This article is commented on by Alloway on pages 203204 of this issue.
The profile of executive function in very preterm children at 4 to 12 years
Article first published online: 29 NOV 2011
© The Authors. Developmental Medicine & Child Neurology © 2011 Mac Keith Press
Developmental Medicine & Child Neurology
Volume 54, Issue 3, pages 247–253, March 2012
How to Cite
AARNOUDSE-MOENS, C. S. H., DUIVENVOORDEN, H. J., WEISGLAS-KUPERUS, N., VAN GOUDOEVER, J. B. and OOSTERLAAN, J. (2012), The profile of executive function in very preterm children at 4 to 12 years. Developmental Medicine & Child Neurology, 54: 247–253. doi: 10.1111/j.1469-8749.2011.04150.x
- Issue published online: 10 FEB 2012
- Article first published online: 29 NOV 2011
- Accepted for publication 22nd August 2011. Published online 29th November 2011.
Aim To examine executive functioning in very preterm (gestational age ≤30wks) children at 4 to 12 years of age.
Method Two-hundred very preterm (106 males, 94 females; mean gestational age 28.1wks, SD 1.4; mean age 8y 2mo, SD 2y 6mo) and 230 term children (106 males, 124 females; mean gestational age 39.9wks, SD 1.2; mean age 8y 4mo, SD 2y 3mo) without severe disabilities, born between 1996 and 2004, were assessed on an executive function battery comprising response inhibition, interference control, switching, verbal fluency, verbal and spatial working memory, and planning. Multiple regression analyses examined group differences while adjusting for effects of parental education, age, sex, and speed indices.
Results Relative to children born at term, very preterm children had significant (ps<0.02; where ps represents p-values) deficits in verbal fluency (0.5 standardized mean differences [SMD]), response inhibition (0.4 SMD), planning (0.4 SMD), and verbal and spatial working memory (0.3 SMD), independent of slow and highly fluctuating processing speed. A significant group by age interaction indicated that group differences for response inhibition decreased between 4 and 12 years.
Interpretation Very preterm birth is associated with a profile of affected and non-affected executive functions independent of impaired speed. Deficits are of small to moderate magnitude and persist over time, except for response inhibition for which very preterm children catch up with peers.
Mean reaction time
Standardized mean differences
What this paper adds
- •Executive dysfunction after very preterm birth comprises a profile of affected and non-affected areas.
- •The executive function profile remains largely persistent over time and is independent of slow and highly fluctuating processing speed.
Improvements in perinatal care have resulted in increased survival rates for children born very preterm (gestational age ≥30wks). The incidence of major disabilities, such as cerebral palsy (CP), intellectual disabilities, deafness, or blindness in these children is relatively low.1 There is growing awareness, however, that most of the survivors with normal IQs are at risk for ‘subtle’ neurocognitive deficits, such as motor impairments,2 academic underachievement, and behavioural problems.3
Executive functioning has been considered one of the crucial mechanisms underlying academic and behavioural problems.4–8 In the past decade, it has therefore been the subject of much research into the outcomes of very preterm children. Executive functioning refers to interrelated neurocognitive processes, such as response inhibition, working memory, switching, planning, and fluency, that control thought and behaviour.8–10 Earlier studies have found executive functioning deficits in very preterm children.3,11 However, because of inclusion of often small numbers of children of restricted age ranges and the use of measures tapping into multiple aspects of executive functioning, literature still diverges on which executive functioning domains are precisely affected in this population and to what extent executive functioning deficits persist over time.
Poor executive functioning after very preterm birth has been related to smaller volumes of basal ganglia and cerebellum, as well as to disruptions of subcortical white matter circuits connecting frontal, striatal, and thalamic regions.12,13 These white matter disruptions affect efficiency of neural signalling, which also results in slow processing speed and highly variable task performance (i.e. moment-to-moment fluctuations in speed).14 It has, therefore, been postulated that poor executive functioning in very preterm children may in fact reflect deficiencies in speed of information processing.15
The aim of this study was to examine a comprehensive range of executive functioning in a large sample of very preterm and term children across the age range of 4 to 12 years with well-defined and validated measures of executive functioning. Response inhibition, interference control, verbal and spatial working memory, switching, verbal fluency, and planning were assessed in a large sample of very preterm and term children who were comparable in age and sex. All children were free of major disabilities.
