INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Although the sleep apnea/hypopnea syndrome (SAHS) is associated with problems in daytime functioning, including excessive sleepiness, cognitive deficits, and psychological impairment (1, 2), the minimal polysomnographic severity at which patients benefit from treatment is not well established. Nevertheless, recent epidemiologic evidence indicates that even minor elevations of sleep-disordered breathing (apnea + hypopnea index [AHI] 5 or more) are statistically associated with increased frequencies of a variety of adverse consequences, including hypertension and cardiovascular events (3) and daytime problems in the form of excessive daytime sleepiness (4), cognitive deficits (5), and road traffic accidents (6). Case-control studies in subjects with mild sleep-disordered breathing, some asymptomatic (7), and in patients with mild SAHS (2, 8) provide further links between these conditions and impaired performance on attention-biased cognitive tests (7) and reduced quality of life (2, 8).

This evolving evidence suggests that consequences from sleep-disordered breathing may occur at low AHI values, often below the thresholds centers use to determine whether patients should be treated. In the case of daytime function, objective (9) and self-report (10) studies conducted in SAHS patients with a wide range of severity have demonstrated few or weak relationships between AHI and subsequent benefits from therapy, weakening the rationale for treatment of severe SAHS only.

The current treatment of choice for SAHS, continuous positive airway pressure (CPAP) therapy, was employed in a pilot placebo-controlled, randomized study in 16 patients with mild SAHS (AHI 5 to 15) (11). Despite the small sample size, improvements to symptom ratings, cognitive performance, and psychological well-being were documented, although no changes in sleepiness were observed. We have thus conducted a second study using a modified daytime testing protocol in a newly recruited, larger, and more symptomatic sample, in an attempt to expand on these preliminary findings and identify whether treatment benefits extended to patients with 5 to 10 apneas + hypopneas per hour slept.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Selection

A prospective series of patients were recruited from new attenders at the outpatient sleep clinic. None were included in any of our previous clinical trials (9). Entry criteria specified an initial complaint of at least two symptoms of the SAHS (14), including significant sleepiness demonstrated by an Epworth sleepiness score of 8 or greater or admitted sleepiness while driving, and a demonstrated AHI on polysomnography in the range 5.0 to 14.9 per hour slept. Apneas were scored when thermistor airflow was absent for 10 s or longer, and hypopneas scored when abdominal or thoracic respiratory movement amplitude was reduced to 50% or less of the preceding stable baseline value for 10 s or longer, during sleep (15). Microarousals were defined by 1.5 s or longer of increased electroencephalogram (EEG) frequency accompanied by a rise in electromyogram (EMG) amplitude (1). Patients residing more than 50 miles from the laboratory, shift workers, and those with other coexisting sleep disorders, neurological or lung disease were excluded.

Study Protocol

The trial adopted a randomized, placebo-controlled crossover design, in which all patients spent 4 wk on CPAP and 4 wk on an oral placebo therapy, with no washout period. Patients were randomized to a treatment order using balanced blocks, half commencing treatment with CPAP and half with the oral placebo as first treatment. Independent randomization blocks were employed for patients with lower (AHI 5.0 to 9.9) and higher (AHI 10.0 to 14.9) polysomnographic disease severity to ensure no severity bias in assigned treatment order.

With the permission of the local ethics subcomittee, patients were told that the placebo treatment (Glaxo, Greenford, UK), prescribed in a dose of two tablets at bedtime, might improve upper airway muscle function in sleep. At the start of the CPAP treatment limb, patients were issued with a Sullivan III CPAP unit and a heated CPAP humidifier (both ResMed Ltd., Abingdon, UK) and advised to use CPAP, with or without humidification, all night and every night and during any daytime naps during that treatment period. Patients were supplied with a contact telephone number in the event of problems or side effects with CPAP, and any problems not prevented by humidification were actively sought in telephone contact made in the second week of treatment, so that these could be managed and compliance reinforced.

