Comorbid insomnia and sleep apnea is associated with worse verbal episodic memory in older females

Abstract

Study Objectives:

To investigate whether comorbid insomnia and sleep apnea (COMISA) is associated with poor verbal memory in older adults, and whether this relationship is moderated by sex.

Methods:

A total of 110 older adults aged (65–83), all diagnosed with obstructive sleep apnea, completed overnight polysomnography and cognitive testing. COMISA was defined as obstructive sleep apnea plus an Insomnia Severity Index score ≥ 11. Verbal memory was assessed via the delayed recall component of the Alzheimer’s Disease Cooperative Study Preclinical Alzheimer’s Cognitive Composite. Moderation analysis examined the interaction between COMISA and sex on verbal memory performance, adjusting for age, body mass index, APOE4 status, and education. Post hoc sleep architecture differences between males and females with COMISA and females with COMISA compared to obstructive sleep apnea only were analyzed using multivariate analysis of covariance.

Results:

COMISA was associated with significantly worse verbal memory performance, with this effect driven by females (b = −2.82, standard error = 0.94, t = −3.01, P = .003) and absent in males (b = 0.62, standard error = 0.97, t = 0.63, P = .528). Post hoc analyses revealed that females with COMISA showed reduced rapid eye movement sleep and increased slow wave sleep compared to males with COMISA.

Conclusions:

COMISA is linked to sex-specific cognitive vulnerability, with older females showing worse verbal memory than males. Post hoc analyses revealed differences in sleep architecture by sex within COMISA, warranting further investigation into stage-specific sleep contributions to cognitive risk. These findings highlight the importance of sex-informed approaches to assessing and managing cognitive risk in aging populations.

Clinical Trial Registration:

Registry: ClinicalTrials.gov; Name: Is Obstructive Sleep Apnea Important in the Development of Alzheimer’s Disease?; URL: https://clinicaltrials.gov/study/NCT05094271; Identifier: NCT05094271.

BRIEF SUMMARY

Current Knowledge/Study Rationale:

Insomnia and obstructive sleep apnea often co-occur in older adults as comorbid insomnia and sleep apnea (COMISA), a condition linked to greater cognitive and health risks than either condition alone. Females are disproportionately affected by insomnia, underdiagnosed for obstructive sleep apnea, and make up most Alzheimer’s disease cases, yet sex-specific cognitive effects of COMISA remain unclear.

Study Impact:

This study shows that older females with COMISA exhibit poorer verbal memory compared to males with COMISA. Findings highlight the need for sex-specific diagnostic and therapeutic strategies for cognitive risk in COMISA.

INTRODUCTION

Insomnia and obstructive sleep apnea (OSA) are among the most prevalent sleep disorders in older adults.^1,2 A growing body of literature has identified a substantial overlap between insomnia and OSA,^3–5 a condition termed comorbid insomnia and sleep apnea (COMISA).⁵ COMISA is highly prevalent, with estimates ranging widely depending on population and risk factors of focus.⁶ For instance, an estimated 30–50% of individuals with OSA report clinically significant insomnia symptoms and 30–40% of those with chronic insomnia meet criteria for OSA.⁷ Compared to individuals with either disorder alone, COMISA is associated with greater sleep fragmentation, worse cognitive and physical health outcomes, and increased all-cause mortality risk.^8,9 Further, females are disproportionately affected by insomnia compared to males¹⁰ and often underreport or go undiagnosed for OSA,¹¹ highlighting the need to consider sex-specific sleep vulnerabilities when assessing cognitive risk.

Beyond its high prevalence, COMISA is particularly concerning due to its potential impact on cognitive function. Sleep fragmentation and intermittent hypoxia, 2 hallmark features of insomnia and OSA,¹² are known contributors to cognitive decline and have been implicated in Alzheimer’s disease (AD) pathology.^13–16 Additionally, disrupted sleep architecture has been associated with poorer cognitive performance and increased risk of neurodegeneration, particularly in older adults.^17–20 Episodic memory, a cognitive domain that is highly sensitive to early AD-related changes,²¹ may be especially vulnerable to the combined effects of sleep disruption and oxygen desaturation^22,23 in COMISA. Given that females typically outperform males in verbal episodic memory,^24–26 it is possible that sleep disturbances in COMISA may attenuate this advantage, placing females at increased risk for memory impairment and possibly neurodegenerative disease.

