Night-to-night consistency of at-home nocturnal pulse oximetry testing for obstructive sleep apnea in children

  • Published on

  • View

  • Download

Embed Size (px)


<ul><li><p>Pediatric Pulmonology</p><p>Night-to-Night Consistency of At-Home NocturnalPulse Oximetry Testing for Obstructive Sleep</p><p>Apnea in Children</p><p>Martino Pavone, MD,1* Renato Cutrera, MD, PhD,1 Elisabetta Verrillo, MD,1 Teresa Salerno, MD,1</p><p>Serena Soldini,1 and Robert T. Brouillette, MD2</p><p>Summary. Rationale: At-home nocturnal pulse oximetry has a high positive predictive value</p><p>(PPV) for polysomnographically-diagnosed obstructive sleep apnea (OSA) but no studies have</p><p>been published testing the night-to-night consistency of at-home nocturnal pulse oximetry for</p><p>the evaluation of suspected OSA in children. We therefore determined the night-to-night con-</p><p>sistency of nocturnal pulse oximetry as a diagnostic test for OSA in children. Methods:</p><p>We prospectively studied 148 children (96 male) aged 4.9 2.4 (1.211.8) years, referred forsuspected OSA. To evaluate night-to-night consistency, we compared an oximetry analysis</p><p>method, the McGill Oximetry Score (MOS), from two consecutive at-home nocturnal pulse</p><p>oximetry recordings. Results: Pulse oximetry metrics were similar on the two nights. The MOS</p><p>on the two nights showed excellent night-to-night consistency when analyzed as positive for</p><p>OSA versus inconclusive, 143/148 (Spearmans correlation coefficient 0.90). A more detailedanalysis using four categories (MOS 1, 2, 3, and 4) of OSA severity showed very good night-</p><p>to-night agreement, 133/148 (Spearmans correlation coefficient 0.91). Variability was in-creased in children younger than 4 years of age compared to older children. Conclusions:</p><p>Night-to-night consistency of nocturnal pulse oximetry as a diagnostic test for OSA showed</p><p>excellent agreement. Night-to-night consistency of pulse oximetry, as analyzed by the MOS, for</p><p>diagnosis and severity evaluation further validates this abbreviated testing method for pediatric</p><p>OSA. Polysomnography (PSG) is required to rule in or rule out OSA in children if a single night</p><p>oximetry testing is inconclusive. Pediatr Pulmonol. 2012 Wiley Periodicals, Inc.</p><p>Key words: night-to-night variability; pulse oximetry; obstructive sleep apnea.</p><p>Funding source: none reported.</p><p>INTRODUCTION</p><p>Obstructive Sleep Apnea (OSA) in children is a dis-order of breathing characterized by partial and/or inter-mittent complete upper airway obstruction that disruptsnormal ventilation during sleep and normal sleep pat-terns.1 OSA affects about 14% of children.2,3 Themain risk factors are adenotonsillar hypertrophy, obesi-ty, craniofacial anomalies, and neuromuscular dis-ease.3,4 Commonly reported symptoms include snoring,labored breathing, witnessed apneas, disturbed sleep,and daytime neurobehavioral problems. Complicationsinclude neurocognitive impairment, behavioral, cardio-vascular, and metabolic problems.5,6 Nocturnal poly-somnography (PSG) in a sleep laboratory represents thegold standard for the diagnosing of OSA but is costlyand time-consuming.1,7</p><p>Due to these difficulties, some pediatric centers begintesting for OSA with an abbreviated modality such aspulse oximetry or home PSG.8 Pulse oximetry is anincreasingly used abbreviated testing modality for theevaluation of children with suspected OSA.911</p><p>Normative data12 and technical comparison betweendifferent devices have been published.13 Brouilletteet al.9 demonstrated that an abnormal nocturnal pulseoximetry had a 97% positive predictive value (PPV) fordetecting moderate to severe OSA when compared with</p><p>1Respiratory Unit, Department of Pediatrics, Bambino Gesu` Childrens</p><p>Hospital, Piazza S. Onofrio, Rome, Italy.