bs_bs_banner

Pain Medicine 2014; 15: 1043–1051 Wiley Periodicals, Inc.

REHABILITATION SECTION Original Research Article Variability in the Relationship Between Sleep and Pain in Patients Undergoing Interdisciplinary Rehabilitation for Chronic Pain Sara Davin, PsyD, MPH, Josh Wilt, MS, Edward Covington, MD, and Judith Scheman, PhD Neurological Center for Pain, Neurological Institute, Cleveland Clinic, Cleveland, Ohio, USA Reprint requests to: Sara Davin, PsyD, MPH, Neurological Center for Pain, Neurological Institute, Cleveland Clinic, 9500 Euclid Avenue, C21, Cleveland, OH 44195, USA. Tel: 216-445-3977; Fax: 216-445-7000; E-mail: [email protected]. Disclosure/Conflict of Interest Statement: There are no conflicts of interest or related disclosures from any of the authors on this manuscript.

Abstract Objective. Chronic pain and sleep disturbance frequently coexist and often complicate the course of treatment. Despite the well-established comorbidity, there are no studies that have investigated concurrent changes in sleep and pain among patients participating in an interdisciplinary chronic pain rehabilitation program (ICPRP). The goal of this study was to investigate the daily changes in sleep and pain among patients participating in an ICPRP. Methods. Multilevel modeling techniques were used to evaluate the daily changes in total sleep time (TST) and pain among a sample of 50 patients with chronic noncancer pain participating in the ICPRP. Results. Increases in TST were predictive of less pain the following treatment day, although daily pain ratings were not predictive of that night’s TST. Time in treatment was a significant predictor of both TST and pain reduction, even while controlling for age, gender, anxiety, and depression. Additional analyses revealed significant individual variability in the relationship between TST and next day pain. Individuals with stronger associations between previ-

ous night’s TST and next day pain were found to experience the greatest treatment benefits overall, in terms of pain reduction and TST. Conclusions. Our results provide compelling support for individual variability of the pain–sleep relationship in patients with intractable pain conditions participating in an ICPRP. Importantly, these findings suggest that when pain and sleep are comorbid, both must be addressed to reap the maximum response to treatment programs such as an ICPRP. Perspective Statement. This study demonstrates the utility of treating sleep problems in patients participating in an interdisciplinary chronic pain rehabilitation program. Results highlight the benefits of accounting for individual variability in the pain-sleep relationship in a clinical setting and targeting sleep interventions for those individuals whose pain and sleep problems are comorbid. Key Words. Chronic Pain; Sleep; Interdisciplinary Rehabilitation; Treatment

Introduction Sleep difficulties are common among chronic pain patients with 50–80% experiencing some sort of sleep disturbance [1,2]. Numerous studies have linked impaired sleep to subsequent increases in pain sensitivity and lower pain tolerance [3,4]. While some studies seem to suggest that the relationship between pain and sleep is bidirectional, the evidence that increased pain interferes with sleep is less convincing [5,6]. Comorbid pain and sleep disturbances complicate the course of treatment of chronic noncancer pain (CNCP) [7]. In spite of the extensive support for the pain–sleep connection, there are few well-established treatment interventions that substantially improve both pain and sleep among CNCP, when it is comorbid. The studies that have explored the effect of nonpharmacological treatments on 1043

