There are currently estimated to be more than 2 million children and adults living in the United States with congenital heart disease (CHD). Reference Hoffman and Kaplan1,Reference Gilboa, Devine and Kucik2 As a result of advances in diagnostic tools, surgical approaches, perioperative care, and longitudinal follow-up, survival after cardiac surgery has improved dramatically over the past few decades. Reference Martin and Jonas3 Alongside this increased survival, there is now a greater recognition that children with CHD are at significant risk of long-term neurodevelopmental impairment. Reference Gaynor, Stopp and Wypij4 In addition to the well-described neurodevelopmental (cognitive, learning, and motor) Reference Brosig, Bear and Allen5 and neurobehavioral (anxiety, depression, attention-deficit/hyperactivity disorder) Reference Calderon and Bellinger6,Reference DeMaso, Calderon and Taylor7 impairments, there has been increasing concern regarding deficits in social cognition and symptoms of autism spectrum disorder in children with CHD. Reference Bellinger8–Reference Wier, Yoshida, Odouli, Grether and Croen11 Autism spectrum disorder is a neurodevelopmental disorder marked by deficits in social interaction and social communication, as well as the presence of restricted interests and/or repetitive behaviours. Reference Hyman, Levy and Myers12 Several factors have been associated with increased risk for autism spectrum disorder such as decreased gestational age, prematurity, and seizures. Reference Kuzniewicz, Wi, Qian, Walsh, Armstrong and Croen13
In the most recent report from the Centers for Disease Control and Prevention autism surveillance programme (Autism and Developmental Disabilities Monitoring Network), autism spectrum disorder prevalence was 17.0 per 1,000 (1 in 59, or 1.7%) for children 4 years of age and 23.0 per 1,000 (1 in 44 or 2.3%) for 8 years of age. Reference Maenner, Shaw and Bakian14,Reference Shaw, Maenner and Bakian15 Recent research on children with CHD reveals that they are more likely to have a positive screening (3.2%), receive a diagnosis (2.6%), or be identified as ‘at risk’ (8.2%) for autism spectrum disorder than children without CHD. Reference Razzaghi, Oster and Reefhuis10,Reference Bean Jaworski, Flynn and Burnham16,Reference Tan, Semmel, Wolf, Hammett and Ilardi17 Furthermore, larger studies using international diagnostic codes to define autism spectrum disorder diagnosis showed higher incidence rates of autism spectrum disorder in CHD patients than in a control (non-CHD, matched for age, sex, and enrolled time) group of patients in Taiwan (3%), while in the United States, a study showed that patients diagnosed with autism spectrum disorder had higher odds of having CHD compared to controls (matched for date of birth, sex, and time of enrolment in the health system – OR 1.85 95% 1.65–2.10 n = 1063). Reference Sigmon, Kelleman, Susi, Nylund and Oster18,Reference Tsao, Lee and Jeng19 Autism spectrum disorder can be reliably diagnosed as early as 18–24 months of age; however in the United States, autism spectrum disorder diagnosis generally tends to occur later and varies by site (e.g. between 36 m and 63 m depending on the geographic location), possibly due to the timing of routine screening. Reference Hyman, Levy and Myers12,Reference Maenner, Shaw and Bakian14,Reference Pierce, Gazestani and Bacon20,Reference Shaw, McArthur and Hughes21 Limited data in the CHD population have suggested a much later age at diagnosis of 60 months. Reference Tsao, Lee and Jeng19 We hypothesised that children with CHD would be at risk of developmental delays including autism than children without CHD. The aim of our current study was to investigate the prevalence, age at diagnosis, and potential risk factors for autism spectrum disorder in a cohort of children who have undergone surgery for CHD at the same medical centre and are followed in a specialised multi-disciplinary cardiac developmental outcomes clinic.
