Linking non-restorative sleep and activity interference through pain catastrophizing and pain severity: An intra-day process model among individuals with fibromyalgia

Abstract

Non-restorative sleep is a key diagnostic feature of the musculoskeletal pain disorder fibromyalgia, and is robustly associated with poor physical functioning, including activity interference. However, the mechanisms through which non-restorative sleep elicits activity interference among individuals with fibromyalgia at the within-person level remain unclear. The present study tested the following three-path mediation model, using data gathered from a 21-day electronic daily diary in 220 individuals with fibromyalgia: previous night non-restorative sleep → morning pain catastrophizing → afternoon pain severity → end-of-day activity interference. Results of multilevel structural equation modeling supported the three-path mediation model. Previous night’s non-restorative sleep and morning pain catastrophizing were also directly related to end-of-day activity interference. Previous night non-restorative sleep did not significantly predict afternoon pain severity while controlling for the effect of morning pain catastrophizing. Greater non-restorative sleep during the previous night and a higher level of morning pain catastrophizing appear to serve as risk factors for experiencing greater daily pain and activity interference later in the day. These findings point to the potential utility of targeted interventions that improve both sleep quality and pain catastrophizing to help individuals with chronic pain engage in important daily activities despite experiencing pain.

Introduction

Fibromyalgia (FM) is a chronic widespread musculoskeletal pain condition that predominantly affects women and interferes with individuals’ daily functioning^58,80. Repeated activity interference can substantially limit one’s successful adjustment to chronic pain through the development of negative self-schema and reduction of rewards that stem from engaging in meaningful and enjoyable activities^23,26. Hence, a central target of chronic pain management is helping individuals increase their engagement in and performance of important daily activities¹⁴.

One of the key diagnostic and clinical features of FM that is modifiable and has implications in modulating individuals’ daily functioning is non-restorative sleep, a subjective perception of poor sleep quality and being unrefreshed upon awakening^63,76. More than 90% of individuals with FM report experiencing non-restorative sleep^5,70, and ample evidence indicates that poor sleep quality is robustly associated with activity interference and decreased daily goal pursuit^2,5,32,41. However, the intra-day processes that link non-restorative sleep with activity interference among individuals with FM are not well understood. Identifying the underlying mechanisms can inform our efforts to develop targeted intervention strategies to diminish the negative impact of non-restorative sleep on daily functioning of individuals with FM.

Both experimental and observational studies have repeatedly demonstrated that sleep can modulate individuals’ pain sensations and perception^20,34,57. Greater pain experience, in turn, is considered one of the strongest precursors of activity interference^33,54. Combined, these findings suggest that non-restorative sleep may promote heightened pain-related activity interference. However, the factors mediating the relation between non-restorative sleep and heightened pain experience have been largely unexplored, especially beyond the laboratory. Pain catastrophizing, a salient maladaptive pain coping style characterized by rumination, helplessness, and magnification⁶⁵, is a promising candidate as a mediator of this relation. Previous studies have found that individuals with FM who are poor sleepers report high levels of pain catastrophizing³⁷, and that poorer sleep quality predicts higher pain catastrophizing^24,77. This may be because poor sleep quality induces biased attentional and interpretative processes that can perpetuate and/or exacerbate pain sensitivity^25,51,52. Hence, when an individual experiences non-restorative sleep, he or she may be more likely to manifest higher pain catastrophizing. Pain catastrophizing, in turn, is a potent predictor of various pain-related experiences, including pain severity and activity interference⁵⁰. In fact, using a cross-lagged model, Campbell and her colleagues found that pain catastrophizing preceded and contributed to subsequent changes in the pain experience of patients with FM⁸. When individuals experience higher than usual non-restorative sleep, they may be more likely to catastrophize, which in turn, may precipitate increased pain.

