Effects of sleep and circadian rest-activity rhythms on daily pain severity in women with temporomandibular disorders

Introduction

The link between sleep disturbance and chronic pain is well-established [16], with 73–75% of individuals with chronic pain experiencing sleep disturbances [53], and there is causal evidence that experimental sleep deprivation and sleep continuity disruption enhance pain sensitivity [16,27,28]. An emerging body of research suggests that disturbances in circadian rhythm—near-24-hour physiological and behavioral oscillations governed by the central clock [2]—are a novel risk factor for chronic pain [37]. Animal literature suggests that circadian rhythm disruptions can modulate pain through multiple mechanisms, including the expression of various pain-related genes and proteins, and the immune and endocrine systems [7,60]. In human studies, night-shift work, a major source of circadian rhythm disturbances, is related to a heightened likelihood of low back pain [15,54,61] and more severe musculoskeletal pain and headaches [36,59]. Additionally, studies have found significantly lower levels of plasma melatonin, a biological marker of the circadian system, in individuals with low back pain compared to healthy controls [56], and altered excretion patterns of urinary 6-sulfatoxymelatonin in patients with fibromyalgia [10].

Although the neurobiological mechanisms underlying sleep and circadian rhythms are closely intertwined [45], their co-assessment is scant in pain research. Hence, the extent to which circadian rhythm disturbances are associated with pain severity, in addition to sleep disturbances, is not well understood [37]. Given that circadian rhythm can fundamentally impact sleep, the moderating role of circadian rhythm disturbances on the association between sleep and pain experiences in individuals with chronic pain warrants further investigation. Additionally, most studies examining the role of circadian rest-activity rhythms¹ on pain have been cross-sectional and focused on the between-person associations [42,46,47].

The present study aims to address these research gaps by investigating a cohort of 140 women with temporomandibular disorders (TMD)—a group of chronic pain conditions disproportionately impacting women [6]—and insomnia symptoms. TMD is one of the most relevant chronic pain conditions for the present study, as it is not only highly prevalent (31% in the general population [58]), but also the majority of individuals with TMD report sleep-related issues [1]. Using both actigraphy and daily diary data, we examined the individual and interaction effects of daily sleep and circadian rest-activity rhythms on the severity of next-day pain. We hypothesized that disturbances in both daily sleep continuity (i.e., total sleep time and duration of wake after sleep onset) and circadian rest-activity rhythms (i.e., intradaily variability and relative amplitude) are associated with next-day pain severity. Additionally, we explored whether circadian rest-activity rhythms moderated the relationship between sleep and next-day pain severity. We hypothesized that on days with greater circadian rest-activity rhythm disturbances, the association between sleep disturbances and pain severity would be intensified.

Methods

This is a secondary data analysis study that utilized the baseline data from a clinical trial focused on evaluating the efficacy and mechanisms of psychological interventions for insomnia and pain catastrophizing in women with TMD and insomnia symptoms (ClinicalTrials.gov Identifier: NCT01794624). The primary outcome paper of this trial is currently under development. We have also published a number of papers using the same dataset [17,22,30,31,40,44]; however, none of these papers utilized circadian rest-activity rhythm data. The objectives of the present study are distinct from the those of our previously published papers.

Participants

Recruitment strategies included distributing fliers, airing radio ads, sending mails, and obtaining referrals from local dental professionals. As the parent study was a mechanistic clinical trial, it had strict eligibility criteria. Inclusion criteria were as follows: (1) female between 18 and 60 years of age; (2) positive TMD determined by a trained research dental hygienist based upon the Diagnostic Criteria for TMD [48]; (3) facial pain of myalgic or arthralgic origin reported for at least 3 months; (4) facial pain reported for at least 10 days out of the past 30 days; (5) past week average pain severity score ≥ 3 on a 0–10 numerical rating scale; (6) the Insomnia Severity Scale [3] and Pain Catastrophizing Scale [52] scores > 8; (7) difficulties in initiating and/or maintaining sleep (> 3 days per week) for at least one month; (8) using the consistent non-opioid medication for pain treatment regimen for the last 30 days; (9) willingness to undergo a 4-week washout period before enrolling in the study, if using an opioid pain medication or a benzodiazepine or sedating tricyclic antidepressant for sleep ≥ 3 days per week; (10) willingness to use contraception throughout the study if of child-bearing potential; (11) if post-menopausal, the participant should have been so for at least 12 consecutive months prior to screening; and (12) ability to comprehend and a willingness to adhere to all study protocols, while also being accessible for the entire study duration.

