Sleep misperception in women with myofascial temporomandibular disorder

Abstract

Study Objectives:

Temporomandibular disorders (TMDs) were linked to poor sleep on the Pittsburgh Sleep Quality Index (PSQI), whereas polysomnography revealed no major sleep disturbances, implying sleep state misperception. This study investigates sleep state misperception in TMD and control participants; correlates sleep state misperception with objective short sleep duration (SSD), depression symptoms, daytime sleepiness, and orofacial pain; and compares objective SSD between the groups.

Methods:

General linear models were used to compare second-night polysomnography total sleep time, sleep latency, sleep efficiency (SE), and wake after sleep onset with homologous PSQI-derived variables in 124 women with myofascial TMD and 46 age and body mass index matched controls. PSQI variables were regressed onto objective SSD, depression symptoms, daytime sleepiness, and pain. Lastly, objective SSD was related to TMD presence.

Results:

Compared to controls, TMD cases misperceived SE (P = .02); depression symptoms explained PSQI-derived SE (P = .002) and mediated the effect of pain (P < .001). PSQI variables were unrelated to respective polysomnography measures or objective SSD, except a significant self-reported-objective correlation in SE among controls only (P = .002). Objective SSD was more frequent in TMD cases (P = .02, odds ratio = 2.95), but it was unrelated to depression symptoms, daytime sleepiness, or prepolysomnography pain.

Conclusions:

The study demonstrates misperception of SE among TMD cases, which was accounted for by depression symptoms. Objective SSD nearly tripled in TMD cases; however, it was unrelated to PSQI variables, depression, daytime sleepiness, or pain, suggesting that sleep state misperception and objective SSD are 2 independent sleep features in TMD.

Citation:

Chan C, Dubrovsky B, Bouchard M, Tartter VC, Raphael KG. Sleep misperception in women with myofascial temporomandibular disorder. J Clin Sleep Med. 2025;21(1):55–64.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Patients with temporomandibular disorder (TMD) report poor sleep, while polysomnography reveals no major disturbance, although one uncontrolled study found a relationship between objective short sleep duration (SSD) and TMD pain. The current study investigates sleep state misperception and objective SSD in TMD cases and controls in relationship to depression symptoms, daytime sleepiness, and pain.

Study Impact: TMD cases evidence (i) misperception of sleep efficiency which relates to depression symptoms, and (ii) elevated frequency of objective SSD which is unrelated to any self-reported variables. These results suggest that objective SSD and sleep misperception are 2 independent facets of sleep in TMD and call for further investigations of physiological and cognitive contributions to sleep disturbance in TMD population.

INTRODUCTION

Temporomandibular disorders (TMDs) are comprised of conditions that affect the masticatory muscles or temporomandibular joints, associated with chronic orofacial pain and disability.¹ TMD affect approximately 5–12% of the population² with higher prevalence in women.³ Women with TMD also report more severe symptoms than men, including myofascial pain and poor sleep.^4–6

Sleep evaluations in TMD show divergent results, depending on whether physiological or self-reported measures are used. Numerous studies using the Pittsburgh Sleep Quality Index (PSQI) found disturbed sleep in patients with TMD.^6–9 However, several evaluations with polysomnography (PSG) and actigraphy failed to provide evidence of major sleep disturbance.^10–12 One study using the Insomnia Severity Index suggested a high prevalence of insomnia in patients seeking care for orofacial pain: out of 952 patients, 296 (31.1%) presented elevated Insomnia Severity Index.¹³ Women in this study had higher Insomnia Severity Index than men across a wider age range.¹³ This study did not employ physiological sleep assessments.¹³

A recent PSG evaluation of 128 women with both temporomandibular joint disorder and chronic insomnia by clinical history and Insomnia Severity Index found a high prevalence (24%) of objectively short sleep duration (SSD),¹⁴ defined as less than 6 hours of sleep in the insomnia literature.¹⁵ Objective SSD was related to higher pain severity in the week preceding the PSG.¹⁴ Thus, in this uncontrolled sample of patients with temporomandibular joint disorder and clinical insomnia, 76% did not have objective SSD.¹⁴

These data suggest a discrepancy between self-reported and PSG sleep measures among patients with TMD, pointing to paradoxical insomnia or sleep state misperception (SSM), in which individuals may experience their sleep as wakefulness.¹⁵ The mismatch between self-reported and PSG-defined sleep may be related to cognitive attribution.¹⁶ In 1 study, participants who were led to believe that their nocturnal sleep was poor reported worse daytime function, regardless of objective sleep quality.¹⁷ A study on SSM in insomnia revealed a relationship between underestimation of sleep duration and MMPI profiles of depression and anxiety.¹⁸ Another study on SSM in depressed patients found a direct relationship between depression scores and overestimation of wake after sleep onset (WASO).¹⁹ A similar pattern may be present in patients with TMD, whose daytime alertness and mood, possibly affected by pain, may give rise to attributions of poor sleep without objective disturbance. In a group of TMD and control participants, PSQI global score was related to depression symptoms rather than PSG variables,²⁰ suggesting that depression symptoms may lead to SSM in patients with TMD; however, this study did not directly analyze for SSM using specific sleep variables.

