OSA and Neurocognitive Impairment in Children With Congenital Heart Disease

Abstract

Background

Children with congenital heart disease (CHD) have an increased risk of neurocognitive impairment. No prior studies have evaluated the role of OSA, which is associated with neurocognitive impairment in children without CHD.

Research Question

Is OSA is associated with neurocognitive impairment in children with CHD?

Study Design and Methods

Children aged 6 to 17 years with corrected moderate to complex CHD without syndromes that may affect neurocognition were recruited from the pediatric cardiology clinic. Participants underwent home sleep testing and neurocognitive testing, including a validated Intellectual Quotient (IQ) test as well as validated tests of memory (Paired Associates Learning test), executive function (Intra-Extra Dimensional set shift test), and attention (Simple Reaction Test) from the CANTAB neurocognitive testing battery.

Results

Complete results were available for 30 children. Seventeen children (57%) were found to have OSA. Total IQ was markedly lower in children with CHD and comorbid OSA compared with children with CHD without comorbid OSA (mean, 86 ± 12 vs 98 ± 11; P = .01). Children with CHD and OSA did significantly worse on the Paired Associates Learning test, with a median of eight total errors (interquartile range [IQR], 2.25-15) compared with children with CHD without OSA (median total errors, 2, IQR, 1-8; P = .02).

Interpretation

Children with CHD and comorbid OSA have impaired neurocognition compared with children with CHD without comorbid OSA. OSA may be a reversible cause of neurocognitive impairment in children with CHD. Further research is needed to evaluate the effects of OSA treatment on neurocognitive impairment in children with CHD.

Congenital heart disease (CHD) is the most prevalent congenital malformation, estimated to occur in nearly one in 100 live births, with one third of children with CHD requiring corrective procedures.¹ In 2012, the American Heart Association, in conjunction with the American Academy of Pediatrics, published a joint statement regarding neurocognitive outcomes in children with CHD and the need for more research addressing the effects of ongoing comorbidities.² In the statement, they recognized that children with CHD are at increased risk for developmental and cognitive abnormalities. The neurocognitive and psychosocial morbidity related to CHD adversely impacts educational achievements and health-related quality of life for many patients and was recently identified as a high-priority future research need by the National Institutes of Health (NIH)-National Heart, Lung, and Blood Institute.³ Although CHD can be associated with comorbid conditions causing neurocognitive impairment, such as neurodevelopmental syndromes, most children with CHD (89%) do not have a non-CHD syndromic diagnosis associated with neurocognitive impairment.⁴ Recognized factors that may contribute to neurocognitive impairment include chronic hypoxia (in cyanotic CHD), surgical procedures requiring cardiopulmonary bypass, anesthesia, and prolonged intensive care stays.² For the 1.5 million current survivors of CHD, limited options are currently available to mitigate the neurocognitive impairment associated with CHD.

OSA can negatively affect neurocognitive outcomes in children.^5,6 Prior studies have shown that individuals with CHD have an elevated apnea-hypopnea index and abnormal overnight oximetry compared with those without CHD.^7,8 No prior studies have evaluated the potential contribution of OSA to neurocognitive impairment in children with CHD. Given that OSA is a known cause of neurocognitive impairment in children without CHD,⁵ and may be prevalent in children with CHD, we hypothesized that OSA may contribute to neurocognitive impairment in children with CHD.

Materials and Methods

Participants were recruited from a local pediatric cardiology clinic that primarily provides care for children with CHD. Inclusion criteria included being age 6 to 17 years and having moderate to complex CHD (per Bethesda criteria⁹) that required prior intervention. Exclusion criteria included prematurity less than 37 weeks’ estimated gestational age; history of seizures; presence of neurodevelopmental syndrome that is associated with neurocognitive impairment independent of CHD, such as Down syndrome, DiGeorge syndrome, or other genetic disorders, or inability to complete neurocognitive testing. Children were eligible to participate in the study with or without a diagnosis of attention deficit hyperactivity disorder. We planned a priori to exclude any participants found to have hypoxemia independent of respiratory events on home sleep study (≥5 minutes with oxygen saturation <90%¹⁰); or presence of central sleep apnea on home sleep study (central apnea hypopnea index ≥ 5¹⁰) from neurocognitive analysis. The study was approved by the University of Arizona Institutional Review Board (#140427757). Parental informed consent and child assent were obtained from all participants before participation.

