A randomized sham-controlled trial on the effect of continuous positive airway pressure treatment on gait control in severe obstructive sleep apnea patients

To determine the effect of continuous positive airway pressure (CPAP), the gold standard treatment for obstructive sleep apnea syndrome (OSAS), on gait control in severe OSAS patients. We conducted a randomized, double-blind, parallel-group, sham-controlled monocentric study in Grenoble Alpes University Hospital, France. Gait parameters were recorded under single and dual-task conditions using a visuo-verbal cognitive task (Stroop test), before and after the 8-week intervention period. Stride-time variability, a marker of gait control, was the primary study endpoint. Changes in the determinants of gait control were the main secondary outcomes. ClinicalTrials.gov Identifier: (NCT02345694). 24 patients [median (Q1; Q3)]: age: 59.5 (46.3; 66.8) years, 87.5% male, body mass index: 28.2 (24.7; 29.8) kg. m−2, apnea–hypopnea index: 51.6 (35.0; 61.4) events/h were randomized to be treated by effective CPAP (n = 12) or by sham-CPAP (n = 12). A complete case analysis was performed, using a mixed linear regression model. CPAP elicited no significant improvement in stride-time variability compared to sham-CPAP. No difference was found regarding the determinants of gait control. This study is the first RCT to investigate the effects of CPAP on gait control. Eight weeks of CPAP treatment did not improve gait control in severe non-obese OSAS patients. These results substantiate the complex OSAS-neurocognitive function relationship.

. Participants characteristics, obstructive sleep apneas characteristics at baseline and after 8 weeks of CPAP or Sham-CPAP. AHI apnea-hypopnea index, BMI body mass index, CPAP continuous positive airway pressure, ESS Epworth Sleepiness Scale, MMSE Mini Mental State Examination, ODI oxygen desaturation index, defined as a drop in SpO 2 > 3% for at least 10 s ; SpO 2 : pulsed dioxygen saturation; SpO 2 < 90%: percentage of recording time spent at a SpO 2 < 90% Past cardiovascular events: defined as a past history of nonfatal stroke, non-fatal myocardial infarction, transient ischemic attack, any cardiac event of coronary origin (including revascularization procedures), hospitalizations due to cardiac or cerebrovascular causes (acute coronary syndrome, stroke, heart failure, heart rhythm disorder including atrial fibrillation, bleeding, etc.). AHI Flow was measured using the CPAP machine's internal microprocessor. Data are presented as median [Q1; Q3] or as number (%) of participants.

Discussion
In this randomized, sham-controlled trial (RCT), eight weeks of CPAP treatment did not improved gait control of severe OSAS patients. We found no improvement in STV, our primary outcome, following eight weeks of CPAP treatment. The STV values obtained at baseline were comparable to those of the severe OSAS patients reported by Celle et al. 8 and in our previous study 10 showing impaired gait control in OSAS patients. This negative result contrasts with that of two recent uncontrolled interventional studies. Allali et al. 9 showed improvements in gait speed and temporal gait parameters specifically in dual tasks. Their study included mostly obese patients, lacked an appropriate control group and the changes in gait performance they reported were subtle with limited clinical relevance. In our previous prospective, non-randomized controlled study 10 , we showed that at baseline OSAS patients had significantly larger STV compared to matched healthy controls. After eight weeks of CPAP treatment, STV was significantly improved and no longer different from controls. The lack of an OSAS group treated by sham-CPAP and the absence of a retest of the controls eight weeks after baseline assessment limited the conclusions of this previous study. The present study, using a stronger methodological design, suggests that eight weeks of CPAP may be insufficient to reverse gait control impairments in severe OSAS patients. This result was supported by the lack of changes after CPAP treatment in the main determinants of gait control (i.e. brain oxygenation and cognitive performance). A key outcome was DT gait assessment using a Stroop test. We found no difference between the two groups but a mutual facilitation 18 reflecting an improvement in DT performance both in gait (i.e. a reduction in STV) and cognition (i.e. an increased Stroop CRR) compared to ST performance. This might be explained by an exercise-induced arousal effect 19 , the positive incidence of acute physical activity on cognitive     16-item free and total recall: for the free and total recall tests, results are presented as the sum of the three consecutive trials, with a maximum score of 48. For the delayed free and total recall tests, maximum score is 16, higher