INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: nasal CPAP; neurobehavioral impairment; obstructive sleep apnea

Obstructive sleep apnea (OSA) is a common condition characterized by repetitive obstruction of the upper airway during sleep with resultant episodic hypoxia and arousal. There is a continuous spectrum of severity of sleep-related upper airway dysfunction ranging from simple snoring and upper airway resistance syndrome through mild to severe obstructive hypopnea and apnea. The prevalence of symptomatic OSA in adult males has been estimated to be between 2 and 4% (1, 2). These patients demonstrate behavioral and neuropsychological consequences to varying degrees, including excessive daytime sleepiness (3), psychomotor deficits (4), an increase in motor vehicle accidents (5), and lost work productivity and absenteeism (6). Cardiorespiratory conditions associated with OSA are myocardial infarction (7), pulmonary hypertension (8, 9), systemic hypertension (10, 11), and ventricular arrhythmias (12).

Although many studies have addressed patients with severe OSA, more recent evidence from epidemiologic studies suggests that subjects with even mild sleep breathing abnormalities (apnea-hypopnea index [AHI] < 15) may have associated hypertension (10), neurocognitive deficits (13-15), and increased motor vehicle accidents (5).

Treatment of patients with severe OSA with nasal continuous positive airway pressure (nCPAP) reverses upper airway obstruction during sleep (16) and reduces many of the consequences of severe OSA, most notably daytime sleepiness and psychological symptoms (17-24). The effect of CPAP treatment on these outcomes in patients mildly infected with OSA is less clear. Three randomized controlled trials (13, 14, 25) indicate that there may be an improvement in some areas of neurobehavioral function and self-reported symptoms in this population, but improvements in objective measurements, such as sleep latency in the Multiple Sleep Latency Test (MSLT) and maintenance of wakefulness tests and cardiovascular function, have not been shown. In the study by Redline and coworkers (25), CPAP was compared with conservative treatment, whereas in the studies by Engleman and coworkers (13, 14) a placebo tablet was used as a control.

This study was designed to assess cardiorespiratory and neurobehavioral function in patients with mild OSA (AHI 5-30/h) and to determine treatment response to nCPAP therapy compared with that of a placebo tablet. Treatment outcomes for both CPAP and placebo were compared with baseline measurements.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Design

A randomized, placebo-controlled, cross-over trial was conducted in two Australian centers (Austin and Repatriation Medical Centre, Heidelberg, Victoria and Repatriation General Hospital, Daw Park, South Australia) to investigate daytime sleepiness, neurobehavioral function, and 24-h systemic blood pressure in patients with mild obstructive sleep apnea (OSA) and to assess the response to 8 wk of treatment with nCPAP and a placebo tablet. The Human Research Ethics Committees of the two centers approved the project and written informed consent was obtained from each patient.

Patients recruited into the study were referred for investigation of symptomatic sleep-disordered breathing (snoring, observed breathing pauses in sleep, and daytime sleepiness). They were eligible for inclusion in the study if they were more than 18 yr of age and if their overnight diagnostic sleep study showed an AHI of between 5 and 30/h. Each diagnostic polysomnographic study required at least 4 h of sleep, at least 30 min of sleep in the supine position, and at least 30 min of rapid eye movement (REM) sleep. Patients with minimum blood oxygen saturation less than 75% in REM and 80% in non-REM were excluded, as were patients with clinically significant coexisting disease (e.g., diabetes, unstable ischemic heart disease) or sleepiness deemed to be unsafe and requiring urgent treatment, for example, history of falling asleep while driving or working, or in some other unsafe situation. To further exclude comorbidities, a clinical examination, full blood cell count, electrolytes, renal function, random blood glucose, and liver function tests were performed. To ensure valid interpretation of the neurobehavioral tests patients were required to be fluent in the English language and to have no history of cerebrovascular disease, closed head injury associated with loss of consciousness greater than 15 min in duration, psychiatric illness, or alcohol or drug abuse.