Several studies evaluated the discrepancy between self-reported and objective sleep measures in patients with insomnia using homologous variables side by side. Measures that allowed direct comparison between self-reported and electroencephalographic data included total sleep time (TST), sleep latency (SL), sleep efficiency (SE), and WASO.^21–24 These variables can be employed for a similar analysis in a TMD sample.

The present study is a secondary analysis of PSG and self-reported data from a large sample of women with TMD and matched controls, with 3 aims. First, we evaluate SSM in TMD cases relative to age and body mass index matched controls using PSG measures of TST, SL, SE, and WASO and homologous self-reported variables. We expect to find correlations between self-reported and objective sleep variables in controls, but not in TMD cases.

Second, self-reported variables evidencing misperception are tested for relationships with objective SSD, depression symptoms, daytime sleepiness, and, in TMD only, orofacial pain. A mediating effect of depression symptoms is analyzed for self-reported sleep variables related to TMD presence or to pain intensity within the TMD group. Self-reported sleep variables are expected to be related to depression symptoms rather than objective SSD, and the effect of pain is expected to be mediated by depression symptoms.

Third, we compare the frequency of objective SSD between TMD cases and controls and determine whether it is related to depression symptoms, daytime sleepiness, and in TMD only, pain intensity. We expect to find higher frequency of objective SSD among TMD cases and its relationship with reports of depression symptoms, daytime sleepiness, and pain.

METHODS

Participants

All women in the experimental group were recruited at the New York University’s College of Dentistry. They were selected based on the complaint of myofascial (masticatory muscle-based) pain and met the research diagnostic criteria for masticatory muscle-based TMD.²⁵ Selection criteria were not based on the history of sleep bruxism or poor sleep. A total of 169 patients were approached; 19 declined participation and 12 did not meet Research Diagnostic Criteria for TMD. Women in the control group were haphazardly recruited from acquaintances of participants and patients at the New York University’s College of Dentistry dental clinics. Out of 63 approached candidates, 6 were excluded due to > 1 tender point, and another 8 dropped out early in the process. Control participants did not have orofacial pain complaints or other signs of TMD; they were matched to the participants with TMD by body mass index and age within 5-year strata.

As the primary interests at the time of data collection were the role of sleep bruxism and sleep architecture, all prospective participants who reported sleeping less than 4 hours per night, and those whose PSG showed less than 4 hours of sleep on the first night were excluded during the screening evaluation to ensure sufficient sleep sampling. Further, participants who were shift workers, pregnant, or previously diagnosed with sleep disorders, such as chronic insomnia and sleep apnea, were excluded. Menopause did not disqualify from the study. Participants were allowed to take their usual medications, including muscle relaxants, nonsteroidal anti-inflammatory drugs, and antidepressants, as long as they met group eligibility criteria. The exclusion rates did not differ between the groups (Fisher exact test, P = .36).

All participants had to be fluent in English. After recruited participants provided their written informed consent, they were examined and interviewed at the New York University’s College of Dentistry, and underwent a 2-night PSG protocol. Participants were asked to maintain their usual sleep schedule for a minimum of 2 weeks prior to the study.

Measures

PSG variables

PSG data were recorded on a SomnoStarPro system (Viasys Healthcare, San Diego, California), as described elsewhere.¹¹ For the present analysis, we selected the following sleep parameters: TST (total amount of sleep time in minutes per night), SL (time in minutes between “lights out” and sleep onset), SE (percentage of TST out of time in bed), and WASO (time in minutes spent awake between sleep onset time and “lights on” in the morning). These sleep variables have been used in previous research on SSM,^21–24 because they can be derived from both objective measures and sleep questionnaires. These variables are also commonly used in clinical practice to describe sleep in patients with chronic insomnia.²⁶ In the present study, TST, SL, SE, and WASO were derived from EEG data manually scored by professionally trained scorers in accordance with the then-current sleep scoring manual.^11,27 The second night’s recordings were used to avoid the first night effect.²⁸ Objective SSD was defined as less than 6 hours, in accordance with the criterion used in insomnia literature.¹⁵

Self-reported sleep variables

Self-reported sleep variables were derived from the PSQI, a standardized questionnaire widely used in TMD research to show poor sleep quality.^7–9 The analysis of self-reported-objective discrepancies within each of the sleep parameters was achieved by the extraction of PSQI variables that have direct PSG counterparts: TST, SL, and SE. The PSQI-sleep disturbance component was used to compare with PSG-derived WASO, as described below.

