The effect of birthweight on childhood cognitive development in a middle-income country

7/29/2019 The effect of birthweight on childhood cognitive development in a middle-income country

http://slidepdf.com/reader/full/the-effect-of-birthweight-on-childhood-cognitive-development-in-a-middle-income 1/11

The effect of birthweight on childhoodcognitive development in a middle-income

countryFlorencia Torche1* and Ghislaine Echevarrıa2

1Department of Sociology, New York University, USA and 2Faculty of Medicine, Universidad Catolica de Chile

*Corresponding author. Department of Sociology, New York University, 295 Lafayette street #4129, New York, NY 10012, USA.

Email: [email protected]

Accepted 24 January 2011

Background Intra-uterine growth is a powerful predictor of infant mortalityand of health, developmental and socioeconomic outcomes in adult-hood. The question about whether this relationship is causal ratherthan driven by unobserved characteristics of low-weight infants is,however, still open. We use twin models to examine the hypothesisthat in utero growth has a detrimental impact on cognitive develop-ment in childhood.

Methods We merge birth registry information on birthweight with standar-dized Math and Spanish test scores for all fourth graders in Chile tocreate a prospective data set. Twin fixed-effects models are used toestimate the causal effect of intra-uterine growth on test scores.Fixed-effect estimates are compared with traditional regression re-sults in a cross-section of births to gauge the omitted variable biasemerging from unobserved genetic, maternal and pregnancy-relatedfactors in cross-sectional models.

Results Birthweight differences within twin pairs have a substantial ef-fect on test scores. A 400-g increase in birthweight results in a15% standard deviation increase in Math scores. The effect islarger among (estimated) monozygotic than dizygotic pairs, reach-ing420% standard deviation. The effect varies across family socio-economic status. It is strong among disadvantaged families but itnearly disappears among advantaged ones.

Conclusion Scarcity of uterine resources resulting in intra-uterine growth

restriction has a detrimental effect on cognitive development inchildhood. This effect interacts with family socioeconomic status(SES), so that low-SES families reinforce the effect of low birthweight and high-SES families fully compensate for it. Findings areparticularly relevant in the developing world, where intra-uterinegrowth restriction is the main determinant of low birthweight.

Keywords Birthweight, twin studies, zygosity, fixed-effects, cognitive devel-opment, developing countries

Published by Oxford University Press on behalf of the International Epidemiological Association

ß The Author 2011; all rights reserved.

International Journal of Epidemiology 2011;1–11

doi:10.1093/ije/dyr030

1

Int. J. Epidemiol. Advance Access published March 8, 2011



IntroductionBirthweight is a powerful marker of early disadvan-tage. It is the main predictor of neonatal and infantmortality1 and is related to health, development and

well-being in childhood and adulthood.2–11 To date,however, most research studying the influence of

birthweight on later outcomes examines differencesin a cross-section of births using regression models.Even if these models control for observed characteris-tics such as family socioeconomic status, maternalsmoking and pregnancy risk factors, they leave openthe possibility of unobserved factors driving the influ-ence of birthweight. If infants born low-weight differin terms of genetic endowments, maternal behavioursor any other factor unobserved by the researcher andconsequential for later outcomes, failure to control forthem will result in a spurious association. For ex-ample, the same genetic traits that determine lowbirthweight may affect individual educational andeconomic attainment, so that a policy increasing

birthweight would have little effect on later achieve-ments. To the extent that spuriousness cannot beruled out, the policy implications of the findings areunclear.

A powerful strategy to address the unobserved se-lectivity of low birthweight is to examine variation in

weight between twins through fixed-effects models.This approach relates within twin-pair differences inbirthweight with differences in the outcome of inter-est to identify a causal influence. Because twins sharethe uterine environment, the model accounts for con-founders related to the mother and the pregnancy.Because twins have, by definition, the same gesta-tional age, the observed effect of birthweight emerges

entirely from intra-uterine growth, rather than fromlength of gestation. Recent studies use twins to exam-ine the effect of in utero growth on perinatal andinfant mortality,12,13 early infancy development14

and adult outcomes such as education, body massindex (BMI), height and wages.15–18 Exploiting atwin fixed-effects methodology, this article addressestwo questions: does intra-uterine growth have aneffect on children’s cognitive development (measuredby test scores)? And, does this effect vary acrossfamily socioeconomic status?

The key assumption in the fixed-effect approach isthat weight differences within twin pairs emerge fromrandom differences in access to nutritional intakes

resulting, e.g. from position in the uterus or umbilicalcord insertion in the placenta. Research supports thisassumption, showing that twins grow at the samerate as singletons up to the 30th week of gestation.19

After that, the positive relationship between gesta-tional age and fetal size cannot be maintained dueto the restricted capability of the uterine environmentto nurture two fetuses in a uniparious species, leadingto weight differences between twins.20,21

Although all twins face scarcity in the womb, themagnitude and etiology of intra-uterine growth

differences depend on zygoticity. Dizygotic twins(DZ) result from the separate fertilization of twoova and share, on average, 50% of their genes, thesame as regular siblings. Monozygotic twins (MZ)are formed from the division of a single ovum afterfertilization. Even though they are not geneticallyidentical,22–24 monozygotic twins share most of their

genes. Zygosity affects also the access to uterine re-sources. DZ virtually always have separate placentas

with no obvious vascular connections.23 In contrast,most monozygotic twins share a single placenta,

which places them at heightened competition foruterine resources.23,25–27 Although access to nutrientsmay differ between DZ because of differences in themass and function of the placenta or placental le-sions, these differences are usually moderate.28–30 Incontrast, among monozygotic twins sharing a pla-centa, severe differences in nutritional intake mayemerge from differential umbilical cord insertionand uneven placental sharing.23,31 Weight discrepan-cies are thus larger and the lighter twin is likely to begenuinely growth restricted among monozygoticpairs.30,32,33 Because of these biological features,monozygotic twinning controls not only for mother-and pregnancy-related confounders but also for gen-etic variation, and offers a setting characterized byrandom—and acute—environmental determinants of intra-uterine growth.

To date, studies using twin models have focused ona handful of industrialized countries. Researchershave questioned the generalizability to other nationalcontexts, but empirical evidence is lacking. In add-ition, research on the influence of birthweight hasfocused on early outcomes such as infant mortality

and morbidity or on outcomes in adulthood suchas completed education and wages. Less evidenceexists on intermediate outcomes such as cognitive de-

velopment in childhood, which likely connect earlydevelopments with adult attainment. Furthermore,researchers usually compare fixed-effect with cross-sectional regression estimates to gauge unobserved se-lectivity bias in cross-sectional models. But becausetwins share the same gestational age, meaningfulcomparison requires controlling for gestational agein cross-sectional samples, so that estimates capturesolely the effect of intra-uterine growth in both cases.

