Inter‐Night Variability of Nocturnal Pulse Oximetry in People Living With Motor Neuron Disease: A Retrospective Observational Study

ABSTRACT

Background and Objective

Nocturnal pulse oximetry (NPO) is a simple and inexpensive assessment tool that has previously been shown to correlate with prognosis and timing of non‐invasive ventilation (NIV) initiation in people living with motor neuron disease (plwMND). However, the optimal number of nights for measuring NPO has not been defined for this population, with other respiratory conditions exhibiting both low and high night‐to‐night variability in NPO parameters. This study aims to determine the inter‐night variability in NPO data over three nights in plwMND.

Methods

We conducted a retrospective analysis of 132 studies in which plwMND underwent three consecutive nights of NPO. Intraclass correlation coefficients (ICC) were used to assess the reliability of key NPO parameters, including mean percentage of total recording time with oxygen saturation (SpO₂) < 90% (T90), oxygen desaturation index (ODI), basal SpO₂ and nadir SpO₂. The proportion of plwMND meeting NIV criteria based on single‐night versus multi‐night assessments was also compared.

Results

Excellent reliability was observed for T90 (ICC(1) = 0.940) and ODI (ICC(1) = 0.901), while basal SpO₂ (ICC(1) = 0.845) and nadir SpO₂ (ICC(1) = 0.768) demonstrated good reliability. However, relying on a single‐night NPO assessment failed to identify 12% of plwMND who met NIV criteria when evaluated over three nights.

Conclusion

Despite good to excellent inter‐night variability of NPO data in plwMND, multi‐night NPO monitoring improves the accuracy of identifying plwMND requiring NIV. These findings support the need for multi‐night assessments to enhance clinical decision‐making in MND management.

Short abstract

We aimed to determine the inter‐night variability of nocturnal pulse oximetry (NPO) parameters in people living with Motor Neuron Disease (plwMND). Although there was good to excellent reliability of NPO data, single‐night NPO assessment failed to identify 12% of plwMND who met NIV criteria when evaluated over three nights.

1. Introduction

Motor neuron disease (MND) encompasses a group of neurodegenerative disorders characterised by the progressive loss of motor neurons [1], leading to muscle weakness, respiratory failure, and death [2]. The initial respiratory manifestation is sleep‐disordered breathing (SDB) [3], caused by impaired respiratory muscle function and reduced upper airway tone [4]. In sleep studies of people living with MND (plwMND), nocturnal hypoventilation is the most frequently observed form of SDB, primarily attributed to periods of non‐obstructive hypopnoeas and central apnoeas [4]. Non‐invasive ventilation (NIV) is commonly used to manage respiratory insufficiency in plwMND, extending survival, improving quality of life and slowing the decline in respiratory function [5, 6, 7, 8, 9, 10, 11]. Additionally, early initiation of NIV has been associated with improved survival and adherence [11, 12].

Although there is no consensus on the most effective modality for predicting prognosis or determining the optimal timing of NIV initiation [13], nocturnal pulse oximetry (NPO) is a widely accepted screening tool for determining the timing of NIV initiation [11]. This is due to a variety of factors including cost‐effectiveness, accessibility, and convenience [14], which is important for plwMND who often have high care needs. Compared to forced vital capacity (FVC), the volume of air one can forcibly exhale following maximal inspiration and one of the main parameters utilised for determining the timing of NIV initiation, NPO offers greater sensitivity in detecting early respiratory insufficiency [15]. NPO has the added advantage of mitigating the confounding effects of variable volitional effort, particularly in the presence of significant bulbar muscle weakness. In this subset of plwMND, slow vital capacity (SVC), the volume of air slowly exhaled after maximal inspiration, may provide a more accurate assessment of respiratory function than FVC [16]. Nocturnal desaturation with > 10% of total sleep time with SpO₂ < 90% using NPO in plwMND was associated with more rapid onset of respiratory failure (described as PaO₂ < 60 mmHg or SpO₂ < 90%) and more frequent hospitalisation [17].

