More and more people are living with dementia worldwide, with no curative treatment in sight. Alzheimer’s disease (AD) is by far the most common form of dementia, representing 60–70% of all dementia cases [1], and is generally preceded by a predementia stage of mild cognitive impairment (MCI). Around 40% of worldwide dementia cases are thought to be attributable to potentially modifiable risk factors, including diabetes, hypertension, obesity, physical inactivity, depression, smoking, low educational attainment, hearing impairment, low social contact, excessive alcohol consumption, and air pollution [2]. In addition, there is a growing body of evidence supporting the role of sleep disorders in the development of MCI and dementia [3,4,5], among which obstructive sleep apnea (OSA) could represent a modifiable risk factor of particular interest [6,7,8,9]. Therapeutic interventions targeting these modifiable risk factors may have the potential to delay dementia onset and slow its progression. Therefore, it is essential to properly identify these modifiable risk factors, and to fully understand the mechanisms by which they may increase the risk of dementia. Moreover, it is also essential to better understand how modifiable risk and protective factors interact to increase or reduce a person’s vulnerability to dementia. Better identifying individuals who could benefit from potential preventive therapies and those who should be included in clinical trials is a worldwide objective [10,11].

2. Definition and Prevalence of OSA

OSA is a respiratory disorder characterized by repeated episodes of partial (hypopnea) and complete (apnea) obstructions of the upper airway during sleep [12,13]. These recurrent respiratory events lead to cortical arousal and sleep fragmentation, intermittent hypoxemia, and increased sympathetic activity, affecting sleep quality and daytime functioning [14]. OSA severity is typically assessed with the apnea–hypopnea index (AHI), which is the average number of apneas and hypopneas per hour of sleep. The severity of OSA is determined according to the following thresholds: mild (AHI ≥ 5 and <15), moderate (AHI ≥ 15 and <30) or severe (AHI ≥ 30) [15]. According to a systematic review in the adult population, 9–38% of individuals present with mild OSA and 6–17% present with moderate to severe OSA [16]. In the elderly, these estimations reach up to 84% for mild OSA and 36% for moderate to severe OSA [16]. Even more worrying, one study observed that 56% of adults aged 65 and older were at high risk of OSA, while only 8% of them had been tested for OSA and diagnosed [17]. Thus, many older adults remain undiagnosed, and therefore not treated. Furthermore, when people do get diagnosed, only about 41% remain adherent to their continuous positive airway pressure (CPAP) treatment after one year [18]. This low CPAP adherence is alarming, especially considering that this treatment could delay cognitive decline in individuals with OSA [9].

3. Potential Mechanisms Linking OSA and Cognitive Decline in Older Adults

It is well established that OSA causes sleep fragmentation and intermittent hypoxemia. More specifically, repeated micro-arousals alter both sleep macroarchitecture (time spent in stage N3 and rapid-eye movement (REM) sleep) [19] and microstructure (slow-wave and spindle characteristics) [20,21,22,23]. Considering the critical roles of sleep continuity, slow-wave sleep, REM sleep, and sleep spindles in neurogenesis, brain plasticity, alertness, and memory formation and consolidation [24,25], chronic sleep changes caused by OSA could negatively affect cognitive health [6,26,27]. REM-dependent OSA could be particularly harmful to the brain, with respiratory events occurring during this sleep stage being associated with reduced daytime regional cerebral blood flow, even in mild OSA [28]. Indeed, muscle atonia during REM sleep can increase the occurrence and the hypoxic levels of respiratory obstructive events, and thus, some individuals present respiratory events mostly in this sleep stage. Further, REM-dependent OSA is more strongly associated with excessive daytime sleepiness than NREM-OSA [29], which is in turn related to cognitive impairment [30].

Moreover, both sleep fragmentation and intermittent hypoxia interfere with brain structure and function, increasing their vulnerability to neurodegenerative diseases. Indeed, it has been suggested that a biphasic pattern of neuroimaging findings could be in play in OSA [31], with acute transitory or compensatory responses (i.e., gray matter hypertrophy, restricted white matter diffusivities) followed by evidence of cellular damage (i.e., gray matter atrophy, higher white matter hyperintensity burden, lower white matter fractional anisotropy, higher water diffusivities). In addition, OSA has been recently associated with increased amyloid and tau burden [21,32,33,34,35,36,37,38,39,40,41,42,43], two proteins involved in AD pathophysiology. Several mechanisms likely underlie these neuroimaging or pathological findings and include inflammation, oxidative stress, metabolic disturbances, cerebral edema and endothelial dysfunction [27]. Indeed, inflammation is involved in neurodegenerative processes, notably by triggering a positive feedback loop that increases amyloid beta production and oxidative stress, facilitating amyloid and tau pathology [26,44].

