INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A growing body of evidence indicates that patients with sleep apnea-hypopnea syndrome (SAHS) have decreased quality of life, as determined by self-reporting questionnaires and decreased performance on neuropsychological testing (1). Several studies report impairments in adults with SAHS compared with control subjects (2). Other studies show a significant correlation between these impairments in patients with SAHS during wakefulness and the severity of the SAHS based on polysomnographic results (10). Previous investigators have further extended these findings by showing a significant improvement in quality of life measures and neuropsychological deficits with treatment of SAHS with nasal continuous positive airway pressure (CPAP) (2, 3, 7, 9, 14). Whereas many of the latter studies did not include a placebo control treatment, the studies by Engleman and colleagues (3, 19) compared CPAP treatment with placebo treatment consisting of an oral tablet taken at bedtime. Three of these placebo-controlled studies had a cross-over design and found improved neuropsychological function after CPAP treatment (3, 20, 21).

The parallel design study by Jenkinson and coworkers (22) randomized 107 men with SAHS to 1 mo of therapeutic or subtherapeutic CPAP (about 1 cm H₂O). Between-group comparisons showed that subjects receiving therapeutic CPAP had a significant improvement in self-reported daytime hypersomnolence, mean maintenance of wakefulness time, and several variables in a quality of life questionnaire. In another parallel, randomized design study, Yu and coworkers (24) administered the Profile of Mood States questionnaire to 34 patients with SAHS before and after 1 wk of treatment with effective and ineffective CPAP. Significant improvements in mood were not observed by Yu and coworkers when sham CPAP was compared with therapeutic CPAP (24).

The primary purpose of the current study was to determine whether treatment with nasal CPAP is associated with improvements in neuropsychological function in SAHS. As in the study of Jenkinson and coworkers (22), ineffective CPAP (0-1 cm H₂O) was used as the placebo treatment. Besides this primary outcome objective, several other questions were addressed. On the basis polysomnographic comparisons, can ineffective nasal CPAP be used as a placebo treatment for SAHS? Are the changes in neuropsychological assessment results related to the polysomnographic severity of SAHS before treatment and/or length of treatment with nasal CPAP?

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subject Selection

A total of 45 subjects (25 males and 20 females) participated in the study. Subjects were recruited into the study between April 1992 and June 1996. During the subject recruitment phase of the study, all patients referred to the sleep laboratory and diagnosed with SAHS (apnea-hypopnea index > 10 events/h plus daytime sleepiness or apnea- hypopnea index > 20 events/h with or without daytime sleepiness) were asked to participate. None of the subjects had received previous treatment for SAHS. The following exclusion criteria were used: oxygen saturation < 85% for > 50% of the sleep time, clinical signs of right-sided congestive heart failure, and claustrophobia or nasal obstruction preventing use of nasal CPAP. Many patients declined to participate in the study. Because of the amount of time commitment the study involved patients declined for two major reasons: either they worked full-time days and were unable to participate, or they lived too far away to travel for the study. None of the patients had a history of drug or alcohol abuse, myocardial infarction within the past 6 mo, cirrhosis, chronic obstructive pulmonary disease, or unstable diabetes or thyroid disorder. The protocol was approved by the Institutional Review Board and informed written consent was obtained from each subject.

Study Design

A randomized, placebo-controlled, partial cross-over, double-blind study was performed (Table 1). At study entry, subjects were randomly assigned to one of two groups. Group 1 subjects received effective CPAP throughout the treatment period. Group 2 subjects received ineffective CPAP (0-1 cm H₂O) during the first half of the treatment period and effective CPAP during the latter half of the treatment period. A partial cross-over design was used, rather than a full reversal, to avoid any potential carryover effects of CPAP after cessation of effective treatment. The study was double blind in that neither the subjects nor the staff who had contact with the subjects or their records knew the group to which the subjects belonged. At study entry, the subjects were informed that during the course of the study they may or may not be receiving effective treatment but that every subject would receive effective treatment for at least part of the study.

