The cognitive cost of being a twin: evidence from comparisons within f
http://www.100md.com
《英国医生杂志》
1 Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London WC1E 7HT
Correspondence to: D A Leon david.leon@lshtm.ac.uk
Objectives To determine whether twins have lower IQ scores in childhood than singletons in the same family and, if so, whether differences in fetal growth explain this deficit.
Design Cohort study.
Setting Scotland.
Participants 9832 singletons and 236 twins born in Aberdeen between 1950 and 1956.
Results At age 7, the mean IQ score of twins was 5.3 points lower (95% confidence interval 1.5 to 9.1) and at age 9, 6.0 points lower (1.7 to 10.2) than that of singletons in the same family. Adjustment for sex, mother's age, and number of older siblings had little effect on these differences. Further adjustment for birth weight and gestational age attenuated the IQ difference between twins and singletons: the difference in mean IQ was 2.6 points (-1.5 to 6.7) at age 7 and 4.1 points (-0.5 to 8.8) at age 9.
Conclusions Twins have substantially lower IQ in childhood than singletons in the same family. This effect cannot be explained by confounding due to socioeconomic, maternal, or other family characteristics, or by recruitment bias. The reduced prenatal growth and shorter gestations of twins may explain an important part of their lower IQ in childhood.
For many years, researchers have been asking whether sharing life with a co-twin in the womb or after birth affects cognitive ability.1 The question is not only of importance to researchers who are interested in twins but may help to explain the determinants of cognitive ability more generally. Childhood cognition is predictive of educational attainment, socioeconomic position, and health in adulthood and therefore has important social and public health implications.2-4
Most previous studies reported that twins have lower cognitive ability than singletons. In a very large study of children born in Birmingham, United Kingdom, between 1950 and 1954, twins had a deficit in verbal reasoning scores at age 11 of 4.4 points on average.5 In the US collaborative perinatal project of hospital births delivered in 1959-65, twins scored lower in cognitive tests at 8 months, 4 years, and 7 years, although substantial loss to follow up had occurred by 7 years.6 In a national sample of Australian schoolchildren born in the 1960s, singletons performed better than twins in tests of word knowledge, reading, and numeracy at ages 10 and 14.7 Similarly, among 10 year olds in Stockholm born in 1953, singletons tended to have higher verbal ability and numerical test scores than twins.8 Most recently, a study was reported that used the Netherlands twin registry to look at differences within families in cognition between 260 adult twins and 98 of their singleton siblings. This found no evidence for a difference in cognitive ability between singletons and twins in the same family.9
Despite these various studies it is still unclear whether something intrinsic to the experience in the womb of being a twin is associated with a cognitive deficit. Maternal characteristics and other aspects of the postnatal family and socioeconomic environment are clearly different between twins and singletons10; many of these aspects are known to be related to cognitive ability.11-15 Much of this potential confounding by familial factors can be dealt with by studying whether twins have a cognitive deficit compared with their singleton brothers or sisters in the same family. However, the only study to date to take this approach did not adjust for factors that vary between siblings in the same family, such as maternal age and order among siblings.9
We used a within family design to investigate the deficit in cognition between twins and singletons. However, instead of using a twin registry we identified families containing twins and singletons from a representative cohort of all people born in Aberdeen, Scotland, and attending primary school there in 1962. We also take our analysis further than others by looking at how far any true twin deficit results from reduced intrauterine growth of twins or shorter gestation.
Methods
Subjects and data
Our study subjects participated in the Aberdeen children of the 1950s study.16 This is comprised of 12 150 individuals born in Aberdeen between 1950 and 1956 and who took part in the Aberdeen child development survey (1962-64) of all children in Aberdeen primary schools in December 1962.17 A large number of participants (5048) have a sibling in the cohort. Siblings were identified by the original Aberdeen child development survey team, using information provided by schools in 1962.
