Background
Approximately one fourth of adults worldwide have chronic obstructive pulmonary disease (COPD) [
1], and half of them suffer from insomnia [
2] (difficulty falling or staying asleep, or poor-quality sleep that interferes with daytime functioning [
3]). Insomnia is related to greater mortality [
4‐
6], with studies showing
four times the mortality risk for sleep times below 300 min [
5,
6]. Insomnia also produces morbidity – insomnia sufferers generate 75 % greater healthcare costs than individuals without insomnia [
7]. Lost productivity from insomnia is estimated at US$63.2 billion annually in the US alone [
8]. People with COPD experience debilitating fatigue and a gradual decline in function that are partly related to insomnia [
9‐
11]. However, insomnia medications are used with caution in COPD due to potential respiratory effects, hypoxia and effects on cognition [
12]. Common features of COPD, such as dyspnea, chronic inflammation and emotional arousal (anxiety and depression), also affect insomnia and can interfere with therapy outcomes. Cognitive behavioral therapy for insomnia (CBT-I), a therapy that provides guidance on changing unhelpful sleep-related beliefs and behavior, is a promising non-pharmacological treatment that has demonstrated effectiveness for insomnia, but the efficacy of components and mechanisms of action of CBT-I have not been thoroughly examined in people with insomnia coexisting with COPD.
Insomnia is prevalent in people with medical disorders, and therapy for insomnia might improve these medical disorders [
13]. However, there are only a limited number of randomized, controlled studies in this area. Although evidence obtained from studies that included participants with cancer [
14,
15], renal failure [
16], chronic pain [
17], osteoarthritis, [
18] and COPD [
19,
20] support the notion that CBT-I is effective in reducing insomnia when coexisting with chronic illness, no studies to date have examined CBT-I specifically for people with COPD. It is yet undetermined whether improving self-efficacy for the medical illness (i.e., COPD) mediates outcomes after CBT-I, but our preliminary data suggest that education on COPD topics, such as management of dyspnea and COPD exacerbations, may improve self-efficacy and reduce depression, indirectly reducing insomnia and fatigue [
20].
Our long-term goal is to help develop safe and effective non-pharmacological interventions to minimize insomnia and its consequences in people with COPD. The purpose of this study is to systematically test the efficacy of two components of therapy for people with coexisting insomnia and COPD – CBT-I and COPD education (COPD-ED) – and to identify the mechanisms responsible for therapeutic outcomes. Our central hypothesis is that both CBT-I and COPD-ED will have positive, lasting effects on objectively and subjectively measured insomnia and fatigue. This hypothesis is consistent with preliminary data from our pilot study [
20] comparing CBT-I and COPD-ED in people with insomnia coexisting with COPD. The pilot study results demonstrated feasibility of the two components and provided preliminary evidence of positive effects on outcomes, which differed in men and women. Given the prevalence of insomnia and its consequences, it is important not only to identify the most effective approach to minimize insomnia for those with COPD but also to identify the mechanisms responsible for the outcomes, as not all patients will benefit from CBT-I. The rationale for this study is that once the efficacy and mechanisms of CBT-I and COPD-ED are known, new and innovatively tailored interventions can be developed to non-pharmacologically minimize insomnia and fatigue, thereby leading to longer, higher-quality and more productive lives for people with COPD, and reduced societal cost due to the effects of insomnia. The clinical gains could be, unlike those of pharmacotherapy, sustained following treatment discontinuation. We are testing our central hypothesis by completing a randomized, parallel-group, 2 × 2 factorial design which results in four groups (
N = 35 each group): CBT-I, COPD-ED, both CBT-I and COPD-ED, and neither [
21,
22].
Results from previous studies of predictors of positive outcomes of CBT-I suggest that mechanisms include changes in sleep-related beliefs, sleep habits, emotional arousal and self-efficacy for sleep [
23]. Thus, our contribution here is expected to be knowledge of the efficacy and mechanisms of the components of a novel insomnia therapy for people with COPD and a detailed understanding of which patients are most likely to benefit from the therapy. This contribution will be significant because it is a necessary step in the development of effective non-pharmacological therapies for insomnia coexisting with COPD. Once such advances in therapy for insomnia coexisting with COPD become available, results can be used to further develop and make available effective and efficient insomnia therapies for people with COPD. Components of therapy found to be most important for positive outcomes can be included in new, more efficient therapies. It is expected that what we learn will be equally applicable to the prevention of insomnia and fatigue in people with COPD, potentially leading to them having longer, higher-quality and more productive lives and to society having to burden lower costs related to insomnia in COPD.
