Technical Report: Diagnosis and Management of Childhood Obstructive Sleep Apnea Syndrome
- Michael S. Schechter, MD, MPH,
- Section on Pediatric Pulmonology, Subcommittee on Obstructive Sleep Apnea Syndrome
Objective. This technical report describes the procedures involved in developing the recommendations of the Subcommittee on Obstructive Sleep Apnea Syndrome in children. The group of primary interest for this report was otherwise healthy children older than 1 year who might have adenotonsillar hypertrophy or obesity as underlying risk factors of obstructive sleep apnea syndrome (OSAS). The goals of the committee were to enhance the primary care clinician’s ability to recognize OSAS, identify the most appropriate procedure for diagnosis of OSAS, identify risks associated with pediatric OSAS, and evaluate management options for OSAS.
Methods. A literature search was initially conducted for 2the years 1966–1999 and then updated to include 2000. The search was limited to English language literature concerning children older than 2 and younger than 18 years. Titles and abstracts were reviewed for relevance, and committee members reviewed in detail any possibly appropriate articles to determine eligibility for inclusion. Additional articles were obtained by a review of literature and committee members’ files. Committee members compiled evidence tables and met to review and discuss the literature that was collected.
Results. A total of 2115 titles were reviewed, of which 113 provided relevant original data for analysis. These articles were mainly case series and cross-sectional studies; overall, very few methodologically strong cohort studies or randomized, controlled trials concerning OSAS have been published. In addition, a minority of studies satisfactorily differentiated primary snoring from true OSAS. Reports of the prevalence of habitual snoring in children ranged from 3.2% to 12.1%, and estimates of OSAS ranged from 0.7% to 10.3%; these studies were too heterogeneous for data pooling. Children with sleep-disordered breathing are at increased risk for hyperactivity and learning problems. The combined odds ratio for neurobehavioral abnormalities in snoring children compared with controls is 2.93 (95% confidence interval: 2.23–3.83). A number of case series have documented decreased somatic growth in children with OSAS; right ventricular dysfunction and systemic hypertension also have been reported in children with OSAS. However, the risk growth and cardiovascular problems cannot be quantified from the published literature. Overnight polysomnography (PSG) is recognized as the gold standard for diagnosis of OSAS, and there are currently no satisfactory alternatives. The diagnostic accuracy of symptom questionnaires and other purely clinical approaches is low. Pulse oximetry appears to be specific but insensitive. Other methods, including audiotaping or videotaping and nap or home overnight PSG, remain investigational. Adenotonsillectomy is curative in 75% to 100% of children with OSAS, including those who are obese. Up to 27% of children undergoing adenotonsillectomy for OSAS have postoperative respiratory complications, but estimates are varied. Risk factors for persistent OSAS after adenotonsillectomy include continued snoring and a high apnea-hypopnea index on the preoperative PSG.
Conclusions. OSAS is common in children and is associated with significant sequelae. Overnight PSG is currently the only reliable diagnostic modality that can differentiate OSAS from primary snoring. However, the PSG criteria for OSAS have not been definitively validated, and it is not clear that primary snoring without PSG-defined OSAS is benign. Adenotonsillectomy is the first-line treatment for OSAS but requires careful postoperative monitoring because of the high risk of respiratory complications. Adenotonsillectomy is usually curative, but children with persistent snoring (and perhaps with severely abnormal preoperative PSG results) should have PSG repeated postoperatively.
- sleep apnea
- sleep disorders
This technical report describes in detail the procedures involved in developing recommendations as given in the accompanying practice guideline on obstructive sleep apnea syndrome (OSAS).1 A description of the process, methods of data compilation and analysis, and summaries of the conclusions of the committee will be given.
FORMULATION AND ARTICULATION OF THE QUESTION ADDRESSED BY THE COMMITTEE
The practice guideline is primarily aimed at office-based pediatricians and other primary care clinicians who treat children (family physicians, nurse practitioners, physician assistants). The secondary audience for the guideline includes pediatric pulmonologists, neurologists, otolaryngologists, and developmental/behavioral pediatricians.
The primary focus of the committee was on OSAS in childhood. The committee agreed to use the definition provided in a statement from the American Thoracic Society2 with some additional elaboration of associated symptoms:
OSAS in children is a disorder of breathing during sleep characterized by prolonged partial upper airway obstruction and/or intermittent complete obstruction (obstructive apnea) that disrupts normal ventilation during sleep and normal sleep patterns. It is associated with symptoms including habitual (nightly) snoring, sleep difficulties, and/or daytime neurobehavioral problems. Complications may include growth abnormalities, neurologic disorders, and cor pulmonale, especially in severe cases. Various risk factors have been identified and are defined.
The committee sought to focus on otherwise healthy children who might have adenotonsillar hypertrophy or obesity as underlying risk factors and to specifically exclude infants younger than 1 year, children with central hypoventilation syndromes, and children at risk because of underlying abnormalities, such as craniofacial disorders; Down syndrome; cerebral palsy; neuromuscular disorders; chronic lung disease; sickle cell disease; genetic, metabolic, and storage diseases; and laryngomalacia.
