Predictors of Behavioral Changes After Adenotonsillectomy in Pediatric Obstructive Sleep Apnea: A Secondary Analysis of a Randomized Clinical Trial

Abstract

Importance: Adenotonsillectomy (AT) is associated with improved behavior in children with obstructive sleep apnea (OSA). However, it is unknown whether polysomnographic parameters are superior to the parent-reported severity of sleep-disordered breathing (SDB) in predicting behavioral changes after AT.

Objective: To ascertain whether polysomnographic parameters vs parent-reported severity of SDB are better predictors of treatment-related behavioral changes in children with OSA.

Design, setting, and participants: This ad hoc secondary analysis of the Childhood Adenotonsillectomy Trial (CHAT) downloaded and analyzed data from January 1 to January 31, 2020. Children aged 5 to 9 years with a polysomnographic diagnosis of OSA were enrolled in the CHAT and subsequently randomized to undergo either early AT or watchful waiting with supportive care. All outcome measures were obtained at baseline and at follow-up (7 months after randomization).

Interventions: Early AT vs watchful waiting with supportive care.

Main outcomes and measures: Postrandomization changes between the baseline and follow-up periods were derived from (1) T scores in 4 validated behavioral assessments (Conners Global Index parent and teacher versions, Behavior Rating Inventory of Executive Function metacognition index, and Child Behavior Checklist of total, internalizing, and externalizing behavior subscales); (2) 8 aggregated polysomnographic parameters representing the severity of obstruction, hypoxemia, sleep quality, and structure; and (3) the parent-reported severity of SDB measured by the Pediatric Sleep Questionnaire-Sleep-Related Breathing Disorder (PSQ-SRBD) scale. The treatment-related changes in each of the behavioral outcomes attributable to changes in SDB severity (represented by the subjective PSQ-SRBD score and objective polysomnographic parameters) were measured and compared using mediation analysis.

Results: A total of 453 children were assessed at baseline, of whom 234 were girls (52%) and the mean (SD) age was 6.6 (1.4) years. The postrandomization changes in 7 of 8 behavioral outcome measures between the baseline and follow-up periods were partially mediated by the changes in PSQ-SRBD scores (range of nonzero causally mediated effects, 2.4-3.5), without contribution from any of the polysomnographic parameters.

Conclusions and relevance: This secondary analysis of a national randomized clinical trial found that most treatment-related behavioral changes in children with OSA were mediated by the changes in parent-reported SDB severity alone. These findings suggest that polysomnographic parameters provide clinicians with limited means to predict the improvement in neurobehavioral morbidity in OSA.

Trial registration: ClinicalTrials.gov Identifier: NCT00560859.

Conflict of interest statement

Conflict of Interest Disclosures: Dr Isaiah reported receiving grants from Burroughs Wellcome Fund and the National Institutes of Health during the conduct of the study as well as owning a patent to sleep apnea imaging, which is issued, is licensed, and has royalties paid, and a patent to sleep apnea treatment, which is pending and licensed. No other disclosures were reported.

Figures

Figure 1.. General Causal Mediation Analysis and Magnitude of Changes in Behavior After Treatment of Obstructive Sleep Apnea

A, The average causal mediation effect (ACME) were the changes in a mediator variable (eg, Apnea Hypopnea Index [AHI] or the Pediatric Sleep Questionnaire–Sleep-Related Breathing Disorder [PSQ-SRBD] score) after the intervention (early adenotonsillectomy ([AT] vs watchful waiting with supportive care) was provided. The magnitude of ACME was obtained from the products of the coefficients of the 2 regressions (mediator modeled by the intervention and outcome modeled by the mediator) after controlling for potential confounders. The average direct effect (ADE) was the outcome of treatment that was not attributable to the putative mediation pathways. The ACME and ADE provided the total effect, which was estimated as the size of the intervention effect. B, The magnitude of the ACME, the ADE, and the total effect is shown with the error bars representing the associated bootstrapped measure of uncertainty (95% CIs).

Figure 2.. Comparison of Polysomnographic Parameters to Parent-Reported Severity of Sleep-Disordered Breathing in Predicting Baseline Behavioral Outcomes

Shown are results of a generalized linear model fitted to 2 representative behavioral outcome measures (Conners Global Index parent and teacher versions) and the Pediatric Sleep Questionnaire–Sleep-Related Breathing Disorder (PSQ-SRBD) score (panels A and C) and the Apnea Hypopnea Index (AHI) (panels B and D), after adjustment for all covariates, including demographic and socioeconomic characteristics. The model was fitted with a smoothing spline, with the error bars spanning 1 SD for the horizontal extent of the distribution. Eight polysomnographic parameters were compared with the PSQ-SRBD score in predicting the various baseline behavioral outcomes (panel E). The null-effect model included a regression comprising the covariates only and the baseline behavioral outcome of interest. The effect size was measured by estimating the change in proportion of variance (adjusted R2) by adding the clinical or polysomnographic parameter to the covariates-only model. BRIEF MI indicates Behavior Rating Inventory of Executive Function metacognition index; CBCL, Child Behavior Checklist; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition); REM, rapid-eye-movement sleep; SpO2, oxygen saturation as measured by pulse oximetry; and TST, total sleep time.

Figure 3.. Comparison of Treatment-Related Changes in Behavioral Outcomes vs Parent-Reported Symptoms of Sleep-Disordered Breathing and Polysomnographic Parameters in Predicting Postrandomization Behavioral Outcomes

Exact values illustrated in panel A are reported in the eTable in the Supplement. ACME indicates average causal mediation effect (ACME is unitless, as it is the change in the mediator variable resulting from a unit change in the outcome variable); AHI, Apnea Hypopnea Index; BRIEF MI, Behavior Rating Inventory of Executive Function metacognition index; CBCL, Child Behavior Checklist; DSM-IV, indicates Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition); PSQ-SRBD, Pediatric Sleep Questionnaire–Sleep-Related Breathing Disorder; REM, rapid-eye-movement sleep; SpO2, oxygen saturation as measured by pulse oximetry; and TST, total sleep time.