We are testing components of a novel therapy (CBT-I and COPD-ED) in an understudied population using a highly efficient study design. Despite the established need for such a therapy for people with COPD, this group is understudied, perhaps because of challenges such as recruiting subjects with COPD and managing periodic exacerbations of COPD that could occur during treatment. Furthermore, people with insomnia coexisting with COPD are subject to exacerbations of their illness that predispose them to recurrent insomnia and may interfere with CBT-I outcomes. The research represents a new and substantive departure from the usual insomnia therapy, specifically by combining traditional CBT-I with disease-specific education (COPD-ED) to enhance self-efficacy for the management of COPD. Enhancing self-efficacy for COPD is likely to reduce insomnia and fatigue by attenuating the anxiety and depressed mood associated with the disease.
Aim 1
To determine the efficacy of individual treatment components, CBT-I and COPD-ED, on insomnia and fatigue. Our hypothesis is that both components will decrease insomnia and fatigue at the end of the six-session treatment period, and that these differential effects will be sustained for at least 3 months post treatment.
Aim 2
To define the mechanistic contributors to the outcomes after CBT-I and COPD-ED. Our hypothesis is that CBT-I and COPD-ED components impact insomnia and fatigue through complementary mechanisms. Positive changes in beliefs about sleep, sleep habits, self-efficacy for management of COPD (SEC) and sleep (SES), and reduced emotional arousal (EA) will mediate the improving conditions in insomnia and fatigue and that gender, inflammation and functional status will moderate the outcomes.
Methods: data collection, data management and analysis
Data collection
Screening procedures include an initial telephone screening to determine eligibility. Potential subjects are then scheduled for their first screening visit, which takes place at the CNSHR. At that visit, after informed consent is obtained, PFTs are performed, followed by questionnaires. Reliability and validity of pulmonary function testing have been previously demonstrated [
62,
63]. The Hospital Anxiety and Depression Scale (HADS) [
64] is administered to screen for major depressive symptoms. The HADS was designed for outpatients with medical illnesses, and it has been widely used in people with COPD [
65‐
67]. Reliability and validity were demonstrated [
68]. A cutoff score of 11 on the HADS depression subscale is used for depression [
64]. A clinical questionnaire is completed at this visit to characterize the sample in terms of demographics and health history. Blood specimens are sent to Quest Diagnostics® for measurement of C-reactive protein and ferritin level. These tests are done in order to rule out anemia and to adequately describe the sample and may be used as covariates in the analyses.
The screening includes a one-night sleep study at home or at the UIC Sleep Science Center or the Hines VA Sleep Laboratory. The sleep study is used to screen potential subjects for sleep apnea, and low SaO2 during the night. Objective evidence of any primary sleep disorder according to standard clinical criteria is used to exclude subjects. The remaining subjects receive instructions for continued actigraphy monitoring by wearing an actigraph on their non-dominant wrist at home for a total of 1 week. They are instructed on how to complete the daily sleep diary.
Excluding the screening, participants have a total of eight or nine laboratory visits (six intervention, pretest, posttest and follow-up). The 3-month follow-up is completed at home or in the laboratory. After the 1-week at-home monitoring period, they are scheduled for the baseline (pretest) visit at either the UIC CON Center for Narcolepsy, Sleep and Health Research (CNSHR) or the Hines VA. Upon completion of the baseline testing, subjects are randomly assigned to group. Subjects recruited from the Hines VA attend the six intervention visits there. Subjects recruited from outside of the VA attend the six intervention visits at the UIC CNSHR. For the 3-month follow-up, subjects either come to UIC or Hines VA for a final visit or they are mailed a packet containing an actigraph, questionnaires and a self-addressed, stamped return envelope. We track attendance using attendance records and the sleep diary. All tests are performed in the same order for each subject. Tests and questionnaires are administered in a private, comfortable room. We make every attempt to minimize selection and performance bias, attrition and missing data. Selection bias is minimized by use of the randomized controlled trial design. Staff performing data analysis are blinded to group assignment. We use strategies to reduce attrition such as good communication between research staff and participants, minimizing participant burden by the use of PROMIS CAT instruments, and monetary incentives. When subjects do not complete the study, we document the reasons data are missing in order to evaluate the missing data using justified assumptions.
