Browse the corpus

Key messages What is the key question? Can electrical current be delivered non-invasively via transcutaneous patches with ongoing stimulation for the entire night? What is the bottom line? The current trial provides evidence that transcutaneous electrical stimulation of the upper airway dilator muscles, a non-invasive approach, delivered throughout the night improves sleep apnoea. Why read on? It is important to further refine transcutaneous electrical stimulation of the upper airway dilator muscles, as this method offers a promising and novel approach in the range of non-CPAP therapies for patients with obstructive sleep apnoea. Introduction Obstructive sleep apnoea (OSA) is the most common form of sleep-disordered breathing1 and is widely recognised as a risk factor for cardiovascular diseases.2 The prevalence of OSA continues to rise, imposing a worldwide burden on public health and currently affecting 10% of middle-aged men and 3% of women aged 30–49 years in the USA.3 The principle evidence-based treatment for OSA, in addition to weight loss, is continuous positive airway pressure (CPAP);4 however, CPAP is frequently not tolerated over longer periods, with a quarter of patients being non-compliant within weeks and half of all patients not using the equipment after 1 year.5 Alternative therapies are needed to reduce symptoms and health risks for patients who fail CPAP treatment.

e airway pressure (CPAP);4 however, CPAP is frequently not tolerated over longer periods, with a quarter of patients being non-compliant within weeks and half of all patients not using the equipment after 1 year.5 Alternative therapies are needed to reduce symptoms and health risks for patients who fail CPAP treatment. In 1978, Remmers et al6 described the pathophysiology of upper airway obstruction in sleep apnoea. Since then several research groups have observed that electrical stimulation of the upper airway could result in an increased tone of the dilator muscles of the upper airway,7 thereby enabling patients to maintain a patent upper airway while asleep.8 9 In 2014, hypoglossal nerve stimulation using an implantable stimulator was approved by the US Food and Drug Administration for the treatment of OSA following earlier publication of the results of the STAR trial.10 Previously, Miki et al11 12 demonstrated that transcutaneous electrical stimulation was effective in reducing the apnoea index and duration and improved oxygen saturation; however, other researchers could not replicate these results.13 14 In 2011, our group showed that continuous transcutaneous electrical stimulation (CTES) was a feasible and effective approach to stimulate the upper airway dilator muscles during short periods while asleep.15 In this trial, the aim was to conduct a randomised, sham-controlled and double-blind clinical trial to test the effectiveness and safety of overnight transcutaneous electrical stimulation of the upper airway muscles in patients with OSA.

Previously, Miki et al11 12 demonstrated that transcutaneous electrical stimulation was effective in reducing the apnoea index and duration and improved oxygen saturation; however, other researchers could not replicate these results.13 14 In 2011, our group showed that continuous transcutaneous electrical stimulation (CTES) was a feasible and effective approach to stimulate the upper airway dilator muscles during short periods while asleep.15 In this trial, the aim was to conduct a randomised, sham-controlled and double-blind clinical trial to test the effectiveness and safety of overnight transcutaneous electrical stimulation of the upper airway muscles in patients with OSA. Patients and methods This sham-controlled and randomised crossover trial was approved by the London ethics committee for clinical trials (London-Dulwich, UK; 12/LO/1428) and was registered in ClinicalTrials.gov (NCT01661712). We enrolled patients referred to Guy’s and St Thomas’ and the Royal Brompton & Harefield NHS Foundation Trusts sleep services (both London, UK) for treatment of OSA between March 2013 and December 2015 when we achieved the full sample size. All patients were provided with a patient information sheet and informed written consent was obtained prior to enrolment (figure 1).16 Figure 1 Consort diagram for the TESLA trial. AHI, apnoea-hypopnoea index; ODI, oxygen desaturation index.

Patients and methods This sham-controlled and randomised crossover trial was approved by the London ethics committee for clinical trials (London-Dulwich, UK; 12/LO/1428) and was registered in ClinicalTrials.gov (NCT01661712). We enrolled patients referred to Guy’s and St Thomas’ and the Royal Brompton & Harefield NHS Foundation Trusts sleep services (both London, UK) for treatment of OSA between March 2013 and December 2015 when we achieved the full sample size. All patients were provided with a patient information sheet and informed written consent was obtained prior to enrolment (figure 1).16 Figure 1 Consort diagram for the TESLA trial. AHI, apnoea-hypopnoea index; ODI, oxygen desaturation index. Inclusion and exclusion criteria The study included patients aged 18–75 years with a body mass index (BMI) between 18 and 40 kg/m2 who had OSA with an oxygen desaturation index (4%ODI) ≥15/hour or with an ODI ≥5/hour plus an Epworth Sleepiness Scale (ESS) >10 points. Exclusion criteria were obesity-hypoventilation syndrome, significant airway obstruction, acute or critical illness. Polysomnography Patients underwent nocturnal full polysomnography (Alice5 equipment, Respironics, Murrysville, Pennsylvania, USA) at baseline and during randomly allocated sham electrical stimulation and active electrical stimulation nights. These three nights were recorded with a gap of at least 3 days to provide ‘wash-out’ periods.

Patients underwent nocturnal full polysomnography (Alice5 equipment, Respironics, Murrysville, Pennsylvania, USA) at baseline and during randomly allocated sham electrical stimulation and active electrical stimulation nights. These three nights were recorded with a gap of at least 3 days to provide ‘wash-out’ periods. Interventional sleep studies Patients were randomised in a crossover design into one of two treatment arms; either they underwent polysomnography plus CTES during the first study night (‘active treatment’) followed by another polysomnography plus sham-CTES during a second night, at least three nights later (‘washout period’) or they underwent the tests in the reverse order (figure 1). Transcutaneous electrical stimulation To deliver CTES, a specifically tailored electrical stimulator was employed (SOSATS device, Morgan Innovation and Technology/MIAT, Petersfield, UK) which was connected to a laptop (Toshiba, Tokyo/Japan) via a standard USB cable. A stimulation lead from the device was attached bilaterally to the patient's neck via two 4×4 cm patches (Verity Medical, Hampshire, UK) placed bilaterally halfway between the chin and the angle of the mandible over the submental area, as previously described (see online supplementary material).15 Sham-CTES included the placement of the stimulation patches and a stimulation marker that was displayed on the computer screen; no electrical current was applied. 10.1136/thoraxjnl-2016-208691.supp1Supplementary material

Transcutaneous electrical stimulation To deliver CTES, a specifically tailored electrical stimulator was employed (SOSATS device, Morgan Innovation and Technology/MIAT, Petersfield, UK) which was connected to a laptop (Toshiba, Tokyo/Japan) via a standard USB cable. A stimulation lead from the device was attached bilaterally to the patient's neck via two 4×4 cm patches (Verity Medical, Hampshire, UK) placed bilaterally halfway between the chin and the angle of the mandible over the submental area, as previously described (see online supplementary material).15 Sham-CTES included the placement of the stimulation patches and a stimulation marker that was displayed on the computer screen; no electrical current was applied. 10.1136/thoraxjnl-2016-208691.supp1Supplementary material Outcome measures The primary outcome measure for this trial was the 4%ODI/hour of sleep (events/hour). The 4%ODI was chosen as the primary outcome parameter in preference to the apnoea-hypopnoea index (AHI). Secondary outcome measures were AHI, nocturnal oxygen saturation levels and nadir oxygenation and sleepiness, as measured by the Stanford Sleepiness Scale. Patient comfort and device acceptance were measured by administering ad-hoc visual analogue scales on waking after sham and stimulation sleep studies (see online supplementary material). Last, a post-hoc analysis was performed to determine responders whose ODI had improved by >25% compared with sham night and/or the total ODI was <5/hour.

device acceptance were measured by administering ad-hoc visual analogue scales on waking after sham and stimulation sleep studies (see online supplementary material). Last, a post-hoc analysis was performed to determine responders whose ODI had improved by >25% compared with sham night and/or the total ODI was <5/hour. Sample size calculation A sample size calculation for the primary outcome, 4%ODI, was performed based on the results of a feasibility study.15 A sample size of 30 patients would achieve 90% power to detect a mean of paired differences between the treatments of 17.9 events/hour with an estimated SD of differences of 18.9 events/hour and with a significance level of 5% using a two-sided Wilcoxon test assuming that the actual distribution was normal. To adjust for the unknown distribution of the primary outcome and based on the lower bound for the asymptotic relative efficiency of the Wilcoxon test, the required sample size was increased by 20% to 36 patients. Further accounting for a dropout and loss-to-follow-up rate of up to 20%, consistent with the experience from previous studies of this type, a total sample size of 44 patients was required for inclusion in the trial.

relative efficiency of the Wilcoxon test, the required sample size was increased by 20% to 36 patients. Further accounting for a dropout and loss-to-follow-up rate of up to 20%, consistent with the experience from previous studies of this type, a total sample size of 44 patients was required for inclusion in the trial. The effect size of the expected difference which the study was powered for was aligned with the range of severity of sleep apnoea: the threshold between mild and moderate OSA is an AHI of 15 events/hour and the threshold between moderate and severe sleep apnoea is an AHI of 30 events/hour. We wanted to demonstrate that electrical stimulation was able to reduce OSA severity by at least 15 events/hour allowing for a drop in severity by one class (eg, the severe to the moderate range or from the moderate to the mild range). Statistical analysis The differences in primary and secondary outcome measures between the ‘active treatment’ and ‘sham intervention’ were compared in a paired design (crossover trial) using SPSS Statistics (V.23; IBM, New York, USA). To compare study groups, the Wilcoxon (4%ODI) and paired t test for continuous paired variables were used. To identify predictors of response a stepwise multiple linear regression including the variables ‘age’, ‘gender’ and OSA severity (‘ODI’) was performed. The McNemar's χ2 test for nominal variables was used for the gender comparison in the responder group. The 95% CI was used to describe the treatment response. The level of significance was selected at p<0.05.

wise multiple linear regression including the variables ‘age’, ‘gender’ and OSA severity (‘ODI’) was performed. The McNemar's χ2 test for nominal variables was used for the gender comparison in the responder group. The 95% CI was used to describe the treatment response. The level of significance was selected at p<0.05. (For additional information on the methods and measurements, please refer to the online supplementary material). Results Patient characteristics A total of 390 patients were assessed in order to determine the eligibility for this trial. Following initial assessment, 44 patients were further screened and underwent baseline polysomnography (first sleep study). Eight patients were excluded as screening failures and 36 patients were randomised in a crossover design and allocated to receive either active or sham treatment first. The participants returned for the first treatment night (second sleep study) after at least 3 days. After a ‘washout period’ of at least another 3 days, participants returned for the opposite treatment during the second treatment night (third sleep study). All 36 patients completed the trial after the third polysomnography and were included in the analysis (figure 1).

night (second sleep study) after at least 3 days. After a ‘washout period’ of at least another 3 days, participants returned for the opposite treatment during the second treatment night (third sleep study). All 36 patients completed the trial after the third polysomnography and were included in the analysis (figure 1). The studied patients were middle-aged, predominantly Caucasian male subjects and overweight-to-obese. The neck circumference was increased, the pharyngeal lumen was narrowed—only eight patients had a Mallampati score of I—and across the cohort there was a neutral waist-to-hip mass distribution. Participants were sleepy, as assessed by the ESS and had well-preserved lung function and normal daytime oxygenation. The past medical history indicated limited use of alcohol and nicotine (table 1). About 27.8% of patients had hypertension, 19.4% dyslipidaemia and 8.3% had type II diabetes. Table 1 Demographic details of the patients included in the study (n=36)

