Comparison of WatchPAT to polysomnography measurement of apnea-hypopnea index and obstructive sleep apnea severity in adolescents with overweight and obesity

Abstract

Study Objectives:

The prevalence of obstructive sleep apnea (OSA) increases dramatically in adolescents with overweight or obesity. The gold standard for diagnosis of OSA is in-laboratory polysomnography (PSG). However, access to PSG can be challenging, necessitating development of alternative devices. This study aimed to compare the accuracy of WatchPAT, a device that indirectly detects apnea and hypopnea events through peripheral arterial tonometry, to PSG in adolescents with overweight and obesity.

Methods:

Thirty-eight participants (15.5 ± 1.6 years; 76% female; 40% Hispanic/Latino; body mass index 97.7 ± 2.0 percentile) were analyzed. Correlation, agreement, sensitivity, specificity, and concordance between WatchPAT with PSG-derived measures of apnea-hypopnea index (AHI) and OSA severity were conducted. A subset (n = 21) underwent self-reported assessment of nasal flow sensor reliability and video characteristics to evaluate possible underestimation of PSG-assessed AHI.

Results:

Mean bias between WatchPAT- and PSG-derived AHI was 16.9 ± 13.4 events/h (95% confidence interval 12.5–21.3). WatchPAT overestimated OSA severity compared to PSG in 89% of participants. Sensitivity was 62–100% and specificity was 5.5–32% across all OSA severity levels. Of the subset with self-reported assessment of PSG data, 38% had nasal flow sensor reliability < 75%, 14% had underestimated AHI, and 28% had probably underestimated AHI. However, these characteristics did not fully explain the discrepancy between WatchPAT and PSG measurements.

Conclusions:

The WatchPAT device showed a significant discrepancy compared to the gold-standard PSG in measurement of AHI and accuracy of OSA severity in adolescents with overweight or obesity. Future research is needed to understand pathophysiological differences to enhance assessment of OSA in this high-risk population.

Citation:

Dobbin ML, Bernard A, Pyle L, et al. Comparison of WatchPAT to polysomnography measurement of apnea-hypopnea index and obstructive sleep apnea severity in adolescents with overweight and obesity. J Clin Sleep Med. 2025;21(4):695–702.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Polysomnography is the gold-standard for diagnosing of obstructive sleep apnea, yet access is limited, highlighting the need for alternative measurement options. WatchPAT indirectly detects apnea and hypopnea events through peripheral arterial tonometry and has been explored as an alternative diagnostic tool, but research is limited on its use in adolescents with overweight and obesity despite an increased prevalence of obstructive sleep apnea and related comorbidities in this population.

Study Impact: The aim of the current study was to compare the performance of WatchPAT for assessing apnea-hypopnea index and determining of obstructive sleep apnea severity in a sample of adolescents with overweight and obesity. Findings revealed a notable discrepancy in accuracy and agreement of WatchPAT compared to polysomnography.

INTRODUCTION

The prevalence of obstructive sleep apnea (OSA), a sleep disorder characterized by apneic and hypopneic episodes associated with oxygen desaturations and arousals from sleep, is estimated at 1–3% in adolescents of healthy weight. However, this figure dramatically rises to estimates of 13–61% in adolescents with overweight or obesity.^1–3 Untreated OSA is associated with significant cardiometabolic morbidity as well as cognitive and behavioral impairment. Therefore, it is imperative to make accurate and timely diagnosis of OSA in adolescents, particularly in high-risk individuals such as those with overweight or obesity.^4–6

The gold standard for diagnosing OSA is in-laboratory polysomnography (PSG).^4,7 However, access to PSG may be challenging due to factors such as cost, geographic access to pediatric sleep laboratories, and extended wait times caused by a shortage of sleep beds. Additionally, PSG requires highly trained personnel to staff, score, and interpret the study. Finally, the multiple sensors used in PSG may pose difficulty for youth to tolerate. Consequently, alternative diagnostic devices for OSA have been developed and investigated. While a variety of home sleep test devices are currently on the market, the American Academy of Sleep Medicine has advised against the use of at-home testing for diagnosis of OSA in youth, in part, due to limited validation studies in this group.³

