|Home | About | Journals | Submit | Contact Us | Français|
Rationale: Although cognitive deficits are well documented in patients with sleep apnea, the impact on memory remains unclear.
Objectives: To test the hypotheses that (1) patients with obstructive sleep apnea have memory impairment and (2) memory impairment is commensurate with disease severity.
Methods: Patients with obstructive sleep apnea and healthy volunteers (apnea–hypopnea index <5 events/h) completed a test battery specially designed to differentiate between aspects of memory (semantic, episodic, and working) versus attention. Sleepiness was measured on the basis of the Epworth Sleepiness Scale and Oxford Sleep Resistance test. Memory performance in patients versus control subjects was compared (Mann-Whitney U test; P < 0.01, Bonferroni corrected for multiple comparisons) and relationships between performance and disease severity were analyzed by linear regression.
Measurements and Main Results: Sixty patients and healthy control subjects matched for age (mean ± SD: patients, 51 ± 9 yr; control subjects, 50 ± 9 yr) and education (patients, 14 ± 3 yr; control subjects, 15 ± 3 yr) participated. Patients demonstrated impaired Logical Memory Test results (immediate recall: patients, median [range], 36 [9–69]; control subjects, 43 [19–64], P = 0.0004; and delayed recall: patients, 22 [6–42]; control subjects, 27 [10–46]; P = 0.0001). There were minimal differences in attention, visual episodic, semantic, or working memory; patients performed better than control subjects on Spatial Span forward and backward. Regression analysis revealed that Logical Memory Test performance was not significantly related to disease severity after controlling for age, education, and sleepiness.
Conclusions: Obstructive sleep apnea is associated with impairment in verbal, but not visual, memory. The impairment was present across a range of disease severity and was not explained by reduced attention. Such verbal memory impairment may affect daytime functioning and performance.
Patients with obstructive sleep apnea are known to have cognitive dysfunction; however, the scope of impairments in different aspects of memory is unclear.
Assessment based on a comprehensive test battery showed that patients with obstructive sleep apnea have specific impairments in verbal episodic memory, whereas visual episodic memory and semantic memory remain unaffected. The verbal memory performance was equivalent to normative data from healthy people 10 years older than our patients and was largely independent of changes in attention. The impact of these impairments on daytime function requires further investigation.
There is increasing evidence to suggest that obstructive sleep apnea (OSA) is associated with cognitive dysfunction (1–6). Cognitive function consists of many factors including perception, attention, language, comprehension, planning, problem solving, reasoning, learning, and memory. If these factors are impaired they are likely to have major public health implications, particularly memory impairments, which may lead to early cognitive decline (7). Although some studies have included memory as part of a wider cognitive assessment (1, 3, 5, 6, 8), a detailed examination of memory function in OSA is required (2, 9).
The complexity of memory function means that to capture the different aspects of memory that may be affected by OSA, study design needs to be carefully considered. It must control for attention deficits associated with excessive daytime sleepiness. Figure 1 shows different aspects of memory. Episodic memory is the memory of specific events associated with times and places, for example, the first day at school or work. Semantic memory is memory of information, or knowledge that is not linked to a specific time or event; for example, the reader may know that the capital of the United States is Washington DC, but it may not be clear when this information was learned.
Where memory has previously been examined in patients with OSA, the findings are inconsistent. Some studies have reported deficits in verbal (2, 3, 6, 8) and visual (1, 3) episodic memory, whereas others have failed to replicate these findings (1, 5, 10–13). Semantic memory impairment has also been reported (1, 8, 12), but again the findings are not consistent on all tests (1, 3, 11, 14). Reports of working memory impairment include maintenance and manipulation of information (2, 5) but not dual task performance (2). Finally, where procedural memory has been investigated, this too has produced inconsistent results (2, 15). Taken together these studies suggest that patients with OSA may have difficulties learning new information, using semantic cues to aid retrieval, and in using memory to guide behavior, such as when following directions.
The aim of the present study was to systematically investigate various aspects of memory and attention in a large group of patients with OSA and healthy volunteers. Specifically, we aimed to test the hypothesis that patients with OSA would show impairment in various aspects of memory compared with healthy volunteers, independent of any changes in attention. Our secondary hypothesis was that memory impairment would be associated with disease severity. Some of the results of these studies have been previously reported in the form of an abstract (16, 17).