Emerging research suggests that COMISA poses a heightened cognitive risk beyond insomnia or OSA alone. For example, Huang et al (2024) found that individuals with COMISA had nearly twice the odds of self-reported cognitive decline compared to those with healthy sleep.²⁷ Neuroimaging studies further indicate that COMISA is linked to gray matter atrophy in memory-related brain regions, including the temporal lobes and fusiform gyrus.²⁸ These structural changes correlate with worse cognitive performance and mood dysregulation. These findings highlight COMISA’s potential role in neurodegeneration and reinforce the importance of studying its impact on objective memory performance. Although insomnia and OSA have each been independently associated with cognitive decline,^15,29–32 their combined effect remains largely unexamined.

Importantly, these cognitive risks may not be evenly distributed. Females are disproportionately affected by insomnia,¹⁰ underreport OSA symptoms and are underdiagnosed with OSA,¹¹ and make up almost two-thirds of AD cases.³³ At the same time, females tend to show better performance than males on verbal episodic memory tasks.^25,26 It remains unclear whether COMISA undermines this cognitive advantage in females, which may reveal early, sex-specific risk for AD. We hypothesized that in older adults, COMISA would be associated with poorer verbal memory performance compared to those with OSA only, and that this relationship would be more pronounced in females than in males. To explore possible physiological correlates of this pattern, we also examined sex differences in sleep architecture using polysomnography (PSG) data in post hoc analyses.

METHODS

Participants

Older adults (ages 65–83 years) were recruited from the San Diego community as part of a larger study that examined the role of OSA as a modifiable factor in the development of AD (ClinicalTrials.gov ID: NCT05094271). Participants were recruited prospectively from electric medical record extraction University of California, San Diego and prior research studies. Inclusion criteria required participants to be cognitively normal males and females aged 65–85 years, as assessed by a Montreal Cognitive Assessment score of ≥ 26, and not currently undergoing treatment for OSA. Exclusion criteria included smoking, a history of chronic obstructive pulmonary disease or asthma, uncontrolled cardiovascular disease, neuromuscular disease, brain tumors, and current drug abuse. Additional exclusion criteria for the present study included absence of OSA (ie, apnea-hypopnea index [AHI] < 5 events/h).

During the study, OSA was either newly diagnosed or in some cases re-diagnosed. Between October 8, 2021 and March 6, 2025, 210 participants consented to participate in the study, with 143 completing their overnight visit. However, 13 participants were excluded from the data analysis due to the absence of OSA, missing Insomnia Severity Index (ISI) questionnaire data (n = 5), Alzheimer’s Disease Cooperative Study Preclinical Alzheimer’s Cognitive Composite testing data (n = 13), AD polygenic hazard score data (n = 6), or education levels (n = 5). As a result, 110 participants with complete data met the criteria for this analysis. The sample was 55.4% females, 83.7% White, 90.2% Non-Hispanic, 20% apolipoprotein E ε4 (APOE4) carriers, and 73.9% with Bachelor’s degree or higher (see Table 1 for full sample characteristics). COMISA, present in 37% of participants, was defined by an AHI ≥ 5 events/h and an ISI score ≥ 11 (see Table 1 for complete characteristics). All participants provided written informed consent, and the study protocol was approved by the University of California, San Diego Institutional Review Board under the Human Research Protections Program.

Table 1.

—Full and COMISA-stratified participant characteristics.

Variable	Full Sample	COMISA (n = 41) Mean (SD) or %	OSA Only (n = 69) Mean (SD) or %	P
Age (years)	69.9 (3.9)	69.5 (4.0)	70.3 (3.9)	ns
Race (% White)	83.6%	77.5	85.7%	ns
Ethnicity (% not Hispanic/Latino)	90.9%	80%	95.7%	ns
Sex (% females)	50%	57.5%	47.1%	ns
Income
0–50,000	23.6%	29.3%	20.3%	ns
50,000–100,000	26.4%	22%	29%	ns
100,000–150,000	17.3%	17.1%	17.4%
Education (% bachelors or higher)	17.8%	69.2%	72.5%	ns
AHI	34.3 (22.9)	37.1 (24.2)	32.2 (21.1)	ns
BMI	27.5 (5.4)	27.5 (5.9)	27.4 (5.2)	ns
MOCA	27.9 (1.4)	27.7 (1.4)	28.0 (1.4)	ns
PACC		135.1 (10.3)	138.7 (13.6)
Verbal memory-delayed	12.0 (3.5)	11.1 (3.5)	12.7 (3.4)	ns
APOE4+	20.9%	20%	25.7%	ns
Insomnia Severity Index	8.1 (5.5)	14.1 (2.7)	4.6 (3.1)	< .001
Total sleep time (minutes)	296.5	288.7	301.2	ns
N1 (minutes)	63.1	53.1	69.0	.038
N2 (minutes)	161.2	161.0	161.3	ns
N3 (minutes)	28.7	29.3	28.3	ns
REM (minutes)	43.6	45.3	42.6	ns
Hypoxic burden	74.1	83.0	68.8	ns
Arousal index	32.0	31.7	32.2	ns

