Behavior Problems Associated with Sleep Disordered Breathing in School-Aged Children—the Tucson Children’s Assessment of Sleep Apnea Study

Abstract

Objective The purpose of the current study was to examine prevalence of and relations between a commonly used measure of nighttime breathing problems, the Respiratory Disturbance Index (RDI), and a range of problem behaviors in community children. Methods Participants were 403 unreferred children aged 6–12 years. Recruitment was completed through public elementary schools. Overnight unattended in-home polysomnography was used to assess sleep and breathing. The RDI was used as the indicator of respiratory events during sleep. The Child Behavior Checklist and the Conners’ Parent Rating Scales-Revised were used to assess behavior. Results Prevalence rates for Attention, Cognitive Problems, Aggression, Oppositional behavior, and Social Problems were greatest for subjects with high RDIs. Prevalence for Internalizing behaviors was not greater for those subjects with high RDIs. Hyperactivity was not strongly related to higher RDIs. Conclusions Behavioral problems may exist in the presence of nocturnal breathing events in unreferred children. Specific patterns of behavioral morbidity have still not been established. Some behaviors, such as hyperactivity, may show differing sensitivity and specificity in relation to the RDI.

Received November 15, 2004; revision received January 20, 2005 and March 24, 2005; accepted April 20, 2005

Introduction

Sleep disordered breathing (SDB) is a condition in which breathing is reduced for brief periods of time during sleep. Sleep apnea, defined as complete cessation of breathing, is the most extreme form of SDB. In children, the minimum duration of these periods is 2 breath cycles or 6 seconds (Goodwin et al., 2001), but some can be longer than 20 seconds. The etiology of the disorder in children has been linked to obesity, central nervous system problems, genetic abnormalities such as Down’s syndrome, structural abnormalities of the crianiofacial area, and enlarged tonsils (Erler & Paditz, 2004). Adults with this disorder exhibit excessive daytime sleepiness, cognitive problems, and loud snoring (Gotfried & Quan, 1984). SDB has been estimated to be present in 2–3% of the pediatric population. (For reviews of SDB in children, see O’Brien, Mervis, Holbrook, Bruner, Smith et al., 2004, and Schecter, 2002).

Recent research has indicated that children with SDB or snoring show higher levels of, and increased risk for, behavioral and cognitive abnormalities (Blunden, Lushington, Kennedy, Martin, & Dawson, 2000; Chervin, Dillon, Bassetti, Gonoczy, & Pituch, 1997; Guilleminault, Korobkin, & Winkle, 1981; O’Brien, Mervis, Holbrook, Bruner, Klaus et al., 2004; Owens, Spirito, Marcotte, McGuinn, & Berkelhammer, 2000; Rosen et al., 2004; Schecter, 2002). Studies to date have been correlational in nature. The strongest causal data available indicate that treatment for SDB can result in significant improvement in behavior and school performance (Gozal, 1998; Stradling, Thomas, Warley, Williams, & Freeland, 1990). Studies that have examined sleep using polysomnography in children with previously diagnosed Attention Deficit Hyperactivity Disorder (ADHD) have found increased sleep related problems such as sleepiness, reduced Rapid Eye Movement (REM) sleep, or awakenings but none have found an increased level of nocturnal breathing events (apneas/hypopneas) during sleep (Cohen-Zion & Ancoli-Israel, 2004; Golan, Shahar, Ravid, & Pillar, 2004; O’Brien et al., 2003).

A wide range of behavioral problems have been associated with SDB or snoring in children including somatic complaints, anxiety/depression, aggression, oppositional behavior, conduct problems, inattention, and hyperactivity (Blunden et al. 2000; Chervin et al., 1997; Chervin et al., 2002; Chervin, Dillon, Archbold, & Ruzicka, 2003; Lewin, Rosen, England, & Dahl, 2002; Rosen et al., 2004). Behaviors associated with ADHD, primarily inattention and hyperactivity, have been the focus of multiple studies. The observation of hyperactive behavior in the context of actual or reported nocturnal breathing problems has led to the hypothesis that a potentially significant proportion of children with SDB are being diagnosed with ADHD and that this association could have a substantial public health impact (Chervin et al., 1997; Lewin et al., 2002).

