Objective

To investigate whether correction for prematurity affects executive function scores in school-aged children born very preterm.

Study design

Executive functions were assessed with standardized neuropsychological tests in 142 very preterm (born at ≤32 weeks of gestational age or with a birth weight of ≤ 1500g) and 391 control children, aged 7 to 13 years. Four-month age-bands were established from the data of control children. Differences between uncorrected and corrected scores were compared against zero difference and between very preterm children born before and after 28 weeks of gestation. Regression models were used to compare very preterm participants’ uncorrected and corrected scores to controls.

Results

For all executive functions, significant, larger-than-zero differences between uncorrected and corrected scores were apparent in children born very preterm. Mean differences ranged from 0.04 to 0.18 standard deviations. Weak evidence was found that the effect of age correction is more pronounced in very preterm children born before 28 weeks of gestation than in those born after. Differences in executive function scores between very preterm and control children were attenuated if scores were corrected for prematurity.

Conclusion

Test scores based on corrected rather than uncorrected age may more accurately determine the developmental stage of very preterm children’s executive functions at school age. Potential consequences for clinical and research practice need to be discussed in the future.

Key WordsTo assess the neurodevelopmental status of very preterm survivors (ie, those born below 32 weeks of gestation), clinical guidelines commonly recommend correcting for prematurity in the first few years of life (e.g.,[1]Doyle L.W. Anderson P.J. Battin M. Bowen J.R. Brown N. Callanan C. et al.Long term follow up of high risk children: who, why and how?.,[2]Adams M. Borradori-Tolsa C. Bickle-Graz M. Grunt S. Weber P. Capone Mori A. et al.Follow-up assessment of high-risk newborns in Switzerland.). Clinicians usually do this by subtracting the number of weeks and days a child was born prematurely from the child’s chronological age. These recommendations are based on and supported by numerous studies that report a substantial difference between test scores based on preterm infants’ uncorrected and corrected age when assessing motor and cognitive development3Morsan V. Fantoni C. Tallandini M.A. Age correction in cognitive, linguistic, and motor domains for infants born preterm: an analysis of the Bayley Scales of Infant and Toddler Development, developmental patterns., 4Harel-Gadassi A. Friedlander E. Yaari M. Bar-Oz B. Eventov-Friedman S. Mankuta D. et al.Developmental assessment of preterm infants: Chronological or corrected age?., 5Formiga C.K.M.R. Vieira M.E.B. Linhares M.B.M. Developmental assessment of infants born preterm: comparison between the chronological and corrected ages., 6Restiffe A.P. Gherpelli J.L.D. Comparison of chronological and corrected ages in the gross motor assessment of low-risk preterm infants during the first year of life., 7Is It Correct to Correct for Prematurity? Theoretic Analysis of the Bayley-4 Normative Data.. The effect of correcting for prematurity has been reported to be largest for those born extremely preterm but was also apparent in moderately and late preterm infants[8]Parekh S.A. Boyle E.M. Guy A. Blaggan S. Manktelow B.N. Wolke D. et al.Correcting for prematurity affects developmental test scores in infants born late and moderately preterm.. Whereas most studies focus on the effects of correcting for prematurity during the first few years of life, these effects may remain relevant well beyond infancy and toddlerhood. In fact, theoretical models and empirical evidence suggest that prematurity should be considered across childhood and into adolescence to accurately estimate cognitive functioning9van Veen S. Aarnoudse-Moens C.S. van Kaam A.H. Oosterlaan J. van Wassenaer-Leemhuis A.G. Consequences of correcting intelligence quotient for prematurity at age 5 years., 10Wilson‐Ching M. Pascoe L. Doyle L.W. Anderson P.J. Effects of correcting for prematurity on cognitive test scores in childhood., 11Roberts R.M. George W.M. Cole C. Marshall P. Ellison V. Fabel H. The Effect of Age-Correction on IQ Scores among School-Aged Children Born Preterm..The need to correct for prematurity may also vary across developmental domains[12]

Romeo D. Correcting for prematurity with the Bayley Scales of Infant Development. 2018.

