CrossTalk proposal: The intermittent hypoxia attending severe obstructive sleep apnoea does lead to alterations in brain structure and function
During the past several decades, a large body of research has attempted to elucidate whether the presence of obstructive sleep apnoea (OSA) is causally associated with an increased risk for cognitive deficits that originate from selective injury to CNS regions. To enable a methodologically sound discussion, I will use Hill's ‘viewpoints’ for causality (Hill, 1965).
Viewpoints 1, 2: strength of association and consistency
First, oversimplification or excessive focus on selected aspects of cognition have led to a plethora of studies that cannot be cross-compared due to their discrepant methodologies. Small sample sizes, and either no or only partially matched control groups accounting for co-existing morbidities and confounders, along with the inability to estimate disease duration, may also play a major role in the signal-to-noise ratio of the putative associations. Notwithstanding, associative findings have emerged whereby deficits in short-term memory, verbal fluency, problem solving, attention, and/or perception were frequently (albeit not always) reported, and correlated with the severity of hypoxaemia (Quan et al. 2011; Bucks et al. 2012; Kielb et al. 2012). The question then reverts to whether an association exists between OSA and the presence of structural and/or functional brain deficits. The initial report by Macey et al. (2002) employed T1-weighted magnetic resonance scans and voxel-based morphometric approaches, and found grey matter losses in specific forebrain regions, including the hippocampus. In follow-up studies from the same group, altered regional brain responses in OSA patients emerged following various challenges, and were more prominent in those patients manifesting depressive symptoms, suggesting that the presence of the latter may underlie inordinately increased susceptibility to neural injury in the context of OSA, particularly within affective, cognitive, respiratory and autonomic control regions (Harper et al. 2003; Cross et al. 2008). Subsequent groups of investigators have since further corroborated the presence of hippocampal lesions in patients with OSA (Morrell et al. 2003; Gale & Hopkins, 2004; Yaouhi et al. 2009). However, not all studies have yielded significant associations (O’Donoghue et al. 2005), and such disparities have been tentatively, albeit not conclusively attributed to differences in image processing and analytical procedures.
Using functional magnetic resonance imaging (fMRI) approaches, deficits in working memory tasks in adults with sleep apnoea have been correlated with decreased activation of brain regions associated with such tasks, such as the frontal lobes (Thomas et al. 2005; Ayalon et al. 2010). However, increased activation of neural regions following specific memory or attentional tasks have also been reported (Archbold et al. 2009; Castronovo et al. 2009), and may be attributable to the need for increased recruitment of more extensive brain regions to perform specifically susceptible tasks, either adequately, or less than optimally. It remains unknown whether transitions from heightened regional recruitment to attenuated functional magnetic resonance signal responses occur in the course of the disease. Furthermore, it remains unclear which patients will develop one of these two response patterns, if these two do not occur one after the other.
Thus, despite the relatively strong associations between the presence of neural deficits in selected brain regions of patients with OSA, the skeptical reader may remind us that although repeated observations of an association could indeed suggest that the results are unlikely to be due to chance, they may also simply be due to the presence of the same confounding factors, such as obesity or hypertension.
Viewpoints 3, 4: specificity and coherence
The criterion of specificity requires that a cause leads to a single effect, not multiple effects. However, the fact that different disorders, e.g. obesity and OSA, can induce hypertension, vascular dysfunction, metabolic derangements, or even a vast array of psychological dysfunctions, should not detract from the validity of the initial premise, i.e. that OSA can induce functional and structural brain alterations.
Viewpoint 5: temporality (i.e. exposure to OSA must precede the CNS morbidity)
Most of the clinically based diagnoses of patients with OSA probably occur very late in the course of the disease, even in children. The relative high prevalence of symptoms of OSA, namely, snoring, and the lack of awareness of the onset of such a cardinal symptom may therefore lead to enormous discrepancies in the duration of OSA before it comes to clinical attention. At such time, the evidence of structural CNS damage may be already present, and cannot therefore be placed in the correct time sequence, i.e. before or after the onset of OSA. The assumption that treatment of the disease will lead to reversal of the end-organ morbidity may not hold true, particularly if the duration of the disease causes irreversible damage, has already maximized its regenerative/recovery ability, or no damage has occurred because of genetically/environmentally determined tolerance (Kheirandish & Gozal, 2006). In all these case scenarios, lack of improvement with therapy would not refute the temporality assumption. These considerations are extremely pertinent, since the response to OSA therapy may not reflect the anticipated improvements in cognitive function or reversal of structural brain deficits.