Goals of the Committee
The committee sought to address several specific goals and questions:
To enhance the primary care clinician’s ability to recognize OSAS. The committee believed that a certain amount of consciousness raising would be appropriate to alert the clinician to suspect the presence of OSAS. Thus, a catalog of associated signs and symptoms is provided in the accompanying practice guideline1 but will not be addressed in this technical report
To identify the most appropriate procedure for diagnosis of OSAS. Approaches evaluated by the committee included history and physical examination, questionnaires, audiotaping or videotaping, nocturnal pulse oximetry, nap polysomnography (PSG), and ambulatory PSG, all of which would be compared with the gold standard, comprehensive overnight PSG, as defined by the American Thoracic Society.2,3 In view of the fact that overnight PSG is not readily available to children in all geographic areas, consideration was given to alternative diagnostic approaches even if their accuracy is suboptimal
To identify risks associated with pediatric OSAS. In adults, OSAS is associated with excessive daytime sleepiness (leading to cognitive defects and increased mortality attributable to susceptibility to motor vehicle crashes), pulmonary hypertension, and systemic hypertension. The committee wished to evaluate the strength of pediatric data in this area
To evaluate management options for OSAS. The committee sought to find data relating to adenotonsillectomy and alternative treatment modalities in the management of OSAS. The committee set out to evaluate the risk of complications after adenotonsillectomy in children with OSAS, especially given the fact that patients not suspected to have OSAS might undergo adenotonsillectomy for other indications. The postoperative complication of particular concern was respiratory compromise. Other complications of surgery, such as bleeding and pain, were not specifically addressed in relation to OSAS. In addition, data on postoperative recurrence and persistence of OSAS was sought
A literature search of the National Library of Medicine’s PubMed database (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed) for the years 1966–1999 was conducted in August 1999 by staff at the American Academy of Pediatrics. The search was limited to English language literature concerning children older than 2 and younger than 18 years. The following search terms were used: sleep apnea syndrome; apnea; sleep disorders; snoring; polysomnography; airway obstruction; adenoidectomy; tonsillectomy (adverse effects, mortality); and sleep-disordered breathing.mp. This search was updated in November 2000 before preparation of this technical report.
Titles and abstracts (when available) of articles found by the literature search were reviewed by committee members, and all those considered to be possibly relevant were marked for detailed review. Recent review articles were included to compare their bibliographies with the result of the automated literature search. Articles deemed possibly relevant were then printed and distributed to committee members for more detailed review. A literature review form was developed for this project to standardize this part of the process (Appendix 1). Because there was a large number of articles requiring evaluation, some committee members recruited residents and fellows to assist in the performance of these reviews under their supervision. Although it became clear at this point that the number of articles that could be considered high quality by conventional epidemiologic standards4–6 was small, a low threshold was used to allow inclusion of any possibly relevant articles into the next level of review. At this point, articles were compiled and divided by the committee chair, additional articles were obtained by a review of literature, committee members’ files were added, and committee members were assigned specific topics (as discussed previously in “Goals of the Committee”) for detailed review and compilation of evidence tables. The findings of committee members were then presented at a follow-up meeting of the entire committee. A final review and compilation into evidence tables was performed by the lead author of this technical report (M.S.S.).
Calculation of prevalence, diagnostic test characteristics, and odds ratios were performed independently of the authors’ reports, using data provided in the original articles. In 2 cases, authors were contacted for clarification of data. Where applicable, odds ratios from different studies were combined, using Mantel-Haenszel weights in stratified tables. Tests for heterogeneity are reported. All statistical calculations were performed using Stata 5.0 software (Stata Corporation, College Station, TX).
The literature search identified 2067 articles for initial review. Titles and abstracts (when available) of these articles were divided among the 7 committee members for perusal as an initial screening, and of those, 278 (16.2%) were retained for more detailed scrutiny. An additional 48 relevant publications were found outside this initial review. Articles were read in full if they appeared to have any relevance to childhood OSAS; included in this group were 70 general reviews and case reports or descriptive case series, which were used to allow committee members to gain a general sense of the literature and access their bibliographies. Among various committee members, the percentage of articles chosen for more detailed review ranged from 12.7% to 23.4%. This variability was statistically significant (P = .007 by Pearson χ2). Excluding the methodologist, whose approach was the most permissive (23.4% acceptance rate), the range was 12.7% to 19.4%, and variability was not statistically significant (P = .143 by Pearson χ2) among remaining committee members.
A total of 113 articles were found that contained original data relevant to the specific aims of this committee. Most of the publications that were reviewed in detail provided little quantitative data for analysis. In particular, most papers that were older than 5 years presented case series or poor-quality cohort studies or omitted important details. These lower-quality studies provide the background for current expert opinion but are otherwise of limited value and will not be listed in detail for this report. Studies that were of quantitative value are tabulated and were given quality ratings4,6 in the tables. Briefly, rating levels of studies on treatment efficacy were assigned as follows4:
Level I—Randomized trials with low rates of false-positive (α) and/or false-negative (β) results (high power).
Level II—Randomized trials with high rates of false-positive (α) and/or false-negative (β) results (low power).
Level III—Nonrandomized concurrent cohort comparisons between contemporaneous patients who did and did not receive an intervention, or case-control or cross-sectional studies with appropriate control group.
Level IV—Nonrandomized historical cohort comparisons between current patients who received an intervention and former patients (from the same institution or from the literature) who did not, or case-control or cross-sectional studies for which control groups were suboptimally chosen.
Level V—Case series without controls.
Rating levels for diagnostic tests were assigned as follows6:
Level 1—Independent blind comparison of patients from an appropriate spectrum of patients, all of whom have undergone both the diagnostic test and the reference standard.
Level 2—Independent blind or objective comparison performed in a set of nonconsecutive patients or confined to a narrow spectrum of study individuals (or both), all of whom have undergone both the diagnostic test and the reference standard.
Level 3—Independent blind or objective comparison of an appropriate spectrum of patients, but the reference standard was not applied to all.
Level 4—Reference standard was unobjective, unblinded, or not independent; positive and negative tests were verified using separate reference standards; or study was performed in an inappropriate spectrum of patients.
