Medium- and Long-Term Effects of Insomnia Severity and Circadian Preference on Pain and Emotional Distress Among Individuals With Chronic Pain.

The present study is a secondary data analysis from a parent study. The parent study aimed to investigate relationships between emotion regulation and pain-related experiences in adults with chronic pain, and the main outcome papers are previously published. 29 , 30 We would like to note that the aims of the current study do not overlap with those in the parent study. In addition, the present study also incorporates the newly collected Time 4 data (described below) that stemmed from additional funding (i.e., PM&R Seed Grant from the Department of Physical Medicine and Rehabilitation at the Johns Hopkins School of Medicine) that we received after we published the main outcome papers. Participants were recruited from Amazon Mechanical Turk (MTurk) at four time points throughout the study: Time 1 (04/22/2020 – 05/15/2020), Time 2 (07/29/2020 – 08/05/2020), Time 3 (05/07/2021 – 05/26/2021), and Time 4 (04/26/2022 – 06/02/2022). Cloud Research, a third-party data collection company, was used to recruit MTurk “workers” (they are not employees of the company Amazon) for the study. Previous studies have found that MTurk responses exhibit strong overall validity and reliability. 31 An initial pain duration screening item was used to determine preliminary study eligibility. Those who reported experiencing pain more than half of the week, every week, over the past 3 months were invited to complete an additional screening questionnaire to determine definitive eligibility. Inclusion criteria were: (a) age ≥18 years; (b) average past week pain severity of ≥3/10; (c) U.S. residence; (d) English proficiency; and (e) willingness to participate in follow-up assessments. The following data quality standards based upon Meade and Craig 32 were implemented to maximize the strength and reliability of survey responses: (1) inclusion of workers with ≥95% approval ratings from other MTurk requesters; (2) exclusion of participants who failed one or more attention check items out of three (e.g., “Please select Never True”); (3) exclusion of participants who took substantially greater (>60 minutes) or less than (<16 minutes) the average completion time of 35 minutes based upon pilot data. A total of 30,096 workers responded to the initial one-item screening question. Of this group, 10,308 (34.3%) indicated the presence of chronic pain, 2,153 met the inclusion criteria, and 1,809 (84.0%) initiated the additional screening survey. A total of 1,484 (82%) individuals met the data quality standards. Of these, we identified 31 duplicate cases which were removed. As a result, the final Time 1 sample size was 1,453. These participants were invited to participate in Time 2 (3 months), 3 (12 months), and 4 (24 months) follow-up assessments. Among 1,453 participants, 884 (61%) were retained in Time 2, 813 (56%) were retained in Time 3, and 784 (54%) were retained in Time 4. As compared to retention rates of other longitudinal MTurk studies 33 , rates for the present study were marginally higher. Note that in the present study, we focus on Time 2, 3, and 4 data, as the circadian preference measure was available from Time 2. Hence, the final sample used for the current study was N = 884 (those who completed Time 2 survey). Among 884 participants, 614 (70%) and 586 (66%) participants completed Time 3 (9-month follow-up from Time 2) and Time 4 (21-month follow-up from Time 2) surveys, respectively. In the present study, we use the term “baseline” to refer to Time 2. In sum, as the baseline of the present study starts from Time 2, the follow-up assessments were 9 (Time 3) and 21 months (Time 4) apart from Time 2. MTurk workers who met eligibility criteria were sent a link to complete self-report questionnaires. Participants were compensated $5–7 (approximately at a rate of $12/hour) for survey completion at each time point. All study procedures were approved by the Johns Hopkins School of Medicine (JHSOM) Institutional Review Board. In compliance with the JHSOM IRB policy for exempt applications, which asserts that the identity of human subjects must be well-protected, each participant was provided with a study outline, contact information for the principal investigator, and the JHSOM IRB number prior to consenting to participate. At baseline, participants reported their age, gender, race, ethnicity, education, income, marital status, and duration of chronic pain. The original items that were used to measure socio-demographic characteristics are available as online supplement material . Note that for statistical analyses, the present study focused on the gender binary (i.e., male and female), given that a very small proportion (0.5%) of individuals identified as non-binary and/or genderqueer. The Insomnia Severity Index (ISI) 34 assessed perceived severity of insomnia. ISI is a well-validated measure for insomnia and has been widely used in studies with individuals who have various types of chronic pain conditions. 