The PSQI was used at the initial daytime interview to collect data on self-reported sleep quality during the preceding month.²⁹ The 19-items of the PSQI are standardly clustered into 7 components measuring various aspects of sleep disturbance on a 0–3 scale, and a global PSQI score measuring the overall disturbance on a 0–21 scale. For the present analysis, standard PSQI components sleep duration, SL, and habitual SE were not used, due to the paucity of responses in the “2” and “3” categories (frequencies varied between 1.6% and 13.7% in the TMD group and between 0% and 6.5% in the control group). Instead, we selected PSQI-items that can be most directly aligned with the PSG variables and provided continuous scale measures. For TST, PSQI-item #4 was selected (“During the past month, how many hours of actual sleep did you get at night?”).²⁹ For SL, PSQI-item #2 was selected (“During the past month, how long (in minutes) has it usually taken you to fall asleep each night?”).²⁹ For SE, TST (PSQI-item #4) was calculated as a percentage of the time in bed defined as the time between the usual bedtime (PSQI-item #1) and getting up time (PSQI-item #3). Two responses with > 100% SE (1 TMD and 1 control) were converted into 100%. The sleep disturbance component was used as a self-reported proxy for the amount of wakefulness during the night. This component is a sum of 9 PSQI-items (#5b through 5j) measuring the frequency of difficulty sleeping due to several factors.²⁹ Each item is measured on a 0–3 scale, and the sum is standardized to a 0–3 scale. Due to the low frequency of category “3” in the standard PSQI scoring (4% in TMD group and 0% in controls), the sum of 9 items was used and treated as a continuous variable for PSQI-derived sleep disturbance.

Self-reported mood and daytime sleepiness measures

Symptoms of depression were assessed using the Symptom Checklist-90 (SCL-90),³⁰ where a total depression score was an average of 13 responses that measure the degree of disturbance in 13 respective domains over the last week on a 5-point scale ranging from 0 (“not at all”) to 4 (“extremely”). For daytime sleepiness, the Epworth Sleepiness Scale (ESS)³¹ was used, which was a sum of 8 items measuring the likelihood of dozing off in various situations that ranges from 0 (“no daytime sleepiness”) to 24 (“severe excessive daytime sleepiness”).

Orofacial pain in TMD cases

After arriving at the sleep laboratory, TMD cases reported characteristic pain intensity (CPI) using the Chronic Pain Grade Scale.³² The scale asks 3 questions on: (1) pain at the time of interview, (2) worst pain in the last 6 months, and (3) the average pain in the last 6 months. Each item is scored on an 11-point scale ranging from 0 (“no pain”) to 10 (“worst pain”), and the average of the 3 scores is defined as CPI. On the evening of the PSG, TMD cases were asked to rate the average pain intensity on a 0–10 scale during the day preceding the PSG and current pain at the time of questioning. The average of these 2 ratings was a measure of pre-PSG pain intensity.

Procedure

Each participant spent 2 consecutive nights in a sleep laboratory affiliated with the New York University School of Medicine, where both nights included all standard PSG recording procedures. The first night was an adaptation night to eliminate “first-night effect” as data from the second night are considered more representative of the typical sleep architecture.²⁸ Data from the second night were used for the present analysis for all participants except for 6 experimental and 1 control whose second night data were lost due to a technical malfunction; for consistency with previously published work on this cohort, the first-night data were used for those participants.

Data analysis

The IBM SPSS Statistical Software–V28.0 (IBM Corp., Armonk, NY) was used for data analyses. For the first aim, we analyzed the relationship between self-reported and objective sleep measures in TMD cases and controls using general linear models. Explanatory variables included TMD presence, a self-report sleep variable (PSQI-derived TST, SL, SE, or sleep disturbance), and the interaction between TMD presence and the self-reported sleep variable, with a respective PSG measure as an outcome variable. For the second aim, multiple regression models were used in each group to predict self-reported sleep variables based on objective SSD, daytime sleepiness, depression symptoms and, in TMD only, orofacial pain. The Sobel test was used for mediation effects.³³ For the third aim, a logistic regression model was used to predict objective SSD based on TMD presence. Further, the relationship between objective SSD and depression symptoms, daytime sleepiness, and, in TMD only, pain intensity was analyzed using multivariate analysis of covariance. All analyses used age and body mass index as covariates known to relate to sleep, mood, sleepiness, and pain. The alpha level of 0.05 was used for significance testing.

RESULTS

A total of 170 participants were included in the study, with 124 TMD cases and 46 controls matched by age and body mass index. Racial/ethnic composition and medication use have been reported elsewhere.^10,11 Self-reported sleep schedules were not significantly different between TMD (Mean_bedtime = 11:37 pm, standard deviation = 58 minutes; Mean_{get-up time} = 7:30 am, standard deviation = 80 minutes) and control participants (Mean_bedtime = 11:34 pm, standard deviation = 66 minutes; Mean_{get-up time} = 7:19 am, standard deviation = 92 minutes; independent samples t_bedtime(168) = 0.29, not significant, and t_{get-up time}(168) = 0.78, not significant). Means and standard deviations for demographic characteristics, PSG, and PSQI variables relevant for further analysis, as well as between-group comparisons, are shown in Table 1.

Table 1.