The WatchPAT (Itimar Medical Ltd, Caesarea, Israel) is a wristwatch-like device equipped with a finger probe and chest monitor designed to indirectly detect apnea and hypopnea events through peripheral arterial tonometry.^8–10 While WatchPAT is approved for use in individuals 12 and older by the Food and Drug Administration, the existing literature presents limited evidence regarding its validity compared to PSG in adolescents.^11–13

Despite the increased risk for OSA in adolescents with overweight and obesity, prior studies include a minority of participants meeting criteria for overweight and obesity. Positive associations have been investigated between adolescents with obesity and endothelial and autonomic dysfunction, finding that metabolic dysfunction not only plays a role in increasing risk of cardiovascular disease but also autonomic imbalance.^14,15 Sympathetic predominance in particular poses a significant concern, as sympathetic abnormalities may interfere with the effectiveness of the existing algorithm used in the WatchPAT system.¹⁰ This raises a concern for its suitability in the adolescent population, particularly among those with an elevated body mass index (BMI).

Given the previous mixed findings and limited emphasis on high-risk adolescents, the objective of the present study was to assess the accuracy and agreement of the WatchPAT in comparison to the gold standard in-laboratory PSG for assessing apnea-hypopnea index (AHI) and identifying OSA severity in adolescents with overweight or obesity. We hypothesized that when compared to PSG, WatchPAT would provide comparable assessments of AHI and accurately categorize participants by OSA severity.

METHODS

Participants were drawn from 2 distinct sources: (1) a retrospective chart review of adolescents who had completed a research study utilizing WatchPAT and who had undergone a clinical PSG within 18 months of study completion, and (2) a prospective study in which participants underwent an in-laboratory PSG while simultaneously wearing WatchPAT. All participants were clinically referred for a PSG due to concern for possible OSA.

Retrospective chart review

A chart review was conducted for participants aged 10–20 years with a BMI ≥ 85th percentile for age and sex who had completed 1 of 4 research studies at Children’s Hospital Colorado where WatchPAT was completed as part of the research protocol (NCT03919929, NCT03717935, NCT03041129, NCT02157974). Data for the current analysis were included if participants underwent a clinical PSG at the Children’s Hospital Colorado Sleep Laboratory within 18 months of their research study participation. Data from the scored WatchPAT and PSG studies were extracted from the research and medical record for analysis. All studies were approved by the Colorado Multiple Institutional Review Board, and all participants and their parent/guardian signed informed consent/assent as appropriate before study participation.

Prospective WatchPAT study

Patients scheduled for a clinical PSG at the Children’s Hospital Colorado Sleep Laboratory between the ages of 13 and 21 years with BMI ≥ 85th percentile for age and sex were recruited for the prospective study. Exclusion criteria included: medical conditions that would affect tonometer readings, such as peripheral vascular disease, systemic hypertension, cardiac disease, severe lung disease, neuromuscular disease, epilepsy, and sickle cell crisis, or treatment with medication that commonly affect readings. Approval for the study was granted by the Colorado Multiple Institutional Review Board. Research staff contacted families ahead of their scheduled PSG to provide information about the study. Upon agreement to participate, informed consent (adolescents > 18 years or parents/guardians of adolescents < 18 years) and assent (adolescents < 18 years) were obtained prior to initiating study procedures. Participants underwent PSG per usual clinical procedure while simultaneously wearing WatchPAT.

