Executive function deficits in children born preterm or at low birthweight: a meta-analysis.

AIM: To investigate the magnitude of executive function deficits and their dependency on gestational age, sex, age at assessment, and year of birth for children born preterm and/or at low birthweight.
METHOD: PubMed, PsychINFO, Web of Science, and ERIC were searched for studies reporting on executive functions in children born preterm/low birthweight and term controls born in 1990 and later, assessed at a mean age of 4 years or higher. Studies were included if five or more studies reported on the same executive function measures.
RESULTS: Thirty-five studies (3360 children born preterm/low birthweight, 2812 controls) were included. Children born preterm/low birthweight performed 0.5 standardized mean difference (SMD) lower on working memory and cognitive flexibility and 0.4 SMD lower on inhibition. SMDs for these executive functions did not significantly differ from each other. Meta-regression showed that heterogeneity in SMDs for working memory and inhibition could not be explained by study differences in gestational age, sex, age at assessment, or year of birth.
INTERPRETATION: Children born preterm/low birthweight since 1990 perform half a SMD below term-born peers on executive function, which does not seem to improve with more recent advances in medical care or with increasing age. WHAT THIS PAPER ADDS: Children born preterm/low birthweight perform below term-born children on core executive functions. Lower gestational age or male sex are not risk factors for poorer executive functions. Executive function difficulties in children born preterm/low birthweight remain stable across childhood. Executive function difficulties are similar for children born recently and children born in earlier eras.

Digit Span Task

Standardized mean difference

Preterm birth frequently occurs all over the world. Of all live‐born <span class="Species">children in the USA in 2015, for instance, 9.6% were born preterm (gestational age <37wks, according to the World Health Organization definition)1 and 8.1% were born with a low birthweight (<2500g, according to the World Health Organization definition).1, 2 Preterm birth and low birthweight co‐occur frequently, with 69.2% of <span class="Species">children born preterm also being born with a low birthweight and 49.8% of <span class="Species">children born with a low birthweight also being born preterm.3 Preterm birth and/or low birthweight survivors are at high risk of adverse cognitive, academic, and behavioural outcomes.1, 4, 5 <span class="Species">Children born preterm/low birthweight have 0.8 standard deviation (SD) lower IQ scores and perform about a 0.5 SD poorer than term‐born peers on mathematics, reading, and spelling tests.4, 5 Also, a two to four times higher risk of being diagnosed with attention‐deficit/hyperactivity disorder than for term‐born peers has been reported.4, 6, 7 A large body of studies has also shown impairments in the so‐called executive functions in <span class="Species">children born preterm/low birthweight.8, 9 ‘Executive functions’ is an umbrella term for a set of higher‐order cognitive functions, with core functions including <span class="Disease">working memory, inhibition, and <span class="Disease">cognitive flexibility, which allow for top‐down, goal‐directed behaviour.10, 11 Executive functions rely upon lower‐order cognitive processes, such as processing speed, to operate effectively.12

Executive functions are increasingly studied because of their crucial role in the onset of academic and behavioural problems.13, 14 Even as early as the toddler and preschool years, executive functions are predictive of both (pre‐)academic skills and behaviour problems.15, 16 Importantly, executive function deficits have also been shown to be key to the behavioural and academic problems observed in <span class="Species">children born preterm/low birthweight.12, 17, 18, 19, 20, 21, 22, 23 Recent research showed that executive function is a substantially better predictor of poor behavioural and academic outcomes in <span class="Species">children born preterm/low birthweight than IQ and motor functions.12, 17, 18, 19, 20, 21, 22, 23, 24 For example, measures of executive functions are highly predictive of mathematic and reading abilities, and attention regulation in <span class="Species">children born very preterm and/or very low birthweight.12, 17, 22, 23

