Neurocognitive function in obstructive sleep apnoea: A meta-review



    Corresponding author
    1. School of Psychology
      Romola S. Bucks, M304, School of Psychology, University of Western Australia, 35 Stirling Hwy, Crawley, WA 6009, Australia. Email:
    Search for more papers by this author

    1. School of Psychology
    Search for more papers by this author

    1. Centre for Sleep Science, School of Anatomy, Physiology and Human Biology, The University of Western Australia
    2. West Australian Sleep Disorders Research Institute, Sir Charles Gairdner Hospital, Perth, Western Australia, Australia
    Search for more papers by this author

  • The Authors: Romola Bucks (MSc. Clinical Psychology and PhD) is a Professor in the School of Psychology at the University of Western Australia and has research interests in the fields of neuropsychology and disorders of ageing, including Parkinson's disease, diabetes and sleep-disordered breathing, in particular the cognitive deficits found in such individuals and their relationship to the development of dementia. Michelle Olaithe is a PhD candidate in her final year at the School of Psychology, University of Western Australia, co-supervised by Professor Romola Bucks and Professor Peter Eastwood, with interests in the fields of sleep, neuropsychology and mental health, in particular the effective measurement of cognitive dysfunction due to sleep disorders. Professor Peter Eastwood (PhD) holds appointments as a National Health and Medical Research Council Senior Research Fellow at the West Australian Sleep Disorders Research Institute at Sir Charles Gairdner Hospital, Winthrop Professor at the University of Western Australia and has research interests in the fields of respiratory and sleep physiology, in particular the pathophysiology of upper airway dysfunction in individuals with sleep-disordered breathing.


Romola S. Bucks, M304, School of Psychology, University of Western Australia, 35 Stirling Hwy, Crawley, WA 6009, Australia. Email:


Adult obstructive sleep apnoea (OSA) is associated with cognitive dysfunction. While many review articles have attempted to summarize the evidence for this association, it remains difficult to determine which domains of cognition are affected by OSA. This is because of marked differences in the nature of these reviews (e.g. many are unsystematic) and the many different tasks and domains assessed. This paper addresses this issue by comparing the results of only systematic reviews or meta-analyses assessing the effects of OSA on cognition, the relationship between OSA severity and cognition, and/or the effects of treatment on cognition in OSA. Electronic databases and hand-searching were undertaken to select reviews that reported on these areas. We found 33 reviews; five reviews met predetermined, stringent selection criteria. The majority of reviews supported deficits in attention/vigilance, delayed long-term visual and verbal memory, visuospatial/constructional abilities, and executive function in individuals with OSA. There is also general agreement that language ability and psychomotor function are unaffected by OSA. Data are equivocal for the effects of OSA on working memory, short-term memory and global cognitive functioning. Attention/vigilance dysfunction appears to be associated with sleep fragmentation and global cognitive function with hypoxaemia. Continuous positive airway pressure treatment of OSA appears to improve executive dysfunction, delayed long-term verbal and visual memory, attention/vigilance and global cognitive functioning. In order to improve our understanding of cognitive dysfunction in OSA, future research should pay particular attention to participant characteristics, measures of disease severity and choice of neuropsychological tests.


intelligence quotient


obstructive sleep apnoea


Obstructive sleep apnoea (OSA) is a common disorder of breathing during sleep1,2 characterized by repeated obstruction of the upper airway. These obstructive events often result in reduced blood oxygen saturation (hypoxaemia), increased blood carbon dioxide (hypercapnia) and increased sympathetic nervous system activity. Resolution of airway obstruction is accompanied by arousal from sleep.3,4 A consequence of these repetitive obstructions and arousals is that normal sleep architecture is disturbed and sleep-fragmented, particularly the time spent in rapid eye movement sleep and slow-wave sleep (stages 3/4), which are often decreased.5

OSA has been associated with a broad range of psychological problems, of which depression6–8 and neurocognitive difficulties, particularly in memory and new learning,9 attention10 and executive function,9,11–13 are the most widely reported. The prevailing, current view is that the neurocognitive impairment seen in OSA is due to the adverse effects of sleep fragmentation and/or intermittent hypoxia.