Before the commencement of treatment, patients underwent a day of familiarization and baseline assessment with all daytime function tests (Table 1) except the maintenance of wakefulness test, and were fitted with a nasal mask and educated in the mechanisms and technique of CPAP treatment. All underwent an overnight CPAP titration study to establish an optimal pressure to abolish breathing irregularities and arousals from sleep. After commencement of treatment, the outcome measures in Table 1 were repeated on the last day of each treatment period. At the final assessment, patients were asked to state their treatment preference, weighing up the benefits and drawbacks of each treatment type.

Assessments

CPAP use. The total time that CPAP units were switched on was logged using the units' integral timeclock. Total time at an effective pressure was logged using a CM-1 compliance meter (ResMed Ltd) attached to the CPAP unit, which sensed CPAP level at units' outflow port. The compliance meter's threshold pressure was adjusted to a level approximately 1 cm H₂O below prescribed pressure, so that time-logging was activated only when the nasal CPAP mask was in place. Threshold pressure was adjusted for each patient's prescribed pressure and mask type, and confirmed as operative both with and without the addition of a humidifier to the CPAP circuit. At the end of each CPAP treatment limb, the CPAP unit timeclock was read and the compliance meter memory downloaded onto a PC, allowing average nightly rates that CPAP units were switched on and were used effectively to be calculated. Patients were also asked to complete a diary of daily CPAP use, which was not returned by 10 patients.

Daytime assessments. Daytime function tests, shown in Table 1 and most previously described (9, 13), were administered in a standardized order over each assessment day, subjects having slept the night before in their own bed. The test battery was modified from that used previously (9, 11, 13) by a reduction in the number of cognitive outcome measurements, a substitution of the maintenance of wakefulness test (MWT) for the multiple sleep latency test (MSLT), and a supplementation of the health and functional status measurements.

The symptom questionnaire was modified from the dichotomous item format (present/absent) used in previous studies (9, 11, 13), and asked patients to rate the frequency of nine symptoms (see Table 1) on a 6-point Likert scale from 0 (never) to 5 (always). This revised format was introduced in an attempt to increase the sensitivity of the questionnaire to subtotal improvements in symptoms.

The tests of sleepiness included subjective ratings of trait (Epworth scale) (16) and state (University of Wales Institute of Science and Technology [UWIST] mode adjective checklist [UMACL] energetic arousal scale) sleepiness, and the objective MWT (17). The Epworth scale (score range 0-24) assessed the likelihood of dozing in eight everyday situations in the preceding month, whereas the energetic arousal scale (score range 8-32) elicited ratings of immediate mood by asking subjects to strongly agree, agree, disagree, or strongly disagree with the adjectives "tired," "sluggish," "unenterprising," "passive," "active," "energetic," "alert," and "vigorous." The MWT consisted of four trials spaced at 2-h intervals through the assessment day, each lasting up to 40 min and performed with the patient semireclined on a bed in a dimly lit sleep laboratory bedroom. Subjects were instructed to remain awake as long as possible, without reading, singing, or making excessive physical movements. Sleep onset was scored from the first 30-s epoch of any sleep stage using standard criteria (18), or as 40 min in the absence of sleep. The MWT was used because the MSLT showed no changes with treatment in a pilot study of patients with mild SAHS (11).

The cognitive battery was reduced to those tests previously showing changes with treatment in controlled trials (9, 11, 19) in order to minimize the possibility of Type 1 errors. These comprised primarily attention-biased performance tests, including the TrailMaking tests. Wechsler Adult Intelligence Scale-Revised (WAIS-R) scale performance IQ subtests (Digit Symbol Substitution and Block Design), and the Paced Auditory Serial Addition Test (PASAT). A 30-min computer-administered vigilance task, SteerClear, has previously shown improvement with CPAP in patients with classic SAHS (20, 21) and in a heterogeneous patient group (9), but no change in patients with mild SAHS (11). It was thus extended to a 60-min run for the present sample to probe for longer-onset performance deficits.