This study contributes to the literature in three re-spects. Firstly, we use twin fixed-effect models to

examine the effect of intra-uterine growth on Mathand Spanish test scores of Chilean fourth graders in2008. Fixed-effect models explicitly account for unob-served genetic, maternal and pregnancy-related char-acteristics related to birthweight. Test scores provide aproxy for early cognitive development and predictlater educational attainment and labour market out-comes,34,35 providing an intermediate variable linkingshort- and long-term outcomes of birthweight.Secondly, a middle-income country with higheconomic inequality, Chile offers a useful setting to

2 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY



test the generalizability of the effect of birthweight.Thirdly by using an adjusted measure of intra-uterinegrowth orthogonal to gestational age, we provide validcomparisons between fixed-effects and cross-sectionalestimates of the effect of in utero growth.

MethodsData

We construct a longitudinal data set by merging stan-dardized test scores with birth registry information onbirthweight for all fourth graders in 2008. Test scores

were obtained from the system of measurement of educational quality (Sistema de Medicion de laCalidad de la Educacion, or SIMCE in Spanish) dataset. SIMCE is a standardized pencil-and-paper testadministered to all fourth graders in the country inNovember 2008 by the Chilean ministry of education.The test includes multiple choice and short answers,and is given to students in their classrooms. Scoresfrom Math and Spanish were used for this analysis.Birthweights were obtained from the Chilean birthregistry database established by the Chilean Ministryof Health, which records all births in the country. Thedatabase is constructed on the basis of live-birth cer-tificates, a form completed by the professional (MD orcertified midwife) who attends the delivery (99.8% of deliveries are attended by a professional in Chile), andmeasured with little error.36 The birth certificate con-tains information on birthweight, gestational age, ma-ternal residence and demographic characteristics. Therestricted-access version of both data sources containsan individual identifying number assigned to allChilean citizens at time of birth. We used identifyingnumbers to merge these sources. Given that the nor-mative age to enter fourth grade is 9 years old inChile, most of our merged sample was born in 1998(59.3%) and 1999 (35.0%). A small percentage wereborn in 1997 (5.8%) and 2000 (0.01%). The resultingmerged data set has 220 062 valid observations. Fromthis data set, we identified 2474 twins with informa-tion on all relevant variables. The use of administra-tive data has important advantages. Firstly, it providesa census of fourth graders in 2008 with no sampleerror. Secondly, it avoids the potentially severe recallbias and measurement error affecting self-reportedbirthweight information. Zygosity information is not

available in the data set, so we estimate the numberof monozygotic twin pairs as explained in detail in thenext subsection.

Variables

Our measure of intra-uterine growth is the stan-dardized value of gestational age-specific birthweight,obtained as the difference between individual birth-

weight and gestational week-specific mean birthweight, and dividing by the gestational week-specificstandard deviation of birthweight. This measure is

similar to the standard definition of ‘small for gesta-tional age’—birthweight less than the 10th percentilefor gestational week37—but it captures variationthroughout the entire distribution rather than estab-lishing a single threshold. Given sex differences infetal growth, standardized measures were separatelyobtained for male and female fetuses. Our standar-

dized measure is orthogonal to length of gestation(with a correlation of À0.007 in our sample). Thestandard deviation of birthweight is $400 g through-out the distribution of gestational age, so a 1-unitincrease in the standardized measure of intra-uterinegrowth involves a gain in birthweight of $400 g. Thedependent variables are Math and Spanish test scores.These variables have, by construction, a mean of 250and a standard deviation of 50. All models include thefollowing controls: maternal age (<19 years of age,20–29 years, 30–34 years, 535 years), education (pri-mary, secondary, post-secondary), urban residence(urban¼1), employment status (mother works atthe time of delivery ¼ 1) and infant’s sex (male¼1).

Analysis

We obtain twin fixed-effect estimates of the effect of birthweight on test scores, and compare them withregression estimates of the cross-sectional sample of both, singleton and twin births. The fixed-effectsmodel is formulated as follows:

tijk ¼ 0 þ bwijk1 þ sijk2 þ X 0

jk3 þ A jk þ "ijk

where t identifies the dependent variable, test scores,of child i born to mother j in birth k. bw refers tobirthweight, s identifies the sex of the child, X jk isa vector of mother- and pregnancy-specific character-

istics such as maternal age, rural residence and edu-cation, A is a vector of mother- and birth-specificunobserved factors (for example, genetic factors orpregnancy behaviours) and " is an idiosyncraticerror term. Cross-sectional estimates may be biasedbecause no controls are included for factors in A,if these factors are correlated with both birthweightand test scores. One strategy to control for factorsincluded in A is to differentiate the data withintwin pairs. For a pair formed by twins 1 and 2, weobtain the differenced equation:

t2jk À t1jk ¼ ð02 À 01Þ þ 1ðbw2jk À bw1jkÞ

þ 2ð s2jk À s1jkÞ þ ð"2jk À "1jkÞ

Because the unobserved mother- and pregnancy-specific factors included in A do not vary withintwin pairs, the term A drops from the differencedequation. Given that we have information on child’ssex, and under the assumption that within-twin dif-ferences in "ijk are independent from within-twindifferences in bwijk, 1 is consistent. This assumptionis more likely to hold among monozygotic thandizygotic because the former share most of theirgenes. We estimate models for the entire twin

THE EFFECT OF BIRTHWEIGHT ON CHILDHOOD COGNITIVE DEVELOPMENT 3



sample, for the same-sex and mixed-sex subsamples,and for the group estimated to be monozygotic. Giventhat monozygotic twins share most their genes andthat they face stronger competition for uterine re-sources than their DZ counterparts, we expect differ-ences in birthweight within monozygotic pairs to be amuch better proxy for individual twin levels of nutri-

tion than differences within DZ pairs.Lacking information about zygosity, we use a

method proposed by Weinberg38 to obtain an estimateof the number of monozygotic pairs. This method isbased on the fact that mixed-sex twins are necessarilyDZ, whereas same-sex twins may be either DZ ormonozygotic. Under the assumption that the probabil-ity of a male birth equals the probability of a femalebirth and that the sexes in a dizygotic pair are inde-pendent, the rate of DZ twinning is twice the rate of mixed-sex births. The monozygotic twinning rate isthe difference between the rates of same-sex andmixed-sex twin pairs. The Weinberg method makessome questionable assumptions, including the rare oc-currence of monozygotic twins of unlike sex, the factthat when monozygotic twins are obviously discord-ant they are judged dizygotic, and the suggestion thatthe actual frequency of same-sex twins is higher thanexpected in DZ pairs.23,39 However, it produces ex-tremely accurate estimates.40 Naturally, it is not pos-sible to identify specific twin pairs by zygosity, but a

weighted average strategy can be used to estimatebirthweight effects by zygosity, as implemented byConley et al.13 In a final step, we examine variationin the effect of birthweight across family’s socioeco-nomic resources, using mother’s education as a proxy.

