Obstructive sleep apnoea (OSA) is a multisystem, debilitating, chronic disorder of breathing during sleep, resulting in a relatively consistent pattern of cognitive deficits (Rosenzweig et al., 2015; Bucks et al., 2017), particularly in attention, executive function and episodic memory (Bucks et al., 2017). Moreover, there is a high prevalence of depression, anxiety and other psychiatric problems, which are only partially remediated by treatment (Rosenzweig et al., 2015; Bucks et al., 2017).

Cognitive functions traditionally comprise broad domains of attention and memory, as well as those of higher order cognitive skills such as planning, problem-solving, and mental flexibility (grouped together as executive function), visuospatial abilities, processing speed, and both expressive and receptive language (Esther Strauss, 2006; Bucks et al., 2017; Rosenzweig et al., 2022). Historically, a body of work has suggested significant impact of OSA on: attention and vigilance, long-term verbal and visual memory, expressive and receptive language (Bucks et al., 2013; Wallace and Bucks, 2013), and visuo-spatial and constructional abilities (Bucks et al., 2013). Similarly, deficits in the executive domain have also been demonstrated (Olaithe and Bucks, 2013; Bucks et al., 2017), with somewhat less uniform evidence for short-term memory deficits (Rosenzweig et al., 2022). However, cognitive domains are not unitary constructs, and only judiciously deconstructed analysis of their different sub-capacities and their vulnerabilities to a range of risks and protective factors specific to OSA can provide a more accurate appraisal of a patient's deficits (Rosenzweig et al., 2017, 2022).

Accordingly, OSA's bidirectional link to neurodegenerative disorders, including Alzheimer's disorder (AD), has similarly highlighted the importance of disentangling some of the major cognitive neuromechanisms at play (Ancoli-Israel et al., 2008; Cooke et al., 2009; Osorio et al., 2015; Bubu et al., 2022). However, the links between severity of OSA, historically indexed by apnea-hypopnea index (AHI) or respiratory disturbance index (RDI), or by indices of hypoxia severity, sleep fragmentation, or sleepiness (Pépin et al., 2005; Bucks et al., 2017), and the severity of the cognitive deficits observed, are far from being well understood (for more in-depth review please refer to Aloia et al., 2004; Olaithe and Bucks, 2013; Wallace and Bucks, 2013; Gagnon et al., 2014; Rosenzweig et al., 2015; Bucks et al., 2017). Similarly poorly understood, is the link between the timing of the hypoxia or arousal during each sleep cycle, and the severity or the cognitive phenotype (Rosenzweig et al., 2015).

Moreover, the link between cognition, OSA, and aging, has proven equally difficult to fully discern (Rosenzweig et al., 2022). For instance, aging is known to be independently linked with physiological changes that may predispose to OSA (Rosenzweig et al., 2022). It has been proposed that this may be in part due to changes in upper airway morphology that can lead to a reduction in upper airway dilator muscle function at sleep onset (Bucks et al., 2017), contributing to an age-related propensity for upper airway collapse in response to negative pressure (Kirkness et al., 2008) independent of body mass index (Eikermann et al., 2007; Rosenzweig et al., 2022). Against this background, it has been argued that some of these aging-associated processes may underlie the failure to find a consistent relationship between the severity of OSA and the risk of cognitive impairment (Bucks et al., 2017; Rosenzweig et al., 2022).

This evident polymorphic picture is further compounded by the fact that not everyone with OSA is cognitively impaired (Quan et al., 2006), with the individual's cognitive reserve (Alchanatis et al., 2005; Sforza et al., 2010; Bucks et al., 2013; Martin et al., 2015; Olaithe et al., 2015; Schembri et al., 2017) and their genetic risk (Cosentino et al., 2008; Nikodemova et al., 2013) for neurodegenerative decline (Bucks et al., 2017) possibly also playing an important role (Rosenzweig et al., 2022).

Taken together, it has been proposed that only middle-aged patients with OSA may demonstrate a consistent pattern of cognitive deficits, otherwise lacking in older patients (Bubu et al., 2020). Notably, it has also been argued that deficits are principally driven by common cardiovascular and metabolic comorbidities, rather than by distinct OSA-processes (Bubu et al., 2020). In further support of this, patients with OSA invariably present with already established comorbidities, such as overweight or obesity, sleepiness in passive situations or while driving, and are often affected by systemic hypertension, type 2 diabetes, and dyslipidemia (Levy et al., 2015), making it in most cases impossible to delineate the specific contribution of all associated risks to the resulting cognitive presentation (Bonsignore et al., 2019; Bubu et al., 2020; Rosenzweig et al., 2022).