Despite its recommendation as an assessment tool for respiratory insufficiency in plwMND, the optimal number of nights for measuring NPO is not specified in guidelines, including the UK National Institute for Health and Care Excellence (NICE) [18], the American Academy of Neurology (AAN) [19], the European Federation of Neurological Societies (EFNS) [20] and the Canadian Amyotrophic Lateral Sclerosis (ALS) Best Practice Recommendations [21]. High night‐to‐night variability in NPO parameters has been reported in conditions such as obstructive sleep apnoea [22] and chronic obstructive pulmonary disease [23], whereas low night‐to‐night variability has been observed in chronic heart failure [24]. To our knowledge, the inter‐night variability of NPO parameters in plwMND has yet to be investigated.

A better understanding of night‐to‐night variability of NPO parameters is important as the findings are used to help determine the timing of NIV initiation. It is particularly important to know the risk of false negative findings and potential treatment delay. Some causes of false negative findings include reduced sleep efficiency, poor quality sleep (e.g., Rapid eye movement sleep restriction) or positional variability, the information for which is not available in a level 4 study. However, a protocol of three nights of recordings reduces the availability of oximeters in a clinical setting. To justify this practice, multiple nights of NPO would need to be shown to improve the detection of individuals who would qualify for NIV compared to a single night of NPO. Government funding for NIV devices requires > 2% of total sleep time with oxygen saturation (SpO₂) < 90% as one criterion to provide equipment for plwMND in our region [25]. In our multidisciplinary MND outpatient clinic, three nights of NPO are usually performed when assessing for SDB, based on precedent in the general sleep apnoea population [26] and pragmatism due to the risk of failed recordings and delays associated with reissuing devices.

In the present study, our primary aim was to assess inter‐night variability in NPO data over three nights in plwMND. Secondary aims were to assess the relationship between this inter‐night variability and other pulmonary function measures (erect and supine SVC, sniff nasal inspiratory pressure (SNIP) and peak cough flow (PCF)) and to explore associations with orthopnoea and reduced PCF. Finally, we evaluated differences in NIV eligibility when using data from a single night versus three nights of NPO.

2. Methods

We conducted a retrospective observational study of plwMND who underwent diagnostic NPO testing between January 2015 and December 2022 at a single tertiary referral centre. Studies were excluded if there was insufficient NPO data (fewer than 3 nights of adequate oximetry quality recordings), if the NPO was performed with the use of supplemental oxygen or any positive airway pressure support, or if the person had an alternative neurodegenerative diagnosis. Adequate oximetry quality was defined as a total recording time of non‐artefactual trace of at least 4 h on each night of oximetry. Up to three separate episodes of NPO monitoring could be included for each participant. If more than three eligible NPO recording periods had been conducted prior to NIV initiation, the three study periods with the most complete data were included for analysis. The study was approved by the Sydney Local Health District Human Ethics Research Committee (RPA zone, X20‐0238 & 2020/STE02849).

2.1. Clinical Data

Baseline demographic data were collected from the participants' clinical history and electronic medical records. The following clinical characteristics were collected at the time of NPO: presence of orthopnoea, erect and supine SpO₂, erect and supine SVC, SNIP, PCF and symptomatic breathlessness measured in the erect and supine positions using a modified Borg dyspnoea scale [27]. Predicted SVC and SNIP were calculated in accordance with established reference ranges [28, 29].

2.2. NPO

NPO was performed using a wrist pulse oximeter (WristOx2, model 3150, Nonin Medical Inc., Plymouth, MN, USA) over three consecutive nights with a sampling rate of one sample per second. Oximetry data was extracted and analysed using the manufacturer's software (Nvision, Nonin Medical). The inter‐night variability of four primary outcomes were of interest in this study: (1) basal SpO₂, defined as the average oxygen saturation not included in any desaturation events; (2) the lowest oxygen saturation measured during the night (nadir SpO₂); (3) the percentage of time spent with a SpO₂ below 90% during the night (T90); (4) oxygen desaturation index (ODI), defined as the number of times per hour that SpO₂ drops by 3% or greater from baseline during sleep.

