Screening for obstructive sleep apnoea in post‐treatment cancer patients

Abstract Background and aims For cancer patients, comorbid obstructive sleep apnea (OSA) poses additional risk to their surgical/anaesthetic outcomes, quality of life, and survival. However, OSA screening is not well‐established in oncology settings. We tested two screening tools (STOP‐Bang questionnaire [SBQ] and the at‐home monitoring device, ApneaLink™Air), for predicting polysomnography (PSG) confirmed OSA in post‐treatment cancer patients. Methods Breast (n = 56), endometrial (n = 37) and melanoma patients (n = 50) were recruited from follow‐up clinics at Westmead Hospital (Sydney, Australia). All underwent overnight PSG, 137 completed SBQ, and 99 completed ApneaLink™Air. Positive (PPV) and negative (NPV) predictive values for PSG‐determined moderate‐to‐severe OSA and severe OSA, were calculated using an SBQ threshold ≥3 au and ApneaLink™Air apnoea‐hypopnea index thresholds of ≥10, ≥15 and ≥30 events/h. Results Both SBQ and ApneaLink™Air had high NPVs (92.7% and 85.2%–95.6% respectively) for severe OSA, but NPVs were lower for moderate‐to‐severe OSA (69.1% and 59.1%–75.5%, respectively). PPV for both tools were relatively low (all <73%). Combining both tools did not improve screening performance. Conclusions These screening tools may help identify cancer patients without severe OSA, but both are limited in identifying those with moderate‐to‐severe or severe OSA. PSG remains optimal for adequately identifying and managing comorbid OSA in cancer patients.

The aetiology of OSA in cancer cohorts is not well understood.
However, a likely candidate is the presence of common risk factors.
Among these, age and obesity are well-known risk factors for both OSA [24][25][26] and a number of cancers. [27][28][29] OSA is a treatable condition, but is often left unrecognised, undiagnosed, and untreated. 30,31 Implementing protocols expediting OSA diagnoses in oncology-clinic settings can facilitate timely OSA treatment, to reduce the burden of comorbid OSA in cancer patients.
An OSA diagnosis is made using overnight, attended, inlaboratory polysomnography (PSG), or unattended home PSG, both of which are finite, labour intensive resources. 32 Alternatively, tools such as questionnaires 33,34 and at-home limited channel monitoring studies, which record between 2 and 8 physiological variables, 35 can be used to identify patients with a high probability of OSA for streamlining the utilisation of finite PSG resources. This approach has been utilised in primary care, 36 and pre-anaesthetic clinics, 33 but has not been widely applied in oncology clinics. The aim of the present study was to investigate the relative ability of two widely used and previously validated community/sleep clinic screening tools to accurately identify cancer patients with a high probability for OSA when deployed in an oncology clinic setting: (1) the STOP-Bang questionnaire (SBQ), 33 and (2) the ApneaLink™Air device (ResMed, Australia). 35 We recruited patients from three different post-treatment oncology clinics: (1) breast cancer; a female cohort with a reported high prevalence of sleep disturbance, 37 (2) endometrial cancer; a female cohort with a high prevalence of obesity (a shared risk factor for OSA), 38 and (3) 11 Patients were eligible to participate if they: (1) were ≥ 18 years of age, (2) had a confirmed diagnosis of either breast cancer, endometrial cancer or melanoma, (3) had completed their treatment regimen (i.e., surgery, chemotherapy, radiotherapy) for a minimum of 2 months (endometrial cancer), 6 months (melanoma), or 12 months (breast cancer) prior to study recruitment, (4) were able to understand instructions relevant to study requirements, and (5) provided informed written consent. Patients were also screened for serious medical conditions via a physician-conducted telephone interview, and were excluded if they had serious respiratory, cardiovascular, hepatic, renal, neurological or psychiatric conditions, or if they were pregnant.