Whether OSA itself, or these common comorbidities drive the cognitive effects has wide ranging clinical implications, and may impact future clinical guidelines with treatment of comorbidities taking precedence over treatment of the core determinants of neuropathological process in OSA, such as sleep fragmentation (Jordan et al., 2014), sleep disruption and blood gas abnormalities (Olaithe et al., 2018).

To this end, and in order to clarify whether, independent of concomitant metabolic or cardiovascular comorbidities, OSA-induced injury (Rosenzweig et al., 2015) may present with abnormal functional outcomes (Bubu et al., 2020; Rosenzweig et al., 2022), we undertook a proof of the concept study and set to define the cognitive pattern in a (rare) group of male, middle-aged patients with untreated OSA who present without comorbidities, compared to matched controls.

Preliminary analysis of cognitive parameters in patients with different OSA severities was undertaken as a part of the multimodal clinical study InCOSA (Clinical.Trials.gov, identifier: NCT02967536). All experimental protocols were approved by the U.K. Research Ethics Committee [Integrated Research Application System (IRAS): IRAS-Project-ID-170912; REC-REF16/L0/0893] and informed consent for study participation was obtained from all participants. Due to the nature of several investigations in the overarching multimodal clinical study, some of which are known to have significant sex and gender differences, only male participants were included. Thus, 27 male adult (35–70 years-old), non-obese [body-mass-index (BMI) < 30 kg/m²], mildly somnolent [Epworth Sleepiness Scale (ESS); 5 > ESS < 15] patients with no current or past co-morbidities, and no current or past alcohol or smoking history, with a de novo diagnosis of OSA according to ICSD criteria (American Academy of Sleep Medicine, 2014), and a group of healthy sex- and education-matched individuals were identified, as previously described (Gnoni et al., 2021) (Supplementary material). All methods were carried out in accordance with relevant UK and international guidelines and regulations.

All participants underwent a domiciliary respiratory testing via WatchPAT system (https://www.itamar-medical.com/), as previously described (Walter et al., 2023). Additionally, patients also underwent a video-polysomnography (vPSG) in the sleep center. Full night vPSG recordings were based on the international 10:20 system; for purposes of the PSG scoring, six EEG channels (i.e., F3, F4, C3, C4, O1, and O2) were referenced to the mastoid, and used along with electrooculography, submental-electromyography, respiratory inductance plethysmography, nasal pressure sensor, oronasal thermistor, pulse-oximeter, two-lead electrocardiogram, body position detector and synchronized audio-visual recording, as previously described (Rosenzweig et al., 2016). The scoring was carried out according to AASM rules (Berry et al., 2017), and as previously described (Rosenzweig et al., 2016). Eleven cognitive domains were pre-selected (Bucks et al., 2017), based on previous reports of OSA- and depression-related deficits, and assessed via 23 automated Cambridge-Neuropsychological-Test-Automated-Battery (CANTAB) tests.

Twenty-seven OSA patients and seven healthy controls (see Table 1) completed the CANTAB (Table 2). Of the 27 patients, 16 were diagnosed with mild OSA (AHI mean ± S.D.:11.7 ± 4.0 events/h; mean age 42.6 ± 8.2 years; mean BMI 26.7 ± 4.1 kg/m²), and 11 with severe OSA (AHI 41.8 ± 20.7 events/h; age: 46.9 ± 10.9 years, BMI: 28.0 ± 3.2 kg/m²), according to ICSD criteria (American Academy of Sleep Medicine, 2014).

The cognitive findings for the whole set of behavioral readouts for the three experimental groups (control, mild OSA, and severe OSA), controlled for age and BMI and grouped into CANTAB sub-tests are shown in Table 3. Distinct deficits were observed in the tests investigating cognitive domains of vigilance, executive functioning, short-term visual recognition memory and social and emotion recognition, with the greatest number of differences between controls and those with severe OSA. Whilst subjects with mild OSA performed better than those with severe OSA on most of those same tasks, they were rarely worse than controls (see Figure 1).

The most significant deficits, by comparison to the control group, were demonstrated in the tests that assess both simultaneous visual matching ability and short-term visual recognition memory for non-verbalizable patterns (Figure 1; DMS), tests of executive functioning and cued attentional set shifting (Figure 1; AST), in vigilance and psychomotor functioning (RTT), and lastly, in social cognition and emotion recognition (ERT).

For full details of the cognitive findings please refer to Tables 2, 3 and Supplementary material.

We report a distinct pattern of circumscribed cognitive deficits in middle-aged male patients with severe OSA, in the absence of any overt neuropsychiatric, cardiovascular or metabolic co-morbidities (Bubu et al., 2020). The findings are largely in keeping with previous studies of OSA patients with associated multiple comorbidities that similarly showed aberrant executive-functioning, visuospatial short-term-memory, deficits in vigilance and psychomotor control (Bucks et al., 2017; Rosenzweig et al., 2022). Thus, arguably, our findings suggest that distinct OSA-driven processes, particularly when OSA is severe, may be sufficient for cognitive changes to occur as early as middle age, in otherwise healthy male individuals.

