Validation of home portable monitoring for the diagnosis of sleep-disordered breathing in adolescents and adults with neuromuscular disorders

Abstract

Study Objectives:

Sleep-disordered breathing (SDB) is common in patients with neuromuscular disorders (NMD), developing before chronic hypercapnia appears. Polysomnography (PSG) is the diagnostic gold standard but is often impractical and poorly accessible for individuals with NMD. We sought to determine the diagnostic accuracy, feasibility, and patient preference of home sleep apnea testing (HSAT) compared with PSG for the detection of SDB in NMD.

Methods:

Participants with NMD at risk for SDB aged ≥ 13 years underwent HSAT followed by overnight PSG with concomitant laboratory sleep apnea testing (same device as HSAT). Sensitivity and specificity were calculated for standard apnea-hypopnea index cutoffs for mild (≥ 5 events/h), moderate (≥ 15 events/h), and severe SDB (≥ 30 events/h) and for an oxygen desaturation index ≥ 5 events/h. Receiver operating characteristic curves were built. A questionnaire assessed patient preference.

Results:

Of 38 participants, 73% had moderate to severe SDB and 79% had technically acceptable HSAT. For an apnea-hypopnea index ≥ 15 events/h, HSAT sensitivity and specificity were 50% and 88%, respectively. For an oxygen desaturation index ≥ 5 events/h, HSAT sensitivity and specificity were 95% and 78%, respectively. The area under the receiver operating characteristic curve for an apnea-hypopnea index ≥ 15 events/h was 0.88 (95% confidence interval, 0.69–1.00) for HSAT. The HSAT underestimated the apnea-hypopnea index from PSG (bias, –10.7 ± 15.9 events/h). HSAT was preferred to PSG by 61% of participants.

Conclusions:

HSAT is feasible, preferred by patients, and reliable for detecting SDB in most patients, although it cannot definitively rule out SDB. Therefore, HSAT is a viable diagnostic approach for SDB in NMD when PSG is not feasible, recognizing that it does not accurately distinguish between upper-airway obstruction and hypoventilation. Additional work is needed to further optimize home sleep testing in NMD.

Citation:

Westenberg JN, Petrof BJ, Noel F, et al. Validation of home portable monitoring for the diagnosis of sleep-disordered breathing in adolescents and adults with neuromuscular disorders. J Clin Sleep Med. 2021;17(8):1579–1590.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Patients with neuromuscular disorders are at increased risk of sleep-disordered breathing (SDB). Whereas full polysomnography is the gold standard for SDB diagnosis, home sleep apnea testing (HSAT) is currently not recommended in patients with neuromuscular disorders; however, if HSAT is diagnostically accurate, then it may improve access to testing, reduce wait times, and decrease patient burden.

Study Impact: HSAT had generally acceptable feasibility and specificity for the detection of SDB in selected individuals with neuromuscular disorders who were primarily at risk for obstructive sleep apnea, but SDB severity was underestimated and sensitivity was insufficient to use HSAT as a “rule out” test, similar to HSAT performance in the general population. Nevertheless, HSAT detected SDB in most participants and was preferred by more participants, suggesting that HSAT could improve access to SDB diagnosis in patients with neuromuscular disorders.

INTRODUCTION

Respiratory muscle weakness causing ventilatory impairment is the leading cause of morbidity and mortality in patients with neuromuscular disorders (NMD). ^1–3 Sleep-disordered breathing (SDB) typically begins with rapid eye movement sleep–related pseudocentral apneas or hypopneas resulting from diaphragmatic weakness and/or associated obstructive apneas or hypopneas, before the development of nocturnal hypoventilation and finally daytime hypercapnic respiratory failure. ^4–6 In addition, obstructive sleep apnea (OSA) is a frequent feature of NMD in both adults and children. ^7–10 Predisposing factors include upper-airway (oropharyngeal) muscle weakness, macroglossia, sedentary lifestyle, and obesity. ^4,11 Some NMD are associated with central apnea. ¹² In patients with Duchenne muscular dystrophy, the most common dystrophy, the prevalence of OSA is likely to increase with the increasing use of corticosteroids in this population and the associated weight gain. ¹¹ OSA has adverse consequences not only on the quality of life but also on clinical and physiologic outcomes, including cardiac function. ^13–15 Therefore, both OSA and nocturnal hypoventilation represent clinically important forms of SDB that can negatively impact patient outcomes in NMD.

Accordingly, early recognition of SDB leading to timely treatment may prevent clinical deterioration and improve quality of life in patients with NMD. ^16,17 Conversely, delayed diagnosis of SDB may be associated with sudden unforeseen and occasionally catastrophic deterioration requiring hospitalization and intensive care unit care, which could potentially be avoided by initiating earlier treatment. Full polysomnography (PSG), or level 1 sleep testing, is currently considered the gold standard for SDB diagnosis. ^18,19 The failure to make a timely diagnosis of SDB in patients with NMD can result from inadequate awareness of SDB manifestations on the part of the patient or physician and because of a lack of suggestive symptoms. ²⁰ However, a major barrier is often the suboptimal access to PSG in many jurisdictions and the high burden for patients and their caregivers in performing overnight testing in a sleep laboratory. ²¹ PSG is a labor-intensive test that typically includes recordings of more than 16 data channels and requires an overnight stay at the sleep laboratory. In Canada, the waiting time to access PSG studies varies from 8–36 months, whereas in the United States the range is 2–10 months. ²² Many sleep laboratories are not wheelchair accessible, lack bathrooms for disabled persons, or may require patients to transfer into bed independently. Moreover, pediatric PSG is even less accessible and is unavailable in some jurisdictions. ²¹

