OUP user menu

Critical periods of brain growth and cognitive function in children

Catharine R. Gale, Finbar J. O’Callaghan, Keith M. Godfrey, Catherine M. Law, Christopher N. Martyn
DOI: http://dx.doi.org/10.1093/brain/awh034 321-329 First published online: 25 November 2003

Summary

There is evidence that IQ tends to be higher in those who were heavier at birth or who grew taller in childhood and adolescence. Although these findings imply that growth in both foetal and postnatal life influences cognitive performance, little is known about the relative importance of brain growth during different periods of development. We investigated the relationship between brain growth in different periods of pre‐ and postnatal life and cognitive function in 221 9‐year‐old children whose mothers had taken part in a study of nutrition in pregnancy and whose head circumference had been measured at 18 weeks gestation, birth and 9 months of age. Cognitive function of the children and their mothers was assessed with the Wechsler Abbreviated Scale of Intelligence. Full‐scale IQ at age 9 years rose by 1.98 points [95% confidence interval (CI) 0.34 to 3.62] for each SD increase in head circumference at 9 months and by 2.87 points (95% CI 1.05 to 4.69) for each SD increase in head circumference at 9 years of age, after adjustment for sex, number of older siblings, maternal IQ, age, education, social class, duration of breastfeeding and history of low mood in the post‐partum period. Postnatal head growth was significantly greater in children whose mothers were educated to degree level or of higher socio‐economic status. There was no relation between IQ and measurements of head size at 18 weeks gestation or at birth. These results suggest that brain growth during infancy and early childhood is more important than growth during foetal life in determining cognitive function.

  • children; cognitive function; foetal growth; head circumference; postnatal growth
  • SDS = standard deviation scores

Introduction

The development of the CNS in man begins in the embryo and continues for several years of postnatal life. Critical early events are the closure of the neural tube around day 22 of embryonic life and neurogenesis, which is complete by 16 weeks. In later gestation there are overlapping phases of neuronal migration, glial cell proliferation and dendritic sprouting. Myelination of the cerebral commisures and long tracts continues into late childhood. Synaptic density in both cerebral and cerebellar cortex increases until early adult life.

Although severe degrees of intrauterine growth retardation are associated with poorer cognitive performance in later life (Strauss, 2000; Hack et al., 2002), it is less clear whether variation in early growth within the normal range affects higher mental function. Some studies have found that IQ tends to be higher in those who were heavier at birth (Sorensen et al., 1997; Matte et al., 2001; Richards et al., 2001), which implies that brain growth during foetal life is important. On the other hand, the finding that height and head size are consistently related to IQ in adults suggests that postnatal growth is also influential (Cox et al., 1987; Andreasen et al., 1993; Rushton and Ankney, 1996; Tuvemo et al., 1999; Richards et al., 2002). Little is known about the relative importance of brain growth during different periods of pre‐ and postnatal development.

Head circumference is known to correlate closely with brain volume (Cooke et al., 1977; Wickett et al., 2000) and can therefore be used to measure brain growth, but a single measurement cannot provide a complete insight into neurological development. Different patterns of early brain growth may result in a similar head size. A child whose brain growth both pre‐ and postnatally followed the 50th centile might attain the same head size as a child whose brain growth was retarded in gestation but who later experienced a period of rapid growth. Different growth trajectories may reflect different experiences during sensitive periods of brain development and have different implications for later cognitive function.

We have investigated whether brain growth during different periods of pre‐ and postnatal development influences later cognitive function in a group of children for whom serial measurements of head growth through foetal life, infancy and childhood were available.

Subjects and methods

Participants

The mothers of the children in this study had all taken part in an earlier study of nutrition during pregnancy between April 1992 and June 1993. The study consisted of singleton children born to Caucasian women aged 16 years or older who registered under two obstetric consultants and who attended the midwives’ antenatal booking clinic at the Princess Anne Maternity Hospital in Southampton at <17 weeks gestation. Anthropometric information was available at three time points: 18 weeks gestation (via foetal ultrasound), birth and 9 months of age. There were routine obstetric data about the pregnancy and delivery, parental social class, and maternal education. At the 9‐month postnatal visit, information on infant feeding was collected and maternal psychological morbidity since the infant’s birth was assessed using the Edinburgh postnatal depression scale (Cox et al., 1987). A score of >12 was taken to indicate low mood in the post‐partum period. In total, 559 children were followed up to the age of 9 months.