Patients, aged 18 to 69 years, were sleep clinic referrals with suspected OSA. Exclusion criteria were as follows: concurrent sleep disorders; a history of neurological or cardiovascular disease; diabetes, or psychiatric illness including clinically diagnosed depression. Lifestyle factors precluding participation were shift work, excessive alcohol consumption (>35 units/wk), or a history of recreational drug use. Healthy volunteers were recruited from local advertising, and through colleagues who were naive to sleep research, and paid for their participation. All volunteers underwent polysomnography to ensure that they did not have undiagnosed OSA (apnea–hypopnea index [AHI] <5 events/h of sleep). Healthy volunteers were matched within 3 years of age of a selected patient from the OSA group. Likewise, volunteers that were within 2 years of education of a selected patient with OSA were recruited. One exception to this rule was a patient with 6 years of education; the closest match we could find was a volunteer with 10 years of education. The study was approved by the Royal Brompton and Harefield Research Ethics Committee and all participants gave written informed consent.
Patients and healthy volunteers completed a 3-hour cognitive testing battery (see online supplement) designed to examine various aspects of memory, including the Logical Memory Test, a verbal test assessing memory for a short story. The battery was conducted in a quiet laboratory between 15:00 and 17:00 in the afternoon before their polysomnography. Several attention tests were also included to account for the potential influence of attention deficit on memory performance (see Figure 1 and the online supplement). Participants were asked to abstain from caffeinated beverages for the duration of their stay. Sleepiness was assessed subjectively using the Epworth Sleepiness Scale (18) and objectively with the Oxford Sleep Resistance (OSLER) Test, which was performed at 09:00 after polysomnography (19). Mood was assessed using the Hospital Anxiety and Depression Scale (HADS) (20).
OSA was diagnosed polysomnographically ((SleepScreen; Hoechberg, Germany and Somnomedics; Tampa, FL); see the online supplement for details). Sleep and arousals were scored according to standard criteria (20–22). The AHI and 4% Oxygen Desaturation Index (ODI) were calculated as markers of disease severity (see the online supplement for scoring criteria).
Statistical analysis was performed with SPSS (Chicago, IL). Summary data are expressed as means ± the standard deviation (SD) or as median (range) when data were nonnormally distributed. Group comparisons (patients vs. control subjects) were made with one-way analysis of variance or Mann-Whitney U test for data that were not normally distributed; the threshold for statistical significance was P < 0.01, Bonferroni corrected for multiple comparisons (24 comparisons, P < 0.0004). For memory tasks that were found to be significantly impaired, a linear regression model was constructed to examine the relationship between functional impairment and disease severity (AHI), sleepiness (Epworth Sleepiness Scale), while controlling for the confounding factors of age and education (years). When a model was able to significantly predict performance on a test (P ≤ 0.05), the individual factors within the model were examined to establish which significantly predicted memory impairments.
Sample size calculations were based on data from a previous study that investigated cognitive function, but not the various aspects of memory (1). A minimal number of 58 patients with OSA and healthy control subjects would be needed to detect a significant difference for the Logical Memory Tests at a threshold P < 0.01 and 80% power (assuming a mean difference in scores of 11, SD 17). For the attentional task Digit Span forward, 48 patients with OSA and healthy control subjects would be needed (assuming a mean difference in scores of 0.5, SD 0.7).
A total of 138 participants completed the protocol, comprising 78 sleep clinic referrals and 60 healthy volunteers. Twelve of the sleep clinic referrals were found not to have OSA (AHI <5 events/h) and were excluded from the analysis. A further six patients were excluded for other reasons: regular heavy alcohol consumption undisclosed during recruitment (n = 4), consumption of alcohol before completion of the testing battery (n = 1), and chronic obstructive pulmonary disease (n = 1). Therefore analysis was performed on 60 patients with OSA and 60 healthy volunteers. Ten of the healthy volunteers were unable to perform the memory tests on the same day as the polysomnography because of time constraints (e.g., unable to leave work). Eight of these people underwent the memory tests after polysomnography, and two performed the tests before polysomnography. The duration of time between the tests and the polysomnography was 25 (range, 2–64) days.
Demographic data for the OSA patient and healthy volunteer groups are presented in Table 1. There were no differences in age, years of education, or caffeine or alcohol consumption between the patient and healthy volunteer groups. The patients with OSA were significantly heavier, smoked more, and had higher HADS scores for possible depression and anxiety compared with the healthy volunteers.
Polysomnographic parameters are presented in Table 2. Patients with OSA had higher levels of intermittent hypoxia (ODI) compared with healthy volunteers. Patients with OSA also reported more subjective daytime sleepiness than did the healthy volunteers, although neither group was found to be objectively sleepy, maintaining wakefulness for the maximal time of 40 minutes on the OSLER Test. Patients with OSA and healthy control subjects had similar levels of sleep efficiency, although the patients were found to have more stage 1 NREM sleep, and the arousal index was higher in patients with OSA.