4 Each of the 7 items is rated on a scale ranging from 0 ( not at all ) to 4 ( very much ). A total score with higher scores indicating greater insomnia severity was used in the present study. Total score can range from 0 to 28. Although some suggested clinical cutoffs of ISI exist, we elected to use the total score as a primary variable because (1) binary categorization of continuous variable can lead to substantial loss of statistical power to detect effects 35 ; and (2) ISI is not a gold standard measure to diagnose individuals for insomnia disorder. Cronbach’s alpha at baseline was .90. Note that ISI was measured across all time points, but given that the present study was focusing on examining the medium- and long-term effects of baseline insomnia levels, we only focused on using the baseline ISI. We would like to note that morningess and eveningness are often inaccurately referred to as ‘chronotype.’ However, chronotype reflects a behavior (i.e., sleep midpoint) rather than a preference. 36 Morningness and eveningness are an individual’s self-reported preference, and thus, using the term ‘circadian preference’ rather than chronotype is more accruate. 36 The 19-item Horne-Östberg Morningness and Eveningness Questionnaire (MEQ) 37 was used to determine individuals’ circadian preference. The MEQ is well-validated and one of the most widely used self-report circadian preference measures. 38 The MEQ features both Likert-type and time-scale items, with each section of the scale being scored from 1–5 to compute the global score. Once global scoring for the scale is determined, participants are categorized into one of the following three groups and corresponding scores: (1) morning type (>58); (2) intermediate (neither) type (scores between 42–58); and (3) evening type (<42). 39 Evening type was used as a reference group in the present study. Cronbach’s alpha at baseline was .79. The Brief Pain Inventory–Short Form 40 was used to assess pain severity and pain interference at all time points (i.e., baseline, 9 months follow-up, and 21 months follow-up). For pain severity, participants rated their current pain, as well as least, worst, and average pain in the prior 24 hours based upon a scale of 0 ( no pain ) to 10 ( pain as bad as you can imagine ). A composite of these 4 items was created by taking the average. For pain interference, participants were asked to indicate the extent to which pain interfered with general activity, mood, walking ability, work, relations with others, sleep, and enjoyment of life based up on a scale of 0 ( does not interfere ) to 10 ( completely interferes ). A composite of these 7 items was created by taking the average. Cronbach’s alphas across time ranged from .90 to .91 for pain severity, and .92 to .93. for pain interference. The 4-item PROMIS Emotional Distress-Depression and Anxiety scales 41 assessed the severity of depressive and anxiety symptoms among participants over the past week. Each item was rated on a scale of 1 ( never ) to 5 ( always ). Total scores were transformed into T-scores with a score of 50 reflecting the average for the general U.S. population. T-scores typically range from 20 to 80. These scales have demonstrated good internal reliability and validity within the chronic pain population. 41 Cronbach’s alphas across time ranged from .92 to .94 for PROMIS Emotional Distress-Depression, and .90 to .91. for PROMIS Emotional Distress-Anxiety. As mentioned above, this is a secondary data analyses of a parent longitudinal study. Hence, we computed the sensitivity to detect an effect in a multiple regression model based upon the Time 2 sample size. G*Power showed that in a multiple regression model with 10 predictors, a sample of 884 with alpha level of .05 (two-tailed) can produce a statistical power of .80 to detect even very small effects (Cohen’s f 2 = .009). Note that Cohen’s f 2 values of .02, .15, and .35 indicate small, medium, and large effect sizes, respectively. First, descriptive statistics were computed to summarize the characteristics of the current sample. Second, attrition analyses were conducted on key sociodemographic variables, as well as predictor and outcome variables at baseline. Third, associations among circadian preferences, insomnia severity, and outcomes at baseline were examined. We conducted an ANOVA to explore mean differences in insomnia severity and outcomes measured at baseline among different circadian preferences. Pearson bi-variate correlations were conducted to examine relationships between insomnia severity and outcomes measured at baseline. Fourth, using the Mplus statistical software (Version 8.3), we conducted two main path models that examined: (1) predictive effects of insomnia severity and evening type (compared to morning and intermediate types) on pain severity, pain interference, and emotional distress (i.e., depressive and anxiety symptoms) assessed at 9-month follow-up; and (2) the same outcomes measured at 21-month follow-up. Given that circadian preference is a categorical predictor, we created two dummy variables by setting the evening type as a priori reference group. For both models, select socio-demographic variables (i.e., age, gender, ethnicity, income, and education) that are known to be related to pain-related outcomes 42 – 44 were included as covariates. Note that race was not included as a covariate because our previous studies using the present data did not reveal any significant association with pain-related outcomes. 