AHI = apnea-hypopnea index, APOE4, apolipoprotein E ε4, BMI = body mass index, COMISA = comorbid insomnia and sleep apnea, MOCA = Montreal Cognitive Assessment, N1 = stage 1 sleep, N2 = stage 2 sleep, N3 = stage 3 sleep, ns = non-significant, OSA = obstructive sleep apnea, PACC = Preclinical Alzheimer’s Cognitive Composite, REM – rapid eye movement, SD = standard deviation.

Procedure

The broader multiarm study protocol has been described else-where.³⁴ In brief, the data presented herein are derived from the initial screening visits, which involved the first daytime and first overnight visits. These visits were designed to screen for cognitive impairment and presence of OSA. During the daytime visit participants were administered a brief cognitive battery, which consisted of the Montreal Cognitive Assessment and the Alzheimer’s Disease Cooperative Study Preclinical Alzheimer’s Cognitive Composite. Participants with normal cognition were eligible for further participation and also completed several questionnaires related to sleep and mood, including the ISI. Additionally, a saliva sample was collected to evaluate genetic risk for AD, measured via Polygenic Hazard Score.³⁵ Following these assessments, participants underwent an overnight in-laboratory PSG visit to diagnose OSA, with an average follow-up time of 3.8 weeks from cognitive assessment. As per standard laboratory procedure, participants were required to sleep between 10:00 PM to 6:00 AM, adhering to a light schedule enforced by night technicians.

Measures

PSG

Participants underwent in-laboratory PSG with continuous monitoring using equipment from Nihon Khoden America (Irvine, CA) or Nox Medical USA (Alpharetta, GA). Electroencephalogram electrodes were placed at left and right frontal, central, and occipital sites along with mastoid leads. Electrooculogram and surface electromyogram electrodes were applied to measure sleep stages and arousals. Respiratory events were monitored using abdominal and chest magnetometers, pulse oximetry, and airflow sensors for both oral and nasal breathing. Participants were instructed to sleep on their backs as much as possible, with their sleep position tracked via onboard accelerometers.

A blinded registered PSG technologist scored the recordings for sleep stages and respiratory events, classifying sleep into non-rapid eye movement (NREM; stages 1–3) and REM based on the American Academy of Sleep Medicine standard criteria.³⁶ Total sleep time (TST) was calculated as the sum of these sleep stages. The AHI was measured following a well-established protocol from the American Academy of Sleep Medicine.³⁶ In this method, hypopneas are identified as a reduction in airflow exceeding 50% for more than 10 seconds, accompanied by either a 3% drop in oxygen saturation and/or a brief arousal from sleep.³⁶ Additionally, the lowest overnight blood oxygen saturation (saturation of peripheral oxygen nadir) and the total burden of respiratory event-related hypoxemia (hypoxic burden) were derived from the pulse oximetry signal.³⁷ To account for the skewed distribution of raw hypoxic burden data, values were log₁₀-transformed based on prior analytical approaches.³⁷

ISI

The ISI is a widely used clinical instrument designed to assess the severity of insomnia symptoms and the impact of insomnia on daily functioning.³⁸ The ISI consists of 7 items, each rated on a 5-point Likert scale ranging from 0 (no problem) to 4 (very severe problem), providing a total score ranging from 0–28. The items cover various aspects of insomnia, including difficulty falling asleep, difficulty staying asleep, waking up too early, satisfaction with current sleep pattern, interference with daily functioning, noticeability of impairment attributed to sleep problems, and the degree of distress caused by sleep problems.³⁸ A higher ISI score indicates more severe insomnia. The ISI is commonly categorized into 4 severity levels: no clinically significant insomnia (0–7), subthreshold insomnia (8–14), moderate insomnia (15–21), and severe insomnia (22–28). This categorization helps in identifying individuals who may require clinical intervention. In the present study, a cutoff score of 11 was selected to minimize false negatives and to balance sensitivity and specificity in clinical populations.³⁹ The ISI has demonstrated good reliability and validity in various populations, including older adults.^38,39