The initial literature on executive functions in <span class="Species">children born preterm/low birthweight was summarized in two meta‐studies published in 2009, reporting a 0.36 standardized mean difference (SMD) for <span class="Disease">working memory, a 0.25 SMD for inhibition, and a 0.49 to 0.50 SMD for <span class="Disease">cognitive flexibility between <span class="Species">children born preterm/low birthweight and term‐born comparison <span class="Species">children.8, 9 Newly published literature on executive functions since the meta‐studies published in 2009 warrants an update of this work. In addition, a meta‐analysis on cognitive function, including executive functions, in <span class="Species">children born very preterm was published recently.25 However, this paper only focused on <span class="Species">children born at less than 32 weeks of gestation and piled the diverse subdomains within the broad concept of executive functions. Including both newly published studies on executive functions since 2009 and studies into executive functions in <span class="Species">children across the entire range of preterm birth (<37wks gestational age) offers the possibility of improving on previous meta‐analytic work by examining the profile of executive function difficulties, i.e. whether specific executive function domains are more severely impaired than others, and whether the magnitude of the effect sizes depend on the degree of preterm birth (gestational age), sex,26 age at assessment,27, 28, 29 and year of birth (as a proxy measure of advances in neonatal care).30, 31 Owing to slightly different inclusion criteria and the large number of studies published on this subject since 2009, only two studies included in this meta‐analysis were included in the 2009 meta‐analyses.32, 33

Using meta‐regression, this quantitative meta‐analysis aimed to aggregate and quantify impairments in executive function domains and assess the impact of gestational age, sex, age at assessment, and year of birth on executive function effect sizes. This unravels the nature and extent of executive function impairments in the population born preterm/low birthweight and contributes to a better understanding of the behavioural and academic problems observed in <span class="Species">children born preterm/low birthweight.

Method

Our (unpublished) protocol was performed according to Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines.34

Search strategy

A literature search was performed by one author (CAvH) in PubMed, PsychINFO, Web of Science, and ERIC, on 16th January 2017, using search terms concerning the birth status (e.g. preterm, low birthweight, small for gestational age), executive function measures (<span class="Disease">working memory, inhibition, <span class="Disease">cognitive flexibility), and age group (<span class="Species">child, adolescent, teenager, young adult, adult, middle aged). An experienced librarian was consulted for construction of the search terms, which are provided in (Appendix S1).

Study selection

Inclusion criteria for study selection were as follows: (1) the study included <span class="Species">participants born preterm (<37wks gestation) and/or with low birthweight (<2500g) and a comparison group of term‐born, typical‐birthweight <span class="Species">participants (>37wks gestation and >2500g); (2) the mean age at assessment of the <span class="Species">participants was at least 4 years (studies with younger <span class="Species">participants were not included in this meta‐analysis as executive functions cannot be reliably assessed before the age of 4y);35 (3) the year of birth of the <span class="Species">participants was 1990 or later (i.e. after the introduction of antenatal <span class="Chemical">steroids and surfactant supplementation); (4) the study reported administration of <span class="Disease">working memory, inhibition, and/or <span class="Disease">cognitive flexibility tasks; and (5) the study was published in an English‐language, peer‐reviewed journal.

There are many different tasks available to measure executive functions and some executive function tasks may have been used in only one or two studies. Meta‐analytic procedures can be applied to a small number of studies; however, the results obtained might then be unstable.36 Therefore, to maximize the robustness of findings, only executive function tasks reported in at least five papers, as was done in a previous meta‐analysis,8 were included. Papers were thorougly checked for overlapping cohorts of <span class="Species">children born preterm/low birthweight. If multiple papers reported on overlapping cohorts of <span class="Species">children born preterm/low birthweight on the same executive function domain, the study with the most complete data for that domain was selected. When multiple papers reported on overlapping cohorts of <span class="Species">children born preterm/low birthweight but the papers differed in the executive function domains described, all these papers were included in analyses. If it was not clear whether cohorts were overlapping, the authors of the studies were contacted. Screening of titles and abstracts and assessment of full‐text articles was performed by two authors (CAvH, CSHAM).

Measures

This meta‐analysis reports the results for the following executive function tasks; results for each of these tasks were reported in at least five papers.

