Use of Oximetry to Determine Need for Adenotonsillectomy for Sleep-Disordered Breathing

Children eligible for participation in this prospective randomized controlled parallel-group study (Chania Community Oximetry-Based Study) (1) were 4 to 10 years old; (2) presented with at least a 6-month history of snoring >3 nights per week to the outpatient ear, nose, and throat (ENT) clinic of a regional hospital in Greece (Chania General Hospital, Chania, Crete); (3) had a tonsillar size >2+; and (4) were considered T/A candidates by an attending ENT surgeon (for exclusion criteria see Supplemental Information).¹⁰ 

Waiting time for T/A at the Chania General Hospital is ∼3 months. Eligible children were recruited and randomly assigned to the following 2 different pairs of oximetry recordings (baseline and 3-month follow-up): (1) an oximetry immediately before T/A and at the 3-month follow-up, which occurred postoperatively (T/A group); or (2) an oximetry at the initial clinic visit and at the end of the usual 3-month waiting period for surgery (control group) (Fig 1). Thus, participants in both groups had baseline (0 month) and follow-up (3 months) oximetry and clinical evaluation.

After recruitment, simple randomization with a 1:1 allocation ratio was performed at the Chania General Hospital by opening sealed envelopes with computer-generated group assignment, which were provided by the coordinating center (University of Athens). The study protocol was approved by the Chania General Hospital Scientific Council (Proceeding 8,May 22, 2013). Written informed consent was obtained from caregivers, and assent was obtained from children 7 years of age or older.

Body weight and height were recorded at baseline and follow-up. Obesity was defined as BMI ≥95th percentile (z score ≥1.65).^11,12 SDB-specific health-related quality of life was assessed by using the Obstructive Sleep Apnea-18 (OSA-18), an 18-item questionnaire, which was filled out by parents during both visits.¹³ Details about clinical evaluation are provided in the Supplemental Information.

Participants underwent continuous oximetry from 8 to 10 pm to 8 am in the ENT ward. To facilitate recognition of major motion artifacts, awakenings were recorded by the nurse and/or parent accompanying the child. A Nonin 7500 pulse oximeter was used with 8000AA sensor applied to 1 finger (Nonin Medical BV, Amsterdam, Netherlands). Nonin PureSAT technology uses an averaging time of <3 seconds and pulse-by-pulse filtering to remove signals caused by motion or low perfusion. Spo₂ recording was downloaded by using the nVISION software (Nonin Medical BV).

Data with motion artifacts or corresponding to awakenings were excluded from tracings. Oximetry was considered technically acceptable if the recording had a duration ≥360 minutes and after removal of motion artifacts and/or awakenings was ≥300 minutes.¹⁴ Number of Spo₂ drops ≥3% or ≥4% per hour of recording (ODI3 or oxygen desaturation [≥4%] of hemoglobin index [ODI4]), baseline Spo₂ (average of Spo₂ readings not included in any saturation drops), and Spo₂ nadir were provided by the software. An MOS of 1 to 4 was assigned by visual inspection of the tracing (See Outcome Measures in Supplemental Information).⁶ For determining MOS, a cluster of desaturation events was defined as 5 or more Spo₂ drops (≥4%) within 10 to 30 minutes.⁵ 

The study was single-blind: (1) Oximetry was scored by 1 investigator at the coordinating center who was blinded to patient randomization status and order of the oximetry study (baseline versus follow-up); (2) investigators who administered the OSA-18 questionnaire were unaware of patient group assignment, and calculation of the total score was completed by a blinded investigator at the coordinating center; (3) data analysis was performed at the coordinating center. Families and patients were not blinded to the randomization assignment.

The primary outcome measures were (1) change in the proportion of subjects with an MOS of 1 (normal or inconclusive oximetry) between follow-up (3 months) and baseline (0 months) visits and (2) the proportion of subjects with oxygenation abnormalities at baseline who improved at follow-up (ie, proportion of subjects who achieved an MOS of 1 at 3 months if they had an MOS >1 at 0 months, and the proportion of subjects who achieved an ODI3 <2 episodes per hour at 3 months, if they had an ODI3 ≥3.5 episodes per hour at 0 months).