Verbal memory

Verbal memory was assessed using the Logical Memory IIa subtest from the Wechsler Memory Scale-Revised (WMS-LM)⁴⁰ in the Alzheimer’s Disease Cooperative Study Preclinical Alzheimer’s Cognitive Composite,⁴¹ which has shown adequate sensitivity for the detection of preclinical AD⁴² as opposed to cognitive changes related to normal aging.⁴¹ The Alzheimer’s Disease Cooperative Study Preclinical Alzheimer’s Cognitive Composite battery is comprised of the total score from the Mini-Mental State Exam (0–30 points),⁴³ the delayed recall score on the WMS-LM (0–25 points),⁴⁰ Digit Symbol Substitution Test (0–90),⁴⁴ and Free and Cued Selective Reminding Test total score (0–48).^45,46

Although other cognitive assessments were included in the larger testing battery, verbal memory was selected as the primary domain of interest as it is a cognitive domain that shows early vulnerability in age-related cognitive decline⁴⁷ and sensitivity in assessing sex differences in verbal memory, particularly in aging populations.^25,26 The WMS-LM was used to assess this domain. In the WMS-LM a short story is read, and the participant is asked to recall the story both immediately and after an approximate 30-minute delay. Verbal memory was assessed using the delayed recall score from the WMS-LM, indicating the correct number of details about the story that were recalled after the approximately 30-minute delay.

Statistical analysis

To examine whether sex moderated the relationship between COMISA status and delayed verbal memory performance, we conducted a moderated regression analysis using the PROCESS macro for SPSS.⁴⁸ Model 1 of the macro was employed, specifying COMISA status as the independent variable (X), delayed memory performance (LMD) as the outcome variable (Y), and sex as the moderator (W). Covariates included age, body mass index, AHI, APOE4+ status, and educational attainment to control for potential confounding factors that may influence memory performance. We also conducted post hoc analyses to explore whether sex differences in sleep features (eg, slow wave sleep [SWS], arousal index) might help contextualize the sex-specific effects of COMISA on verbal memory. PSG metrics included TST, NREM/REM sleep distribution, arousal index, and hypoxic burden.

Post hoc analysis

Given the observed interaction indicating that COMISA was associated with poorer verbal memory among females but not males, we conducted exploratory post hoc analyses to determine whether sleep features differed between males and females within the COMISA group. A multivariate analysis of covariance was performed to identify potential explanatory mechanisms underlying this sex-specific cognitive vulnerability. The analysis examined sex differences in PSG parameters, including sleep architecture (TST, NREM, and REM sleep), hypoxic burden, and arousal index among individuals with COMISA. Covariates included age, body mass index, AHI, APOE4+ status, and education to maintain consistency with the primary regression model.

However, given the small sample size within the COMISA group (n = 41), these analyses were exploratory in nature and underpowered to detect nuanced sex effects. Results should be interpreted with caution, and no definitive conclusions can be drawn about causality or generalizability. These preliminary findings may suggest sex-specific variation in sleep architecture among individuals with COMISA, which could provide context for the observed differences in verbal memory. Future studies with larger and more diverse samples are needed to validate these associations.

RESULTS

Participant characteristics

Demographics and participant characteristics are presented in Table 1. On average, the sample was 70 years old, predominantly non-Hispanic (90.9%), White (83.6%), and half were female (50%), 20.9% of participants were APOE4 carriers. All participants had OSA based on an AHI criterion of ≥ 5 events/h, with the majority of cases (78.2%) being moderate to severe in severity (AHI ≥ 15 events/h). The mean AHI of participants in this study was 34.2 events/h, above the severe threshold (≥ 30 events/h). Participants had a mean ISI of 8.1, with 35.5% of the sample reported subthreshold insomnia symptoms and (8–14) 15.5% reported clinical insomnia symptoms (15–21). Approximately 37.3% of participants had comorbid OSA (AHI ≥ 5 events/h) and ISI (> 11). Participants slept 4.9 ± 1.3 hours during the overnight visit between the testing sessions and displayed a standard sleep architecture for older adults characterized by a high proportion of NREM stages 1/2 and low proportion of NREM stage 3 and REM.

Males and females in our sample were demographically and clinically similar, except for expected physiological differences in sleep architecture⁴⁹ (eg, higher AHI and arousal index in males, more stage 3 sleep (N3) in females), supporting the validity of sex-stratified interpretations (see Table S1 in the supplemental material).

A moderation analysis was conducted to examine whether sex moderated the relationship between COMISA and verbal memory adjusting for age, body mass index, AHI, APOE4+ status, and educational attainment. The overall model was statistically significant, F(7,102) = 2.24, P = .037, explaining approximately 13.3% of the variance in verbal memory. A significant main effect of COMISA status was observed, such that individuals with COMISA demonstrated poorer verbal memory performance compared to those with OSA only, b = −2.82, standard error [SE] = 0.94, t = −3.01, P = .003. The main effect of sex was non-significant, b = −1.49, SE = 0.83, t = −1.80, P = .075. Importantly, however, a significant COMISA by sex interaction was observed, b = 3.43, SE = 1.35, t = 2.55, P = .012. Conditional effects analysis indicated that COMISA was significantly associated with worse verbal memory performance in females, b = −2.82, SE = 0.94, t = −3.01, P = .003, but not in males, b = 0.62, SE = 0.97, t = 0.63, P = .528 (see Figure 1).