Exclusion criteria were as follows: (1) body mass index (BMI) > 35; (2) resting systolic blood pressure > 140 mm Hg and diastolic blood pressure > 90 mm Hg; (3) history of temporomandibular joint surgery or temporomandibular joint growth disturbances, neoplasm or injury to the temporomandibular joint area in the past 6 months; (4) surgery scheduled for TMD during study period; (5) history of health conditions impacting sleep, the central nervous system, or peripheral neuropathy (e.g., COPD, seizure disorder, cancer); (6) Raynaud’s Syndrome; (7) unstable severe psychiatric disorder; (8) problematic substance or alcohol use in the past 6 months; (9) regular use (≥ 3 times per week) of opioids, benzodiazepines, or sedating tricyclic antidepressants; (10) stable sleep phase between 10 am and 10 pm (i.e., night workers) or self-reported significant variability in sleep due to changes in work shifts; (11) Center for Epidemiologic Studies of Depression Scale (CES-D) scores ≥ 27 [43] or current suicidal ideation; (12) positive urine test result for barbiturates, THC, alcohol, cocaine, and other recreational drugs; (13) positive urine pregnancy test result; (14) Respiratory Disturbance Index (RDI) > 15 as determined from the baseline polysomnography (PSG); (15) periodic limb movement index with arousals > 15 as determined from the baseline PSG; and (16) any other factors that can hinder a participant’s completion of the study. Note that the presence of other co-morbid non-cancer chronic pain conditions was permitted in the present study, as TMD is recognized as one of the key chronic overlapping pain conditions [35].

The study enrolled participants from 2013 to 2018, with 1,642 individuals initially screened by phone. Of these, 496 met the preliminary eligibility criteria and underwent further in-person screening, ultimately resulting in 151 participants who fully qualified and consented to join the study. Of these participants, our study included a sample of 140 who provided concurrent actigraphy and daily diary data. Ethical approval was secured from the Institutional Review Boards at both the Johns Hopkins School of Medicine and the University of Maryland School of Dentistry. The decision to exclude male participants was made due to the predominant occurrence of TMD in women within clinical dental settings. The inclusion of male participants would introduce significant heterogeneity, potentially affecting the statistical power and our ability to investigate the underlying biological mechanisms of the parent mechanistic randomized controlled trial, which was focused on investigating mechanisms.

Procedures

First, individuals expressing interest in the study underwent a phone-based eligibility screening. Those who successfully passed this initial phone screening progressed to an in-person screening visit. During this in-person session conducted by dental hygienist research faculty, participants provided self-reports of their medical history and facial pain. Additionally, they underwent an examination by the dental hygienist to confirm alignment with research-case criteria according to the Diagnostic Criteria for TMD [48]. Eligible individuals subsequently completed a baseline study session. Then, research staff provided training on the use of wrist actigraphy and an interactive voice response (IVR) system for daily diary entries that participants completed for 14 days immediately following the visit.

Interactive Voice Response (IVR) Assessment

Participants received training on how to complete a diary twice a day for two weeks using the IVR assessment system. The initial diary assessment was available and permitted responses between 4 AM and 4 PM, immediately after participants woke up. This broad time window for the first diary assessment was implemented for two primary reasons: (1) to maximize the likelihood of participants completing their diaries, and (2) to accommodate participants with irregular wake-up times, as insomnia symptoms were part of the inclusion criteria for this study. Upon awakening, participants were instructed to dial a dedicated toll-free phone number, and the IVR system presented a series of pre-recorded questions related to their sleep, pain, emotions, and other relevant experiences. Participants entered their responses to these questions using the keypad on their phones. The end-of-day diary assessment took place between 8 PM and 12 AM before participants went to bed for the night. In cases where participants accidentally disconnected or were unable to complete a survey, they were permitted to log back into the IVR system and resume from where they had left off. To maintain data quality and ensure compliance, our research staff regularly reviewed the data on a weekly basis. Automatic reminder calls were made if the participant did not initiate the call on time. If any participant encountered difficulties with the IVR system, our research staff reached out to them and addressed any questions or concerns they may have had.