Changes in ODI4 and OSA-18 score between follow-up and baseline were secondary outcome measures ([value at follow-up]−[value at baseline]).¹³ Additional secondary outcomes unrelated to the current hypothesis have been included in the study protocol and will be presented in future publications (Supplemental Information).

Assuming a 20% rate of technically nonacceptable oximetry tracings and 20% rate of participants that would be lost at follow-up or discontinue their participation, the target sample size of 188 recruited subjects was calculated (Supplemental Information). Participants in the 2 study groups were compared regarding primary outcome measures by using the χ² test. Odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were computed by logistic regression. Multivariable logistic regression analysis was applied to adjust ORs for the effects of baseline MOS or ODI3, age, sex, and presence of obesity. The number of children with elevated ODI3 that underwent T/A to prevent persistence of ODI3 ≥2 episodes per hour in 1 patient at follow-up was calculated (number needed to treat).

The 2 groups were compared in terms of ODI4 change by Mann–Whitney U test and regarding OSA-18 score change by Student’s t test. The relationship between study group and change in OSA-18 score was adjusted for other factors by using a general linear model that incorporated age, sex, and presence of obesity as covariates. The interaction between SDB severity and group assignment regarding the effect on OSA-18 change was also tested.

Between June 2013 and April 2016, families of 196 eligible children were approached for recruitment to the study. Of those, 93 subjects were randomly assigned to the T/A group and 93 to the control group (Fig 2). Follow-up evaluation of the last participant was completed in August 2016. None of the participants in the control group required T/A on an urgent basis while on the surgical waiting list, and there were no major adverse events.

The frequency of technically nonacceptable oximetry at baseline or follow-up was 22.6% in the T/A group and 15.1% in the control group. The reason of nonacceptable oximetry was no sensor contact or poor contact with the child’s finger. A total of 4.3% of children randomly assigned to the T/A group and 7.5% of those assigned to the control group did not undergo follow-up oximetry because they did not return for evaluation (Fig 2).

A total of 140 children (68 in the T/A group and 72 in the control group) had technically acceptable oximetry tracings both at baseline and at follow-up and were included in the primary outcome measures analysis. The 2 study groups were similar in demographic characteristics, tonsillar size, somatometrics, OSA-18 scores, and oximetry indices (Tables 1 and 2).

Children randomly assigned to the T/A group with technically acceptable baseline oximetry who completed (n = 68) or who did not complete (n = 14) the study did not differ significantly in the frequency of an MOS of 1 or proportion of subjects with ODI3 ≥3.5 episodes per hour. Similarly, there were no significant differences regarding oxygenation indices between children in the control group with acceptable baseline oximetry who completed (n = 72) or did not complete (n = 14) the study.

The T/A and control groups were not different regarding change in the proportion of subjects with an MOS of 1 between the follow-up and baseline visits (P = .54; Table 3). OR (95% CI) for achieving an MOS of 1 at follow-up in the T/A relative to the control group was 1.3 (0.5–3.3). The absence of an association persisted after adjustment for baseline MOS, age, sex, and presence of obesity (Supplemental Table 5).

Changes in the MOS between follow-up and baseline visits according to study group are presented in Supplemental Tables 6 and 7. Over the 3-month study period, 12 of 17 (70.6%) children with an MOS >1 in the T/A group and 10 of 21 (47.6%) children with an MOS >1 in the control group achieved an MOS of 1 at follow-up (P = .14; OR 2.6 [0.7–10.2]; adjusted OR 2.5 [0.6–10.7]) (Supplemental Table 8).

Individual changes in ODI3 for all subjects in the 2 groups are shown in Fig 3 (details in Supplemental Fig 5). The median (quartile 25, quartile 75) ODI3 change between follow-up and baseline was −3.2 (−6.5, −1.7) episodes per hour in the T/A group and −1.7 (−5.7, 0.9) episodes per hour in the control group.