As in previous studies (9, 11), minor psychiatric morbidity was assessed using the Hospital Anxiety and Depression Scale (HADS). The Nottingham Health Profile (NHP) Part 2 was retained as a quality of life measure due to its previous sensitivity to therapeutic intervention (9, 11) and was supplemented by the SF-36 scale, a self-rating of health and functional status (22). Scores for the PASAT were missing for two subjects who declined to attempt the task during both treatment assessments. One of these also declined to complete the SF-36 scale during the placebo treatment.

Statistics

Ordinal, interval, and non-normally distributed variables included effective CPAP use, baseline microarousal index, and minimal oxygen saturation, daytime outcomes from SteerClear and TrailMaking B, and the HADS, NHP Part 2, and SF-36. All other data were continuous and normally distributed.

The effectiveness of treatments was assessed by comparing assessments conducted on placebo and CPAP, using Wilcoxon tests for ordinal, interval and non-normally distributed variables, and two-way analysis of variance, with treatment order as a between-subjects factor and treatment type as a within-subject factor, for normally distributed and continuous data. Treatment preference was assessed using the binomial test. As an indicator of the magnitude of therapeutic benefit, the effect sizes of changes between placebo and CPAP were calculated (difference/SD of difference) with effect sizes of 0.20 considered as small, 0.50 as moderate, and 0.80 large, after Jenkinson and coworkers (23).

The influence of putative covariates, comprising effective CPAP use and baseline AHI, microarousal index, and minimal oxygen saturation, was evaluated by entering these potential determinant variables (all except AHI normalized by log transformation) into the significant analyses of treatment effects. This analysis investigated the effects of CPAP use and polysomnographic severity on the scale of therapeutic benefit. Additionally, Mann-Whitney tests were employed to seek differences in treatment response between better and poorer CPAP users.

A subanalysis was conducted within the lower randomized severity group (AHI 5.0 to 9.9) to separately assess whether significant treatment effects extended to this mildest group. All analyses were conducted using SPSS for Macintosh (24). For the sample size of 34 patients, statistical power to show significant differences (p  0.05) between treatments for the observed differences in variables was 95% for symptoms, 25% for MWT, > 99% for Epworth, 60% for PASAT, and 85% for HADS depression scores.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Letters inviting participation were sent to 59 patients fulfilling selection criteria, of whom 11 did not respond. A further 11 patients declined participation (five with difficulties in obtaining time off work and two difficulties with the travel required, two patients unwilling to take tablets, one unwilling to use CPAP, and one patient failing to keep appointments). Thirty-seven patients commenced the trial. Three of these withdrew after treatment commencement, one being unwilling to persist with CPAP therapy as a first treatment, one repeatedly failing to attend for final assessment on placebo, and the last finding the traveling for assessment in the final, CPAP limb too demanding.

Thirty-four patients (13 women) completed the study, with an average age of 44 ± 8 yr (mean ± SD) and mean body mass index 30 ± 5 kg/m². Polysomnography in these showed an average AHI of 10 ± 3 and microarousal index of 24 ± 10 per hour slept, with a mean minimal oxygen saturation of 89 ± 5%. Twenty-seven of the patients were automobile drivers, of whom 13 at baseline admitted sleepiness while driving, occurring monthly or more often.

Use of CPAP

CPAP units were switched on for an average 3.2 ± 2.4 (mean + SD) (range, 0.1 to 7.7) h per night, and used effectively for an average 2.8 ± 2.1 (range, 0.0 to 7.4) h per night. In the 24 patients returning a CPAP use diary, self-reported use averaged 4.5 ± 2.5 h per night compared with a mean objectively monitored effective use rate of 3.5 ± 2.0 h per night (p < 0.001).

Side Effects of Treatments

Twenty-three of the 34 patients reported some degree of adverse effects attributed to CPAP use, including early wakenings from sleep (n = 4), sleep disturbance to patient or partner caused by noise from CPAP generator or humidifier (n = 8), mask or headgear problems (n = 8), or dry or open mouth during CPAP use (n = 4). Other problems volunteered were waking with the mask off (n = 2), continued snoring on CPAP (n = 1), or inability to fall asleep with prescribed pressure (n = 1). No nasal stuffiness was reported. Each of these problems was managed on an individual basis, with advice on noise reduction, issuing of chinstraps and alternative sizes of masks and headgear, empirical alterations to CPAP pressure, and encouragement.

