clinical assessment of pediatric obstructive sleep apnea: a systematic review and meta-analysis

The LaryngoscopeVC 2012 The American Laryngological,Rhinological and Otological Society, Inc.

Clinical Assessment of Pediatric Obstructive Sleep Apnea: ASystematic Review and Meta-Analysis

Victor Certal, MD; Emanuel Catumbela, MD; Joao C. Winck, MD, PhD; Ines Azevedo, MD, PhD;

Armando Teixeira-Pinto, PhD; Altamiro Costa-Pereira, MD, PhD

Objectives/Hypothesis: Clinical symptoms and signs are routinely used to investigate pediatric obstructive sleep apnea(OSA). This study aimed to systematically assess the evidence for the diagnostic accuracy of individual or combined clinicalsymptoms and signs in predicting pediatric OSA.

Study Design: A systematic review of the literature and diagnostic meta-analysis.Methods: Four medical databases were searched (from inception to August 2011). Studies were included that compared

the clinical assessment with the current gold standard (full polysomnography). The study quality was assessed using thequality assessment tool for diagnostic accuracy studies. Summary estimates of diagnostic accuracy were determined using thesensitivity, specificity, positive and negative likelihood ratios, diagnostic odds ratio, and hierarchical summary receiver operat-ing characteristic (HSROC) model for meta-analyses.

Results: Ten diagnostic studies with 1,525 patients were included in the review. There was substantial variation in thesensitivity and specificity among different symptoms and signs, as well as across studies. Tonsillar size and snoring reportedby parents or caregivers had high sensitivity but low specificity. In contrast, excessive daytime somnolence, observed apnea,and difficulty in breathing during sleep had high specificity but low sensitivity. Seven models of a combination of symptomsand signs presented moderate sensitivity (range, 0.04–0.94) and specificity (range, 0.28–0.99). The HSROC indicates poordiagnostic performance of the symptoms and signs in predicting pediatric OSA.

Conclusions: Neither single nor combined symptoms and signs have satisfactory performance in predicting pediatricOSA. Alternative diagnostic models are necessary to improve the accuracy.

Key Words: Pediatric sleep apnea, systematic review, meta-analysis, diagnostic accuracy.Laryngoscope, 000:000–000, 2012

INTRODUCTIONPediatric obstructive sleep apnea (OSA) is charac-

terized by recurring episodes of complete and/or partialobstruction of the upper airway during sleep, resultingin intermittent hypoxemia and hypercapnia, frequentarousals, and sleep fragmentation.1,2 OSA is a severecondition in children and differs from its adultcounterpart in its physiology, clinical presentation, poly-somnographic characteristics, and outcomes. Theestimated prevalence in children is 1% to 3%3; however,the prevalence is difficult to measure because of subdiag-nosis.2,4 Lack of community awareness about the

negative sleep-related effects on the daily functioning ofchildren, together with the parents’ underestimation ofthe problem when they talk to the physician, are factorsthat contribute to this underestimation.5,6

Polysomnography (PSG) is the gold standard fordiagnosing and quantifying OSA.6–9 Nocturnal PSGrecordings provide unbiased and objective informationon various sleep-related characteristics such as sleeparchitecture, cardiac and respiratory patterns, andgas exchange. However, several factors have ham-pered a more extensive implementation of suchdiagnostic procedures, including the inconvenience forboth parents and child spending the night in the lab-oratory, the rather onerous and labor-intensive natureof this diagnostic procedure, the relative scarcity oflaboratories with expertise in children’s sleep disor-ders, and as a corollary to this, the extended waitingperiod between referral and actual testing. Thus, thediagnosis of pediatric OSA is still often made on aclinical basis.10

Numerous studies have assessed the accuracy ofclinical symptoms and signs in detecting pediatric OSA.The most commonly assessed symptoms and signsinclude snoring, observed apnea, mouth breathing, ex-cessive daytime somnolence (EDS), and large tonsil size,but the diagnostic accuracy varies significantly for dif-ferent symptoms and signs as well as across studies.

From the Department of Otorhinolaryngology (V.C.), Hospital SaoSebastiao, Santa Maria da Feira; Center for Research in HealthTechnologies and Information Systems (V.C., E.C., A.T.-P., A.C.-P.),University of Porto, Porto; Department of Pulmonology (J.C.W.),Department of Pediatrics (I.A.), and Department of Biostatistics andMedical Informatics (A.T.-P.), Faculty of Medicine, University of Porto,Porto, Portugal.

Editor’s Note: This Manuscript was accepted for publication May4, 2012.

The authors have no funding, financial relationships, or conflictsof interest to disclose.

