Combination of dysfunctional beliefs about sleep and excessive daytime sleepiness as a psychobehavioral characteristic of comorbid insomnia and sleep apnea

Abstract

Comorbid insomnia and sleep apnea (COMISA) is associated with poorer clinical outcomes than chronic insomnia disorder (CID) or obstructive sleep apnea (OSA) alone. However, the psychobehavioral characteristics that differentiate COMISA remain unclear. We aimed to examine the combined roles of dysfunctional beliefs and attitudes about sleep (DBAS) and excessive daytime sleepiness (EDS) in COMISA. We analyzed data from 283 patients (42 with CID alone, 166 with OSA alone, 75 with COMISA) from the National Center of Neurology and Psychiatry database, a component of a polysomnography databank for sleep disorders in Japan. DBAS was assessed using the 16-item DBAS scale (0–160 points), and EDS was measured using the Epworth Sleepiness Scale (ESS; 0–24 points). Participants were categorized into four groups based on median ESS and DBAS scale scores (medians: 90 and 9, respectively): low ESS/low DBAS, low ESS/high DBAS, high ESS/low DBAS, and high ESS/high DBAS. Logistic regression analyses, adjusting for multiple covariates, such as age, sex, and body mass index, were conducted. The high ESS/high DBAS phenotype was significantly associated with COMISA relative to the low ESS/low DBAS reference group in comparisons with both CID and OSA. The low ESS/high DBAS phenotype was also significantly associated with COMISA compared with OSA. Coexisting elevated DBAS and EDS appear to represent a distinct psychobehavioral phenotype distinguishing COMISA from CID or OSA alone. Assessing DBAS in patients with OSA and evaluating daytime sleepiness in those with CID may support earlier identification of COMISA and inform targeted management.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-50792-9.

Introduction

Chronic insomnia disorder (CID) and obstructive sleep apnea (OSA) are highly prevalent sleep disorders. CID affects approximately 6–10% of the general population^1,2, whereas OSA affects approximately 26%³. Notably, 29–67% of individuals with CID have coexisting OSA, and 39–58% of those with OSA report insomnia symptoms⁴, indicating that comorbid insomnia and sleep apnea (COMISA) is common.

COMISA is associated with poorer clinical courses and outcomes than either disorder alone. Patients with COMISA show reduced therapeutic responses to hypnotic medications^5,6, limited improvements in objective sleep parameters, such as wake after sleep onset, slow-wave sleep, and rapid eye movement sleep following cognitive behavioral therapy for insomnia (CBT-I)⁷, and marked heterogeneity in insomnia severity⁸. Furthermore, COMISA exhibits more severe psychological symptoms, such as dysfunctional beliefs and attitudes about sleep (DBAS), depression, and anxiety^9,10 and poorer adherence to continuous positive airway pressure (CPAP) therapy, relative to patients with OSA alone^11,12. Despite these reported clinical features, studies that systematically characterize the psychobehavioral profile of COMISA are limited¹³. The psychological and behavioral patterns that differentiate COMISA from CID or OSA alone, therefore, remain poorly understood. Clarifying how these features manifest when both disorders co-occur may improve our understanding of the unique COMISA phenotype and support more accurate clinical identification and management.

The psychobehavioral characteristics of CID commonly include elevated DBAS¹⁴, and efforts to intentionally initiate sleep often exacerbate insomnia symptoms¹⁵. Such cognitive processes are central therapeutic targets in CBT-I. However, not all individuals with CID endorse strong DBAS¹⁶, and cognitive intervention alone shows smaller effects on insomnia severity than do behavioral interventions, such as sleep restriction and stimulus control¹⁷. Moreover, the total DBAS score showed only a moderate correlation with insomnia severity when assessed by the Insomnia Severity Index (ISI) in CID¹⁸. These observations suggest that although DBAS reflect an important theoretical mechanism of CID, they do not consistently serve as a reliable phenotypic marker across all patients with CID.

In contrast, the primary pathophysiology of OSA involves intermittent hypoxemia and sleep fragmentation caused by recurrent upper-airway obstruction. These disruptions impair restorative sleep processes and lead to excessive daytime sleepiness (EDS)¹⁹. EDS is widely recognized as a hallmark symptom of OSA and constitutes an important psychobehavioral component of this disorder²⁰. However, subjective sleepiness does not necessarily correlate with objective indices, such as the apnea–hypopnea index (AHI)²¹ or mean sleep latency on the multiple sleep latency test²². In addition, only approximately 40% of patients with OSA report EDS²³, indicating substantial heterogeneity in its presentation. These findings suggest that although EDS is a pathophysiologically expected consequence of OSA, it does not consistently function as a phenotypic marker across all patients with OSA.

Although DBAS and EDS do not clearly emerge as phenotypic markers when CID and OSA occur separately, their coexistence may reinforce these features more prominently through their mutual interaction. Specifically, sleep fragmentation in OSA may sustain DBAS in patients with CID by creating repeated nocturnal experiences that reinforce maladaptive sleep-related cognitions. Conversely, OSA-related sleep disruption may maintain daytime sleepiness at levels comparable to those observed in OSA alone, even in the context of coexisting insomnia. Patients with COMISA may therefore exhibit OSA-related EDS to a degree comparable to or even greater than that in patients with OSA, while simultaneously endorsing CID-related DBAS to a degree comparable to or even greater than that in patients with CID. If this holds true, neither DBAS nor EDS alone can distinguish COMISA from one of its constituent disorders; however, the co-occurrence of both elevated DBAS and EDS—a pattern unlikely in either CID or OSA alone—may serve as a distinctive marker of COMISA. Therefore, we aimed to investigate whether DBAS and EDS together constitute a distinctive psychobehavioral phenotype that differentiates COMISA from CID or OSA alone.