Potential side effects attributable to the placebo tablets were reported by eight patients, and included muscle tightness (n = 1), more frequent awakenings from sleep (n = 1), paresthesia in limbs (n = 1) or throat (n = 1), headaches (n = 3), delayed sleep onset (n = 1), stomach cramps (n = 1), "hungover" and tired sensation in mornings (n = 3), and an episode of chest and arm pain (n = 1). No patients discontinued placebo treatment as a result of these.

Changes in Daytime Function

Treatment order and learning effects. There were no significant treatment-order effects, which could have indicated a breakdown in randomization or carryover effects (p > 0.1) (All changes in daytime function are shown in Table 2.). Treatment × treatment-order interactions, representing learning effects, but controlled by randomization of treatment order, were demonstrated for the cognitive TrailMaking A, digit symbol, performance IQ, and PASAT tasks.

Symptoms. Total symptom score was reduced after CPAP (effect size [ES] = 0.63, p < 0.01), with individual improvement in ratings of severity of snoring and breathing pauses (p < 0.01).

Sleepiness. Recent daytime sleepiness, self-rated on the Epworth scale, fell (ES = 0.75, p < 0.01), but the state rating of sleepiness supplied by the UMACL energetic arousal score showed only a trend (ES = 0.33, p = 0.06) for improvement. Objective sleepiness measured by the mean sleep onset latency from the maintenance of wakefulness test showed no differences between placebo and CPAP treatments (ES = 0.19, p < 0.2).

Cognitive performance. Performances for Digit Symbol Substitution task (ES = 0.14, p < 0.01) and two-second rate PASAT (ES = 0.36, p = 0.02) were significantly enhanced on CPAP compared with placebo, and a trend for improved TrailMaking A performance (ES = 0.33, p = 0.06) with CPAP was observed.

Psychological well-being and health and functional status. The depression score of the HADS scale was significantly lowered after CPAP (ES = 0.41, p < 0.01), reflecting improved psychological well-being, and a trend for improved HADS anxiety score (ES = 0.38, p = 0.07) emerged. Health and functional status, as measured by Part 2 of the NHP, showed no change with CPAP (ES = 0.24, p > 0.3), but when rated using the SF-36 questionnaire showed significant enhancements with CPAP in five of the nine subscales, including health transition, rolephysical, bodily pain, social function, and vitality (ES = 0.44-0.67, p  0.03). A trend for improvement in the mental health subscale was also seen (ES = 0.29, p = 0.09).

Treatment Preference

Fourteen of 34 patients preferred CPAP, when asked to weigh the benefits of treatments against their inconvenience (p > 0.3).

Analysis of Covariance

Both effective CPAP use and baseline microarousal index were significant covariates in the treatment effects of symptom total ( = 0.57 [Figure 1a] and  = 0.41, respectively, p  0.02) and Epworth score ( = 0.35 and  = 0.40 [Figure 1b], p  0.04). Neither AHI nor minimal oxygen saturation showed any significant regression with treatment effects.

View larger version (16K): [in this window] [in a new window]   Figure 1.   Covariance of (a) effective CPAP use with treatment effect for symptom score and (b) microarousal index with treatment effect for Epworth sleepiness score.

Better versus Poorer CPAP Users

Patients were split into better and poorer CPAP users, using a cutpoint at median effective CPAP use (2.5 h per night). Ten of 17 better CPAP users started treatments with CPAP and 10 of 17 poorer CPAP users with placebo treatment. Comparisons of the differences in score between placebo and CPAP in these two groups showed significantly larger improvement in better CPAP users in outcome measures of symptoms total score (better users 8 ± 7 versus poorer users 1 ± 8; p = 0.01), TrailMaking A (6 ± 11 versus 0 ± 6 s; p = 0.04), NHP Part 2 (3.1 ± 6.5 versus +0.7 ± 2.1; p = 0.04), SF-36 social function (+20 ± 28 versus +4 ± 18; p = 0.05), and SF-36 vitality score (+21 ± 16 versus +5 ± 16; p = 0.01). Trends toward larger improvements in MWT (+4.9 ± 10.8 versus 1.2 ± 6.8 min; p = 0.07) and Epworth score were also observed (4 ± 5 versus 1 ± 4; p = 0.07) in the better CPAP users.