Home sleep apnea testing (HSAT), or level 3 sleep testing, involves a more limited number of channels than PSG and can be performed in an unattended setting such as at home. In patients with a high pretest probability of moderate to severe OSA and without major comorbidities or other sleep disorders, sensitivity and specificity are reportedly 79%–93% and 60%–90%, respectively, including in children. ^19,23,24 Detection of central apneas has also been shown to be accurate with portable monitors. ²⁵ Cost-effectiveness studies have suggested a substantially decreased cost for HSAT compared to PSG. ²⁶ However, despite the potential of HSAT to reduce wait times, logistical burden, and cost for patients with NMD, the American Academy of Sleep Medicine and the Canadian Thoracic Society recommend against using HSAT for the diagnosis of SDB in this patient population. ^18,19 This recommendation stems primarily from a lack of data to support such use rather than any clear evidence against this approach. ^18,19

Accordingly, the overall goal of this study was to explore whether HSAT using a portable monitoring device might be a viable alternative to PSG for the diagnosis of SDB in NMD. Our primary objective was to validate the Alice PDX portable recording device (Philips Respironics, Markham, ON) for the home diagnosis of SDB in pediatric (aged 13–17 years) and adult (aged ≥ 18) patients with NMD at risk for OSA or hypoventilation. Our secondary objective was to evaluate the feasibility of HSAT in our target population along with patient preference.

METHODS

Study design

This was a prospective study, approved by the institutional review ethics board (REB# 15-381-MUHC/2016-755). Eligible participants (or their legal guardian if minors) provided written consent. The neuromuscular diagnosis and other medical history, capillary blood gases and pulmonary function testing including spirometry, maximal inspiratory pressure, maximal expiratory pressure, and sniff nasal inspiratory pressure results were extracted from medical records when available within the preceding 3 months. Participants first underwent HSAT, followed within 1–3 months by overnight in-laboratory PSG. To prevent familiarization with sleep testing before HSAT and avoid biasing feasibility, HSAT was always performed first. The in-laboratory PSG was performed with a concomitant laboratory sleep apnea test (LSAT) performed using the same device as with HSAT. Participants were asked to fill out questionnaires querying their preference (HSAT vs PSG) along with the pros and cons they perceived with each type of test.

Recruitment

Participants were patients aged ≥ 13 years with different NMD. This age cutoff was chosen because adult PSG scoring criteria can be used in this age group. ²⁷ Recruitment occurred between April 2016 and September 2018 through pediatric respiratory neuromuscular clinics and adult respiratory, sleep, and “home ventilation” clinics at the McGill University Health Centre (Montreal, Quebec), an academic and tertiary care center that receives referrals from the Montreal Neurological Hospital and outside centers.

The inclusion criteria for adult participants were the presence of NMD and either a forced vital capacity (FVC) < 50% of predicted or a positive STOP questionnaire. ²⁸ In adolescents, the inclusion criteria were the presence of NMD and a FVC < 70% or a positive I’M SLEEPY questionnaire. ²⁹ Patients were considered at high risk for OSA with ≥ 2 positive answers on the STOP questionnaire for adults and with ≥ 3 of the 8 questions on the I’M SLEEPY questionnaire for adolescents. The parent version of the I’M SLEEPY tool was also completed but did not impact inclusion or exclusion. ²⁹

Exclusion criteria were as follows: living at too great a distance for completion of the study procedures, apparent inability to install the portable monitoring device independently, tracheostomy, requiring urgent initiation of noninvasive ventilation based on clinical assessment, and clinical deterioration during the period between the HSAT and PSG studies. Patients who had previously performed sleep testing using a portable monitoring device were also excluded to avoid biasing feasibility.

Measurements

PSG

Patients underwent standard overnight PSG (Nihon Kohden, Irvine, CA) at the sleep laboratory of the Research Institute of the McGill University Health Centre. Standard full PSG channels were recorded in the laboratory as per American Academy of Sleep Medicine recommendations (The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications, Version 2.2), ²⁷ including 6 electroencephalogram (EEG) channels (C3, C4, F3, F4, O1, O2) for sleep staging, bilateral tibialis anterior electromyogram to assess periodic leg movements, and digital audio and video; respiratory inductance plethysmography was used for thoracoabdominal motion, and oronasal pressure and thermistor were used for airflow. Pulse oximetry was used for monitoring hemoglobin oxygen saturation. Scoring of PSG recordings was done by a single experienced, certified PSG technologist using standard American Academy of Sleep Medicine criteria (AASM Scoring Manual). ²⁷ Respiratory events were summarized as the apnea-hypopnea index (AHI), the respiratory disturbance index (RDI; including respiratory effort–related arousals [RERAs]), and the 3% oxygen desaturation index (ODI).

Portable monitor

The Alice PDX (Philips Respironics, Markham, ON) is a widely used level 3 sleep testing device, categorized as S₀ C₄ O₁ P₂ E₁ R₁ according to the SCOPER system. ¹⁹ It records pulse oximetry, providing oxygen saturation and pulse rate, airflow (pressure-based airflow with snore detection), thoracic and abdominal movements (inductance plethysmography), and body position. For the home measurement, the participants were provided with verbal and written instructions and were given an in-person demonstration by a research assistant, who was also available by telephone the evening of the test for any questions related to installation. For the LSAT, the same device was installed for simultaneous recording during the PSG by the PSG technician. The nasal pressure recording was split using a Y connector with 1 limb going to the PSG system and the other to the LSAT system.

