You are on page 1of 7

Sleep Medicine Reviews 17 (2013) 341e347

Contents lists available at SciVerse ScienceDirect

Sleep Medicine Reviews

journal homepage:


Neuropsychological functioning after CPAP treatment in

obstructive sleep apnea: A meta-analysis
Wytske A. Kylstra a, e, f, Justine A. Aaronson b, c, *, f, Winni F. Hofman c, Ben A. Schmand c, d
Heliomare Rehabilitation, Wijk aan Zee, Netherlands
Heliomare Research & Development, Wijk aan Zee, Netherlands
Department of Psychology, Brain and Cognition group, University of Amsterdam, Amsterdam, Netherlands
Department of Neurology, Academic Medical Center, Amsterdam, Netherlands

a r t i c l e i n f o s u m m a r y

Article history: The generally held clinical view is that treatment with continuous positive airway pressure (CPAP)
Received 27 April 2012 improves cognition in patients with obstructive sleep apnea (OSA). However, the cognitive domains in
Received in revised form which recovery is found differ between studies. A meta-analysis was conducted to quantify the effect of
4 September 2012
CPAP treatment in OSA on neuropsychological functioning. A literature search of studies published from
Accepted 4 September 2012
Available online 12 October 2012
January 1990 to July 2012 was performed. The inclusion criteria were: randomized controlled trial,
diagnosis of OSA by poly(somno)graphy, apnea/hypopnea index, duration and compliance of CPAP
treatment reported, use of one or more standardized neuropsychological tests. Mean weighted effect
Obstructive sleep apnea
sizes of CPAP treatment for seven cognitive domains were calculated, including processing speed,
Continuous positive airway pressure attention, vigilance, working memory, memory, verbal fluency and visuoconstruction. Thirteen studies
Neuropsychological functioning encompassing 554 OSA patients were included. A small, significant effect on attention was observed in
Cognition favor of CPAP (d ¼ 0.19). For the other cognitive domains the effect sizes did not reach significance.
Improvement on measures of sleepiness was modest (d ¼ 0.30e0.53) and comparable to prior research.
In conclusion, this meta-analysis indicates that the effect of CPAP on cognition is small and limited to
attention. Contrary to the general assumption, only slight improvement of neuropsychological func-
tioning after CPAP treatment can be expected.
Ó 2012 Elsevier Ltd. All rights reserved.

Introduction Extensive research on neuropsychological functioning among

adults with untreated obstructive sleep apnea has shown that OSA
Obstructive sleep apnea (OSA) is characterized by complete negatively affects cognitive and psychological functioning.4 Vigi-
cessations (apneas) and partial decreases (hypopneas) in respira- lance, attention, executive functioning, memory and motor coor-
tion caused by pharyngeal collapse during sleep. The reduction of dination have been found to be moderately to markedly affected.
airflow causes oxygen desaturation, which can lead to sleep frag- No substantial effects on intelligence, verbal functioning, or visual
mentation and hypoxemia. Due to this, patients with OSA often perception have been reported.1,5,6
wake up feeling tired, with excessive daytime sleepiness being the The treatment of choice for OSA is continuous positive airway
most reported complaint. In turn, this can hamper daily func- pressure (CPAP). CPAP corrects the respiratory disturbances and the
tioning.1 OSA has also been associated with an increased risk for resultant transient desaturation, leading to less sleep fragmenta-
serious medical conditions, particularly cardiovascular diseases, tion during sleep.7 Consequently, it is expected that when sleep is
such as hypertension, heart disease and stroke.2,3 normalized by CPAP treatment, functioning in daily living, cognitive
functioning, and psychological wellbeing of OSA patients will
improve. Previous meta-analyses demonstrated significant
* Corresponding author. Heliomare Research & Development, PO Box 78, 1940 AB improvement in sleepiness and self-reported health status with
Beverwijk, Netherlands. Tel.: þ31 88 920 8013. CPAP when compared to placebo treatment or conservative
E-mail addresses: (W.A. Kylstra), j.aaronson@ management.8e10 A number of reviews on cognitive functioning in (J.A. Aaronson), (W.F. Hofman), b.a.schmand@ OSA patients after CPAP treatment have documented partial (B.A. Schmand).
Tel.: þ31 88 920 8328.
reversibility of cognitive dysfunction.1,11,12 The cognitive domains in
These authors contributed equally to this paper. which recovery was noted, as well as the extent of recovery,