Sample characteristics and between-group differences.*

	Controls (n = 46)	TMD (n = 124)	t Test**	MANCOVA***
	Mean ± SD	Mean ± SD
Sample Characteristics
Age (years)	36.1 ± 13.5	40.3 ± 14.8	.10
BMI	25.0 ± 5.5	25.0 ± 5.0	.96
Myofascial Pain
Mean pain duration (# months since onset)	n/a	126.1 ± 127.1	n/a
Characteristic pain intensity (0–10)	n/a	5.2 ± 1.7	n/a
Functional effect of pain (0–10)	n/a	1.8 ± 2.2	n/a
PSG Variables
SL (min)	9.0 ± 10.1	11.6 ± 16.5	.32
TST (hours)	6.7 ± 0.8	6.4 ± 0.9	.07
SE (= TST/time in bed × 100%)	92.3 ± 6.4	89.7 ± 8.7	.07
WASO (min)	24.6 ± 28.8	32.7 ± 31.8	.13
PSQI Variable
SL (min, PSQI item #2)	17.7 ± 16.5	25.3 ± 25.6	.06
TST (hours, PSQI item #4)	7.3 ± 1.0	7.1 ± 1.1	.15
SE (PSQI-reported TST/time in bed × 100%)	94.4 ± 5.8	89.7 ± 10.5	< .001	F = 7.3, P = .008
PSQI sleep disturbance (sum of items #5b–5j)	4.0 ± 3.5	8.9 ± 5.2	< .001	F = 32.2, P < .001
Depressive Symptomatology (SCL-90-D)	0.34 ± 0.33	1.00 ± 0.78	< .001	F = 28.3, P < .001
ESS	6.5 ± 4.2	7.1 ± 4.3	.39

*Only PSG and PSQI variables used for further analyses are presented in this table. **Independent samples t tests, significant P values are in bold. ***MANCOVA model comparing 3 variables that showed between-group differences on independent samples t tests: PSQI-SE, PSQI-sleep disturbance, and SCL-90-D between the groups. BMI and age were used as covariates. BMI = body mass index, ESS = Epworth Sleepiness Scale, MANCOVA = multivariate analysis of covariance, n/a = not applicable, PSG = polysomnography, PSQI = Pittsburgh Sleep Quality Index, SCL-90-D = Symptom Checklist-90-Depression, SD = standard deviation, SE = sleep efficiency, SL = sleep latency, TMD = temporomandibular disorder, TST = total sleep time, WASO = wake after sleep onset.

Aim 1: relationship between self-reported and objective sleep variables in TMD and control participants

General linear models were used to analyze PSG-derived variables as outcomes of TMD presence, the respective PSQI variable, and their interaction. The results are summarized in Table 2. PSG-derived TST was not significantly related to TMD presence (P = .13), PSQI-derived TST (P = .45), or their interaction (P = .07), showing a similar lack of correspondence between self-reported and objective TST in both groups.

Table 2.

General linear models analyzing PSG variables as outcomes of respective PSQI variables in TMD (n = 124) and control participants (n = 46).^*

General Linear Models		F (1,164)	P
Outcome	Predictors
PSG-TST	TMD vs Controls	2.3	.13
	PSQI-TST (item #4)	0.6	.45
	TMD by PSQI-TST interaction	3.2	.07
PSG-SL	TMD vs Controls	0.7	.39
	PSQI-SL (item #2)	1.8	.18
	TMD by PSQI-SL interaction	0.1	.76
PSG-SE	TMD vs Controls	4.8	.03
	PSQI-SE (PSQI-reported TST/time in bed × 100%)	4.8	.03
	TMD by PSQI-SE interaction	5.3	.02
PSG-WASO	TMD vs Controls	0.5	.47
	PSQI sleep disturbance (sum of items #5b–5j)	1.1	.29
	TMD by PSQI-sleep disturbance interaction	0.2	.66

^*General linear models used BMI and age as covariates. BMI = body mass index, PSG = polysomnography, PSQI = Pittsburgh Sleep Quality Index, TMD = temporomandibular disorder, TST = total sleep time, SE = sleep efficiency, SL = sleep latency, WASO = wake after sleep onset.

PSG-derived SL was not significantly related to TMD presence (P = .39), PSQI-derived SL (P = .18), or their interaction (P = .76), showing a similar lack of correspondence between self-reported and objective SL in both groups.

The analysis of PSG-derived SE revealed significant main effects of TMD presences (P = .03), PSQI-derived SE (P = .03), and a significant interaction (P = .02). Further multiple regression analysis revealed a significant positive correlation between PSQI-derived and PSG-derived SE among controls (partial r = .437, P = .002, shared variance = 19.1%) but not among TMD cases (partial r = .01, P = .89). These results are also presented in Figure 1.

Figure 1. PSG-SE% vs PSQI-derived SE% in the control vs TMD groups.

Multiple regression analysis revealed a significant positive correlation between PSQI-derived and PSG SE among controls (partial r = .437, P = .002, shared variance = 19.1%) but not among TMD cases (partial r = .01, P = .89) with age and BMI covariates. BMI = body mass index, PSG = polysomnography, PSQI = Pittsburgh Sleep Quality Index, SE = sleep efficiency, TMD = temporomandibular disorder.