The very preterm (gestational age not more than 30wks) sample was derived from all (n=706) very preterm surviving singletons admitted between 1996 and 2004 to the neonatal intensive care unit of the Erasmus University Medical Center, Sophia Children’s Hospital Rotterdam, the Netherlands. For an elaborate description of the inclusion procedure and neonatal characteristics of very preterm children we refer to an earlier publication.16 Briefly, twins were excluded as inclusion of these children would violate the assumption of independence of observations. Very preterm children with a severe disability (one that was likely to put the child in need of physical assistance to perform daily activities)17 would not be able to perform tests as used in the present study and were therefore not invited. The present study was performed in 2007 and 2008. The comparison group was recruited from three regular primary schools located in the same neighbourhoods as schools attended by the very preterm children; it included children without histories of prematurity (gestational age >37wks), perinatal complications, and neurological disorders.
Minor neurosensory dysfunctions as observed in participating children are presented in Table I and included (1) vision corrected to normal with contact lenses or glasses, (2) hearing loss corrected to normal with hearing aids, and (3) spastic unilateral CP, classified according to standards of the Surveillance of Cerebral Palsy in Europe.
|Very preterm group (n=200)||Term comparison group (n=230)|
|Gestational age, wks||28.1||1.4||24.5–30.0||39.9||1.2||37.0–43.0|
|<28wks, n (%)||87.0||43.5||0.0||0.0|
|<1500g, n (%)||191.0||95.5||0.0||0.0|
|Males, n (%)||106.0||53.0||106.0||46.1|
|Parental education,cn (%)|
|Minor neurosensory dysfunction, n (%)||37.0||18.5||13.0||5.6|
|Minor vision loss or corrected with contact lenses or glasses||26.0||13.0||13.0||5.6|
|Minor hearing loss or corrected with hearing aids||5.0||2.5||0.0||0.0|
|Spastic unilateral cerebral palsy||6.0||3.0||0.0||0.0|
Response inhibition was measured with the stop task, which requires a child to respond as quickly and accurately as possible to a go-stimulus (cartoon airplane presented for 1000ms) and to inhibit the response if a stop-stimulus (cross presented for 50ms) is presented. The initial delay between the go-signal and stop-signal was 250 milliseconds and was increased by 50 milliseconds if the child inhibited the response, and decreased by 50 milliseconds if the child did not succeed in inhibiting the response. Twenty-five per cent of the trials were stop-trials. The intertrial interval was 1500 milliseconds. Two practice blocks of 24 trials, of which the first included go-trials, and the second go-trials and stop-trials, preceded four experimental blocks of 48 trials of go-trials and stop-trials. Dependent variables derived included errors of commission and omission, and stop signal reaction time,18 an estimate of the time a child needed to stop his or her response (defined as mean reaction time [MRT] minus the mean delay).
Interference control was assessed using an Eriksen flanker task,19 which involves neutral, congruent, and incongruent trials. A neutral trial consisted of a target arrow flanked by rectangles (-->-- or --<--). A congruent trial consisted of a target arrow flanked by arrows that pointed in the same direction as the target (>>>>> or <<<<<). An incongruent trial consisted of a target arrow flanked by arrows pointing in the opposite direction (incongruent) as the target (>><>> or <<><<), which causes interference.19 Children were required to inhibit responses to these interfering stimuli. Stimuli disappeared after the child responded and were presented with a maximum duration of 3000 milliseconds. The intertrial interval was 1500 milliseconds. A practice block of 12 trials (four trials per type) preceded two experimental blocks, consisting of 36 trials each (24 trials per type). Incongruent trials induced slower reaction times and more omission and commission errors than congruent trials (ps<0.001; where ps represents p-values). Dependent variables were an interference score for MRT (i.e. MRT on incongruent trials minus MRT on congruent trials), and interference scores for errors of omission and commission.
Switching was measured using a stimulus–response compatibility task. Target stimuli, arrows, differed in colour with a green arrow indicating that the child had to respond with a spatially compatible response (left arrow mapping onto left response button), and a red arrow indicating that the child had to respond with a spatially incompatible response (left arrow mapping onto right response button). Stimuli disappeared after the child responded and were presented with a maximum duration of 3000 milliseconds. The intertrial interval was 1500 milliseconds. Two practice blocks of six trials each (six compatible and six incompatible trials) preceded an experimental block consisting of 48 trials (24 compatible and 24 incompatible trials). Incompatible trials induced slower reaction times and more omission and commission errors than compatible trials (ps<0.01). Dependent variables were a switch score for MRT (i.e. MRT on incompatible trials minus MRT on compatible trials), and switch scores for errors of commission and omission.