All subjects received a neuropsychological assessment at prebaseline (T1). The initial polysomnograms of the subjects, establishing the diagnosis of SAHS were used for the prebaseline polysomnographic measures. To minimize nonspecific learning effects with repeated testing, the neuropsychological evaluation was repeated before any intervention (T2), and these latter results were used as the baseline for the subjects. The time interval between T1 and T2 was at least 7 d for all but five subjects (mean 47.3 ± 42.4 [SD] d; range, 1-184 d). The T2 neuropsychological evaluation was followed later that evening by a polysomnogram. During the T2 polysomnogram in Group 1 subjects, the amount of CPAP needed to eliminate snoring and prevent apneas and hypopneas was determined. Group 2 subjects performed the T2 polysomnogram on 0-1 cm H₂O CPAP with face masks into which a large hole had been drilled. The resulting level of CPAP in the nose mask was measured with a water manometer. The next morning, subjects assigned to Group 1 were given a nasal CPAP apparatus set at the optimal level determined the previous night (mean, 10 ± 2.4 cm H₂O; range 6-16 cm H₂O), whereas subjects in Group 2 were given a nasal CPAP apparatus set at 0-1 cm H₂O. The subjects were told to use their nasal CPAP apparatus at home whenever they slept for the remainder of the study. Subjects in Group 1 received effective treatment for an average of 16.1 ± 7.3 d (range, 13-38 d) between T2 and T3. Subjects in Group 2 received ineffective treatment for an average of 13.9 ± 4.7 d (range, 7-27 d) between T2 and T3.

All subjects received a third (T3) neuropsychological assessment followed that evening by a polysomnogram. During the T3 polysomnogram, Group 1 subjects were maintained on their optimal CPAP levels. A titration CPAP study was performed with Group 2 subjects to determine the optimal pressure settings needed to treat their SAHS. The next morning, Group 2 subjects had the CPAP levels for their home machines adjusted to the optimal therapeutic levels determined the previous night. All subjects then remained on effective CPAP levels for a minimum of 12 d (mean, 20.1 ± 13.6 d; range, 12-82 d). A neuropsychological assessment and polysomnogram on effective CPAP were performed in all subjects at the conclusion of the study (T4). During the periods between T2-T3 and T3-T4, adherence to CPAP treatment at home was assessed by the time meter on the CPAP machines. Total power-on time in hours was divided by the time interval in days to calculate mean power-on time in hours per day.

Forty-five subjects completed the assessments at T1, T2, and T3. Thirty-nine completed the entire study. Of the subjects completing the entire study, 24 subjects had been assigned to Group 1 and 15 subjects to Group 2. In four subjects in Group 2, the T2 polysomnogram was mistakenly performed on effective CPAP rather than ineffective CPAP. Without objective documentation of the effect of ineffective CPAP on the polysomnographic measures, it was retrospectively elected to drop these subjects even though they had completed the entire protocol.

Assessments

Neuropsychological assessment. The neuropsychological tests were administered by a clinical psychologist, who was blinded to group assignment. Wechsler Adult Intelligence Scale-Revised (WAIS-R) information raw and scaled scores were determined at prebaseline (T1) to further characterize the study participants. The following tests were administered at each of the four time points during the study: Trail Making A and B, Digital Span Test Forward and Backward, Epworth Sleepiness Scale, SteerClear, Digit Symbol, Controlled Oral Word Association, Medical College of Georgia Complex Figure Recall, Geriatric Depression Scale, and Selective Reminding Test Total Words Recalled (six trials) (4, 25, 26). These tests required 2.0-3.5 h to complete, depending on the subject. To control for diurnal variation, all neuropsychological evaluations were performed between 2:00 and 6:00 P.M. Test-specific learning effects were minimized by administration of four parallel packets on a randomized, counterbalanced schedule.

