Cognitive Behavioral Therapy for Insomnia May Improve Sleep and Pain in Crohn’s Disease: A Waitlist Control Pilot Trial

Abstract

Background

Poor sleep is common in Crohn’s disease (CD), prospectively predicts worse disease course, and is often attributable to insomnia. Cognitive behavioral therapy for insomnia (CBT-I) is the recommended treatment for chronic insomnia disorder. CBT-I improves sleep and may improve pain intensity, pain interference, and inflammation. We sought to investigate whether CBT-I impacts these factors in patients with active CD.

Method

We recruited patients with insomnia and mild-to-moderate CD symptoms from an inflammatory bowel disease center. Exclusion criteria were other sleep disorders, significant psychiatric concerns, and presence of other common influences on sleep. Participants completed baseline assessments of sleep, pain, and inflammation then were randomized to receive CBT-I immediately, or wait 12 weeks and then repeat the baseline assessment and complete CBT-I. Similar assessments occurred immediately post–CBT-I and 1 month later. CBT-I included sleep restriction, stimulus control, sleep hygiene, arousal reduction, and cognitive therapy.

Results

A total of 26 participants completed the study. In group × time analyses, CBT-I led to greater reductions in insomnia severity (P < .001) and wake after sleep onset (P = .02) than waitlist. In pre- to post-treatment analyses, participants reported significant improvements in subjective measures of sleep continuity, CD symptom severity, pain intensity, and pain interference. C-reactive protein trended toward improvement.

Discussion

This study provides preliminary evidence of efficacy of CBT-I in people with CD. CBT-I improved self-reported sleep and may improve pain and CD symptoms. The results highlight the importance of addressing sleep concerns in inflammatory bowel disease, particularly in people with persistent pain or fatigue. Future trials powered to detect changes in pain and inflammation are warranted.

Key Messages.

What is already known?

Insomnia is common in Crohn’s disease (CD) and people with CD are interested in insomnia treatment.

What is new here?

Behavioral treatment of insomnia improves sleep problems, and may improve CD symptoms, pain intensity, pain interference, and other psychosocial concerns, including fatigue.

How can this study help patient care?

Pain and fatigue are often hard to treat in CD. Insomnia may be an important modifiable treatment target for people experiencing this triad of overlapping concerns.

Introduction

Poor sleep is present in at least half of people with inflammatory bowel disease (IBD),¹ which prospectively predicts greater likelihood of symptom flares, surgery, and/or hospitalization.^2-4 While sleep disturbances may be attributable to a number of different factors, the most common sleep disorder present in people with IBD is insomnia disorder.⁵ Insomnia disorder is characterized by chronic difficulty initiating and/or maintaining sleep combined with daytime impairment and is the most common sleep disorder in people with Crohn’s disease.⁶ Approximately two-thirds of people with active symptoms of Crohn’s disease experience symptoms of insomnia,⁵ and most people with Crohn’s disease believe that sleep is closely related to their Crohn’s symptoms.^7,8 Further, people with Crohn’s disease are interested in insomnia treatment as part of their overall disease management.⁵

Insomnia also has the potential to maintain and/or worsen pain, likely via direct alterations in central pain processing circuits that increases pain sensitivity.^9-11 Microlongitudinal data in people with inflammatory bowel disease has demonstrated that sleep quality predicts next day pain.^12,13 Similarly, recent research using actigraphy has demonstrated an association between abdominal pain and sleep fragmentation in IBD.¹⁴ Given that approximately 70% of people with Crohn’s disease experience abdominal pain,^15,16 the sleep/pain relationship is likely an important intervention target.