1087-0792/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
342 W.A. Kylstra et al. / Sleep Medicine Reviews 17 (2013) 341e347

information was not originally reported, we contacted the

Abbreviations study authors to obtain relevant data.
B Assessment using at least one standardized neuropsychological
AHI apnea/hypopnea index test was employed as a dependent variable.
BMI body mass index B Test scores were reported for both the experimental and
CPAP continuous positive airway pressure control group at baseline and after treatment (mean and
EF executive functioning standard deviation), or other statistics that could be converted
ESS Epworth sleepiness scale to effect sizes. In case this information was not originally re-
HADS hospital anxiety and depression scale ported, we contacted the study authors to obtain relevant
MSLT multiple sleep latency test statistics.
MWT maintenance of wakefulness test
OSA obstructive sleep apnea When articles reported overlapping samples of participants, the
PASAT paced auditory serial addition task article with the largest sample size was included.
PG polygraphy The quality of the randomized controlled trial (RCT) in the final
PSG polysomnography selection was judged using the Jadad rating score assessing
RCT randomized controlled trial randomization procedure, blinding and description of dropout.13
RDI respiratory disturbance index
TMT trail making test Exclusion criteria

Studies were eliminated according to the following exclusion

differed widely across the studies reviewed. This might be due to
criteria: monographs, letters, book chapters, commentaries, review
differences in methods, neuropsychological measures, clinical
articles, case studies, dissertations, abstracts, studies within pedi-
characteristics and treatment compliance. Importantly, these
atric (<18 years) or elderly populations (>65 years) and studies
reviews were qualitative and their conclusions were based on re-
within special medical populations with OSA (e.g., dementia, stroke
ported statistical significance levels without consideration of the
or Down’s syndrome).
magnitude of any observed effects.
In order to quantify the magnitude of the overall effect of CPAP
treatment in OSA on neuropsychological functioning, we carried Outcome measures
out a meta-analysis of the randomized controlled trials of CPAP
treatment in OSA. Primary outcome measures

Methods Multiple neuropsychological tests were used across studies to

assess cognitive functioning. All neuropsychological tests were
Literature search classified into seven cognitive domains following two standard
textbooks of neuropsychological assessment14,15: processing speed,
We initially performed a literature search of Medline, PsycInfo, attention, vigilance, working memory, memory, verbal fluency and
Embase and Cochrane Library covering the period from January visuoconstruction. Appendix 1 lists the included tests per cognitive
1990 to December 2011. In August 2012 we updated our search. domain.
Search terms for OSA were apnea (MeSH), apn*ea, obstructive sleep
apn*ea, OSA, hypopn*ea and SAHS. Treatment search terms included Secondary outcome measures
positive pressure respiration (MeSH), continuous positive airway*
pressure, CPAP, bilevel positive airway pressure, BiPAP, positive pres- For subjective sleepiness, the Epworth sleepiness scale (ESS),
sure therapy and nocturnal ventilation. Search terms for cognitive a self-rating of recent sleepiness behavior, was used, as it was
measures were mental processes (MeSH), neuropsychol*, mental employed in the majority of studies reviewed. For similar reasons,
status, cogniti*, memory, attention, vigilance, executive and psycho- objective sleepiness was quantified by the multiple sleep latency
motor. Searches for all possible combinations of OSA, treatment and test (MSLT), a measure of the time taken to fall asleep, and the
cognitive measures were conducted. We identified additional maintenance of wakefulness test (MWT), a measure of the ability to
published studies by scanning the reference lists of the identified stay awake. Mood was assessed by the hospital anxiety and
papers and checking for journal publications of conference depression scale (HADS), a self-report questionnaire consisting of
abstracts. Two independent assessors identified relevant studies an anxiety and a depression subscale.
based on title and abstract that included empirical data related to
the treatment effect on neuropsychological functioning in OSA.
Participant and study variables
Inclusion criteria
We recorded variables that are considered important risk factors
Studies had to meet the following criteria to be included in the for OSA such as age and body mass index (BMI). BMI was classified
meta-analysis: according to the World Health Organization criteria as normal (BMI
20.0e24.9 kg/m2), overweight (BMI 25.0e29.9 kg/m2) or obese
B The diagnosis OSA was made by polysomnography (PSG) or (BMI 30 kg/m2).16 We registered the AHI in order to quantify the
polygraphy (PG) and the number of apneas and hypopneas per severity of OSA. By convention OSA severity is classified as mild,
hour sleep was stated by apnea/hypopnea index (AHI) or moderate or severe (AHI, >5; >15; >30 events/h, respectively).17
respiratory disturbance index (RDI). We recorded the average number of hours usage of CPAP per
B The treatment of CPAP was investigated within a randomized night. Patients were considered compliant when using CPAP for
controlled design. more than five days a week and for more than 4 h a night, as
B Duration and compliance of CPAP treatment for both the defined by Kribbs et al.18 Duration of treatment was registered in
experimental and control group was reported. In case this weeks as noted in the study designs.
W.A. Kylstra et al. / Sleep Medicine Reviews 17 (2013) 341e347 343