PSG-derived WASO was not significantly related to TMD presence (P = .47), PSQI-derived sleep disturbance (P = .29), or their interaction (P = .66), showing a similar lack of correspondence between PSQI-measured sleep disturbance and objective WASO in both groups.

Aim 2: relationship between self-reported sleep variables and objective SSD, depression symptoms, daytime sleepiness, and pain

Multiple regression models were used to analyze the effects of objective SSD, SCL-90-Depression (SCL-90-D), ESS, and CPI on PSQI-derived TST, SL, SE, and sleep disturbance scores. The results are summarized in Table 3. Shorter self-reported TST was predicted by higher ESS in TMD cases only (P = .001); no other predictor variables were significant in either group. Longer self-reported SL was predicted by higher SCL-90-D in TMD cases only (P = .02); no other predictor variables were significant in either group. Lower self-reported SE was predicted by higher SCL-90-D in both TMD (P = .002) and controls (P = .01). Finally, higher PSQI-sleep disturbance score was predicted by higher SCL-90-D (P < .001) and higher CPI (P = .01) in TMD cases only; no other predictor variables were significant in either group.

Table 3.

Multiple regression models analyzing PSQI variables in TMD and control participants.^*

		Controls (n = 46)		TMD Cases (n = 124)
		0-Order Correlations	Multiple Regressions	0-Order Correlations	Multiple Regressions
Outcome Variables	Predictor Variables	P	B ± Std. Error, P	P	B ± Std. Error, P
PSQI-TST	Objective SSD	.86	0.12 ± 0.44, P = .80	.28	−0.10 ± 0.20, P = .61
	SCL-90-D	.91	0.44 ± 0.53, P = .41	.04	−0.16 ± 0.12, P = .20
	ESS	.48	−0.03 ± 0.04, P = .48	.001	−0.07 ± 0.02, P = .001
	CPI	n/a	n/a	.39	0.06 ± 0.06, P = .29
PSQI-SL	Objective SSD	.88	2.29 ± 7.26, P = .75	.43	3.81 ± 4.95, P = .44
	SCL-90-D	.04	15.47 ± 8.70, P = .08	.02	7.52 ± 3.08, P = .02
	ESS	1.00	−0.49 ± 0.62, P = .44	.13	−1.02 ± 0.53, P = .06
	CPI	n/a	n/a	.45	0.81 ± 1.48, P = .58
PSQI-SE	Objective SSD	.60	0.63 ± 2.50, P = .8	.61	−0.02 ± 1.95, P = .99
	SCL-90-D	.02	−7.95 ± 2.98, P = .01	< .001	−3.82 ± 1.21, P = .002
	ESS	.90	0.23 ± 0.22, P = .3	.03	−0.37 ± 0.21, P = .08
	CPI	n/a	n/a	.03	−0.33 ± 0.58, P = .57
PSQI-SD	Objective SSD	.47	1.50 ± 1.51, P = .33	.95	−0.81 ± 0.83, P = .33
	SCL-90-D	.02	3.26 ± 1.80, P = .08	< .001	2.60 ± 0.52, P < .001
	ESS	.22	0.06 ± 0.13, P = .66	.02	0.16 ± 0.09, P = .07
	CPI	n/a	n/a	< .001	0.63 ± 0.25, P = .01

^*Regression models analyzed PSQI variables as outcomes of objective SSD, ESS, and used BMI and age as covariates. BMI = body mass index, CPI = characteristic pain intensity, ESS = Epworth Sleepiness Scale, n/a = not applicable, PSQI = Pittsburgh Sleep Quality Index, SCL-90-D = Symptom Checklist-90-Depression, SSD = short sleep duration, SE = sleep efficiency, SL = sleep latency, Std. Error = standard error, TMD = temporomandibular disorder, TST = total sleep time, WASO = wake after sleep onset.

A mediating role of depressive symptoms in the effect of orofacial pain on PSQI-derived SE was investigated, as the presence of TMD affected the correspondence between self-reported and objective SE (Table 2), a significant 0-order correlation was found between CPI and self-reported SE among TMD cases (Table 3), and SCL-90-D was significantly related to self-reported SE in both groups. First, the mediating effect of SCL-90-D was analyzed between the TMD and control groups. A total effect between TMD presence and PSQI-derived SE was significant (B = −4.72 ± 1.63, t = −2.90, P = .004). TMD presence significantly increased SCL-90-D (Path a: B = 0.6 ± 0.12, t = 5.53, P < .001) and SCL-90-D reduced PSQI-derived SE (Path b: B = −4.63 ± 1.00, t = −4.62, P < .001). After controlling for SCL-90-D, the effect of TMD presence on PSQI-derived SE was no longer significant (Path c’: B = −1.69 ± 1.67, t = −1.01, P = .31). The Sobel Test revealed that SCL-90-D significantly mediates the relationship between TMD presence and PSQI-derived SE (t = −3.55, standard error = 0.85, P < .001). Second, in the analysis of the mediating effect of SCL-90-D within the TMD group, a total effect between CPI and PSQI-derived SE was significant (B = −1.15 ± 0.54, t = −2.14, P = .03). Higher CPI significantly increased SCL-90-D (Path a: B = 0.13 ± 0.04, t = 3.30, P = .001), and SCL-90-D significantly decreased PSQI-derived SE (Path b: B = −4.13 ± 1.20, t = −3.45, P = .001). After controlling for SCL-90-D, the effect of CPI on PSQI-derived SE was not significant (Path c’: B = −0.62 ± 0.54, t = −1.15, P = .25). The Sobel test revealed that SCL-90-D significantly mediated the relationship between CPI and PSQI-derived SE (t = 2.39, standard error = 0.22, P = .02). Thus, SCL-90-D mediated the effect of orofacial pain on self-reported SE between the groups and within the TMD group.