Prevalence of Snoring and OSAS
We found 7 studies that attempted to establish prevalence of snoring in childhood.7–13 These studies came from a variety of European countries, and all ascertained data via parent questionnaire. The prevalence of snoring in these studies ranged from 3.2% to 12.1%, which was significantly heterogeneous (P < .0001). The study by Gislason and Benediktsdottir8 seemed to be somewhat of an outlier, especially because the frequency of OSAS they reported was nearly the same as that of snoring, but with omission of that study, the heterogeneity remains significant (P < .001). Characteristics of the studies are shown in Table 1, and prevalence estimates are seen in Fig 1.
Three studies reported on prevalence of OSAS, and estimates ranged from 0.7% to 10.3%. All used very different criteria, including 1 that gave estimates based on 2 different criteria.14 The variability of these estimates was so great that no attempt was made to combine the data, which are summarized in Table 1.
Sequelae of OSAS
Most published articles on complications of OSAS are reports of retrospective case series or prospectively collected, uncontrolled data comparing measures before and after surgical treatment. These articles are summarized in this report and in evidence tables (Tables 2–4).
Cognitive and Behavioral Abnormalities
The committee found 12 publications that evaluated the association of behavioral problems, especially hyperactivity or attention-deficit/hyperactivity disorder (ADHD), with sleep-disordered breathing (Table 2). In an early case series15 of 50 children with OSAS documented by PSG, 84% had excessive daytime sleepiness, 76% had some behavior disturbance, 42% were hyperactive, and 16% had decreased school performance. A number of cross-sectional studies have been done that compare the risk of behavioral problems in children who snore with that of a control population. None of these studies distinguish children with OSAS from those with primary snoring (PS). For the study of Weissbluth et al,16 parents of children attending a general pediatric practice were surveyed, and 71 children were reported to have behavioral or academic problems. Snoring, mouth breathing, and labored breathing when asleep were reported to be more than twice as common in these children as in the comparison group. The study of Chervin et al17 reported that 33% of children with ADHD were habitual snorers, compared with 11% of children attending a general psychiatric clinic and 9% attending a general pediatric clinic. A previously mentioned prevalence study by Ali et al7 found that children who were reported to snore during most nights were also reported to have more daytime sleepiness and hyperactivity than were children in a comparison group. Children were evaluated using Conners scales, and those with more severe sleep disturbance were also more likely to be at the 95th percentile on Conners subscales relating to hyperactive, inattentive, and aggressive behaviors. In a follow-up report,18 29 of 60 children reported to snore 2 years previously no longer snored (weighted κ 0.52), although habitual snoring was again found to be associated with daytime sleepiness and hyperactivity. In another publication, the same authors identified 12 children with sleep-disordered breathing (by overnight pulse oximetry and videotaping), 11 snoring children without sleep-disordered breathing, and a control group not undergoing adenotonsillectomy and administered Conners scales, the Continuous Performance Test (a test of attention), and the Matching Familiar Figures Test (a test of vigilance).19 Significant improvement was found in Conners parent scale scores for aggressive, inattentive, and hyperactive behaviors; attention; and vigilance in the sleep-disordered breathing group after adenotonsillectomy, and similar changes were found in the primary snorers postoperatively, but not in control groups. In a recent population-based cross-sectional study of 988 Portuguese children, Ferreira et al13 found that habitual snorers were twice as likely as nonsnorers to have an abnormal score on the Children’s Behavioral Questionnaire. Finally, Blunden et al20 compared 16 children referred for adenotonsillectomy or snoring with a control group and found impaired selective and sustained attention scores only in the children who snored. They also reported that the snoring children had significantly lower average IQ scores. PSG was done in these children, but for purposes of analysis, the PS and OSAS groups were combined because preliminary analyses revealed no significant group differences on any neurocognitive or behavioral parameter (all P > .05). The authors pointed out that even patients with OSAS had very mild abnormalities (mean respiratory disturbance index [RDI] <1). Although it is possible that lack of power and selection bias may have contributed to their findings, the implication is that snoring without overt OSAS might be associated with neurocognitive abnormalities.
Although the 6 cross-sectional studies described previously all reported on slightly different behavioral and cognitive phenomena, their findings were pooled, and the results of this are shown in Fig 2. The Mantel-Haenszel test for heterogeneity was not significant (P = .4577). The combined odds ratio for neurobehavioral abnormalities in snoring children is 2.93 (95% confidence interval [CI], 2.23–3.83).
Several other papers are of interest but cannot be quantitatively combined. Goldstein et al21 had parents of 36 children who were referred for adenotonsillectomy because of clinically significant obstructive symptoms complete the Child Behavior Checklist and found that 10 (28%) had abnormal results. Postoperatively, 15 had a repeat evaluation, and only 2 (13%) still had abnormal results. This finding was not clinically significant, presumably because of lack of statistical power, but there was a clinically significant improvement in mean test scores postoperatively (P < .001).
Four studies used overnight PSG testing to establish a diagnosis of OSAS. Rosen22 reported on a series of 326 children referred for evaluation of snoring, of whom 59% met PSG criteria for OSAS. Daily tiredness was reported in 19%; excessive daytime sleepiness was reported in 10%; and behavior, school, or mood problems were reported in 9%, with no difference between the OSAS and non-OSAS groups. Gozal23 conducted an interesting study in which 297 first graders who were in the lowest 10th percentile academically were evaluated for sleep-disordered breathing by parent questionnaire combined with overnight (home) oximetry. Adenotonsillectomy was recommended in the 54 children (18.1%) with abnormal test results; 24 accepted surgery, and 30 did not. Mean grades increased from 2.43 ± 0.17 to 2.87 ± 0.19 in the children who had surgery, with no change in the untreated OSAS group or the non-OSAS group (P < .001).