ResultsDescriptive analysis

Table 1 presents the descriptive statistics of all variables in the twin and singleton samples. Asexpected, birthweight, intra-uterine growth and ges-tational age in twins is significantly less than insingletons. Singletons perform better, on average, inMath and Spanish tests. Women giving birth to twinsare, on average, more socioeconomically advantaged,but the differences are small. The secondary sex ratiois lower among twins (47.3% vs 50.3% for singletons),

probably related to a well-documented excess of fe-males in spontaneous monozygotic twins.23

An important requirement of twin models is thatthere is sufficient within twin-pair variation in birth-

weight to identify its effect. Figure 1 displays the fre-quency distribution of differences among same-sexand mixed-sex twin pairs, and shows substantial vari-ation. The average difference within twin pairs is283 g, with about half of the pairs having differencesof 5250 g, and about 10% displaying differences of 5600g.

The effect of intra-uterine growth ontest scores

Table 2 examines the effect of intra-uterine growth onMath and Spanish test scores using linear regressionmodels on cross-sectional samples of singletons andtwins and fixed-effect models in the twin sample.Models include controls for sex, maternal age, educa-

tion, urban status and employment status at time of birth. All cross-sectional specifications adjust for theclustering of standard errors within twin pairs in thecalculation of 95% confidence intervals (95% CIs) and

P-values. Cross-sectional results for Math test scoresindicate that a one-unit increase in intra-uterinegrowth is associated with 2.2 (95% CI: 2.01–2.46)points increase in test scores among singletons, and2.6 (95% CI: 0.26–4.84) points among twins, net of sociodemographic covariates. This effect is substan-tially very small. Given that one unit of the independ-ent variable is equivalent to $400 g and that testscores have a standard deviation of 50 by con-struction, a 400-g increase would result in test scoregain of about 5% (2.5/50) of a standard deviation.Comparable figures for Spanish test scores are 1.2(95% CI: 0.94–1.38) and 2.6 (95% CI: 0.37–4.80), cor-responding to 2.4% and 5.2% of a standard deviation,respectively.

When twin fixed effects are used to control for un-observed characteristics of the child, the effect ismuch larger, reaching 15% standard deviation forMath and 11% for Spanish (7.41/50 and 5.31/50, re-spectively). The larger fixed-effects estimate has twoalternative interpretations. Firstly, it suggests that thescarcity of nutrients and oxygen resulting from uter-ine competition, rather than unobserved factors, hasa substantial effect on test scores. Secondly, intra-uterine growth may be negatively correlated with en-dowments determining test scores, or nutrient intakesmay be provided less to fetuses with higher endow-ments in the cross-section.15 To the best of our know-ledge, no empirical evidence suggests the latteralternative, therefore we interpret this finding as evi-dence for a detrimental effect of resource scarcityin utero.

Exploiting zygosity to examine effectof uterine environment

As a next step, we use sex differences within twinpairs to explore the factors driving the effect of

in utero growth. In our sample, 75% of twin pairsare of the same sex and the 25% remaining aremixed-sex. Because monozygotic twins are virtuallyalways of the same sex, they should be overrepre-sented in the same-sex subsample. If the effect of birthweight on test scores is driven by environmentalfactors in utero, rather than unobserved genetic vari-ation at the individual level, we expect the effect to begreater among same-sex twins.

This hypothesis is supported in Table 3, whichshows a larger effect of intra-uterine growth among




same-sex twins. A 400-g increase results in a gain of 8.7 (95% CI: 5.32–12.12) points in Math test scoresand 6.9 (95% CI: 3.33–10.53) points in Spanish,equivalent to 17% and 14% standard deviation,

respectively. The effect is much smaller amongmixed-sex twin pairs. Because mixed-sex pairs arenecessarily DZ and DZ are much less subject to com-petition for uterine resources than MZ pairs (but are

Table 1 Means (standard deviations) and percentage distributions by type of birth, and P-values for means/proportionscomparison between twins and singletons, Chile 1997–99

Twins Singletons P-value differencemeans/proportions

Birthweight 2500.21 (487.62) 3373.38 (499.42) <0.001

Weeks of gestation 36.29 (2.09) 38.92 (1.47) <0.001

Intra-uterine growth À0.677 (0.883) 0.028 (0.987) <0.001

Small for gestational agea 22.89% (0.42) 5.30% (0.22) <0.001

Math test scores 242.79 (56.2) 248.6 (54.8) <0.001

Spanish test scores 257.1 (53.3) 261.7 (53.4) <0.001

Age at test (years) 10.07 (0.433) 10.12 (0.466) <0.001

Mother’s age (years), (%) <0.001

<19 (1) 8.2 16.1

19–29 (2) 46.1 49.9

30–34 (3) 26.0 20.4

35–39 (4) 16.4 11.0

40 or more (5) 3.3 2.7

Mother’s education (%) <0.001

Primary (1) 24.5 26.5

Secondary (2) 55.3 58.6

Tertiary (3) 20.2 14.9

Urban residence 88.9 87.6 0.018

Mother works 32.4 26.8 <0.001

Male 47.3 50.3 0.001

n 2474 217 588

aBelow the 10th percentile of gestational age-specific birthweight.

Figure 1 Absolute value of birthweight difference among same-sex and mixed-sex twins. Twins born in Chile in 1997–99and who took the fourth grade standardized test in 2008




T a b l e 2 E f f e c t o f i n t r a - u t e r i n e g r o

w t h o n M a t h a n d S p a n i s h t e s t s c o r e s , O L S m o d e l i n s i n g l e t o n a n d t w i n d a t a b a s e s a n d t w i n f i x e d - e f f e c t s m o d e l