Calculation of effect sizes and statistical analysis Ninety-five articles were initially selected by the independent
assessors with substantial inter-observer agreement (97%; Cohen’s
For parallel studies, we calculated the effect size Hedges’ g by kappa ¼ 0.78). Of these, nine articles did not describe treatment
dividing the mean difference between the CPAP treatment and studies and two articles were case studies. Of the remaining 84
control intervention (i.e., sham, placebo or no treatment) by the articles, 26 were excluded because no control group was included,
pooled standard deviation. In case of cross-over trials, we con- 16 were excluded because the comparison group was made up of
ducted a paired analysis using the mean difference, p-values from healthy individuals, 21 studies were excluded because CPAP was
a paired t-test, or confidence intervals from paired analyses. When compared to other treatments or CPAP withdrawal was investi-
the data required to include a paired analysis were not reported, gated, and five studies did not use standardized neuropsychological
available data were analyzed as if the trial was a parallel group tests. Of the remaining 16 randomized controlled studies that met
design with treatment versus control intervention. We considered most of our criteria, three had to be excluded because it was neither
this to be a conservative approach in that studies were under- possible to extract data to calculate effect sizes from the article nor
weighted; the advantage of the reduced influence of confounding to obtain relevant data directly from the authors. Thus the meta-
covariates in cross-over designs was disregarded. In order to pool analysis was based on 13 RCTs.
the results across studies, we calculated a pooled d-value for all We calculated the Jadad rating score, assessing randomization
seven cognitive domains, weighted for the sample sizes of the procedure, blinding and description of dropout for the final set of 13
individual studies. studies. Four studies reached a maximum score of 5, one a score of 4,
By convention, an effect size of 0.2 was considered small, 0.5 two a score of 3, and six had a poor score of 2 or lower. The poor
moderate, and 0.8 large.19 A positive direction of effect sizes implies scores were due to a lack of information on the randomization
improved performance. When studies used more than one measure method, an inappropriate method of blinding, and no report on
in a cognitive domain, we computed an averaged effect size to avoid withdrawals and dropouts. Study characteristics are shown in Table 1.
one study over-influencing the results for any given domain. We note that the majority of the studies reviewed used a cross-over
Heterogeneity in the results could be expected, given the design. Demographic and clinical characteristics of the samples of all
diversity of both clinical variables and cognitive assessments. studies included in the meta-analysis are displayed in Table 2.
Therefore, we applied the chi-square statistic Q and the I2 index.
The Q-statistic quantifies the degree to which the studies contrib- Participant and study characteristics
uting to each respective mean effect size can be regarded as
homogeneous. A significant Q-value indicates heterogeneity among In total, neuropsychological test results of 533 OSA patients who
studies contributing to the particular mean, in which case a further received CPAP treatment and 497 OSA patients who participated in
search for potential moderator variables is needed. When meta- a control condition were included in the meta-analysis. The average
analyses encompass a relatively small number of studies, the Q- sample size of the 13 studies was 54.2. As can be seen in Table 3,
statistic is believed to have limited value in detecting true hetero- patients had a mean age of 50.5 years and an average BMI of 30.4
geneity among studies because of its rather low power. Therefore, kg/m2, which is qualified as obese by the World Health Organiza-
the I2 was also calculated to evaluate the degree of inconsistency in tion.16 On average, the OSA severity was mild (AHI < 15 events/h) in
the study results. I2 is calculated using the equation (Q  df)/ three studies,22e24 moderate (15  AHI<30 events/h) in four
Q  100%, where df signifies the degrees of freedom (¼number of studies25e28 and severe (AHI  30 events/h) in six studies.29e34 The
studies  1). The I2 index reflects the percentage of total variation overall use of CPAP was 4.5 h per night, which can be considered as
across studies caused by heterogeneity rather than by chance. A sufficiently compliant (>4 h/night for >5 days/week equals
value of 0% indicates no observed heterogeneity and larger values >2.9 h/night). However, CPAP usage was slightly below the defined
indicate increasing heterogeneity. level of compliance in three studies.23,24,31 The treatment duration
For all effect sizes, the statistical uncertainty was accounted for varied widely, ranging from 1 to 13 weeks.
in a 90% confidence interval. We chose a 90% confidence interval
because the treatment effect was expected to be in a positive
Effect sizes
direction. All analyses were conducted using Comprehensive Meta-
Analysis software 2.0 (Englewood, NJ, USA, 2005).
We calculated mean weighted effect sizes (d-values) of CPAP
treatment for each cognitive domain (Table 4). A significant but
Publication bias
Table 1
Design, quality and duration of studies included in the meta-analysis.
To assess publication bias, we inspected funnel plots. The
distribution of these plots should approximate a symmetric funnel Author Year RCT-design Control Jadad Duration
shape, in which the small studies with relatively large variance
scatter at the bottom and the larger studies converge, forming Barbé et al.29 2001 Parallel Sham CPAP 5 6 weeks
Bardwell et al.30 2001 Parallel Sham CPAP 1 1 week
a peak at the average effect size. Asymmetry of the funnel plot may
Barnes et al.22 2002 Cross-over Oral placebo 5 8 weeks
indicate publication bias.20 In addition, we used Orwin’s fail-safe N Barnes et al.25 2004 Cross-over Oral placebo 5 3 months
formula to estimate the number of studies that would be theoret- Engleman et al.26 1994 Cross-over Oral placebo 2 4 weeks
ically needed to overturn the obtained effect and yield a non- Engleman et al.23 1997 Cross-over Oral placebo 2 4 weeks
Engleman et al.31 1998 Cross-over Oral placebo 2 4 weeks
significant effect.21
Engleman et al.24 1999 Cross-over Oral placebo 2 4 weeks
Gast et al.27 2006 Parallel No treatment 2 1 week
Lee et al.32 2012 Parallel Sham CPAP 4 3 weeks
Marshall et al.28 2005 Cross-over Sham CPAP 5 3 weeks
Monasterio et al.33 2001 Parallel Conservative 3 3/6 months
Identification of studies Pelletier-Fleury et al.34 2004 Parallel No treatment 3 6 months

