Minimally invasive surgical approach in children treated for oesophageal atresia is associated with attention problems at school age: a prospective cohort study

The long-term neurodevelopment of children born with oesophageal atresia (OA) is unclear. Therefore, we assessed the neurocognitive domains and their predictors. Patients born with OA between February 2006 and December 2014, who were routinely seen at eight years as part of a structured prospective longitudinal follow-up program, were included. Main outcome measures were need for school support, performance in various neurocognitive domains and potential predictors of neurocognitive problems. We analysed data of 65 children with a mean (SD) age of 8.1 (0.2) years, of whom 89% with OA type C. Thirty-five (54%) surgical corrections were minimally invasive; the median (interquartile range) duration of exposure to anaesthetics in the first 24 months was 398 (296 – 710) minutes. Forty-four (68%) attended regular education without extra support and intelligence was within normal range (99–108). More than 50% had z-scores ≤ -2 on one or more neurocognitive domains, of which attention was the most frequently affected domain. The speed on the sustained attention task was significantly below normal (z-score -1.48 (2.12), p < .001), as was fluctuation of sustained attention (z-score -3.19 (3.80), p < .001). The minimally invasive approach and a lower socio-economic status (both p = 0.006) proved significant predictors for sustained attention problems in multivariable analyses. Conclusion: Children who undergo minimally invasive surgery for OA correction are at risk for sustained attention problems at school age. Future studies unravelling the effects of perioperative events on neurodevelopment should lead to optimal surgical, anaesthesiological, and intensive care management in the neonatal period. What is Known: • School-aged children born with oesophageal atresia have normal intelligence but problems with sustained attention at eight years. What is New: • Oesophageal atresia patients, who undergo minimally invasive surgery or who have a background of lower socioeconomic status are at serious risk for sustained attention problems at school age. • Moreover, those who have been intubated for a longer period are at risk for stronger fluctuations in sustained attention. Supplementary Information The online version contains supplementary material available at 10.1007/s00431-024-05449-y


Introduction
Mortality and morbidity in children born with oesophageal atresia (OA) are primarily determined by major comorbidities and a complicated clinical course [1,2].Multiple surgeries and complications in the neonatal period necessitating prolonged hospitalisation affect development and daily functioning during childhood [3].
Neurodevelopment can be negatively affected by neonatal surgery and any acquired brain injury [4,5].A study in children with OA found impaired expressive language at preschool age [6].Other studies found normal intelligence at school age but impaired attention and working memory [7,8].However, for the children born with OA, the potential causative factors behind these impairments have not yet been identified.
Minimally invasive surgery (MIS) provides better visualisation of the surgical field, with less surgical damage and better long-term outcome expectancy concerning scoliosis [9].However, the effects of artificial CO 2 -pneumothorax to create a larger surgical workspace and the resulting hypercapnia and acidosis on the neonatal brain remain unclear.Moreover, this approach is technically more demanding than an open approach and often involves longer anaesthetic exposure [10][11][12].
After our previous assessment of neurocognitive functioning of school-aged children born with OA [7], we have introduced more extensive neurocognitive assessments since 2014 in our prospective, standardised follow-up program for children with congenital anatomical malformations [3].Consistent with studies in survivors of neonatal extracorporeal membrane oxygenation and those with congenital diaphragmatic hernia [13], we hypothesised that school-aged children born with OA, who all have undergone interventions during the first years of life, are at risk for long-term neurocognitive problems that could potentially affect their school functioning.Hence, our objective was to assess the requirement for school support, evaluate performance in various neurocognitive domains, and identify predictors of neurocognitive problems.

Population
We included data collected between April 2015 and June 2023 of 8-year-old children born with OA, seen in the context of the standardised prospective longitudinal follow-up program at the Erasmus MC -Sophia Children's Hospital [3].
Exclusion criteria were genetic syndromes known to affect neurodevelopment, leading to an inability to complete the neurocognitive test battery.The children's parents were informed that data were used for research purposes.This study was performed in line with the principles of the Declaration of Helsinki and was approved by the Medical Ethics Committee of the Erasmus University Medical Centre, the Netherlands (MEC-2017-185).Written informed consent was formally waived as there is no additional patient burden and no privacy concern.No funding has been received for this article.