Davin et al. pain and sleep are limited to those using cognitive behavioral therapy (CBT) for insomnia (CBT-I) and those that use CBT for pain (CBT-P) while measuring changes in sleep [5,8,9]. A summary of findings from these studies generally shows that CBT-I improves insomnia, with minimal pain reduction, and CBT-P decreases pain, but not insomnia. There are few treatment options that address both chronic pain and insomnia concurrently. One study included a brief (one session) intervention on catastrophic thinking, but it failed to specifically address pain catastrophizing, a unique construct measuring thinking and rumination related specifically to pain [10]. In one recent pilot study of combined CBT-I/CBT-P, researchers found that the combined approach offered a greater benefit in terms of insomnia reduction, over solo CBT-P or CBT-I, but was unrelated to greater improvements in pain [11]. Scientific evidence for the best combination of pharmacological agents to treat comorbid pain and insomnia is also lacking. While benzodiazepines are one of the most commonly prescribed medications, their benefit is transient at best, and the associated risks from use are unclear [12]. Additionally, many commonly prescribed analgesics, such as opiates and tricyclic antidepressants, interfere with rapid eye movement sleep [13]. In other words, medications used to treat pain may contribute to sleep disturbances. This is a particularly important consideration for clinicians who treat individuals with comorbid chronic pain and sleep problems. Interdisciplinary chronic pain rehabilitation programs (ICPRPs) have consistently received support from studies demonstrating their efficacy in the treatment of chronic pain [14]. ICPRPs typically emphasize restoration of function and reduction of suffering in spite of pain and utilize a variety of approaches including education, medical management, physical therapy/occupational therapy, and psychological interventions. To our knowledge, there are no studies that look at the changes in sleep and pain among individuals participating in an ICPRP. ICPRPs offer a unique setting to evaluate the pain–sleep connection, as patients receive multimodal, intensive daily intervention, and changes in pain and sleep are evaluated on a daily basis. The present study aims to explore the effects of treatment in an ICPRP on daily changes in pain and total sleep time (TST). While it is implausible to discern what specific component of an ICPRP may be attributed to changes in sleep and pain (as such programs inherently provide a variety of interventions at one time), the present study aims to address the lack of knowledge if such a treatment approach is a valuable option. A unique component to our study is the ability to monitor daily changes in sleep, pain, and mood over the course of 3–4 weeks. Methods Participants and Treatment Setting Participants were 50 consecutive patients (36 female) ages 20–80 (M = 45.96, SD = 13.94) obtained from a convenience sample of individuals with chronic noncancer 1044

pain who were admitted to the Chronic Pain Rehabilitation Program (CPRP) at Cleveland Clinic Foundation. The average duration of pain from the sample of patients was 14.6 years (SD = 10.74). Participants demonstrated a high rate of pain and psychiatric comorbidities, with the mean number of pain related and psychiatric diagnoses being 6.8 (SD = 2.17). The most common presenting pain complaints were low back pain (36%), total body pain/ fibromyalgia (28%), neck pain (8%), and knee pain (8%). The majority of the participants (58%) were not working due to pain. The CPRP, within the Neurological Center for Pain, Neurologic Institute at the Cleveland Clinic, is a comprehensive, interdisciplinary program designed to treat patients with disabling chronic pain. Patients attend the CPRP Monday through Friday from 7:30 AM to 5:00 PM. The CPRP is an outpatient treatment program that combines physical therapy, occupational therapy, psychodynamic group psychotherapy, individual psychotherapy incorporating relaxation training and biofeedback, substance use education, CBT, vocational counseling, family education, and medication management. All psychological interventions are delivered by counselors, social workers, or psychologists trained in diverse psychotherapy approaches including CBT and psychodynamic therapy. Medication management occurs daily and includes weaning from opiate analgesics, sedatives, and benzodiazepines during the CPRP. The program’s goals emphasize functional restoration and self-management of pain, depression, anxiety, and associated disability. All participants in this study completed the entirety of the CPRP, which ranges from 3 to 4 weeks. The average length treatment among the 50 patients in this study was 15.02 treatment days (SD = 3.43) (max = 22; min = 8). Exclusion criteria for the CPRP are active psychosis, untreated comorbid medical conditions, active illicit drug use or alcohol dependence, or patients who are of imminent suicidal risk. This study was approved by the Cleveland Clinic Institutional Review Board. Procedure and Measures Daily Measures of Pain, Mood, and Sleep At the beginning of each treatment day, patients estimated their TST from the previous night and their current pain intensity, depression, and anxiety. TST was reported as total hours slept. Depression and anxiety were rated on separate 11-point Likert scales ranging from 0 to 10, with 0 representing no mood disturbance and 10 representing severe mood disturbance. Daily ratings of pain intensity were obtained using the Numerical Rating Scale (NRS), a 0–10 Likert scale, with 0 representing no pain and 10 representing the worst pain imaginable. Admission and Discharge Measures of Pain, Mood, and Function Patients completed measures of pain, depression, anxiety and pain related function upon admission and discharge