Polysomnograms. Four overnight polysomnograms were performed in each subject, using standard techniques (27). The following measures were recorded during each polysomnogram: electroencephalogram (C3/A2 and O2A1, or C4/A2 and O1/A2), bilateral electrooculograms, submental electromyogram, nasal and oral airflow, rib cage and abdominal movement and their sum via impedance plethysmography (Ambulatory Monitoring, Ardsley, NY), electrocardiogram, tibialis anterior electromyogram, body position, and oxygen saturation by pulse oximetry (Ohmeda, Louisville, CO). The amplified signals (AstroMed-Grass, West Warwick, RI) were collected and stored on a computerized sleep system (Mallinckrodt, Plymouth, MN). All studies were scored manually with the aid of computer software by well trained polysomnography technicians under the scrutiny of the laboratory supervisor. The following primary outcome measures were used to assess the severity of the SAHS: apnea-hypopnea index, apnea index, hypopnea index, minimum oxygen saturation, desaturation index, average minimum O₂ desaturation during abnormal respiratory events, and percent sleep time below 90% oxygen saturation. The following secondary outcome measures were used to compare the quality of the polysomnograms: total recording time, total sleep time, sleep efficiency, sleep maintenance efficiency, percentage of time in stage 1, 2, 3, and 4 nonrapid eye movement (NREM) sleep, percentage of time in REM sleep stage, and amount of time in supine and lateral recumbent positions.

Statistical Methods

Assessments of the two groups at baseline for differences among demographic variables (e.g., age, weight, body mass index [BMI], ethnicity, education) were made with two-sample t tests and ² tests. Within-group changes over time were assessed with paired t tests. For example, the extent to which neuropsychological test results improved after the initial period of effective CPAP compared with baseline (T2-T3 for Group 1) was a within-group assessment. Between-group comparisons of means were made with two-sample t tests. The comparison of CPAP adherence between the two groups is an example of a between-group comparison made with two-sample t tests. The Pearson correlation, the Fisher exact test, and analysis of covariance were also used as described in RESULTS. All statistical tests were performed with Sigmastat software (Jandel, San Rafael, CA). A two-sided  level of significance of 0.05 was used for all tests. Unless otherwise specified, all means are presented with standard deviations.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subject Characteristics

A comparison of subjects in Group 1 and 2 at T1 is shown in Table 2. No significant differences in age or BMI were found between the two groups at entry into the study (T1), using two-sample t tests (all p  0.87). In addition, no significant differences were apparent in sex or ethnicity between the two groups by ² tests (both p  0.40). Body weight at T1 was 117.7 ± 30.4 kg in Group 1 and 118.2 ± 30.2 kg in Group 2. No significant changes in body weight were detected within either group between T1 and T4.

Polysomnographic Results

Outcome measures from the four polysomnograms in each group are shown in Table 3. As expected, no significant differences were present between groups in any of the seven primary or seven secondary outcome measures from the T1 polysomnograms (diagnostic study), using two-sample t tests (all p  0.11). Using paired t tests in Group 2, no significant differences were present in any of the outcome measures from the polysomnogram without CPAP (T1) compared with those from the polysomnogram with ineffective CPAP (T2) (all p  0.08), that is, the placebo did not affect any measured sleep parameter. No significant differences in polysomnographic measures between groups were present at T4 (both groups on effective CPAP) except for the desaturation index (p = 0.03). Three subjects in Group 2 had desaturation indices at T4 of 20, 10, and 10 events/h. Despite this finding, effective CPAP was associated with a near elimination of apneas and hypopneas in both groups, and a dramatic improvement in the primary polysomnographic outcome measures assessing severity of SAHS.

CPAP Adherence

Individual means for power-on time were available in 32 subjects (13 in Group 1, 19 in Group 2) for the period between T2 and T3 and in 23 subjects (12 in Group 1, 11 in Group 2) for the T3-T4 period (Table 4). For Group 1, mean power-on time was 5.9 ± 1.8 h/d in the T2-T3 period and 5.8 ± 2.0 h/d in the T3-T4 period. For Group 2, mean power-on time was 5.2 ± 2.2 h/d in the T2-T3 period and 4.9 ± 2.4 h/d in the T3-T4 period. No significant differences were present in CPAP adherence between the two groups in either of the time periods, using two-sample t tests (both p  0.30). No significant differences in CPAP adherence within groups between time periods were found using paired t tests (both p  0.63). When interviewed at the end of the study by a physician who was not directly involved with data collection and analysis, the subjects in Group 2 were unaware that they had been on ineffective CPAP.

