ScreenOX - An Automated At-home Screening Test for Adult Sleep Apnea Using Nocturnal Oximetry
Study Details
Study Description
Brief Summary
The sleep apnea-hypopnea syndrome (SAHS) is a respiratory disorder characterized by frequent breathing cessations (apneas) or partial collapses (hypopneas) during sleep. SAHS is linked with the most important causes of death in adults from industrialized countries. Metabolic deregulation and cardiovascular and cerebrovascular diseases, such as atrial fibrillation, stroke, myocardial infarction and sudden cardiac death, could affect people having untreated SAHS. The gold standard method for SAHS diagnosis is in-hospital, technician-attended nocturnal polysomnography (PSG). Nevertheless, this methodology is labor-intensive, time-consuming, and relatively unavailable, especially in low-resource settings. These drawbacks have led to large waiting lists, which delay diagnosis and treatment and limits its effectiveness as single diagnostic method for SAHS. Blood oxygen saturation (SpO2) and pulse rate (PR) from nocturnal pulse oximetry (NPO) provide relevant and essential information to detect apneas. In addition, it is significantly less intrusive for patients and it can be easily recorded at patients' home. In the same way, automated signal processing and pattern recognition techniques have demonstrated to provide accurate tools able to detect and effectively use this information. Therefore, the investigators hypothesize that automated pattern recognition of at-home NPO recordings could provide reliable and efficient tools able to simplify the management of SAHS. The aim of this study is two-fold: 1) to prospectively assess the reliability and effectiveness of at-home NPO in the context of adult SAHS; 2) to design, optimize and extensively assess the diagnostic performance of automated NPO-based screening tools for SAHS. In order to achieve these goals, both PSG and NPO recordings are carried out ambulatory and simultaneously at patient's home. A portable polysomnograph (Embletta MPR, Natus) is used for standard PSG at home, whereas a portable wrist-worn pulse oximeter (WristOX2 3150, Nonin) is used for ambulatory NPO. In addition, conventional in-lab PSG and attended pulse oximetry are also performed simultaneously in the hospital facilities.
| Condition or Disease | Intervention/Treatment | Phase |
|---|---|---|
|
|
Detailed Description
Participants are recruited from the specialized sleep outpatient facilities of the Río Hortega University Hospital from Valladolid (Spain). All patients are referred from primary care due to moderate-to-high clinical suspicion of suffering from sleep apnea-hypopnea syndrome (SAHS). The final population is randomly split into two independent datasets: 1) training set (50%), which is used to design and build/train the screening algorithms; and 2) the test set (remaining 50%), which is used to further assess performance using unseen data.
The American Academy of Sleep Medicine rules are used to score respiratory events and to obtain the apnea-hypopnea index (AHI) from ambulatory PSG at home, which is used to definitively diagnose SAHS.
A portable wrist-worn pulse oximeter (WristOX2 3150, Nonin) is used for at-home NPO. Portable NPO is carried out simultaneously to ambulatory PSG (Embletta MPR, Natus) at patient's home. In addition, attended portable in-lab NPO (WristOX2 3150, Nonin) and in-lab PSG (E-Series, Compumedics) are performed simultaneously in the hospital in a different consecutive/previous night for comparison purposes. Participants are randomly assigned to carry out unattended sleep studies at home before or after in-hospital recordings.
SpO2 and PR from NPO are recorded simultaneously at a sampling rate of 1 Hz (1 sample every second). All recordings are saved to separate files and processed offline. An automatic signal pre-processing stage is carried out to remove artifacts due to patient movements (signal loss).
The signal processing methodology is divided into three automated stages: (i) feature extraction, (ii) feature selection, and (iii) pattern recognition.
Firstly, NPO recordings are parameterized by means of a wide set of variables, which previously demonstrated a high discriminative power in the context of SAHS detection. All features are computed for each whole portable overnight recording. The following feature subsets are composed:
-
Time domain statistics from SpO2 recordings. First to fourth-order statistical moments in the time domain, i.e., arithmetic mean, variance, skewness and kurtosis, which quantify central tendency, amount of dispersion, asymmetry, and peakedness, respectively.
-
Time domain features from PR recordings: average, standard deviation, and root mean square of standard deviation of the pulse-to-pulse interval time series.
-
Frequency domain statistics from SpO2 recordings. First to fourth-order statistical moments, median frequency, and Shannon spectral entropy from the power spectral density function.
-
Frequency domain statistics from PR recordings. First to fourth-order statistical moments, median frequency, and Shannon spectral entropy from the power spectral density function.
-
Conventional spectral measures from SpO2 recordings. Total signal power as well as peak amplitude and relative power in the frequency range 0.014 - 0.033 Hz.
-
Conventional spectral measures from PR recordings. Normalized power in the low (0.04 - 0.15 Hz) and in the high (0.15 - 0.40 Hz) frequency bands, as well as the low frequency to high frequency ratio (sympathovagal balance).