M a t h t e s t s c o r e s

S p a n i s h t e s t s c o r e s

S i n g l e t o n s O L

S

T w i n s O L S

T w i n s F E

S i n g l e t o n s O L S

T w i n s O L S

T w i n s F E

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

I n t r a - u t e r i n e g r o w t h

2 . 2 3

2 . 0 1 t o 2 . 4 6

< 0 . 0 0 1

2 . 5 5

0 . 2 6 t o 4 . 8 4

0 . 0 2 9

7 . 4 1

4 . 3 3 t o 1 0 . 4 8

< 0 . 0 0 1

1 . 1 6

0 . 9 4 t o 1 . 3

8

< 0 . 0 0 1

2 . 5 8

0 . 3 7 t o 4 . 8 0

0 . 0 2 2

5 . 3 1

2 . 1 2 t o 8 . 5 0

0 . 0 0 1

M o t h e r ’ s s e c o n d . e d . a

1 6 . 4 8

1 5 . 9 3 t o 1 7 . 0 2

< 0 . 0 0 1

1 7 . 9 6

1 2 . 4 8 t o 2 3 . 4 3

< 0 . 0 0 1

1 3 . 9 9

1 3 . 4 6 t o 1 4 . 5 2

< 0 . 0 0 1

1 7 . 2 0

1 1 . 9 0 t o 2 2 . 4 9

< 0 . 0 0 1

M o t h e r ’ s t e r t i a r y e d . a

4 2 . 4 6

4 1 . 6 6 t o 4 3 . 2 6

< 0 . 0 0 1

5 3 . 6 3

4 6 . 5 4 t o 6 0 . 7 2

< 0 . 0 0 1

3 5 . 2 4

3 4 . 4 5 t o 3 6 . 0 2

< 0 . 0 0 1

4 2 . 7 8

3 5 . 9 3 t o 4 9 . 6 4

< 0 . 0 0 1

M o t h e r ’ s a g e 2 0 – 2 9 b

0 . 9 6

0 . 3 2 t o 1 . 6 2

0 . 0 0 3

0 . 6 9

À 7 . 2 6 t o 8 . 6 3

0 . 8 6 6

1 . 1 0

0 . 4 6 t o 1 . 7

5

0 . 0 0 1

0 . 8 3

À 6 . 8 3 , 8 . 5 2

0 . 8 3 2

M o t h e r ’ s a g e 3 0 – 3 4 b

3 . 5 8

2 . 8 2 t o 4 . 3 5

< 0 . 0 0 1

3 . 8 4

À 4 . 6 1 t o 1 2 . 2 9

0 . 3 7 2

3 . 6 8

2 . 9 3 t o 4 . 4

3

< 0 . 0 0 1

5 . 7 7

À 2 . 4 1 t o 1 3 . 9 4

0 . 1 6 7

M o t h e r ’ s a g e 3 5 – 3 9 b

3 . 9 9

3 . 1 1 t o 4 . 8 8

< 0 . 0 0 1

8 . 2 2

À 0 . 7 0 t o 1 7 . 1 3

0 . 0 7 1

4 . 7 9

3 . 9 2 t o 5 . 6

5

< 0 . 0 0 1

6 . 6 8

À 1 . 9 4 t o 1 5 . 3 0

0 . 1 2 9

M o t h e r ’ s a g e 5 4 0 b

4 . 0 9

2 . 6 2 t o 5 . 5 6

< 0 . 0 0 1

5 . 6 9

À 7 . 5 9 t o 1 8 . 9 8

0 . 4 0 1

5 . 0 4

3 . 5 9 t o 6 . 4

8

< 0 . 0 0 1

1 1 . 1 0

À 1 . 7 5 t o 2 3 . 9 5

0 . 0 9 0

U r b a n

1 . 8 9

1 . 2 0 t o 2 . 5 9

< 0 . 0 0 1

À 1 . 2 9

À 8 . 2 6 t o 5 . 6 8

0 . 7 1 7

À 0 . 9 6

À 1 . 6 4 t o À 0

. 2 7

0 . 0 0 6

À 7 . 2 3

À 1 3 . 9 8 t o À 0 . 4 9

0 . 0 3 6

M o t h e r w o r k s

8 . 7 6

8 . 2 0 t o 9 . 3 1

< 0 . 0 0 1

9 . 7 8

4 . 8 9 t o 1 4 . 6 8

< 0 . 0 0 1

7 . 3 5

6 . 8 1 t o 7 . 8

9

< 0 . 0 0 1

8 . 3 2

3 . 5 9 t o 1 3 . 0 5

0 . 0 0 1

M a l e

4 . 3 0

3 . 8 6 t o 4 . 7 4

< 0 . 0 0 1

4 . 7 4

0 . 6 5 t o 8 . 8 3

0 . 0 2 3

9 . 8 3

4 . 4 6 t o 1 5 . 2 0

< 0 . 0 0 1

À 1 2 . 4 0

À 1 2 . 8 3 t o À 1

1 . 9 7

< 0 . 0 0 1

À 1 2 . 3 8

À 1 6 . 3 3 t o À 8 . 4 2

< 0 . 0 0 1

À 6 . 2 7

À 1 1 . 8 4 t o À 0 . 7 0

0 . 0 2 7

C o n s t a n t

2 2 4 . 6

2 2 3 . 7 t o 2 2 5 . 4

< 0 . 0 0 1

2 1 9 . 1

2 0 9 . 2 t o 2 2 9 . 0

< 0 . 0 0 1

2 4 7 . 0

2 4 3 . 4 t o 2 5 0 . 5

< 0 . 0 0 1

2 5 1 . 4

2 5 0 . 6 t o 2 5 2 . 2

< 0 . 0 0 1

2 4 9 . 8

2 4 0 . 3 t o 2 5 9 . 4

< 0 . 0 0 1

2 6 7 . 7

2 6 4 . 0 t o 2 7 1 . 4

< 0 . 0 0 1

O b s e r v a t i o n s

2 1 7 5 8 8

2 4 7 4 ( 1 2 3 7 p a i r s )

2 4 7 4 ( 1 2 3 7 p a i r s )

2 1 7 5 8 8

2 4 7 4 ( 1 2 3 7 p a i r s )

2 4 7 4 ( 1 2 3 7 p a i r s )

R 2

0 . 0 8 4

0 . 1 4 6

0 . 0 2 8

0 . 0 7 2

0 . 1 1 2

0 . 0 1 2

C h i l d r e n b o r n i n C h i l e i n 1 9 9 7 – 9 9 a n

d w h o t o o k t h e f o u r t h g r a d e s t a n d a r d i z e d t e s t i n 2 0 0 8 ( 9 5 % C I s i n p a r e n t h e s e s ) . C I s a n d P - v a l u e s i n O L S m o d e l s a r e a d j u s t e d f o r c l u s t e r i n g o f s t a n d a r d

e r r o r s a t t h e m o t h e r ’ s l e v e l .

a R e f e r e n c e g r o u p f o r m o t h e r ’ s e d u c a t i o n i s p r i m a r y e d u c a t i o n .

b R e f e r e n c e g r o u p f o r m o t h e r ’ s a g e i s

< 2 0 y e a r s o l d .

O L S : o r d i n a r y l e a s t s q u a r e m o d e l ; F E : f i x e d e f f e c t s m o d e l ; b : r e g r e s s i o n c o e f f i c i e n t .



genetically different), this finding suggests thatthe factor accounting for the effect of intra-uterinegrowth on test scores is access to nutrition in uterorather than unobserved genetic factors.

These estimates, in association with the Weinbergrule, allow us to gauge the effect among monozygotictwins. According to Weinberg’s formulation, the par-

ameter estimate for MZ twins can be calculated asfollows:

bMZ ¼bss À bMSð1 À pMZÞ

pMZ

where bMZ is the estimate among monozygotic twins,bSS and bMS are, respectively, the estimates forsame-sex and mixed-sex pairs presented in Table 3,and pMZ is the proportion of monozygotic twins inthe same-sex group; 602 twins in the sample belongto mixed-sex pairs and 1872, to same-sex pairs, for atotal sample of 2474. According to Weinberg’s rule,602 twins in the same-sex group are DZ and the re-

maining 1270 are MZ, yielding a pMZ value of 0.678. According to these calculations nearly half of the twinpairs (1270/2474) are monozygotic. This percentage ishigher than reported for some regions of the world,41

but very similar to previous findings for Chile.42 Usingthe parameter estimates from Table 3 we obtainbMZ¼ [8.721 – 3.721 (0.322)]/0.678¼ 11.10 (95% CI:5.16–17.04) for Math test scores and bMZ¼ [6.929 –0.754 (0.322)]/0.678¼9.86 (95% CI: 3.69–16.03) forSpanish (calculation of standard errors, 95% CI and

P-values for these parameter estimates are provided inSupplementary data). Thus 400g increase in birth-

weight results in a very large increase of 22%(11.10/50) standard deviation in Math test scores

and 20% (9.86/50) standard deviation in Spanishtest scores among monozygotic twins. This result fur-ther supports the assertion that the main factor ac-counting for the beneficial effect of fetal growth ontest scores is the availability of uterine resourcesrather than unobserved genetic attributes.