During the primary school careers of our subjects, all children at Scottish primary schools were routinely given written cognitive tests at around the ages of 7 and 9. The original survey team in 1962 ed information about the test scores from school records. Tests were administered within six months of the child's 7th or 9th birthday. Age standardised IQ scores were derived from the test results and were normalised to a national mean of 100 (SD 15). Of the study participants, 11 669 took the test at age 7 and 11 376 at age 9. The Moray House picture intelligence test No 1 or 2 was used at age 7. This is a test based on recognition and understanding of differences between sets of line drawn pictures. At age 9, children took the Schonell and Adams essential intelligence test form A or B, which primarily measured reading ability.
Data on singleton or multiplet status, mother's age at delivery (in five year age bands), birth weight (to the nearest 0.5 lb), gestational age (in completed weeks based on date of last menstrual period), and father's occupational social class at the time of delivery were ed from the Aberdeen maternity and neonatal databank18 at the time of the original survey in 1962. We have taken each child's birth weight for gestational age z score as a measure of their intrauterine growth rate. We used sex specific mean and standard deviations of birth weight for each completed week of gestation among all births in the study for our calculation. We obtained information on the number of older siblings (excluding any co-twin) from a questionnaire administered in 1962 by the original study team. We do not consider information on order of delivery of each twin to be reliable in these data. As a result for the purposes of this analysis we assigned both members of a twin pair to the birth order of the firstborn twin.
Statistical methods
We compared the mean IQ values measured at ages 7 and 9 years of singletons and twins overall and across categories of the available potential confounders. We used analysis of variance to assess the significance of differences between groups and linear trends in mean IQ values.19
To account for sibling correlations we used random effects linear regression models to calculate crude overall mean differences in IQ between twins and singletons.20 We used fixed effects linear regression to estimate differences within families in mean IQ of twins and singletons.20 These differ from those obtained from the random effects models because they control directly for fixed family characteristics. We adjusted for observed potential confounders by introducing them into the models, at first separately and then jointly.
We modelled IQ as a continuous dependent variable. Birth weight, gestational age, and birth weight z score seemed to have non-linear associations with IQ. We therefore included these variables as either categorical variables (gestational age 33-37 weeks, 38-39 weeks, 40-41 weeks, and 42 weeks; birth weight < 5.5 lb (< 2495 g), 5.5-6.0 lb (2495-2947 g), 6.5 lb (2948-3174 g), 7.0-7.5 lb (3175-3628 g), 8.0 lb ( 3629 g); birth weight z score fourths 0.61, -0.61, -0.04, -0.05, 0.63, 0.64) or as continuous variables transformed according to the best fitting fractional polynomial function, best to represent the relation between IQ and the continuous explanatory variables, including repeated powers.21 With this latter approach the regression model might include more than one transformation of the variables. For example, with two such transformations for birth weight—say, one linear and the other square root—there would be two corresponding powers: a power of 1 for the linear and a power of 1/2 for the square root transformation.
In the results we present effects adjusted for the continuous versions of birth weight, gestational age, and birth weight z score, because without exception these seem to be stronger confounders of the difference between twins and singletons in IQ than their categorical equivalents. We categorised the remaining variables as number of older siblings (0, 1, 2, 3); maternal age at delivery (15-24 years, 25-29 years, 30-34 years, 35 years); registrar general's occupational social class of father at child's birth (I, II, III, IV, V); and a category for unemployed, disabled, or deceased.
Results
According to maternity hospital records, of the 12 150 members of the children of the 1950s cohort, 10 were triplets, 306 were twins, and 11 834 were singletons. Excluding the triplets, data on maternal age, father's social class, birth weight, and number of older siblings were available for more than 12 000 (about 99%) of the study sample (respectively 12 134, 12 139, 12 118 and 12 050). Gestational age was available for 10 883 (90%), while IQ at 7 and 9 was available for 11 669 (93%) of the subjects. To obtain comparable estimates we excluded 910 participants because they were missing one or both IQ scores. Of the remaining 11 230 twins and singletons, 1132 had missing data on gestational age and we also excluded these, together with a further 30 who had missing data on other covariates. This left 10 068 participants with complete data, whom we included in multivariable regression analyses (9832 singletons and 236 twins, who together belonged to 8160 families).