Cognitive behavioral therapy for insomnia (CBT-I) is the recommended first line treatment for chronic insomnia disorder by the American College of Physicians¹⁷ due to its comparable effectiveness, enhanced durability, and improved side-effect profile compared with pharmacotherapy.^18,19 While we have demonstrated that CBT-I adapted to IBD is feasible and acceptable,²⁰ additional research is needed to better understand the impact of CBT-I on both Crohn’s disease symptoms and related concerns. Herein, we sought to investigate the preliminary impact of CBT-I on not only insomnia symptoms, but also self-reported and objective measures of sleep continuity, pain, and inflammation in a sample of participants with self-reported active symptoms. We specifically sought to investigate these relationships in Crohn’s disease (as opposed to ulcerative colitis) because pain is a key symptom of Crohn’s disease and poor sleep, including circadian rhythm dysrhythmia, appears to be particularly linked with CD.^2,7,21,22

Method

Trial design

Participants with self-reported mild-to-moderate symptoms of Crohn’s disease and insomnia were recruited for a clinical trial on CBT-I (NCT05034159). Following consent and screening, participants completed a comprehensive baseline assessment. Once this assessment was complete, participants were randomized by into either a 12-week waitlist period or CBT-I using block randomization (blocks of 4). Randomization was completed by the research director for the department, who was otherwise not directly involved in this trial. Randomization was stratified by insomnia symptom severity and Crohn’s disease symptom severity. Participants who were randomized to the waitlist completed a second baseline after 12 weeks, following which they received CBT-I. All participants completed another comprehensive assessment 1 month after completing CBT-I to assess for changes that occurred following intervention. Due to the waitlist control design, neither the participants nor the study coordinator were blinded. See Figure 1 for study design. This study was approved by the Dartmouth Health Human Research Protection Program (#02001191), and informed consent was provided by all participants.

Figure 1.

Trial CONSORT diagram. CBT-I, cognitive behavioral therapy for insomnia.

The trial was also designed to reduce barriers to participation, with measures taken including scheduling the majority of study visits through telehealth, requiring limited travel to the study site (2 or 3 visits over 4-7 months), scheduling visits outside normal business hours, and timing study visits with other required medical appointments to reduce travel.

Participants

Participants were recruited from the Dartmouth-Hitchcock Medical Center Inflammatory Bowel Disease Center. Recruitment occurred between October 2021 and September 2022, and inclusion criteria were as follows: (1) mild-to-moderate Crohn’s disease activity as assessed by the Patient Reported Outcomes-3 (PRO-3),²³ (2) Insomnia Severity Index²⁴ score ≥8, (3) average weekly sleep onset latency and/or wake after sleep onset ≥30 minutes, (4) willingness to not change sleep medications over the course of the trial, and (5) access to a device and internet or cell phone service sufficient for telehealth visits. Exclusion criteria were (1) Patient Health Questionnaire-9²⁵ depression score ≥20, (2) Generalized Anxiety Disorder-7²⁶ anxiety score ≥20; (3) unstable major psychiatric condition; (4) current alcohol or substance abuse; (5) current opioid use for pain control, (6) current smoker (tobacco, nicotine); (7) current systemic corticosteroid use; (8) current pregnancy or nursing, (9) ileostomy or colostomy; (10) diagnosis of seizure disorder; (11) diagnosis of sleep apnea or positive STOP-Bang²⁷ screen; (12) diagnosis of restless leg syndrome or positive Cambridge-Hopkins Restless Leg Syndrome Questionnaire²⁸ screen; and (13) night shift, rotating shift work, or frequent travel outside of primary time zone.

Intervention

A 5-session multicomponent CBT-I intervention was delivered individually via telehealth over the course of 7 to 8 weeks. CBT-I components included sleep restriction, stimulus control, sleep hygiene, arousal reduction (worry time and relaxation), and cognitive therapy. We previously adapted a traditional CBT-I interventionist outline and patient manual to people with IBD.²⁰ All intervention sessions were delivered by J.K.S.-D., who is a clinical psychologist and a diplomate in behavioral sleep medicine. Sessions 1 and 2 occurred 1 week apart, whereas the others were separated by 2 weeks. In between sessions 2 and 3, 3 and 4, and 4 and 5, brief phone calls occurred to make changes to sleep opportunities and/or problem-solve any barriers to skill practice. Treatment sessions lasted around 45 minutes, while phone calls were around 10 minutes.