Nocturnal breathing events such as apneas and hypopneas (abnormal breathing events during sleep that are less severe than apneas) have been regularly related to behavior problems, but only inconsistently related to specific behavior problems. For example, one study found no relationship between nocturnal breathing events and hyperactivity (Chervin et al., 2002), and two studies using small samples found behavioral problems to actually be lower for mild versus moderate SDB (Owens et al., 2000; Lewin et al., 2002). A recent prospective study of school-aged children (with a sample heavily weighted with formerly preterm subjects and including subjects with primary snoring) found evidence of increased hyperactivity along with a broad range of behaviors including internalizing behaviors (Rosen et al., 2004). Thus, the studies to date provide variable empirical support for the hypothesis that children with SDB exhibit specific patterns of behavior problems. There are several possible methodological reasons for a lack of consistency or pattern. Early studies did not utilize standardized behavioral measures. Moreover, some studies have used only parent report for sleep and breathing problems, and others included subjects with primary snoring. While research has positively related nocturnal breathing problems to daytime behaviors in children, few have used polysomnograms (PSGs) together with standardized behavioral measures.

Regardless of the specificity of behavioral consequences related to nocturnal breathing problems, clinicians and researchers will be limited in determining their importance until rates of relevant behaviors, nocturnal breathing events, and their joint prevalence or comorbidity are established. Base rates need to be derived from large community-based samples. Rates of pediatric behavioral problems across a wide range of severity of breathing events during sleep in a healthy sample are useful for comparison with clinical samples. Finally, examination of these relationships in healthy subjects is important because the clinical threshold for the number of apneas and hypopneas has not been established for children.

The analyses presented here are derived from the Tucson Children’s Assessment of Sleep Apnea (TuCASA) study, a cross sectional cohort of healthy children. Previous analyses from the TuCASA study have examined the feasibility of in-home polysomnography (Goodwin et al., 2001), the impact of SDB on learning (Kaemingk et al., 2003), association of parasomnias and SDB (Goodwin et al., 2004), relationship between ventilatory drive and SDB (Fregosi et al., 2004), correlation of pharyngeal size with SDB (Fregosi et al., 2003), prevalence of clinical symptoms of SDB (Goodwin et al., 2003), relationship of SDB to blood pressure (Enright, Goodwin, Sherrill, Quan, & Quan, 2003), and normative values for sleep (Quan et al., 2003).¹ The current analyses are the first from TuCASA to document the association between nocturnal breathing events and comorbid behavioral problems. We hypothesized that the occurrence of hypopneas and apneas as defined by an elevated Respiratory Disturbance Index (RDI) would be associated with higher levels of reported problems with attention and a broad range of externalizing behaviors.

Methods

Participants

The Tucson Unified School District (TUSD) and University of Arizona Institutional Review Boards approved the study protocol. All families completed approved parental consent and child assent forms before participating in the study. The design of the TuCASA study specified recruitment of 6–12-year-old Hispanic and Caucasian children to undergo an unattended home PSG and perform a neuropsychological and behavioral assessment (Goodwin et al., 2001). Subjects were recruited through the TUSD, a large district with a population representative of children living in Southern Arizona. At the time of initial contact parents were asked to complete a short sleep habits questionnaire (created for this study), related to typical bedtime behaviors, and to provide their contact information if they would allow study personnel to call them for further participation. Approximately, 33% (2,327/7,055) of those families who were sent questionnaires returned them. Of those families, 52% provided information to allow for additional contact.

Children were excluded from the study if there was a history of head injury, tonsillectomy, mental retardation, or asthma. Children with known sleep problems or emotional–behavioral disorders (e.g., ADHD) were included in the study. Approximately 1 month after the PSG, each child underwent a neuropsychological evaluation. Parents completed a behavioral assessment by filling out the Child Behavior Checklist (CBCL) and the Conners’ Parent Rating Scales-Revised (CPRS-R) at the time of the neuropsychological evaluation. Approximately, 95% of children who completed PSG studies also participated in the behavioral assessment portion of the study. Families were paid $25 for participation in the evaluation. The present analyses utilize the first 403 subjects who met inclusion/exclusion criteria, whose overnight PSG was technically acceptable, and who completed behavioral assessment measures.

Measures

The CBCL (Achenbach, 1991) allows assessment of 118 parent-reported behavioral and emotional problems of children aged 4–18. Parents rate their child’s behavior on a three-point scale (Not True, Somewhat True, or Very/Often True). In addition to eight syndrome scales, the CBCL includes a Total problem score, and higher order Internalizing and Externalizing scales. Internalizing scales include Anxious/Depressed, Withdrawn and Somatic Complaints. Externalizing scales include Aggressive Behavior and Delinquent Behavior. Three other syndrome scales, Social Problems, Thought Problems, and Attention Problems are also included in the measure. A T-score (M = 50, SD = 10) over 65 was considered to indicate moderate to severe clinical impairment.