2.3. Statistical Analysis

Data were analysed using descriptive statistics for participant demographics, baseline characteristics at the time of NPO and NPO recording data. Primary outcome data were analysed using one‐way repeated measures analysis of variance (ANOVA) and intraclass correlation coefficient one‐way random effects model (ICC(1)). We defined an ICC(1) value less than 0.5 as poor reliability, between 0.5 and 0.75 as moderate reliability, between 0.75 and 0.90 as good reliability and greater than 0.90 as excellent reliability [30]. Secondary outcome data to assess associations between variables were analysed using unadjusted odds ratio (OR) and simple linear regression models. These data was presented with 95% confidence intervals, and a value of p < 0.05 was considered statistically significant. All statistical analyses were performed using Statistical Package for Social Science (SPSS), version 28.0 (SPSS Inc., Chicago, IL, USA).

3. Results

One hundred and twenty‐three participants were identified for review from the clinic database. From these participants, we identified 181 studies of NPO. Forty‐nine studies were excluded: 19 due to fewer than three nights of NPO performed, 22 due to inadequate oximetry quality, three due to NPO performed on CPAP in the setting of concomitant obstructive sleep apnoea and four due to participants already having three eligible studies included. Following exclusion, there were 132 eligible studies from 100 participants. Seventy‐eight participants had one study of three‐night NPO, 12 participants had two studies of three‐night NPO, and 10 participants had three studies of three‐night NPO included for analysis. The flow diagram demonstrating NPO study inclusion for analysis is shown in Figure 1.

FIGURE 1.

Flow diagram of NPO study inclusion. CPAP, continuous positive airway pressure; MND, motor neuron disease; NPO, nocturnal pulse oximetry.

The mean ± SD age of participants at diagnosis was 59.5 ± 11 years. Among this group, 63% of participants were male and the mean BMI was 25.8 kg/m². Limb‐onset MND was the predominant phenotype (56%), followed by bulbar‐onset MND (28%), primary lateral sclerosis (PLS) (5%) and respiratory onset (3%). The median time from symptom onset to diagnosis was 10 months. These findings are summarised in Table 1.

TABLE 1.

Participant demographics.

	N = 100
Age at diagnosis, mean (SD) (years)	59.5 ± 11.3
Sex (n)
Male	63
Female	37
BMI at time of first NPO, mean (SD) (kg/m2)	25.8 ± 6.2
Phenotype (n)
Limb‐onset	56
Bulbar‐onset	28
PLS	5
Respiratory‐onset	3
Not recorded	8
Time from symptoms to diagnosis (months), median	10 (6–24.5)

Note: Values are means ± SD or medians (interquartile range).

Abbreviations: BMI, body mass index; N, sample size; NPO, nocturnal pulse oximetry; PLS, primary lateral sclerosis; SD, standard deviation.

The median duration of symptoms of MND at the time of the NPO study was 27 months and the median duration from diagnosis was 14 months. In 69% of the studies, participants did not report orthopnoea. These are presented alongside the baseline pulmonary function parameters in Table 2. Mean NPO parameters were similar across the three consecutive nights of monitoring and are presented in Table 3.

TABLE 2.

Baseline measurements at time of NPO.

	N
Age (years)	132	62.5 ± 12.2
Time since symptoms of MND (months)	112	27 (16–56)
Time since diagnosis of MND (months)	124	14 (8–27)
Orthopnoea
Yes	28
No	91
Not recorded	13
Clinic SpO2 erect (%)	109	96.3 ± 1.77
Modified Borg erect	31	0.44 ± 1.02
Clinic SpO2 supine (%)	53	95.6 ± 1.68
Modified Borg supine	30	0.92 ± 1.55
SVC predicted, erect (%)	109	64.8 ± 21.3
SVC predicted, supine (%)	81	58.0 ± 22.5
Drop in SVC from erect to supine (%)	81	16.0 ± 19.6
SNIP predicted (%)	87	33.8 ± 16.8
PCF (L/min)	102	234 ± 97.9

Note: Values are means ± SD or medians (interquartile range).