Working memory

<span class="Disease">Working memory is the ability to hold information in mind and actively manipulate this information.10 <span class="Disease">Working memory comprises a verbal and a visual–spatial subsystem. For both subsystems, tasks were reported in at least five papers. Both the visual–spatial and verbal <span class="Disease">working memory tasks reported below have been shown to activate brain areas that are important for <span class="Disease">working memory.37, 38, 39

Visual–spatial working memory

In the Cambridge Neuropsychological Test Automated Battery Spatial <span class="Disease">Working Memory task,40 a number of coloured boxes are shown on a screen. <span class="Species">Children were asked to find a yellow token in these boxes without selecting boxes that have already been found to be empty or revisiting boxes that have already been found to contain a token. Raw scores were used in the analyses.

Verbal working memory

In the Digit Span Task (DST),41, 42, 43 <span class="Species">children are asked to repeat a number of digits, first in the same order and then in reverse order. The reverse part measures verbal <span class="Disease">working memory. When the DST reverse score was not available, the DST total score was used instead. For the reversed DST, either raw scores or scaled scores (mean 10 [SD 3]) were used (as indicated in Table SI, online supporting information). For the DST total, scaled scores were used (mean 100 [SD 15]).

In the Letter Number Sequencing task,42, 43 the test administrator reads a sequence of numbers and letters out loud and asks the <span class="Species">child to repeat the numbers in ascending order, followed by the letters in alphabetical order. For Letter Number Sequencing, scaled scores were used (mean 10 [SD 3]).

When DST and Letter Number Sequencing scores were not available but the <span class="Disease">Working Memory Index (i.e. DST, Letter Number Sequencing, and Arithmetic subtests of the fourth and fifth editions of the Wechsler Intelligence Scales for <span class="Species">Children)42, 43 was, the <span class="Disease">Working Memory Index was used instead (scaled scores, mean 100 [SD 15]).

Inhibition

Inhibition is the ability to deliberately inhibit a prepotent response or stop an ongoing response (response inhibition) or suppress disruption by competing responses (interference control).44 For response inhibition and interference control, tasks were reported in at least five papers. Both the response inhibition and interference control tasks reported below (or tasks very similar to those) have been shown to activate brain areas that are important for inhibition.45, 46, 47, 48, 49 For all tasks, raw scores were used in analyses.

Response inhibition

In the Go/No‐Go task, <span class="Species">children have to press a button in case of a go‐trial and have to withhold from responding in case of a no‐go trial.50, 51

In the Test of Everyday Attention for <span class="Species">Children Opposite Worlds task,52 <span class="Species">children have to read aloud a series of numbers twice. In the ‘same world’ condition they read the numbers aloud as they appear; in the ‘opposite world’ condition they are asked to say the opposite of each digit (i.e. if they read ‘1’, they need to respond by saying ‘2’ and vice versa).

Interference control

In the Test of Everyday Attention for <span class="Species">Children Sky Search task,52 <span class="Species">children are asked to find and circle target spaceships as quickly as possible on a sheet filled with similar but not exactly the same distractor spaceships.

Cognitive flexibility

<span class="Disease">Cognitive flexibility is the ability to shift between multiple tasks or mental sets.10 For all tasks, raw scores were used in analyses.

In the first part of the <span class="Gene">Trail Making Test/Trails Preschool Revised,53, 54, 55 <span class="Species">children are asked to connect numbers in the correct order (1–2–3). In the second part, <span class="Species">children are asked to connect numbers and letters in the correct order (1–A–2–B). The Trails Preschool Revised Test is a version of the <span class="Gene">Trail Making Test adapted for the use in younger <span class="Species">children. In the first part, <span class="Species">children are asked to connect <span class="Species">dogs in order of increasing size. In the second part, <span class="Species">children are asked to alternate connecting <span class="Species">dogs and bones in order of increasing size.

