A new wearable diagnostic home sleep testing platform: comparison with available systems and benefits of multinight assessments

Abstract

Study Objectives:

We assessed the real-world performance of the ANNE Sleep system against 2 Food and Drug Administration–cleared home sleep testing platforms and the intraindividual night-to-night variability of respiratory event index measured by ANNE Sleep.

Methods:

We evaluated the home performance of the ANNE Sleep system compared with 2 Food and Drug Administration–cleared home sleep testing platforms (WatchPAT: n = 29 and Alice NightOne: n = 46) during a synchronous night with unsupervised patient application. Additionally, we evaluated night-to-night variability of respiratory event index and total sleep time using the ANNE Sleep system (n = 30).

Results:

For the diagnosis of moderate and severe obstructive sleep apnea, the ANNE Sleep system had a positive percent agreement of 58% (95% confidence interval, 28–85%) and a negative percent agreement of 100% (95% confidence interval, 80–100%) compared to WatchPAT. The positive and negative percent agreement for ANNE Sleep vs Alice NightOne was 85% (95% confidence interval, 66–96%) and 95% (95% confidence interval, 74–100%). There were no differences in mean total sleep time or respiratory event index across multiple nights of monitoring with ANNE. There were no differences consistent with a first-night effect but testing multiple nights reclassified obstructive sleep apnea severity in 5 (17%) individuals and detected 3 additional cases of moderate disease, with only a 12% (standard deviation, 28%) mean fluctuation in respiratory event index from the first night of testing compared to a mean of multiple nights. Overall, 80% of users found ANNE comfortable and easy to use.

Conclusions:

ANNE Sleep exhibited stronger concordance with Alice NightOne compared to WatchPAT. While we illustrated low night-to-night variability for ANNE Sleep, the results suggest multiple nights increased detection of moderate or severe obstructive sleep apnea.

Clinical Trial Registration:

Registry: ClinicalTrials.gov; Name: ANNE Diagnostic Agreement With Home Sleep Testing; URL: https://clinicaltrials.gov/ct2/show/NCT05421754; Identifier: NCT05421754.

Citation:

Walter J, Lee JY, Blake S, et al. A new wearable diagnostic home sleep testing platform: comparison with available systems and benefits of multinight assessments. J Clin Sleep Med. 2023;19(5):865–872.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Obstructive sleep apnea is often diagnosed with a single night of home-based sleep testing, despite clinically significant intraindividual night-to-night variation in the apnea-hypopnea index.

Study Impact: We performed home-based, real-world testing of ANNE Sleep compared to other home sleep testing platforms and demonstrated comparable performance to diagnose moderate and severe obstructive sleep apnea in adults. We assessed intraindividual respiratory event index night-to-night variability with ANNE Sleep and found no evidence of a first-night effect but an increase in detection of moderate and severe obstructive sleep apnea with multiple nights of testing.

INTRODUCTION

Obstructive sleep apnea (OSA) is a common but underdiagnosed chronic disease.^1,2 OSA is typically diagnosed by either polysomnography (PSG) or a home-based sleep apnea test (HSAT).³ The current paradigm relies on the PSG-derived apnea-hypopnea index or HSAT-derived respiratory event index (REI), the average number of apnea or hypopnea events per hour, calculated from a single night of sleep monitoring. However, clinically significant intraindividual night-to-night variability in both apnea-hypopnea index and REI has been documented in the literature.^2,4–8 For instance, apnea-hypopnea index fluctuation exceeded 10 events/h in 15–65% of patients undergoing PSG across multiple nights of testing.^4,5 This magnitude of change is large enough to recategorize a patient’s OSA severity, which informs prognosis and guides both treatment decisions and payor coverage for services.^8,9