The HSAT and LSAT were scored by a single scorer, in a blinded fashion. The AHI was defined as the number of apneas and hypopneas (based on 3% desaturation) per hour of recording. In addition, autonomic arousals, defined as a transient pulse rise of ≥ 6 beats per minute, were scored as a surrogate marker of EEG arousals. ^30,31 The RDI was calculated and included both AHI events and events associated with autonomic arousal (hypopneas and respiratory effort–related arousals) as previously described. ³²

Data analysis

Descriptive statistics were used to summarize the baseline demographics, medical history, and PSG, LSAT, and HSAT study results. For normally distributed continuous variables, mean and standard deviation were calculated, whereas median and interquartile range were determined for nonnormally distributed continuous variables. The Shapiro-Wilk test was used to measure normality. The Student t test was used to determine significance between normally distributed data, and the Mann-Whitney U test was performed when data were not normally distributed. Spearman correlations were performed between SDB variables and clinical parameters of respiratory function.

Correlation plots were constructed to visually compare respiratory event indices from HSAT or LSAT vs PSG. Standard severity cutoffs for the AHI as measured on PSG were used to evaluate the diagnostic performance of HSAT and LSAT: AHI ≥ 5 events/h (at least mild OSA), AHI ≥ 15 events/h (at least moderate OSA), and AHI ≥ 30 events/h (severe OSA). Similar RDI cutoffs were also used, along with an ODI cutoff of ≥ 5 events/h. Sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of HSAT and the LSAT were calculated using these same cutoff values. Receiver operating characteristic curves were constructed using PSG AHI thresholds of 5, 15, and 30 events/h as described above. Receiver operating characteristic curves were also constructed for the same RDI cutoffs, using the RDI from HSAT or LSAT as the predictor and for the ODI cutoff of 5 events/h. For each PSG index, the best predictive threshold value from HSAT was calculated that optimized sensitivity and specificity for the detection of PSG-based threshold severity (eg, AHI ≥ 15 events/h). Modified Bland-Altman plots were used as a graphical representation of the observed differences between the paired measurements and the PSG-based values. Mean bias and standard deviation were calculated.

The feasibility assessment of the HSAT studies was based on the technical adequacy of recordings. A recording was deemed to have failed if no signals were recorded at all or if there were < 4 hours of adequate pulse oximetry, effort, and airflow signals available. If this was the case, then the test was attempted a second time when possible. Recordings were classified as partially successful when some signals were suboptimal or missing (eg, fluctuating quality of airflow, absent signal on 1 of the 2 inductance plethysmography bands) but the study was nonetheless interpretable. Questionnaire responses were summarized as proportions. For open questions regarding the perceived pros and cons of home vs laboratory testing, responses were grouped into themes and summarized. All analyses were performed using SAS statistics version 9.4 (SAS Institute, Cary, North Carolina) and RStudio version 1.1.383 (RStudio, Boston, Massachusetts). Statistical significance was set at P < .05.

RESULTS

Participant characteristics

From the 54 adult patients identified for this study, 5 did not meet the inclusion criteria and 17 declined participation, resulting in 32 enrolled adult patients. From the pediatric clinic, 18 patients were initially identified with NMD who were in the relevant age range. Of those, 12 declined participation, which left 6 adolescent participants eligible who enrolled in the study. A summary of participant characteristics is shown in Table 1 . The majority of adult patients had myotonic dystrophies (n = 13) or postpolio syndrome (n = 5), and the most frequent diagnosis among pediatric patients was Duchenne muscular dystrophy (n = 3). Of the 32 adult patients, 27 had positive STOP scores (1 patient did not complete the questionnaire) and 7 had subthreshold FVC (< 50%), with 2 patients meeting both criteria. Of the 6 adolescent patients, 5 had a positive questionnaire and 3 had subthreshold FVC (< 70%), with 3 meeting both criteria. Of the 38 enrolled participants, 2 adult patients and 1 pediatric patient did not use the HSAT monitor and dropped out of the study; thus 35 (30 adults, 5 children) patients attempted the HSAT and underwent PSG. Among those, based on the PSG results and using the standard OSA severity cutoff values, 2 (5.7%) patients showed no evidence of OSA, 7 (20.0%) patients had mild OSA, 11 (31.4%) patients had moderate OSA, and 15 (42.9%) patients had severe OSA.

Table 1.

Participant baseline characteristics.

	Adult (n = 32)	Pediatric (n = 6)	Total (n = 38)	Total with Interpretable HSAT (n = 30)
Male sex, n (%)	17 (53.1)	6 (100)	23 (60.5)	18 (60.0)
Age (y)	40.3 (28.4–57.6)	16.0 (14.6–17.3)	36.1 (20.4–55.5)	36.8 (21.8–55.5)
Body mass index (kg/m2)	25.1 (21.2–30.7)	23.3 (21.7–27.8)	24.7 (21.1–30.2)	26.1 (22.5–31.2)
Neuromuscular disorders, n (%)
Myotonic dystrophies	13	1	14	12
Postpolio syndrome	5	-	5	5
Inclusion-body myositis	2	-	2	1
Spinal muscular atrophy type 2	1	1	2	1
Duchenne muscular dystrophy	2	3	5	2
Friedrich ataxia	-	1	1	1
Othera	9	-	9	8
Wheelchair dependent, n (%)	10 (31.3)	4 (66.7)	14 (36.8)	13 (43.3)
FVC (L)	2.8 (1.0); n = 29	2.4 (1.4); n = 5	2.7 (1.1); n = 34	2.9 (1.1); n = 28
FVC (% predicted)	72.3 (24.4); n = 29	72.9 (22.1); n = 5	72.4 (23.8); n = 34	75.9 (22.0); n = 28
FVC% < threshold,b n	7	3	10	7
MIP (cm H2O)	–69.9 (26.4); n = 23	–64.5 (20.2); n = 5	–68.9 (25.2); n = 28	–72.9 (26.3); n = 22
MEP	82.3 (33.5); n = 23	67.0 (6.4); n = 5	79.5 (30.9); n = 28	81.0 (32.8); n = 22
SNIP	62.1 (21.6); n = 10	—	62.1 (21.6); n = 10	63.1 (23.4); n = 8
pCO2 c	40.0 (5.7); n = 29	36.8 (2.4); n = 5	39.5 (5.4); n = 34	39.2 (4.4); n = 29
STOP scores	2.0 (2.0–3.0)	3.0d	2.5 (2.0–3)	3.0 (2.0–3.0)
Positive STOP scores, n	27	1d	28	27
I’M SLEEPY adult version scores	-	2.0 (1.0–3.3)	1 (1.0–3.3)	1 (1.0–2.0)
I’M SLEEPY child version scores	-	3.0 (2.5–3.3)	3.0 (2.5–3.3)	3.0 (2.0–3.0)
Positive I’M SLEEPY scores, n	-	4	4	3