Figure 1. —Interaction between COMISA status and sex on verbal memory performance.

Among females, those with COMISA (right blue box) exhibited significantly lower verbal memory scores compared to females with OSA only (left blue box), reflecting a clear cognitive disadvantage. In contrast, males did not show a meaningful difference in memory performance based on COMISA status. This pattern illustrates a significant COMISA by sex interaction, driven by poorer verbal memory performance in females with COMISA. Individual data points are overlaid for clarity. COMISA = comorbid insomnia and sleep apnea, OSA = obstructive sleep apnea.

Post hoc analysis

While all subgroups are shown in Figure 1 for completeness, the primary contrast of interest was between females with COMISA and males with COMISA. Univariate analyses revealed a significant effect of sex on REM sleep F(1,34) = 4.25, P = .047, η² = .111 and a marginal effect on N3 sleep F(1,34) = 3.98, P = .054, η² = .105 (see Table 2). Females had significantly more N3 sleep (mean [M] = 44.95 minutes, standard deviation = 46.62) than males (M = 11.13 minutes, standard deviation = 16.05), while males had more REM sleep (M = 52.65 minutes, standard deviation = 37.79) than females (M = 38.95 minutes, standard deviation = 31.62). A Welch’s test confirmed that the N3 difference remained significant despite unequal variances, F(1,26.57) = 10.18, P = .004. No significant differences were observed between males and females for stage 1 sleep, stage 2 sleep, TST, hypoxic burden, or arousal index after covariate adjustment. The overall multivariate model was significant for sex, F(6,34) = 2.88, P = .022, η² = .337, indicating a multivariate effect of sex on sleep-related outcomes.

Table 2.

—Post hoc MANCOVA results for males and females with COMISA.

Dependent Variable	Females (n = 22) M (SD)	Males (n = 19) M (SD)	F(1,34)	P	Partial η2	Direction of Effect
TST (minutes)	297.9 (58.0)	278.1(103.3)	0.07	.796	0.002	ns
N1 (minutes)	48.7 (29.0)	58.2 (34.0)	0.01	.918	0.000	ns
% of TST	17%	23%
N2 (minutes)	165.2 (52.7)	156.1 (67.0)	0.15	.700	0.004	ns
% of TST	55%	55%
N3 (minutes)	45.0 (46.6)	11.1 (16.0)	3.98	.054	0.105	↑ in females
% of TST	15%	4%
REM (minutes)	39.0 (31.6)	52.6 (37.8)	4.25	.047	0.111	↓ in females
% of TST	13%	18%
Hypoxic burden	57.3 (45.2)	112.8 (90.9)	0.97	.332	0.028	ns
Arousal index	26.8 (17.2)	31.7 (20.2)	0.64	.430	0.018	ns

COMISA, M = mean, MANCOVA = multivariate analysis of covariance, N1 = stage 1 sleep, N2 = stage 2 sleep, N3 = stage 3 sleep, ns = non-significant, REM = rapid eye movement, SD = standard deviation, TST = total sleep time.

DISCUSSION

This study provides novel evidence that COMISA is associated with poorer verbal memory performance and that this effect is significantly moderated by sex. Specifically, COMISA, rather than OSA-only, was linked to worse memory outcomes in older females but not older males. This interaction remained significant even after adjusting for age, body mass index, AHI, APOE4+ status, and education, and was further contextualized by differences in sleep architecture. Within the COMISA group, females demonstrated significantly more SWS and less REM sleep than males.⁵⁰ Several of these differences, including increased SWS in females and greater stage 1 sleep and arousals in males, are consistent with previously reported sex-based patterns in sleep among individuals with OSA (eg,⁴⁹). However, the finding of reduced REM sleep in females with COMISA compared to males is not typical of normative sex differences, as older females generally have similar or greater REM sleep than males. Although comparisons between females with COMISA and OSA only revealed no significant difference in PSG parameters, including REM or SWS, this unexpected pattern may indicate a sex-specific vulnerability in the context of COMISA, though further research is needed to determine whether this reflects a true deviation or sample-specific variation. Overall, the observed differences in sleep architecture within the COMISA group may reflect broader sex-related patterns, but our data cannot establish whether these differences are specific to females with COMISA. These findings tentatively highlight a possible sex-specific cognitive vulnerability in COMISA, though further investigation is needed to determine the role of sleep architecture in this association.

