Pediatric obstructive sleep apnea (OSA) has a prevalence of up to 5% and is the leading sleep disorder in children (1, 2). The condition is characterized by repeated intermittent upper airway obstruction during sleep, which results in episodic alveolar hypoventilation and hypoxemia, as well as sleep fragmentation (2). The condition leads to multiple daytime symptoms and is associated with an increased risk for severity-dependent cognitive, behavioral, and cardiovascular and metabolic morbidities, all of which likely impact optimal behavioral and health development (3–11). In addition, the cognitive and behavioral deficits frequently encountered in pediatric OSA likely lead to poor school performance (12–15). Brain injury is a possible mechanism underlying these symptoms, as shown in the recently reported gray matter changes in pediatric OSA (16). These earlier findings showed general regions with lower gray matter volume, but many structures of specific interest to cognitive function and school performance were not assessed. In particular, the methodology was suited to detecting changes deep in the brain, whereas much executive function and cognitive processing is centered in cortical structures. This lack of understanding of whether and how cortical regions are affected limits selection of possible interventional targets, particularly for the cognitive difficulties, which are especially important for children in relation to their scholastic development.

Previous studies examining OSA in adults have shown brain injury and physiological and psychological symptoms similar to those detected in pediatric OSA. Neuroimaging studies in OSA report alterations of white and gray matter, as well as changes in neurochemicals within various brain regions, raising the possibility that pediatric patients may also exhibit such brain changes (17–22). Our earlier study in children with OSA focused on regional gray matter volume assessment showed selected limbic, subcortical, and brainstem regions with reduced gray matter (16). These brain regions, along with other sites, control cognition and mood, such that OSA-induced injury may underlie the cognitive deficits frequently seen in pediatric OSA. Given that more frequent behavioral symptoms are seen in even milder forms of OSA (23, 24), more widespread cortical areas may be impacted. However, the “voxel-based morphometry” technique used for the earlier gray matter quantitation is less sensitive to changes in outer cortical regions that in deeper brains structures, thus raising the need to examine regional cortical sites with a different approach.

Cortical thickness is one measure that is highly sensitive to changes only in neuronal cortical areas (25), and both the cognitive and behavioral findings suggest that cortex-specific regions may be affected, perhaps due to injury from the intermittent hypoxia or sleep fragmentation in pediatric OSA (7, 11, 26). Cortical thickness examination can be performed with standard image processing techniques, such as FreeSurfer software (27). This approach involves detecting the inner and outer surfaces of the cortical gray matter, and calculating the thickness as the distance between the two surfaces. Specialized statistics are used to assess thinning or thickening of the cortex, and this methodology has been applied to many conditions.

The purpose of this study was to assess potential injury to the brain cortical regions in children diagnosed with OSA using polysomnography. Building on earlier findings of gray matter injury in mixed cortical and subcortical areas (16), we hypothesized that regional cortical thickness alterations would occur in many areas in pediatric OSA subjects, and more specifically in those regions that are involved in cognition.

Sixteen children with OSA and nine age, sex, ethnicity, and BMI Z-score matched controls underwent sleep studies, neurocognitive testing, and MR imaging. Demographic and polysomnography data for those subjects are listed in Table 1, showing any pertinent differences in sleep characteristics and BMI Z scores.

Whole-brain group comparisons of the 16 OSA patients with a control group consisting of the nine controls and the 138 Pediatric MRI subjects revealed cortical thinning associated with OSA in various areas including the superior and medial frontal, prefrontal and parietal cortices, and the occipital cortex (FDR = 0.05). Regions of significantly thicker or thinner cortex in OSA are shown in Table 2 (left hemisphere) and Table 3 (right hemisphere). Representations of these cortical thickness differences between control and OSA groups in the left and right hemispheres are shown in Figure 1 from multiple views overlaid onto the FreeSurfer template pial surface, and a lateral, inflated view to illustrate otherwise hidden structures (i.e., insular cortex, and sulci).

There were increases in cortical thickness in the bilateral precentral gyrus, the left central gyrus, and regions in bilateral posterior-mid and right anterior insular cortices (Figure 1; Tables 2 and 3). The posterior cingulate and sub-genu of the anterior cingulate cortices, extending into medial prefrontal areas also showed increased thickness in OSA children, as did the temporal cortex and poles, the latter especially on the right side. The left side showed thickening in the precentral gyrus.

The analysis of DAS scores and cortical thickness performed in the 25 children studies at the University of Chicago did not reveal any significant relationships. The small number of subjects meant that the statistical power and hence sensitivity was low for this analysis.

Children with OSA exhibit evidence of significant cortical thinning in the frontal (including the superior frontal gyrus), prefrontal, parietal, temporal, and occipital cortices, all of which are consistent with the gray matter decreases previously reported (16, 36–38). Thus, the findings further attest to the presence of neural injury in pediatric OSA, extending earlier indications of injury to cortex-specific areas. These cortical regions play important roles in cognition, executive function, and memory. The frontal lobe is the central region for motor function, problem solving, memory, language, impulse control, and social behavior (39). The prefrontal cortex plays a large role in executive function, attention, and personality development (39). The prefrontal cortex continues to mature into late adolescent years and even early 1920s (40, 41), such that damage at a young age could greatly hinder the normative development processes of higher cognitive functions. Both the parietal and occipital cortices also contribute to cognition and behavior; the parietal cortex through spoken and written word comprehension, encoding, consolidating and retrieving memory, and manipulating working memory (42–44) and the occipital cortex through memory, visuoconstruction (the ability to manipulate spatial information), counting and arithmetic, and executive tasks (45–47). Therefore if damaged, as cortical thinning would suggest, the frontal, parietal, and occipital cortex could exhibit altered function, affecting neurobehavioral development and functioning.