45 , 46 Both education and income variables were dichotomized to aid the interpretability of the findings. Education was coded 0 = “Less than College Education” and 1 = “College or Above.” Income was coded 0 = “$49,999 or below” and 1 = “$50,000 or above.” In addition, as recommended previously 47 , we included the baseline measure of a given outcome (e.g., pain severity at baseline) as a covariate in these models. This approach also allowed for a change score interpretation associated with the outcome. Lastly, we examined insomnia severity and circadian preference interaction effects in these two models. Note that path model is an extension of linear regression in multivariate modeling. Hence, the path estimates can be interpreted the same way as regression estimates (e.g., B = .50 indicates that one unit change of independent variable results in .50 increase in the outcome). To determine model fit of the path models, we employed several fit indices including the Comparative Fit Index (CFI; critical value ≥ .90 48 ), Standardized Root Mean Residual (SRMR; critical value, ≤ .10 49 ), and the Root Mean Squared Estimate of Approximation (RMSEA; critical value ≤ .08 50 ). Missing data were handled by the full information maximum likelihood (FIML) method, which provides unbiased parameter estimates under the missing at random (MAR) assumption (i.e., the missingness is related to some of the observed data). In terms of effect sizes, we reported η 2 (eta-squared) for ANOVA (small effect = .01, medium effect = 0.06, and large effect = 0.14) and standardized betas for path models. Lastly, a number of sensitivity analyses (i.e., excluding the sleep disturbance item from the pain interference measure, dichotomizing ISI variable based upon clinical cutoff score, and re-running analysis using total score of the MEQ) were conducted to further demonstrate the robustness of study findings.

Table 1 presents a summary of the participant characteristics. The mean age of participants was 44.2 years, and the majority of them were women (65.4%), White (87.3%), had at least some college education (87.8%), were working part- or full-time (63%), and were married (46.4%). The median income category range was $50,000-$74,999. In terms of the clinical characteristics, participants reported an overall moderate level of pain severity (M=3.8), pain interference (M=4.5), and depressive (M=56.2) and anxiety (M=58.1) symptoms at baseline. About 22 percent of the participants were categorized as evening type. We conducted a series of attrition analyses (i.e., t-tests for continuous variables and chi-square tests for categorical variables) based upon select socio-demographic variables (i.e., age, gender, race, ethnicity, income, education, and working status), baseline predictor, and outcome variables (i.e., insomnia severity, circadian preference, pain severity, pain interference, depressive symptoms, and anxiety symptoms). Results showed that only age was significantly associated with missingness at 21-month follow-up ( p < .001), such that participants of a younger age were more likely to drop out at 21-month follow-up. These findings lend further support that the missingness in the present study is likely to meet the MAR assumption. Hence, use of FIML to handle the missing data is justified. Table 2 shows the results of the ANOVA employed to examine mean differences in baseline insomnia severity and baseline outcome levels across the three circadian preferences (i.e., Evening, Intermediate, and Morning types). Insomnia severity ( p < .001; η 2 = .04), pain interference ( p = .043; η 2 = .01), depressive symptoms ( p < .001; η 2 = .05), and anxiety symptoms ( p < .001; η 2 = .03) significantly differed by circadian preferences. The Evening type had the highest level of insomnia severity, pain interference, and depressive and anxiety symptoms. There were no significant circadian preference differences in the levels of pain severity at baseline. In terms of bi-variate correlations, results showed that insomnia severity was moderately and positively correlated with all outcome variables at baseline: (1) r = .32 with pain severity; (2) r = .45 with pain interference; (3) r = .47 with depressive symptoms; and (4) r = .47 with anxiety symptoms. The model fit of this path model was good overall, RMSEA = .07, CFI = .98, and SRMR = .03. Table 3 shows the detailed parameter estimates of this model. Even after