score indicating better performance. Paced Auditory Serial Addition test: A pre-recorded tape delivered a random series of 61 numbers from 1 to 9, at a constant rate of 1 number every 4 s. Participants were instructed to add pairs of numbers such that each number was added to the one that immediately preceded it on the recording: the second was added to the first, the third to the second, the fourth to the third, and so on. The response had to be given before the presentation of the next stimulus (4 s later). The sum of any given pair never exceeded 15 www.nature.com/scientificreports/ performance. However, such an effect is expected to occur only in patients with morphologically and functionally unaltered prefrontal cortices suggesting only slight neurological consequences of OSAS prior to treatment in the included patients 18 . All patients underwent a comprehensive neurocognitive evaluation, covering the domains supposed to be most impaired in OSAS 5,20 . CPAP treatment elicited no significant change in cognitive performance, except an improvement in the Stroop Interference test. We had included patients without significant cognitive deficiency at baseline and most of them were highly educated, with at least half having more than 11 years of formal education. Thus, included patients were normal or supra normal, free of comorbidities, and with probably a high cognitive reserve 6 , representing a preserved adaptability of cognitive processes explaining their reduced susceptibility to cognitive deterioration with aging or pathologies 21 .
During cognitive or motor tasks, cerebral hemodynamic regulation is essential so as to deliver the adequate amount of oxygen and substrates required for brain metabolism 22 . Severe OSAS might have impaired cerebral hemodynamic regulation 23,24 . CPAP treatment, through suppression of nocturnal intermittent hypoxia and fragmented sleep, is thought to limit OSAS-related neuroinflammation and improve cerebrovascular hemodynamics 3 . CPAP has a clear acute impact in suppressing acute cerebral hemodynamic abnormalities occurring concurrently with sleep obstructive events 25 . Some studies have suggested a partial improvement in cerebral blood flow (CBF) in the frontal areas during wakefulness in OSAS patients with a dose-response effect-size related to CPAP adherence 26,27 . However, in our well controlled RCT no chronic impact of CPAP treatment on prefrontal cortices hemodynamics was found, as assessed by fNIRS during daytime walking. This result is consistent with those of a previous study from our group 28 , showing no improvement in prefrontal cortex hemodynamics while performing an incremental cycling test following eight weeks of CPAP treatment.
The major strength of our study is its randomized controlled design which is unique in the field of gait control and OSAS. We acknowledge that the duration of the CPAP intervention might not have been long enough for CPAP to have any effect. Although we note that most of the existing studies on neurocognitive function or brain imaging in OSAS patients have used similar timeframes. Moreover, one third of the patients of the CPAP group showed low CPAP compliance (< 4 h of use per night). Altogether, this may have dampened the expected positive effect of CPAP on gait control.
We included only non-obese OSAS patients. This choice was driven by the fact that obesity is a cofounding factor when assessing gait control and postural control, as overweight and even more obesity directly affect gait control and gait kinetics 29,30 . Another limitation could be that the gait and postural tasks might have been too www.nature.com/scientificreports/ easy for participants, with parameters already satisfactory at baseline, which reduced the chances to conclude.
Lastly, the small sample size of highly selected patients, together with three loss of follow-up patients in the CPAP group may have impacted our results. As we included only severe OSAS patients, further generalization of the study findings should be restricted to OSAS patients with an AHI > 30 events. h −1 , with a low burden of comorbidities.

Conclusion
This first randomized controlled trial in the field showed that eight weeks of CPAP treatment did not improve gait control in severe non-obese OSAS patients. Long-term real-life observational data on different clusters of phenotypes 31 might help to identify specific subgroups for whom CPAP could have a significant impact on gait control, which could then be tested in further RCTs. An external data quality control was performed systematically for the following criteria: informed consent, complications, adverse events, and case report forms.

Methods
The primary study endpoint was the change in STV of OSAS patients after eight weeks of CPAP treatment, in comparison with sham-CPAP treatment.