Abbreviations: N, sample size; NPO, nocturnal pulse oximetry; PCF, peak cough flow; SD, standard deviation; SNIP, sniff nasal inspiratory pressure; SpO₂, oxygen saturation; SVC, slow vital capacity.

TABLE 3.

NPO recording data.

		N = 132
Night 1	Duration of recording (h:min)	08:31 ± 01:37
	Basal SpO2 (%)	92.3 ± 2.25
	Nadir SpO2 (%)	81.3 ± 6.63
	T90 (%)	12.3 ± 21.7
	ODI (/h)	13.7 ± 10.9
Night 2	Duration of recording (h:min)	08:42 ± 01:25
	Basal SpO2 (%)	92.4 ± 2.24
	Nadir SpO2 (%)	81.4 ± 6.53
	T90 (%)	12.0 ± 22.0
	ODI (/h)	13.7 ± 11.0
Night 3	Duration of recording (h:min)	08:25 ± 01:30
	Basal SpO2 (%)	92.4 ± 2.24
	Nadir SpO2 (%)	81.0 ± 7.54
	T90 (%)	12.4 ± 21.7
	ODI (/h)	13.0 ± 10.4

Note: Values are means ± SD.

Abbreviations: N, sample size; ODI, the number of times per hour that SpO₂ drops by 3% or greater; SD, standard deviation; SpO₂, oxygen saturation; T90, the percentage of time spent with a SpO₂ below 90% during the night.

3.1. Primary Outcomes

Basal SpO₂ and nadir SpO₂ demonstrated good reliability across three nights of NPO (ICC(1), 0.845; 95% CI, 0.800–0.883 and ICC(1), 0.768; 95% CI, 0.705–0.821, respectively). There was excellent reliability of T90 and ODI across three nights of NPO (ICC(1), 0.940; 95% CI, 0.921–0.955) and (ICC(1), 0.901; 95% CI, 0.871–0.926, respectively) (Figure 2).

FIGURE 2.

Primary outcome: The relationship between the inter‐night variability of the four NPO parameters. ICC(1), intraclass correlation coefficient using a one‐way random‐effects model across all three nights; ODI, the number of times per hour that SpO₂ drops by 3% or greater; SpO₂, oxygen saturation; T90, the percentage of time spent with a SpO₂ below 90% during the night.

3.2. Secondary Outcomes

Simple linear regression analyses of the covariates of SVC % predicted (erect and supine), SNIP % predicted, and PCF did not demonstrate any statistically significant association with the coefficient of variation (CV) of any of the NPO parameters. These findings are presented in Figure S1 and Table S1.

Participants with orthopnoea at the time of NPO testing had a 3.25‐fold higher likelihood of qualifying for NIV based on oximetry alone (85.7% vs. 64.8% among those without orthopnoea; OR = 3.25; 95% CI, 1.04–10.2). There was no increased risk of qualifying for NIV associated with a PCF below 270 L/min (71.0% vs. 62.5%; OR = 1.47; 95% CI, 0.63–3.41). These findings are presented in Table S2. Among the 94 NPO studies that met NIV criteria (T90 > 2% on at least one night), 11 (12%) would not have qualified based on first‐night data alone and 5 (5%) would not have qualified based on the first two nights of data alone. These findings are presented in Table 4.

TABLE 4.

Comparison of NIV qualification based on three‐night vs. one‐night NPO.