Patients were randomized to receive CPAP (Sullivan Elite; ResMed, Sydney, Australia) for 8 wk and a placebo lactose tablet for 8 wk. Randomization was conducted by picking a piece of paper with a treatment order written on it out of a box, and then that piece of paper was placed back in the box. There was no intervening washout period, as the onset and offset of benefits from CPAP occurs within 1 or 2 d (26). Patients were told that the tablet was intended to improve airway function during sleep and were instructed to take it immediately before going to bed. We chose to use an oral placebo rather than sham CPAP because new patients in our clinics are often familiar with the principles of CPAP treatment and in a cross-over trial may have been aware that they were not receiving therapeutic pressure. Before commencement of CPAP or placebo, and at the end of each 8-wk treatment period, patients underwent the MSLT, 24-h blood pressure measurements, and neurobehavioral assessment and completed the Epworth Sleepiness Scale (ESS) and Quality of Life and Worsnop Symptom Questionnaires. The latter was developed in-house and is given in the Appendix (see online data supplement).

Standardization and Calibration of Measurements between the Two Centers

Before commencement of the project, protocols for overnight polysomnography, the administration of the MSLT, and the administration of the neuropsychological tests were standardized between the two centers. One trained researcher at each institution administered the neurobehavioral tests in a blinded manner and one person at each site performed sleep study scoring. Calibration of sleep study scoring between scorers at the two centers was achieved by the exchange of five overnight studies, and an initial concordance rate of 86% for sleep stages and 100% for respiratory events was achieved. During the course of the study, this calibration process was repeated twice; sleep stage concordance was 94 and 92%, and respiratory event concordance was 100% on each occasion.

Polysomnography

The Sleepwatch (Compumedics, Melbourne, Australia) sleep monitoring system was used to record electroencephalogram activity (standard central leads C₃-A₂/C₄-A₁), left and right electro-oculogram, submental electromyogram, body position (mercury switch), and right and left leg movements (peizoelectric movement sensors). Monitored cardiopulmonary variables were as follows: oxygen saturation (OxiRadometer; Radiometer, Copenhagen, Denmark), airflow with oral and nasal thermistors; electrocardiogram using a Lead I chest lead; thoracic and abdominal movements by respiratory inductance plethysmography; and CPAP mask pressure. Data were recorded and stored on a magnetooptical disc for analysis.

The MSLT was performed according to standard guidelines (27). Polysomnography was not performed on the night immediately preceding the MSLT. Instead, normal sleep-wake schedules in patients were confirmed for the week before MSLT by patients keeping a sleep diary and by Gaehweiler wrist actigraphy readings (activity monitor model Z80-32K V1; Gaehweiler Electronic, Stafa, Switzerland).

Analysis of Sleep Data

All polysomnograms and MSLTs were manually staged according to standard criteria (28) with 30-s epochs. A hypopnea was defined as an event of at least 10 s in duration, with a > 50% reduction from baseline in at least two of the following three signals: airflow, thoracic movement, abdominal movement. An apnea was defined as < 20% airflow lasting at least 10 s and was classified as obstructive, central, or mixed depending on whether there were ongoing respiratory efforts. The AHI was computed by dividing the number of events by the hours of sleep.

Other Measurements

Age, sex, medications, resting blood pressure, height, weight, level of education, and pulmonary function tests were recorded. Systemic blood pressure over 24 h was measured every 20 min by day and hourly overnight, using an Accutracker ambulatory blood pressure device (Accutracker 11/86E7; Suntech Medical Instruments, Raleigh, NC) in Victoria, and the Ambulatory NIBP machine (ambulatory NIBP; Spacelabs Medical Products, Redmond, WA) in South Australia.

Daytime sleepiness was assessed using the MSLT, the ESS (29), and our in-house symptom questionnaire, rating commonly recognized OSA symptoms on a visual analog scale (see Appendix in the online data supplement). The Epworth Sleepiness Scale (29) is an eight-item self-administered questionnaire that asks patients to rate their likelihood of falling asleep in given situations in his/her "usual way of life in recent times." A higher score indicates increased subjective daytime sleepiness.