Group median performances of patients with OSA and healthy volunteers on each of the memory tests are given in Table 3. Investigation of episodic memory performance revealed a deficit in immediate and delayed recall from the Logical Memory Test, but normal recognition memory and retention of information over time on this test, suggesting that patients with OSA have difficulty assimilating information, but do not forget learned information more readily than healthy volunteers. That is, they have a reduced capacity to acquire new information but no difficulty in retaining previously learned memories.
Examination of the semantic memory performance revealed that patients with OSA tended to perform less well on the two tests of verbal fluency (Semantic Fluency and Phonemic Fluency), but these did not meet our significance threshold adjusted for multiple comparisons. Working memory and attention appeared to be unaffected in patients with OSA, with the patients performing equally as well as the healthy volunteers on four of the five tests of attention, and significantly better than the healthy volunteers on the Spatial Span test backward and forward. Spatial Span forward is an attention task, whereas Spatial Span backward is considered a working memory task as it requires holding information in memory. This involves attention, but the two tests are considered to map onto different frontal lobe circuits. The good performance of the patients with OSA on the spatial tasks is consistent with the patients with OSA achieving similar scores on the Paired Associate Learning test, another spatial task.
We constructed linear regression models to investigate the relationship between verbal episodic memory performance (immediate and delayed recall from Logical Memory, respectively) and measures of disease severity (AHI, Epworth Sleepiness Scale score) in the patients with OSA, while controlling for age and education. The proportion of the overall variance explained by these models was 13% (immediate recall) and 13% (delayed recall). Examination of the individual components of the models revealed that years of education significantly predicted performance on both immediate (β-coefficient, 0.28; P = 0.002) and delayed (β-coefficient, 0.23, P = 0.008) recall from logical memory. AHI and sleepiness did not contribute significantly to the model.
We also investigated whether the cognitive tests were correlated with one another. Not surprisingly, several tests such as Semantic Fluency and Phonemic Fluency were related, as were Semantic Fluency and Graded Naming, and Spatial Span forward and backward (see the online supplement for details of the analysis).
The patients with OSA had higher scores on the HADS. Seven patients (12%) had a score greater than 11, although none reported any depressive symptoms. Therefore, analysis was undertaken to investigate whether the high HADS scores were associated with memory impairment. A cutoff of 8 was used to reorganize the patients with OSA as follows: possibly mildly depressed patients (n = 18; mean [SD], 10 [2.0] HADS) and nondepressed patients (n = 42; 3.7 [2.2] HADS). The groups were matched for OSA severity (AHI), body mass index, years of education, and caffeine and alcohol intake. However, the nondepressed group was significantly older than the mildly depressed group. No significant differences were found between any of the memory scores on any of the test battery in the mildly depressed versus nondepressed patients with OSA (see the online supplement for details of the analysis).
The main finding of our study was that patients with OSA, compared with a matched sample of control subjects, displayed reduced performance on verbal episodic memory tasks, whereas visual episodic, semantic, and working memory remained intact. The absence of any impairment in attention precludes the notion that these findings were due to a general attention deficit. Further, no relationship between memory impairment and OSA severity was found, indicating that memory impairment may also be observed in patients with mild OSA. We suggest that patients with OSA have specific difficulties assimilating and later recalling information presented to them verbally, but have no problems in recalling visual information, and these deficits do not show further decline with increasing disease severity.
Using a specially designed comprehensive test battery we were able to distinguish between verbal and visual memory impairment as well as between different types of memory, including episodic versus semantic and short-term (working) memory. To our knowledge, only one other study has employed a similar approach (2). However, this study by Naegele and colleagues included just one test of verbal episodic memory; consequently they were unable to conclude whether these impairments were specific to verbal memory, or whether patients also had deficits in visual episodic memory. Consistent with the study by Naegele and colleagues, we found that patients with OSA had specific difficulties in retrieval from verbal episodic memory, without any associated deficits in recognition memory or difficulties in retaining information over a short period of time. Normative data from standardized Z scores provided with the test suggest that the OSA patient group had verbal memory performance equivalent to that of healthy volunteers 10 years older than themselves (23).
Patients with OSA did not show a visual episodic memory deficit, despite having specific verbal episodic memory impairments. To our knowledge, only the Rey-Osterrieth Complex Figure Test has been previously used with patients with OSA to test visual episodic memory (1, 14). Some of these studies have reported deficits in immediate and delayed recall (14), which may be explained by a failure to adequately encode the material to allow a representation to form in memory; reduced performance has also been reported in copying the figure, for example (1, 14). Our study controlled for this encoding issue by expressing the recall score as a percentage of the copy score. Therefore, we suggest that patients with OSA have intact visual episodic memory performance after controlling for encoding.