The studied patients were middle-aged, predominantly Caucasian male subjects and overweight-to-obese. The neck circumference was increased, the pharyngeal lumen was narrowed—only eight patients had a Mallampati score of I—and across the cohort there was a neutral waist-to-hip mass distribution. Participants were sleepy, as assessed by the ESS and had well-preserved lung function and normal daytime oxygenation. The past medical history indicated limited use of alcohol and nicotine (table 1). About 27.8% of patients had hypertension, 19.4% dyslipidaemia and 8.3% had type II diabetes. Table 1 Demographic details of the patients included in the study (n=36) Parameters Data Treatment first Sham first p Value Age (years) 50.8 (11.2) 51.2 (11.9) 50.6 (11) 0.89 Sex (males/females) 30/6 13/1 17/5 0.22 White British, n (%) 29 (80.5) 12 (85.7) 17 (77.2) <0.05 Caribbean, n (%) 3 (8.3) 1 (7.1) 2 (9.1) 0.83 African, n (%) 2 (5.5) 1 (7.1) 1 (4.5) 0.74 Indian, n (%) 1 (2.7) 0 1 (4.5) 0.41 White other, n (%) 1 (2.7) 0 1 (4.5) 0.41 Height (cm) 175.3 (8.4) 175.6 (6.4) 175.1 (9.6) 0.83 Weight (kg) 95.8 (17.7) 93.8 (17.7) 97.1 (18.1) 0.59 BMI (kg/m2)* 29.7 (26.9–34.9) 28.4 (26.6–35) 30.6 (27.4–34.9) 0.44 Neck circumference (cm) 42.6 (3.8) 42.0 (2.9) 42.9 (4.2) 0.42 Waist circumference (cm) 103.8 (16.5) 104.4 (14.1) 103.4 (18.2) 0.85 Hip circumference (cm)* 105.0 (99.0–111.0) 107.2 (99.8–111.0) 104.5 (98.3–108.0) 0.26 Waist:hip ratio 0.99 (0.08) 0.98 (0.07) 1.00 (0.08) 0.29 Mallampati score I, n (%) 8 (22.2) 4 (28.5) 4 (18.1) 0.56 Mallampati score II, n (%) 15 (41.6) 4 (28.5) 11 (50) 0.2 Mallampati score III, n (%) 8 (22.2) 5 (35.7) 3 (13.6) 0.12 Mallampati score IV, n (%) 5 (13.8) 1 (7.1) 4 (18.1) 0.35 ESS (points) 10.5 (4.6) 11.4 (4.6) 10 (4.5) 0.39 FEV1 (L) 3.13 (0.82) 2.94 (0.73) 3.25 (0.86) 0.26 FVC (L) 4.08 (0.89) 3.89 (0.86) 4.20 (1.07) 0.34 FEV1/FVC (%) 76.8 (8.2) 75.7 (9.9) 77.4 (7.0) 0.58 SpO2 awake (%) 94.6 (1.0) 94.7 (0.6) 94.5 (1.2) 0.43 Previous treatment: none/CPAP, n (%) 18 (50.0)/18 (50.0) 7 (50.0)/7 (50.0) 11 (50.0)/11 (50.0) 1.0 Alcohol consumption (units/week)* 2.0 (0.0–10.5) 1 (0.0–5.9) 4 (0–11.5) 0.49 Pack-years (years)* (12 lifelong non-smokers, 24 current or ex-smokers) 10.0 (5.0–14.3) 10.0 (6.5–12.1) 10.0 (5–15) 0.55 *Data were non-normally distributed and are expressed as median and IQR. Pack-years are calculated for current and ex-smokers only (n=24).

ion (units/week)* 2.0 (0.0–10.5) 1 (0.0–5.9) 4 (0–11.5) 0.49 Pack-years (years)* (12 lifelong non-smokers, 24 current or ex-smokers) 10.0 (5.0–14.3) 10.0 (6.5–12.1) 10.0 (5–15) 0.55 *Data were non-normally distributed and are expressed as median and IQR. Pack-years are calculated for current and ex-smokers only (n=24). BMI, body mass index; ESS, Epworth Sleepiness Scale; SpO2: oxygen saturation. Baseline polysomnography Participants had predominantly moderate-to-severe OSA with more severe upper airway obstruction in the supine posture, the majority of respiratory events being obstructive apnoea. Sleep was fragmented with an increased time of ‘wakefulness after sleep onset’ and reduced sleep efficiency. Normal sleep architecture was preserved, although rapid eye movement (REM) sleep latency was delayed. Snoring was observed during more than 10% of the night (table 2). Table 2 Baseline polysomnography data of the studied patients (n=36)

Baseline polysomnography Participants had predominantly moderate-to-severe OSA with more severe upper airway obstruction in the supine posture, the majority of respiratory events being obstructive apnoea. Sleep was fragmented with an increased time of ‘wakefulness after sleep onset’ and reduced sleep efficiency. Normal sleep architecture was preserved, although rapid eye movement (REM) sleep latency was delayed. Snoring was observed during more than 10% of the night (table 2). Table 2 Baseline polysomnography data of the studied patients (n=36) Parameters Results ODI (events/hour)* 25.7 (16.0–49.1) AHI (events/hour)* 28.1 (19.0–57.0) Obstructive apnoea (events/hour)* 15.2 (6.7–31.4) Central apnoea (events/hour)* 0.1 (0.0–0.5) Mixed apnoea (events/hour)* 0.2 (0.0–1.9) Obstructive hypopnoea (events/hour)* 7.8 (1.2–14.3) Supine AHI (events/hour) 43.2 (27.0) REM AHI (events/hour) 36.7 (24.9) Arousal index (events/hour) 28.7 (14.8) SpO2 asleep (%) 93.3 (1.6) Nadir SpO2 asleep (%)* 80.5 (74.0–85.0) Total sleep time (min) 337.5 (75.3) Time in bed (min) 448.4 (51.8) Sleep efficiency (%) 74.9 (15.1) Sleep onset (min)* 18.8 (9.0–39.1) Wake after sleep onset (min)* 79.0 (35.9–117.2) REM latency (min)* 131.0 (69.5–162.5) Sleep stage N1 (%)* 13.0 (8.3–18.7) Sleep stage N2 (%) 50.7 (15.1) Sleep stage N3 (%) 16.3 (12.8) Sleep stage REM (%) 15.7 (9.3) Snoring time (min)* 43.3 (15.2–72.4) Snoring time (%)* 13.2 (5.5–22.3) *Data were non-normally distributed and are expressed as median and IQR. Sleep stages and snoring time are expressed as percentage of time asleep.

Parameters Results ODI (events/hour)* 25.7 (16.0–49.1) AHI (events/hour)* 28.1 (19.0–57.0) Obstructive apnoea (events/hour)* 15.2 (6.7–31.4) Central apnoea (events/hour)* 0.1 (0.0–0.5) Mixed apnoea (events/hour)* 0.2 (0.0–1.9) Obstructive hypopnoea (events/hour)* 7.8 (1.2–14.3) Supine AHI (events/hour) 43.2 (27.0) REM AHI (events/hour) 36.7 (24.9) Arousal index (events/hour) 28.7 (14.8) SpO2 asleep (%) 93.3 (1.6) Nadir SpO2 asleep (%)* 80.5 (74.0–85.0) Total sleep time (min) 337.5 (75.3) Time in bed (min) 448.4 (51.8) Sleep efficiency (%) 74.9 (15.1) Sleep onset (min)* 18.8 (9.0–39.1) Wake after sleep onset (min)* 79.0 (35.9–117.2) REM latency (min)* 131.0 (69.5–162.5) Sleep stage N1 (%)* 13.0 (8.3–18.7) Sleep stage N2 (%) 50.7 (15.1) Sleep stage N3 (%) 16.3 (12.8) Sleep stage REM (%) 15.7 (9.3) Snoring time (min)* 43.3 (15.2–72.4) Snoring time (%)* 13.2 (5.5–22.3) *Data were non-normally distributed and are expressed as median and IQR. Sleep stages and snoring time are expressed as percentage of time asleep. AHI, apnoea-hypopnoea index; N1–N3, non-REM sleep stages 1–3; ODI, 4% oxygen desaturation index; REM, rapid eye movement; SpO2, oxygen saturation.

Parameters Results ODI (events/hour)* 25.7 (16.0–49.1) AHI (events/hour)* 28.1 (19.0–57.0) Obstructive apnoea (events/hour)* 15.2 (6.7–31.4) Central apnoea (events/hour)* 0.1 (0.0–0.5) Mixed apnoea (events/hour)* 0.2 (0.0–1.9) Obstructive hypopnoea (events/hour)* 7.8 (1.2–14.3) Supine AHI (events/hour) 43.2 (27.0) REM AHI (events/hour) 36.7 (24.9) Arousal index (events/hour) 28.7 (14.8) SpO2 asleep (%) 93.3 (1.6) Nadir SpO2 asleep (%)* 80.5 (74.0–85.0) Total sleep time (min) 337.5 (75.3) Time in bed (min) 448.4 (51.8) Sleep efficiency (%) 74.9 (15.1) Sleep onset (min)* 18.8 (9.0–39.1) Wake after sleep onset (min)* 79.0 (35.9–117.2) REM latency (min)* 131.0 (69.5–162.5) Sleep stage N1 (%)* 13.0 (8.3–18.7) Sleep stage N2 (%) 50.7 (15.1) Sleep stage N3 (%) 16.3 (12.8) Sleep stage REM (%) 15.7 (9.3) Snoring time (min)* 43.3 (15.2–72.4) Snoring time (%)* 13.2 (5.5–22.3) *Data were non-normally distributed and are expressed as median and IQR. Sleep stages and snoring time are expressed as percentage of time asleep. AHI, apnoea-hypopnoea index; N1–N3, non-REM sleep stages 1–3; ODI, 4% oxygen desaturation index; REM, rapid eye movement; SpO2, oxygen saturation. Primary and physiological outcome parameters During active treatment (current of 626.1 µA (409.8 µA)), the primary outcome of the trial, the ODI, improved modestly, with a mean of 4.1/hour (95% CI −0.6 to 8.9) for the whole group, when comparing with sham stimulation (figures 2 and 3). No differences were observed in the oxygenation levels and there were no significant improvements in the AHI or the supine AHI. Polysomnographic data revealed a similar sleep architecture and duration as observed at baseline with a reduction in N1 during the active treatment (table 3). The analysis of the apnoea:hypopnoea ratio revealed that the ratio was 1.59 (0.75–17.12) during the baseline sleep study and 0.82 (0.25–2.81) during the stimulation night (p<0.001), indicating a higher contribution of apnoea to the AHI without treatment.

ine with a reduction in N1 during the active treatment (table 3). The analysis of the apnoea:hypopnoea ratio revealed that the ratio was 1.59 (0.75–17.12) during the baseline sleep study and 0.82 (0.25–2.81) during the stimulation night (p<0.001), indicating a higher contribution of apnoea to the AHI without treatment. Table 3 Respiratory and polysomnography data during sham treatment night compared with active treatment

ine with a reduction in N1 during the active treatment (table 3). The analysis of the apnoea:hypopnoea ratio revealed that the ratio was 1.59 (0.75–17.12) during the baseline sleep study and 0.82 (0.25–2.81) during the stimulation night (p<0.001), indicating a higher contribution of apnoea to the AHI without treatment. Table 3 Respiratory and polysomnography data during sham treatment night compared with active treatment Parameters Sham stimulation Active treatment p Value ODI (events/hour)* 26.9 (17.5–39.5) 19.5 (11.6–40.0) 0.026 AHI (events/hour)* 33.8 (16.6–46.1) 23.7 (11.4–47.6) 0.20 Obstructive apnoea (events/hour)* 9.9 (3.8–32.4) 7.6 (3.4–30.0) 0.21 Central apnoea (events/hour)* 0.5 (0.0–1.8) 0.4 (0.0–0.9) 0.53 Mixed apnoea (events/hour)* 0.5 (0.0–2.1) 0.1 (0.0–1.7) 0.68 Obstructive hypopnoea (events/hour)* 12.7 (3.0–23.6) 7.8 (4.5–15.6) 0.42 Supine AHI (events/hour) 44.9 (24.2) 38.6 (25.9) 0.09 REM AHI (events/hour) 35.2 (26.0) 31.3 (23.9) 0.50 Arousal index (events/hour) 28.8 (16.9) 22.6 (16.9) 0.007 SpO2 asleep (%) 93.2 (2.0) 93.2 (2.2) 0.48 Nadir SpO2 asleep (%)* 80.5 (74.5–86.0) 81.0 (74.0–84.0) 0.58 Total sleep time (min)* 366.3 (323.6–409.0) 356.8 (340.8–396.4) 0.41 Time in bed (min)* 452.0 (424.4–475.0) 447.3 (406.6–483.2) 0.77 Sleep efficiency (%)* 83.8 (70.9–89.3) 84.2 (69.1–89.0) 0.60 Sleep onset (min)* 14.5 (5.6–39.8) 17.0 (3.9–47.9) 0.17 Wake after sleep onset (min)* 51.5 (29.4–97.3) 52.3 (29.4–89.8) 0.52 REM latency (min)* 94.0 (64.1–156.6) 89.0 (64.0–132.0) 0.85 Sleep stage N1 (%)* 10.2 (7.0–20.3) 9.8 (6.3–15.0) 0.039 Sleep stage N2 (%)* 49.9 (40.0–58.8) 51.4 (43.1–65.6) 0.32 Sleep stage N3 (%) 15.7 (11.6) 15.4 (10.0) 0.76 Sleep stage REM (%) 16.9 (9.2) 16.0 (8.4) 0.50 Snoring time (min)* 64.6 (23.0–140.8) 46.3 (26.0–125.4) 0.45 Snoring time (%)* 19.9 (6.1–35.9) 17.0 (7.6–28.7) 0.47 *Data were non-normally distributed and are expressed as median and IQR.