4. Cohort Studies Investigating the Association between OSA and Cognitive Decline in Older Adults

4.1. Cross-Sectional Studies

The majority of large cross-sectional cohort studies investigating the association between OSA and cognitive impairment in middle-aged and older adults have used objective sleep measures, i.e., polysomnography or portable devices, to evaluate OSA [45,46,47,48,49,50,51,52], and one study used a questionnaire to screen for OSA [53]. While some studies have investigated cognitive performance using comprehensive neuropsychological batteries [51,53,54], others used a limited number of neuropsychological tests targeting specific cognitive functions or global functioning [45,46,49,52]. In studies showing that OSA is associated with poorer cognitive functioning, associations were found for long-term verbal memory [47], working memory [47] and global cognition [52]. The OSA severity markers or OSA-related symptoms associated with cognition were highly heterogeneous across studies, including snoring [53], self-reported apneas [53], hypoxemia [45,54] and AHI [50,51]. Other studies have not found a significant association between any OSA severity marker and cognition [46,48,49].

Despite the heterogenous results emerging from cross-sectional studies, associations between OSA and cognitive functioning seem better established in the domains of attention, memory, and processing speed, while less evidence supports an association between OSA and working memory, executive functions, language and visual abilities in middle-aged and older adults. Interestingly, in younger adults, the most affected cognitive domains in OSA are attention, episodic memory, working memory, and executive functions [14]. This suggests that while some cognitive domains appear to be affected by OSA regardless of the age (namely, memory and attention), other cognitive domains seem less impacted by OSA with increasing age, such as executive function and working memory.

4.2. Longitudinal Studies

Longitudinal cohort studies have the advantage of quantifying cognitive decline over time. They used self-reported diagnoses (e.g., [9,55]) or in-home polysomnography or portable devices (e.g., [7,56,57,58,59]) to identify OSA cases. Furthermore, the majority of these studies used global cognitive measures or screening tools, such as the Mini-Mental State examination (MMSE; e.g., [7,59]), while a comprehensive neuropsychological battery was used in one study [58].

Among the main longitudinal cohort studies, the Study of Osteoporotic Fractures included 298 82-year-old women and found that 45% of women with OSA developed MCI or dementia at five-year follow-up, compared to 31% of women without OSA [59]. However, neither of the two neuropsychological tests used to assess global cognition and executive functions could identify impairment of specific cognitive functions related to OSA. This result was also obtained in the Atherosclerosis Risk in Communities Study that included men and women aged 45–64 years and failed to show an association between OSA and specific cognitive tests at 15-year follow-up [56], while highlighting an increased risk of dementia in individuals with severe OSA [57]. In the Osteoporotic Fractures in Men Sleep Study, including men aged 65 and older, another team showed a negative relationship between baseline nocturnal hypoxemia and global cognitive functioning after three years, while executive functions were not associated with OSA [7]. On the other hand, the Proof-SYNAPSE study used a wide selection of neuropsychological tests in men and women aged 65 years and showed only a slight decline in attention related to OSA after eight years, without any change in executive functions and memory [58].

In summary, studies using measures of global cognition and clinical diagnosis of cognitive impairment or dementia were more susceptible to highlight cognitive decline associated with OSA over time than those using comprehensive neuropsychological batteries [56,57,58]. This suggests that longitudinal studies are more susceptible to identify major OSA-related cognitive decline over time, but not milder cognitive decline in specific domains.

4.3. Meta-Analyses, Meta-Reviews and Systematic Reviews

Most meta-analyses and systematic reviews concluded that a significant association between OSA and cognition exists [8,60,61,62] and that OSA increases the risk of AD [63,64,65]. Interestingly, these meta-analyses and systematic reviews highlighted the fact that small cohorts and controlled case studies from sleep clinics have shown effects of OSA on cognition [66,67], while most studies of large community-based cohort studies failed to show significant associations [45,56]. Another systematic review concluded that the link between OSA and cognition is weak [68], possibly due to the age range used (50 years and over, while others included studies of participants aged 30 years and over [8,60,62]). Similarly, another systematic review showed a significant association between OSA and cognition only in adults younger than 60 years [60]. In addition to age, the variability in the strength of the association could also be due to the study types and designs, the recruitment methods and/or that only more severe OSA cases are associated with cognitive dysfunction.