Cognitive Effects of Adenotonsillectomy for Obstructive Sleep Apnea
OBJECTIVE: Research reveals mixed evidence for the effects of adenotonsillectomy (AT) on cognitive tests in children with obstructive sleep apnea syndrome (OSAS). The primary aim of the study was to investigate effects of AT on cognitive test scores in the randomized Childhood Adenotonsillectomy Trial.
METHODS: Children ages 5 to 9 years with OSAS without prolonged oxyhemoglobin desaturation were randomly assigned to watchful waiting with supportive care (n = 227) or early AT (eAT, n = 226). Neuropsychological tests were administered before the intervention and 7 months after the intervention. Mixed model analysis compared the groups on changes in test scores across follow-up, and regression analysis examined associations of these changes in the eAT group with changes in sleep measures.
RESULTS: Mean test scores were within the average range for both groups. Scores improved significantly (P < .05) more across follow-up for the eAT group than for the watchful waiting group. These differences were found only on measures of nonverbal reasoning, fine motor skills, and selective attention and had small effects sizes (Cohen’s d, 0.20–0.24). As additional evidence for AT-related effects on scores, gains in test scores for the eAT group were associated with improvements in sleep measures.
CONCLUSIONS: Small and selective effects of AT were observed on cognitive tests in children with OSAS without prolonged desaturation. Relative to evidence from Childhood Adenotonsillectomy Trial for larger effects of surgery on sleep, behavior, and quality of life, AT may have limited benefits in reversing any cognitive effects of OSAS, or these benefits may require more extended follow-up to become manifest.
- AT —
- CHAT —
- Childhood Adenotonsillectomy Trial
- DAS-II —
- Differential Abilities Scales, 2nd edition
- eAT —
- early adenotonsillectomy
- mESS —
- Epworth Sleepiness Scale modified for children
- NEPSY —
- A Developmental Neuropsychological Assessment
- NEPSY-II —
- NEPSY 2nd edition
- OSAS —
- obstructive sleep apnea syndrome
- PSQ-SRBD —
- Pediatric Sleep Questionnaire Sleep Related Breathing Disorder Scale
- WRAML2 —
- Wide Range Assessment of Memory and Learning, 2nd edition
- WWSC —
- watchful waiting with supportive care
What’s Known on This Subject:
Research indicates variable but possibly selective effects of adenotonsillectomy (AT) on cognitive test scores in children with obstructive sleep apnea syndrome. However, few if any studies have examined changes after AT in a randomized trial assessing diverse cognitive skills.
What This Study Adds:
Findings confirm small, selective effects of AT on cognitive test scores in a randomized trial of AT compared with nonsurgical management, as well as associations of pre-AT to post-AT gains in scores with improvement on measures of sleep disturbance.
Childhood obstructive sleep apnea syndrome (OSAS) is characterized by intermittent upper airway obstruction that disrupts normal ventilation during sleep and sleep patterns.1 The prevalence of OSAS is ∼1% to 6%, with higher rates in African Americans and children from families of lower socioeconomic status.2–4 Children with untreated OSAS are at risk for adverse outcomes ranging from daytime sleepiness and compromised cardiovascular health to behavior problems and impairments in cognition and academic performance.5–11 Problems in behavior and emotional regulation are common in children with OSAS compared with healthy controls, but evidence for adverse effects of OSAS on children’s cognitive abilities is more mixed.6 Some studies fail to find differences between children with OSAS and healthy controls,12,13 and those that do report variable associations of measures of sleep disturbance with cognitive test scores.7,14–18 Similarly, studies of outcomes of adenotonsillectomy (AT) in children with OSAS indicate variable benefits on such tests, with little evidence for associations of these effects with the severity of OSAS and sleep disruption.8,13,14,16,19–26
In the recently completed multicenter Childhood Adenotonsillectomy Trial (CHAT), children with OSAS without prolonged oxyhemoglobin desaturation assigned to early AT (eAT) improved more than those assigned to watchful waiting with supportive care (WWSC) on key secondary outcomes.27–29 Specifically, the eAT group improved more than the WWSC group from a baseline preintervention assessment to a 7-month postintervention follow-up on polysomnographic indices and symptoms of OSAS, global indices of behavior, and quality of life, but not on the primary cognitive outcome measure (A Developmental Neuropsychological Assessment [NEPSY] Attention and Executive Function Domain score) or on global cognitive ability. However, the individual tests that make up these 2 composite measures and other tests predesignated as secondary measures of outcome were not examined for their sensitivity to the effects of AT. Examination of these measures was warranted to determine potential benefits of AT on specific cognitive skills and identify measures sensitive to the effects of AT on children’s functioning but more objective than child behavior ratings.5,10
The primary aim of this study was to determine whether the eAT group improved more than the WWSC group on select measures of cognitive function. Despite variability in the cognitive tests that best discriminate children with OSAS from healthy controls, OSAS-related weaknesses are most evident on tests of sustained and selective attention, response inhibition, nonverbal reasoning, phonological processing, verbal fluency, and fine motor and visual–motor skills.6,7,9,14,16–18,30 Findings from nonrandomized trials of AT in children with snoring or OSAS suggest beneficial effects of surgery on attention and nonverbal problem solving.8,14,19,20 Based on this evidence we hypothesized that the eAT group would improve more across follow-up than the WWSC group on tests of these skills. Within the eAT group we also investigated increases in test scores across follow-up in relation to the degree of improvement in OSAS as measured by overt symptoms and polysomnography. Finally, we explored whether measures of more severe sleep disturbance at baseline were associated with lower baseline test scores.