Data collection
Data from the medical records included: sex, gestational age (GA), prematurity (yes if GA < 37 weeks), birthweight, small for gestational age (SGA; i.e. birthweight < -2 SD) [14], type of OA according to Gross [15], congenital cardiac anomaly (yes, if surgery or follow-up by a cardiologist), VACTER-L association according to Solomon [16], surgical approach for OA correction, duration of anaesthetic exposure and number of procedures under general anaesthesia within the first 24 months, need for tracheotomy or gastrostomy, duration of intubation (pre-and postoperative, including re-intubation), sepsis during first hospital admission (positive blood culture), duration of initial hospital stay, feeding type at discharge, history of fundoplication surgery.
Additionally, we categorised type of education (regular, regular with extra help or special education), and socioeconomic status (SES) according to highest maternal education level (International Standard Classification of Education (ISCED) levels low (0-4) and high (5-8) [17].

Neurocognitive assessment
An experienced paediatric psychologist performed the following assessments in the outpatient clinic (see Supplemental File 1 for details): Neurocognitive test scores were converted into z-scores and -if applicable -inverted so that a higher score always equals better performance.Z-scores ≤ -2 were regarded as reflecting impaired functioning; z-scores > -2 and ≤ -1 were regarded as reflecting borderline functioning (general population: mean z-score = 0; SD = 1) [19].

Statistical analysis
The Shapiro-Wilk Test was used to assess normality of continuous variables.Parametric tests were used to evaluate differences in normally distributed continuous patient characteristics between participating and non-participating patients (independent samples t-test) and to assess whether the normally distributed neurocognitive test outcomes (one-sample t-test) differed from population norms.Nonparametric tests were used for categorical (Fisher's exact test) and non-normally distributed patient characteristics (Mann-Whitney U test) and neurocognitive test outcomes (one-sample Wilcoxon signed-rank test).All data are shown as mean (SD), median (interquartile range (IQR)), or proportions (%).The children who did not attend the followup program and those who had not undergone assessment of intelligence with WISC-III or V, were considered as non-participants.
To evaluate the predictor variables for impaired neurocognitive outcome, we conducted univariable and multivariable linear regression analyses for the outcome measures that scored significantly below the norm.Based on clinical experience and earlier studies, we chose five predictor variables, reflecting background (GA and SES), initial treatment (MIS (yes/no) and duration of intubation), and clinical course (duration of anaesthetic exposure).The duration of intubation was coded as a categorical variable with categories 1-2 days and ≥ 3 days.The predictor variables were checked for outliers.The multivariable linear regression included all five predictor variables.The amount of and reasons for missing data were evaluated.In case data were considered missing at random, multiple imputation of missing values was performed using a fully conditional specification [20].The missing values for type of surgical approach were however not imputed, as we did not consider this appropriate.Patients with unknown surgical approach were not included in multivariable analysis.Thirty imputed data sets were generated for each outcome, and the results were pooled using Rubin's rules.The assumptions for linear regression analysis were assessed using normal probability plots of the residuals and by calculating variance inflation factors.Multicollinearity was assumed if variance inflation factors exceeded 2.5 [21].
Analyses were performed with SPSS 25.0 (IBM, Armonk, NY, USA), and a two-sided p-value < 0.05 was considered statistically significant.Due to the study's explorative nature, correction for multiple comparisons was not performed.

Patient characteristics
Of 110 children born with OA between February 2006 and December 2014, six had died before the age of eight, and twelve were excluded based on genetic syndromes or psychomotor retardation.Eventually, data from 65 children were analysed (Fig. 1).Their mean age was 8.1 (0.2) years.The baseline characteristics of participants and nonparticipants did not significantly differ (Table 1).
Fifty-eight participants had Gross type C. The overall median gestational age was 38 (IQR 36-39) weeks.Twenty-seven children were born prematurely, with a range from 29 up to 37 weeks.In 37 cases (57%), the primary OA correction was minimally invasive, with two conversions to open surgery.Minimally invasive surgery was performed when the patient was cardiopulmonarystable and when deemed surgically feasible.The median duration of exposure to anaesthetics within the first 24 months was 398 (IQR 269-710) minutes in 3 (IQR 2-6) surgical and/or diagnostic procedures under anaesthesia.The median duration of intubation was 2 (IQR 1-3) days, and the initial hospital stay was 20 (IQR 12-35) days.For GA, two missing values were imputed, for anaesthetic exposure twelve, for duration of intubation ten and for SES eight patients.Missing data for these variables were due to initial treatment in other centres.

