TABLE VI.F. 4.

Evidence for oximetry for diagnosis of OSA

Study	Year	LOE	Study design	Study groups	Clinical end point	Conclusion
CHF
Sharma et al.619	2017	1b	Prospective single center, controlled trial of patients admitted with CHF.	Simultaneous measurement of apnea link and high-resolution pulse oximetry (HRPO) for a single night. N = 105 61 (58%) M	HRPO-derived ODI (oxygen desaturation index) was compared with PM-derived respiratory event index (REI) using receiver operator characteristic (ROC) curve analysis and a Bland–Altman plot.	1) ROC area under curve (AUC) was 0.89 for REI > 5 events/h. AUC ranged from 0.84 (REI ≤ 10 events/h) to 0.89 (REI ≤ 5 events/h and REI ≤ 20 events/h). 2) The Bland–Altman plot had good agreement. 3) 88% of the REI in moderate–severe category were correctly classified. 4) Cannot differentiate CSA vs. OSA.
Sharma et al.624	2015	3b	Prospective, single center cohort of consecutively admitted acute decompensated heart failure patients with high clinical suspicion of SDB.	Overnight (ON) inpatient oximetry (photoplethysmography) compared with outpatient PSG apnea hypopnea index (AHI). N = 105 subjects had ON oximetry and 68 underwent outpatient PSG within 4 weeks of discharge.	PSG defined hypopneas as some drop in flow with 4% oxygen desaturation compared with ON oximetryODI of 4% desaturation using ROC analysis and Bland–Altman plot.	1) ODI correlated with AHI with AUC of 0.82 on ROC for AHI ≥5. 2) The Bland–Altman plot had no major bias.
Ward et al.623	2012	1b	Prospective, single center cohort of CHF patients from cardiology clinics.	Simultaneous unattended PSG, ambulatory electrocardiography, and ON pulse oximetry at home or hospital N = 173 86% M.	Compared oximetry % ODI cutoff of >7.5 desaturations/h to PSG AHI >15/h. ODI used 3% desaturation. PSG defined hypopneas as 50% drop in flow with 3% desaturation or arousal.	1) At a cutoff of >7.5 desaturations/h, the ODI3% had sensitivity 97%, specificity 32%, negative likelihood ratio (LR) 0.08, and positive LR 1.42. 2) At a cut-off of 12.5 desaturations/h, ODI3% sensitivity was 93% and specificity was 73%. 3) The 3% ODI had an AUC under ROC curve of 0.92 for detection of SDB in CHF, at the cutoff of >7.5 desaturations/h.
COPD
Andrés-Blanco et al.625	2017	1b	Prospective single center cohorts.	Simultaneous portable ON oximetry at home and in-hospital PSG; and unsupervised portable ON oximetry at home. Two independent validation datasets were analyzed: COPD versusnon-COPD. N = 110 non-COPD test set (69% M) and 68 COPD test group (88% M).	A regression-based multilayer perceptron (MLP) artificial neural network (ANN) was trained to estimate AHI from portable oximetry recordings. Two independent validation datasets were analyzed: COPD vs. non-COPD.	1. Portable ON oximetry-based ANN reached similar ICC values between the estimated and actual AHI for the non-COPD and the COPD groups either in the hospital (non-COPD: 0.937, COPD: 0.936) and at home (non-COPD: 0.731, COPD: 0.788) setting. 2. No significant differences in ROC between COPD and non-COPD groups in both settings.
Lajoie et al.626	2020	3b	Prospective cohort recruited from an ongoing multicenter trial.	Compared home ON oximetry and laboratory-based PSG in patients with moderate-to-severe COPD. N = 90 45 had OSA, 71% M 45 did not have OSA, 87% M	ODI3% used for oximetry. AHI hypopnea definition not stated.	1. Oxygen desaturation indices obtained with nocturnal oximetry and during PSG were not correlated (r = –0.27; p = 0.1). 2. Diagnosis of OSA in COPD should not be based solely on oximetry. OSA was confirmed in only 50% of subjects with oximetry tracings suggestive of OSA.