The neurobehavioral tests used consisted of the Neuropsychological Assessment Battery (NAB) (30) and paper-based tests as shown in Table 1. The complete session required approximately 1 h of patient time. The NAB test was administered three times during the day of the MSLT and a mean score was computed to account for any time of day effects. The other neurobehavioral tests were administered only in the first session. To minimize any learning effect due to repetition of these tests, patients attended a familiarization session 1 wk before their baseline assessment and performed an abbreviated version of the neurobehavioral tests. Quality of life was assessed with two self-administered questionnaires completed by patients on the day of their MSLT. The first was the generic 36-item Short Form Medical Outcomes Survey (SF-36) (31), a general health and quality of life questionnaire. This has eight main domains: physical functioning, role limitation due to physical problems, role limitation due to emotional problems, social functioning, mental health, energy/vitality, bodily pain, and general health perception. The second questionnaire was the Functional Outcomes of Sleep Questionnaire (FOSQ) (32), which has been developed specifically for patients with sleep disorders leading to excessive sleepiness. It has five main domains: activity level, vigilance, general productivity, social outcome, and intimacy and sexual relationships. The SF-36 and FOSQ are both lists of questions that require a response grading the severity of that symptom or condition. In the SF-36, each dimension item score is coded, summed, and transformed onto a scale from 0 to 100 (worst to best possible health). For the FOSQ the dimension items are similarly scored, summed, and then averaged, with a maximum possible score of 4 for each question indicating best possible health. The total summed score was then divided by the maximum possible score, to give an overall FOSQ score with a potential maximum of 1. Thus, a higher score indicates better health in both the FOSQ and the SF-36.

Compliance with both treatments was measured. The CPAP pump had a built-in compliance meter that measured how long (in hours) the CPAP pump was used by the subject each time it was switched on. A "SmartStop" automatically switched off the machine after 7 s of zero pressure, for example, if the patient pulled the mask off his/her face during the night. In the placebo arm, each patient was issued a bottle containing 70 capsules for the 8-wk treatment period. Compliance was estimated by counting the remaining capsules when patients returned them at the end of this treatment period. To maximize compliance, each patient was reviewed in person 2 wk after starting each treatment and then telephoned on two further occasions during the ensuing 6 wk to discuss and resolve problems and to encourage nightly use of the treatments.

Statistics

The Statistical Package for Social Sciences (SPSS, Chicago, IL) database program was used. Continuous variables were expressed as means with standard deviations. For each of the response variables, the data have been analyzed by analysis of variance with allowance made for both subjects and periods. Because each subject had a baseline value, it was possible to test for interaction between period and treatment, using an F-test with 1 and the error degrees of freedom. In a small number of cases, the test for interaction was significant, which indicates the existence of a carryover effect. Given the subjective nature of many of the response variables, it was felt that the carryover effect was more psychological than physiological. When there was a significant interaction between period and treatment, changes between baseline, CPAP, and placebo values are listed separately for those who had CPAP before placebo and those who had placebo before CPAP. When residuals were found to have affected normality, a square root transformation was performed (as the mean was proportional to the variance). Multiple regression analysis was used to assess the relationship between variables. A p value of < 0.05 was taken as significant. When interpreting the SF-36 quality of life results, the changes from baseline with CPAP and placebo are also quoted as "effect sizes" (33). This identifies changes that are clinically significant rather than statistically significant; however, in that larger effect sizes mean larger differences, they will tend to correspond to smaller p values in a test of equality of means. An effect size of 0.20 is considered small, 0.50 is considered moderate, and 0.80 is considered large. Clinical interpretation of effect sizes should also take into account the mean values and standard deviations. Power and sample size calculations were done as described by Bach and Sharpe (34). The number of subjects required for a study with paired data, power of 90%, and  = 1% is 17 subjects, so our recruitment should have been more than adequate. Post-hoc power calculations were performed by using the lowest difference of clinical significance and the standard deviation of this difference. The power to detect a significant difference at the 5% level for each of the major variables was then calculated. The only variables that failed to reach a power of at least 98% were psychomotor vigilance task (PVT) 1/RT (reaction time) for slowest 10% of responses (power of 76%), Profile of Moods States total mood disorder (25%), MSLT sleep latency (76%), and Stroop interference score (76%). We would have required 100 subjects for the chance of a Type 11 error in these variables to be less than 10% (a power of 90% for a test with significance level of 0.05).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subject Selection and Retention

Of 52 patients approached to participate in the study, 42 agreed: 19 from Victoria and 23 from South Australia. Refusals were due to an inability to attend for testing because of work (8), family commitments (1), or for unspecified reasons (1). During the course of the trial, 14 patients dropped out. This was due to work commitments (6), intolerance of CPAP (5), unrelated surgery (1), subsequent diagnosis of periodic limb movement syndrome (1), and loss of interest (1).