Some night-to-night variability may be secondary to the first-night effect, a phenomenon initially described in the early 1960s.¹⁰ The first-night effect, attributable to sleeping in an unfamiliar environment, restricted movement, anxiety, or uncomfortable testing equipment, shortens total sleep time, prolongs sleep onset latency, lowers sleep efficiency, and increases awakening frequency.^11–14 A recent meta-analysis found the first-night effect is not a homogenous phenomenon but varies by underlying sleep pathology and patient characteristics. The first-night effect was more pronounced among middle-aged and older patients, who demonstrated a 5.3- to 6.5-fold greater difference in total sleep time between first and second nights of testing compared to individuals 20–24 years of age. Rapid eye movement sleep latency was similarly prolonged among middle-aged and older patients, with a 3- to 4-fold greater increase in the weighted mean difference between middle-aged and older patients and those 20–24 years of age.¹¹

A large study of more than 40,000 adults found a nightly fluctuation of 5.5 events/h in REI measured by HSAT, and more than 30% of the sample had fluctuations across diagnostic cut points for OSA.⁶ A recent study of WatchPAT (Itamar Medical, Caesarea, Israel), a single-use, wrist-mounted sleep diagnostic device, found no evidence of a first-night effect but did demonstrate 1 out of every 4 patients had misclassified OSA severity if only 1 night of monitoring was used as opposed to a mean of serial nights.⁷ Night-to-night variability in these instances may be secondary to the amount of time spent in the supine position or other factors.^11,15

Thus, there is a need for diagnostic platforms that demonstrate high internal precision over multiple nights of wear in a patient’s natural sleeping environment. Recently, we reported 90% sensitivity and 98% specificity for the diagnosis of moderate and severe OSA by a wireless wearable dual-sensor system (ANNE Sleep, Sibel Health, Niles, Illinois) compared to PSG in a multicenter pivotal trial of 225 high-risk patients; the system is also Food and Drug Administration–cleared to aid in the diagnosis of sleep-related breathing disorders.¹⁶ Given the need for rigorously tested home-based diagnostic tools and the marked intraindividual variability in REI, we report the comprehensive evaluation of ANNE Sleep at home. We performed a single-arm prospective real-world study comparing the performance and usability of ANNE Sleep with WatchPAT and a Type III HSAT (Alice NightOne) on synchronous sleep nights in 2 different high-risk patient groups. We then evaluated intraindividual night-to-night variability of REI measured by ANNE Sleep.

METHODS

Equipment

Validation of the ANNE sensors, a Food and Drug Administration–cleared general physiological monitor, has been previously published.^17–19 Briefly, the ANNE Sleep system consists of 2 small, flexible wireless patches that are time-synchronized. One sensor is placed at the suprasternal notch collecting heart rate, temperature, accelerometry, chest wall movements, and respiratory rate and the second sensor wraps around the index finger collecting pulse oxygenation, peripheral temperature, pulse transit and arrival time, and peripheral arterial tonometry, as shown in Figure 1. The sensors are reusable and rechargeable with the capacity for on-board data storage and real-time streaming to companion devices for repeated nights of use.

Figure 1. The ANNE Sleep system.

(A) Illustration of the ANNE Sleep system. The ANNE Sleep system consists of 2 wireless patches placed at the suprasternal notch and the index finger. (B) Sample output of multiple channels and vital signs collected by the system that can be reviewed by a health care provider or sleep technician to identify apneic nighttime events. ECG = electrocardiograph, PAT = peripheral arterial tone, SpO₂ = blood-oxygen saturation.

The ANNE Sleep comparators for this study were WatchPAT (Elevance Health, Indianapolis, Indiana), a single-use, wrist-mounted wireless device determining REI based on peripheral arterial tone (PAT)-based measurements, and the Philips Alice NightOne Home Sleep Test (Koninklijke Philips N.V., Amsterdam, The Netherlands). The Alice NightOne is a Type III HSAT system with a nasal cannula component, central chest unit, a thoracic effort belt, and a wired pulse oximeter. This system measures nasal air flow, snoring, thoracic effort, SpO₂, heart rate, and body position.