In the present study, we proposed an intra-day process model that links the relation between non-restorative sleep and activity interference. Although the associations between components (i.e., non-restorative sleep, pain catastrophizing, pain severity, and activity interference) have been examined previously, a model that integrates all these factors simultaneously at the within-person level has not been explored. Investigating this intra-day process model can yield a better understanding of “how” previous night non-restorative sleep is associated with activity interference during the day, providing additional insight for researchers and clinicians who aim to help individuals with FM improve their daily physical function. Specifically, in the present study we tested a model in which morning pain catastrophizing and afternoon pain severity are sequential mediators of the prospective association between morning reports of the previous night’s non-restorative sleep and end-of-day activity interference. We hypothesized that on days when individuals report greater than usual non-restorative sleep for the prior night, they experience more activity interference at the end of the day through higher morning pain catastrophizing and afternoon pain severity. This hypothesized intra-day process model is tested by means of a three-path mediation model⁶⁷, which is illustrated in Figure 1. Lastly, we conducted sensitivity analyses to evaluate if this model is robust from individual differences including overall sleep quality, depression status, and pain management medication (i.e., opioids and tricyclics) use that may modulate individuals’ daily experience of sleep, pain, catastrophizing, and physical function.

Figure 1.

Visual illustration of hypothesized multilevel three-path mediated model.

Method

This is a secondary data analysis of a parent study which collected daily diary data at both pre-and post-intervention trials. The focus of the parent study was to compare the benefits of cognitive, acceptance-based, and educational approaches for functional and affective health in individuals with fibromyalgia. The main intervention outcome paper is currently in preparation. There have been a number of studies that were published based upon the pre-intervention diary data^69,78,79,83. However, none of these papers overlap with the aims of the present study.

Participants

Participants were recruited from the Phoenix, Arizona metropolitan area as part of a larger randomized clinical trial for fibromyalgia (FM) between 2008 and 2013. Recruitment method included physician referrals, newspaper advertisements, online positing, fliers posted in local physicians’ offices, and FM support groups. Inclusion criteria were: 1) ages between 18 and 72 years; 2) speaks English; 3) have had pain lasting three or more months in at least four quadrants of the body, or in two quadrants of the body with substantial sleep disturbance and fatigue; and 4) report pain in at least 11 of 18 tender points during a physical examination at a home visit, consistent with diagnostic criteria for fibromyalgia established by the American College of Rheumatology⁸¹. Exclusion criteria were: 1) pursuing litigation related to pain; 2) participating in a psychosocial treatment for pain or mood disturbance; and 3) having autoimmune or neuropathic pain conditions. Participants were not allowed to engage in psychological treatment for mood and/or pain symptoms throughout the duration of the study. However, they were not refrained from using some other complementary and alternative medicines (e.g., acupuncture) or physical therapy during the study. Two hundred twenty participants who consented to participate, met the inclusion and exclusion criteria, and participated in pre-intervention daily diary assessment were used in the present study.

Participant screening and survey data assessment.

Individuals who expressed interest in the study were first screened by phone to determine their eligibility. Those who screened to be eligible then received a tender point examination performed by a trained nurse during a home visit. For this exam, a dolorimeter was used to administer 4 kilograms of pressure to 18 tender point and 3 control point sites on the body. Once enrolled into the study, those participants who signed the informed consent form received an initial packet of questionnaires that measured their pain-related experiences and physical and psychosocial health. Trained research staff members also conducted phone interviews to assess participants’ history of depression, trauma, and other significant life events. All the study procedures were approved by the Institutional Review Board at Arizona State University prior to initiating the study.

Daily diary assessment.

Prior to starting the pre-intervention diary assessment, a research staff member met with participants individually to provide them with a cell phone and instructions on how to complete the diary through the phone. During the 21-day diary, an automated phone system (i.e., Interactive Voice Response system) called each participant on their phone four times a day and delivered audio recorded questions. Participants were asked to indicate their responses to questions using the phone keypad. The four assessment time points per day were: 1) in the early-morning, 20 minutes after a wake up time specified by the participant (the time that participant set for the early-morning call was consistent across the 21 days and the call assessed only sleep quality); 2) in the late-morning at 11:00 am (subsequently referred to as “morning”); 3) in the afternoon at 4:00 pm; and 4) at the end of the day shortly before going to bed. If a participant missed a call, he or she was instructed to call into the system within three hours to complete the assessment. Participants were not allowed to call into the system to complete the early-morning sleep survey after 10:30 AM. Participants’ calls were regularly monitored by research staff members and if a participant missed all diary assessments two days in a row, a research staff contacted the individual to address any barriers to completing diary. Participants were paid $2/day for completing diaries with a bonus of $1/day for a 50% or greater diary completion rate.