Measures

WatchPAT

The WatchPAT is a wristwatch-like device comprising a finger probe and a chest monitor with 3 points of contact that utilizes peripheral arterial tonometry, actigraphy, and pulse oximetry to capture data for estimating sleep and respiratory disturbances. Peripheral arterial tonometry signals are generated from the transient changes in sympathetic tone in vascular smooth muscle that is detected by finger plethysmograph using arterial volume changes and increases in heart rate that result from the sympathetically-driven response of a hypopneic or apneic episode.⁸ Following recording, the zzzPAT software (Itimar Medical Ltd, Caesarea, Israel), which employs a proprietary WatchPAT algorithm, determines respiratory events as defined by 3% or 4% pulse oximetry desaturation thresholds. The AHI, an assessment of the number of apneas and hypopneas occurring per hour of sleep using a 3% desaturation threshold, served as the primary variable in the current analysis, following clinical convention as the gold standard for diagnosing OSA.¹⁶ Additional sleep characteristics collected include oxygen desaturation index (the average number of oxygen desaturation episodes > 3% and > 10-second duration per hour of sleep time), percent time spent in rapid eye movement sleep, light sleep, and deep sleep, and total sleep time (TST).

The current study utilized both the WatchPAT 200 and 300 models, both of which are compatible with the zzzPAT software. The WatchPAT200 was discontinued in early 2020 upon the 300 model’s release. Data from the oximetry channels are collected from either the built-in pulse oximeter sensor present in the WatchPAT200 or the uPAT probe in the WatchPAT300 model. Raw data generated by WatchPAT was analyzed by the proprietary scoring algorithm.

PSG

PSG was conducted following standard care procedures and scored and interpreted by a board-certified sleep physician in accordance with the American Academy of Sleep Medicine guidelines.¹⁷ Data extracted from the scored PSG report included AHI based on a 3% oxygen desaturation threshold as well oxygen desaturation index, percent time spent in rapid eye movement (REM), and non-REM stage 1 and 2 (light sleep), non-REM stage 3 (deep sleep), and TST. Coauthor N.R.F. also reviewed the PSG data and video characteristics and coded nasal pressure flow sensor reliability as < 25%, 50%, < 75%, and > 75%. Breathing characteristics from the video were coded as presence of normal breathing, snoring, increased work of breathing, and/or retractions. Based on the self-reported quality of measurements, each study was coded as either accurate, possibly underestimated, or underestimated concerning the AHI from PSG.

Demographic characteristics

Demographic information was obtained separately for the prospective study and retrospective study. For the participants in the prospective research study, a medical chart review was conducted at the time of their PSG appointment, and the demographics reported nearest to the date of their PSG appointment were recorded. For the participants whose retrospective data was obtained, demographic characteristics were collected during their participation in the research study, and medical comorbidities were obtained via chart review at the time of data extraction. Weight classes were defined as follows: overweight (85th ≤ BMI < 95th percentile) and obesity (BMI ≥ 95th percentile).

Statistical analysis

Summary statistics for PSG and WatchPAT are presented as mean ± standard deviation or N (%). An independent samples t test examined differences between studies completed with WatchPAT200 and WatchPAT300 devices, and a paired samples t test compared differences in BMI percentile on the date of the WatchPAT and PSG studies for the retrospective study only. To assess the agreement of AHI obtained by WatchPAT compared to PSG, a Bland-Altman plot was generated. The Bland-Altman plot provides an estimate of bias, which is used to evaluate the level of agreement between 2 measurement methods. Higher degrees of bias indicate less agreement.^18,19 The tolerated level of bias for a measurement in comparison to a gold standard is self-reported and depends on the clinical context. For this analysis, we a priori determined that a difference of 10 points would be clinically significant as it could alter the classification of OSA severity.

