OBJECTIVE: The aim was to study the impact of a range of gestational ages (GAs) on cognitive competence in late adolescence and how this effect is modified by contextual social adversity in childhood.
METHODS: This was a register study based on a national cohort of 119664 men born in Sweden from 1973 to 1976. Data on GA and other perinatal factors were obtained from the Medical Birth Register, and information on cognitive test scores was extracted from military conscription at the ages of 18 to 19 years. Test scores were analyzed as z scores on a 9-point stanine scale, whereby each unit is equivalent to 0.5 SD. Socioeconomic indicators of the childhood household were obtained from the Population and Housing Census of 1990. The data were analyzed by multivariate linear regression.
RESULTS: The mean cognitive test scores decreased in a stepwise manner with GA. In unadjusted analysis, the test scores were 0.63 stanine unit lower in men who were born after 24 to 32 gestational weeks than in those who were born at term. The difference in global scores between the lowest and highest category of socioeconomic status was 1.57. Adjusting the analysis for the childhood socioeconomic indicators decreased the effect of GA on cognitive test scores by 26% to 33%. There was also a multiplicative interaction effect of social adversity and moderately preterm birth on cognitive test scores.
CONCLUSIONS: This study confirms previous claims of an incremental association of cognitive competence with GA. Socioeconomic indicators in childhood modified this effect at all levels of preterm birth.
Increasing numbers of children who are born preterm survive into adulthood as a result of advances in prenatal and neonatal care. A considerable number of cohorts of former patients from NICUs have been followed prospectively to adolescence.1,2 At school age, 10% to 12% of preterm infants have considerable impairment because of a neurologic disability.3–5 Apart from these more obvious disabilities, there are also indications of more subtle effects of very preterm birth on cognitive competence, with reports of learning problems in school1 and mean IQ scores 5 to 27 points below those of children who are born at term.2 There are some indications that other kinds of psychiatric disorders may be associated with preterm birth, particularly for children who are raised in socially disadvantaged families,6 whereas antisocial behavior and drug use has not been reported to be more common in children who are born very preterm.7
In contrast to the abundance of follow-up studies of former patients from NICUs, the long-term influence of moderately preterm and early term birth on cognitive competence has been much less studied.8 Another less explored area is the variation of the effect of preterm birth on cognitive competence in various socioeconomic contexts, despite indications that socioeconomic factors modify cognitive outcomes of preterm birth9 as well as intrauterine growth retardation.10 In a recent Dutch study of 19-year-olds, socioeconomic factors were even found to be more important for intelligence than very preterm birth.11 Wolke and Meyer9 reported the effect of low socioeconomic status (SES) and preterm birth on cognitive development to be additive, whereas Bendersky and Lewis12 found evidence of a multiplicative effect (ie, that SES enhances the effect of preterm birth on cognitive development).
In this study, we used a window in the Swedish military conscription history, when the same intellectual performance test was administered to 4 birth cohorts of Swedish 18-year-old men, to study the impact of various levels of preterm birth on cognitive competence in the late teens. Do socioeconomic variables modify the effect of gestational age (GA) on cognitive competence in adolescence?
This study was based on data from the Swedish Medical Birth Register, the Multi-Generation Register, and the Military Service Conscription Register linked through each individual's unique personal identification number. Additional information on the household was obtained through linkage to the Population and Housing Census of 1990 and the Swedish Register of Education.
The Military Service Conscription Register comprises information about male residents in Sweden who have been conscripted for military service. Conscription is mandatory and enforced by law, except for individuals with severe somatic and social handicaps and residents with a foreign citizenship. The conscription procedure takes place at 6 regional centers in Sweden.
In the Medical Birth Register, all male individuals who were born in Sweden between 1973 and 1976 with a reported GA ≤41 weeks were identified (n = 180719). A total of 1376 children were excluded because of a registered birth weight >3 SD or less than −6 SD, according to the growth chart developed by Marsal et al,13 indicating a probable coding error.14 A total of 2344 individuals died before conscription, most of them during the neonatal period. The mortality rate was highest (62%) among those who were born at GAs of 24 to 28 weeks and decreased gradually by GA to 0.9% in the term group. Of the remaining 176699 individuals, 140279 had a record in the Military Service Conscription Register of conscription from January 1991 to September 1994. For 119664 of these individuals, complete data were available for all 4 subtests of intellectual capacity. These individuals constitute the study population. The proportion of individuals who were available for military conscription and were included in the final study population varied in a predictable manner with GA at birth from 52% in those who were born in weeks 24 to 28, 62% for weeks 29 to 32, 64% for weeks 33 to 34, and 66% to 68% for weeks 35 to 41, consistent with the higher rates of disabilities reported for individuals who were born preterm.15
Pregnancy, Birth, and Perinatal Variables
Maternal age at delivery, date of birth, and birth order of the offspring and relevant perinatal characteristics were identified in the Swedish Medical Birth Register. GA, estimated as completed gestational weeks on the basis of the reported last menstrual period, was used to group individuals according to GA (24–28, 29–32, 33–34, 35–36, 37–38, and 39–41 weeks). Small for GA (SGA; less than −2 SDs) was coded according to Marsal et al.13 Low Apgar was defined as an Apgar score <7 at 5 minutes.