Our findings suggest that co-occurring insomnia symptoms alongside OSA may be associated with additional cognitive vulnerability beyond OSA alone, particularly among older females. This finding aligns with a growing body of literature suggesting that the combined effects of insomnia and OSA may be more detrimental to health than either condition alone. For example, Wulterkens et al (2023) described COMISA as involving the additive burden of both conditions, resulting in more fragmented sleep and potentially worse clinical outcomes than either disorder alone.⁵¹ This characterization aligns with our findings, which revealed sex differences in sleep architecture within COMISA and cognitive deficits in females with COMISA. Liu et al (2022) further reinforce this view, reporting greater reductions in N3, or SWS and REM sleep and heightened psychiatric burden in COMISA populations.¹² Together, these studies suggest that the co-occurrence of OSA and insomnia may intensify disruptions to sleep homeostasis and magnify their impact on memory and cognition, rather than acting through fully independent mechanisms.

Central to understanding the cognitive vulnerability observed in females with COMISA is the sex difference in sleep architecture. Females in our sample exhibited more SWS and less REM sleep compared to males. Although Redline et al (2004) reported that females generally have more SWS and REM sleep than males,⁴⁹ our findings suggest that this advantage, as it relates to REM sleep, may be reversed in the context of COMISA. One possibility is that this pattern reflects the higher prevalence of REM-predominant OSA, which is a form of sleep apnea in which respiratory events are disproportionately concentrated during REM sleep.⁵² Bahammam et al (2020) demonstrated that female sex is a strong independent predictor of REM-predominant OSA. In this context, the REM sleep stage in females may be especially vulnerable to disruption from both apneic events and insomnia-related hyperarousal, leading to suppressed or fragmented REM sleep. Thus, while females are biologically predisposed to REM-predominant OSA, the co-occurrence of insomnia may exacerbate REM sleep disruption. In our sample, although REM sleep was reduced in females relative to males with COMISA and co-occurred with poorer memory performance, this difference was not observed when comparing females with COMISA to females with OSA only, limiting our ability to draw definitive conclusions about REM disruption as a mechanism.

Although SWS has traditionally been considered beneficial for memory consolidation, particularly in younger adults,⁵³ its functional role in older adults appears more nuanced. In the present study, females with COMISA exhibited approximately 34 minutes more and 13 minutes less REM sleep than males with COMISA, suggesting that this finding was not simply a one-to-one reallocation of time between stages. While this pattern may reflect broader shifts in sleep regulation or compensatory responses, it may also reflect a broader shift in sleep regulation or a compensatory increase in SWS that ultimately fails to offset the cognitive consequences of reduced REM sleep. While we did not directly assess the relationship between SWS and memory, this co-occurrence raises the possibility that SWS alone may be insufficient for memory support in aging populations, particularly when REM sleep, critical for emotional regulation and integrative memory processes, is concurrently reduced.^54,55 In fact, some research suggests that in older adults, the well-established positive relationship between SWS and episodic memory seen in younger individuals was often diminished or even reversed.⁵³ Similarly, others have suggested that while SWS supported prospective memory in younger adults, no such relationship was observed in older adults, suggesting an age-related attenuation of the relationship between SWS and memory consolidation.⁵⁶ Taken together, these findings raise the possibility that increased SWS co-occurs with reduced REM sleep, though the functional implications of this pattern remain unclear. While one interpretation is that this reflects a compensatory response to REM disruption, it is also possible that increased SWS reflects broader age- or sex-related changes in sleep physiology, such as electroencephalogram slowing, rather than a COMISA-specific effect. Indeed, some studies have found more SWS in older females with greater cognitive impairment,⁵⁷ suggesting that higher SWS is not always indicative of better cognitive function. Whether this shift is adaptive or insufficient to preserve cognitive function remains an open question. This result underscores the importance of considering the broader architecture and quality of sleep, rather than individual sleep stages in isolation, when assessing cognitive risk in aging populations.