</p><p>2Department of Pediatrics, McGill University and the McGill University</p><p>Health Center Research Institute at the Montreal Childrens Hospital</p><p>Montreal, Canada.</p><p>*Correspondence to: Martino Pavone, MD, Respiratory Unit, Department</p><p>of Pediatrics, Bambino Gesu` Research Institute, Piazza S. Onofrio 4,</p><p>00165 Rome, Italy. E-mail:</p><p>Received 5 June 2012; Revised 28 August 2012; Accepted 29 August</p><p>2012.</p><p>DOI 10.1002/ppul.22685</p><p>Published online in Wiley Online Library</p><p>(</p><p> 2012 Wiley Periodicals, Inc.</p></li><li><p>PSG. However, children with inconclusive oximetrystudies required PSG for definitive testing to rule in andrule out OSA. Nixon et al.14 subsequently developedthe McGill Oximetry Score (MOS) to classify the sever-ity of OSA and plan appropriate perioperativemanagement.</p><p>Night-to-night variability in sleep quality and respira-tory parameters has been previously studied both inchildren1518 and in adults19 undergoing PSG. Thisnight-to-night variability has also been defined as firstnight effect.15,19 Pediatric guidelines suggested that asingle night PSG is sufficient to rule out OSA in chil-dren, but there is need for more studies to confirm thisrecommendation.7 Urschitz et al.20 reported excellenttest-retest correlation for oximetry parameters in 34third grade children drawn from a community sample.However, no previous data have been published aboutthe night-to-night variability of at-home nocturnal pulseoximetry in children evaluated for suspected OSA. Theprimary aims of this study were therefore (1) to exam-ine the night-to-night consistency of pulse oximetry asa diagnostic test for pediatric OSA and (2) to comparethe night-to-night variability of the MOS for OSA se-verity evaluation. A secondary aim was to evaluatenight-to-night variability in individual oximetry metrics.</p><p>METHODS</p><p>We performed a prospective test-retest study compar-ing data from two consecutive nights of at-home noc-turnal pulse oximetry recordings in children beingevaluated for suspected OSA. The study was approvedby the Bambino Gesu` Childrens Hospital ScientificBoard (Rome, Italy) and parents signed an informedconsent. We compared night-to-night consistency for in-dividual oximetry metrics and compared night-to-nightagreement of an analysis system used for diagnosisand severity scoring of OSAthe MOS. Consecutivechildren aged 118 years old referred to our pediatricsleep laboratory in Rome for suspected OSA betweenMarch 2010 and June 2011 were included. We</p><p>excluded children with congenital or genetic abnormali-ties, significant cardiorespiratory disease, neurologicalor neuromuscular conditions, and/or global develop-mental delay. Children whose oximetry recordingswere less than 6 hr duration on each night were alsoexcluded.</p><p>Pulse Oximetry</p><p>Pulse oximetry was performed as described byBrouillette et al.9 and Nixon et al.14 Parents and chil-dren came to our sleep laboratory. Parents were brieflyinstructed on how to perform oximetry testing and com-pleted a questionnaire that included demographic andclinical data including questions regarding their childsusual sleep and breathing pattern. We performed a com-plete physical examination on each child. Weight wasassessed by a digital scale; height by a stadiometer.BMI was calculated from the ratio of weight/height2</p><p>(kg/m2). BMI percentiles were calculated from theWHO standards.21,22</p><p>The parents took the oximeter home, performed thestudy on two consecutive nights, and then returned theoximeter to our sleep laboratory. A motion-resistantpulse oximeter set for a 2-sec averaging time for hemo-globin saturation (SpO2) (RAD 5, Masimo, Irvine, CA)was used. The pulse oximeter used a fixed 7-sec averag-ing time for pulse rate (PR) and stored in memory newSpO2 and PR values every 2 sec.