.9 (40.0–58.8) 51.4 (43.1–65.6) 0.32 Sleep stage N3 (%) 15.7 (11.6) 15.4 (10.0) 0.76 Sleep stage REM (%) 16.9 (9.2) 16.0 (8.4) 0.50 Snoring time (min)* 64.6 (23.0–140.8) 46.3 (26.0–125.4) 0.45 Snoring time (%)* 19.9 (6.1–35.9) 17.0 (7.6–28.7) 0.47 *Data were non-normally distributed and are expressed as median and IQR. AHI, apnoea-hypopnoea index; N1–N3, sleep stages; ODI, 4% oxygen desaturation index; REM, rapid eye movement sleep; SpO2, oxygen saturation. Figure 2 Box-and-whisker plot for the 4% oxygen desaturation index (4%ODI) in all studied patients (n=36). Figure 3 Box-and-whisker plot for the apnoea-hypopnoea index (AHI) in all studied patients (n=36). Secondary outcome parameters The patients' device acceptance was good with patients reporting no skin discomfort or unpleasant sensations at night. There was no difference in patients' perceived sleep quality between the sham stimulation and the active treatment, but patients reported an improvement of their dry mouth after active treatment (table 4). The only significant side effect observed was one patient who complained about claustrophobia at night; this was during both treatment nights. The total count of mild side effects occurred in 2.8% of the studied cohort and there were no severe adverse events. Table 4 Assessment of symptoms and side effects when waking after the sham stimulation and active treatment nights, as measured by a visual analogue scale (0–10 points)—higher scores indicate an improvement

Secondary outcome parameters The patients' device acceptance was good with patients reporting no skin discomfort or unpleasant sensations at night. There was no difference in patients' perceived sleep quality between the sham stimulation and the active treatment, but patients reported an improvement of their dry mouth after active treatment (table 4). The only significant side effect observed was one patient who complained about claustrophobia at night; this was during both treatment nights. The total count of mild side effects occurred in 2.8% of the studied cohort and there were no severe adverse events. Table 4 Assessment of symptoms and side effects when waking after the sham stimulation and active treatment nights, as measured by a visual analogue scale (0–10 points)—higher scores indicate an improvement Parameters Sham stimulation Active treatment p Value Feeling refreshed 5.7 (2.7–7.2) 6.6 (2.2–8.5) 0.40 Sleep quality 5.6 (2.9–7.1) 6.4 (2.4–8.0) 0.28 Mouth dryness 4.4 (2.2–8.5) 7.4 (4.9–9.7) 0.007 Tongue unpleasant sensation 9.9 (9.4–10.0) 9.9 (9.4–10.0) 0.63 Morning headache 9.4 (6.3–10.0) 9.9 (8.1–10.0) 0.27 Skin discomfort 9.9 (9.5–10.0) 9.9 (9.7–10.0) 0.95 Sleepiness* 3.0 (2.0–3.5) 3.0 (2.0–3.0) 0.29 *Sleepiness was assessed in the mornings using the Stanford Sleepiness Scale to pick up ad-hoc changes. All variables are presented as median and IQR.

) 9.9 (9.4–10.0) 0.63 Morning headache 9.4 (6.3–10.0) 9.9 (8.1–10.0) 0.27 Skin discomfort 9.9 (9.5–10.0) 9.9 (9.7–10.0) 0.95 Sleepiness* 3.0 (2.0–3.5) 3.0 (2.0–3.0) 0.29 *Sleepiness was assessed in the mornings using the Stanford Sleepiness Scale to pick up ad-hoc changes. All variables are presented as median and IQR. Responder group analysis While the primary outcome parameter improved modestly for the whole cohort, in 17 of 36 patients the ODI improved by >25% compared with sham night and/or the total ODI became <5/hour. This group had similar baseline characteristics as the whole cohort, but subjects were more likely to have mild-to-moderate OSA and to be of female gender (table 5). In these ‘responders’, 4%ODI improved by 10.0/hour (95% CI 3.9 to 16.0) (figure 4) and the AHI by 9.1/hour (2.0–16.2) (figure 5). Table 5 Main characteristics of responders and non-responders Parameters Responders (n=17) Non-responders (n=19) p Value Age (years) 48.4 (11.4) 53.0 (10.8) 0.22 Sex (males/females) 13/4 17/2 0.007 BMI (kg/m2) 32.3 (5.4) 30.0 (3.0) 0.19 Neck size (cm) 42.6 (2.8) 42.5 (4.5) 0.97 Mallampati score I, n (%) 3 (17.6) 5 (26.3) 0.53 Mallampati score II, n (%) 5 (29.4) 10 (52.6) 0.26 Mallampati score III, n (%) 6 (35.2) 2 (10.5) 0.07 Mallampati score IV, n (%) 3 (17.6) 4 (21.0) 0.63 Waist:hip ratio 0.98 (0.06) 0.99 (0.08) 0.67 ESS (points)* 10.0 (8.0–13.0) 13.0 (7.5–15.0) 0.14 *Data expressed as median and IQR. BMI, body mass index; ESS, Epworth Sleepiness Scale; n, number.

Parameters Responders (n=17) Non-responders (n=19) p Value Age (years) 48.4 (11.4) 53.0 (10.8) 0.22 Sex (males/females) 13/4 17/2 0.007 BMI (kg/m2) 32.3 (5.4) 30.0 (3.0) 0.19 Neck size (cm) 42.6 (2.8) 42.5 (4.5) 0.97 Mallampati score I, n (%) 3 (17.6) 5 (26.3) 0.53 Mallampati score II, n (%) 5 (29.4) 10 (52.6) 0.26 Mallampati score III, n (%) 6 (35.2) 2 (10.5) 0.07 Mallampati score IV, n (%) 3 (17.6) 4 (21.0) 0.63 Waist:hip ratio 0.98 (0.06) 0.99 (0.08) 0.67 ESS (points)* 10.0 (8.0–13.0) 13.0 (7.5–15.0) 0.14 *Data expressed as median and IQR. BMI, body mass index; ESS, Epworth Sleepiness Scale; n, number. Figure 4 Box-and-whisker plot for the 4% oxygen desaturation index (4%ODI) among ‘responders’ (n=17). There is a significant improvement in the primary outcome between sham stimulation night and active treatment night. Figure 5 Box-and-whisker plot for the apnoea-hypopnoea index (AHI) among ‘responders’ (n=17). There is a significant improvement in the primary outcome between sham stimulation night and active treatment night.

Figure 4 Box-and-whisker plot for the 4% oxygen desaturation index (4%ODI) among ‘responders’ (n=17). There is a significant improvement in the primary outcome between sham stimulation night and active treatment night. Figure 5 Box-and-whisker plot for the apnoea-hypopnoea index (AHI) among ‘responders’ (n=17). There is a significant improvement in the primary outcome between sham stimulation night and active treatment night. In the total study cohort, there were 6 patients with mild OSA, all of whom responded; 13 patients with moderate OSA, with six responders (46.2%) and 17 patients with severe OSA, of whom five were responders (29.4%). There was a low negative correlation between ODI and the response to stimulation and the AHI and the response, respectively (r=−0.334, p=0.023 and r=−0.320, p=0.029, respectively). A stepwise multiple linear regression including the variables ‘age’, ‘gender’ and OSA severity (‘ODI’) at baseline identified only the ODI as an independent predictor for response (standardised β=0.348; 95% CI 0.012 to 0.372). The regression model with ODI severity at baseline predicted about 10% of the response variance (R2=0.121, adjusted R2=0.095, p=0.037). Age (p=0.851) and gender (p=0.720) were excluded from the model (see online supplementary material). There were no further significant correlations or predictors to identify responders.

he regression model with ODI severity at baseline predicted about 10% of the response variance (R2=0.121, adjusted R2=0.095, p=0.037). Age (p=0.851) and gender (p=0.720) were excluded from the model (see online supplementary material). There were no further significant correlations or predictors to identify responders. A further subgroup analysis using more stringent criteria for treatment response revealed that 7 of the 36 patients (19.4%) had a reduction in the AHI of >50% when compared with sham stimulation. Out of this group, 2 of the 36 patients (5.6%) had a residual AHI <10/h during electrical stimulation and a total of 6 of the 36 patients (16.7%) had an AHI <15/hour, which would indicate mild OSA. In the responder group, 2 of the 17 patients (11.7%) had an ODI <5 events/hour during sham treatment. When taking the data of these two patients out the Wilcoxon signed-rank test, there was still a significant difference between the two groups (Z-value −3.4078, p=0.00064).

n AHI <15/hour, which would indicate mild OSA. In the responder group, 2 of the 17 patients (11.7%) had an ODI <5 events/hour during sham treatment. When taking the data of these two patients out the Wilcoxon signed-rank test, there was still a significant difference between the two groups (Z-value −3.4078, p=0.00064). Discussion This is the first randomised, sham-controlled and double-blind clinical trial testing the feasibility of transcutaneous electrical stimulation of the pharyngeal dilator muscles in OSA for a whole night. The primary outcome, improvement in the ODI, revealed a modest improvement for the whole group compared with sham stimulation. The majority of the participants improved their sleep-disordered breathing and almost half of the studied participants were identified as responders with an improvement by a clinically relevant margin. Although the total AHI did not change during stimulation, there was a shift in the ratio of obstructive apnoea:hypopnoea when electrical current was applied. This observation indicates a resolution of the upper airway obstruction during apnoeas with electrical stimulation.

n improvement by a clinically relevant margin. Although the total AHI did not change during stimulation, there was a shift in the ratio of obstructive apnoea:hypopnoea when electrical current was applied. This observation indicates a resolution of the upper airway obstruction during apnoeas with electrical stimulation. Secondary outcome parameters were related to patients' perception of stimulation and sleepiness, snoring, oxygen saturations and side effects. Patients' sleep quality, polysomnographically determined sleep architecture and sleep duration were not adversely affected by the use of electrical stimulation. There were no significant adverse events during the trial and patients did not experience any pain or skin discomfort. Claustrophobia was observed during two nights in a single patient (sham and true stimulation night). Snoring, average and nadir oxygen saturations were not altered by stimulation in this cohort; however, participants experienced fewer problems with a dry mouth following the night with electrical stimulation.

iscomfort. Claustrophobia was observed during two nights in a single patient (sham and true stimulation night). Snoring, average and nadir oxygen saturations were not altered by stimulation in this cohort; however, participants experienced fewer problems with a dry mouth following the night with electrical stimulation. Clinical relevance of transcutaneous stimulation Electrical stimulation has long been studied as a technique to influence upper airway dilator function in sleep. Prior studies have reported both positive11 12 17 and negative13 14 results, likely to reflect the heterogeneous populations studied and different approaches used.7 An important point when using this technique is to avoid arousals from sleep. For many years, no further reports were published using the transcutaneous technique. Following on from our own feasibility study,15 the current trial used a titration algorithm during wakefulness to define individual skin sensation thresholds of electrical current to enable unproblematic nocturnal use of currents within the range observed to be comfortable while awake.

ing the transcutaneous technique. Following on from our own feasibility study,15 the current trial used a titration algorithm during wakefulness to define individual skin sensation thresholds of electrical current to enable unproblematic nocturnal use of currents within the range observed to be comfortable while awake. CTES rather than intermittent or inspiratory-triggered stimulation bursts is less likely to trigger uncomfortable skin sensation, as changes in current intensity activate specific skin receptors; CTES has the advantage of working with a single-channel device (‘pacing’) without the need for ‘sensing’. The use of continuous low current requires less force to maintain the neuromuscular tone in the upper airway than intermittent stimulation as it is easier to maintain upper airway patency than to initiate the reopening of an occluded airway. It is important that this approach can be used for prolonged periods in patients with sleep apnoea with no apparent adverse effect on sleep quality or sleep duration.

in the upper airway than intermittent stimulation as it is easier to maintain upper airway patency than to initiate the reopening of an occluded airway. It is important that this approach can be used for prolonged periods in patients with sleep apnoea with no apparent adverse effect on sleep quality or sleep duration. Hypoglossal neural stimulation Several recent studies have investigated the use of an implantable electrical stimulator to target the distal branch of the hypoglossal nerve that innervates the genioglossus.8–10 18 19 The Clinical Trial by Apnex Medical (Roseville, Minnesota, USA; NCT01446601) was terminated prematurely because the primary outcome, a reduction in OSA severity (defined as AHI reduction >50% and AHI <20 and ODI 4% reduction ≥25% or ODI 4% <5 from baseline to 6-month follow-up) was not met. In contrast, the STAR trial achieved positive results.10 An important feature of the STAR trial was sophisticated screening to identify potential ‘responders’ to the treatment. The investigators excluded patients with pronounced anatomical abnormalities of the upper airway and those with concentric collapse of the retropalatal airway, as assessed by endoscopy during drug-induced sleep.10 After screening 929 patients, 126 participants had a device implanted and 124 completed the trial. At 12 months the median AHI was reduced by 68% from 29.3/hour to 9.0/hour. In a randomised therapy-withdrawal arm the features of sleep apnoea were again observed when the treatment was discontinued (AHI 25.8/hour; ODI 23.0/hour). In the STAR trial, 56 patients were excluded from implantation following endoscopy during drug-induced sleep and 13 patients were excluded because of anatomical abnormalities of the upper airway following surgical consultation.