Measures

Demographics

Participants self-reported their age, sex, race, ethnicity, home address, and Crohn’s disease history (age at symptom onset, age at diagnosis, surgical history). Crohn’s disease phenotype and medications were abstracted from participants’ electronic medical record.

Given the possible influence of social determinants of health on sleep and overall disease outcomes,^29,30 we also assessed this important demographic variable. The Social Deprivation Index,³¹ developed by the Robert Graham Center, is a zip code–based measures of social determinants of health. Scores range from 1-100 and greater scores are suggestive of worse overall area deprivation. Social Deprivation Index scores were calculated based on zip code level data from the 2019 American Community Survey.

Questionnaires

The following questionnaires were used: the Insomnia Severity Index, a 7-item questionnaire designed to assess symptoms of insomnia disorder, with a higher score indicating greater severity²⁴; the PRO-3, a 3-item questionnaire designed to assess severity of Crohn’s symptoms, including frequency of liquid or soft stools, severity of abdominal pain, and overall well-being²³; the Brief Pain Inventory, a questionnaire that assesses pain intensity (4 items) and pain interference (10 items) in the past week,³² with higher scores indicating greater pain intensity and pain interference; the Generalized Anxiety Disorder-7, a 7-item questionnaire designed to assess symptoms of generalized anxiety disorder, with higher scores indicating greater symptom severity²⁶; the Patient Health Questionnaire-9, a 9-item questionnaire designed to assess symptoms of depression, with higher scores indicating greater symptom severity²⁵; the Pain Catastrophizing Scale, a 13-item questionnaire designed to assess the severity of catastrophic thoughts about pain experiences, with higher scores indicating worse overall catastrophizing³³; and PROMIS (Patient-Reported Outcomes Measurement Information System) Fatigue, a 4-item questionnaire developed as part of PROMIS and assesses overall severity of fatigue, in which raw scores are transformed into T scores, and higher scores indicate worse fatigue.³⁴

Daily diary

Participants completed a structured electronic diary twice per day, morning, and evening, as both a study measure and intervention component. The morning diary was modeled after a consensus sleep diary.³⁵ Variables of interest include average weekly sleep onset latency (minutes it initially took to fall asleep); number of awakenings; wake after sleep onset (minutes spent awake during the night after falling asleep); total sleep time (time spent in bed minus minutes awake), sleep efficiency (total sleep time/time in bed * 100); sleep quality, rated from 0 (worst) to 10 (best); and nighttime bowel movements. The evening diary included items on stress, anxiety, fatigue, number of minutes napped, whether a participant was menstruating that day, and the 3 items from the PRO-3. We considered sleep onset latency >30 minutes, wake after sleep onset >30 minutes, sleep efficiency <85%, and total sleep time <6.5 hours to be clinically significant.

Objective measures

The ActiGraph GT9X is a well-validated wrist-worn medical-grade triaxial accelerometer.³⁶ Actigraphy data were scored using a polysomnography-validated algorithm (Cole-Kripke) embedded in the CentrePoint platform.³⁷ Based on standard guidelines,³⁸ participants wore the ActiGraph on their nondominant wrist for 1 week during each assessment period. Following study completion, actigraphy data were examined for validity and extreme deviations from diary values. Changes were made to the automatically detected sleep period onset and offset time based on standardized procedures.³⁶

The Sleep Profiler is a compact, wireless, ambulatory sleep monitor that measures sleep architecture via 3 channels of frontal electroencephalography and a head-movement sensor. Sleep continuity and architecture was scored via a standardized approach developed by our team, which is consistent with American Academy of Sleep Medicine guidelines.³⁹ Validity for the measurement of sleep architecture is comparable to the gold standard for assessing sleep architecture, polysomnography, while less costly and burdensome than laboratory testing.³⁹ Participants wore the Sleep Profiler for 3 nights during each assessment period.