When these 559 children approached their ninth birthday, we asked the Community Paediatric Service in Southampton to write to their parents with an invitation to take part in a further follow‐up study. All the children in the cohort had previously been flagged on the Service’s child health computer. Letters were sent to all 461 families who were still living in the Southampton area; 226 (49%) agreed to take part in the follow‐up study. In total, 221 families (48%) were willing to allow their child’s cognitive function to be measured. The children who took part in the 9‐year follow‐up did not differ significantly from the non‐responders or from those who had moved out of the area in terms of social class distribution (P = 0.142), mean birthweight (P = 0.10), or mean head circumference at 18 weeks gestation (P = 0.36), at birth (P = 0.17) or at 9 months of age (P = 0.56).

Measurements

At 18 weeks’ gestation, foetal head circumference was measured around the outer table of the skull using standard sonographic landmarks (Campbell and Thoms, 1977). Occipito‐frontal circumference was measured within 48 h of birth, at 9 months and at 9 years of age by passing a tape measure around the head, placing it on the most anterior protuberance of the forehead and the most posterior protuberance of the back of the head. The tape measure was pulled tight to compress the hair and measurements made to the nearest 0.1 cm. Weight was measured at birth, at 9 months and at 9 years using digital scales. Crown–heel length at birth and at 9 months was measured using a neonatal stadiometer. Height at 9 years of age was measured using a portable stadiometer. Inter‐observer tests of repeatability were conducted regularly during each phase of the study to ensure that any discrepancies in measurements made by the research nurses were very small.

As part of the 9‐year follow‐up study, the cognitive function of the child and his/her mother was assessed by a member of the research team who visited the family at home. Cognitive function was measured using the Wechsler Abbreviated Intelligence Scale (Wechsler, 1999) This provides age‐adjusted IQ scores for full‐scale, verbal and performance intelligence.

The protocol for this study was approved by the Southampton and South West Hants Joint Local Research Ethics Committee. The children and their mothers gave written informed consent.

Statistical analysis

Anthropometric measurements made at the 9‐month and 9‐year examinations were adjusted for age at the time of examination. We used the British 1990 growth reference data on weight, height and head circumference (Freeman et al., 1995; Cole et al., 1998), obtained from the Child Growth Foundation, in combination with our measurements and data on gestation and sex to estimate for each child what his or her weight, length and head circumference would have been at exactly 9 months and 9 years of age. Comparisons of means and proportions was performed using analysis of variance (ANOVA), t‐tests or χ2‐tests as appropriate. Linear regression was used to examine the relation between anthropometric measurements, expressed as standard deviation scores (SDS), and cognitive function, taking account of other factors. Analysis of the relationship between head circumference at 18 weeks gestation and cognitive function was restricted to the 133 children whose mothers had been confident of the date of their last menstrual period and were not taking the oral contraceptive pill immediately prior to conception. As girls tended to be lighter, shorter and to have a smaller head circumference than boys at each time of measurement, all linear regression models were adjusted for sex.

Results

Characteristics of the 221 study participants and their mothers recorded at 18 weeks’ gestation, at birth, at 9 months and 9 years of age are shown in Table 1.

View this table:
Table 1

Characteristics of the 221 study participants and their mothers recorded at around 18 weeks gestation, at birth, at 9 months and at 9 years of age

Characteristics 18 weeks gestation Birth 9 months 9 years
Child
Head circumference (cm) 15.3 (1.1) 35.0 (1.4) 45.6 (1.4) 53.3 (1.6)
Weight (kg) 3.3 (0.6) 9.0 (1.2) 29.5 (5.6)
Length/height (cm) 50.0 (2.4) 70.8 (3.0) 130.9 (6.1)
Premature [n (%)] 17 (7.7)
Female [n (%)] 105 (47.5)
No. of older siblings [n (%)]
0 120 (54.8)
1 71 (32.4)
2 or more 28 (12.8)
Mother
Social class [n (%)]
I/II 56 (25.6)
III 126 (57.5)
IV/V 37 (16.9)
Age (years) 27.0 (4.8)
Education
None or CSE 48 (21.7)
O level 82 (37.1)
A level 60 (27.1)
Degree or HND 31 (14.0)
Low mood post‐partum [n (%)] 89 (40.3)
IQ 103.9 (13.2)
Breastfeeding duration
Never 64 (30.3)
<1 month 48 (22.7)
1–4 months 45 (21.3)
>4 months 54 (25.6)

Values are means (SD) unless stated otherwise.