Sleep and Pain to the CPRP. Admission and discharge pain severity was obtained using the NRS. Depression and anxiety were assessed using the Depression, Anxiety and Stress Scale [15]. Pain-related functional impairment was measured using the Pain Disability Index [16]. Data Analysis We employed multilevel modeling (MLM) techniques using maximum likelihood estimation as our primary analytic strategy. MLMs are preferred over traditional regression methods because the data in this study are assessed at multiple levels, with daily reports (level 1) nested within individuals (level 2). Avoiding the use of conglomerated group data eliminates the inference that group data is descriptive of the individual [17]. MLM obtains the associations between the independent variables and dependent variables for each participant and then pools those results across participants to find the association for the typical individual. The association between variables is estimated by unstandardized b coefficients, which are partial regression coefficients that quantify the magnitude and direction of association in changes in the dependent variable with changes in the independent variable. In one set of MLMs, our primary dependent variable was current pain, and our primary independent variable was previous night’s sleep. Pain was entered as the dependent variable, and previous night’s sleep (mean centered around each participant’s mean) was entered as the predictor variable. Follow-up analyses controlled for gender and age, as well as current depression and anxiety. In order to examine the reciprocal association between pain and TST, a second set of MLMs following the same structure described above treated same night’s sleep as the dependent variable and pain as the independent variable. In each MLM, we obtained the average association across individuals as well as the b coefficients describing each individual’s association between TST and pain. MLMs were also used to estimate changes in TST and pain over the course of treatment. In one set of MLMs, TST was

used as the dependent variable and in another set pain was used as the dependent variable. The primary independent variable in each analysis was time, measured as days in treatment. Again, follow-up analyses controlled for age, gender, depression, and anxiety. We also obtained each individual’s average change in pain per day and average change in TST over the course of treatment. For example, a b of 0.10 relating days in treatment to TST would indicate that, on average, that particular individual’s TST improved by 0.10 hours per day in treatment. We then calculated the correlations between patients’ average change in pain per day and average change in TST to determine whether individuals who get remission from pain also get remission from sleep difficulties. Multiple regression analyses were conducted to clarify the associations between average change in pain per day and average change in TST. These analyses examined whether patients’ within-person associations between pain and TST moderated the association between average change in pain per day and average change in TST. Descriptive statistics, correlations, and regressions were computed in the statistical program R using the psych package. MLM analyses were done in R using the multilevel and nlme packages [18]. Results Descriptive Statistics Table 1 shows descriptive statistics of the daily reports of TST, pain, anxiety, and depression. The median score of daily sleep was 5.64. The SDs of the aggregate mean estimates shows that there was greater between-person variability in depression and anxiety than in pain and greater variability in pain than in sleep. Within-person SDs showed that patients varied around their own means over the course of the study; it appears that TST, pain, and depression varied more than did anxiety. The intraclass correlation (ICC1) values showed that a substantial percentage of variability in pain (49%), depression (58%), and anxiety (60%) was due to between-person variation, whereas most of the variability in TST was attributable to within-person variation (100% − 24% = 76%). High ICC2 values indicated that the participants’ aggregate means

Table 1 Descriptive statistics of the daily reports of previous night’s sleep, pain, anxiety, and depression (SD = standard deviation; ICC = intraclass correlation)

Sleep Pain Depression Anxiety

Aggregate Mean

SD of Aggregate Mean

WithinPerson SD

ICC1

ICC2

5.67 5.28 3.21 4.21

0.90 1.64 2.18 2.29

1.32 1.46 1.51 0.76

0.26 0.49 0.58 0.60

0.83 0.93 0.95 0.95

1045

Davin et al.

Table 2 Admission and discharge pain, mood, and function Pain

Admission Discharge

Depression

Anxiety

Function

Mean

Median

SD

Mean

Median

SD

Mean

Median

SD

Mean

Median

SD

4.46 1.54

7.0 3.0

15.07 14.69

18.06 6.86

23.0 6.0

20.8 18.1

9.58 6.18

9.0 6.0

18.95 17.38

39.84 16.81

40.0 16.0

23.13 20.96

Key: SD = Standard deviation. Depression Score: Normal = 0–9, Mild = 10–13, Moderate = 14–20, Severe = 21–27, Extremely Severe = 28+. Anxiety Score: Normal = 0–7, Mild = 8–9, Moderate = 10–14, Severe = 15–19, Extremely Severe = 20+. Function as measured by the Pain Disability Index: Disability due to pain rating: 0–10 = None to Mild, 10–30 = Mild to Moderate, 30–50 = Moderate to Severe, 50–70 = Severe.