noea were again observed when the treatment was discontinued (AHI 25.8/hour; ODI 23.0/hour). In the STAR trial, 56 patients were excluded from implantation following endoscopy during drug-induced sleep and 13 patients were excluded because of anatomical abnormalities of the upper airway following surgical consultation. Compared with hypoglossal nerve stimulation, the effect size of CTES in the current trial is smaller. Several factors can explain this observation: the delivery of effective electrical current to the muscles is affected by skin and soft tissue resistance; in patients with large neck circumference, this can be a significant limitation. Stimulation frequency, duration and waveform need to be considered when stimulating for longer periods to maintain force generation while avoiding muscle fatigue. Last, the intensity of the electrical current is dependent on individual comfort and perception and titration of the current needs to be tailored to patient's individual perception of skin discomfort to avoid arousal from sleep. Future studies using transcutaneous electrical stimulation will need to consider the level of obstruction of the upper airway to identify potential responders prior to enrolment in trials. Transcutaneous electrical stimulation might also be used to test patients prior to implanting hypoglossal nerve stimulators to test the individual response.

scutaneous electrical stimulation will need to consider the level of obstruction of the upper airway to identify potential responders prior to enrolment in trials. Transcutaneous electrical stimulation might also be used to test patients prior to implanting hypoglossal nerve stimulators to test the individual response. Responders to transcutaneous electrical stimulation In the current trial, 17 of the 36 participants (47.2%) had improvement in OSA severity during the electrical stimulation night, as defined by the ODI and the AHI.10 From the studied cohort, it appears that this method is more suitable for milder disease and in female subjects. In the multivariate analysis none of the studied characteristics or demographics seemed to be associated with responsiveness. The neck anatomy and the distance between dilator muscles and dermal patches as well as the threshold of comfort for the effective current are likely to be further determinants of effectiveness. Effect on REM sleep AHI In REM sleep, the peripheral skeletal muscles are in a state of physiological atonia during which the work of breathing is predominantly delivered by the diaphragm. The tone of the upper airway dilator muscles is reduced and the critical occlusion pressure can significantly increase making it more likely that the subjects experience obstruction and collapse of the upper airway as the positive intramural and the negative inspiratory pressure gradients favour unopposed occlusion.

m. The tone of the upper airway dilator muscles is reduced and the critical occlusion pressure can significantly increase making it more likely that the subjects experience obstruction and collapse of the upper airway as the positive intramural and the negative inspiratory pressure gradients favour unopposed occlusion. It is therefore of interest to understand how transcutaneous electrical stimulation impacts on the stability of the upper airway, as indicated by the AHI. For the whole group, the REM-sleep related AHI did not significantly improve (REM AHI for sham stimulation 35.3 (26.0) versus active treatment 31.3 (23.9)/hour, p=0.50) and this might indicate that the force to maintain upper airway patency is not strong enough in REM sleep due to the increased load and reduced endogenous neuromuscular tone. However, neither ‘responders’ nor ‘non-responders’ improved their REM AHI significantly with stimulation (p=0.282 for responders, p=0.725 for non-responders), but responders had a lower REM AHI than non-responders with active stimulation (non-responders REM AHI 36.9 (13.3–56.8) versus responders REM AHI 13.3 (6.7–32.7)/hour, p=0.044). This is consistent with the observation that patients with milder forms of sleep apnoea who were less obese were more likely to be responders.

but responders had a lower REM AHI than non-responders with active stimulation (non-responders REM AHI 36.9 (13.3–56.8) versus responders REM AHI 13.3 (6.7–32.7)/hour, p=0.044). This is consistent with the observation that patients with milder forms of sleep apnoea who were less obese were more likely to be responders. Subjective perception of transcutaneous electrical stimulation Previous studies using transcutaneous electrical stimulation for the treatment of sleep apnoea did not apply electrical current for the whole night. The TESLA data on patients' perception importantly highlight the feasibility of the method, with patients remaining asleep with no difference in their sleep profile compared with sham stimulation. Sleep quality was not adversely affected. Following the sham stimulation, patients complained more about a dry mouth in the morning than after active treatment. Whether this is a feature of a more patent upper airway at night or the mouth being closed remains to be elucidated.

ep profile compared with sham stimulation. Sleep quality was not adversely affected. Following the sham stimulation, patients complained more about a dry mouth in the morning than after active treatment. Whether this is a feature of a more patent upper airway at night or the mouth being closed remains to be elucidated. Snoring in patients with sleep apnoea The current trial failed to reduce snoring duration during stimulation nights in patients with sleep apnoea. The breakdown of the AHI shifted during electrical stimulation, the obstructive apnoea:hypopnoea ratio during the baseline sleep study was 1.59 vs 0.82 during active treatment. This indicates a preferential resolution of apnoea with electrical stimulation. However, continued flow-limited breathing during ongoing hypopnoea is likely to contribute to the total time of snoring. In comparison, there was absence of airflow during obstructive apnoea and no snoring sound. Whether snoring in patients with milder sleep apnoea, patients with upper airway resistance syndrome or snorers without flow limitation would improve using this treatment remains to be studied.

bute to the total time of snoring. In comparison, there was absence of airflow during obstructive apnoea and no snoring sound. Whether snoring in patients with milder sleep apnoea, patients with upper airway resistance syndrome or snorers without flow limitation would improve using this treatment remains to be studied. Stimulation pattern The way electrical current can be used to stimulate the upper airway dilator muscles transcutaneously is influenced by multiple factors. Waveform (rectangle, triangular, single impulse, rounded), polarity (unipolar, bipolar), frequency, intensity and duration (continuous, intermittent, triggered) are important factors to generate efficient force and avoid muscle fatigue. Following on from previous work, the current trial used bipolar current, individually titrated by daytime in awake patients to define the lower and upper current thresholds of skin sensation at a frequency of 30 Hz.15 Inspiratory-triggered stimulation has the advantage of avoiding fatigue but requires additional recording for sensing of a physiological signal to identify inspiration (eg, flow, sound, inspiratory effort measuring either movement or electromyography activity), live analysis and sophisticated algorithms to stimulate the muscle at the right time. In contrast, continuous stimulation, as previously used, is effective in avoiding these problems; similar to CPAP therapy, the treatment is provided all night but it is likely to cause muscle fatigue. The stimulation pattern used for the TESLA trial (5 s on/5 s off) was chosen following initial studies to address these points; it does not require additional sensing and its duty cycle will guarantee stimulation during any potentially occurring apnoea which is defined as absence of airflow for >10 s. This pattern also provides sufficient rest time for the muscle to avoid fatigue.

f) was chosen following initial studies to address these points; it does not require additional sensing and its duty cycle will guarantee stimulation during any potentially occurring apnoea which is defined as absence of airflow for >10 s. This pattern also provides sufficient rest time for the muscle to avoid fatigue. Limitations Upper airway endoscopy during drug-induced sedation is presently not standard practice in the UK, but this technique might be of theoretical value in selecting patients for this therapy by describing the level of upper airway obstruction. Endoscopy was not considered when screening patients for the current trial and a better characterisation of the upper airway would likely have led to a greater number of ‘non-responders’ being prospectively excluded. Future trials using transcutaneous electrical stimulation could screen for likely ‘responders’, define patient phenotypes, test the feasibility and effectiveness of this method in the community and test whether it is sufficient to treat REM-related events. Endoscopy could also describe the site of upper airway collapse during delivery of transcutaneous electrical current. It is important to identify the impact of posture in this method and whether a treatment effect is observed in the non-supine posture. Determinants of effectiveness like posture and neck flexion should be studied in future. These points are important in the context that 2 of the 17 responders had a normal ODI in the sham night. Whether this improvement compared with the baseline sleep study is due to the taping of the submental region is unclear, but changes in the neuromuscular tone due to increased afferent feedback could, in part, contribute to our findings. This study was set up to test transcutaneous electrical stimulation for a single night only and it remains to be shown whether use of dermal patches and transcutaneous stimulation over longer follow-up periods is a feasible method and whether responders benefit symptomatically.

uld, in part, contribute to our findings. This study was set up to test transcutaneous electrical stimulation for a single night only and it remains to be shown whether use of dermal patches and transcutaneous stimulation over longer follow-up periods is a feasible method and whether responders benefit symptomatically. The double-blind study design of the TESLA trial was chosen to minimise bias caused by patients' or research staff perception of the method. A computer mode provided random selection of active treatment or sham stimulation. Once the mode was selected, the computer indicated at night that stimulation was delivered, independent of whether it was sham stimulation or active treatment, to simulate a potential stimulation artefact. Independent experienced technicians from the sleep laboratory were assigned for the offline analysis after the patients had been studied over all three nights. The data were then inserted in the database by a different investigator, still in the randomised order and without access to the respective raw data of the polysomnography. After the last patient had been studied and data acquisition had concluded, the data were unblinded in that the sham stimulation nights' data were separated from the active treatment nights' for analysis by the medical statistician. Any potentially small stimulation artefact picked up in the technicians' analysis of the original polysomnographies would not have been reported to the trial's team or the patient and is therefore unlikely to have impacted on the trial's outcome.

om the active treatment nights' for analysis by the medical statistician. Any potentially small stimulation artefact picked up in the technicians' analysis of the original polysomnographies would not have been reported to the trial's team or the patient and is therefore unlikely to have impacted on the trial's outcome. Only about 1 in 10 screened patients underwent randomisation and, therefore, the sample of the randomised patients does not fully represent the whole cohort of patients with OSA. However, it is not uncommon that strict adherence to a protocol requires screening of a large number of patients who, eventually, turn out to be excluded. In the STAR trial, Strollo et al10 conducted a randomised controlled trial of invasive electrical stimulation in patients with OSA and they recruited 929 subjects, of whom only 126 were randomised and included in the treatment group (13.6%). Broader inclusion criteria could help the generalisability of observed effects. However, particularly in studies testing a novel therapeutic approach, the inclusion of inappropriate patients could dilute the effect size, as it is likely that these patients are less responsive. It should be pointed out that this approach will not replace CPAP as the standard first choice therapy for OSA, but might benefit some patients who are unable to wear CPAP masks. Importantly, even patients who are well established on standard treatment and compliant with CPAP may wish to try novel and non-invasive alternatives.20 The prevalence of OSA is rising and even a modest response using electrical stimulation could be helpful in avoiding symptoms and long-term risks in some of the patients who fail CPAP therapy. However, it is important to note that the present study was not powered to identify predictors of response and, therefore, the post-hoc results of the subgroup analysis must be interpreted with caution.

stimulation could be helpful in avoiding symptoms and long-term risks in some of the patients who fail CPAP therapy. However, it is important to note that the present study was not powered to identify predictors of response and, therefore, the post-hoc results of the subgroup analysis must be interpreted with caution. Conclusion Transcutaneous electrical stimulation of the upper airway dilator muscles in OSA can be safely delivered throughout the whole night. Although we observed only a modest improvement in the whole study cohort, approximately half of the studied population were ‘responders’—predominantly those with mild-to-moderate disease. Future use of this method should focus on the prospective identification of responders. Defining upper airway features of those who could benefit and assessing long-term symptomatic improvements and feasibility of the method in the domiciliary setting will be important to further develop this approach.

oderate disease. Future use of this method should focus on the prospective identification of responders. Defining upper airway features of those who could benefit and assessing long-term symptomatic improvements and feasibility of the method in the domiciliary setting will be important to further develop this approach. The authors are grateful for the support of Eskinder Solomon, Tariq Sethi, Gill Arbane and the clinical team at the Sleep Disorders Centre at Guy's and St Thomas’ NHS Foundation Trust. They gratefully acknowledge the hardware support provided by Morgan IAT, Petersfield, UK, and Medical Physics at Guy's and St Thomas’ NHS Foundation Trust. Dr Pengo's salary was partially funded by the Italian Hypertension Society. Professor Moxham's and Dr Steier's contributions were partially supported by the National Institute for Health Research (NIHR) Biomedical Research Centre based at Guy’s and St Thomas' NHS Foundation Trust and King’s College London, UK. Professor Polkey's contribution to this work was supported by the NIHR Respiratory Biomedical Research Unit at the Royal Brompton and Harefield NHS Foundation Trust and Imperial College who part funded his salary. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. Twitter: Follow Nicholas Hart at @NickHartThorax

The authors are grateful for the support of Eskinder Solomon, Tariq Sethi, Gill Arbane and the clinical team at the Sleep Disorders Centre at Guy's and St Thomas’ NHS Foundation Trust. They gratefully acknowledge the hardware support provided by Morgan IAT, Petersfield, UK, and Medical Physics at Guy's and St Thomas’ NHS Foundation Trust. Dr Pengo's salary was partially funded by the Italian Hypertension Society. Professor Moxham's and Dr Steier's contributions were partially supported by the National Institute for Health Research (NIHR) Biomedical Research Centre based at Guy’s and St Thomas' NHS Foundation Trust and King’s College London, UK. Professor Polkey's contribution to this work was supported by the NIHR Respiratory Biomedical Research Unit at the Royal Brompton and Harefield NHS Foundation Trust and Imperial College who part funded his salary. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. Twitter: Follow Nicholas Hart at @NickHartThorax Contributors: All listed authors confirm that they have fulfilled the following criteria, as required by the International Committee of Medical Journal Editors (www.icmje.org): substantial contributions to the conception or design of the work; or the acquisition, analysis or interpretation of data for the work and drafting the work or revising it critically for important intellectual content and final approval of the version to be published and agreeing to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

drafting the work or revising it critically for important intellectual content and final approval of the version to be published and agreeing to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Funding: Italian Hypertension Society, NIHR Biomedical Research Centre based at Guy’s and St Thomas’ NHS Foundation Trust and King’s College London, ResMed Foundation (10.13039/100003305). Competing interests: None declared. Patient consent: Obtained. Ethics approval: Ethics Committee for Clinical Trials London-Dulwich, UK (12/LO/1428). Provenance and peer review: Not commissioned; externally peer reviewed.