The CPRS-R is a well validated 80-item behavior rating scale that measures symptoms of ADHD (Hyperactivity, Impulsivity, and Inattention) as well as comorbid behaviors such as Oppositional behavior, Anxiety, and Somatic Complaints (Conners, Sitarenios, Parker, & Epstein, 1998). CPRS-R scales focus on behaviors central to a diagnosis of ADHD such as Cognitive Problems and Hyperactivity. Additionally, several scales measure behaviors that are commonly comorbid with Inattention and Hyperactivity. Three scales on the CPRS-R are considered internalizing correlates of ADHD (Anxious-Shy, Perfectionism, and Psychosomatic Complaints). Behaviors are rated on a four-point scale that ranges from “Very True” to “Not True.” Seven of the scales on the CPRS-R are derived directly from the Diagnostic and Statistical Manual-IV (DSM) criteria for ADHD (American Psychiatric Association, 1994). The remaining scales were developed empirically. A T-score is derived for each scale, based on a large age and gender specific normative sample. A T-score over 65 is considered to indicate moderate to severe clinical impairment (Conners, 1997).

Polysomnography

An unattended home-based PSG was scheduled as soon as possible after recruitment. Methods for obtaining polysomnographic data have been described previously (Goodwin et al., 2001). In adults, home-based PSG has been shown to provide results similar to PSG studies performed in the laboratory (Iber et al., 2004). Briefly, a two-person, mixed gender team arrived at the home approximately 1 hour before the child’s normal bedtime. Questionnaires were administered, and anthropometric and other physiological measurements were completed. Unattended PSGs were obtained using the Compumedics PS-2 system (Abbotsford, Victoria, Australia). The following signals were acquired as part of the montage: C3/A₂ and C₄/A₁ electroencephalogram, right and left electrooculogram, a bipolar submental electromyogram, thoracic and abdominal displacement (inductive plethysmography bands), airflow (nasal/oral thermister), finger pulse oximetry, electrocardiogram (single bipolar lead), snoring microphone, body position (Hg gauge sensor), and ambient light (sensor attached to the vest to record on/off).

The Compumedics software system was used to process all PSGs. Scoring has been described in detail previously (Goodwin et al., 2001). Briefly, sleep stages were scored according to Rechtshaffen & Kales, 1968) criteria. Apneas were scored if the amplitude (peak to trough) of the airflow signal using the thermister decreased below at least 25% of the amplitude of “baseline” breathing (identified during a period of regular breathing with stable oxygen levels), if this change lasted for > 6 seconds or 2 breath cycles. Hypopneas were designated if the amplitude of any respiratory signal decreased below (approximately) 70% of the amplitude of “baseline” and if the thermister signal did not meet the criterion for apnea. For this study, the RDI was defined as the number of respiratory events (apneas and hypopneas) per hour of the total sleep time irrespective of any associated oxygen desaturation or arousal.

Polysomnograms with less than 4 hours of scorable oximetry were classified as failed studies and not used in analyses. A single registered polysomnographic technologist, who was required to demonstrate a complete understanding of the study’s scoring rules and to articulate reasons for assigning epoch-by-epoch codes for sleep and respiratory scoring, scored all studies. Similar to the quality control procedures used for the multicenter Sleep Heart Health Study, approximately 5% of studies were rescored by the same scorer on a blinded basis to determine consistency in scoring (Redline et al., 1998). No systematic differences were observed between initial and rescored studies.

Analyses

All statistical procedures were completed using SPSS Version 11. Results were summarized as M ± SD or by percentages. For each behavioral measure, raw scores were converted to age and gender standardized T-scores (M = 50, SD = 10). Results are generally presented as those pertaining to the entire sample followed by results related to comparisons between subgroups of the RDI score distribution. One-way analysis of variance was used to test differences between means. Cohen’s d effect sizes (Cohen, 1988) and 95% confidence intervals were calculated as indicators of standardized mean differences between groups. Odds ratios were used to indicate increased probability of behavioral morbidity given membership in the high RDI subgroup. Cut points for continuous variables used in odds ratios were based on generally agreed upon levels of indicated variables, such as using at least a “moderate to severe” level of behavioral problems or based on the most extreme portion of the sample for RDI. A conservative cut point using the upper 15% of the RDI score distribution was utilized here. Diagnostic criteria for SDB using RDI have not been established in children (Schecter, 2002).