Thirty-two of 68 participants in the T/A group and 38 of 72 subjects in the control group had ODI3 ≥3.5 episodes per hour at baseline. Children with elevated ODI3 in the 2 groups did not differ in baseline characteristics (Supplemental Table 9). Significantly more children in the T/A group than in the control group with an elevated ODI3 at baseline had an ODI3 <2 episodes per hour at follow-up (OR 14.0 [95% CI 2.9 to 68.4]; P < .001; Table 4). Subjects in the T/A group were significantly more likely to normalize their ODI3 at follow-up after considering ODI3 at baseline, age, sex, and presence of obesity (P < .01; Supplemental Table 10). Three children with elevated ODI3 underwent T/A to prevent an ODI3 ≥2 episodes per hour in 1 child at the 3-month follow-up. Comparisons of the 2 study groups regarding changes in ODI4 between follow-up and baseline are presented in the Supplemental Information.

Children in the T/A group had significantly larger improvement in OSA-18 total score between 3 and 0 months than participants in the control group (−32.4 ± 16.9 vs −0.8 ± 9.1; P < .001; Fig 4). The association between study group and change in the OSA-18 score remained significant after adjustment for age, sex, and presence of obesity (P < .01; Supplemental Table 11). There was no interaction between SDB severity (MOS >1 versus MOS of 1 or ODI3 ≥3.5 episodes per hour versus ODI3 <3.5 episodes per hour) and treatment group regarding change in OSA-18 (P = .68 and P = .20, respectively).

The present randomized controlled study is the first to reveal that ODI3 from nocturnal oximetry performed in a community setting can be used to identify otherwise healthy children with snoring and tonsillar hypertrophy who are likely to have resolution of hypoxemia related to SDB after T/A. A preoperative ODI3 ≥3.5 episodes per hour was associated with attainment of normal ODI3 in ∼40% of patients, postoperatively. In contrast, an MOS >1 was not responsive to T/A, because the frequency of a normal and/or inconclusive score (MOS = 1) increased similarly in both the T/A and control groups at the end of the follow-up period. Disease-specific, health-related quality of life improved significantly more in the T/A than in the control group.

Preoperative objective assessment of SDB severity is important.³ Moderate-to-severe SDB (ie, AHI >5 episodes per hour) diagnosed by polysomnography in the presence of adenotonsillar hypertrophy is an indication for T/A.³ This recommendation is supported by results of the Childhood Adenotonsillectomy Trial (CHAT), in which the efficacy of T/A has been assessed by polysomnography.¹⁵ For every 3 children with AHI >4.7 episodes per hour undergoing surgery, persistence of abnormal AHI was prevented in 1 child, whereas frequency of SDB resolution without treatment was only 26%.^15,16 Because of the high spontaneous resolution frequency of mild SDB (AHI: 2–5 episodes per hour), 5 children with mild disease underwent T/A to prevent persistence of abnormal AHI in 1 child.

Additionally, prevalence of major respiratory complications in the post-T/A period increases with severity of SDB as measured by polysomnography.¹⁷ Because the availability of polysomnography is limited and more so in community settings, it has been suggested that children at risk for SDB should be offered at least 1 of the alternative diagnostic methods.³ Nocturnal oximetry is such an economical substitute. An MOS of 1 is associated with low risk of major respiratory complications postoperatively, whereas an MOS of 4 predicts high risk of complications requiring airway intervention, which can be reduced by limiting the total dose of administered opioids for pain.^18,19 

The response of abnormal oximetry parameters to T/A has not been defined.^20,21 In healthy 4- to 10-year-old children, the 90th and 97.7th percentile values for ODI3 are 1.2 and 2.04 episodes per hour, respectively.^9,22 Saito et al⁹ have proposed that an ODI3 cutoff value of 3.5 episodes per hour, which corresponds to 2 SDs above the 97.7th percentile or 4 SDs above the mean ODI3 value in asymptomatic children, allows the detection of SDB-related hypoxemia, which is responsive to T/A. In the current study, 3 children with ODI3 ≥3.5 episodes per hour had surgery to prevent the persistence of abnormal ODI3 (≥2 episodes per hour) in 1 child postoperatively, a surgical success rate similar to that reported for patients with moderate-to-severe SDB (AHI >4.7 episodes per hour) in the CHAT.¹⁵ Hence, when oximetry is used as a diagnostic tool, an ODI3 ≥3.5 episodes per hour defines a subgroup of patients with SDB who respond to T/A as favorably as those with an AHI >5 episodes per hour.

