Neuroimaging and Neurodevelopmental Outcomes Among Individuals With Complex Congenital Heart Disease

Abstract

Although neuroimaging advances have deepened our understanding of brain health in individuals with congenital heart disease (CHD), it is less clear how neuroimaging findings relate to neurodevelopmental and mental health outcomes across the lifespan. We systematically synthesized and critically evaluated evidence on associations between neuroimaging and neurodevelopmental, neurocognitive, psychiatric, or behavioral outcomes among individuals with transposition of great arteries or single-ventricle CHD (Protocol CRD42021229617). Six databases were searched and 45 papers from 25 unique studies were identified. Structural brain injury was generally linked to poorer neurodevelopment in infancy. Brain volumes and microstructural and functional brain changes appear linked to neurocognitive outcomes, including deficits in attention, learning, memory, and executive function in children and adolescents. Fetal neuroimaging studies were limited. Four papers investigated psychiatric outcomes; none found associations with neuroimaging. Multicenter, longitudinal studies incorporating functional neuroimaging and mental health outcomes are much-needed to inform early neuroprotective and therapeutic strategies in CHD.

With increased survival of infants with complex congenital heart disease (CHD),¹ focus has sharpened on the neurodevelopmental sequalae and mental health of individuals with these conditions.² Vulnerability to adverse neurodevelopmental outcomes is well-documented in this population, including delays or deficits in fine and gross motor skills,^3,4 language,^5,6 attention,^6,7 visual-spatial integration,⁶ social cognition,^6,8 executive functioning,^9,10 and academic achievement and adaptive skills^8,11 across the lifespan.¹² Causes of neurodevelopmental disability are multifactorial, additive, and interactive,¹² including genetic and epigenetic factors,^13,14 disease complexity and comorbidity,¹⁴ prematurity, history of mechanical support, cardiopulmonary resuscitation, perioperative seizures or stroke,^6,15 prolonged intensive care unit and total hospital stay,¹⁵ poor nutrition,¹⁶ severe psychological stress,¹⁷ and socioeconomic factors.¹⁸ Compared with other forms of CHD, individuals with single-ventricle disease or transposition of the great arteries (TGA) are reported to have a greater vulnerability to neurological sequalae, with longer-term implications for neurocognitive functioning, mental health, and overall quality of life.⁶ The lifetime prevalence of anxiety and depression among individuals with CHD, eg, is estimated at 50%; this rate far exceeds that in the general population.¹⁹

Advances in neuroimaging techniques, such as electroencephalography (EEG), transcranial Doppler, near-infrared spectroscopy (NIRS), and multimodal magnetic resonance imaging (MRI), have expanded our understanding of brain health and development in the context of complex CHD.^6,20 Studies using these techniques have yielded evidence of restricted oxygen delivery during the fetal and preoperative periods,^21,22 delayed brain maturation,²³ alterations in structural and functional brain networks compared with typical development,²⁴ and brain injuries sustained perioperatively.²⁵ Meta-analytic neuroimaging data indicate that one-third of patients with TGA (34%) and one-half of patients with left-sided heart lesions (49%) sustain preoperative brain injury, including white matter injury (WMI), ischemic lesions, and ventriculomegaly.²⁰ Recently, meta-analytic data showed brain abnormalities, including reduced cortical volumes and disruptions to microstructural integrity, were approximately 8 times more likely in adolescents and adults with complex CHD (ie, CHD requiring open-heart surgery) than typically developing individuals without CHD.²⁶ What is less clear is how neuroimaging findings relate to neurodevelopmental and mental health outcomes in the complex CHD population.

To optimize clinical care and guide future research, this review aimed to systematically identify, synthesize, and critically evaluate evidence on associations between neuroimaging findings and neurodevelopmental, neurocognitive, psychiatric, and behavioral outcomes among individuals with single-ventricle CHD or TGA across the lifespan. We were especially interested in investigating the potential utility of various neuroimaging techniques, including MRI and NIRS, in predicting subsequent neurodevelopmental and mental health outcomes in this population. A secondary aim was to investigate potential differences in findings across patient ages and developmental stages, including infancy, early and later childhood, adolescence, and adulthood, as well as CHD types (ie, TGA vs single-ventricle CHD).

METHODS

SEARCH STRATEGY AND SOURCES.

The review protocol was prospectively registered with PROSPERO (CRD42021229617). The Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines²⁷ were followed to identify, extract, and synthesize data from eligible papers. Six electronic databases were systematically searched (MEDLINE, PsycINFO, the Cumulative Index to Nursing and Allied Health Literature, EMBASE, Scopus, Emcare) to identify relevant literature from inception to June 2021. The search strategy was developed in MEDLINE and adapted for use in each database. Search terms were designed to identify the target population (eg, transposition of the great arteries, hypoplastic left heart syndrome, congenital heart disease) and outcomes of interest (ie, neuroimaging, neurodevelopment, behavior, emotion, mental health) (Supplemental Table 1). To maximize the number of papers returned, date range restrictions were not applied. Searches were limited to human studies published in the English language in a peer-reviewed journal. Ancestry methods, citation chaining, and ancestry and prolific author searching were used to identify additional papers. Titles were screened by 1 reviewer (K.P.) and all remaining abstracts and full-texts were independently screened by 2 reviewers (K.P., F.A.). Any discrepancies in screening were resolved through discussion with a third reviewer (N.K.). Auto-alerts were created using the same unique search algorithm for each database, with findings incorporated in the review until August 2022.

STUDY SELECTION CRITERIA.

Studies were eligible for review if the following criteria were met:

1. Study design: Cross-sectional, case control, retrospective, prospective cohort, and intervention studies, including longitudinal studies, were included. Studies with a control or comparator sample, or no control group were eligible for inclusion. Case studies, reviews, and nonpeer-reviewed or gray literature were excluded. Although excluded from analysis, review papers identified during the search (21 papers) were collected and screened to ensure concept originality, which was confirmed; we found no review papers presenting results on associations between neuroimaging and neurodevelopmental or mental health outcomes in this population.

2. Participants: Individuals of all ages, from fetus to adult, with a diagnosis of TGA or single-ventricle CHD. Studies reporting on preoperative and/or postoperative findings were included. Studies including participants with mixed congenital heart anomalies or control samples were included if results for patients with TGA or single-ventricle CHD were reported separately or if the total sample comprised ≥80% of participants with these CHD types.

3. Outcomes: Studies reporting on neuroimaging data (any form) and neurodevelopmental, neurocognitive, psychiatric, emotional, or behavioral outcomes (any form) were included. In terms of neuroimaging, structural findings obtained via MRI or computed tomography, microstructural findings obtained via diffusion tensor imaging (DTI), and functional or physiological results obtained via functional MRI (evoked or resting-state), fetal cranial ultrasound (Doppler), or NIRS were included. Neurodevelopmental, neurocognitive, psychiatric, emotional, or behavioral outcomes included any functional outcomes, such as IQ or general cognitive functioning, fine and gross motor skills, speech, language, attention or visuo-spatial integration skills, social cognition, executive function, neurological function, emotion or behavior regulation, or psychological symptoms, assessed using a standardized or validated measure. Outcomes also included diagnosis of a psychiatric or neurodevelopmental disorder, such as anxiety, depression, post-traumatic stress disorder, attention deficit/hyperactivity disorder (ADHD), or autism spectrum disorder. Studies were included only if they reported on both neuroimaging and neurodevelopmental or mental health outcomes, and directly tested associations between these (eg, using regression or correlation analyses). For this study, neuroimaging findings were reported only if these data were used to examine associations with the neurodevelopmental or mental health outcomes of interest.

DATA EXTRACTION.

Data were initially extracted by 1 reviewer (K.P.) using a standardized, prepiloted form to collect the following: publication characteristics (eg, authors, year), study characteristics (eg, design, setting), sample characteristics (eg, sample size, demographic factors such as patient age, sex), outcome measurement (eg, assessment methods, scoring), main results, and methodological strengths and limitations. All data were checked for completeness and accuracy by a second reviewer (F.A.); the level of agreement between reviewers was extremely high (11 differences identified, all typographical errors).

RISK OF BIAS ASSESSMENT.

Risk of bias assessment was independently assessed at the paper level by 2 reviewers (K.P., F.A.) using the appropriate National Institutes of Health Quality Assessment tools: 1 tool for observational cohort and cross-sectional studies, and 1 for controlled intervention studies.²⁸ Any conflicts were resolved through consultation with a third reviewer (N.K.) until consensus was reached. The National Institutes of Health Quality Assessment tools include 14 items to appraise quantitative research in terms of study design, methodology, data analysis, and outcome reporting. Each item is rated as “yes,” “no,” or “other” (eg, “cannot determine”) and studies are rated as “good” (ie, low risk of bias), “fair” (ie, moderate risk of bias), or “poor” (ie, high risk of bias). An a priori decision was made to exclude from data synthesis studies rated as “poor.”

DATA SYNTHESIS AND ANALYSIS.

Narrative synthesis was the primary reporting method, with findings presented by patient age and outcome assessed. Statistical synthesis using meta-analysis was considered but was not possible because of the heterogeneity in study designs, outcome assessment and timing, and sample characteristics.