To evaluate the accuracy of WatchPAT in categorizing OSA severity compared to PSG, sensitivity, specificity, and area under the curve (AUC) values were calculated for each severity level with a 95% confidence interval. In the current study, severity categories were defined as mild (5–15 events/h), moderate (> 15–30 events/h), and severe (> 30 events/h), as used in adults.²⁰ These criteria were used because the participants were adult-sized and physically mature.^16,20,21

To assess the concordance between WatchPAT and PSG, a contingency table was created.¹⁹ The contingency table compared the OSA severity categories as determined by each method and ascertained the frequency of agreement at each severity level. Concordance was then determined by dividing the frequency of agreement by the total number of participants, with higher values indicating greater concordance. Statistical analyses were computed using the Statistical Package for the Social Sciences and R version 4.2 (R Core Team, Vienna, Austria).

RESULTS

Participants

A total of 23 participants were identified via retrospective chart review with WatchPAT from a previous research study and a clinical PSG completed within 18 months of study participation. The average time interval between the research WatchPAT and the clinical PSG was on average 4.68 ± 5.19 months. There were no statistically significant differences in BMI percentile as calculated on the date of the WatchPAT and the PSG (98.21 ± 1.86 and 98.43 ± 1.42, respectively; P > .05). All participants in the retrospective study were assessed using the WatchPAT200.

An additional 22 participants consented to the prospective study; of these, 7 participants were excluded from analysis due to missing data. Missing data were due to WatchPAT failure (not recording any data or recording stopping after a few hours; n = 2), technician errors (not turning on the device; n = 3), or missing or poor quality PSG data that precluded accurate interpretation (n = 2). Three participants in the prospective study were assessed with the WatchPAT 200 while the remaining 12 were studied with the WatchPAT300. There were no significant differences in any demographic or WatchPAT or PSG variables between those who had their study on different models of the device (all P > .05).

The final pool for data analysis comprised 38 participants. Their characteristics are presented in Table 1. There were no statistically significant differences between the retrospective or prospective groups in their demographic or medical characteristics (all P > .05); therefore, all participants were analyzed together.

Table 1.

Participant characteristics.

Variable	Total Sample (n = 38)	Retrospective Sample (n = 23)	Prospective Sample (n = 15)
Sex, n (%)
Male	9 (24)	0 (0)	9 (60)
Female	29 (76)	23 (100)	6 (40)
Age, years, mean ± SD	15.5 ± 1.63	15.5 ± 1.50	15.5 ± 1.82
Race, n (%)
White	27 (71.1)	15 (65.2)	12 (80)
Black/African American	3 (7.9)	2 (8.7)	1 (6.7)
Asian	1 (2.6)	1 (4.3)	0 (0)
Other/not reported	7 (18.4)	5 (21.7)	2 (13.3)
Ethnicity, n (%)
Hispanic or Latino	15 (39.5)	12 (52.2)	3 (20)
Not Hispanic or Latino	21 (55.3)	10 (43.5)	11 (73.3)
Not reported	2 (5.3)	1 (4.3)	1 (6.7)
BMI percentile, PSG, mean ± SD	97.7 ± 2.03	98.21 ± 1.86	97.16 ± 2.73
BMI percentile, WP, mean ± SD	97.9 ± 2.12	98.43 ± 1.42	97.16 ± 2.73
BMI category, n (%)
Overweight	3 (7.9)	1 (4.3)	2 (13.3)
Obese	35 (92.1)	22 (95.6)	13 (86.7)
Medical comorbidities, n (%)
Polycystic ovary syndrome	23 (60.5)	22 (95.6)	1 (6.7)
Prediabetes	6 (15.8)	5 (21.7)	1 (6.7)
Type 2 diabetes	3 (7.9)	3 (13.0)	0 (0)

Weight classification is based on the CDC’s definition for BMI percentile for adolescents. Comorbidities are not mutually exclusive, and some participants were diagnosed with more than 1 comorbidity. Independent-samples t tests revealed no statistically significant differences between demographic characteristics for participants in the retrospective and prospective cohorts (all P > .05). Paired samples t tests found no statistically significant differences in BMI measured at the PSG and WatchPAT assessments for the retrospective cohort (P > .05). BMI = body mass index, CDC = Centers for Disease Control and Prevention, PSG = polysomnography, SD = standard deviation, WP = WatchPAT.