Methods

Participants

This study used data from the National Center of Neurology and Psychiatry database, which is part of a polysomnography (PSG) databank for sleep disorders in Japan²⁴. The provided data were from 879 patients who underwent PSG between January 2020 and December 2023. From these data, those from 370 patients who had not received diagnoses of sleep disorders other than CID or OSA were excluded (remaining n = 509). Subsequently, patients with comorbid sleep disorders other than CID or OSA were excluded (24 patients with hypersomnia, 17 with circadian rhythm sleep–wake disorders, 56 with parasomnias, and 89 with sleep-related movement disorders; remaining n = 346). Finally, 283 patients were included after excluding those who did not complete the DBAS scale (n = 23) and Epworth Sleepiness Scale (ESS) scale as a measure of EDS (n = 50), including 10 patients with missing data on both scales. Subsequently, we divided the 283 patients into three groups: CID alone (n = 42), OSA alone (n = 166), and COMISA (n = 75), defined as the concurrent diagnoses of CID and OSA (n = 75) (Supplementary Fig S1). Specifically, the exclusion criteria were as follows: (1) absence of a diagnosis of CID or OSA; (2) presence of comorbid sleep disorders other than CID or OSA, including hypersomnia, circadian rhythm sleep–wake disorders, parasomnias, and sleep-related movement disorders; and (3) incomplete DBAS scale or ESS.

This study was approved by the Ethics Committee of the National Center of Neurology and Psychiatry, Tokyo, Japan and conducted in accordance with the principles of the Declaration of Helsinki. This study utilized anonymized data from a PSG databank for sleep disorders in Japan for scientific research purposes; therefore, the informed consent procedure was waived. Nevertheless, in the original study of a PSG databank for sleep disorders in Japan, informed consent was obtained using two different approaches according to the study design: (1) for retrospective data collection, an opt-out consent method was employed through notification on our institutional website, and (2) for prospective data collection, written informed consent was obtained from all individual participants prior to study enrollment.

Measures

Primary outcome

The primary outcome was COMISA, defined as concurrent diagnoses of CID and OSA in the absence of other sleep disorders, compared with CID or OSA alone. Diagnoses of CID and OSA were based on the International Classification of Sleep Disorders, Third Edition (ICSD-3)²⁵. All clinical diagnoses were established by the board-certified physicians of the Japanese Society of Sleep Research and the Japanese Society of Psychiatry and Neurology, based on comprehensive clinical evaluation, including PSG findings. Specifically, CID was diagnosed when patients reported difficulties initiating sleep, maintaining sleep, or early morning awakening occurring at least three nights per week for at least 3 months, accompanied by daytime functional impairment, consistent with ICSD-3 criteria. OSA was diagnosed based on an AHI of ≥ 5 events per hour with associated daytime functional impairment symptoms or an AHI of ≥ 15 events per hour regardless of symptoms, in accordance with ICSD-3 criteria.

Psychological and behavioral predictors

The DBAS scale²⁶ was used as one exposure variable. It consists of 16 items, each scored from 0 to 10 points, with a maximum total score of 160. Higher scores indicate stronger dysfunctional beliefs and attitudes about sleep.

The ESS, which assesses EDS, was used as the second exposure variable. It evaluates the likelihood of dozing off in eight everyday situations, with each item rated from 0 to 4 points. The maximum score is 24, with higher scores indicating stronger daytime sleepiness²⁷. These questionnaires were evaluated within a period extending from 3 months before to 1 month after the PSG recording date.

Both the DBAS and ESS scores were divided into high and low categories based on the overall median values (90 and 9, respectively). These categories were then combined into four groups: low ESS/low DBAS, low ESS/high DBAS, high ESS/low DBAS, and high ESS/high DBAS.

Other variables

Demographic data included age, sex, body mass index (BMI), alcohol consumption habits (current habitual alcohol consumption–yes/no), and smoking habits (current habitual cigarette smoking–yes/no). Clinical information included the presence of current conditions, such as depressive disorders, anxiety disorders, neurocognitive disorders, dyslipidemia, hypertension, and diabetes mellitus, as well as the use of medications, including sleep medications, antidepressants, and antipsychotics. Participants completed several self-administered questionnaires during outpatient visits. These included the ISI²⁸ to assess insomnia symptom severity, Patient Health Questionnaire-9 (PHQ-9)²⁹ to assess depressive symptoms, and Generalized Anxiety Disorder-7 (GAD-7)³⁰ to assess anxiety symptoms. Subjective sleep duration was self-reported by patients in hours and minutes on the morning following the PSG assessment.

Objective sleep parameters were obtained from overnight PSG, including sleep duration, sleep onset latency, sleep efficiency, wake after sleep onset, and the AHI. PSG data were collected using digital polygraphy systems (Alice 5 or Alice 6, Respironics, Murrysville, PA, USA) in accordance with recommendations from the American Academy of Sleep Medicine (AASM)³¹. Recordings included electroencephalography (F3–M2, F4–M1, C3–M2, C4–M1, O1–M2, O2–M1), bilateral electrooculography (E1–M2, E2–M1), electrocardiography, and surface electromyography of the chin and submental muscles and bilateral anterior tibial muscles. Respiratory signals included nasal airflow recorded with thermistors and nasal pressure sensors, thoracoabdominal respiratory effort, oxygen saturation measured by transcutaneous fingertip pulse oximetry, and snoring recorded with microphones.

All PSG studies were manually scored by certified clinical laboratory technicians of the Japanese Society of Sleep Research. Sleep stages and sleep-related events were scored according to the AASM guidelines³¹. The AHI was defined as the total number of apnea and hypopnea events divided by total sleep time in hours. The oxygen desaturation index was defined as the number of arterial oxygen saturation decreases of ≥ 3% per hour of sleep.

Data analysis

Patient demographic variables, questionnaire scores, and PSG parameters were summarized as the mean ± standard deviation or number (%) for each disease group (COMISA, CID, and OSA) and for each ESS and DBAS category (low ESS/low DBAS, low ESS/high DBAS, high ESS/low DBAS, and high ESS/high DBAS). Group differences were examined using analysis of variance, followed by pairwise comparisons with Bonferroni correction.