Enactment and fidelity
The clinician interventionalists are trained in CBT-I, COPD-ED, CBT-I + COPD-ED, and the AC education topics by experts. They observe several sessions being conducted by a clinician experienced in the interventions and practice the interventions under supervision. They use a manual developed during the pilot research to guide the sessions. In order to assure high-quality sessions, samples are reviewed and scored for adherence to the protocol. Subject adherence is also tracked. Adherence to stimulus control and sleep restriction is assessed at each session using the sleep diary and/or actigraph. Evaluation of adherence to stimulus control is assessed by examining responses to a sleep diary question that asks how the participant managed periods awake during the night. Evaluation of adherence to sleep restriction is assessed by comparing the recommended time in bed to the reported time in bed.
Data management
Data are entered onto a REDCap application and stored on a secure server with daily back-ups. The raw data are stored in locked file cabinets that are housed in our research laboratory protected with a security lock. Only the immediate research staff (PI, key personnel) have access to these data files. The computerized data files are password-protected and available only to the immediate research staff (PI, key personnel). Identifiable elements are needed during the study, for instance in order to contact subjects for data collection appointments. Data are coded so that subject identifiers are on data sheets. All data are checked for completeness and validity. The quality of the data is monitored at least annually by the principal investigator (PI) assisted by the research specialists.
Statistical analysis
Test of specific hypotheses
Descriptive statistics (means, standard deviation, frequency) and plots are used to screen the data prior to our main analyses. Necessary transformation and imputations are conducted based on the raw data distribution. Baseline adjustment for covariates (e.g., age, COPD severity (PFT), and gender) will be incorporated into the main effect analyses to reduce error variance and improve statistical power [
69]. Data analyses is performed using SAS 9.3 statistical software (SAS Inc., Cary, NC, USA) and Mplus 6.0 (Muthén & Muthén, Los Angeles, CA, USA). All tests will be two-sided and an error rate of
α <0.05 will be considered statistically significant. The test of each specific aim is described below, and all statistical analysis will employ an intent-to-treat approach. A fully specified statistical analysis plan will be written before unmasking.
Aim 1: to determine the efficacy of individual treatment components, CBT-I and COPD-ED, on insomnia and fatigue. In this 2 × 2 design, participants are randomized into four groups created by crossing the two factors, CBT-I versus AC1 and COPD-ED versus AC2 (see Table
3). We hypothesize that both components will improve insomnia and decrease fatigue following the six-session program, and that participants will maintain these gains at 3 months post intervention. Our factorial design permits the test of an interaction between the two treatment components, and we will assess for the presence of a strong interaction effect. However, we do not anticipate a strong synergy of the treatments since our pilot work showed that both had within-group improvement, and the hypothesized mechanisms for change differ. If the interaction is negligible, we will proceed with the interpretation of averaged main effects as shown in Table
3. The statistical tests for both components will be more robust than tests against a simple control condition because they are averaged across the levels of the other treatment. That is, the effect of each component will be tested controlling for the effect of the other. This is an efficient design because all subjects will be used to test both components. Finally, it may be more acceptable to participants since all receive supportive contact and only one out of four conditions does not receive a component hypothesized as an active treatment for insomnia in COPD patients. We will employ mixed-effects models using insomnia and fatigue as time-varying dependent variables and treatment groups as fixed effects. Demographic and clinical characteristics will be entered as time-invariant covariates if baseline group differences are observed despite randomization. Individuals’ baseline sleep measures and change over time (i.e., intercept and slope) will be modeled as random effects. The null hypothesis will be rejected if a significant treatment × time interaction is observed. For example, a significant CBT-I × time effect in the hypothesized direction would mean that participants receiving CBT-I improved more than those who did not, averaged across the other conditions. The mixed models will be run using PROC MIXED SAS 9.3 (SAS Inc., Cary, NC, USA) and estimated by residual maximum likelihood (REML). Covariance pattern structures (compound symmetry and unstructured) will also be examined, and models will be compared using likelihood ratio tests or Akaike’s Information Criteria (AIC) which is a function of the log likelihood and can be compared across models.
COPD-ED | 35 | 35 | 70 |
AC2 | 35 | 35 | 70 |
Main effect CBT-I | 70 | 70 | |
Aim 2: to define the mechanistic contributors to the outcomes after CBT-I and COPD-ED. We hypothesize that CBT-I and COPD-ED components impact insomnia and fatigue through complementary mechanisms. We have previously shown that CBT-I was related to beliefs about sleep, sleep habits, self-efficacy for sleep, and that COPD-ED was related to management of COPD and emotional arousal, thus making them both appropriate candidates for an insomnia treatment approach for COPD patients. We will employ path analysis to test the conceptual model illustrated in Fig.
