Background: Adolescence is a critical period of brain structural reorganisation and maturation of cognitive abilities. This relatively late developmental reorganisation may be altered in individuals who were born preterm.
Methods: We carried out longitudinal neuropsychological testing in 94 very preterm individuals (VPT; before 33 weeks’ gestation) and 44 term born individuals at mean ages of 15.3 years (adolescence) and 19.5 years (young adulthood).
Results: Full scale, verbal and performance IQ and phonological verbal fluency were significantly lower in the VPT group than the term group at both ages. Repeated measures ANOVA showed only one group by time point interaction for semantic verbal fluency (F = 10.25; df = 107; p = 0.002). Paired-sample t tests showed that semantic verbal fluency increased significantly in the term group over adolescence (t = −5.10; df = 42; p<0.001), but did not increase in the VPT group (t = 0.141; df = 69; p = 0.889). For verbal IQ, there was a significant interaction between time point and sex (F = 4.48; df = 1; p = 0.036) with paired-sample t tests showing that verbal IQ decreased in males between adolescence and adulthood (t = 3.35; df = 71; p = 0.001), but did not change significantly in females (t = 0.20; df = 52; p = 0.845).
Conclusion: Decrements of intellectual functioning in VPT individuals persist into adulthood. Additionally, there is a deficit in the adolescent maturation of semantic verbal fluency in individuals born VPT.
Statistics from Altmetric.com
Adolescence is a time of complex changes in behaviour, cognition, social interaction and brain structure. Neuropsychological abilities such as verbal fluency mature during this period,1 associated with a refining and restricting of fMRI activation patterns in the frontal lobe.2 Executive functions are generally refined with age, with “mature” levels of performance being attained in adolescence and young adulthood.3–5 Longitudinal imaging studies6–10 show that white matter generally increases through adolescence, most likely the result of ongoing myelination, whereas grey matter is generally decreasing, probably a consequence of synaptic pruning or of increased myelination in its underlying white matter.11 Maturation of specific areas of temporo-parietal white matter revealed by diffusion tensor imaging (DTI) may underlie the maturation of reading ability.12
Less is known about how adolescent maturation is altered in individuals who have recovered from early neurodevelopmental insults—such as those born very preterm (VPT; before 33 weeks’ gestation) or of very low birth weight (VLBW; less than 1500 g). These individuals are at increased risk of perinatal brain damage and subsequent developmental delay, but little is known about their adolescent development. Neuropsychological deficits in these groups certainly persist into the teenage years—for example, Rushe et al13 found impairments of cognitive flexibility and phonological verbal fluency in a group of VPT 15-year-olds. There are also some reports that neuropsychological function may deteriorate with maturation. Botting et al. (1998)14 found a decline in cognitive measures between 6 and 12 years in a cohort of VLBW children. Saigal et al15 found a decline in mathematical ability between 8 years and early teens in extremely low birth weight (ELBW) individuals, although global cognitive function (measured by IQ) remained stable over this time. A recent study by O’Brien et al16 reported that IQ declines significantly between 8 and 14 years in VPT individuals. There is evidence that VLBW males are more susceptible than females to adverse developmental sequelae,17 but little is known about sex differences in cognitive maturation in VPT adolescents.
In spite of such reports, few studies have yet followed up preterm individuals through adolescence and into adulthood. In this study we carried out neuropsychological evaluations in two groups of individuals: one VPT and the other term born, in adolescence and then again in young adulthood. We hypothesised that preterm individuals would show a decline in intellectual performance or a failure to make age-appropriate gains in performance, over this interval. We further hypothesised that this would be more evident in VPT males than females.