Logistic regression analyses were conducted to evaluate whether the coexistence of DBAS and EDS functions as a distinctive psychobehavioral phenotype that differentiates COMISA from CID or OSA. The dependent variable was COMISA versus either OSA or CID, and the exposure variable was the combined DBAS and ESS classification. Three logistic regression models were estimated using established covariates. The first model was adjusted for age and sex because both factors may influence the pathophysiology of OSA and CID^1,32–34. The second model added BMI to account for the potential role of body composition in CID, OSA, and COMISA ³⁵. Because insomnia severity, depressive symptoms, and anxiety symptoms are known to potentially act as confounders, the ISI, PHQ-9, and GAD-7 were included as additional covariates when comparing CID and COMISA. When comparing OSA and COMISA, the AHI was included because OSA severity could act as a confounder. To assess the robustness of the findings, the bootstrap method with 1,000 simple resamples was used to estimate regression coefficients and percentile-based 95% confidence intervals (CIs). Bonferroni correction was applied to the logistic regression analyses, with statistical significance defined as a corrected p-value of < 0.05 for two-sided tests. Two primary comparisons were conducted: COMISA versus OSA and COMISA versus CID. All analyses were performed using SPSS Statistics, version 23 (IBM Corp, Armonk, NY, USA).

Several sensitivity analyses were conducted to assess the robustness of the primary findings. First, to directly compare DBAS and ESS levels between COMISA and each comparison group, multiple regression analyses were conducted with DBAS or ESS as dependent variables and diagnostic group as the independent variable, using the same covariate structure as the primary analyses. Second, to address potential contamination of the OSA group by psychiatric comorbidities, two complementary approaches were employed: (a) hypnotics, antidepressants, or/and use antipsychotics was included as an additional covariate, and (b) hypnotics, antidepressants, or/and use antipsychotics users were excluded from the OSA group. Third, to evaluate whether the observed associations were dependent on the median-based categorization, the analyses were repeated using established clinical cutoff values (ESS ≥ 10 for excessive daytime sleepiness; DBAS scale score ≥ 68.8, corresponding to a per-item average of 4.3 on the DBAS-16, which has been proposed as the clinical threshold for dysfunctional beliefs associated with insomnia)³⁶. Fourth, to address the potential confounding effect of OSA severity heterogeneity between the COMISA and OSA groups, the analyses were repeated after excluding patients with mild OSA (AHI < 15 events/hour) from both the COMISA and OSA groups, restricting the comparison to patients with moderate-to-severe OSA (AHI ≥ 15 events/hour).

Results

Patient characteristics

Table 1 presents demographic and clinical characteristics of each disease group. No significant differences were observed in smoking status, alcohol consumption habits, objective sleep duration, sleep onset latency, or wake after sleep onset. Post hoc analyses showed that, compared with the CID group, the COMISA group had significantly fewer female patients and a higher age, AHI, and prevalence of hypertension. Compared with the OSA group, the COMISA group had significantly greater use of sleep medications, greater age, ISI, and DBAS scale scores, and a lower BMI and AHI. There was neither CPAP nor oral appliance users in both OSA and COMISA patients.

Table 1.

Demographic and clinical characteristics of study patients by disease group.

	All patients (n = 283)	COMISA (n = 75)	CID (n = 42)	OSA (n = 166)	F/χ2 (p-value)
Age, mean ± SD, y	51.7 (16.9)	57.0 (18.1)	44.9 (15.8)**	51.1 (15.9)***	7.483 (0.001)*
Female, n (%)	111 (39.4)	24 (32.4)	24 (57.1)**	63 (38.0)	7.190 (0.027)*
BMI, mean ± SD, kg/m2	24.6 (4.9)	23.5 (4.8)	22.7 (3.8)	25.7 (4.9)***	9.636 (< 0.001)*
Current smoking, n (%)	37 (13.3)	4 (5.5)	5 (12.2)	28 (17.0)	5.856 (0.053)
Daily alcohol consumption, n (%)	37 (13.3)	7 (9.6)	7 (17.1)	23 (13.9)	1.440 (0.487)
Depressive disorders, n (%)	51 (18.0)	16 (21.3)	10 (23.8)	25 (15.1)	2.495 (0.287)
Anxiety disorders, n (%)	16 (5.7)	6 (8.0)	3 (7.1)	7 (4.2)	1.591 (0.451)
Neurocognitive disorders, n (%)	5 (1.8)	0 (0.0)	0 (0.0)	5 (3.0)	3.587 (0.166)
Diabetes, n (%)	26 (9.2)	4 (5.3)	1 (2.4)	21 (12.7)	6.054 (0.048)*
Dyslipidemia, n (%)	56 (19.8)	14 (18.7)	3 (7.1)	39 (23.5)	5.727 (0.057)
Hypertension, n (%)	61 (21.6)	18 (24.0)	2 (4.8)**	41 (24.7)	8.240 (0.016)*
Hypnotic use, n (%)	90 (31.8)	34 (45.3)	22 (52.4)	34 (20.5)***	24.341 (< 0.001)*
Antidepressant use, n (%)	47 (16.6)	10 (13.3)	7 (16.7)	30 (18.1)	0.838 (0.658)
Antipsychotic use, n (%)	34 (12.0)	11 (14.7)	6 (14.3)	17 (10.2)	1.198 (0.549)
ESS score (0–24), mean ± SD	9.8 (6.0)	9.4 (6.1)	8.3 (5.5)	10.3 (6.0)	1.982 (0.140)
DBAS score (0–24), mean ± SD	84.9 (32.8)	95.6 (26.0)	94.2 (25.9)	77.7 (35.3)***	10.238 (< 0.001)*
ISI score (0–15), mean ± SD	16.2 (6.2)	19.5 (4.1)	20.6 (4.4)	13.5 (6.1)***	48.835 (< 0.001)*
PHQ-9 score (0–27), mean ± SD	10.7 (6.4)	10.7 (6.0)	13.3 (6.4)	10.0 (6.5)	4.533 (0.012)*
GAD-7 score (0–21), mean ± SD	7.7 (5.2)	7.9 (4.9)	8.1 (5.1)	7.4 (5.3)	0.320 (0.727)
Subjective TST, mean ± SD, min	353.8 (109.8)	335.7 (105.8)	322.5 (102.8)	369.6 (111.0)	3.966 (0.020)*
Objective TST, mean ± SD, min	397.8 (76.9)	391.3 (75.6)	389.2 (70.1)	402.8 (79.1)	0.883 (0.415)
SL, mean ± SD, min	14.5 (22.8)	13.8 (16.0)	20.1 (28.0)	13.4 (23.8)	1.490 (0.227)
WASO, mean ± SD, %	15.3 (11.8)	16.6 (11.5)	12.1 (9.6)	15.4 (12.3)	2.042 (0.132)
SE, mean ± SD, %	80.4 (12.9)	79.7 (12.5)	80.7 (12.5)	80.7 (13.3)	0.162 (0.850)
AHI, events/h, mean ± SD	22.0 (21.0)	19.6 (18.0)	2.3 (1.5)**	28.1 (21.6)***	31.506 (< 0.001)*
AI, events/h, mean ± SD	7.6 (12.9)	5.4 (9.7)	0.5 (0.5)	10.4 (14.8)***	12.1 (< 0.001)*
HI, events/h, mean ± SD	14.4 (13.3)	14.2 (10.3)	1.8 (1.3)**	17.7 (14.2)	28.7 (< 0.001)*
SpO2 90% total duration, min, mean ± SD	7.7 (33.1)	5.0 (15.5)	0.7 (2.0)	10.7 (41.8)	1.9 (0.153)
ESS/DBAS combination					19.244 (0.004)*
Low ESS/Low DBAS, n (%)	88 (31.1)	17 (22.7)	13 (31.0)	58 (34.9)
Low ESS/high DBAS, n (%)	65 (23.0)	25 (33.3)	14 (33.3)	26 (15.7)
High ESS/Low DBAS, n (%)	59 (20.8)	9 (12.0)	7 (16.7)	43 (25.9)
High ESS/High DBAS, n (%)	71 (25.1)	24 (32.0)	8 (19.0)	39 (23.5)