	Qualified for NIV a (three nights of oximetry) N = 94
Qualified for NIV using the first night of oximetry	83 (88%)
Did not qualify for NIV using the first night of oximetry	11 (12%)
SVC < 50%	5 (5%)
50% ≤ SVC < 80%	2 (2%)
SVC ≥ 80%	1 (1%)
SVC data not available	3 (3%)
Did not qualify for NIV using the first and second night of oximetry	5 (5%)

^a Definition of qualifying for NIV was determined on the basis of T90 > 2% on at least a single night of NPO.

4. Discussion

While various international guidelines recommend the use of NPO to screen for respiratory insufficiency in plwMND, there is no consensus on the optimal number of nights required for accurate assessment. In this study, our novel finding is that T90 and ODI exhibited excellent reliability, while basal SpO₂ and nadir SpO₂ showed good reliability in plwMND. Additionally, the reliability of these parameters was not significantly influenced by pulmonary function measures, including erect and supine SVC, SNIP and PCF. Nadir SpO₂ represents a single time point which can be attributed to artefact or a true nadir by software, which without expert review is inherently more susceptible to inter‐night fluctuations, consistent with our findings. Nadir SpO₂ adds clinical context regarding the severity of overnight desaturation; however, it requires expertise to interpret from NPO as it can be misleading due to artefact error.

Although the Canadian ALS Best Practice Recommendations, AAN and EFNS recommend NIV initiation based on ‘abnormal nocturnal oximetry’ and ‘nocturnal desaturation’, this term is not explicitly defined in their guidelines [21, 31]. Using our local funding criterion threshold of T90 > 2% of the night for NIV eligibility, a single‐night NPO assessment would have failed to identify 11 of 94 eligible participants (12%) for NIV initiation. Thus, despite our findings of strong reliability across NPO measures, relying solely on a single night of NPO may lead to under‐detection of respiratory insufficiency in plwMND, potentially delaying access to NIV. However, this under‐detection would have been mitigated by incorporating pulmonary function parameters, as recommended by AAN, EFNS and NICE; whose cutoffs vary between FVC/SVC < 50% and FVC/SVC < 80% [31]. For these cases, SVC < 50% threshold would have reduced the number of missed cases to three, while an SVC < 80% threshold would have identified all but one missed case. This threshold of SVC < 80% is further supported by a randomised controlled trial suggesting that early NIV initiation in the stages of respiratory failure may improve survival [32]. Additionally, 22 out of the initial 181 NPO studies identified in our clinic database were excluded due to inadequate recording quality, further underscoring the limitations of a single‐night NPO approach. This variability in data quality may result in delayed NIV initiation, or unnecessary cost and distress when repeated testing is required. It also raises the importance of expert clinician review of recordings, specifically in identifying periods of artefact to ensure interpretability of NPO parameters.

Our study further describes the association between orthopnoea and increased likelihood of qualifying for NIV [33, 34]. This aligns with current recommendations from the Canadian ALS Best Practice Recommendations, AAN, and NICE guidelines [18, 19, 21], which suggest incorporating orthopnoea as a criterion for NIV initiation. This finding strengthens the argument for including orthopnoea in clinical screening tools for respiratory insufficiency in plwMND [35]. In contrast, a reduced PCF was not found to correlate with increased NIV eligibility, possibly due to the volitional nature of cough and the potential for confounding factors such as an inadequate interface seal, particularly in people with bulbar dysfunction.

A maximum of three eligible NPO studies taken at individual time points were used per participant. This threshold was determined as a balance between temporal changes given the rapidly progressive nature of MND and the risk of sampling bias. Each NPO study was treated as an independent observation for analysis; however, we do not expect this would significantly bias results given the small proportion of people for whom repeated screening was performed. Additionally, the broad eligibility criteria enabled us to assemble a diverse cohort that closely mirrors the general MND population [36], enhancing the generalisability of our findings.

One of the study's strengths is the use of a single pulse oximeter device across all participants, eliminating potential variability related to different device models which has been observed in other populations [37, 38, 39]. However, this may limit the generalisability of our findings, as results could differ with other oximeter models and brands. Additionally, although oximetry data were automatically scored by software, all results were manually checked by two experienced sleep and respiratory physicians, with manual exclusion of obvious artefact.