Materials and methods
Patients
This was a retrospective, single-centre cohort study of patients born between January 1st 2013 and December 31st, 2016 who were followed at the Cardiac Developmental Outcomes Programme clinic. For purposes of this study, only clinic visits until July 2022 were included in the analysis. Given that part of the period of follow-up included the early phase of the COVID-19 pandemic, clinic visits were a combination of telemedicine and in-person. For those patients for which concerns for autism spectrum disorder were reported or observed during a telemedicine visit, an in-person assessment was offered. The study was approved (with a waiver of consent) by the Institutional Review Board. The Cardiac Developmental Outcomes Programme Clinic was established in 2013 and offers routine longitudinal neurodevelopmental assessments at 6, 12, 18, and 24 months of age and yearly thereafter for children (from infancy to adolescence) who underwent cardiac procedures during the first 6 months of life. Reference Monteiro, Serrano and Tsang22 In addition, the clinic offers a routine follow-up to all infants and children with a history of extracorporeal membrane oxygenation or ventricular assist device (VAD) support; children undergoing transplant; and other children with CHD and developmental/learning/behavioural concerns. Before the patient is discharged from the hospital, the Cardiac Developmental Outcomes Programme clinic coordinator visits the patient, provides information to the parents about the programme, and offers to schedule the 6 months evaluation. Outpatient evaluations are completed by a team of dedicated developmental-behavioural paediatricians and a psychologist. Patient demographic, diagnostic, clinical data relating to the infant’s hospitalisation, and the results of their neurodevelopmental assessments are all recorded in a dedicated Institutional Review Board-approved Cardiac Developmental Outcomes Programme clinical database. During the study period, there were 657 patients referred to the Cardiac Developmental Outcomes Programme clinic. Of those, 240 (37%) did not attend the programme and thus were not included in the study and 417 (63%) attended the programme. The main criteria for inclusion in this study were surgery during the first 6 months of life and at least one clinic assessment at 24 months of age or older. Patients were excluded from the study if they did not have surgery or if surgery occurred after the age of 6 months (56 patients excluded). We identified 361 who fulfilled the inclusion criteria for this study.
Demographic and clinical data
Data were collected from the Cardiac Developmental Outcomes Programme outcome database and from the Electronic Medical Record (EPIC Systems Corporation, November 2019). Data included age, sex, gestational age, maternal age, prenatal diagnosis, prematurity (<37 weeks), race, ethnicity, language, insurance type, surgical complexity (The Society of Thoracic Surgeon-European for Cardiothoracic Surgery (STAT) score) for index surgery, history of non-febrile seizures or a diagnosed seizure disorder, stroke, and history of extracorporeal membrane oxygenation.
Autism diagnosis
The diagnoses of autism spectrum disorder were made through the clinical neurodevelopmental assessments that are a component of the routine clinical care provided in the Cardiac Developmental Outcomes Programme clinic. During the routine clinical care visit, each infant undergoes a comprehensive clinical and neurological examination with standardised developmental testing. Experienced clinicians based their autism spectrum disorder diagnosis on Diagnostic and Statistical Manual of Mental Disorders-5 criteria 23 and by either the Childhood Autism Rating Scale 2-Standard Reference Schopler, Van Bourgondien, Wellman and Love24 or the Autism Diagnostic Observation Schedule, 2nd Edition (ADOS-2). Reference Lord, DiLavore, Risi, Gotham and Bishop25,Reference Luyster, Gotham and Guthrie26 Developmental-behavioural paediatricians utilised the Childhood Autism Rating Scale 2-Standard for their diagnosis and if further evaluations were needed, patients were referred to a neuropsychologist and the Autism Diagnostic Observation Schedule, 2nd Edition was utilised.
Statistical analysis
All baseline characteristics were summarised using mean (standard deviation), median (25th and 75th percentiles), or frequency (percentage). The prevalence of autism spectrum disorder was calculated with an exact binomial 95% confidence interval. Characteristics were summarised by autism spectrum disorder and compared using an independent two-sided t-test, Wilcoxon rank-sum test, chi-squared test, or Fisher’s exact test as appropriate. Logistic regression was used to assess the association between characteristics and the odds of autism spectrum disorder. A multivariable logistic regression model included all characteristics that were associated with the odds of autism spectrum disorder with a p-value less than 0.1 in univariable analysis. Any variable with greater than 10% missing data was not included in the logistic regression analysis. Due to collinearity between the timing variables, only age at surgery was included in the final multivariable model but not age at the first Cardiac Developmental Outcomes Programme visit and time between surgery and the first Cardiac Developmental Outcomes Programme visit. All statistical analyses were performed using Stata v 15.1 (StataCorp). A p-value of less than 0.05 was considered statistically significant.