The marital status of the mother (married/unmarried) and the SES of the father were identified in the Swedish Population and Housing Census of 1990. Socioeconomic groups were defined according to a 6-category variable created by Statistics Sweden, which is based on occupation but also takes into account the level of education, type of production, and position at work of the head of the household.16 A dichotomized simplified SES was created for a graphic illustration of the relation between SES and GA on test scores (Fig 1). “High SES” consisted of white collar 2 and 3 households in which the head of household has a qualified nonmanual employment, and “low SES” of all other households. The highest completed number of years in education of the mother was identified in the Swedish Register of Education in 2001, 1995, or 1990 (in that order) and categorized as 9, 10 to 11, 12 to 13, or ≥14 years.
At conscription, the young men underwent a standardized physical health examination that included height measurement and a general intellectual performance test. Scores from 1990 to August 1994, which were based on the “Enlistment Battery 80,” were used to create our outcome variables. During the conscription procedure, intellectual performance is measured by 4 subtests, each consisting of 40 items, representing logical, spatial, verbal, and technical capabilities. The correlations of the 4 tests were moderately high (0.45–0.68). A summary score from the 4 subtests was standardized against an entire conscript cohort from a previous year and categorized into 9 bands, to give a Gaussian distributed score between 1 and 9. Higher values indicate enhanced intellectual capacity, with each unit increase on the stanine scale corresponding to an increase of 0.5 SD.
Because of military secrecy, the tests are not available for people outside the Swedish conscription authority; however, a construct validity analysis of the global scale has been published.17 In a confirmatory factor analysis, it was demonstrated that the global score could be “seen as a good estimate of general intellectual capacity defined as an ability to solve complex problems.”17 The logical test measures the capacity for understanding written instructions and applying them for solving a problem. In the spatial test, the task is to determine which 3-dimensional object (of examples presented) will result from folding up a given 1-dimensional object, which has marked lines, indicating where to fold (the “folding” is performed mentally). The verbal test measures the knowledge of synonyms; the subject should determine which of 4 alternatives is the synonym of a given word (40 words are presented as such key words). The aim of the verbal test is to measure “linguistic understanding and ability to use oral and written language.” The technical test, “technical comprehension,” also measures knowledge of chemistry and physics and implies a component of general knowledge. All tests are presented in succession to the conscripts as written questionnaires.
The associations between test scores and GA were analyzed by multiple linear regression whereby the effect on test scores of socioeconomic and perinatal variables were accounted for in a sequence of 3 models. In this analysis, GA was entered as a 5-category dummy variable (24–32, 33–34, 35–36, 37–38, and 39–41 weeks). The dependent variable was units of decrease in stanine score in various categories of preterm birth in relation to term births (ie, the reference group). Model 1 was adjusted for year of birth, age at conscription, and testing center only. Childhood SES, birth order, and maternal age and education were added in model 2, and Apgar score and SGA were added as potential perinatal mediators of GA in model 3 (Table 1).
Possible multiplicative interaction effects of SES and preterm birth on test scores were investigated in an age- and gender-adjusted linear regression with interaction terms based on the dichotomized SES variable described already. This analysis was made with only 4 categories of GA—weeks 24 to 32, weeks 33 to 36, and weeks 37 to 38—because of the lower frequencies in each cell in this multivariate analysis. Statistical analyses were performed using SPSS 14.0 for Windows software package (SPSS, Inc, Chicago, IL).
Sociodemographic and perinatal indicators of the study population in relation to GA are presented in Table 2. Children who were born preterm tended to have parents with less advantageous socioeconomic situations and lower educational levels in 1990 than those who were born at term. Compared with term-born infants, preterm infants more frequently had low Apgar scores at 5 minutes and were more often SGA than those who were born at term.