Send correspondence to Victor Certal, MD, Departamento deCiencias da Informacao e da Decisao em Sa�ude, Faculdade de Medicina,Universidade do Porto, Al. Prof. Hernani Monteiro, 4200-319 Porto,Portugal. E-mail: [email protected]

DOI: 10.1002/lary.23465

Laryngoscope 000: Month 2012 Certal et al.: Clinical Assessment of Pediatric OSA

1

Two previous reviews addressed this variability.11,12

However, both used a poor methodological approach andpresented conflicting results.

We conducted a systematic review and meta-analy-sis to assess the accuracy of clinical symptoms and signsin predicting pediatric OSA. The secondary aim was toassess possible sources of heterogeneity.

MATERIALS AND METHODS

Study DesignA systematic review and meta-analysis of studies focusing

on the clinical evaluation of pediatric OSA was undertaken. Themethodological approach included the development of selectioncriteria, definition of search strategies, assessment of the qual-ity of the studies, data abstraction, and statistical dataanalysis.

Selection CriteriaFor proper identification of studies eligible for the analy-

sis, the study selection criteria were defined before datacollection. Only articles whose primary objective was to evalu-ate the ability of clinical evaluation (i.e., the clinical historyand/or physical examination) to accurately diagnose pediatricOSA were selected. Articles that predominantly evaluated otherdiagnostic modalities (e.g., radiographs, pulse oximetry, electro-cardiography, laboratory testing) were excluded. Only articlesusing full multichannel PSG (with electroencephalography) asthe gold standard reference test for comparison with clinicalevaluation were selected. The included studies provided infor-mation such as sensitivity and specificity or sufficientinformation to allow the construction of the diagnostic 2 � 2table with four cells for true positives, false negatives, false pos-itives, and true negatives.

Search StrategyOur primary method to locate potentially eligible studies

was a computerized literature search in the MEDLINE data-base (through PubMed), from inception to August 2011, withoutany restriction on the language of publication, using the follow-ing search key words and MeSH terms: ‘‘sleep apnea ORsnoring OR sleep disordered breathing OR obstructive sleepapnea syndrome OR sleep apnea/diagnosis’’ AND ‘‘polysomnog-raphy OR sleep monitoring’’ AND ‘‘children OR child.’’Literature searches were also undertaken, using the samesearch key words, in the following databases: the Cochrane col-laboration, SCOPUS, and ProQuest Dissertations and ThesisDatabase.

In defining the search strategies, we prioritized formatswith higher sensitivity to increase the probability of identifyingall relevant articles.

Conference proceedings of the American Thoracic Society,Associated Professional Sleep Societies, European SleepResearch Society, and European Respiratory Society for theyears 2002 to 2007 were hand-searched for eligible studies. Ref-erence lists from eligible studies and review articles were cross-checked to identify additional trials. A grey literature researchwas also performed using OpenSIGLE Database. Both Englishand non-English studies were considered.

Study Quality Assessment and Data AbstractionThe titles and abstracts of the retrieved studies were inde-

pendently screened for relevance by two reviewers (V.C. andE.C.). As currently recommended for systematic reviews of diag-nostic accuracy studies, the reviewers evaluated themethodology of the selected studies using the 14-item QualityAssessment Tool for Diagnostic Accuracy Studies (QUADAS).All disagreements were resolved by consensus.

The data were used to construct 2 � 2 contingency tables,from which sensitivity and specificity were calculated for each

Fig. 1. Flow diagram of study identi-fication and selection. PSG ¼polysomnography.


2

study. When the raw data were presented in 3 � 3 or 4 � 4tables (e.g., when the reference apnea-hypopnea index [AHI]was defined by >2 categories), we reconstructed 2 � 2 contin-gency tables by considering AHI >1 as the positive state ofpediatric OSA.

Statistical AnalysisWe used the sensitivity, specificity, positive (LRþ) and

negative likelihood ratios (LR�), and diagnostic odds ratio(DOR) as measures of diagnostic accuracy. Individual diag-nostic studies usually use different cutoff criteria (explicitlyor implicitly) to define an abnormal result. This may be par-ticularly true for diagnostic studies that evaluate clinicalmanifestations, where detection of the same symptom orsign can vary significantly among observers. For this reason,we used a hierarchical summary receiver operating charac-teristic (HSROC) model for meta-analyses of diagnosticaccuracy that model the accuracy, threshold effect, and de-pendence of accuracy on threshold.13 The random effectsmodel was applied to account for the within-study andbetween-study variation.

The heterogeneity in results across studies was investi-gated through a visual examination of forest plots of sensitivityand specificity, as well as receiver operating characteristic(ROC) plots. We did not use statistical tests of heterogeneitybecause of expected variations arising from the interdependenceof sensitivity and specificity. We anticipated a substantial heter-ogeneity between studies and used a random-effects model formeta-analyses that allowed the heterogeneity to be taken intoaccount. We identified outliers (studies with data points fallingoutside the 95% confidence region of the summary operatingpoint) in the ROC plots, and the characteristics of these studieswere investigated. We also performed subgroup analyses toexplore the possible causes of heterogeneity, such as AHI crite-ria for OSA (�1 or >1 episode/hour) and study setting (sleepcenters or non-sleep centers).