The findings of an Australian study24 contrasted with the aforementioned papers. Thirty-nine children with PSG evidence of OSAS were followed 6 months after the initial PSG, and 24 had received adenotonsillectomy. These were compared with children who were waiting for intervention (n = 5; median apnea-hypopnea index [AHI] = 5.5) or didn’t require intervention (n = 10; median AHI = 3.1). Information on AHI was not given for the surgical group. At follow-up, children in the surgical and nonsurgical groups had improved sleep behavior. Intervention did not result in any statistically significant improvement in development or temperament, although the study was probably underpowered.
In summary, studies generally show a nearly threefold increase in behavior and neurocognitive abnormalities in children with sleep-disordered breathing. Most of these studies did not definitively differentiate children with PS from those with OSAS, so the true prevalence of behavior and learning problems in children with OSAS versus PS is not clear. It is possible, however, that PS, even in the absence of clear-cut OSAS, might place children at risk.
Four studies evaluating growth and OSAS were found. Marcus et al27 evaluated 14 prepubertal children with a mean age of 4 years ± 1 standard deviation who had OSAS documented by overnight PSG and measured caloric intake and sleeping energy expenditure as well as anthropomorphic measurements before and after adenotonsillectomy. Average sleeping energy expenditure decreased, and mean weight z score increased postoperatively without any change in caloric intake. Bar et al28 evaluated changes in growth and also measured insulin-like growth factor (IGF)-I and IGF-binding protein-3 levels before and 18 months after adenotonsillectomy. Both studies showed statistically significant increases in weight but not height; IGF-I levels increased and IGF-binding protein levels did not. An interesting report on the effect of adenotonsillectomy on growth included a group of obese and morbidly obese children29 and documented postoperative increases in weight and height, even in those children who were initially obese. The other 2 studies,25,26 which had poorer documentation of OSAS, reported similar results. None of these studies reported a comparison with a nonsurgical control group, comparison with children operated on for indications other than OSAS, or comparison with children who had PS.
Eight case reports or small series were found that documented cor pulmonale or hypertension, which reversed with adenotonsillectomy or other surgical correction in patients with clinically diagnosed OSAS.35–42 Two case series30,31 reported children with adenotonsillar hypertrophy with or without clinical airway obstruction who were found to have right ventricular dysfunction that reversed after adenotonsillectomy. One study32 described pulsus paradoxus and leftward shift of the interventricular septum secondary to snoring in 3 of 6 children with OSAS. This correlated with negative esophageal pressures but not with oxygen desaturation, and it reversed with nasal continuous positive airway pressure (CPAP). Tal et al33 used radionuclide ventriculography to evaluate ventricular function in 27 children referred for oropharyngeal obstruction who had abnormal Brouillette questionnaire scores for OSAS; PSG was not performed. They found decreased right ventricular ejection fraction in 37% of these children and abnormal wall motion in 67%. All of the 11 patients who had a repeat evaluation after adenotonsillectomy showed improvement. Systemic blood pressure was evaluated in a study of children referred for PSG.34 Higher diastolic pressures (adjusted for body mass index and age) were found in children with OSAS, compared with those with PS. The prevalence of blood pressure measurements >95th percentile was high in both groups (32% vs 19%, respectively), with a nonsignificant difference that may have been attributable to low power. The response after adenotonsillectomy was not reported. This was the only study that compared cardiovascular complications in children with OSAS versus those with PS.
One study43 reported on 115 enuretic children undergoing adenotonsillectomy for any indication. There was a 66% reduction in enuretic nights 1 month after surgery and a 77% decrease 6 months after surgery. In the group with secondary enuresis, 100% were dry 6 months after surgery.
Diagnosis of OSAS
One of the problems in evaluating various methods of diagnosing OSAS in children is that the gold standard, overnight PSG, has not been well standardized in its performance or interpretation. Although recent consensus statements pertaining to standards and normative data2,3 should lessen this problem, the question of definition remains problematic. Pediatric sleep specialists use the adult model in describing a continuum of sleep-disordered breathing from PS to upper airway resistance syndrome to obstructive hypoventilation and OSAS. It is assumed that PS is a benign condition and OSAS is associated with undesirable complications. Normative standards for their polysomnographic determination have been chosen on the basis of statistical distribution of data,44,45 but it has not been established that those standards have any validity as predictors of the occurrence of complications. In other words:
“On the basis of normative data, an obstructive apnea index of 1 is often chosen as the cutoff for normality. However, while an apnea index of 1 is statistically significant (ie, at the 97.5th percentile for an asymptomatic, normative population), it is not known what level is clinically significant.”3
Of the few studies that compare children with polysomnographically defined OSAS with those with PS in regard to prevalence of complications,19,22,30,34 only 1 found a clear difference between the 2 groups.34 This is an important point, because with a poorly validated gold standard, statements regarding diagnostic accuracy of alternative methods of diagnosis become dubious.46 Finally, the test-retest reliability of overnight PSG, which in adults is no greater than 91% and possibly somewhat lower,47 has never been evaluated in children.
Having stated these points, additional analysis of the validity of alternative diagnostic approaches will be done assuming PSG as the gold standard. One additional benefit of overnight PSG is that in addition to establishing the diagnosis of OSAS, PSG also may be used to determine its severity. It has been suggested that the severity of OSAS is an important predictor of complications, particularly in the immediate postoperative period.48 None of the alternative diagnostic techniques discussed below have been evaluated for this purpose.