Results
The median (IQR) age of the 361 children in this cohort at the time of the study analysis was 7.7 (6.6, 8.6) years. The median (IQR) age at first cardiac surgery was 15 (7.0, 51.0) days, median (IQR) age at the time of the first clinic visit was 7 (6.2, 9.4) months, and the median (IQR) number of visits to the clinic was 4 (2.0, 6.0) (Table 1). The majority of patients were male (59.8%); 42.9% were non-Hispanic Whites; 83.9% were English-speaking, and 47.4% had private insurance. Furthermore, less than 20% of patients were premature, and more than half had an antenatal diagnosis of CHD (53.7%). The majority (61.5%) of the patients in this cohort showed major surgical complexity (The Society of Thoracic Surgeon-European for Cardiothoracic Surgery 4–5) for first surgery and had surgery occurring < 30 days of age (64%). A history of seizures, stroke, or the need for extracorporeal membrane oxygenation was seen in less than 10% of patients in this cohort (Table 1).
Table 1. Summary statistics of all Cardiac Developmental Outcomes Programme clinic patients.
ECMO=Extracorporeal membrane oxygenation; GA=gestational age; IQR=interquartile range; STAT category=the Society of Thoracic Surgeon-European for cardiothoracic surgery score.
Of the 361 children in the cohort, 33 (9.1%) of the patients have been diagnosed with autism spectrum disorder (95% CI: 6.48, 12.60) by the end of the follow-up period, at a median age of diagnosis of 34 (25.0, 52.0) months. Table 2 summarises demographic and patient characteristics during infancy. Median gestational age was lower among those with autism spectrum disorder (p = 0.035). The majority of those diagnosed with autism spectrum disorder were male (n = 29; 87.9%) and the proportion of males with autism spectrum disorder is higher compared to those without (p ≤ 0.001). Patients with autism spectrum disorder were older at first surgery, 54 days versus 13.5 days (p = 0.002), and were older at their first clinic visit, 10.4 versus 6.8 months (p ≤ 0.001). Further stratification, according to first surgery before 30 days of age (64% of the cohort) or older (36%), revealed that the majority of those with autism spectrum disorder (54.5%) underwent cardiac surgery ≥ 30 days of age (p = 0.023). Furthermore, there was an increased likelihood of an autism spectrum disorder diagnosis for those born premature (i.e. < 37 weeks, p = 0.013), who were placed on extracorporeal membrane oxygenation (p = 0.023), or had history of non-febrile seizures (p = 0.001). There were no significant differences in maternal age, race, ethnicity, language, insurance type, antenatal diagnosis, the Society of Thoracic Surgeon-European for Cardiothoracic Surgery category, or history of stroke between patients with or without autism spectrum disorder.
Table 2. Demographics and infant characteristics of CHD patients.
ASD=autism spectrum disorder; ECMO=extracorporeal membrane oxygenation; IQR=interquartile range; SD=standard deviation; STAT category=the Society of Thoracic Surgeon-European for cardiothoracic surgery score.
Unadjusted logistic regression analysis showed that the odds of autism spectrum disorder were 5.5 times higher for males compared to females (95% CI 1.88, 15.93; p = 0.002). The odds of autism spectrum disorder were 2.31 higher in infants who had surgery at ≥ 30 days of age (95% CI 1.12, 4.77; p = 0.023) than < 30 days. Furthermore, the odds of autism spectrum disorder were 2.8 times higher for babies who were born prematurely (95% CI 1.29, 6.22; p = 0.009), 4.89 times higher in those on extracorporeal membrane oxygenation (95% CI 1.42, 16.88; p = 0.012), and 5.20 times higher (95% CI 2.07, 13.09 p=<0.001) in patients with a history of non-febrile seizures. Multivariate logistic regression analysis including only variables associated with autism spectrum disorder showed that adjusting for sex (OR 4.85; 95% CI 1.62, 14.53; p = 0.005), age at surgery ≥ 30 days of age (OR = 2.46; 95% CI 1.12, 5.59; p = 0.025), extracorporeal membrane oxygenation (OR = 4.91; 95% CI 1.23, 19.57; p = 0.024), and seizures (OR = 4.32; 95% CI 1.63, 11.44; p = 0.003) odds of autism spectrum disorder were still significantly higher (Table 3).
