Abstract
Objective
To examine the relation between birth weight and measured intelligence at age 7 years in children within the normal range of birth weight and in siblings.
Design
Cohort study of siblings of the same sex.
Setting
12 cities in the United States.
Subjects
3484 children of 1683 mothers in a birth cohort study during the years 1959 through 1966. The sample was restricted to children born at ⩾37 weeks gestation and with birth weights of 1500-3999 g.
Main outcome measure
Full scale IQ at age 7 years.
Results
Mean IQ increased monotonically with birth weight in both sexes across the range of birth weight in a linear regression analysis of one randomly selected sibling per family (n= 1683) with adjustment for maternal age, race, education, socioeconomic status, and birth order. Within same sex sibling pairs, differences in birth weight were directly associated with differences in IQ in boys (812 pairs, predicted IQ difference per 100 g change in birth weight =0.50, 95% confidence interval 0.28 to 0.71) but not girls (871 pairs, 0.10, −0.09 to 0.30). The effect in boys remained after differences in birth order, maternal smoking, and head circumference were adjusted for and in an analysis restricted to children with birth weight ⩾ 2500 g.
Conclusion
The increase in childhood IQ with birth weight continues well into the normal birth weight range. For boys this relation holds within same sex sibships and therefore cannot be explained by confounding from family social environment.
What is already known on this topic
IQ at school age is linked to birth weight among low birthweight babies
Some evidence suggests the association might also apply to children of normal birth weight
What this study adds
IQ at age 7 years is linearly related to birth weight among children of normal birth weight
The relation was not due to confounding by maternal or socioeconomic factors
IQ is also associated with differences in birth weight between boy sibling pairs but not girls
Introduction
Many studies have shown that children born at low birth weight (<2500 g) have deficits in average intelligence test scores at school age.1 Within the low birth weight range, children who are smaller at birth have larger deficits than those closer to normal birth weight.2,3 The effect seems to be similar for both performance and verbal IQ.1
Recent studies, as well as some earlier reports, have suggested that the direct relation of birth weight to measured intelligence continues well into the normal range of birth weight.4–8 However, these studies have not fully controlled for potential confounders, the most important of which is family social environment. Moreover, two early studies raise questions about the relation. Record et al found an association between birth weight and verbal reasoning scores in a birth cohort overall, but not within sibships, suggesting that the association was confounded by family environment.7 A study of the Dutch famine showed that maternal starvation during pregnancy lowered birth weight but not IQ.9
Although the reported effects of variation within normal birth weight on IQ are modest and of no clinical importance for individual children, they could be important at a population level because of the large proportion of children born of normal weight. In addition, these effects could shed light on links between fetal growth and brain development. We therefore sought to confirm the association between individual birth weight and childhood IQ using data from the National Collaborative Perinatal Project.10 We also examined the relation of birth weight to IQ within families.
Methods
The National Collaborative Perinatal Project was established to study the relation between prenatal factors, labour and delivery, and child development through to age 7 years. About 40 000 women were enrolled, and follow up of about 58 000 pregnancies was initiated from 1959 to 1966 at 12 medical centres across the United States. The methods have been previously described.11 We used the archival public data set for this analysis.12
Sample
To examine the relation between variation in birth weight and IQ, we constructed a sibship sample from children meeting specific inclusion criteria. The criteria were singleton birth, birth order less than 5, birth weight 1500-3999 g, gestational age ⩾37 weeks, and living at age 7. These restrictions aimed to reduce the influence of factors other than fetal growth (for example, birth order, gestational age) on any association between birth weight and IQ. The sibship sample was further restricted to those families that could contribute at least two eligible siblings of the same sex. Families that could contribute sibships of both sexes (two boys and two girls) were dropped to preserve independence among sibships (36 children from nine families). We also excluded families in which any eligible sibling had a serious malformation of the cardiovascular or central nervous system.
The full sibship sample included 3484 children from 1683 families. Of these 1683 families, 1567 contributed two siblings, 114 contributed three siblings, and two contributed four siblings of the same sex. Compared with all births in the database, children in the sibship sample were from families with a slightly higher than average socioeconomic index (50 v 47 on a scale of 0 to 100 that reflects household income and education and occupation of the head of household13), were more likely to have white mothers (61% v 45%), and were more likely to be born to mothers under 20 years of age (27% v 23%).