RCT, randomized controlled trial; Jadad, rating score for the quality of a randomized
The final searches yielded 938 articles for consideration (January controlled trial; CPAP, continuous positive airway pressure; duration, duration of
1990 - July 2012). CPAP treatment.
344 W.A. Kylstra et al. / Sleep Medicine Reviews 17 (2013) 341e347

Table 2 Table 4
Demographic and clinical characteristics of studies included in the meta-analysis. Mean weighted effect sizes, confidence interval and heterogeneity for each cognitive
Author N (E/C) Age BMI AHI CPAP usage
(y) (kg/m2) (events/h) (h/night) Cognitive domain k N d 90% CI Q p (Q) I2 Fail-safe
Barbé et al.29
54 (20/25) 53.1 29.0 55.4 4.5b N
Bardwell et al.30 36 (20/16) 47.4 31.2 51.2 5.4b Processing speed 10 466 0.10 0.03 to 0.22 9.87 0.36 9 e
Barnes et al.22 28 45.5 30.2 13.0 3.5b Attention 12 666 0.19* 0.08 to 0.31 2.30 1.00 0 13
Barnes et al.25 80 46.4 31.0 21.0 3.6b Vigilance 10 467 0.05 0.07 to 0.16 2.12 0.99 0 e
Engleman et al.26 32 49.0 33.0 28.0 3.4b Memory 9 425 0.09 0.03 to 0.22 6.56 0.58 0 e
Engleman et al.23 16 52.0 29.8 11.0 2.8 Working memory 6 362 0.18 0.34 to 0.02 2.41 0.79 0 e
Engleman et al.31 23 47.0 30.0 43.0 2.8 Verbal fluency 8 378 0.02 0.14 to 0.11 1.55 0.98 0 e
Engleman et al.24 34 44.0 30.0 10.0 2.8 Visuoconstruction 5 250 0.01 0.17 to 0.16 5.08 0.28 21 e
Gast et al.27 29 (17/12) 52.3 35.9 20.4 ea
k, number of studies; N, number of patients; d, mean effect size; CI, confidence
Lee et al.32 38 (21/17) 48.6 28.6 30.9 6.1b
interval; Q, within domain heterogeneity; p(Q), p-value for heterogeneity; I2,
Marshall et al.28 29 50.5 31.5 22.0 4.9b
percentage of heterogeneity due to study differences; fail-safe N, number of studies
Monasterio et al.33 125 (66/59) 53.5 29.4 58.1 4.8b
theoretically needed to yield the observed effect to non-significance; *p < 0.05.
Pelletier-Fleury et al.34 171 (82/89) 52.9 30.9 53.2 5.4b

N, number of subjects (E, in experimental condition/C, in control condition); BMI,