A home sleep study was performed for all participants, using an ApneaLink Air device (Resmed). This device measures airflow through nasal cannula, respiratory effort via a thoracoabdominal belt and heart rate, and oxygen saturation through finger pulse oximetry. The device has been previously validated in children.¹¹ Participants underwent one night of testing, with a repeat night performed if insufficient valid data were recorded. Studies were considered adequate if there were more than 2 hours with simultaneous adequate flow, effort, and oximetry data. Studies were scored based on American Academy of Sleep Medicine pediatric scoring rules. Briefly, an obstructive apnea was scored if there were a ≥90% reduction in airflow for at least two breaths’ duration, and persistent respiratory effort was seen throughout the event. A central apnea was scored if there were a ≥90% reduction in airflow with absent respiratory effort and the event lasted ≥20 seconds, or the event lasted at least two breaths’ duration and was associated with 3% or higher oxygen desaturation. An obstructive hypopnea was scored if there were 30% or more (but ≤90%) reduction in airflow for at least two breaths’ duration associated with a 3% desaturation as well as one of the following: persistent or increased respiratory effort, increased inspiratory flattening of the airflow signal, or snoring. A central hypopnea was scored if there were a 30% or more (but ≤90%) reduction in airflow for at least two breaths’ duration associated with a 3% desaturation and none of the following: persistent or increased respiratory effort, increased inspiratory flattening of the airflow signal, or snoring. In keeping with previous pediatric literature12, 13, 14, 15 and to improve interpretability, results are reported using apnea-hypopnea index rather than respiratory event index. An obstructive apnea-hypopnea index ≥ 1 was considered diagnostic for OSA, a central apnea-hypopnea index ≥ 5 was diagnostic of central sleep apnea, and oxygen saturation < 90% for greater than 5 minutes independent of respiratory events was diagnostic of nocturnal hypoxemia.¹⁰ Studies were scored by a board-certified pediatric sleep medicine physician blinded to neurocognitive testing results.

Participants underwent a neurocognitive evaluation within 1 week of home sleep study. The neurocognitive testing included Intellectual Quotient (IQ) testing, computerized neurocognitive assessment, and parent questionnaires. IQ testing was performed using the Kaufman Brief Intelligence Test, 2nd edition (K-BIT, Pearson). The K-BIT has been validated for a wide age range, from 4 to 90 years of age, and provides an age-adjusted total IQ, as well as verbal and performance subscale IQs. Computerized neurocognitive assessment was performed using three measures from CANTAB (Cambridge Cognition). All of the tests are validated in children and are validated for use in clinical trials.¹⁶ Previous research has shown differences in the selected tests between children with and without OSA.¹⁷ To minimize the risk of bias, the neurocognitive evaluation was performed blinded to home sleep study results.

The CANTAB Paired-Associates Learning task measures the learning of associations between nonverbalizable stimuli and their location within a spatial array on a touch-screen computer. The participant is asked to remember the spatial location associated with each pattern and to touch that place in response to the presentation of the pattern (one to eight patterns are displayed per trial, with eight attempts to solve each level until the participant is correct). The outcome measure was the mean number of errors until success across trials.