Very preterm group
A total of 406 infants born before 33 weeks’ gestation between 1982 and 1984 were admitted to the neonatal unit of University College Hospital (UCH) London within five days of birth and later discharged (VPT group). From this population, 302 survived and were recruited as part of a long-term follow-up study. The cohort was assessed at 1, 4, 8 and 15 years old using a battery of neuropsychological tests. These results have been extensively published.18 19 At 15 years, 111 of these individuals were assessed (the adolescent assessment). For the young adult assessment, these individuals were re-contacted. Ninety-four (85%) agreed to participate in further assessment. Preterm individuals who were not assessed did not differ significantly from those who were assessed in their gestational age (t = 1.55, df = 469, p = 0.12), Apgar scores at 1 minute (t = 0.84, df = 145, p = 0.40) and 5 minutes (t = 0.31, df = 145, p = 0.76), sex (χ2 = 0.82, df = 1, p = 0.37) or social class (χ2 = 1.3, df = 4, p = 0.87).
A group of 71 individuals born between 38 and 42 weeks’ gestation was recruited by advertisement in the local (South London) press for the adolescent assessments to act as controls. These same individuals were invited for the young adult assessments. Forty-four individuals (62%) were assessed at both time points.
General intellectual functioning
An IQ measure was performed at both adolescent and young adult assessments. In adolescence the age-appropriate 8-item Wechsler Intelligence Scale for Children – Revised (WISC-R)20 was used. In young adulthood, the Wechsler Abbreviated Scale of Intelligence (WASI),21 a 4-subtest measure, was used. At each age, measures of verbal, performance and full-scale IQ were available.
Two domains of verbal fluency were assessed:
phonological or letter fluency, using the Controlled Oral Word Association Test (COWAT),22 where participants are requested to produce words beginning with a given letter (in this case “F”, “A” and “S”) in 60 seconds.
semantic (category) fluency, using the Animal Naming test, in which participants are required to produce as many names of animals as they can, in a trial lasting 60 seconds.23
Verbal fluency assesses the executive system that enables initiation of response, mental flexibility and also the ability to use strategies such as clustering, where words are produced in subcategories, either phonological or semantic.24 In addition to executive functions, components of verbal fluency production include short-term memory of phonological information and simple word retrieval processes. In the analyses the following scores were used: (1) phonological verbal fluency (COWAT score calculated as the total number of words produced during the “F”, “A” and “S” trials); (2) category fluency (score calculated as the total number of words produced during the Animal Naming test). The COWAT and Animal Naming tests were administered at both adolescent and young adult assessments. Harrison et al25 reported test-retest reliability of 0.82 for phonological (FAS) and 0.68 semantic components of the COWAT.
Neonatal cerebral ultrasound classification
Ultrasound ratings were performed in the perinatal period using a linear ultrasound array. VPT subjects were divided into three groups on the basis of these ratings (0 = normal; 1 = uncomplicated periventricular haemorrhage; 2 = periventricular haemorrhage and ventricular dilatation)26 (see table 1). Ultrasound ratings were available for the VPT group only.
Data were analysed using Statistical Package for the Social Sciences (SPSS) v 12.0.1. Demographic data were analysed using independent-sample t tests or χ2 tests for categorical data. Longitudinal data were analysed using repeated-measures ANOVA, with time point (adolescent and young adult) (twofold) as within-subject independent variable, and sex and group as between-subject factors, covarying for subjects’ ages at both time points and for their socioeconomic status. Main effects were tested using models that did not include the interactions. Partial eta squared is reported as a measure of effect size. This measure provides an estimate of the proportion of the total variance accounted for by the factor under consideration. For example, a partial eta squared of 0.1 would indicate that the factor accounts for 10% of the overall variance. Longitudinal neuropsychological changes within groups were explored further using paired sample t tests. For some analyses, a measure of change in neuropsychological function was calculated using the formula: [adult score] – [adolescent score]. Relationships with perinatal variables were sought using Spearman correlations. To assess the effects of neonatal brain injury on adolescent cognitive maturation, a separate repeated measures ANOVA was performed using neonatal ultrasound severity as between-subjects factor, and time point as within-subject independent variable in the VPT group only. Statistical assumptions were checked and the data were screened for outlying values.