Of the 42 subjects who agreed to be involved in the study, 5 completed the familiarization session only, 5 completed baseline assessment only, 4 completed baseline and placebo assessments, and 28 completed both treatment arms.

Clinical Features of All Enrolled Study Patients

Anthropometric data and baseline polysomnographic indices of the patients are given in Table E1 (see online data supplement).

Of the 42 patients enrolled into the study, 35 were male and 7 were female. In general, they were middle-aged and overweight. Analysis of baseline polysomnographic data showed a mean AHI of 12.9/h and a mean arousal index of 19.0/h. After CPAP implementation, the mean (SD) AHI was 4.24 (2.9), and 72% of patients had an AHI < 5.0 (baseline AHI versus CPAP AHI, p < 0.001). There was no significant difference in any baseline parameters between those subjects who did not complete the study and those who completed it (Table E2, see online data supplement), apart from the treatment order; patients were more likely to drop out if they received CPAP in the second treatment arm (p < 0.05).

Daytime Sleepiness and Symptoms

Results for the assessment of sleepiness at baseline and after each treatment period are given in Table 2. On study entry, mean ESS score was 11.2 (± 5.0) and the mean MSLT sleep latency was 12.5 min (± 4.8). Twenty-four patients (65%) had an ESS score  10, suggesting significant subjective daytime sleepiness (29). Multiple regression analysis failed to show any relationship between baseline MSLT sleep latency score and baseline AHI or arousal index, or between ESS score and baseline AHI or arousal index. Thus, the commonly used polysomnographic indicators of the severity of sleep apnea were not significantly related to either subjective or objective assessments of daytime sleepiness. Ten patients had a mean MSLT sleep latency < 10 min, 9 of whom also had an ESS score  10, and there was a significant correlation between MSLT and ESS results for our study population (R = 0.5, p = 0.001; Figure 1). A significant improvement was seen in the ESS score with both CPAP and placebo, but there was no significant difference between the two treatment effects (Table 2). When the period effect was assessed, those subjects who received CPAP first, and then placebo, showed a significant improvement with both treatments, whereas those who used the placebo first did not show a significant improvement with either. Thus, there was a significant interaction between treatment and period (F = 7.442), indicating a possible carryover effect of the CPAP into the placebo period, but not from placebo into the CPAP period. The MSLT sleep latency showed no significant improvement with either CPAP or placebo. Our in-house symptom questionnaire (see Appendix in online data supplement) showed that CPAP was significantly more effective in improving the overall symptoms of subjects than was placebo (p = 0.04). We further analyzed this by grouping the questions into four main areas looking at sleepiness (questions 3, 7, 9, 11, 12, and 14), sleep fragmentation (questions 1, 2, 4, and 8), personality change (questions 6 and 13), and morning confusion/headache (questions 5 and 10). There were marked CPAP effects on sleepiness that were not seen with placebo, and the difference in response reached statistical significance in the areas of sleepiness and morning confusion/headache.

View larger version (17K): [in this window] [in a new window]   Figure 1.   Relationship between Epworth Sleepiness Scale and Multiple Sleep Latency Test scores. R = 0.5, p = 0.001.