controlling for baseline levels of the outcomes and socio-demographic covariates, insomnia severity at baseline was significantly associated with changes in pain severity ( B = .03, SE = .01, p = .003), pain interference ( B = .05, SE = .01, p < .001), depressive symptoms ( B = .18, SE = .04, p < .001), and anxiety symptoms ( B = .21, SE = .05, p < .001) at 9-month follow-up. In terms of interpretability of the path estimates, for every one unit increase of insomnia severity (from a total score that ranges 0–28), pain severity increased by .03 (from a 0–10 scale), pain interference increased by .05 (from a 0–10 scale), depressive symptoms increased by .18 (from T-score that typically ranges from 20 to 80), and anxiety symptoms increased by .21 (from T-score that typically ranges from 20 to 80). Individuals with greater levels of insomnia severity at baseline reported worsening pain-related symptoms at 9-month follow-up. Dummy variables that compared Evening vs. Intermediate and Evening vs. Morning types demonstrated no significant mean differences in the change of pain-related outcomes at 9-month follow-up. The model (RMSEA = .07, CFI = .98, and SRMR = .03) that examined insomnia severity and circadian preference moderation revealed that there were no significant interactions ( p -values ranging from .07 to .95) across all four outcomes. The model fit of this long-term outcome model was also good overall, RMSEA = .06, CFI = .98, and SRMR = .03. Table 4 shows the detailed parameter estimates of this model. Even controlling for baseline levels of the outcomes and socio-demographic covariates, insomnia severity at baseline was significantly associated with changes in pain interference ( B = .05, SE = .01, p < .001), depressive symptoms ( B = .17, SE = .05, p = .001), and anxiety symptoms ( B = .23, SE = .05, p < .001), but not pain severity, at 21-month follow-up. Dummy variables that compared Evening vs. Intermediate and Evening vs. Morning types revealed no significant mean differences in the change of pain-related outcomes at 21-month follow-up, apart from a significant difference between Evening and Intermediate types in terms of depressive symptoms ( B = −2.4, SE = .74, p = .001). Individuals who were Evening types showed worsening depressive symptoms than those who were Intermediate types. The model (RMSEA = .06, CFI = .98, and SRMR = .03) that examined insomnia severity and circadian preference moderation demonstrated that there were no significant interactions ( p -values ranging from .23 to .96) across all four outcomes. The BPI Pain Interference subscale includes an item relating to pain’s interference with sleep. It is possible that the robust longitudinal association between insomnia severity and pain interference is mainly driven by this sleep interference item in the BPI. Hence, we created a composite of the pain interference variable, excluding this sleep interference item, and re-analyzed the main path models. The sensitivity analysis revealed that even after excluding the sleep interference item from the pain interference composite, insomnia severity at baseline significantly predicted both 9-month ( B = .04, SE = .01, p = .001) and 21-month ( B = .05, SE = .01, p = .001) follow-up pain interference, while controlling for covariates that were mentioned above. This demonstrates that the sleep interference item did not drive the significant association between insomnia severity and pain interference. We elected to use total ISI scores in the present study due to a methodological advantage (i.e., better statistical power than when dichotomizing a continuous variable) and potential false diagnosis of insomnia disorder using a clinical cutoff for ISI. However, some may still be interested in whether clinical level of insomnia (rather than insomnia severity) predicts pain-related outcomes longitudinally. Hence, we conducted an additional sensitivity analysis by creating a binary ISI variable (i.e., clinical-level insomnia vs. not clinical-level insomnia) using ISI cutoff score of 14. 51 , 52 Based upon this cutoff score, we found that 41% of our participants are categorized as having clinical level insomnia. We re-analyzed all of the main models switching the continuous ISI variable to this binary ISI variable. Results showed that all of the findings were replicated using the dichotomized ISI variable. The summary of these sensitivity analyses is available as online supplement Table S1 and S2 . It is possible that one of the main reasons that we could not observe the effects of eveningness on pain-related outcomes is due to reduction in statistical power that is caused by categorization of a continuous variable (i.e., the MEQ). We re-analyzed models by including a total MEQ score instead of dummy MEQ variables. However, consistent with the dummy variables, the total score of MEQ did not significantly predict both 9- and 21-month pain-related outcomes, while controlling for all other covariates including the insomnia severity.