Demographic characteristics of the study groups
Ninety-four VPT individuals (41 female; 53 male) and 44 term individuals (18 female; 26 male) were assessed at two time points: in adolescence (at a mean age of 15.3 years (SD 0.70; range 14–17) and in young adulthood (at a mean age of 19.5 years (SD 1.08; range 18–25)). Sex distribution did not differ between the groups (χ2 = 0.09; df = 1; p = 0.854). The groups were significantly different in social class distribution (Registrar General’s classification) (χ2 = 11.44; df = 4; p = 0.022). VPT individuals were slightly, but significantly, older than term individuals at both adolescent (t = −3.77; df = 135; p<0.001) and adult assessments (t = −3.62; df = 132; p<0.001) (see table 1). The neonatal characteristics of the VPT group have been published elsewhere.18 19 The results presented below include all patients on whom an assessment at both ages was available. Denominators may vary between different tests.
Results of IQ and verbal fluency tests in adolescence and adulthood in VPT and term groups are shown in table 2. For the following repeated-measures ANOVAs, the assumption that covariance matrices were equal across groups was tested using Box’s M test. This was significant only for verbal IQ (M = 19.34; F = 2.07; p = 0.028). The repeated measures ANOVA for vIQ was therefore adjusted using the Greenhouse-Geisser correction.
Full scale IQ
Repeated measures ANOVA for full scale IQ (at two time points, covarying for social class and age at assessments) revealed a significant main effect of group (F = 10.89; df = 108; p = 0.001; partial eta squared = 0.092), but no significant main effect of sex or time point. There was a significant interaction between time point and sex (F = 5.52; df = 108; p = 0.021; partial eta squared = 0.049), but no significant interaction between time point and group, or between group and sex. There was no significant three-way interaction between time point, group and sex (see table 3).
In adolescence, full scale IQ was 11.1 points lower in the VPT group (t = 3.25; df = 124; p = 0.001). In adulthood, full-scale IQ was 8.9 points lower in the VPT group (t = 4.25; df = 138; p = 0.001) (see table 2). Among males, full-scale IQ decreased by 5.0 points between adolescence and adulthood (t = −3.75; df = 70; p<0.001). There was no significant change in full-scale IQ over this time in the females.
Repeated measures ANOVA for verbal IQ revealed a main effect of group (F = 9.66; df = 115; p = 0.002; partial eta squared = 0.077), but no significant main effect of sex or time point. There was a significant interaction between time point and sex (F = 4.48; df = 1; p = 0.036; partial eta squared = 0.038), but no significant interactions between time point and group or group and sex, and no significant three-way interaction between time point, group and sex (see table 3).
Verbal IQ was significantly lower in the VPT group in adolescence (t = 3.63; df = 124; p<0.001) and adulthood (t = 3.44; df = 137; p = 0.002). Paired sample t tests showed that verbal IQ decreased in males between adolescence and adulthood (t = 3.35; df = 71; p = 0.001), but did not change significantly in females (t = 0.20; df = 52; p = 0.845).
Repeated measures ANOVA for performance IQ revealed a main effect of group (F = 5.58; df = 112; p = 0.020; partial eta squared = 0.047), but no significant main effects of time point of sex. There were no significant interactions between time point and group; time point and sex; group and sex; or time point, group and sex (table 3).
Performance IQ was lower in the VPT group in both adolescence (t = 2.52; df = 122; p = 0.013) and adulthood (t = 3.44; df = 138; p = 0.001).
Phonological verbal fluency
Repeated measures ANOVA for phonological verbal fluency revealed a significant main effect of group (F = 6.91; df = 116; p = 0.010; partial eta squared = 0.056) but no main effect of sex or time point. There were no significant interactions between time point and group; or between time point and sex; or between group and sex; or between time point, group and sex (see table 3).