Sleep characteristics

Sleep characteristics for the PSG and WatchPAT are described in Table 2. Notably, the mean AHI recorded by the WatchPAT device was approximately 3-fold greater than that derived from PSG. Paired samples t tests displayed in Table 2 compared differences between PSG and WatchPAT data. Significant differences were found for AHI (P < .001), rapid eye movement (P < .001), light sleep (P < .001), deep sleep (P = .021), and TST (P = .002). AHI, percent time spent in rapid eye movement and deep sleep, and TST were higher when measured with WatchPAT compared to PSG, while percent of time spent in light sleep was lower when measured with WatchPAT compared to PSG. Oxygen desaturation index was comparable between the 2 devices.

Table 2.

Descriptive sleep characteristics from polysomnography and WatchPAT.

	PSG	WatchPAT	95% CI	P
	M ± SD	M ± SD
Apnea-hypopnea index	10.09 ± 17.37	27.00 ± 18.37	−21.31 to −12.50	< .001
Oxygen desaturation index	15.26 ± 20.56	14.13 ± 15.79	−2.94 to 4.51	.672
% REM sleep	13.00 ± 7.00	22.00 ± 8.00	−0.12 to −0.05	< .001
% Light sleep (stage N1 and N2)	68.00 ± 8.00	57.00 ± 9.00	0.07 to 0.15	< .001
% Deep sleep (stage N3)	18.00 ± 7.00	22.00 ± 7.00	−0.07 to −0.006	.021
Total sleep time (hours)	5.31 ± 1.87	6.49 ± 1.31	−113.27 to −28.5	.002

CI and P values represent differences between PSG and WatchPAT values from t tests. CI = confidence interval, M = mean, N1 = non-rapid eye movement stage 1 sleep, N2 = non-rapid eye movement stage 2 sleep, N3 = non-rapid eye movement stage 3 sleep, PSG = polysomnography, REM = rapid eye movement, SD = standard deviation.

AHI agreement between WatchPAT and PSG

The agreement of AHI between WatchPAT and PSG was evaluated using a Bland-Altman plot, depicted in Figure 1. The mean bias between the 2 methods was 16.92 ± 13.4 events/h (95% confidence interval 12.5–21.3), surpassing the predetermined acceptable difference of 10, indicating a bias greater than acceptable. The limits of agreement, representing 2 standard deviations away from the mean bias, signify where 95% of differences between the measurement methods fall. Ideally, narrow limits suggest high precision. However, in this study, the limits of agreement ranged from −9.34 to 43.18 events/h, reflecting a relatively wide range. Notably, increased bias was observed at higher and lower levels of AHI.

Figure 1. Bland-Altman plot comparing WatchPAT and polysomnography.

Estimated mean difference 16.92 ± 13.4 events/h (95% Cl 12.5–21.3). The greatest bias occurs at the highest and lowest mean AHI. The upper limit of agreement is 43.18 and the lower limit of agreement is −9.34. AHI = apnea-hypopnea index, CI = confidence interval.

Accuracy of OSA severity between WatchPAT and PSG

Table 3 presents the accuracy of WatchPAT compared to PSG at different OSA severity cutoffs. Sensitivity ranged from 62–100% across all severity levels, while specificity varied between 5.5% and 32%. The AUC was 0.53 and 0.66 for mild and moderate severities, respectively. Specificity and AUC were not calculated for severe OSA (AHI ≥ 30 events/h) due to lack of observations.

Table 3.

Sensitivity, specificity, and AUC for OSA severity classifications from WatchPAT compared to PSG.