In 1984, Brouillette et al49 reported high accuracy for a diagnostic questionnaire for OSAS in children with adenotonsillar hypertrophy. This questionnaire was initially tested on 23 children with OSAS and 46 controls. On the basis of this questionnaire, a 3-variable discriminant function was calculated as follows:
where D is difficulty during sleep, A is apnea observed during sleep, and S is snoring. Values assigned to D and S were: 0 = never; 1 = occasionally; 2 = frequently; and 3 = always. Values assigned to A were: 0 = no; and 1 = yes. This system was then applied to a prospective group of 23 patients referred for evaluation of possible OSAS. The authors demonstrated that a score of >3.5 perfectly predicted the presence of OSAS by PSG; a score of <−1 perfectly predicted absence of OSAS; and a score in between was indeterminate. Unfortunately, there were 5 children who were believed to have a borderline PSG, confusing the issue somewhat. It appears that the choices of 3.5 and −1 as breakpoints in the score were made posthoc and were, thus, somewhat arbitrary. Since the initial publication of the questionnaire, 3 additional studies have been published detailing the results of its use. The results of these are provided in Table 5. All of these studies prospectively evaluated similar groups of pediatric patients with a similar prevalence of OSAS; PSG with similar evaluation criteria was performed on all subjects, and all completed the same questionnaire applied in similar ways. This scoring system is sufficiently simple and straightforward, so its application can be expected to be fairly standard and replicable. Thus, data from these studies was combined, and conclusions were drawn accordingly.
As can be seen from Table 5, the OSAS questionnaire by Brouillette et al performed much less well in subsequent applications. The 4 studies (including a later study by the same authors) included a total of 765 patients with an overall prevalence of OSAS confirmed by PSG of 60%. Applied to these patients, the score was indeterminate in 47%; in subjects who were categorized (ie, not indeterminate), its positive predictive value (PPV) was 65% and negative predictive value (NPV) was 46%. Using the pooled data for calculation, the likelihood ratio of positive questionnaire results is 1.24, and the likelihood ratio of negative questionnaire results is 0.78. Overall, the use of the questionnaire by Brouillette et al as a substitute for PSG would clearly be fraught with error, leading to numerous false-positive and false-negative results in the diagnosis of OSAS.
Other publications reporting attempts at creating questionnaires or developing other purely clinical criteria to substitute for PSG are uninterpretable because of their failure to compare their criteria with PSG50–53 or unsuccessful in developing any reliable predictive criteria.48,54–57 They are tabulated in Table 5.
Audiotaping and Videotaping
Two studies have evaluated the use of home audiotaping, and 1 evaluated home videotaping, as a screening test for OSAS. The methods used to evaluate these techniques were different, so the data do not lend themselves to pooling. Sivan et al60 scored a 30-minute videotape in 58 children using 7 variables, including loudness and type of inspiratory noise, movements during sleep, number of waking episodes, number of apneas, chest retractions, and mouth breathing. The PSG results were abnormal in 62%. They reported a sensitivity of 94%, specificity of 68%, PPV of 83%, and NPV of 88%. Posthoc analysis (similar to what was done in the 1984 study by Brouillette et al) was performed in 2 ways, leading to the development of an indeterminate score and better test characteristics in the categorizable group. As might be expected, a scoring system that places a greater number of subjects into the indeterminate group leads to better NPV and PPV. The results of all 3 approaches are shown in Table 6.
Goldstein et al61 developed a 7-item predictive score that considered the presence of snoring, respiratory pauses, gasping, sleeping with neck extended, daytime sleepiness, adenoid facies, and the presence of pauses in breathing of at least 5 seconds on an audiotape recorded by the parents. The criteria used to score each category were not precisely described, and there seemed to be significant variability in the evaluation of the audiotapes, which were reviewed for “at least 2 minutes… for each child, and an average of 10 minutes were generally reviewed. Various parts of the tape were sampled.” Patients were categorized as definitely, possibly, or not likely having OSAS on the basis of these items, but no description was provided of how these items were scored and combined, and no measure of interobserver variability was attempted, so reproducibility is unknown. A total of 30 children were studied prospectively, of whom 13 (43%) had OSAS confirmed by PSG. The authors reported a sensitivity of 92.3%, specificity of 29.4%, PPV of 50.0%, and NPV of 83.3%, which they calculated by combining the “definite” and “possible” groups into a positive screening category. If their possible group was eliminated from consideration (analogous to the way Brouillette et al treated indeterminate scores49), a mild decrease in sensitivity (91%) and mild increases in specificity (38%) and PPV (56%) are seen.
Goldstein et al concluded that children whose results of evaluation for sleep apnea (as performed using their technique, including audiotaping) are negative do not need PSG, because the sensitivity of their clinical assessment is high. They recommended PSG for children who appear to have OSAS, because the specificity of clinical assessment is low. It is important to note that the percentage of positive results of PSG in their study (43%) was somewhat lower than the prevalence of approximately 60% reported in most studies of children referred for evaluation of possible OSAS. This is probably (at least in part) because they used more restrictive PSG criteria for diagnosing OSAS (AHI >15). If a population with a higher prevalence of OSAS were studied, it is likely that the PPV of the clinical evaluation by Goldstein et al would be higher and the NPV would be lower.a In addition, the higher AHI as a diagnostic criterion might have biased the study toward the more severe end of the OSAS spectrum. The possibility of spectrum bias62 and the undocumented reproducibility63 of the tape evaluation raise the question of whether test characteristics will be as good if applied to a large, general population.
