To assess the relative contribution of potential risk factors for adverse neurobehavioral outcomes in children referred for evaluation of sleep-disordered breathing (SDB), including weight, mean sleep duration, and comorbid sleep disorders.
Medical record review.
Academic pediatric medical center.
Clinical sample of 235 children aged 3 to 18 years undergoing overnight polysomnography for symptoms of SDB.
History of behavioral, emotional, and academic problems and Child Behavior Checklist (CBCL) scores.
More than half (56%) of the sample was overweight or at risk for overweight, more than one-third (36%) was classified as being short sleepers, and almost half (49%) had at least 1 additional sleep diagnosis. Forty-seven perent had a history of behavioral problems and 23% had a reported diagnosis of attention-deficit/hyperactivity disorder. There were no significant differences in CBCL scores based on any measure of SDB disease severity. Increased weight was associated with increased internalizing CBCL scores in a dose-dependent fashion (P = .003), while short sleepers were more likely to have elevated externalizing scores (P < .001). Overall, the strongest predictor of adverse behavioral outcomes was the presence of at least 1 additional sleep diagnosis (P < .001).
The relationship between SDB and parent-reported behavioral outcomes in children is complex. In addition to SDB-related impairments,
clinicians should consider the relative contributions of being overweight,
insufficient sleep, and comorbid sleep disorders when assessing behavior in these children.
Sleep-disordered breathing (SDB) is a common childhood condition that encompasses a broad pathologic spectrum, ranging from primary snoring (ie, snoring without evidence of ventilatory abnormalities)
to complete or partial airway occlusion (obstructive apneas or hypopneas).
The clinical sequelae of childhood SDB range from cardiopulmonary complications in those with severe disease1 to more subtle daytime neurobehavioral symptoms, including affect changes and behavioral issues, such as hyperactivity and poor impulse control.2- 4 Deficits in attention, memory, and general cognition and impairments in academic performance have also been associated with childhood SDB.5- 13
However, not all studies have found a relationship between SDB in children and deficits in specific neurobehavioral or neurocognitive domains.14,15 Some studies,
for example, have failed to find a direct correlation between polysomnographically derived parameters of disease severity, such as the number of apneic or hypopneic episodes per hour, and cognitive and behavioral problems.16,17 Several recent studies have also found neurobehavioral deficits associated with primary snoring in children that are similar to those found in children with obstructive sleep apnea (OSA), suggesting that disease severity–related variables alone may not account for the impairments observed.18- 20
As a result, attention has recently turned to the role of other factors frequently related to SDB (including obesity and insufficient sleep duration) as potential contributors to neurobehavioral outcomes in these children.21 Obesity, which has been demonstrated to be the most significant risk factor for the development of OSA in adults, has also been identified as an increasingly important risk factor for childhood OSA.22,23 In addition, childhood obesity,
independent of OSA, is associated with a significant number of short-
and long-term health consequences and psychosocial risks, including decreased health-related quality of life and self-esteem and psychosomatic and internalizing symptoms.24- 27
Insufficient sleep has been linked to a variety of neurocognitive impairments, such as executive function deficits as well as behavioral problems and affect dysregulation in children.28- 32 This could potentially exacerbate the effect of SDB. For example,
a recent study suggested that children referred for SDB are more likely to have impaired cognitive function if they have reduced time in bed.33 Short sleep duration has also been identified as a potential risk factor for childhood obesity,34- 37 further complicating the interrelationships among these contributing factors.
Finally, a few studies have also suggested that comorbid sleep disorders, such as behaviorally based insomnias or periodic limb movement disorder, may contribute to the impact of disease in children with SDB.8,38,39Our previous work39,40 has suggested that not only do these sleep disorders commonly cooccur in children with OSA, but their presence also appears to increase the likelihood of behavioral problems associated with increased sleepiness.
To our knowledge, no studies have simultaneously and comprehensively assessed the effect of other sleep-related risk factors, such as comorbid sleep problems and short sleep duration, in a large sample of children with polysomnographically documented SDB.3,11,18,40- 45Furthermore, relatively few studies of neurobehavioral outcomes of childhood SDB have been conducted with clinical samples of children referred for suspected SDB, and these generally have been limited to small samples. Yet, an increased understanding of the relative contribution of potential risk factors to adverse neurobehavioral outcomes in children undergoing diagnostic evaluation for SDB could have practical value in assisting the clinician to triage, target appropriate interventions, and develop more comprehensive treatment plans. The use of a clinical sample also allows for identification of comorbid sleep problems in children with SDB and assessment of their relative contribution to behavioral outcomes.