Moreover, a growing body of evidence underscores the importance of REM sleep in memory consolidation, particularly for verbal information. Researchers have demonstrated that REM-specific oxygen desaturation was linked to worse verbal memory, especially in individuals at elevated risk for AD (eg, APOE4 carriers).^58,59 Although sex did not significantly moderate this effect in their sample, the recognition that REM-predominant OSA is more common in females supports the idea that REM disruption may carry disproportionate cognitive consequences for females. Similarly, Casey et al (2016) showed that selective REM sleep deprivation, more than SWS deprivation, significantly impaired verbal episodic memory recall in healthy adults, underscoring REM’s unique role in memory consolidation.⁶⁰ Extending these findings, researchers demonstrated in a large longitudinal cohort that each 1% reduction in REM sleep was associated with a 9% increase in risk of incident dementia,²⁰ highlighting the relevance of REM sleep not only for memory but also for long-term cognitive health. Our finding that females with COMISA had reduced REM sleep relative to males with COMISA is consistent with prior framework emphasizing the importance of REM sleep for cognition, though causal pathways cannot be established here. Moreover, given that PSG was derived from a single night and no REM differences were observed between females with COMISA and females with OSA only, we caution against over-interpreting this finding as evidence of REM-specific vulnerability.

Limitations

Despite the strength of these findings, several limitations should be acknowledged. The cross-sectional nature of the study precludes causal inference regarding the directionality of sleep disturbances and memory impairment. An additional limitation of this study is the relatively low TST observed across participants (M = 4.9 hours), which may reflect the constraints of the in-laboratory protocol rather than participants’ habitual sleep duration. Additionally, sleep architecture was assessed using a single night of PSG, which may not fully capture participants’ typical sleep patterns due to first-night effects or night-to-night variability. Future research may benefit from incorporating home-based sleep monitoring to reflect better real-world sleep patterns and circadian influences. These controlled conditions, while necessary for standardization, may have restricted our ability to capture individual differences in chronotype or habitual sleep timing. Our sample was predominantly non-Hispanic White and highly educated, which may limit the generalizability of our findings to more diverse populations. Similarly, this was a secondary analysis of a cohort of individuals with OSA, and we were not able to include an insomnia-only or healthy control group. As such, we cannot determine whether the cognitive or PSG differences observed in females with COMISA are specific to the interaction between insomnia and OSA, or whether similar effects might also be seen in individuals with insomnia alone. Importantly, the COMISA group sample size (n = 41) limits the statistical power to detect sex differences, generalizability, and findings related to sex should be interpreted with caution. While post hoc analyses revealed sex differences in sleep architecture among individuals with COMISA, we did not test a formal mediated moderation model. As such, the role of SWS or REM sleep as mechanisms underlying the COMISA-by-sex interaction on memory remains speculative and warrants future investigation.

Implications and future directions

The present findings have important implications for both clinical care and research on cognitive aging. First, they add to a growing body of evidence suggesting that COMISA may represent a distinct clinical phenotype rather than simply the co-occurrence of insomnia and OSA,¹² particularly among older females, but larger samples are needed to confirm this association. Future work should also evaluate whether the patterns observed here extend to individuals with insomnia alone, without co-occurring OSA. While this study focused on older adults with OSA, prior research suggests that verbal memory deficits may also be present in individuals with insomnia alone and even in comparisons between poor and healthy sleepers. Experimental and meta-analytic work has shown that adults with primary insomnia exhibit small-tomoderate impairments in episodic memory relative to good sleepers, with such deficits linked to poor sleep continuity and increased hyperarousal.^61,62 Similarly, in a large population-based study, Zhao et al (2023) found that insomnia disorder was associated with a higher likelihood of self-reported memory decline in middle-aged and older adults.⁶³ In contrast, older adults without insomnia or OSA who report 6–8 hours of sleep tend to show better episodic memory performance, and reduced slow-wave sleep has been linked to poorer memory across samples.⁶⁴ Additionally, Gervais et al (2024) found that self-reported sleep disturbances were associated with reduced verbal memory and smaller entorhinal cortex volume in middle-aged females who underwent early ovarian removal, further suggesting broader cognitive consequences of poor sleep.⁶⁵ These findings indicate that while COMISA may represent a higher-risk phenotype, memory impairment associated with sleep disruption is not exclusive to this group. Future work that includes insomnia-only and healthy control groups will be essential to determine whether COMISA confers distinct or additive effects on memory and sleep physiology. While this study was limited to an OSA sample, prior research suggests that COMISA may exert additive or synergistic effects on cognitive function and sleep architecture beyond either condition independently.^8,9,27,51

Understanding whether COMISA represents a distinct phenotype or merely reflects the cumulative burden of 2 common disorders will be critical to guiding both clinical diagnosis and targeted treatment development. Including insomnia-only and healthy control groups in future research will help clarify whether the observed verbal memory deficits and sleep architecture patterns are unique to COMISA or generalizable across sleep disorders. Identifying sex-specific alterations in sleep architecture, specifically, reduced REM and increased SWS in females with COMISA, suggests that stage-specific sleep disruptions may be important in understanding memory function in this population and should be considered in diagnostic and treatment strategies. This reinforces the clinical utility of incorporating REM-sensitive indices (eg, REM AHI, REM oxygen desaturation index) into sleep evaluations, especially in older adults. Moreover, because REM tends to occur more in the latter half of the night and many individuals with OSA discontinue continuous positive airway pressure use partway through the night, REM-related events may go untreated. This issue may limit the effectiveness of therapy and contribute to persistent cognitive risk, especially in females.