Sleep fragmentation and sleep deprivation, and the associated excessive daytime sleepiness, have been proposed as mechanisms underlying cognitive impairment in OSA via their impact on attention.3,14,15 The basis of this argument is the strong similarity between cognitive deficits seen in OSA and those seen in healthy individuals who have been experimentally deprived of sleep.14 Such sleep-deprived individuals show increased daytime sleepiness and reduced activity in the prefrontal and posterior parietal cortices and in the thalamus.16–18 These functional, central, neural changes have been associated with reductions in attention and vigilance.19 Treatment of OSA reverses the daytime sleepiness experienced in OSA patients;20 however, it remains unclear whether cognitive deficits also return to normal.

Sleep fragmentation might also mediate the cognitive deficits seen in OSA via dysfunction in neural networks, especially in the frontal lobes.12 The basis of this hypothesis is that sleep disruption reduces the efficacy of restorative processes in the prefrontal cortex leading to cellular and biochemical stress.21–23 These stresses, in turn, disrupt functional homeostasis, altering glial and neuronal viability. Sleep fragmentation, as seen in OSA, contributes to neurocognitive dysfunction, specifically decrements in attention,19 memory dysfunction24 and sleepiness.25 The underlying mechanisms are thought to be slowing of cognitive processing19,26 and/or disruption of the restorative processes of sleep.12,27

Intermittent hypoxia, as often accompanies obstructive events during sleep,8,28 could also be an important contributor to cognitive dysfunction in OSA.29–31 Experimental research and neuroimaging studies show effects of intermittent hypoxia on sleepiness, memory and executive dysfunction.12,27 In individuals with OSA, such changes have been attributed to reduced cell neurogenesis and density of the hippocampus,32–35 the frontal cortex36 and generalized grey matter.37

Many review articles have sought to collate and summarize the accumulating evidence regarding the cognitive deficits of adult OSA. Despite this, it remains difficult to determine which domains of cognition are affected by OSA. This is because of marked differences in the nature of previous reviews (e.g. many are unsystematic or methods insufficiently specified), and the many different tasks and domains used to report cognitive function. Our paper seeks to address this issue by comparing the results of only high-quality, systematic reviews and/or meta-analyses that assessed the effects of OSA on cognition and its domains. In this sense, our paper provides a ‘meta-review’ of the current evidence for cognitive dysfunction in OSA, the relationship between OSA severity and cognitive dysfunction, and the effects of treating OSA on cognitive function.


Selection criteria

For the present paper, we considered all review articles that reported summaries for comparisons of cognitive ability in adults with OSA either to controls or to normative data, summarized the evidence relating to the relationship between OSA severity and cognitive performance, or summarized treatment response in OSA. Only English language articles were considered, and reviews could be either systematic or meta-analytic. A summary of the selection criteria, search strategy, study selection and data extraction are presented in Figure 1.

Figure 1.

Flow chart of search strategy and study selection.

Search strategy

The following electronic databases were searched: Medline 1946 to 15.3.2012, PsycInfo 1806 to March Week 1 2012, Trip Database to 15.3.2012, Cochrane Reviews, plus hand-searching of reference lists and analysis of all citations in seminal papers in the field.12,29,38–41

Search terms for Medline and PsycInfo are listed in Figure 1 and produced 20 and 85 papers, respectively. Preliminary searches revealed that adding the terms ‘treatment’ or ‘(C/A)PAP’ did not increase the number of papers found. Searching of Trip Database produced 11 additional articles, and hand-searching produced a further 31. Thus, a total of 147 papers were available for further analysis.

Study selection

Inspection of the titles and abstracts led to 114 papers being removed because they were not review papers, were not in English or were related to children. Of the remaining 33 papers, five focused on imaging, four did not have cognition as a primary focus, and one collapsed the cognitive outcomes into a single category combining them with mood measures.42 These papers were also removed.

Of the remaining 23 papers, 18 did not report the search strategy and were removed because they may have been selective reviews of the literature. For example, one review20 reported a systematic assessment of the effect of continuous positive airway pressure or behavioural therapies on cognitive function, searching Medline, PsycINFO and Cochrane Reviews (1994–2007) but failed to report the search terms used or the inclusion/exclusion criteria applied nor did it provide summaries within cognitive domains.