Remarkably, we also report, for the first time, diminished social cognition in this group of middle-aged severe OSA patients. Social and emotional cognition is an important ability to interpret and identify socially relevant information, known to be impaired in several psychiatric conditions, including major depressive disorder, and thought to be strongly associated with sleep physiology (Gujar et al., 2011; Weightman et al., 2014). In past studies, sleep deprivation has been shown to selectively impair the accurate judgment of human facial emotions, especially threat relevant and reward relevant categories (van der Helm et al., 2010). Significant deficits in emotional facial recognition have been previously also reported following a night of sleep fragmentation, without significant reduction of total sleep time (Lee et al., 2022). In keeping, it has been suggested that the disruption of normal sleep process, and not the reduction of sleep time, may likewise play the role (Lee et al., 2022). Moreover, there is evidence to suggest that emotional facial recognition can be sleep-stage dependent, with REM sleep known to play a critical role on both emotional and neutral face recognition (Cunningham and Payne, 2017; Lee et al., 2022). Thus, it is likely that sleep fragmentation and associated sleep loss in our OSA patients, particularly REM-related fragmentation (Cunningham and Payne, 2017), may act to impair discrete affective neural systems, disrupting the identification of salient affective social cues (van der Helm et al., 2010).

More recently, in a thought provoking set of studies, sleep loss and sleep's diminished quality and or quantity, which indeed present one of the important features of OSA, have also been linked to diminished altruism (Ben Simon et al., 2022). Specifically, the authors argued that sleep loss represents one previously unrecognized factor that may dictate whether humans choose to help each other, which they based on their observations at three different scales, within individuals, across individuals, and across societies (Ben Simon et al., 2022). For instance, in one of the studies, one night of sleep loss was shown to trigger the withdrawal of help from one individual to another, with the associated fMRI findings showing deactivation of key nodes within the social cognition brain network that facilitate prosociality (Ben Simon et al., 2022).

Following this argument and our findings, as well as taking into account that currently around one-seventh of the world's adult population, or approximately one billion people, are estimated to have OSA (Lyons et al., 2020), the clinical and societal impact of OSA's effects on cognition, even in the absence of any associated co-morbidities, dictates urgent attention and a joint multidisciplinary effort. It is increasingly evident that OSA's functional neuropsychiatric impact may go well beyond OSA's currently best recognized role in increasing driving and occupational accidents risks (Bucks et al., 2017; Rosenzweig et al., 2022).

We believe that our pilot study, despite limitations, including its size, a small control group, multiple comparisons, and a cross-sectional design, significantly contributes to understanding of the complex interplay between OSA-severity and cognitive problems. Critically, our data also reveal a threshold effect in the cognitive domain of executive functioning. Furthermore, it appears that cognitive deficits in this age group are greatest in male patients with severe OSA, likely suggestive of already existent widespread intricate physiologic central nervous changes, and in further support of early treatment for this patient group (Rosenzweig et al., 2015; Gnoni et al., 2021).

Finally, another important limitation to any direct translational generalization of our findings lies in inclusion of male participants only. Whilst this enabled controlling for possible effects of the oestrous cycle, it also prevents us from generalizing to female patients. Moreover, over the last decade, pioneering new findings suggest a spectrum of changes in the brain metabolism during the pre-, peri-, and post-menopausal period (Mosconi et al., 2021), all of which may arguably interplay with OSA pathomechanisms (Driver et al., 2005; Saaresranta et al., 2015), as well as underlie its links with neurodegenerative processes and cognitive deficits (Polsek et al., 2018) in female patients with OSA.

In conclusion, future multi-center multi-modal longitudinal studies should confirm these findings, as well as decipher how these cognitive deficits may interplay in men and women with other comorbidity-driven impairments over time.

The datasets presented in this article are not readily available because all data that support the findings of this study will be made available upon reasonable request from the corresponding author, once the appropriate ethics amendments are sought to accommodate this request. Requests to access the datasets should be directed to the U.K. Research Ethics Committee [Integrated Research Application System (IRAS)].

The studies involving human participants were reviewed and approved by the U.K. Research Ethics Committee [Integrated Research Application System (IRAS): IRAS-Project-ID-170912; REC-REF16/L0/0893]. The patients/participants provided their written informed consent to participate in this study.

VG and IR designed the study. VG, DO'R, MJ, RB, and MM analyzed the data. All authors contributed equally to the writing and revision of this manuscript.