Scott et al.627	2014	2b	Consecutive chart review of the inpatient pulmonary rehabilitation service.	Subjects with moderate–severe COPD who were clinically prescribed oximetry and PSG. N = 59 46% M	Criteria consisted of visually identified desaturation “events” (sustained desaturation ≥4%, 1 h time scale), “patterns” (≥3 similar desaturation/saturation cycles, 15 min time scale) and the automated oxygen desaturation index. Compared using AUC.	1) Thirty-five were correctly identified as having OSA/no OSA with accuracy of 59%, a sensitivity and specificity of 59% and 60%, respectively; AUC 0.57. 2) Using software-computed desaturation events (hypoxemia ≥4% for ≥10 s) indexed at ≥15 events/h of sleep as diagnostic criteria, sensitivity was 60%, specificity was 63%, and the AUC 0.64.
Atrial Fibrillation
Linz et al.25	2018	2b	Prospectively single center cohort in patients with atrial fibrillation (AF) who underwent PSG.	Subjects with documented AF. N = 439 69% M	ON oximetry from the PSG was used to determine the ODI. ODI was validated against PSG AHI. ODI4% used for oximetry, which came off PSG. PSG hypopnea definition was a 30% drop in flow for 10 s with either a 3% desaturation or an arousal.	1) ODI was able to detect moderate-to-severe SDB (AHI ≥ 15/h) AUC: 0.951; severe SDB (AHI ≥ 30/h) AUC 0.932. 2) An ODI cut-off of 4.1/h had 91% sensitivity and 83% specificity in patients with and without AHI ≥ 15/h. 3) An ODI of 7.6/h yielded a sensitivity and specificity for AHI ≥ 30/h of 89% and 83%, respectively. 4) Cannot differentiate CSA vs. OSA.
CVA
Lin et al.629	2018	2b	Retrospective chart analysis.	Subjects with acute stroke or TIA underwent ON oximetry and HSAT. N = 254 50.7% M 232 (91.3%) were ischemic or TIA.	ODI from pulse oximetry channel was compared to respiratory event index (REI) obtained from HSAT devices. ODI3% used for oximetry. REI3% used for HSAT.	1) ODI3% had correlation (r = 0.902) and agreement with REI3%. 2) ODI3% was accurate in predicting SDB at different REI thresholds (REI ≥ 5, REI ≥ 15, and REI ≥ 30 events/h) with AUC of 0.965, 0.974, and 0.951, respectively. 3) An ODI3% ≥ 5 events/h rules in the presence of SDB (specificity 91.7%, PPV 96.3%). 4) An ODI3% ≥ 15 events/h rules in moderate to severe SDB (specificity 96.4%, PPV 95%) and an ODI3% < 5 events/h rules out moderate to severe SDB (sensitivity 100%, NPV 100%).
Aaronson et al.628	2012	2b	Retrospective study of stroke patients.	Compared polygraphy and oximetry from HSAT in stroke subjects. N = 56 62% male, 46% of the stroke patients had OSA. 69% with OSA were ischemic strokes.	Compared REI to ODI. REI hypopneas defined as 50% drop in flow with a 4% desaturation. ODI used 4% desaturation.	1) Sensitivity, specificity, and PPV and NPV for the ODI4% ≥15 were, respectively, 77%, 100%, 100%, and 83%. 2) ODI4% predicted 87% of the variance in the REI. 3) Given a 46% prevalence of OSA in stroke, the PPV of oximetry was 100% with an NPV of 83%
Ryan et al.630	2017	3b	Prospective cohort of patients with acute stroke in a stroke rehabilitation unit (SRU).	Compared testing with BresoDx – a portable single-channel acoustic device – both simultaneously during attended PSG in lab and unattended on the SRU. N = 23 48% M 78% had OSA (defined by AHI ≥15) on PSG. 74% of subjects were ischemic strokes.	Compared PSG AHI to BresoDx AHI. PSG hypopneas defined by a 30% drop in flow with 3% desaturation or arousal. Determined AUC and Bland–Altman plot.	1) Using cutoff AHI of ≥15 by PSG to diagnose OSA in-lab BresoDx had sensitivity of 90.0%, specificity of 84.6%, and accuracy of 87.0%. 2) Bland–Altman plot: good agreement, but BresoDx overestimated AHI by 4.4. 3) The AUCs for AHI in lab Breos vs. in-lab PSG at thresholds of ≥5, ≥10, and ≥15 were 0.90, 0.91, and 1.00, respectively. 4) For home BresoDx vs. in-lab PSG, at an AHI threshold of ≥15 had a sensitivity of 100%, specificity of 85.7%, and accuracy of 91.3%.