The Cold Pressor Test is a behavioral measure of pain tolerance⁴⁰ that took place during a telehealth visit. Participants were mailed a standardized, submersible thermometer that was accurate to 1 °F. Briefly, participants were instructed to fill one-quarter of a large container with ice and then add cold water until it was three-quarters full. Participants took the water temperature, and if >40 °F, added more ice. Participants were then instructed to submerge their nondominant hand in ice water until intolerable and rate their discomfort from 0 (absolutely fine) to 10 (worst imaginable) at 20-second intervals. Participants were instructed to remove their hand when their pain or discomfort became intolerable, with a maximum submersion time of 2 minutes. The full protocol for this remote adaptation of the cold pressor test, including feasibility and acceptability metrics, is described in detail elsewhere.⁴¹

To measure C-reactive protein (CRP), participants had their blood drawn at the hospital lab between the hours of 8:00 am and 10:00 am via venipuncture. Blood samples were then isolated and aliquoted, and the plasma was cryopreserved at −80 °C using a PAX gene tube. Following study completion, plasma was extracted and analyzed based on standard hospital procedures. Values were reported to the nearest 0.1 mg/L, and values >4.9 mg/L were considered indicative of active inflammation.

For fecal calprotectin, participants were mailed standardized Buhlmann Diagnostics stool collection kits. They collected a small stool sample at home and refrigerated the sample overnight. The stool sample was brought to the hospital-based lab the following day and analyzed according to Buhlmann Diagnostics’ standardized procedures. Values were detectable down to 30 µg/g. Values >160 µg/g were considered elevated and were winsorized at the 25th and 75th percentiles to reduce the impact of extreme outliers in the upper range.

Outcomes

Primary outcomes

We assessed efficacy of CBT-I based on pre- to post-treatment change in self-reported insomnia severity (Insomnia Severity Index) and diary-based sleep onset latency and wake after sleep onset.

Secondary outcomes

We investigated the pre- to post-CBT-I changes in (1) diary-based sleep efficiency and total sleep time, (2) objectively measured sleep continuity variables (sleep onset latency, wake after sleep onset, sleep efficiency, and total sleep time), (3) self-reported pain severity and interference (Brief Pain Inventory), (4) self-reported Crohn’s disease symptom severity (PRO-3), (5) behaviorally assessed pain tolerance (Cold Pressor Test), (6) time spent in slow wave sleep, and (7) inflammation (CRP and fecal calprotectin). We also sought to investigate whether changes in CRP could be explained by increased time spent in slow wave sleep.

Exploratory outcomes

We explored the impact of CBT-I on diary-based number of awakenings, diary-based nighttime bowel movements, and other related psychosocial outcomes, including fatigue, pain catastrophizing, anxiety, and depression.

Sample size calculations

Across recent meta-analyses of CBT-I vs control,^42-45 sleep onset latency typically evidenced the smallest effect size (0.67-0.80) of our primary outcomes. Based on an effect size of 0.67, a significance level of P = .05, and power of 80%, we performed a power analysis to calculate sample size based on a 2-sided, paired-samples t test (pre-post). Accounting for an attrition rate of 21%,⁴⁶ an initial sample size of 25 was estimated to be sufficient to detect effect sizes in the range expected for sleep onset latency changes. Thus, while we were powered to detect changes in our primary outcomes from pre– to post–CBT-I, this study was not sufficiently powered to compare CBT-I to the waitlist period, nor was it fully powered to investigate CBT-I related changes in pain or inflammation. Secondary outcomes were exploratory in nature and results are intended to be used to inform a subsequent trial, including both sample size calculations and likely utility of various measures.