Further, since PSQI-sleep disturbance scores were significantly related to CPI and SCL-90-D, a mediating effect of SCL-90-D in the relationship between CPI and PSQI-sleep disturbance was analyzed within the TMD group. A total effect of CPI and sleep disturbance was significant (B = 1.26 ± 0.25, t = 5.03, P < .001). Higher CPI significantly increased SCL-90-D (Path a: B = 0.13 ± 0.04, t = 3.30, P = .001) and SCL-90-D significantly increased PSQI-sleep disturbance (Path b: B = 2.80 ± 0.52, t = 5.33, P < .001). When controlling for SCL-90-D, the effect of CPI on PSQI-sleep disturbance was significant (Path c’: B = 0.90 ± 0.24, t = 3.81, P < .001). The Sobel test revealed that SCL-90-D significantly mediated the relationship between CPI and PSQI-sleep disturbance (t = 2.81, standard error = 0.13, P = .005).

Aim 3: frequency of objective SSD in TMD and control groups and its relationship with SCL-90-D, ESS, and pre-PSG orofacial pain

Objective SSD (< 6 hours of TST) was found in 13% of controls (6 out of 46) and 30% of TMD cases (37 out of 124). In a logistic regression model, TMD presence was a significant predictor of objective SSD (B = 1.16 ± 0.49, Wald = 5.71, P = .02), with an odds ratio of 2.95. The analysis of objective SSD vs normal sleep duration (≥ 6 hours) in relationship to SCL-90-D, ESS, and, in TMD only, pre-PSG pain using multivariate analysis of covariance revealed no significant relationships (all F values < 1.0, all P values > .40).

DISCUSSION

The present study included a large group of TMD cases and matched controls undergoing a 2-night PSG protocol. Previously, these TMD participants showed significant sleep disturbance on the global PSQI, but only minimal evidence of disturbance on multiple PSG variables.^10,11,20 The present analysis examined specific relationships between 4 PSQI variables (TST, SL, SE, and sleep disturbance) and their respective second night PSG measures, as well as roles of objective SSD, depression symptoms, daytime sleepiness, and pain.

The first aim was to investigate whether TMD cases presented a greater self-reported-objective discrepancy than controls. In both groups, all PSQI-derived measures showed a lack of correspondence with their PSG counterparts, with 1 notable exception. Controls perceived SE more accurately, with self-reported habitual SE accounting for 19% of the variance in the PSG-derived SE, while TMD cases showed no self-reported-objective correspondence. This finding supports the hypothesis of increased SSM in the TMD population. Since SE is inversely proportional to the sum of SL and WASO, and the present analysis revealed no between-group differences in SL perception, the misperception of SE in the TMD group appears to result from increased perception of wakefulness throughout the night.

The second aim was to investigate the relationship between self-reported sleep variables and objective SSD, depression symptoms, daytime sleepiness, and orofacial pain intensity. Objective SSD was not related to any of the self-reported sleep variables in either group. In controls, SCL-90-D had significant 0-order correlations with PSQI-derived SL, SE, and sleep disturbance; however, in multiple regression models, only a relationship with self-reported sleep disturbance remained significant. Among TMD cases, 0-order correlations were found between SCL-90-D and all 4 PSQI variables, between ESS and 3 PSQI variables, and between CPI and 2 PSQI variables. In multiple regression models, ESS significantly predicted PSQI-derived TST, SCL-90-D significantly predicted PSQI-derived SL, SE, and sleep disturbance, while CPI predicted sleep disturbances (see Table 3). These results indicate the absence of relationship between objective SSD and self-reported sleep variables and emphasize the role of depression symptoms in sleep perception, particularly in TMD. The relationship between ESS and self-reported sleep duration in TMD cases also supports the notion of daytime sleepiness being a factor in the perception of sleep in this group.