Abbreviations: BMI, body mass index; ESS, Epworth Sleepiness Scale; DBAS, Dysfunctional Beliefs and Attitudes about Sleep; ISI, Insomnia Severity Index; PHQ-9, Patient Health Questionnaire-9; GAD-7, Generalized Anxiety Disorder-7; TST, total sleep time; SL, sleep latency; WASO, wake after sleep onset; SE, sleep efficiency; AHI, apnea–hypopnea index.

Body mass index is calculated as weight in kilograms divided by height in meters squared. * p < 0.05 in ANOVA, ** corrected p < 0.05 in post-hoc analysis between COMISA and CID with Bonferroni correction, *** p < 0.05 in post-hoc analysis between COMISA and OSA with Bonferroni correction.

Table 2 presents demographic and clinical characteristics across the ESS and DBAS combinations. No significant differences were observed in the proportion of female patients, alcohol consumption habits, objective sleep duration, or the AHI. Post hoc analyses indicated that the high ESS/high DBAS group had significantly lower age and wake after sleep onset percentage; higher proportion of sleep efficiency, and prevalence of depression; greater use of sleep medications, antidepressants, and antipsychotics; higher ISI, PHQ-9, and GAD-7 scores; and a longer objective sleep time, compared with the low ESS/low DBAS group.

Table 2.

Demographic and clinical characteristics of study patients by ESS/DBAS combination.