Participants. Patients were recruited from a consecutive sample at the Sleep Laboratory of Grenoble Alpes
University Hospital and from the Outpatient Sleep Clinic in a French tertiary-care university hospital (Grenoble, France) between February 2015 and December 2018. Inclusion criteria were: (1) age ≥ 18 years and < 70 years, (2) non-obese patients (body-mass index (BMI) < 30 kg. m −2 ), to control for obesity-related gait kinetic changes 29 (3) severe OSAS (apnea-hypopnea index, AHI > 30. events h −1 ) on polysomnography or respiratory polygraphy, (4) to be naive of CPAP treatment and (5) to present with a strictly normal neurological examination. Patients were not included if they declined to participate or were unable to give their informed consent. Patients with any of the following criteria were also not included: (1) the presence of any medical condition supposed to interfere with gait, (2) cognitive impairment, defined as a score on the Mini-Mental State Examination (MMSE 32 ) < 24/30, (3) current pregnancy, (4) ongoing hypnotic or central nervous system targeted medication, (5) chronic alcohol consumption and (6) a profession requiring an effective CPAP treatment (e.g. public transport or truck drivers).

Overnight sleep studies. Obstructive sleep apnea syndrome (OSAS) diagnosis was based on an overnight
sleep study (respiratory polygraphy or polysomnography), performed before inclusion and treatment by continuous positive airway pressure (CPAP) or sham-CPAP according to standard procedures 33 . Polysomnographic recordings were undertaken with electroencephalography (EEG) electrode positions C3/A2-C4/A1-Cz/01 of the international 10-20 Electrode Placement System, eye movements, chin electromyogram and ECG with a modified V2 lead. Sleep was scored manually according to standard criteria. For polysomnographic and polygraphic recordings, airflow was measured with nasal pressure prongs, together with the sum of oral and nasal thermistor signals. Respiratory effort was monitored using abdominal and thoracic bands. Oxygen saturation was measured using a pulse oximeter and oxygen desaturation index (ODI), mean nocturnal SpO 2 , and percentage of recording time spent at a SpO 2 < 90% were also calculated. Respiratory events were scored manually by a trained sleep specialist and the apnea-hypopnea index (AHI) was calculated from the number of apneas and hypopneas per hour of sleep according to international guidelines 34 . An apnea was defined as the complete cessation or a reduction of at least 90% of airflow for at least 10 s and hypopnea as a reduction of at least 30% in the nasal pressure signal associated with either oxygen desaturation of at least 3% or an EEG arousal, both lasting for at least 10 s 35 . Apnea was classified as obstructive, central or mixed, according to the presence or absence of respiratory efforts. The classification of hypopnea as obstructive or central was based on the thoraco-abdominal band signal and the shape of the respiratory nasal pressure curve (flow limited aspect or not). Severe OSAS was defined as an AHI > 30 events h −1 .
Evaluation protocol. At baseline and at the end of the 8-week interventional period the same parameters were recorded following the same protocol for evaluation, performed in the same order at the same hour of the day: (1) baseline clinical assessment, (2) single task (ST) and dual task (DT) overground gait assessments, using a dual-task paradigm similar to our previous work 10 , (3) comprehensive neuropsychological assessment, (4) postural control assessment during ST and DT, using a dual-task paradigm similar to our previous work 10 and (5) treadmill gait assessment in ST and DT, with continuous recording of prefrontal cortex functional activation using functional Near-Infrared Spectroscopy (fNIRS). Following the collection of baseline parameters patients were randomized to be treated by CPAP or sham-CPAP for eight weeks (the duration of interventions previously used in studies investigating the effects of CPAP treatment on cognition 15 and gait 9,10  www.nature.com/scientificreports/ was homogenously set between 6 and 14 cmH 2 O in the effective CPAP group. Patients receiving sham-CPAP had a similar device, delivering a pressure that was too low to suppress sleep respiratory events, as previously validated [36][37][38] . As a double-blind protocol, investigators and the study team as well as patients were blinded to treatment allocation. CPAP compliance and residual AHI 39 under CPAP were downloaded from the CPAP device software. Baseline clinical assessment. Prior to all gait and postural assessments, patients underwent a screening history and physical examination to ensure that they were free of significant orthopedic, neurological and visual disorder which could interfere with the outcomes of the present study. Cardiovascular risk factors were recorded, including past or actual smoking, systemic hypertension, diabetes and dyslipidemia (Table 1). Past cardiovascular events, including non-fatal stroke and non-fatal myocardial infarction, transient ischemic attack, any cardiac event of coronary origin (including vascular recanalization procedures), hospitalization due to cardiac or cerebrovascular causes (acute coronary syndrome, stroke, heart failure, heart rhythm disorder including atrial fibrillation, bleeding, etc.) were retrieved from medical interviews. Patients were then tested for their ability to distinguish the four colors (red, yellow, blue, green) used in our Stroop test appropriately by using the same screen setting as during the dual-task assessments. They completed an Epworth Sleepiness Scale (ESS).