Participants excluded from our study had a significantly (P < 0.001) lower mean score at age 7 than those included. The IQ difference was 7.4 points (95% confidence interval 6.6 to 8.3). They also had a significantly (P < 0.001) lower mean score at age 9 (IQ difference 5.5 points, 4.6 to 6.5)). This was partly explained by the fact that those with very low IQs at age 7 or 9 may have been less likely to be tested at either or both ages, as a proportion would have been in special schools where the standard test was not routinely given. In addition, subjects with missing gestational age included a disproportionate number with low birth weight, which is associated with impaired later cognition.22 23
As expected, twins tended to be born smaller and earlier in gestation than singletons and also to be born smaller for their gestational age (table 1). Twins had a greater number of older siblings than singletons on average, indicating that mothers with higher parity (and hence older age) were more likely to deliver twins. The difference in the distribution of father's social class between twins and singletons did not reach significance (P = 0.35).
Table 1 Frequency distribution and mean (SD) IQ at age 7 and 9, by categories of birth and childhood characteristics, separately for singletons and twins. Values are numbers (percentages) of children unless otherwise indicated
We found that singletons had significantly higher mean IQ scores than twins (P < 0.001 at both ages). The crude overall difference was 6.6 points at age 7 (4.4 to 8.8) and 6.9 points at age 9 (4.5 to 9.2). IQ at both ages showed positive linear trends, and for both singletons and twins, across categories of birth weight and gestational age while we found a graded inverse trend with number of older siblings and paternal social class (although this was weaker among twins, table 1).
As shown in table 2, sex, maternal age, number of older siblings, and father's social class at birth explain little of the mean IQ difference between twins and singleton at either age. However, adjustment for birth weight, gestational age, or birth weight z score reduces these differences substantially. The results, however, are still potentially confounded by unmeasured shared maternal and family characteristics that can be controlled for when computing effects within families. Table 3 shows unadjusted differences within families in mean IQ between twins and singletons that are only slightly smaller than the equivalent ones of table 2. As was the case there, sex, maternal age, and number of older siblings have very few confounding effects (paternal social class was effectively fixed in families and therefore not adjusted for). The difference in mean IQ between singletons and twins was again reduced substantially when controlled for birth weight. This adjusted estimate was similar to that produced by simultaneous adjustment for birth weight and gestational age.
Table 2 Estimates of crude and adjusted overall mean differences in IQ between singletons and twins, at ages 7 and 9
Table 3 Mean differences (95 confidence intervals) in IQ within families between singletons and twins, at ages 7 and 9
Discussion
Mehotra SN, Maxwell J. The intelligence of twins. A comparative study of eleven-year-old twins. Population Studies 1949;3: 295-302.
Batty GD, Deary IJ. Early life intelligence and adult health. BMJ 2004;329: 585-6.
Deary IJ, Whiteman MC, Starr JM, Whalley LJ, Fox HC. The impact of childhood intelligence on later life: following up the Scottish mental surveys of 1932 and 1947. J Pers Soc Psychol 2004;86: 130-47.
Hart CL, Taylor MD, Davey SG, Whalley LJ, Starr JM, Hole DJ, et al. Childhood IQ, social class, deprivation, and their relationships with mortality and morbidity risk in later life: prospective observational study linking the Scottish Mental Survey 1932 and the Midspan studies. Psychosom Med 2003;65: 877-83.
Record RG, McKeown T, Edwards JH. An investigation of the difference in measured intelligence between twins and single births. Ann Hum Genet 1970;34: 11-20.
Myrianthopoulos NC, Nichols PL, Broman SH. Intellectual development of twins—comparison with singletons. Acta Genet Med Gemellol (Roma) 1976;25: 376-80.
Hay DA, O'Brien PJ, Johnston CJ, Prior M. The high incidence of reading disability in twin boys and its implications for genetic analyses. Acta Genet Med Gemellol (Roma) 1984;33: 223-36.