The CANTAB Intra-Extra Dimensional Set Shift test is a test of cognitive flexibility. In the initial stages, participants are first presented with two colored shapes, and they must learn which shape is “correct” through trial and error. After several trials of recognizing the correct rule, the “correct” shape is reversed. In later stages, a second shape is transposed onto each shape, so that the participant must take another dimension into consideration when determining which shape is “correct.” The task progresses from rule shifts within a dimension (ie, to a different stimulus of the same type) to responses outside of the trained dimension (ie, between shapes in which one has never been rewarded) across nine stages of increasing difficulty. The outcome measure was the number of stages completed.

In the CANTAB Simple Reaction Time test, participants press a button when a stimulus (a white box with variable onset) appears on a computer screen. The outcome measure was the mean correct latency.

In addition to this neurocognitive assessment, parents completed three surveys. Subjective neurocognitive function was measured using the Behavior Rating Inventory of Executive Function (BRIEF, Psychological Assessment Resources). The outcome measure was global executive composite score (t-scores adjusted for age and sex). T-scores for the BRIEF are computed from raw scores, using published normative values that account for age and sex from the BRIEF manual. Additionally, parents filled out a validated questionnaire on behavior (Conners 3rd edition, Pearson). The Conners ADHD (attention-deficit hyperactivity disorder) index was extracted from this and used to determine the likelihood of comorbid ADHD. Finally, the PedsQL family information form (Mapi Research Trust) was used to collect socioeconomic data from families in a subset of participants. The Hollingshead socioeconomic status index was used to stratify participant socioeconomic status. Given that not all participants provided data to compute their Hollingshead socioeconomic status index, we also used median income for each participant’s reported home zip code as an alternate measure of socioeconomic status. Data for median income by zip code were obtained from the American Community Survey 5-year estimate from 2013 to 2017.¹⁸

Medical records were reviewed to confirm eligibility and extract covariates, including age, sex, BMI, and CHD diagnosis. A board-certified pediatric cardiologist used Bethesda criteria to classify CHD as mild, moderate, or complex for all participants.⁹ BMI was converted to age- and sex-adjusted percentiles based on the Centers for Disease Control data.¹⁹

Statistical analysis was done using SPSS 25 for all analyses except linear regression, which was performed with Stata 14. The primary outcome was difference in total IQ between participants with and without OSA. Outcomes data were assessed for normality using the Shapiro-Wilk test. IQ, simple reaction test, and BRIEF global executive composite results were found to be normally distributed. The Paired Associates Learning test and Intra-Extra Dimensional set shift test were non-normally distributed. Unadjusted analysis was performed using χ² tests or Fisher exact tests (if more than 20% of cells had an expected value < 5) for categorical data and t-tests for continuous data if the normality assumption was not violated. For t-tests, equal variance was checked using Levene’s test of equality of variances. In cases in which variance was unequal, Satterthwaite’s approximation for the degrees of freedom was used for the t-test. For non-normally distributed continuous data, the Wilcoxon sum-rank test was used. All statistical tests were two-sided, and P < .05 was considered significant.

Given that CHD is a heterogeneous phenotype, we also examined the contribution of CHD subtype to neurocognitive impairment in our sample. Evidence has suggested that cyanotic CHD may be associated with more neurocognitive deficits.^2,20,21 Some evidence has suggested that increased CHD complexity is associated with worse neurocognition,²² although this study did not include cyanosis as a covariate, and others have not found this same association.²³ Cardiopulmonary bypass has been implicated as well, but there appears to be no difference in cognitive impairment in children who underwent cardiopulmonary bypass compared with children that underwent surgery without bypass.²⁴ Given this, we evaluated CHD-related covariates that may affect cognition, specifically age at first cardiac surgery, history of cyanotic CHD, CHD severity (Bethesda criteria⁹), number of cardiac surgeries, and Fontan circulation as possible covariates. We performed linear regression with stepwise variable selection (variables included for P < .1), evaluating these factors as predictors of total IQ. After this, we performed regression analysis evaluating the association between OSA and total IQ, with adjustment for CHD-related factors found to be significant predictors of total IQ.

