive ability.9

Most studies on the effects of breastfeeding focused on cognitive development using test batteries such as the Wechsler Intelligence Scale for Children (WISC-R),4,6 the Kaufman Assessment Battery for Children,8 the Bayley Scales of Infant Development,7 and the McCarthy Scales of Children’s Abilities General Cognitive Index.3

The available evidence therefore suggests that the positive effect of continued breastfeeding on intelligence in healthy term-born infants is controversial, and neurodevelopmental outcomes other than IQ have rarely been investigated.

A large cohort of children previously participating in a clinical trial on the efficacy of pertussis vaccines was enrolled in a study that aimed to detect the effect on neuropsychological development at 10 to 12 years of age of thimerosal, a mercury compound used as a preservative in vaccines. The results of this study have been published elsewhere.10 Children were selected by criteria set for the original clinical trial, which excluded infants with a history of major diseases in the neonatal period and those whose weight was below the third centile at the time of first immunization.11 In this cohort, duration of breastfeeding and information on potential confounding variables were precisely collected. The present study aims to investigate the effect of breastfeeding duration on neuropsychological development in this cohort of healthy children as selected at 2 months of age and evaluated at 10 to 12 years of age.

Method

Participants

Children originally included in an Italian clinical trial on acellular pertussis vaccines were recruited at vaccination clinics at 6 to 12 weeks of age in 1992 to 1993.11 Inclusion criteria included the following: (1) weight above the third centile for age; (2) no history of seizures or other central nervous system diseases; (3) no major congenital abnormalities, failure to thrive, or renal failure; (4) no known or suspected immune deficiency; and (5) no previous illness compatible with pertussis. Other inclusion criteria were an Italian-speaking mother, accessibility to the family by telephone, and family planning to remain in place for at least 12 months after enrolment.11 During the trial, families were followed up with monthly telephone calls, during which clinical data were prospectively collected. The study on vaccines continued until 1998.12 In 2003, we randomly selected a sample from 3399 individuals enrolled in the original trial and resident in one of the four regions where the trial was conducted, with the aim of comparing the neuropsychological outcomes of children exposed to different amounts of thimerosal.10 The same cohort served to assess the effect of breastfeeding on neuropsychological development. Children were aged 10 to 12 years when the neuropsychological assessment was conducted. Parents’ written informed consent was obtained before the assessment. The study was approved by the ethics review boards of the Istituto Superiore di Sanità (Italian National Health Institute) and the Veneto Region.

Definitions and collection of data on breastfeeding

Data on breastfeeding duration were prospectively collected throughout the first year of life during the follow-up in the original clinical trial through monthly telephone interviews with mothers until 1998.11 Duration of exclusive breastfeeding was defined as the time during which children received breast milk without receiving any supplemental formula or food. Total breastfeeding duration was also measured, irrespective of supplementation with other foods.

Abstraction of potential confounding variables

Personal and clinical information collected during the trial was used for the analysis. The following information was abstracted and used in the analysis as potential confounders: sex, birthweight, gestational age, mother’s age at birth, type of delivery, family composition, parents’ education, presence of central nervous system diseases or other chronic diseases, current prescription of antihistamines or antiepileptic drugs, and the amount of thimerosal to which children were exposed through vaccines. Medical conditions and relevant therapies were verified with the family paediatricians through review of local medical charts.

Neuropsychological assessment

Neuropsychological assessment was conducted between 2003 and 2005 during school hours by 10 psychologists who had previously been extensively trained in test administration and scoring.13

The following cognitive domains were assessed: memory and learning, using the California Verbal Learning Test–Children’s Version14 and the digit span test;13 attention, using the continuous performance test – a computer-assisted test that evaluates sustained attention, inhibition, and impulsivity;15 executive functions, using the coding test of the WISC-R16 and the phonemic verbal fluency test, which evaluates the ability to access the lexicon through a phonemic cue by setting up an adequate verbal search strategy;13 visuospatial functions, using the WISC-R block design test;16 language, using the WISC-R vocabulary and similarities tests,16 the Boston naming test,17 and the semantic verbal fluency test, which evaluates the ability to access the lexicon according to specific semantic categories;13 and motor skills, using the finger tapping test, which provides an assessment of fine motor speed,18 and the writing praxis test, which evaluates visuomotor coordination and writing speed.19

The entire test battery required nearly 2 hours to complete. An estimation of IQ was obtained from the scores of the four WISC-R subtests, i.e. vocabulary, similarities, block design and coding, as described by Sattler and Dumont.20

Statistical analysis

We explored the effect of exclusive breastfeeding duration, both as a continuous variable and as a categorical variable (≤6mo, >6mo) on the estimated IQ and neuropsychological outcomes. The scores of neuropsychological tests were treated as continuous variables.