One systematic review43 and one meta-analysis44 were conducted by the same authors. Given that the meta-analysis was the more recent publication and that the search strategy yielded four more source papers, only the meta-analysis study was chosen for selection for the present review. This left four systematic reviews or meta-analyses.29,39,44,45 To this list, we added one recently accepted meta-analytic review,46 leaving a total of five eligible studies. A summary of the methodologies and main findings of these five studies is presented in Table 1. A more detailed version of Table 1 is presented online in Table S1 in the online supporting information.

Table 1. Descriptive details of the five eligible reviews chosen for this meta-review
AuthorCitationMethodology and main findings
  1. EF, executive function; OSA, obstructive sleep apnoea; OSAHS, OSA—hypopnoea syndrome; WCST, Wisconsin Card Sorting Test.

Fulda and Schulz44Sleep Medicine Reviews (2001), 5: 423–45 Studies: 24 studies between 1966 and 2000 were reviewed that explored cognitive dysfunction in people with sleep-related breathing disorders, insomnia or narcolepsy.
Findings: Impairments were found for driving ability, attention span, divided attention and sustained attention.
Beebe et al.29Sleep (2003), 26: 298–307 Studies: 25 articles between (no start date given)—2001
 All studies retrieved were pretreatment studies, and the authors reported both norm- and control-referenced comparisons.
Findings: Untreated OSA was found to have a negligible impact on intellectual and verbal functioning but a substantial impact upon vigilance and executive functioning.
Wallace and Bucks46Sleep (2012), submitted and in review Studies: 42 studies: control-referenced, 26 samples; norm-referenced, 39 samples from 1861 to 2011
Findings: Individuals with OSA are significantly impaired compared with norms and controls on verbal immediate and delayed recall measures. Visual memory appeared intact, while evidence of the effect of OSA on visuospatial memory, verbal learning and verbal recognition memory was inconclusive.
Aloia et al.39Journal of the International Neuropsychological Society (2004), 10: 772–85 Studies: 37 peer-reviewed articles between 1985 and 2002
 Studies meeting the inclusion criteria were then divided into three non-exclusive categories: pretreatment group comparison studies, treatment efficacy studies and correlational studies.
Findings: Findings were equivocal for most cognitive domains
 Treatment improves attention and vigilance
 Argue for the battery approach to study cognition in OSA
 Propose a potential neurofunctional theory to account for the aetiology of cognitive deficits in OSAHS.
Saunamäki and Jehkonen45Journal of Acta Neurologica Scandinavica (2007), 115: 1–11 Studies: 40 studies were included although most were used for demographic and descriptive purposes, only 13 articles comparing healthy controls with sleep apnoea were used to report EF dysfunction
Findings: Findings were inconclusive for many tests, however some demonstrated impairment (e.g. mazes) and others no impairment (e.g. Trails A).
Treatment did improve performance on certain EF tests (e.g. mazes, WCST preservative errors)
Argue for more stringent reporting of nocturnal, demographic and treatment information, and choice of neuropsychological tests

Data extraction

Data extracted from the final reviews included: author/s, period of review, databases searched, search terms used, exclusion criteria employed, the number of studies identified and retained, cognitive domains reported, effects reported within each cognitive domain, methodological issues identified by the review's authors, and methodological issues we noted with the review.

Overall effects were scored as positive (+) if they provided evidence of a deficit in OSA compared with controls or norms, or of an association with severity of OSA, or showed improvement with treatment. Effects were negative (–) if no such relationships were found. Cognitive function was reported within the following domains: executive function (working memory reported separately), attention/vigilance, short-term memory, immediate long-term memory (verbal, visual, visuospatial), delayed long-term memory (verbal, visual, visuospatial), language; visuospatial/constructional function, psychomotor function, and general cognitive functioning.

The nature of these domains is summarized later. Further information on how the data from each review were mapped onto these domains, and key points of discrepancy is available online in Table S2 in the online supporting information.