Data extraction

Data on <span class="Disease">working memory, inhibition, and/or <span class="Disease">cognitive flexibility and the moderators gestational age, sex, age at assessment, and year of birth were extracted from the studies by one author (CAvH) and entered in the database. A second <span class="Species">person, not involved in the design and writing of this meta‐analysis, independently confirmed the data extracted from the studies. If necessary, authors were contacted for additional data. If studies reported on subgroups of <span class="Species">children born preterm/low birthweight, subgroup data were pooled using the formulas 56 Pooled data were used in subsequent analyses.

Study quality

Study quality was assessed with an adapted version (i.e. maximum of 7 points) of the Newcastle–Ottowa Scale for cohort studies and were assessed independently by two authors (CAvH, CSHAM).57 Inconsistencies between raters were discussed and consensus was reached for all studies.

Statistical analyses

Comprehensive Meta‐Analysis V3.0 (Biostat Inc., Englewood, NJ, USA) was used to perform this meta‐analysis. Hedges’ g was used as measure for the <span class="Disease">SMDs in executive function between <span class="Species">children born preterm/low birthweight and controls. Hedges’ g corrects for the bias in Cohen's d, which becomes increasingly more apparent in smaller sample sizes.58 A SMD of 0.2 translates into a small effect, a SMD of 0.5 is a medium effect, and a SMD of 0.8 is a large effect.59

Random effects meta‐analyses were used to calculate <span class="Disease">SMDs and to investigate whether <span class="Disease">SMDs differed significantly between executive function tasks, executive function subdomains, and executive function domains.

To rule out any dependency of data, data for one executive function task or for one executive function subdomain per study were entered in the analysis. The selection was based on maximizing the number of studies per executive function task or per executive function subdomain. In case no significant differences in <span class="Disease">SMDs for the diverse executive function tasks or for the diverse executive function subdomains were found, in subsequent analyses the mean SMD aggregated across all executive function tasks assessing a subdomain or aggregated across executive function subdomains was used respectively.

Variation in <span class="Disease">SMDs that were used to calculate a mean SMD across studies was tested using Cochran's Q. The percentage of variation across studies due to heterogeneity rather than chance was expressed at I 2, with 30% to 60% representing moderate heterogeneity; 50% to 90% representing substantial heterogeneity; and 75% to 100% representing considerable heterogeneity.60

Random‐effects meta‐regressions were performed to explore whether heterogeneity in <span class="Disease">SMDs between studies was explained by between study differences in gestational age, sex, age at assessment, and year of birth of the <span class="Species">children born preterm/low birthweight. Meta‐regressions were only performed for those executive function domains of which more than 10 studies were included in the analyses. Associations between study quality and <span class="Disease">SMDs were assessed with random‐effects meta‐regressions. Publication bias was assessed by Egger tests.

Results

In total, 2079 articles were retrieved from PubMed, 538 articles from PsychINFO, and 2109 articles from Web of Science. After the removal of duplicates, 3030 articles remained. After initial screening of titles and abstracts, 475 full‐text articles were assessed. Of all articles that fulfilled the inclusion criteria, executive function tasks used were checked to extract which tasks had been reported on in at least five papers. All articles not reporting on those tasks were excluded. Furthermore, the remaining articles were checked for overlapping cohorts. The selection process is depicted in detail in (Figure S1). A total of 45 studies met all the inclusion criteria. Executive function data were provided by 35 of these studies (3360 <span class="Species">children born preterm/low birthweight and 2812 term‐born controls) either in the study paper or after a request for additional data sent to the authors.18, 21, 24, 32, 33, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89 The characteristics and main study results are given in (Tables SI, SII, and SIII, online supporting information; <span class="Disease">working memory, inhibition, and <span class="Disease">cognitive flexibility respectively). A total of 25 studies reported data on <span class="Disease">working memory (2272 <span class="Species">children born preterm/low birthweight and 2021 term‐born controls), 13 studies reported data on inhibition (1258 <span class="Species">children born preterm/low birthweight and 984 term‐born controls), and five studies reported data on <span class="Disease">cognitive flexibility (287 <span class="Species">children born preterm/low birthweight and 307 term‐born controls). Eight studies reported data on more than one executive function domain. Study quality ranged between 3 and 7, with a mean of 5.7 (SD 1.2). Overall, study quality was considered fair to good (see Tables SI, SII, and SIII).