We drew two samples from the full sibship sample. The one sib sample comprised one sibling, chosen at random, from each family (n=1683; 871 girls and 812 boys). The two sib sample comprised all sibling pairs from families contributing only two children, plus pairs chosen at random from larger sibships (n=3366; 1742 girls and 1624 boys).
Measurement of intelligence
Children enrolled in the National Collaborative Perinatal Project had periodic developmental and neuropsychological assessments from birth through to age 7 years. The intelligence tests administered at age 7 included four of five verbal and three of five performance tests from the Wechsler intelligence scale for children.14 We used the full scale IQ score that was constructed from the seven subtests administered, with equal weight given to verbal and performance tests.
Statistical analysis
We used linear regression to model the relation of birth weight to IQ in the one sib sample. Birth weight was treated first as a continuous variable and then as a categorical variable, with categories defined as 1500-2499 g, 2500-2999 g, 3000-3499 g, and 3500-3999 g. Because the distribution of birth weight differs between boys and girls, we fitted separate models for each sex. We included the following potential confounders as covariates: maternal race (white/non-white), family socioeconomic index (average score reported across pregnancies), maternal education (highest number of years maternal education reported across pregnancies), maternal age at birth, and birth order.
We used the two sib sample to assess the relation of birth weight to IQ within sibling pairs. We fitted a series of linear regression models, specifying IQ difference between siblings as a function of their difference in birth weight. (IQ and birthweight differences were computed after randomly assigning the order of the siblings in each pair.) Difference in birth weight was considered first as a continuous variable and then as a categorical variable (<−500 g; −500 to −200 g; −200 to 200 g; 200 to 500 g; and >500 g). To reduce potential confounding by pregnancy specific factors, we fitted separate models with adjustments for differences in birth order, maternal smoking, and head circumference. Difference in birth order was defined as a binary variable in two ways: younger versus older, and firstborn versus not. Difference in maternal smoking during pregnancy was defined as concordant (+/+, −/−) or discordant in two distinct categories (+/− and −/+). Finally, to ensure that any associations observed were not accounted for by the inclusion of low birthweight children, we repeated the analyses using only those sibling pairs in which both siblings had birth weights of ⩾2500 g.
Results
Tables 1 and 2 provide descriptive data for the sibship sample on all variables used in the analyses and their crude relations with birth weight and IQ. Birth weight and IQ relate to measures of social disadvantage in the expected direction. Lower birth weights and lower IQs were seen in mothers of younger age, lower education, non-white race, and lower socioeconomic index. Mean IQ increased in roughly even increments across the birthweight categories examined.
Table 1.
Categories | No (%) of children | Mean birth weight (g) | Mean full scale IQ |
---|---|---|---|
Sex: | |||
Male | 1690 (49) | 3233 | 97.4 |
Female | 1794 (51) | 3143 | 97.2 |
Birth order: | |||
1 | 907 (26) | 3144 | 97.7 |
2 | 1269 (36) | 3208 | 98.2 |
3 | 891 (26) | 3179 | 96.2 |
4 | 417 (12) | 3229 | 96.1 |
Maternal age: | |||
11-19 | 953 (27) | 3144 | 92.2 |
20-29 | 2234 (64) | 3204 | 99.0 |
⩾30 | 297 (9) | 3194 | 100.8 |
Maternal education: | |||
Below high school | 1715 (49) | 3135 | 92.3 |
High school graduate | 1367 (39) | 3228 | 99.7 |
Some after high school | 253 (7) | 3267 | 107.6 |
College graduate | 149 (4) | 3269 | 115.3 |
Race: | |||
White | 2108 (61) | 3236 | 102.4 |
All others* | 1376 (39) | 3111 | 89.5 |
Birth weight (g) | |||
<2500 | 187 (5) | 2303 | 91.9 |
2500-999 | 892 (26) | 2798 | 94.5 |
3000-499 | 1578 (45) | 3240 | 97.7 |
3500-999 | 827 (24) | 3705 | 100.7 |
98.2% are black; the remainder are Hispanic, Asian, other, or unknown.