Results

A total of 30 children had complete data available and were included in analysis. Participant recruitment information is presented in Figure 1. Briefly, a total of 109 potential participants were approached. Fifty-four potential participants declined, five because of distance from clinic, seven reported their child did not want to do the home sleep study, four parents could not take time off from work, one declined because they felt their child did not have sleep apnea, and 37 did not provide a reason for declining. Thirteen children did not meet inclusion criteria, five because of a comorbid syndrome associated with neurocognitive impairment, four because of simple CHD, two because of premature birth, one because of a seizure disorder, and one child was already on CPAP therapy for OSA. A total of 42 subjects were enrolled, and 12 subjects were excluded because of missing data (five without adequate home sleep test or neurocognitive testing, five without adequate home sleep test but complete neurocognitive testing, and two with successful home sleep study but who did not complete neurocognitive testing). One of the 30 included participants did not have valid IQ results because of poor test cooperation and was excluded from analysis for IQ results. Sensitivity analysis including this child did not materially change the results.

Figure 1.

Participant recruitment. Potential participants were approached at a pediatric cardiology clinic with a focus on congenital heart disease.

Seventeen of 30 (57%) of children were found to have OSA. OSA was mild (obstructive apnea-hypopnea index [AHI] of 1-5 events/hour) in 88% of cases, and moderate (obstructive AHI of 5-10 events/hour) in 12% of cases. No participants were found to have central sleep apnea or nocturnal hypoxemia unrelated to OSA. No significant differences were seen in age, BMI percentile, Conners ADHD Index score, or socioeconomic status (available for 16 participants). Female sex was more common in children with OSA (77% compared with 23%; P = .003). No differences were seen in CHD severity, number of cardiac surgeries, or history of cyanotic CHD (Table 1).

Table 1.

Characteristics of Children With CHD With and Without OSA

Characteristic	No OSA (n = 13)	OSA (n = 17)	P
Age, y	10.9 ± 3.7	9.8 ± 3.9	.41
Female sex	3 (23%)	13 (77%)	.004
Ethnicity
Latino	10 (77%)	13 (77%)	.35
White, Non-Latino	3 (23%)	2 (12%)
Native American	0 (0%)	2 (12%)
BMI percentilea	53 ± 29	70 ± 31	.14
Obese	0 (0%)	5 (29%)	.052
Hollingshead socioeconomic status	30 ± 7	31 ± 10	.69
Median income	$42,282 ± $12,723	$49,420 ± $18,951	.26
ADHD indexb	48 ± 38	42 ± 30	.65
CHD severity
Moderate	6 (46%)	9 (53%)	.57
Complex	7 (54%)	8 (47%)
No. cardiac surgeries	1.4 ± 1.1	1.6 ± 1.1	.62
History of cyanotic CHD	8 (62%)	13 (77%)	.44
Total apnea-hypopnea index	1.4 ± 1.2	3.3 ± 1.9	.003
Obstructive apnea-hypopnea index	0.7 ± 0.5	2.9 ± 1.6	< .001
Central apnea-hypopnea index	0.7 ± 1.0	0.4 ± 0.6	.26
Oxygen saturation nadir	89 ± 4.5	88 ± 2.6	.36
Minutes with oxygen <90%	0.0 ± 0.0	0.1± 0.5	.39

Data are presented as mean ± standard deviation. Obesity is defined as >95th percentile BMI for age and sex. Bold signifies P value < .05. ADHD = attention-deficit hyperactivity disorder; CHD = congenital heart disease.

^aBMI, age-, and sex-adjusted percentile.

^bConners attention-deficit hyperactivity disorder index. P values are derived using t-tests for continuous variables. For categorical variables, χ² tests were used for sex and ethnicity, and Fisher exact tests were used for obesity, CHD severity, and history of cyanotic CHD.