Alin AB, Fischbein S. Twins: are they at risk? A longitudinal study of twins and nontwins from birth to 18 years of age. Acta Genet Med Gemellol (Roma) 1991;40: 29-40.
Posthuma D, de Geus EJ, Bleichrodt N, Boomsma DI. Twin-singleton differences in intelligence? Twin Res 2000;3: 83-7.
Bulmer MG. The biology of twinning in man. Oxford: Clarendon Press, 1970.
Record RG, McKeown T, Edwards JH. The relation of measured intelligence to birth order and maternal age. Ann Hum Genet 1969;33: 61-9.
Najman JM, Aird R, Bor W, O'Callaghan M, Williams GM, Shuttlewood GJ. The generational transmission of socioeconomic inequalities in child cognitive development and emotional health. Soc Sci Med 2004;58: 1147-58.
Osler M, Andersen AM, Due P, Lund R, Damsgaard MT, Holstein BE. Socioeconomic position in early life, birth weight, childhood cognitive function, and adult mortality. A longitudinal study of Danish men born in 1953. J Epidemiol Community Health 2003;57: 681-6.
Shenkin SD, Starr JM, Pattie A, Rush MA, Whalley LJ, Deary IJ. Birth weight and cognitive function at age 11 years: the Scottish mental survey 1932. Arch Dis Child 2001;85: 189-96.
Lawlor DA, Batty GD, Morton SM, Deary IJ, Macintyre S, Ronalds G, et al. Early life predictors of childhood intelligence: evidence from the Aberdeen children of the 1950s study. J Epidemiol Community Health 2005;59: 656-63.
Batty GD, Morton SMB, Campbell D, Clark H, Davey-Smith G, Hall M, et al. The Aberdeen children of the 1950s cohort study: background, methods and follow-up information on a new resource for the study of life-course and intergenerational effects on health. Paed Perinat Epidemiol 2004;18: 221-39.
Birch HG, Richardson SA, Baird D, Horobin G, Illsley R. Mental subnormality in the community: a clinical and epidemiologic study. Baltimore: Williams and Wilkins, 1970.
Samphier ML, Thompson B. Longitudinal studies in Aberdeen, Scotland. B. The Aberdeen maternity and neonatal databank. In: Mednick S, Baert A, Bachmann B, eds. Prospective longitudinal research. An empirical basis for the primary prevention of psychosocial disorders. Oxford: Oxford University Press, 1981: 61-5.
Altman DG. Practical statistics for medical research. London: Chapman and Hall, 1991.
Baltagi BH, Chang Y. Incomplete panels: a comparative study of alternative estimators for the unbalanced one-way error component regression model. J Econom 1994;62: 67-89.
Royston P, Altman DG. Regression using fractional polynomials of continuous covariates: parsimonious parametric modelling (with discussion). Applied Stat 1994;43: 429-67.
Hack M, Weissman B, Breslau N, Klein N, Borawski Clark E, Fanaroff AA. Health of very low birth weight children during their first eight years. J Pediatr 1993;122: 887-92.
Shenkin SD, Starr JM, Deary IJ. Birth weight and cognitive ability in childhood: a systematic review. Psychol Bull 2004;130: 989-1013.
Boomsma DI, Vink JM, van Beijsterveldt TC, de Geus EJ, Beem AL, Mulder EJ, et al. Netherlands twin register: a focus on longitudinal research. Twin Res 2002;5: 401-6.
Breslau N. Psychiatric sequelae of low birth weight. Epidemiol Rev 1995;17: 96-106.
Matte TD, Bresnahan M, Begg MD, Susser E. Influence of variation in birth weight within normal range and within sibships on IQ at age 7 years: cohort study. BMJ 2001;323: 310-4.