Introduction Patients with emphysema are breathless because of gas trapping and hyperinflation due to the loss of lung elastic tissue and resultant expiratory airways collapse. Surgical lung volume reduction (LVRS) can improve survival, lung function and quality of life in selected patients with exercise limitation and heterogeneous emphysema.1 2 The placement of endobronchial valves (BLVR) as a means to reduce lung volume is a potential alternative to LVRS. BLVR has been shown to improve lung function, reduce chest wall asynchrony and reduce the work of breathing.3 4 Atelectasis following BLVR is associated with improved survival.5 6 The BeLieVeR-HIFi study, a double-blind sham-controlled trial,7 8 found that BLVR led to significant improvements in lung function, exercise capacity and health status at 3 months when performed in patients with a higher chance of developing atelectasis—those with intact interlobar fissures and heterogeneously distributed emphysema. In this research letter we present data from the control patients in the BeLieVeR-HIFi study who went on to have open label endobronchial valve treatment after completion of the clinical trial. We also combine these data with patients from the original treatment arm who had been found to be collateral ventilation negative (CV−) using the Chartis catheter system and completed trial follow-up.

eLieVeR-HIFi study who went on to have open label endobronchial valve treatment after completion of the clinical trial. We also combine these data with patients from the original treatment arm who had been found to be collateral ventilation negative (CV−) using the Chartis catheter system and completed trial follow-up. Methods The study protocol, design, randomisation, assessments, procedure and details of the participants have been previously published8 and further details of the methods and statistical analyses are in online supplementary panel S1. 10.1136/thoraxjnl-2016-208865.supp1supplementary data

eLieVeR-HIFi study who went on to have open label endobronchial valve treatment after completion of the clinical trial. We also combine these data with patients from the original treatment arm who had been found to be collateral ventilation negative (CV−) using the Chartis catheter system and completed trial follow-up. Methods The study protocol, design, randomisation, assessments, procedure and details of the participants have been previously published8 and further details of the methods and statistical analyses are in online supplementary panel S1. 10.1136/thoraxjnl-2016-208865.supp1supplementary data Results Baseline characteristics of the open label treated patients (n=14) are detailed in online supplementary table S1. Three-month follow-up data were available for 12 open label patients. One died 4 days following treatment due to a pneumothorax occurring at their home; one developed a persistent intractable cough necessitating valve removal and did not return for follow-up evaluation. Clinical outcomes are detailed in table 1 and online supplementary table S2, and illustrated in figure 1A–D and online supplementary figure S1A–D. FEV1 increased by 24.2 (27.3)% from baseline following endobronchial valve treatment. The patients also experienced statistically significant improvements in carbon monoxide transfer factor and COPD assessment test score as well as measures of exercise capacity. Table 1 also includes data from the 19 CV− patients from the original treatment arm of the BeLieVeR-HIFi trial for whom follow-up data were available (‘original CV− treatment arm patients’) and for the two groups combined (‘all CV− treated patients’) (n=31).

sessment test score as well as measures of exercise capacity. Table 1 also includes data from the 19 CV− patients from the original treatment arm of the BeLieVeR-HIFi trial for whom follow-up data were available (‘original CV− treatment arm patients’) and for the two groups combined (‘all CV− treated patients’) (n=31). Table 1 Change in lung function, health status and exercise tolerance at 90 days Open label valve treated patients (n=12) p Value Original Chartis CV− treatment arm patients (n=19) p Value All CV− treated patients (per Chartis) (n=31) p Value %ΔFEV1 24.2 (27.3) 0.06 28.9 (40.1) 0.001 27.3 (36.4) 0.0002 ΔFEV1 (l) 0.14 (0.20) 0.06 0.23 (0.28) 0.001 0.19 (0.25) 0.0002 %ΔFVC 5.1 (13.0) 0.5 7.51 (16.9) 0.03 6.5 (15.6) 0.02 ΔTLC (l) −0.23 (0.49) 0.13 −0.37 (0.56) 0.01 −0.33 (0.53) 0.002 ΔRV (l) −0.42 (0.80) 0.41 −0.54 (0.76) 0.01 −0.49 (0.76) 0.007 ΔRV/TLC % −3.50 (6.77) 0.10 −4.6 (6.9) 0.03 −4.3 (6.85) 0.004 ΔFRC (l) −0.28 (0.83) 0.27 −0.42 (0.69) 0.04 −0.38 (0.75) 0.009 ΔTLco (absolute percentage points) 3.5 (6.77) 0.005 3.45 (6.2) 0.02 3.62 (5.16) 0.0007 ΔKco (mmol/min/kPa/l) 0.10 (0.07) 0.007 0.05 (0.07) 0.009 0.07 (0.07) <0.0001 ΔCAT −3.9 (5.5) 0.05 −4.2 (10.1) 0.20 −4.1 (8.5) 0.03 ΔSGRQc total −7.5 (14.9) 0.08 −7.5 (20.8) 0.3 −8.5 (20.2) 0.05 Δ6MWD 29 (48) 0.16 33.2 (80.2) 0.02 32.6 (68.7) 0.01 ΔTLim 138 (312) 0.08 165 (260) 0.07 155 (275) 0.01 Data are presented as mean (SD). The p values are for the Wilcoxon signed-rank test.

(0.07) <0.0001 ΔCAT −3.9 (5.5) 0.05 −4.2 (10.1) 0.20 −4.1 (8.5) 0.03 ΔSGRQc total −7.5 (14.9) 0.08 −7.5 (20.8) 0.3 −8.5 (20.2) 0.05 Δ6MWD 29 (48) 0.16 33.2 (80.2) 0.02 32.6 (68.7) 0.01 ΔTLim 138 (312) 0.08 165 (260) 0.07 155 (275) 0.01 Data are presented as mean (SD). The p values are for the Wilcoxon signed-rank test. 6MWD, 6-min walk distance; CAT, COPD assessment test score; Chartis CV−, no interlobar collateral ventilation on Chartis assessment; CV−, collateral ventilation negative; FRC, functional residual capacity; Kco, carbon monoxide transfer coefficient; RV, residual volume; SGRQc, St George's Respiratory Questionnaire for COPD; TLC, total lung capacity; TLco, carbon monoxide transfer factor; Tlim, endurance time on cycle ergometry at 70% of peak workload. Figure 1 Response to bronchoscopic lung volume reduction in open label treated patients, in the original BeLieVeR-HIFi treated patients who were collateral ventilation negative (CV−) and in both groups combined. (A) FEV1; (B) endurance time on cycle ergometry at 70% maximal work rate (Tlim); (C) St George's Respiratory Questionnaire for COPD (SGRQc); (D) Residual volume (RV) assessed by body plethysmography. The p values are for the Wilcoxon signed-rank test. *p<0.05, **p<0.01, ***p<0.001.

ve (CV−) and in both groups combined. (A) FEV1; (B) endurance time on cycle ergometry at 70% maximal work rate (Tlim); (C) St George's Respiratory Questionnaire for COPD (SGRQc); (D) Residual volume (RV) assessed by body plethysmography. The p values are for the Wilcoxon signed-rank test. *p<0.05, **p<0.01, ***p<0.001. Responder rates for achievement of minimal clinically important differences were similar in the open label patients to those in the original treatment group of the trial (see online supplementary table S3). Eight of 12 patients treated with valves developed atelectasis or complete lobar collapse on CT, and another two had significant volume loss. Details of adverse events are in online supplementary panel S2. Discussion These data further support the view that treating patients with heterogeneous emphysema and without interlobar CV with endobronchial valves leads to improved lung function, exercise capacity and quality of life. The benefits are more impressive where stricter patient selection criteria are employed, although there is still significant variability in response. The improvement in gas transfer is of particular interest as this is the lung function measure most strongly associated with survival in people with COPD.9

ality of life. The benefits are more impressive where stricter patient selection criteria are employed, although there is still significant variability in response. The improvement in gas transfer is of particular interest as this is the lung function measure most strongly associated with survival in people with COPD.9 In the original BeLieVeR-HIFi trial, eligibility for valves was based on the results of CT fissure analysis. CV was measured directly using the Chartis system, but by design patients in the intervention arm were still treated even if they were CV-positive. In this open label follow-up however, patients had had a previous bronchoscopy which confirmed airway anatomy suitable for adequate valve placement and prior Chartis measurements confirming the absence of CV. The proportion of open label treated patients with radiological evidence of volume loss was 83% (10 of 12), higher than the 65% in the original treatment cohort (15 of 23) with rates of responders achieving minimum clinically important differences (MCIDs) in the various outcomes broadly similar to the original group.

. The proportion of open label treated patients with radiological evidence of volume loss was 83% (10 of 12), higher than the 65% in the original treatment cohort (15 of 23) with rates of responders achieving minimum clinically important differences (MCIDs) in the various outcomes broadly similar to the original group. The development of bronchoscopic lung volume reduction techniques has been driven by the desire to offer patients safer and cheaper alternatives to LVRS. As patient selection improves, increasing the likelihood of successful lung volume reduction, there is a significantly higher rate of pneumothorax than the 5% reported in earlier trials. In this series the rate was 10.3%, including one fatal event, though others have reported rates of 20–25%.10 Although a marker of procedural effectiveness, with eventual clinical benefit in the majority after treatment of the pneumothorax,11 pneumothorax can be fatal in people with advanced lung disease and little respiratory reserve. Thus, the mortality risk of BLVR may not be lower than that of surgical intervention, especially when compared with unilateral LVRS.1 Pooled data suggest that 70% of pneumothoraces occur within 72 hours12 and therefore it may be prudent to observe patients in hospital for 4 days post treatment. Patients who do suffer pneumothoraces are initially managed conservatively and patiently in hospital but may ultimately require valve removal or surgery.10 This also reduces the advantage of BLVR over LVRS in terms of hospital length of stay, though the level of dependency during this observation period is low.

atment. Patients who do suffer pneumothoraces are initially managed conservatively and patiently in hospital but may ultimately require valve removal or surgery.10 This also reduces the advantage of BLVR over LVRS in terms of hospital length of stay, though the level of dependency during this observation period is low. Our series highlight that fissure completeness, assessed visually on CT thorax (even by dedicated thoracic radiologists) is not a perfect surrogate for the absence of interlobar CV. Out of 50 patients enrolled in the original trial and all judged to have intact fissures, 8 patients (16%) had positive CV on Chartis assessment. In conclusion, bronchoscopic lung volume reduction using endobronchial valves leads to clinically significant improvements in lung function, exercise capacity and quality of life in the majority of patients when appropriately selected. The risk of pneumothorax needs to be considered and a period of close observation is recommended. Longer follow-up to assess durability of clinical benefits and effect on survival is needed as well as direct comparison of endobronchial valve placement against surgical approaches. Twitter: Follow Nicholas Hopkinson at @COPDdoc Contributors: NSH, MIP, SJ, PLS developed the study. NSH, MIP, SJ, DHC were involved in patient selection. PLS, ZZ and WHM performed procedures. CD performed assessments. ZZ performed statistical analyses. ZZ and NSH prepared the first draft of this paper which all authors subsequently contributed to and approved. NSH is the guarantor.

, MIP, SJ, PLS developed the study. NSH, MIP, SJ, DHC were involved in patient selection. PLS, ZZ and WHM performed procedures. CD performed assessments. ZZ performed statistical analyses. ZZ and NSH prepared the first draft of this paper which all authors subsequently contributed to and approved. NSH is the guarantor. Funding: This project was funded by the Efficacy and Mechanism Evaluation (EME) Programme (Grant number: EME 10/90/10) and is funded by the Medical Research Council (MRC) and managed by the National Institute for Health Research (NIHR) on behalf of the MRC-NIHR partnership. The EME Programme is funded by the MRC and NIHR, with contributions from the CSO in Scotland and NISCHR in Wales and the HSC R&D Division, Public Health Agency in Northern Ireland. It is managed by the NIHR Evaluation, Trials and Studies Coordinating Centre (NETSCC) based at the University of Southampton. The study was supported by the NIHR Respiratory Biomedical Research Unit at the Royal Brompton and Harefield NHS Foundation Trust and Imperial College, London who part funded MIP's salary and wholly funded ZZ's salary. The endobronchial valves were provided free of charge by the manufacturers PulmonX Ltd. PulmonX had no input into the trial design, data analysis or presentation. The views expressed in this publication are those of the authors and not necessarily those of the MRC, NHS, NIHR or the UK Department of Health.