Values are reported as mean (SD) for normally distributed continuous variables and median (Q1–Q3) for nonnormally distributed continuous variables unless otherwise stated. ^aOther diagnoses: sarcoglucanopathy, occulopharyngeal muscular dystrophy, multiple sclerosis, familial polyneuropathy, distal arthrogryposis, amyotrophic lateral sclerosis, fascio-scapulo-humeral muscular dystrophy, myasthenia gravis, and cerebral palsy. ^bThreshold was defined as 50% of pred. for adults and 70% of pred. for adolescents. ^cFrom arterialized capillary blood gas. ^dOne patient aged 18.1 years was enrolled as an adolescent but completed the STOP. FVC = forced vital capacity, HSAT = home sleep apnea testing, MEP = maximal expiratory pressure, MIP = maximal inspiratory pressure, pCO₂ = partial pressure of carbon dioxide, SD = standard deviation, SNIP = sniff nasal inspiratory pressure.

Feasibility of HSAT in patients with NMDs

Of the 30 adult patients who underwent HSAT, 18 (60%) had a successful recording on the first attempt of the HSAT, 7 (23.3%) had a partially successful recording, and 5 (16.7%) experienced a recording failure. Of the 5 adults with failed recordings, 2 participants underwent a second HSAT and both were successful. Of the 5 adolescent patients who underwent HSAT, only 1 (16.7%) had a successful recording on the first attempt. From the 4 adolescents who failed the first test and underwent repeat testing, 2 (50%) were successful on the second attempt. In total, 30 (78.9%) of the 38 enrolled participants, or 30 of 35 (85.7%) who attempted testing, provided interpretable HSAT that could be used in the analyses.

Detection of SDB by HSAT in patients with NMD

Results of the HSAT as compared to PSG are summarized in Table 2 . LSAT, performed on the same night as PSG, served to directly compare the technical accuracy of the device, eliminating night-to-night variability and issues related to self-installation of the device. Results are also included in Table 2 . The median AHI in our study population was 24.5 events/h (interquartile range, 11.6–38.8). Our study population had a high prevalence of SDB, with 73% of participants meeting the criteria for moderate or severe OSA. Correlation plots for AHI, RDI, and ODI are shown in Figure 1 .

Table 2.

Sleep study results for participants (n = 30).

	PSG	LSAT	P a	HSAT	Pb
Sleep architecture
TST (min)	367.3 (285.5–384.0)
Total analysis time		439.4 (40.2)		447.6 (97.3)
Sleep efficiency (%)	84.5 (76.0–89.0)
WASO (min)	50.5 (24.0–112.0)
Sleep changes	121.5 (101.0–144.0)
Stage N1 sleep (% TST)	11.4 (7.9–13.7)
Stage N2 sleep (% TST)	42.1 (38.9–49.5)
Stage N3 sleep (% TST)	30.5 (12.5)
REM sleep (% TST)	11.6 (6.6)
Supine position (% TST)	74.4 (31.0–100.0)	77.3 (32.4–100.0)	.94	59.2 (23.0–99.9)	.41
Total arousal index (events/h)	41.0 (16.6)
Respiratory arousal index (events/h)	19.1 (15.6–34.8)
Periodic limb movement arousal index (events/h)	0.4 (0.0–3.0)
Periodic limb movement sleep index (events/h)	3.3 (0.0–12.8)
OSA variables
AHI (events/h)	24.5 (11.6–38.8)	13.5 (4.3–24.5)	.011	11.6 (4.8–19.6)	.007
Participants with AHI ≥ 5 events/h, n (%)	28 (93.3)	22 (73.3)	.038	22 (73.3)	.038
Participants with AHI ≥ 15 events/h, n (%)	22 (73.3)	14 (46.7)	.035	12 (40.0)	.009
Participants with AHI ≥ 30 events/h, n (%)	13 (43.3)	4 (13.3)	.010	6 (20.0)	.052
Central apnea index (events/h)	0.4 (0.0–1.2)	0.9 (0.3, 2.6)	.017	0.9 (0.4, 2.6)	.029
Obstructive apnea index (events/h)	0.4 (0.0–1.9)	0.1 (0.0–2.2)	.65	0.3 (0.0–2.2)	.78
Hypopnea index (events/h)	21.6 (11.6–31.1)	12.7 (2.8–17.9)	.003	8.8 (4.7–16.8)	< .001
RDI (events/h)	29.5 (23.4–46.0)	26.5 (12.1–37.1)	.055	22.4 (14.0–34.0)	.04
Supine RDI (events/h)	35.6 (24.9–50.3)	30.4 (14.5–38.0)	.021	23.5 (15.9–34.0)	.005
Nonsupine RDI (events/h)	13.6 (0.0–25.1)	5.8 (0.0–14.5)	.24	10.7 (0.0–19.7)	.93
ODI (events/h)	11.3 (4.0–23.8)	13.0 (3.3–21.0)	.80	9.2 (4.3–18.4)	.88
NREM sleep RDI (events/h)	28.1 (21.7–43.7)
REM sleep RDI (events/h)	34.6 (19.6–65.5)
Mean SpO2 (%)	94.0 (92.0–95.0)	94.0 (91.0–96.0)	.87	94.5 (93.0–95.0)	.47
Minimum SpO2 (%)	87.5 (82.0–91.0)	85.5 (81.0–89.0)	.23	85.0 (77.0–89.0)	.14
T90 ≥ 5 consecutive min (n)	4	7	.50	5	.99