General
Pataka et al.616	2019	1b	Prospective study in a sleep clinic.	Compared sleep questionnaires STOP-BANG (SB), Berlin (BQ), Epworth Sleepiness Scale (ESS) completed by subjects with home oximetry and in laboratory PSG, to determine predictive value of test for CPAP initiation. N = 204 77.5% M	Determine correlations and accuracy. Compared PSG and oximetry values as well. PSG hypopneas defined by a 30% drop in flow with 4% desaturation or arousal. ODI used 4% desaturations.	1. Good correlation between oximetry ODI (ODIox) and PSG ODI (r = 0.95, p < 0.0001) and between ODIox and AHI (r = 0.811, p < 0.0001). ODIox ≥ 15 had sensitivity 89.3%, specificity 83.5%, PPV 87%, and NPV 86.4% for CPAP initiation. 2) Among questionnaires, ESS had highest specificity (68.6%) and PPV (68.6%) and SB had the highest sensitivity (98%) and NPV (80%) but the lowest specificity (11%) for CPAP initiation. Oximetry was superior to questionnaires for predicting CPAP treatment initiation.
Christensson et al.617	2018	2b	Prospective, observational multicenter trial of sleep clinic patients.	Subjects underwent HSAT (Nox-T3), ON oximetry, and STOP-BANG (SB) questionnaires. N = 449 subjects with suspected OSA. 61.5% M	Compared REI to ODI. Compared REI to SB questionnaire scores. HSAT hypopneas defined by 30% drop in flow and 30% desaturation. ODI used 3% for oximetry.	1) Strong correlation between REI and ODI3%, Spearman 0.96. 2) Positive correlation between SB score and ODI3%, Spearman ρ 0.50; An SB score of <2 almost excludes moderate to severe OSA, whereas nearly all OSA patients with an SB score ≥6 had OSA.
Sharma et al.619	2017	3b	Retrospective review of a large database of hospitalized inpatients. Only those high ODI on ON oximetry were offered PSG.	Compared in-hospital ON HRPO to PSG post-discharge. N = 1410 underwent in-hospital HRPO with 1092 having and ODI4% ≥ 5. Of these, 680 underwent PSG post-discharge. 54% M (of HRPO group).	Determined accuracy, AUC, and Bland–Altman plot of HRPO-determined ODI vs. AHI. ODI used 4% for oximetry. PSG hypopneas defined by a 30% drop in flow with 4% desaturation.	1) ODI4% ≥5 had sensitivity 0.89 and pecificity 0.48. 2) ODI4% ≥ 15 had a sensitivity 0.65 and specificity 0.90. 3) ODI4% ≥5 had an AUC of 0.83 for an AHI ≥5 and 0.76 for an AHI ≥15. 4) Bland–Altman plot showed no significant bias when using ODI vs. AHI to define SDB.
Hang et al.618	2015	2b	Prospective study of sleep clinic patients undergoing PSG for suspected OSA.	Oximeter from PSG was used for ODI calculation without considering other PSG information. N = 699 (though only analyzed 544 with adequate TST and acceptable PSG signals) 77.1% M. PSG results: 20.6% had an AHI 5–15, 21.4% had an AHI >15–30, 46.3% had an AHI >30.	Compared accuracy and AUC of ODI from PSG oximetry to AHI from same PSG. ODI used 3% and 4% desaturations. PSG defined hypopneas as 30% drop in flow with a 4% desaturation.	1) For AHI ≥ 15, ODI3% had sensitivity, specificity, and accuracy of 86.1%, 92.4%, and 89.5%. 2) For AHI ≥15, ODI4% had sensitivity, specificity, and accuracy of 85.7%, 89.7%, and 87.8%. 3) AUC for severe OSA: 0.953–0.957; AUC of 0.921–0.924 for moderate to severe OSA patients. 4) Limitation due to removal of those with low TST on PSG.