ANNE Sleep vs WatchPAT

We performed a prospective single-arm clinical trial to evaluate the home-based performance of ANNE Sleep against WatchPAT on a synchronous night. As a secondary analysis, we performed multiple sequential nights of home-based monitoring with the ANNE Sleep system in this cohort to measure any device- or patient-related variability in key sleep metrics. Adults with a high clinical suspicion of OSA based on medical history and/or physical examination were eligible. After giving written consent participants provided demographic information and medical history and completed the Epworth Sleepiness Scale.²⁰ Subjects were excluded if they had significant chronic cardiopulmonary disease. Enrolled participants completed 1 night of concurrent monitoring, self-applying both WatchPAT and the ANNE Sleep system. Participants then wore the ANNE system for 3 additional nights. At the conclusion of the study individuals completed questions regarding their preferences and experiences using the sensors.

ANNE Sleep outputs were independently scored by a trained physician grader. Automatic reports were generated by the WatchPAT system and approved by a physician grader. All reviewers were blinded to patient characteristics and results of the other testing modality. The study was approved by Northwestern University’s Institutional Review Board (no. STU00214445) and was registered on ClinicalTrials.gov (NCT05421754). All participants provided written informed consent prior to enrollment.

ANNE Sleep vs Alice NightOne

We performed a second prospective single-arm clinical trial enrolling a cohort of patients from beneficiaries of a large private insurer. Participants were at least 18 years old with a high risk of OSA as determined by the STOP-BANG questionnaire. The exclusion criteria included self-reported diagnosis of (1) chronic obstructive pulmonary disease, (2) congestive heart failure, (3) obesity hypoventilation syndrome, (4) sleeping disorder other than sleep apnea, (5) and physical or cognitive inability to appropriately use study equipment. Individuals with current oxygen requirements, actively being treated for OSA, or class III obesity defined by body mass index greater than 40 kg/m² were also excluded. This protocol was approved by the Salus Institutional Review Board (protocol no. AI.SS.01.2019) and all participants provided signed informed consent. Enrolled participants were mailed a study kit with 2 OSA testing systems, a Philips Alice NightOne Home Sleep Test and the ANNE Sleep system. Participants used both systems on the same night of sleep. Study coordinators performed a pre- and posttest phone call with participants to review use of both devices and experiences with the systems. Participants completed an anonymized Systems Usability Index Score to evaluate usability of both systems.

Statistical methods

The primary outcome of the study was diagnostic concordance of moderate or severe OSA by REI determined by the ANNE Sleep system with either WatchPAT or Alice NightOne. An REI between 15 and 30 was defined as moderate OSA and an REI greater than 30 events/h was defined as severe disease per the American Academy of Sleep Medicine definitions. Bland-Altman plots were generated to assess for systematic bias, estimate the mean bias, and 95% interval of differences between ANNE Sleep, WatchPAT, and Alice NightOne REI.²¹ Scatter plots of ANNE Sleep-derived REI plotted against WatchPAT- and Alice NightOne-derived REI, with determination of the Pearson’s correlation coefficient, were generated.

We also evaluated the performance of ANNE Sleep as a diagnostic tool compared to the WatchPAT. Based on the first night of synchronous testing, a second or third night of testing, a mean of serial nights, and highest REI after multiple nights of testing, we calculated positive and negative percent agreement and comparisons of the receiver operating characteristic curves area under the curve. Secondary outcomes included an analysis of variation between REI and total sleep time as determined by the ANNE Sleep system to assess for a first-night effect. Analysis of variance tests were used to compare the means of REI and total sleep time by patient across each night of testing. Correlation coefficients were also determined for paired nights. Systems Usability Index Scores were compared with the Mann-Whitney U test. All statistical programming and analyses were performed with STATA version 15.1 (StataCorp LLC, College Station, Texas).

RESULTS

ANNE Sleep vs WatchPAT

A total of 38 patients were enrolled in the study between January 31, 2022 and March 25, 2022. Eight participants withdrew before completing any nights, and 1 of the participants had a technical failure with WatchPAT. For this portion of the results, there were n = 29 participants in the comparison of ANNE Sleep vs WatchPAT and n = 30 for the multiple nights portion. Each participant wore ANNE Sleep for a median of 3 nights. The mean age of participants was 48.0 years (standard deviation [SD] = 15.3), and 30% were women. The cohort included 63% (n = 19) self-identified as White, 27% (n = 8) as Asian, and 3% (n = 1) as Black. The mean body mass index was 28.0 kg/m² (SD = 3.9). Overall, 27% of participants had hypertension (n = 8) and 7% (n = 2) had type 2 diabetes. The median Epworth Sleepiness Scale score was 8 (Table 1).