Data analytic strategy

Data preparation

All data were initially reviewed for completeness. We elected to include all available observations in the analysis instead of employing an imputation method, which may overfit in a small dataset and can introduce bias when data is not missing at random. The only exception was for actigraphy data, in which participants with <3 observations were excluded. We accounted for participant-level differences as a random effect in mixed-effects linear regression models, which may mitigate bias resulting from incomplete data. The number of nonmissing observations obtained at each time point are reported.

Given the small sample size, deviations from normality were expected. The Shapiro-Wilk test was used to assess the distribution of raw data for each outcome by time point and the residuals of mixed-effects linear regression models. When residual diagnostics indicated non-normality, a sensitivity analysis was conducted by refitting the model with robust standard errors.

Pre-post analyses

Mean ± SD was calculated for each measure at baseline, 8 weeks (immediately post-treatment), and 12 weeks. A standard effect size (Hedges’s g) with 95% confidence intervals was calculated to evaluate the change in mean scores 8 and 12 weeks after baseline. Mixed-effects linear regression was used to assess the impact of treatment on primary, secondary, and exploratory outcomes, with time point included as a fixed effect and participant-level variation accounted for as a random effect. Unadjusted analyses included only the random effect for participant-level variation, while adjusted models included age, sex, and baseline inflammation as fixed-effect covariates. Adjusted models were compared using the likelihood ratio test, with the model providing the best fit selected based on Akaike information criteria, Bayesian information criteria, and log-likelihood. Overall significance of fixed effects was evaluated using a type III analysis of variance, with F statistics and corresponding P values presented to represent the contribution of fixed effects to explaining variance in the outcomes. Standardized model estimates are presented for interpretability and to facilitate comparison of effect sizes across variables with different scales.

Group comparison

To evaluate whether CBT-I produced changes in primary outcome measures, analyses were repeated to compare the intervention group with the waitlist control group. Group means at baseline were compared using t tests for normally distributed data and Wilcoxon rank sum tests for non-normally distributed data. Mixed-effects linear regression models included an interaction term for group assignment and time as a fixed effect to assess differential changes over time.

Results

Ninety-three patients were referred to the study and 41 enrolled. As is displayed in the CONSORT diagram (Figure 1), 13 participants did not proceed to the baseline assessment due to not meeting study criteria or electing to withdraw from the study due to reasons outside of their control. In total, 28 participants initiated the baseline assessment, 27 were randomized, and 26 completed the study in its entirety (93% retention rate). For both participants who dropped out of the study, the participant and research team collaboratively agreed that withdrawal was indicated.

As is detailed in Table 1, study completers were 65% female, 88.5% White, and the average age was 44.46 ± 12.81 years. Participants’ average age at Crohn’s disease diagnosis was 29.18 ± 11.98 years and 53.8% had ever been hospitalized for Crohn’s disease. Participants were recruited from 3 states and lived an average distance of 64.04 ± 43.08 miles from the medical center. Social deprivation was variable, with 11.5% living in high deprivation areas, 46.2% in intermediate deprivation areas, and 42.3% in low deprivation areas. Participants were prescribed a variety of medications for Crohn’s disease and for sleep and/or mental health at baseline; 88.5% of participants were taking Crohn’s disease-specific medications and 42.3% of participants were taking medication for sleep and/or mental health. The presence/absence of medication did not differ significantly between groups. Post-treatment, with regard to Crohn’s disease treatments, 1 participant had discontinued their baseline Crohn’s disease medication and another had switched from ustekinumab to certolizumab pegol. With regard to sleep and mental health medications, 3 participants discontinued sleep/mental health medications (mirtazapine, clonazepam, and amitriptyline), 1 participant changed from no medication to zolpidem, and 1 participant added mirtazapine to their regimen (bupropion and trazadone). The CBT-I and waitlist groups did not differ significantly on baseline age, insomnia severity, Crohn’s disease symptom severity, pain intensity, or pain interference. Groups did differ significantly by gender (Fisher’s exact P = .04). As determined by an elevated CRP and/or fecal calprotectin 42.3% of participants (n = 11) evidenced active disease at baseline. Participants with and without biochemically active disease did not differ significantly in self-reported Crohn’s disease symptom severity (P = .36) or insomnia severity (P = .59) at baseline.