. To date, the effect of age correction in the domain of executive functions, a set of higher-order cognitive abilities that help direct goal-oriented behaviour[13]Assessment and development of executive function (EF) during childhood., has not yet been assessed systematically. This is despite executive functions being among the cognitive abilities most frequently affected in children born very preterm[14]

Brydges CR, Landes JK, Reid CL, Campbell C, French N, Anderson M. Cognitive outcomes in children and adolescents born very preterm: a meta‐analysis. Developmental Medicine & Child Neurology. 2018.

,[15]van Houdt C.A. Oosterlaan J. van Wassenaer‐Leemhuis A.G. van Kaam A.H. Aarnoudse‐Moens C.S. Executive function deficits in children born preterm or at low birthweight: a meta‐analysis.. Moreover, executive functions, which include inhibition, cognitive flexibility, working memory, and processing speed, develop rapidly across childhood and into adolescence[13]Assessment and development of executive function (EF) during childhood.,[16]A developmental perspective on executive function., so correction for prematurity may remain relevant throughout child development.

Consequently, this study aims to investigate the effect of age correction on executive function scores at school age. Larger-than-zero differences between uncorrected and corrected test scores are expected for very preterm children, and this difference is hypothesized to be larger in very preterm children born before 28 weeks of gestation than in those born after. Further, differences between children born very preterm and typically developing children are hypothesized to diminish once prematurity has been corrected for.

Method

This study was a secondary data analysis drawing on data assessed in various clinical studies conducted at the University Children's Hospital Zurich and the Children’s University Hospital Bern and on data assessed as part of a community sample study. The pooling of data from several studies allowed the assembly of a large dataset of control participants—a prerequisite for the establishment of the narrow age-bands necessary to investigate the effect of correcting for prematurity.

The data on children born very preterm stem from 2 Swiss studies. The EpoKids study is an ongoing follow-up study investigating the long-term effect of early erythropoietin administration on executive functions in children born between 26 and 32 weeks of gestational age[17]Wehrle F.M. Held U. O’Gorman R.T. Disselhoff V. Schnider B. Fauchère J.-C. et al.Long-term neuroprotective effect of erythropoietin on executive functions in very preterm children (EpoKids): protocol of a prospective follow-up study.. The current analysis uses data on the subgroup of children who were assessed between 2017 and 2019 (for details, see[18]Schnider B. Tuura R. Disselhoff V. Latal B. Wehrle F.M. Hagmann C.F. Altered brain metabolism contributes to executive function deficits in school-aged children born very preterm.). The Neuropsychology and Memory (NEMO) research programme was run between 2011 and 2015 and assessed cognitive development and training-induced improvements of cognitive abilities in school-aged children born before 32 weeks of gestational age or with a birth weight below 1500 grams[19]Ritter B.C. Nelle M. Perrig W. Steinlin M. Everts R. Executive functions of children born very preterm - deficit or delay?.. The current study includes pre-training data. In total, 142 children born very preterm or with low birth weight were included (EpoKids study: 86, NEMO research program: 56). All very preterm children were between 8 years 0 months and 13 years 5 months at the time of the assessment (uncorrected age; corrected ages: between 7 years 10 months and 13 years 2 months).For the control group, data were assembled from several ongoing and completed studies conducted at the University Children's Hospital Zurich and the Children’s University Hospital Bern. Control participants were selected if they were within the age range of the very preterm participants and if the parents reported birth at term (i.e., ≥37 weeks and 0 days of gestation), no neonatal complications, no neurodevelopmental or neurologic illness past or present (e.g., attention-deficit/hyperactivity disorder, epilepsy), and no learning disabilities. Inclusion criteria were met by 108 control children from the EpoKids study, 46 from the NEMO research programme, 36 from the Brainfit study[20]Benzing V. Eggenberger N. Spitzhüttl J. Siegwart V. Pastore-Wapp M. Kiefer C. et al.The Brainfit study: efficacy of cognitive training and exergaming in pediatric cancer survivors–a randomized controlled trial., 19 from the Hemispheric Reorganization (HERO) study[21]Kornfeld S. Rodríguez J.A.D. Everts R. Kaelin-Lang A. Wiest R. Weisstanner C. et al.Cortical reorganisation of cerebral networks after childhood stroke: impact on outcome., 19 from the TeenHeart study[22]Ehrler M. Naef N. Tuura R.O.G. Latal B. Executive function and brain development in adolescents with severe congenital heart disease (Teen Heart Study): protocol of a prospective cohort study., and 23 from unpublished datasets assessed at the University Children’s Hospital Zurich.Additionally, data from the Nathan Kline Institute–Rockland Sample (NKI–RS) was included in the dataset. The NKI–RS is an ongoing, institutionally-centered community sample of participants across the lifespan[23]Nooner K.B. Colcombe S. Tobe R. Mennes M. Benedict M. Moreno A. et al.The NKI-Rockland sample: a model for accelerating the pace of discovery science in psychiatry.. From the NKI–RS sample, data was selected on children who were within the same age range as the very preterm children. Data was excluded on children for whom birth defects, serious head injury, migraine, meningitis, genetic disorders (e.g., Huntington disease), psychiatric disorders (e.g., anorexia, autism or ADHD), heart diseases, cancer, learning difficulties (e.g., reading problems), or an IQ below 85 was reported. The NKI–RS dataset does not document gestational age. In total, 140 participants of the NKI–RS were selected. The final control group thus consisted of 391 participants.