Day-level variables

Previous Night’s Non-Restorative Sleep:

Each morning, 20 minutes after participants’ usual wake up time, they were asked to rate the quality of their sleep on the previous night on a 101-point scale (0 = “extremely poor sleep” to 100 = “extremely good sleep”), and how refreshed they felt upon awakening on a 101-point scale (0 = “not at all refreshed” to 100 = “extremely refreshed”). These sleep diary items are analogous to those in the PROMIS Sleep Disturbance scale⁶. They were reverse-coded so as to indicate that higher values represent greater non-restorative sleep. The within-person correlation between the two items was .60, and we computed the mean of these two items.

Morning Pain Catastrophizing:

Participants were asked to rate the degree to which they experienced pain catastrophizing (i.e., “you felt your pain was so bad you couldn’t stand it anymore”) in the past two to three hours on a five-point numerical scale ranging from 1 (not at all) to 5 (completely). The item was adapted from the Pain Catastrophizing Scale⁶⁶.

Afternoon Pain Severity:

Participants were asked to rate their average level of pain in the past two to three hours using the standard Numerical Rating Scale²⁷ that ranges from 0 (no pain) to 100 (pain as bad as it can be). Note that the pain severity rating was also measured in the morning and was used as a covariate in our model.

End-of-Day Activity Interference:

End-of-day activity interference was measured by four items drawn from SF-36 Role Physical subscale⁷⁵. Participants were asked to rate the following items using a three-point numerical scale ranging from 1 (no) to 3 (yes, very much): (1) “Did you cut down on the amount of time spent on work or other activities?”, (2) “Did you accomplish less than you would have liked?”, (3) “Were you limited in the kind of work or other activities you did?”, and (4) “Did you have difficulty performing work or other activities?” We computed means of these four items. The within-person reliability of these items was 0.74.

Person-level variables for sensitivity analyses

Poor Sleepers vs. Non-Poor Sleepers.

By averaging daily non-restorative sleep assessments across 21-days, we categorized participants into poor sleepers vs. non-poor sleepers. Those we scored more than 1.5 standard deviation above the mean of average non-restorative sleep (i.e., > 65.11 out of 100) classified as poor sleepers (n = 37), and the remaining 177 participants were categorized as non-poor sleepers.

Individuals with Clinical Level Depression vs. Non-Clinical Level Depression.

The Hamilton Depression Inventory-Short Form (HDI-SF; Reynolds & Kobak, 1995) was used to examine participants’ clinical level depression status. There are 15 items in HDI-SF and scores can range between 0 and 52; scores greater than 13 are thought to signify clinical level of depression. Based upon this clinical cutoff, we classified participants into either clinical level depression group (n = 66) or non-depressed group (n = 124). HDI-SF is known for its strong reliability and validity⁵³. Cronbach’s alpha for the HDI-SF was .88.

Tricyclics and Opioid Medication Use.

Participants were asked to list the tricyclics and opioid medications they were currently using. A total 21 participants reported using tricyclics and 109 reported using opioids.

Data Analytic Plan

Prior to conducting the main analyses, diary completion rates, descriptive statistics including means, standard deviations, and intra-class correlations (ICCs) were calculated. As our data structure was hierarchical (i.e., repeated daily assessments nested in each participant), a multilevel structural equation model (MSEM) was employed to test the hypothesized model. MSEM allows for testing complex multilevel models such as the three-path mediation model evaluated in the current study^43,49. As it is possible that the effect of morning pain catastrophizing on afternoon pain severity is confounded by the participants’ experience of morning pain severity, morning pain severity (i.e., auto-correlation) was included as a control variable in the final model.

MSEM partitions the within-and between-person level variances, and therefore, path coefficients are directly interpreted at the corresponding levels of analysis⁴⁹. In the present study, we report the results of only the within-person level model because (1) the research questions and hypotheses of the present study are all at the within-person level, and (2) the associations between study variables in model at the between-person level are all cross-sectional, i.e., they represent average of 21-day diary assessments.

In terms of centering, all within-person predictors in MSEM were person-mean centered, which is the conventional method employed in multilevel modeling¹⁹. The person-centered scores represent day-to-day deviations from an individual’s own mean score over the entire 21-day diary period for that variable. Hence, person-centered predictors can address research questions such as, “On days when individuals experience greater than their usual level of non-restorative sleep, do they experience higher levels of morning pain catastrophizing?”