body mass index; AHI, apnea/hypopnea index pre-treatment; CPAP, continuous asymmetry toward positive effects for any of the outcome
positive airway pressure.
Usage was >5 h/night for 9 patients and 5 for 8 patients.
measures; thus there was no indication of publication bias in the
CPAP usage considered to be compliant (>4 h/night for >5 days/week). selection of studies included in the current analysis. Fig. 1 displays
the funnel plot for attention, the only cognitive domain in which
CPAP treatment had a significant, positive effect.
small effect on attention was observed in favor of CPAP treatment We calculated the fail-safe N value for attention, sleepiness and
(d ¼ 0.19, 90% CI 0.08e0.31, p ¼ 0.005). The Q-statistic and the I2 depression, the only measures that showed significant difference
index indicated that there was no significant heterogeneity for between CPAP treatment and control conditions. Thirteen studies
attention. There were no significant effects of CPAP in the other would be required to reduce the observed effect on attention to
cognitive domains. Unexpectedly, working memory worsened with a non-significant level. For objective and subjective sleepiness, 13
CPAP treatment, although not significantly (d ¼ 0.18, 90% CI 0.34 and 75 studies would be needed, respectively. Ten studies would be
to 0.02, p ¼ 0.006). sufficient to reach non-significance in depression. Although, all fail-
In an exploratory analysis, the effect sizes for the subtests of safe N values exceeded the number of included studies per outcome
attention were calculated. A small, significant treatment effect was measure (k), they are considered to be relatively small for attention,
found for the paced auditory serial addition task (PASAT; d ¼ 0.39, objective sleepiness and depression.
90% CI 0.24e0.55, p < 0.001) and the trail making test B (TMT B;
d ¼ 0.18, 90% CI 0.006e0.30, p ¼ 0.02). No significant effect of CPAP Discussion
treatment was observed for the trail making test A (TMT A; d ¼ 0.15,
90% CI 0.01e0.29, p ¼ 0.07) nor for the Stroop color-word test The current meta-analysis of neuropsychological functioning
(d ¼ 0.11, 90% CI 0.14 to 0.36, p ¼ 0.47). after CPAP showed a small improvement in cognitive functioning,
The mean weighted effect of CPAP treatment on sleepiness and specifically in the attention domain, in patients with OSA. Only the
mood was calculated from the data reported in the selected studies divided attention tasks PASAT and TMT B showed significant
(Table 5). Objective sleepiness was assessed in six studies and improvement, with the PASAT, the most demanding attention task,
showed a small, significant effect (d ¼ 0.30; 90% CI 0.16e0.44, revealing the largest effect. Within the selected studies a small
p < 0.001). The effect of subjective sleepiness calculated from decrease in the objective sleepiness after CPAP was observed.
eight studies was moderate (d ¼ 0.53; 90% CI 0.27e0.78, p < 0.001). Although a moderate improvement of subjective sleepiness and
The effect of CPAP on depression reached significance, showing a small decrease of depressive symptoms was found, these findings
a small effect (d ¼ 0.35; 90% CI 0.12e0.44, p ¼ 0.04). However, there are of limited value due to moderate heterogeneity.
was moderate statistical heterogeneity in the analyses for subjective The generally held clinical view is optimistic: CPAP improves
sleepiness and depression, limiting the value of these effect sizes. cognition substantially. Several qualitative reviews have reported
a wide range of positive effects of CPAP on cognition. The most
Publication bias frequently reported improvement is in the domain of attention and
vigilance, but changes in the domains of memory and executive
We conducted funnel plot analyses for all cognitive domains,
measures of sleepiness and mood. We did not observe any Table 5
Mean weighted effect sizes, confidence interval and heterogeneity for measures of
sleepiness and mood.
Table 3
k N d 90% CI Q p (Q) I2 Fail-safe
Summary of demographic and clinical characteristics of studies included in the
Variables M SD Range N Objective 6 214 0.30* 0.16 to 0.44 3.49 0.63 0 13
Sample size 54.2 43.7 16e172 13 Subjective 8 478 0.53* 0.27 to 0.78 22.60 0.01 56 75
Age (y) 50.5 6.8 44.0e53.5 13
BMI (kg/m2) 30.4 3.8 28.6e38.2 13 Mood
AHI (events/h) 34.6 11.5 10.0e55.5 13 Depression 5 102 0.35* 0.07 to 0.64 11.75 0.01 66 10
CPAP usage (h/night) 4.5 1.4 2.8e6.1 12 Anxiety 5 102 0.10 0.16 to 0.36 7.10 0.13 43 e
Treatment duration (weeks) 8.2 e 1e13 13
k, number of studies; N, number of patients; d, mean effect size; CI, confidence
M, mean; SD, standard deviation; N, number of studies in which the variable was interval; Q, within domain heterogeneity; p (Q), p-value for heterogeneity; I2,
reported; BMI, body mass index; AHI, apnea/hypopnea index pre-treatment; CPAP, percentage of heterogeneity due to study differences; fail-safe N, number of studies
continuous positive airway pressure. theoretically needed to yield the observed effect to non-significance; *p < 0.05.
W.A. Kylstra et al. / Sleep Medicine Reviews 17 (2013) 341e347 345

Fig. 1. Funnel plot for attention.