All studies were approved by either the ethical committees of the Cantons of Zurich or Bern, Switzerland, the ethical committee of the Children’s University Hospital Bern, Switzerland, or the Nathan Kline Institute and Montclair State University, NY, USA. Parents provided written informed consent before participation.

 Executive function assessment

The majority of the primary studies used the same tasks to assess four executive functions: inhibition, cognitive flexibility, working memory, and processing speed. These were thus the tasks selected for the analyses reported here.

In the Color–Word Interference Tasks (CWIT [Delis-Kaplan Executive Function System; D-KEFS][24]

Delis DC, Kaplan E, Kramer JH. Delis-Kaplan executive function system (D-KEFS): Psychological Corporation; 2001.

), the child is asked to name colour patches (Condition [C] 1) and read colour words (C2). C1 and C2 assess processing speed. Then, the child is asked to name the ink colour of an incongruent colour word (C3: inhibition) and to switch between naming the ink colour and reading the colour word if indicated by a box around the word (C4: cognitive flexibility).In the Trail Making Task (TMT [D-KEFS][24]

Delis DC, Kaplan E, Kramer JH. Delis-Kaplan executive function system (D-KEFS): Psychological Corporation; 2001.

), the child is asked to connect numbers distributed across a sheet of paper in ascending order from 1 to 16 (C2: processing speed) and to switch between connecting numbers and letters in ascending order (C4: cognitive flexibility). For all D-KEFS tasks, the completion time (in seconds) was used as dependent variable.The Digit Span subtest of the Wechsler Intelligence Scale for Children, Fourth Edition (WISC-IV, German Version[25]

Petermann F, Petermann U. Hamburg-Wechsler Intelligenztest für Kinder-IV (HAWIK-IV) [Hamburg-Wechsler-Intelligence Test for children (HAWIK-IV)]. Bern: Huber; 2006.

) was used to assess working memory. The child repeats increasing sequences of numbers in the same or reversed order as presented by the examiner. Sum scores of the forward and backward conditions were both used as dependent variables.In the primary studies, the tasks that assess executive functions were applied as components of more comprehensive assessment protocols of neurodevelopmental outcome. Some of the study protocols also included MR imaging of the brain or an EEG assessment. All tasks were administered by trained examiners and in accordance with the standardized instructions provided by the test manuals[24]

Delis DC, Kaplan E, Kramer JH. Delis-Kaplan executive function system (D-KEFS): Psychological Corporation; 2001.

,[25]

Petermann F, Petermann U. Hamburg-Wechsler Intelligenztest für Kinder-IV (HAWIK-IV) [Hamburg-Wechsler-Intelligence Test for children (HAWIK-IV)]. Bern: Huber; 2006.

to ensure reliable data assessment.

Mean education level of the mother and father at the time of assessment was used as a proxy for socio-economic status because this information was assessed in all the primary studies. The scales to assess parental education level were harmonized between studies by defining the following categories: 1 = no high school graduation, 2 = high school graduation/apprenticeship, 3 = college graduation, and 4 = university degree.