Table 3. Multivariable logistic regression for autism.
CI=confidence interval.
Discussion
Our study demonstrates that early, routine neurodevelopmental follow-up after cardiac surgery in infancy can lead to earlier identification of autism spectrum disorder than has previously been described. The average age of diagnosis in the general population is 4 years, and in previous reports of children with CHD – the age at diagnosis was 60 months. Reference Tsao, Lee and Jeng19 In our patients, who were seen in the Cardiac Developmental Outcomes Programme clinic from infancy onwards, the median age at diagnosis was 34 months. It is important to note that for telemedicine visits during the pandemic due to the need for a second (in-person) visit to make or confirm a diagnosis, this may have added to the age at diagnosis for some patients. Furthermore, in our clinic population, we found a higher rate of autism spectrum disorder diagnosis that has been previously reported in the general population as well as in studies limited to children with CHD. We also have identified several important risk factors for this including prematurity, the need for extracorporeal membrane oxygenation, and a history of seizures. Additionally, in contrast with the accepted ‘norm’ regarding neurodevelopmental delay after surgery for CHD in infancy, autism spectrum disorder was not more prevalent in children who had undergone more complex cardiac surgery and/or surgery at a younger age. In this cohort, only 36% of the patients had their first surgery ≥ 30 days; nevertheless, they had a significantly higher autism spectrum disorder diagnosis than those with first surgery < 30 days (54% versus 45.5%, respectively). This underscores the message that an outcome programme that limits follow-up to traditionally ‘high-risk patients’ (undergoing surgery during the neonatal period) would miss the opportunity for early diagnosis of autism spectrum disorder in an important proportion of patients. In our Cardiac Developmental Outcomes Programme clinic, the median age of autism spectrum disorder diagnosis in children with CHD was 34 months, significantly earlier than the 60 months of age reported in children with CHD, Reference Tsao, Lee and Jeng19 and this offers the opportunity for earlier intervention before typical school age. A national survey of parents caring for children with autism showed that obtaining an autism spectrum disorder diagnosis is often delayed, despite parents often having concerns early on but then experiencing long wait times and in some cases multiple appointments with several specialists before receiving a diagnosis. Reference MacKenzie, Mazefsky and Eack27 Our findings are consistent with recommendations from the Cardiac Neurodevelopmental Outcomes Collaborative, which recommends routine screening for autism from infancy through adolescence. Reference Tan, Semmel, Wolf, Hammett and Ilardi17 Furthermore, an earlier diagnosis of autism spectrum disorder can facilitate the introduction of therapeutic interventions that may lead to improved developmental outcomes for these children. Reference Dawson, Rogers and Munson28–Reference Reichow31
In addition to our study demonstrating that autism spectrum disorder can be diagnosed earlier in children with CHD, we also found that the rate of autism spectrum disorder was significantly higher compared to the general population. In the most recent report from the CDC autism surveillance programme, autism spectrum disorder prevalence was between 1.7% and 2.3% for children 4 years of age and 8 years of age, respectively. Reference Maenner, Shaw and Bakian14,Reference Shaw, Maenner and Bakian15 Although several studies suggest that children with CHD are at a higher risk of having autism spectrum disorder, Reference Hultman, Sparen and Cnattingius9,Reference Razzaghi, Oster and Reefhuis10,Reference Bean Jaworski, Flynn and Burnham16–Reference Tsao, Lee and Jeng19 our single-centre study relied on confirmed autism spectrum disorder diagnosis by clinic providers using autism spectrum disorder diagnostic assessments and found that 9.1% of the CHD patients attending the clinic have autism spectrum disorder. Consistent with studies in the general population, we found several factors associated with higher odds of autism spectrum disorder in the CHD population. For instance, studies in pre-term infants have shown a higher prevalence of autism spectrum disorder, with each week of shorter gestation associated with higher risk. Reference Kuzniewicz, Wi, Qian, Walsh, Armstrong and Croen13 In our premature group (17%), we found higher odds for autism spectrum disorder in the unadjusted analysis; however, when adjusting for several factors, this association was no longer significant. One possible