The data processing and statistical analysis were per-formed using the Cochrane Collaboration’s Review Managersoftware version 4.2, RevMan Analyses software version1.0, and SAS software (SAS 9.2 procNlmixed, SAS Institute,Cary, NC).

RESULTS

Search and Study SelectionAll database searches were performed in August

2011. A flow chart of the process of study identificationand inclusion/exclusion is shown in Figure 1.

In total, 810 articles were identified using thesearch strategy and sources listed. After the titles andabstracts were screened for relevance, 780 articles wereexcluded (the reasons for exclusion are presented in Fig.1). The remaining 30 articles were retrieved for moredetailed full-text evaluation, and 21 were excluded forthe following reasons: six studies14–19 used a resumedPSG as the reference gold standard and thus lackedelectroencephalography in their reference PSG; twostudies included pediatric and adult populations in theirresults,8,20 four studies did not provide sufficient infor-mation to build 2 � 2 tables;21–24 and nine studiesevaluated other diagnostic modalities.25–33 One study34

was included after a hand-search of references of theincluded studies. Finally, 10 studies were included in thereview.34–43

TABLEI.

CharacteristicsofIncludedStudies.

Author,Year,Reference

Country

Setting

Sample

Size(n)

Age,Male

(%)

ClinicalCriteriaforOSA

Polysomnographic

CriteriaforOSA

Carrolletal.,

199535

USA

Pediatric

sleepandbreathing

disorders

clinic

83

5.4

months–14.8

years

(58)

Individualsymptomsand

combinationofsym

ptoms

AHI>1

Chauetal.,

200336

Singapore

Sleepclinic

62

2–13years

(58)

Individualsymptoms

AHI>1

Chervin

etal.,

200738

USA

University-basedsleep

disorders

laboratory

105

5–12.9

years

(57)

BasedonthePediatric

Sleep

Questionnaire

AHI>1

Goldstein

etal.,

199438

USA

Pediatric

otolaryngology

outpatientclinic

30

1–14years

(60)

Combinationofsymptoms

AHI>15

Goodwin

etal.,

200539

USA

Generalpopulation

480

6–8years

(50)

Individualsymptomsandcombination

ofsymptoms

AHI>1

Leachetal.,

199140

USA

Sleepdisorders

center

93

18months–1

2years

(59)

Individualsymptoms

NR

Lietal.,

200641

China

Sleepclinic

229

5–15years

(65)

Combinationofsymptoms

AHI>5

Rosenetal.,

199934

USA

Child

ren’s

sleeplaboratory

326

1–12years

(56)

Individualsymptomsandcombination

ofsymptoms

AHI>1

Sprosonetal.,

200942

UK

Communityhospital

67

3–8years

(60)

BasedonthePediatric

SleepQuestionnaire

AHI>1

Xuetal.,

200643

China

Sleepclinic

50

4–18years

(76)

Individualsymptomsand

combinationofsym

ptoms

AHI>5

OSA¼

obstructivesleepapnea;AHI¼

apnea/hypopneaindex;NR

¼notreported.


3

Methodological Quality of the Included StudiesThe main characteristics of the included studies are

presented in Table I. Overall, 1,525 patients wereincluded, with a mean of 152.5 patients per study(range, 30–480 patients). Despite the American Associa-tion of Sleep Medicine (AASM) recommendations,44 onlysix studies34–37,39,42 defined AHI >1/hour as the diagno-sis of pediatric OSA (range, 1–15/hour). One study40 didnot report any AHI threshold.

All studies satisfied seven of the 14 items in the QUA-DAS checklist for the assessment of methodologicalquality, namely, selection criteria described, complete veri-fication using reference standard, same reference standardused, reference standard independent of the index test,details of execution of the reference standard, similar clini-cal data available for interpretation of the test as that inpractice, and withdrawals explained (Table II). The mainmethodological limitations of the studies were related topoor reporting of items four (period between referencestandard and index test), eight (details about the executionof the index test), 10 and 11 (blind interpretation of thereference and index tests), and 13 (reporting of uninter-pretable/intermediate test results). In all studies, noexplicit criterion for identifying symptoms and signswas presented, and interobserver agreement was notassessed.

Sensitivity and SpecificitySensitivity and specificity for different symptoms

and signs varied substantially; this variation was alsoobserved across the studies included (Fig. 2). We there-fore did not report the pooled sensitivity and specificityof each symptom or sign.