Table 1.

Participant demographics at baseline (N = 26).

	Full sample	CBT-I	Waitlist
Age, y	44.46 ± 12.81	42.86 ± 13.00	46.33 ± 12.89
Sex, %
Female	65.4	85.7	41.7
Male	34.6	14.3	58.3
Race/ethnicity, %
White	88.5	85.7	91.7
Asian	3.8	7.1	0
Biracial	7.7	7.1	8.3
Distance from medical center, miles	64.04 ± 43.08	68.14 ± 53.55	59.25 ± 27.96
Social Deprivation Index	30.31 ± 23.80	32.07 ± 23.39	28.25 ± 25.14
Age at Crohn’s diagnosis, y	29.18 ± 11.98	27.75 ± 10.38	30.90 ± 14.04
Crohn’s disease phenotype
Location
Ileal only	10	6	4
Ileocolonic	13	6	7
Colonic only	3 y	2	1
Behavior
Inflammatory	11	6	5
Stricturing	9	6	3
Penetrating	3	2	1
Stricturing and Penetrating	3	0	3
Perianal involvement
No	17	10	7
Yes	9	4	5
Crohn’s disease medications
Adalimumab	6	1	5
Azathioprine	1	0	1
Balsalazide	1	0	1
Budesonide	1	0	1
Infliximab	7	5	2
6-Mercaptopurine	2	2	0
Ustekinumab	8	6	2
Vedolizumab	2	1	1
None	3	2	1
Patient Reported Outcomes-3	23.32 ± 12.52	25.36 ± 13.53	20.00 ± 9.56
C-reactive protein, mg/L	2.73  ± 4.05	2.14  ± 2.61	3.39  ± 5.23
Fecal calprotectin, μg/g	223.92 ± 213.42	269.04 ± 219.42	171.29 ± 202.50
Ever hospitalized for Crohn’s, %
Yes	53.8	64.3	41.7
No	30.8	21.4	41.7
Missing	15.4	14.3	16.7
Medications for sleep and/or mood
Amitriptyline	2	2	0
Bupropion	1	0	1
Citalopram	1	1	0
Clonazepam	1	1	0
Escitalopram	1	1	0
Mirtazapine	2	1	1
Propranolol	1	1	0
Sertraline	2	2	0
Trazodone	2	1	1
Zaleplon	1	0	1
Zolpidem	1	1	0
None	15	6	9
Insomnia Severity Index	15.46 ± 4.58	16.64 ± 4.05	14.08 ± 4.94

Values are mean ± SD or n, unless otherwise indicated. Group differences in medication use were not assessed due to low numbers.

Abbreviation: CBT-I, cognitive behavioral therapy for insomnia.

^aSignificantly different at P < .05.

Study measure completion

All participants completed questionnaires at baseline and following the waitlist period. In follow-up periods, 1 or 2 participants did not complete questionnaires and 1 participant had only partial completion. At baseline, all participants completed at least 5 days of actigraphy data. Completion rates ranged from 66% to 73% at subsequent visits. Sleep profiler data were available for 85% of participants at baseline and 75% to 79% at subsequent visits. A total of 100% of participants completed blood draws at baseline and 92% to 100% completed follow-up blood draws at subsequent visits. A total of 100% of participants completed their initial and second baseline stool collection and 88% completed the final stool collection. Of import, no participant had missing data on all assessments at any time point, and most missing data were due to factors unrelated to the study (eg, death in the family).