In unadjusted analysis, OSA was associated with a 12-point lower total IQ (95% CI, –21 to –3; P = .01). Similar results were seen for verbal IQ (12 points lower; 95% CI, –2 to –20; P = .02) and nonverbal IQ (10 points lower; 95% CI, –0.4, to –19; P = .04). Participants with neurocognitive impairment, defined as an IQ < 85 (1 SD below the population mean) had a significantly worse obstructive and total AHI compared with participants without neurocognitive impairment (Table 2). Complete neurocognitive results are summarized in Table 3.

Table 2.

Characteristics of Children With CHD With and Without Neurocognitive Impairment

Characteristic	No Cognitive Impairment (n = 21)	Cognitive Impairment (n = 8)	P
Age, y	10.8 ± 3.8	8.5 ± 3.2	.15
Female sex	48%	75%	.18
Ethnicity
Latino	76%	75%	.56
White, Non-Latino	14%	25%
Native American	10%	0%
BMI percentilea	63 ± 30	58 ± 37	.71
Hollingshead Socioeconomic status	31 ± 7	29 ± 10	.62
Median income	$47,263 ± $17,237	$43,902 ± $15,731	.62
ADHD indexb	45 ± 38	46 ± 22	.94
CHD severity
Moderate	48%	50%	1
Complex	52%	50%
No. cardiac surgeries	1.6 ± 1.1	1.3 ± 1.0	.43
History of cyanotic CHD	67%	88%	.26
Total apnea-hypopnea index	1.8 ± 1.5	4.0 ± 2.1	.004
Obstructive apnea-hypopnea index	1.4 ± 1.2	3.5 ± 1.9	.002
OSA	48%	88%	.06
Central apnea-hypopnea index	0.4 ± 0.6	0.8 ± 1.2	.40
Oxygen saturation nadir	88 ± 4.0	89 ± 2.2	.46
Minutes with oxygen < 90%	0.0 ± 0.0	0.2± 0.7	.35

Neurocognitive impairment is defined as total Intellectual Quotient score <85 (>1 standard deviation below the mean population Intellectual Quotient). Data are presented as mean ± standard deviation or percentage as appropriate. Bold signifies P value < .05. See Table 1 legend for expansion of abbreviations.

^aBMI, age, and sex-adjusted percentile.

^bConners attention-deficit hyperactivity disorder index. Median income is reported by participant zip code.

Table 3.

Neurocognitive Results for Children With CHD With and Without Comorbid OSA

Measurement	No OSA (n = 13)	OSA (n = 17)	P
Total IQ	98 ± 11	86 ± 12	.01
Verbal IQ	97 ± 11	86 ± 12	.02
Nonverbal IQ	99 ± 11	89 ± 13	.04
Simple Reaction Test Correct Latency (ms)	391 ± 137	523 ± 205	.09
Paired Associates Learning Test (total errors)	2 (1-8)	8 (2.25-15)	.02
Intra-Extra Dimensional set shift (stages completed)	7 (7-8)	7 (7-8.75)	.94
BRIEF Global Executive Composite (t-score)	55 ± 14	52 ± 11	.63

IQ is presented as standard score. An IQ standard score of 100 is equivalent to the 50th percentile, whereas an IQ score of 85 is equivalent to the 15th percentile. Data are presented as mean ± standard deviation for normally distributed data, and as median (interquartile range) for non-normally distributed data. Bold signifies P value < .05. IQ = Intellectual Quotient. See Table 1 legend for expansion of other abbreviation.

We performed linear regression with a stepwise variable selection for CHD-related variables that are likely to affect cognition. Evaluated variables included age at first cardiac surgery, history of cyanotic CHD, CHD severity, number of cardiac surgeries, and Fontan circulation as possible covariates. Only history of cyanotic CHD was an important predictor of total IQ (β, –14; 95% CI, –23 to –4; P = .009). We found that the presence of OSA was associated with a 10-point decrease (95% CI, –19 to −2) in total IQ after adjustment for history of cyanotic CHD (P = .01). In this analysis, OSA and cyanotic CHD were both evaluated as categorical variables. No interaction was found between cyanotic CHD and OSA (P = .48).