Growing evidence suggests that insomnia and evening type are associated with both pain-specific and emotional health-related outcomes. However, there has been a dearth of studies that consider both insomnia and circadian preference in the context of chronic pain maintenance and progression. The present study is one of the first to investigate prospective effects of both insomnia severity and evening type on pain-related outcomes using a large dataset that followed individuals with chronic pain longitudinally. Contrary to our expectations, only insomnia severity was significantly associated with changes in both medium- and long-term pain-related outcomes. In addition, we found no significant interaction effects between insomnia severity and circadian preference on pain-related outcomes. Consistent with previous observational and well-controlled laboratory-based sleep disruption studies 1 , 5 , 6 , our findings demonstrate that insomnia may serve an important role in the progression of chronic pain. Greater insomnia severity at baseline was associated with worsening pain interference and emotional distress over nearly two years. However, unlike pain interference, we found that insomnia severity was not significantly associated with long-term (21 months) changes in pain severity. It is unlikely that insomnia severity is more strongly associated with pain interference because of the inclusion of the sleep interference item in the BPI. As we demonstrated from our sensitivity analysis above, even after excluding this item from pain interference composite, we found significant predictive effects of insomnia severity on pain interference at both 9 and 21 months. In fact, a recent meta-analysis of CBT-I on chronic pain showed that the effect size of CBT-I on pain severity (standardized mean difference = 0.19) is much smaller than that on pain interference (standardized mean difference = 0.75). 7 Also, at least two previous studies revealed that the CBT-I improved only pain interference, not pain severity. 53 , 54 We speculate that the negative impact of insomnia on chronic pain is more likely manifested by difficulties in ‘coping’ with pain (e.g., increase in pain catastrophizing 55 , 56 and negative affect 57 ), as opposed to a significant elevation in pain severity, thus resulting in greater pain interference and emotional distress. Future studies that investigate the underlying biobehavioral mechanisms linking insomnia and the progression of chronic pain are strongly warranted. Similar to previous cross-sectional studies 9 , 12 – 14 , 58 , we found that evening types showed greater pain interference and emotional distress at baseline compared to intermediate and morning types. However, we found that evening types were not at a higher risk of experiencing worsening pain-related outcomes, while controlling for the effects of baseline insomnia severity, than intermediate and morning types. Furthermore, there were no significant moderating effects of insomnia severity and circadian preference on pain-related outcomes, such that evening types with greater insomnia severity did not show worsening pain-related outcomes compared to other types with lower insomnia severity. Overall, our findings suggest no clinically meaningful associations between evening types and an exacerbation of pain-related outcomes over time, and levels of insomnia severity do not modulate its effects on the outcomes. Yet, our results should be replicated and further extended in future studies with use of more sophisticated biobehavioral circadian markers and by employing a well-controlled laboratory design. Assessment of objective circadian markers via actigraphy, dim light melatonin onset (DLMO), and/or 6-sulfatoxymelatonin (aMT6s) in urine that provides reliable measures of endogenous circadian rhythms 59 , 60 , may help us better understand how circadian misalignment or blunted circadian rhythms may be associated with maintenance and exacerbation of chronic pain. 61 Notably, a number of animal studies have consistently demonstrated that experimental disruption of circadian rhythms (e.g., exposure to dim light at night, mistimed feeding, and simulated jet lag) cause amplification of pain sensitivity. 62 – 64 Human laboratory-based chronobiological studies, such as employing inverted sleep schedules, simulated night shifts, or forced desynchrony protocols 65 , may also shed further light on understanding the effects of circadian misalignment on pain-related outcomes. 8 A few clinical and research implications can also be gleaned from the findings of the current study. First, our results suggest that untreated insomnia can facilitate maintenance and progression of chronic pain over a long period of time. Hence, symptoms of insomnia should be routinely examined in clinical settings when working with individuals with chronic pain. Second, as shown in a recent meta-analysis 7 , CBT-I is efficacious in improving not only sleep, but also pain-related outcomes, particularly pain interference. However, consistent with findings of the present study that showed overall significant but small magnitude of effects of insomnia on pain-related outcomes, the meta-analysis suggests that the overall effect size of CBT-I on pain-related outcomes are small to moderate at most. 7 Hence, a sensible future goal may be to examine whether a combination of CBT-I and another evidence-based intervention for chronic pain (e.g., Mindfulness-Based Stress Reduction, Acceptance Commitment Therapy) can produce a more powerful and enduring treatment effects for chronic pain. The present study had a number of strengths, including the use of a large sample, long-term (21 months) follow-up, and integration of both insomnia and circadian preferences in the same model. However, there were also several limitations. First, circadian preference was not assessed at Time 1, and thus the final sample was limited to those who completed Time 2 assessments. There may have also been a sample selection bias, potentially limiting the generalizability of our findings. Second, the data of our study are based upon an online crowdsourcing (i.e., MTurk) sample and the sample did not represent minoritized individuals well, especially those who identify as Black/African American and Hispanic/Latino/a/x. Hence, it may be difficult to generalize findings of our study to the chronic pain population in the U.S. in general and minoritized individuals in particular. Future studies should incorporate a better sampling strategy, such as using random digit dialing to collect data from a nationally representative sample or oversampling minoritized racial and ethnic groups. Third, although we assessed both insomnia and circadian preference using well-validated self-report measures, we did not collect any objective markers of sleep or insomnia diagnoses evaluated by a structured clinical interview. Also, we did not assess social jet lag (i.e., discrepancy between biological and social clocks) 66 , which may have shown more powerful association with pain-related outcomes. These data could have provided us with greater insight into our current findings. Lastly, we did not assess whether participants received any treatment for chronic pain before and/or during the observational period. The present study suggest that the severity of insomnia is relatively more important in predicting changes in pain-related outcomes in both medium- and long-term when compared to eveningness. Greater insomnia severity is robustly associated with not only worsening pain severity and interference, but also emotional distress. Therefore, insomnia remains an important clinical target for chronic pain management. However, to better understand potentially unique roles of sleep and circadian rhythm disturbances in chronic pain maintenance and progression, future studies should assess sleep and circadian rhythm using more sophisticated biobehavioral markers. In addition, conducting human laboratory studies may shed further light on the effects of circadian misalignment on chronic pain.

