OBJECTIVES: Children with congenital heart disease (CHD) often have neurocognitive deficits, sometimes with a detrimental impact on daily and school functioning. These deficits may increase through childhood. In this study, we investigated whether children with CHD, who underwent heart surgery as infants, show more neurocognitive deficits, especially in the executive functions, as they get older, compared with healthy controls.
METHODS: In this longitudinal follow-up study, 107 children with CHD and 77 healthy control children underwent extensive neurocognitive testing at 4 years of age. Ninety-three percent of the children (100 patients with CHDs and 72 controls) underwent a second neurocognitive testing 3 years later. Intelligence, visual-motor integration (VMI), alertness, motor coordination, executive functions, and psychosocial functioning were assessed.
RESULTS: IQ scores were consistently lower in the CHD group (P < .001); however, the difference of 11.7 IQ points between both groups at follow-up 1 decreased to 7 IQ points at follow-up 2 (P = .003). Inhibition reaction time had improved in both study groups at follow-up 2 (P < .001) and did not differ between both groups from follow-up 1 to follow-up 2 (P = .849). Deficits in VMI, alertness, motor coordination, and psychosocial functioning also did not worsen for patients with CHDs at follow-up 2, compared with healthy controls.
CONCLUSIONS: Children with CHD, who underwent heart surgery as infants, do not show an increase of neurocognitive deficits between the ages of 4 and 7 years, compared with healthy controls. Patients with CHDs keep deficits in intelligence, VMI, and psychosocial functioning, but seem to partially grow out of their deficits.
- ANT —
- Amsterdam Neuropsychological Tasks
- CBCL —
- Child Behavior Checklist
- CHD —
- congenital heart disease
- CI —
- confidence interval
- EFs —
- executive functions
- GLM —
- general linear model
- IQR —
- interquartile range
- LGC-trial —
- Leuven Glucose Control Trial
- RACHS —
- risk adjustment for congenital heart surgery
- TBI —
- traumatic brain injury
- VMI —
- visual-motor integration
- WPPSI-R —
- Wechsler Preschool and Primary Scale of Intelligence Revised
What’s Known on This Subject:
Previous research suggested that children with congenital heart disease (CHD), who underwent heart surgery, may have neurocognitive deficits that worsen with age. However, this has not been confirmed in longitudinal follow-up studies beyond early childhood in the population with CHD.
What This Study Adds:
Patients with various CHDs, who underwent heart surgery as infants, seem to partially grow out of their deficits when tested together with matched healthy controls at the age of 4 and 7.
Congenital heart disease (CHD) affects nearly 1% of births per year.1 Approximately 25% of these children with CHD need surgery or other procedures in the first year of their lives, called critical CHD. Due to medical progress, the survival rate of these children with CHD has strongly improved over the years. However, an important proportion of these children with CHD are confronted with a developmental delay and learning difficulties. Certainly children with critical CHD are more at risk for neurocognitive deficits. These can have a negative impact on daily and academic functioning and may persist into adulthood.2–5
Neurocognitive deficits can occur in intelligence, but also in more specific functions, such as attention and executive functions (EFs).2,6 EFs cover a variety of higher cognitive functions, such as inhibition and cognitive flexibility. Attention-deficit/hyperactivity disorder, with impaired inhibition as a key feature, is more prevalent in children with CHD.7 These EFs especially depend on the developing prefrontal cortex, which continues to mature until young adulthood.8 Because attention and EFs are therefore also constantly maturing functions, studying neurodevelopmental patterns in children with CHD when they are still young may offer a better understanding of how the children function in the long run.8,9 Moreover, neurocognitive deficits in children with CHD may increase through childhood.10,11 This phenomenon, called “growing-into-deficit,” stands for the increasing deficits that are emerging through childhood when more complex functions are developing and required brain structures are disturbed.12 For instance, dyslexia can be diagnosed only after a period of reading education, although the brain structures necessary for reading are dysfunctional earlier.
However, in children with CHD, prospective, longitudinal studies beyond the age of early childhood are scarce.5,13 Variation in age at surgery and age at testing,14 requiring different assessment tools,11 have been important confounders in the investigation of developmental patterns in children with CHD.