Wrist Actigraphy Assessment

Participants wore a wrist accelerometer (Actigraph GT3X+; ActiGraph LLC, Pensacola, FL, USA) on their non-dominant wrist continuously for 24 hours a day over a span of 14 days. Movement data were recorded at 60-second intervals. Concurrent with actigraphy, they completed morning and evening sleep diaries. At the conclusion of the two-week monitoring period, participants returned the device, and the data were processed using ActiLife software (Version 6.7.3; ActiGraph). We utilized the Wear Time Validation function in the ActiLife software to detect extended periods of inactivity that could indicate an off-wrist period. These identified periods were visually inspected and reviewed individually. If it was determined that an off-wrist period occurred during a given night, the record was marked as missing. In accordance with the actigraphy-based sleep monitoring guidelines outlined by the Society of Behavioral Sleep Medicine [1], the software initially auto-scored sleep periods using the Cole-Kripke algorithm. The algorithm reviews the initial minute-by-minute scoring within each time-in-bed window. Sleep onset time is determined as the start of the first sequence of five or more consecutive minutes scored as sleep. Similarly, the sleep offset time was determined by identifying the end of the last sequence of five or more consecutive minutes of sleep. Trained research staff manually verified these sleep periods by cross-referencing them with the data from the daily sleep diaries, if available. This manual verification aided in determining sleep periods and in excluding data points when the device was confirmed to be not on the wrist. Adjustments to the auto-scored sleep periods were made using a standardized process as follows: (1) If the auto-scored sleep time deviated by ≤ 60 minutes from the diary-reported “lights out” or “time out of bed,” the auto-scored time was used. (2) If the deviation exceeded 60 minutes, the actigraphy data were adjusted to align with the sleep diary. (3) In cases where the auto-scored data split a night into two separate sleep periods, the actigraphy was rescored to include both periods, provided this was consistent with the sleep diary.

Measures

Actigraphy-Derived Daily Sleep Variables:

We focus on two key sleep variables, the Total Sleep Time (TST) and Wake After Sleep Onset (WASO) measured in hours. TST refers to the total amount of time spent asleep, calculated as the sleep offset time minus the sleep onset time. WASO refers to the total duration of time spent awake after sleep onset, until the sleep offset. Sleep onset latency was not used in the present study, as it is known to be underestimated by actigraphy [49].

Actigraphy-Derived Daily Circadian Rest-Activity Rhythm Variables:

Using the ‘ActCR’ package [12] in R, we extracted non-parametric circadian rest-activity rhythm (nPCRA) variables from the actigraphy data. The ActCR package utilizes Axis 1 data from actigraphy to calculate nPCRA variables, and it requires complete 24-hour actigraphy data for analysis. Although several nPCRA variables are available, we focused on two variables: (1) Intradaily Variability, which measures the frequency and extent of transitions between rest and activity periods within a 24-hour cycle. Higher Intradaily Variability indicates a more fragmented rest-activity rhythm, reflecting frequent shifts between rest and activity states throughout the day; (2) Relative Amplitude, which quantifies the contrast between the most and least active phases over a 24-hour period. A greater Relative Amplitude signifies a more pronounced difference between periods of activity and rest, indicating a stronger and more robust (clearer) daily rest-activity rhythm. Examining both Relative Amplitude and Intradaily Variability provides a comprehensive understanding of an individual's daily circadian rest-activity rhythm by highlighting different aspects of circadian function that complement each other. Relative Amplitude measures the difference between the most active and least active periods within a 24-hour cycle, reflecting the overall strength and robustness of the circadian rhythm. Intradaily Variability, in contrast, measures the fragmentation and stability of the circadian rhythm by quantifying the frequency of transitions between rest and activity throughout the day. In short, Relative Amplitude highlights the overall strength and definition of the circadian rhythm, whereas Intradaily Variability provides insights into the consistency of this rhythm. Together, these measures help us understand both the robustness of the circadian signal and the potential disruptions or irregularities within it. Note that we have focused on these two daily nPCRA variables (i.e., Relative Amplitude and Intradaily Variability) computed by the ActCR package because the main goal of this study is to examine the prospective association at the daily level between circadian rest-activity rhythm disturbances and pain. Although several other measures are generated by the ActCR package, they are aggregates or averages across time, and thus serve as trait-like measures. Hence, these aggregated circadian measures (e.g., interdaily stability) can only allow for assessing cross-sectional between-person associations rather than prospective within-person longitudinal associations in relation to pain intensity. This is why we have focused on these daily circadian measures (i.e., Relative Amplitude and Intradaily Variability) in the present study.