At age 9 years, the children’s mean full‐scale IQ was 106.6 (SD 14.4). Mean performance IQ was 104.7 (SD 13.8) and mean verbal IQ was 106.9 (SD 14.4). On average, boys gained higher scores than girls for full‐scale IQ (108.7 compared with 104.2; P = 0.022) and for performance IQ (107.4 compared with 101.7; P = 0.005), although there was no statistically significant difference between the sexes in verbal IQ. As might be expected, the children’s IQ test performance was strongly associated with maternal characteristics. In univariate analysis, full‐scale IQ, performance IQ and verbal IQ were significantly greater in those whose mothers were older, of higher social class, more educated, had a higher IQ or had continued breastfeeding for >1 month. Full‐scale and performance IQ tended to be lower in those whose mothers had scored >12 on the Edinburgh postnatal depression scale at the 9‐month visit, suggesting low mood in the post‐partum period. Verbal IQ tended to be lower in children who had one or more older siblings.

We found no statistically significant associations between head circumference at 18 weeks’ gestation or head circumference at birth SDS and IQ at the age of 9 years. Table 2 shows the relationships between these two measures of foetal head growth and full‐scale IQ, verbal IQ and performance IQ, adjusted for sex and gestation, then with further adjustment for number of older siblings, maternal age, social class, education, IQ, duration of breastfeeding and history of low mood post‐partum. When we repeated our analysis excluding the 17 children who had been born before 37 weeks’ completed gestation, the relationships between head circumference at birth SDS and IQ changed little. We examined whether birthweight or length at birth were associated with IQ at age 9 years, but these relationships were not statistically significant. After adjustment for sex and gestational age at birth, full‐scale IQ fell by 0.27 points [95% confidence interval (CI) –2.89 to 2.36] for each SD increase in birthweight and rose by 0.51 points (95% CI –1.92 to 2.95) for each SD increase in length. The relationships between birthweight or length at birth and verbal or performance IQ were very similar (data not shown).

View this table:
Table 2

Results of multivariate linear regression analyses into the relation between measures of head growth during foetal life and full‐scale, verbal and performance IQ at age 9 years (each measure of head growth has been analysed separately)

Regression coefficient (95% CI), adjusted for sex and gestation* P Regression coefficient (95% CI), further adjusted for other factors P
Full‐scale IQ
Head circumference at 18 weeks gestation SDS§ 0.93 (–3.09 to 4.94) 0.649 1.05 (–2.84 to 4.93) 0.594
Head circumference at birth SDS 0.81 (–1.60 to 3.23) 0.507 0.16 (–2.00 to 2.33) 0.883
Verbal IQ
Head circumference at 18 weeks gestation SDS§ 1.98 (–2.10 to 6.07) 0.339 1.27 (–2.83 to 5.36) 0.541
Head circumference at birth SDS 1.72 (–0.75 to 4.19) 0.171 1.20 (–1.09 to 3.49) 0.303
Performance IQ
Head circumference at 18 weeks gestation SDS§ –0.17 (–4.48 to 4.14) 0.938 0.75 (–3.50 to 4.99) 0.727
Head circumference at birth SDS –0.35 (–2.87 to 2.18) 0.787 –1.02 (–3.45 to 1.41) 0.407

*Length of gestation at time of head measurement; maternal age, social class, education, IQ, duration of breastfeeding, history of low mood in post‐partum period and number of older siblings; §based on 133 children whose mothers were confident of the date of their last menstrual period and were not taking oral contraceptives immediately prior to conception.

In contrast, there were strong statistically significant associations between measures of postnatal head growth and IQ. After adjustment for sex, full‐scale IQ rose by 2.59 points (95% CI 0.87 to 4.32) for each SD increase in head circumference at 9 months of age, and by 3.85 points (95% CI 1.96 to 5.73) points for each SD increase in head circumference at 9 years; verbal IQ rose by 2.66 points (95% CI 0.49 to 4.83) for each SD increase in head circumference at 9 months of age, and by 3.76 points (95% CI 1.81 to 5.72) for each SD increase in head circumference at 9 years; performance IQ rose by 2.88 points (95% CI 0.659 to 5.11) for each SD increase in head circumference at 9 months of age, and by 3.16 points (95% CI 1.16 to 5.16) for each SD increase in head circumference at 9 years. We found no significant associations between IQ and weight or length at 9 months or weight or height at 9 years.