Boomsma DI, van Beijsterveldt CE, Rietveld MJ, Bartels M, van Baal GC. Genetics mediate relation of birth weight to childhood IQ. BMJ 2001;323: 1426-7.(Georgina A Ronalds, research assistant1,)
Correspondence to: D A Leon david.leon@lshtm.ac.uk
Objectives To determine whether twins have lower IQ scores in childhood than singletons in the same family and, if so, whether differences in fetal growth explain this deficit.
Design Cohort study.
Setting Scotland.
Participants 9832 singletons and 236 twins born in Aberdeen between 1950 and 1956.
Results At age 7, the mean IQ score of twins was 5.3 points lower (95% confidence interval 1.5 to 9.1) and at age 9, 6.0 points lower (1.7 to 10.2) than that of singletons in the same family. Adjustment for sex, mother's age, and number of older siblings had little effect on these differences. Further adjustment for birth weight and gestational age attenuated the IQ difference between twins and singletons: the difference in mean IQ was 2.6 points (-1.5 to 6.7) at age 7 and 4.1 points (-0.5 to 8.8) at age 9.
Conclusions Twins have substantially lower IQ in childhood than singletons in the same family. This effect cannot be explained by confounding due to socioeconomic, maternal, or other family characteristics, or by recruitment bias. The reduced prenatal growth and shorter gestations of twins may explain an important part of their lower IQ in childhood.
For many years, researchers have been asking whether sharing life with a co-twin in the womb or after birth affects cognitive ability.1 The question is not only of importance to researchers who are interested in twins but may help to explain the determinants of cognitive ability more generally. Childhood cognition is predictive of educational attainment, socioeconomic position, and health in adulthood and therefore has important social and public health implications.2-4
Most previous studies reported that twins have lower cognitive ability than singletons. In a very large study of children born in Birmingham, United Kingdom, between 1950 and 1954, twins had a deficit in verbal reasoning scores at age 11 of 4.4 points on average.5 In the US collaborative perinatal project of hospital births delivered in 1959-65, twins scored lower in cognitive tests at 8 months, 4 years, and 7 years, although substantial loss to follow up had occurred by 7 years.6 In a national sample of Australian schoolchildren born in the 1960s, singletons performed better than twins in tests of word knowledge, reading, and numeracy at ages 10 and 14.7 Similarly, among 10 year olds in Stockholm born in 1953, singletons tended to have higher verbal ability and numerical test scores than twins.8 Most recently, a study was reported that used the Netherlands twin registry to look at differences within families in cognition between 260 adult twins and 98 of their singleton siblings. This found no evidence for a difference in cognitive ability between singletons and twins in the same family.9
Despite these various studies it is still unclear whether something intrinsic to the experience in the womb of being a twin is associated with a cognitive deficit. Maternal characteristics and other aspects of the postnatal family and socioeconomic environment are clearly different between twins and singletons10; many of these aspects are known to be related to cognitive ability.11-15 Much of this potential confounding by familial factors can be dealt with by studying whether twins have a cognitive deficit compared with their singleton brothers or sisters in the same family. However, the only study to date to take this approach did not adjust for factors that vary between siblings in the same family, such as maternal age and order among siblings.9
We used a within family design to investigate the deficit in cognition between twins and singletons. However, instead of using a twin registry we identified families containing twins and singletons from a representative cohort of all people born in Aberdeen, Scotland, and attending primary school there in 1962. We also take our analysis further than others by looking at how far any true twin deficit results from reduced intrauterine growth of twins or shorter gestation.
Methods
Subjects and data
Our study subjects participated in the Aberdeen children of the 1950s study.16 This is comprised of 12 150 individuals born in Aberdeen between 1950 and 1956 and who took part in the Aberdeen child development survey (1962-64) of all children in Aberdeen primary schools in December 1962.17 A large number of participants (5048) have a sibling in the cohort. Siblings were identified by the original Aberdeen child development survey team, using information provided by schools in 1962.