Overground gait assessment. Spatiotemporal gait parameters were recorded using a modular optoelectronic floor-based system (OptoGait, Microgate, Bolzano, Italy), which consists of two parallel 1-m bars (one emitting, containing 96 lights diodes and one receiving bar). The system demonstrated high reliability for the assessment of spatiotemporal gait parameters 40 . Ten emitting and ten receiving bars were disposed parallel to each other to build a 10-m long, 1.5-m width walkway. Patients walked barefoot, at their self-selected comfortable speed, continuously around an oval circuit 10 (Fig. 4a, b). Each participant completed three familiarization loops prior to the 20 evaluation loops, alternating ST (10 evaluation loops, gait only) and DT (10 evaluation loops, gait and Stroop test performed simultaneously). Gait kinetic parameters were recorded at 1 kHz sample frequency and analyzed using the OptoGait software (version 1.10.7.0, Microgate). The average value of all recorded steps was used for data analysis. The following gait kinetic parameters were recorded and calculated: speed (m.s −1 ), cadence (step.s −1 ), stride time (s), total double support (percentage of total gait cycle time), step width (cm). The walk ratio (WR, cm/(steps/min)) 41 , a speed-independent index of the overall neuromotor gait control, which reflects balance, between-step variability, and attentional demand was calculated as follows: WR = Step Length (cm)/cadence (steps/min). Stride time variability (STV), our primary study endpoint, is a marker of gait control related to cerebral integrity and greater STV values reflect impaired gait control 42 . STV was calculated as the mean of the coefficient of variation of stride time as follows 42 : Dual-task paradigm. The DT paradigm consists in the assessment of the interferences occurring when a motor and a cognitive task are performed simultaneously 12,43 . This can result in performance decrements in one or both of the tasks, suggesting the simultaneous engagement of the same functional brain subsystems 44 . We used the same DT paradigm as in our previous work 10 , combining gait and postural assessments with a Stroop color-word interference test (Fig. 4c), a cognitive task considered to specifically assess executive functions. Stroop test consists in color names (blue, red, green and yellow) written in a conflicting font color. Participants were instructed to name the word font color and to inhibit reading the word (e.g., the word "green" written in red font color). To avoid learning effects, we used 30 different versions of the Stroop test, presented randomly to the participants throughout the different assessments. The Stroop test was displayed on a black background screen installed at the end of the corridor, which height was adjusted for each participant and for each evaluation. Words were presented one by one, and the evaluator skipped manually to the next one after the subject gave an oral response. The number of correct answers and errors were recorded by a trained evaluator. In ST, a red sight was displayed on the black background screen. In DT, participants were asked to perform both the motor and cognitive tasks at the best of their capacity without any task prioritization. The correct response rate for Stroop test performance in ST and DT gait assessments 45 was calculated as follows: To deepen our understanding of the complex interactions between gait, posture and cognition, we calculated the dual task effect (DTE) 18 for each gait, postural and cognitive parameter, as follows: DTE quantifies the dual-task interference, i.e. the relative change in performance associated with dual-tasking. Indeed, as attention or executive functions are limited resources, dividing attention between two concurrent tasks can result in a decrement in performance in one or both of the tasks, relative to when each task is performed alone 18 . Therefore, the magnitude of DT interference is influenced by the interaction between the two tasks.