RESULTS

Following screening of 2,723 unique titles, 626 papers remained for abstract review, and 207 papers were retained for full-text review. Agreement between reviewers was 83% (Cohen’s Kappa = 0.62, indicating substantial agreement). Six additional papers were identified and reviewed at the full-text level after citation chaining and prolific author searching, but none met inclusion criteria. A total of 56 papers were assessed for risk of bias: 10 papers were scored as “good,” 35 rated “fair,” and 11 rated “poor” and were excluded from further analysis, yielding a total of 45 papers (25 unique studies/samples) for review (Figure 1). Across the 56 papers identified, common methodological problems included lack of reported power calculations (87.5%), small samples (48% with a sample <30), inadequate descriptions of sample characteristics (26.7%) or statistical methods (32.1%), and high attrition (23.2% with attrition >20%).

FIGURE 1. PRISMA Flow Diagram of Systematic Review Process.

Flow diagram depicts study selection process in the present systematic review. Papers were identified through database searching, then systematic screening at the title and abstract level. Full-text papers were then screened, with potentially eligible papers assessed for risk of bias. There were 45 papers included in the quantitative synthesis. PRISMA = Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

STUDY CHARACTERISTICS.

Across studies, a total of 1,462 individuals with TGA or single-ventricle CHD were studied, with sample sizes ranging from 16 to 171 participants. Seven samples included TGA patients only, 9 included single-ventricle CHD patients, and 9 included both TGA and single-ventricle CHD patients (Table 1, Supplemental Table 2). Most studies were performed at a single center (n = 19, 76%), used a cross-sectional design (n = 18, 72%), and were carried out in North America (n = 15, 60%), followed by Europe (n = 7, 28%), Australia and New Zealand (n = 2, 8%), and Japan (n = 1, 4%). A total of 21 papers included a control or comparison group; of these, 20 included a CHD-free comparison group and 1 included patients with other CHD diagnoses. Throughout this next section, findings are presented by patient age, neuroimaging technique, and outcome examined (Central Illustration).

TABLE 1.

Characteristics of Included Studies

Study characteristics (number of unique samples = 25)
Type of CHD
Dextro-TGA	7
Single-ventricle CHD	9
Mixed TGA and single-ventricle CHD	9
Study design
Cross-sectional	18
Longitudinal	7
Recruitment
Single center	19
Multicenter	6
Sample size (at first study assessment)
<30	5
30–50	11
51–100	4
>100	5
Brain imaging modality
MRI
Structural MRI	14
fMRI	1
DTI	5
EEG	4
NIRS	5
Transcranial Doppler	2
Fetal Doppler	2
Outcome assessed
Infant or toddler neurodevelopment	17
Intelligence	9
Executive function	5
Attention	5
Learning and memory	6
Visual-perception	2
Academic outcomes	4
Neurological examination	4
Psychiatric symptoms or diagnoses	3
Social cognition/emotion recognition	3
Othera	4
Article characteristics (number of total papers = 45)
Year of publication
Before 2015	16
2015 or later	29
Age group at outcome measurement
Infant (≤12 mo)	8
Toddler (13 mo to 3 y)	13
Pre-school age (4–5 y)	3
School-age child (6–12 y)	3b
Adolescent (13–19 y)	14
Adult (≥20 y)	0
Mixed agesc	4

Values are n.

^aStudies including “other” outcomes were: Bellinger et al,⁵² 1999 (apraxia of speech and motor outcomes), Hansen et al,⁵³ 2016 (global and verbal cognitive abilities), Hoffman et al,⁵⁴ 2013 (visual-motor integration), and Neukomm et al,⁵⁹ 2022 (motor outcomes and visual-motor integration).

^b3 studies in childhood plus 2 studies (Rollins et al,⁵⁶ 2014, and Panigrahy et al,⁵⁵ 2015) where magnetic resonance imaging (MRI) done in adolescence was retrospectively associated with childhood outcome assessment.

^cStudies including mixed ages were: Peyvandi et al,⁵⁰ 2018 (longitudinal neurodevelopmental assessment in infancy and toddlerhood), Rappaport et al,⁵¹ 1998 (longitudinal neurodevelopmental assessment in infancy and toddlerhood), Verrall et al,⁶⁷ 2021 (sample included adolescents and adults), and Verrall et al,⁷² 2022 (sample included adolescents and adults).

CHD = congenital heart disease; DTI = diffusion tensor imaging; EEG = electroencephalography; fMRI = functional magnetic resonance imaging; NIRS = near-infrared spectroscopy; TGA = transposition of great arteries.

CENTRAL ILLUSTRATION. Neurodevelopmental Outcomes Associated With Neuroimaging Across the Lifespan in Congenital Heart Disease.

This figure illustrates the different neuroimaging techniques used in congenital heart disease (CHD) research with patients with transposition of the great arteries or single-ventricle CHD and associations with neurodevelopmental, neurocognitive, or mental health outcomes reported in the literature. Mental health conditions, such as anxiety, depression, and post-traumatic stress, are notably missing, given the lack of examined or identified associations with neuroimaging. Gray shading indicates no associations examined or reported in the published literature. ADHD = attention-deficit/hyperactivity disorder; MRI = magnetic resonance imaging.

INFANT AND TODDLER NEURODEVELOPMENT (AGE 0–36 MONTHS; 23 PAPERS).

A total of 23 papers reported on associations between neuroimaging and neurodevelopment in infancy or toddlerhood: 8 papers (7 unique samples) in infancy,^29–36 13 (9 unique samples) in toddlerhood,^37–49 and 2 papers^50,51 including patients during both infancy and toddlerhood (Table 2, Supplemental Table 3). Of those assessing outcomes in infancy, 4 papers included only infants with TGA^30,34,35,51 and 6 included a mixed sample (TGA and single-ventricle CHD)^{29,31–33,36,50}; most (7 of 10) measured outcomes at age 12 months. Five papers (4 samples) assessed neurodevelopmental outcomes using the Bayley Scales of Infant and Toddler Development Second Edition (BSID-II),^{34–36,50,51} 3 used the BSID Third Edition (BSID-III),^29,30,33 and 2 assessed neurological dysfunction using a standardized clinical examination.^31,32

TABLE 2.

Associations Between Neuroimaging and Neurodevelopmental Outcomes Among Infants and Toddlers With TGA or Single-Ventricle CHD