Executive function

Executive functions collectively manage other cognitive processes, including memory and attention, and are responsible for volition, planning, purposeful action and monitoring effective performance.47,48

Working memory

Working memory is often regarded as a specific type of executive function and involves actively holding information in one's mind, while dynamically manipulating this information for a task at hand. One such task of working memory is ‘digit span backwards’, which requires an individual to repeat a sequence of numbers in reverse order.47,48


Attention is an umbrella term referring to several capacities that enable an individual to become aware of, receptive to and concentrate on a particular stimulus, while ignoring other aspects of the environment.47,48 Vigilance refers to one's ability to maintain this alertness over time.47,48 Tasks assessing attention and vigilance typically gauge reaction times and responsiveness to stimuli or measure evoked potentials.49


Memory refers to an individual's capacity to receive information from the outside world, mentally store this information and recall this information to consciousness.47,48

Short-term memory

Short-term memory refers to the ability to recall a small amount of information within about 30 s.47,48 This system is functionally and neurologically separate from the ability to store and recall information for more than 30 s, which is supported by encoding into long-term memory stores.50,51

Long-term memory

Long-term memory refers to the acquisition and consolidation of new information, that is, learning.47,48 For the present review, this was further broken into:

Immediate long-term memory (verbal, visual and visuospatial), which reflects the requirement of an individual to recall information immediately (e.g. after a duration longer than 30 s), and;

Delayed long-term memory (verbal, visual and visuospatial), which reflects the requirement of an individual to recall information after some time had passed (e.g. after a duration of 20 min).


Language refers to the capacity to acquire, comprehend and produce complex symbolic systems for communication.47,48,52 Tests measure the capacity to produce and understand language under structured conditions, such as naming or verbal fluency.47,48

Visuospatial/constructional function

Visuospatial/constructional ability is the capacity to reconstruct a picture or object from an original.47,48,53 This domain is most often measured through copying a complex figure or reconstructing a model.53

Psychomotor function

Psychomotor function denotes the relationship between cognitive ability and physical function, and is demonstrated through the measurement of an individual's manual dexterity and coordination.47,48 Psychomotor function is most often measured by two hand coordination or reaction times.

General cognitive functioning

Cognitive functioning has historically been attributed to a single function, commonly labelled general cognitive ability or intelligence.47,48 Intelligence refers to an individual's general capacity for intellectual function and is most commonly measured through the summation of scores across a range of specific tests.47,48


The effects of OSA on cognition from each of the five selected reviews assessing each of seven cognitive domains are presented in Table 2. The majority of reviews reported deficits in attention/vigilance, executive function and in some subdomains of memory function. Language ability, delayed long-term visuospatial memory and psychomotor function were unaffected by OSA. Results were equivocal for global cognitive function, immediate long-term memory and working memory.

Table 2. Effects of obstructive sleep apnoea (OSA) on the domains of neurocognitive function
DomainSubdomainModalityOSA comparison studies (with norms or controls)Number of comparisons finding deficit in OSA (control or norm)Relation between cognition and OSA severityOSA treatment studies (pretreatment–post-treatment)
Meta-analysesSystematic reviewsSystematic reviewsSystematic reviews
Fulda and Schulz44Beebe et al.29Wallace and Bucks46Aloia et al.39Saunamäki and Jehkonen45Aloia et al.39Aloia et al.39Saunamäki and Jehkonen45
  • NA indicates that this domain was not assessed in the review; Summary indicates whether the reviewers concluded that, on balance, there was a significant effect in the domain being tested. N+, OSA significantly different from norms; C+, OSA significantly different from controls; N−, no difference between OSA and norms; C−, no difference between OSA and controls; S+, poorer cognitive performance in more severe OSA; S−, no relationship between cognitive performance and severity of OSA; T+, performance significantly improved with treatment; T−, performance did not improve with treatment.

  • † 

    The authors concluded that executive function was improved with treatment yet reported only 7 of 15 studies showing improvement; our view is the evidence is against improvement with treatment.

  • ‡ 

    In the opinion of the authors, this conclusion is problematic (refer to the text for individual explanations).

  • § 

    Aloia et al.39 did not separate out visual and verbal memory, combining them together in all calculations of effect size.

  • ¶ 

    There were only five reviews included in the present paper; however, this category included six comparisons as both Beebe et al.29 and Wallace and Bucks46 examined comparisons with norms and with controls separately.