Treatment Effects on Neuropsychological Function and Mood

Treatment effects of CPAP on neuropsychological function were not universal or consistent (Table 3). There was a significant improvement after CPAP in verbal fluency (Controlled Oral Word Association Test, COWAT) and vigilance (Psychomotor Vigilance Task 1/RT for slowest 10% of responses), but only for the COWAT was the CPAP effect significantly different from the placebo effect. When the subjects were analyzed for period effects, it was found that the placebo improvement occurred only in those subjects who received CPAP first. Those who used the placebo first, followed by CPAP, had no significant change with either treatment. There was a statistically significant improvement in short-term memory (Word Pair Memory Recall) and Digit Symbol Substitution Test with placebo, but these were not significantly different from the CPAP effect. The Trail Making B Test improved with both CPAP and placebo, but neither was significantly better than baseline.

The Profile of Moods States (POMS) test (Total Mood Disorders) showed a significant improvement with both CPAP but not with placebo therapy, and there was no significant difference between the two. When the score was broken down into the component categories (tension-anxiety, depression-dejection [DD], anger-hostility, vigor-activity [V], fatigue-inertia [F], and confusion-bewilderment [CB]), CPAP improved DD, V, F, and CB and placebo improved DD and CB, but there was no difference in any of the components between CPAP and placebo. Of the 37 patients who had baseline assessments, 6 patients (16%) had a baseline Beck Depression Index (BDI) score that suggested clinical depression that was in the mild- moderate or severe range. Five of these patients completed the study, and one dropped out before any treatment could be commenced. Of the five who completed the study, the BDI returned to within the normal range in three subjects (60%) after both CPAP and placebo; in the remaining two patients, there was no improvement with either treatment modality.

Quality of Life

Results from the FOSQ (Table 4) show that there was significant baseline morbidity (32). CPAP improved all areas except that of intimate relationships and sexual activity; a placebo effect was seen in the same areas, apart from general productivity. The difference between CPAP and placebo effects was significant at the 0.06 level for vigilance, and there was no statistically significant difference in any other domain.

The SF-36 questionnaire revealed a significant placebo effect on patient perception of physical functioning, role limitation due to emotional problems, mental health, and energy/ vitality (Table 5). CPAP also had a widespread effect, with improvement in three of the eight domains (social functioning, mental health, and energy/vitality), but in no area was there a significant difference between the improvement seen with CPAP and that with placebo. Effect sizes confirm the results as being clinically important in the dimension of energy/vitality with both CPAP and placebo; role limitation due to emotional problems improved with placebo but not with CPAP.

Blood Pressure

Baseline and posttreatment blood pressure data are given in Table E3 (see online data supplement).

Seven of 24 patients (25%) were hypertensive at baseline, with either a mean 24-h systolic blood pressure > 140 mm Hg, or a mean 24-h diastolic blood pressure > 90 mm Hg. Of these, four became normotensive after CPAP and two were normotensive after both CPAP and placebo. In the group data, there were no significant changes in any of the indices derived from 24-h blood pressure recordings after CPAP or placebo compared with baseline and no differences between these treatments.

Compliance and Patient Treatment Preference

Use of placebo capsules was 93%. CPAP compliance data were not available for 5 patients because of failed recording of machine "on time"; in the remaining 23 patients, mean nightly CPAP use (including nights when the machine was not switched on at all) was 3.53 (± 2.13) h. CPAP compliance was found to be bimodal (Figure 2): one group of 12 subjects used the pump for less than 4 h per night on average (mean, 1.72 h/night), the other 11 patients used it for more than 4 h per night (mean, 5.5 h/night). Indicators of disease severity measured at baseline, such as AHI, arousal index, ESS score, MSLT result, and symptom score, did not predict a higher use of CPAP (Table E4; see online data supplement). There was no statistically significant difference in the number of hours CPAP was used by those with high versus low AHI, arousal index, ESS score, or MSLT sleep latency. A separate statistical analysis of the 11 patients who used CPAP for greater than 4 h per night failed to reveal any new differences in treatment effectiveness between CPAP and placebo. There was a tendency for those with better compliance to be sleepier at baseline (high ESS score, 12.6 versus 10.4) and short MSLT sleep latency (10 versus 13), but these differences failed to reach statistical significance.

View larger version (48K): [in this window] [in a new window]   Figure 2.   Use of CPAP. Median = 3.14, n = 23.