We performed separate simple linear regression analyses and ordinary least-squares multivariate linear regression analyses, using the raw score of each test as the dependent variable, to assess the association between breastfeeding and the test results. The multivariate analyses adjusted for sex, birthweight, gestational age, mother’s age at birth, type of delivery, family composition, parents’ education, presence of central nervous system diseases or other chronic diseases, current prescription of antihistamines or antiepileptic drugs, and the amount of thimerosal to which children were exposed through vaccines. Standardized regression coefficients were used as a measure of effect size, with the standardized regression coefficients representing the quantitative difference in the outcome scores, expressed in SD units. We did not correct for multiple comparisons.

In all phases of the study, until the end of the final analysis, psychologists, coordinators and statisticians were blinded to the duration of breastfeeding.

Results

We invited 1979 children, randomly selected out of the 3399 originally recruited in one of the regions participating in the original trial since 1992/1993, to participate. A total of 1403 (70.9%; 693 females; 710 males; mean age 11y 9mo [SD 6mo], range: 10y 3mo–12y 8mo) completed the neuropsychological evaluation and were included in the analysis. Of the 576 children who were excluded, 160 (8.1%) were not traceable, 415 (21.0%) declined to participate or did not undergo the neuropsychological assessment, and one (0.05%) died. No significant differences were found with regard to birthweight, gestational age, and breastfeeding duration between the participants and the children who did not undergo the neuropsychological evaluation. However, the two groups of children differed significantly with respect to mother’s education (p=0.009). Indeed, although the proportion of mothers with a degree was similar between participants and excluded children (9.1% vs 9.9% respectively), there was a difference in the proportion who held a high school diploma (42.9% for participants vs 36.5% for those excluded).

The characteristics of the 1403 children included in the analysis are reported in Table I.

Table I.   Description of the study population
CharacteristicValue
  1. aMean (SD). bInformation available on 1391 children. cn (%). dMedian (range). eInformation available on 1400 children. CNS, central nervous system.

Age at neuropsychological assessment (y)11.76 (0.47)a
Females/Males693/710
Birthweight (g)3349 (483)a
Gestational age (wks)36.92 (0.41)a
Caesarean sectionb307 (22.1)c
Birth order1 (1–7)d
Number of siblings2 (1–8)d
Single parent119 (8.5)c
Age of mother (y)e30.15 (4.47)a
Mothers with degree or higher qualification128 (9.1)c
Fathers with degree or higher qualification141 (10.1)c
History of CNS disease20 (1.4)c
History of other severe diseases63 (4.5)c
History of hospitalization253 (18)c
Current antihistaminic prescription27 (1.9)c
Current antiepileptic prescription4 (0.3)c

The population enrolled in the study shared the characteristics of the general population (as per the inclusion criteria): 66 (4.7%) children were born preterm (<37 gestational weeks) and 53 (3.8%) had a birthweight of less than 2.5kg.

Table II shows absolute and percentage frequencies of children stratified by sex and breast- or bottle-feeding.

Table II.   Absolute (percentage) frequencies of children stratified by sex and breast- or bottle-feeding
SexInformation not availableFeedingTotal
Exclusively bottle-fedBreast- and bottle-fedExclusively breastfed for ≤6moExclusively breastfed for >6mo
  1. Pearson χ2(4)=1.9822; p=0.739.

Male10 (1.4)155 (21.8)329 (46.3)185 (26.1)31 (4.4)710
Female6 (0.9)167 (24.1)318 (45.9)175 (25.2)27 (3.9)693
Total16 (1.1)322 (23.0)647 (46.1)360 (25.7)58 (4.1)1403

Table III illustrates the effect of exclusive breastfeeding duration on single scores of the neuropsychological battery and on the estimated IQ, estimated by multiple linear regression.

Table III.   Duration of exclusive breastfeeding (continuous) and neuropsychological performance (multiple linear regression)
DomainTestSubcategoryCoefficient95% CI p
LowerUpper
  1. The analysis adjusted for the following confounders: sex, birthweight, gestational age, mother’s age at birth, type of delivery, family composition, parents’ education, presence of central nervous system diseases or other chronic diseases, current prescription of antihistamines or antiepileptic drugs, and amount of thimerosal to which children were exposed through vaccines. CI, confidence interval; RT, reaction time; ISI, interstimulus interval.