A second study of the use of home audiotaping as an abbreviated test for OSAS64 used 7 observers to analyze audiotapes of 29 children referred for evaluation; 48% were subsequently found to have positive PSG. Observers listened to 15 minutes of audiotape and specifically scored the presence of struggle sounds and respiratory pauses. A mean κ statistic of 0.70 (range, 0.50–0.93) was calculated, indicating moderately good interobserver agreement. The presence of a struggle sound on the audiotape gave the best posthoc test characteristics, with a sensitivity of 0.71, specificity of 80%, NPV of 73%, and PPV of 75%.
To summarize, the use of home audiotaping and videotaping has been inadequately investigated. Additional studies are necessary. It should be pointed out that there was no consensus of the committee regarding acceptable rates of false-negative and false-positive results for tests used as an alternative to PSG.
Seven studies were found that reported on pulse oximetry in children suspected of having OSAS.9,19,23,51,59,67,68 However, only 1 compared pulse oximetry to PSG. In this study59 involving 349 children, pulse oximetry was performed during PSG and was evaluated independently of the PSG interpretation, with well-defined criteria and excellent interobserver agreement. There were 89 PSGs (25.5%) performed in a sleep lab; the others were done at home, so the gold standard was not identical for all subjects. In this group, with a 60.2% prevalence of OSAS, the PPV was 97% (90 of 93). However, the NPV of the test (calculated by the authors by combining subjects with either inconclusive or negative tests) was only 53%. When the analysis was limited to subjects without any medical diagnoses other than adenotonsillar hypertrophy, the PPV was 100%, with an insubstantial change in NPV.
Given the test characteristic described, it appears that overnight pulse oximetry could provide an accurate screen for OSAS, insofar as a positive result may be a good predictor of an abnormal PSG result. However, the findings of the single study described in this report need to be replicated.
Two papers from the same institution have evaluated the utility of brief (1 hour) daytime nap studies in comparison with full overnight PSGs (Table 6). The conclusions of both are generalizable to only a limited degree, however. Marcus et al65 studied 40 children referred for evaluation of possible OSAS, but this group was not representative of the type of patient addressed by this practice guideline, because only 35% had adenotonsillar hypertrophy as the underlying cause of their sleep disturbances. Other diagnoses included Down syndrome (40% of subjects), various upper airway abnormalities, and other neurologic and respiratory problems. Furthermore, 95% (38 of 40) of the patients studied had abnormal overnight PSG results, providing little opportunity to evaluate the test performance of nap studies in children with normal PSG results. The study by Saeed et al66 limited itself to children addressed by this practice guideline (age, 1–18 years; adenotonsillar hypertrophy; absence of other significant disease). They reported on the results of overnight PSG in children with normal and mildly abnormal nap study results. Patients with severely abnormal nap study results were excluded; they were assumed to have significant OSAS and, therefore, referred directly for tonsillectomy without overnight PSG (S.D. Ward, personal communication). In this group, for which prevalence of PSG-documented OSAS was 66%, the nap studies had a PPV of 77% and NPV of 49%. In fact, if children with more severely abnormal nap study results were included in the analysis and the investigators are correct in their assumption that these children all have abnormal overnight PSG results, then the sensitivity and PPV of nap studies is actually higher than that reported in this paper. Thus, it is possible that abnormal nap study results might provide a predictive value adequate to allow the recommendation for surgery without corroborative overnight PSG, but confirmation of this conclusion (asserted by the authors) is lacking. On the other hand, a nap study with negative results would still require a follow-up overnight PSG for confirmation.
One group has published data comparing the results of PSG performed in children at home with those performed at the sleep laboratory.69,70 In a report of 21 children between the ages of 2 and 12 years who were studied in both environments, the sensitivity and specificity of home PSG varied depending on the severity of OSA. When an AHI >1 was used as the criterion for diagnosing OSAS, the sensitivity of home PSG was 100% and the specificity was 62%; for AHI >3, sensitivity was 88% and specificity was 77%; for AHI >5, sensitivity and specificity were both 100%. This group uses a sophisticated type of ambulatory PSG that is not commercially available and is not analogous to commercial systems. Also, their system did not allow for detection of obstructive hypoventilation. Furthermore, the subjects used in their report were not chosen sequentially or at random, and the authors describe a complex process for specifically selecting children for inclusion in the study. Nonetheless, the comparability of the results of home and sleep laboratory overnight PSG appears good; additional study using commercially available equipment in a more representative population would be helpful.
Treatment of OSAS
Tonsillectomy and/or Adenoidectomy
There are many published papers, primarily case reports and case series, that support the efficacy of tonsillectomy with or without adenoidectomy as treatment for OSAS. Most of these studies use relief of snoring and other clinical symptoms as their endpoint.56,68,71–77 Others cite improvement in growth,25–29 behavior,19 cardiovascular complications,31,33 or enuresis78 after surgery. Several papers suggest that adenotonsillectomy is effective treatment of OSAS even in children who are morbidly obese.29,79,80 Many of these studies are anecdotal and methodologically uninterpretable; those that used PSG to document OSAS are summarized below and in Table 7.