Our test battery allowed us to examine the two processes considered to be important for working memory: those associated with maintenance of information in short-term memory, and those that allow successful division of attention between two tasks simultaneously. The patients performed similarly, or better than healthy volunteers on both of these processes, indicating that OSA was not associated with working memory deficits in our patient group. Although our findings of normal dual task performance are consistent with Naegele and colleagues (2), our failure to find deficits in maintenance of working memory is in contrast to the work of others (2, 3, 6). We suggest that a basic attention deficit could explain poor performance on this aspect of working memory (24). Indeed, this may explain the discrepancy between our findings and those of others, because our patients did not have any overt attentional deficits.
Our second aim was to examine the relationship between severity of OSA and memory function. Despite the finding of significant verbal episodic memory defects in patients with OSA, we were unable to detect a relationship with disease severity measures (AHI). However, our data are consistent with other studies that have failed to find a significant relationship between AHI and memory dysfunction (13). Importantly, this means that even mild OSA (AHI ≥5) may be sufficient to produce the memory deficits.
Verbal episodic memory impairment has been shown to be associated with distinct neural correlates located bilaterally in the prefrontal cortex and in the left hippocampus (25). Both of these areas are vulnerable to the effects of intermittent hypoxia (26, 27) and some (28–30) but not all (31) studies have shown a reduction in brain volume in the hippocampus in patients with OSA compared with healthy volunteers. An alternative suggestion is that the memory impairments observed in patients with OSA are produced by sleep fragmentation, particularly during REM sleep. In support of this idea, one night of sleep deprivation can result in a reduction in hippocampal activity during episodic memory encoding, and subsequent deficits in memory retention (32). Data from a study in rats also suggest that fragmenting sleep every 2 minutes is sufficient to cause a significant reduction in synaptic plasticity, which would affect memory consolidation (33).
It is notable that our patient group had significantly greater scores on the depression and anxiety subscales of the HADS than did the healthy volunteers, although in both groups the mean scores were within the normal range for depression (score <7) (20). Our analysis revealed no differences between patients with high and lower HADS depression scores. However, this analysis needs to be interpreted with caution because the nondepressed group was older than the mildly depressed group. In addition, the HADS score is a crude measurement of depression, and the cutoff for mild depression is not well defined.
It may be argued that our long testing battery was unsuitable for use with patients with OSA. This issue has been previously investigated and no differences were found when memory was investigated with longer and shorter testing batteries (2). When planning the test battery, we also paid particular attention to the order in which the tests were administered. This minimized the risk of interference between tests. We also took care to allow all volunteers and patients to have a restful break halfway through the battery. Therefore, we do not believe that our findings are attributable to a test order effect.
In the present study we have shown that patients with OSA have specific difficulties in assimilating and recalling information presented verbally, whereas their ability to process visual information remains intact. The memory deficits were present across the spectrum of disease severity, and even patients with mild disease were affected. The impairments were equivalent to a verbal memory performance of healthy volunteers 10 years older than our patients according to normalized scores provided with the test results. We suggest that such memory impairments may have significant detrimental effects on daytime functional performance in patients with OSA.
The authors thank Dr. Michael Polkey for allowing recruitment of patients from his sleep clinic; Mr. Michael Roughton and Mr. Winston Banya, Medical Statistician (Royal Brompton Hospital NHS Foundation Trust and Imperial College) for statistical assistance; Dr. Hilary Green, Ms. Tina Emery, Dr. Jennifer Clealand, and Dr. Richard Wise for assistance during this project; and Ms. Jennifer Jardine for outstanding help in recruiting healthy volunteers.
Supported by a National Heart and Lung Institute Foundation PhD Studentship, Wellcome Trust, the NIHR Respiratory Disease Biomedical Research Unit at the Royal Brompton and Harefield NHS Foundation Trust and Imperial College London.
This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org
Originally Published in Press as DOI: 10.1164/rccm.200901-0065OC on March 18, 2010
Conflict of Interest Statement: G.L.T. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; I.P. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; M.J. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; R.G. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; Z.S. received more than $100,001 from NMT Medical in industry-sponsored grants as previous employment as a research fellow on a grant sponsored by NMT from May 2007 to February 2009; J.J. was employed by Intus Ltd, reporting on overnight pulse oximetry studies, received up to $1,000 from Intus Ltd in advisory board fees, and more than $100,001 from ResMed in industry-sponsored grants; K.S.G. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; A.K.S. received more than $100,001 from ResMed in industry-sponsored grants; M.J.M. received more than $100,001 from ResMed (as a collaborator on the SERVE-HF trial), more than $100,001 from ResMed (CPAP machines for the PREDICT trial are supplied by ResMed), more than $100,001 from Embla (sleep monitors for the PREDICT trial are loaned from Embla), and $10,001–$50,000 from NMT Medical (collaborator on the SPACE Study, the cost of the sleep studies) in industry-sponsored grants.