Table 3 shows the distribution of test scores in relation to GA. The scores increased gradually from 4.46 in gestational weeks 24 to 28 to 5.07 in weeks 39 to 41. There was a consistent socioeconomic gradient with higher household SES and maternal education linked to higher scores in all tests, with a span from 4.46 in conscripts in the lowest SES category to 6.01 in the highest. Firstborn young men generally had higher test scores than the subsequent children in the same household. In addition, men who were born to younger mothers had lower scores than those with mothers aged 25 to 34 at childbirth.
Table 2 presents the multivariate analysis of GA and test scores. The gradual association remained in all 3 models. The difference in test score between the most preterm and term birth was −0.63. The magnitude of the effect of preterm birth on the test score decreased by 26% to 33% after adjustment for the socioeconomic variables and another 5% to 20% after adjustment for the perinatal mediators.
Figure 1 shows test scores according to GA by a dichotomized SES variable. All GA categories, including very preterm, from high-SES households had higher mean scores than the term category from low-SES households. When this dichotomized SES variable was investigated in linear regression, a multiplicative interaction effect was found for being born in weeks 33 to 36 (β = .16, P < .01), whereas no such effects were found for weeks 24 to 32 or for weeks 37 to 38.
In this national cohort study of Swedish men who did not have major disabilities, were born in the 1970s, and were followed up at 18 to 19 years of age at military conscription, there was a gradual incremental association of GA with test scores of intellectual performance all the way from very preterm to early term birth (weeks 37–38) and term birth. Approximately one third of the magnitude of the association of preterm birth with test scores was explained by the socioeconomic indicators in childhood in the multivariate analysis. Because the socioeconomic variables that were available in this study must be considered crude proxy measures for complex contextual living conditions during childhood, there is probably considerable unmeasured social confounding and/or effect modification in the analyses, suggesting that this estimate is an underestimation of the true effect of social adversity.
This study suggests that factors related to the family environment during childhood are important mediators of the more subtle cognitive deficits that have repeatedly been described in school children and adolescents who are born preterm without major disabilities. The socioeconomic indicators that were used in this study should be considered as markers for a myriad of risk and protective factors that are involved in the interaction of the child with his environment,18 including genetic factors nested into family variables.19 One may speculate that 1 key factor in this particular cohort of Swedish children may have been access to preschool activities, which were more readily available for children in high-SES families in Sweden in these birth cohorts.20
The results of this study suggest that lack of appropriate socioeconomic matching or adjustment, as was the case with the majority of the studies in the aforementioned meta-analysis by Bhutta et al,2 will create a consistent overestimation of the true effect of GA on cognitive competence. The magnitude of this bias was ∼30% in this study but will probably be greater in many other countries, because Sweden is a country with comparatively limited socioeconomic differences in health in an international perspective.15
Recent studies with structural MRI have suggested that the effect of GA on cognitive competence in adolescents is mediated by a gradual decrease in gray matter and a parallel increase in white matter by degree of preterm birth.21 This indicates that GA at birth to a certain degree determines the long-term development of the brain structure.22 The results of this study are consistent with this hypothesis but add new research questions by pointing to the importance of the childhood environment for these effects.
Generalizations to later cohorts of preterm infants should be made with caution. This is particularly important for the conclusions that can be drawn about children who are born before 27 weeks of gestation, for whom the survival rate was very low in the birth cohorts in this study. It seems probable that the increased survival rates of smaller and more vulnerable children will have a negative effect on cognitive competence in this GA group. To a certain extent, however, this effect can be expected to be counterbalanced by the dramatic improvement of obstetric and neonatal care since the 1970s. The latter is illustrated by a steadily decreasing frequency of cerebral palsy, despite higher survival rates, in preterm infants in Europe since the 1980s.23,24
Another important limitation is that only young men were investigated and, furthermore, only young men without severe disabilities. Weisglas-Kuperus et al11 did not find any important gender differences of the effect of very preterm birth on intelligence in 19-year-olds. Neither did the exclusion of individuals with overt handicaps change the effect of very preterm birth on intelligence in their study, but we cannot be certain that the results from this more recent Dutch neonatal context are valid for our study population as well.
The SES indicator of the childhood household was obtained in 1990, when the study population was a mean age of 15. The SES of some families can be expected to have changed considerably since the early childhood of the study population. It seems quite possible that SES obtained early in the life of the study population, when family influences on cognition are at its peak, would have demonstrated an even greater impact of SES on cognitive competence.
