INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sleep apnea syndrome (SAS) has a high prevalence estimated as 4% in males and 2% in females aged 30 to 60 yr (1). SAS is associated with increased morbidity and probably with significant mortality. Treatment of SAS improves patient excessive daytime sleepiness and their survival (2). Diagnosis assessment has been based on standard polysomnography (PSG) during a night in a laboratory (LabPSG) (2, 3). However it is labor intensive and expensive. Long waiting lists in sleep laboratories contribute to lack of diagnosis and untreated SAS. It would be more practical, less expensive, and less time consuming to carry out these studies under home or unattended conditions (4). Cardiorespiratory studies may be an acceptable alternative to LabPSG for patients in a high pretest-probability stratification group (based on clinical history and examination) (2) and some devices have been validated at home under unattended conditions (4).

Home complete PSG (HoPSG) could represent an alternative to LabPSG: it allows identification of sleep stages and calculation of total sleep time (TST), and therefore permits accurate assessment of the respiratory disturbance index (RDI). Technological progress and volume reduction of portable devices make it possible to carry out unattended home recordings; their use has considerably increased. Ambulatory PSG devices have been validated under laboratory conditions (10, 11). With the exception of the study by Fry and colleagues (12), few PSG data are available under actual home conditions (13). Questions remain concerning the feasibility and quality of data obtained under realistic home conditions. Furthermore, during home night studies, patients sleep in their normal environment and thus should have a better quality of sleep. Reliability of SAS diagnosis by HoPSG remains to be evaluated. Data concerning the acceptance and tolerance of HoPSG by patients must be collected and considered in the development of this method. HoPSG should be less expensive than LabPSG because it avoids a night stay in hospital and does not require the attendance of nocturnal technicians.

The aim of this study was to evaluate, under realistic conditions, HoPSG compared with LabPSG based on these criteria.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

A total of 333 new patients was referred to the sleep laboratory of the Pneumology Department (Centre Hospitalo-Universitaire de Rouen, Rouen, France) for possible SAS. Two hundred thirty of these patients (69%) were not included in the study: 49 patients (14.7%) were living too far away, 63 others (18.9%) were judged unable to undergo HoPSG, i.e., were not able to give informed consent or had a disability that prevented their cooperation; 118 patients (35.4%) refused to take part in the study. One hundred three patients (31%) gave their informed consent and were included in the study (Table 1).

Home and Laboratory Polysomnographies

All patients were allocated to two PSGs: at home and in the laboratory in a random fashion. The mean delay between the two studies was 14 ± 18 d.

At home, the recording device was portable, weighing 660 g, and was able to collect and store 8 h of data from 10 to 18 channels (Minisomno; Mallinkrodt, Les Ulis Courtaboeuf, France) (home studied device polysomnography, HoSD-PSG). Patients came to the laboratory for about 1 h in the late afternoon for device setting, slept at home, and returned to the laboratory the next morning. Before initiating the study, we tested the ambulatory recording system under home conditions in order to adjust sensor fixation, make recommendations to patients, and deal with other practical details. In the laboratory, the Respisomnographe (Mallinkrodt) monitor was used. Patients were hospitalized under normal conditions during a night and monitored from 10:30 P.M. to 6:00 A.M. Both devices recorded two electroencephalogram (EEG) derivations (C₃-A₂ and C₄-A₁), one chin electromyogram (EMG), two electrooculograms (EOGs), air flow (mouth and nose thermistors), thoracic and abdominal movements (strain gauges), and oximetry (finger probe). Sensors were fixed by the exact same procedure. The Minisomno was the ambulatory version of the Respisomnographe.

Data Quality Evaluation

Raw data were analyzed with the same software and were reviewed by trained specialists in both neurophysiology and pneumology. To evaluate data quality, two criteria were considered for each night study: percentage of total recording time (TRT) with poor airflow signal and neurophysiological data quality, classified as sleep staging impossible (loss of all channels) or as very poor, poor, correct, or good quality data (i.e., one, two, three, or four channels available, respectively).