Spatial working memory was assessed using the Spatial Span subtest of the Cambridge Neuropsychological Testing Automated Battery.20 This test measures the capacity to store and manipulate spatial information temporarily. Children viewed a lighted sequence of squares and were required to reproduce the sequence by touching items on a touch-screen in the same order as originally illuminated. The dependent variable was the maximum span.
Verbal working memory was assessed using the backwards condition of the Digit Span subtest of the Wechsler Intelligence Scale for Children–III.21 This test measures the capacity to store and manipulate verbal information temporarily. In the backwards condition, digits that were read by the examiner (one digit per second) were to be repeated in the reverse order. Children received one point for each correct response. The dependent variable was the total number of correct sequences.
Verbal fluency was measured in a task that required children to name as many examples of two specific categories: ‘animals’ and ‘things you can eat or drink’ within a 40-second time frame.5 Two examples of each category were provided before the beginning of the task. An item named for the second time was scored as incorrect. The dependent variable was the total number of correct responses.
Planning was assessed using the Cambridge Neuropsychological Testing Automated Battery subtest Stockings of Cambridge.20 The Stockings of Cambridge is a touch-screen-adapted version of the Tower of London task. Children were instructed to solve problems by moving coloured circles between three locations in a prescribed number of moves. Problems were graded in ascending difficulty, involving two to five moves required per problem. Dependent variables derived were number of problems solved, planning time, and execution time. Analyses were performed on trials with five moves, taking performance on two-move trials into account to examine effects of increasing difficulty levels.
Processing speed was measured with the MRT on go-trials of the Stop task (only correct trials).
Fluctuations in speed were measured using the SD of the MRT on go-trials of the Stop task divided by MRT (SD of MRT/MRT22).
IQ was estimated using the subtests Vocabulary and Block Design of the Wechsler Intelligence Scale for Children-III,21 or the Wechsler Primary and Preschool Scale Intelligence–Revised23 (depending on the child’s age). Subtest scores were converted into a composite score that was used to calculate an estimated IQ, which correlates highly (0.9 range) with Full-scale IQ.24
Assessments of executive functioning and IQ for very preterm children took place at the Erasmus University Medical Centre Rotterdam, Sophia Children’s Hospital Rotterdam. Children in the comparison group were assessed at their schools. All assessments were performed by specifically trained experimenters using standardized instructions. Written informed consent was obtained from all parents of the participating children. The medical ethics review board of the Erasmus University Medical Centre Rotterdam approved the study protocol.
Multiple linear regression analyses tested group differences between very preterm and comparison children for executive-functioning-dependent variables. Raw scores were used in all analyses. Missing data were handled by casewise deletion. We examined assumptions of normality, linearity, and homoscedasticity, by visual inspection of the residual scatterplots.25 For errors of commission and omission on the Stop task, and the Flanker task MRT interference score, the residual scatterplots deviated from a normal distribution owing to heteroscedasticity. However, the widest spread in SDs of residuals was not greater than three times the most narrow spread.25,26
Parental education (highest of the two parents), sex, and age, might have correlated with the executive functioning measures27,28 and were therefore entered as covariates in the analyses. Interaction effects with group were also inspected. Interaction effects with a significant change in the value of R2 (ΔR2) that did not reach the threshold for a small effect (0.01)29 were not interpreted. Analyses were conducted with and without adjustment for processing speed and fluctuations in speed, and IQ, and with and without inclusion of children with minor neurosensory dysfunctions. We calculated effect sizes in terms of standardized mean differences (SMD), which is the difference between two group means divided by an estimate of the within-group SD. Effect sizes of 0.2, 0.5, and 0.8, refer to small, medium, and large effects respectively.29p-values <0.05 (two-tailed) were considered significant. Analyses were performed with SPSS 17.0 (SPSS, Chicago, IL, USA).
Table I presents sample characteristics for the very preterm and term comparison group. Very preterm children had a significantly lower mean gestational age (p<0.001), lower mean birthweight (p<0.001), lower mean IQ (0.80 SMD, p<0.001), lower mean level of parental education (p<0.001), and more minor neurosensory dysfunctions (p<0.001) than comparison children. There were no group differences for sex (p=0.29) or age at assessment (p=0.81). One hundred and three children were 4 to 6 years of age, 79 children were 6 to 8 years of age, 107 children were 8 to 10 years of age, and 115 children were 10 to 12 years of age.