We carried out multivariate analyses examining these measures of postnatal head growth separately and including in the models all the factors that had been significantly associated with IQ in univariate analysis. In these analyses, head circumference SDS either at 9 months (model 1) or at 9 years (model 2) remained a statistically significant predictor of full‐scale, verbal and performance IQ (Tables 35 ). After adjustment for sex and other factors, full‐scale IQ rose by 1.98 points (95% CI 0.34 to 3.62) for each SD increase in head circumference at 9 months and by 2.87 points (95% CI 1.05 to 4.69) for each SD increase in head circumference at 9 years. Maternal IQ, maternal education and duration of breastfeeding were the only other factors that remained significantly associated with full‐scale IQ in multivariate analysis (Table 3). In multivariate analyses of factors associated with verbal IQ, verbal IQ rose by 1.92 points (95% CI 0.19 to 3.64) for each SD increase in head circumference at 9 months and by 2.82 points (95% CI 0.90 to 4.74) for each SD increase in head circumference at 9 years. Maternal IQ and education remained significant predictors of the child’s verbal IQ, but there were no associations between IQ score and maternal age, social class, duration of breastfeeding, history of low mood post‐partum or number of older siblings (Table 4). In multivariate analyses of factors associated with performance IQ, performance IQ rose by 1.85 points (95% CI 0.09 to 3.61) for each SD increase in head circumference at 9 months and by 2.39 points (95% CI 0.35 to 4.43) for each SD increase in head circumference at 9 years. Duration of breastfeeding and maternal education were the only other factors that remained significantly associated with performance IQ in multivariate analysis (Table 5).

View this table:
Table 3

Results of multivariate linear regression analyses into the relation between measures of head growth during postnatal life and full‐scale IQ at age 9 years; model 1 includes head circumference at 9 months and model 2 includes head circumference at 9 years

Model 1: regression coefficient (95% CI), adjusted for sex and other variables in the table P Model 2: regression coefficient (95% CI), adjusted for sex and other variables in the table P
Head circumference at 9 months SDS 1.98 (0.34 to 3.62) 0.018
Head circumference at 9 years SDS 2.87 (1.05 to 4.69) 0.002
Maternal IQ (per point) 0.24 (0.06 to 0.40) 0.007 0.24 (0.07 to 0.40) 0.006
Maternal age (per year) –0.18 (–0.61 to 0.25) 0.419 –0.21 (–0.63 to 0.22) 0.342
Maternal education*
O level 4.04 (–0.97 to 9.05) 0.114 4.15 (–0.81 to 9.12) 0.100
A level 6.12 (0.70 to 11.65) 0.027 6.51 (1.08 to 11.93) 0.019
Degree or HND 11.73 (3.80 to 19.65) 0.004 11.18 (3.31 to 19.05) 0.006
Maternal social class*
III –0.36 (–5.20 to 4.46) 0.882 0.09 (–4.70 to 4.89) 0.969
IV/V 0.12 (–6.01 to 6.33) 0.970 0.48 (–5.69 to 6.63) 0.879
Duration of breastfeeding*
Up to 1 month 0.001 (–4.85 to 4.85) 0.999 0.07 (–4.74 to 4.89) 0.976
1 to 4 months 5.12 (0.01 to 10.24) 0.049 4.82 (–0.24 to 9.88) 0.062
Over 4 months 6.66 (1.42 to 11.89) 0.013 6.73 (1.57 to 11.90) 0.011
Low mood post‐partum –2.59 (–6.20 to 1.03) 0.160 –2.58 (–6.15 to 0.99) 0.157
No. of older siblings*
1 –1.49 (–5.54 to 2.56) 0.470 –1.02 (–5.04 to 2.30) 0.617
2 or more –1.30 (–7.22 to 4.62) 0.665 –0.72 (–6.51 to 5.07) 0.805

*Comparison groups for categorical variables are as follows. maternal education: no qualifications or CSE; maternal social class: classes I/II; duration of breastfeeding: never breastfed; number of older siblings: none.