Tonsillar size and snoring reported by parents orcaregivers had relatively high sensitivity but low speci-ficity. In contrast, EDS, observed apnea, and difficulty

breathing during sleep had relatively high specificity butlow sensitivity.

Several models of combined symptoms and signswere assessed (Table III). Seven models/algorithms pre-sented a moderate sensitivity (range, 0.04–0.94) andspecificity (range, 0.28–0.99): the Pediatric Sleep Ques-tionnaire, the OSA score initially proposed by Brouilletteet al.14 (snoring, difficulty breathing, EDS and behavior,personality or school performance), the modified OSAscore of Carroll et al.35 (observed apnea, difficultybreathing, and parents watching child during sleep), themodel of Xu et al.43 (observed apnea, nocturnal enuresis,intrusive naps, mouth breathing, and moderate to severetonsil hypertrophy), the model of Goldstein et al.38 (snor-ing, pauses, difficulty breathing, sleep with neckextended, EDS, and adenoid face), and Goodwin et al.39

presented two models of symptoms (snoring and learningproblems/snoring and EDS).

Positive Likelihood Ratio, Negative LikelihoodRatio, and Diagnostic Odds Ratio

Table IV shows the pooled LRþ, LR�, and DOR.None of the symptoms or signs had a pooled LRþ >10,which indicates that the presence of these symptomsand signs may not provide convincing evidence to iden-tify OSA among children. Moreover, none of thesymptoms or signs had pooled LR� <0.1, which suggeststhat the absence of these symptoms or signs cannotaccurately exclude a diagnosis of pediatric OSA. Snoringand observed apnea had a higher pooled DOR than theother symptoms or signs.

HSROCFigure 3 displays the ROC plots for each of the five

symptoms and signs assessed by at least four studies.

TABLE II.Assessment of the Methodological Quality of the Studies Included According to the 14-Item QUADAS Checklist*

Author, Year, Reference

Question

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Carroll et al., 199535 Yes Yes Yes Unclear Yes Yes Yes No Yes Yes Yes Yes Unclear Yes

Chau et al., 200336 Yes Yes Yes Unclear Yes Yes Yes No Yes Unclear No Yes Unclear Yes

Chervin et al., 200737 Yes Yes Yes Unclear Yes Yes Yes Yes Yes Yes Yes Yes Unclear Yes

Goldstein et al., 199438 Yes Yes Yes Unclear Yes Yes Yes No Yes Unclear Unclear Yes Unclear Yes

Goodwin et al., 200539 Unclear Yes Unclear Unclear Yes Yes Yes Yes Yes Yes Yes Yes Unclear Yes

Leach et al., 199240 Yes Yes Unclear Unclear Yes Yes Yes No Yes Unclear Unclear Yes Unclear Yes

Li et al., 200641 Yes Yes Yes Unclear Yes Yes Yes Yes Yes Unclear Unclear Yes Unclear Yes

Rosen et al., 199934 Yes Yes Yes Unclear Yes Yes Yes No Yes Unclear No Yes Unclear Yes

Sproson et al., 200942 Unclear Yes Yes Unclear Yes Yes Yes Yes Yes Unclear Unclear Yes Unclear Yes

Xu et al., 200643 Yes Yes Yes Unclear Yes Yes Yes No Yes Yes Yes Yes Unclear Yes

*Quality Assessment of Diagnostic Accuracy Studies (QUADAS) checklist: 1) Was the spectrum of patients representative of the patients who will receivethe test in practice? 2) Were selection criteria clearly described? 3) Is the reference standard likely to classify the target condition correctly? 4) Is the periodbetween reference standard and index test short enough to be reasonably sure that the target condition did not change between the two tests? 5) Did a wholesample or random selection of the sample receive verification using a reference standard? 6) Did patients receive the same reference standard regardless of theindex test result? 7) Was the reference standard independent of the index test (i.e., the index test did not form part of the reference standard)? 8) Was the execu-tion of the index test described in sufficient detail to permit replication of the test? 9) Was the execution of the reference standard described in sufficient detail topermit its replication? 10) Were the index test results interpreted without knowledge of the results of the reference standard? 11) Were the reference standardresults interpreted without knowledge of the results of the index test? 12) Were the same clinical data available when test results were interpreted as would beavailable when the test is used in practice? 13) Were uninterpretable/intermediate test results reported? 14) Were withdrawals from the study explained?


4

The shape of the summary ROC curves differed betweensymptoms and signs; however, none of the summaryROC curves was close to the upper left corner of thegraph, which indicates poor diagnostic performance ofthe symptoms and signs in predicting OSA amongchildren.