Executive function  C+N+, C+NAC+C+5/5S− T− T+
Executive functionWorking memory C−NANANAC+1/2NANAT−
MemoryShort-term memory C−NANANAC+1/2NANAT−
Immediate long-term memoryVerbalC− N−, C−N+, C+NANA2/5NANANA
Visual C− N−, C+N+, C−NANA2/5NANANA
Visuo-spatialNANAN−, C+NANA1/2NANANA
Delayed long-term memoryVerbalC+N+, C−N−, C+ § C+NA 4/6 § S− § T+NA
Visual C+N−, C+NA § C+NA3/4 § S− § T+NA
Visuo-spatialNANAN−, C−NANA0/2NANANA
Attention/vigilance  C+C+NA C+NA3/3 S+ T+NA
Language  C+N+, C−NAC−C−2/5S−T−NA
Visuo-spatial/construction  C−N−, C+NAC+C+3/5S−T−T−
Psychomotor function  C−N−, C+NAC+C−2/5S−T−NA
Global cognitive functioning  C+N+, C−NA C−NA2/4 S+ T+NA

Only one systematic review assessed the relationship between OSA severity and cognitive performance.39 With the exception of attention/vigilance and global cognitive function, more severe OSA did not produce poorer cognition (executive function, delayed long-term visual and verbal memory language, visuospatial/construction, or psychomotor function) than milder OSA.

Two systematic reviews assessed the effects of treatment of OSA on cognition. Both reported improvements in executive functioning, while one reported improvements in delayed long-term memory and global cognitive function.39


Three of the five reviews assessed attention/vigilance function; all reported that OSA impairs performance relative to controls or norms.29,39,44 Only one review explored OSA severity and found a significant increase in impairment of attention/vigilance with increasing disease severity,39 although a potential problem was the use of a measure of attention that included measures of short-term and working memory via the digit span test (discussed later). This review also reported that sleep fragmentation was more closely related to attention/vigilance than was hypoxaemia.39 This was also the only review to explore OSA treatment effects and reported improvements in attention/vigilance.39

Executive function

Four reviews assessed whether executive function was impaired in individuals with OSA compared with controls or norms.29,39,44,45 All reported poorer executive function in OSA. Working memory was explored in a separable way by two of these reviews,44,45 with conflicting results. Only one review explored the effects of OSA severity and found no relationship between executive function and any nocturnal measure of disease severity.39 The two reviews that examined treatment effects both reported improvements in executive function, although these conclusions were problematic as the conclusion was based on a minority of studies in the case of the review by Aloia et al.,39 while the review by Saunamäki and Jehkonen45 combined tasks that might not be considered executive in nature (see later). Only one review examined the effect of treatment of OSA on working memory and found no effect.45


The effect of OSA on short-term memory was reported in a separable way by two reviews,44,45 of which only one reported impairments.45 Four reviews assessed verbal and visual delayed long-term memory in OSA,29,39,44,46 and found deficits in both. There were no consistent findings on the effect of OSA on immediate long-term memory (verbal, visual or visuospatial) or visuospatial delayed long-term memory. Only one review explored the effects of severity on memory39 and showed no effect. Only two reviews examined the effects of treatment of OSA on memory: one on short-term memory45 and the other on delayed long-term memory (visual and verbal combined).39 The former showed no effect of treatment, while the latter showed improvements with treatment.

Visuospatial/constructional ability

Visuospatial/constructional capacity was reviewed in five reviews (cf norms or controls) and appeared to be impaired in three of the reviews.29,39,45 There were no consistent relationships reported between disease severity and visuospatial/constructional capacity, and no effect of continuous positive airway pressure treatment.39,45

Language ability and psychomotor function

No consistent effects were found for OSA on language ability or psychomotor function. Of the five reviews, language ability was impaired in two29,44 and psychomotor function in two.29,39 Likewise, there was no clear link to disease severity or effect of treatment of OSA in either of these domains.

Global cognitive function

Contradictory findings were reported from the four reviews assessing this function on the effect of OSA on global cognitive function, with two of the four reporting deficits in OSA.29,44 Only one review explored the effects of OSA severity and of treatment of OSA on global cognitive function and found that global cognition might be more affected by hypoxaemia than other nocturnal disturbances (e.g. sleep fragmentation) and was improved with OSA treatment.39


This metareview provides support for the notion of OSA being associated with deficits in attention/vigilance, delayed verbal and visual long-term memory, visuospatial/constructional abilities, and executive function. OSA does not appear to result in deficits in language ability or psychomotor function. Current results are equivocal for the effects of OSA on working memory, short-term memory and global cognitive functioning. The current metareview also supports the presence of relationships between attention/vigilance dysfunction and sleep fragmentation, and between hypoxaemia and global cognitive function. Furthermore, treatment of OSA with continuous positive airway pressure appears to improve executive dysfunction, delayed long-term verbal and visual memory, attention/vigilance, and global cognitive functioning.