Daily Pain Severity:

Participants were asked to rate the following diary item using the numerical rating scale (NRS) that ranges from 0 (no pain) to 10 (worst pain imaginable) [24] at the end-of-day diary assessment: “Please rate your average (typical) level of pain experienced throughout the entire day.” When we mentioned pain severity, we were referring to any form of pain experienced, not limited to TMD-related pain.” This broader approach was adopted because we anticipated a high likelihood of participants experiencing concurrent pain from other sources, such as headaches and/or neck pain, which are common in TMD. In addition, other chronic pain conditions such as fibromyalgia, irritable bowel syndrome, and chronic low back pain, frequently co-occur with TMD [34].

Data Analytic Plan

Demographics, actigraphy and daily diary completion rates, and descriptive statistics (i.e., means and standard deviations) and bivariate correlations among predictor variables were first examined. Note that TST and WASO were originally measured in minutes but were rescaled to hours to reduce computational burden. Using minutes as units resulted in extremely large variance estimates and very small coefficients and standard errors, making the model outputs difficult to interpret. To properly analyze the nested data (i.e., days nested in people), we employed linear mixed-effects regression models. These models allowed us to estimate the extent to which disturbances in sleep and circadian rest-activity rhythms from the previous day are associated with overall pain severity on the following day, while taking into account within-participant correlations (see Figure 1 that provides an overview of the analysis timeline). We selected linear mixed-effects regression modeling over other more sophisticated approaches (e.g., dynamic structural equation modeling) because of its robustness, simplicity in implementation, and suitability for directly assessing within-person associations in our study. Sleep and circadian variables were included as fixed effects, and participants were included as random effects to account for within-participant correlations. All predictors (TST, WASO, Intradaily Variability, and Relative Amplitude) in the model were person-mean centered, by subtracting the individual’s mean value from each observation, enabling us to interpret regression coefficients as pure within-person (level-1) effects [14]. As the investigation of interaction effects between sleep and circadian variables is exploratory in nature, the first linear mixed-effects regression (Model 1) only tested for main effects without considering the interaction effects. Next, we examined the interaction between sleep and circadian variables, focusing on one interaction at a time across four distinct models: (1) Model 2a: TST × Intradaily Variability; (2) Model 2b: TST × Relative Amplitude; (3) Model 2c: WASO × Intradaily Variability; and (4) Model 2d: WASO × Relative Amplitude. A total of four separate interaction models were conducted. To address the issue of family-wise error arising from conducting multiple comparisons, the Benjamini-Hochberg procedure was employed in the analysis interaction effects within our models [55]. This approach was chosen over the more conservative Bonferroni correction to more effectively manage the trade-off between detecting true effects and reducing the risk of false positives [41]. Lastly, we conducted a series of sensitivity analyses by excluding participants who provided fewer than 3, 5, and 7 days of actigraphy data, respectively. All statistical analyses were performed in R 4.3.0, and linear mixed-effects regression models were conducted using the ‘lme4’ package [4]. In mixed-effects modeling, missing data are handled using maximum likelihood estimation, which allows for the inclusion of all available data, even with some missing observations.

Figure 1.

The visual overview of lagged analysis timeline when integrating sleep and circadian data with daily diary data.