There are several limitations to our study. This dataset includes all people with MND meeting inclusion criteria and did not exclude people with a non‐neuromuscular diagnosis. As a result, desaturations detected by NPO could be influenced by other comorbid conditions such as underlying lung disease or congestive heart failure [40]. This may contribute to more variability but is a reflection of the clinical cohort. As a retrospective study, there was significant incomplete data with regards to lung function parameters such as supine FVC and Borg scale, reducing our statistical power and ability to perform additional regression analyses. Furthermore, the inclusion of additional pulmonary function measures such as maximal inspiratory pressure (MIP) would have been valuable, as it is recommended as a screening parameter by the NICE and AAN guidelines for NIV initiation. In addition, MIP has demonstrated a correlation with nocturnal hypoventilation and prognosis in MND [13]. However, due to the presence of bulbar dysfunction or weak oral musculature, it was not always easy for participants to perform a MIP manoeuvre. Consequently, SNIPs were performed instead. Notably, 28% of our cohort had bulbar‐onset MND, which typically complicates pulmonary function testing, including PCF. These factors contributed to missing lung function data, which was further exacerbated by the increase of telehealth during the COVID‐19 pandemic. Thus, our ability to perform robust regression analyses across all NPO and respiratory data was limited.

In conclusion, our study demonstrates that the inter‐night variability of NPO parameters in plwMND shows good to excellent reliability across three consecutive nights. However, relying on a single‐night NPO assessment would risk under‐detection of respiratory failure in 12% of people eligible for NIV therapy. While we advocate for a multi‐night NPO approach in NIV decision‐making for people with MND, judicious use of single‐night oximetry may be useful when the pretest probability is very high based on muscle strength or SVC measures or when it is not feasible to do more than one night recording. Diagnosis and management of respiratory failure in MND require considered multimodal assessment due to significant clinical heterogeneity.

Author Contributions

Steven Luu: data curation (supporting), formal analysis (lead), investigation (equal), visualization (lead), writing – original draft (lead), writing – review and editing (equal). Olivia McGuiness: conceptualization (supporting), data curation (lead), investigation (equal), visualization (supporting), writing – review and editing (equal). Collette Menadue: data curation (supporting), investigation (supporting), writing – review and editing (equal). Amanda J. Piper: project administration (supporting), supervision (supporting), writing – review and editing (equal). Keith Wong: formal analysis (supporting), methodology (supporting), project administration (supporting), supervision (supporting), writing – review and editing (equal). Brendon J. Yee: conceptualization (supporting), project administration (supporting), supervision (supporting), visualization (supporting), writing – review and editing (equal). Emma L. Gray: conceptualization (lead), methodology (lead), project administration (lead), supervision (lead), writing – review and editing (equal).

Ethics Statement

This human study was approved by Sydney Local Health District Human Ethics Research Committee—approval: RPA zone, X20‐0238 & 2020/STE02849. Adult participant consent was not required because this study is a retrospective analysis of existing medical records. Due to the size of the records to be accessed, consent is impracticable. We have taken appropriate measures to protect the privacy and confidentiality of the data, including de‐identification of all personal information.

Conflicts of Interest

A.J.P. is an Editorial Board member of Respirology and a co‐author of this article. She was excluded from all editorial decision‐making related to the acceptance of this article for publication. The other authors declare no conflicts of interest.

Supporting information

Data S1. Supporting Information.

Luu S., McGuiness O., Menadue C., et al., “Inter‐Night Variability of Nocturnal Pulse Oximetry in People Living With Motor Neuron Disease: A Retrospective Observational Study,” Respirology 30, no. 10 (2025): 995–1002, 10.1111/resp.70072.

Data Availability Statement

The data that support the findings of this study are openly available in figshare at http://doi.org/10.6084/m9.figshare.28765754, [41].