We studied a group of patients suffering from mild-to-severe OSA, using a comprehensive neuropsychological test battery and, for the first time, both GM density and resting-state brain glucose utilization examinations in the same group of patients, including (i) an optimized VBM procedure using a customized template from both patients and controls; (ii) the correction of PET data for PVE and (iii) the same threshold for assessing both atrophy and hypometabolism statistics.
Compared with controls, our patients showed relatively preserved attentional performance. We did not find any group effect in our patients, although some displayed scattered attentional impairments (reaction time and/or errors). Mazza et al. (2005) reported vigilance and/or attentional impairment in the majority of their patients. These authors used an attentional test battery comprising three long-lasting tasks: (i) Oxford sleep resistance, which is an objective daytime somnolence test resembling the MWT condition; (ii) a driving simulator task and (iii) the continuous performance test. Our longest attention test was ‘bimodal vigilance’, which lasted 15 min, whereas all the attentional tasks in the study conducted by Mazza et al. (2005) lasted more than 20 min. In addition, our patient group fell asleep within approximately 20 min on average (MWT score), suggesting that potential attentional impairments might have been revealed by long-lasting tasks (Decary et al., 2000).
Alchanatis et al. (2005) showed that high-intelligence patients had the same attention/alertness performances as high-intelligence controls, and assumed that high intelligence offers some protection against OSA-related cognitive decline, possibly in the form of increased cognitive reserve. In our study, most of the patients were highly educated (graduate and postgraduate levels) and had professions involving permanent intellectual and/or attentional constraints (e.g. entrepreneurs, directors, a corporate lawyer, two medical professionals and three subjects occupying a position of responsibility). There is reason to believe that their professional experience may have allowed them to cope somewhat better with routine attentional requirements. However, they did acknowledge having to struggle to maintain their attention in monotonous and long-lasting situations.
In the executive function assessment, our patients performed similarly to controls, in agreement with some authors (Redline et al., 1997; Verstraeten et al., 2004) but in contrast to others (Bedard et al., 1991; Montplaisir et al., 1992). A critical review (Verstraeten and Cluydts, 2004) regarding the executive control of attention in apneics suggested that several findings in the literature, interpreted as evidence of dysexecutive performance, should be viewed with caution, given the lack of careful methodology (e.g. basic attentional performances were only infrequently monitored in executive tasks). Our findings, i.e. only vigilance decrements, as evidenced by MWT scores, suggest a probable decrease in the basic tonic attention level in apneics (Lis et al., 2008; Verstraeten and Cluydts, 2004).
In terms of memory, our patients displayed minor impairment in word list learning, confirming the findings of Salorio et al. (2002). They also displayed slightly decreased recall in a visual episodic memory task. In their group of patients, Naëgeléet al. (2006) observed a free recall deficit in episodic memory but normal maintenance, recognition and forgetfulness, suggesting that memory impairment in OSA is mild (Fulda and Schulz, 2001) and does not affect all memory processes (Naëgeléet al., 1995, 2006; Salorio et al., 2002). Thus, in our study, a memory task, namely ‘paired associates’, evidenced some learning difficulties and, another one, namely ‘family pictures’, showed a mild free recall impairment, in accordance with the literature, these two tasks being usually used to assess the same cognitive function, i.e. episodic memory.
In sum, our apneic sample displayed only minor impairment when they underwent a comprehensive neuropsychological test battery, in agreement with other studies (Kim et al., 1997; Redline et al., 1997). Our findings do not corroborate the presumed dysexecutive profile (Beebe and Gozal, 2002).
It should be pointed out that OSA without notable cognitive impairments has been fortuitously discovered in about half a non-clinical population (Quan et al., 2006). The daytime neurobehavioral manifestations of OSA may simply be an epiphenomenon of respiratory disturbances during sleep, and many resilience factors may temporarily prevent apneics from sustaining substantial neurobehavioral deficits (Beebe, 2005). These factors include sleep deprivation tolerance (Mu et al., 2005) and cognitive reserve that allows patients to cope relatively well with daytime challenges (Stern, 2002).