functioning have also been noted.1,11,12 However, the minimal determine whether CPAP improves sleepiness.8e10 These meta-
treatment effect observed in the current, quantitative meta- analyses showed that subjective sleepiness (ESS) as well as objec-
analysis does not support widely held clinical beliefs or the rela- tive sleepiness measured by the MWT were significantly reduced by
tively positive conclusions drawn in previous reviews. The differ- CPAP. Possibly, the minimal treatment effect we found on cognition
ence between our findings and those of previous reviews is might be due to a smaller improvement in sleepiness and mood
probably due to less stringent criteria for study inclusion and the symptoms than is generally observed. In order to test this hypoth-
absence of appropriate weighing of the study effects. As a result of esis, we compared the observed effect of CPAP treatment on
the latter, small positive effects may be overestimated, while sleepiness and mood symptoms to the findings of the meta-analysis
negative results may be underexposed. by Giles et al.8 They reported significant improvement in subjective
A drawback of administering stringent inclusion criteria in the sleepiness (ESS; d ¼ 0.37, 90% CI 0.32e0.42, p < 0.001), while
current meta-analysis is the loss of possibly valuable information objective sleepiness improved partially (MWT; d ¼ 0.19, 90% CI
from studies that compare CPAP with other treatments. Examples 0.05e0.33, p ¼ 0.02 and MSLT; d ¼ 0.11, 90% CI 0.07 to 0.29,
of such treatments are dental devices, postural therapy and other p ¼ 0.31). The available data in our study, showing overlap with
positive airway therapies. Other studies that were excluded aim to Giles et al., showed a clear but modest improvement of CPAP
improve CPAP compliance using a behavioral approach compared treatment in objective as well as subjective sleepiness (Table 5). Our
to the regular administration of CPAP therapy. findings are comparable to those of the more extensive meta-
It seems plausible to assume that improvement of cognitive analysis by Giles et al. In our meta-analysis, we observed a decline
function after CPAP treatment will be most pronounced in the in depressive symptoms after CPAP treatment, but no change in
domains affected by OSA. One could question whether the present anxiety level (HADS). In a smaller study sample, Giles et al. found
results are in line with this expectation. In a meta-analysis of neu- a comparable small effect on depression favoring CPAP (d ¼ 0.20).
ropsychological effects of untreated OSA, Beebe et al.6 reported When corrected for heterogeneity, the effect was reduced to a non-
a substantial negative impact upon vigilance, executive functioning significant level. Considering the similar improvement on sleepi-
(EF) and motor coordination. Vigilance and motor coordination are ness and mood symptoms compared to prior findings, the minimal
relatively one-dimensional concepts, whereas EF is an umbrella treatment effect on cognition in our meta-analysis is unlikely to be
concept, composed of a number of cognitive functions, including due to a lack of improvement in sleepiness or mood.
working memory, mental flexibility, planning, problem-solving, There are some methodological limitations to this meta-analysis
inhibition and verbal fluency. In our study, EF was divided in the that could have influenced our findings. First, we used uncorrected
subdomains of attention, working memory and verbal fluency in test scores. In clinical practice, test results are often corrected
order to evaluate the treatment effect more precisely. Our explor- demographically using normative data,14 because test performance
atory analysis of the attention domain showed improvement on the is strongly influenced by age and level of education. Despite the use
PASATand TMT B after CPAP treatment. Both tests are included within of uncorrected data, the possible influence of age was considered to
the cognitive domain of EF according to Beebe et al. Unexpectedly, be negligible as all selected studies included middle-aged patient
a small negative effect of CPAP on working memory was seen, albeit samples. The effect of level of education could not be included in
not significant. There was no notable effect on verbal fluency. our analyses, because only two studies provided these data.27,29 It is
Although Beebe et al. did not report which tests accounted for the conceivable that larger effect sizes would have emerged if studies
significant deterioration of EF in OSA, our findings at least partly had used demographically corrected test scores, because this
support the assumption that the most affected domains improve correction removes sources of variance that are irrelevant to the
most with treatment. In addition, despite similar operationalization treatment effect.
of the vigilance domain in both Beebe et al.’s and our studies, we did Second, no moderator analysis was conducted, because it was
not find significant improvement after CPAP treatment. No statement considered statistically inappropriate due to the limited number of
on the treatment effect on motor functioning can be made, as it was studies included in the meta-analysis per domain and the minimal
not investigated in the studies included in our meta-analysis. treatment effect found. Nevertheless, the severity of OSA, treat-
Daytime sleepiness is one of the most frequently reported ment compliance and treatment duration could have moderated
complaints of patients with OSA and is expected to improve with the effect of CPAP treatment on cognition, as these factors differed
CPAP treatment. Several meta-analyses have been performed to substantially across studies. To inspect whether studies with mild
346 W.A. Kylstra et al. / Sleep Medicine Reviews 17 (2013) 341e347

OSA and low compliance diminished the effects, we recalculated Conflict of interest
effect sizes of cognition for studies with moderate to severe
OSA (AHI15 events/h, k ¼ 10) and for studies with sufficient There is no conflict of interest.
compliance (>4 h/night for >5 days, k ¼ 9). Neither recalculation
had a significant influence on the observed effect sizes. Appendix 1
Third, we used a conservative approach to calculate effect sizes
for three of the cross-over studies.23,24,26 This may have resulted in
a reduced contribution of these studies to the overall results. An Domain Test k
additional analysis, excluding these three studies, did not yield Processing speed 10
different effect sizes for the attention domain. This suggests that Digit symbol substitution (WAIS) 9
the relatively modest treatment effect we found on attention was Symbol search (WAIS) 1
8-choice reaction time 3
not due to our conservative approach of data analysis.
Cross-over studies have the advantage over parallel studies of Attention 12
reducing the influence of confounding covariates. However, a carry- Trail making test A/B 12
over effect can distort the results in a cross-over study. One strategy Stroop color-word 5
for minimizing carry-over effects is to include a washout period Paced auditory serial addition task 8
between treatment arms. Only two of the seven cross-over studies
Vigilance 10
in our meta-analysis included a washout period (both of two
Psychomotor vigilance task 2
weeks).25,28 Yet, in all of the cross-over studies a statistical analysis Steer clear 6
was conducted to check whether a significant interaction of treat- Rapid visual information processing 3
ment had occurred. Only one study reported a significant carry- Digit vigilance 2
over effect on a cognitive outcome measure, resulting in adjust-
Memory 9
ment of the analysis.26 Thus, because the cross-over studies
Verbal paired associated (WMS) 2
included in our analysis either employed a design to minimize California verbal learning test 1
possible carry-over effects, or adjusted for such possible effects Word pair memory recall 2
statistically, we have no reason to believe that this design issue had Hopkins verbal learning test 1
an effect on our results. Logical memory 1&2 (RBMT) 1
Visual reproduction (WMS) 1
In conclusion, in this meta-analysis we observed only a minimal
Benton-R visual retention test 3
effect of CPAP treatment on cognition in patients with OSA. The only Brief visuospatial memory test 1
cognitive domain in which a treatment effect was found was
attention, and the improvement seen was modest. Future studies Working memory 6
should investigate the possible moderating effect of OSA severity, Digit span (WAIS) 6
Letter-number sequencing (WAIS) 1
daytime sleepiness, compliance, and treatment duration on study
Digit ordering test 1
outcomes, preferably in the context of a randomized controlled trial. Mental control (WMS) 2