The rationale and methods of the CHAT trial are detailed in previous reports.4,28 In brief, between January 2008 and September 2011, 453 participants were recruited by screening children 5.0 to 9.9 years of age referred from sleep programs, pediatric and otolaryngology clinics, and the surrounding communities of 6 academic medical centers. Study procedures were approved by the institutional review boards of each center. Informed consent was obtained from parents or guardians, and assent was obtained from children ≥7 years old. Eligible children were otherwise healthy and had a history of snoring, tonsillar hypertrophy, and polysomnography indicating OSAS without prolonged oxyhemoglobin desaturation (<2% of total sleep time with pulse oxygen saturation <90%) and an obstructive apnea index (apneas per hour of sleep) of 1 to 20 or an obstructive apnea hypopnea index (apneas or hypopneas per hour of sleep) of 2 to 30. Children with extreme obesity (BMI z score ≥3) or on psychotropic medications were excluded, including those treated for attention-deficit/hyperactivity disorder.
Procedures and Measures
Before group assignment, participants completed polysomnography and a baseline assessment that included parent ratings of sleep symptoms and child neuropsychological testing.4 Children were then randomly assigned by the data coordinating center to WWSC (n = 227) or eAT (n = 226). Assignment was stratified by site, age (5–7 or 8–10 years), race (African American or other), and overweight status (BMI age- and gender-adjusted z score ≤85% or >85%), with the eAT group receiving surgery within 4 weeks of randomization. All assessments were readministered after 7 months (mean [SD] = 7.1 [0.9]). The follow-up period was chosen as one that would be acceptable to parents and referring physicians while also sufficient to detect post-AT changes in cognitive test scores.7,26,31 Measures are listed in Table 1 and included indices of sleep disturbance as assessed by polysomnography, parent ratings, and neuropsychological tests of verbal skills, nonverbal reasoning, attention and executive function, perceptual–motor and visual–spatial skills, and verbal learning and memory (for test descriptions see Supplemental Table 6). Tests were individually administered in 2 fixed sequences counterbalanced across participants by examiners who were uninformed of group assignment.
Repeated-measures mixed-effects models were fit to assess group differences in change in age-adjusted standard scores from baseline to the 7-month follow-up. Factors were group (WWSC vs eAT), visit (baseline, follow-up), and the group × visit interaction. Stratification factors and maternal education level were included as covariates. All children with valid test scores were included in the analysis. An intention-to-treat approach was used in the primary analyses, followed by analyses that excluded 20 children (13 WWSC, 7 eAT) who did not receive their assigned treatment (ie, crossovers).
To examine the relationship of changes in the sleep measures to changes in cognitive tests across follow-up for the eAT group, we estimated gains in scores related to practice effects (ie, greater familiarity of children with the tests at follow-up) by using data from the WWSC group. For each test, follow-up scores for these children were regressed on their corresponding baseline scores. The regression equations were then applied to the eAT group to estimate expected follow-up scores. Cognitive change was defined as the standardized difference between the expected and observed scores at follow-up, reflecting the degree to which the follow-up scores differed from those predicted by the baseline scores and practice effects. Subsequent regression models examined changes in the sleep measures as predictors of these change scores, controlling for stratification factors and maternal education. All polysomnography measures except percentage sleep time in rapid eye movement sleep were log transformed to provide more normal distributions. Regression analysis controlling for these same factors was also used to examine associations of baseline neuropsychological test scores for the total sample with baseline sleep measures.