All analyses were conducted by Mplus version 8.0⁴⁵. To test the three-path mediated effect, we used the joint significance test. According to Taylor et al’s simulation study⁶⁷, the joint significance test provides the best control of Type I error rates and adequate statistical power. We were not able to use the percentile bootstrap method that provides 95% confidence limits of a three-path mediated effect because this function is not yet available in the MSEM framework. Use of the joint significance test for examining the three-path mediated effect requires three paths to be statistically significant at an alpha level of 0.05. These paths are: (1) the path from the independent variable to the first mediator (i.e., previous night non-restorative sleep → morning pain catastrophizing), (2) the path from the first mediator to the second mediator (i.e., morning pain catastrophizing → afternoon pain severity), and (3) the path from the second mediator to the dependent variable (i.e., afternoon pain severity → end-of-day activity interference). If all three paths are significant at p-value less than 0.05, then a significant three-path mediated effect can be inferred.

Model fit was evaluated by using comparative fit index (CFI), root mean square of approximation (RMSEA), and the standardized root mean squared residual (SRMR). Values greater than 0.90 for CFI, less than 0.08 for RMSEA, and below 0.10 for SRMR indicate an acceptable model fit^7,31. We used the full information maximum likelihood (FIML) under the missing at random assumption to handle missing data³⁸. FIML is an advanced and one of the most powerful methods to handle missing data. Instead of imputing missing values, FIML estimates a maximum likelihood function for each individual case using all the observed data and provide the most plausible model parameters¹⁷. FIML is superior to other conventional missing data strategies including listwise deletion, pairwise deletion, and mean imputation, and it provide unbiased regression estimates under MAR assumption¹⁸. In calculating effect sizes, we used R² (variance explained by predictors). Unfortunately, Mplus statistical software cannot calculate how each of the predictors (i.e., exogenous variables) contribute to the R-square of the dependent (i.e., endogenous) variables. As an alternative, we provided standardized regression path coefficients for each of the paths in Figure 2.

Figure 2.

Morning pain catastrophizing and afternoon pain severity as sequential mediators of the relationship between previous night non-restorative sleep and end-of-day activity interference. Note. Both unstandardized and standardized regression coefficients are provided. Values in brackets are standardized path coefficients. *p < .05, **p < .01, ***p < .001

For sensitivity analyses, we used multiple group analyses. Multiple group analysis⁴⁴ allows for examining whether regression paths in the within-person model significantly differ by subgroups of individuals with fibromyalgia who exhibit different personal characteristics. There are two major steps in conducting multiple group analysis. First, two models are compared using a chi-square difference test: (1) a configurable model that allows all the within-person level regression paths to be freely estimated across groups, and (2) a constrained model that restricts all paths to be equal across groups. Basically, this analytic approach is similar to the omnibus F-test in ANOVA, which tests the overall difference between groups (factors). Second, if there is a significant chi-square difference between the configurable vs. constrained model, then each regression path is examined to see if there are any significant differences across groups. However, if there is no statistically significant chi-square difference between the two models, we did not proceed to this second step (similar to the conventional approach to ANOVA) as it can significantly elevate Type I errors.

Results

Diary Completion Rates

Overall, participants completed 3,796 of 4,620 (82%) diary assessments possible across 21 days. More specifically, participants completed on average 17.5 days (SD = 4.24) of early-morning assessments of previous night non-restorative sleep, 17.8 days (SD = 4.55) of morning pain catastrophizing, 16.5 days (SD = 5.24) of afternoon pain severity, 17.1 days (SD = 4.87) of end-of-day activity interference.

Sample Characteristics

Table 1 summarizes demographic characteristics of our sample. Participants were predominantly European White, middle-aged women. Approximately 64% of the participants were either married or partnered. About half (53.1%) of the participants were employed and currently working at least part-time. Participants’ median education level was 1 to 3 years of college and their median annual household income level was in the range of $40,000 to $59,999.

Table 1.