Group differences in executive functioning task performance
Missing data resulted from examiner error or child non-compliance. It varied from 2% for the Verbal Fluency task to 12% for the Switch task. Hardware problems resulted in missing data for the Spatial Span (<18%) and for the Stockings of Cambridge (<7%). Error scores were analysed for all participating children; however, for several children speed scores could not be interpreted reliably because of high error rates.30
Table II presents, for each dependent variable, the number of children included in the analyses, the means and standard errors for the very preterm and term comparison children, and group effects, in terms of unstandardized regression coefficients (B) and accompanying standard errors.
|Very preterm group (n=200)||Term comparison group (n=230)|
|Stop signal reaction time||179||316.3||7.6||211||82.1||5.5||37.1||8.5||<0.001|
|IS omission errors||184||0.8||0.2||219||0.3||0.1||0.5||0.2||0.063|
|IS commission errors||184||2.0||0.2||219||1.4||0.2||0.5||0.3||0.067|
|SS omission errors||189||0.3||0.2||224||0.8||0.1||0.1||0.1||0.356|
|SS commission errors||189||−0.6||0.4||224||−0.4||0.2||−0.1||0.4||0.747|
|Verbal working memory|
|Total correct sequencesa||200||3.7||0.1||222||4.1||0.1||−0.5||0.2||0.001|
|Spatial working memory|
|Total problems solved||187||5.9||0.2||213||6.3||0.1||−0.5||0.2||0.019|
There were no significant main effects of parental education. Main effects of sex were significant for the Stop task stop signal reaction time, omission and commission errors, and Stockings of Cambridge planning time (ts>2.28, ps<0.01; where ts represents t-values), with females outperforming males in both the very preterm and term comparison groups. There were no significant interactions between group and sex (ts<0.64, ps>0.05). Main effects of age were significant for all executive functioning dependent variables (ts>2.54, ps<0.02), indicating better performance with increasing age. Age interacted with group for stop signal reaction time (t=−2.37, p=0.02, ΔR2=0.02), showing a decrease of the group difference of 0.70 SMD, p<0.001, to 0.15 SMD, p>0.12, between 4 and 12 years of age.
Very preterm children had significantly poorer scores on the Stop task stop signal reaction time, omission and commission errors, on the Verbal Fluency total correct, Digit Span total correct sequences, Spatial Span maximum span, and Stockings of Cambridge planning time and problems solved (ts>−2.72, ps<0.007). Groups did not differ in Stockings of Cambridge execution time (t=0.20, p=0.84), Flanker task interference scores for MRT, errors of omission and errors of commission (ts<1.68, ps>0.10), and Switch task switch scores for MRT, errors of omission and errors of commission (ts<−1.51, ps>0.13).
Basic processing speed was significantly slower (0.40 SMD, t=5.06, p<0.001) and showed significantly greater fluctuations (0.70 SMD, t=7.00, p<0.001) in very preterm children than in term comparisons. There were no interaction effects between group and these speed indices. Except for omission errors on the Stop task (t=1.56, p=0.12), group differences remained unchanged if processing speed and fluctuations in speed were taken into account. In the analyses with IQ, group differences for dependent variables of the Digit Span and Spatial Span, however, were no longer significant. Analyses with and without inclusion of children with neurosensory dysfunctions revealed similar results.
Figure 1 displays the SMDs for executive functioning adjusted for covariates and speed indices, in a profile with the comparison group as the reference group (SMD 0.0).
This study assessed executive functioning in a large sample of very preterm and term comparison children aged 4 to 12 years to study how executive functioning deficits in this sample are in the proportion of each other, whether these deficits are persistent over time, and their dependency on processing speed and fluctuations in speed.
The results show that, consistent with previous research,3,11 very preterm children perform poorer than term children on executive functioning measures with effect sizes ranging from small (0.3 SMD for working memory) to moderate (0.5 SMD for verbal fluency). Results add to our previous study on this issue31 as well as to studies by other researchers (for an overview see Aarnoudse-Moens et al.3 and Mulder et al.11) in that we found that very preterm children catch up with peers in response inhibition, but stay behind in neurocognitive functions such as fluency, planning, and working memory. In addition, we once more demonstrated that executive functioning deficits cannot be explained by slow and highly fluctuating processing speed nor by lower IQ.31 Results remained unchanged if very preterm children with neurosensory dysfunctions were excluded from the analyses.