Table 3 summarises the regression results from the analysis using the one sib sample. In both boys and girls, the association between IQ and birth weight is significant, although it is reduced by adjustment for familial factors. The association is stronger in boys than girls. For example, a 1000 g increase in birth weight relates to a 4.6 point increase in IQ among boys but only 2.8 points in girls. The associations remained significant when birth weight was treated as a categorical variable, even after important covariates were adjusted for (P=0.0012 in boys; P=0.007 in girls).
Table 3.
Birth weight variable (g) | Estimated coefficient
|
|
---|---|---|
Crude | Adjusted* (95% CI) | |
Boys | ||
Continuous† | 0.77 | 0.46 (0.25 to 0.66) |
Categorical‡: | ||
1500-2499 | −6.6 | −3.7 (−8.0 to 0.6) |
2500-2999 | −4.9 | −2.2 (−4.4 to −0.07) |
3000-3499 | 1.0 | 1.0 |
3500-3999 | 3.6 | 2.2 (0.2 to 4.2) |
Girls | ||
Continuous† | 0.63 | 0.28 (0.09 to 0.47) |
Categorical‡: | ||
1500-2499 | −5.7 | −2.8 (−5.8 to 0.3) |
2500-2999 | −3.6 | −2.1 (−4.0 to −0.3) |
3000-3499 | 1.0 | 1.0 |
3500-3999 | 4.2 | 1.5 (−0.6 to 3.5) |
Adjusted for race (white, non-white), mother's education, mother's age, family socioeconomic index, and birth order.
IQ difference per 100 g difference in birth weight.
Estimated average IQ difference compared with reference birth weight category.
Table 4 shows the results of the analysis on the two sib sample. Within sibling pairs of the same sex, IQ differences were directly related to differences in birth weight (heavier sibling having, on average, the higher IQ) when treated as a continuous variable. However, the slope of IQ difference by birthweight difference was about five times greater in boys than in girls, and the association was significant only in boys. To formally test whether sex modifies the association between birthweight difference and IQ difference, we fitted an interaction model using data from all 1683 sibling pairs. The interaction between sex and birthweight difference was significant (P=0.008).
Table 4.
Difference in birth weight (g) | Estimated coefficient (95% CI)* |
---|---|
Boys (812 pairs) | |
Continuous* | 0.50 (0.28 to 0.71) |
Categorical†: | |
−1900 to −500 | −6.2 (−9.3 to −3.1) |
−500 to −200 | −1.7 (−4.3 to 0.9) |
−200 to 200 | 1.0 |
200 to 500 | −1.3 (−3.8 to 1.2) |
500 to 1502 | 3.6 (0.5 to 6.7) |
Girls (871 pairs) | |
Continuous* | 0.10 (−0.09 to 0.30) |
Categorical†: | |
−1900 to −500 | 0.5 (−2.3 to 3.3) |
−500 to −200 | −0.9 (−3.2 to 1.5) |
−200 to 200 | 1.0 |
200 to 500 | −1.2 (−3.5 to 1.2) |
500 to 1502 | 2.4 (−0.2 to 5.1) |
IQ difference per birth weight difference of 100 g.
Estimated average IQ difference compared with reference category of birthweight difference.
Adjustment for differences in birth order and maternal smoking had no effect on the estimated effect of birth weight for either sex. Adjustment for differences in head circumference between siblings slightly reduced the effect of birth weight in boys (0.44, 95% confidence interval 0.17 to 0.70) and girls (−0.03, −0.28 to 0.21).
Table 4 also gives estimated differences in IQ by categories of differences in birth weight. These results re-enforce findings from the continuous analysis, showing a significant difference across categories of birthweight difference in boys (P<0.001) but not in girls (P=0.17).
All analyses in table 4 were repeated in the subsample of pairs where both siblings had birth weights ⩾2500 g. The findings were essentially identical to those based on the entire two sib sample.
Discussion
Our results add to the evidence that the association between IQ and birth weight continues well into the normal range of birth weight.4–7 We have shown that this association is not explained by confounding from unmeasured or poorly measured social or maternal factors. The relation within sibships of birth weight to IQ we observed in boys was tightly controlled for such factors.