Of the 28 patients who completed the study, 16 stated that they preferred placebo and 12 preferred CPAP.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Although there is a clear role for nCPAP in the treatment of severe sleep apnea, this is still the subject of debate for patients with mild to moderate disease. We have studied 42 patients with mild to moderate OSA (AHI, 5-30/h) in a cross-over design randomized placebo-controlled trial (RCT). The outcome variables were blood pressure and neurobehavioral morbidity, measured at baseline and after 8 wk of treatment with CPAP and oral placebo tablet. Nasal CPAP improved self-reported symptoms of disturbed sleep quality; however, it was no better than placebo for improving measures of objective or subjective daytime sleepiness. Only for verbal fluency (COWAT) was there a positive treatment effect of CPAP over placebo in tests of neurobehavioral function; there was no benefit of CPAP over placebo in generic (SF-36) or sleep-specific (FOSQ) quality of life questionnaires (although vigilance was almost significant), mood score (POMS and BDI), or 24-h blood pressure.

There are now many reports of randomized controlled studies of CPAP in patients with OSA of varying severity (13, 14, 17- 21, 23-25). The three most relevant to the current study are those of Engleman and coworkers (13, 14), studies of 16 patients and of 34 patients with mild OSA (AHI, 5-14.9 plus 2 symptoms), and that of Redline and coworkers (25), a study of 97 patients with mild OSA (AHI, 5-30). These studies measured objective and subjective sleepiness, cognitive function, and psychological well-being. The cognitive function data from the study of Redline and coworkers (25) have yet to be reported.

There has been one randomized controlled study from the Edinburgh group (23) of the blood pressure response to CPAP in patients with a wide range of sleep-disordered breathing. We found there to be no group change with CPAP in day or nighttime systolic or diastolic blood pressure; Faccenda and coworkers (23) found a small but significant drop in diastolic blood pressure with CPAP, and the effect of CPAP was increased in the subgroup that had high CPAP use ( 3.5 h per night) and in those who had more than twenty 4% oxygen desaturations per hour.

Subjective sleepiness (ESS) improved with both CPAP and placebo in our study, although this improvement was only in those subjects who had CPAP first and there was a significant interaction. ESS improved with CPAP alone in the study by Redline and coworkers (25) and the second Edinburgh study (13). The earlier study by Engleman and coworkers (14) showed no treatment response in subjective sleepiness, and none of the four studies of mild OSA have demonstrated a response in objective sleepiness (MSLT or Maintenance of Wakefulness Test) to CPAP or placebo. Studies that have shown improvement in objective measures of sleepiness with CPAP have been in patients with more severe OSA (17, 18, 21, 22). Both the Edinburgh studies and our own used questionnaires that assessed general symptoms of OSA (e.g., snoring, breathing pauses, nighttime choking arousals, and daytime sleepiness) and showed a significant benefit of CPAP over placebo treatment.

The only cognitive tests we found to be improved after CPAP were verbal fluency and vigilance, but there was no difference between CPAP and placebo in the vigilance task. Engleman and coworkers found an improvement with CPAP compared with placebo in the Digit Symbol Substitution Test in one group of patients with mild OSA (13) but not in the other (14). We found some improvement with placebo in this test, but the difference was significant at a level of 0.07 only. Engleman and coworkers also documented an improvement with CPAP in two other measures of complex cognitive function, Trails B (14) and the Paced Auditory Serial Addition Test (2-s rate) (13).

Psychological well-being was improved by CPAP in the previous studies of mild OSA, as measured by the Hospital Anxiety Depression Scale (13, 14) and in several subscales of the SF-36 (13, 25). We found that initial BDI scores indicated that 16% of our patients were depressed, and 56% had a total mood disorder score above the predicted range in the POMS. Both measures improved with both treatments and there was no significant difference between CPAP and placebo, although there was potentially a Type 11 error here.

The FOSQ data showed that many (72%) of our patients had an impaired quality of life before treatment; there were significant improvements with both CPAP and placebo in most areas of the FOSQ and of the SF-36. The only dimension where there was a trend to a significant difference in quality of life between the placebo and CPAP effects was in the FOSQ domain of vigilance. Engleman and coworkers (14) found an improvement (CPAP versus placebo) in the Nottingham Health Profile Part 2 only among patients with mild OSA described as "high compliers" in their earlier study and improvements in five of the SF-36 domains in the later study (13). Redline and coworkers (25) found an improvement with CPAP in the energy-fatigue subscale of the SF-36, an area in which the Engleman group also found a CPAP effect.