Neuropsychological assessment. A comprehensive neuropsychological assessment, covering the most frequently impaired domains in OSAS 5,20 (i.e. episodic memory and executive functions) was administered by a certified neuropsychologist, blind for treatment allocation, always at the same time of the day, in the same order: Neuropsychological assessment lasted approximatively 1 h, with short breaks between each test to avoid fatigue. Raw performances at the different evaluations are displayed in Table 2.
Standing posture. Posture and gait are interrelated functions and their control is anatomically and functionally intertwined 46 . Moreover, postural control is altered in OSAS 47 . We assessed standing balance using a posturographic platform (Feetest 6, TechnoConcept, Céreste, France), composed of two dynamometric posturographic clogs (with a total of 12 strain gauges). Participants stood upright barefoot on the clogs, in a conventional manner (feet side by side, forming an angle of 30° with both heels separated by 4 cm), with their arms alongside the body. To ensure participants security, the platform was settled in the middle of handlebars and an evaluator stood behind the participants to avoid them falling. The examination took place in a dedicated quiet room with standardized lighting conditions. Assessments were alternatively performed in ST (posture only, 4 trials) and in DT (posture and Stroop test performed simultaneously, 4 trials). Each trial lasted 30 s and a minimal 30 s-period of rest sitting was systematic between trials. In ST, subjects were instructed to maintain their balance while looking straight-ahead at a fixed red sight displayed on the screen installed 1.5 m ahead of each participant. In DT, subjects were asked to maintain the erect posture as still as possible and to perform the Stroop test at the best of their capacity without any task prioritization. Data were recorded with a sampling rate of 40 Hz and calculated using the Posturewin 4 software. As a marker of an efficient standing postural control, the amount of sway was assessed by calculating the center of pressure (CoP) area (90% confidence ellipse, mm 2 ). The smaller the area, the better the postural control 48 . Beyond center of pressure (CoP) Area, the following postural kinetic parameters were recorded and calculated: mediolateral instability, defined as one standard deviation of the CoP displacement along the mediolateral axis, anteroposterior instability, defined as one standard deviation of the CoP displacement along the anteroposterior axis and mean speed of CoP displacement (mm.s −1 ).
Treadmill dual-task gait and cerebral oxygenation. To investigate the link between brain oxygenation and cognitive and gait performances, we designed a DT gait test on a treadmill (Gait Trainer 3, Biodex Medical System, Patients were instructed to name the words font color and to inhibit reading the word (Correct answers here: "red" then "yellow" then "blue"). Words were presented one by one, and the evaluator skipped manually to the next one after the subject gave an oral response. www.nature.com/scientificreports/ NY, USA), with continuous recording of cerebral oxygenation using bilateral fNIRS on the two prefrontal cortices. fNIRS is a non-invasive, optical neuroimaging technique based on neurovascular coupling to infer changes in neuronal activity 49 . fNIRS provides reproducible measurements for investigating functional activation of the human cerebral cortex by tracking changes in cerebral O 2 status while performing cognitive tasks or walking 50 . Left and right pre-frontal cortices hemodynamics were assessed respectively between Fp1 and F3 (left prefrontal cortex) and Fp2 and F4 (right prefrontal cortex) locations according to the international 10-20 EEG system with a 3.5-cm inter-optodes distance. The probe holders were secured to the skin with double-sided tape and maintained with Velcro headbands. Oxyhemoglobin ([HbO 2 ]) and deoxyhemoglobin ([HHb]) concentration changes and the tissue saturation index (TSI) were measured throughout the testing sessions using a twowavelength (780 and 850 nm) multichannel, continuous wave NIRS system (Oxymon MkIII, Artinis Medical Systems, the Netherlands). Total hemoglobin concentration ([HbTot]) was calculated as the sum of [HbO 2 ] and [HHb] and reflects changes in tissue blood volume within the illuminated area. Data were recorded continuously at 10 Hz and filtered with a 1-s width moving Gaussian smoothing algorithm before analysis. Each measurement was visually and manually checked by a trained evaluator and only valid recordings (with at least > 90% of valid signal after having removed artefacts) were kept for final analysis.