Socioeconomic heterogeneity in the effectof intra-uterine growth

So far we have assumed that the effect of in uterogrowth is homogeneous across the population. How-ever, the homogeneity assumption is likely not satis-fied. In particular, the literature on intra-household

resource allocation suggests that parents may investdifferentially depending on offspring’s birthweight.Parents may reinforce initial disadvantage by allocat-ing more resources to the high-weight twin. Alterna-tively, they may compensate by investing more in thelow-weight twin. Theoretical models suggest parentalbehaviour depends on the degree of equal concernacross children, preference for equity vs productivity,distribution of endowments across children and char-acteristics of the offspring’s outcome (test scores, inthis case) function.43,44 Empirical evidence mostly

favours reinforcement in the USA 44–47 as well as developing contexts.48–51

Parental behaviours are not directly observable,but it is plausible that the extent of compensation

varies according to the socioeconomic resources of the family. Several studies in the industrialized

world have considered this possibility but have

found no substantial variation across socioeconomicstatus.16,47 Variation may be more pronounced in anunequal middle-income country, however. Givenlower levels of income and higher levels of inequalityin Chile, we hypothesize that resource-constrainedfamilies may be more likely to target resources onthe twin with higher chances of succeeding, whereasless constrained families may be more able to com-pensate for observed disadvantages.

In order to assess heterogeneity in the effect of birthweight, Table 4 presents a model with inter-actions between intra-uterine growth and our bestproxy for family resources—mother’s education—dis-tinguishing primary, secondary and tertiary levels of

schooling. The variation across maternal education issubstantial. For Math test scores, a 400 g increasein birthweight raises scores by about 12.4 (95% CI:6.21–18.50) points, i.e. by about 25% (12.4/50) of astandard deviation among low-education mothers.The effect declines to only 14% for mothers with sec-ondary schooling and a negligible 3.7% amongmothers with tertiary schooling. The pattern is thesame for Spanish scores although effects are some-

what weaker and is replicated also across same-sexpairs. The variation in the effect of intra-uterinegrowth by maternal education is consistent with thehypothesis of reinforcing behaviour among poorfamilies and compensatory behaviour among better-

off families—to the extent that the disadvantage asso-ciated with low intra-uterine growth fully disappearsamong advantaged families.

Discussion Although the amount of variation explained by ourmodels is small, our analysis indicates that intra-uterine growth has a substantial effect on children’scognitive development, as measured by test scores inprimary school. The observed effect is larger amongtwin pairs estimated to be monozygotic than amongthose estimated to be dizygotic. Combined, these re-

sults suggest that the factor driving the effect of intra-uterine growth on cognitive development is thescarcity of environmental resources in utero, ratherthan unobserved confounders, including geneticones. These findings are particularly relevant fordeveloping countries, not only because they wereobtained in Chile, but also because the source of

variation in birthweight analysed—intra-uterinegrowth—is the main determinant of low birthweightin the developing world. In fact, intra-uterine growthrestriction emerging from factors such as deficient


http://ije.oxfordjournals.org/cgi/content/full/dyr030/DC1





w t h ( I G ) o n M a t h a n d S p a n i s h t e s t s

c o r e s


S p a n i s h - t e s t s c o r e s

A l l t w i n s

M i x e d - s e x t w i n s

S a m e - s e x t w i n s

A l l t w i n s

M i x e d - s e x t w i n s

S a m e - s e x t w i n s

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

I n t r a - u t e r i n e

g r o w t h

7 . 4 1

4 . 3 3 t o 1 0 . 4 8

< 0 . 0

0 1

3 . 7 2

À 3 . 1 4 t o 1 0 . 6 0

0 . 2 8 7

8 . 7 2

5 . 3 2 t o 1 2 . 1 2

< 0 . 0 0 1

5 . 3 1

2 . 1 2 t o 8 . 4 9

0 . 0 0 1

0 . 7 5 4

À 6 . 0 3 t o 7 . 5 4

0 . 8 2 7

6 . 9 3

3 . 3 3 t o 1 0 . 5 3

< 0 . 0 0 1

M a l e

9 . 8 3

4 . 4 6 t o 1 5 . 2 0

< 0 . 0

0 1

9 . 9 9

3 . 8 4 t o 1 6 . 1 4

0 . 0 0 2

À 6 . 2 7

À 1 1 . 8 4 t o À 0 . 7 0

0 . 0 2 7

À 6 . 0 7 5

À 1 2 . 1 6 t o 0 . 0 1

0 . 0 5 1

C o n s t a n t

2 4 7 . 0

2 4 3 . 4 t o 2 5 0 . 5

< 0 . 0

0 1

2 4 8 . 2

2 4 2 . 0 t o 2 5 4 . 3

< 0 . 0 0 1

2 5 1 . 2 2

4 8 . 4 t o 2 5 3 . 9

< 0 . 0 0 1

2 6 7 . 7

2 6 4 . 0 t o 2 7 1 . 4

< 0 . 0 0 1

2 6 8 . 4

2 6 2 . 3 t o 2 7 4 . 5

< 0 . 0 0 1

2 6 4 . 8

2 6 1 . 9 t o 2 6 7 . 7

< 0 . 0 0 1


2 4 7 4 ( 1 2 3 7 p a i r s )

6 0 2 ( 3 0 1 p a i r s )

1

8 7 2 ( 9 3 6 p a i r s )

2 4 7 4 ( 1 2 3 7 p a i r s )

6 0 2 ( 3 0 1 p a i r s )

1 8 7 2 ( 9 3 6 p a i r s )

R 2

0 . 0 2 8

0 . 0 3 8

0 . 0 2 6

0 . 0 1 2

0 . 0 1 3

0 . 0 1 5

T w i n f i x e d - e f f e c t s f o r a l l t w i n s , s a m e

- s e x a n d m i x e d - s e x t w i n p a i r s . C h i l d r e n

b o r n i n C h i l e i n 1 9 9 7 – 9 9 a n d w h o t o o k

t h e f o u r t h g r a d e s t a n d a r d i z e d t e s t i n 2 0 0 8 ( 9 5 % C I s i n p a r e n t h e s e s ) .