Thus, our study is similar to previous RCTs of CPAP in patients with mild OSA, in that we were able to demonstrate an advantage for CPAP over a placebo in general symptoms of OSA (including snoring, choking arousals, breathing pauses, and daytime sleepiness). As in the previous studies, we were unable to demonstrate any treatment effect of CPAP on an objective measurement of sleepiness. However, our study differs from these other studies in that we did not find any treatment advantage of CPAP versus placebo in the trait of subjective sleepiness (ESS), in all but one neuropsychological assessment, or in any of the quality of life measures. The reasons for these differences are not clear. One possibility is different patient selection. In the present study, we included any patient referred to our laboratories who was found to have an AHI between 5 and 30/h. Although they had symptoms generally acknowledged to be those of the OSA syndrome, a threshold level of symptom severity was not required. The Edinburgh studies of mild OSA, which were in some respects similar to the current study (cross-over study design, placebo capsule, and similar sample sizes), required patients to have two or more symptoms and an ESS of 8 or more. It is possible therefore that patients in these studies may have been more neurobehaviorally impaired for the same or similar levels of AHI severity. In fact, the Edinburgh mean baseline AHI values were slightly lower (11/h [14] and 10/h [13] compared with 13/h for the present study) and mean baseline ESS values were higher (13 in the 1999 study [13] and no baseline assessment done in the 1997 study [14], compared with 11.2 in the current study).

A second difference between the current and earlier studies of mild OSA is the sex ratio of subjects. Both Edinburgh trials (13, 14) and the Redline trial (25) had relatively large numbers of females (female-to-male ratios of 1:1.6, 1:3, and 1:1.1, respectively) whereas the current trial had a female-to-male ratio of 1:5, which is closer to that reported in other RCTs of CPAP in OSA (19-21, 23). It is not clear whether there is any sex bias in performing tests of neurobehavioral function, although several are standardized by sex. Different CPAP compliance rates would not appear to explain the difference between our CPAP response and those reported previously. CPAP use in this study (mean, 3.5 h per night) is similar to that across a number of other studies of patients with mild to moderate OSA (13, 14, 25). Although two studies have demonstrated a better effect on psychological well-being among good CPAP compliers (14, 18, 22), we found no such effect, and we did not find that baseline indicators of disease severity (AHI, arousal index, objective and subjective sleepiness, symptoms) were predictive of greater CPAP use.

A major finding of this study was that an oral placebo tablet conferred a significant treatment benefit in a wide range of functional variables. Patients were told (with the approval of our Ethics Committees) that the tablet had been developed to improve their breathing at night. Statistically significant placebo effects were observed in subjective daytime sleepiness (ESS), memory and cognition, four of eight domains in the SF-36 assessment of quality of life, and three of five domains in the FOSQ. Baseline measures of sleepiness and cognitive function have been performed in most of the previous RCTs of CPAP in OSA, but until now the statistical significance of the placebo response has been reported in only four studies. Engleman and coworkers (18), Jenkinson and coworkers (21), and Montserrat and coworkers (24) reported these data in patients with moderate to severe sleep apnea, and Redline and coworkers (25) reported a response in the Epworth score and MSLT sleep latency to a nasal dilator strip. There has been no comprehensive analysis of the extent of the placebo effect on neurobehavioral outcomes and psychological well-being in mild sleep apnea. In our study population, positive placebo effects were widespread. They probably explain the failure to demonstrate more than just one or two benefits of CPAP over placebo in the present study. It is possible that some of the improvements observed clinically with CPAP treatment in patients with mild disease could be due to placebo effects.