 Calculation of age-band specific scoresThe 1-year age-bands provided by the D-KEFS[24]

Delis DC, Kaplan E, Kramer JH. Delis-Kaplan executive function system (D-KEFS): Psychological Corporation; 2001.

are too broad to investigate the effect of age correction, because only preterm children tested very close to their birthdays fall into younger age categories. In contrast, two-month age bands would place all children born very preterm (i.e., n = 391), the respective comparison groups would have been very small with an average of 11.8 control children per category. Pragmatically, and in accordance with the age-bands provided by the WISC-IV[25]

Petermann F, Petermann U. Hamburg-Wechsler Intelligenztest für Kinder-IV (HAWIK-IV) [Hamburg-Wechsler-Intelligence Test for children (HAWIK-IV)]. Bern: Huber; 2006.

, 4-month age-bands were constructed for the D-KEFS tasks from the data of the control participants. Each control participant was assigned to the appropriate age category, resulting in an average of 21.7 controls per category (number of control participants per age category presented in Figure 1 [available at www.jpeds.com]). Individual raw scores were then transformed into age-band specific scores as follows:

(raw score of each participant) – (mean of all control participants of the same age category)

(SD of all control participants of the same age category)

z =

For the very preterm participants, this transformation was performed twice: once with the scores based on uncorrected age and once with scores based on corrected age at assessment.

The WISC-IV Digit Span test was only applied in a subset of the primary studies, leading to a considerably smaller control group for this task (n = 196). This precluded the establishment of new age-band specific scores from the data on the control group. Instead, the norms of the WISC-IV manual [25]

Petermann F, Petermann U. Hamburg-Wechsler Intelligenztest für Kinder-IV (HAWIK-IV) [Hamburg-Wechsler-Intelligence Test for children (HAWIK-IV)]. Bern: Huber; 2006.

were used: The sum scores of the forward and backward condition were each transformed into scaled scores according to the age norms provided by the WISC-IV manual. For the very preterm children, again, this was done once with the scores based on uncorrected age and once with the scores based on corrected age at assessment. For better comparison with the D-KEFS tasks, the scaled scores were then transformed into z-scores. Statistical analyses

Demographic variables were compared between groups by Chi-square and independent t-tests as appropriate. For very preterm individuals, the differences between uncorrected and corrected scores were calculated (i.e., corrected – uncorrected score), with positive values indicating that scores based on corrected age are higher than scores based on uncorrected age. These difference scores were compared against a difference of zero by one-sample t-tests. Further, they were compared between children born before and after 28 weeks of gestation by independent t-tests to investigate whether the effect of age correction is associated with the degree of prematurity. To assess whether the effect of correcting for prematurity varies with age at assessment, the difference scores were correlated with age at assessment using Spearman correlation.

Group differences in executive function scores between children born very preterm and controls were investigated using linear regression models, adjusting for parental education level. Age at assessment was not included into the models because executive function z-scores were calculated separately for each age band, zeroing out the effect of age at assessment. Each executive function task was defined as dependent variable in a separate model, once using scores based on uncorrected age and once using scores based on corrected age. To account for missing data in the covariate parental education level, multiple imputation was conducted with Multivariate Imputation by Chained Equation[26]

Buuren Sv, Groothuis-Oudshoorn K. mice: Multivariate imputation by chained equations in R. Journal of statistical software. 2010:1-68.

with five imputations and 50 iterations. Regression models were then performed with pooled mean and standard error of parental education level. Effect sizes of group differences between children born very preterm and controls were calculated by converting the F-statistics of the linear regression models to d, i.e., standardized mean difference between groups while taking into account parental education [27]

Del Re A. compute. es: Compute effect sizes. R package version. 2013:0.2-.

. Effect sizes of 0.2, 0.5, and 0.8 were interpreted as small, medium, and large effects, respectively. To investigate whether group differences between children born very preterm and controls are present in younger and older children, the analyses were repeated in two subgroups: Subgroups were split at 10 years of age (uncorrected age) because in one of the primary studies, namely, the NEMO research program, it has previously been shown that group differences in executive functions were only present in children younger than age 10 years and no differences were apparent after 10 years of age[19]Ritter B.C. Nelle M. Perrig W. Steinlin M. Everts R. Executive functions of children born very preterm - deficit or delay?.. The significance level was set at p ≤ .05. Statistical analyses were performed with IBM SPSS Statistics 24 and R statistical software, Version 4.0.229.Results Participant characteristicsThe very preterm and control participant groups were comparable with regard to sex and uncorrected and corrected age at assessment. By design, gestational age was lower in very preterm participants. Parental education level was higher in families of control participants (Table 1 and Table II [available at www.jpeds.com]) for sample characteristics of very preterm children below and above 28 weeks of gestation). In the very preterm group, year of birth was associated with processing speed (TMT C2; r = .22, p = .013) and working memory (Digit Span forward: r = -.19, p = .031; Digit Span backward: r = .27, p = .002) but not with the other measures of executive functions (r-values ranging from -.08 to .09, all p > .05).