Cognitive function and OSA severity

Of the five reviews assessed in this paper, two examined the effect of OSA severity on cognitive function by separating those with OSA into a ‘mild/moderate’ group and a ‘severe’ group based on apnoea-hypopnoea index (apnoeas and hypopnoeas per hour of sleep) or apnoea index (apnoeas per hour of sleep).29,46 Compared with controls/norms, both reviews reported deficits in cognitive function in OSA irrespective of disease severity.29,46

Only one review directly explored the relationship between nocturnal measures of OSA severity and cognitive performance.39 The effect of OSA on attention/vigilance was found to be more strongly influenced by sleep fragmentation than hypoxaemia. This finding was based on the proposition that apnoea-hypopnoea index, respiratory disturbance index or apnoea index can be considered measures of sleep fragmentation, whereas nadir of blood oxygen saturation or time blood oxygen saturation ≤ 80% represent measures of hypoxaemia. While the latter proposal is commonly accepted, regarding apnoea-hypopnoea index, apnoea index or RDI as specific measures of sleep fragmentation, while convenient, is an oversimplification.

It remains unclear, which measures of oxygen desaturation or sleep fragmentation may be the best predictors of cognitive function.30,54,55 While most studies have used minimum oxyhaemoglobin saturation and cumulative time below 90% saturation when examining associations with cognitive function and excessive daytime sleepiness,56–58 Findley et al.10 and Alchanatis et al.59 suggest that mean blood oxygen saturation is better correlated with cognitive impairment. Despite considerable efforts to delineate the effects of sleep fragmentation and hypoxaemia on cognitive function in OSA, it remains unclear whether more severe OSA is associated with poorer cognition.

Methodological issues

Measuring cognitive performance

Measuring cognitive performance is extremely complex. Many different tests are available; each is capable of measuring a different aspect of cognition. The appropriateness of any given test should be supported by neurological or neuropsychological/theoretical evidence. This has not always been the case in studies examining neurocognitive function in OSA.

Classification of digit span tasks.  The present metareview supports the current view that attention/vigilance is impaired in OSA. However, the robustness of this finding is critically dependent on the measure used to define it. For example, two authors39,44 found a significant effect of OSA on attention/vigilance effects; however, their category of attention included short-term and working memory tasks (as measured using digit span forwards and backwards): our view is that these domains are separable. The digit span task requires participants to recall strings of digits of increasing length either in their original order or in reverse order. As such, digit span forwards tests assess short-term memory, and digit span backwards tests additionally assess working memory.48 Similarly, Saunamäki and Jehkonen45 grouped digit span forwards with digit span backwards, but they considered these to be both measures of executive function. Because both of these reviews reported individual study effects, it was possible to calculate the number of studies reporting positive effects within forwards and backwards digit span separately; these data were utilized in the current metareview (see Table 1).

A slightly different grouping problem of the digit span tasks appears in the review by Beebe et al.29 These authors grouped digit span tasks within a subdomain of memory: immediate recall measures (the ability to recall a list of words or a story immediately after hearing them). Memory theory proposes that the ability to recall a small amount of information within about 30 s is supported by a short-term memory store that is functionally and neurologically separate from the ability to store and recall information for more than 30 s, which is supported by encoding into long-term stores.50,51 In our view, the domain generated by Beebe et al.29 was problematic because it put together short-term/working memory task findings with immediate recall findings and thus may have confounded any true effects. It was not possible to separate these different effects for the purpose of the current review.

Classification of executive function tasks.  Another problematic grouping exists with the Trail Making test, which was classified as a ‘focused attention’ measure by Fulda and Schulz.44 The task is in two parts: part A involves joining numbers in sequence, and part B involves joining letters and numbers in alternating sequence as quickly as possible (1—A—2—B, etc.). This second part is generally regarded as an executive measure, given that it relies on abstraction, set-shifting, inhibition of previously joined numbers or letters, and working memory to recall where in each string (1, 2, 3 or A, B, C. . .) the person needs to go next.48 However, because the authors reported effects on Trails separately, these were considered within the Executive domain in the present review.