explanation could be that our cohort consisted mostly of moderate to late pre-term (32– < 37 weeks) infants. As infants with CHD are at risk of being born prematurely and prematurity is a known risk factor for neurodevelopmental impairment, this population should be longitudinally followed. Reference Goff, Luan and Gerdes32,Reference Pierrat, Marchand-Martin and Marret33 Several co-occurring medical conditions, including seizures, are common in children with autism spectrum disorder. Reference Hyman, Levy and Myers12 A population-based study found that prevalence of autism spectrum disorder is higher in children with a history of seizures than in the general population. Reference Saemundsen, Ludvigsson, Hilmarsdottir and Rafnsson34 Furthermore, in the general population, it has been reported that 20% of individuals with epilepsy have autism and thus this association is not surprising. Reference Besag35 We found odds of autism spectrum disorder higher in patients with history of seizures, and it has been well documented that the occurrence of seizures in patients with CHD is consistent with worse neurodevelopmental outcomes, again demonstrating the importance of continuous surveillance. Reference Bellinger, Wypij and Rivkin36,Reference Rappaport, Wypij and Bellinger37 Although the group of patients placed on extracorporeal membrane oxygenation in our cohort was small, the odds of autism spectrum disorder were still significant after adjusting for several factors. Extracorporeal membrane oxygenation is lifesaving for many critically ill cardiac patients; however, there is an increased risk for neurologic complications (i.e. stroke, seizures) and long-term neurodevelopmental delays associated with it. Reference Boyle, Felling and Yiu38,Reference Sadhwani, Cheng and Stopp39 To our knowledge, this is the first study demonstrating an association between extracorporeal membrane oxygenation and autism spectrum disorder in the CHD population, and thus, it will be of importance to corroborate in a larger cohort of patients.
The association between autism spectrum disorder and older age at surgery deserves further discussion, particularly given that surgery during the neonatal period has generally been considered to be associated with a higher risk of developmental impairment. Sigmon et al (2019) compared autism spectrum disorder with multiple CHD subtypes and found a higher risk of autism spectrum disorder among children with less severe forms of CHD, such as atrial septal defects and ventricular septal defects, than more complex types of CHD. Although we did not find a relationship between a diagnosis of autism spectrum disorder and surgical complexity according to the Society of Thoracic Surgeon-European for Cardiothoracic Surgery category, our clinic population is currently limited to patients who had surgery for CHD during early infancy, the majority of whom were of higher surgical complexity. However, our finding that those undergoing surgery beyond 30 days of age had a threefold increased risk for autism spectrum disorder supports the observation of Sigmon et al. Reference Sigmon, Kelleman, Susi, Nylund and Oster18
There are some important limitations that should be considered. First, although we have a relatively large sample size, this is a single-centre cohort that may not be representative of other outcomes programmes. Second, it is possible that our higher-than-expected incidence of autism spectrum disorder could be related to sampling biases as patients with concerns for possible delays or existing impairments may be more likely to participate than patients who appear to be developmentally appropriate and whose parents are not concerned. However, our clinic schedule, with the initial visit being scheduled at a relatively early stage in development (6 months of age), may make this less likely. Third, data regarding the rate of autism diagnosis in patients who did not attend our follow-up programme were not available for review. Fourth, although a genetic link between autism spectrum disorder and CHD has been suggested, it was beyond the scope of this study. Reference Rosenthal, Willsey and Xu40 Finally, further studies will be needed in order to postulate a mechanism for our findings or for the apparent relationship with age at surgery.
Acknowledgements
The authors acknowledge the efforts of Estrella Mazarico De Thomas for her assistance in collecting data and assisting families during their clinic visits. To the TCH outcomes/IS team and in particular to Karla Dyer for her assistance in collecting data for this manuscript.
Financial support
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Competing interests
The authors have no conflicts of interest relevant to this article to disclose.