</p><p>Pulse Oximetry Variables</p><p>Saturation and PR data were extracted and analyzedusing Profox Oximetry Software, Version Masimo0706.05D. The program provided the following metrics:Mean oxygen saturation (mean SpO2), lowest SpO2,number of SpO2 dips 4%/hr of study (DI4), mean,minimum, maximum PR and standard deviation of PR,and total effective recording time (TERT). The standarddeviation of the PR was used to estimate PR variability(PRV) over the entire recording.23 Periods of oximetryrecording were excluded from analysis if the oximeterquality signal indicated low signal IQTM (Masimo), lowperfusion, unrecognized, defective or no sensor, inter-ference, or ambient light.</p><p>MOS Classification</p><p>For each night, oximetry recordings were evaluatedfor patterns suggestive of pediatric OSA and for OSAseverity grading. Oximetry recordings that had at leastthree clusters of desaturation events and at least threedips in saturation to less than 90% were regarded asdiagnostic of OSA.9 Recordings that did not meet thesediagnostic criteria were regarded as inconclusive forOSA. This method has been previously shown to have a97% PPV against PSG as the reference standard.9 The</p><p>ABBREVIATIONS:</p><p>OSA obstructive sleep apnea</p><p>PSG polysomnography</p><p>PPV positive predictive value</p><p>MOS McGill oximetry score</p><p>NPV negative predictive value</p><p>BMI body mass index</p><p>SpO2 oxygen saturation</p><p>DI4 number of SpO2 falls 4%/hr of studyPR pulse rate</p><p>PRV pulse rate variability</p><p>TERT total effective recording time</p><p>T&amp;A adenotonsillectomy</p><p>SDB sleep disordered breathing</p><p>2 Pavone et al.</p><p>Pediatric Pulmonology</p></li><li><p>MOS was used to assess severity of OSA as previouslydescribed by Nixon et al.14 MOS of 2, 3, and 4were considered diagnostic for increasing severity ofOSA; a MOS of 1 was considered as inconclusive andunable to rule out OSA.9 Briefly, MOS category 2 hadat least three clusters of desaturation events and at leastthree dips in saturation to less than 90%; MOS category3 had at least three clusters of desaturation eventsand at least four dips in saturation to less than 85%;MOS category 4 had at least three clusters of desatura-tion events and at least four dips in saturation to lessthan 80%.14</p><p>Data Analysis</p><p>Statistical analysis was performed with Statistica 7.0.(StatSoft) and IBM SPSS Statistics 20. Continuous datawere summarized as mean SD. Correlation for con-tinuous oximetry metrics across the two recordingnights were expressed using Pearson correlation coeffi-cients, r. Spearmans correlation coefficient, rho, with95 percentile limits was used to analyze the levels ofagreement between inconclusive (MOS 1) versus diag-nostic (MOS 24) studies and between the differentMOS categories in the first and second night recordings.During data analysis we noted different levels of agree-ment between younger and older children and thereforereport a post-hoc exploratory subgroup analysis by age.For all analyzed parameters, P &lt; 0.05 was consideredstatistically significant.</p><p>RESULTS</p><p>For this study, 211 children were recruited. Sixty-three children were excluded because one (36) or both(27) recordings did not reach the required duration of aminimum of 6 hr per night. Amongst 36 children witha single night sufficient duration recording, 33 childrenhad two inconclusive tests (MOS 1) and three childrenhad a single MOS 2 result. Excluded children did notdiffer statistically from included children on age(4.5 2.7 years vs. 4.9 2.4 years) or gender distribu-tion (59% vs. 65% males). Thus, our study sampleincluded 148 children (96 males) aged 4.9 2.4 (1.211.8) years. Body mass index percentile averaged56.8 34.7. Twenty-six (17.6%) were obese (&gt;95thBMI percentile) and 22 children (14.9%) were at riskfor overweight (8595th BMI percentile). Average noc-turnal pulse oximetry metrics were similar on nights 1and 2 (Table 1). Individual oximetry metrics variedwidely in the degree of correlation across nights(Table 1 and Fig. 1). The frequency of desaturations4%, DI4, was most highly correlated across nights,r 0.95; saturation nadir across nights showed moder-ate correlation, r 0.80; TERT was not closely corre-lated across nights, r 0.23. An example of nocturnal</p><p>pulse oximetry trend graphs from one subject recordedon two consecutive nights is shown (Fig. 2).