In this growing-into-deficit study, we therefore hypothesized that children with various types of critical CHD and who had undergone heart surgery as infants, show worse neurocognitive deficits over time, notably the later maturing EFs, compared with matched, healthy children. Hence, the gap in neurocognitive function between children with CHD and healthy children was hypothesized to widen, when using exactly the same neurocognitive test battery at the ages of 4 and 7.
This study is a longitudinal neurocognitive follow-up of the subgroup of patients with CHDs who were enrolled as infants after heart surgery in the Leuven glucose control trial (LGC-trial) between October 2004 and December 2007 at the PICU of the University Hospitals Leuven, Belgium.15 A first follow-up of all patients took place between August 2008 and January 2012 (Neurocognitive development of children 4 years after critical illness and treatment with tight glucose control, NCT00214916).16 For the second follow-up (NCT01632813), we included 7-year-old children with CHD and healthy control children who were 4 years old when they participated in the LGC follow-up study (ie, first follow-up time point). Healthy control children consisted of patients’ siblings (n = 14) and unrelated healthy children (n = 63) recruited via schools and word-of-mouth advertising all over Flanders, during the first follow-up.16 Exclusion criteria for the second follow-up were lack of baseline neurocognitive measurements of IQ and inhibition data during the first follow-up and date of birth before February 2005, because the children would be too old to perform the same IQ test again.
Written informed consent was obtained from the parents or the legal guardian. A copy of the signed informed consents was given to the parents. The study protocol and consent forms were approved by the institutional ethical review board (ML8351). Our study thus complied with the Declaration of Helsinki and the guidelines of Good Clinical Practice and the International Conference on Harmonisation.
The same psychologist who tested most of the children at the first follow-up, tested all participating children at the second follow-up (CS). If parents decided not to come to the University Hospitals Leuven, the psychologist examined the child at home.
The assessment tools for the second follow-up were the same neurocognitive tests taken from 4-year-old participants in the first follow-up (LGC follow-up-study).16 The neurocognitive test battery consisted of tests measuring general intellectual functioning, visual-motor integration (VMI), attention, motor coordination, and EFs. The Dutch version of the Revised Wechsler Preschool and Primary Scale of Intelligence (WPPSI-R) was used to assess intelligence.17 Total IQ, Verbal IQ, and Performance IQ were analyzed. VMI was measured with the Beery-Buktenica Developmental Test of Visual-Motor Integration.18 Total VMI standard score was obtained. To better gauge the effects of aging on IQ and VMI development, also raw IQ and VMI scores without age correction were analyzed. To assess attention, motor coordination, and EFs, 4 computerized tasks of the Amsterdam Neuropsychological Tasks (ANT) were taken.19,20 The main advantage of the ANT, as compared with paper-and-pencil tasks, is that the measurement of reaction times in combination with accuracy (error rate) contributes to the sensitivity to detect problems in these neurocognitive domains.21 Baseline speed assesses alertness by measuring simple reaction time to 32 visual stimuli, expressed in milliseconds. Mean reaction time and SD of the reaction time were obtained for the dominant and nondominant hands. The tapping task measures motor coordination by counting the number of taps for the nondominant hand, dominant hand, bimanual alternation, and bimanual synchronous. The Response Organization Objects test measures inhibitory control and cognitive flexibility, by calculating the differences in reaction time and the differences in number of errors between tests of increasing demand.22,23 The Memory Search Objects–2 Keys assesses working memory by calculating the increase in reaction time and error rate during higher memory load.24,25 The completion of this full neurocognitive test battery took ∼2 to 3 hours.
To assess psychosocial functioning, the Child Behavior Checklist (CBCL) was filled out by the parents or the child’s legal guardian (CBCL/1.5–5 at follow-up 126 and the CBCL/6–18 at follow-up 227). T-scores for total problems and internalizing and externalizing problems were analyzed.