Given the significantly higher percentage of girls in the OSA group, we additionally performed linear regression, including sex and OSA as covariates. After adjustment for sex, OSA was associated with a 12-point decrease (95% CI, –23 to –1) in total IQ compared with participants without OSA (P = .03).

Children with CHD and OSA did significantly worse on the Paired Associates Learning test, with a median of 8 total errors (interquartile range [IQR], 2.25-15) compared with children with CHD without OSA (median total errors, 2; IQR, 1-8); P = .02. This difference was more pronounced when only performance on the most difficult stage (eight targets) was evaluated. Children with CHD and OSA had a median of 4 errors on this stage (IQR, 2-7.75) compared with a median of 1 error (IQR, 0-2) in participants without OSA; P = .003.

Participants with OSA had a delayed mean correct latency on the CANTAB Simple Reaction Test compared with those without OSA (mean latency of 523 ± 205 ms compared with 391 ± 137 ms), but this did not reach statistical significance (P = .09). No difference in performance on the CANTAB Intra-Extra Dimensional set shift test was seen. Participants with OSA completed a median of 7 (IQR, 7-8.75) stages, compared with a median of 7 (IQR, 7-8) stages for participants without OSA; P = .94.

No difference was seen in parent-reported neurocognition, as measured by the BRIEF global executive composite. Participants with OSA had a mean t-score of 52 ± 11 compared with 55 ± 14 in participants without OSA; P = .63 (lower scores indicate better neurocognitive performance).

Discussion

We found that OSA was common in children with CHD, with an incidence of 57% in our sample, compared with a typical community prevalence of 1% to 5% in children.²⁵ Additionally, although central sleep apnea has been found to be common in children with cardiomyopathies²⁶ and adults with heart failure,^27,28 we found no cases of central sleep apnea in this sample of children with CHD. Similarly, we found no cases of nocturnal hypoxemia in our sample. This is likely because all children in our sample had already undergone surgical procedures for their CHD. Likely hypoxemia would be much more common in younger children/infants who had not yet undergone surgical correction for cyanotic CHD or children who had undergone a Glenn or Norwood procedure but not yet had a Fontan procedure performed.

Mechanistic studies are needed to explore why children with CHD are at increased risk for OSA. Adenotonsillar hypertrophy is the most common cause of OSA in children, if children with CHD were at higher risk of adenotonsillar hypertrophy, this may explain the increased incidence of OSA in our study. To our knowledge, no data have been published on the subject of adenotonsillar hypertrophy in children with CHD. Some evidence suggests that infants with CHD associated with high pulmonary artery flow or pulmonary congestion may be at higher risk for atopic asthma,^29,30 which is associated with adenotonsillar hypertrophy.³¹ In our study, a diagnosis of asthma in their medical record was more common in children with CHD and comorbid OSA compared with children with CHD without comorbid OSA (21% vs 0%), but given our small sample size, this was not significant (P = .24 by Fisher exact test). Childrenwith CHD may have underlying defects in neural crest cell migration that underlie both craniofacial developmental abnormalities and CHD.³² Neural crest cells are critical to brain, craniofacial, and heart development. Therefore, possibly children with the phenotype of CHD also may have underlying craniofacial abnormalities predisposing them to OSA.³³ An example of this possibility is Digeorge syndrome, which is associated with neural crest abnormalities resulting in both increased risk of CHD and OSA.³⁴ Our study is limited by the absence of a craniofacial assessment as well as the absence of an assessment of adenoid or tonsillar hypertrophy. Additionally, children with Digeorge syndrome were excluded from this study because of the known association of Digeorge syndrome with neurocognitive impairment.³⁵ Finally, possibly underlying abnormalities of brain development in children with CHD may lead to both the increased risk of neurocognitive impairment and increased risk of OSA. A study on the effects of OSA treatment on neurocognition in children CHD is needed to address this possibility.