</p><p>Night-to-Night Agreement of Pulse Oximetry as aDiagnostic Test for OSA</p><p>There was excellent night-to-night agreement foroximetry testing as diagnostic versus inconclusive forOSA, (143 of 148 (97%) subjects), Spearmans correla-tion coefficient 0.90 (0.81,0.99), (Table 2). Of 34children found to have an abnormal oximetry diagnosticof OSA on either night, 32 were found to have an ab-normal oximetry on night 1 and 31 were found to havean abnormal oximetry on night 2. The 34 childrenfound to have abnormal studies on either night weresignificantly younger than the 114 children with twoinconclusive studies, 3.6 1.6 years versus 5.2 2.4years, P &lt; 0.001, t-test. Amongst children less than4 years old, night-to-night agreement was 0.93; 17 chil-dren had positive tests on both nights, 46 children hadan inconclusive test on both nights and five childrenhad a positive test on one of the two nights. Amongstchildren 4 years or older, there were no discrepanciesbetween the two nights; 12 children had abnormalresults on both nights and 68 children had inconclusiveresults on both nights.</p><p>Night-to-Night Agreement of Pulse Oximetry forAssessing Severity of OSA</p><p>There was very good night-to-night agreement (133/148) for oximetry as a method to assess severity ofOSA, Spearmans correlation coefficient 0.91 (0.84,0.98), (Table 2). Amongst 80 children older than4 years, the MOSs were concordant in 78; the MOS</p><p>TABLE 1 Night-to-Night Consistency of Pulse OximetryMetrics</p><p>Oximetry metrics (148 pts)</p><p>First night Second night r</p><p>TERT (hr) 8.3 1.3 8.1 1.1 0.23Mean SpO2 (%) 97.7 1.1 97.8 1.0 0.74Lowest SpO2 (%) 87.7 11.4 88.3 10.2 0.80SpO2 </p></li><li><p>increased from level 3 to level 4 in two children fromnight 1 to night 2. Amongst 68 children 4 years old orless, the MOSs were concordant in 55; nine subjectshad a difference in MOSs of one severity category andfour subjects had a difference of two severity catego-ries. An example of oximetry recordings concordant forresults suggestive of OSA but discordant for OSA se-verity scoring is provided in Figure 2.</p><p>DISCUSSION</p><p>In this study we investigated the night-to-night agree-ment of at-home nocturnal pulse oximetry in childrenwith suspected OSA. When evaluated as a diagnostictest for OSA, the MOS showed excellent night-to-nightconsistency. When evaluated as a severity assessment</p><p>tool, the MOS showed very good night-to-night consis-tency. Individual oximetry metrics showed variable cor-relations but the DI4 was highly correlated acrossnights, r 0.95.</p><p>When evaluated as a diagnostic test for OSA, theMOS showed excellent night-to-night consistency. Asthe two recordings were made on separate nights, somedegree of night-to-night variability would be expected.In our study the first night pulse oximetry identified 32of 148 (21.6%) patients as having OSA. In earlier seriesfrom Montreal the prevalence of oximetry studies diag-nostic for OSA was much the same, 22 and 27%.9,14 Ittherefore seems likely that other pediatric sleep labora-tories would have similar proportions of diagnosticstudies and thereby could avoid the need for PSG inmany of their referrals. Our results suggest that there</p><p>Fig. 1. Oximetry metrics: Degree of correlation across two nights. These graphs demonstrate</p><p>the variable correlation between individual oximetry metrics from night 1, shown on the x</p><p>axes, and night two, shown on the y axes. Filled circles show the 34 children with an oximetry</p><p>diagnostic of OSA; X shows those wi...</p></li></ul>