OSA Severity	Sensitivity	Specificity	AUC
Mild (AHI 5–15 events/h)	1.0	0.06	0.53
Moderate (AHI > 15–30 events/h)	1.0	0.32	0.66
Severe (AHI > 30 events/h)	0.62	1.0	N/A

Higher values indicate better sensitivity and specificity; an AUC above 0.80 indicates diagnostic accuracy. AHI = apnea-hypopnea index, AUC = area under the curve, N/A = not applicable, OSA = obstructive sleep apnea, PSG = polysomnography.

Concordance of OSA severity between WatchPAT and PSG

Table 4 illustrates the concordance between WatchPAT and PSG in estimating OSA severity. WatchPAT matched the same OSA severity as PSG for 4 out of 38 participants (10.5%). These instances include 1 participant in each severity level classification (no OSA, mild, moderate, and severe). In the remaining 34 participants (89.5%), WatchPAT overestimated OSA severity compared to PSG.

Table 4.

Concordance of OSA severity classifications from WatchPAT and polysomnography.

		Polysomnography				Total
		AHI < 5 events/h	AHI ≥ 5 events/h	AHI ≥ 15 events/h	AHI ≥ 30 events/h
WatchPAT	AHI 0–5 events/h	1	0	0	0	1
	AHI > 5–15 events/h	8	1	0	0	9
	AHI > 15–30 events/h	5	7	1	0	13
	AHI > 30 events/h	4	4	6	1	15
Total		18	12	7	1	38

AHI = apnea-hypopnea index, OSA = obstructive sleep apnea.

Reliability of PSG

Twenty-one PSG studies were available and coded based on PSG video characteristics and nasal flow sensor reliability. Among these, 8 participants (38%) had flow sensor reliability less than 75%. Increased work of breathing was observed in 5 participants (24%), and retractions were noted in 2 participants (9.5%). Three studies (14%) were identified as underestimates of AHI and another 6 (28.5%) were considered possible underestimates. Notably, WatchPAT AHI was higher than PSG AHI in 8 out of the 9 studies identified as underestimated or likely underestimated. The characteristics of the underestimated studies compared to WatchPAT are presented in Table S1 in the supplemental material.

DISCUSSION

Contrary to our initial hypothesis, our study revealed a significant discrepancy in agreement, accuracy, and concordance of WatchPAT compared to PSG in the current sample of adolescents with overweight and obesity. The mean AHI recorded by the WatchPAT device was approximately 3-fold greater than that obtained from PSG. Statistically significant differences between PSG and WatchPAT were seen in AHI and other sleep characteristics, such as percentage of time spent in sleep stages and TST. Previously, few studies examined the utility of WatchPAT in this high-risk demographic. Our results fill an important research gap and suggest a limited utility of WatchPAT in this patient population. The importance of studying adolescents with overweight and obesity is underscored by the increased prevalence of OSA for this group and the significant health risks associated with untreated OSA.

Bland-Altman plot analysis found poor agreement of AHI as determined by WatchPAT compared to PSG. The mean bias in the current sample was 2–16 times greater than that observed in other studies comparing WatchPAT and PSG.^8,12,13 While some studies have found high correlation and good agreement of AHI between WatchPAT and PSG, the populations studied were adults or adolescents both with and without obesity.^12,13,22,23 For example, 1 prior study found a high correlation and good agreement with mean AHI as well as good concordance for determining OSA severity between WatchPAT and PSG in a sample of adolescents without obesity (n = 38, mean age = 15.1 ± 1.4 years).¹² A study by Tanphaichitr et al done of children of mixed weight status (n = 36, mean age 10.2 ± 1.8 years) reported excellent WatchPAT agreement (intraclass correlation coefficient = .89) and moderate to good correlation for AHI using linear regression (r = .64). However, the study also found WatchPAT-determined AHI of 10 events/h was solely statistically significant for diagnostic capability (AUC = 0.82).¹³

In line with our findings, a 2022 meta-analysis of studies performed in adults reported poor agreement between WatchPAT and PSG, with significant misclassification by WatchPAT for adults at mild and moderate OSA severity classes.²⁴ Similarly, the study by Tanphaichitr et al reported consistent discrepancy of WatchPAT underreporting AHI events, represented by linear regression analysis.¹³ It is worth noting a distinction in our results as we found the greatest degree of misclassification occurring at the extreme ends of AHI values.