Neurocognitive outcome
Forty-four children (68%) attended regular education without extra support.Seventeen children (26%) received support to keep up with regular education.Four children attended special education (6%).This distribution is close to the general Dutch population's [22].
The total IQ and verbal comprehension were above general population norms (Table 2; Fig. 2).The parents rated the daily executive functioning favourably (Table 2).
In the neuropsychological assessment, the scores for short-and long-term visuospatial recall (RCFT-immediate and delayed), visuospatial working memory (WNV Spatial Span), and executive functioning subdomain planning (Modified Six Elements) were significantly below normal (Table 2).
Figure 2 shows performance in different neurocognitive domains.Over 50% of the children had z-scores ≤ -2 on one or more tests (Fig. 2).For fourteen tests, the proportion of children with results in the borderline range was higher than expected.For twelve tests -three assessing attention -the proportion of children with results in the impaired range was higher than expected.
In multivariable regression analyses with multiple imputation to account for data missing at random, MIS and lower SES (both p = 0.006) were associated with sustained attention problems (Table 3).The MIS and lower SES were also significantly associated with sustained attention fluctuation (both p = 0.008) (Table 3).The remaining univariable and multivariable regression analyses are provided in Supplemental File 2. All variance inflation factors were below 2.5, thus no multicollinearity was assumed.

Discussion
This study is -to our knowledge -the first to extensively evaluate multiple domains of neurocognitive outcome and its association with MIS and other potential predictors in school-aged children born with oesophageal atresia.We showed that these children had normal school performance and intelligence within normal ranges.Still, sustained attention problems were noted, suggesting that school-aged children with OA can focus briefly on a task but do not sustain attention easily.Significant independent predictors for impaired sustained attention and its fluctuation were MIS and a background of lower SES.Test scores slightly below the norm are mainly within the neurocognitive domains (verbal and visuospatial memory) where attention is required for adequate performance.In univariable analyses, lower SES and longer duration of intubation were predictors of sustained attention fluctuation.However, in multivariable analysis, the duration of intubation did not remain a significant predictor.Children with low SES are also at risk for impaired visuospatial working memory and delayed visuospatial memory.
Several previous studies on neurocognitive outcomes in children born with OA showed impairments throughout different stages of childhood.Burnett and co-workers longitudinally studied the cognitive, academic, and behavioural functioning of 71 five-year-old and 72 eight-year-old OA patients [8].The children were at risk for cognitive difficulties, particularly in attention and working memory, not significantly associated with additional congenital anomalies, duration of hospitalisation, or prematurity [8].Earlier, we demonstrated sustained attention problems at school age in a cohort of OA children born between 1999 and 2006; before introducing the minimally invasive approach in our centre [7].Two studies reported a higher prevalence of intellectual disabilities with special educational needs than we did: in 22% of school-aged OA children and 33% of adolescents, respectively [23,24].A Swedish national registry-based study showed a higher risk of autism spectrum disorders and intellectual disability in 735 adults born with OA [25].Other than this Swedish study, no data on more profound psychiatric diagnostics has been published to date.Previous studies have not identified possible predictors for impaired neurocognitive outcomes in individuals born with OA.However, in the broader perspective of non-cardiac congenital anomalies, multiple risk factors for developmental delay in infants have been identified, such as longer duration of mechanical ventilation, parental education level [4], and undergoing numerous surgical interventions in the first 24 months [3,4].
In this study, MIS was negatively associated with the speed and fluctuation of sustained attention.MIS is technically more demanding and generally takes longer to complete than the open approach [10][11][12].Moreover, the   artificial CO 2 -pneumothorax used to create surgical workspace results in a significantly higher pCO 2 level in the patient [26,27].This additional CO 2 load may result in hypercapnia and acidosis [28,29], the effects of which on the neonatal brain are still unclear.This study suggests a possible association between MIS and neurocognitive outcomes at school age.However, the cohort examined in this study underwent surgery 8-17 years ago.Since then, surgeons and anaesthesiologists have gained more experience with the surgical and anaesthesiological aspects of MIS, and the techniques have been further developed.Therefore, future research with new cohorts is needed to investigate whether single or multiple surgeries using these techniques affect the neonatal brain.
The reduced sustained attention should be seen in the context of this cohort's relatively high intelligence scores.The high intelligence in our cohort could be associated with the high SES [30].Concomitant with the high intelligence scores, at least average attention scores would be anticipated.This further emphasises the discrepancy between intelligence and impaired attention in our cohort.Therefore, parents of children with attentional issues need counselling at school age [31].In addition, the current study must be considered when deliberating treatment options for these patients, both in infancy and school age.
Our study -which contributes to knowledge on neurocognition and potential risk factors for adverse outcomes in school-aged children with OA -underlines the importance of previously advocated standardised follow-up programs for these children [8,32,33].Further research into predictors of neurocognitive outcomes is critical to improve care for individuals born with OA [7,8,24,32,33].
Strengths of our study are the relatively large cohort, who prospectively attended the same structured follow-up program, and the high participation rate (71% of eligible candidates).We found no evidence of selection bias, with similarity in baseline characteristics in participants and nonparticipants.Still, some potential drawbacks associated with our study should be mentioned.The duration of anaesthetic exposure was considered as one of the predictors, but we did not account for the choice of anaesthetics and the occurrence of perioperative respiratory or hemodynamic events.Unfortunately, our medical records' data did not permit such detailed analyses.