Neuroimaging Modality	Number of Papers	Timing of Neuroimaging	Key Neuroimaging Findings	Neurodevelopmental Correlates	CHD Type and Sample Size
Infant studies (n = 10)a
Structural MRI	2	Preoperative MRI at age 1–12 d Postoperative MRI at 6–30 d after surgery	Preoperative, but not incident postoperative, brain injury (WMI, infarction, hemorrhage	Poorer BSID-III language and motor scores, or poorer neuromotor scores on neurological examination	TGA (n = 20)30 TGA or SVCHD (n = 22)31
Structural MRI	1	Preoperative MRI immediately presurgery (age <30 d) Postoperative MRI at median 7 d after surgery	Postoperative, but not preoperative, brain injury	Poorer BSID-III language and cognitive scores	TGA or SVCHD (n = 59)29
Structural MRI	1	Preoperative MRI at mean age2.8 d	Preoperative brain injury score in moderate to severe range	No relationship to moderate or severe neurological deficits	TGA or SVCHD (n = 53)32
Structural MRI	1	Preoperative MRI at median age 7 d Postoperative MRI at median 15 d after surgery	Preoperative or postoperative strokeb	Higher BSID-II psychomotor scores	TGA or SVCHD (n = 104)50
Structural MRI	1	Preoperative MRI at approximately 41 wks GA Postoperative MRI at approximately 43 wks GA	Higher brain growth rate in left posterior perisylvian region	Higher BSID-III language scores	TGA or SVCHD (n = 44)33
Video EEG monitoring	2 (1 sample)	Mean age at surgery 10 d, EEG monitoring for first 48 h after surgery	Seizure activity detected via EEG up to 48 h postoperative	Poorer BSID-II psychomotor scores	TGA (n = 155)34,51
Transcranial Doppler	1	Median age at surgery 5 d, perioperative monitoring	Lower systolic cerebral blood flow velocity in middle cerebral artery at 18 h postoperative	Poorer BSID-II psychomotor scores	TGA (n = 100)35
Transcranial Doppler	1	Mean age at surgery 15.9 d, perioperative monitoring	Higher resistive index in the anterior and middle cerebral arteries immediately postoperative	Higher BSID-II cognitive, language, and motor scores	TGA or SVCHD (n = 16)36
Toddler studies (n = 15)a
Structural MRI	1	Mean age at MRI 38.8 ± 17.4 mo	Larger frontal lobe volumes	Higher BSID-II mental and psychomotor scores	TGA or SVCHD (n = 33)37
			Larger whole and temporal lobe volumes	Higher BSID-II psychomotor scores
Structural MRI	3 (1 sample)	Mean age at MRI 25.9 ± 3.4 mo	Any abnormality on MRI	Lower BSID-III language and motor scores	47 with SVCHD41
			Subarachnoid space enlargement WMI or stroke	Lower BSID-III language, motor, cognitive scores No relationship to outcome
		Mean age at MRI 26.7 ± 3.9 mo	Larger cerebrospinal fluid volumes Intracranial, GM, and WM volumes	Lower BSID-III language, motor, and cognitive scores and higher likelihood of abnormal neurological examination No relationship to BSID-III scores	44 with SVCHD39,40
Structural MRI and DTI	1	Preoperative MRI at median age 7 d Postoperative MRI at median 15 d postsurgery	Preoperative or postoperative moderate to severe white matter injury	Lower BSID-II psychomotor scores	TGA or SVCHD (n = 70)50
			Larger WMI volumes	Lower BSID-II psychomotor scores
			FA scores	No relationship to BSID-II outcome
Structural MRI	1	Preoperative MRI at median 5 d (range 1–26 d) Postoperative MRI at median 21 d (range 4–70 d)	Perioperative brain weight z-scores	No relationship to BSID-III outcome	TGA (n = 24)42
Fetal MCA ultrasound	2 (1 sample)	Mean 26.6 ± 5.1 wks GA at initial ultrasound	Higher MCA pulsatility index z-score at initial ultrasound	Lower BSID-II psychomotor scores	SVCHD (n = 82)43
			MCA pulsatility index z-score <−2 at any point	Higher BSID-II psychomotor scores
		Mean 27.2 ± 5.3 wks GA at any ultrasound	Change in MCA pulsatility index across GA periods	No relationship to BSID-II outcome	SVCHD (n = 82)44
Fetal MCA ultrasound	1	Mean 33.8 ± 3.5 GA wks	MCA pulsatility index	No relationship to BSID-III outcome	TGA or SVCHD (n = 60)45
Amplitude integrated EEG	1	Mean age at surgery 3 d (Q1, Q3: 2, 5 d)	Seizure activity detected via EEG for up to 48 h postoperative	No relationship to BSID-III outcome	SVCHD (n = 25)46
			Recovery to continuous background pattern by 48 h postoperative	14-point increase in BSID-III motor score
Video EEG monitoring	1	Mean age at surgery 10 d	Seizure activity detected via EEG up to 48 h postoperative	Worse performance on the 3/Bc scales of Minnesota Child Development Inventory and 5/ld scales of MacArthur Communicative Development Inventory	TGA (n = 90)51
EEG	1	Median 25.3 mo (range 19.2–34.4 mo)	Abnormal EEG findings (epileptiform activity, diffuse background abnormality, asymmetry)	Lower BSID-II mental scores and more frequent neurological impairment	SVCHD (n = 36)47
NIRS	1	Mean 8 ± 8.5 d at surgery	Lower cerebral tissue oxygenation variability postoperative	Lower BSID-II mental and psychomotor scores	TGA or SVCHD (n = 44)48
NIRS	1	Median 4 mo (Q1, Q3: 3, 4 mo)	Perioperative cerebral NIRS parameters (nadir saturation and AUC)	No relationship to BSID-III scores	SVCHD (n = 19)38
NIRS	1	Median 8 d (range 6–125 d)	Preoperative cerebral NIRS scores below 35%	No relationship to BSID-II scores	TGA (n = 20)49

^a2 papers (Peyvandi et al,⁵⁰ 2018, and Rappaport et al,⁵¹ 1998) performed longitudinal neurodevelopmental assessment in infancy and toddlerhood.

^bThese infants were also more likely to have had a balloon atrial septostomy.

^cEEG seizures were associated with lower scores on the general development, expressive language, and personal-social scales of the Minnesota Child Development Inventory.

^dEEG seizures were associated with lower scores on the vocabulary production, irregular forms, over-regularization, sentence complexity, and the mean length of utterance in morphemes of the 3 longest sentences heard recently scales of the MacArthur Communicative Development Inventory.

AUC = area under the curve; BSID = Bayley Scales of Infant Development; FA = fractional anisotropy; GA = gestational age; GM = gray matter; MCA = middle cerebral artery; SV = single ventricle; SVCHD = single ventricle congenital heart disease; WM = white matter; WMI = white matter injury; other abbreviations as in Table 1.

Of those assessing outcomes in toddlerhood, 3 papers included only toddlers with TGA,^42,49,51 8 papers (4 samples) reported on outcomes in single-ventricle CHD,^{38–41,43,44,46,47} and 4 papers (4 samples) reported results from mixed samples. ^37,45,48,51 A total of 7 papers assessed outcomes using the BSID-II,^{37,43,44,47–50} 7 used the BSID-III,^{38–42,45,46} 1 assessed neurological dysfunction,⁴⁰ and 1 assessed neurodevelopment and language via parent-report questionnaire.⁵¹

BSID OUTCOMES (21 PAPERS).

Overall, we identified 21 papers examining associations between neuroimaging and BSID scores in infants (9 papers) or toddlers (13 papers), with 1 paper including both infants and toddlers. Four of these papers examined structural brain injury^29,30,41,50; 5 focused on brain volumes, weight, or growth^{33,37,39,40,42}; 4 used EEG^34,46,47,51; 3 examined fetal middle cerebral artery (MCA) readings^43–45; 3 investigated cerebral tissue oxygenation with NIRS^38,48,49; 2 investigated transcranial Doppler findings^35,36; and 1 examined DTI metrics.⁵⁰ Overall, structural brain injury tended to be associated with lower (poorer) BSID scores, but there was no clear pattern between type of injury and BSID domain. While total brain volume and weight were not associated with BSID outcomes in the included studies, larger volumes and greater growth in specific brain regions were associated with higher (better) BSID scores. Seizures or other abnormalities detected by EEG tended to be associated with lower BSID scores, whereas transcranial Doppler, fetal MCA, and NIRS findings were mixed.

Structural MRI (9 papers) and DTI (1 paper).

In total, 9 papers used MRI: one used a 3-T scanner and 8 used a 1.5-T scanner. Of the 4 papers examining associations between structural brain injury on MRI and BSID scores in infants and toddlers, 3 papers^29,30,41 found structural brain injury (any form) was associated with lower BSID-III motor, language, and cognitive scores. The remaining paper found moderate to severe WMI and larger WMI volumes were associated with lower BSID-II psychomotor scores in toddlers.⁵⁰ In contrast, Knirsch et al⁴¹ found no association between WMI and BSID-III motor, language, or cognitive scores; however, subarachnoid space enlargement was associated with lower BSID-III scores on all 3 domains.

Of the 2 studies examining associations between cerebral volumes and BSID scores, Reich et al³⁹ and Heye et al⁴⁰ (2 papers, 1 sample) found no association between total, gray, and white matter volumes and BSID scores, but did find larger cerebrospinal fluid volumes were associated with poorer BSID-III motor, language, and cognitive outcomes. In contrast, Ibuki et al³⁷ found that larger frontal lobe volumes were associated with higher BSID-II mental and psychomotor scores, and larger whole and temporal lobe volumes were associated with higher psychomotor scores. Brain weight z-score was not associated with BSID scores.⁴² On fetal MRI, higher brain growth rate in the left posterior perisylvian region was associated with higher BSID-III language scores.³³ In the one DTI study, there was association between fractional anisotropy (FA) values and BSID scores.⁵⁰

Electroencephalography (4 papers).

Of the 4 papers investigating EEG and BSID scores, 3 found that seizure activity^34,51 and other abnormal findings, such as asymmetry,⁴⁷ were associated with lower BSID-II psychomotor and mental scores.^34,47,51 The fourth paper found no link between seizure activity and BSID outcomes, but did find recovery of the EEG to a continuous background (normal) pattern by 48 hours postoperative was associated with higher BSID-III motor scores.⁴⁶

Transcranial Doppler (2 papers).

In the 2 transcranial Doppler studies, higher postoperative systolic cerebral blood flow velocity in the middle cerebral artery³⁵ and higher resistive index in the anterior and middle cerebral arteries³⁶ were associated with higher BSID-II psychomotor scores³⁵ and higher BSID-III scores (all domains).³⁶

Near-infrared spectroscopy (3 papers).

Only 1 of 3 papers⁴⁸ found a link between cerebral tissue oxygenation measured by NIRS and BSID outcomes; lower post-operative cerebral tissue oxygenation index variability was associated with lower BSID-II mental and psychomotor scores.

Fetal ultrasound of cranial arteries (3 papers).

The fetal MCA pulsatility index z-score is a key parameter used in Doppler assessments, with lower MCA pulsatility index z-scores indicative of redistribution of cardiac output to the brain. In the 3 papers (2 samples) examining fetal MCA readings,^43–45 1 found that lower MCA pulsatility index z-scores at initial fetal head ultrasound (adjusting for gestational age) were associated with higher BSID-II psychomotor scores.^43,44

OTHER INFANT AND TODDLER NEURODEVELOPMENTAL OUTCOMES. Neurological dysfunction (4 papers).