The cohort in this study was born long before ultrasound became a routine procedure to measure GA in early pregnancy in Sweden. Ultrasound measures tend to lead to higher rates of preterm birth than estimates that are based on maternal report of last menstrual period,25 indicating that some of the early term births in this study would have been recategorized as preterm if ultrasound measures had been available. Register data also include a certain number of coding errors that may create outliers that are falsely labeled preterm births.14 We excluded individuals with implausible combinations of birth weight and GA to minimize this problem. Still, a few individuals who were born at term may have been coded as preterm, thereby possibly diluting the true associations of negative outcomes with preterm birth. To single out further the impact of preterm versus other perinatal complications, we were able to adjust our analyses for SGA and low Apgar score. Still, some residual confounding probably remains, because we were not able to adjust the analysis for all kinds of relevant perinatal problems.
Figure 1 confirms the recent results of Weisglas-Kuperus et al11 that social disadvantage is a more important determinant of cognitive deficits than moderately preterm birth per se. Children who carry the double burden of being born preterm into a socially disadvantaged family, however, should be of particular concern for developmental interventions. This in line with previous findings by Brooks-Gunn et al26 that early developmental interventions for preterm and low birth weight offspring are most effective in families in which the parents lack a college education.
The multiplicative interaction effect of social adversity and moderately preterm birth on cognitive test scores suggests that the negative effect of preterm birth on cognitive competence to a considerable extent may be prevented by a favorable environment later in life. This finding needs to be investigated further for very preterm birth in more recent birth cohorts in which survival rates of very preterm births were much higher.
There is a gradual incremental association of GA with cognitive competence. Socioeconomic context modifies this effect at all levels of preterm birth.
This study was funded by the Swedish National Board of Health and Welfare and the Swedish Council for Working Life and Social Research.
- Accepted July 18, 2009.
- Address correspondence to Anders Hjern, MD, PhD, Centre for Epidemiology, Swedish National Board of Health and Welfare, 106 30 Stockholm, Sweden. E-mail:
Financial Disclosure: The authors have indicated they have no financial relationships relevant to this article to disclose.
What's Known on This Subject:
Survivors of very preterm birth often have cognitive deficits.
What This Study Adds:
This study demonstrates that low SES is a more powerful determinant of cognitive competence/IQ than moderately preterm birth. Low SES enhances the effect of moderately preterm birth on cognitive competence.
- ↵Ross G, Lipper EG, Auld PA. Educational status and school-related abilities of very low birth weight premature children. Pediatrics.1991;88 (6):1125– 1134
- ↵Lindstrom K, Lindblad F, Hjern A. Psychiatric morbidity in adolescents and young adults born preterm: a Swedish national cohort study. Pediatrics.2009;123 (1). Available at:www.pediatrics.org/cgi/content/full/123/1/e47
- ↵Escobar GJ, McCormick MC, Zupancic JA, et al. Unstudied infants: outcomes of moderately premature infants in the neonatal intensive care unit. Arch Dis Child Fetal Neonatal Ed.2006;91 (4):F238– F244
- ↵Bergvall N, Iliadou A, Johansson S, Tuvemo T, Cnattingius S. Risks for low intellectual performance related to being born small for gestational age are modified by gestational age. Pediatrics.2006;117 (3). Available at: www.pediatrics.org/cgi/content/full/117/3/e460
- ↵Weisglas-Kuperus N, Hille ET, Duivenvoorden HJ, et al. Intelligence of very preterm or very low birthweight infants in young adulthood. Arch Dis Child Fetal Neonatal Ed.2009;94 (3):F196– F200
- ↵Wilkinson RW. Unhealthy Societies: The Afflictions of Inequality. London, England: Routledge; 1996
- ↵Statistics Sweden. Socio-economic Classification (SEI). Stockholm, Sweden: Statistics Sweden; 1982
- ↵Garbarino J. Children and Families in the Social Environment. New York, NY: Aldine de Gruyter; 1992
- ↵Nosarti C, Giouroukou E, Healy E, et al. Grey and white matter distribution in very preterm adolescents mediates neurodevelopmental outcome. Brain.2008;131 (pt 1):205– 217
- ↵Brooks-Gunn J, Gross RT, Kraemer HC, Spiker D, Shapiro S. Enhancing the cognitive outcomes of low birth weight, premature infants: for whom is the intervention most effective? Pediatrics.1992;89 (6 pt 2):1209– 1215
- Copyright © 2010 by the American Academy of Pediatrics