Neuropsychological Test Results

No significant differences in neuropsychological test results were found between the two groups at entry into the study (T1) (all p  0.30). In addition, no significant differences were detected in neuropsychological test results obtained at T1 versus T2 using paired t tests of data from all subjects, suggesting that the results of these tests were not associated with a significant learning effect (all p  0.86). The placebo treatment did not affect any of the neuropsychological test results. Using paired t tests to compare the neuropsychological test results for all subjects during their initial period of effective treatment (T2-T3 in Group 1 and T3-T4 in Group 2), significant improvements in performance were present in the Digit Span Test Forward and Backward, Complex Figure Recall, and Digit Symbol tests (all p < 0.03) (Figure 1). Favorable improvements also occurred in the Epworth Sleepiness Scale (p < 0.001) and SteerClear scores (p = 0.007) based on paired t tests (Figure 1).

View larger version (22K): [in this window] [in a new window]   Figure 1.   Mean ± SE of results in Group 1 (closed symbols) and Group 2 (open symbols) during the four assessment times on the Digit Span Test Forward and Backward, Complex Figure Recall, Digit Symbol, Epworth Sleepiness Scale, and SteerClear. For all tests except for the Epworth Sleepiness Scale, higher scores indicate improvement. Thick line after time 2 indicates period of effective CPAP treatment, and dashed line indicates period of ineffective CPAP treatment (0-1 cm H2O). Significant improvements in all tests occurred during the initial treatment period in both groups (T2-T3 in Group 1 and T3-T4 in Group 2); however, no significant differences between groups were present in the T2-T3 period, that is, when Group 1 was receiving effective CPAP and Group 2 was receiving ineffective CPAP.

During the total time on effective CPAP in Group 1 (T2 to T4; mean, 31.3 ± 11.6 d), significant improvements occurred in the following tests: Controlled Oral Word Association, Digit Symbol, Epworth Sleepiness Scale, and SteerClear (all p values  0.025). The improvement in Trail Making Test B data approached significance (p = 0.069). During the time on effective CPAP treatment in Group 2 (T3 to T4; mean, 19.6 ± 12.2 d), significant improvements occurred only in the Epworth Sleepiness Scale, SteerClear, and Trail Making Test B (all p  0.01). The greater number of significant differences in the group with longer treatment suggests that improvement continued with the longer treatment.

To further assess whether the improvements in neuropsychological test results were related to the length of time on effective CPAP treatment, changes in scores on the 11 neuropsychological measures during the period of effective treatment were correlated with the number of days on effective CPAP in Group 1 (T2 to T4) and Group 2 (T3 to T4). No significant correlations were found (all p > 0.08, 10 of the 11 tests p > 0.32), indicating that longer treatment was not associated with greater improvements.

We also wanted to assess whether changes in the neuropsychological tests from T2 to T3 differed between groups because the two groups were undergoing different experimental conditions (i.e., Group 1 on effective CPAP, Group 2 on ineffective CPAP). We created binary variables for each test categorized as improved (+ change) versus not improved (zero or  change). Using the Fisher exact test, we found that the amount of improvement was not significantly different between the two groups (all p  0.09). As an alternative approach, analysis of covariance was used to assess group differences in the test scores at T3, using the scores at T2 as the covariate. The scores at T3 were highly correlated with the T2 scores; but the group differences were all nonsignificant (all p  0.13). Therefore, the improvement in some of the neuropsychological test results with therapeutic CPAP treatment does not appear to be a specific effect of therapeutic CPAP. Comparison of the neuropsychological test results over comparable time periods on effective CPAP versus placebo treatment showed no significant differences.