Heterogeneity and Subgroup AnalysesAs expected, the forest plots and ROC plots demon-

strated substantial heterogeneity across studies, asshown in Figures 2 and 3. No outlier was identified.Because only two studies were performed outside sleepcenters, no subgroup analysis was possible for the cova-riate setting (sleep centers vs. general population). Thesubgroup analysis was only performed for covariate AHI(AHI ¼ 1 vs. AHI >1); for all symptoms but one (diffi-culty in breathing), no statistical evidence is provided

for a difference in the diagnostic accuracy of symptomsand signs according to the AHI threshold (Figure 4).

DISCUSSION

Methodological LimitationsSome methodological limitations of this review

should be considered. First, as expected, there was sub-stantial heterogeneity in results across studies. Severalfactors could contribute to this situation. In the 2005edition of the International Classification of Sleep Disor-ders,44 the AASM defined an AHI greater than 1/hour asabnormal in children, but three included studies38,41,43

had other thresholds in their polysomnograph analysis,and one study40 did not specify any threshold. None ofthe studies clearly described how to identify symptomsand signs, and it is therefore unclear whether there wasa substantial variation across studies with regard to the

Fig. 2. Diagnostic accuracy (sensi-tivity, specificity, and 95% confi-dence interval) of clinical symptomsand signs from each study. TP ¼true positive; FP ¼ false positive; FN¼ false negative; TN¼ true negative.


5

explicit criteria for the definition of these symptoms andsigns. However, a substantial variation in the implicitthreshold could be expected among studies, as the testresults (presence or absence of symptoms and signs)depend on the perceptions, interpretation, and judgmentof observers. In this review, we used the HSROC modelfor meta-analyses of diagnostic accuracy, which takesthreshold effects into account. The subgroup analysis forthe covariate AHI (AHI ¼ 1 vs. AHI >1) did not appearto influence the diagnostic accuracy.

In addition, eight34–39,40,42,44 of the 10 studies wereconducted in sleep clinics or sleep centers, where theprevalence of pediatric OSA is generally higher than inprimary health services. A further decrease in the diag-nostic yield may thus be expected when clinicalsymptoms and signs are used to predict pediatric OSAamong children in primary care settings.

Diagnostic Accuracy of Symptoms and Signs forPredicting Pediatric OSA

Despite the substantial variation in results amongstudies, this review demonstrates a poor overall diagnos-tic accuracy for single symptoms or signs in predictingpediatric OSA. Tonsillar size and snoring reported byparents or caregivers have relatively high sensitivity.Generally, when a test has high sensitivity, a negative

result can be used to rule out the target condition. Thus,the absence of these symptoms and signs may be usefulfor excluding a diagnosis of OSA. However, the low spec-ificity of these symptoms and signs may lead to a largenumber of false diagnoses of OSA.

In contrast, EDS, observed apnea, and difficultybreathing during sleep have relatively high specificitybut low sensitivity. Because a positive result in a high-specificity test can be used to confirm a diagnosis of thetarget condition, the presence of the above-mentionedsymptoms or signs may be useful for identifying OSA.

Several models for the combination of symptomsand signs have been proposed; however, the diagnosticperformance of these combinations has not yet been wellassessed. Eight studies34,35,37–39,41–43 evaluated the diag-nostic accuracy of combined symptoms and signs, andthe wide variation in models makes it impossible to poolthe results. None of the models present reasonable sensi-tivity and specificity, and there appears to be substantialvariation in the diagnostic performance among thesemodels.

Clinical Implications of the FindingsThis review demonstrates that single and combined

symptoms and signs do not have satisfactory diagnosticperformance in predicting pediatric OSA. These results,

TABLE III.Models of Combinations of Symptoms and Signs and Their Diagnostic Accuracy in Predicting Pediatric OSA.

Models of Combinations Author, Year, Reference Sensitivity Specificity

Pediatric sleep questionnaire Sproson et al., 200942;Chervin et al., 200737

0.53;0.78

0.67;0.72

Modified OSA score: observed apnea,difficulty breathing and parents watchingchild during sleep

Carroll et al., 199535 0.40 0.92

OSA score: snoring, difficulty in breathing,EDS, and behavior, personality orschool performance

Rosen et al., 199934 0.47 0.28

Night sweating, mouth breathing and snoring Li et al., 200641 0.81 0.57

Observed apnea, nocturnal enuresis, intrusivenaps, mouth breathing, and moderate tosevere tonsillar hypertrophy

Xu et al., 200643 0.94 0.42

Snoring and learning problems Goodwin et al., 200539 0.04 0.99

Snoring and EDS Goodwin et al., 200539 0.09 0.97

Snoring, pauses, difficulty in breathing, sleepwith neck extended, EDS, and adenoid face

Goldstein et al., 199438 0.92 0.29

OSA ¼ obstructive sleep apnea; EDS ¼ excessive daytime somnolence.

TABLE IV.Summary Estimates of the Diagnostic Accuracy of Clinical Symptoms and Signs for Pediatric Obstructive Sleep Apnea.