Memory and learningCalifornia verbal learningTotal five trials0.00−0.050.060.91
Tuesday list0.01−0.040.060.74
Short-term free recall0.00−0.050.060.87
Short-term cued recall−0.01−0.060.050.80
Long-term free recall0.03−0.050.060.82
Long-term cued recall0.01−0.040.060.76
Total recognition0.03−0.030.080.30
Digit spanForward0.00−0.050.060.93
Backward0.02−0.030.070.47
Total0.02−0.040.070.52
AttentionContinuous performanceOmissions−0.04−0.090.010.14
Commissions−0.01−0.060.050.80
Hit RT−0.01−0.060.050.83
Variability−0.03−0.080.030.32
Detectability0.00−0.060.050.90
Response style0.00−0.060.050.97
Perseverations0.00−0.050.060.95
Hit RT block change0.04−0.010.100.12
Hit RT ISI change0.00−0.050.060.87
Executive functionsCoding−0.02−0.060.030.53
Verbal fluencyPhonemic0.03−0.020.080.20
Visuospatial functionsBlock design0.02−0.030.070.35
LanguageVocabulary0.050.000.100.04
Similarities0.060.010.110.03
Boston naming0.050.000.100.06
Verbal fluencySemantic−0.01−0.060.040.73
Motor skillsFinger tappingDominant hand0.00−0.060.050.86
Non-dominant hand0.03−0.020.080.31
Writing praxisNumbers−0.02−0.070.030.48
‘Uno’0.00−0.050.050.88
‘Le’−0.06−0.11−0.010.03
IntelligenceEstimated IQ0.04−0.010.100.09

According to the results of the analysis, scores of only a few neuropsychological tests were affected by exclusive breastfeeding duration. In particular, mean scores on vocabulary, similarities, the Boston naming test, and estimated IQ improved with the duration of breastfeeding, whereas performance in one of the subtests for writing praxis decreased with breastfeeding duration. When adjusting for potential confounding variables, the multiple linear regression analysis still yielded a significant association between exclusive breastfeeding duration and vocabulary, similarities, and marginally the Boston naming test in the language domain; the negative association with the writing praxis test was confirmed.

We also tested potential differences in neuropsychological performance between those who were breastfed for 6 months or less and those who were breastfed for more than 6 months. In the multiple linear regression analysis, the score on one subcategory of the California verbal learning test was negatively associated with breastfeeding for longer than 6 months. No difference was detected in any of the other test scores included in the analysis (see Table III).

When studying the effect of total duration of breastfeeding irrespective of the consumption of other food on neuropsychological performance at the univariate level, none of the scores were associated with duration of breastfeeding except the ‘Le’ subtest of the writing praxis test, aimed at evaluating fine visuomotor coordination, which was negatively correlated with breastfeeding duration (see Table III). These results were confirmed in the multiple linear regression analysis.

Regarding the effect of potential confounding variables, when considering continuous duration of exclusive breastfeeding, the multiple linear regression analysis revealed that other variables explaining a lower estimated IQ were older siblings in the family, male sex, and maternal and paternal education. Importantly, many neuropsychological test scores increased with increasing levels of parental education. Some test scores were also positively associated with birthweight. In the multiple linear regression models, gestational age did not affect any test score (data not shown).

Discussion

This large study allowed focusing on the long-term effects of breastfeeding on neuropsychological outcomes and estimated IQ in a population of healthy children, selected at birth and tested at the age of 10 to 12 years. We did not find significant or clinically relevant associations between duration of either exclusive or non-exclusive breastfeeding and neuropsychological outcomes. Although the battery of tests we performed did not allow for the calculation of a full-scale IQ score, the estimated IQ was not significantly associated with breastfeeding duration.

We included breastfeeding duration as a continuous variable in our analysis to look for even small associations with neuropsychological outcomes per time unit, assuming a linear relationship between breastfeeding duration and neuropsychological performance. We also specifically investigated the effect on neuropsychological performance of breastfeeding duration longer than 6 months. We finally tested the effect of non-exclusive breastfeeding as a continuous variable on the neuropsychological outcomes. The only statistically significant associations were weak.