Four papers investigated neurological dysfunction in infants or toddlers; 3 used MRI and 1 used EEG. Of these, 2 papers (mixed samples) found no evidence of association between preoperative or postoperative brain lesions and functional deficits according to postoperative neurological examination.^31,32 Infants with preoperative cerebral lesions (WMI, infarction, hemorrhage, or stroke) demonstrated poorer preoperative neuromotor scores (especially poorer posture and vision or hearing on neurological examination) at age 7-days (mixed sample).³¹ Cerebrospinal fluid volumes were higher in patients with an abnormal neurological examination compared with patients with a normal examination.³⁹ In another study (single-ventricle CHD), abnormal EEG findings, including seizure activity at age 24 months, were associated with lower BSID-II mental but not psychomotor scores and more frequent diagnosis of neurological impairment.⁴⁷

Parent-report questionnaire (1 paper).

The one study⁵¹ using parent-report questionnaire (Minnesota Child Development Inventory and MacArthur Communicative Development Inventory) found that presence of perioperative seizure activity detected via EEG was associated with lower parent-reported expressive language and communication scores for toddlers with TGA.

CHILD, ADOLESCENT, AND ADULT NEUROCOGNITION (AGE ≥4 YEARS; 22 PAPERS).

A total of 8 papers investigated associations between neuroimaging and neurodevelopment in children (Table 3, Supplemental Table 4): 3 examined findings in children aged 4 to 5 years^52–54 and 5 included school-aged children (mean age 6 to 11 years).^55–59 Of these, 2 papers assessed neurodevelopment at age 8 years and then performed neuroimaging during adolescence.^55,56 Four papers (2 samples) included children with TGA,^52,55–57 2 included children with single-ventricle CHD,^53,54 and 2 studied a mixed sample.^58,59 A total of 16 papers, reporting on data from 6 unique samples, examined associations between neuroimaging and neurocognitive, behavioral, or mental health outcomes in adolescents or adults (Table 4, Supplemental Table 5).^{8,55,56,60–72} Only 2 papers (using the same sample) had a mean sample age above 18 years and included adults aged ≥21 years.^67,72 Six papers (2 samples) included individuals with TGA,^{8,55,56,60,64,65} 8 papers (3 samples) included individuals with single-ventricle CHD,^{61–63,66,67,70–72} and 2 papers (1 sample) described findings from a mixed sample.^68,69 Two of these papers spanned childhood and adolescence, reporting longitudinal assessments from 1 sample.^55,56 Of the papers reporting results from MRI, 7 (3 samples) used a 1.5-T scanner,^{8,55–58,64,65} 6 used a 3-T scanner,^{59,60,67–69,71} and 5 (1 sample)61-63,66,70 used a 3-T scanner where possible or 1.5-T scanner for participants with an implanted cardiovascular device or coils.

TABLE 3.

Associations Between Neuroimaging and Neurodevelopmental or Mental Health Outcomes n Children (Ages 3–12 Years)

Neuroimaging Modality	Number of Papers	Timing of Neuroimaging	Key Neuroimaging Findings	Neurodevelopmental Correlates	CHD Type and Sample Size
Preschool child studies (n = 3)
Video EEG monitoring	1	Perioperative EEG at mean age 9.8 d	Seizure activity detected via EEG up to 48 h postoperative	Lower FSIQ, verbal IQ, and performance IQ	TGA (n = 158)52
				Definite or probable neurological abnormalities
				No relationship to apraxia of speech or motor outcomes
NIRS	1	Perioperative NIRS at median 5 d (range 2–30 d)	Higher cerebral oxygen at baseline (preoperative) but not postoperative	Higher FSIQ, verbal IQ and nonverbal IQ Higher global and verbal cognitive abilities	SVCHD (n = 43)53
NIRS	1	Perioperative NIRS at mean 6.5 ± 2.2 d	Regional oxygen saturations below a breakpoint of 49.2% perioperative	Below breakpoint, positive linear relationship between oxygen saturations and matrix reasoning	SVCHD (n = 21)54
			Lower average cerebral oxygenation in the first 48 h postoperative	Low or abnormal visual-motor integration scores
			More time at thresholds <45% and <55%	Low or abnormal visual-motor integration scores
			Any hourly regional oxygenation <45% in patients without stroke	Lower visual-motor integration scores and lower composite neurodevelopmental scores
			Higher hourly regional cerebral oxygenation saturation	Higher visual-motor integration scores
School-age child studies (n = 5)
Structural MRI	1	Mean age TGA = 113.4 ± 21.5 mo Mean age SVCHD = 113.0 ± 19.3mo	Larger total brain volumes and larger cerebellum and cerebral cortex volumes	Higher FSIQ scores	TGA or SVCHD (n = 2l)58
			Larger temporal cortex volumes	Higher FSIQ, processing speed, working memory scores
			Larger posterior WM volumes	Higher perceptual reasoning scores
Structural MRI	1	11.4 ± 2.7 y	Larger hippocampal volumes	Higher memory scores (CMS) No relationship to verbal fluency, academic outcomes, or FSIQ	TGA (n = 40)5l
Structural MRI	1	Preoperative MRI at median 7 d (Q1, Q3: 5.8, 9.0 d) Postoperative MRI at median 25.0 d (Q1, Q3: 20.0–32.0 d)	Preoperative or postoperative brain injury (stroke, WMI, hemorrhage) or total brain volume	No relationship to WPPSI-III scores or motor function	TGA or SVCHD (n = 4l)59
DTI	2	16.1 ± 0.5 y	Higher FA in the right PLIC Decreased global efficiency and increased small-worldness	Higher verbal IQ scores at 8 y of age Mediated relationship between TGA status and FSIQ at 8 y of age	TGA (n = 49)55,56

CMS = Children’s Memory Scale; FSIQ = full scale IQ; PLIC = posterior limb of the internal capsule; SV = single-ventricle; WPPSI-III = Weschler Preschool and Primary Scale of Intelligence–Third Edition; other abbreviations as in Tables 1 and 2.

TABLE 4.

Associations Between Neuroimaging and Neurocognitive or Mental Health Outcomes in Adolescents and Adults