of each variable could be reliably differentiated from each other. Table 2 shows descriptive statistics (mean, median, and SD) of admission pain, depression, anxiety, and painrelated function. Upon admission, participants demonstrated moderate levels of pain, moderate depression, mild-moderate anxiety and moderate-severe pain related disability. Figure 1 further displays these findings. Daily Associations Between TST and Pain Results from this MLM predicting next day pain from TST showed a negative association (b = −0.27, P < 0.001), such that increases in TST were predictive of less pain the following day. That is, for the typical individual, an increase in TST of 1 hour the previous night predicted a decrease of 0.27 points of pain the following day. Results from models controlling for gender, age, depression, and anxiety showed that previous night’s TST maintained its negative association with next day pain (b = −0.14, P < 0.001). Current depression (b = 0.10, P < 0.01) and anxiety (b = 0.29, P < 0.001) were positively associated with pain, whereas age (b = 0.02, P = 0.25) and gender (b = 0.16, P = 0.76) were unrelated to pain. Daily Associations Between Depression, Anxiety, TST, and Pain Current depression (b = 0.06, P = 0.21) was not associated with next day pain but was positively associated with

that night’s TST (b = 0.10, P = 0.02). Current anxiety was significantly associated with next day pain (b = 0.20, P < 0.001), and weakly associated with the same night’s TST (b = −0.08, P = 0.06). These results tell us that higher daily anxiety ratings predict increased next day pain and that daily depression ratings predict increases in the subsequent night’s TST. In order to examine the reciprocal association between pain and TST, a second set of MLMs following the same structure described above treated TST as the dependent variable and pain as the independent variable. The initial model showed that increased pain during the day was not related to decreased that night’s TST (b = −0.08, P < 0.10). Additionally, there was no effect of pain on same night’s TST in the model controlling for gender, age, depression, and anxiety. Combining these results with the results from the first set of MLMs, yields the conclusion that TST is a robust predictor of next day’s pain, but that same day pain does not predict that night’s TST. Do the Associations Between TST and Pain Vary Across Individuals? Although the results reported above indicate that the typical individual experiences reduced pain following a night of increased TST, it is possible that this association is stronger for some patients than for others. To examine this possibility, we compared the MLM that allowed the association between pain and TST to vary across individuals to a model that did not allow this association to vary across

Figure 1 Mean pain, mood, and function at admission and discharge. 1046

Sleep and Pain individuals. Results showed that the model allowing the association to vary across individuals fit the data better than the fixed model (l.ratio = 16.38, P < 0.001), indicating that the effect of previous night’s TST on next day pain varied across individuals. We therefore extracted each individual’s b coefficient relating TST to next day pain and found that they varied from −0.79 to 0.16 (M = −0.27, SD = 0.21). These results showed that some patients had very strong negative associations between TST sleep and next day pain, whereas others had weaker negative associations, and a few even had positive associations between TST and pain. Thus, previous night’s TST sleep does not associate with next day pain uniformly across all individuals. We also examined whether the association between same day pain and that night’s TST varied across individuals. However, the model allowing the association to vary across individuals did not fit the data better than the fixed model (l.ratio = 0.48, P = 0.79), indicating that the null association reported above held for all individuals.

Changes in TST, Mood, and Pain over the Course of Treatment One set of MLMs assessed whether TST was predicted by time in treatment, and another set of MLMs assessed whether pain was predicted from time in treatment. Results showed that time in treatment positively predicted TST (b = 0.05, P < 0.001) and was associated with pain reduction (b = −0.14, P < 0.001). Time in treatment also remained a significant predictor of both TST and pain reduction while controlling for age, gender, anxiety, and depression (for sleep, b = 0.04 P < 0.01; for pain, b = −0.10, P < 0.001). Thus, the typical patient experienced increases in TST and decreases in pain over the course of treatment. Time in treatment was also related to decreases in depression (b = −0.10, P < 0.001) and anxiety (b = −0.10, P < 0.001). The 95% confidence intervals relating time in treatment to pain, TST, depression, and anxiety were: pain (−0.18, −0.09), sleep (0.05, 0.11), depression (−0.15, −0.04), and anxiety (−0.15, −0.05). We next examined whether average changes in TST and pain per day over the course of treatment varied across individuals. The MLMs allowing the effect of time on TST and time on pain to vary across individuals fit the data better than each of the respective models that constrained the effects of time to be identical across individuals (for TST, l.ratio = 14.85, P < 0.001; for pain, l.ratio = 71.05, P < 0.001), indicating that average changes in TST and pain per day during treatment varied across individuals. Patients’ b coefficients relating treatment duration to TST ranged from −0.08 to 0.15 (M = 0.05, SD = 0.05), and b coefficients relating time to pain ranged from −0.49 to 0.13 (M = −0.14, SD = 0.13). In summary, most patients experienced increased TST and decreased pain intensity over the course of treatment, yet the magnitude of these changes varied across individuals.