Viewpoint 6: biological gradient (dose–response relationship)
A linear or exponential relationship between the severity of OSA and the magnitude of regional brain losses would support causality. Unfortunately, there is a paucity of data regarding this specific point. In adults with OSA, the high probability for co-existing medical conditions that can contribute to or independently elicit the same end-organ injury inevitably creates a quasi impossible situation that does not permit such assessments to be explored with confidence. This situation is, however, much improved in children, in whom we have documented that the probability and magnitude of cognitive dysfunction follows an OSA-severity gradient, even if at any given level of OSA severity, cognitive function may be either preserved or severely affected (Gozal et al. 2012). The existence of such association between severity markers of OSA and grey matter losses has been documented, however (Canessa et al. 2011).
Viewpoint 7: biological plausibility
There is very little doubt that the current biological knowledge about the disease is consistent with the plausibility of the putative causal link between OSA and brain injury. Experimental models mimicking components of the perturbations encountered in sleeping patients with OSA, such as intermittent hypoxia (IH) and sleep fragmentation (SF), clearly and consistently demonstrate the occurrence of regionally selective brain injury (Gozal et al. 2001; Nair et al. 2011a). Indeed, varying durations of exposures to IH during the habitual sleep times of adult and developing rodents (a) induced major impairments in both hippocampal-dependent learning tasks and long-term potentiation and working memory tasks, (b) induced increases in neuronal apoptosis in the CA1 region of the hippocampus and frontotemporal cortex, and (c) further identified oxidative stress and inflammatory-related pathways accounting for such deficits (Gozal et al. 2001; Payne et al. 2004; Row et al. 2007; Wang et al. 2010; Nair et al. 2011b). Of note, exposure to long-term IH led to impaired wakefulness and selective losses in wake promoting locus coeruleus catecholaminergic neurons (Zhu et al. 2007). Collectively, IH is capable of replicating the cardinal neurocognitive morbidities associated with OSA, and therefore, chronic exposures to IH constitute an underlying cause of cognitive and behavioural dysfunction via selective and regional injury within the CNS. Notably, parallel findings and mechanisms are also applicable to SF (Nair et al. 2011a; Ramesh et al. 2012), raising the possibility of additive or even synergistic interactions between IH and SF may occur (Kaushal et al. 2012). Finally, neither acute nor chronic sustained hypoxic exposures elicit the magnitude of cellular losses associated with IH (Li et al. 2011).
Viewpoint 8: experimental evidence
Randomized controlled trials (RCT) are currently considered to provide strong evidence of cause and effect, and potential ethical issues addressing the implementation of RCT in the context of OSA have been recently reviewed (Redline et al. 2011). To the best of my knowledge, there are no published RCT on the neurocognitive effects of treatment of OSA, except for a selected small cohort of patients with Alzheimer's disease and OSA (Ancoli-Israel et al. 2008). Similarly, only one study has thus far assessed changes in grey matter after continuous positive airway pressure (CPAP) intervention in adults with OSA, and showed improvements in grey matter within selected regions after intervention (Canessa et al. 2011).
Viewpoint 9: analogy
Here, I would argue that the best potential analogous condition would be intermittent exposures to high altitude, whereby the evidence appears to favour the occurrence of selected and transient cognitive declines with altitude exposures (de Aquino Lemos et al. 2012), even if evidence opposing such conclusions has also been reported (Richardson et al. 2011).
Summary
Clearly, a major challenge in human studies is to establish causality between a disease process such as OSA to a slowly progressive, potentially irreversible process, such as neural injury and attendant cognitive dysfunction. From the current evidence, many of the arguments in favour of causality appear to converge, and support the concept that OSA may cause both reversible and irreversible neural injury and functional impairments. However, the contribution of co-morbidities of OSA along with other genetically and environmentally determined modifiers of susceptibility will have to be extensively characterized. In addition, improved insights into the temporal and severity determinants of the neural injury phenotype will require much more extensive investigation.
Call for comments
Readers are invited to give their views on this and the accompanying CrossTalk articles in this issue by submitting a brief comment. Comments may be posted up to 6 weeks after publication of the article, at which point the discussion will close and authors will be invited to submit a ‘final word’.
To submit a comment, go to http://jp.physoc.org/letters/submit/jphysiol;591/2/379
Appendix
Acknowledgements
The author is supported by National Institutes of Health grants HL-086662, HL-65270 and HL107160.