Results

Sample Characteristics

Table 1 displays the demographic characteristics of the current sample, consisting of 140 participants. The average age of the participants was 37 years (SD = 11.4). The majority of the participants (75%) identified themselves as White, while 17.1% identified as Black/African American, 5% identified as Asian, and the remaining 2.9% belonged to other racial groups/unknown. A small proportion of the participants (5.7%) identified as of Hispanic origin. In terms of their socio-economic status, a significant portion of the sample (68%) had completed a college degree or higher education, and 72% reported an annual household income of ≤ $50,000. Additionally, nearly half of the participants (45%) were either married or living with a partner.

Table 1.

Descriptive statistics of select demographic characteristics of the analytic sample (N = 140).

		M (SD)	Range
Age		37 (11)	19 – 61
		N	%
Race	White	105	75.0
	Black	24	17.1
	Asian	7	5.0
	Other/unknown	4	2.9
Ethnicity	Hispanic	8	5.7
Marital status	Single	62	46.3
	Living with partner	6	4.5
	Married	54	40.3
	Separated	4	3.0
	Divorced	7	5.2
	Widow/Widower	1	0.8
Education	Some high school	1	0.7
	High school graduate/GED	13	9.4
	Some college	30	21.6
	Technical school/College degree	56	40.0
	Masters/Doctoral degree	39	28.0
Income	≤ $25000	53	41
	$25001 – $50000	40	31
	$50001 – $75000	20	15
	> $75000	17	13

Note. There is missing data for some of the demographic variables, therefore the total does not necessarily equal N = 140.M = Mean. SD = Standard deviation.

Actigraphy and Diary Completion Rates

Participants who completed at least 7 days of actigraphy and evening diary assessment were 83% and 69%, respectively. 66% had at least 7 days of aligned data (i.e., t and t+1).

Descriptive Statistics and Bivariate Correlations Among Predictor Variables

Table 2 presents the descriptive statistics and bivariate correlations among the sleep and circadian predictor variables. Statistically significant correlations were observed among all sleep and circadian rest-activity rhythm variables, ranging in magnitude from small to moderate. These findings suggest that each of the sleep and circadian variable measures distinct aspects, and there is no evidence of multicollinearity. Supplementary Table S1 provides additional descriptive statistics for the L5 (lowest 5-hour activity) and M10 (most active 10-hour activity) variables derived from actigraphy data, along with their correlations with other sleep and circadian variables.

Table 2.

Bivariate associations between sleep and circadian variables of interest (N = 140).

	M (SD) [IQR]	TST	WASO	RA
TST	7.1 (1.6) [6.1, 8.1]
WASO	.97 (.92) [.42, 1.22]	−.07 [−.12, −.01]*
RA	.92 (.09) [.91, .97]	.43 [.38, .48]***	−.38 [−.43, −.33]***
IV	.54 (.12) [.47, .60]	−.16 [−.22, −.11]***	.07 [.02, .13]*	−.21 [−.26, −.16]***

^*p < .05.

^**p < .01.

^***p < .001.

Correlations are presented as Pearson correlation coefficients; 95% CI are presented in brackets. M = Mean. SD = Standard deviation. IQR = Interquartile range. TST = Total Sleep Time (in hours). WASO = Wake After Sleep Onset (in hours). RA = Relative Amplitude. IV = Intradaily Variability.

Findings of Linear Mixed-Effects Regression Model Without Including Interaction Terms

As indicated in Table 3, when all sleep and circadian variables were included in the same model, both TST (b = -.11, 95% CI = -.18, -.03, p = .006) and WASO (b = .18, 95% CI = .05, .32 p = .006) from the previous night, along with Relative Amplitude (b = -2.56, 95% CI = -4.03, -1.09, p = .001) from the past 24 hours, significantly predicted next-day pain severity. Specifically, shorter sleep duration, greater sleep continuity disturbance, and less robustness of 24h circadian rest-activity rhythm were each associated with greater next-day pain severity. The magnitude of these significant effects was small. For instance, in the case of TST, a one-hour increase is associated with a 0.11 unit decrease in next-day pain severity (measured on a 0–10 NRS scale), whereas in the case of WASO, a one-hour increase is associated with a 0.18 unit increase in next-day pain severity.