Comparison with other studies
Our results for the one sib sample are similar to those in recent reports. Richards et al found an association between birth weight and cognitive ability at age 8 across the normal range of birth weight in the British 1946 birth cohort.4 Breslau et al also found birth weight was associated with IQ at age 6 in a sample of several hundred low and normal birthweight children and that the association extended into the range of normal birth weights.5
In the only other study of birth weight and intelligence in siblings who were mostly of normal birth weight, Record et al did not find the associations we report.7 Differences between their result and ours may be due to differences in the intelligence score used (we used full scale IQ and they used verbal reasoning scores from school examinations) or in the exclusion criteria (Record et al did not exclude preterm births or high birth order siblings).
The difference between boys and girls is puzzling and needs replication. A recent report noted sex differences in the relation of body measurements at birth to coronary artery disease in adult life. The authors postulated that sex differences in fetal growth rates (with growth generally being slower among girls) could lead to different responses to fetal undernutrition.15 Hormonally mediated sex differences in prenatal brain development could also have a role.16
The National Collaborative Perinatal Project is a thoroughly analysed data set, and our findings should also be compared with previous analyses. An unadjusted analysis found an association of IQ with birth weight well into the normal range, consistent with our adjusted one sib analysis.8 In an analysis of a smaller subsample of sibling pairs, however, 148 low birthweight children had slightly lower mean IQ than their same sex, normal birthweight siblings; the difference was not significant except in low birthweight children with small head circumference at birth.17 We found that differences in head circumference did not explain the association between birth weight and IQ. One explanation of our finding is that head circumference is not sensitive to any subtle influence of fetal growth restriction on brain development. In addition, compared with birth weight, head circumference varies less between individuals and is less precisely measured.
Importance of association
At least two studies suggest that the relation we observed between birth weight and IQ might persist into adult life. Richards et al found a gradient relation of birth weight to covariate adjusted cognitive test scores from childhood through age 26 in the 1946 birth cohort. Sorensen et al linked birth records to cognitive test results among 4300 Danish men being evaluated for military service and found a positive relation for birth weights up to about 4200 g.6 These results suggest an effect size of about a 0.2 standard deviation increase in mean adjusted cognitive test score per 1 kg increase in birth weight. This is similar to our findings. In each of these studies, however, control for potentially confounding influences, such as the home environment, relied on proxy measures such as social class and maternal education and may therefore have been incomplete.
From a clinical point of view, the effect in our study of even a 1 kg difference in birth weight is trivial. At the population level, however, more than seven times as many children are born with birth weights between 2500 and 3499 g than are born with low birth weight.18 Small shifts in the distribution of birth weight among normal birthweight babies might have greater impact on the population distribution of intelligence than larger shifts in the percentage of babies born at low birth weight. It may therefore be important to identify the sources of variation in birth weight that are both amenable to intervention and linked to improved cognitive ability.
From a scientific point of view, our findings may have important implications for future research on the connections between fetal growth and brain development. Considerable efforts continue to be directed at understanding causes of low birth weight, its neurodevelopmental consequences, and the mechanisms linking the two. Evidence from this and other studies suggests that these efforts should be broadened to examine the influences on fetal growth among those with normal birth weight and the neurodevelopmental implications of such variation.
Table 2.
Mean | Correlation with birth weight | Correlation with IQ | |
---|---|---|---|
Socioeconomic index* | 50 | 0.12 | 0.48 |
Birth weight (g) | 3187 | 1.00 | 0.17 |
Full scale IQ | 97.3 | 0.17 | 1.00 |
Socioeconomic index was missing for six children (three families).
Acknowledgments
We thank Marcus Richards and Sander Greenland for their comments.
Footnotes
Funding: New York Community Trust.
Competing interests: None declared.