Conclusion
We demonstrated that children born with OA are at risk for sustained attention problems at school age, highlighting the need for counselling and timely guidance in educational settings.MIS was identified as an important independent risk factor for sustained attention problems.Future research should employ multimodal neuromonitoring with larger sample sizes and include more detailed perioperative and anaesthesiological parameters linked to neurodevelopmental outcomes.This will enable a better understanding and guide both surgeons and anaesthesiologists in their choices regarding treatment.

Fig. 2
Fig. 2 Distribution of test scores in the study population.Red: percentage of patients with a z-score ≤ -2 (impaired), orange: -2 ≤ -1 (borderline), and green: > -1 (normal) on each of the neurocognitive tests.The dotted line at 16% of the population represents the expected proportion of children having z-scores ≤ -1.Abbreviations: TIQ = Total IQ; VC = Verbal Comprehension; PO = Perceptual Organisation; PS = Processing Speed; TMT = trial making test; Stroop DCT-Time representing sustained attention, DCT-SD the fluctuation of sustained attention and WNV Spatial Span the visuospatial working memory DCT dot cancellation test, SD standard deviation, WNV Wechsler Nonverbal Scale of Ability, SES socio-economic status a 0 = minimally invasive surgery/converted and 1 = thoracotomy b 0 = 1-2 days and 1 = ≥ 3 days c 0 = low and 1 = high

Table 2
Overview of neurocognitive outcome compared to norm scores Mean (standard deviation) = average IQ score or z-score of neurocognitive test.Compared to the norm population with a population IQ mean of 100 (SD = 15) or mean z-score of 0 (SD = 1).One-sample t-test was used for normally distributed data, and one-sample Wil-NL Wechsler Intelligence Scale for Children, third, Dutch version, TIQ total IQ, VC verbal comprehension, PO perceptual organisation, PS processing speed, TMT trial making test, Stroop Stroop colour word test, DCT dot cancellation test, SD standard deviation, RAVLT Rey auditory verbal learning test, RCFT Rey complex figure test, WNV Spatial Span Wechsler Nonverbal Scale of Ability, subtest Spatial Span, BRIEF Behaviour Rating Inventory of Executive Functioning (see Supplemental File 1 for neurocognitive test descriptions) *Missing data due to hygiene protocol during the COVID-19 pandemic

Table 3
Univariable and multivariable linear regression analyses with predefined variables for the DCT -Time, DCT -SD and WNV Spatial Span