Frank et al81 were the first to use PSG to analyze the effect of surgery on OSAS in children. Of an initial group of 32 children referred for suspected OSAS, they reported on 7 who had PSG before and after adenotonsillectomy. These children had an average of 194 obstructive apneas per night preoperatively and 7 postoperatively (P < .025). They provide no breakdown of individual cure rate. Zucconi et al,82 using nocturnal or nap PSG, reported a 100% cure rate of OSAS in 29 children receiving adenotonsillectomy or adenoidectomy and monotonsillectomy and a 0% cure rate in 5 children receiving only adenoidectomy. Two more recent studies were methodologically superior regarding diagnosis of OSAS. Suen et al48 reported on 69 children referred for evaluation of possible OSAS; 35 (51%) had a RDI >5 and were referred for adenotonsillectomy, and 30 had the procedure. Follow-up PSG was performed in 26; all showed improvement, although 4 (15%) still had an RDI >5. All children with persistently high RDIs continued to snore, although 3 children with RDIs that had normalized continued to snore. Thus, adenotonsillectomy resulted in a cure rate of 85%, and the absence of postoperative snoring was associated with no treatment failures (NPV of postoperative snoring = 100%), whereas 57% of children who still snored continued to have abnormal PSG results (PPV = 57%). The authors of that paper emphasized that a high preoperative RDI was a strong predictor of abnormal postoperative RDI and suggested 19.1 as a cutoff. However, their data shows that the PPV of preoperative RDI ≥19.1 for a postoperative RDI >5 was 43% and the NPV was 95%, neither of which are as high as the predictive values afforded by the presence of persistent snoring postoperatively. The findings of Nieminen56 were similar, although their criteria for positive PSG results were slightly different (AHI >1). They reported a 95% cure rate for a group of 21 children after adenotonsillectomy or tonsillectomy; 1 of 5 children who continued to snore had postoperative PSG results that remained abnormal (PPV = 20%), and none of the children who stopped snoring had abnormal PSG results (NPV = 100%). The authors pointed out that 73% of this group had previously had their adenoids removed, implying confirmation of the lack of efficacy of adenoidectomy alone for relief of OSAS. This paper also mentioned in passing that 2 children with abnormal results of PSG did not have surgery; in 1, the follow-up PSG results were unchanged, and in the other, the results had normalized. Although no generalizations can be made on the basis of these data, it represents the only published report of follow-up PSG in children with OSAS who were not treated.
Several other papers reported PSG results in association with adenotonsillectomy, but these reports were somewhat less clearly written. Wiet et al80 reported a series of 48 patients in whom sleep studies were performed because of unclear history or physical findings, or complicated OSA. An AHI >5 was considered abnormal. Thirteen patients had no complicating medical factors, and of the 35 remaining, 20 were morbidly obese. All 13 uncomplicated patients had adenotonsillectomy. They had a significant decrease in mean AHI (from 23 to 6 [P < .01]); it was not stated whether any had residual abnormal postoperative PSG results. Of the obese patients, 12 of 20 had adenotonsillectomy alone, and the rest had uvulopharyngopalatoplasty in addition. It was not specified how the decision to perform uvulopharyngopalatoplasty was made, and the report of results for this group was not broken down by surgical procedure. Mean AHI in the obese group decreased from 33 to 4 (P < .001). Agren et al83 reported on a group of 20 children with “unequivocal anamnestic nocturnal obstructive breathing.” The preoperative AHI was >5 in 10 children, and the apnea index (AI) was >1 in 17. Five of these patients had an adenoidectomy in the past. The terminology used in that paper was confusing, and it was not entirely clear whether AI meant apnea index or apnea-hypopnea index. Postoperatively, no AHI (or AI) was reported; it was stated that 5 patients still had some partial obstruction postoperatively, but all had a normal oxygen desaturation index (which had been abnormal in 13 preoperatively). Shintani et al84 described 134 children referred for snoring and clinical sleep apnea; 74 had a preoperative AHI >10, but for the rest of the group, the AHI was unspecified. Of this group, 114 had adenotonsillectomy, 13 had adenoidectomy, 4 had adenoidectomy with monotonsillectomy, and 3 had tonsillectomy alone, all presumably at the discretion of the surgeon. Using the authors’ criterion for improvement of a postoperative decrease in AHI by 50%, 84.5% of children who had adenoidectomy and 75.4% of those who had adenotonsillectomy were improved postoperatively (difference between adenotonsillectomy and adenoidectomy, P = .732). In contrast to the findings of Suen et al,48 the preoperative AHI in this report did not predict the likelihood of treatment failure.
To summarize these studies, all of which were case series that were reported with variable rigor, it appears that adenotonsillectomy is curative in 75% to 100% of children, even if obese. The role of adenoidectomy alone is unclear. Postoperatively, children should be retested for OSAS if they continue to snore and possibly if the preoperative AHI was high.
Postoperative Complications and the Need for Inpatient Monitoring
A number of publications have catalogued postoperative complications of adenotonsillectomy in large series of patients,86–98 but these will not be discussed further here. An additional large group of papers have described the risk of complications associated with outpatient adenotonsillectomy in the general population99–110; these case series have generally excluded children with upper airway obstruction from consideration and also will not be discussed further. However, several papers provide data pertaining to complications of surgery in children undergoing adenotonsillectomy for upper airway obstruction, all specifically addressing the risk of postoperative respiratory obstruction.111–118 These are listed in Table 8. These authors define respiratory compromise in various ways but generally consider the need for supplemental oxygen as a minimum criterion. The papers report a wide range for the rate of postoperative respiratory complications (0%–27%), primarily because their populations include different proportions of children with neuromuscular, chromosomal, and craniofacial disorders. This variation makes the study groups too heterogeneous for pooling of the data, and their inclusion of complex patients makes them less valid in estimating the risk of postoperative respiratory compromise in the population being addressed by this practice guideline. Young age (younger than 3 years) and associated medical problems were found in most papers to define the highest risk groups. High preoperative RDI also seems to be a risk factor for postoperative complications.112,113 Time to onset of respiratory compromise appears to be brief, although McColley et al112 reported that 1 patient took 14 hours to manifest respiratory symptoms.