Table 1.

The demographic information is provided for participants in the ANNE vs WatchPAT study and for the ANNE vs Alice NightOne study.

Factor	Value
ANNE vs WatchPAT
n	30
Age (years), mean (SD)	48.4 (15.3)
Sex, n (%)
Female	10 (30.0)
Male	17 (60.0)
Did not answer	3 (10.0)
Race, n (%)
Asian	8 (26.7)
Black or African American	1 (3.3)
White	19 (63.3)
Did not answer	3 (10.0)
Hispanic or Latino, n (%)	2 (7.0)
Body mass index (kg/m2), mean (SD)	28.2 (3.9)
Epworth Sleepiness Scale score, median (IQR)	8 (6, 14)
Hypertension, n (%)	8 (26.7)
Type II Diabetes, n (%)	2 (6.7)
Asthma, n (%)	2 (6.7)
Depression, n (%)	6 (20.0)
Anxiety, n (%)	4 (13.0)
ANNE vs Alice NightOne
n	46
Age (years), mean (SD)	49.1 (15.7)
Sex, n (%)
Female	25 (54.3)
Male	18 (39.1)
Did not answer	3 (6.5)
Body mass index (kg/m2), mean (SD)	33.8 (4.0)

IQR = interquartile range, SD = standard deviation.

The mean WatchPAT REI was 18.8 (SD = 19.2) and 44.4% (n = 12) of participants had moderate or severe disease, compared to mean ANNE REI of 11.0 (SD = 11.4), and 23% (n = 7) of participants had moderate or severe disease. If REI was greater than 15 events/h on any night of monitoring with the ANNE system the proportion of patients with moderate or severe disease rose to 32% (n = 10). ANNE-derived REI was highly associated with the WatchPAT REI on synchronous nights given a correlation coefficient of 0.82 (95% confidence interval [CI], 0.64–0.91). A Bland-Altman comparison of ANNE Sleep vs WatchPAT REI demonstrated a mean bias of 7.8 (95% CI, 3.27–12.3) events/h (95% limits of agreement: −15.5 to 31.0) (Figure 2). If a single night of monitoring was used, ANNE Sleep had a positive percent agreement of 58% (95% CI, 28–85%) and a specificity of 100% (95% CI, 80–100%). This increased to a positive percent agreement of 75% (95% CI, 42.8–94.5%) and negative percent agreement of 94.2% (95% CI, 71.3–99.9%) if multiple nights of testing were used and the highest REI was used to categorize participants (Table 2).

Figure 2. Comparison of WatchPAT and ANNE.

Presented is the (A) 2 × 2 matrix comparing the diagnosis of moderate and severe OSA (defined as an REI greater than or equal to 15 events per hour) as diagnosed by ANNE Sleep vs WatchPAT on synchronous nights of testing, (B) scatter plot and comparing absolute REI values derived by both systems, and (C) the Bland-Altman plot demonstrating a mean bias of 7.8 events/h. CI = confidence interval, NPA = negative percent agreement, OSA = obstructive sleep apnea, PPA = positive percent agreement, REI = respiratory event index.

Table 2.

The diagnostic performance of ANNE Sleep compared to Alice NightOne (Type III HSAT) and WatchPAT (home-based test) including the area under of the curve of the receiving operating characteristic curve and positive and negative percent agreements.