-
Nonlinear features from SpO2 recordings. Sample entropy, central tendency measure, and Lempel - Ziv complexity, which measure irregularity, variability, and complexity of SpO2 recordings.
-
Nonlinear features from PR recordings. Sample entropy, central tendency measure, and Lempel - Ziv complexity, which measure irregularity, variability, and complexity of PR recordings.
Then, the optimum feature subset composed of the most relevant as well as complementary variables are composed. In order to achieve this goal, the following feature selection methods are applied:
-
Forward stepwise feature selection
-
Genetic algorithms
-
Fast correlation-based filter
-
Minimal-redundancy maximal-relevance criterion
Finally, the third stage corresponds to patter recognition. The aim of this stage is two-fold: (i) to design and optimize binary classification-oriented models trained to discern between SAHS negative and SAHS positive subjects using optimum features from NPO; (ii) to design and optimize regression-oriented models trained to estimate the AHI using optimum features from NPO. In order to achieve this goal, the following pattern recognition algorithms are assessed:
-
Binary classification: logistic regression, artificial neural networks, Bayesian networks, decision trees, ensemble learning (AdaBoost).
-
Regression models: multiple linear regression, artificial neural networks, Bayesian networks, ensemble learning (least squares boosting).
These models are subsequently combined to optimize the following 2-stage screening protocol:
stage-1) true negative screening stage, which is aimed at detecting the maximum number of non-SAHS subjects while minimizing the number of false negative patients (ideally 0% false positive rate); stage-2) true positive screening stage, which is aimed at detecting (among patients not identified as true negative in the first stage) the maximum number of true positive patients while minimizing the number of false positive cases (ideally 0% false positive rate). Both stages are complementary and they are implemented consecutively, such that:
-
Patients identified as true negative in the first stage are referred to the sleep specialist to finally discard SAHS taken into account symptoms, comorbidities and past clinical history. These patients are no longer derived to the sleep unit unless requested by the sleep specialist due to persistent and/or additional symptoms.
-
Patients identified as true positive in the second stage are referred to the sleep specialist to finally confirm SAHS and decide the most suitable treatment option. These patients are no longer derived to the sleep unit unless requested by the sleep specialist.
-
Non-conclusive cases are finally derived to the sleep unit for a standard PSG in order to confirm/discard SAHS.
Study Design
Outcome Measures
Primary Outcome Measures
- Percentage of patients correctly classified [6 months after the inclusion of the last patient]
Percentage of patients (%) correctly classified/screened by the automated NPO-based screening test. At-home ambulatory PSG is used as the gold standard method for positive SAHS. Subjects with apnea-hypopnea index (AHI) <5 are considered no-SAHS subjects, with 5<=AHI<15 as mild SAHS patients, with 15<=AHI<30 moderate SAHS patients, and AHI>=30 as severe SAHS patients.
Secondary Outcome Measures
- Body mass index [6 months after the inclusion of the last patient]
Average (median and interquartile range) body mass index (kg/m2) of the cohort.
- Patients with chronic obstructive pulmonary disease [6 months after the inclusion of the last patient]
Number of patients (n) with comorbid chronic obstructive pulmonary disease (COPD), according to standard definitions.
- Patients with hypertension [6 months after the inclusion of the last patient]
Number of patients (n) with comorbid arterial hypertension (HT), according to standard definitions.
- At-home PSG-derived AHI [6 months after the inclusion of the last patient]
Apnea-hypopnea index (events per hour) derived from unattended PSG at patients' home.
- At-home PSG-derived time in REM sleep [6 months after the inclusion of the last patient]
Percentage of time (%) in rapid eye movement (REM) sleep to the total sleep time derived from unattended PSG at patients' home.
- At-home PSG-derived sleep efficiency [6 months after the inclusion of the last patient]
Sleep efficiency (%) measured as the percentage of total sleep time to the total recording time derived from unattended PSG at patients' home.
- At-home PSG-derived arousal index [6 months after the inclusion of the last patient]
Number of arousals per hour of sleep (events per hour) derived from unattended PSG at patients' home.
- At-home PSG-derived time in supine position [6 months after the inclusion of the last patient]
Percentage of time (%) in supine position to the total sleep time derived from unattended PSG at patients' home.
- At-home PSG-derived average SpO2 [6 months after the inclusion of the last patient]
Average overnight SpO2 (%) from unattended PSG at patients' home.
- At-home PSG-derived minimum SpO2 [6 months after the inclusion of the last patient]
Minimum overnight SpO2 (%) from unattended PSG at patients' home.
- At-home PSG-derived oxygen desaturation index of 3% (ODI3) [6 months after the inclusion of the last patient]
Number of desaturations greater than or equal to 3% from baseline per hour of sleep (events per hor) from unattended PSG at patients' home.