We found minor effects of breastfeeding on test scores in the linear regression analysis. However, most of these effects disappeared when adjusting for confounding variables in the multiple linear regression models, with the majority of outcomes being associated with differences in maternal education. Evidence suggests that maternal IQ is a major confounding variable when studying the effect of breastfeeding on neurocognitive outcomes.5 We did not measure maternal IQ, but we used other variables as a proxy, including maternal education. The choice of breastfeeding and breastfeeding duration are both positively affected by maternal education,21 and recent studies have confirmed the association of maternal education with neurocognitive outcomes.22 However, we cannot exclude a residual confounding effect of maternal IQ on our results, which would possibly further reduce the association between breastfeeding and the test scores.

We found a statistically significant dose–response effect of breastfeeding duration on the verbal WISC-R subtests (vocabulary and similarities), which remains after correction for confounding variables and confirms previous evidence.4 This result, however, should be treated with caution, and the effect may not have a relevant impact on language development.

The negative association between breastfeeding duration and a subtest of the writing praxis test, which evaluates fine visuomotor coordination, is not easily explained and should be further investigated. However, as we did not adjust for multiple comparisons in the analysis, we cannot exclude the fact that significant associations may be merely the effect of chance.

In the original trial on pertussis vaccines, the cohort was selected to exclude low birthweight or unhealthy children. The major benefits of breastfeeding on IQ have been described in children born preterm or with very low birthweight, whereas healthy term-born infants seem to have few or no advantages.8,23

Long-chain polyunsaturated fatty acids involved in neurocognitive development (docosahexaenoic acid and arachidonic acid) are contained in human breast milk. Nevertheless, the amount of these bioactive compounds in milk may not affect children’s intelligence.24 However, other authors have shown an association between breastfeeding and IQ in individuals who carry a genetic polymorphism in the FASD2 gene, which is involved in fatty acid metabolism.25

Other biological factors associated with breastfeeding, such as parental characteristics (i.e. education and socioeconomic status), and the psychological interaction of the breastfed child with the mother may play a central role in cognitive development.

As very few studies have investigated the long-term neuropsychological benefits of breastfeeding, some advantages of the present study should be underscored. The sample size was large; the framework of a clinical trial and the frequent contacts with families allowed the investigators to collect precisely clinical and sociodemographic information throughout the study.

Trial nurses were responsible for collecting several pieces of information, including breastfeeding duration, and were regularly in touch with mothers through monthly telephone calls.

Another strength of this study is that we took mother’s education and other environmental factors into account as potential confounding variables and adjusted the outcome measures accordingly. Moreover, the setting in which the study was conducted allowed for the evaluation of neuropsychological development in a population exposed to a standard school curriculum. Finally, children were evaluated at 10 to 12 years of age, at a later age than in many of the published studies, and the proportion of children lost to follow-up was reasonably low.

Our study has several limitations. Of the original sample, a total of 576 children were excluded from the analysis. One of the variables we measured (mother’s education) differed significantly between the samples. It is possible that this difference affected our results and that there was a bias towards the selection of children with more highly educated parents, who are therefore more likely to perform well in neuropsychological tests. Although this could limit the generalization of our results, we do not feel that the difference was large enough to create a major bias.

The large number of statistical tests performed may have led to potential false-positive results. In order to avoid this effect, we could have applied corrections for multiple comparisons by lowering the threshold of the statistical significance test. However, this might have reduced the probability of type I errors, yet increased the probability of type II errors, possibly causing us to miss truly important differences. We, therefore, decided to adopt a more cautious attitude in order to identify all potentially significant associations, but, even so, apart from a few weak associations, we actually did not find any relevant effects of breastfeeding on neuropsychological development in healthy children.

We did not perform a complete IQ assessment but rather a wide neuropsychological assessment, using domain-specific tests. The study, in fact, was designed with the aim of measuring the association between breastfeeding and neuropsychological development.

Recall bias can be considered irrelevant because data on breastfeeding duration were collected during the follow-up to the original clinical trial through monthly telephone interviews with mothers throughout the first year of life, i.e. at the same time of breastfeeding.

In conclusion, breastfeeding has largely demonstrated advantages for child health and development, including neuropsychological benefits, mainly in preterm infants; however, according to the results of our study, the effect of duration of breastfeeding on neuropsychological performance in healthy children may be of limited clinical relevance and is confounded by parental education.

Acknowledgements

The study was supported in part by the US Centers for Disease Control and Prevention, through contract 2002-N-00448 with the Istituto Superiore di Sanità.

Ancillary