There is no universally accepted threshold for an acceptable degree of bias as indicated by a Bland-Altman plot. The determination of an acceptable bias depends on the study’s clinical context and its population. In this study, we pre-established a bias limit of 10 events/h as an a priori value, considering that a difference of 10 may represent a clinically significant difference in severity that could have implications for treatment recommendations.

Our findings revealed high levels of sensitivity and low levels of specificity for WatchPAT in this sample of adolescents with overweight and obesity. While other studies similarly reported high sensitivity, the specificities observed in those studies differed significantly from the current study. For our sample, the specificity for categorizing AHI ≥ 5 events/h was 19 times lower than the specificity reported by Choi et al in their sample of adolescents with and without obesity.¹² Similarly, our results found a specificity 2.5 times lower for AHI ≥ 15 events/h compared to a study of youth aged 10–15 years.¹³ Overall, our study found better sensitivities were observed in the mild and moderate groups of OSA severities while there was better specificity for the severe range.

In the present study, AUC for all AHI severity cutoffs fell below the 0.8 threshold for diagnostic accuracy. In comparison, other studies found AUCs ranging from 0.82–0.94 for WatchPAT compared to PSG in youth aged 10–15 years with and without overweight and obesity and in adults with chronic obstructive pulmonary disease (mean BMI = 28 ± 7).^13,22 It is likely that more studies will be required to determine the true diagnostic accuracy of WatchPAT in this population, however this data indicates cautions should be used when using WatchPAT as a diagnostic tool for OSA in adolescents with overweight and obesity.

The agreement of AHI severity between WatchPAT and PSG was further assessed using a contingency table. In all but 4 participants, WatchPAT overestimated the severity of OSA compared to PSG. The pattern of misclassification of OSA severity by WatchPAT has been reported in other studies. For instance, in adults with chronic obstructive pulmonary disease, WatchPAT overestimated severities for those with mild OSA while underestimating severity for those with severe OSA.²² In youth with and without overweight and obesity, WatchPAT underestimated the severity of OSA.¹³

Unique from other studies, we also considered whether PSG characteristics could explain the discrepancy between PSG and WatchPAT in the current sample. Recognizing that not all respiratory events are captured solely with oxyhemoglobin desaturation by oximetry, we considered the relevance of flow sensor reliability. Notably, most participants assessed had adequate flow sensor reliability. Three participants were noted by physician review to have had an underestimate of AHI on PSG per video characteristics based on the presence of increased work of breathing and/or retractions. While these factors could potentially impact comparisons with WatchPAT, they represent only a small percentage of the overall sample, making it unlikely that they explain all discrepancies. Therefore, PSG characteristics do not provide a comprehensive explanation for the differences between WatchPAT and PSG measurements.

Several hypotheses may explain the discrepancy between WatchPAT and PSG in adolescents with overweight and obesity. Firstly, the use of adult-sized measurement components (ie, finger probe) and an algorithm developed for adults may limit the ability to accurately detect sleep-disordered breathing parameters in the youth population.^11,13 However, in the current study, this reason for discrepancy is unlikely given that our sample had a mean age of 15.5 years and more closely resembled an adult phenotype, and therefore we utilized adult AHI scoring criteria. Relatedly, adolescents are a unique group with phenotypes consistent with both pediatric (eg, adenotonsilar hypertrophy) and adult (eg, central adiposity and increased load of the chest wall) OSA.^6,25,26 The endothelial dysfunction often found in adolescents with overweight and obesity may introduce confounding factors in WatchPAT measurements.²⁷ Studies of children with OSA have reported attenuated peripheral arterial tonometry signals with wakefulness and altered vascular reactivity, indicating an overall elevated sympathetic activity that might affect an adolescent’s ability to “down-regulate” their vascular tone after the challenge of an apneic episode.²⁸ Moreover, differences in heart rate variability seen in youth with OSA may impact WatchPAT’s ability to measure AHI based on their current algorithmic parameters.²⁸ Youth with more severe OSA exhibit significantly reduced heart rate variability at fast and intermediate heart rates, but enhanced variation at slow heart rates. Altered cardiac autonomic regulation may be the cause of a higher sympathetic response and a reduction in parasympathetic modulation.²⁸ However, adults with OSA often have endothelial dysfunction and dysautonomia as well.²⁹ Consequently, it is possible that there are other, unknown, complex morphological and pathophysiological differences underlying OSA in adolescents with overweight and obesity compared to other patient populations, potentially disrupting the accuracy of WatchPAT assessments.