CHAT was designed to detect an effect size of ≥0.32 with 90% power for group differences in the primary outcome of attention and executive function.28 For the exploratory analyses presented here, corrections were not made for multiple comparisons. We computed effect sizes by using Cohen’s d for group differences from mixed models and f2 for regressions, defining small, medium, and large effects, respectively, as 0.2, 0.5, and 0.8 for d and 0.02, 0.15, and 0.35 for f2.42 We analyzed data by using SAS Proprietary Software 9.3 (TS1M0; SAS Institute, Inc, Cary, NC) and IBM SPSS Statistics Version 23 (IBM SPSS Statistics, IBM Corporation).
Table 2 presents group demographic and sleep characteristics and Table 3 test scores on the neuropsychological battery at baseline and follow-up. Although mean scores at baseline were within the average range relative to normative standards, means for 2 NEPSY 2nd edition (NEPSY-II) Inhibition conditions (Inhibition and Switching) were somewhat reduced relative to other scores (scaled scores = 8, 25th percentile). The WWSC and eAT groups differed significantly in only 1 of the tests at baseline.
Neuropsychological assessments were available at the 7-month follow-up for 203 (89.4%) children in the WWSC group and 196 (86.7%) in the eAT group. Slight differences in this sample compared with that examined in the original study28 reflect our inclusion of 2 children with partial test data who were excluded from that study because of missing data for the primary outcome. Compared with the children who completed the study, those without follow-up data included proportionally more black than white participants (38 [15%] vs 16 [8%], P < .05), had lower sleep efficiency, had lower scores on NEPSY-II Inhibition Switching and NEPSY Arrows, and had higher scores on Purdue Pegboard Both Hands (Ps < .05), but none of these differences varied by group.
Group Differences in Change in Test Scores From Baseline to 7-Month Follow-Up
Results from the intention-to-treat analysis are presented in Table 4. Analysis revealed significant group × visit interactions for Differential Abilities Scales, 2nd edition (DAS-II) Sequential and Quantitative Reasoning and Purdue Pegboard Both Hands. Increases in both scores were larger for the eAT group than for the WWSC group, but effect sizes were small (d = 0.20 for both measures). Figure 1 depicts group differences in change on these 2 tests. When crossovers were excluded, group differences with small effect sizes were found for change on Purdue Pegboard Both Hands, unstandardized β (SE) = 0.21 (0.08), P = .013, d = 0.23, and on NEPSY Visual Attention, β (SE) = 0.65 (0.31), P = .040, d = 0.24. Additional exploratory analyses failed to reveal evidence that group differences in change varied in relation to weight status, age, or race, although children who were overweight had significantly lower scores than those not overweight on several measures (data not shown). Practice effects were suggested by significant increases in multiple scores across follow-up for both groups.
Associations of Changes in Test Scores With Changes in Sleep Measures for Children in the eAT Group
Regression analysis revealed several associations of improved scores with positive changes in sleep parameters as measured by polysomnography and sleep questionnaires (Table 5). The associations were weak (partial rs −0.15 to −0.30) and had small effect sizes (f20.022–0.088). The associations tended to cluster around select tests and were evident on 2 of the 3 tests on which the eAT group made greater gains across follow-up than the WWSC group (Purdue Pegboard Non-dominant or Both Hands, NEPSY Visual Attention). Similar associations were found for DAS-II Pattern Construction, NEPSY Auditory Attention and Response Set, NEPSY-II Inhibition Naming Condition, NEPSY-II Word Generation Semantic Condition, and Wide Range Assessment of Memory and Learning, 2nd edition (WRAML2) Verbal Learning. Contrary to expectations, improved scores on Purdue Pegboard Non-dominant Hand were associated with increases in the arousal index, and improved scores on WRAML2 Verbal Learning Recognition were associated with decreased sleep efficiency. Findings were similar when we excluded crossovers.
Associations of Test Scores With Sleep Measures at Baseline
Regressions of baseline test scores on sleep measures for the total sample revealed 3 significant associations. Lower scores on WRAML2 Verbal Learning, DAS-II Word Definitions, and NEPSY-II Word Generation Initial Letter Condition were associated, respectively, with more sleep problems on the Pediatric Sleep Questionnaire Sleep Related Breathing Disorder Scale (PSQ-SRBD), greater sleepiness on the Epworth Sleepiness Scale modified for children (mESS), and higher percentage sleep time in stage 1 sleep. These associations were also weak (partial rs −0.15 to −0.17) with small effect sizes (f2 0.021–0.025).