To determine whether the improvements in neuropsychological test results with effective CPAP treatment were related to the polysomnographic severity of SAHS at study onset, baseline AHI was correlated with change in scores on the 11 neuropsychological measures during the first 2 wk on effective CPAP (T2 to T3 in Group 1, and T3 to T4 in Group 2). The only significant correlation found was between baseline AHI and the Epworth score (p = 0.03). Subjects with the higher baseline AHI had a greater decrease in Epworth score during the first 2 wk on effective CPAP.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This randomized, placebo-controlled, partial cross-over study found that neuropsychological function in patients with SAHS improves with therapeutic CPAP treatment but that these changes were not different from those that occurred with placebo CPAP treatment. Thus, the results of the study do not indicate that treatment specifically with CPAP has a favorable impact on at least those neuropsychological parameters that were assessed. The study also demonstrates the importance of using placebo CPAP when looking at neuropsychological outcomes after CPAP treatment.

Our results also support those of previous investigators indicating the feasibility of using sham CPAP as a placebo treatment (22, 24, 28, 29). The current study demonstrates that sham CPAP can be used without changing any measured polysomnographic outcome measure. Adherence to ineffective CPAP was similar to use of effective CPAP as was also reported by Yu and coworkers (24). No significant differences in power-on time during ineffective and effective CPAP treatments were detected either within the cross-over group or between the two groups. During the exit interview at the end of the study, subjects in Group 2 were unaware that they had been on ineffective treatment for part of the study.

In the current study, CPAP compliance was measured using power-on time. The mean run time  4.9 h, compared favorably with that reported in previous outcome treatment studies (3, 19). The CPAP machines used in this study were not equipped with a covert pressure monitor to measure the amount of pressure delivered when the machines were turned on. Previous investigators have compared these two measures of CPAP compliance and report that the desired pressure is delivered for about 80-90% of the power-on time (30).

The value of CPAP therapy across the range of SAHS severity and daytime sleepiness is currently an issue of debate. Randomized controlled studies have compared the efficacy of CPAP with conservative treatment such as sleep hygiene, weight loss, and nasal dilator (9, 15). Additional randomized controlled studies have compared efficacy of CPAP with placebo oral tablet (3, 18, 20, 21) and subtherapeutic CPAP (22, 24, 28). Collectively, these studies demonstrate improvement in symptoms (3, 15, 20), quality of life (3, 9, 15, 20, 22), sleepiness (3, 9, 15, 20, 22), mood (3, 20, 22), and neuropsychological performance (3, 20). These studies have generally been performed with patients with severe obstructive sleep apnea (OSA). Two groups of investigators have specifically evaluated the impact of treatment in the largest segment of patients with OSA, that is, those with mild to moderate disease (3, 9, 21). In these latter studies, treatment benefit was present, but not as marked as that with more severe disease.

In contrast to the above-described studies, other randomized, placebo-controlled studies have failed to show an improvement in neuropsychological function with CPAP treatment. The parallel design study of Engleman and coworkers (19) compared CPAP treatment with conservative treatment in 37 patients with SAHS and found significant within-group improvements in cognitive function between baseline and 3 mo. However, when comparisons were made between groups, the only significant difference was a greater improvement in multiple sleep latency with CPAP. Similarly, Yu and coworkers (24) found no significant difference in improvement in mood symptoms between patients with SAHS treated for 1 wk with therapeutic versus sham CPAP. Comparisons of changes in neuropsychological function on effective and ineffective CPAP in the current study also failed to disprove the null hypothesis. Comparison of the current study with published studies also reveals that the subjects in the current study had one of the highest mean apnea-hypopnea index values. However, one would predict that the greater severity of SAHS in our subjects would only maximize the ability to detect changes with treatment.