C o n t r o l s f o r m a t e r n a l a g e , m a t e r n a l e d u c a t i o n , m a t e r n a l e m p l o y m e n t s t a t u s

a n d u r b a n r e s i d e n c e d r o p p e d b e c a u s e v a r i a b l e s d o n o t v a r y w i t h i n m o t h e r s .

b : r e g r e s s i o n c o e f f i c i e n t .


w t h ( I G ) o n M a t h a n d S p a n i s h t e s t s

c o r e s a c r o s s l e v e l s o f m o t h e r ’ s e d u c a t i o n


S p a n i s h t e s t s c o r e s

A l l t w i n s

D i f f e r e n t s e x

S a m e s e x

A l l t w i n s

D i f f e r e n t s e x

S a m e s e x

b

9 5 % C I

P - v

a l u e

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

b

9 5 % C I

P - v a l u e

I G * P r i m a r y

e d u c a t i o n

1 2 . 3 5

6 . 2 0 t o 1 8 . 5 0

< 0

. 0 0 1

9 . 8 5

À 6 . 0 2 t o 2 5 . 7 2

0 . 2 2 3

1 2 . 9 7

6 . 4 5 t o 1 9 . 4 9

< 0 . 0 0 1

5 . 7 3

À 0 . 6 5 t o 1 2 . 1

2

0 . 0 7 8

3 . 0 5

À 1 2 . 6 4 t o 1 8 . 7 4

0 . 7 0 2

6 . 4 1

À 0 . 5 0 t o 1 3 . 3 1

0 . 0 6 9

I G * S e c o n d a r y

e d u c a t i o n

6 . 8 2

2 . 4 7 t o 1 1 . 1 7

0 . 0 0 2

1 . 9 2

À 7 . 5 1 t o 1 1 . 3 5

0 . 6 8 9

8 . 7 3

3 . 8 5 t o 1 3 . 6 0

< 0 . 0 0 1

6 . 1 7

1 . 6 6 t o 1 0 . 6 9

0 . 0 0 7

À 2 . 9 6

À 1 2 . 2 8 t o 6 . 3 7

0 . 5 3 3

9 . 7 3

4 . 5 7 t o 1 4 . 8 9

< 0 . 0 0 1

I G * T e r t i a r y

e d u c a t i o n

3 . 6 8

À 2 . 4 4 t o 9 . 8 0

0 . 2 3 9

3 . 0 6

À 9 . 9 1 t o 1 6 . 0 3

0 . 6 4 3

3 . 9 3

À 2 . 9 8 t o 1 0 . 8 5

0 . 2 6 5

3 . 1 6

À 3 . 2 0 t o 9 . 5 2

0 . 3 3 0

6 . 2 6

À 6 . 5 6 t o 1 9 . 0 8

0 . 3 3 7

1 . 8 8

À 5 . 4 5 t o 9 . 2 0

0 . 6 1 5

M a l e

9 . 8 2

4 . 4 5 t o 1 5 . 2 0

< 0

. 0 0 1

9 . 8 3

3 . 6 4 t o 1 6 . 0 2

0 . 0 0 2

À 6 . 1 9

À 1 1 . 7 7 t o À 0 . 6 1

0 . 0 3 0

À 6 . 3 8

À 1 2 . 5 0 t o À 0 . 2 7

0 . 0 4 1

C o n s t a n t

2 4 7 . 0

2 4 3 . 4 t o 2 5 0 . 5

< 0

. 0 0 1

2 4 8 . 3

2 4 2 . 2 t o 2 5 4 . 5

< 0 . 0 0 1

2 5 1 . 1

2 4 8 . 4 t o 2 5 3 . 9

< 0 . 0 0 1

2 6 7 . 7

2 6 4 . 1 t o 2 7 1 . 4

< 0 . 0 0 1

2 6 8 . 4

2 6 2 . 2 t o 2 7 4 . 5

< 0 . 0 0 1

2 6 5 . 0

2 6 2 . 1 t o 2 6 8 . 0

< 0 . 0 0 1


2 4 7 4 ( 1 2 3 7 p a i r s )

6 0 2 ( 3 0 1 p a i r s )

1 8 7 2 ( 9 3 6 p a i r s )

2 4 7 4 ( 1 2 3 7 p a

i r s )

6 0 2 ( 3 0 1 p a i r s )

1 8 7 2 ( 9 3 6 p a i r s )

R - s q u a r e d

0 . 0 3 2

0 . 0 4 0

0 . 0 3 0

0 . 0 1 3

0 . 0 1 7

0 . 0 1 8

T w i n f i x e d - e f f e c t s f o r a l l t w i n s , s a m e

- s e x a n d m i x e d - s e x t w i n p a i r s . C h i l d r e n

b o r n i n C h i l e i n 1 9 9 7 – 9 9 a n d w h o t o o k

t h e f o u r t h g r a d e s t a n d a r d i z e d t e s t i n 2 0 0 8 ( 9 5 % C I s i n p a r e n t h e s e s ) .

b : r e g r e s s i o n c o e f f i c i e n t .



maternal nutrition, low pre-pregnancy BMI, shortstature and infectious diseases such as malaria affectsup to one-third of births in less-developed coun-tries.53,54 Because these factors are likely to result inthe type of uterine scarcity acutely experienced by

monozygotic twins, our findings are potentially relevant for most of the low-weight infants in less advan-taged regions of the world.

But twins are only about 1.5% of all births. Arethese findings generalizable to the larger populationof singletons? Twins are on average much smallerthan singletons at birth. In our sample, for instance,singletons’ weight on average was 3373 g, whereastwins’ weight was only 2500 g. The gap itself is notan issue as long as the factors accounting for thesmallness of twins do not reflect omitted variablesbeyond the scarcity of uterine space and nutrients ina uniparious species.55 In other words, we rely on the

assumption that the smallness of twins at birthshould be ‘the result of twinning rather than the se-lectivity of parents who produce twins’(p. 591).15

Although the literature supports this assumption,this may not be the case for singletons. Given thatsingleton pregnancies are not constrained by uterinespace or nutrient scarcity; is a singleton weighing2500 g comparable with a twin of the same weight?Research suggests that in utero growth restrictionamong singletons may be related to congenital mal-formations, chromosomal anomalies or inborn errorsin metabolism.37,56 To the extent that these negativefactors are correlated with test scores, findings fortwins will not generalizable to singletons of compar-

able birthweight. Furthermore, given that twins are atmuch higher risk of perinatal and infant mortality27

than singletons, surviving twins may be positively se-lected in factors affecting test scores, further reducingcomparability. Negative selectivity of twins is alsopossible. Even if ‘the smallness of twins results fromtwinning’, twins may still be selected on endowmentsrelated to cognitive outcomes such as test scores.