Table 1Sample characteristics of very preterm and control participants.

Note. VPT: very preterm; t: independent t-test (continuous variables); χ[2]Adams M. Borradori-Tolsa C. Bickle-Graz M. Grunt S. Weber P. Capone Mori A. et al.Follow-up assessment of high-risk newborns in Switzerland.: chi square (categorical variable). aExact gestational age was available for control participants of the EpoKids study only (n = 106). bMean education level of the mother and the father (1 = no high school graduation, 2 = high school graduation/apprenticeship, 3 = college graduation, 4 =university degree). For two VPT participants and of 26 control participants, only the mother’s or the father’s education level was available and was used alone for analyses. cFor four VPT participants, information on parental education level was missing. dFor 11 control participants, information on parental education level was missing.

Difference between scores based on uncorrected and corrected age in children born very preterm

When age at assessment was corrected for prematurity, 56.7% of very preterm children fell into a younger age category. Figure 2 presents uncorrected and corrected scores (for exact means and standard deviations, please refer to Table 4, columns ‘VPT group’). Across the whole very preterm group, the differences between scores based on uncorrected and corrected age were significantly larger than zero in all executive function tasks except for one processing speed task (TMT C2; Table 3). The mean differences ranged from z = .04 to z = .18, with z ± 1.0 equalling one standard deviation. The difference score for working memory (Digit Span backward) was significantly correlated with age at assessment (r = .18, p = .04). None of the other difference scores were associated with age at assessment with r-values ranging from -.121 to .119 (all P > .05).

Figure 2Executive function scores based on uncorrected and corrected age in children born very preterm. Note: The symbols represent mean z-scores of the very preterm group in the different tasks (left: scores based on uncorrected age, right: scores based on corrected age; the dotted connecting line is for better visualization only). Horizontal lines illustrate the means of the control group. For the working memory scores, the norms of the WISC-IV test manual were used and transformed into z-scores for better comparison with the other tasks. The mean of the control group is -.03 for Digit Span forward and .34 for Digit Span backward. For all other tasks, the mean of the control group equals zero. Please refer to the text for further details. C: Condition; CWIT: Color–Word Interference Task; TMT: Trail Making Task; VPT: Very preterm.

Table 4Executive functions (z-scores) of very preterm and control participants and results of group comparisons using linear regression models.

Note: C: Condition; CWIT: Color–Word Interference Task; VPT: very preterm; TMT: Trail Making Task. acontrol/very preterm. bB (standard error) and p- values of predictor group in the linear regression models accounting for parental education level. cd = standardized mean difference between groups while taking into account parental education level. dDigit Span data was available only for a subgroup of control children. Consequently, scores were transformed into scal

Table 3Difference between executive function scores based on uncorrected and corrected age in children born very preterm (total group and children born before and after 28 weeks of gestation separately).

Note. C: Condition; CWIT: Color–Word Interference Task; GA: gestational age; VPT: very preterm; TMT: Trail Making Task. aDifference between z-scores based on uncorrected and corrected age (i.e., corrected – uncorrected scores; positive values indicate that scores based on corrected age are higher than scores based on uncorrected age); btotal sample is compared against zero (one-sample t-test); children born before and after 28 weeks of gestation are compared against each other (independent sample t-test).

The differences between uncorrected and corrected scores were significantly larger in children born before 28 weeks of gestation than in children born after 28 weeks of gestation in two of the three processing speed tasks (CWIT C1 and C2). In all other tasks, the difference scores were comparable between the two prematurity groups (p > .05, Table 3).