View this table:
Table 4

Results of multivariate linear regression analyses into the relation between measures of head growth during postnatal life and verbal IQ at age 9 years; model 1 includes head circumference at 9 months and model 2 includes head circumference at 9 years

Model 1: regression coefficient (95% CI), adjusted for sex and other variables in the table P Model 2: regression coefficient (95% CI), adjusted for sex and other variables in the table P
Head circumference at 9 months SDS 1.92 (0.19 to 3.64) 0.030
Head circumference at 9 years SDS 2.82 (0.90 to 4.74) 0.004
Maternal IQ (per point) 0.25 (0.07 to 0.43) 0.006 0.25 (0.08 to 0.43) 0.005
Maternal age (per year) –0.06 (–0.51 to 0.39) 0.790 –0.09 (–0.54 to 0.36) 0.694
Maternal education*
O level 4.54 (–0.74 to 9.82) 0.091 4.66 (–0.56 to 9.89) 0.080
A level 6.68 (0.92 to 12.44) 0.023 7.01 (1.29 to 12.72) 0.017
Degree or HND 11.91 (3.57 to 20.26) 0.005 11.37 (3.08 to 19.66) 0.007
Maternal social class*
III –0.13 (–5.22 to 4.95) 0.959 0.32 (–4.74 to 5.37) 0.901
IV/V 2.05 (–4.48 to 8.60) 0.536 2.42 (–4.08 to 8.91) 0.464
Duration of breastfeeding*
Up to 1 month 1.16 (–3.95 to 6.27) 0.656 1.23 (–3.84 to 6.29) 0.633
1–4 months 3.09 (–2.29 to 8.47) 0.259 2.79 (–2.54 to 8.13) 0.304
Over 4 months 4.83 (–0.68 to 10.33) 0.085 4.91 (–0.53 to 10.35) 0.077
Low mood post‐partum –1.59 (–5.40 to 2.21) 0.410 –1.58 (–5.34 to 2.19) 0.410
No. of older siblings*
1 –3.34 (–7.60 to 0.92) 0.124 –2.89 (–7.12 to 1.34) 0.180
2 or more –4.58 (–10.81 to 1.65) 0.149 –4.03 (–10.13 to 2.07) 0.194

*Comparison groups for categorical variables are as follows: maternal education: no qualifications or CSE; maternal social class: classes I/II; duration of breastfeeding: never breastfed; number of older siblings: none.

View this table:
Table 5

Results of multivariate linear regression analyses into the relation between measures of head growth during postnatal life and performance IQ at age 9 years; model 1 includes head circumference at 9 months and model 2 includes head circumference at 9 years

Model 1: regression coefficient (95% CI), adjusted for sex and other variables in the table P Model 2: regression coefficient (95% CI), adjusted for sex and other variables in the table P
Head circumference at 9 months SDS 1.85 (0.09 to 3.61) 0.040
Head circumference at 9 years SDS 2.39 (0.35 to 4.43) 0.022
Maternal IQ (per point) 0.17 (–0.01 to 0.35) 0.063 0.16 (–0.24 to 0.35) 0.088
Maternal age (per year) –0.26 (–0.73 to 0.20) 0.271 –0.30 (–0.78 to 0.17) 0.210
Maternal education*
O level 3.01 (–2.47 to 0.86) 0.275 2.90 (–2.66 to 8.45) 0.305
A level 3.91 (–2.14 to 9.96) 0.204 4.70 (–1.37 to 10.78) 0.129
Degree or HND 9.05 (0.36 to 17.73) 0.041 8.67 (–0.15 to 17.48) 0.054
Maternal social class*
III –0.56 (–5.89 to 4.77) 0.836 0.06 (–5.31 to 5.43) 0.982
IV/V –1.08 (–7.80 to 5.65) 0.753 –1.12 (–8.03 to 5.77) 0.748
Duration of breastfeeding*
Up to 1 month –0.57 (–5.90 to 4.76) 0.834 –0.60 (–5.98 to 4.78) 0.827
1–4 months 6.90 (1.30 to 12.51) 0.016 6.58 (0.91 to 12.25) 0.023
Over 4 months 7.60 (1.95 to 13.26) 0.009 7.57 (1.78 to 13.36) 0.011
Low mood post‐partum –3.30 (–7.33 to 0.73) 0.108 –3.30 (–7.30 to 0.70) 0.105
No of older siblings*
1 1.87 (–2.47 to 6.22) 0.396 1.75 (–2.75 to 6.25) 0.443
2 or more 3.16 (–3.31 to 9.63) 0.336 3.46 (–3.03 to 9.94) 0.294

*Comparison groups for categorical variables are as follows: maternal education: no qualifications or CSE; maternal social class: classes I/II; duration of breastfeeding: never breastfed; number of older siblings: none.