Values are reported as mean (SD) for normally distributed continuous variables and median (interquartile range, Q1–Q3) for nonnormally distributed continuous variables unless otherwise stated. A Mann-Whitney U test was used for nonnormally distributed data and a chi-square test was used for category variables. ^aBetween LSAT and PSG. ^bBetween HSAT and PSG. AHI = apnea-hypopnea index, HSAT = home sleep apnea testing, LSAT = laboratory sleep apnea testing, NREM = nonrapid eye movement, ODI = oxygen desaturation index, OSA = obstructive sleep apnea, PSG = polysomnography, RDI = respiratory disturbance index, REM = rapid eye movement, SpO₂ = oxygen saturation, T90 = time with SpO₂ below 90%, TST = total sleep time.

Figure 1. Correlation plots for respiratory indices—values from HSAT or LSAT compared with PSG.

(A) AHI from HSAT. (B) AHI from LSAT. (C) RDI from HSAT. (D) RDI from LSAT. (E) ODI from HSAT. (F) ODI from LSAT. AHI = apnea-hypopnea index, HSAT = home sleep apnea testing, LSAT = laboratory sleep apnea testing, ODI = oxygen desaturation index, PSG = polysomnography, RDI = respiratory disturbance index.

The HSAT for detecting mild SDB (AHI_PSG ≥ 5 events/h) showed a sensitivity of 79% and a specificity of 100% ( Table 3 ). The LSAT performed similarly. For patients with AHI_PSG ≥ 15 events/h, the sensitivity of HSAT was 50% and was 64% for the LSAT. For patients with AHI_PSG ≥ 30 events/h, the sensitivity of both HSAT and LSAT was poor. The specificity and positive predictive value were generally high for the HSAT at all AHI cutoffs in our study population, but the negative predictive value was low.

Table 3.

Performance of HSAT and LSAT compared to PSG for at least mild, at least moderate, and severe OSA.

PSG	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)	Accuracy (95% CI)
	HSAT
AHI ≥ 5 events/h	AHIHSAT ≥ 5 events/h
	0.79 (0.64–0.94)	1.00 (1.00–1.00)	1.00 (1.00–1.00)	0.25 (0–0.55)	0.80 (0.66–0.94)
AHI ≥ 15 events/h	AHIHSAT ≥ 15 events/h
	0.50 (0.29–0.71)	0.88 (0.65–1.00)	0.92 (0.77–1.00)	0.39 (0.16–0.62)	0.60 (0.42–0.78)
AHI ≥ 30 events/h	AHIHSAT ≥ 30 events/h
	0.38 (0.12–0.64)	0.94 (0.83–1.00)	0.83 (0.53–1.00)	0.67 (0.48–0.86)	0.70 (0.54–0.86)
	LSAT
AHI ≥ 5 events/h	AHILSAT ≥ 5 events/h
	0.79 (0.64–0.94)	1.00 (1.00–1.00)	1.00 (1.00–1.00)	0.25 (0–0.55)	0.80 (0.66–0.94)
AHI ≥ 15 events/h	AHILSAT ≥ 15 events/h
	0.64 (0.44–0.84)	1.00 (1.00–1.00)	1.00 (1.00–1.00)	0.50 (0.26–0.75)	0.73 (0.57–0.89)
AHI ≥ 30 events/h	AHILSAT ≥ 30 events/h
	0.31 (0.06–0.56)	1.00 (1.00–1.00)	1.00 (1.00–1.00)	0.65 (0.47–0.83)	0.70 (0.54–0.86)

AHI = apnea-hypopnea index, CI = confidence interval, HSAT = home sleep apnea testing, LSAT = laboratory sleep apnea testing, NPV = negative predictive value, OSA = obstructive sleep apnea, PPV = positive predictive value, PSG = polysomnography.

The receiver operating characteristic curve analysis for patients with AHI_PSG ≥ 15 events/h showed an area under the curve (AUC) for HSAT of 0.88 (95% confidence interval [CI], 0.69–1.00; Figure 2A ), just slightly lower than that for the LSAT (0.95; 95% CI, 0.88–1.00; Figure 2B ).

Figure 2. ROC curves for AHI and ODI from HSAT and LSAT.

(A) HSAT. (B) LSAT. AHI = apnea-hypopnea index, AUC = area under the curve, CI = confidence interval, HSAT = home sleep apnea testing, LSAT = laboratory sleep apnea testing, ODI = oxygen desaturation index, ROC = receiver operating characteristic.