Therefore, the primary goal of this medical record review was to systematically assess which of a number of potential risk factors are most strongly associated with adverse neurobehavioral outcomes in children referred for evaluation of SDB. We hypothesized that children referred for assessment of SDB symptoms who met polysomnographic diagnostic criteria for OSA would have more parent-reported behavioral problems compared with children who did not meet diagnostic criteria and that the relationship between SDB severity and adverse neurobehavioral outcomes would be influenced by a number of other confounding factors commonly present in children with SDB, specifically, an increased body mass index (BMI) (calculated as weight in kilograms divided by height in meters squared), shortened sleep duration, and comorbid sleep problems.
All children and adolescents between the ages of 3 and 18 years referred for overnight polysomnography between 1999 and 2005 at a tertiary care children's hospital for SDB symptoms were initially screened for our sample. Patients were either evaluated in the Pediatric Sleep Disorders Clinic, Providence, Rhode Island, before undergoing overnight polysomnography or were first directly referred to the sleep laboratory by subspecialists in the community (ie, otolaryngologists)
and subsequently seen for follow-up in the clinic. The Pediatric Sleep Disorders Clinic is staffed by a multidisciplinary team of pediatricians and psychologists. As part of the routine clinical evaluation, parents completed intake packets including sleep, medical, developmental,
and family history questionnaires developed by the clinic staff as well as several standardized questionnaires. A nurse or trained technician measured and recorded the weight and height of all of the children at the time of the sleep clinic evaluation or polysomnography. Extensive clinical evaluations, including diagnostic interviews, physical examinations,
and laboratory and radiologic tests, when appropriate, were conducted by members of the multidisciplinary team, who determined the final diagnoses. Evaluators were blind to the results of the Child Behavior Checklist (CBCL). Only children who had complete data for overnight polysomnography, concurrent growth parameters, the CBCL, and parent-reported sleep duration and a complete evaluation for comorbid sleep disorders were retained in the final sample. Children with significant medical conditions (eg, diabetes) or with significant developmental delays were excluded. The hospital institutional review board approved the study.
All night, attended polysomnographic evaluation consisted of the following: 2 electrooculography leads, 2 electroencephalography leads, a submental electromyography, a nasal/oral airflow thermistor and/or nasal pressure transducer, a snoring microphone, chest and abdominal inductance plethysmography bands, a single-channel electrocardiograph,
pulse oximeter, and bilateral anterior tibial leg electromyography leads. Sleep architecture was scored in 30-second epochs according to Rechtschaffen and Kales.46 Apnea was defined as a complete cessation of airflow for 2 or more respiratory cycles. Hypopnea was defined as an event that met at least 2 of the following criteria: (1) decrease in airflow of 50% or more, (2) an electroencephalograph arousal, or (3) oxygen desaturation of 2% or greater. The respiratory disturbance index (RDI) was defined as the number of obstructive and/or mixed apneas and hypopneas per hour of total sleep. Respiratory disturbance index electroencephalograph arousals were those occurring in association with an apneic or hypopneic event.
Nadir oxygen saturation was defined as the lowest oxygen saturation measured by pulse oximeter.
Participants were divided into 3 weight groups based on sex-
and age-adjusted norms for BMI, as recently published by the Centers for Disease Control and Prevention.47 Children between the 15th and 84th BMI percentiles for their age and sex were classified as having average weight; children between the 85th and 94th percentiles were classified as at risk for overweight;
and children above the 95th BMI percentile for their age and sex were classified as overweight. In the current sample, there were very few children (n = 6) who presented below the 15th percentile for BMI. Owing to the small population size, these children were excluded from the data set.
Three open-ended items from the Children's Sleep Habits Questionnaire,
a 33-item parent-reported sleep measure48 (average bedtime, wake time, and nap duration, if applicable),
were used to compute parent-reported mean 24-hour sleep duration.