Primary outcomes

Immediately postintervention, 58.3% of participants evidenced clinically meaningful improvement in their insomnia and 56% had achieved complete remission. At the 1-month follow-up visit, 69.6% of participants evidenced clinically meaningful improvement in their insomnia and 62.5% had achieved complete remission.

Linear mixed-effects models indicated that CBT-I resulted in significantly greater change in insomnia severity than the waitlist control, with the group-by-time interaction P < .001 (Table 2, Figures 2 and 3). Similarly, wake after sleep onset significantly reduced in the CBT-I group, with the group-by-time interaction P = .02. While sleep onset latency did reduce more in the CBT-I group compared with control, this difference did not rise to the level of significance (P = .10).

Table 2.

Results of linear mixed-effects models comparing pre- and postintervention sleep, pain, and inflammation.

	Pre–CBT-I	Post–CBT-I	1-mo follow-up	Model results
Primary outcomes
Insomnia severity (ISI)	15.32 ± 4.50	8.12 ± 6.88 −0.99 (−1.37 to −0.61)	7.42 ± 6.28 −1.05 (−1.45 to −0.66)	Group by time: F = 18.56, P < .001 Pre-post: F = 18.53, P < .001
Sleep onset latency (diary), min	32.47 ± 20.83	20.51 ± 18.58 −0.63 (−0.96 to −0.30)	16.44 ± 9.54 −0.69 (−1.04 to −0.35)	Group by time: F = 3.20, P = .09 Pre-post: F = 10.85, P < .001
Wake after sleep onset (diary), min	44.36 ± 33.95	17.71 ± 20.72 −0.89 (−1.28 to −0.50)	19.78 ± 18.78 −0.79 (−1.20 to −0.38)	Group by time: F = 6.31, P = .02 Pre-Post: F = 12.63, P < .001
Secondary outcomes (all pre– to post–CBT-I)
Total sleep time (diary), h	6.71 ± 1.14	7.16 ± 1.39 0.35 (0.10 to 0.60)	7.16 ± 1.28 0.29 (0.03 to 0.55)	F = 4.51, P = .02
Sleep efficiency (diary)	63.51 ± 28.50	74.82 ± 34.34 0.34 (0.24 to 0.44)	77.12 ± 32.38 0.35 (0.24 to 0.45)	F = 31.85, P < .001
Wake after sleep onset (actigraphy), min	77.98 ± 39.72	—	71 ± 36.31 −0.22 (−0.66 to 0.23)	F = 0.96, P = .34
Total sleep time (actigraphy), h	6.32 ± 0.83	—	6.27 ± 1.11 −0.06 (−0.49 to 0.37)	F = 0.07, P = .79
Sleep efficiency (actigraphy)b	82.83 ± 7.16	—	83.51 ± 8.19 0.10 (−0.26 to 0.46)	F = 0.33, P = .57
Sleep onset latency (sleep profiler), min	29.25 ± 36.20	—	19.99 ± 20.09 −0.19 (−0.81 to 0.42)	F = 0.42, P = .53
Wake after sleep onset (sleep profiler), min	38.07 ± 31.97	—	28.85 ± 19.85 −0.38 (−0.82 to 0.06)	F = 3.09, P = .09
Total sleep time (sleep profiler), h	6.1 ± 1.85	—	6.13 ± 1.73 0.01 (−0.37 to 0.39)	F = 0.0, P = .97
Sleep efficiency (sleep profiler)	83.69 ± 10.88	—	85.76 ± 9.22 0.12 (−0.45 to 0.68)	F = 0.17, P = .68
Slow wave sleep (stage 3), min	64.32 ± 34.37		80.65 ± 32.89 0.47 (−0.06 to 1.01)	F = 3.17, P = .09