Neuroimaging Modality	Number of Papers	Timing of Neuroimaging	Key Neuroimaging Findings	Neurodevelopmental Correlates	CHD Type and Sample Size
Adolescent and adult studies (n = 16)
Intelligence or general cognitive functioning
MRI	5 (4samples)	14.5 ± 2.9 y	Structural abnormalities (diffuse abnormality, brain mineralization, focal infarction or atrophy)	No relationship to WISC or WAIS scores	SVCHD (n = 144)70
		14.7 ± 2.9 y	Lower left and right frontal sulcal pattern similarity	Lower processing speed index	SVCHD (n = 115)61
			Lower left-right parietal sulcal pattern symmetry	Lower perceptual reasoning index Scores
		16.9 ± 1.7 y	Reduced brain volume	Lower verbal IQ, performance IQ, FSIQ	TGA (n = 60)60
			Higher total MRI score (degree of abnormality)	Lower FSIQ Lower analytical thinking score
		Median 16 y (range 15–11 y)	Higher left and right hippocampal volume	Higher total MoCA scores	SVCHD (n = 25)71
		22.3 ± 8 y	Structural abnormalities (infarct, WMI, microhemorrhage)	No association with processing speed or psychomotor function (Cogstate)	SVCHD (n = l9)67
			Larger cerebellar GM volume	Higher psychomotor functioning scores (Cogstate)
DTI	2	23.1 ± 7.8 y	Higher mean FA in the right AF, SLF-II and SLF-III	Slower processing speed	SVCHD (n = 92)72
		14.7 ± 3.0 y	Higher FA in the bilateral forceps minor, LH+RH cerebral peduncle, LH+RH SCP, LH corticospinal tract, LH+RH external capsule, LH+RH IFOF, LH+RH ILF, LH SLF, RH ATR	Higher FSIQ scores	SVCHD (n = 102)62
			Higher FA in the bilateral forceps minor and forceps major, LH+RH cerebral peduncle, LH corticospinal tract, LH+RH IFOF, LF+RH ILF, LH+RH SCP, RH external capsule	Higher processing speed index scores
Academic abilities
MRI	3	16.1 ± 0.5 y	Imaging abnormality (infarction, mineralization, iron deposition, myelination delay, ventriculomegaly, abnormal T2 hyperintensity)	No association with academic outcome	TGA (n = 111)8
		14.5 ± 2.9 y	Structural abnormalities. (diffuse abnormality, brain mineralization, focal infarction or atrophy)	No association with academic outcome	SVCHD (n = 144)70
		16.9 ± 1.7 y	Higher total MRI score (degree of abnormality)	Lower spelling scores	TGA (n = 60)60
DTI   1.5T scanner	2 (1 sample)	16.1 ± 0.5 y	Increased small-worldness and increased modularity in WM topology	Mediated the relationship between TGA status and poorer academic outcome	TGA (n = 49)55,56
			Higher FA in the left parietal lobe and left temporal isthmus	Higher mathematics composite (WIAT)
			No significant correlations with FA	Reading composite (WIAT)
Executive functioning
MRI	5 (3 samples)	16.1 ± 0.5 y	Imaging abnormality (infarction, mineralization, iron deposition, myelination delay, ventriculomegaly, abnormal T2 hyperintensity)	No association with executive function (BRIEF, D-KEFS)	TGA (n = 111)8
		14.5 ± 2.9 y	Focal abnormalities (infarction or atrophy)	Higher (worse) parent BRIEF scores	SVCHD (n = 144)63,70
		14.7 ± 2.9 y	Higher left temporal sulcal pattern similarity	Higher executive functioning (D-KEFS)	SVCHD (n = 115)61
		22.3 ± 8 y	Larger cerebellar GM volume	Higher executive functioning scores (Cogstate)	SVCHD (n = 79)67
DTI	2 (1 sample)	16.1 ± 0.5 y	Increased small-worldness and increased modularity in WM topology	Mediated the relationship between TGA status and lower executive functioning (D-KEFS)	TGA (n = 49)55,56
			Higher FA in the right precentral gyrus	Higher executive functioning (BRIEF)
			Higher FA in the left temporal isthmus	Higher executive functioning (D-KEFS)
fMRI	1	17.76 ± 1.72 y	BOLD signal change in the right precentral gyrus	No relationship to executive functioning (BRIEF)	TGA or SVCHD (n = 17)68
Attention or ADHD symptoms or diagnosis
MRI	7 (4 samples)	16.1 ± 0.5 y	Imaging abnormality (infarction, mineralization, iron deposition, myelination delay, ventriculomegaly, abnormal T2 hyperintensity)	No relationship to ADHD symptoms, ADHD diagnosis	TGA (n = 111)8,65
		14.5 ± 2.9 y	Focal abnormalities	Higher parent-reported ADHD symptoms	SVCHD (n = 144)63,66,70
			Structural abnormalities (diffuse abnormality, brain mineralization, focal infarction or atrophy)	No relationship to ADHD diagnosis
		Median 16 y (range 15–17 y)	Higher left hippocampal volume	Higher MoCA attention scores	SVCHD (n = 25)71
		22.3 ± 8 y	Larger cerebellar GM volume	Higher attention scores (Cogstate)	SVCHD (n = 19)67
DTI	4(3 samples)	16.1 ± 0.5 y	Lower FA in right precentral gyrus, left parietal lobe	Higher parent-reported ADHD symptoms (CADS)	TGA (n = 49)56,64
			Higher FA in anterior cingulate cortex	Higher parent-reported ADHD symptoms (CADS)
			Increased cost in WM topology	Mediated the relationship between TGA status and higher CADS scores (SR, PR, and TR)
			Increased transitivity and increased global efficiency in WM topology	Mediated the relationship between TGA status and higher CADS scores (PR and TR)
		23.1 ± 7.8 y	Tract-specific DTI metrics (FA, MD, AD, RD)	No significant associations with attention scores	SVCHD (n = 61)72
		17.82 ± 1.65 y	Higher FA along the middle cerebellar peduncle	Higher attention span but not once statistically corrected	TGA or SVCHD (n = 22)69
Learning and memory
MRI	5 (4 samples)	16.1 ± 0.5 y	Imaging abnormality (infarction, mineralization, iron deposition, myelination delay, ventriculomegaly, abnormal T2 hyperintensity)	No association with memory (CMS)	TGA (n = 111)8
		14.5 ± 2.9 y	Structural abnormalities (diffuse abnormality, brain mineralization, focal infarction or atrophy)	No association with memory (CMS or WMS)	SVCHD (n = 144)70
		14.7 ± 2.9 y	Lower left frontal sulcal pattern similarity	Lower working memory	SVCHD (n = 115)61
		22.3 ± 8.0 y	Presence but not severity of WMI	Lower paired associate learning z-scores	SVCHD (n = 92)67
			Higher total brain, intracranial, total GM, cortex GM, subcortical GM, cerebral WM, cerebellar GM volumes	Higher visual learning scores
			Total brain, intracranial, GM, cortex GM, subcortical GM, cerebellar GM, cerebellar WM volumes	Higher paired associate learning scores
			Cerebellar WM	Higher working memory scores
		Median 16 y (range 15–17 y)	Larger left and right hippocampal volume	Higher WRAML-2 Global Memory Index scores, higher MoCA delayed memory recall	SVCHD (n = 25)71
DTI	3 (2 samples)	16.1 ± 0.5 y	Higher FA in the right PLIC	Higher general memory index (CMS and WMS)	TGA (n = 49)55,56
			Decreased global efficiency and increased modularity of WM topology	Mediated the relationship between TGA status and lower immediate and delayed visual memory (CMS)
		17.82 ± 1.65 y (range 16–21 y)	Higher FA along the uncinate fasciculus	Higher verbal memory performance (uncorrected) but not once corrected	TGA or CHD (n = 22)59
fMRI	1	17.76 ± 1.72 y	BOLD signal change in the right precentral gyrus	Lower working memory scores	TGA or SVCHD (n = 11)68
Visual-spatial perception
MRI	2 (2 samples)	16.1 ± 0.5 y	Imaging abnormality (infarction, mineralization, iron deposition, myelination delay, ventriculomegaly, abnormal T2 hyperintensity)	No association with visual-perception outcome (TVPS, Sense of Direction test, Rey-Osterrieth Complex Figure)	TGA (n = 111)8
		14.5 ± 2.9 y	Structural abnormalities (diffuse abnormality, brain mineralization, focal infarction or atrophy)	No association with visual-perception outcome (TVPS, Sense of Direction test, Rey-Osterrieth Complex Figure)	SVCHD (n = 144)70
DTI	2 (1 sample)	16.1 ± 0.5 y	Higher FA in the right frontal region	Higher visual closure subscale score	TGA (n = 49)55,56
			Decreased global efficiency of WM topology	Mediated association between TGA status and lower TVPS visual-spatial relationships score
			Increased small-worldness of WM topology	Mediated association between TGA status and lower Rey-Osterrieth Complex Figure copy structural element score
Social cognition and emotion recognition
MRI	3	16.1 ± 0.5 y	Imaging abnormality (infarction, mineralization, iron deposition, myelination delay, ventriculomegaly, abnormal T2 hyperintensity)	No association with social cognition	TGA (n = 111)8
		14.5 ± 2.9 y	Focal infarction or atrophy	Higher Autism Spectrum Quotient scores	SVCHD (n = 144)70
		22.3 ± 8 y	Cerebral volumes (including WM and GM) and structural abnormalities	No association with emotion recognition (Cogstate)	SVCHD (n = 19)67
DTI	1	16.1 ± 0.5 y	Lower FA in the right precentral white matter	Higher social cognition scores	TGA (n = 49)56
Psychological or psychiatric functioning
MRI	4 (3 samples)	16.1 ± 0.5 y	Imaging abnormality (infarction, mineralization, iron deposition, myelination delay, ventriculomegaly, abnormal T2 hyperintensity)	No association with global psychosocial functioning	TGA (n = 111)65
		14.5 ± 2.9 y	Structural abnormalities. (diffuse abnormality, brain mineralization, focal infarction or atrophy)	No association with scores on Brief Psychiatric Rating Scale, lifetime anxiety diagnosis, or global psychosocial functioning	SVCHD (n = 121)63,66
		Median 16 y (range 15–1 y7)	Left and right hippocampal volumes	No association with anxiety or depressive symptoms	SVCHD (n = 25)71

AD = axial diffusivity; ADHD = Attention deficit/hyperactivity disorder; AF = arcuate fasciculus; ATR = Anterior thalamic radiation; BOLD = blood oxygen level-dependent; BRIEF = Behavioral Rating Inventory of Executive Function; CADS = Conners ADHD/DSM-IV Scale; CMS = Children’s Memory Scale; D-KEFS = Delis-Kaplan Executive Function System; IFOF = Inferior fronto-occipital fasciculus; ILF = Inferior longitudinal fasciculus; LH = left hemisphere; MD = mean diffusivity; MoCA = Montreal Cognitive Assessment; PR = parent report; RD = radial diffusivity; RH = right hemisphere; SCP = superior cerebellar peduncle; SLF = superior longitudinal fasciculus; SLF-II or III = superior longitudinal fasciculus components II or III; SR = self-report; TR = teacher report; TVPS = Test of Visual Perceptual Skills; WAIS = Weschler Adult Intelligence Scale; WIAT = Weschler Individual Achievement Test; WISC = Weschler Intelligence Scale for Children; WMS = Weschler Memory Scale; WRAML-2 = Wide Range Assessment of Memory and Learning - 2^nd cedition; other abbreviations as in Tables 1,2, and3.

CHILD AND ADOLESCENT WESCHLER INTELLIGENCE SCALE SCORES (12 PAPERS).

We identified 12 papers examining associations between neuroimaging and Weschler Intelligence Scale for Children scores (full-scale IQ, verbal and nonverbal/performance IQ, processing speed, working memory, and matrix reasoning) in children (8 papers) and adolescents (4 papers), representing 8 different samples. Three of these papers examined structural brain injury,^59,60,70 4 focused on brain volumes,^57–60 3 on diffusion tensor imaging (DTI) metrics,^55,56,62 2 on NIRS parameters,^53,54 1 on sulcal pattern symmetry,⁶¹ and 1 used EEG.⁵² Overall, MRI findings were mixed but suggested that brain volumes may be a more robust correlate of Weschler scores than structural brain injury in children and adolescents. Using DTI, white matter microstructure and topology were associated with Weschler scores. Seizure activity detected on EEG and higher perioperative cerebral oxygen saturations were also correlated with Weschler scores.