The CBCL for patients aged 4 to 18 years is a 113-item age- and sex-adjusted measure of global behavioral functioning with good reliability and validity data.49T scores were derived for internalizing, externalizing, and total behavior problems as well as for the following measures: withdrawn,
somatic, and/or anxious/depressed behavior, social problems, thought problems, attention problems, delinquent behavior, and aggressive behavior. The cutoff for at-risk clinical range scores was set at a t score of 60 or higher.
Several single items regarding the child's mental health history from the Pediatric Sleep Disorders Clinic intake questionnaire were also included as study variables, including history of (1) emotional,
behavioral problems, or academic problems and/or previous mental health counseling and (2) attention-deficit/hyperactivity disorder (ADHD).
Finally, children evaluated in the Pediatric Sleep Disorders Clinic who met International Classification of Sleep Disorders50 diagnostic criteria for 1 or more sleep disorders, exclusive of SDB-related disorders, were included in the additional sleep disorder group. Additional sleep diagnoses, grouped by International Classification of Sleep Disorders categories, included movement, parasomnia, hypersomnia, insomnia, and circadian rhythm disorders.
All variables were plotted to determine normal distribution.
Variables with a skewed distribution (eg, BMI and RDI) were log transformed,
and the log-transformed variables were used in all analyses. To account for RDI values of 0, a new variable, RDI + 1, was created and log transformed for all analyses. However, in the Tables, original values for any log-transformed variable are provided to allow for clinically relevant interpretation. χ2 Analyses were run to examine differences in basic demographics as well as categorical variables between groups. A series of 1-way analyses of variance were run to explore the effect of SDB group, weight, and sleep duration on behavioral symptoms. Significant differences were followed up with the Tukey test. The relationships between variables, including BMI and RDI, were explored through Pearson r correlations.
Finally, a series of hierarchical multiple regression equations were run to examine the predictors of severity of behavioral symptoms.
Given the exploratory nature of these analyses, P ≤ .05 was considered statistically significant and a Bonferroni correction was not used. Data were analyzed with a commercially available statistical software package (SPSS; SPSS Inc, Chicago, Illinois).
A total of 235 participants met inclusion criteria. Most (77%)
were initially evaluated in the Pediatric Sleep Disorders Clinic.
Because there were no clinically significant differences in demographic or behavioral variables between children evaluated first in the sleep clinic and those directly referred for polysomnography, the 2 groups were combined in the analyses.
Demographics for the sample, polysomnographically defined SDB disease-severity variables, mean BMI and weight categories, and behavioral and sleep variables are presented in Table 1. More than half (56.2%) of the sample was in the at risk for overweight or overweight categories. Because there are little empirically based data to define ideal sleep amounts in children,
we used a definition of short sleep that was based on a sleep duration of less than the 10th percentile for reported norms by age.51 The sample was divided into short vs average sleepers based on cutoffs of less than 10 hours for children aged 3 to 5.9 years, less than 9 hours for children aged 6 to 12.9 years,
and less than 8 hours for adolescents aged 13 years and older; more than one-third (36.2%) of the participants overall were short sleepers.
Almost half (n = 116 [49%]) of the children met International Classification of Sleep Disorders diagnostic criteria for at least 1 comorbid sleep disorder. Mean CBCL total and subscale scores and the percentage of participants falling in the clinical range are also presented in Table 1.
Because there are currently no universally accepted disease cutoffs for diagnosing SDB in children and because of the wide age range of the sample, we used several different definitions. Table 1 presents the results of both a more conservative (RDI score cutoff of ≥ 5) and a broader definition (RDI score cutoff of ≥ 1) of RDI to categorize the sample as having OSA or primary snoring. It should be noted that the largest percentage of the sample fell into a borderline zone of an RDI between 1 and 4.99 (n = 96 [40.9%]).
Using either OSA RDI definition and comparing the 3 SDB groups (RDI score < 1, 1 ≤ RDI score < 5,
and RDI score ≥ 5), we found no significant differences in the percentage of children with a history of developmental, emotional,
or school problems or ADHD diagnosis; in CBCL scores; nor the percentage of children in the at-risk/clinical range on these scores. We also examined correlations between CBCL subscales and RDI and other polysomnographic measures of disease severity (ie, arousal index and baseline nadir oxygen saturation) expressed as continuous variables. The only significant correlations were modest ones between RDI and somatic complaints (r = 0.14, P = .03)
and between the social problems subscale and RDI (r = 0.23, P = .