- At-home NPO-derived ODI3 [6 months after the inclusion of the last patient]
Number of desaturations greater than or equal to 3% from baseline per hour of recording (events per hor) from unattended pulse oximetry at patients' home.
- At-home NPO-derived cumulative time below 90% (CT90) [6 months after the inclusion of the last patient]
Percentage (%) of cumulative time with a saturation below 90% from unattended pulse oximetry at patients' home.
- At-home NPO-derived average SpO2 [6 months after the inclusion of the last patient]
Average saturation (%) from unattended pulse oximetry at patients' home.
- At-home NPO-derived minimum SpO2 [6 months after the inclusion of the last patient]
Minimum saturation (%) from unattended pulse oximetry at patients' home.
- At-home NPO-derived average pulse rate [6 months after the inclusion of the last patient]
Average pulse rate (beats per minute) from unattended pulse oximetry at patients' home.
- At-home NPO-derived minimum pulse rate [6 months after the inclusion of the last patient]
Minimum pulse rate (beats per minute) from unattended pulse oximetry at patients' home.
- Prevalence of SAHS [6 months after the inclusion of the last patient]
Prevalence of SAHS (%) in the population under study according to at-home PSG.
- Severity of SAHS [6 months after the inclusion of the last patient]
Number of patients (n) with moderate-to-severe SAHS according to the at-home PSG-derived patient's AHI.
- NPO-derived ODI3 agreement [6 months after the inclusion of the last patient]
Mean difference (mean +/- 1.96 standard deviation interval) from the Bland and Altman agreement plot between unattended ODI3 from at-home NPO and supervised ODI3 from in-hospital NPO.
- PSG-derived AHI agreement [6 months after the inclusion of the last patient]
Mean difference (mean +/- 1.96 standard deviation interval) from the Bland and Altman agreement plot between unattended AHI from at-home PSG and supervised AHI from in-hospital PSG.
- Optimum diagnostic performance - Area under the ROC curve [6 months after the inclusion of the last patient]
Area under the receiver operating characteristics (ROC) curve of the optimum NPO-based binary classifier compared to standard at-home PSG.
- Optimum diagnostic performance - Accuracy [6 months after the inclusion of the last patient]
Accuracy (percentage, %) of the optimum NPO-based binary classifier compared to standard at-home PSG.
- Optimum agreement - Intra-class correlation coefficient [6 months after the inclusion of the last patient]
Intra-class correlation coefficient (ICC) between the optimum NPO-based estimated AHI and the actual AHI derived from at-home PSG.
- Patient's Sleep quality [6 months after the inclusion of the last patient]
Patients' sleep quality assessment using the Pittsburg questionnaire.
- Patient's somnolence [6 months after the inclusion of the last patient]
Patients' somnolence assessment using the Epworth questionnaire.
- Patients' quality of life [6 months after the inclusion of the last patient]
Patients' quality of life assessment using the Quebec sleep questionnaire (QSQ).
- Percentage of unsatisfactory recordings [6 months after the inclusion of the last patient]
Number of recordings (n) removed from the study due to reasons (either technical or human) related to unattended portable oximetry.
Eligibility Criteria
Criteria
Inclusion Criteria:
-
Men and women over 18 years old
-
Subjects derived from primary care to the sleep specialized outpatient facilities showing moderate-to-high clinical suspicion of suffering from sleep apnea (daytime hypersomnolence, loud snoring, nocturnal choking and awakenings, and/or apneic events)
-
Written informed consent signed
Exclusion Criteria:
-
Subjects under 18 years old
-
Subjects not signing the informed consent
-
Presence of any previously diagnosed sleep disorder: narcolepsy, insomnia, chronic sleep deprivation, regular use of hypnotic or sedative medications and/or restless leg syndrome.
-
Patients with the following chronic diseases: congestive heart failure, renal failure, neuromuscular diseases, chronic respiratory failure.
-
Patients with >50% of central apneas or the presence of Cheyne-Stokes respiration.
-
Previous continuous positive airway pressure (CPAP) treatment for SAHS diagnosis
-
A medical history that may interfere with the study objectives or, in the opinion of the investigator, compromise the conclusions
Contacts and Locations
Locations
| Site | City | State | Country | Postal Code | |
|---|---|---|---|---|---|
| 1 | Río Hortega University Hospital | Valladolid | Spain | 47012 |
Sponsors and Collaborators
- Hospital del Río Hortega
- OXIGEN salud
- University of Valladolid
- Five Flames Mobile
Investigators
- Principal Investigator: Félix Del Campo, PhD,MD, Río Hortega University Hospital
Study Documents (Full-Text)
None provided.More Information
Additional Information:
- Web of the ScreenOX clinical trial
- Web of the Rio Hortega University Hospital, Spain
- Web of OXIGEN salud. Respiratory therapy provider, Spain
- Web of the Biomedical Engineering Group (GIB), University of Valladolid, Spain
Publications
None provided.- RTC-2015-3446-1