	AUC (95% CI)	Positive Percent Agreement (95% CI)	Negative Percent Agreement (95% CI)
ANNE vs Type III HSAT*	0.90 (0.81–0.99)	0.85 (0.66–0.96)	0.95 (0.74–1.0)
ANNE Night 1 vs WatchPAT*	0.79 (0.65–0.94)	0.58 (0.28–0.85)	1.00 (0.80–1.0)
ANNE Night 2 vs WatchPAT	0.76 (0.59–0.92)	0.58 (0.28–0.85)	0.93 (0.66–1.0)
ANNE Night 3 vs WatchPAT	0.85 (0.70–0.99)	0.70 (0.35–0.93)	0.87 (0.59–1.0)
Mean ANNE vs WatchPAT	0.76 (0.61–0.92)	0.58 (0.27–0.85)	0.94 (0.71–1.0)
Highest ANNE REI vs WatchPAT	0.85 (0.71–0.94)	0.75 (0.43–0.95)	0.94 (0.71–1.0)

^*Worn concurrently. The performance of ANNE Sleep defined by either the synchronous night of testing, second or third nights, mean of serial nights, or the highest REI compared to WatchPAT is also presented. AUC = area under the curve, CI = confidence interval, HSAT = home sleep apnea test, REI = respiratory event index.

ANNE Sleep and detection of night-to-night variability

ANNE-derived REIs across multiple nights were strongly associated; night 1 vs night 2 demonstrated a correlation coefficient of 0.90 (95% CI, 0.79–0.95), night 1 vs night 3 had a correlation coefficient of 0.92 (95% CI, 0.82–0.96), and the night 2 vs night 3 correlation coefficient was 0.82 (95% CI, 0.62–0.92). There was no statistically significant difference in the mean REI or total sleep times across multiple nights of testing (Figure 3).

Figure 3. Variability of total sleep time and REI.

The box plots demonstrate the variability of total sleep time (A) and REI (B) across multiple nights of monitoring with the ANNE Sleep system. There was no statistical difference in the mean total sleep time or mean REI consistent with an absence of a first-night effect. REI = respiratory event index.

Overall, this study demonstrated a mean change of 12% (SD 28%) in REI from the first night compared to the mean REI determined across multiple nights. If use of multiple nights to determine a mean REI were used, instead of a single night of testing, the OSA severity classification would have changed in 17% (n = 5) of patients. Two patients increased from mild to moderate disease (7%) and 1 patient changed from moderate to severe disease (3%). Additionally, 1 patient would have been reclassified as mild rather than moderate and 1 patient as moderate rather than severe. Alternatively, if the highest or worst REI on any given night was used to classify disease as opposed to the first night of testing, 3 patients would have been reclassified as having moderate rather than mild disease (10%) and 2 patients would have been reclassified as having severe rather than mild disease (7%).

At the study’s conclusion, 80% of participants reported agreeing or strongly agreeing that the ANNE Sleep system was easy to sleep with during use, and 57% of users reported WatchPAT was easy to sleep with during use. Additionally, 80% (n = 24) of users found the ANNE Sleep system easy to apply, all found it easy to remove, and 87% (n = 26) reported use of the system at home was straightforward.

ANNE Sleep vs Alice NightOne

A total of 60 participants were initially enrolled in this study. Fourteen patients were excluded from the final analysis (4 did not complete sleep monitoring with either device, 2 had inconclusive Alice NightOne results, 4 had less than 4 hours of high-quality data collected by ANNE Sleep, 2 withdrew consent, 1 reported a technical issue with Alice NightOne, and 1 was lost to follow-up). The final cohort (n = 46) had a mean age of 46 (SD = 10) years and 33% were male. Overall, the mean body mass index was 33.8 kg/m² (Table 1). The mean HSAT REI was 19.1 (SD = 13.7) and 59% (n = 27) of participants had moderate or severe disease; the mean ANNE Sleep REI was 19.8 (SD = 13.8) events/h and 52% (n = 24) of participants had moderate or severe disease. A blinded comparison of REI for the experimental system exhibited a mean bias of 0.5 (95% CI, −2.3 to 3.3) events/h with the 95% limits of agreement −18.2 to 19.2 (Figure 4). REI determined by the experimental system and HSAT showed high correlation (r = .78; 95% CI, 0.64–0.87). For the diagnosis of moderate and severe OSA, the positive percent agreement and negative percent agreement for the ANNE Sleep vs HSAT for the diagnosis of moderate to severe OSA was 85% (95% CI, 66–96%) and 95% (95% CI, 74–100%), respectively. The average Systems Usability Index Score was 61 for ANNE Sleep (95% CI, 57–65) and 46 for the HSAT system (95% CI, 41–51) (P < .001), whereby a higher score denotes greater user-reported usability. Moreover, 84% of users preferred the ANNE Sleep system compared to the Alice NightOne system, citing comfort and natural sleeping positions as the primary reasons.