Verbal fluency 8
Verbal fluency 1
Category fluency 1
Practice points COWAT letter fluency 2
Borkowski letter fluency 3
 The results of this meta-analysis indicate that CPAP has
a small, positive effect on attention, and no effect on Visuoconstructive skills 5
Block design (WAIS) 5
other cognitive domains. These results do not support
the conclusion drawn in previous qualitative reviews k, number of studies.
that CPAP improves functioning in several cognitive
 In accordance with previous quantitative reviews References
sleepiness and mood moderately improve after CPAP.
*1. Aloia MS, Arnedt JT, Davis JD, Riggs RL, Byrd D. Neuropsychological sequelae of
obstructive sleep apnea-hypopnea syndrome: a critical review. J Int Neuro-
psychol Soc 2004;10:772e85.
2. Yaggi H, Mohsenin V. Obstructive sleep apnoea and stroke. Lancet Neurol
3. Parish JM, Somers VK. Obstructive sleep apnea and cardiovascular disease.
Research agenda Mayo Clin Proc 2004;79:1036e46.
4. Alchanatis M, Zias N, Deligiorgis N, Liappas I, Chroneou A, Soldatos C, et al.
Comparison of cognitive performance among different age groups in patients
 Additional RCTs are required to quantify the influence with obstructive sleep apnea. Sleep Breath 2008;12:17e24.
of OSA severity, treatment duration and compliance 5. Engleman HM, Kingshott RN, Martin SE, Douglas NJ. Cognitive function in the
with CPAP treatment on cognitive functioning. sleep apnea/hypopnea syndrome (SAHS). Sleep 2000;23:S102e8.
 Studies should follow the CONSORT guidelines for *6. Beebe D, Groesz L, Wells C, Nichols A, McGee K. The neuropsychological effects
standardized reporting of randomized trials to improve of obstructive sleep apnea: a meta-analysis of norm-referenced and case-
controlled data. Sleep 2003;26:298e307.
their methodology and to augment their suitability for
7. Lamphere J, Roehrs T, Zorick F, Conway W, Roth T. Recovery of alertness after
meta-analyses. CPAP in apnea. Chest 1989;96:1364e7.
 Neuropsychological test batteries should include *8. Giles TL, Lasserson TJ, Smith BH, White J, Wright J, Cates CJ. Continuous
demanding attention tasks, such as PASAT and TMT B. positive airways pressure for obstructive sleep apnoea in adults. Cochrane Lib
 Further research of the effect of CPAP on cognition in 2006;4:1e107.
the elderly and neurological patients with OSA (e.g.,
patients with dementia or stroke) is required.
* The most important references are denoted by an asterisk.
W.A. Kylstra et al. / Sleep Medicine Reviews 17 (2013) 341e347 347