Treatment Effects in Milder Group

A subanalysis of treatment effects within the milder severity group (AHI 5 to 10, n = 14, seven starting with CPAP and seven with placebo treatment) showed significant differences between treatments, all reflecting better function with CPAP, in total symptom score (placebo 16 ± 8 versus CPAP 10 ± 7; ES = 0.75, p = 0.02), in performances for SteerClear (CPAP 217 ± 167 versus placebo 174 ± 50; ES = 0.52, p = 0.05) and 2-s PASAT (37 ± 14 versus 41 ± 12; ES = 0.80, p < 0.01), depression score (6.1 ± 4.1 versus 3.8 ± 3.2; ES = 0.66, p = 0.04), and SF-36 scores for physical function (76 ± 31 versus 81 ± 29; ES = 0.86, p = 0.02), social function (69 ± 31 versus 89 ± 15; ES = 0.70, p = 0.03), mental health (73 ± 15 versus 82 ± 13; ES = 0.50, p = 0.05), and vitality (39 ± 24 versus 55 ± 20; ES = 0.94, p < 0.01). A trend for improved digit symbol score was also observed (53 ± 15 versus 55 ± 12; ES = 0.50, p = 0.07), but no significant changes in sleepiness outcomes were identified (Epworth ES = 0.40, MWT ES = 0.13, both p < 0.10).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This second randomized, placebo-controlled crossover study in a larger and more symptomatic sample of patients with mild SAHS (AHI 5-15) corroborates and extends previous findings of improved cognitive function and psychological well-being following CPAP (11). Additional clinically and statistically significant improvements in subjective sleepiness and in health and functional status with CPAP were documented in the current study. The study design employed controlled both for learning effects and the expectation of benefit from therapy, leading us to believe that its results are comparatively robust, and provide a valuable addition to the growing body of evidence for the efficacy of CPAP requested in a recent systematic review (25).

The only determinants of daytime outcomes with CPAP identified by analysis of covariance were moderate correlations involving microarousal index and effective CPAP use. These determinant variables were themselves intercorrelated (Spearman correlation, r = 0.44, p = 0.01), suggesting that some 19% of the variability in CPAP use might be related to microarousal frequency. The results of the analysis of covariance were supported by the comparison of better and poorer CPAP users, showing larger improvements for symptom ratings, the TrailMaking A test of visuomotor tracking skill, and health and functional status measures including NHP Part 2 score and SF-36 subscores for social function and vitality, in better CPAP users. Trends toward better improvements in sleepiness in better CPAP users were also seen, supporting the hypothesis that greater compliance with CPAP was associated with larger treatment response. These results are consistent with subanalyses in previous studies (11, 12) which showed additional improvements in well-being and functional status in better adherers to CPAP treatment, and occurred despite an imbalance in the treatment order within the two groups which weighed against finding larger improvements, particularly for learning-influenced cognitive variables.

The absence of significant association between AHI and either CPAP use or changes in daytime functions, although investigated in a sample of limited variability, suggests that AHI may be less clinically informative than microarousal index in this sleepy but polysomnographically mild group. Such an inference is also supported by the subanalysis in patients with AHI in the range 5 to 10, who showed significant within-group improvements with CPAP in all areas of daytime function except sleepiness, despite much-reduced statistical power (n = 14). Some outcome measures significantly improved by CPAP in these mildest patients varied from those improving in the full group, and the effect sizes of some improved variables were larger for the AHI 5 to 10 group than for the AHI 10 to 15 group. In keeping with the results of the analysis of covariance, these may in part be due to the nonsignificantly greater effective CPAP use in the milder group (3.2 ± 2.3 versus 2.5 ± 2.1 h per night; p > 0.4), but may equally represent random interindividual variability and changes in statistical power with varying sample size.