In this study, MOS was not responsive to T/A. Although an MOS >1 predicts AHI >5 episodes per hour in >90% of cases, one-third of children with clinical suspicion of SDB and an MOS of 1 also have an AHI >5 episodes per hour.⁷ Similarly to an ODI3 ≥3.5 episodes per hour, an MOS of 1 can also be abnormal because it frequently incorporates oximetry tracings revealing hypoxemia (eg, those with only 2 clusters of desaturation events or fewer than 3 Spo₂ drops <90%).^5,6 In a study by Nixon et al,⁶ an MOS of 1 corresponded to a mean AHI of 4.1 episodes per hour and was characterized as a “normal/inconclusive” score for excluding SDB. Misclassification of SDB severity when using MOS probably explains why the 2 study groups did not differ in the frequency of achieving a normal or inconclusive score. In summary, an ODI3 ≥3.5 episodes per hour can be used to recognize cases of SDB with hypoxemia responsive to T/A, whereas an MOS >1 identifies those children who should be monitored in the hospital overnight post-T/A.¹⁸ 

Health-related quality of life as measured by an OSA-18 score was similar in both participants of the CHAT and of the current study and was affected in a mild-to-moderate degree.²³ In both studies, improvement in quality of life was not predicted by preoperative SDB severity.²⁴ The improvement in OSA-18 contrasts with the response of polysomnography indices to T/A, which may remain abnormal in an appreciable proportion of children postoperatively.²⁵ Therefore, impaired quality of life is an indication for treatment regardless of the severity of nocturnal hypoxemia.³ 

The usual waiting time for T/A at Chania General Hospital is ∼3 months, and surgery has not been delayed for the purposes of this study. T/A has been the preferred treatment of SDB in all participants because tonsillectomy alone may not suffice because of residual adenoid tissue increasing upper airway resistance.²⁶ Participants did not have a trial of nasal corticosteroid before proceeding to surgery because no improvement in SDB related to tonsillar hypertrophy was expected.²⁷ 

A study limitation is the appreciable frequency of technically nonacceptable oximetry tracings, which could decrease by using disposable (adhesive) sensors rather than reusable finger clips for cost reduction. Another limitation is that oximetry has been conducted in hospital, but the same evaluation could be completed at the patient’s home. Although the analysis based on ODI3 ≥3.5 episodes per hour was performed in subsets of the T/A and control groups, subjects of the former group were significantly more likely to normalize their ODI3 at follow-up even after adjustment for factors potentially affecting efficacy of T/A.

The use of a control group minimized improvement in abnormal ODI3 or MOS because of the “regression toward the mean” phenomenon. Improvement in the outcome measures in the control group provides an estimate of changes attributed to regression toward the mean and to placebo effect. Of note, this study was not designed to recognize subgroups of patients at risk for residual nocturnal hypoxemia post-T/A who would require repeat oximetry.

The present randomized controlled study reveals that nocturnal oximetry can be used in community settings to identify children with snoring and tonsillar hypertrophy who are likely to have resolution of nocturnal intermittent hypoxemia after undergoing T/A. An ODI3 that is not elevated does not necessarily exclude the need for treatment because impaired quality of life improves post-T/A regardless of the severity of nocturnal hypoxemia and obstructive apneas or hypopneas are not always accompanied by Spo₂ drops.

AHI

apnea-hypopnea index

CHAT

Childhood Adenotonsillectomy Trial

CI

confidence interval

ENT

ear, nose, and throat

MOS

McGill oximetry score

ODI3

oxygen desaturation (≥3%) of hemoglobin index

ODI4

oxygen desaturation (≥4%) of hemoglobin index

OR

odds ratio

OSA-18

Obstructive Sleep Apnea-18

SDB

sleep-disordered breathing

SpO₂

oxygen saturation of hemoglobin measured by pulse oximetry

T/A

adenotonsillectomy