Figure 4. Comparison of Alice NightOne and ANNE.

Presented is the (A) 2 × 2 matrix comparing the diagnosis of moderate and severe OSA (defined as an REI greater than or equal to 15 events/h) as diagnosed by ANNE Sleep vs Alice NightOne on synchronous nights of testing, (B) scatter plot and comparing absolute REI values derived by both systems, and (C) the Bland-Altman plot demonstrating a mean bias of 0.5 events/h. CI = confidence interval, NPA = negative percent agreement, OSA = obstructive sleep apnea, PPA = positive percent agreement, REI = respiratory event index.

DISCUSSION

In a single-arm prospective study we demonstrated ANNE Sleep had good agreement with WatchPAT for the detection of moderate and severe OSA performed on a single synchronous night, with a positive percent agreement of 58% and negative percent agreement of 100%. In a second single-arm prospective study, ANNE Sleep and Alice NightOne showed greater concordance for the diagnosis of moderate to severe OSA in a cohort of high-risk patients with a positive percent agreement of 85% and a negative percent agreement of 95%. This performance is largely analogous to a previous multicenter clinical trial where ANNE Sleep was compared to in-laboratory PSG.¹⁶

No evidence of a first-night effect among users of the ANNE Sleep system at home was observed. There was no significant, systematic change in total sleep time, REI, or diagnostic categorization across multiple nights of testing. Although there was no consistent change in either directionality or magnitude of REI change across multiple nights, this study adds to the growing body of literature demonstrating the additive value of additional nights of monitoring because of inherent intraindividual night-to-night variability of REI. Night-to-night variability of REI and consequently severity of OSA have been demonstrated independent of the diagnostic tool.^2,4–8

Fluctuation of apnea-hypopnea index and REI has been documented in repeated PSG studies, HSATs, and now the ANNE Sleep system.^2,4–8 The night-to-night variability noted in this study agrees with previous reports in the literature of home-based sleep apnea testing. A large, retrospective study of patients undergoing 2 nights of HSAT found traditional risk factors associated with OSA, such as age, sex, body mass index, neck circumference, and sleepiness, were not reliably predictive of night-to-night variability. However, variability was clinically significant for 1 in 3 patients as defined by a change in OSA severity classification.⁶ Similarly, in a study of 51 patients with a high pretest probability of OSA who repeatedly wore the WatchPAT device, 1 in 4 patients had reclassification of OSA severity and a mean difference of 57% was recorded in REI across nights. Similarly, our findings demonstrated no systematic differences in REI suggestive of an absent first-night effect but persistent, measurable, and clinically meaningful night-to-night variability leading to changes in severity classification.⁷

Additionally, if the mean REI of multiple nights was used as opposed to a single night of testing, there was reclassification of OSA severity (mild, moderate, or severe) in 17% (n = 5) of the cohort. Of these 5 patients, 3 OSA classifications increased in severity and diagnosis based on 1 night of testing would have led to suboptimal or contraindicated treatments. With the addition of multiple nights, 2 patients would have been recommended to initiate treatment regardless of symptoms, as they were stratified as having moderate rather than mild disease, and 1 patient would no longer be a candidate for oral appliance therapy, given an elevation in risk to severe disease (REI > 30 events/h).³ The remaining 2 patients had a drop in their risk stratification, though mean REI was still abnormal, and each remained eligible for treatment.