ZZ's salary. The endobronchial valves were provided free of charge by the manufacturers PulmonX Ltd. PulmonX had no input into the trial design, data analysis or presentation. The views expressed in this publication are those of the authors and not necessarily those of the MRC, NHS, NIHR or the UK Department of Health. Competing interests: PLS, SJ, MIP, ZZ, WHM and NSH have been investigators in trials of endobronchial valves, coils, thermal ablation and the airway bypass procedure, and the authors' institution reimbursed for trial expenses by the device manufacturers. Ethics approval: London—Bentham Research Ethics Committee. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: Requests for anonymised individual patient data can be made to the corresponding author.

Key messages What is the key question? The mechanisms underlying sleep-related hypoventilation in patients with coexistent COPD and obstructive sleep apnoea (OSA), an overlap syndrome, are unknown. What is the bottom line? This study shows that sleep-related hypoventilation in patients with overlap syndrome is due to an increase in upper airway resistance associated with OSA rather than reduction of neural respiratory drive associated with COPD. Why read on? Sleep-related hypoventilation in patients with COPD alone mainly occurs because of a decrease in neural respiratory drive whereas it is mainly a result of an increase in upper airway resistance in patients with overlap syndrome. Introduction COPD is a common condition and patients with COPD are subject to hypoxaemia or even respiratory failure during sleep because of hypoventilation.1 2 In prior reports we showed that this hypoventilation, in the absence of upper airway obstruction was due to a reduction in neural respiratory drive to the respiratory muscles as measured by the diaphragm electromyogram (EMGdi).2 3 Obstructive sleep apnoea (OSA) is characterised by repeated partial or complete collapse of the upper airway leading to increased upper airway resistance and arousal from sleep which are associated with increased neural respiratory drive.4 5

he respiratory muscles as measured by the diaphragm electromyogram (EMGdi).2 3 Obstructive sleep apnoea (OSA) is characterised by repeated partial or complete collapse of the upper airway leading to increased upper airway resistance and arousal from sleep which are associated with increased neural respiratory drive.4 5 It has been previously reported that nocturnal oxygen desaturation is more severe in patients with COPD and OSA, a phenotype termed the ‘overlap syndrome’, than those with COPD alone.6 However, this view is mainly based on the results derived from patients with mild or moderate COPD and predominant OSA who are often obese7 8 and may not be the case for patients with more severe COPD.9 In contrast to the hypoventilation in patients with COPD, OSA is characterised by increased upper airway resistance and is associated with an increase in neural respiratory drive.2 4 5 10 If patients with severe COPD develop OSA, the sleep-related reduction in neural respiratory drive associated with COPD could be offset by an increase in neural respiratory drive due to increased upper airway resistance. Consequently, ventilation in patients with severe COPD may not further decrease when COPD and mild or modest OSA occur together.

develop OSA, the sleep-related reduction in neural respiratory drive associated with COPD could be offset by an increase in neural respiratory drive due to increased upper airway resistance. Consequently, ventilation in patients with severe COPD may not further decrease when COPD and mild or modest OSA occur together. The EMGdi recorded from a multipair oesophageal electrode can be used to assess neural respiratory drive,2 4 5 10 11 and upper airway resistance can be inferred by the ratio of the tidal volume (VT) to EMGdi (VT/EMGdi) assuming that lung mechanics and lower airway resistance remain constant in wakefulness and sleep.2 Here we aimed to investigate the underlying mechanisms of hypoventilation during sleep in patients with COPD alone and patients with overlap syndrome by comparing neural respiratory drive and ventilation. The data from 10 of the patients with COPD alone and 10 healthy subjects has previously been reported; these were the participants with complete datasets from that report.2

hypoventilation during sleep in patients with COPD alone and patients with overlap syndrome by comparing neural respiratory drive and ventilation. The data from 10 of the patients with COPD alone and 10 healthy subjects has previously been reported; these were the participants with complete datasets from that report.2 Methods Subjects Thirty-nine consecutive patients with COPD from the outpatient clinic of Guangzhou Institute of Respiratory Disease were studied, including 10 patients whose data were reported previously.2 Exclusion criteria were a known diagnosis of OSA before the study, clinically significant coexisting diseases including cardiovascular and neuromuscular disease, an acute exacerbation of COPD in the preceding month, FEV1<20% predicted, and use of long-term oxygen therapy. Usual medications including inhaled bronchodilators were allowed for all the patients. Fourteen healthy subjects and 14 patients with OSA were also studied; as noted 10 of the healthy subjects had appeared in our prior report.2 The study was approved by the Ethics Committee of the First Affiliated Hospital of Guangzhou Medical University and all subjects gave written informed consent. Lung function tests Spirometry (Cosmed Micro Quark, Cosmed, Italy) was performed on the same night as polysomnography. Measurements were repeated until maximal reproducible values of FEV1 were achieved with variation of less than 150 ml between tests. The modified Medical Research Council dyspnoea scale (mMRC) was also recorded before polysomnography.

osmed Micro Quark, Cosmed, Italy) was performed on the same night as polysomnography. Measurements were repeated until maximal reproducible values of FEV1 were achieved with variation of less than 150 ml between tests. The modified Medical Research Council dyspnoea scale (mMRC) was also recorded before polysomnography. Oesophageal electrode and its positioning A soft fine catheter with an external diameter of 1.6 mm was used to record the EMGdi. The catheter had 10 metal coils which provided five pairs of recording electrodes was used to record the EMGdi. The oesophageal electrode was passed through the nose into the stomach and was carefully positioned based on the EMGdi amplitude recorded simultaneously from the five pairs of electrodes, as reported previously.4 12 Briefly the position of the electrode catheter was judged to be optimal when electrode 5 was located at the level of the diaphragm confirmed by the amplitude of EMGdi activity being greatest in pairs 1 and 5, and smallest in pair 3 during inspiration. When the electrode catheter was in the optimal position, it was securely taped at the nose. The EMGdi signals were amplified and band-pass filtered between 20 Hz and 1 kHz (RA-8, Yinghui Medical Technology Co., Guangzhou, China).

EMGdi activity being greatest in pairs 1 and 5, and smallest in pair 3 during inspiration. When the electrode catheter was in the optimal position, it was securely taped at the nose. The EMGdi signals were amplified and band-pass filtered between 20 Hz and 1 kHz (RA-8, Yinghui Medical Technology Co., Guangzhou, China). Measurement of maximal EMGdi Maximal EMGdi was recorded during two manoeuvres: maximal inspiration to total lung capacity (TLC) and maximal inspiration against a closed airway at functional residual capacity for 3 s. Each manoeuvre was repeatedly performed until subjects were able to master the techniques. More than three maximal efforts, with an interval of 30 s or more between them, were recorded for analysis and the largest single value was considered maximal. Polysomnography Full overnight polysomnograms including the EEG (C3-A2, C4-A1), left and right electro-oculograms (EOG), submental electromyogram (EMGchin), airflow, snoring, body position, oxygen saturation and end-tidal CO2 were recorded. Airflow was recorded with a pneumotachgraph connected to a full facemask, and integrated to produce volume. All signals were recorded simultaneously using a Powerlab recording system (ADInstruments, Castle Hill, Australia). The sampling rate was 2 kHz for EMGdi and 200 Hz for other signals. EMGdi was recorded before and during sleep.

orded with a pneumotachgraph connected to a full facemask, and integrated to produce volume. All signals were recorded simultaneously using a Powerlab recording system (ADInstruments, Castle Hill, Australia). The sampling rate was 2 kHz for EMGdi and 200 Hz for other signals. EMGdi was recorded before and during sleep. Analysis of data Conventional polysomnography was manually analysed based on standard criteria.13 An obstructive apnoea event was defined as absence of airflow for longer than 10 s, while there was phasic inspiratory EMGdi.10 13 A hypopnoea event was defined as reduction of airflow of more than 30% for longer than 10 s associated with ≥3% desaturation or the event being associated with arousal.13 Overlap syndrome was defined as apnoea hypopnoea index (AHI) ≥5 events/hour in the presence of COPD. Clinically significant sleep-related desaturation was defined as SaO2<90% which lasted longer than 5 min. The root mean square of the EMGdi was calculated by computer with a time constant of 100 ms. The root mean square reported was that from the electrode pair with the largest EMGdi amplitude for each breathing cycle. To avoid the influence of the electrocardiogram on the EMGdi, root mean square was measured from segments between QRS complexes. A ratio of VT to root mean square of EMGdi (VT/EMGdi) of each breath was used to assess upper airway resistance.2 Data were selected during stable breathing without respiratory events unless stated otherwise. Ten minutes of data before sleep and at least 14 min during stage 2 in the supine position were selected for analysis.

of VT to root mean square of EMGdi (VT/EMGdi) of each breath was used to assess upper airway resistance.2 Data were selected during stable breathing without respiratory events unless stated otherwise. Ten minutes of data before sleep and at least 14 min during stage 2 in the supine position were selected for analysis. To further confirm that the VT/EMGdi could be used to assess change in upper airway resistance from wakefulness to sleep, we measured VT/EMGdi during wakefulness, during snoring and hypopnoea events in patients with overlap syndrome. At least 50 breathing cycles were measured for each condition. Data were presented as mean±SD. The Wilcoxon rank sum test and Kruskal–Wallis test followed by Dunn's test were used to test for differences between groups as appropriate. Statistical significance was determined by p<0.05.

To further confirm that the VT/EMGdi could be used to assess change in upper airway resistance from wakefulness to sleep, we measured VT/EMGdi during wakefulness, during snoring and hypopnoea events in patients with overlap syndrome. At least 50 breathing cycles were measured for each condition. Data were presented as mean±SD. The Wilcoxon rank sum test and Kruskal–Wallis test followed by Dunn's test were used to test for differences between groups as appropriate. Statistical significance was determined by p<0.05. Results Four patients with COPD and two healthy subjects could not tolerate the oesophageal catheter or full face mask connected to the pneumotachgraph, therefore final data were derived from 35 patients with COPD, 12 healthy subjects and 14 patients with OSA (see table 1 and online supplementary table E-1). Using the above criteria 16 of the patients with COPD were found to have overlap syndrome (FEV1 47.5±16.2%; AHI 20.5±14.1 events/hour) and 19 patients had COPD alone (FEV1 38.5±16.3%; AHI 1.9±1.6). The body mass index (BMI) in patients with COPD alone (19.9±2.7 kg/m2) was significantly lower than that in patients with overlap syndrome (23.5±3.8 kg/m2) (p<0.05). The FEV1 in patients with COPD alone was not significantly different from that in patients with overlap syndrome (38.5%±16.3% vs 47.5%±16.2%, p>0.05). The respiratory events in patients with overlap syndrome were predominantly hypopnoea rather than apnoea, except for subject 1 (see table 1 and online supplementary table E-2). Overall, the ratio of hypopnoea to apnoea events in patients with overlap syndrome was 4.4:1. COPD alone and overlap syndrome groups were similar to each other but, as expected, had significantly greater dyspnoea compared with healthy subjects and patients with OSA (p<0.001). The maximal EMGdi was similar between groups and was 176.9±72.1, 192.3±56.6, 156.1±50.3, 166.2±40.8 μV for COPD alone, patients with overlap syndrome, healthy subjects and patients with OSA (p>0.05), respectively.

pected, had significantly greater dyspnoea compared with healthy subjects and patients with OSA (p<0.001). The maximal EMGdi was similar between groups and was 176.9±72.1, 192.3±56.6, 156.1±50.3, 166.2±40.8 μV for COPD alone, patients with overlap syndrome, healthy subjects and patients with OSA (p>0.05), respectively. Table 1 Basic information for all subjects Characteristics COPD (n=19) Overlap (n=16) Normal (n=12) OSA (n=14) Age (years) 62.0±9.2 61.5±10.2 58.1±8.3 55.8±8.9 BMI (kg/m2) 19.9±2.7 23.5±3.8 22.3±2.8 25.8±3.7 FVC%pred 73.7±19.9 68.5±17.4 98.6±9.6 100±12.6 FEV1%pred 38.5±16.3 47.5±16.2 98.1±9.2 99.1±14.6 FEV1/FVC% 41.3±11.2 54.6±11.4 79.6±3.4 80.4±6.3 mMRC 1.8±0.9 1.6±0.5 0±0 0±0 EMGdi-max (μV) 176.9±72.1 192.3±56.6 156.1±50.3 166.2±40.8 EMGdi-rest (μV) 29.5±13.3 29.3±15.8 12.6±3.7 14.2±7.9 AHI (events/hour) 1.9±1.6 20.5±14.1 1.6±1.4 25.6±18 TST (hours) 4.4±2.0 3.9±1.8 4.2±1.4 5.6±1.8 Rest SaO2 (%) 96.8±1.4 97.1±1.3 98.0±1.0 96.9±0.7 Mean SaO2 (%) 96.0±1.6 95.8±1.4 97.6±0.9 96.1±1.4 Mini SaO2 (%) 90.6±5.3 90.9±3.9 93.6±2.4 85.3±8.7 TB90% 4.7±5.5 0.4±0.5 – 2.2±5.5 AHI, Apnoea Hypopnoea Index; BMI, body mass index; EMGdi-max, the maximum of the diaphragm electromyogram; mean SaO2, mean nocturnal oxygen saturation; mini SaO2, minimum nocturnal oxygen saturation; mMRC, modified British Medical Research Council; TB90%, time spent with saturation below 90%; TST, total sleep time. 10.1136/thoraxjnl-2016-208467.supp1supplementary data