Furthermore, when differences in scores between placebo and CPAP in the full current study group (AHI 5 to 15) and in a pooled sample of 48 patients with AHI of 15 or more (20) were compared, no significant differences in improvements with CPAP were revealed except a greater shortening in TrailMaking A completion time in the mild sample (3 s) than in the more severe patients (+1 s; Mann-Whitney test, p < 0.03). This lack of evidence for greater benefits from CPAP in more severe SAHS patients is congruent with previous observations (9, 10), and has relevance to clinicians and health providers using high AHI thresholds during the diagnostic assessment and management of suspected SAHS patients. Together these findings undermine the clinical validity of recent proposals to restrict treatment to patients with severe SAHS (25, 26). Along with epidemiologic evidence of adverse consequences from mildly elevated sleep-disordered breathing (AHI 5 to 15) (3), they instead support a role for treatment in sleepy patients with mild SAHS.

Effect Sizes of Changes in Daytime Function

The effect sizes of all significant improvements in daytime function scores were small or moderate (23) by the employed method of calculation. However, these were minimized by being based on the difference between placebo and CPAP scores, and not between baseline scores and CPAP, as used in other longitudinal trials (23). If baseline and CPAP were compared, large effect sizes of 1.83 for symptoms, 1.67 for Epworth sleepiness score, 0.75 for the cognitive PASAT score, and 0.83 for HADS depression score would be produced.

Changes in Symptoms

The effect size of the improvement in total symptom rating with CPAP was moderate, but limited in statistical significance to the individual questionnaire items on snoring and breathing pauses rather than daytime symptoms.

Changes in Sleepiness

The magnitude of improvement in Epworth sleepiness score was moderate, with mean baseline and placebo values reflecting excessive daytime sleepiness falling into the normal range (16) with CPAP. This finding contrasted with the sleep onset latency from the MWT, which did not improve with CPAP. MWT values remained in a range associated with excessive sleepiness on improvement with CPAP. MWT values remained in a range associated with excessive sleepiness on both treatments, in contrast to previous uncontrolled studies of CPAP in SAHS (17, 27). The MWT was included instead of the MSLT in the expectation that it might yield a more clinically relevant outcome (27). The lack of statistical improvement in MWT in sleepy patients with mild SAHS might be a result of lesser disease severity, differences in study design, low CPAP use, or the limited statistical power afforded by this study for this variable. The MWT, by evaluating the ability to resist sleep, may also be vulnerable to patient motivation factors and placebo effects.

Changes in Cognitive Performance

Cognitive improvements with CPAP (on digit symbol substitution and PASAT, with a trend in TrailMaking A) were small in size and obtained from tests drawing on attention skills. Other longitudinal studies of treatment in SAHS have also suggested selective enhancement to vigilance and attention skills (28), although learning effects have been inconsistently controlled.

Changes in Psychological Well-Being and Health and Functional Status

As in our previous therapeutic studies (9, 11, 12), the HADS depression score showed significant and modestly sized improvements to psychological well-being with CPAP. Using a cutoff score of 8 or more to identify anxiety and depression "cases" (31), 23 patients (67%) had elevated scores for anxiety and 14 (41%) for depression at baseline, suggesting a high prevalence of psychological morbidity in this polysomnographically mild but sleepy patient group. As an indication of the clinical effect of active treatment, the number of depression cases fell from 14 on placebo to five on CPAP (McNemar test; p < 0.01).

Use of the SF-36 questionnaire indicated enhanced health and functional status for physical activity, pain, social function, and energy levels with active treatment, as well as improvement to recent health changes. Prior use of the SF-36 within sleep clinic patients by other research groups in the United Kingdom (2, 23) and North American (8) allows a useful comparison of the degree of functional impairment, and benefit from therapy, gained by this sample of sleepy patients with mild SAHS. Baseline SF-36 scores (Table 2 and Figure 2) showed significant and substantial impairments in all subscales (unpaired t tests; p < 0.001) excepting general health perceptions, compared with United Kingdom normative values for adults of working age (32). These baseline scores were very similar in profile to those in more severe SAHS patients with significant desaturation from other United Kingdom studies (2, 23), and reflected greater health and functional impairment than patient samples characterized as mild in other clinical samples (2, 8, 23). This may well result from our selection criterion of significant sleepiness, but attests to significant health and functional limitations in symptomatic patients with mild SAHS.