*9. Marshall NS, Barnes M, Travier N, Campbell AJ, Pierce RJ, McEvoy RD, et al. 24. Engleman HM, Kingshott RN, Wraith PK, Mackay TW, Deary IJ, Douglas NJ.
Continuous positive airway pressure reduces daytime sleepiness in mild to Randomized placebo-controlled crossover trial of continuous positive airway
moderate obstructive sleep apnoea: a meta-analysis. Thorax 2006;61:430e4. pressure for mild sleep apnea/hypopnea syndrome. Am J Respir Crit Care Med
*10. McDaid C, Griffin S, Weatherly H, Duree K, van der BM, van HS, et al. 1999;159:461e7.
Continuous positive airway pressure devices for the treatment of obstructive 25. Barnes M, McEvoy RD, Banks S, Tarquinio N, Murray CG, Vowles N, et al.
sleep apnoea-hypopnoea syndrome: a systematic review and economic Efficacy of positive airway pressure and oral appliance in mild to moderate
analysis. Hlth Technol Assess 2009;13:1e162. obstructive sleep apnea. Am J Respir Crit Care Med 2004;170:565e664.
*11. Weaver TE, Chasens ER. Continuous positive airway pressure treatment for 26. Engleman HM, Martin SE, Deary IJ, Douglas NJ. Effect of continuous positive
sleep apnea in older adults. Sleep Med Rev 2007;11:99e111. airway pressure treatment on daytime function in sleep apnoea/hypopnoea
*12. Sánchez AI, Martínez P, Miró E, Bardwell WA, Buela-Casal G. CPAP and syndrome. Lancet 1994;343:572e5.
behavioral therapies in patients with obstructive sleep apnea: effects on 27. Gast H, Schwalen S, Ringendahl H, Jorg J, Hirshkowitz M. Sleep-related
daytime sleepiness, mood, and cognitive function. Sleep Med Rev 2009;13: breathing disorders and continuous positive airway pressure-related changes
223e33. in cognition. Sleep Med Clin 2006;1:499e511.
13. Jadad A, Moore R, Carroll D, Jenkinson C, Reynolds D, Gavaghan D, et al. 28. Marshall NS, Neill AM, Campbell AJ, Sheppard DS. Randomised controlled
Assessing the quality of reports of randomized clinical trials: is blinding crossover trial of humidified continuous positive airway pressure in mild
necessary? Control Clin Trials 1996;17:1e12. obstructive sleep apnoea. Thorax 2005;60:427e32.
14. Lezak M. In: Neuropsychological assessment. 2nd ed. New York: Oxford 29. Barbe F, Mayoralas LR, Duran J, Masa JF, Maimo A, Montserrat JM, et al.
University Press; 2004. Treatment with continuous positive airway pressure is not effective in
15. Strauss E, Sherman E, Spreen O. A compendium of neuropsychological tests: patients with sleep apnea but no daytime sleepiness: a randomized,
administration, norms and, commentary. New York: Oxford University Press; 2006. controlled trial. Ann Int Med 2001;134:1015e23.
16. World Health Organization. Obesity: preventing and managing the global 30. Bardwell WA, Ancoli-Israel S, Berry CC, Dimsdale JE. Neuropsychological
epidemic. Report of a WHO consultation. WHO; 2000. effects of one-week continuous positive airway pressure treatment in patients
17. American Academy of Sleep Medicine. International classification of sleep with obstructive sleep apnea: a placebo-controlled study. Psycho Med
disorders. In: Diagnostic and coding manual. 2nd ed. Westchester: ASSM; 2005. 2001;63:579e84.
18. Kribbs NB, Pack AI, Kline LR, Getsy JE, Schuett JS, Henry JN, et al. Effects of one 31. Engleman HM, Martin SE, Kingshott RN, Mackay TW, Deary IJ, Douglas NJ.
night without nasal CPAP treatment on sleep and sleepiness in patients with Randomised placebo controlled trial of daytime function after continuous
obstructive sleep apnea. Am Rev Resp Dis 1993;147:1162e8. positive airway pressure (CPAP) therapy for the sleep apnoea/hypopnoea
19. Cohen J. In: Statistical power analysis for the behavioral sciences. 2nd ed. New syndrome. Thorax 1998;53:341e5.
York: Academic Press; 1988. 32. Lee I, Bardwell WA, Kamat R, Tomfohr L, Heaton RK, Ancoli-Israel S, et al.
20. Borenstein M, Hedges L, Higgins J, Rothstein H. Introduction to meta-analysis. A model for studying neuropsychological effects of sleep intervention: the
Hoboken, NJ: John Wiley & Sons Inc; 2009. effect of three-week continuous positive airway pressure treatment. Drug Dis
21. Orwin RG. A fail-safe N for effect size in meta-analysis. J Educ Stat 1983;8: Today Dis Mod 2011;8:147e54.
157e9. 33. Monasterio C, Vidal S, Duran J, Ferrer M, Carmona C, Barbe F, et al. Effec-
22. Barnes M, Houston D, Worsnop CJ, Neill AM, Mykytyn IJ, Kay A, et al. tiveness of continuous positive airway pressure in mild sleep apnea-hypopnea
A randomized controlled trial of continuous positive airway pressure in mild syndrome. Am J Respir Crit Care Med 2001;164:939e43.
obstructive sleep apnea. Am J Respir Crit Care Med 2002;165:773e80. 34. Pelletier-Fleury N, Meslier N, Gagnadoux F, Person C, Rakotonanahary D,
23. Engleman HM, Martin SE, Deary IJ, Douglas NJ. Effect of CPAP therapy on Ouksel H, et al. Economic arguments for the immediate management of
daytime function in patients with mild sleep apnoea/hypopnoea syndrome. moderate-to-severe obstructive sleep apnoea syndrome. Eur Resp J 2004;23:
Thorax 1997;52:114e9. 53e60.