The current study adds to the previous findings by suggesting small effects of AT on selective cognitive tests. Specifically, children randomly assigned to eAT made more gains than those in the WWSC group on tests of nonverbal reasoning and fine motor skills. In secondary analysis that excluded crossovers, the eAT group also made significantly greater gains on a timed measure of selective attention and visual scanning (NEPSY Visual Attention). Improvements in similar cognitive domains (fine motor coordination, nonverbal reasoning, and attention and impulse regulation) were associated with positive changes in sleep after eAT. The pattern of associations is in line with previous research suggesting that both respiratory disturbances and sleep quality contribute to cognitive functioning in OSAS.6
Cognitive weaknesses in children with OSAS are often reported in the domains of attention, executive function, and nonverbal reasoning.6,7,14,16–18 Weaknesses in motor dexterity have also been reported in children with snoring or OSAS and adults with OSAS.24,43,44 The present results offer support for small effects of treatment in these same domains. The effects of sleep disturbance on cognition and behavior have been attributed to sleepiness and to adverse effects of intermittent hypoxia and sleep fragmentation on neural development and brain functioning.6 Little is known about the effects of these processes on brain development, but frontal, subcortical, hippocampal, and cerebellar regions are especially vulnerable.7,11,17,44
Nonrandomized clinical trials of AT in children with OSAS or snoring have documented improved test performance after surgery.8,14,19,22–26 Several of these studies found greater gains on tests of attention and executive function, visual–motor and spatial skills, nonverbal reasoning, or memory in children receiving AT for OSAS compared with controls, although others have failed to document these effects.13,16 Post-AT associations between increased attention or nonverbal reasoning scores and improvements in sleep have also been reported.8,20 However, these studies are limited by their nonrandomized design, which could lead to an overestimation of effects. The current study suggests that cognitive benefits of AT over a 7-month period in children with OSAS without significant hypoxemia are probably small and selective. It is unclear whether such minor effects led to improvements in school performance or other aspects of daily functioning, but some children may have benefited more than others.
Our study failed to find any benefit of AT on tests of language, visual perceptual skills, or global cognitive ability.6 These negative findings and mean baseline scores that were comparable to normative means for age are in keeping with past evidence for average global cognitive abilities in children with OSAS.7,12,13 A relative weakness at baseline on NEPSY-II Inhibition is consistent with the vulnerability of children with OSAS to deficits in specific aspects of cognitive ability.6 However, CHAT was not designed to evaluate the effects of OSAS, and any differences between test means of CHAT participants and normative values may reflect differences in background characteristics between the participants and samples used to establish national standards.
The small effects of AT on cognitive test scores contrast with the more pronounced effects of surgery on child behavior and quality of life.27,28 One explanation for these small effects is that sleep-related cognitive weaknesses may be less evident on highly structured tests than under “free-living” conditions in which children have to regulate their own behavior according to environmental demands.45,46 Other possibilities are that the effects of chronic sleep disturbances on brain function are more difficult to reverse than responses to environmental conditions or that longer follow-up is needed to detect more substantial effects of AT on test performance.11,13,46 The tests used in this study may also be suboptimal for detecting effects of AT; measures placing greater demands on sustained attention and novel problem solving may have been more sensitive to the effects of AT.11,14,22,26,43,47 Although OSAS measures in the study were those routinely used in clinical settings and scored using rigorous approaches, alternative measures of OSAS or sleep disruption may also provide more sensitive indices of the effects of AT on sleep.7,8,16,17,20,48
A secondary aim was to explore associations of baseline test scores with baseline measures of sleep disturbance. Although several past studies failed to identify such associations in samples of children with OSAS or snoring and their controls,6,8,14,16,21,25,49 other studies report associations of a variety of indices of sleep disturbance with scores on tests of IQ, nonverbal reasoning, vigilance, executive function, and memory.13,15,17,45,50,51 In agreement with these findings, more symptoms of sleep disruption, greater sleepiness, and a greater percentage of stage 1 sleep were each associated with lower scores on 1 of the cognitive tests. Although the results must be interpreted with caution in view of small effect sizes, they accord with other reports of associations between better sleep and higher neurocognitive functioning.46
The design of CHAT conferred several methodological advantages for examining neuropsychological effects of AT and associations of test scores with sleep measures.3 OSAS was confirmed by standardized polysomnography to ensure uniformity of participant selection and quantification of sleep parameters. Because group assignment was random, potential biases in assessing neuropsychological consequences of AT were minimized. Assessing test score change across follow-up in WWSC group provided an opportunity to take effects of repeat testing into account in assessing the relationship of cognitive changes in the eAT group to changes in the sleep measures. Finally, recruitment from multiple centers yielded a large and diverse sample, and cognitive assessments were comprehensive and administered by examiners naive to group assignment.