Association Between Changes in Sleep and Pain over the Course of Treatment Do individuals who experience greater increases in TST also experience greater remission from pain over the course of treatment? We examined this question by correlating individuals’ b coefficients relating time and TST with their b coefficients relating time and pain. The correlation (r = −0.13, P = 0.37) suggested that individuals’ average changes in TST and pain per day were unrelated to one another. Although most individuals experienced increases in TST and decreases in pain, those who experienced greater increases in TST did not necessarily experience greater remission from pain. Given the earlier findings that 1) previous night’s TST associated with next day pain and 2) both TST and pain improved over the course of treatment, we hypothesized that the association between average changes in TST and daily pain ratings may be moderated by individuals’ withinperson associations between previous night’s TST and next day pain. Specifically, we reasoned that individuals with stronger negative associations between previous night’s TST and next day pain should experience greater correspondence between their improvements in TST and pain over the course of treatment. The rationale for this hypothesis was straightforward. At the level of the individual, a person whose pain is tied to his/her TST should experience a greater overall reduction in pain (compared with a person whose pain is not tied to their TST) given the same amount of sleep improvement. That is, one would intuitively expect that increasing TST would help pain only in those individuals whose pain was exacerbated by insomnia. We tested this hypothesis by conducting a simultaneous multiple regression predicting 1) individuals’ average change in pain per day from average change in TST per day, 2) individuals’ within-person associations between previous night’s TST and next day pain, and 3) the interaction term of the two predictors. Results showed a positive and significant interaction term (b = 4.75, P < 0.001), indicating that the association between average changes in TST and pain per day was moderated by within-person associations between previous night’s TST and next day pain. This effect can be understood by examining the interaction plot in Figure 2. The plot shows the predicted associations between average changes in TST and pain per day for an individual with a weak association between previous night’s TST and next day pain (+1 SD, b = −0.06), as well as for an individual with a strong negative association (−1 SD, b = −0.48). For an individual with a more negative association (i.e., previous night’s TST was closely tied to pain the following day), greater increases in TST per day were predicted to associate with greater decreases in pain per day. In contrast, for an individual who had a relatively weak (i.e., close to 0) association, increases in TST per day are predicted to relate to slight increases in pain per day over the course of treatment. 1047

0.1 0. 16

0. 12

0. 08

0 0. 04

.0 4

-0

-0.1

.0 8

0 -0

Treatment Effect for Pain

Davin et al.

-0.2 +1 SD Within-Person Association between Pain and Sleep

-0.3 -0.4

-1 SD Within-Person Association between Pain and Sleep

-0.5

Treatment Effect for Sleep

Discussion We examined 1) the daily changes of pain and estimated TST among patients participating in an ICPRP and 2) the reciprocality of the pain–sleep relationship through analysis of individual ratings of pain, mood, and TST across a ∼15-day treatment period. We found that both pain and TST are responsive to treatment in an ICPRP, with the majority of patients experiencing increased TST and decreased pain intensity over the course of treatment. This is an important finding, given the lack of previously noted improvements in pain, from interventions that specifically target sleep only (CBT-I) and lack of improvement in insomnia from interventions that target pain only (CBT-P), as well as equivocal evidence for the pharmacological treatment of comorbid pain and insomnia. Additionally, time in treatment predicted increased improvement for pain and TST. Greater duration of treatment yielded greater increases in TST and decreases in pain. This finding remained true even when accounting for mood, gender, and age. When exploring the day-to-day sleep and pain associations, we found that a night of poor sleep (defined as less TST) predicted increased pain the following day, even when controlling for age, gender, depression, and anxiety. However, daily pain ratings were not predictive of subsequent TST. Our study also explored individual variations in the associations between TST and pain. While it is valuable to know that increased TST predicts lowered next day pain, in a clinical context, it is important to know how much this relationship varies from individual to individual. Our results showed that in some patients a night of less TST was very strongly associated with heightened next day pain, whereas others had weaker associations. On the other hand, we found a lack of individual variability in association between day of pain and that night’s TST. In other words, for the majority of individuals, there was minimal association between daily pain rating and that night’s TST. These findings are noteworthy, as they underscore the importance of accounting for individual differences in the 1048