SMDs for executive functions

Based on the pooled analysis, <span class="Species">children born preterm/low birthweight performed 0.52 SMD (95% confidence interval [CI] 0.65–0.38) lower than controls on <span class="Disease">working memory (Fig. 1), 0.39 SMD (95% CI 0.55–0.23) lower than controls on inhibition (Fig. 2), and 0.51 SMD (95% CI 0.72–0.31) lower than controls on <span class="Disease">cognitive flexibility (Fig. 3). <span class="Disease">SMDs between domains did not differ significantly (Q=1.19, p=0.55). Substantial heterogeneity among studies was found for <span class="Disease">working memory (I =70.22, p<0.001) and inhibition (I =68.2, p<0.001). <span class="Disease">SMDs for the different tasks for verbal <span class="Disease">working memory and response inhibition did not differ significantly (Q=0.73 [p=0.87] and Q=0.34 [p=0.56] respectively). Also, <span class="Disease">SMDs did not differ significantly between the subdomains verbal and spatial <span class="Disease">working memory, or between the subdomains response inhibition and interference control (Q=0.38 [p=0.54] and Q=1.1 [p=0.30] respectively).

Figure 1

Forest plot of the studies on working memory. CI, confidence interval.

Figure 2

Forest plot of the studies on inhibition. CI, confidence interval.

Figure 3

Forest plot for studies on cognitive flexibility. CI, confidence interval.

Forest plot of the studies on <span class="Disease">working memory. CI, confidence interval.

Forest plot for studies on <span class="Disease">cognitive flexibility. CI, confidence interval.

Meta‐regression analyses

Meta‐regression analyses were carried out for <span class="Disease">working memory and inhibition, as only five studies reported data on <span class="Disease">cognitive flexibility. In Table 1, the range of each moderator (gestational age, sex, age at assessment, year of birth, and study quality) is depicted for <span class="Disease">working memory and inhibition separately. Meta‐regression analyses performed for <span class="Disease">working memory showed a significant relationship with gestational age (β=0.07; 95% CI 0.01–0.13; R 2=0.15, p=0.02). Visual inspection of the scatterplots, however, indicated that this result relied on two studies featuring <span class="Species">children at the extreme ends of the gestational age distribution (i.e. 24.4wks and 35.6 wks respectively) (Fig. 4a).61, 62 Rerunning analyses without these two studies yielded a non‐significant relationship between gestational age and <span class="Disease">working memory (β=0.03; 95% CI –0.05 to 0.11; R 2=0.00, p=0.48 (Fig. 4b). No significant relationships with <span class="Disease">working memory <span class="Disease">SMDs were found for sex, age at assessment, and year of birth. None of the moderators were significantly associated with <span class="Disease">SMDs for inhibition. Study quality was not significantly associated with <span class="Disease">SMDs for working memory or inhibition.

Table 1

Moderator ranges for the working memory and inhibition subdomains

Executive function domain	Gestational age range (wks)	Percentage of males (%)	Age range at assessment (y:mo)	Year of birth (range)	Study quality (range)
Working memory	24.4–35.6	31–65	4:6–14:10	1991–2007	4–7
Inhibition	26.0–35.8	45–65	4:6–11:2	1996–2011	3–7

Figure 4

(a) Meta‐regression of gestational age on working memory. (b) Meta‐regression of gestational age on working memory after excluding two outliers with gestational ages of 24.4 and 35.6 weeks respectively. [Colour figure can be viewed at http://www.wileyonlinelibrary.com/].

Moderator ranges for the <span class="Disease">working memory and inhibition subdomains

(a) Meta‐regression of gestational age on <span class="Disease">working memory. (b) Meta‐regression of gestational age on <span class="Disease">working memory after excluding two outliers with gestational ages of 24.4 and 35.6 weeks respectively. [Colour figure can be viewed at http://www.wileyonlinelibrary.com/].