During the primary school careers of our subjects, all children at Scottish primary schools were routinely given written cognitive tests at around the ages of 7 and 9. The original survey team in 1962 ed information about the test scores from school records. Tests were administered within six months of the child's 7th or 9th birthday. Age standardised IQ scores were derived from the test results and were normalised to a national mean of 100 (SD 15). Of the study participants, 11 669 took the test at age 7 and 11 376 at age 9. The Moray House picture intelligence test No 1 or 2 was used at age 7. This is a test based on recognition and understanding of differences between sets of line drawn pictures. At age 9, children took the Schonell and Adams essential intelligence test form A or B, which primarily measured reading ability.
Data on singleton or multiplet status, mother's age at delivery (in five year age bands), birth weight (to the nearest 0.5 lb), gestational age (in completed weeks based on date of last menstrual period), and father's occupational social class at the time of delivery were ed from the Aberdeen maternity and neonatal databank18 at the time of the original survey in 1962. We have taken each child's birth weight for gestational age z score as a measure of their intrauterine growth rate. We used sex specific mean and standard deviations of birth weight for each completed week of gestation among all births in the study for our calculation. We obtained information on the number of older siblings (excluding any co-twin) from a questionnaire administered in 1962 by the original study team. We do not consider information on order of delivery of each twin to be reliable in these data. As a result for the purposes of this analysis we assigned both members of a twin pair to the birth order of the firstborn twin.
Statistical methods
We compared the mean IQ values measured at ages 7 and 9 years of singletons and twins overall and across categories of the available potential confounders. We used analysis of variance to assess the significance of differences between groups and linear trends in mean IQ values.19
To account for sibling correlations we used random effects linear regression models to calculate crude overall mean differences in IQ between twins and singletons.20 We used fixed effects linear regression to estimate differences within families in mean IQ of twins and singletons.20 These differ from those obtained from the random effects models because they control directly for fixed family characteristics. We adjusted for observed potential confounders by introducing them into the models, at first separately and then jointly.
We modelled IQ as a continuous dependent variable. Birth weight, gestational age, and birth weight z score seemed to have non-linear associations with IQ. We therefore included these variables as either categorical variables (gestational age 33-37 weeks, 38-39 weeks, 40-41 weeks, and 42 weeks; birth weight < 5.5 lb (< 2495 g), 5.5-6.0 lb (2495-2947 g), 6.5 lb (2948-3174 g), 7.0-7.5 lb (3175-3628 g), 8.0 lb ( 3629 g); birth weight z score fourths 0.61, -0.61, -0.04, -0.05, 0.63, 0.64) or as continuous variables transformed according to the best fitting fractional polynomial function, best to represent the relation between IQ and the continuous explanatory variables, including repeated powers.21 With this latter approach the regression model might include more than one transformation of the variables. For example, with two such transformations for birth weight—say, one linear and the other square root—there would be two corresponding powers: a power of 1 for the linear and a power of 1/2 for the square root transformation.
In the results we present effects adjusted for the continuous versions of birth weight, gestational age, and birth weight z score, because without exception these seem to be stronger confounders of the difference between twins and singletons in IQ than their categorical equivalents. We categorised the remaining variables as number of older siblings (0, 1, 2, 3); maternal age at delivery (15-24 years, 25-29 years, 30-34 years, 35 years); registrar general's occupational social class of father at child's birth (I, II, III, IV, V); and a category for unemployed, disabled, or deceased.
Results
According to maternity hospital records, of the 12 150 members of the children of the 1950s cohort, 10 were triplets, 306 were twins, and 11 834 were singletons. Excluding the triplets, data on maternal age, father's social class, birth weight, and number of older siblings were available for more than 12 000 (about 99%) of the study sample (respectively 12 134, 12 139, 12 118 and 12 050). Gestational age was available for 10 883 (90%), while IQ at 7 and 9 was available for 11 669 (93%) of the subjects. To obtain comparable estimates we excluded 910 participants because they were missing one or both IQ scores. Of the remaining 11 230 twins and singletons, 1132 had missing data on gestational age and we also excluded these, together with a further 30 who had missing data on other covariates. This left 10 068 participants with complete data, whom we included in multivariable regression analyses (9832 singletons and 236 twins, who together belonged to 8160 families).