Clinical CriteriaNo. of Patients

(Studies)SensitivityRange

SpecificityRange

Pooled DOR(95% CI)

Pooled LRþ(95% CI)

Pooled LR�(95% CI)

Snoring 753 (5) 0.30–0.97 0.26–0.90 4.94 (1.93–12.64) 1.76 (1.09–2.82) 0.36 (0.17–0.76)

Observed apnea 273 (4) 0.35–0.74 0.48–0.95 3.73 (1.21–11.54) 2.29 (0.94–5.60) 0.61 (0.45–0.84)

Difficulty breathing 272 (4) 0.12–0.89 0.42–0.95 3.62 (1.56–8.37) 2.37 (1.49–3.76) 0.65 (0.37–1.17)

EDS 691 (4) 0.18–0.45 0.58–0.86 1.59 (1.02–2.47) 1.49 (1.01–2.06) 0.91 (0.82–1.00)

Tonsillar size 516 (4) 0.48–0.82 0.43–0.84 3.34 (1.97–5.66) 1.73 (1.22–2.45) 0.52 (0.40–0.67)

DOR ¼ diagnostic odds ratio; CI ¼ confidence interval; LRþ ¼ positive likelihood ratio; LR� ¼ negative likelihood ratio; EDS ¼ excessive daytimesomnolence.


6

taken together with the findings from a previousreview,9 suggest that the diagnosis of pediatric OSAshould be based on more reliable noninvasive meth-ods. In recent years, several diagnostic tools havebeen developed, which enable some degree of predict-ability in the detection of OSA or some of its

consequences.9,25,26,33,45,46 However, we should empha-size that such tools have not been extensivelyevaluated beyond the program where they were devel-oped, and their validity has not been assessed acrossdifferent countries or settings (i.e., community vs.referral populations).

Fig. 3. Hierarchical summary receiver operating characteristic (HSROC) curves for the diagnostic performance of each symptom and sign.EDS ¼ excessive daytime somnolence.


7

A recent study,47 based on the conceptual frame-work that OSAS will result in unique signatures in theexpression of genes or proteins, analyzed morning urinesamples from 120 children using differential in-gel elec-trophoresis (2D-DIGE) and found that unique sets ofproteins were either increased or decreased in the urineof children with OSA. In addition, ROC analyses using

more than one of the putative biomarkers showed that ifall four proteins were employed, the diagnostic accuracyyielded 100% sensitivity and 96.5% specificity to predictOSA. Further research is needed to explore the diagnos-tic value of this approach and its value to clinicaldecision making, but if this approach is validated, itshould permit the delineation of specific criteria for

Fig. 4. Hierarchical summary receiver operating characteristic (HSROC) curves for the diagnostic performance of each symptom and signfor covariate apnea/hypopnea index (AHI) (AHI ¼ 1 vs. AHI >1). EDS ¼ excessive daytime somnolence.


8

population screening (low cost, high sensitivity, andspecificity) and enable priority assignments for treat-ment and outcome monitoring in children considered tobe at increased risk.

In the interim, most physicians have to rely on clin-ical symptoms and signs for identifying possible OSApatients who need further treatment. However, giventhe substantial variation in the models of combinedsymptoms and signs reported by the studies included inthis analysis, as well as the limited number of studieswith each model, further prospective studies are stillnecessary to compare the diagnostic performance ofthese models, as well as to develop and validate moreeffective models for the prediction of pediatric OSA. Thenew models should include the symptoms and signs withhigh specificity identified by this systematic review, suchas EDS, observed apnea, and difficulty breathing duringsleep, and symptoms with high sensitivity such as snor-ing and tonsillar size. This combination strategy maylead to an increase in sensitivity with acceptable speci-ficity. Different models should be established andvalidated in various age groups, as many of the symp-toms and signs are age dependent. Further studiesshould define explicit criteria for identifying symptomsand signs, interpret the index tests (symptoms andsigns) and reference tests blindly, and include patientsfrom primary care settings.

CONCLUSIONThis meta-analysis provides a comprehensive criti-

cal review of the literature to date and a statisticalanalysis of the current diagnostic accuracy of clinicalsymptoms and signs for investigating pediatric OSA.The major finding of this study is that both single andcombinations of symptoms and signs have poor diagnos-tic accuracy in predicting pediatric OSA. It is importantto critically evaluate the current evidence to appropri-ately guide future efforts in this research field. High-quality diagnostic studies that address alternative non-invasive methods in the evaluation of pediatric OSA arestill necessary.

AcknowledgmentThe authors gratefully acknowledge the support pro-

vided by Helder Silva, MD, from the hospital SaoSebastiao.

BIBLIOGRAPHY

1. Goldstein NA, Pugazhendhi V, Rao SM, et al. Clinical assessment of pedi-atric obstructive sleep apnea. Pediatrics 2004;114:33–43.