Chung et al.499	2012	1b	Prospective study of patients presenting to presurgical clinic for elective surgery.	Subjects underwent unattended PSG and ON oximetry on the same night. N = 475 45.7% M	Compared PSG AHI and ON oximetry ODI. Hypopnea definition was 30% drop in flow and 4% desaturation. ODI used 4% desaturation.	1) ODI4% > 5 had a sensitivity, specificity, accuracy of 0.96, 0.67, 87% for an AHI >5; and 0.99, 0.39, 61.7% for AHI >15. 2) ODI4%>15 had a sensitivity, specificity, accuracy of 0.45, 0.98, 62.1% for an AHI >5, and 0.70, 0.93, 84% for AHI >15. 3) The AUC for ODI to predict AHI >5, AHI >15, and AHI >30 was 0.908, 0.931, and 0.958, respectively.
del Campo et al.620	2006	2b	Prospective study of cohort of patients undergoing PSG for suspected OSA.	Oximetry and PSG done at the same time. Approximate entropy (ApEn) (a mathematical tool) was calculated off oximetry and compared with PSG data. N = 187 (22.5% had COPD) 79% M	Determined accuracy between PSG and ApEN. PSG hypopneas defined by 30% drop in flow with a 3% desaturation. ODI used 3% and 4% desaturations.	1) AHI correlated with ApEn (r = 0.607; p < 0.001). 2) For AHI > 10, ApEn at 0.679 had sensitivity, specificity, PPV, and NPV of 88.3%, 82.9%, 88.3%, and 82.9%, respectively.
Erdenebayar et al.621	2017	2b	Prospective cross-sectional study of patients referred to a sleep clinic.	Subjects underwent an in-lab PSG and piezo-electric sensor at the same time. The piezo-electric sensor detected snoring and heartbeat information, and snoring index (SI) and features based on pulse rate variability (PRV) analysis. A support vector machine (SVM) was used as a classifier to detect OSA events. N = 45 70% M	Compared accuracy of piezo-electric sensor with PSG. PSG scored per “AASM standards” but not defined further.	1) Mild OSA detection: sensitivity, specificity, and accuracy of 72.5%, 74.2%, and 71.5%; moderate OSA detection: 85.8%, 80.5%, and 80.0%; and severe OSA: 70.3%, 77.1%, and 71.9%. 2) Automatic snoring detection had sensitivity, specificity, and accuracy of 88.5%, 96.1%, and 95.6%. 3) Heartbeat detection had sensitivity and PPV of 94.3% and 87.1%, all respectively.
Alakuijala et al.622	2016	4	Prospective cross-sectional study of patients referred to a sleep clinic.	Subjects underwent a HSAT (Nox T3) at home. Periodic snoring data was collected from the same HSAT. N = 211 61% M There was no separate validation group.	Analyzed the percentage of periodic snoring during HSAT and compared to the AHI from the HSAT. Correlations and Bland–Altman plot were analyzed. The HSAT defined hypopneas by 3% desaturations.	1) AHI ranged from 0.1 to 116 events/h, and % of periodic snoring from 1% to 97%. 2) Positive correlation (r = 0.727, p < 0.001) between periodic snoring and AHI. 3) Sensitivity was 93.3%, specificity 35.1%, and NPV 75.0%. 4) Bland–Altman plot showed that periodic snoring percentage, and AHI agreed within range of various grades of OSA.
Neuromuscular disease
	2018	2b	Prospective cross-sectional study of patients followed for chronic respiratory failure due to neuromuscular disease, treated with chronic noninvasive ventilation (NIV).	All patients underwent the screening test panel (clinical evaluation, daytime arterial blood gas [ABG], nocturnal pulse oximetry [SpO2], and data from ventilator software), HSAT (Embletta Gold) and nocturnal transcutaneous CO2 (while on their NIV). N = 67	Compared accuracy among the tests. HSAT used 4% desaturation criteria. ODI3% used for oximetry.	1. Nocturnal SpO2 and daytime ABG all failed to accurately detect nocturnal hypoventilation (NH). 2. ODI3% had a high sensitivity but low specificity for identifying obstructive events on NIV.