Differences in study design seem unlikely to explain the differences in results. Placebo-controlled studies reporting a significant improvement in neuropsychological function on CPAP have used either a cross-over or parallel design. The analysis of the data in our partial cross-over study during the first 2 wk of treatment would be analogous to the parallel design. It is possible that differences with regard to the information given to the participants regarding the purpose of the study might influence the results. In the current study, subjects were informed that they might or might not be given ineffective treatment during some part of the study. In most studies that showed significant improvements in neuropsychological function with CPAP, subjects were told that the study was comparing the effectiveness of two types of treatment (oral tablet versus CPAP or two different CPAP pressures).

The absence of a significant difference in changes in neuropsychological function after effective CPAP treatment in the current study may have been due to the fact that the selected neuropsychological tests were inappropriate to assess the deficits secondary to SAHS. However, some of the same tests reported to show significant improvements in the oral placebo or conservative treatment versus CPAP studies of Engleman and coworkers (3, 19) were used in the current study, that is, Trailmaking Test B, Digit Symbol, and SteerClear. Of interest, in the current study, subjects receiving effective CPAP showed significant improvements by paired t test in all three of these tests. However, no differences were detected when results in subjects receiving effective treatment were compared with those in subjects receiving ineffective CPAP.

Using subtherapeutic CPAP as their placebo-control, Jenkinson and coworkers (22) performed a randomized, parallel study of 107 men with SAHS and found significant improvements in Epworth score, mean maintenance of wakefulness time, and several variables on the SF-36 quality of life questionnaire after 1 mo of treatment. Our partial cross-over study, which treated both men and women for approximately 2 wk, found that the improvements in Epworth score were similar on ineffective and effective CPAP (Figure 1). These different findings may be explained by the selection of patients with a baseline Epworth score > 10 in the Jenkinson study compared with inclusion of subjects regardless of Epworth score in the current study. Differences in design, duration of treatment, and sex are other differences between these two particular studies that may explain the conflicting results.

The absence of an improvement in neuropsychological outcome measures on effective CPAP when compared with results on ineffective CPAP may have been influenced by an inadequate treatment duration and/or insufficient power due to the relatively low number of subjects. It is of note that most previous studies reporting significant changes in neuropsychological function used a minimum CPAP treatment period of 4 wk. In contrast, Yu and colleagues, using a treatment period of only 7 d, did not find any significant improvement in mood state when comparing CPAP with sham CPAP treatment (24). Although no correlation was found in the current study between length of time on effective CPAP and change in score on the neuropsychological assessments, the possibility that length of treatment may have been insufficient between the T2-T3 time periods to allow detection of group differences is supported by the finding that the subjects receiving effective CPAP throughout the treatment period (T2-T4 in Group 1, mean of 34.0 ± 10.1 d) achieved significant improvements on a greater number of the neuropsychological tests than did subjects receiving effective CPAP for half that time (T3-T4 in Group 2, mean of 19.6 ± 12.2 d) when the results were compared by paired t test with each subject's baseline.

Much larger sample sizes might have enabled us to document more significant findings between the groups for the neuropsychological tests. A post-hoc power analysis using means and standard deviations from this study (Group 1, T2-T4; Group 2, T3-T4) showed that whereas two of the tests had adequate power (Trail Making Test B and Complex Figure Recall, both > 80), the other nine measures had power ranging from 5 to 28% Total sample size requirements for the two groups of 87 to more than 1,000 would have been necessary to find significance. This then raises the question of whether these changes would be clinically significant if so many subjects would be required to detect a significant improvement?

In summary, the findings of this study demonstrate both the feasibility and the importance of using a CPAP placebo when assessing neuropsychological outcomes related to CPAP treatment in OSA. In agreement with previous investigators, we found significant improvements in neuropsychological outcomes with effective CPAP treatment. However, these improvements were no different when comparing the subjects on effective versus ineffective CPAP. Adequately powered studies comparing quality of life assessments in addition to neuropsychological assessments with effective and placebo CPAP are needed to resolve the issue of who derives benefit from CPAP therapy and who, therefore, should receive this intervention, which is neither clinically nor monetarily inconsequential.