The primary study goal was to test the hypothesis of whether children with CHD show more neurocognitive deficits than healthy controls at the second follow-up, compared with follow-up 1. The maximum sample size was determined by the first follow-up.16 Taking into account inclusion criteria, there was a maximum of 107 eligible patients with CHDs and 77 eligible control children. Although our maximum sample size was fixed, an a priori statistical power calculation was performed. A mean difference in Total IQ of 3.16 IQ points, the standard measurement error,17 was considered to be the minimal clinically relevant effect size. The a priori calculated statistical power to detect a difference of 3.16 IQ points (SD 7) in the CHD group and no difference (SD 7) in the control group was 85.6% for the maximum sample size in both study groups. For the inhibition reaction time measure, the primary outcome measure, we based our power calculation on a longitudinal study in which children with transposition of the great arteries operated on as neonates were tested twice.10 At follow-up 1, double the rate of impairment was found compared with follow-up 2. Consequently, for our study we hypothesized that the difference in inhibition reaction time between the CHD and the control groups would double from 60 ms to 120 ms. The a priori calculated statistical power (α error 5% and SD, in both groups of 250 ms) thus to detect a minimal effect size of 120 ms between both groups was 89.5% for the maximum number of eligible children in both groups.
Discrete demographic and clinical variables were summarized by counts and percentages and differences in proportions between the CHD and control group were analyzed by the χ2 or Fisher’s Exact test. Not normally distributed continuous variables were summarized as median and interquartile range (IQR), and differences between the CHD and control group were analyzed by the Wilcoxon rank-sum test. JMP version 11.2.0 (SAS Institute Inc, Cary, North Carolina) was used for these analyses.
For each neurocognitive outcome separately, a general linear model (GLM) for longitudinal measures (with an unstructured covariance structure for the repeated measures) was used to compare the evolution between both groups. The model contained group and time as factors. Because the study goal was to verify whether children with CHD develop more neurocognitive deficits compared with controls, primary interest was in the interaction between both factors. Observed means (with SD) and the estimated means (with 95% confidence interval [CI]) obtained from the GLMs were reported (Supplemental Figures 7–22). The GLMs were fitted by using IBM SPSS Statistics Version 22 (IBM SPSS Statistics, IBM Corporation, Chicago, IL). Because there was a difference in socioeconomic status16 and age at follow-up 1 between both groups, socioeconomic status and age (in case of non–age-adjusted data) were included as covariates in all models. Because some neurocognitive variables were rightly skewed, the GLM was also fitted on log-transformed data (natural logarithm, after adding a constant in case zero values occur) as a sensitivity analysis to test the robustness of the results. Exploratory subgroup analyses were done for high-risk patients (risk adjustment for congenital heart surgery [RACHS] ≥3) and neonates (heart surgery in first 30 days of life). To assess whether there was a difference between patients/controls who did and did not receive remedial therapies between the 2 time points, a GLM analysis was done with group, time, and help for school functioning as factors. For all endpoints, a 2-sided P < .05 was considered statistically significant. No corrections were performed for multiple comparisons and therefore a single significant P value was interpreted with caution.
Demographic and clinical data of the CHD group and the healthy control group at follow-up 1 are presented in Table 1. Median age at follow-up 1 was 4.16 years (IQR 4.08–4.30) for the CHD group and 4.55 (IQR 4.28–4.77) for the control group. Follow-up 2 was performed between July 2012 and September 2014 at a median of 2.93 years (IQR 2.64–3.08) after follow-up 1. Median age at follow-up 2 was 7.20 years (IQR 7.09–7.33) for the CHD group and 7.17 (IQR 7.09–7.27) for the control group. The flowchart is shown in Fig 1. A total of 172 children (100 patients with CHDs and 72 controls) participated in the second follow-up, representing a dropout rate of only 7%.
Most data were comparable between both groups, except for socioeconomic status and age (both P < .001). Also, the biometric measures of height, weight, and head circumference were smaller in the CHD group. However, these differences are typical of children with CHD and do not affect neurocognitive function directly.30 There were no differences in demographic and clinical data between participating and nonparticipating patients/controls at follow-up 2 (data not shown).