Participants excluded from our study had a significantly (P < 0.001) lower mean score at age 7 than those included. The IQ difference was 7.4 points (95% confidence interval 6.6 to 8.3). They also had a significantly (P < 0.001) lower mean score at age 9 (IQ difference 5.5 points, 4.6 to 6.5)). This was partly explained by the fact that those with very low IQs at age 7 or 9 may have been less likely to be tested at either or both ages, as a proportion would have been in special schools where the standard test was not routinely given. In addition, subjects with missing gestational age included a disproportionate number with low birth weight, which is associated with impaired later cognition.22 23
As expected, twins tended to be born smaller and earlier in gestation than singletons and also to be born smaller for their gestational age (table 1). Twins had a greater number of older siblings than singletons on average, indicating that mothers with higher parity (and hence older age) were more likely to deliver twins. The difference in the distribution of father's social class between twins and singletons did not reach significance (P = 0.35).
Table 1 Frequency distribution and mean (SD) IQ at age 7 and 9, by categories of birth and childhood characteristics, separately for singletons and twins. Values are numbers (percentages) of children unless otherwise indicated
We found that singletons had significantly higher mean IQ scores than twins (P < 0.001 at both ages). The crude overall difference was 6.6 points at age 7 (4.4 to 8.8) and 6.9 points at age 9 (4.5 to 9.2). IQ at both ages showed positive linear trends, and for both singletons and twins, across categories of birth weight and gestational age while we found a graded inverse trend with number of older siblings and paternal social class (although this was weaker among twins, table 1).
As shown in table 2, sex, maternal age, number of older siblings, and father's social class at birth explain little of the mean IQ difference between twins and singleton at either age. However, adjustment for birth weight, gestational age, or birth weight z score reduces these differences substantially. The results, however, are still potentially confounded by unmeasured shared maternal and family characteristics that can be controlled for when computing effects within families. Table 3 shows unadjusted differences within families in mean IQ between twins and singletons that are only slightly smaller than the equivalent ones of table 2. As was the case there, sex, maternal age, and number of older siblings have very few confounding effects (paternal social class was effectively fixed in families and therefore not adjusted for). The difference in mean IQ between singletons and twins was again reduced substantially when controlled for birth weight. This adjusted estimate was similar to that produced by simultaneous adjustment for birth weight and gestational age.
Table 2 Estimates of crude and adjusted overall mean differences in IQ between singletons and twins, at ages 7 and 9
Table 3 Mean differences (95 confidence intervals) in IQ within families between singletons and twins, at ages 7 and 9
Discussion
Mehotra SN, Maxwell J. The intelligence of twins. A comparative study of eleven-year-old twins. Population Studies 1949;3: 295-302.
Batty GD, Deary IJ. Early life intelligence and adult health. BMJ 2004;329: 585-6.
Deary IJ, Whiteman MC, Starr JM, Whalley LJ, Fox HC. The impact of childhood intelligence on later life: following up the Scottish mental surveys of 1932 and 1947. J Pers Soc Psychol 2004;86: 130-47.
Hart CL, Taylor MD, Davey SG, Whalley LJ, Starr JM, Hole DJ, et al. Childhood IQ, social class, deprivation, and their relationships with mortality and morbidity risk in later life: prospective observational study linking the Scottish Mental Survey 1932 and the Midspan studies. Psychosom Med 2003;65: 877-83.
Record RG, McKeown T, Edwards JH. An investigation of the difference in measured intelligence between twins and single births. Ann Hum Genet 1970;34: 11-20.
Myrianthopoulos NC, Nichols PL, Broman SH. Intellectual development of twins—comparison with singletons. Acta Genet Med Gemellol (Roma) 1976;25: 376-80.
Hay DA, O'Brien PJ, Johnston CJ, Prior M. The high incidence of reading disability in twin boys and its implications for genetic analyses. Acta Genet Med Gemellol (Roma) 1984;33: 223-36.