We found that OSA was associated with a markedly lower IQ in children with CHD. Children with CHD and comorbid IQ had a mean IQ 12 points lower than children with CHD without comorbid OSA. This difference was present even after adjustment for cyanotic heart disease and suggests that OSA is an independent risk factor for neurocognitive impairment in children with CHD. We found a significantly greater decrease in total IQ compared with prior research in children without CHD. For comparison, in the TuCASA study of children without CHD in the same geographic area as our sample, OSA was associated with only a three-point decrease in IQ.³⁶ This suggests that children with CHD may experience worse neurocognitive impairment related to OSA compared with children without CHD (Fig 2).

Figure 2.

Comparison of neurocognitive impairment associated with OSA with and without CHD. Comparative data for children without CHD is from the TuCASA study.³⁶ IQ scores for children without CHD were adjusted per K-BIT manual to account for differences between IQ tests used (Kaufman Brief Intelligence Test, 2nd edition vs Wechsler Abbreviated Scale of Intelligence). CHD = congenital heart disease; IQ = Intellectual Quotient.

OSA was also associated with memory impairment in children with CHD. OSA is known to affect the development of the hippocampus,³⁷ a critical structure for learning and memory, and particularly spatial associative memory tasks such as the Paired Associates Learning test.³⁸ Because the hippocampus is sensitive to OSA and hypoxia,³⁹ this structure may be particularly vulnerable to the combined effects of CHD and OSA. Research has shown that hippocampal changes are predictive of neurocognitive deficits in children with CHD.⁴⁰

Although we found significant objective neurocognitive impairment related to OSA in children with CHD, we did not find a difference in parent-reported subjective neurocognitive performance as measured on the BRIEF. The BRIEF is a measure of executive function, a subjective correlate to the objective results of the CANTAB Intra-Extra Dimensional set shift test, which also assesses executive function. Given that we did not see any difference in CANTAB Intra-Extra Dimensional set shift test performance in participants with and without OSA, possibly that OSA may not be associated with specific impairment to executive function in children with CHD. Additionally, it may be possible that an effect is present, but the magnitude of this effect is smaller than the effect of other CHD-related factors that have a greater influence on executive function. Given that results from the childhood adenotonsillectomy study showed improvement in BRIEF scores in children with OSA who underwent adenotonsillectomy compared with those who delayed treatment even though baseline BRIEF scores were normal, possibly OSA treatment may improve executive function in children with CHD and comorbid OSA. A subsequent study evaluating the neurocognitive effects of OSA treatment would be needed to determine this.

The primary limitation to our study is the use of home sleep studies in place of laboratory-based polysomnography. Home sleep studies were used to reduce participant burden and increase study participation. In-laboratory polysomnography would greatly increase the burden of participation for children in the study. Given the lack of prior studies of OSA in children with CHD, it was thought that this would reduce study participation. Although in-lab polysomnography is recommended for clinical practice, home sleep studies are used for research in children.12, 13, 14, 15 Three prior studies have validated the ApneaLink device used in our study for use in children.^11,41 The first study included 60 participants aged 0 to 22 years and, dependent on an AHI cutoff of one or five, found a sensitivity of 79% to 94% and specificity of 29% to 94% for the diagnosis of OSA compared with in-laboratory polysomnography. The authors noted that test characteristics were worse in young children and also reported results when only participants older than 10 years old were included. In older children, sensitivity was improved to 80% to 100% and specificity was improved to 50% to 64%. The second study included 35 children aged 4 to 18 years and reported a sensitivity of 94% and specificity of 61% compared with in-laboratory polysomnography. The authors also noted that ApneaLink-reported AHI was highly correlated with in-laboratory polysomnogram AHI (r = 0.89; P < .001). The final study included 25 children aged 9 to 18 years and, dependent on obstructive AHI cutoff of 1.5, 5, or 10, found a sensitivity of 80% to 100% and specificity of 46% to 100% compared with in-lab polysomnography. Based on these prior validation studies, the ApneaLink device appears to be sensitive for the detection of OSA in children, although the lower specificity may lead to the risk of overdiagnosis of OSA in children. In relationship to our study, this suggests the possibility that the high prevalence of OSA (57%) in our study may be an overestimation. However, any potential misclassification of participants would be expected to decrease the magnitude of any differences between the two groups. Therefore, our reported association of OSA with neurocognitive impairment in children with CHD is unlikely to be attributed to any potential misclassification (and may potentially be greater than our study estimates). Future studies would benefit from use of laboratory-based polysomnography, the standard of care for OSA diagnosis in children.⁴² Additionally, the requirement for only 2 hours of usable data could theoretically lead to inaccurate AHI values compared with an entire night of usable data. However, the mean length of useable data from our home sleep studies was 5.6 hours, reducing the risk of this possibility.