The results from the GLMs are presented in Table 2, Figures 2–6 and Supplemental Figures 7-22. Observed means are shown in Supplemental Table 3. Individual raw data of both groups at both time points can be found in Supplemental Figures 23–43.
Total IQ scores were consistently lower in the CHD group (group P < .001) (Fig 2 and Table 2). However, the difference of 11.7 IQ points between both groups at follow-up 1 decreased to 7 IQ points at follow-up 2 (interaction P = .003). The gap in Verbal and Performance IQ also decreased (Table 2). The individual raw intelligence data show a steeper increase in raw IQ scores in the CHD group, compared with healthy controls (Supplemental Figures 23–25).
VMI was poorer in the CHD group (P < .001), but the gap between the 2 study groups did not change (interaction P = .207) (Fig 3).
The alertness reaction time of the nondominant hand improved in both study groups between follow-up 1 and 2 (time P < .001). Despite a steeper improvement in the CHD group compared with the control group (interaction P = .005 and Supplemental Figure 28), the nondominant reaction time was overall slower in the CHD group (group P = .016) (Table 2 and Fig 4). Reaction time of the dominant hand, a less difficult task, and the within-subject SD of both dominant and nondominant hand showed a greater improvement over time in the CHD group, compared with the controls (for all interactions P ≤ .003) (Table 2).
The easier (unimanual tapping dominant and nondominant hand), as well as the more complex motor coordination tasks (synchronous and alternating [Fig 5] tapping) were better in and comparable between both study groups at follow-up 2 (Table 2).
Inhibition and flexibility reaction time (Table 2 and Fig 6) had improved by 50% and 15%, respectively, in both groups at follow-up 2 (P < .001), but did not differ between the CHD and the control group (P > .1).
Psychosocial functioning (total problem score CBCL) was consistently lower in the CHD group (group P = .013) (Table 2). Parents reported more problems at follow-up 2 than at follow-up 1 (time P < .001).
Estimated means of log-transformed (natural logarithm, Ln) and back-transformed data are presented in Supplemental Figures 44–51. GLM yielded similar results for original and log-transformed data.
No change in IQ was found for high-risk patients (RACHS ≥3) (n = 57) and neonates (n = 38) compared with the control group (for all IQ scores: interaction P > .1) (Supplemental Tables 4 and 5). Greater improvement of alertness over time was found for both high-risk groups (interaction for all reaction time measures: P < .04) (Supplemental Tables 4 and 5). No difference in IQ and alertness evolution was found between patients who received help for school functioning (n = 37) and patients who did not receive help for school functioning (n = 63) between the 2 time points, or between control children with (n = 20) and without help for school functioning (n = 52) (for all IQ scores: interaction P ≥ .2 and for all reaction time measures: interaction P ≥ .3) (Supplemental Table 6).
This longitudinal study demonstrated that children with critical CHD do not show worsening of neurocognitive deficits between the ages of 4 and 7, compared with healthy controls. Although children with critical CHD seem to partially grow out of their deficits, they are still confronted with deficits in intelligence, VMI, and psychosocial functioning. Unexpectedly, the longer maturing EFs, inhibition, flexibility, and working memory of children with CHD improved as they got older and were comparable with healthy children.
It has been postulated that recovery after early brain injury could be explained within a recovery continuum of plasticity and vulnerability.31 Our results fit well in this framework. Evidence of plasticity of neurocognitive problems in children with CHD is found for attention, motor coordination, and EFs. On the other hand, the consistent deficits in intelligence, VMI, and psychosocial functioning also suggest vulnerability.
Our results are in line with a study in which EFs were tested in 45 children with transposition of the great arteries and 45 healthy controls for 3 consecutive years between the ages of 5 and 7.32 They also did not find a worsening of EFs in this specific group of children with CHD. The latter study and our study contradict the study by Hövels-Gürich and colleagues,10 in which they detected severely worsened neurocognitive deficits over time in children who underwent a neonatal arterial switch operation. The discrepancy may be explained by the assessment of more cognitive domains and the use of other, possibly more sensitive testing tools at the second follow-up. Moreover, the lack of a control group of healthy children, who could have been tested in parallel, may have overestimated the worsening of the cognitive deficit.