001), total arousal index (r = 0.16, P = . 01), and log baseline nadir oxygen saturation (r = 0.21, P = .002).
There was no difference in the percentage of children in each weight group with a history of developmental, emotional, or school problems or ADHD diagnosis. Significant weight group differences emerged in total and internalizing CBCL scores and in the percentage of each group in the clinical/at-risk range on these scores (Table 2). Using BMI percentile (adjusted for age and sex) as a continuous variable, the significant correlations with functional outcomes yielded a similar pattern. Increased BMI percentile was associated with total (r = 0.22, P = .001) and internalizing (r = 0.22, P = .001)
CBCL scores and with the social problems (r = 0.35, P < .001),
thought problems (r = 0.25, P < .001), withdrawn behavior (r = 0.24, P < .001),
anxious behavior (r = 0.16, P = .01), somatic complaints (r = 0.15, P = .02),
and aggressive behavior (r = 0.13, P = .05) subscales.
Short sleepers were significantly more likely to have a reported diagnosis of ADHD (P = .007) and to have higher total CBCL scores (Table 3). However, in contrast with the weight groups, the pattern of differences in subscale scores reflected more externalizing behaviors.
Mean parent-reported sleep duration as a continuous variable was also modestly negatively correlated with a number of functional outcomes (total [r = − 0.21, P = .001] and externalizing [r = − 0.22, P = .001] scores) and the following subscales: aggressive behavior (r = − 0.22, P = .001), delinquent behavior (r = − 0.22, P = .001), social problems (r = − 0.15, P = .03),
attention problems (r = − 0.15, P = .03), and anxious behavior (r = − 0.14, P = .03).
The number and percentage of children with 1 or more additional sleep diagnoses and the types of diagnoses are presented in Table 4. Compared with children without an additional sleep diagnosis, those with an additional sleep disorder were significantly more likely to have a history of both behavioral,
emotional, and/or academic problems and ADHD diagnosis. The additional sleep disorder group also had significantly higher total, externalizing,
and internalizing CBCL scores (Table 4). Table 5 presents results from a series of independent t tests examining the relative contribution of different sleep diagnoses to behavioral outcomes.
Table 6 presents the results of the hierarchical multiple regressions analyses with demographic variables (age, race, sex, and socioeconomic status) and the following continuous variables, which were entered with total, internalizing,
and externalizing CBCL scores: BMI, RDI, parent-reported sleep duration,
and comorbid sleep disorder status (yes or no). While BMI predicted both total and internalizing CBCL scores, sleep duration predicted CBCL externalizing scores; however, the strongest predictor of all 3 CBCL scores was the presence of an additional sleep diagnosis.
The results of this study suggest that the relationship between SDB and parent-reported behavioral outcomes is a complex one, at least in this clinical sample of children referred for overnight polysomnography for suspected SDB. Contrary to our original hypothesis, there was little significant correlation between adverse behavioral outcomes (defined both by a history of behavioral or emotional problems and by concurrent parental ratings of behavior) and any of the measures used to denote SDB disease severity. In contrast, weight group appeared to be more closely associated with poor behavioral outcomes, specifically internalizing problems, in a dose-response fashion, with the most consistent differences between the overweight and average weight groups.
As predicted, a shorter mean sleep duration was associated with worse behavioral outcomes, particularly externalizing concerns. Finally,
and somewhat surprisingly, the presence of at least 1 other sleep disorder diagnosis, particularly insomnia, was the most significant predictor of CBCL scores (though it should be noted that the correlations in the regression were modest). This suggests that, in clinical settings,
a combination of risk factors for adverse behavioral outcomes may be most important in identifying those children who are most likely to have behavioral and emotional problems and thus may be most in need of aggressive intervention and close follow-up.
The lack of association between SDB severity and behavioral outcomes in this study should not in any way be interpreted as implying that SDB does not have a deleterious effect on behavior in children.
Because ours was a referral population, all children had some presenting symptoms suggestive of OSA (eg, snoring), and thus, by definition,
fell within the clinical spectrum of SDB. Our results may be interpreted as suggesting that, within that spectrum, severity may have a relatively weaker influence on behavioral outcomes. Furthermore, children with primary snoring may be at a similar risk for adverse outcomes compared with children who meet diagnostic criteria for OSA, echoing the finding of several other recent studies.3,16,17,45,46 The possibility also exists that the polysomnography parameters currently used to denote disease severity may not reflect more subtle alterations in sleep microarchitecture, for example, that are potentially better predictors of neurobehavioral outcomes. Finally, other factors in addition to severity, such as duration of disease and timing of exposure,2 may have a relatively greater effect on specific outcome measures in these children.