Table 3.

Unstandardized coefficients from linear mixed effects regression models examining the associations between sleep and circadian variables and next-day overall pain severity.

	Model 1				Model 2a				Model 2b				Model 2c				Model 2d
	(No interaction term added)				(Tests TST × IV Interaction)				(Tests TST × RA Interaction)				(Tests WASO × IV Interaction)				(Tests WASO × RA Interaction)
	Est	SE	95% CI	p	Est	SE	95% CI	p	Est	SE	95% CI	p	Est	SE	95% CI	p	Est	SE	95% CI	p
Intercept	3.26	.17	2.92, 3.60	<.001	3.27	.17	2.93, 3.61	<.001	3.25	.17	2.91, 3.59	<.001	3.25	.17	2.92, 3.59	<.001	3.24	.17	2.91, 3.57	<.001
TST	−.11	.04	−.18, −.03	.006	−.14	.04	−.22, −.07	<.001	−.09	.04	−.16, −.01	.023
WASO	.18	.07	.05, .32	.006									.23	.07	.10, .36	<.001	.11	.07	−.02, .25	.103
IV	−.43	.49	−1.38, .52	.375	.01	.48	−.93, .96	.981					.16	.48	−.78, 1.09	.746
RA	−2.56	.75	−4.03, -1.09	.001					−2.73	.79	−4.28, -1.18	.001					−2.76	.70	−4.13, -1.39	<.001
TST × IV					.54	.37	−.19, 1.27	.147
TST × RA									.41	.40	−.37, 1.20	.305
WASO × IV													1.52	.72	.11, 2.92	.034
WASO × RA																	−1.52	,50	−2.50, −.53	.003
σ2		2.32				2.38				2.34				2.38				2.33
τID		3.57				3.55				3.52				3.53				3.47
ICC		.60				.60				.60				.60				.60
R 2		.611				.602				.606				.600				.605

Note. Participants are included as random effects. Est = Unstandardized regression coefficients. 95% CI = 95% confidence intervals. TST = Total sleep time (in hours). WASO = Wake after sleep onset (in hours). RA = Relative amplitude. IV = Intradaily variability. σ² = residual variance. τ_ID = variance of random intercept. ICC = intraclass correlation.

Findings of Linear Mixed-Effects Regression Models Including Interaction Terms

Among the four interaction terms tested (see Table 3), significant effects were observed for two interactions: (1) between WASO and Relative Amplitude (Model 2d: b = -1.52, 95% CI = -2.50, -0.53, p = .003) and (2) between WASO and Intradaily Variability (Model 2c: b = 1.52, 95% CI = 0.11, 2.92, p = .034). Figure 2 illustrates the interaction between WASO and Relative Amplitude on next-day pain severity. When a participant’s WASO is 42 minutes (0.7 hours) shorter than their usual WASO, there is no significant interaction between WASO and Relative Amplitude in predicting next-day pain severity. However, when WASO is not at least 42 minutes shorter than their typical level, the interaction becomes significant and more pronounced. Specifically, on days with a more robust circadian rest-activity rhythm (i.e., higher Relative Amplitude), the positive association between previous night’s WASO and next-day pain severity is attenuated.

Figure 2.

Estimated pain severity by the interaction between person-centered Wake After Sleep Onset (hours) and person-centered Relative Amplitude (0 and ± 1.5 SD).

Note. Shaded area denotes 95% confidence intervals. Vertical long dashed line denotes the threshold at which the interaction effect is statistically significant at p < .05, as per Johnson-Neyman approach. Horizontal thick solid line denotes the range of observed person-centered Wake After Sleep Onset (hours) data.

Regarding the interaction between WASO and Intradaily Variability, as shown in Figure 3, there is a general positive association between WASO and pain severity, with greater WASO linked to increased next-day pain severity. The interaction between WASO and Intradaily Variability on pain severity becomes evidence when WASO is either significantly shorter than usual (more than 2.4 hours shorter) or higher than usual (more than 1 hour longer). On days with greater circadian rest-activity rhythm fragmentation (i.e., higher Intradaily Variability), the positive association between previous night’s WASO and next-day pain severity is intensified.