In adult OSA, cortical thinning is often attributed to cell death or neuronal injury, as thinning constitutes the most likely imaging correlate of dramatic volume reduction in a mature brain (17, 48–51). Earlier findings of reduced gray matter volume (16) could reflect both atrophy and vascular or cell damage, including injury with atrophy, whereas cortical thinning is specific to atrophy per se. While direct neuronal injury is a likely event in pediatric OSA due to the intermittent hypoxia and sleep fragmentation that characterize this condition, the effect of OSA on brain maturational processes must also be taken into account. In children ages 5–11 years undergoing normal neural development, the frontal and occipital lobes increase in volume by 0.4–1.5 mm every year (52). Therefore, a disruption of neural developmental processes could be the underlying cause of volume reduction in pediatric OSA as opposed to implicitly signifying the presence of neurodegeneration. Thus, OSA could prevent normal developmental brain volume increases through the same aforementioned mechanisms implicated in neurodegeneration, resulting in thinner cortices in children with OSA than in controls.

Cortical thickening was seen in the precentral and left central gyrus, the medial prefrontal cortex, the mid and anterior insular cortex, the posterior and subgenu of the anterior cingulate cortex, and the medial temporal lobe, areas that are involved in cognitive, emotional, autonomic, pain, and motor function. The insular cortex is involved in emotional control, self-awareness and cognitive functioning and motor control (53–58). The anterior cingulate cortex also plays a role in emotions as well as reward, decision-making, and autonomic regulation including blood pressure and heart rate (59–61), while the posterior cingulate is involved in human awareness and pain and episodic memory retrieval (62–64). The medial temporal lobe is a key region for long-term memory (along with the hippocampus region) (65) and the precentral gyrus is the site of the primary motor cortex and the motor strip and controls voluntary skeletal muscle movement (66–68). Cortical alterations in these regions could result in the cognitive, autonomic, memory, and psychological symptoms experienced in OSA, including worse performance on cognitive tasks, hypertension and high sympathetic tone, which could eventually cause cardiovascular issues, memory loss, and depressive symptoms, respectively. Nonetheless, the relatively small number of OSA subjects and the fact that some children had evidence of cognitive deficits while others did not may account for the absence of a significant association between cortical thickness and cognitive test scores.

Increases in cortical thickness could also indicate abnormal neural development in pediatric OSA patients. Synaptic pruning is a normally occurring process that is actively regulated through childhood and adolescence, whereby synapses are removed in order to optimize brain function (52, 69). In children ages 5–11 years, cortical thinning occurs in the right dorsal frontal and bilateral parietals regions by about 0.15–0.30 mm per year as a result of synaptic pruning (52, 69). In pediatric OSA, the present findings suggest the normal process of synaptic pruning may be disrupted. In addition, cortical thickness may also increase due to hypoxia-induced neuro-inflammation and glial activation, indicating an immune response in affected cortical areas. Identifying the exact cause will obviously require further investigation with additional complementary imaging approaches such as diffusion tensor imaging, MR spectroscopy, or functional MRI, which would provide information on alterations in regional neurochemical substrates among pediatric OSA subjects when compared to controls.

The cognitive and behavioral symptoms reported in children with OSA are consistent with symptoms reported by sleep deprivation and disruption studies in children (70). Altered sleep duration and continuity may induce behavioral perturbations in children with sleep-disordered breathing, ultimately favoring the occurrence of dis-maturational processes in the brain and finally manifesting as poor performance in school (71). Although we found no significant associations between the structural findings and the cognitive test performances, the absence of such associations was not surprising considering the large heterogeneity in the prevalence of a cognitive deficit phenotype, which would, therefore, require a markedly larger sample size than the present 25 to enable the detection of such correlation. Neuroimaging studies in particular benefit from greater sample sizes (34). Another consideration is that some cortical structures showed multiple subregions affected; in terms of whether one cortical structure can have meaningfully distinct subregions, we have shown that for one area, the insular cortex, that subregions are functionally distinct (72). Thus, within one structure, the different subregions affected in pediatric OSA may have different neuropsychological consequences. Furthermore, cortical thickness changes may reflect a late-stage of brain alterations (since atrophy occurs over many years), and other measures of structure may detect earlier pathology. Since OSA can go for years undetected, it is unclear how long the participants in our study had the disorder, or how long they neuropsychological performance may have been affected. For example, within the hippocampus, changes in structure measured by water diffusivity are correlated with neuropsychological function (37). Whether such subtle structural changes extend to the cortex is unclear, but assessment of water diffusion specifically within the cortex is warranted for future studies.

In summary, we present evidence of regionally defined cortical thinning and thickening in children with OSA compared to healthy controls. These findings clearly indicate the presence of injury to the brain that reflects either the consequences of neuronal cell losses, the imposition of perturbed maturation brain processes, or a combination thereof in those sites with cortical thinning and hypoxia-induced inflammatory changes in sites with cortical thickening. Further studies evaluating the effect of intervention on these findings are needed.

The study was approved by the human ethics committee at the University of Chicago (IRB Protocol # 11-0280-CR004), and written informed consent was obtained from the legal caregiver of each participant in accordance with the Declaration of Helsinki. Assent was obtained from children >7 years old.

PM performed major components of image data analysis and contributed to manuscript editing. LK-G conceptualized the study, recruited subjects, coordinated the database, analyzed data, and drafted portions of the manuscript. JP performed data analyses. RM performed data analyses. RK performed components of data analyses and interpretation and contributed to manuscript editing. MP performed sleep study scoring and interpretation, coordinated and performed the MRI scans. DG participated in the conceptualizing the study and edited the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.