Phonological verbal fluency was significantly lower in the VPT group than the term group at both adolescent (t = 2.60; df = 133; p = 0.010) and adult (t = 2.73; df = 131; p = 0.007) assessments (see table 2).
Semantic verbal fluency
Repeated measures ANOVA for semantic verbal fluency revealed no significant main effect of group, sex or time point. There was a significant interaction between time point and group (F = 10.25; df = 107; p = 0.002; partial eta squared = 0.077) (see fig 1), but no significant interactions between time point and sex; or group and sex; or time point, group and sex (see table 3). Paired-sample t tests showed that semantic verbal fluency increased significantly in term group over adolescence (mean difference = −2.86; 95% confidence intervals (CI) −3.99, −1.73; t = −5.10; df = 42; p<0.001), but did not increase in the VPT group (mean difference = 0.086; 95% CI −1.13, 1.30; t = 0.141; df = 69; p = 0.889).
Associations between perinatal variables and cognitive maturation in the VPT group
A measure of change in neuropsychological function over time was derived by subtracting the adolescent test results from the adult test results. Relationships between change scores and perinatal variables were sought using Spearman’s correlations as the change scores did not all follow a normal distribution. There were no significant correlations between neuropsychological change and birth weight and gestational age. To assess relationships between neonatal brain injury and neuropsychological maturation, we performed repeated measures ANOVA in VPT individuals with time point as within-subject independent variable and neonatal ultrasound classification as between-subject factor. There were no significant main effects of ultrasound severity, and no significant interactions between ultrasound classification and time point for any of the neuropsychological tests.
We have shown that decrements in neuropsychological performance are still present as VPT individuals enter early adulthood, with VPT individuals having lower verbal, performance and full-scale IQ, and phonological verbal fluency scores, at both adolescent and adult assessments. This indicates that neuropsychological deficits in VPT individuals persist into adulthood rather than being attenuated by maturation. There is little published evidence that is directly comparable to this study, and other longitudinal studies on this group have generally included participants at younger ages. For example, O’Brien et al16 followed up VPT individuals at 8 years and 15 years and found that the mean full-scale IQ (measured by the WISC-R) fell from 104 to 95 in a VPT group between 8 and 15 years. Similarly, Botting et al14 found a decline in cognitive performance between 6 and 12 years. Our results would indicate that further decline does not occur between 15 and 19 years. Methodological considerations compel us to treat these results with some caution. Comparison of neuropsychological testing at different ages, as reported here, is often complicated by the use of different assessment instruments at different ages. This is partly due to, and is further confounded by, the fact that the range of neuropsychological tests that can be performed increases greatly with age and maturation of the central nervous system. For example, an age-appropriate assessment for an 8 year old may differ markedly from an age-appropriate test for an 18 year old.
Additionally, we have demonstrated an altered pattern of maturation of semantic verbal fluency in VPT individuals who did not make the same age-related gains that were seen in the term group. Little is known about normal adolescent cognitive maturation and the maturation of verbal fluency. Cohen et al27 reported that verbal fluency increases in children between 6 and 12 years. Tombaugh et al28 in a cross-sectional study, found that semantic fluency was more affected by the age of the subject than phonological fluency, whereas Matute et al1 also found age-related increases in both phonological and semantic verbal fluency. Here we provide evidence of maturation of both phonological and semantic verbal fluency, with performance on these tasks increasing between 15 and 19 years in normal adolescents. A similar pattern of maturation was found in the VPT group for phonological fluency, but VPT individuals did not show the age-related increases in semantic fluency that we found in the term group.