Similar to other studies, we found high base rates of behavioral and academic problems, eg, the percentage of children in our sample with a reported diagnosis of ADHD was 2 to 4 times greater than in the general population. In comparison with other studies of children with SDB that have used the CBCL as an outcome measure, our study population had mean scores within a similar range,3,8,9,16,18,25,47- 51 though, not surprisingly, the percentage of children in this referral sample within the clinical/at-risk range appears to be significantly higher than most of the nonclinical samples in other studies.
The high rate of comorbid sleep problems in the sample (49%)
also deserves comment. A robust association was found in this study between these comorbid sleep disorders and daytime behavioral problems.
One explanation for this association is that the insufficient sleep and/or sleep disruption associated with comorbid sleep problems may lead to increased behavioral morbidity. For example, a recent study38 of children with both a history of snoring and behavioral sleep problems reported more problematic behavior than children with snoring alone. Of course, cross-sectional descriptive studies such as these cannot directly demonstrate causality. Rather than sleep disorders leading to daytime behavioral disruption, it may be that children with daytime behavioral issues are more likely to have nighttime ones as well. In any event, these results suggest that clinicians should systematically assess children presenting primarily with symptoms of SDB for other sleep disorders.
It should be noted that there are a number of limitations in our study design that may increase the likelihood of type II errors2 and affect the generalizability of the results.
The limitations associated with the nature of our study sample as a referral population have been noted previously. Furthermore, only parent-reported measures were used to assess behavioral outcomes,
without independent observer corroboration, such as teacher ratings or objective measures of neurobehavioral function. This may also have resulted in a spurious overestimation of the association between behavioral outcomes and SDB, because parents may be more likely to report behavioral problems in children with known symptoms of SDB and attribute them to SDB. Finally, we did not include a control group of children without SDB in this descriptive study, but chose rather to compare groups within the sample and with established norms for the main outcome variable (CBCL).
Despite these limitations, the results of this study add to the current body of literature on the consequences of SDB in the pediatric population by emphasizing and further delineating the multifactorial and complex etiology of the relationship between SDB and behavioral outcomes. That SDB in children results in cognitive and behavioral dysfunction across a number of domains and that treatment of SDB reverses at least some of these deficits is no longer at issue and the findings of this study do not in any way obviate the need for continued routine screening for SDB in clinical settings. What is now needed is a more sophisticated understanding of the nature and relative contribution of the various causes of sleep disruption that occur, both as a result of and in association with SDB. In particular, their role in compromising daytime alertness and the subsequent link between decreased alertness and behavior highly deserve additional research. From a pragmatic standpoint, it seems clear that clinicians also need to consider the relative contributions of overweight, insufficient sleep, and comorbid sleep disorders in screening, triaging, evaluating, and designing interventions for children with SDB.
Correspondence: Judith A. Owens,
MD, MPH, Rhode Island Hospital, 593 Eddy St, Potter Bldg, Ste 200,
Providence, RI 02903 (firstname.lastname@example.org).
Accepted for Publication: November 25, 2007.
Author Contributions: Dr Owens had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Owens and Mehlenbeck. Acquisition of data: Owens, Lee, and King. Analysis and interpretation of data: Owens, Mehlenbeck,
Lee, and King. Drafting of the manuscript: Owens, Mehlenbeck, Lee, and King. Critical revision of the manuscript for important intellectual content: Owens and Mehlenbeck. Statistical analysis: Mehlenbeck,
Lee, and King. Administrative, technical, or material support: Mehlenbeck and King. Study supervision: Owens and Mehlenbeck.
Financial Disclosure: None reported.
Owens JA, Mehlenbeck R, Lee J, King MM. Effect of Weight, Sleep Duration, and Comorbid Sleep Disorders on Behavioral Outcomes in Children With Sleep-Disordered Breathing. Arch Pediatr Adolesc Med. 2008;162(4):313-321. doi:10.1001/archpedi.162.4.313