Results

Characteristics of the sample are shown in Table I. Obesity was defined using age, gender, and ethnicity-specific norms (Rosner, Prineas, Loggie, & Daniels, 1998). Average sleep architecture values from the overnight polysomnography are in Table II. The mean RDI for the sample was 5.29 (SD = 4.80), the median was 4.30, and the range was .10–72.4. There was no difference between the average RDI for Caucasian (RDI = 5.36, SD = 5.49) and Hispanic (RDI = 5.19, SD = 3.39) subjects (F = .123, p = .725). Sixty-three subjects were classified as the “high RDI” group that was defined as the upper 15% of the RDI score distribution. This left 340 subjects in the lower end of the score distribution. The average RDI for the low RDI group was 4.05 (SD = 1.83) and for the high RDI group was 12.4 (SD = 8.78).

Table III shows CBCL means, F ratio from the analysis of variance, and Cohen’s d effect sizes. Most mean CBCL scale scores for the high RDI group were in or just above the average range (45–55). Mean CBCL scores were significantly higher on the Aggressive, Attention Problems, Social Problems, and Thought Problems for those subjects in the upper 15% of the RDI. In addition, the higher order scale scores, Total and Externalizing were significantly higher for the high RDI subgroup. Effect sizes for scales with significant Fratios were small and ranged from .22–.35 (Cohen, 1988). Effect sizes over 0.2 were noted in 8 of the scales with the two largest effect sizes associated with the Aggressive (.35) and Social Problems (.33) scales.

Table IV shows the proportion of children in the high RDI group who scored in the clinical range for CBCL scales in comparison to the total cohort. Significant odds ratios, indicating a greater probability of having a CBCL score in the clinical range given a high RDI, were found for Attention Problems (24% prevalence), Social Problems (21% prevalence), and Externalizing Behaviors (16% prevalence). The differences in prevalence rates between the two RDI subgroups ranged from 2 to 10 with Attention Problems and Social Problems showing the highest values.

Table V shows mean CPRS-R scale scores by RDI group, F ratios, and Cohen’s d effect sizes. Most mean Conners’ scores for the high RDI group were in or slightly above the average range (45–55). For eight of the 12 scales, mean Conners’ scores for the group representing the upper 15% of the RDI distribution were significantly greater than those in the lower 85% of the distribution. Effect sizes for scales with significant F ratios were small and ranged from .27 to .42. Nine of the scales showed effect sizes over >.20. The two largest effect sizes were associated with the Diagnostic and Statistical Manual-IV (DSM) Inattentive (.42) and the Cognitive Problems (.40) scales.

Table VI shows prevalence rates and odds ratios for the CPRS-R scales. Significant odds ratios (>1) were found for the Oppositional, Cognitive Problems, Social Problems, Psychosomatic, ADHD Index, and DSM Total scales. The most prevalent behavior problems for the high RDI group were found on the DSM Total (25%), Conners’ Oppositional (24%) and Cognitive Problems (24%) scales. The differences in prevalence rates between the two RDI subgroups ranged from –1 to 14 with Oppositional and DSM Total scales showing the largest values.

Discussion

This descriptive cross-sectional analysis of unreferred school children showed positive associations between the number of apneas and hypopneas per hour of sleep and a range of daytime behavior problems. An increased frequency of aggressive or oppositional behavior, as well as cognitive/attention and social problems, were apparent for those subjects with a relatively high RDI across two standardized behavioral measures. Although the overall scale means were in the average range, almost twice as many children with elevated RDIs were identified as having these kinds of moderate to severe behavioral problems on the CBCL and the Conners’ when compared to the rates of those problems for the remainder of the sample.

No clear pattern of behavioral morbidity has been associated with pediatric sleep-related breathing problems. The types of behaviors that were most prevalent in the context of high RDI were cognitive, externalizing, and social behaviors. The types of externalizing behaviors most prevalent were aggressive and oppositional, but not delinquent. Hyperactivity was not strongly associated with higher RDI. It is unclear whether the increased ADHD Index prevalence rate we observed could be attributed to either hyperactivity or inattention because that scale measures both behaviors. Overall, behaviors such as withdrawal, perfectionism, and anxiety/depression were not more prevalent in the context of a high RDI. Social problems, consistently related to RDI across the two behavioral measures, are measured by items related to being liked by other children, shyness/preferring to be alone, and interactions with peers. Problems with social behavior may be a corollary of aggressive and oppositional behaviors as well as attention or cognitive problems.