Addressing this question requires examination of thecauses of twinning. Research identifies genetic fac-tors, maternal age, race, height, BMI, fertility-

enhancing drugs and folic acid supplementation ascauses of DZ twinning; and problems during transportthrough the fallopian tube, conception in close prox-imity to oral contraception and assisted reproductivetechnology as causes of MZ twinning.57–59

To assess the comparability of twins and singletonsof similar birthweight and gestational age, we create apooled sample over an area of common supportdefined in terms of birth outcomes and maternalcharacteristics by means of propensity score weight-ing. This strategy weights the singleton data set toensure balance with the twin data set according tothe following variables: birthweight, gestational age,fetal growth, sex, maternal age, education, urbanstatus and work status. Using this matched data set,

we predict Math and Spanish test scores using anindicator for twin. If singletons comparable withtwins are ‘negatively’ selected on factors affecting

test scores, we should observe that being a twin hasa positive effect on test scores. If singletons are in-stead ‘positively’ selected, we should observe thatbeing a twin has a negative effect. Findings, reportedin Table 5, reject both hypotheses. They show no sig-nificant differences in test scores between twins andcomparable singletons, suggesting that growth-restricted singletons are not negatively (or positively)selected and that findings from this study could po-tentially be extended to the larger singleton popula-tion affected by intra-uterine growth restriction.

Another concern refers to the determinants of in utero growth differences within twin pairs, and

whether they are similar to cross-sectional differences

across singletons. We have argued that these factorsrefer to nutrient scarcity in utero, which is more acuteamong MZ than DZ twins. However, in some cases of monozygotic twinning, heavier birthweight is not amarker of advantage, because the twin-to-twin vascu-lar exchange results in poor outcomes for both, thelighter and the heavier twin. This is the case, in par-ticular, with the twin-to-twin transfusion syndrome,

which affects $15% of monozygotic twins that sharea placenta.26 To the extent that twin-to-twin ex-change results in detrimental outcomes for the

Table 5 Propensity score matching estimates of the effect of being a singleton (vs a twin) on Math and Spanish test scores

Math test scores Spanish test scores

b 95% CI P-value b 95% CI P-value

Twin À0.62 À3.16 to 1.92 0.633 À1.72 À4.14 to 0.71 0.165

Constant 247.2 245.9 to 248.4 <0.001 263.0 261.7 to 264.2 <0.001

Observations 220 062 220 062Propensity score matching produces a matched sample of twins and singletons, in which singletons are similar to twins in terms of birthweight, gestational age, uterine growth, sex, maternal age, education and work status. Matching method is nearest neighbour

with replacement and five neighbours in the singleton sample for each twin observation. Alternative methods to calculate matchedoutcome included single nearest neighbour with and without replacement, kernel matching with Epanechnikov kernel and locallinear regression matching provide similar results. Estimates obtained using routine developed by Leuven and Sianesi.60




heavier twin too, the observed effect will be attenu-ated and our estimates will underestimate the trueeffect of uterine growth.

In sum, our analysis shows that in utero growth hasa substantial impact on cognitive development duringchildhood among twins and likely among singletonstoo. Given that intra-uterine growth is the main de-

terminant of low birthweight in the developing world,the analysis is particularly relevant to less-advantagedregions of the world. The variation in the effect acrosslevels of maternal education suggests that, amongdisadvantaged families, parental behaviour may re-inforce the influence of birthweight on educationalachievement, increasing the vulnerability of thelow-birthweight children; whereas high-resourcefamilies may be able to fully compensate for earlydisadvantage. This pattern replicates recent findingsin behavioural genetics showing that the effect of shared environment on twins’ intelligence quotient(IQ) is much stronger in disadvantaged familiesthan in affluent ones.52 More research accounting

for socioeconomic heterogeneity of birthweight effectsin diverse national contexts is necessary to furtherunderstand the influence of the ‘first injustice’ overthe life cycle.

Supplementary DataSupplementary data are available at IJE online.

AcknowledgementsThe authors would like to thank the Ministry of

Education, Chile, for providing the restricted-access version of the SIMCE test scores data set; andDanuta Rajs, Head of the Department of Statisticsand Information, Ministry of Health, for providingaccess to the Chilean birth registry data set. The

views expressed in this article are those of the authorsand do not necessarily reflect those of the Ministries.The authors would also like to thank FranciscoMardones for help in locating the data sets, andJere Behrman and Alejandro Corvalan for helpfulcomments.

Conflict of Interest: None declared.

References1 Mathews TJ, MacDorman MF. Infant mortality statistics

from the 2005 period linked birth/infant death data set. Natl Vital Stat Rep 2008;57:1–32.

2 Hales CN, Barker DJ, Clark PM et al. Fetal and infantgrowth and impaired glucose tolerance at age 64. BMJ 1991;303:1019–22.

3 Barker DJ, Gluckman PD, Godfrey KM, Harding JE,Owens JA, Robinson JS. Fetal nutrition and cardiovascu-lar disease in adult life. Lancet 1993;341:938–41.

4 Paneth NS. The problem of low birth weight. FutureChild. Spring 1995;5:19–34.

5 Pallotto EK, Kilbride HW. Perinatal outcome and laterimplications of intrauterine growth restriction. ClinObstet Gynecol 2006;49:257–69.

6 Currie J, Hyson R. Is the impact of health shocks cush-ioned by socioeconomic status? The case of low birth-

weight. Am Economic Rev 1999;89:245–50.7 Strauss RS. Adult functional outcome of those born smallfor gestational age: twenty-six-year follow-up of the 1970British Birth Cohort. JAMA 2000;283:625–32.

8 Brooks AM, Byrd RS, Weitzman M, Auinger P, McBride JT.Impact of low birth weight on early childhood asthmain the United States. Arch Pediatr Adolesc Med 2001;155:401–06.

9 Matte TD, Bresnahan M, Begg MD, Susser E. Influence of variation in birth weight within normal range and withinsibships on IQ at age 7 years: cohort study. BMJ 2001;323:310–14.

10 Boardman JD, Powers DA, Padilla YC, Hummer RA. Lowbirth weight, social factors, and developmental outcomesamong children in the United States. Demography 2002;39:353–68.

11 Case A, Fertig A, Paxson C. The lasting impact of child-hood health and circumstance. J Health Econ 2005;24:365–89.

12 Almond D, Chay KY, Lee DS. The costs of low birth weight. Q J Econ 2005;120:1031–83.

13 Conley D, Strully KW, Bennett NG. Twin differencesin birth weight: the effects of genotype and prenatal environment on neonatal and post-neonatal mortality. Econ Hum Biol 2006;4:151–83.

14 Datar A, Jacknowitz A. Birth weight effects on children’smental, motor, and physical development: evidence fromtwins data. Matern Child Health J 2009;13:780–94.

15 Behrman J, Rosenzweig M. Returns to birthweight. Rev Econ Stats 2004;86:586–601.

16 Black SE, Devereux PJ, Salvanes KG. From the cradle tothe labor market? The effect of birth weight on adultoutcomes. Q J Econ 2007;122:409–39.

17 Oreopoulos P, Stabile M, Walld R, Roos LL. Short-,medium-, and long-term consequences of poor infanthealth - An analysis using siblings and twins. J Hum Resources 2008;43:88–138.

18 Royer H. Separated at girth: US Twin estimates of theeffects of birth weight. Am Econ J: Appl Econ 2009;1:49–85.

19 Cleary-Goldman J, D’Alton ME. Growth abnormalitiesand multiple gestations. Semin Perinatol 2008;32:206–12.

20 Blickstein I. Is it normal for multiples to be smaller than

singletons? Best Pract Res Clin Obstet Gynaecol 2004;18:613–23.