Structural MRI (6 papers).

Of the 3 studies examining structural brain injury and Weschler scores,^59,60,70 only 1 found an association; greater abnormality detected on MRI was associated with lower full-scale IQ (FSIQ).⁶⁰ Of the 4 papers investigating brain volumes and Weschler scores, 2 found that larger total brain volumes were associated with higher FSIQ^58,60 and 1 found no association.⁵⁹ Investigating specific brain regions, hippocampal volumes were not associated with FSIQ⁵⁷; larger cerebellum and cerebral cortex volumes were linked with higher FSIQ; larger temporal cortex volumes were linked with higher FSIQ, processing speed, and working memory scores; and larger posterior white matter (volumes were associated with higher perceptual reasoning scores.⁵⁸ Finally, lower left-right sulcal pattern symmetry in the frontal and parietal regions was associated with lower processing speed and perceptual reasoning scores, respectively.⁶¹

DTI (3 papers).

Of the 3 papers investigating DTI metrics, 1 found that higher FA in the right posterior limb of the internal capsule was associated with higher verbal IQ⁵⁶ and higher FA in several white matter tracts, such as the forceps minor, external capsule, and anterior thalamic radiation (see Table 4 for full list), was associated with higher FSIQ and processing speed.⁶² When examining white matter topology, graph metrics (specifically, decreased global efficiency, increased modularity [segregation], and increased small-worldness [ratio between segregation and integration in the network]) mediated the relationship between TGA status (ie, whether individuals had TGA or not) and lower FSIQ scores.⁵⁵

EEG (1 paper) and NIRS (2 papers).

Only 1 study examined EEG and Weschler outcome; seizure activity detected up to 48 hours postoperatively was associated with lower FSIQ, verbal, and nonverbal IQ.⁵² In both NIRS studies, higher oxygen saturations both preoperative⁵³ and perioperative⁵⁴ were associated with higher FSIQ, verbal, and nonverbal IQ,⁵³ and matrix reasoning (below a breakpoint of 49.2%; above this breakpoint, there was no association).⁵⁴

OTHER CHILD NEURODEVELOPMENTAL OUTCOMES. Structural MRI (2 papers).

In 1 MRI study (TGA sample), larger hippocampal volume was not correlated with academic abilities or verbal fluency but was correlated with better memory assessed using the Children’s Memory Scale and Rivermead Behavioral Memory Test.⁵⁷ In a mixed sample, there were no associations found between motor function and either brain volumes or brain injury.⁵⁹

NIRS (2 papers).

The NIRS studies included single-ventricle CHD patients. Hansen et al⁵³ found that higher preoperative cerebral oxygenation was associated with better global and verbal cognitive abilities, including visual perception, auditory and visual memory, expressive and receptive communication skills, attention, and psychomotor performance. Hoffman et al⁵⁴ found that lower average cerebral oxygenation in the first 48 hours after Stage 1 palliation and greater time spent in a desaturated state were associated with low or “abnormal” visual-motor integration scores in childhood. Again, Hoffman et al⁵⁴ performed a breakpoint analysis (ie, segmented linear regression) between cerebral oxygen saturation and multiple outcomes and reported linear associations of breakpoints of 62.2%, 56.9%, and 55.8% language, visual-motor integration, and a neurodevelopmental composite score, respectively.

EEG (1 paper).

Among children with TGA, no differences in fine or gross motor scores were identified for those with or without seizures detected via EEG. There was also no association between EEG seizures and apraxia of speech; however, seizures were associated with a higher likelihood of definite or probable neurological abnormalities identified via clinical examination at age 4 years.⁵²

OTHER ADOLESCENT AND ADULT NEUROCOGNITIVE OUTCOMES. General cognitive functioning (3 papers).

Three papers (all with single-ventricle CHD patients) investigated general cognitive functioning: 1 used a computerized cognitive assessment⁶⁷ and 2 papers (1 sample) used a cognitive screening test (Montreal Cognitive Assessment [MoCA]).^71,72 Structural abnormalities detected on MRI were not associated with psychomotor function or processing speed.⁶⁷ In patients with single-ventricle CHD, larger left and right hippocampal volumes were correlated with higher total MoCA scores,⁷¹ and larger cerebellar gray matter was associated with greater psychomotor function.⁶⁷ In the 1 DTI study, higher FA in the right arcuate fasciculus and superior longitudinal fasciculus was associated with slower processing speed.⁷²

Academic abilities (5 papers).

Five papers (3 samples) examined links between neuroimaging and academic abilities. When classifying brain injury using structural MRI, higher grade of injury (scale of 0–2) was associated with lower spelling scores in 1 TGA sample,⁶⁰ but academic ability was not associated with injury in another TGA sample⁸ or in single-ventricle CHD patients.⁷⁰ For white matter structural network topology investigated using DTI, poorer scores on the Weschler Individual Achievement Test mathematics composite among TGA patients relative to healthy control subjects were mediated by increased small-worldness (ie, the balance of integration and segregation in white matter) and increased modularity, a measure of network segregation.⁵⁵ Finally, FA in the left parietal and left temporal isthmus was also positively associated with Weschler Individual Achievement Test mathematics composite scores in individuals with TGA.⁵⁶

Executive functioning (8 papers).

Eight papers (4 samples) examined associations between MRI findings of structure or white matter connectivity (DTI) and executive functioning; 3 papers reported results from 1 TGA sample,^8,55,56 4 reported on 2 single-ventricle CHD samples,^61,63,67,70 and 1 reported on findings from a mixed sample.⁶⁸ Among TGA patients, Bellinger et al⁸ found no association between structural abnormalities detected using MRI and executive functioning assessed via the Delis-Kaplan Executive Function Scale (D-KEFS) or Behavioral Rating of Executive Function scores. In single-ventricle CHD patients, Bellinger et al found focal infarction or atrophy predicted worse parent-reported Behavioral Rating of Executive Function scores but not D-KEFS scores,⁷⁰ a result also reported by Calderon et al,⁶³ and in the same sample, lower left temporal sulcal pattern symmetry predicted poorer executive functioning.⁶¹ In a study from the Australian and New Zealand Fontan Registry, larger cerebellar gray matter volume was associated with higher executive functioning scores.⁶⁷

Examining microstructural brain differences using DTI, Panigrahy et al⁵⁵ found that lower executive functioning (D-KEFS scores) in TGA patients relative to healthy referent controls was mediated by increased modularity and small-worldness; increased transitivity (a measure of network segregation at the nodal level) and increased small-worldness mediated the association between greater total length of hospital stay and poorer executive functioning.⁵⁵ This group also found poorer executive functioning was associated with higher FA in the left temporal isthmus and right precentral.⁵⁶

Attention and ADHD symptoms (11 papers).

Eleven papers (5 samples: 1 TGA,^8,56,64,65 3 single-ventricle CHD,^{63,66,67,70–72} 1 mixed)⁶⁹ investigated associations between structural MRI or white matter connectivity (DTI) data and attention. In the Boston Circulatory Arrest Study, structural brain abnormalities on MRI were not associated with self- or parent-reported attentional difficulties or ADHD diagnosis.^8,65 In a single-ventricle CHD sample, focal infarction or atrophy on structural MRI predicted greater parent-reported ADHD symptoms^63,70 but not ADHD diagnosis.^63,66 In another single-ventricle sample, larger left hippocampal volume was associated with higher MoCA attention scores,⁷¹ and larger cerebellar gray matter was associated with better attention scores in a third sample.⁶⁷ Parent-reported ADHD symptoms were associated with lower FA in the precentral and left parietal white matter and higher FA in the anterior cingulate cortex.⁵⁶ FA in the middle cerebellar peduncle (hypothesized to be associated with attention) was also positively associated with performance on an auditory attention span task, but did not meet a predefined false discovery rate threshold.⁶⁹ In a single-ventricle CHD sample, no associations between tract-specific DTI metrics and attention scores were found.⁷² Higher (worse) ADHD scores among individuals with TGA compared with control subjects were mediated by differences in global white matter structural topology.⁶⁴

Memory and learning (9 papers).

Nine papers (5 samples) investigated memory and structural MRI or white matter connectivity (DTI) results: 1 TGA sample (3 papers), 3 single-ventricle CHD samples (4 papers), and 1 mixed sample (2 papers). Among those with TGA, structural abnormalities on MRI (any, focal/multifocal, or mineralization) were not associated with memory functioning assessed using the general memory index of the Children’s Memory Scale.⁸ This was also the case in 2 single-ventricle CHD samples; one using the same scale⁷⁰ and the other using a paired associate learning test, which measures visual memory whereby one stimulus promotes recall of a second stimuli.⁶⁷ In the latter study, presence but not severity of WMI predicted lower paired associate learning scores and cerebellar white matter volume was positively correlated with working memory performance.⁶⁷ In addition, total brain volume, intracranial volume, gray matter volume (total, cortex, subcortical, and cerebellar), and cerebral white matter volume were positively associated with visual learning and memory.⁶⁷ In another single-ventricle CHD sample, higher global memory scores on the Wide Range Assessment of Memory and Learning, Second Edition, as well as other MoCA and Wide Range Assessment of Memory and Learning, Second Edition memory subtests, were positively correlated with left and right hippocampal volumes.⁷¹ Lower left frontal sulcal pattern similarity was associated with lower working memory scores.⁶¹

In terms of white matter connectivity differences assessed using DTI, higher FA in the right posterior limb of the internal capsule was correlated with better global memory performance in a TGA sample.⁵⁶ Increased modularity (segregation) of white matter networks mediated lower immediate and delayed visual memory scores among adolescents with TGA relative to healthy referent adolescents.⁵⁵ Working memory performance was positively correlated with blood oxygenation level dependent signal change in the right precentral area.⁶⁸

Visual-spatial perception (4 papers).