Following a 10-min period of habituation to the treadmill 51 , each participant's preferred walking speed was determined according to a standardized protocol (updated from 52 ): participants were instructed to walk on the treadmill at an initial speed of 1.5 km.h −1 . Speed was progressively increased manually by the investigator in increments of 0.2 .km h −1 every 30 s until subjects reported that they were walking at their preferred walking speed. Then 1 km.h −1 was added to the current speed, followed by a decrease of 0.2 km.h −1 to confirm preferred walking speed. This procedure was repeated until a ± 0.4 km.h −1 agreement was obtained in preferred walking speed, as recommended 52 . The same speed was retained for the post-intervention evaluation.
The DT gait evaluation protocol was divided in 3 consecutive phases (Supplementary Figure S1): (1) Standing phase composed of 5 min of quiet standing immediately followed by 2 different conditions of cognitive assessment performed in ST (cognition only, baseline cognitive performance) and ending with 3 min of rest. The two different cognitive assessments were: a serial-7 test (S-7 test), consisting in subtracting 7 from a random 3-digit number and a Stroop test (performed according to the methodology previously described 10 ). Each cognitive test consisted in 3 consecutive 1-min blocks interspersed by 1-min rest periods. CRR accounted for cognitive performance in ST and in DT; (2) Walking phase composed of 5 min of walk in ST (gait only, baseline gait performance) immediately followed by the 2 cognitive assessments (S-7 test and Stroop test) performed in DT (gait and cognition) and ending with 3 min of walk in ST. In ST gait, participants were instructed to walk according to their natural pattern, arms moving freely by their sides, while looking straight-ahead at a fixed red sight displayed on the screen. In DT, subjects were asked to walk as naturally as possible and to perform the S-7 test or Stroop test at the best of their capacity without any task prioritization; (3) Recovery phase consisting in 5 min of quiet standing. The DT gait evaluation protocol is further illustrated in Supplementary Figure S1.
Data and statistical analysis. Study design and data are reported in accordance with the Consolidated Standards of Reporting Trials (CONSORT) criteria 53 . Patients were randomized by an independent statistician according to a computer-generated list following a 1:1 ratio. Data were analyzed in complete case analysis. Except where clearly stated, results are presented as Median [Quartile 1 (Q1); Quartile 3 (Q3)]. All data have been tested for normality prior to further analysis. For non-normally distributed data, a logarithmic transformation has been applied. Baseline data were compared (CPAP vs. sham-CPAP) by a non-parametric Mann-Whitney test for continuous variables, and χ 2 test or Fisher exact test for categorical variables. For the analysis of data evolution between baseline (pre) and eight weeks (post-intervention), a linear mixed effect model was used with a patient random effect. Group effects (CPAP vs. sham-CPAP) and period effects (pre vs. post CPAP/sham-CPAP) were added as fixed effects. Interaction terms (Group * Period) were tested for all variables, but no significant effect was observed. Therefore, interaction terms were not included in the final model. For complete information of the readers, results of the linear mixed effect model with interaction terms are displayed in Supplementary Tables S4  and S5. Deltas (Pre-Post intervention in Sham-CPAP and CPAP group) for primary outcome were also calculated and between group comparisons using independent sample t-tests were performed. Results are displayed in Supplementary Table S6. A p value < 0.05 was considered significant. Sample size calculation was based on a previous study from our group 10 . The sample size needed to observe a significant difference in STV of 0.8, with a standard deviation of 0.6, a power of 80% and an α value of 0.05 was 10 patients per arm. Moreover, we expected that approximatively 20% of OSAS patients would not be compliant with CPAP therapy. Consequently, we estimated that 12 patients in each group would be sufficient to show differences before and after effective or sham-CPAP therapy. Statistical analyses were performed using SAS (V.9.4, SAS Institute, Cary, NC, USA).

Data availability
Study protocol as well as data collected for the study, including deidentified individual participant data will be made available to others following the publication of this article, for academic purposes (e.g. meta-analyzes) on request to the corresponding author.