2. Kotagal S. Sleep disorders in childhood. Neurol Clin 2003;21:961–981.3. Ali NJ, Pitson DJ, Stradling JR. Snoring, sleep disturbance, and behaviour

in 4–5 year olds. Arch Dis Child 1993;68:360–366.4. Marcus CL. Sleep-disordered breathing in children. Curr Opin Pediatr

2000;12:208–212.5. Reuveni H, Simon T, Tal A, Elhayany A, Tarasiuk A. Health care services

utilization in children with obstructive sleep apnea syndrome. Pediatrics2002;110:68–72.

6. Rembold CM, Suratt PM. Children with obstructive sleep-disorderedbreathing generate high-frequency inspiratory sounds during sleep.Sleep 2004;27:1154–1161.

7. Schechter MS. Technical report: diagnosis and management of childhoodobstructive sleep apnea syndrome. Pediatrics 2002;109:e69.

8. de Almeida FR, Ayas NT, Otsuka R, et al. Nasal pressure recordings todetect obstructive sleep apnea. Sleep Breath 2006;10:62–69.

9. Morielli A, Ladan S, Ducharme FM, Brouillette RT. Can sleep and wake-fulness be distinguished in children by cardiorespiratory and videotaperecordings? Chest 1996;109:680–687.

10. Gozal D, Kheirandish-Gozal L. New approaches to the diagnosis of sleep-disordered breathing in children. Sleep Med 2010;11:708–713.

11. Brietzke SE, Katz ES, Roberson DW. Can history and physical examina-tion reliably diagnose pediatric obstructive sleep apnea/hypopnea syn-drome? A systematic review of the literature. Otolaryngol Head NeckSurg 2004;131:827–832.

12. Messner AH. Evaluation of obstructive sleep apnea by polysomnographyprior to pediatric adenotonsillectomy. Arch Otolaryngol Head Neck Surg1999;125:353–356.

13. Leeflang MM, Deeks JJ, Gatsonis C, Bossuyt PM. Systematic reviews ofdiagnostic test accuracy. Ann Intern Med 2008;149:889–897.

14. Brouilette R, Hanson D, David R, et al. A diagnostic approach to suspectedobstructive sleep apnea in children. J Pediatr 1984;105:10–14.

15. Nieminen P, Tolonen U, Lopponen H. Snoring and obstructive sleep apneain children: a 6-month follow-up study. Arch Otolaryngol Head NeckSurg 2000;126:481–486.

16. Otsu SY, Koike Y, Hori Y, Abe K. Hypertrophy of the palatine tonsils andsleep respiratory disorders. Acta Otolaryngol Suppl 1996;523:219–221.

17. Preutthipan A, Chantarojanasiri T, Suwanjutha S, Udomsubpayakul U.Can parents predict the severity of childhood obstructive sleep apnoea?Acta Paediatr 2000;89:708–712.

18. Constantin E, Tewfik TL, Brouillette RT. Can the OSA-18 quality-of-lifequestionnaire detect obstructive sleep apnea in children? Pediatrics2010;125:e162–168.

19. Valera FC, Avelino MA, Pettermann MB, et al. OSAS in children: correla-tion between endoscopic and polysomnographic findings. OtolaryngolHead Neck Surg 2005;132:268–272.

20. Erdamar B, Suoglu Y, Cuhadaroglu C, Katircioglu S, Guven M. Evaluationof clinical parameters in patients with obstructive sleep apnea and pos-sible correlation with the severity of the disease. Eur Arch Otorhinolar-yngol 2001;258:492–495.

21. Li AM, Wong E, Kew J, Hui S, Fok TF. Use of tonsil size in the evaluationof obstructive sleep apnoea. Arch Dis Child 2002;87:156–159.

22. Brooks LJ, Stephens BM, Bacevice AM. Adenoid size is related to severitybut not the number of episodes of obstructive apnea in children. JPediatr 1998;132:682–686.

23. Suen JS, Arnold JE, Brooks LJ. Adenotonsillectomy for treatment of ob-structive sleep apnea in children. Arch Otolaryngol Head Neck Surg1995;121:525–530.

24. Jain A, Sahni JK. Polysomnographic studies in children undergoing ade-noidectomy and/or tonsillectomy. J Laryngol Otol 2002;116:711–715.

25. Nigro CA, Aimaretti S, Gonzalez S, Rhodius E. Validation of the WristOx3100 oximeter for the diagnosis of sleep apnea/hypopnea syndrome.Sleep Breath 2009;13:127–136.

26. Grover SS, Pittman SD. Automated detection of sleep disordered breathingusing a nasal pressure monitoring device. Sleep Breath 2008;12:339–345.