| Polysomnography
All patients underwent a standard clinical practice (single-night) 32 overnight in-laboratory PSG session in the Sleep Research Facility, at the Westmead Institute for Medical Research (Sydney, Australia). The following signals were measured: nasal pressure, oronasal thermistor signals, snoring auditory signals, thoracic and abdominal inductive plethysmography, pulse oximetry (SpO 2 ), EEG, EOG, and chin, diaphragm, and pre-tibial EMG.

| STOP-Bang questionnaire
The SBQ is an eight-item binary (yes/no) validated tool, measuring recognised OSA phenotypic characteristics. 33 The 'STOP' section contains four questions, each capturing a self-reported OSA symptom (i.e., snoring, tiredness, observed apnoea and high blood pressure). The 'Bang' section records four additional demographic conditions (BMI ≥35 kg/m 2 , age ≥ 50 years, neck circumference ≥40 cm and male gender). Patients self-completed the questionnaire, prior to undergoing PSG.

| ApneaLink™Air
Patients received an at-home limited channel monitoring device kit (ApneaLink™Air kit; ResMed, Sydney, Australia) with printed instructions, either during recruitment or upon PSG completion. The device records respiratory airflow from a nasal cannula, respiratory effort via a thoracic movement 'effort' sensor, and blood oxygen saturation and pulse rate using a finger pulse oximeter. After self-administered usage at home for one night, the kit was returned in-person or via a pre-paid postal envelope.

| STOP-Bang questionnaire
'Yes' responses were scored '1,' and 'No' responses were scored '0.' A total SBQ score was determined by summing the 'yes' responses. The total SBQ score can range between 0 and 8 arbitrary units (au). 33

| ApneaLink™Air data
ApneaLink™Air data were downloaded onto a computer for analysis.
Data were analysed and scored automatically using commercially available software (ApneaLink™Air Application Software Multilingual, ResMed Australia). The evaluation period for each recording was the total recording period minus the first 10 min, and sections with poor quality signals. If the evaluation period was <120 min, the recording was excluded. The recordings were reviewed by an experienced sleep technician for acceptable technical quality and to confirm all identified events met AASM-2012 criteria. 40 Events were re-scored by the sleep technician if required.

| Patient characteristics
Anthropometric and demographic data for all patients are displayed in

| ApneaLink™Air data
Of the 117 patients who received an ApneaLink™Air device,

| SBQ scores
An SBQ≥3 au, produced modest PPVs for predicting both moderateto-severe and severe OSA, but a high NPV for severe OSA (see Table 3).

| ApneaLink™Air data
Across all threshold levels, PPV values were at the most modest for predicting both moderate-to-severe and severe OSA. NPVs were highest for predicting severe OSA, but lower for predicting moderateto-severe OSA (Table 3).  18,19 and is higher than for the general female population (13.2%-23.4%). 18,19 Indeeed, prevalence rates arising from PSG data in the present study, approach those reported in sleep-clinic cohorts, 20-23 suggesting OSA prevalence in cancer patients may be as high as in the high-pretest probability environment of a sleep physician referral clinic. Furthermore, given that OSA was not previously clinically recognised in any of the cancer patients included in the present study, these findings emphasise the nature of the unmet need for OSA screening in cancer patients.

| STOP-Bang questionnaire
The SBQ is a straightforward, easy-to-administer OSA screening tool with a simple scoring system. 33 The tool has been validated across   ApneaLink™Air being also less effective in the older age group ( Figure 3). The reason for these outcomes is not clear from our data set.

| CONCLUSION
In a cohort of breast, endometrial and melanoma cancer patients, SBQ data effectively identified patients unlikely to have severe OSA, and shared similar performance characteristics to ApneaLink™Air data.
Both tools were less effective at positively identifying patients with moderate-to-severe or severe OSA. There was no improvement with a 2-step combined tool approach. We conclude that PSG remains the optimal tool for the positive diagnosis and management of comorbid OSA in cancer patient cohorts.