Sample characteristics (N = 220)

Variables	Mean or % (SD)
Age (years)	51.25 (11.02)
Gender
Male	11.2
Female	87.0
Education
Less than high school	2.2
Completed high school	13.0
Post high school	13.4
1–3 years of college	33.2
4 years of college	17.5
Post graduate	17.0
Marital Status
Never married	8.1
Married/partnered	55.3
Widowed	5.8
Divorced	27.4
Separated	1.4
Employment
Employed	53.1
Not working	46.9
Race/Ethnicity
Caucasian	78.0
Black/African American	2.7
Asian	1.3
Hispanic	14.3
Native American	4.0
Native Hawaiian/Pacific Islander	0.9
Other	3.6
Income
Under $3,000-$20,999	25.6
$21,000-$39,999	22.0
$40,000-$59,999	17.9
$60,000-$99,999	19.7
$100,000 and over	8.1

Preliminary Findings

Table 2 displays descriptive statistics and ICCs. Participants reported moderate levels of previous night non-restorative sleep, morning pain catastrophizing, morning pain severity, and end-of-day activity interference. ICCs ranged from .30 (previous night non-restorative sleep) to .51 (afternoon pain severity), indicating use of multi-level modeling is appropriate given that there was substantial within-person variation. For instance, only 30% of the variation was explained by between-person differences in previous night’s non-restorative sleep and the rest of the variation was explained by within-person changes. All study variables were significantly correlated in expected directions. For instance, greater previous night non-restorative sleep was associated with higher morning pain catastrophizing, afternoon pain severity, and end-of-day activity interference.

Table 2.

Descriptive statistics and bi-variate within-person correlations of study variables

Variables	1	2	3	4
1. Previous night non-restorative sleep	–	.18**	.06*	.17**
2. Morning pain catastrophizing		–	.48**	.28**
3. Afternoon pain severity			–	.28**
4. End-of-day activity interference				–
Mean	51.37	2.12	50.22	2.00
SD	23.50	1.16	24.27	.63
Observed Range	0–100	1–5	0–100	1–3
ICC	.30	.45	.51	.37

Note.

^*p < .01,

^**p < .001

Test of the three-path mediation model

Figure 2 presents a summary of the MSEM findings, including the unstandardized and standardized path estimates. Model fit indices showed good fit to the present data, CFI = .98, RMSEA = .05, SRMR-within = .03, and SRMR-between = .05.

On days with greater than average previous night non-restorative sleep, participants reported higher morning pain catastrophizing (Unstandardized B = .11, SE = .01, p < .001). However, previous night non-restorative sleep did not significantly predict afternoon pain severity (B = .47, SE = .30, p = .12), controlling for morning pain catastrophizing. Higher morning pain catastrophizing was associated with higher afternoon pain severity (B = 2.43, SE = .48, p < .001), controlling for morning pain severity and previous night non-restorative sleep. Morning pain catastrophizing also predicted end-of-day activity interference (B = .08, SE = .01, p < .001), over and above afternoon pain severity and previous night’s non-restorative sleep. On days when individuals reported higher afternoon pain severity, they reported greater evening activity interference (B = .01, SE = .00, p < .001), beyond morning pain catastrophizing and previous night non-restorative sleep. Interestingly, previous night’s non-restorative sleep significantly predicted end-of-day activity interference (B = .03, SE = .01, p < .01) even when controlling for morning pain catastrophizing and afternoon pain severity. The joint significance test revealed a significant three-path mediation as predicted: Previous night non-restorative sleep → Morning pain catastrophizing → Afternoon pain severity → End-of-day activity interference.

With regard to effect sizes, previous night non-restorative sleep explained 1.6% variance of morning pain catastrophizing at the within-person level. In terms of afternoon pain severity, a total of 18% of the variance was explained by morning pain severity, morning pain catastrophizing, and previous night non-restorative sleep at the within-person level. Finally, for evening activity interference, 6.1% of the variance was explained by morning pain catastrophizing, afternoon pain severity, and previous night non-restorative sleep.

Findings of sensitivity analyses

Four multiple group analyses were conducted to test potential moderation effects of subgroups [poor sleepers vs. good sleepers; clinically depressed vs. non-depressed groups; individuals using tricyclics vs. non-users; individuals using opioids vs. non-users] on regression paths at the within-person level. Findings of chi-square difference tests showed that the difference of model fit between the freely estimated (configurable model) and the constrained models across these moderators were not statistically significant, [χ²(7) = ranged from .05 to .25, p = 1.00]. These results indicate that overall there were no subgroup differences in the intra-day process model. Thus, we did not proceed to explore whether each of the regression paths differed by these subgroups.

Discussion

In the present study we collected ecologically sensitive daily diary data several times a day from individuals with FM to investigate whether the prospective relation between previous night’s non-restorative sleep and end-of-day activity interference is sequentially mediated through morning pain catastrophizing and afternoon pain severity. As we hypothesized, a statistically significant three-path mediated effect was observed.