Publication bias

Egger's tests were non‐significant for all three executive function domains, as well as for subdomains within <span class="Disease">working memory and inhibition, indicating that there is no evidence for publication bias.

Discussion

This meta‐analysis aggregated the literature on the three core executive functions (<span class="Disease">working memory, inhibition, and <span class="Disease">cognitive flexibility), in <span class="Species">children born preterm/low birthweight after the introduction of antenatal <span class="Chemical">steroids and surfactant supplementation. Results show that, compared with term‐born peers, <span class="Species">children born preterm/low birthweight perform 0.5 SMD lower on <span class="Disease">working memory and <span class="Disease">cognitive flexibility measures (medium effect) and 0.4 SMD lower on inhibition measures (small‐to‐medium effect). Analysis indicated no significant differences between the <span class="Disease">SMDs for working memory, <span class="Disease">cognitive flexibility, and inhibition measures, indicating that all three executive functions seem to be affected to a similar degree. There was significant heterogeneity in effect sizes for <span class="Disease">working memory and inhibition. Heterogeneity in <span class="Disease">working memory, but not inhibition, could partly be explained by study differences in gestational age; however, this effect was driven by two studies at the extreme ends of the gestational age range. Heterogeneity could not be explained by study differences in sex, age at assessment, or year of birth.

In the literature, executive function deficits have been described to be proportional to decreasing gestational age.8, 9, 63, 90 The magnitude of the difficulties in <span class="Disease">working memory as observed in our meta‐analysis was dependent on gestational age; however, we are cautious in interpreting this result as it was driven by two studies featuring <span class="Species">children at the extreme ends of the gestational age range, and with these studies removed the gestational age of included studies only ranged between 26 and 30 weeks. Also, for inhibition, there were only a small number of studies investigating the ends of the gestational age range. To be able to draw robust conclusions on whether there is an effect of gestational age on executive function, more studies in <span class="Species">children born extremely preterm (<26wks gestational age) and moderate‐to‐late preterm (32–37wks gestational age) are clearly needed. In our analysis, executive function difficulties were not related to male sex, even though male sex is a risk factor for multiple medical adverse outcomes.26 This suggests that male as well as female <span class="Species">children born preterm/low birthweight are at substantial risk for poor executive function.

There is debate about whether executive function deficits in <span class="Species">children born preterm/low birthweight represent a stable deficit, a deficit that increases during development (growing into deficit), or a delay in maturation in which <span class="Species">children ‘catch up’ over time.27, 28, 29 In our meta‐analysis, there was no significant association between age at assessment and the <span class="Disease">SMDs for both working memory and inhibition, suggesting that difficulties in these areas in <span class="Species">children born preterm/low birthweight are stable and do not deteriorate or diminish as these <span class="Species">children grow older. It should be noted that in our meta‐analysis age at assessment ranged between 4 years 6 months and 14 years 10 months for <span class="Disease">working memory studies, and between 4 years 6 months and 11 years 2 months for the inhibition studies. There might be catch‐up in these executive functions after this age, but at least until primary school age (inhibition) or secondary school age (<span class="Disease">working memory), we found no evidence for this.

Year of birth of <span class="Species">children born preterm/low birthweight was not a factor of relevance for the size of the executive function difficulties. This finding suggests that for <span class="Disease">working memory and inhibition, outcomes are not improving with advances in medical care. Burnett et al.65 have recently investigated whether executive function outcomes of <span class="Species">children born extremely preterm (<28wks gestational age) have improved by comparing three cohorts born respectively in 1991 to 1992, 1997, and 2005. They found that the outcomes of the three cohorts did not improve and that some outcome measures were even deteriorating. Although the study of Burnett et al. relied on a questionnaire to assess executive function, their outcomes on year of <span class="Disease">birth for working memory and inhibition are in accordance with the results of our meta‐regression analyses.92 To ensure that gestational age‐related survival bias did not explain the lack of executive function improvement in more recent years, we examined the relationship between mean gestational age and year of birth for the studies included in our meta‐analysis (data available upon request). No significant relationship was found, suggesting that studies reporting on more recent cohorts of <span class="Species">children did not contain a larger number of <span class="Species">children at the lower end of the gestational age range.