Further analysis supported the expected mediation effects of depression symptoms on the relationship between orofacial pain and PSQI-derived SE between the groups and within the TMD group. After accounting for SCL-90-D, the difference between TMD and controls participants in self-reported SE was no longer significant. Similarly, SCL-90-D fully accounted for the relationship between CPI and self-reported SE in the TMD group. SCL-90-D was also a significant mediator of the effect of CPI on PSQI-sleep disturbance, although CPI remained a significant predictor of sleep disturbance after accounting for SCL-90-D. As the prior analysis of this sample revealed no relationship between CPI and PSG variables,¹¹ the results suggest that mismatch between self-reported and PSG variables in TMD cases is related to both the direct effect of pain and the mediating effect of depression symptoms.

The third aim was to compare frequency of objective SSD between the groups and to test for the relationship between objective SSD and depression symptoms, daytime sleepiness, and pre-PSG pain. Objective SSD was significantly more prevalent in the TMD group relative to controls, with an odds ratio showing nearly 3 times the occurrence. The percentage of TMD participants with objective SSD in the present sample (30%) was similar to that reported by Lerman et al¹⁴ (24%). This finding is particularly noteworthy, as Lerman et al¹⁴ selected TMD participants with a clinical history of insomnia, whereas the present sample excluded those with history of insomnia. This pattern strongly suggests 2 distinct phenotypes within the TMD population: those with objective SSD and those with self-reported sleep disturbance or insomnia complaints without objective SSD. The dissociation between objective SSD and self-reported sleep disturbance is further supported by the absence of relationships between objective SSD and PSQI-sleep variables, depression symptoms, ESS, or CPI in the present analysis. Notably, Lerman et al¹⁴ found a relationship between objective SSD and self-reported pain intensity during the week prior to the PSG. However, the present sample failed to show a relationship between objective SSD and pain intensity during the day and evening preceding the PSG. This discrepancy may be due to self-reported variables, as the current sample was selected based on orofacial muscle pain in the absence of clinical insomnia history, whereas Lerman et al¹⁴ sample was selected based on joint pain and insomnia presence.

The present study results are consistent with the hypothesis that sleep perception in TMD may be affected by daytime symptoms, similar to chronic insomnia. Insomnia is a self-reported condition without requirement of an objective sleep disturbance, and a subtype of insomnia with objectively normal sleep duration and architecture has been identified and termed paradoxical insomnia.³⁴ A recent study on sleep misperception in patients with depression and insomnia symptoms found that higher depression scores were associated with higher self-reported amount of wakefulness during the night, regardless of objective sleep duration.¹⁹ Another study showed that underestimation of TST on a PSG night was related to morning tiredness and higher PSQI scores.³⁵ This pattern is similar to the present findings in TMD cases. The cognitive model of insomnia proposes that cognitive processes may contribute to sleep misperception.¹⁶ Kim et al³⁶ suggested that individuals with sleep misperception may be more sensitive to insomnia-related cues (eg, clock-monitoring, daytime signs of tiredness, and fatigue) as evidenced by increased brain reactivity on fMRI in multiple areas involved in cognition and stress response, compared to controls and patients with insomnia without sleep misperception. Semler and Harvey’s study¹⁷ on patients with insomnia showed a misattribution of the daytime symptomatology to self-reportedly poor sleep in the absence of objective SSD. In the present study, misperception of sleep among TMD cases may reflect a cognitive bias of explaining daytime symptoms with poor sleep.

Although exact mechanisms linking paradoxical insomnia and TMD are currently unknown, there are 2 possibilities worth considering. First, research on the bidirectional relationship between sleep and pain suggests that the effects of chronic pain on mood and cognitive attributions may account for self-reportedly poor sleep.^37,38 This postulates a misattribution of daytime consequences of chronic pain to poor sleep even in the absence of objective sleep disturbance. One study found that cognitive behavioral therapy for pain and antidepressant medications improved PSQI-assessed sleep in women with TMD, with more sustained improvement in the group that received both treatments.³⁹ This study did not use PSG to establish objective sleep duration. Future studies might examine the relationship between periods of exacerbation vs remission of orofacial pain and associated daytime symptoms with self-reported sleep measures, specifically in patients with TMD whose objective sleep is already within normal limits. Moreover, to improve sleep experience and chronic insomnia, cognitive behavioral therapy for insomnia specifically addresses maladaptive attributions, sleep-interfering behaviors, and dysfunctional beliefs about sleep.⁴⁰ Future investigations could focus on the effect of cognitive behavioral therapy for insomnia on self-reported sleep and pain experience in patients with TMD and insomnia, with and without the misperception component.