	All patients (n = 283)	Low ESS		High ESS		F/χ2 (p-value)
		Low DBAS (n = 88)	High DBAS (n = 65)	Low DBAS (n = 59)	High DBAS (n = 71)
Age, mean ± SD, y	51.7 (16.9)	59.6 (16.1)**	50.3 (16.4)	49.8 (15.1)	44.9 (16.1)	11.8 (< 0.001)*
Female, n (%)	111.0 (39.4)	40 (45.5)	25 (38.5)	18 (30.5)	28 (40.0)	3.340 (0.342)
BMI, mean ± SD, kg/m2	24.6 (4.9)	23.9 (4.0)	23.7 (4.9)	25.3 (4.8)	25.8 (5.6)	3.234 (0.023)*
Current smoking, n (%)	37 (13.3)	7 (8.1)	14 (21.9)	4 (6.8)	12 (17.1)	9.161 (0.027)*
Daily alcohol consumption, n (%)	37 (13.3)	14 (15.9)	9 (14.3)	7 (11.9)	7 (10.1)	1.276 (0.735)
Depressive disorders, n (%)	51 (18.0)	10 (11.4)**	12 (18.5)	9 (15.3)	20 (28.2)	7.903 (0.048)*
Anxiety disorders, n (%)	16 (5.7)	7 (8.0)	1 (1.5)	5 (8.5)	3 (4.2)	4.089 (0.252)
Neurocognitive disorders, n (%)	5 (1.8)	3 (3.4)	0 (0.0)	2 (3.4)	0 (0.0)	4.709 (0.194)
Diabetes, n (%)	26 (9.2)	11 (12.5)	6 (9.2)	2 (3.4)	7 (9.9)	3.573 (0.311)
Dyslipidemia, n (%)	56 (19.8)	22 (25.0)	14 (21.5)	7 (11.9)	13 (18.3)	4.063 (0.255)
Hypertension, n (%)	61 (21.6)	25 (28.4)	11 (16.9)	14 (23.7)	11 (15.5)	4.978 (0.173)
Hypnotic use, n (%)	90 (31.8)	19 (21.6)**	27 (41.5)	14 (23.7)	30 (42.3)	12.421 (0.006)*
Antidepressant use, n (%)	47 (16.6)	5 (5.7)**	12 (18.5)	9 (15.3)	21 (29.6)	16.448 (0.001)*
Antipsychotic use, n (%)	34 (12.0)	3 (3.4)**	11 (16.9)	8 (13.6)	12 (16.9)	9.384 (0.025)*
ESS score (0–24), mean ± SD	9.8 (6.0)	5.0 (2.6)**	5.4 (2.5)	14.1 (3.5)	16.1 (4.1)	235.540 (< 0.001)*
DBAS score (0–24), mean ± SD	84.9 (32.8)	54.6 (24.3)**	111.6 (16.1)	68.1 (20.0)	112.0 (15.6)	168.336 (< 0.001)*
ISI score (0–15), mean ± SD	16.2 (6.2)	13.4 (6.8)**	18.7 (5.1)	14.2 (5.6)	19.0 (4.6)	19.479 (< 0.001)*
PHQ-9 score (0–27), mean ± SD	10.7 (6.4)	7.0 (5.0)**	11.6 (6.0)	10.7 (5.8)	13.9 (6.7)	17.9 (< 0.001)*
GAD-7 score (0–21), mean ± SD	7.7 (5.2)	4.8 (3.7)**	8.7 (5.0)	7.0 (5.1)	10.2 (5.2)	16.5 (< 0.001)*
Subjective TST, mean ± SD), min	353.8 (109.8)	327.6 (114.5)**	350.3 (108.6)	365.7 (113.0)	377.8 (97.9)	2.730 (0.045)*
Objective TST, mean ± SD, min	397.8 (76.9)	376.3 (87.6)**	397.3 (67.4)	406.6 (65.6)	417.5 (74.3)	13.48 (0.260)
SL, mean ± SD, min	14.5 (22.8)	15.2 (19.2)	22.3 (37.7)	10.0 (11.4)	10.3 (11.6)	4.316 (0.005)*
WASO, mean ± SD, %	15.3 (11.8)	18.7 (12.7)**	12.8 (9.6)	15.1 (11.0)	13.2 (12.1)	4.369 (0.005)*
SE, mean ± SD, %	80.4 (12.9)	76.6 (14.8)**	81.2 (10.6)	81.8 (11.4)	83.3 (12.7)	4.217 (0.006) *
AHI, events/h, mean ± SD	22.0 (21.0)	23.5 (18.0)	21.0 (24.3)	22.4 (19.0)	20.8 (22.9)	0.276 (0.843)
AI, events/h, mean ± SD	7.6 (12.9)	9.0 (13.2)	6.6 (13.3)	7.0 (11.6)	7.2 (13.3)	0.557 (0.644)
HI, events/h, mean ± SD	14.4 (13.3)	14.4 (9.7)	14.4 (14.2)	15.4 (15.5)	13.6 (14.4)	0.196 (0.899)
SpO2 90% total duration, min, mean ± SD	7.7 (33.1)	5.1 (12.2)	7.5 (21.7)	13.2 (64.7)	6.6 (17.3)	0.748 (0.524)
Each disorder						19.244 (0.004) *
CID, n (%)	42 (14.8)	13 (14.8)	14 (21.5)	7 (11.9)	8 (11.3)
OSA, n (%)	166 (58.7)	58 (65.9)	26 (40.0)	43 (72.9)	39 (54.9)
COMISA, n (%)	75 (26.5)	17 (19.3)	25 (38.5)	9 (15.3)	24 (33.8)

Abbreviations: BMI, body mass index; ESS, Epworth Sleepiness Scale; DBAS, Dysfunctional Beliefs and Attitudes about Sleep; ISI, Insomnia Severity Index; PHQ-9, Patient Health Questionnaire-9; GAD-7, Generalized Anxiety Disorder-7; TST, total sleep time; SL, sleep latency; WASO, wake after sleep onset; SE, sleep efficiency; AHI, apnea–hypopnea index; CID, chronic insomnia disorder; OSA, obstructive sleep apnea; COMISA, comorbid insomnia and sleep apnea.

Body mass index is calculated as weight in kilograms divided by height in meters squared. * p < 0.05 in ANOVA, ** corrected p < 0.05 in post-hoc analysis between Low ESS/Low DBAS and High ESS/High DBAS with Bonferroni correction.

Combination of EDS and DBAS that distinguishes COMISA from CID

Table 3 shows the logistic regression results for the ESS and DBAS combinations using CID as the reference group. In the model adjusted for age and sex, the high ESS/high DBAS group was significantly associated with COMISA, compared with the low ESS/low DBAS group (odds ratio [OR], 14.33; 95% CI 3.06–67.06; Bonferroni-corrected p = 0.0022). After additional adjustment for BMI and ISI, PHQ-9, and GAD-7 scores, the high ESS/high DBAS group remained significantly associated with COMISA (OR, 9.48; 95% CI 1.64–54.71; Bonferroni-corrected p = 0.024). Sensitivity analyses using 1000 bootstrap resamples yielded consistent results (Online Resource 1, Supplementary Table S1).

Table 3.

Logistic regression analysis of the ESS/DBAS combinations in the COMISA group with the CID group serving as a reference.

Predictor	COMISA vs. CID (ref)
	OR (95% CI, corrected p-value)
	Age- and sex-adjusted	Model 1a	Model 2b
Low ESS/low DBAS	Reference	Reference	Reference
Low ESS/high DBAS	2.67 (0.79–9.02, 0.230)	2.67 (0.80–9.10, 0.224)	3.34 (0.85–13.08, 0.166)
High ESS/low DBAS	1.63 (0.34–7.81, 1.00)	1.61 (0.34–7.66, 1.00)	2.14 (0.40–11.46, 0.752)
High ESS/high DBAS	14.33 (3.06–67.06, 0.002) *	13.76 (2.92–64.94, 0.002) *	19.69 (3.36–115.28, 0.002) *

^aModel 1 includes age, sex, and BMI.

^bModel 2 includes model 1 variables as well as the ISI, PHQ-9, and GAD-7 scores.