Fu et al29 in an fMRI study found that phonological verbal fluency activates a network comprising the anterior cingulate, left inferior frontal, middle frontal and parietal cortex and the right cerebellum. This would suggest that the classical neuropsychological explanation of verbal fluency as a property of the dominant frontal lobe may be an over-simplification. Given the widespread reorganisation of brain structure and function in VPT and VLBW individuals, it is reasonable to suggest that tasks requiring the recruitment and coordination of several neural “modules” may be performed less well by these individuals. Previous work in a similar cohort of VPT individuals has shown a relationship between volume decrement of the cerebellum and intellectual functioning.30
An explanation for the dissociation between semantic and phonological fluency maturation in the VPT group could be that semantic tasks require recruitment of more neural resources than letter fluency. Both tasks require activation of a phonological system, but the semantic task requires the additional activation of a “meaning” pathway.31 Converging evidence from structural26 32 and functional33 imaging in VPT and VLBW adolescents and adults suggests that such individuals have undergone considerable plastic reorganisation of the brain. This neural plasticity is beneficial in that it protects against gross functional disturbance, but it may give rise to a brain that has difficulty coping with increasing task demands.34
Adolescent maturation of cognitive function may be related to the dynamic changes in brain structure that have been demonstrated over adolescence and early adulthood. Longitudinal imaging studies such as that of Sowell et al8 have demonstrated changes in Broca’s and Wernicke’s areas, which may underlie the rapidly changing cognitive and linguistic abilities. Early brain lesions acquired by VPT and VLBW individuals have been shown to cause focal and distributed abnormalities of structure.32 How this interacts with normal processes of brain maturation is currently not determined, but it may be that early lesions impede these relatively late neurodevelopmental changes such that the brain is unable to maintain a normal developmental trajectory, and neurodevelopment lags behind increasing cognitive demands. An alternative possibility is that the altered maturation of verbal fluency that we report is related to reduced speed of neural processing. Rose et al35 reported slower processing of information about faces in preterm infants up to 1 year when compared to term infants. It has even been suggested that processing speed reduction may account for as much as 60% of the IQ differences between VPT and term 11 year olds.36 Further follow up would be necessary to distinguish these two potential aetiologies.
We should acknowledge several potential limitations of this study. Chief among these is the problem of comparing neuropsychological measures taken at different ages, using different versions of the Wechsler intelligence scales (the WISC-R in adolescence and the WASI in adulthood). To some extent this is an unavoidable consequence of longitudinal studies—the appropriate test may be different at different stages of cognitive maturation. We have used well-standardised measures to mitigate this. Additionally, there is a well-known tendency for IQ scores to increase over time37 38 and we used the WISC-R towards the end of its life (when it may have been overestimating IQ). Conversely, the WASI was relatively newly introduced and may have underestimated IQ in early adulthood. However, we would expect that such effects on IQ measurement would apply equally to males and females, whereas we have demonstrated differential effects. The identical test of verbal fluency was administered at both ages, and differences in the maturation of this cognitive domain are therefore less likely to be an artefact of testing. A further limitation is the slight (but statistically significant) difference in ages between the groups, with the VPT group being slightly older at both testing phases. However, this should have allowed the VPT group more “maturation time”, which would tend to attenuate rather than exaggerate our findings, and additionally we controlled for the effects of age in the statistical analyses.
VPT birth is associated with a decrement in general intellectual ability, which we have shown to persist into adulthood. Further, we have shown that normal maturational trajectory of semantic verbal fluency is not followed in VPT individuals. The course of maturation of other cognitive abilities that we did not test remains unclear. It is also not yet clear whether VPT individuals’ verbal fluency development lags behind that of their peers, and may therefore be able to “catch up”, or whether it represents an enduring deficit. Further follow up of VPT individuals into adulthood will be necessary to answer this question.
We thank the preterm and term born individuals who took part in this study.
Funding: The study was funded by the Wellcome Trust. Dr Allin is supported by the Peggy Pollak Fellowship in Developmental Psychiatry, Psychiatry Research Trust.
Competing interests: None.
Ethics approval: This study was approved at both adolescent and adult phases by the Medical Ethical Committee of the Institute of Psychiatry, King’s College London.
Patient consent: At adolescence, written informed consent was obtained from a parent or guardian. All participants provided further written informed consent in adulthood.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.