Although all participants in this study were cognitively normal, verbal memory is a domain sensitive to age-related cognitive decline and can serve as an early marker of vulnerability. REM sleep, in particular, has been implicated in memory consolidation and neural restoration, and its disruption has been associated with poorer cognitive outcomes in older adults. Our finding that females with COMISA exhibit reduced REM sleep raises the possibility that REM-related vulnerability may be one pathway through which sleep fragmentation contributes to subtle cognitive changes, though we cannot infer causality or directionality from the current data. As discussed, prior research has shown that REM-specific oxygen desaturation can negatively affect verbal memory, particularly in individuals with genetic risk for AD.⁵⁸ While our study did not assess biomarkers or longitudinal outcomes, these findings highlight the value of further examining whether sleep architecture predicts trajectories of memory performance in at-risk populations.

Future research should employ longitudinal and multimodal designs to assess whether changes in REM and SWS over time predict cognitive decline, and whether these associations differ by sex or APOE genotype. Neuroimaging approaches could help elucidate how altered sleep stages influence memory-related brain structures, such as the hippocampus and medial temporal lobes.²⁸ Additionally, more comprehensive cognitive batteries should be used to determine whether COMISA affects other domains beyond verbal memory, including executive function or emotional regulation. Finally, future studies should formally test whether disrupted sleep stages mediate the association between COMISA and memory outcomes, while considering psychosocial and hormonal factors, particularly in females.⁶⁶

CONCLUSIONS

Overall, this study contributes to the growing body of literature suggesting that COMISA may be associated with cognitive vulnerability in aging populations, with potential sex differences in how this risk manifests. Specifically, poorer verbal memory performance was observed in females with COMISA compared to females with OSA only. While we did not see differences in sleep architecture between females with COMISA compared to females with OSA only, we did observe differences between males and females with COMISA, such that the females showed reduced REM and increased SWS compared to their male counterparts. Rather than drawing mechanistic conclusions, these findings highlight the importance of considering sex-specific patterns in sleep and cognition when evaluating COMISA. Future work should build on these observations to clarify whether stage-specific sleep disruptions contribute meaningfully to memory outcomes. A more precise understanding of these pathways could ultimately inform more tailored diagnostic and therapeutic approaches for older adults with COMISA.

Table 3.

—Post hoc MANCOVA results for females with COMISA and OSA only.

Dependent Variable	COMISA (n = 22) M (SD)	OSA only (n = 33) M(SD)	F(1,48)	P	Partial η2	Direction of Effect
TST (minutes)	297.9 (58.0)	291.9 (88.78)	0.26	.616	0.005	ns
N1 (minutes)	48.73 (29.0)	56.0 (34.0)	0.71	.403	0.015	ns
% of TST	17%	22%
N2 (minutes)	165.2 (52.7)	152.8 (65.8)	1.88	.176	0.038	ns
% of TST	55%	51%
N3 (minutes)	45.0 (46.7)	39.7 (42.2)	0.00	.994	0.000	ns
% of TST	15%	13%
REM (minutes)	39.0 (31.6)	43.4 (28.3)	0.27	.608	0.006	ns
% of TST	13%	14%
Hypoxic burden	57.3 (45.2)	56.0 (48.9)	0.00	.990	0.000	ns
Arousal index	26.8 (2)	29.7 (15.7)	0.88	.352	0.018	ns

COMISA, M = mean, MANCOVA = multivariate analysis of covariance, N1 = stage 1 sleep, N2 = stage 2 sleep, N3 = stage 3 sleep, ns = non-significant, OSA = obstructive sleep apnea, REM = rapid eye movement, SD = standard deviation, TST = total sleep time.

ABBREVIATIONS

AD

Alzheimer’s disease

AHI

apnea-hypopnea index

APOE4

apolipoprotein E ε4

COMISA

comorbid insomnia and sleep apnea

ISI

Insomnia Severity Index

M

mean

NREM

non-rapid eye movement

OSA

obstructive sleep apnea

PSG

polysomnography

REM

rapid eye movement

SE

standard error

SWS

slow wave sleep

TST

total sleep time

WMS-LM

Logical Memory IIa subtest from the Wechsler Memory Scale-Revised