Abbreviations: COMISA, comorbid insomnia and sleep apnea CID; chronic insomnia disorder OR, odds ratio; CI, confidence interval; BMI, body mass index; ESS, Epworth Sleepiness Scale; DBAS, Dysfunctional Beliefs and Attitudes about Sleep; ISI, Insomnia Severity Index; PHQ-9, Patient Health Questionnaire-9; GAD-7, Generalized Anxiety Disorder-7.

To control for type I error inflation due to multiple comparisons, the Bonferroni correction is applied. Corrected p-values > 1.00 are truncated and reported as 1.00. * Corrected p-value < 0.05.

Combination of EDS and DBAS that distinguishes COMISA from OSA

Table 4 shows the logistic regression results using OSA as the reference group. In the model adjusted for age and sex, both the high ESS/high DBAS group (OR, 3.38; 95% CI 1.46–7.86; Bonferroni-corrected p = 0.01) and low ESS/high DBAS group (OR, 4.49; 95% CI 1.97–10.23; Bonferroni-corrected p < 0.001) were significantly associated with COMISA, compared with the low ESS/low DBAS group. These associations remained significant after additional adjustment for the BMI and AHI. In the fully adjusted model, the high ESS/high DBAS group showed an OR of 3.13 (95% CI 1.22–8.06; Bonferroni-corrected p = 0.036) and the low ESS/high DBAS group an OR of 4.95 (95% CI 2.02–12.16; Bonferroni-corrected p < 0.001). Bootstrap sensitivity analyses produced consistent findings (Online Resource 1, Supplementary Table S2).

Table 4.

Logistic regression analysis of the ESS/DBAS combinations in the COMISA group with the OSA group serving as a reference.

Predictor	COMISA vs. OSA (ref)
	OR (95% CI, corrected p-value)
	Age- and sex-adjusted	Model 1a	Model 2b
Low ESS/low DBAS	Reference	Reference	Reference
Low ESS/high DBAS	4.49 (1.97–10.23, < 0.001) *	4.55 (1.95–10.58, < 0.001) *	4.32 (1.83–10.20, 0.002) *
High ESS/low DBAS	0.92 (0.36–2.34, 1.00)	1.00 (0.39–2.59, 1.00)	0.92 (0.35–2.41, 1.00)
High ESS/high DBAS	3.38 (1.46–7.86, 0.01) *	3.72 (1.55–8.94, 0.006) *	3.20 (1.31–7.79, 0.022) *

^aModel 1 includes age, sex, and BMI.

^bModel 2 includes model 1 variables plus AHI.

Abbreviations: COMISA, comorbid insomnia and sleep apnea; OSA, obstructive sleep apnea; OR, odds ratio; CI, confidence interval; BMI, body mass index; ESS, Epworth Sleepiness Scale; DBAS, Dysfunctional Beliefs and Attitudes about Sleep.

To control for type I error inflation due to multiple comparisons, the Bonferroni correction is applied. Corrected p-values > 1.00 ware truncated and reported as 1.00. * Corrected p-value < 0.05.

Overall, the sensitivity analyses consistently supported the original findings, confirming that the co-occurrence of elevated ESS and DBAS rather than either variable alone distinguished COMISA across different analytic approaches. Multiple regression analyses with DBAS or ESS as the dependent variable (Supplementary Tables S3–S6) revealed that, compared with COMISA, CID was associated with significantly lower ESS scores (Supplementary Table S4; fully adjusted β, −3.01; 95% CI −5.34–0.69; uncorrected p = 0.012), without significantly lower DBAS scale scores (Supplementary Table S3; fully adjusted β, −9.44; 95% CI −18.96–0.07; uncorrected p = 0.052). In contrast, compared with COMISA, OSA showed significantly lower DBAS scale scores (Supplementary Table S5; fully adjusted β, −22.52; 95% CI, −31.50–13.70; uncorrected p < 0.001), without significantly lower ESS scale scores (Supplementary Table S6; fully adjusted β, 0.37; 95% CI −1.26–2.00; uncorrected p = 0.653). These findings indicate that neither DBAS nor ESS individually distinguished COMISA from both comparison groups after covariate adjustment, consistent with the original hypothesis for examining the role of their combined pattern.

Sensitivity analyses addressing potential contamination of the OSA group yielded consistent results. When psychotropic medication use was included as an additional covariate, both the high ESS/high DBAS (fully adjusted OR = 2.60; 95% CI 1.02–6.62; uncorrected p = 0.045) and the low ESS/high DBAS (OR = 3.42; 95% CI 1.39–8.39; uncorrected p = 0.007) remained significantly associated with COMISA (Supplementary Table S7). When psychotropic medication users were excluded from the OSA group, the associations were strengthened, with the high ESS/high DBAS phenotype showing an OR of 4.53 (95% CI 1.70–12.07; uncorrected p = 0.002) and the low ESS/high DBAS phenotype showing an OR of 7.82 (95% CI 2.88– 21.20; uncorrected p < 0.001) (Supplementary Table S8).

When established clinical cutoff values (ESS > 10; DBAS scale score > 68.8) were used instead of median-based categorization, the high ESS/high DBAS was significantly associated with COMISA compared with both CID (fully adjusted OR = 10.23; 95% CI 1.74– 60.23; uncorrected p = 0.010) and OSA (fully adjusted OR = 3.26; 95% CI 1.24–8.54; uncorrected p = 0.016) (Supplementary Tables S9–S10).

To address the potential confounding effect of OSA severity heterogeneity, analyses were repeated after restricting the sample to patients with moderate-to-severe OSA (AHI ≥ 15 events/hour). The high ESS/high DBAS remained significantly associated with COMISA in comparisons with both CID (Supplementary Table S11; fully adjusted OR = 170.8; 95% CI 5.12–5697.54; uncorrected p = 0.004) and OSA (Supplementary Table S12; fully adjusted OR = 5.55; 95% CI 1.33–23.11; uncorrected p = 0.019). The low ESS/high DBAS phenotype was also significant in comparisons with OSA (Supplementary Table S12: fully adjusted OR = 11.32; 95% CI 2.94–43.58; uncorrected p < 0.001).