Similarly, if the highest REI on any given night of testing was used (meaning treatment was indicated if moderate or severe level REI was detected on any night of testing), 5 patients would have increased in OSA severity classification (3 rose from mild to moderate disease and 2 from moderate to severe disease). Notably, 3 additional patients regardless of symptoms (10%) would have qualified for empiric treatment based on an REI > 15 events/h on a subsequent night of monitoring. As the patients in the cohort were high-risk based on history and clinical presentation, a negative HSAT would likely prompt a confirmatory PSG to avoid missing OSA.^3,22,23 Thus, the use of ANNE Sleep across multiple nights by 30 patients may have avoided the need for 3 confirmatory PSGs. Based on these findings, the number needed to screen (eg, with multiple nights of sleep testing) to identify 1 additional case of moderate to severe OSA is modest at 10 patients. Cumulatively, these findings suggest a potential beneficial diagnostic yield of repeated nights of monitoring sleep at home to maximize disease detection without requiring expensive and labor-intensive confirmatory PSG studies. This may be particularly relevant for patients who exhibit a borderline REI on the first night, between mild or moderate disease.

There are important limitations to this study to consider when interpreting these findings. This was a small study of only 76 total participants (n = 30 in ANNE Sleep vs WatchPAT and n = 46 in ANNE Sleep vs Alice NightOne). Thus, the sample size may not have been adequately powered to detect a smaller first-night effect. Thus, future studies with a larger number of patients would be important to further validate these results. Although the study was designed to include individuals with a high pretest probability of OSA, the demographics for the WatchPAT vs ANNE Sleep study suggest a population with lower baseline risk—the mean body mass index was 28 kg/m² and the mean Epworth Sleepiness Scale score of 8 is consistent with an average amount of sleepiness.²⁰

Given an inability to reliably predict which patients will experience night-to-night variability, variability persisting even with wireless low-profile sensors, and the modest number needed to treat, this study adds to a growing body of literature in support of a paradigm shift away from single to multiple nights of testing. The significant morbidity and all-cause mortality associated with untreated OSA demonstrate the need for high negative predictive value of diagnostic testing systems. The ANNE Sleep system is cloud-enabled with automated data quality checking after each night, rechargeable with battery life of up to 9 nights on a single charge, and reusable, easily permitting multiple nights of testing that could theoretically capture more clinically significant disease with minimal added cost or burden to the patient or health care system. The low-profile, flexible design of the device helps to reduce night-to-night variability resulting from the first-night effect and allows more natural sleeping for high-risk patients.

Collectively, these findings suggest that the ANNE Sleep system exhibits good agreement with both WatchPAT and Alice NightOne in the home setting and greater user-reported comfort. Given these results and previous clinical validation of ANNE Sleep vs PSG, the ANNE Sleep system may be a valuable addition for reliable at-home diagnosis of OSA in appropriately selected patients. Although there was no statistically significant difference in the discriminatory performance across serial nights of ANNE Sleep monitoring, there may be a benefit of multiple nights of testing given intraindividual night-to-night variability leading to clinical disease severity reclassification that could prevent the need for some confirmatory testing with PSG. There is an increasing call to move away from the paradigm of a single night of sleep testing to a multiple-night approach.^5,6 Use of the highest REI to guide treatment would represent the optimal way to prevent undertreatment and underdiagnosis, given the status quo with unacceptably high rates of undiagnosed OSA and the health and broader societal risks of underdiagnosis or undertreatment.

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. Work for this study was performed at Northwestern Memorial Hospital. J.R.W. reports a spouse with equity/royalty interest in the ANNE Sleep system. J.Y.L., L.Y., S.C., A.S., J.L., A.B., M.B., and S.X. are employees of Sibel Health, commercializing the ANNE Sleep system. S.X. reports royalty interest in the ANNE Sleep system. This study was funded by Elevance Health (S.B. and L.K.) and the National Institute of Health’s National Heart, Lung, and Blood Institute (1R43HL151549-01 to J.Y.L. and S.X.). No other authors report conflicts of interest.

ABBREVIATIONS

CI

confidence interval

HSAT

home sleep apnea test

OSA

obstructive sleep apnea

PSG

polysomnography

REI

respiratory event index

SD

standard deviation