We determined optimal cutoff values from HSAT to predict at least mild, at least moderate, or severe SDB based on the AHI_PSG. For AHI_PSG ≥ 5 events/h, the optimal cutoff of the AHI_HSAT was 4.4 events/h (sensitivity 0.89, specificity 1). For AHI_PSG ≥ 15 events/h, the optimal cutoff of the AHI_HSAT was 4.8 events/h (sensitivity 0.96, specificity 0.75). For AHI_PSG ≥ 30 events/h, the optimal cutoff of the AHI_HSAT was 19.8 events/h (sensitivity 0.54, specificity 0.88). These modified cutoffs resulted in improved sensitivity at the cost of lower specificity, but the sensitivity remained low for severe SDB.

We performed an analysis assessing the RDI from HSAT as a predictor of the RDI_PSG ( Table 4 ), using the same severity cutoff values as for the AHI. We found that the sensitivity of the RDI was higher at all cutoffs compared with the AHI. However, the specificity was reduced particularly for the cutoff value of 15 events/h. The AUC was therefore lower than for the AHI, both for HSAT (0.75; 95% CI, 0.49–1.00) and for LSAT (0.91; 95% CI, 0.80–1.00; see Figure S1 ^{(201.1KB, pdf)} and Table S3 ^{(201.1KB, pdf)} in the supplemental material).

Table 4.

Performance of HSAT compared to PSG using RDI cutoffs.

PSG	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)	Accuracy (95% CI)
RDI ≥ 5 events/h	RDIHSAT ≥ 5 events/h
	1.00 (1.00–1.00)	NA	1.00 (1.00–1.00)	NA	1.00 (1.00–1.00)
RDI ≥ 15 events/h	RDIHSAT ≥ 15 events/h
	0.84 (0.70–0.98)	0.60 (0.17–1.00)	0.91 (0.79–1.00)	0.43 (0.06–0.80)	0.80 (0.66–0.94)
RDI ≥ 30 events/h	RDIHSAT ≥ 30 events/h
	0.57 (0.31–0.83)	0.81 (0.62–1.00)	0.73 (0.47–0.99)	0.68 (0.47–0.89)	0.70 (0.54–0.86)

CI = confidence interval, HSAT = home sleep apnea testing, LSAT = laboratory sleep apnea testing, NPV = negative predictive value, PPV = positive predictive value, PSG = polysomnography, RDI = respiratory disturbance index.

The performance analysis of ODI ≥ 5 events/h is summarized in Table 5 for HSAT (see Table S4 ^{(201.1KB, pdf)} in the supplemental material for LSAT). The overall accuracy of HSAT was high, at 90% (95% CI, 79%–100%), with a sensitivity and specificity of 95% and 78%, respectively. The AUC for the ODI from HSAT was 0.94 (95% CI, 0.85–1.00; Figure 1 ).

Table 5.

Performance of HSAT compared to PSG using ODI.

PSG	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)	Accuracy (95% CI)
ODI ≥ 5 events/h	ODIHSAT ≥ 5 events/h
	0.95 (0.86–1.00)	0.78 (0.51–1.00)	0.91 (0.79–1.00)	0.88 (0.65–1.00)	0.90 (0.79–1.00)

CI = confidence interval, HSAT = home sleep apnea testing, LSAT = laboratory sleep apnea testing, ODI = oxygen desaturation index, NPV = negative predictive value, PPV = positive predictive value, PSG = polysomnography.

Variability between HSAT and PSG

The bias in the AHI between PSG and HSAT was –10.8 ± 24.7 events/h ( Figure 3 ), with HSAT consistently underestimating the AHI. The difference between RDI_PSG and RDI_HSAT was –7.4 ± 28.5 events/h. The ODI showed no clinically relevant bias. The 95% limits of agreement were wide for all 3 measures (AHI, RDI, ODI), with the RDI_HSAT showing the most variability. Visual inspection of the graphs suggests that HSAT underestimated the AHI and RDI more at higher values. Results for LSAT were similar ( Figure 3 ).

Figure 3. Modified Bland-Altman plots for HSAT and LSAT values compared to PSG values.

(A) AHI from HSAT. (B) AHI from LSAT. (C) RDI from HSAT. (D) RDI from LSAT. (E) ODI from HSAT. (F) ODI from LSAT. AHI = apnea-hypopnea index, HSAT = home sleep apnea testing, LSAT = laboratory sleep apnea testing, ODI = oxygen desaturation index, PSG = polysomnography, RDI = respiratory disturbance index.

Relationship between SDB and lung function

We assessed correlations between SDB variables (AHI, ODI, time with oxygen saturation below 90%) and measures of respiratory function (FVC%, maximal inspiratory pressure, and the partial pressure of CO₂). Although the AHI and ODI were not significantly correlated with measures of respiratory function, time with oxygen saturation below 90% on both PSG and HSAT showed an association with daytime capillary blood pCO₂ and lower FVC% (see Table S1 ^{(201.1KB, pdf)} , Table S2 ^{(201.1KB, pdf)} , Figure S1 ^{(201.1KB, pdf)} , and discussion in the supplemental material).

Patient questionnaires

When participants or their parents were asked if they had difficulty installing the HSAT device, nine (30%) answered yes and 21 (70%) answered no (30 valid responses). The majority of participants (60.6%) also indicated that they slept worse or much worse than usual with the HSAT device. Nevertheless, 60.6% answered that they would prefer repeating HSAT rather than undergoing an in-laboratory PSG.

Of the 27 participants who provided comments about HSAT, positive comments included comfort (70.3%, n = 19), home environment/not having to sleep in hospital (48.1%, n = 13), less nervous (14.8%, n = 4), easy to install (14.8%, n = 4), and the ability to have their family with them (14.8%, n = 4). Negative comments identified by participants included not being certain about the installation (25.9%, n = 7), trouble sleeping with machine (14.8%, n = 4), too many wires (7.4%, n = 2), needing to take machine back to the sleep lab (3.7%, n = 1), the machine moving (3.7%, n = 1), not being able to sit (3.7%, n = 1), having to repeat the test (3.7%, n = 1), and not being able to lay on stomach (3.7%, n = 1).