There are a number of other possible reasons for the failure of CPAP treatment to confer more benefit than placebo. The first to consider is whether the prescribed CPAP pressure or CPAP use was insufficient to enable a treatment effect to be observed. It is unlikely that prescribed pressure was inadequate because median (SD) AHI on treatment was 3.65 (2.9) events/h, a result similar to other reports of CPAP treatment in mild to moderate OSA. In the CPAP titration study, the AHI was below 5 in 80% of patients and only one patient had an AHI above 10 (14.5). Average nightly CPAP use in the 23 patients for whom data were available was 3.5 h/night, and suboptimal use may therefore be a possible reason for the lack of treatment response. However, this use compares favorably with average CPAP compliance in the other RCTs of CPAP in mild to moderate OSA (13, 14, 25) (3.1, 3.2, and 2.8 h, respectively), although it is not as good as the CPAP compliance of 4.5 h/night (24), 5.9 h/night (18), and 5.2 h/night (20) reported for severe OSA. Improving CPAP acceptance and use in patients with mild OSA is likely to be difficult (35, 36), but could result in improved treatment outcomes. However, in this study we found no evidence of improved outcomes among those with greater CPAP use (> 4 h per night), and commonly used indicators of OSA severity (AHI, arousal index, objective and subjective sleepiness, symptoms) did not predict better compliance.

Another possible reason for the failure of CPAP to confer a greater benefit than placebo in this population would be if there was little morbidity. This seems an unlikely explanation. Despite the mildness of their sleep apnea as defined by AHI, many of our subjects had significant although mild baseline morbidity. Pathological sleepiness (MSLT latency < 10 min or ESS score  10) was seen in 60% of study patients, 16% had clinically significant BDI depression scores, 25% were hypertensive, and 72% had an abnormal overall score on the FOSQ quality of life assessment (32). Only 5 of 37 patients tested at baseline had all their neurobehavioral test results within the normal range, suggesting significant neurobehavioral morbidity in our population. The measured abnormalities could, of course, have an etiology separate from their sleep apnea, in which case we would not expect CPAP to be better than placebo in treating them. If daytime sleepiness, for example, were due to chronic sleep restriction (which is common in the general population) it would not respond to CPAP. Thus, it remains possible that low patient numbers, inadequate CPAP use, the overall mild degree of neurobehavioral impairment in many patients, or morbidity unassociated with OSA may be responsible for the failure to detect positive treatment effects of CPAP over placebo in many of the variables measured.

Two potential methodological limitations of the study should be acknowledged. The first is the high rate of withdrawal of patients during the course of the study. In part this was due to the time commitment required of subjects, most of whom were in full-time employment. However, intolerance of CPAP was also a significant cause of failure to complete the study, and previous studies (35-39) indicate that intolerance of CPAP is likely to result in early rather than late withdrawal from treatment. We consider it unlikely that the subjects who failed to complete the study biased the results in favor of a negative result in that they did not demonstrate significantly more severe sleep-disordered breathing or daytime sleepiness at baseline. A second methodological factor is the choice of placebo. In theory, a nasal mask with CPAP pump pressure set at zero or at a subtherapeutic level would seem to be an appropriate placebo. This has been used in two previous parallel design RCTs of CPAP in patients with more severe sleep apnea (21, 40) and in two partial cross-over studies of patients with moderate to severe sleep apnea (24, 41). However, knowledge of the principles of CPAP treatment is common among patients presenting to our clinics. We considered it therefore inappropriate to use a mask at almost zero pressure in a cross-over study, as this would be obvious to many patients and compromise the placebo effect. Another group has employed the positive placebo (capsule) approach that we used in RCTs of CPAP versus placebo in OSA (13, 14, 17, 19), allowing direct comparison with their results. Further, this study has been instructive in demonstrating the susceptibility of patients with mild OSA to placebo treatment effects.

In conclusion, we have studied a group of patients with mild to moderate OSA who were given nCPAP and placebo treatment for 2-mo periods. Despite adequate treatment, as documented by a fall in AHI and compliance comparable to previous studies, we found benefit from CPAP therapy over placebo only in the areas of verbal fluency and global symptom score, with a trend seen in one domain of the FOSQ (vigilance). An important finding was that there are large placebo effects for subjective sleepiness, mood, and quality of life, underscoring the need for adequate control comparison and baseline assessments in patients with mild disease in future studies of this type.