Four papers (1 TGA, 1 single-ventricle sample) reported on structural MRI or white matter connectivity (DTI) data and visual-perceptual outcomes.^8,55,56,70 Structural abnormalities on MRI were not associated with visual-perceptual scores among TGA⁸ or single-ventricle CHD patients.⁷⁰ In a TGA sample, higher FA in the right frontal region was associated with better visual-spatial function on the Test of Visual-Perceptual Skills.⁵⁶ Finally, some elements of network topology (eg, increased global efficiency for visual-spatial relationships) mediated the poorer visual-perceptual scores found in the TGA group compared with control subjects.⁵⁵

Social cognition (4 papers).

Four papers (1 TGA and 2 single-ventricle CHD samples) investigated associations between social cognition and structural MRI or white matter connectivity (DTI) findings. ^8,56,67,70 Structural abnormalities or brain volumes were not associated with social cognition among adolescents with TGA⁸ or single-ventricle CHD,⁶⁷ except for focal infarction or atrophy, which led to higher Autism Quotient scores.⁷⁰ Among adolescents with TGA, lower FA in the right precentral white matter was associated with better social cognition.⁵⁶

Mental health (4 papers).

Four papers (3 samples) examined psychiatric symptoms or diagnoses in adolescents or adults.^63,65,66,71 Among adolescents with TGA, no associations were found between structural brain abnormalities on MRI and clinician-rated psychosocial functioning.⁶⁵ For adolescents with single-ventricle CHD, there was also no association between structural abnormalities and psychiatric outcomes, such as anxiety, depression, or post-traumatic stress symptoms, according to parent or adolescent interview, or self- and parent-reported questionnaires.^63,66 Finally, among adolescents with single-ventricle CHD, left and right hippocampal volumes were not associated with self-reported anxiety or depressive symptoms.⁷¹ A summary of main findings from studies of adolescents and adults is shown in Table 4.

DISCUSSION

This review evaluated evidence of associations between structural and functional neuroimaging and neurodevelopmental, neurocognitive, behavioral, and mental health outcomes among individuals with TGA or single-ventricle CHD. We reviewed 45 papers from 25 unique samples and found numerous associations between neuroimaging and functional outcomes from infancy through adulthood. Most studies were performed in the United States at a single center, were cross-sectional, and included samples of <50 CHD patients. Most reported associations between structural MRI findings and neurodevelopmental outcomes. Only 6 of the 25 studies examined outcomes in adolescence and only 1 study included patients with a mean age >21 years,^67,72 limiting conclusions regarding adults. Methodologies varied widely across studies, contributing to the mixed findings reported. Very few studies investigated mental health outcomes (eg, anxiety, depression), or used microstructural or functional imaging modalities.

INTERPRETATION OF KEY FINDINGS.

In infants and toddlers aged 0 to 36 months, associations between brain injury found on structural MRI and neurodevelopment were mixed. Most studies found that either preoperative or postoperative cerebral lesions were associated with adverse neurodevelopmental outcomes, such as lower BSID scores^{29–31,41,50}; however, other studies found no association.³² A single study examined brain injury and functional outcomes in school-aged children, reporting no association between brain injury and neurodevelopment.⁵⁹ In adolescents and adults, structural abnormalities on MRI were not linked to academic outcomes, memory, social cognition, or visual-perception.^8,63,70 Among adolescents, only 1 of 4 studies reported associations between structural brain abnormalities and IQ scores⁶⁰ and focal infarction was predictive of lower attention and executive function in a single-ventricle CHD, but not TGA, sample.^8,70 These results suggest that during infancy, both preoperative and postoperative brain injury may be associated with adverse neurodevelopmental outcomes; however, in adolescence, these associations are less clear. Moreover, most studies investigated the relationship between any form of brain injury and neurodevelopmental outcomes, rather than examining how specific types of brain injury are related to outcomes, barring a few exceptions.^33,43,56 While some studies not included in this review have found associations of specific brain injury with outcomes, we are not aware of studies that have assessed all types of brain injury and examined their relative contributions to neurodevelopment. Thus, we were unable to tease out the percent variance in outcome contributed by various types of brain injury. For example, in a sample of 221 very preterm neonates without CHD, cerebellar hemorrhage in the deeper parts of the posterior lobe were more likely to be associated with motor and visuomotor dysfunction, but not cognitive function, suggesting a location-dependent relationship between cerebellar hemorrhage and specific developmental outcomes.⁷³

Associations between brain volumes and neurodevelopment were also found. Among children aged 4 to 12 years, there was some evidence that various brain volumes detected on structural MRI were positively associated with neurodevelopment; eg, larger hippocampal volumes were associated with better memory.⁵⁷ Other structural changes, such as greater sulcal pattern symmetry and cerebral volumes, were also associated with better neurocognitive outcomes in adolescents, including IQ,⁶⁰ executive functioning, and attention^61,67,71; however, findings were rarely replicated across studies due at least in part to heterogeneous methodologies.

Microstructural brain changes are also associated with neurodevelopmental outcomes in CHD. Although we did not find any functional MRI studies and only 1 DTI study with infants or toddlers, increased white matter microstructure (FA) and more efficient white matter topology identified using DTI with adolescents were linked retrospectively to better cognitive outcomes in childhood, such as verbal and full-scale IQ.^55,56 Studies from 2 Boston-based cohorts found changes in white matter microstructure, such as disrupted white matter network topology and reduced FA, were associated with functional outcomes, such as poorer attention, memory, and executive functioning.^{55,56,62,64,70} This is consistent with studies in other CHD samples showing widespread bilateral reduction in FA in adolescents with CHD compared with control subjects; lower FA in the frontal lobes is associated with working memory deficits,⁷⁴ and lower FA in the left superior corona radiata and the corticospinal tract is associated with executive function impairments.⁷⁵ Similarly, a recent study including toddlers with CHD and hypoxic-ischemic encephalopathy found lower global efficiency of white matter networks was associated with lower BSID-III motor outcomes.⁷⁶ In contrast, one paper in an older cohort⁷² did not find associations between microstructure and attention but did report paradoxical associations between higher white matter microstructure and slower processing speed. In another study,⁵⁶ higher FA in the right precentral white matter was associated with poorer social cognition.

Previous non-CHD research has shown that processing speed at age 25 years predicts social functioning at age 43 years in individuals with mood disorders,⁷⁷ with processing speed hypothesized to be related to social cognition because of the speed of social interactions.⁷⁸ Interpretation of microstructural associations with neurocognitive outcomes may not be uniform across CHD populations and compensatory, but potentially maladaptive white matter remodeling may be present. Brain regions associated with processing speed or social cognition, which may be closely related, might be especially susceptible to compensatory white matter remodeling. Regardless, in this review, we found microstructural changes were more consistently related to functional outcomes compared with structural brain abnormalities.^{55,56,62,64,70} This is important, as previous research has demonstrated that reductions in typical white matter density in adolescents with TGA are not related to structural changes found on MRI and are related to potentially modifiable medical factors, such as duration of hypothermic arrest.⁷⁹

Overall, results across structural, functional, and microstructural MRI studies suggest that although a substantial subset of individuals with TGA or single-ventricle CHD are found to have brain lesions on neuroimaging, these injuries are not always associated with subsequent adverse neurocognitive outcomes. Instead, volumetric, microstructural, and functional changes may more likely underpin adverse functional outcomes, such as working memory, executive function, and attention difficulties, with early alterations in structural and functional brain development potentially leading to downstream changes in brain microstructure and functional organization. Longitudinal studies investigating trajectories of brain development and functional outcomes over time are much-needed.

Findings regarding associations between cerebral oxygenation, seizure activity, and neurodevelopmental outcomes are mixed. Fetal transcranial Doppler studies provide evidence that lower MCA pulsatility index is linked to better BSID scores in infancy and toddlerhood, suggesting that redistribution of cardiac output to the brain may have a potentially “brain-sparing” effect.⁴³ Turning to NIRS, one study found that higher preoperative cerebral oxygen levels were associated with better post-operative IQ, memory, and attention,⁵³ and another study reported a positive correlation between post-operative cerebral tissue oxygenation and BSID-II scores.⁴⁸ Finally, one study found a linear relationship between oxygenation and neurocognitive outcomes only when cerebral oxygenation fell below specific thresholds,⁵⁴ whereas other studies found no relationship.^38,49 The discrepant findings may be attributed to variable NIRS methodologies, differences in patient age at follow-up, and heterogeneity in CHD type. The mixed findings we report are consistent with research demonstrating that suboptimal perioperative hemodynamics are associated with adverse neurodevelopmental outcomes in some⁸⁰ but not all⁸¹ studies. Indeed, studies have found NIRS parameters have no relationship with subsequent cerebral injury,⁸² brain volumes,⁸¹ or neurodevelopment, whereas EEG seizure activity is correlated with new brain injury and subsequent neurodevelopmental and neurological outcomes.^{34,47,51,52,82} Overall, the review findings lend support to the role of cerebral oxygenation and seizure activity during fetal and perioperative periods, as measured by cerebral Doppler and EEG; however, future research should clarify critical oxygenation thresholds and the best neuromonitoring methods for perioperative identification of patients at risk of subsequent adverse outcomes. Longitudinal studies should also clarify whether suboptimal perioperative hemodynamics lead to disrupted white matter microstructure, because studies have so far focused on the relationship between hemodynamics and brain injury.