All studies that fulfilled at least one of the following criteria were judged as unreliable and excluded from further analysis:

 More than 70% of the data lost in the recording because of computer errors: hardware (improper study setup or initiation, storage error) or software malfunction

 More than 80% of TRT with a poor airflow signal

 Sleep staging impossible

 Insufficient total sleep time (TST), defined as TST less than or equal to 180 min

HoSD-PSG and LabPSG Evaluation for SAS Diagnosis

Apnea and hypopnea were defined as a reduction in airflow amplitude of 75 to 100% and of 25 to 50%, respectively, for at least 10 s. Sa_O2 and thoracoabdominal movements were simultaneously analyzed. Sleep staging was performed according to Rechtschaffen and Kales criteria (17). Arousals of more than 10 s were scored. SAS was defined in patients with a respiratory disturbance index (RDI)  15, and severe SAS as an RDI  30. Agreement between HoSD-PSG and LabPSG for RDI measurement was evaluated according to the Bland and Altman method (18). The RDI difference between the two methods was considered as significantly discordant according to two criteria: an RDI difference of more than ± 5 or more than ± 10 events per hour of sleep. It was also considered if the RDI values obtained by the two methods were in discrepancy when compared with the normal threshold (15 events per hour of sleep).

Patient Perception Evaluation

In the morning, patients completed a self-questionnaire about their perception of each night study.

Sleep quality with both devices was evaluated according to time spent in bed (hours), sleep time (hours), quality of sleep (using a visual analog scale from 0 ["very good"] to 10 ["very bad"]), and number of perceived awakenings. Sleep was considered as same, better, or worse than usual. Last, patients were asked about their recording device tolerance and preference.

Other Statistical Methods

Quantitative data measured in the same patients were compared with a paired Student t test (or Kruskall-Wallis nonparametric test when applied) and qualitative data were compared with a paired ² test.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Quality of Polysomnographic Data

Of 103 patients, 21 (20%) were excluded from analysis because of poor quality data for HoSD-PSG and 5 (5%) for LabPSG, including 1 patient for both types of PSG (p < 0.001). Data lost because of computer errors occurred in six patients (6%) with HoSD-PSG and none with LabPSG. A poor airflow signal was observed in 11 (11%) patients with HoSD-PSG versus 2 (2%) patients with LabPSG, including 1 patient for both types of PSG (p < 0.01). Sleep staging was not possible for one patient (1%) with HoSD-PSG and none with LabPSG. Insufficient TST was observed in three patients (3%) with HoSD-PSG and in three other patients (3%) with LabPSG.

HoSD-PSG and LabPSG Evaluation

In 78 patients (76%) both HoSD-PSG and LabPSG data were available for analysis. Their characteristics did not differ from those of the total sample.

Sleep data. TRT and TST were longer at home than in the laboratory, but sleep architecture was similar at home and in the laboratory (Table 2). Sleep data did not differ between the first and second nights, whatever the method used. Neurophysiological data quality was quantified as correct or good (i.e., three or four channels available) in 59 patients (76%) with HoSD-PSG versus 65 patients (83%) with LabPSG (p = 0.09), including 56 patients with data of correct or good quality by both methods.

SAS diagnosis. Using LabPSG, the mean (± SD) RDI was 25.7 (± 30.6) versus 22.8 (± 31.5) using HoSD-PSG. Using LabPSG, 37 patients (47%) were considered as having SAS, as were 31 patients (40%) with HoSD-PSG, including 30 patients (38%) by both methods (p < 0.05). The agreement in RDI measurement between LabPSG and HoSD-PSG is shown in Figure 1. The mean of RDI differences between HoSD-PSG and LabPSG (mean, 2.9; 95% confidence interval, 30 to 36) was not different from 0 (p = 0.13). In 45 patients (57%), the RDI difference between the two types of PSG was more than ± 5 events per hour of sleep (in 27 patients (35%), when considering a difference greater than ± 10 events per hour). Eight of the 45 patients had a different diagnosis status obtained by HoSD-PSG or by LabPSG (eight patients when considering a difference greater than 10). Moreover, discordant results were associated with severe SAS: discordant RDI (more than ± 10 events per hour of sleep difference) was observed in 12 of 54 patients (22%) with normal RDI or moderate SAS, versus 15 of 24 patients (63%) with severe SAS (p < 0.001). Considering an RDI difference of more than ± 5 events per hour of sleep as discordant, 27 of 54 patients with normal or moderate SAS (50%) had discordant RDI, and 18 of 24 patients with severe SAS (75%) had discordant RDI (p < 0.02). Moreover, higher RDI differences were associated with poor airflow signal at home: in 19 patients with more than 20% (and less than 80%) of recording time with poor airflow signal at home, RDI differences were higher (mean, 9.8; 95% confidence interval, 40 to 20.4) than in 59 patients (mean, 0.7; 95% confidence interval, 19.6 to 32.6) with less than 20% of recording time with poor airflow signal at home (p < 0.04).