1 [
70]. The model will test whether each treatment component is related to change in the hypothesized mediator, and whether that change is associated with improvement in the outcome. Direct and indirect treatment effects will be estimated. We will conduct path analyses as follows: (1) specification, (2) identification, (3) estimation, (4) testing of fit, and (5) respecification [
71]. Potential moderators suggested by our previous work include gender, initial insomnia and fatigue severity, and inflammation and will be tested as interactions of the effect of each treatment component on the outcomes. Variables found to be a significant moderator of treatment will be tested further to determine if the effect is due to a moderated impact on significant mediator variables. Path models assume that variables used to describe relationships are manifest variables and measured without error. While we recognize that it would be preferable to estimate a measurement model, the feasible sample size for this study precludes more complex modeling. Thus, we acknowledge this limitation and will restrict path analysis to reliably measured variables, that is, with internal consistency of 0.80 or higher. Mplus (version 6) will be used to estimate the path models. The root-mean-square error of approximation (RMSEA) and Bentler’s Comparative Fit Index (CFI) will be reviewed to assess model fit. An adequate fit of the data to the model is indicated by a RMSEA value less than 0.08 and a CFI greater than 0.90.
Missing data
For missing data, we will determine whether missing data are MCAR (missing completely at random), MAR (missing at random), or NMAR (not missing at random). If MCAR or MAR, the standard multivariate computations will not likely result in biased standard error estimates, and full information maximum likelihood (FIML) estimation will be used. If NMAR, we will use the “pattern mixture” approach to compute a “weighted average” of the parameters associated with the missing data to estimate treatment effects.
Ethics and dissemination
Ethics
The study protocol is approved by the Institutional Review Board Office for the Protection of Research Subjects of the University of Illinois at Chicago and the Institutional Review Board of the Edward Hines, Jr. VA Hospital. The informed consent process begins when potential subjects are contacted. The researcher explains the study over the telephone. During the telephone consent, research staff explain the purpose of the study, study procedures, benefits, risks, confidentiality, and research subject’s rights. For potential subjects who want to come in for a screening appointment in the CNSHR, the informed consent process continues with a face-to-face explanation and discussion of the study. After all questions have been answered and the subject verbally agrees to participate, the subject signs the written informed consent and a copy of the document is provided to the subject. All staff attend the UIC IRB training program and continuing IRB education programs.
Strict procedures are in place to minimize the risk of breach of confidentiality. All subjects are assigned a code. The master list of the subject’s name and the linked code is kept in a password-protected computer in the PI’s office. All information provided by subjects is kept strictly confidential and is not be reported on an individual basis. None of the information provided by subjects becomes part of the medical record. Hard copy data are stored in a locked office, and electronic data are stored on a password-protected computer. Hard copy data and electronic data are coded, with the master list kept separately in a secure file in the PI’s office. A Health Insurance Portability and Accountability Act (HIPAA) consent form was developed for this study to use/disclose protected health information. Subjects are asked to sign the HIPAA consent form and those who refuse to agree to the HIPAA consent are not able to participate in this study.
Dissemination
Investigators will communicate trial results to the public and healthcare professionals through publications and presentations. The final report will follow the main Consolidated Standards of Reporting Trials (CONSORT) 2010 guideline; as well as their extension to non-pharmacological interventions and to PRO outcomes.
Abbreviations
AC, Attention Control; COPD, chronic obstructive pulmonary disease; CBTI,cognitive behavioral therapy for insomnia; COPD-ED, COPD education;COPD SES, self-efficacy for COPD; CRQ, Chronic Respiratory DiseaseQuestionnaire; DBAS, Dysfunctional Beliefs and Attitudes about Sleep;ESS, Epworth Sleepiness Scale; FIML, full information maximum likelihoodestimation; HADS, Hospital Anxiety and Depression Scale; MCAR, missingcompletely at random; MAR, missing at random; NA, number of awakeningsafter sleep onset; NMAR, not missing at random; PFT, pulmonary functiontest; PROMIS, Patient Reported Outcomes Measurement Information System;RMSEA, root mean square error of approximation; SE, sleep efficiency;SES, self-efficacy about sleep; SII, Sleep Impairment Index; SL, sleep onsetlatency; TST, total sleep time; VA, Veterans Administration; WASO, wake aftersleep onset.