View larger version (46K): [in this window] [in a new window]   Figure 2.   Mean SF-36 subscale scores in patients at baseline (white bars), patients on CPAP (black bars), and U.K. normative sample (shaded bars). *p < 0.001 compared with normative sample. (From Reference 32.)

SF-36 changes from baseline values to CPAP (Table 2 and Figure 2) show strong congruence with the large improvements demonstrated in more severe patients with SAHS (2, 23). The number of significant improvements between placebo and CPAP, and their magnitude, were smaller in this study than in the Cambridge (2) and Oxford (23) samples, which incorporated the expectation of benefit from therapy. However, even with a placebo control, ES of some health and functional improvements were moderate, indicating substantial benefits from therapy, and on CPAP no SF-36 scores differed significantly from UK norms (32) (unpaired t tests p > 0.05), although these were not matched for age or sex.

Treatment Preference

The frequency of patient preference for CPAP over the placebo treatment (14 of 34 patients) was low, as was CPAP use rate in this prospective, "intention to treat" trial. Despite these, multiple significant improvements to daytime function, all in the direction of better function on CPAP were documented. A previous trial of CPAP therapy in a heterogeneous sample of snoring patients, some sleepy, found that none took up long-term CPAP treatment (33). The low use and preference rates for CPAP in our study may reflect patients' acceptance of our representation of the placebo treatment as potentially effective, but must also highlight the relatively great inconvenience of CPAP therapy.

Methodological Issues

Although placebo-controlled and randomized, this study design has previously been criticized on methodological grounds (25), for absence of a washout period and its choice of placebo. No washout period was included because of evidence suggesting latencies as short as 1 d for the onset (34) and offset (30) of daytime benefits from CPAP. Further, any persisting effects would reduce, not increase, the chances of finding significant benefits from CPAP. An oral placebo treatment was selected because of reported dangers in administering subtherapeutic CPAP (35), because subtherapeutic CPAP would interfere with sleep thus impairing daytime function and potentially erroneously showing benefit for CPAP, and because in a crossover design the relative strengths of therapeutic and subtherapeutic CPAP would be apparent to subjects, thus unblinding the study. A further potential criticism is that the study included multiple end-points, raising the possibility of Type 1 statistical errors, but 10 of 23 main outcome measurements were significantly improved, and all in the direction of better function with CPAP. We also failed to recruit 22 nonresponders and decliners who were invited to participate, with a potential selection bias in the remaining recruits. Because the most common reason for declining participation related to obtaining time off work for the 3 d of testing required, we may have selected a relatively less healthy or employed population for the trial, but this was an unavoidable consequence of a time-consuming protocol. The daytime assessments conducted on placebo and CPAP were administered in a standardized, and not randomized, order, so that a bias from circadian effects on well-being and performance measurements may exist. A standardized test order was deliberately designed in, with a prospective decision to administer well-being questionnaires in the morning, when mood may be most affected by sleepiness, and cognitive assessments in the afternoon, during the "post-lunch dip" in performance. Such standardization across treatments and patients removed the variability associated with circadian factors, providing greater statistical power to detect the subtle deficits traditionally associated with sleep loss.

In summary, this randomized, placebo-controlled study provides evidence of both objective and subjective benefits to daytime function from CPAP in patients with mild SAHS (AHI 5 to 15), although low use rates and preference rates for CPAP were documented. Thus, further robust evidence exists to support a clinical role for treatment of symptomatic SAHS patients with low indices of severity. Further studies employing alternative treatments such as oral appliances in SAHS patients with mild illness may provide enhanced therapeutic acceptability, although efficacy will need to be proven.