Characteristics COPD (n=19) Overlap (n=16) Normal (n=12) OSA (n=14) Age (years) 62.0±9.2 61.5±10.2 58.1±8.3 55.8±8.9 BMI (kg/m2) 19.9±2.7 23.5±3.8 22.3±2.8 25.8±3.7 FVC%pred 73.7±19.9 68.5±17.4 98.6±9.6 100±12.6 FEV1%pred 38.5±16.3 47.5±16.2 98.1±9.2 99.1±14.6 FEV1/FVC% 41.3±11.2 54.6±11.4 79.6±3.4 80.4±6.3 mMRC 1.8±0.9 1.6±0.5 0±0 0±0 EMGdi-max (μV) 176.9±72.1 192.3±56.6 156.1±50.3 166.2±40.8 EMGdi-rest (μV) 29.5±13.3 29.3±15.8 12.6±3.7 14.2±7.9 AHI (events/hour) 1.9±1.6 20.5±14.1 1.6±1.4 25.6±18 TST (hours) 4.4±2.0 3.9±1.8 4.2±1.4 5.6±1.8 Rest SaO2 (%) 96.8±1.4 97.1±1.3 98.0±1.0 96.9±0.7 Mean SaO2 (%) 96.0±1.6 95.8±1.4 97.6±0.9 96.1±1.4 Mini SaO2 (%) 90.6±5.3 90.9±3.9 93.6±2.4 85.3±8.7 TB90% 4.7±5.5 0.4±0.5 – 2.2±5.5 AHI, Apnoea Hypopnoea Index; BMI, body mass index; EMGdi-max, the maximum of the diaphragm electromyogram; mean SaO2, mean nocturnal oxygen saturation; mini SaO2, minimum nocturnal oxygen saturation; mMRC, modified British Medical Research Council; TB90%, time spent with saturation below 90%; TST, total sleep time. 10.1136/thoraxjnl-2016-208467.supp1supplementary data The mean SaO2% over the entire sleep period was 96.8±1.4, 97.1±1.3, 98.0±1.0 and 96.9±0.7 during wakefulness and 96.0±1.6, 95.8±1.4, 97.6±0.9 and 96.1±1.4 during sleep for patients with COPD alone, patients with overlap syndrome, healthy subjects and patients with OSA, respectively. COPD alone and overlap syndrome groups were similar to each other but had a lower mean minimal SaO2% during overnight sleep compared with healthy subjects (see table 1 and online supplementary table E-2). Mean minimal SaO2% in patients with OSA was lower than those in the other three groups. Three of 19 (16%) patients with COPD alone developed SaO2<90% for longer than 5 min, whereas no subjects in both the patients with overlap syndrome and the healthy subjects developed significant oxygen desaturation (see online supplementary table E-2).

n patients with OSA was lower than those in the other three groups. Three of 19 (16%) patients with COPD alone developed SaO2<90% for longer than 5 min, whereas no subjects in both the patients with overlap syndrome and the healthy subjects developed significant oxygen desaturation (see online supplementary table E-2). Ventilation in patients with COPD, patients with overlap syndromes, healthy subjects and those with OSA during wakefulness and sleep VE was similar between groups and was 8.6±2.0, 8.3±2.0, 8.3±1.6 and 8.0±2.7 L/min for patients with COPD alone, patients with overlap syndromes, healthy subjects and patients with OSA during wakefulness (p>0.05). VE decreased significantly from wakefulness to non-rapid eye movement (NREM) sleep for patients with COPD alone (8.6±2.0 to 6.5±1.5 L/min, p<0.001), patients with overlap syndrome (8.3±2.0 to 6.1±1.8 L/min, p<0.001) and patients with OSA (8.0±2.7 to 6.3±1.9 L/min, p<0.01), although the change in VE failed to attain statistical significance between wakefulness and NREM sleep in healthy subjects (8.3±1.6 to 7.5±1.3 L/min, p=0.07). The decrease in VE from wakefulness to sleep was similar between patients with COPD and patients with overlap syndrome (24% vs 27%, p>0.05), but it was greater in both groups than that observed in healthy subjects (10%), p<0.05. Change in VE was almost proportional to change in VT in patients with COPD, patients with overlap syndrome and patients with OSA (table 2, figures 1 and 2). Table 2 Diaphragm EMG, ventilation during wakefulness and NREM sleep

Ventilation in patients with COPD, patients with overlap syndromes, healthy subjects and those with OSA during wakefulness and sleep VE was similar between groups and was 8.6±2.0, 8.3±2.0, 8.3±1.6 and 8.0±2.7 L/min for patients with COPD alone, patients with overlap syndromes, healthy subjects and patients with OSA during wakefulness (p>0.05). VE decreased significantly from wakefulness to non-rapid eye movement (NREM) sleep for patients with COPD alone (8.6±2.0 to 6.5±1.5 L/min, p<0.001), patients with overlap syndrome (8.3±2.0 to 6.1±1.8 L/min, p<0.001) and patients with OSA (8.0±2.7 to 6.3±1.9 L/min, p<0.01), although the change in VE failed to attain statistical significance between wakefulness and NREM sleep in healthy subjects (8.3±1.6 to 7.5±1.3 L/min, p=0.07). The decrease in VE from wakefulness to sleep was similar between patients with COPD and patients with overlap syndrome (24% vs 27%, p>0.05), but it was greater in both groups than that observed in healthy subjects (10%), p<0.05. Change in VE was almost proportional to change in VT in patients with COPD, patients with overlap syndrome and patients with OSA (table 2, figures 1 and 2). Table 2 Diaphragm EMG, ventilation during wakefulness and NREM sleep COPD (n=19) Overlap (n=16) Characteristics Wakefulness NREM △% p Value Wakefulness NREM △% p Value EMGdi%max 29.5±13.3 23.0±8.9 22 0.006 29.3±15.8 27.3±14.7 7 0.38 VE (L) 8.6±2.0 6.5±1.5 24 <0.001 8.3±2.0 6.1±1.8 27 0.001 VT (L) 0.47±0.11 0.37±0.09 22 0.001 0.47±0.11 0.36±0.10 24 0.001 VT/EMGdi 0.39±0.23 0.37±0.21 5 0.92 0.35±0.21 0.28±0.19 20 0.041 ETCO2 (%) 4.5±0.5 4.5±0.6 0 0.89 4.7±0.7 4.7±0.8 0 0.48 RR 18.4±3.1 17.8±2.1 3 0.33 17.6±1.7 17.2±2.9 2 0.35 Normal (n=12) OSA (n=14) EMGdi%max 12.6±3.7 12.2±3.3 3 0.84 14.2±7.9 23.4±10.8 65 0.001 VE (L) 8.3±1.6 7.5±1.3 10 0.07 8.0±2.7 6.3±1.9 21 0.006 VT (L) 0.50±0.09 0.49±0.10 2 0.56 0.48±0.16 0.39±0.10 19 0.008 VT/EMGdi 0.69±0.18 0.69±0.20 0 0.91 0.56±0.24 0.28±0.13 49 0.001 ETCO2 (%) 4.8±0.3 5.0±0.3 4 0.002 4.5±0.8 4.6±0.9 3 0.18 RR 16.7±2.3 15.6±2.3 7 0.10 16.6±3.0 16.0±2.3 4 0.27 △%, percentage change compared with wakefulness; EMGdi%max, percentage of maximal EMGdi; ETCO2, end-tidal CO2; NREM, non-rapid eye movement; p value, comparison of wakefulness and NREM; RR, respiratory rate; VE, minute ventilation; VT, tidal volume; VT/EMGdi, the ratio of tidal volume to peak root mean square of EMGdi.

△%, percentage change compared with wakefulness; EMGdi%max, percentage of maximal EMGdi; ETCO2, end-tidal CO2; NREM, non-rapid eye movement; p value, comparison of wakefulness and NREM; RR, respiratory rate; VE, minute ventilation; VT, tidal volume; VT/EMGdi, the ratio of tidal volume to peak root mean square of EMGdi. Figure 1 Diaphragm electromyogram (EMG) recording from five pairs of oesophageal electrodes and airflow from pneumotachography during polysomnography in patients with COPD alone, patients with overlap syndrome, normal subjects and patients with obstructive sleep apnoea (OSA). Diaphragm EMG decreases in patients with COPD alone but increases in patients with OSA from wakefulness to sleep, whereas it changes little in patients with overlap syndrome and in normal subjects.

atients with COPD alone, patients with overlap syndrome, normal subjects and patients with obstructive sleep apnoea (OSA). Diaphragm EMG decreases in patients with COPD alone but increases in patients with OSA from wakefulness to sleep, whereas it changes little in patients with overlap syndrome and in normal subjects. Figure 2 Electromyogram (EMG)di% (left panel), ventilation (middle panel) and the VT/EMGdi (right panel) in patients with COPD alone, overlap syndrome, normal subjects and patients with obstructive sleep apnoea (OSA). Ventilation decreases from wakefulness to sleep in all four groups but only in patients with COPD, overlap syndrome and OSA is the reduction statistically significant. EMGdi decreases in patients with COPD alone but increases in patients with OSA from wakefulness to non-rapid eye movement (NREM) sleep, whereas it remains the same in normal subjects and patients with overlap syndrome. VT/EMGdi decreases from wakefulness to sleep in patients with overlap syndrome and those with OSA but it changes little in normal subjects and patients with COPD alone. The decrease in ventilation is associated with a reduction of EMGdi in patients with COPD alone whereas ventilation reduction is associated with decreased VT/EMGdi in patients with overlap syndrome and those with OSA.

lap syndrome and those with OSA but it changes little in normal subjects and patients with COPD alone. The decrease in ventilation is associated with a reduction of EMGdi in patients with COPD alone whereas ventilation reduction is associated with decreased VT/EMGdi in patients with overlap syndrome and those with OSA. Neural respiratory drive decreased significantly in patients with COPD alone (29.5±13.3% vs 23.0±8.9% of maximal EMGdi, p<0.01), but increased significantly in patients with OSA (14.2±7.9% vs 23.4±10.8% of maximal EMGdi) from wakefulness to NREM sleep. However, neural respiratory drive changed little from wakefulness to stage 2 sleep in patients with overlap syndrome (29.3±15.8% vs 27.3±14.7% of maximal EMGdi, p>0.05) and in healthy subjects (12.6±3.7% vs 12.2±3.3% of maximal EMGdi, p>0.05). Respiratory rate also changed little between wakefulness and NREM sleep in all four groups (p>0.05). VT/EMGdi was similar between wakefulness and NREM sleep in both healthy subjects (0.69±0.18 vs 0.69±0.20) and patients with COPD alone (0.39±0.23 vs 0.37±0.21) but it decreased significantly from wakefulness to sleep in patients with overlap syndrome (0.35±0.21 vs 0.28±0.19, p<0.05) and patients with OSA (0.56±0.24 vs 0.28±0.13, p<0.01) (see table 2 and online supplementary tables E-3 and E-4, and supplementary figure E-1).

nd patients with COPD alone (0.39±0.23 vs 0.37±0.21) but it decreased significantly from wakefulness to sleep in patients with overlap syndrome (0.35±0.21 vs 0.28±0.19, p<0.05) and patients with OSA (0.56±0.24 vs 0.28±0.13, p<0.01) (see table 2 and online supplementary tables E-3 and E-4, and supplementary figure E-1). In patients with overlap syndrome the VT/EMGdi decreased from the state of wakefulness, to sleep in which snoring was recorded (0.35±0.21 vs 0.24±0.14, p<0.001) and further decreased during stage 2 hypopnoea events (0.15±0.12, p<0.001) (table 3; data for stage 2 sleep with or without snoring are shown in online supplementary table E-4). Table 3 Diaphragm EMG (EMGdi), tidal volume (VT) and the efficacy of neural respiratory drive (VT/EMGdi) during wakefulness, snoring and hypopnoea in patients with overlap syndrome