Figure 2 Treatment effect for pain as a function of treatment effect for total sleep time and within-person association between pain and total sleep time.

treatment of sleep and pain and directing treatment based on such. Should the magnitude of the association between previous night’s TST and next day pain garner clinicians’ attention? Certain individuals may have little room to improve in sleep (e.g., an individual who is already sleeping 6 hours of night), also raising the possibility of a “ceiling effect,” with regards to pain improvements, when treating sleep in these individuals. However, our findings suggest a steady improvement in TST would lead to meaningful reductions in pain for an individual with a larger discrepancy between current sleep to needed total sleep, provided that they do have a within-person association between TST and pain. Indeed, this is where the finding that people who have a within-person association between previous night’s TST and next day pain also show correlated improvements in TST and pain over the duration of treatment is of relevance. If TST and pain are correlated within the individual, then improvements in TST over the course of treatment will relate to reductions in pain over the course of treatment. This is of marked clinical utility. Additionally, one must consider previous research that demonstrates a curvilinear relationship in the pain–sleep relationship, such as that of Edwards et al. [19], who demonstrated that individuals who slept less than 6 hours or 9 or more hours reported greater next day pain. Taken together, these findings suggest that from a clinical standpoint, it is important to account for not only a ceiling effect, but also an inverse relationship between total hours slept and next day pain, when patients are sleeping greater than ∼9 hours. Another important finding from our study was that the strength of the association between previous night’s TST and next day pain was predictive of greater correspondence between changes in TST and changes in pain over the course of treatment. Individuals whose TST was strongly tied to their next-day pain experienced greater treatment benefits in terms of pain reduction across the 3–4 week treatment in the ICPRP, given a corresponding improvement in TST. This suggests that, for individuals whose TST is tied to next day pain, maximum

Sleep and Pain improvement in pain may be achieved by treating sleep difficulties. Our study adds to the growing empirical support that individuals with sleep disturbance are likely to have increased next day pain. This finding has been repeatedly demonstrated in the aforementioned investigations and was further supported in a recent large-scale prospective study of the daily associations between pain and sleep in the general population [19]. However, when exploring the bidirectionality of the sleep–pain relationship, previous research is less clear. Studies utilizing sleep diaries demonstrated that a night of poor sleep predicts increased next day pain and that day of increased pain predicts night of poor sleep in a cross-sectional analysis of 30 females with fibromyalgia, although such findings were not significant when accounting for pain attention [5]. Other studies found a weak or insignificant association between day of pain and subsequent sleep [11]. A recent study by Tang and colleagues aimed to address this gap in the literature through an analysis of daily pain, sleep, mood, and presleep arousal for a 1-week period using self-report and actigraphy. Similar to our findings, the study challenged previous conceptualizations of reciprocality in the pain–sleep relationship through the analysis of individual-level data across a 1-week period. Specifically, the authors found that while sleep quality predicts subsequent pain, daily pain did not predict that night of sleep, but rather was predicted by presleep cognitive arousal [6]. From a methodological standpoint, our study stresses the importance of examining individual differences in withinperson associations. While previous cross-sectional research demonstrated that the pain–sleep relationship is bidirectional, there has been limited research utilizing prospective daily ratings of pain and sleep to investigate the reciprocal nature of the relationship. Most studies thus far have relied solely upon statistical analyses that display the between person effects or the average of within person effects. Common regression statistics assume that patients’ scores across time are independent (i.e., a patient’s pain on day 1 is independent from that patient’s pain on day 2), which is almost definitely not the case. To the degree that this assumption is violated, common regression methods yield misleading results. The findings of Edwards et al. [19] and Tang et al. [6] provide important steps in utilizing more complex statistical modeling required to fully understand daily sleep and pain associations. Edwards et al. utilized structural equation modeling to analyze the prospective associations between pain and sleep in the general population. Similar to our study, they found a significant association between night of sleep and next day pain. In contrast to our study, they found an association (albeit weak) between day of pain and subsequent night of sleep. MLM was chosen over structural equation modeling (SEM) in our study because in MLM time is treated as a single variable, whereas SEM treats each time point as a separate variable. The consequence of this is that MLM treats time more flexibly and allows for measurements at different times (and at different time