The effect of correcting for prematurity on differences in executive function scores between very preterm and control children

Without correcting for prematurity, regression models adjusted for parental education level revealed significantly lower scores in very preterm participants than in controls in two of the three processing speed tasks (CWIT C1: d = .29, p = .001; TMT C2: d = .51, pd = .28, p = .030) and one of the two cognitive flexibility tasks (TMT C4: d = .40, p = .001). Groups did not significantly differ in any of the other tasks (d = .17 - .38, all p > .05, Table 4). Repeating the analyses for younger and older children separately revealed similar results compared with the whole group analyses (Table 5; available at www.jpeds.com).When the scores based on corrected age were compared between groups, very preterm participants showed significantly lower scores in one of the three processing speed tasks (TMT C2: d = .47, p = .001) and in one of the two cognitive flexibility tasks (TMT C4: d = .36, p = .012). In all other executive functions, scores were comparable between groups (d = .16 – .36, p > .05, Table 4).DiscussionChildren born very preterm exhibited larger-than-zero mean difference between test scores based on uncorrected and corrected age in all the executive functions assessed: inhibition, cognitive flexibility, working memory, and processing speed. The difference ranged from 0.04 to 0.18 of a standard deviation. This effect size is comparable with what has previously been reported for estimates of cognitive functioning: At school age, one theoretical study and empirical results found a mean difference of 1 to 5 IQ points depending on whether prematurity was corrected for or not[10]Wilson‐Ching M. Pascoe L. Doyle L.W. Anderson P.J. Effects of correcting for prematurity on cognitive test scores in childhood.,[11]Roberts R.M. George W.M. Cole C. Marshall P. Ellison V. Fabel H. The Effect of Age-Correction on IQ Scores among School-Aged Children Born Preterm.. The relevance of such small effects has been debated: It has been argued that a clinical decision to correct for prematurity will depend on the specific context and on the potential benefit for the child, whereas for research purposes, age should always be corrected for to take a known bias into account[30]Do we need to correct age for prematurity when assessing children?..The current study found weak evidence that the effect of age correction is more pronounced in very preterm children born before 28 weeks of gestation than in those born after. This is in contrast to previous studies that report greater differences between test scores based on uncorrected and corrected age at lower gestational ages when assessing motor and cognitive functioning of preterm individuals[3]Morsan V. Fantoni C. Tallandini M.A. Age correction in cognitive, linguistic, and motor domains for infants born preterm: an analysis of the Bayley Scales of Infant and Toddler Development, developmental patterns.,[4]Harel-Gadassi A. Friedlander E. Yaari M. Bar-Oz B. Eventov-Friedman S. Mankuta D. et al.Developmental assessment of preterm infants: Chronological or corrected age?.,[9]van Veen S. Aarnoudse-Moens C.S. van Kaam A.H. Oosterlaan J. van Wassenaer-Leemhuis A.G. Consequences of correcting intelligence quotient for prematurity at age 5 years.. These studies reported effects to be largest in extremely premature children born at and below 28 weeks of gestation. In the current study, two thirds of the very preterm individuals were born between 28 and 32 weeks of gestation, and only very few were born before 26 weeks of gestation. This likely reduced the association between the degree of prematurity and the effect of age correction.Executive function scores of very preterm children were found to be lower compared with typically developing children if chronological age was used. This is in line with previous studies that report executive functions to be impaired in school-aged children born very preterm[14]

Brydges CR, Landes JK, Reid CL, Campbell C, French N, Anderson M. Cognitive outcomes in children and adolescents born very preterm: a meta‐analysis. Developmental Medicine & Child Neurology. 2018.

,[31]Stålnacke J. Lundequist A. Böhm B. Forssberg H. Smedler A.-C. A longitudinal model of executive function development from birth through adolescence in children born very or extremely preterm.. Importantly, the current findings provide evidence that these group differences were attenuated if the age at assessment was corrected for prematurity. Although the findings suggest a need to consider prematurity throughout the development of very preterm children to adequately assess executive functions, further studies are needed to confirm this. Meanwhile, studies may report test scores based on uncorrected age alongside test scores based on corrected age. This may support the interpretation of group differences between children born very preterm and at term. Ultimately, such data will provide the evidence needed to develop recommendations for clinicians, schools and others involved in the care of children born very preterm. Besides correcting executive function test scores for prematurity, employing motor-free tests may also contribute to a more accurate assessment as underlying motor impairments may