27. Griffiths A, Maul J, Wilson A, Stick S. Improved detection of obstructiveevents in childhood sleep apnoea with the use of the nasal cannula andthe differentiated sum signal. J Sleep Res 2005;14:431–436.

28. Jacob SV, Morielli A, Mograss MA, Ducharme FM, Schloss MD, Brouil-lette, RT. Home testing for pediatric obstructive sleep apnea syndromesecondary to adenotonsillar hypertrophy. Pediatr Pulmonol 1995;20:241–252.

29. Gozal D, Kheirandish-Gozal L, Serpero LD, Capdevila OS, Dayyat E. Ob-structive sleep apnea and endothelial function in school-aged nonobesechildren: effect of adenotonsillectomy. Circulation 2007;116:2307–2314.

30. Saeed MM, Keens TG, Stabile MW, Bolokowicz J, Davisson-Ward SL.Should children with suspected obstructive sleep apnea syndrome andnormal nap sleep studies have overnight sleep studies? Chest 2000;118:360–365.

31. Sivan Y, Kornecki A, Schonfeld T. Screening obstructive sleep apnoea syn-drome by home videotape recording in children. Eur Respir J 1996;9:2127–2131.

32. Uliel S, Tauman R, Greenfeld M, Sivan Y. Normal polysomnographic respi-ratory values in children and adolescents. Chest 2004;125:872–878.

33. Ghaemmaghami H, Abeyratne UR, Hukins C. Normal probability testingof snore signals for diagnosis of obstructive sleep apnea. Conf ProcIEEE Eng Med Biol Soc 2009;2009:5551–5554.

34. Rosen CL. Clinical features of obstructive sleep apnea hypoventilation syn-drome in otherwise healthy children. Pediatr Pulmonol 1999;27:403–409.

35. Carroll JL, McColley SA, Marcus CL, Curtis S, Loughlin GM. Inability ofclinical history to distinguish primary snoring from obstructive sleepapnea syndrome in children. Chest 1995;108:610–618.

36. Chau KW, Ng DK, Kwok CK, Chow PY, Ho JC. Clinical risk factors for ob-structive sleep apnoea in children. Singapore Med J 2003;44:570–573.

37. Chervin RD, Weatherly RA, Garetz SL, et al. Pediatric sleep question-naire: prediction of sleep apnea and outcomes. Arch Otolaryngol HeadNeck Surg 2007;133:216–222.

38. Goldstein NA, Sculerati N, Walsleben JA, Bhatia N, Friedman DM, Rapo-port DM. Clinical diagnosis of pediatric obstructive sleep apnea vali-dated by polysomnography. Otolaryngol Head Neck Surg 1994;111:611–617.

39. Goodwin JL, Kaemingk KL, Mulvaney SA, Morgan WJ, Quan SF. Clinicalscreening of school children for polysomnography to detect sleep-


9

disordered breathing—the Tucson Children’s Assessment of Sleep Apneastudy (TuCASA). J Clin Sleep Med 2005;1:247–254.

40. Leach J, Olson J, Hermann J, Manning S. Polysomnographic and clinicalfindings in children with obstructive sleep apnea. Arch OtolaryngolHead Neck Surg 1992;118:741–744.

41. Li AM, Cheung A, Chan D, et al. Validation of a questionnaire instrumentfor prediction of obstructive sleep apnea in Hong Kong Chinese children.Pediatr Pulmonol 2006;41:1153–1160.

42. Sproson EL, Hogan AM, Hill CM. Accuracy of clinical assessment of paedi-atric obstructive sleep apnoea in two English centres. J Laryngol Otol2009;123:1002–1009.

43. Xu Z, Cheuk DK, Lee SL. Clinical evaluation in predicting childhoodobstructive sleep apnea. Chest 2006;130:1765–1771.

44. American Academy of Sleep Medicine. International Classification ofSleep Disorders. Diagnostic and Coding Manual. 2nd ed. Westchester,IL: American Academy of Sleep Medicine; 2005.

45. Mason DG, Iyer K, Terrill PI, Wilson SJ, Suresh S. Pediatric obstructivesleep apnea assessment using pulse oximetry and dual RIP bands. ConfProc IEEE Eng Med Biol Soc 2010;2010:6154–6157.

46. Okun MN, Hajiangelis N, Green D, Hedli LC, Lee KC, Krieger AC.Acoustic rhinometry in pediatric sleep apnea. Sleep Breath 2010;14:43–49.

47. Gozal D, Jortani S, Snow AB, et al. Two-dimensional differential in-gelelectrophoresis proteomic approaches reveal urine candidate biomarkersin pediatric obstructive sleep apnea. Am J Respir Crit Care Med 2009;180:1253–1261.


10

clinical assessment of pediatric obstructive sleep apnea: a systematic review and meta-analysis

Documents