Discussion

In the present study, we found that the high ESS/high DBAS combination was significantly associated with COMISA when either CID or OSA served as the comparison group. This association remained robust in all sensitivity analyses and was independent of covariates. These findings suggest that the coexistence of elevated DBAS and EDS may represent a distinctive phenotype that differentiates COMISA from CID or OSA. Sensitivity analyses further clarified the nature of this combined phenotype. When DBAS and ESS were examined individually as continuous variables, DBAS differentiated COMISA from OSA but not from CID, whereas ESS differentiated COMISA from CID but not from OSA. Thus, neither variable alone distinguished COMISA from both comparison groups; nevertheless, the categorical high ESS/high DBAS phenotype remained significant in both comparisons. A concern remains that a substantial proportion of the OSA group was receiving psychotropic medications, which may suggest the presence of insomnia symptoms or psychiatric comorbidities in these patients. To address this concern, we employed two complementary approaches: including psychotropic medication use as an additional covariate and excluding psychotropic medication users from the OSA group. Neither approach altered the original results substantially. Moreover, neither restricting the sample to patients with moderate-to-severe OSA nor using established clinical cutoff values for ESS and DBAS substantially altered the main findings, suggesting that the confounding effect of OSA severity heterogeneity is likely minimal. Taken together, these results—in which individual variables fail to consistently distinguish COMISA, but their categorical co-occurrence pattern does—suggest that COMISA is characterized by a threshold-based co-occurrence of elevated DBAS and EDS.

The pattern of associations differed according to the reference group, which has important implications for how clinicians conceptualize COMISA. When OSA served as the comparison group, both the high ESS/high DBAS and low ESS/high DBAS profiles were associated with COMISA. In contrast, when CID served as the comparison group, only the high ESS/high DBAS profile showed significant associations. These patterns imply different clinical presentations. For clinicians who initially diagnose OSA, COMISA may appear as OSA accompanied by additional insomnia-related features. For clinicians who first diagnose CID, COMISA may reflect a more complex interaction between insomnia characteristics and EDS.

The comparison with OSA highlights key aspects of COMISA heterogeneity. Elevated DBAS scale scores distinguished COMISA from OSA across both the high ESS/high DBAS and low ESS/high DBAS groups, whereas lower DBAS scale levels corresponded to the OSA alone. Variation in the ESS scores among patients with COMISA with high DBAS scale scores may reflect distinct clinical subtypes. The high ESS/high DBAS profile may represent patients in whom both OSA-related sleepiness and insomnia-related cognitive processes are prominent³⁵. The low ESS/high DBAS profile may reflect presentations that are more strongly influenced by insomnia-related features, in which cognitive processes shape subjective daytime symptoms despite the presence of OSA. This presentation aligns with previous descriptions of COMISA cases in which CID characteristics are strongly expressed¹⁰.

From the perspective of CID, our findings suggest that DBAS may be expressed more prominently in COMISA when combined with EDS related to OSA. According to Harvey’s cognitive model of insomnia¹⁴, reflecting on daytime consequences of poor sleep, such as fatigue, concentration problems, and tiredness, increases monitoring of sleep impairment, which perpetuates distress and exacerbates sleep difficulties³⁷. In COMISA, chronic sleepiness associated with OSA may lead to recurrent experiences that intensify rumination about sleep and reinforce maladaptive sleep-related beliefs. This mechanism may strengthen DBAS to a degree that is not consistently observed with CID alone. Importantly, the association between the high ESS/high DBAS phenotype and COMISA remained significant after adjustment for insomnia severity measured using the ISI. This finding indicates that elevated DBAS in COMISA is not merely a reflection of insomnia severity but captures additional psychobehavioral characteristics that emerge from co-existing insomnia and OSA. When examined as an individual continuous predictor, DBAS did not reliably differentiate COMISA from CID (Supplementary Table S3). However, its discriminative value emerged specifically when combined with ESS. This pattern may support that DBAS may be expressed more prominently in COMISA when combined with EDS related to OSA.

These findings may provide new insights into the clinical management of OSA and CID. Insomnia is often underestimated during OSA treatment, with OSA being similarly underestimated in insomnia treatment¹³, which may delay COMISA identification and worsen outcomes, including reduced CBT-I effectiveness and lower CPAP adherence. When patients with OSA present with sleep disturbances, clinicians are generally prompted to consider COMISA³⁸. However, when patients with CID report EDS, this symptom does not necessarily indicate the absence of hyperarousal; rather, it may reflect coexisting OSA, which may indicate a more severe underlying condition. Given that both CID and OSA are associated with adverse health outcomes^39–44 and that COMISA carries increased cardiovascular and hypertension risks beyond either disorder alone⁴⁵, appropriate assessment of DBAS in OSA management and evaluation of daytime sleepiness in CID management may support earlier detection of COMISA and prevent underestimation of comorbid disease severity. The robustness of these findings across sensitivity analyses using established clinical cutoff values (ESS ≥ 10; DBAS scale score ≥ 68.8) further supports their potential clinical applicability. Clinicians may consider incorporating these thresholds into routine screening of patients presenting with either OSA or CID symptoms.