Children who had a larger head at 9 months of age tended to have a larger head at 9 years (r = 0.77, P < 0.001). In order to examine whether the relation between postnatal head size and intelligence differed according to period of growth, we derived standardized residuals from a linear regression first of head circumference at 9 months on head circumference at birth and secondly of head circumference at 9 years on head circumference at 9 months. These residuals provided measures of head growth during the periods between birth and 9 months and between 9 months and 9 years that took account of head size at the start of each period. The residuals were also independent of each other (r = –0.03, P = 0.66) and so could be used simultaneously in linear regression (Table 6). Head growth during both periods was associated with full‐scale IQ, after adjustment for sex and other factors. Full‐scale IQ rose by 2.30 points (95% CI 0.56 to 4.03) for each SD increase in head growth between birth and 9 months, and by 2.12 points (95% CI 0.39 to 3.86) for each SD increase in head growth between 9 months and 9 years. Thus the highest full‐scale IQs were seen in children who had experienced a large increase in head circumference between birth and 9 months of age and a further large increase in head circumference between the ages of 9 months and 9 years. There was some evidence that the two periods of head growth differed in importance as regards verbal and performance IQ. Verbal IQ was associated with head growth between 9 months and 9 years of age, while performance IQ was more strongly associated with head growth in the first 9 months of life.

View this table:
Table 6

Results of simultaneous multivariate linear regression analyses into the relation between head growth during two periods of postnatal life and full‐scale, verbal and performance IQ at age 9 years

Full‐scale IQ Verbal IQ Performance IQ
Regression coefficient (95% CI), adjusted for sex and other factors* P Regression coefficient (95% CI), adjusted for sex and other factors* P Regression coefficient (95% CI), adjusted for sex and other factors* P
Head growth between birth and 9 months 2.30 (0.56 to 4.03) 0.010 1.72 (–0.16 to 3.60) 0.072 2.49 (0.57 to 4.40) 0.011
Head growth between 9 months and 9 years§ 2.12 (0.39 to 3.86) 0.017 2.08 (0.21 to 3.95) 0.030 1.76 (–0.15 to 3.67) 0.070

*Maternal age, social class, education, IQ, duration of breastfeeding, history of low mood in post‐partum period and number of older siblings. Expressed as standardized residuals of regression of head size at 9 months on head size at birth. § Expressed as standardized residuals of regression of head size at 9 years on head size at 9 months.

Finally, we examined whether factors in the child’s home environment might influence IQ test performance through an effect on head growth. Table 7 shows the relationships between the extent of head growth during these two periods of postnatal life, as measured by the standardized residuals described above, and maternal and family characteristics. We found no statistically significant associations between head growth in either period and maternal IQ, age, duration of breastfeeding, or history of low mood post‐partum. There were also no significant associations between head growth and number of older siblings. However, children whose mothers were educated to degree level experienced a significantly greater increase in head circumference both between birth and 9 months, and between 9 months and 9 years, than those whose mothers were less educated or who had no qualifications. There was also evidence to link social class and head growth. Between birth and 9 months of age, head growth tended to be poorer in children whose mothers came from social classes III, IV or V than in those whose mothers came from social classes I or II, although this trend was not statistically significant. But during the period from 9 months to 9 years of age, there was a significant difference in head growth between the classes. Children whose mothers came from social classes III, IV or V experienced markedly poorer head growth than children whose mothers came from social classes I or II.

View this table:
Table 7

Results of separate linear regression analyses into the relation between maternal and family characteristics and head growth during two periods of postnatal life