Alin AB, Fischbein S. Twins: are they at risk? A longitudinal study of twins and nontwins from birth to 18 years of age. Acta Genet Med Gemellol (Roma) 1991;40: 29-40.
Posthuma D, de Geus EJ, Bleichrodt N, Boomsma DI. Twin-singleton differences in intelligence? Twin Res 2000;3: 83-7.
Bulmer MG. The biology of twinning in man. Oxford: Clarendon Press, 1970.
Record RG, McKeown T, Edwards JH. The relation of measured intelligence to birth order and maternal age. Ann Hum Genet 1969;33: 61-9.
Najman JM, Aird R, Bor W, O'Callaghan M, Williams GM, Shuttlewood GJ. The generational transmission of socioeconomic inequalities in child cognitive development and emotional health. Soc Sci Med 2004;58: 1147-58.
Osler M, Andersen AM, Due P, Lund R, Damsgaard MT, Holstein BE. Socioeconomic position in early life, birth weight, childhood cognitive function, and adult mortality. A longitudinal study of Danish men born in 1953. J Epidemiol Community Health 2003;57: 681-6.
Shenkin SD, Starr JM, Pattie A, Rush MA, Whalley LJ, Deary IJ. Birth weight and cognitive function at age 11 years: the Scottish mental survey 1932. Arch Dis Child 2001;85: 189-96.
Lawlor DA, Batty GD, Morton SM, Deary IJ, Macintyre S, Ronalds G, et al. Early life predictors of childhood intelligence: evidence from the Aberdeen children of the 1950s study. J Epidemiol Community Health 2005;59: 656-63.
Batty GD, Morton SMB, Campbell D, Clark H, Davey-Smith G, Hall M, et al. The Aberdeen children of the 1950s cohort study: background, methods and follow-up information on a new resource for the study of life-course and intergenerational effects on health. Paed Perinat Epidemiol 2004;18: 221-39.
Birch HG, Richardson SA, Baird D, Horobin G, Illsley R. Mental subnormality in the community: a clinical and epidemiologic study. Baltimore: Williams and Wilkins, 1970.
Samphier ML, Thompson B. Longitudinal studies in Aberdeen, Scotland. B. The Aberdeen maternity and neonatal databank. In: Mednick S, Baert A, Bachmann B, eds. Prospective longitudinal research. An empirical basis for the primary prevention of psychosocial disorders. Oxford: Oxford University Press, 1981: 61-5.
Altman DG. Practical statistics for medical research. London: Chapman and Hall, 1991.
Baltagi BH, Chang Y. Incomplete panels: a comparative study of alternative estimators for the unbalanced one-way error component regression model. J Econom 1994;62: 67-89.
Royston P, Altman DG. Regression using fractional polynomials of continuous covariates: parsimonious parametric modelling (with discussion). Applied Stat 1994;43: 429-67.
Hack M, Weissman B, Breslau N, Klein N, Borawski Clark E, Fanaroff AA. Health of very low birth weight children during their first eight years. J Pediatr 1993;122: 887-92.
Shenkin SD, Starr JM, Deary IJ. Birth weight and cognitive ability in childhood: a systematic review. Psychol Bull 2004;130: 989-1013.
Boomsma DI, Vink JM, van Beijsterveldt TC, de Geus EJ, Beem AL, Mulder EJ, et al. Netherlands twin register: a focus on longitudinal research. Twin Res 2002;5: 401-6.
Breslau N. Psychiatric sequelae of low birth weight. Epidemiol Rev 1995;17: 96-106.
Matte TD, Bresnahan M, Begg MD, Susser E. Influence of variation in birth weight within normal range and within sibships on IQ at age 7 years: cohort study. BMJ 2001;323: 310-4.
Boomsma DI, van Beijsterveldt CE, Rietveld MJ, Bartels M, van Baal GC. Genetics mediate relation of birth weight to childhood IQ. BMJ 2001;323: 1426-7.(Georgina A Ronalds, research assistant1,)