Combining tests that, in our view, are not largely executive in nature (e.g. Corsi block tapping, Trails A, Digit span forwards) was adopted in the review by Saunamäki and Jehkonen.45 For example, the Corsi block tapping task requires participants to mimic a researcher by tapping a sequence of up to nine blocks in order. This test is a spatial analogue of digit span forwards and assesses aspects of psychomotor speed, visuoconstructional skills and short-term memory. Thus, treating the Corsi task as an executive measure is problematic. Again, because individual test data were provided, the number of studies reporting executive effects could be calculated separately for the present review.

Classification of general cognitive function.  Aloia et al.39 grouped estimates of premorbid intelligence (intelligence quotient (IQ)) based on word-reading tasks, with other measures of global cognitive functioning, such as Wechsler Adult Intelligence Scale scores. While premorbid IQ measures are important in OSA, they should be seen as an index of cognitive reserve60 rather than as a measure of current ability. This is because the ability to read words is learned by early adulthood, correlates well with IQ and is relatively resistant to decline due to neurological disease.61 For this reason, word-reading tasks are used to estimate premorbid ability. Arguably, therefore, this domain contains tasks that should not be combined. However, there was no way to remove the contribution of premorbid IQ from the overall effects calculated. Thus, caution should be observed when interpreting such data because the addition of premorbid IQ estimates, which might not be sensitive to OSA, may increase the risk of a Type 2 error when comparing OSA with controls or norms.

Dealing with participant characteristics

A major challenge in collating the findings from many reviews of cognition and OSA are the potentially confounding effects of age, cognitive reserve and selection of controls.

Age.  Advancing age independently alters sleep architecture, and the prevalence and presentation of OSA.62 As individuals advance from middle to older age, they experience a decrease in slow-wave sleep, increased sleep fragmentation and more established sleep patterns independent of the presence of any sleep disorder.63 Matthieu et al.64 showed that cognitive functioning declines with age independent of OSA. As such, age is an independent risk factor for cognitive impairment in OSA.65 Despite this, the two reviews that considered age when comparing OSA and controls/norms reported that cognitive dysfunction did not increase with increasing age.29,46

Cognitive reserve.  Cognitive reserve is the concept that a high level of premorbid cognitive ability acts to ‘buffer’ the effect of neurocognitive trauma.66 None of the reviews assessed in this metareview evaluated the impact of cognitive reserve either directly or indirectly on cognitive function in OSA. This is because few of the original studies provided such data. Future studies are required, which compare individuals with OSA with controls matched for premorbid IQ.

Selection of controls.  When selecting control participants, a number of the reviews utilized results from studies that had recruited control participants who were likely to have had sleeping problems, for example non-apneoic snorers,41 or individuals with insomnia.67,68 Further, few original studies conducted sleep studies formally to determine the absence of OSA and instead selected individuals on the basis of screening questionnaires. Given that estimates of undiagnosed OSA are as high as 82%,2 this presents a problem both for control studies and for normative comparisons and, by extension, for meta-analyses. Any differences between OSA and a reference sample may be markedly diluted due to the presence of undiagnosed OSA in the control or norm samples. Ideally, future studies should be as stringent in their recruitment and screening of controls as for the selection of OSA patients.

Other methodological issues

Of the reviews examined in the present paper, only those of Wallace and Bucks46 and Saunamäki and Jehkonen45 were from the last 5 years. The other reviews examined in this paper were published between 2003 and 2004. As Wallace and Bucks46 exclusively examined the domain of memory and Saunamäki and Jehkonen45 exclusively examine executive function, evidence from other domains examined in this metareview may not capture the most recent and relevant primary investigations.


The overall picture of cognitive deficits in OSA is complex. On balance, there appear to be negative effects of OSA on cognition, most likely in the domains of attention/vigilance, verbal and visual delayed long-term memory, visuospatial/constructional abilities, and executive dysfunction. The severity of sleep fragmentation appears to be linked to deficits in attention/vigilance, while impairments in global cognitive function appear linked to hypoxia. Effective continuous positive airway pressure treatment appears to improve attention/vigilance and global cognitive function.