Our very preterm sample did not perform poorer than comparisons on measures of interference control and stimulus–response switching. The results for interference control converge with earlier research showing that very preterm children do not perform slower and do not make more errors if faced with interfering information.32,33 However, the results for switching contrast with previous studies. For instance, across studies with very preterm children assessing switching with the Trail Making Test part B, a moderate effect size has been described, whereas we did not find a significant effect of very preterm birth.3 However, differences between these studies and our results are likely due to differences in measures used. The Trail Making Test part B, in contrast to our switch measure, draws heavily on visual–spatial abilities that are frequently observed to be impaired in very preterm children.34,35 and thereby may bias switching effects. We also assessed inhibitory control as it has been considered the core deficit underlying attention disorders,5 one of the major adverse outcomes of very preterm birth,3 nevertheless only scarcely examined in this population. The Stop task allows measurement of the covert inhibitory process in the brain (i.e. stop signal reaction time) isolated from basic measures of information processing. Findings showed that, at early school age, very preterm children have significantly poorer inhibitory processes than same-aged children born at term, but that group differences between very preterm and term children disappear at middle school age. These findings suggest that poor inhibitory skills in very preterm children represent a maturational lag, although future research should replicate this finding.
The large sample size across the wide age range of 4 to 12 years included is not often seen in studies of executive functioning in very preterm children. Nevertheless, including 4- and 5-year-olds in such a study means assessing executive functions that have just begun to emerge. Several of our preschoolers did not comply with task requirements or were impacted by difficulties with response buttons and touch-screen technology. However, more than two-thirds of the very preterm and comparison children were able to accomplish the tasks, which makes our findings on the progress of executive functioning development in very preterm children compared with that in term children reliable.
A limitation was that, although term children were recruited from the same schools as those attended by very preterm children to control for educational environmental characteristics, level of parental education was higher for term children than for very preterm children, possibly because highly educated parents are more willing to participate. Because there were no interactions between group and parental education, we adjusted for the influence of parental education in the analyses. Another limitation was that assessments were done by experimenters who were not blinded to preterm birth status. However, the experimenters were specifically trained for the purposes of the study and used standardized instructions.
In conclusion, relative to term peers, very preterm children who are free of major disabilities and with IQs in the average range performed normally on interference control and switching measures, but poorly on measures tapping into response inhibition, verbal and spatial working memory, verbal fluency, and planning; these deficits could not be explained by these children’s slow and highly fluctuating processing speed nor by their lower IQ. The important ‘take-home’ message is that executive dysfunction in these children is not a global deficit; rather, it constitutes a unique profile of affected and non-affected areas, which remains largely consistent between 4 and 12 years. It is the limited capacity or span to store and flexibly use information temporarily, yet on top of slow and highly fluctuating speed, that hinders these children and may cause a cascade of other neurocognitive deficits. For instance, the inattentiveness so frequently observed in very preterm children in classrooms,3 or their lack of cognitive flexibility, may thus rather reflect their limited speed and stability to process and manipulate incoming stimuli than real interference control or switching problems. Applying the present results, clinicians and researchers working with very preterm children might ensure that executive functions are tapped as ‘purely’ as possible and select executive functioning tasks that are minimally dependent on other neurocognitive skills such as visual-spatial skills or processing speed. In addition, using IQ scores as an indicator of a child’s neurocognitive functioning may not provide sufficient insight into the child’s strengths and weaknesses.
The executive functioning profile associated with very preterm birth as highlighted in this study supports remediation programmes tailored to children of this population. These children’s deficits in executive functioning, in addition to their slow and highly fluctuating response style, may affect their academic achievement, as well as cause attention disorders, which is a subject of our future research. Timely intervention, such as preschool programme ‘Tools of the Mind’,36 trying to help very preterm children overcome their executive functioning difficulties is necessary to prevent the onset of academic and behavioural problems.
- 4Precursors to mathematical skills: examining the roles of visual–spatial skills, executive processes, and parenting factors. Appl Dev Sci 2003; 7: 27–38., , , , .
- 16Development of preschool and academic skills in children born very preterm. J Pediatr 2010; 11: 51–6., , , , .
- 18Impulsivity and inhibitory control. Psychol Sci 1997; 8: 60–4., , .Direct Link:
- 21Wechsler Intelligence Scale for Children-III-NL. 3rd edn. Amsterdam, The Netherlands: Harcourt Test Publishers, 2005..
- 23Wechsler Preschool and Primary Scale of Intelligence-Revised (WPPSI-R). San Antonio, TX: The Psychological Corporation, 1997..
- 24Assessment of Children, WISC-III and WPPSI-R Supplement. San Diego, CA: JM Sattler, 1992..
- 25Using Multivariate Statistics. 4th edn. Needham Heights, MA: Allyn & Bacon, 2001., .
- 26Regression Diagnostics. Newbury Park, CA: Sage Publications, 1991..
- 30Preparation, Adjustment, and Inhibition of Responses. the Netherlands: University of Amsterdam, 1997..