These results are consistent with some clinical studies that found greater mean levels of behavioral problems in children with SDB compared to normal subjects, or an increased percentage of children with SDB showing behavior problems in a clinically relevant range. Our results extend these findings by demonstrating that healthy children in the general population with relatively greater severity of PSG-documented nocturnal breathing events are more likely to exhibit behavior problems. The pattern of behavioral problems found here are somewhat similar to Rosen et al., 2004. Although that study utilized the same measures and analytic approach, direct comparison to that cohort is difficult given the over-sampling of formerly preterm subjects and inclusion of primary snorers. The results presented here do not strongly confirm some previous research indicating a positive relationship between SDB (or suspected SDB) and hyperactivity in children (Ali et al., 1993; Chervin et al., 1997; Chervin et al., 2002; Guilleminault et al., 1981). However, the results are more consistent with Chervin and Archbold (2001) who did not find positive linear or nonlinear relations between CPRS-R Hyperactivity and components of PSG-documented SDB.

Several caveats should be considered in the interpretation of this study. First, the study is correlational and does not provide confirmation of a causal relationship between RDI and behavior problems. Prospective longitudinal studies will allow strong causal inferences regarding the developmental relationship between these two problems. Second, behavioral measures were dependent upon parent report only. We did not engage teachers or young subjects in evaluating behavior. Multiple informants and multiple methods of measuring behavior will provide the best convergent (or divergent) evidence of behavioral problems. Research in pediatric SDB and behavior will benefit from direct observation of behavior over time to capture qualitative characteristics and within-subject variability. Finally, diagnostic criteria for SDB in children are not yet well defined. The RDI is the most common objective index used to define severity of SDB, but definitive diagnostic or risk-related cut points have not yet been established. Research on the specificity and sensitivity of behavior problems across a wide range of RDI severity may help determine the usefulness of behavioral measures in setting pediatric diagnostic criteria.

Documentation of relationships between behavior, cognition, sleep, and breathing are important, as abnormal cognition and behavior are the symptoms that may be most readily noticed by school professionals or caregivers. The challenge to identification of SDB through behavioral problems may be finding specific patterns of associated behaviors. Behavior problems in children are final common pathways related to multiple etiologies. One etiology may be breathing problems during sleep. Unfortunately, the behavioral and cognitive problems associated with nocturnal breathing events may not fit precisely into current behavioral dimensions or categories. For example, nocturnal breathing events may actually lead to a disregulation of behavior, not a particular tendency toward specific behavior problems. Disregulation or disinhibition may be the best labels for the behaviors found here as it could account for aggression, inattention, and social problems. Although hyperactivity may be a result of disinhibition in ADHD and other disorders, it is an intermittent sign in this range of RDI severity.

In conclusion, we found that children sampled from an unreferred general population with a high RDI have a greater prevalence of abnormal behavior as assessed on two validated and commonly used behavioral assessment instruments. These data provide additional evidence that evaluation of SDB should be considered in children exhibiting moderate to severe levels of some abnormal behaviors, specifically cognitive problems, aggression/opposition, and/or social problems. Behaviors may have differential specificity and sensitivity in relation to SDB. Hyperactivity may be more reliably associated with more severe SDB. Additional data will be collected from the TuCASA study to further explore these relationships within a longitudinal analytic approach. Finally, further exploration of a wide range of behavioral symptoms as well as potential mediators and moderators of behavioral morbidity associated with SDB in children should continue to be pursued.

Acknowledgments

This study was supported by grant HL62373 from the National Heart, Lung, & Blood Institute, Bethesda, MD (Dr. Quan).

References

Author notes

1Center for Evaluation and Program Improvement, Vanderbilt University, 2Arizona Respiratory Center and Department of Medicine, University of Arizona, 3Arizona Respiratory Center and Department of Pediatrics, University of Arizona, 4Department of Pediatrics, University of Minnesota, 5Arizona Respiratory Center and Department of Medicine, University of Arizona, and 6Department of Pediatrics & Steele Memorial Children’s Research Center, University of Arizona