The relationship between neuropsychological deficits and OSA features is rarely strong and consistent (Adams et al., 2001; Kingshott et al., 1998; Sauter et al., 2000), and there were no significant correlations in our sample between the deficient neuropsychological scores, the PSG and the MWT variables. Such links would, however, exist in more severely affected patients, as suggested by a number of studies highlighting the impact of nocturnal hypoxemia on vigilance, executive and psychomotor tasks and the impact of vigilance impairment on attention and memory functions (Bedard et al., 1991; Montplaisir et al., 1992). Unlike the aforementioned studies that comprised patients with profound hypoxemia, our sample did not seem to be severely hypoxemic (very few patients were under 80% in terms of the percentage of time spent under this degree of desaturation). Furthermore, our patients were mostly in their mid-50s, and possibly at disease onset. In most cases, they estimated that changes in their quality of life had occurred within the last couple of years. Hence, untreated OSA is likely to have a more substantial effect on cognitive functioning with advancing years. In fact, a recent study showed that aging apneics demonstrate cognitive decline, while younger patients with the same disease severity are somehow able to compensate for this effect (Alchanatis et al., 2008). Moreover, elderly patients may have a risk of dementia comorbidity (Bliwise, 2002; O’Hara et al., 2005).
Taken together, our findings revealed cerebral changes (mainly right-lateralized) regarding both GM density and metabolism. PET measurements of resting-state brain glucose utilization reflect local baseline integrated synaptic activity and therefore both neuronal lesions and synaptic dysfunction, whereas VBM only assesses GM loss. The latter was located in scattered sites, i.e. the frontal and temporo–parieto–occipital cortices, the thalamus, some of the basal ganglia and the cerebellar regions, whereas the decrease in brain metabolism was more localized than GM density changes, restricted to the cuneus and precuneus, temporo–parieto–occipital cortices and middle and posterior cingulate area, as well as the prefrontal cortex (as a statistical trend).
Our patients showed significant GM loss in several brain areas previously identified by Macey et al. (2002). Interestingly, some of the affected brain areas happen to be particularly vulnerable to anoxia, such as the basal ganglia, the cerebellum and the hippocampal region (Gozal et al., 2001; Konaka et al., 2007). Hypoxemia may cause brain structural changes, as shown in animal models (Gozal et al., 2001). That said, our patients did not seem to be severely hypoxemic and the controlled hypoxia of rodent models in a laboratory setting (e.g. Gozal and coworkers) cannot be likened to our patients’ condition, in terms of exposure to hypoxia. Thus, the decrease in GM density may also be induced by factors other than apneic events, i.e. brain lesions that are congenital in nature, acquired (Macey et al., 2002) or subsequent to cardiovascular comorbidities (Macey and Harper, 2005; O’Donoghue et al., 2005). As in the study conducted by Macey et al., it is noteworthy that our sample encompassed patients with cardiovascular comorbidities (controlled hypertension, dyslipidemia and prior but no recent history of myocardial infarction).
The GM density changes seen in our patients do partially corroborate the prefrontal hypothesis, given the changes found not only in the PFC region but also in the basal ganglia and thalamus that are linked to this region (Alexander and Crutcher, 1990; MacDonald et al., 2000).
These GM density changes may account for certain neurobehavioral impairments seen in our patients, as many of the atrophic sites (right frontoparietal circuit extending to basal ganglia) are involved in attentional abilities (Sturm and Willmes, 2001; Sturm et al., 1999). The cerebral damage (especially the prefrontal and thalamic ones) may also account for our patients’ sleepiness (Coull, 1998; Portas et al., 1998) and for the residual somnolence reported in treatment studies (Marshall et al., 2006).
In addition, some brain areas in which GM was significantly decreased in our patients (hippocampus and thalamus which belong to the Papez circuit, as well as frontoparietal regions) play an important role in episodic memory (Eustache and Desgranges, 2008), which is in agreement with the mild memory deficit seen in our patients. Lastly, the motor speed and fine motor coordination impairments displayed by our patients may be linked with cerebellar changes (Ramnani et al., 2001; Verleger et al., 1999).
Hypometabolism was more restricted than the GM density changes and was found in the precuneus, the parieto–occipital cortex and the cingulate gyrus (middle and posterior areas), as well as in the prefrontal cortex (as a statistical trend). Some of these structures were also atrophic, in particular the right inferior parietal cortex and prefrontal cortex, the latter finding reinforcing the frontal hypothesis. The functional impairment of the precuneus and cingulate gyrus, which were not atrophic, may have been caused partly by remote effects originating from morphologically impaired areas with decreased connectivity (Chételat et al., 2003), although we cannot exclude the possibility that minor structural changes in these areas may also have been partly responsible for this hypometabolism (at a lenient threshold of P < 0.01 uncorrected, GM loss was observed in these structures).