Preterm/low birthweight birth is associated with substantial reductions in cognitive outcomes, as measured by IQ,4 deficits in academic performance,4, 5 and <span class="Species">children born preterm/low birthweight have a two to four times higher risk of receiving a diagnosis of attention‐deficit/hyperactivity disorder than term‐born peers.6, 7 Importantly, deficits in executive function may underlie these adverse outcomes in <span class="Species">children born preterm/low birthweight. The neuroanatomical sequels of preterm/low birthweight support the idea that executive function might be crucially involved in the adverse outcome of <span class="Species">children born preterm/low birthweight. Executive function is highly dependent on white matter network integrity, which is often compromised after preterm birth.93, 94, 95, 96 Furthermore, research has shown that hypoxic‐ischaemic events lead to damage in the striatum and its connections with the prefrontal cortex.97 These brain structures are known to be very important for executive functions,98, 99, 100, 101 and <span class="Species">children born preterm/low birthweight are vulnerable to damage in these brain areas as repeated hypoxic‐ischaemic events are common in these <span class="Species">children.97

Early interventions are warranted to improve outcomes for <span class="Species">children born preterm/low birthweight. There is evidence that effects of a computerized executive function training programme do not generalize to other functions than the trained executive function.102, 103, 104 However, this literature is based on solely training <span class="Disease">working memory, while other executive functions, such as inhibitory control and <span class="Disease">cognitive flexibility, are also impaired in this population and may benefit from computerized interventions. Given the fact that executive function remains a vulnerable area of cognitive function in the population born preterm, future studies should be conducted on which type of interventions may be effective to diminish the encountered difficulties in executive function.103, 105, 106, 107, 108, 109, 110

Limitations

First, because of the limited number of studies on executive function in <span class="Species">children born moderate‐to‐late preterm/low birthweight (i.e. >32wks gestational age), analyses for <span class="Disease">working memory included only one study within this gestational age range. Second, not enough studies presented data on <span class="Disease">cognitive flexibility to perform meta‐regressions. Third, we were not able to obtain executive function data for 10 of the 45 studies that met all inclusion criteria and could not include these studies in our analyses. Of those 10 studies, three reported on inhibition and three reported on <span class="Disease">working memory in <span class="Species">children and adolescents with a mean gestational age above 30 weeks. Lastly, unwelcome and potentially biasing heterogeneity could be introduced when the different instruments that are summarized with meta‐analytic techniques are, in fact, not all measuring the same construct.111 Therefore, we excluded executive function tasks that were used in less than five papers and we analysed the executive function tasks separately at first. However, as there were no significant differences in effect sizes between the separate executive function tasks within one executive function domain, and as there is empirical evidence that similar brain regions are activated by these tasks, we combined the effect sizes of the tasks into aggregated effect sizes for each specific executive function domain.

Conclusion

<span class="Species">Children born preterm/low birthweight since the 1990s perform poorer than term‐born peers on the three core executive function of <span class="Disease">working memory, inhibition, and <span class="Disease">cognitive flexibility, and none of these three core executive functions are more severely affected than the other. The magnitude of executive function difficulties was not associated with gestational age, and male sex was not a specific risk factor for poor executive function. Executive function difficulties remained persistent during transition to adolescence and did not improve with more recent year of birth. Given that executive function deficits are associated with worse academic performance at school age, executive functions should be assessed at early schoolage in <span class="Species">children born preterm/low birthweight to initiate early intervention targeted at improving these executive functions.

Appendix S1: Search strategy.

Click here for additional data file.

Table SI: Details on studies included for <span class="Disease">working memory

Click here for additional data file.

Table SII: Details on studies included for inhibition

Click here for additional data file.

Table SIII: Details on studies included for <span class="Disease">cognitive flexibility

Click here for additional data file.

Figure S1: Preferred Reporting Items for Systematic Reviews and Meta‐Analyses flowchart of the study selection procedure.

Click here for additional data file.