More than usual previous night non-restorative sleep was associated with higher morning pain catastrophizing. Previous literature suggests that poor sleep quality can enhance individuals’ threat misinterpretation^51,52,56. Thus, when individuals’ sleep is disturbed during the previous night, they may be more likely to interpret pain experience as a salient threat, and therefore, experience increased maladaptive catastrophic thoughts about pain. We also observed a direct effect from previous night’s non-restorative sleep on activity interference at the end-of-day assessment. This robust effect of non-restorative sleep is consistent with a recent meta-analysis suggesting that sleep disturbance is strongly associated with poor self-reported physical functioning³. Regularly assessing daily changes in non-restorative sleep could provide important clinical information on how individuals interpret and react to pain and function on a day-to-day level.

Morning pain catastrophizing predicted afternoon pain severity even after controlling for auto-correlation (i.e., afternoon pain severity regressed on morning pain severity). This finding replicates and further extends previous experimental studies that revealed changes in state pain catastrophizing can precede pain experience and shape individuals’ subsequent pain experience^8,9. A daily diary study of individuals with rheumatoid arthritis also reported that more than usual level of pain catastrophizing from the previous day significantly predicted next-day pain severity, while controlling for the previous day’s pain severity⁶⁴. Importantly, higher morning pain catastrophizing was also associated with greater end-of-day activity interference over and above the effect of afternoon pain severity and previous night’s non-restorative sleep. Copious findings at the between-person level indicate that pain catastrophizing is one of the most important psychosocial factors that modulate individuals’ experience of both pain and pain-related disability^15,65. Furthermore, the Fear-Avoidance Model of Chronic Pain—one of the most widely known theoretical frameworks that explains the development and exacerbation of chronic pain— posits pain catastrophizing as a core mechanism underlying pain-induced disability^11,35,73. Our within-person intra-day process findings further support previous claims made based upon between-person level findings that helping individuals effectively de-catastrophize momentarily can be an important pain management target^28,71,72.

Afternoon pain severity, controlling for morning pain catastrophizing and previous night’s non-restorative sleep predicted activity interference later in the day. Pain is understood to be a salient internal threat signal that distracts individuals from ongoing activities or goal pursuits, and draws selective attention to painful stimuli or related cues^12,48. Thus, when pain is elevated, individuals are more likely to experience higher interruption in their engagement in and performance of daily activities. Hence, it may be important to help prevent individuals from experiencing more than typical level of pain during the day. Findings of our intra-day process model indicate that jointly improving previous night’s non-restorative sleep and pain catastrophizing in the morning may potentially mitigate the effect of afternoon pain severity on individuals’ daily physical functioning.

The results of sensitivity analyses were also notable. We found that individual difference factors including overall poor sleep, clinical level of depression, and tricyclics and opioid use did not moderate the link between daily non-restorative sleep and activity interference. In other words, the present intra-day process model was quite robust in relation to these individual differences. Although other potential individual difference variables such as presence of sleep disorders and exercise level need to be tested in the future, our findings indicate the potential that the intra-day process model may be generalizable to quite a wide range of individuals with fibromyalgia.

Implications for Future Studies

Replication of our findings to other chronic pain conditions is needed. Although non-restorative sleep is one of the key diagnostic features of FM, reports of sleep disturbance are quite prevalent for many other chronic pain conditions as well^4,68,74. For instance, a recent study revealed that although reports of sleep disturbance were higher in patients with FM than in individuals with osteoarthritis, there was no significant difference in their objective sleep profiles measured by polysomnography⁸⁴. It is unknown, however, whether the relations between non-restorative sleep and daily pain catastrophizing, pain severity, and activity interference are similar between individuals with FM and those with other chronic pain conditions. Such investigations would provide important information relevant to developing intervention strategies targeted for specific chronic pain conditions.

Our findings also suggest that effective management of daily pain severity and activity interference could be enhanced by improving sleep during the previous night and reducing pain catastrophizing. Evidence-based sleep interventions such as Cognitive Behavioral Therapy for Insomnia (CBTi) have been tested multiple times among various chronic pain samples, providing some evidence that such interventions decrease pain severity, pain catastrophizing, and pain-induced activity interference^40,59,60. We suspect that integrating a brief targeted intervention that focuses on reduction of pain catastrophizing could further increase the therapeutic benefit of CBTi on pain-related outcomes. For instance, Darnall and colleagues¹³ have recently tested a brief (single session; 2–3 hour long) pain catastrophizing intervention that has shown some promise in improving pain-related outcomes. Such investigations could be a viable future research avenue.