Head growth between birth and 9 months* Head growth between 9 months and 9 years
Regression coefficient (95% CI) P Regression coefficient (95% CI) P
Maternal IQ (per point) 0.01 (–0.002 to 0.02) 0.136 0.01 (–0.002 to 0.08) 0.138
Maternal age (per year) 0.01 (–0.02 to 0.04) 0.474 0.02 (–0.01 to 0.05) 0.534
Maternal education§
O level 0.12 (–0.25 to 0.47) 0.556 0.03 (–0.32 to 0.38) 0.870
A level 0.29 (–0.10 to 0.68) 0.140 0.01 (–0.36 to 0.39) 0.943
Degree or HND 0.45 (0.002 to 0.91) 0.049 0.56 (0.11 to 1.02) 0.014
Maternal social class§
III –0.23 (–0.55 to 0.09) 0.151 –0.45 (–0.76 to –0.14) 0.005
IV/V –0.35 (–0.78 to 0.08) 0.107 –0.46 (–0.87 to –0.06) 0.026
Duration of breastfeeding§
Up to 1 month 0.03 (–0.36 to 0.41) 0.900 0.15 (–0.23 to 0.52) 0.444
1 to 4 months 0.01 (–0.30 to 0.50) 0.623 0.44 (0.06 to 0.82) 0.025
Over 4 months –0.07 (–0.44 to 0.29) 0.699 0.23 (–0.13 to 0.60) 0.207
Low mood post‐partum 0.01 (–0.27 to 0.28) 0.964 0.09 (–0.18 to 0.35) 0.534
No of older siblings§
1 –0.27 (–0.57 to 0.03) 0.076 –0.22 (–0.51 to 0.07) 0.139
2 or more 0.14 (–0.29 to 0.56) 0.520 –0.37 (–0.78 to 0.04) 0.079

*Expressed as standardized residuals of regression of head size at 9 months on head size at birth. Expressed as standardized residuals of regression of head size at 9 years on head size at 9 months. §Comparison groups for categorical variables are as follows: maternal education: no qualifications or CSE; maternal social class: classes I/II; duration of breastfeeding: never breastfed; number of older siblings: none.

Discussion

In this study of 9‐year‐old children, performance on a test of cognitive function was related to current head size and head size at 9 months of age but not to head size at birth or at 18 weeks gestation, after adjustment for sex, number of older siblings, duration of breastfeeding, and maternal age, IQ, education, social class and history of low mood in the post‐partum period. As head circumference is a close correlate of brain volume (Cooke et al., 1977; Wickett et al., 2000), we interpret these findings as evidence that postnatal brain growth is more important than prenatal brain growth in determining higher mental function. This interpretation is supported by the finding that head growth in the first 9 months of life and head growth between 9 months and 9 years of age are also related to cognitive function, regardless of head size at the beginning of these periods. Postnatal head growth was significantly greater in children whose mothers were educated to degree level or of higher socio‐economic status.

In common with many other studies, we found that children who had been breastfed for longer periods gained higher scores on the test of cognitive function (Anderson et al., 1999; Jain et al., 2002). Interpretation of such findings has been complicated by the failure of most studies to control adequately for social and environmental confounding factors, such as maternal intelligence and quality of parenting (Anderson et al., 1999; Jacobson and Jacobson, 2002; Jain et al, 2002). The association found in our study between duration of breastfeeding and IQ at age 9 years persisted after adjustment for mother’s intelligence, age, education, social class and history of low mood in the post‐partum period, but we had no data on quality of parenting in early life. Several previous studies have shown lower IQs in children of mothers who were depressed postnatally (Cogill et al., 1986; Sharp et al., 1995; Hay et al., 2001). In our study, maternal low mood in the first few months of life was associated with lower IQ in univariate analysis, but this relation was no longer statistically significant after adjustment for other risk factors.

One of the limitations of this study was our inability to follow up all children in the original cohort. This was partly because some had moved away from the area in which the original study took place and it was not possible to trace them, and partly because some declined to participate. Mean head size at birth and at 9 months of age, however, was similar in the groups who did and who did not take part in the current phase of the study, and we think that it is unlikely that non‐response or our inability to follow up children who had moved away will have introduced bias. Another limitation is that data on head circumference at 18 weeks gestation was only available for a proportion of the children in the study—those whose mothers could be sure of the gestational age of their baby from the dates of their last menstrual period. Incompleteness of these data will have reduced the statistical power of the study and our ability to detect associations between early head growth and later cognitive function. A further weakness is that we had no information on the nature and quality of the child’s home environment, so were unable to take account of parenting style or level of cognitive stimulation, both of which are known to affect intellectual development (Guo and Harris, 2000).

The strengths of the study are the longitudinal measures of head size and therefore the ability to estimate the influence of brain growth during different periods of pre‐ and postnatal development. A further advantage was that we had information on other potentially confounding factors that are known to influence intelligence, including the mother’s IQ, educational level and social class, whether she had suffered from low mood in the post‐partum period, duration of breastfeeding, and the child’s birth order.