Figure 3.

Estimated pain severity by the interaction between person-centered Wake After Sleep Onset (hours) and person-centered Intradaily Variability (0 and ± 1.5 SD).

Note. Shaded area denotes 95% confidence intervals. Vertical long dashed line denotes the threshold at which the interaction effect is statistically significant at p < .05, as per Johnson-Neyman approach. Horizontal thick solid line denotes the range of observed person-centered Wake After Sleep Onset (hours) data.

Results of the Sensitivity Analyses

To test the robustness of our study findings, we conducted sensitivity analyses by excluding participants who provided fewer than 3, 5, and 7 days of actigraphy data, respectively. All of the linear mixed-effects model findings were very similar to those obtained with the full sample. Thus, we provide the summary of this sensitivity analyses in Supplement Tables S2-S4.

Discussion

The present study aimed to enhance our nascent understanding of the interplay between sleep and circadian rest-activity rhythm disturbances on pain severity on a day-to-day basis among individuals with chronic pain. Our findings affirm the importance of the roles that sleep duration (i.e., TST) and sleep maintenance (i.e., WASO), as well as the robustness of the circadian rest-activity rhythm (i.e., Relative Amplitude), play in regulating next-day pain severity. Furthermore, we found that both circadian rest-activity rhythm robustness (i.e., Relative Amplitude) and fragmentation (i.e., Intradaily Variability) can modulate the link between daily sleep maintenance disturbance and pain severity. Overall, these findings align well with and extend existing literature that underscores the role of sleep and circadian rhythm disturbances in pain-related experiences in both healthy adults and those with chronic pain [16,21,37,51].

Notably, not only did bivariate correlation analysis reveal that sleep and circadian rest-activity rhythm variables were only moderately correlated, but the linear mixed-effects regression model also showed that Relative Amplitude was independently associated with next-day pain severity, above and beyond the effects of TST and WASO. This finding is consistent with recent cross-sectional studies showing the unique effects of late chronotype (or eveningness)—a trait marker of circadian preference that is misaligned with societal clock, and hence often linked to adverse health outcomes and lower circadian amplitude [19]—on pain outcomes independent of sleep-related factors (e.g., sleep quality, insomnia symptoms, and sleep duration) [20,25,57]. Importantly, a recent small study employing machine learning found that circadian rest-activity rhythm features extracted from 5-days of actigraphy data among 25 individuals with chronic pain more accurately predicted pain severity compared to using just sleep or activity features derived from actigraphy data [46]. Our findings, in the context of previous research, suggest that disturbances in circadian rest-activity rhythms, alongside sleep disturbances, may play a unique role in shaping daily pain experiences.

Interestingly, both Relative Amplitude and Intradaily Variability moderated the relationship between the previous nighťs wake after sleep onset (WASO) and next-day pain severity, whereas no such moderation effect was observed in the association between total sleep time (TST) and pain severity. These interaction effects suggest that disruptions in sleep maintenance, in particular, are crucial sleep components that interact with circadian rest-activity rhythms to exacerbate the perception of clinical pain. Although we cannot provide a definitive explanation for the lack of significant interactions between TST and circadian rest-activity rhythm indices, it is noteworthy that our previous laboratory study comparing the effects of sleep maintenance disruptions versus simple sleep loss among healthy adult women found that sleep maintenance disturbance (i.e., increase in WASO), but not simple sleep loss (i.e., decrease in TST), impaired endogenous pain-inhibitory function, increased spontaneous pain, and enhanced pain facilitatory processes (mechanical temporal summation) in women [50]. This discrepancy may point to potential underlying neurobiological pathways that render the effects of sleep maintenance disturbances on pain particularly detrimental and susceptible to circadian disturbances. Future laboratory studies directly manipulating different types of sleep and circadian disruptions would provide deeper insights into how sleep and circadian rhythms interactively impact pain experiences.