The current study has notable strengths including the focused investigation of a high-risk sample comprising adolescents with overweight and obesity, a demographic with documented elevated prevalence of OSA. Our assessment was thorough and novel, encompassing the evaluation of flow sensor reliability and video characteristics from PSG to determine if PSG inaccuracies contributed to the observed discrepancies with WatchPAT measurements.

It is crucial to acknowledge certain limitations in this study, such as the relatively small but heterogeneous sample collected from 2 different data sources. This includes a retrospective review of participants who completed WatchPAT as part of a research study and then subsequently had a clinical PSG. For these participants, an assumption was made that in the interval between WatchPAT and PSG, there was no significant change in an individual’s sleep-disordered breathing. One longitudinal, prospective study found relative stability in OSA despite intervention for obesity across a year’s duration.² It is unknown whether our participants were actively engaging in treatment for obesity; however, comparison of BMI percentiles measured at the time of the PSG and WatchPAT assessments found no statistically significant differences. Similarly, it is unknown whether participants pursued any intervention, such as intranasal steroids in the interim between WatchPAT and PSG assessments which may have impacted AHI.

Additionally, all participants were clinically referred to have PSG due to suspected OSA, raising the possibility that our results may not be generalizable to adolescents with overweight and obesity who do not exhibit symptoms of sleep-disordered breathing. Furthermore, the analysis of disease severity was limited by the presence of only 1 participant with PSG-confirmed severe OSA among the recruited participants.

Lastly, unequal numbers of participants used either WatchPAT200 or the WatchPAT300 models in our study. Although the 2 models have different oximetry data collection modalities, they share the same software and algorithm for analysis. While no statistically significant differences were found for demographic, WatchPAT, or PSG variables between participants studied on the different models, there is still a concern for potential differences that may have impacted findings.

CONCLUSIONS

In this group of adolescents with overweight and obesity, WatchPAT dramatically overestimated AHI compared to PSG, and accurately classified OSA severity in only 10% of participants. These findings were contrary to our initial hypotheses and diverge from results observed in previously published studies in adults and children with and without obesity. Our findings may be attributed to potential pathophysiological differences specific to adolescents with overweight and obesity affecting the measurement and scoring accuracy of WatchPAT. Collectively, these outcomes suggest that caution should be taken when using WatchPAT as a diagnostic tool for assessing OSA in adolescents with overweight and obesity. Future research investigating the underlying the pathophysiological differences in this population is warranted to enhance assessment tools, facilitating timely and appropriate intervention to mitigate the adverse health consequences associated with OSA in this high-risk group.

DISCLOSURE STATEMENT

All authors have seen and approved this manuscript. Work was performed at Children’s Hospital Colorado and the University of Colorado Anschutz Medical Campus. This work was supported by the Department of Pediatrics, University of Colorado Anschutz Medical Campus. All authors report no conflicts of interest.

ABBREVIATIONS

AHI

apnea-hypopnea index

AUC

area under the curve

BMI

body mass index

OSA

obstructive sleep apnea

PSG

polysomnography

TST

total sleep time