intervals) between individuals [20]. The research of Tang et al. is the only other study to our knowledge that utilized MLM. Our study expands upon this previous research by exploring the relations between daily associations of pain and sleep, over an extended duration and in the context of a concurrent treatment program. In summary, our study suggests that individuals with both pain and sleep difficulties can be treated successfully in the context of an ICPRP. Notably, for those individuals whose daily pain is strongly tied to a previous night of poor sleep (or less TST), interventions focused on sleep may result in improvements in pain over the course of treatment. Thus, our findings highlight the importance of tailoring treatment plans to each individual. Certainly, as a whole, sleep problems are common in the chronic pain population. Our study shows that the degree to which sleep difficulties relate to pain differently across individuals may have implications in terms of treatment planning. It is interesting that for individuals in our study, a day of increased pain was unrelated to a night of less TST. Given previous findings supporting the role of presleep cognitive rumination and presleep pain-related thoughts as a predictor of sleep disturbance in chronic pain, it would be valuable to further explore the unique factors that may mediate the relationship between night of sleep and next day pain. Our findings accentuate the value of preidentifying individuals for more intensive intervention surrounding sleep and pain issues. The use of sleep and pain diaries or other pain and sleep screening measures prior to start of the ICPRP would help clinicians to further elucidate the extent of the pain–sleep difficulties in ICPRP candidates. In turn, such candidates could be appropriately placed in targeted pain and sleep treatment groups Our findings suggest that treatment of comorbid pain and sleep problems in a subset of the patient population is vital to ensure optimal treatment success. However, a significant challenge to date has been establishing empirically supported interventions that provide mutual relief of sleep disturbance and pain in the chronic pain population. Our study reinforces the need for intensive education and intervention programs that address the unique interplay between sleep and pain. Further research, into the efficacy of a combined CBT-I/CBT-P approach for patients with CNCP is needed. Another potentially fruitful area of research would be to compare the ICPRP outcomes of patients’ who are preidentified to participate in a sleep– pain intervention such as CBT-I/CBT-P to those who are not. Additional implications of our study are the benefits of treating both pain and sleep disturbance in the context of an ICPRP. The efficacy and cost-effectiveness of ICPRPs has been repeatedly demonstrated in the literature, with evidence of sustained benefits of treatment for up to 13 years [21]. To our knowledge, this is the first study that has looked at the impact of an ICPRP treatment approach on both sleep and pain. We found that, as a whole, all 1049

Davin et al. patients improved in both pain and TST during treatment. Time in treatment was a predictor of decreased pain and TST, and such findings were maintained even when controlling for age, gender, and mood. Thus, these findings suggest that patients who drop out of treatment early or do not complete an extended period of treatment (∼15 days) may not reap such benefits of improved TST and pain during treatment.

7 Chen G, Guilleminault C. Sleep disorders that can exacerbate pain. In: Lavigne G, Sessle BJ, et al., eds. Sleep and Pain. Seattle: IASP Press; 2007:311–40.

Limitations of our study include a small sample size and reliance upon self-report data of pain, mood, and sleep duration. The self-report rating scales for pain, TST, anxiety, and depression were used to provide a quick and clinically practical snapshot of each patient at the start of the treatment day. However, the single-item rating scale for TST is limited in scope, such that it does not account for other common sleep complaints such as sleep latency or number of awakenings. Future research should utilize objective measures of sleep, such as actigraphy, as well as measures of sleep quality.

9 Vitiello MV, Rybarczyk B, Von Korff M, Stepanski EJ. Cognitive behavioral therapy for insomnia improves sleep and decreases pain in older adults with co-morbid insomnia and osteoarthritis. J Clin Sleep Med 2009;5(4):355–62.