This study has several limitations. Effect sizes were small. Moreover, we did not correct for the multiple comparisons, which accords with our exploratory approach52 but increases the risk of type I error. Findings indicating positive effects of AT on cognition thus require confirmation. Additionally, 2 unanticipated associations of increases in the eAT group’s scores across follow-up with negative changes in sleep are difficult to interpret. However, the majority of associations of changes in scores across follow-up with changes in sleep were in the expected direction and were evident for 2 of the 3 cognitive measures in which the eAT group improved more than the WWSC group. Another limitation is that the sample was restricted to children ≥5 years of age with OSAS without prolonged desaturation who were otherwise healthy.
Additional research is needed to investigate the effects of AT on academic learning and determine whether test performance is more affected for some subsets of children than for others. Study of the cognitive effects of AT in children <5 years of age and in those with more severe desaturation or comorbid conditions is likewise warranted. Another important research goal is to identify the types of cognitive skills most affected by AT. The findings suggest that tests of novel problem solving, attention, and motor dexterity are worthy of consideration in future trials. However, future studies might examine ways to increase test sensitivity by assessing speed of decision-making, lengthening tasks, or imposing greater demands on inhibitory control.
The findings suggest that, on average, AT confers small positive effects on cognitive test scores in children with OSAS without prolonged desaturation and with overall average cognitive functioning. The results provide impetus for more research on the cognitive and neurobiological effects of AT for pediatric OSAS.5,10,29,44 The findings are also consistent with previous research suggesting that tests of nonverbal reasoning, attention, and fine motor skills are selectively affected by OSAS and thus more likely to improve after AT.
The authors thank their collaborators on the CHAT. Appreciation is also extended to participating families and the members of the data and safety monitoring board. We also acknowledge the assistance of Nori Minich and CHAT research staff.
- Accepted May 24, 2016.
- Address correspondence to H. Gerry Taylor, PhD, Rainbow Child Development Center, W.O. Walker Bldg, Suite 3150, 10524 Euclid Ave, Cleveland, OH 44106. E-mail:
FINANCIAL DISCLOSURE: Dr Chervin is named in or has developed patented and copyrighted materials owned by the University of Michigan and designed to assist with assessment or treatment of sleep disorders; these materials include the Pediatric Sleep Questionnaire Sleep-Related Breathing Disorder scale, used in the research reported here. This questionnaire is licensed online by the University of Michigan to appropriate users at no charge and (for electronic use) to Zansors. The other authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Funded by grants HL083075, HL083129, UL1RR024134, UL1TR000003, and UL1RR024989 from the National Institutes of Health (NIH). Funded by the National Institutes of Health (NIH).
POTENTIAL CONFLICT OF INTEREST: Dr Carol Rosen has consulted for Natus, Advance–Medical and is a consultant for Jazz Pharmaceuticals. Relevant to this work, Dr Chervin is named in or has developed patented and copyrighted materials owned by the University of Michigan and designed to assist with assessment or treatment of sleep disorders. These materials include the Pediatric Sleep Questionnaire Sleep-Related Breathing Related Disorder scale, used in the research reported here. Dr Chervin serves on the boards of the American Academy of Sleep Medicine and the International Pediatric Sleep Society, is an editor for UpToDate, has edited a book for Cambridge University Press, has received support for research and education from Philips Respironics and Fisher Paykel, and has consulted for MC3 and Zansors. The other authors have indicated they have no potential conflicts of interests to disclose.
- Garetz SL
- Marcus CL,
- Brooks LJ,
- Draper KA, et al; American Academy of Pediatrics
- Montgomery-Downs HE,
- Crabtree VM,
- Gozal D
- Melendres MC,
- Lutz JM,
- Rubin ED,
- Marcus CL
- Elliott CD
- Korkman N,
- Kirk U,
- Kemp S
- Korkman M,
- Kirk U,
- Kemp S
- Gardner R
- Baron IS
- Beery KE,
- Buktenica NA,
- Beery NA
- Sheslow D,
- Adams W
- Avior G,
- Fishman G,
- Leor A,
- Sivan Y,
- Kaysar N,
- Derowe A
- O’Brien LM,
- Holbrook CR,
- Mervis CB, et al
- Copyright © 2016 by the American Academy of Pediatrics