The present study focused largely on a cognitive-affective pathway (i.e., pain catastrophizing) that links daily sleep, pain, and activity interference. A potentially fruitful focus for future research would be to evaluate potential biological mechanisms in this link. Some findings suggest that both poor sleep and pain catastrophizing can contribute aberrant glial activation and a subsequent neuroinflammatory state which is characterized by elevation in brain-derived neurotrophic factor (BDNF), interleunkin-1β (IL-1β), and tumor necrosis factor alpha (TNF-α) levels⁴⁶. In pre-clinical studies, there is some evidence that these pro-inflammatory markers can induce central sensitization^29,30, a process characterized by increased neural signaling in the central nervous system that leads to pain hypersensitivity^47,82. It has also been hypothesized that central sensitization influences the pathophysiology of fibromyalgia^1,55,62. To our knowledge, the associations among non-restorative sleep, pain catastrophizing and pro-inflammatory biomarkers at the day-to-day level have not been well investigated. Such studies would provide a better insight in understanding the development and maintenance of fibromyalgia.

Limitations

There are a number of limitations in our study that should be addressed in future research. First, morning pain catastrophizing was measured by a single diary item that does not capture all domains of pain catastrophizing. Second, our daily sleep diary only measured non-restorative sleep. Inclusion of additional sleep diary items (e.g., bed time, final awake time, out of bed time) and/or using actigraphy to measure various sleep continuity indices (e.g., wake after sleep onset, sleep onset latency, total sleep time, and sleep efficiency) should be considered in future studies. Additionally, ambulatory electroencephalographic sleep monitoring may be used to assess sleep architecture in the home environment^21,36,39, which may offer other clinically important findings. Third, we did not measure the experience of early morning pain catastrophizing and pain severity immediately after participants’ awakening. Reports of previous night’s non-restorative sleep may have been influenced by participants’ early morning pain severity and catastrophizing level. Fourth, as is quite common for daily diary studies^10,22,43, the effect sizes in the present study were small overall. Future studies need to investigate whether these small effects in the within-day or within-person context can become more clinically meaningful as they accrue over time (i.e., months or years later). Finally, a number of person-level factors that may modulate the present intra-day process model were not considered. For instance, it has been suggested that sleep disorders such as chronic insomnia, sleep apnea, and restless leg syndrome can influence the perception of sleep and are associated with higher pain sensitivity^16,42,61. Some other individual difference factors such as exercise level may have also emerged as a moderating variable of our intra-day process model. Future studies need to assess these important between-person moderating variables and evaluate whether the present model is generalizable across various individuals with FM.

Conclusion

In spite of these limitations, the present study has some notable strengths. This is one of the few studies that thoroughly examined the intra-day process model that explains the relationship between non-restorative sleep and activity interference among individuals with FM. The sample size was quite large, and participants were assessed four times per day for 21 days with overall good diary completion rates. We also used a state-of-the-art statistical approach (i.e., MSEM and three-path mediation) to evaluate our hypothesized model. Our study demonstrated that on days when individuals with FM experienced greater non-restorative sleep during the previous night, they reported greater end-of-day activity interference, mediated at least partially through higher morning pain catastrophizing and afternoon pain severity. Findings of our study emphasize the importance of evaluating daily non-restorative sleep among individuals with FM. Our intra-day process model also points out a potential need for developing accessible pain interventions that efficiently target improving daily sleep quality and attenuating pain catastrophizing to help individuals with FM more fully engage in their important daily activities despite having pain.

Highlights.

• Daily non-restorative sleep and activity interference are sequentially mediated

• Greater non-restorative sleep is associated with higher morning pain catastrophizing

• Morning pain catastrophizing significantly predicts afternoon pain severity

• Higher afternoon pain severity is related to greater end-of-day activity interference

• Non-restorative sleep and catastrophizing also uniquely predict activity interference

Perspectives:

This study provides a better understanding of how non-restorative sleep is associated with daily activity interference among individuals with fibromyalgia. An intervention that targets attenuating non-restorative sleep and pain catastrophizing may help improve daily physical functioning of this population.