21 Blickstein I. Growth aberration in multiple pregnancy.Obstet Gynecol Clin North Am 2005;32:39–54.

22 Gringras P, Chen W. Mechanisms for differences inmonozygous twins. Early Hum Dev 2001;64:105–17.

23 Hall JG. Twinning. Lancet 2003;362:735–43.24 Machin G. Non-identical monozygotic twins, intermedi-

ate twin types, zygosity testing, and the non-randomnature of monozygotic twinning: a review. Am J MedGenet C Semin Med Genet 2009;151C:110–27.






25 Hanley ML, Ananth CV, Shen-Schwarz S, Smulian JC,Lai YL, Vintzileos AM. Placental cord insertion andbirth weight discordancy in twin gestations. ObstetGynecol 2002;99:477–82.

26 Sakata M, Utsu M, Maeda K. Fetal circulation and pla-cental blood flow in monochorionic twins. Ultrasound RevObstet Gynecol 2006;6:135–40.

27

Nikkels PG, Hack KE, van Gemert MJ. Pathology of twinplacentas with special attention to monochorionic twinplacentas. J Clin Pathol 2008;61:1247–53.

28 Victoria A, Mora G, Arias F. Perinatal outcome, placentalpathology, and severity of discordance in monochorionicand dichorionic twins. Obstet Gynecol 2001;97:310–15.

29 Bajoria R, Sooranna SR, Ward S, Hancock M. Placenta asa link between amino acids, insulin-IGF axis, and lowbirth weight: evidence from twin studies. J Clin Endocrinol Metab 2002;87:308–15.

30 Blickstein I. Intrauterine growth. In: Blickstein I, Keith L(eds). Multiple Pregnancy. 2nd edn. Oxford, UK: Taylor &Francis, 2005, pp. 503–13.

31 De Paepe ME, Shapiro S, Young L, Luks FI. Placentalcharacteristics of selective birth weight discordance indiamniotic-monochorionic twin gestations. Placenta 2010;31:380–86.

32 Sherer DM, Divon MY. Fetal growth in multifetal gesta-tion. Clin Obstet Gynecol 1997;40:764–70.

33 Hack KE, Derks JB, Elias SG et al. Increased perinatalmortality and morbidity in monochorionic versus dichor-ionic twin pregnancies: clinical implications of a largeDutch cohort study. BJOG 2008;115:58–67.

34 Currie J, Thomas D. Early test scores, socioeconomicstatus and future outcomes. Res Labor Econ 2001;20:103–32.

35 Rose H. Do gains in test scores explain labor market out-comes? Econ Education Rev 2006;25:430–46.

36 Mardones F, Villarroel L, Karzulovic L et al. Association of perinatal factors and obesity in 6- to 8-year-old Chileanchildren. Int J Epidemiol 2008;37:902–10.

37 Resnik R. Intrauterine growth restriction. Obstet Gynecol2002;99:490–96.

38 Weinberg W. Beitrage zur physiologie und pathologie dermehrlingsgebuten beim menschen. Archiv fur die Gesamte Psysiologie des Menschen und der Tiere 1902;88:346–430.

39 Hall JG. Twinning: mechanisms and genetic implications.Curr Opin Genet Dev 1996;6:343–47.

40 Fellman J, Eriksson AW. Weinberg’s differential rulereconsidered. Hum Biol 2006;78:253–75.

41 Imaizumi Y. A comparative study of zygotic twinning andtriplet rates in eight countries, 1972–1999. J Biosoc Sci2003;35:287–302.

42 Ivanovic DM, Llop ER, Alvear JA et al. Chilean school-agechildren twin registry: planning, sampling and implica-tions. Twin Res Hum Genet 2006;9:743–46.

43 Becker G, Tomes N. An equilibrium theory of the distri-bution of income and intergenerational mobility. J Polit Econ 1979;87:1153–89.

44 Behrman J, Pollak R, Taubman P. Parental preferencesand provision for progeny. J Polit Econ 1982;90:52–73.

45 Griliches Z. Sibling models and data in economics: begin-nings of a survey. J Polit Econ 1979;87:S37–64.

46

Behrman J, Rosenzweig M, Taubman P. Endowmentsand the allocation of schooling in the family and in themarriage market: the twins experiment. J Polit Econ 1994;102:1131.

47 Datar A, Kilburn MR, Loughran DS. Endowments andparental investments in infancy and early childhood. Demography 2010;47:145–62.

48 Rosenzweig M, Schultz T. Market opportunities, geneticendowments, and intrafamily resource distribution: childsurvival in rural India. Am Econ Rev 1982;72:803.

49 Rosenzweig M, Wolpin K. Heterogeneity, intrafamily dis-tribution and child health. J Hum Resources 1988;23:437–61.

50 Pitt M, Rosenzweig M, Hassan M. Productivity, health,and inequality in the intrahousehold distribution of food in low-income countries. Am Econ Rev 1990;80:1139.

51 Rosenzweig M, Zhang J. Do population control policiesinduce more human capital investment? twins, birth weight and China’s ’one-child’ policy. Rev Econ Studies2009;76:1149–74.

52 Turkheimer E, Haley A, Waldron M, D’Onofrio B,Gottesman II. Socioeconomic status modifies heritabilityof IQ in young children. Psychol Sci 2003;14:623–28.

53 Kramer M, Victora C. Low birth weight and perinatalmortality. In: Semba RD, Bloem MW (eds). Nutrition and Health in Developing countries. Totowa, NJ: HumanaPress, 2001, pp. 57–69.

54 Bale JR, Stoll BJ, Lucas AO. Institute of Medicine (U.S.).Committee on Improving Birth Outcomes. Improving birth out-

comes: meeting the challenges in the developing world.Washington, DC: National Academies Press, 2003.

55 Blickstein I, Kalish RB. Birthweight discordance in mul-tiple pregnancy. Twin Res 2003;6:526–31.

56 de Bie HM, Oostrom KJ, Delemarre-van de Waal HA.Brain development, intelligence and cognitive outcomein children born small for gestational age. Horm Res Paediatr 2010;73:6–14.

57 Scott L. The origin of monozygotic twinning. Reprod Biomed Online 2002;5:276–84.

58 Hankins GV, Saade GR. Factors influencing twins andzygosity. Paediatr Perinat Epidemiol 2005;19(Suppl. 1):8–9.

59 Hoekstra C, Zhao ZZ, Lambalk CB et al. Dizygotic twin-ning. Hum Reprod Update 2008;14:37–47.

60

Leuven E, Sianesi B. PSMATCH2: stata module to per-form full Mahalanobis and propensity score matching,common support graphing, and covariate imbalance test-ing. 2003.


The effect of birthweight on childhood cognitive development in a middle-income country

Documents

Transcript of The effect of birthweight on childhood cognitive development in a middle-income country