View larger version (39K): [in this window] [in a new window]   Figure 1.   Comparison of respiratory disturbance index (RDI) measured at home (HoRDI) and in the laboratory (LabRDI). The dotted lines represent the mean bias of the individual differences and the limits of agreement (mean bias ± 2 SD of the differences). In the shaded area are the patients with an absolute 2-RDI difference less than 10 events per hour.

Patient Perception Evaluation

Sleep quality (see Table 3). Patients spent more time in bed at home than in the laboratory. Quality of sleep and the number of nocturnal awakenings did not differ between the two methods.

Recording tolerance. Fifty-three patients (68%) felt uncomfortable with HoSD-PSG versus 54 (69%) with LabPSG, including 42 patients by both methods. The disturbance was mainly due to the sensors (54% with LabPSG and 57% with HoSD-PSG).

Patient's preference. Thirty-seven patients (48%) preferred LabPSG because the device was less cumbersome (38%), they felt less anxious and in safer conditions (30%), or they slept better in the laboratory (22%). Twenty-two other patients (28%) preferred HoSD-PSG because they slept better (50%) or they were in their normal environment (36%). Last, 19 patients (24%) had no preference.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Feasibility

HoSD-PSG was not considered an appropriate method for one-third of our new patients either because of disability or because of difficulties related to transportation. There are obvious limitations in performing HoPSG for patients with a disability that prevents their cooperation.This percentage depends on the social level of the population recruited in a sleep center. The other factor is the population density surrounding the sleep center. In our experience, patients living more than 40 km from the hospital refused two return trips and preferred to spend a night in the laboratory. In the study by Fry and colleagues, 26% of preselected patients did not have HoPSG either because of disability (18%) or transportation difficulties (8%) (12). Our results suggest that it would be important for a sleep center to take this percentage into account when an SAS diagnosis procedure including HoPSG is chosen.

Data Quality

Our results emphasize a rather high failure rate (20% at home compared with 5% in the laboratory) and difficulties in obtaining technically accurate home night studies. In our opinion, prospective evaluation of data quality is most important to validate unattended recording devices. Poor quality or lost data lead to repeated night studies and increasing cost, and may lead to false or inaccurate results. Discordant results have been observed in our patients with poor airflow signal at home (11%). This was because our patients had to set up the thermistors themselves after dinner, some did not adjust the device as recommended, or others damaged the equipment. Therefore, sensors must be designed to be solid, easy to handle, and suited to home studies. For simplified cardiorespiratory portable recorders tested at home, data loss was evaluated as 4 to 24% (5, 9, 19). There are few reported studies that have evaluated HoPSG data quality; Pelletier-Fleury and colleagues, using the same equipment as ours (Minisomno), lost 23.4% of their night studies at home (22). However, Fry and coworkers did not lose any HoPSG recordings and each parameter was scorable in more than 95% of all epochs (12). Quality and strength of the sensors and their fixation, and quality of the education and information given to patients are factors that probably explain the difference. For Ninane and colleagues, data quality could be improved if the technician set up the device at the patient's home (14). However, this method is time consuming for the technician, particularly if patients live far away, which increases the cost of home studies. Comparison data between these two methods (at home or in a sleep center set-up) are not yet available (23).

Reliability of batteries, software, and hardware is also an important consideration in the selection of the ambulatory device, to reduce lost data occurrence.