The exploratory subgroup analyses indicated that children with a high risk for neurocognitive impairment, such as neonates and the children undergoing the most invasive procedures, did not show a neurocognitive worsening over time, compared with the control children. More research is necessary to investigate whether these high-risk children have another neurocognitive profile. Furthermore, no difference was found between patients/controls who did and did not receive remedial therapies between the 2 time points.
In a longitudinal study in children with different severities of traumatic brain injury (TBI), it was demonstrated that, although children with severe TBI had the poorest outcome, all TBI groups stabilize 10 years after injury, regardless of TBI severity.33 Such a long-term follow-up study would be useful to further study neurocognitive development in children with CHD, as specific functions like EFs and related prefrontal brain regions continue to mature.8 Moreover, EFs also have their own developmental trajectories,34 which can vary from age to age.
The increase of the behavior problem scores as they get older, also reported in children who had TBI or after cardiac arrest,35,36 may reflect increasing demands on children in general. It may also be questionnaire-related, as the one, taken at the age of 7, inquires more items than the questionnaire at the age of 4.
Our study was one of the very few with a longitudinal twofold neurocognitive follow-up of a large group of patients with various CHDs, who underwent heart surgery at the same age, and healthy controls, at the same ages. Both groups performed the same neurocognitive test battery twice, so that developmental patterns could be investigated. Bias due to loss to follow-up was minimal.
Our study has some limitations. First, because our follow-up consisted only of assessment at the age of 4 and 7, our findings cannot be generalized to further development in later school ages. Second, the subgroup analyses should be interpreted with caution, because of likely collinearity with severity of CHD, small sample sizes and consequently loss of power. Finally, although children with CHD are known to risk early brain injury,2 no neurologic correlates, such as MRI, were included in our study.
Contrary to our hypothesis, there is no evidence that children with CHD who underwent heart surgery as infants grow into their neurocognitive deficits over a time horizon of 3 years. Future research should include a long-term follow-up to investigate further neurocognitive development and should examine which factors predict neurocognitive outcome.
We thank Marijke Gielen and Kirsten Claessens for acquisition of data at the first follow-up. We thank Steffen Fieuws (MSc, PhD, Department of Biostatistics, KU Leuven, Leuven, Belgium) for his help with the statistical analysis.
- Accepted March 24, 2016.
- Address correspondence to Caroline Sterken, MPsych, Department of Intensive Care Medicine, KU Leuven, Herestraat 49, B-3000 Leuven, Belgium. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: The study was supported by the Methusalem Program from the Flemish government (METH/08/07 and METH/14/06) via KU Leuven to Dr Van den Berghe; the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007–2013)/ERC Advanced Grant Agreement AdG-2012-321670 to Dr Van den Berghe, and a Senior Clinical Investigator fellowship of the Research Foundation Flanders to Dr Mesotten. The sponsor/funder had no role design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, and approval of the manuscript.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Marino BS,
- Lipkin PH,
- Newburger JW, et al; American Heart Association Congenital Heart Defects Committee, Council on Cardiovascular Disease in the Young, Council on Cardiovascular Nursing, and Stroke Council
- Sterken C,
- Lemiere J,
- Vanhorebeek I,
- Van den Berghe G,
- Mesotten D
- Hansen E,
- Poole TA,
- Nguyen V, et al.
- McGrath E,
- Wypij D,
- Rappaport LA,
- Newburger JW,
- Bellinger DC
- Anderson V,
- Northam E,
- Hendy J,
- Wrennal J
- van der Rijken R,
- Hulstijn-Dirkmaat G,
- Kraaimaat F, et al
- Vander Steene G,
- Bos A
- Beery K,
- Beery N
- de Sonneville L
- de Sonneville L
- Achenbach T,
- Rescorla L
- Achenbach T,
- Rescorla L
- Licht DJ,
- Shera DM,
- Clancy RR, et al.
- Anderson V,
- Spencer-Smith M,
- Wood A
- Copyright © 2016 by the American Academy of Pediatrics