Conversely, participants mentioned the presence of a technician (48.1%, n = 13), the thoroughness of the test (7.4%, n = 2), and an obligation to do the test (7.4%, n = 2; responses from adolescent patients’ parents) as the benefits with respect to PSG. Drawbacks to doing PSG testing were too many wires (55.5%, n = 15), not being in own bed (40.7%, n = 11), difficulty falling asleep (25.9%, n = 7), inability to use the restroom (22.2%, n = 6), issues with the sleep lab environment (7.4%, n = 2), not usual bedtime (3.7%, n = 1), and “unhappy with the sleep technologist (3.7%, n = 1).

DISCUSSION

We found that HSAT using the Alice PDX monitor is feasible in a selected group of patients with NMD at risk for SDB. In total, 30 of the 35 participants who attempted the test provided interpretable HSAT results on the first or second try. Feasibility in adolescent patients was poorer, with only 1 of 6 successful recordings on the first try and 2 of 4 attempted repeats. We found high specificity at all 3 predetermined cutoff values of the AHI but a generally low sensitivity for an AHI ≥ 15 events/h and for an AHI ≥ 30 events/h ( Table 3 ). Hence, the AHI measured using HSAT is effective at ruling in SDB but less effective at ruling out SDB. However, the sensitivity was higher when using the RDI, such that this metric might be used for screening but PSG would be required for final diagnosis given the lower specificity of the RDI for AHI ≥ 15 events/h.

Performance

As expected, the accuracy of LSAT was higher than that of HSAT. The LSAT was done on the same night as PSG, ³³ acting as an internal control to assess performance of the Alice PDX device under technically “ideal” conditions in patients with NMD. In this regard, the LSAT studies benefited from the presence of a PSG technologist who installed the device and supervised the recording. In addition, given that HSAT was done on a separate night from the PSG, some differences were expected because of night-to-night variability. ³⁴ The AUCs of LSAT for AHI thresholds of 5, 15, and 30 events/h were between 91% and 98% (highest for AHI ≥ 5 events/h), and those of HSAT were between 79% and 91% (highest for AHI ≥ 5 events/h). these are within the reported AUC ranges published in the literature for HSAT in the general population. ³⁵

The portable monitor tests consistently underestimated the AHI compared with PSG. This finding was expected because of the absence of EEG recording and sleep staging with the portable device such that total recording time rather than total sleep time was used for AHI calculations. Moreover, because EEG arousals cannot be directly evaluated with the portable device, hypopneas associated with arousals but without oxygen desaturation could not be scored. Bland-Altman analyses showed that HSAT (and LSAT) on average underestimated the AHI, and the limits of agreement were fairly wide particularly at higher AHI values, which is akin to general population data. ^23,26 On the other hand, because the RDI_HSAT included events with autonomic arousals (surrogates of EEG arousals), it was a closer estimate of RDI_PSG. However, specificity was affected such that the AUC was lower than for the AHI. The more sensitive RDI_HSAT measure could be useful as a screening test, to be confirmed with full PSG when AHI_HSAT results are inconclusive.

HSAT showed the highest AUC for the ODI. Mean ODI values were similar between tests (no bias). However, there was both over- and underestimation of the ODI by HSAT, which was more apparent at higher ODI values ( Figure 3 ). Because the ODI was considerably lower than the AHI as a result of nondesaturating respiratory events, the ODI alone was not an adequate measure to rule in or rule out SDB in this specific population.

Feasibility and patient preference

The failure rate of HSAT on the first try was relatively high (25.7%). If a second attempt was included, then the failure rate was lower (14.2%). Although the American Academy of Sleep Medicine recommends a full PSG if HSAT is inconclusive or unsuccessful, ³⁶ for those who are unable or unwilling to undergo PSG, a second attempt remains a practical option, and our data suggest that it may provide a diagnosis. The failure rate in this study was higher than the rate of 10.3% previously reported in the general sleep clinic population, ²³ and in our clinical laboratory, at 1%–7%. ³⁷ It is possible that motor symptoms or anatomical changes present in NMD could impede the proper installation of the device by the patients themselves or by the caregiver. Sensors may also be more prone to displacement during the night. In adolescent patients, success on the first attempt was low (16.7%) and repeat testing was only successful in 50% of patients. A possible reason for the high failure rate may be the adolescent patients’ discomfort during the night and removal of the device, consciously or not. Results for adolescents should be considered preliminary because of the limited sample size and may be related to specific participant characteristics. To increase the feasibility of HSAT in this population, strategies will need to be studied that might include less-intrusive devices, direct patient education to increase self-efficacy, or increased supervision during the night. ³⁸ Nevertheless, globally, the success rate in this study suggests that HSAT is feasible in selected patients with NMD.

The majority of participants reported no difficulty in installing the device and stated that they would rather use an HSAT device than undergo laboratory PSG. The available literature on patient preferences is consistent with our results. Because HSAT is more patient-centered and comfortable for patients, it is often the preferred choice. ³⁹ However, if a first HSAT is negative, inconclusive, or technically inadequate, then patients may prefer to proceed directly to PSG rather than repeat the HSAT to avoid inconvenience, because it is more likely to produce a definitive result. ³⁶ When PSG remains impractical in a patient with NMD, a repeat HSAT could be a reasonable option.