The changes in pain severity indicated by our linear mixed-effects regression model were statistically significant, but represent small changes (e.g., a one-hour increase TST is associated with a 0.11 decrease in next-day pain severity measured on a 0–10 NRS scale). However, it is important to note that although a 2-point reduction in pain on a 0–10 NRS is commonly accepted as a clinically meaningful cutoff in the context of treatment and clinical trials [13], there is no established guideline or consensus for interpreting daily or momentary changes in pain in observational studies [38]. Additionally, studies have shown that within-person effects on pain typically have substantially smaller magnitudes than between-person effects [11,18,26,39]. This may be due to the fact that day-to-day variations in an individual's pain level are generally less pronounced than differences in pain levels between individuals or differences resulting from exposure to some intervention. Furthermore, despite their small size, these daily effects should not be completely discounted, as they may accumulate over time and provide valuable clinical insights.

Although preliminary, our findings suggest important implications for clinical practice. Interventions designed to both improve sleep and restore circadian rhythms may offer enhanced benefits for chronic pain management. For instance, there may be value in further optimizing CBT-I, the gold standard treatment for insomnia, so that it places more emphasis on regularly assessing patients’ rest-activity rhythm and also helps them maintain a fixed sleep-wake schedule to reduce circadian rhythm fragmentation. CBT-I may also be combined with other evidence-based chronobiological therapies, such as bright light therapy, which are safe, low-cost, and low-burden for patients to enhance pain outcomes. At least three studies to date suggest bright light therapy can improve clinical pain outcomes, as well as pain threshold and tolerance measured by quantitative sensory testing, among individuals with chronic back pain and fibromyalgia [8,9,29]. In fact, outside of pain literature, several researchers have begun examining the effects of combining CBT-I with a chronobiological therapy (e.g., bright light therapy, evening blue light filtration) [5,23,32,33]. There is now a need for rigorous randomized-controlled trials (RCTs) to evaluate whether CBT-I combined with a chronobiological therapy has additive or synergistic effects on improving chronic pain outcomes. For instance, an RCT to examine whether CBT-I combined with bright light therapy shows greater efficacy in improving chronic pain outcomes compared to when CBT-I or bright light therapy are administered alone.

The present study had several important limitations. First, the generalizability of our study findings may be limited, as our participants were women with TMDs and insomnia symptoms. Also, participants were predominantly of White race, and there were numerous inclusion and exclusion criteria. Second, although the present study was longitudinal, our findings provide no causal inference. Future laboratory studies that directly manipulate sleep and circadian rhythms could provide causal evidence, as well as greater insights into their unique and interactive effects on pain. Third, daily diary adherence in the present study was quite low. However, we addressed this limitation by employing maximum likelihood estimation in linear mixed-effects modeling, a robust method for handling missing data that uses all available data to estimate the model parameters. Fourth, the assessment of sleep and circadian rhythms was only based on actigraphy, which relies solely on individuals’ activity levels rather than multiple physiological parameters. Objective assessments of sleep, on the other hand, such as ambulatory polysomnography, can be conducted in participants' home settings over multiple nights, but lack the ecological validity and cost-effectiveness required for intensive longitudinal designs. Additionally, assessments of circadian rhythms, using urinary 6-sulfatoxymelatonin or dim light melatonin onset (DLMO), may provide deeper insights into our findings.

Conclusion

The present study contributes to the nuanced understanding of how circadian rest-activity rhythms interact with sleep in affecting chronic pain, particularly among women with TMD co-occurring with insomnia symptoms. We found that the robustness of circadian rest-activity rhythms in the past 24 hours was associated with next-day pain severity, over and above previous night sleep duration and sleep continuity disturbance. We also found that on days with less robust and greater fragmentation of circadian rest-activity rhythms, the positive association between sleep continuity disturbance and next-day pain severity was intensified. These findings highlight the importance of considering circadian rest-activity rhythms when examining the association between sleep and pain. Further research is needed to explore these relationships in other chronic pain conditions using more rigorous experimental designs and more objective assessments of sleep and circadian rhythms. Although preliminary, our findings suggest that integrating evidence-based sleep and circadian treatment strategies may offer a novel non-pharmacological approach to chronic pain management. Future studies should undertake rigorous RCTs to determine the full potential and efficacy of combining sleep and chronobiological therapies.