This study has limitations. Diagnoses of CID, OSA, and COMISA were based on the ICSD-3²⁵, which may have resulted in different patient characteristics compared with those reported in studies that relied on the ISI or AHI criteria⁴⁶. In the present study, the COMISA group had significantly fewer female patients and higher age, prevalence of hypertension, and AHI, compared with the CID group. Compared with the patients with OSA, those with COMISA had lower AHI, BMI, and prevalence of diabetes, greater use of sleep medications, and higher age and ISI and DBAS scale scores. Previous studies have also reported that COMISA is associated with a lower proportion of female patients and a higher age and prevalence of hypertension,⁴⁷ compared with CID⁴⁸, and that it is associated with higher age⁴⁷ and ISI and DBAS scale scores, compared with OSA¹⁰; these findings are consistent with those of the present study. However, the lower prevalence of diabetes and AHI observed in our COMISA group differ from those observed in some prior reports. This discrepancy may be explained using ICSD-3 criteria rather than the use of an ISI cutoff or an AHI > 15 for COMISA classification. Our approach included patients with an AHI of ≥ 5 and daytime functional impairment, which may have resulted in the inclusion of patients with milder OSA⁴⁹. As a result, this study may yield findings more applicable to clinical practice, because it includes patients with mild OSA. Second, the ESS and DBAS scores were dichotomized using overall median values (ESS = 9; DBAS scale score = 90) in the primary analysis. Although median-based categorization has been criticized for information loss and sample-specific thresholds, this approach was selected because our study aimed to examine the combined pattern of elevated EDS and DBAS rather than to establish clinical cutoff values. Importantly, sensitivity analyses using established clinical cutoff values (ESS ≥ 10; DBAS scale score ≥ 68.8) produced results consistent with the primary analyses (Supplementary Tables S9 and S10), suggesting that the findings are not dependent on the specific categorization method. Third, the CID group was relatively small (n = 42), reflecting the referral characteristics of our PSG database, in which PSG is more frequently indicated for suspected OSA than for isolated insomnia. Although bootstrap sensitivity analyses supported the robustness of the findings, the results of the COMISA versus CID comparison should be interpreted with this limitation in mind. Fourth, habitual daily sleep duration was not systematically collected in the database; therefore, the potential influence of these effects on ESS and DBAS scores could not be assessed. Fifth, the relatively high rates of psychotropic medication use in the OSA group reflect the referral characteristics of our database from the National Center of Neurology and Psychiatry. These patients had most likely been prescribed psychotropic medications, including hypnotics, as part of the management of their pre-existing psychiatric conditions; however, at the time of sleep medicine evaluation, a diagnosis of CID was not made for these patients. Since these medications were prescribed for pre-existing psychiatric conditions, the possibility that their use may have influenced DBAS and ESS scores cannot be entirely excluded. Although sensitivity analyses demonstrated that the primary findings were robust to both statistical adjustment and exclusion of psychotropic medication users (Supplementary Tables S7 and S8), generalizability to general sleep clinic populations should be confirmed in future studies.

In conclusion, DBAS and EDS, which do not consistently appear as prominent features in CID or OSA when examined separately, may function as characteristic psychobehavioral phenotypes in COMISA through their coexistence. These findings support the potential identification of COMISA among patients presenting with either OSA or CID symptoms and provide a basis for targeted screening approaches in clinical practice.

Author contributions

Conceptualization: Kenichi Kuriyama, Takuya Yoshiike, Tomohiro Utsumi; Investigation: Kenichi Kuriyama, Takuya Yoshiike, Tomohiro Utsumi, Aoi Kawamura, Kentaro Nagao, Momo Fushimi, Ayano Kimura, Megumi Hazumi, Kentaro Matsui; Writing – original draft: Tomohiro Utsumi; Writing – review and editing: Kenichi Kuriyama. Takuya Yoshiike, Aoi Kawamura, Kentaro Nagao, Kentaro Matsui; Funding acquisition: Kenichi Kuriyama; Supervision: Kenichi Kuriyama, Takuya Yoshiike; Formal analysis: Takuya Yoshiike, Tomohiro Utsumi

Funding

This work was supported by the Intramural Research Grant for Neurological and Psychiatric Disorders by the National Center of Neurology and Psychiatry, Tokyo, Japan (grant numbers 2–1: Establishment of a Sleep Disorders and Polysomnography Databank, 5–1: Development of the Sleep Disorders and Polysomnography Databank, and its utility for elucidating the pathophysiology, developing diagnostic markers and treatment methods of sleep disorders and neuropsychiatric disorders).

Data availability

The data analyzed in this article are available from a polysomnography database for sleep disorders in Japan upon request and approval.

Declarations

Competing interests

Dr. Yoshiike reports payments for several items outside the submitted work, namely consulting fees from Eisai, and honoraria for lectures from MSD, Daiichi Sankyo, Eisai, and Mitsubishi Tanabe Pharma. Dr. Kawamura reports personal fees from Takeda Pharmaceutical Co. Ltd. and MSD, outside the submitted work. Dr. Nagao reports payments for lectures outside the submitted work, including services under speaker’s honorarium from Sumitomo Pharma Co. Ltd. Dr. Matsui reports payments for advisory services or honoraria for lectures from Chugai Pharmaceutical, Daiichi Sankyo, Eisai, Japan Blood Products Organization, Meiji Seika Pharma, Mitsubishi Tanabe Pharma, mMEDICI, Mochida Pharmaceutical, MSD, Nobelpharma, Otsuka Pharmaceutical, Sumitomo Pharma, Takeda Pharmaceutical, and Nemielu, outside the submitted work. Dr. Kuriyama reports grant support from Otsuka Pharmaceutical Co. Ltd. and Eisai; personal fees from Meiji Seika Pharma, Eisai, MSD, Daiichi Sankyo, Takeda Pharmaceutical Co. Ltd., Mitsubishi Tanabe Pharma, and Sumitomo Pharma; and advisory fees from Eisai, Shionogi Pharma, Meiji Seika Pharma, Taisho Pharmaceutical Co. Ltd., Nexera Pharma, Idorsia Pharmaceuticals Ltd. And Kissei Pharmaceutical Co., Ltd, outside the submitted work. The other authors have no relevant financial or non-financial interests to disclose.

Ethics approval

This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committee of the National Center of Neurology and Psychiatry, Tokyo, Japan (March 08, 2023/No. A2022-100).

Consent to participate

This study utilized anonymized data from a PSG databank for sleep disorders in Japan for research purposes; therefore, the informed consent procedure was waived.