In patients with overlap syndrome the VT/EMGdi decreased from the state of wakefulness, to sleep in which snoring was recorded (0.35±0.21 vs 0.24±0.14, p<0.001) and further decreased during stage 2 hypopnoea events (0.15±0.12, p<0.001) (table 3; data for stage 2 sleep with or without snoring are shown in online supplementary table E-4). Table 3 Diaphragm EMG (EMGdi), tidal volume (VT) and the efficacy of neural respiratory drive (VT/EMGdi) during wakefulness, snoring and hypopnoea in patients with overlap syndrome EMGdi/max% VT (mL/kg) VT/EMGdi Overlap subjects Wakefulness Snoring Hypopnoea Wakefulness Snoring Hypopnoea Wakefulness Snoring Hypopnoea 1 49.0 54.7 48.1 5.1 4.0 1.2 0.10 0.07 0.02 2 23.1 25.2 25.4 5.8 4.1 2.6 0.25 0.16 0.10 3 28.1 37.8 35.3 10.2 7.5 3.2 0.36 0.20 0.09 4 58.3 50.8 52.8 5.6 4.3 1.9 0.10 0.08 0.04 5 21.0 18.2 14.8 9.3 6.9 5.1 0.44 0.38 0.34 6 31.3 22.3 12.1 5.9 8.3 2.4 0.19 0.37 0.20 7 17.1 23.4 22.8 9.5 7.5 4.1 0.56 0.32 0.18 8 48.9 38.7 39.1 7.0 3.6 2.3 0.14 0.09 0.06 9 18.2 23.3 18.1 8.5 6.4 4.5 0.47 0.27 0.25 10 12.3 18.1 12.9 9.3 9.3 4.9 0.76 0.51 0.38 11 14.9 22.0 18.7 8.5 4.3 1.8 0.57 0.19 0.10 12 17.1 19.5 20.2 8.9 4.3 1.7 0.52 0.22 0.08 13 25.4 29.2 24.2 6.3 4.4 2.1 0.25 0.15 0.09 14 14.0 13.0 12.6 7.3 6.1 4.4 0.52 0.47 0.35 15 58.7 55.3 46.2 5.8 3.7 2.8 0.10 0.07 0.06 16 30.9 24.2 28.8 6.1 4.9 2.2 0.20 0.20 0.08 Mean 29.3 29.7 27.0 7.4 5.6 2.9 0.35 0.24 0.15 SD 15.8 13.6 13.4 1.7 1.8 1.3 0.21 0.14 0.12 EMGdi%max, percentage of maximal diaphragm electromyogram.

.2 24.2 6.3 4.4 2.1 0.25 0.15 0.09 14 14.0 13.0 12.6 7.3 6.1 4.4 0.52 0.47 0.35 15 58.7 55.3 46.2 5.8 3.7 2.8 0.10 0.07 0.06 16 30.9 24.2 28.8 6.1 4.9 2.2 0.20 0.20 0.08 Mean 29.3 29.7 27.0 7.4 5.6 2.9 0.35 0.24 0.15 SD 15.8 13.6 13.4 1.7 1.8 1.3 0.21 0.14 0.12 EMGdi%max, percentage of maximal diaphragm electromyogram. Discussion In the present study we show that unlike healthy subjects who exhibited a small (approximately 10%) difference between NREM sleep and wakefulness, VE was lower in stage 2 sleep than wakefulness in patients with COPD alone and those with overlap syndrome or OSA. In patients with overlap syndrome neural respiratory drive was similar between NREM sleep and wakefulness. However, neural respiratory drive from wakefulness to stage 2 sleep increased in patients with OSA but decreased in those with COPD alone, suggesting that mild or moderate OSA can partially compensate for reduction of neural respiratory drive inherent to COPD and certainly does not seem to worsen hypoventilation associated with COPD.

respiratory drive from wakefulness to stage 2 sleep increased in patients with OSA but decreased in those with COPD alone, suggesting that mild or moderate OSA can partially compensate for reduction of neural respiratory drive inherent to COPD and certainly does not seem to worsen hypoventilation associated with COPD. Methodological issues This is the first study to simultaneously record ventilation and neural respiratory drive from patients with COPD alone and patients with an overlap syndrome. Some studies have previously assessed ventilation during sleep using respiratory inductance plethysmography,14–17 which can be inaccurate particularly in patients with COPD.18 19 Becker et al18 used a pneumotachograph to measure ventilation during sleep in patients with COPD, but the pneumotachograph in their study was connected with a nasal mask rather than a full-face mask, leading to a potential underestimate of ventilation if patients breathed through the mouth. Using a full face mask to quantify ventilation we found a larger reduction of ventilation in NREM than that reported by Becker et al18 using a nasal mask.

their study was connected with a nasal mask rather than a full-face mask, leading to a potential underestimate of ventilation if patients breathed through the mouth. Using a full face mask to quantify ventilation we found a larger reduction of ventilation in NREM than that reported by Becker et al18 using a nasal mask. The FEV1 in patients with overlap syndrome was numerically, although not statistically, higher than that in patients with COPD alone. Because obesity is an important factor contributing to development of OSA and because COPD is a chronic wasting disease sometimes associated with weight loss as disease progresses, it could be argued that patients with severe COPD are less likely to develop OSA because of weight loss. However the current dataset (by showing patients with COPD with polysomnographically proven OSA) shows that this conjecture is insufficient to prevent OSA in patients with COPD.20 21 However we note that while there was no statistically significant difference in FEV1 between patients with COPD alone and those with overlap syndrome, type II error cannot be absolutely excluded given the small numbers dictated by a physiological study of this nature. We would also point out that the similarity in mMRC score and EMGdi at rest between patients with COPD alone and those with overlap syndrome also suggests the severity of airway obstruction between the two groups is similar. In the present study, the prevalence of overlap syndrome (46%) seems to be high for a COPD cohort study, in particular for those with low BMI. However, a high prevalence of overlap syndrome has recently been reported.9 21 Unlike Caucasian patients with COPD,7–9 the BMI in Chinese patients with COPD is usually low as previously reported.2 22 A high prevalence of overlap syndrome (46%) in the COPD cohort study, despite the low BMI, may thus be attributable to Asiatic craniofacial morphology.23

rlap syndrome has recently been reported.9 21 Unlike Caucasian patients with COPD,7–9 the BMI in Chinese patients with COPD is usually low as previously reported.2 22 A high prevalence of overlap syndrome (46%) in the COPD cohort study, despite the low BMI, may thus be attributable to Asiatic craniofacial morphology.23 Oesophageal pressure combined with airflow recordings has recently been used to quantify upper airway resistance.24 However, oesophageal pressure is influenced by lung volume and airflow and thus has a potential limitation in the assessment of upper airway resistance in patients with OSA which is characterised by changes in lung volume and airflow. Classically a catheter positioned in the pharynx has been used to measure upper airway resistance by measurement of pharyngeal pressure.25 26 However, upper airway resistance derived from measurement of pharyngeal pressure is a difficult technique and may be variable during both wakefulness19 and sleep.26 Besides providing data regarding neural respiratory drive, EMGdi has an advantage over oesophageal pressure in the assessment of upper airway resistance because it is independent of change in airflow and lung volume.27 28 Normally, VT is achieved in response to neural respiratory drive reflected by EMGdi after overcoming total respiratory resistance, including elastic recoil of the lung, lower airway resistance and upper airway resistance. VT/EMGdi reflects changes in upper airway resistance if lung mechanics and lower airway resistance remain the same;2 one limitation of our study which should therefore be acknowledged is that lung volumes were not measured during sleep. OSA is characterised by an increase in upper airway resistance from wakefulness to snoring and a further increase to hypopnoea. In line with this concept, VT/EMGdi decreases significantly from wakefulness to snoring and further decreases to hypopnoea in patients with overlap syndrome, which supports using change in VT/EMGdi to assess changes in upper airway resistance.

resistance from wakefulness to snoring and a further increase to hypopnoea. In line with this concept, VT/EMGdi decreases significantly from wakefulness to snoring and further decreases to hypopnoea in patients with overlap syndrome, which supports using change in VT/EMGdi to assess changes in upper airway resistance. Significance of findings This study shows that neural drive decreases from the state of wakefulness to sleep in patients with COPD alone, as previously reported.2 3 19 It may be expected that neural respiratory drive in patients with overlap syndrome would also decrease from wakefulness to sleep, but we found that this is not the case, consistent with our prior report that obstructive respiratory events are associated with increased neural respiratory drive.10 Several studies suggest that the increased neural respiratory drive during sleep is directly related to the presence of increased upper airway resistance since neural respiratory drive decreases when airway resistance is offset by, for example, treatment with continuous positive airway pressure29 or inhalation of Heliox (76%He/24%O2).25 The present study suggests that sleep-related reduction of neural respiratory drive characteristic of COPD could be offset by an increase in upper airway resistance as a consequence of coexistent OSA, in an adaptation potentially protective in nature.

positive airway pressure29 or inhalation of Heliox (76%He/24%O2).25 The present study suggests that sleep-related reduction of neural respiratory drive characteristic of COPD could be offset by an increase in upper airway resistance as a consequence of coexistent OSA, in an adaptation potentially protective in nature. It has been hypothesised that desaturation in patients with overlap syndrome would be more severe than that in patients with COPD alone because COPD and OSA can cause desaturation.6 7 30 However, this view is mainly derived from studies of patients who had predominantly mild or moderate COPD or who were recruited from patients with OSA and obesity,7 8 20 30 and thus may not represent a clinical cohort of patients with severe COPD. However, a recent cohort study of non-obese patients with severe COPD showed that the number of patients with COPD alone who required oxygen supplementation was similar to that in those with overlap syndrome,9 suggesting the prevalence of oxygen desaturation is similar between patients with COPD alone and patients with overlap syndrome. In the present study we found that mean oxygen saturation and minimal oxygen saturation during overnight sleep were similar in patients with or without overlap syndrome. Moreover, although patients with coexistent OSA and severe COPD usually have brief periods of desaturation, prolonged desaturation (SaO2<90 for longer than 5 min) occurred more often in patients with severe COPD alone than those with overlap syndrome (see online supplementary table E-2). This finding may be clinically significant for the management of patients with overlap syndrome. If sleep-related hypoventilation or desaturation in patients with severe COPD had been worsened by coexistent OSA, one would have to be cautious with nutritional supplements which may raise BMI in patients with COPD. In contrast, the present study suggests that when patients with severe COPD develop mild or moderate OSA, a sleep-related reduction of neural respiratory drive as a consequence of COPD would be relieved because of increased upper airway resistance, preventing ventilation from further decreasing. The observation that 16% of patients with COPD alone developed significant oxygen desaturation but none of patients with an overlap syndrome developed a significant desaturation may give a hint that if patients with severe COPD develop mild or moderate OSA, oxygen desaturation would not necessarily be worsen.

creasing. The observation that 16% of patients with COPD alone developed significant oxygen desaturation but none of patients with an overlap syndrome developed a significant desaturation may give a hint that if patients with severe COPD develop mild or moderate OSA, oxygen desaturation would not necessarily be worsen. This interesting finding is in line with the recent report that the clinical outcome in end-stage patients with overlap syndrome is better than those with COPD alone.31 Nevertheless, we note that pathophysiological change in patients with overlap syndrome of severe COPD and mild OSA might differ from those with mild COPD and severe OSA; in addition variation on this spectrum may differ between western patients in whom OSA is more driven by obesity and Asian patients (as studied here) in whom craniofacial morphometry may be more relevant. In conclusion the mechanism underlying the reduction of ventilation at stage 2 sleep in patients with COPD alone differs from that in patients with overlap syndrome. Ventilation reduction in patients with COPD alone is mainly because of a decrease in neural respiratory drive whereas it is mainly a result of an increase in upper airway resistance in patients with overlap syndrome. Contributors: Conception and design: BH, GL, SX, JS, JM, MP, YL; analysis and interpretation: BH, GL, SX, RC; drafting the manuscript for important intellectual content: BH, GL, SX, RC, JS, JM, MP, YL.

In conclusion the mechanism underlying the reduction of ventilation at stage 2 sleep in patients with COPD alone differs from that in patients with overlap syndrome. Ventilation reduction in patients with COPD alone is mainly because of a decrease in neural respiratory drive whereas it is mainly a result of an increase in upper airway resistance in patients with overlap syndrome. Contributors: Conception and design: BH, GL, SX, JS, JM, MP, YL; analysis and interpretation: BH, GL, SX, RC; drafting the manuscript for important intellectual content: BH, GL, SX, RC, JS, JM, MP, YL. Funding: The work was supported by National Natural Science Foundation of China (NSFC No. 81120108001 and 81270143). Professor Polkey's contribution to this project was supported by the NIHR Respiratory Biomedical Research Unit at the Royal Brompton and Harefield NHS Foundation Trust and Imperial College, London UK, who part funded his salary. Competing interests: None declared. Patient consent: Obtained. Ethics approval: The Ethics Committee of the First Affiliated Hospital of Guangzhou Medical University. Provenance and peer review: Not commissioned; externally peer reviewed.