Some of the brain areas found to be hypometabolic in our sample (i.e. the precuneus, cingulate areas and inferior parietal cortex) belong to the ‘default mode’ circuit, a ‘novel and only recently appreciated brain system that participates in internal modes of cognition’, i.e. a neural circuit of ‘normal’ mental activity in the conscious resting state (Buckner et al., 2008; Cavanna and Trimble, 2006; Raichle et al., 2001). Although the wakefulness of our patients was monitored during the PET data acquisition period using a portable EEG device, unlike healthy subjects, they probably had disturbed neural activity in brain regions that are usually active during the normal resting state. Similarly, using quantitative EEG analysis, Morisson et al. (1998) recorded an EEG slowing in awake apneics, in all the cortical regions they examined. This suggests that, during wakefulness, the apneics’ functional neural abnormalities may disturb the mental activity needed in the resting condition.
All the significant correlations between the PET data and MWT scores and the deficient neuropsychological scores were found in cerebellar regions. Our findings not only underline the well-known links between motor function and cerebellum, but also point to the involvement of this region in cognitive functions, notably in episodic memory (Andreasen et al., 1999; Leiner et al., 1991; Schmahmann and Pandya, 1997). In addition, we found significant correlations between the MI index and the metabolism of the bilateral superior motor area and the left superior parietal cortex, which may reflect a negative effect of sleep fragmentation, and could be linked to psychomotor vigilance decrements seen in healthy subjects after sleep deprivation (Drummond et al., 2005a).
One of the limitations of this study is the relatively small sample of patients, which prevent us from generalizing our findings to all apneics, notably the neuropsychological results. Given the small size of this group, we chose not to correct for multiple comparisons as far as the neuropsychological data were concerned, in order to highlight cognitive abnormalities, even mild ones, and clarify their links with the brain imaging results. Regarding the structural imaging analysis, the whole-brain VBM method has sometimes been criticized because of the normalization process (inducing some deformations) and has been regarded as less sensitive than the region-of-interest method when it comes to detecting abnormalities in small subcortical structures. However, the use of an optimized VBM technique, i.e. obtaining a customized template from both samples (patients and their controls), has proved to be effective in studies of several pathologies (Chételat et al., 2003; Desgranges et al., 2007; Mevel et al., 2007; Salmon et al., 2008; for a review) as well as of healthy aging (Kalpouzos et al., 2009). Furthermore, this method presents notable advantages and only minor drawbacks. While the region-of-interest method leaves large brain areas unexplored and is very time-consuming and expertise-dependent, the automatic time-saving voxel-by-voxel method provides a prospective and fully comprehensive assessment of the brain, without bias. Regarding functional imaging, the use of a resting-state PET assessment allowed us to study the functional cerebral changes throughout the entire brain, while an activation study would have been informative in terms of the cerebral structures involved in a given neuropsychological function but would have been confined to that function.
In summary, our patients had significant metabolic and GM density changes, contrasting with minor neuropsychological deficits, suggesting that cognitive reserve may have averted a more significant impact of OSA. Some of these cerebral changes partially corroborate the prefrontal hypothesis. However, the neuropsychological assessment failed to reveal significant attentional or executive deficits. Nonetheless, physiological measurements (MWT latencies) did show decreased ability in our patients to resist sleepiness, evidencing to some extent a decline in vigilance. Thus, the relatively intact behavioral performances of our patients may have been due to compensatory mechanisms (Stern, 2002). In line with this idea, an fMRI study found that, relative to controls, apneics demonstrated intact performances on a verbal learning task, but with greater brain activation (Ayalon et al., 2006). Likewise, after sleep deprivation, healthy subjects have been found to display additional brain activation during demanding cognitive tasks (Drummond et al., 2005b). In the resting-state condition, our patients showed a metabolic defect in the default-mode circuit. Further studies are needed to improve our understanding of the neurobehavioral sequelae of this functional impairment. Our findings also highlight the contribution of cerebellar changes to impairments in the memory and motor domains.
To conclude, in apneics without notable cognitive impairments, cerebral imaging may be more sensitive when it comes to detecting the impact of silent OSA on brain morphology and function. The disease probably starts many years before it is clinically expressed, slowly becoming more severe and apparent in terms of significant neurobehavioral manifestations.