Our findings have additional limitations. First, a selection bias may have occurred in our study population. Parents with concerns about their child’s sleep may have been more likely to participate; possibly the prevalence of OSA reported in our study (57%) may be higher than the true population prevalence in children with CHD. However, this selection bias would have no impact on our results showing an association between OSA and neurocognitive impairment. Second, because of sample size limitation and risk of model overfitting, we did not perform multivariate analyses, accounting for a broad variety of covariates that may affect neurocognition in children. However, we restricted our sample population to eliminate common comorbidities (Down syndrome, prematurity, seizure disorders, and so forth) that would be likely to affect neurocognitive performance. Our limited sample size likely also limited our study’s sensitivity to detect OSA-related differences with a smaller effect size. An example of this may include our Cantab Simple Reaction Test results, in which the OSA group had a longer reaction time, but this did not reach statistical significance (P = .09). Additionally, the OSA and no-OSA groups were demographically homogenous, with only a difference in sex distribution between the two groups. Third, 44% of eligible participants approached agreed to participate in the study, which may potentially reduce the generalizability of our findings. However, this enrollment percentage is actually higher than that reported by several landmark pediatric sleep investigations. For example, the Tucson Children’s Assessment of Sleep Apnea (TuCASA) study conducted in the same geographic area enrolled approximately 30% of potential participants contacted about the study.⁴³ Finally, CHD is a heterogeneous group of diagnoses that may be associated with different degrees of neurocognitive impairment. To account for this, we restricted our population to only moderate to complex CHD. We additionally performed regression analyses to identify and account for the most significant CHD-related factor for neurocognition, history of cyanotic CHD.

Our study has several strengths as well. Participants are likely representative of children with CHD in general, because they were recruited from a cardiology clinic, not a sleep medicine clinic. Additionally, we used a robust neurocognitive battery measuring IQ as well as specific targeted tests of neurocognition. This is the first study to identify both that OSA appears common in children with CHD and that OSA is independently associated with neurocognitive impairment in children with CHD.

Conclusions

Our results suggest that OSA is common in children with CHD and is associated with neurocognitive impairment. Unlike other causes of neurocognitive impairment in children with CHD, OSA can be treated via adenotonsillectomy, medications, or positive airway pressure therapy. OSA treatment has been shown to improve neurocognitive outcomes in children without CHD.^44,45 Given that OSA is amenable to treatment, possibly OSA treatment may lead to improved neurocognitive outcomes in children with CHD. Further research is needed to investigate the effects of OSA treatment on neurocognitive impairment in children with CHD.

Take-home Point.

Study Question: Is OSA is associated with neurocognitive impairment in children with CHD?

Results: In children with CHD, OSA is associated with neurocognitive deficits including reduced IQ and worse performance on a test of memory.

Interpretation: OSA may be a potentially reversible cause of neurocognitive impairment in children with CHD, research on the neurocognitive effects of OSA treatment in children with CHD are needed.