LIMITATIONS OF THE LITERATURE AND AREAS FOR FUTURE RESEARCH.

There is a paucity of longitudinal neuroimaging studies examining trajectories of brain development, neurodevelopment, and mental health across the life course in CHD. Although some studies have linked neurodevelopmental outcomes over time with neuroimaging at a single timepoint,^8,70 we identified only 1 study that examined brain changes longitudinally (during infancy only) and neurodevelopment,²⁹ and only 1 study used fetal brain MRI³³ as a predictor of later neurodevelopmental outcomes. Demonstrating the importance of longitudinal research, a recent study (not eligible for inclusion in this review because of the heterogenous CHD sample), found fetal brain volume independently predicted 10% to 21% of the variance in child neurodevelopment (ie, BSID-III composite scores, adaptive skills) at age 2 years,⁸³ after adjusting for important socioeconomic and medical factors. This study suggests small fetal brain volumes may be an important biomarker of later neurodevelopmental risk, above and beyond typically-examined factors, such as cardiac anatomy or length of hospital stay. Although the mechanism(s) underlying associations between impaired fetal brain growth and adverse neurodevelopmental outcomes remain unclear, this finding highlights the importance of the in utero environment and factors such as impaired oxygen delivery,^22,84 placental dysfunction,⁸⁵ genetics,⁸⁶ maternal psychological stress,^87,88 and socioeconomic factors.^83,89 Future research must continue to examine trajectories of brain development and neurodevelopment over time. Prematurity research, eg, shows that although at age 8 years, children born preterm exhibit similar functional connectivity on resting-state functional MRI to children born at term, significant differences in connectivity are detected between groups by age 16 years, accounting for poorer full-scale IQ scores.⁹⁰ Further highlighting the importance of longitudinal studies, Young et al⁹¹ conducted a series of diffusion MRI scans with very preterm children (gestational age <32 weeks) and demonstrated slower rates of change in white matter microstructure in the internal and external capsules from birth to age 4 years were associated with lower IQ and language scores.

Much of the existing research was designed to examine associations between cerebral abnormalities found on structural MRI and neurodevelopment; however, the results presented in this review indicate that microstructural or functional changes may be more closely related to outcomes, although these studies predominantly included adolescents, limiting generalizability to other developmental periods. Studies also tended to be exploratory, testing multiple, unplanned associations without statistical correction for multiple comparisons,⁶⁹ increasing the risk of type I error. Hypothesis-driven studies are much-needed to generate explanatory models for observed changes and associations with specific functional outcomes.²⁶ Of the 28 papers utilizing MRI, 15 used a 1.5-T scanner, limiting resolution of deeper brain structures and potentially biasing results. Moreover, samples tended to be small, generally <50 participants. In underpowered studies, population sampling variability and inflated effect sizes likely yield irreproducible brain-behavior associations.⁹² Reproducible associations may require samples numbering in the thousands,⁹² highlighting a clear need for multicenter, collaborative, and registry-based studies.

Finally, very few of the identified studies examined associations between neuroimaging and mental health outcomes in this population. There is a growing body of evidence demonstrating associations between neuroimaging markers and depression and anxiety in non-CHD populations. In a 2020 meta-analysis (92 studies) examining data from 2,928 patients with major depressive disorder, structural and functional abnormalities in regions including the hippocampus, amygdala, and putamen were found in patients with major depressive disorder compared with control subjects.⁹³ A 2016 multicenter mega-analysis of 1,546 adults across 19 centers found alterations to the organization of cortical networks involved in sensory processing were a robust biomarker for major depressive disorder.⁹⁴ Similarly, for anxiety, a 2019 meta-analysis found evidence of altered functional connectivity in brain networks associated with emotion processing (affective network) and cognitive control (executive control and default mode networks) in children, adolescents, and adults with anxiety.⁹⁵ Individuals with complex CHD experience a range of early adversities, such as exposure to the highly stressful intensive care unit environment, multiple invasive procedures, and separation from caregivers.⁹⁶ Alterations to brain architecture are a critically important factor linking early adverse experiences to later mental health diagnoses.⁹⁷ Previously-institutionalized children who experienced maternal deprivation, eg, exhibit atypical (accelerated) amygdala to medial prefrontal cortex connectivity, conferring an initially lower risk of anxiety, but with the potential for downstream consequences for brain development and emotion regulation later in life.⁹⁷ Alterations in white matter integrity have also been found, including in regions hypothesized to be involved in emotion processing (eg, limbic circuitry).⁹⁸ Importantly, emerging evidence suggests nurturing caregiving may be a potential avenue for intervention, including remediation of differences in brain development.⁹⁸ It is not clear whether similar changes to emotion processing brain networks are present in individuals with CHD, and studies are needed to address this knowledge gap.

STRENGTHS AND LIMITATIONS OF THE REVIEW.

Key strengths of this review are the synthesis of data from individuals across the lifespan and inclusion of studies using a variety of neuroimaging techniques and assessing a range of functional outcomes. We included only studies of individuals with TGA or single-ventricle CHD to reduce potential variability in outcomes based on CHD type and rigorously assessed study risk of bias. This review is not, however, without limitations. Across the included studies, heterogeneity in neuroimaging methods and outcomes measurement precluded meta-analysis. The results of regression analyses in case series were almost certainly affected by the covariates included, particularly because many variables in cardiac neurodevelopment are highly intercorrelated. Results may have also been affected by unmeasured patient management factors which can vary among institutions. For studies with longitudinal follow-up, the effect size of an early brain MRI finding on neurodevelopment may change over time and be affected both by mitigating factors, such as high socioeconomic status or conversely, by new neurologic insult, especially in the single-ventricle population. As harmonization techniques evolve, it will become increasingly possible to combine neuroimaging data across sites and samples, facilitating new analyses. Where appropriate, studies should focus on harmonizing with past work to enable future meta-analyses, particularly considering that studies in CHD populations tend to be restricted in terms of sample size. As a priority, next steps should include hypothesis-driven structural, microstructural, and functional MRI studies. Optimizing reciprocity between neuroimaging studies conducted for research and in clinical practice will further inform neuroprotective and therapeutic approaches in CHD.⁹⁹ Also, although it was beyond the scope of this review to examine genetic factors, future studies could study gene associations and patient populations with chromosomal abnormalities associated with risk of adverse neurodevelopmental outcomes.^18,100

CONCLUSIONS

Advances in neuroimaging have deepened our understanding of brain health and development among individuals with TGA or single-ventricle CHD. In this systematic review, we identified 45 papers from 25 unique studies; most assessed outcomes among infants and children (73%), were cross-sectional (72%), and were single-center (76%), and two-thirds had a sample size of <50 participants. Although findings varied markedly across studies, structural brain injury (eg, hemorrhage, infarction) was often linked to poorer neurodevelopmental outcomes in infancy, and brain volumes were linked to outcomes in children and adolescents. Microstructural (eg, white matter) and functional changes (eg, seizure activity on electroencephalography) also appear linked to a variety of neurocognitive outcomes, including deficits in attention, learning, memory, social cognition, and executive function in adolescence. It was difficult, however, to draw definitive conclusions because of disparate study methodologies and findings, and replication is strongly recommended. Fetal studies, functional neuroimaging studies, and studies examining mental health outcomes were limited. There is a critical need for multicenter, longitudinal studies using standardized MRI protocols and including data on patient, medical, and environmental factors, samples for genetic analysis, and testing of neurodevelopment and mental health to expose the brain structural, microstructural, and functional underpinnings of neurocognitive and psychological morbidities in individuals with CHD across the lifespan.

HIGHLIGHTS.

• Advances in neuroimaging have improved understanding of brain health and development in patients with CHD.

• Structural, microstructural, and functional neuroimaging findings are linked to neurocognitive outcomes across the lifespan in patients with CHD.

• Multicenter, longitudinal studies incorporating mental health metrics are needed to inform neuroprotective and therapeutic strategies in patients with CHD.

ABBREVIATIONS AND ACRONYMS

ADHD

attention deficit/hyperactivity disorder

BSID

Bayley Scales of Infant and Toddler Development

CHD

congenital heart disease

EEG

electroencephalography

FA

fractional anisotropy

MCA

middle cerebral artery

MRI

magnetic resonance imaging

NIRS

near infrared spectroscopy

TGA

transposition of the great arteries

WMI

white matter injury