Additionally, our study did not account for comorbid diagnosed sleep disorders (such as sleep apnea), or other notable conditions, such as addiction, which may impact sleep quality and duration. Further research should account for potentially confounding comorbid diagnoses. Additionally, longitudinal assessment could provide more compelling evidence for the sustainability of changes in sleep and pain in the chronic pain population.

References 1 Smith MT, Huang MI, Manber R. Cognitive behavior therapy for chronic insomnia occurring within the context of medical and psychiatric disorders. Clin Psychol Rev 2005;25(5):559–92. 2 Stiefel F, Stagno D. Management of insomnia in patients with chronic pain conditions. CNS Drugs 2004;18(5):285–96. 3 Hakki OS, Alloui A, Jourdan D, Eschalier A, Dubray C. Effects of rapid eye movement (REM) sleep deprivation on pain sensitivity in the rat. Brain Res 2001; 900(2):261–7. 4 Raymond I, Nielsen TA, Lavigne G, Manzini C, Choinière M. Quality of sleep and its daily relationship to pain intensity in hospitalized adult burn patients. Pain 2001;92(3):381–8. 5 Affleck G, Urrows S, Tennen H, Higgins P, Abeles M. Sequential daily relations of sleep, pain intensity, and attention to pain among women with fibromyalgia. Pain 1996;68:363–8. 6 Tang NK, Goodchild CE, Sanborn AN, Howard J, Salkovskis PM. Deciphering the temporal link between pain and sleep in a heterogeneous chronic pain patient sample: A multilevel daily process study. Sleep 2012;35(5):675–87A. 1050

8 Currie SR, Wilson KG, Ponteffract AJ, deLaplante L. Cognitive-behavioral treatment of insomnia secondary to chronic pain. J Consult Clin Psychol 2000;68(3): 407–16.

10 Jungquist CR, O’Brien C, Matteson-Rusby S, et al. The efficacy of cognitive-behavioral therapy for insomnia in patients with chronic pain. Sleep Med 2010; 11(3):302–9. 11 Pigeon WR, Moynihan J, Matteson-Rusby S, et al. Comparative effectiveness of CBT interventions for co-morbid chronic pain & insomnia: A pilot study. Behav Res Ther 2012;50(11):685–9. 12 Beaulieu P, Walczak JS. Pharmacological management of sleep and pain interactions. In: Lavigne G, et al., eds. Sleep and Pain. Seattle: IASP Press; 2007:391–415. 13 Cairns BE. Alteration of sleep quality by pain medication: An overview. In: Lavigne G, et al., eds. Sleep and Pain. Seattle: IASP Press; 2007:371–90. 14 Covington E, Davin S, Scheman J. Interdisciplinary chronic pain rehabilitation clinics. In: Daroff RB, Aminoff MJ, ed. Encyclopedia of Neurological Sciences, 2nd edition. Salt Lake City, UT: Academic Press; Forthcoming June 2014. 15 Lovibond PF, Lovibond SH. The structure of negative emotional states: Comparison of the depression anxiety stress scales (DASS) with the beck depression and anxiety inventories. Behav Res Ther 1995;33: 335–43. 16 Tait RC, Pollard CA, Margolis RB, Duckro PN, Krause SJ. The pain disability index: Psychometric and validity data. Arch Phys Med Rehabil 1987;68(7):438–41. 17 Steele F Module 5: Introduction to multilevel modeling concepts. Centre for Multileveling Modeling. 2008. Obtained from http://www.bristol.ac.uk/cmm/ learning/course-topics.html#m05 (accessed February 2014). 18 Chambers JM. Software for Data Analysis: Programming with R. New York: Springer; 2008. 19 Edwards RR, Almeida DM, Klick B, Haythornthwaite JA, Smith MT. Duration of sleep contributes to next-

Sleep and Pain day pain report in the general population. Pain 2008;137:202–7.

Multilevel Analysis. New York: Taylor & Francis Group; 2011:97–111.

20 Stoel RD, Garre FG. Growth curve analysis using multilevel regression and structural equation modeling. In: Hox JJ, Roberts JK, eds. Handbook of Advanced

21 Patrick LE, Altmaier EM, Found EM. Long-term outcomes in multidisciplinary treatment of chronic low back pain. Spine 2004;29(8):850–5.

1051