PSG was the preferred test for a nonnegligible proportion of our participants (40%). The presence of the technician and the thoroughness of the test were identified as positive aspects. Parents of adolescents appreciated knowing the test would be fruitful once their child was onsite. However, there were multiple drawbacks reported by most patients, mainly related to discomfort from the equipment or the sleep laboratory environment. Sleep laboratories are typically not well-equipped or adapted for patients with disabilities, making PSG a particular challenge in patients with NMD.

Clinical relevance

HSAT carries a lower testing burden compared with PSG and could allow for improved access to diagnostic testing in patients otherwise unable to get tested because of a remote location or lack of adapted sleep testing facilities. For these reasons, there is increasing interest in HSAT in patients with NMD. ^19,40,41 In the general population, guidelines suggest that using HSAT in conjunction with a comprehensive sleep evaluation may be an alternative to PSG for the diagnosis of OSA in patients with a high pretest probability of moderate to severe OSA. ¹⁹ Guidelines currently do not recommend HSAT for the diagnosis of SDB in patients with significant comorbid medical conditions such as NMD, primarily because of insufficient data. ¹⁹ A review on SDB in Duchenne muscular dystrophy found 3 studies reporting HSAT in Duchenne muscular dystrophy and other NMD, although a level 2 device was used in 1 of the studies. ⁴² Only 1 study compared HSAT to PSG and was limited to 10 patients, of whom 3 had an AHI > 3 events/h. ²⁰ A recent study of children aged 6–18 years with NMD found a low failure rate but a somewhat lower sensitivity and specificity of a level 3 HSAT for SDB than we found in this study. ⁴³ Hypoventilation, present in 25% of the cohort in that study, was not detected by HSAT.

Note that individuals with NMD who already have daytime hypercapnia should be considered for noninvasive ventilation without diagnostic sleep testing. ^44,45 However, when daytime pCO₂ is normal, sleep testing may indicate SDB in the form of upper-airway obstruction or hypoventilation because of respiratory muscle weakness, both of which may appear as hypopneas, particularly in rapid eye movement sleep. ⁴ Although we found that nocturnal hypoxemia was correlated with daytime hypercapnia and reduced FVC%, some individuals with hypercapnia had very little sleep-related hypoxemia (Figure S1 ^{(201.1KB, pdf)} ). Hence, an abnormal HSAT may not be able to differentiate OSA from hypoventilation, and specific treatment decisions should consider the global clinical picture, such as the degree of underlying respiratory muscle weakness and restrictive pulmonary impairment. In most instances, noninvasive ventilation rather than continuous positive airway pressure will be the therapy of choice if subclinical hypoventilation is a possibility, particularly in NMD characterized by progressive respiratory muscle weakness over time.

Limitations

This study is not without limitations. First, the sample size was limited, especially for the adolescent subgroup. Second, although SDB was prevalent in our study population, the majority of the participants were included based on clinical risk of OSA and few had a significant restrictive pulmonary syndrome putting them at risk for sleep-related hypoventilation. Hence, our results may not be applicable to those with more severe respiratory impairment. However, in patients with the most severe impairment, sleep testing is unnecessary, as noted above. Another limitation is that patients who were unable to install the HSAT device independently, or who could not obtain help from a caregiver, were excluded. Our results are therefore only applicable to a selected population of patients with NMD who may have less physical impairment from their NMD or who have help from a caregiver. Moreover, this study tested a specific device (Alice PDX), and therefore findings from our study may not be generalizable to all home testing devices, particularly those with fewer or different channels (eg, single-effort band or peripheral arterial tonometry signal–based HSAT devices). The device used has the advantage of having 2 respiratory effort bands, allowing the detection of paradoxical breathing and upper-airway obstruction with more precision, but it is fairly bulky, which can potentially be uncomfortable. Finally, we have not evaluated clinical outcomes related to the detection of SDB in our participants. The testing of an algorithm including treatment initiation based on HSAT results will be needed to ascertain adequacy in that regard in comparison to PSG.

CONCLUSIONS

Our study shows the general feasibility of HSAT among adults with NMD, but results for adolescents were less conclusive. Patient preference was in favor of HSAT. Providing patients with the option of HSAT may improve testing uptake and the timely diagnosis and management of SDB. Although the AHI measured by HSAT showed high specificity with lower sensitivity, the RDI had the opposite characteristics for detecting RDI ≥ 15 events/h. Hence, the RDI can be used to complement the AHI to gain the maximal amount of information from HSAT. For example, when the AHI is nondiagnostic but the RDI is elevated, the patient could be prioritized for PSG to clarify the diagnosis. Additional studies with a larger cohort are needed to confirm our findings. Further work in this area should focus on novel HSAT technology to optimize the diagnosis of OSA and sleep-related hypoventilation in children and adults with NMD.

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. This study was funded by Muscular Dystrophy Canada. MK is a member of the advisory board at Biron Soins du Sommeil and receives unrestricted research support from VitalAire, Philips Respironics, and Fisher Paykel. BJP is a consultant for Sanofi Genzyme. MO has served as site principal investigator for clinical trials in spinal muscular atrophy funded by Ionis, Biogen, Roche, and Cytokinetics, and has been reimbursed for travel related to these trials. She has also served on the Avexis data safety monitoring board. The other authors report no conflicts of interest.

ABBREVIATIONS

AHI

apnea-hypopnea index

AUC

area under the curve

CI

confidence interval

EEG

electroencephalogram

FVC

forced vital capacity

HSAT

home sleep apnea testing

LSAT

laboratory sleep apnea testing

NMD

neuromuscular disorders

ODI

oxygen desaturation index

OSA

obstructive sleep apnea

PSG

polysomnography

RDI

respiratory disturbance index

SDB

sleep-disordered breathing