References
- 1.Breslau N. Psychiatric sequelae of low birth weight. Epidemiol Rev. 1995;17:96–106. doi: 10.1093/oxfordjournals.epirev.a036191. [DOI] [PubMed] [Google Scholar]
- 2.McCormick MC, Brooks-Gunn J, Workman-Daniels K, Turner J, Peckam GJ. The health and developmental status of very low-birth-weight children at school age. JAMA. 1992;267:2204–2208. [PubMed] [Google Scholar]
- 3.Breslau N, DelDotto JE, Brown GG, Kumar S, Ezhuthachan S, Hufnagle KG, et al. A gradient relationship between low birth weight and IQ at age 6 years. Arch Pediatr Adolesc Med. 1994;148:377–383. doi: 10.1001/archpedi.1994.02170040043007. [DOI] [PubMed] [Google Scholar]
- 4.Richards M, Hardy R, Kuh D, Wadsworth MEJ. Birthweight and cognitive function in the British 1946 birth cohort: longitudinal population based study. BMJ. 2001;322:199–203. doi: 10.1136/bmj.322.7280.199. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Breslau N, Chilcoat H, DelDotto J, Andreski P, Brown G. Low birth weight and neurocognitive status at six years of age. Biol Psychiatry. 1996;40:389–397. doi: 10.1016/0006-3223(95)00399-1. [DOI] [PubMed] [Google Scholar]
- 6.Sorensen HT, Sabroe S, Olsen J, Rothman KJ, Gillman MW, Fisher P. Birth weight and cognitive function in young adult life: historical cohort study. BMJ. 1997;315:401–403. doi: 10.1136/bmj.315.7105.401. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Record RG, McKeown T, Edwards JH. The relation of measured intelligence to birth weight and duration of gestation. Ann Hum Genet Lond. 1969;33:71–79. doi: 10.1111/j.1469-1809.1969.tb01631.x. [DOI] [PubMed] [Google Scholar]
- 8.Hardy JB, Mellits ED. Relationship of low birth weight to maternal characteristics of age, parity, education , and body size. In: Reed DM, editor. The epidemiology of prematurity: epidemiology workshop, National Institute of Child Health and Human Development,1976. Baltimore: Urban and Schwarzenberg; 1977. [Google Scholar]
- 9.Stein ZA, Susser M, Saenger G, Marolla F. Famine and human development: the Dutch hunger winter of 1944/45. Oxford University Press: New York; 1975. [Google Scholar]
- 10.Niswander NR, Gordon M. The women and their pregnancies. Philadelphia: WB Saunders; 1972. [Google Scholar]
- 11.Broman S. The collaborative perinatal project: an overview. In: Mednick SA, Finello KM, editors. . Handbook of longitudinal research. Vol. 1. New York: Prager Publishers; 1984. [Google Scholar]
- 12.Collaborative Perinatal Project Electronic Records. Records of the National Institute of Neurological Disorders and Stroke 1950-88. College Park, MD: National Archives at College Park; 1958-84. (record group 443.8). [Google Scholar]
- 13.Myrianthopoulos NC, French KS. An application of the US Bureau of the Census socioeconomic index to a large diversified patient population. Soc Sci Med. 1968;2:283–299. doi: 10.1016/0037-7856(68)90004-8. [DOI] [PubMed] [Google Scholar]
- 14.Wechsler D. Wechsler intelligence scale for children—manual. New York: Psychological Corporation; 1949. [Google Scholar]
- 15.Forsen T, Ericksson JG, Osmond C, Barker DJP. Growth in utero and during childhood among women who develop coronary heart disease: longitudinal study. BMJ. 1999;319:1403–1407. doi: 10.1136/bmj.319.7222.1403. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Hutchison JB. Gender-specific steroid metabolism in neural differentiation. Cell Mol Neurobiol. 1997;17:603–626. doi: 10.1023/a:1022581902880. [DOI] [PubMed] [Google Scholar]
- 17.Strauss RS, Dietz WH. Growth and development of term children born with low birth weight: effects of genetic and environmental factors. J Pediatr. 1998;133:67–72. doi: 10.1016/s0022-3476(98)70180-5. [DOI] [PubMed] [Google Scholar]
- 18.Ventura SJ, Martin JA, Curtin SC, Matthews TJ. Hyattsville, MD: National Center for Health Statistics; 1999. Births: final data for 1997. . (National vital statistics reports vol 47, No 18.) [PubMed] [Google Scholar]