All in all, children with OSAS clearly seem to be at high risk of postoperative respiratory compromise, and increased vigilance in postoperative monitoring is warranted.
Several papers report on the successful use of CPAP in childhood.119–124 In children, CPAP is usually used when adenotonsillectomy is unsuccessful or contraindicated rather than as a primary treatment. Thus, most cases in the above reports describe children with complicated OSAS who are not the target group for this practice guideline. For example, of 80 children reported by Waters et al,120 70 had previous adenotonsillar surgery; the 10 who did not were younger than 6 months or had other significant medical conditions. Of 94 patients reported by Marcus et al,119 only 2 of 18 patients whose OSAS was idiopathic (ie, not associated with another predisposing cause) had not had previous adenotonsillectomy; 1 of these patients had cystic fibrosis. All of the patients described by Guilleminault et al (1995)124 were younger than 1 year. All of the patients described by Rains et al122 and by Guilleminault et al (1986)121 had underlying predisposing abnormalities. All of the subjects described by Tirosh et al123 had previous adenotonsillectomy. These studies do confirm, however, that CPAP is efficacious in children.
Prevalence of Childhood OSAS
Snoring is a common occurrence in childhood, with reported prevalence between 3.2% and 12.1%. The prevalence of childhood OSAS is difficult to estimate, largely because published studies use different PSG criteria for its ascertainment. Reports range from 0.7% to 10.3%.
Sequelae of Childhood OSAS
Childhood OSAS is associated with several important sequelae and complications for which prevalence is unclear because of a lack of population-based cohort studies.
Cross-sectional studies suggest a nearly threefold increase in behavior problems and neurocognitive abnormalities in children with sleep-disordered breathing. Most of these studies did not definitively differentiate children with PS from those with OSAS, so the true prevalence of behavior and learning problems in children with OSAS versus PS is not clear.
No systematic studies exist, but case series suggest that growth (especially weight gain) accelerates after surgery for OSAS, even in children with preexisting obesity, so it appears that OSAS has an inhibitory effect on growth. One study suggests that this effect is attributable to increased metabolic expenditures associated with OSAS.
Cor pulmonale, right ventricular dysfunction, and pulmonary hypertension all have been reported in case reports and series, but their prevalence is unknown. These appear to be reversible after adenotonsillectomy. Systemic hypertension is a known complication of adult OSAS, and elevated diastolic blood pressure has been found in children with OSAS.
Diagnosis of OSAS
The gold standard for diagnosis of OSAS is overnight PSG performed in a sleep lab. Methodologic standards and population-based normal ranges have recently been published, so although older published studies reflect a problem of variability in methods and interpretation, this has diminished in recent years. However, current normative standards for PSG determination of OSAS have been chosen on the basis of statistical distribution of data, and it has not been established that those standards have any validity as predictors of the occurrence of complications. Nonetheless, at the very least, it appears that the severity of PSG abnormality is an important predictor of complications in the immediate postoperative period after adenotonsillectomy.
Alternatives to PSG
Clinical evaluation, including the use of questionnaires such as the one published by Brouillette et al,49 has unacceptably low sensitivity and specificity for predicting OSAS. The use of home audiotaping and videotaping to supplement the clinical evaluation has been inadequately investigated. Additional studies are necessary before any statements about their validity can be made. Pulse oximetry and nap PSG appear to have high specificity and low sensitivity, meaning that positive test results are probably true, but negative test results would need to be confirmed using overnight PSG. The comparability of the results of home and sleep laboratory overnight PSG appears good, but additional study using commercially available equipment in a representative population is necessary for confirmation.
Treatment of OSAS
On the basis of case series that were reported with variable rigor, it appears that adenotonsillectomy is curative in 75% to 100% of children, even if the children are obese. The role of adenoidectomy alone is unclear. Postoperatively, children should be retested for OSAS if they continue to snore and possibly if the preoperative AHI was high. Children with OSAS clearly seem to be at high risk of postoperative respiratory compromise, and increased vigilance in postoperative monitoring is warranted, particularly in those with a high preoperative RDI. CPAP is effective in children, but it is usually used when adenotonsillectomy is delayed, contraindicated, or unsuccessful rather than as a primary treatment.
Subcommittee on Obstructive Sleep Apnea Syndrome
Carole L. Marcus MBBCh, Chairperson
Dale Chapman MD
Sally Davidson Ward, MD
Susanna A. McColley, MD
Michael S. Schechter MD, MPH
Lee J. Brooks MD
American College of Chest Physicians
Jacqueline Jones, MD
Section on Otolaryngology and Bronchoesophagology
Carla T. Herrerias, MPH
Section on Pediatric Pulmonology, 2001–2002
Paul C. Stillwell MD, Chairperson
Dale L. Chapman, MD
Sally L. Davidson Ward, MD
Michelle Howenstine, MD
Michael J. Light, MD
Susanna A. McColley, MD
David A. Schaeffer, MD
Jeffrey S. Wagener, MD
↵a Using the authors’ calculations for sensitivity and specificity but a pretest probability of 60% rather than 43% to calculate likelihood ratios, the post-test probability of having OSAS with a negative score using the technique of Goldstein et al is 26%, higher than their paper indicates for a lower risk group. The post-test probability of a positive result in a higher prevalence group is 69% (ie, 31% rate of false positivity).
likelihood ratio of a positive test result = (sensitivity)/(1 − specificity)
likelihood ratio of a negative test result = (1 − sensitivity)/(specificity)
pretest odds = (prevalence)/(1 − prevalence)
post-test odds = pretest odds × likelihood ratio
post-test probability = (post-test odds)/(post-test odds + 1)
- Copyright © 2002 by the American Academy of Pediatrics