Insomnia is an important modifiable risk factor for the development and progression of chronic pain. 1 – 3 A recent meta-analysis revealed that the overall prevalence of clinically meaningful sleep disturbances ranged from 73–75% among individuals with chronic pain. 4 Although sleep and pain are bi-directionally related, evidence from longitudinal studies suggests that the effect of sleep disturbances on pain is stronger than that of pain on sleep disturbances. 1 In fact, a number of laboratory studies of healthy adults also provide causal evidence that sleep deprivation and fragmentation that mimic severe insomnia contribute to greater pain sensitivity. 1 , 5 , 6 Relatedly, a recent meta-analysis of clinical trials of cognitive behavioral therapy for insomnia (CBT-I) also suggests that CBT-I improves not only sleep, but also chronic pain. 7 Emerging evidence suggests that eveningness or evening type (i.e., circadian preference for staying up late and waking up late), may also serve an important role in chronic pain development and progression, independent from sleep disturbances. 8 Heikkala and colleagues 9 found that even after controlling for sleep duration, evening types were more likely to report disabling pain among 4,961 individuals with musculoskeletal pain residing in Finland. The same research group published a 15-year follow-up study using the same cohort data suggesting that evening types, compared to morning types, were more likely to show persistent multisite musculoskeletal pain between ages 31 and 46 years. 10 Furthermore, Heikkala and colleagues also found that evening types report a higher reduction in health-related quality of life when experiencing worsening musculoskeletal pain, compared to morning types. 11 Importantly, studies on fibromyalgia and irritable bowel syndrome have shown that greater eveningness is associated with more severe symptom profiles, disease complications, and a lower quality of life. 12 – 14 Emotional distress, which refers to anxiety and depressive symptoms, is considered an important pain-related outcome 15 , 16 , as pain can impact one’s emotional well-being. 16 , 17 Although studies that longitudinally examine the effects of insomnia and circadian preference on emotional distress among individuals with chronic pain are scarce 8 , the deleterious effects of both insomnia and eveningness on emotional distress, and vice versa, in the general population are well-established. Overall, insomnia and eveningness are associated with development and maintenance of various mental health conditions, including mood and anxiety disorders. 18 – 23 Despite the fact that mechanisms underlying insomnia and eveningness are closely intertwined (e.g., evening types are more likely to experience misalignment between the internal biological rhythm and the societal rhythm than morning or intermediate types, which can cause significant sleep loss and insomnia) 24 , 25 and each may play a unique role in pain and emotional distress, their co-evaluation is scant in the literature. Also, to our knowledge, none of the previous studies have examined the effects of both insomnia severity and evening type (compared to morning and intermediate types) on pain-related outcomes (i.e., pain severity, pain interference, and emotional distress) 15 , 26 prospectively among individuals with chronic pain. The present study tries to address these important gaps in the literature with use of a large dataset that longitudinally examined the effects of insomnia and eveningness on pain-related outcomes over 21 months. We hypothesized that individuals with greater insomnia severity and eveningness will exhibit worsened pain-related outcomes both in medium- (at 9-months) and long-term (at 21-months) follow-ups (see Figure 1 ). Although some studies suggested that evening-type insomniacs present greater levels of depression, lower positive affect, and blunted neural function in brain regions associated with positive affect compared to morning- or intermediate-types 27 , 28 , other studies showed no significant moderating effects of sleep disturbance in the association between eveningness and musculoskeletal pain. 9 , 10 Given these mixed previous findings, we also explored the effects of an interaction between eveningness and insomnia severity on outcomes.