A number of other investigators have examined associations between foetal or childhood head growth and subsequent cognitive function. Results of studies into the influence of head size at birth have been conflicting. In a follow‐up study of over 14 000 children born at term whose head circumference at birth had been above the 10th percentile for gestational age, there were no differences in IQ at age 4 or 7 years between those whose head circumference had been small relative to their birthweight and the rest of the population (Brennan et al., 1985). But in a study of around 248 000 young men, risk of poor performance on a test of cognitive function was significantly higher in those whose head circumference at birth had been more than 2 SDs below the mean (Lundgren et al., 2001). Although a statistically significant association was found in this very large study, the size of the effect exerted by head size at birth was small; the difference in mean cognitive function test scores between those whose head circumference at birth had been more than 2 SDs below the mean and those whose head circumference had been more than 2 SDs above the mean was only 0.5 of a point.

Evidence that postnatal head growth may influence cognitive function has been more consistent. In a cohort of 249 children who weighed <1.5 kg at birth, those whose head circumference was >2 SD below the mean at 8 months of age had poorer cognitive function at the age of 8 years than those whose head size was normal (Hack et al., 1991). Among 365 children aged 7 years, larger head circumference was associated with higher scores on measures of verbal ability and practical reasoning (Ounsted et al., 1988). A similar relationship was found between head circumference and IQ in 334 boys aged 8 to 9 years (Weinberg et al., 1974). Only one previous study has examined how head growth at various stages during postnatal development relates to cognitive performance. In a cohort of 2023 children followed up from birth, those who had an IQ in the superior range (≥120) at the age of 7 years had a larger head circumference at age 1 year than children whose IQs were average (80–119) or low (≤79), and this difference in head size persisted at 4 and 7 years of age (Fisch et al., 1976).

Findings in our study that the extent of postnatal head growth was greater in the children of women who were educated to degree level or who were of higher social class suggest that part of the influence of maternal education and socio‐economic status on a child’s intellectual development may be due to their effect on brain growth. It may be that the extent and quality of cognitive stimulation and the style of parenting provided by such mothers help promote brain growth, and with it, intellectual development. Several studies have demonstrated that provision of a cognitively enriched environment in early life can lead to improvements in intellectual performance (Anderson et al., 2003; Eickmann et al., 2003; Hill et al., 2003), but whether this is partly mediated by increases in head growth is unclear.

In contrast to some recent studies, we found no significant association between cognitive function and birthweight in these 9‐year‐old children. Lower birthweight has been linked to poorer cognitive test performance in childhood and adolescence in the 1946 and 1958 British birth cohorts (Richards et al., 2001; Jefferis et al., 2002) in 4300 Danish conscripts (Sorensen et al., 1997), in 3484 children aged 7 years enrolled in the US National Collaborative Perinatal Project (Matte et al., 2001) and in 449 Scottish 11 year olds (Shenkin et al., 2001). The effect size in all these studies was small; in the US study, for example, a 1 kg increase in birthweight was associated with an increase in IQ of 4.6 points in boys and 2.8 points in girls (Matte et al., 2001). The lack of such an association in our data may be due to the smaller size of our study.

Findings in our study linking full‐scale, performance and verbal IQ in 9‐year‐old children to measurements of head circumference made currently and at 9 months of age, but not those made at 18 weeks gestation or at birth, suggest that postnatal brain growth is more important than foetal growth in determining cognitive function. Intelligence in children of this age tends to remain stable into adulthood, so the fact that head growth in the first 9 months of life and between 9 months and 9 years were both associated with IQ, regardless of head size at the start of each period, suggests that maximizing growth during both infancy and childhood is critical for the attainment of peak cognitive capacity in adult life. Work is now needed to investigate the mechanisms whereby maternal education and socio‐economic status influence postnatal head growth and to explore which other factors in the child’s environment help to determine eventual brain size.

Acknowledgements

We wish to thank the children and their families for their help with this study, the research nurses, Tracey Tudball, Anne Abel and Glynis Bousfield, for collecting data for the 9‐year follow‐up, Chris Newsome for helping to organize the study, Susan Knight for advice on IQ testing and Vanessa Cox and Patsy Coakley for computing assistance. The study was funded by the Medical Research Council, the Dunhill Medical Trust and Children Nationwide.

References

View Abstract