HoSD-PSG and LabPSG Evaluation for SAS Diagnosis

Sleep data. TRT and TST were longer at home because our patients had the opportunity to adapt their home recording session to their sleeping habits, as they woke up earlier in the laboratory. Despite this difference in TST, we did not observe any difference in sleep architecture and in subjective evaluation of sleep quality at home compared with the laboratory setting. This could be explained by the fact that most apneic patients have the ability to fall asleep easily anywhere, and their sleep architecture disorders are much more influenced by their respiratory disorders than by their sleep environment. The fact that sleep architecture and efficiency could be modified by the unfamiliar surroundings of a sleep laboratory has not been proven. Fry and colleagues observed a longer TRT and TST and a better subjective evaluation of sleep in the laboratory (12). Whether LabPSG will continue to be the gold standard still remains controversial.

SAS diagnosis. A positive agreement between HoRDI and LabRDI was observed, particularly for nonapneic patients and patients with moderate SAS. Greater variability of RDI was found in 35% of patients, primarily in cases of severe SAS and poor quality data. However, only 8 of 78 patients were considered normal by one method and abnormal by the other. On the basis of the same criteria, Bliwise and coworkers found 18% of patients with a high variability, i.e., two-night RDI difference greater than 10 events per hour (24). In our study, discrepancies observed between HoRDI and LabRDI could be explained either by differences in the two devices or differences in recording conditions (home versus laboratory), or by night-to-night respiratory abnormality variations.

Differences between the two monitors were kept to a minimum because we used similar devices, sensors, and software produced by the same company. Also, all raw data were assessed by the same physicians.

The main differences were due to recording conditions: home versus laboratory and attended versus unattended. It is obvious that home conditions were not standardized and controlled. Patients were advised not to drink alcohol, not to take sleeping tablets, not to sleep with the television on, etc., as these factors could have influenced respiratory abnormalities. Ballester and coworkers found a substantial difference between the behavior position in the sleep laboratory and at home (25). On the other hand, standardized laboratory conditions could be different from a patient's daily habits, and the basic question is probably to determine the respiratory abnormalities in the normal sleep environment.

Also, unattended home recording conditions were associated in our study with poor quality data. Poor airflow signal was associated with higher RDI differences, because the respiratory events count was inaccurate. During attended LabPSG the mean number of technician interventions to adjust the signals was four per night. Pelletier-Fleury and coworkers compared HoSD-PSG with out-laboratory PSG under telesurveillance and were able to improve quality of data by this second technique (22).

Night-to-night variability of respiratory abnormalities has been well documented (9, 24, 26, 27). It raises the question of the validity of a single night study as a diagnostic test, and has an impact on the results of all studies that compare different night studies. Many factors have been involved in nightly variability: differences in posture and body position, medication and alcohol use, nasal congestion, sleep stage composition, and pulmonary status (9, 24).

We did not assess either respiratory effort with microarousals or periodic leg movements, which could be associated with SAS, and could perhaps in part explain night-to-night variability. However, it is impossible to predict the specific role of these factors for each patient. Wittig and coworkers found that patients with mild SAS had a higher level of variability than patients with severe SAS (26). This was not confirmed by other studies (9, 24, 27). We also observed a higher variability for severe SAS. This is less significant for therapeutic behavior, which is generally the same for severe SAS.

In our opinion, the discordance between the two methods that we observed was explained mainly by the poor reliability of the home studied device.

Patient Evaluation

A majority of our patients preferred LabPSG. In their subjective assessment of home versus laboratory PSG, Fry and colleagues reported that their patients also favored attended study. They preferred the equipment and availability of a technician and, contrary to popular belief, perceived better quality of sleep in the laboratory (12). However, simplified cardiorespiratory home studies are probably better accepted by patients than HoPSG studies, which are judged too cumbersome because of the presence of more sensors on the head.

Cost Evaluation

The mean hospital cost of standard LabPSG was evaluated at 307.6 Euros (28). By subtracting the cost of the technician monitoring at night, and the cost of the hospital stay, the cost of HoPSG was 152.5 Euros. This 50% reduction in cost must be considered when considering the 20% equipement failure at home. The financial benefit of HoPSG could be significantly improved if home quality data were reliable.

In conclusion, HoSD-PSG is a useful SAS diagnosic procedure because it is less expensive and could replace LabPSG under selected conditions. HoSD-PSG is not feasible for all patients and its reliability will clearly depend on the quality of data obtained under unattended conditions. There is also no evidence of a better quality of sleep at home or of a better tolerance of this method. Further investigations are needed to define clearly the type of patients who could benefit from HoPSG and the optimal conditions under which to carry it out.