An alternative interpretation of paradoxical insomnia suggests that the standard sleep EEG analysis does not capture the physiological mechanisms that may degrade sleep quality in insomnia and chronic pain conditions.⁴¹ For example, inconsistency between self-reported and actigraphy measures of sleep has been associated with inflammatory markers, particularly in women,⁴² whereas cognitive behavioral therapy for insomnia was shown to reduce inflammation.⁴³ Further, symptoms associated with anxiety disorder were predictive of self-reported-objective mismatch in SE.⁴⁴ Perceived wakefulness during PSG-defined sleep may be due to overactivity of brain regions involved with self-monitoring and emotional response to pain, such as increased pain vigilance which has been associated with decreased volume and reduced connectivity between insular and cingulate cortices.^45,46 Alterations of glucose metabolism in the insula and cingulate have been linked to self-reported-objective sleep discrepancies in both paradoxical insomnia patients and good sleepers.⁴⁷ The cingulate cortex and insula have rich connections with the prefrontal cortex,^48,49 and higher-EEG frequency power in the prefrontal cortex during sleep has been correlated with self-reported-objective discrepancy of sleep assessments among healthy sleepers.⁵⁰ Thus, physiological mechanisms involved in chronic pain, such as inflammation, activation of the hypothalamo-pituitary-adrenal axis and alterations in the functionality of several cortical regions may contribute to elevated levels of awareness during standard PSG-defined sleep among patients with TMD.

Future studies utilizing EEG and brain imaging measures are needed to elucidate the interplay between physiological and psychological parameters that may contribute to the self-reported sleep disruption in patients with TMD without abnormal PSG findings. Concurrent fMRI and EEG measures with pain vigilance and sleep perception measures would further clarify the role of connectivity between different brain regions in the pain and sleep experiences of patients with TMD.

Limitations

There are several limitations in the present study. First, due to the nature of secondary analysis, participants with low TST (< 4 hours) or previously diagnosed insomnia were excluded to ensure sufficient PSG sleep data for the primary research’s aim during data collection. This precludes the generalizability of the present findings to patients with TMD who may have particularly extreme objectively SSD. However, the TMD sample still included objectively short sleepers (TST < 6 hours), with a higher prevalence of objective SSD among TMD cases (30%) compared to controls (13%). The percentage of TMD cases with objective SSD was similar to the percentage reported by Lerman et al,¹⁴ who did not use the TST exclusion criterion yet found 24% of their TMD participants being objectively short sleepers. With a similar percentage of objectively short sleepers, our TMD sample revealed no relationship between objective SSD and self-reported measures of sleep, mood, daytime sleepiness, and pain. Therefore, sampling bias notwithstanding, the present evidence for sleep misperception in TMD appears to be valid.

Second, the PSQI may not be the best measure to analyze SSM. The PSQI ask respondents to provide estimates of sleep based on a preceding 30-day period, whereas in prior sleep misperception research, self-reported estimates of a single-night’s sleep were compared to respective night’s PSG recordings. However, the 2-night protocol employed in the present study, with the second night’s data used for analysis, presented an opportunity to eliminate the “first-night effect” and compare a PSG recording that is presumed to capture a participant’s typical sleep pattern at home with a self-reported measure widely used to assess sleep in patients with TMD. Additionally, in the development of the PSQI, single-night PSG recordings were used to test the validity of the questionnaire.²⁹ Therefore, lining up the PSQI data with second-night PSG data may have better ecological validity in the TMD population than 1 night self-reported-objective discrepancy.

Third, clinical evaluations for depression or chronic insomnia were not performed. The present study includes only questionnaires that assessed experience of sleep, mood, or daytime alertness. In future studies, clinical assessments of insomnia and depression could be used alongside self-reported measures.

Lastly, there is a limitation to external validity due to sampling from a care-seeking population and the availability-based recruitment of controls.

CONCLUSIONS

Overall, this study confirms an elevated prevalence of objective SSD, and provides evidence of SE misperception in TMD relative to controls. The misperception of SE is likely due to increased perception of wakefulness during the night, as SL perception was similar between the groups. Further, self-reported experience of sleep variables in TMD is consistently related to (1) depression symptoms, (2) pain intensity, in part mediated by depression symptoms, and (3) daytime sleepiness, albeit to a lesser extent. Self-reported sleep, measured by the PSQI, does not relate to objective SSD. These results show 2 independent sleep patterns among patients with TMD: objective SSD and sleep misperception in conjunction with depression symptoms and daytime sleepiness. Thus, paradoxical insomnia and SSD appear to be 2 independent facets of sleep experience in TMD. As PSQI variables generally do not align well with PSG variables in either controls or TMD cases, the PSQI should be used with caution to evaluate sleep in this population. Future studies are needed to elucidate relative contributions of brain pathophysiology, cognitive processes, and other factors to paradoxical insomnia in TMD.

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. Data collection was performed at the New York University College of Dentistry. The New York University School of Medicine approved the study. This research was supported in part by the National Institutes of Health, Bethesda, Maryland (grant number: R01 DE018569). The authors report no conflicts of interest.

ABBREVIATIONS

CPI

characteristic pain intensity

ESS

Epworth Sleepiness Scale

PSG

polysomnography

PSQI

Pittsburgh Sleep Quality Index

SCL-90

Symptom Checklist-90

SCL-90-D

Symptom Checklist-90-Depression

SE

sleep efficiency

SL

sleep latency

SSD

short sleep duration

SSM

sleep state misperception

TMD

temporomandibular disorder

TST

total sleep time

WASO

wake after sleep onset