Slow wave sleep (stage 3), % time	19.17 ± 13.07		25.51 ± 13.58 0.55 (0.17 to 0.92)	F = 8.80, P = .007
Crohn’s symptom severity (PRO-3)	23.87 ± 12.26	19.88 ± 13.78 −0.30 (−0.54 to −0.05)	17.05 ± 13.16 −0.37 (−0.63 to −0.12)	F = 4.88, P = .01
Pain intensity (BPI [past week])	4.04 ± 2.10	3.14 ± 2.16 −0.45 (−0.78 to −0.11)	3.1 ± 2.17 −0.41 (−0.76 to −0.06)	F = 4.15, P = .02
Pain interference (BPI [past week])	4.02 ± 2.93	2.56 ± 2.26 −0.56 (−0.97 to −0.15)	2.5 ± 2.48 −0.61 (−1.04 to −0.17)	F = 5.05, P = .01
Cold pressor test max pain (0-10)	7.92 ± 2.0	—	8.42± 1.79 0.29 (−0.01 to 0.58)	F = 3.89, P = .06
Cold pressor test time submerged, s	100.83 ± 35.55	—	100.17 ± 32.32 −0.10 (−0.40 to 0.20)	F = 0.44, P = .51
C-reactive protein, mg/L	4.42 ± 6.96	—	2.2 ± 3.27 −0.17 (−0.38 to 0.05)	F = 2.45, P = .13
Fecal calprotectin, μg/ga	210.38 ± 193.87	—	199.83 ± 180.77 −0.08 (−0.22 to 0.07)	F = 1.12, P = .30
Exploratory outcomes (all pre– to post–CBT-I)
Awakenings (diary)	2.48 ± 1.54	1.56 ± 1.33 −0.58 (−0.84 to −0.32)	1.77 ± 1.52 −0.47 (−0.74 to −0.19)	F = 11.32, P < .001
Nighttime bowel movements (diary)	0.32 ± 0.57	0.16 ± 0.36 −0.35 (−0.69 to 0.00)	0.16 ± 0.45 −0.31 (−0.67 to 0.06)	F = 2.38, P = .10
Fatigue (PROMIS short form)	62.82 ± 6.35	54.8 ± 11.5 −0.74 (−1.09 to −0.38)	53.9 ± 10.33 −0.77 (−1.14 to −0.40)	F = 11.44, P < .001
Pain catastrophizing	14.92 ± 13.49	10.96 ± 10.59 −0.33 (−0.63 to −0.03)	10.64 ± 12.94 −0.40 (−0.72 to −0.09)	F = 3.87, P = .03
Anxiety (GAD-7)	8.44 ± 6.29	6.04 ± 5.54 −0.41 (−0.72 to −0.09)	5.41 ± 5.52 −0.59 (−0.93 to −0.26)	F = 6.72, P = .004
Depression (PHQ-9)	9.84 ± 5.31	6.52 ± 5.63 −0.58 (−0.88 to −0.29)	5 ± 5.61 −0.88 (−1.19 to −0.57)	F = 16.89, P < .001

Values are mean ± SD or standardized coefficient (95% confidence interval). Only primary outcomes were evaluated comparing cognitive behavioral therapy for insomnia with the waitlist period (group by time interaction). Primary, secondary, and exploratory outcomes were evaluated within the CBT-I period, comparing preintervention values to immediately postintervention values and 1-month postintervention values. Daily diary values are self-reported and actigraphy and sleep profiler values are objective measures of sleep continuity and sleep architecture.

Abbreviations: BPI, Brief Pain Inventory; CBT-I, cognitive behavioral therapy for insomnia; GAD-7, Generalized Anxiety Disorder-7; ISI, Insomnia Severity Index; PHQ-9, Patient Health Questionnaire-9; PRO-3,  Patient-Reported Outcomes-3; PROMIS, Patient-Reported Outcomes Measurement Information System.

^aUnadjusted model was a better fit.

Figure 2.

Changes in insomnia severity following either cognitive behavioral therapy for insomnia (CBT-I) or waitlist control (WLC). Change in insomnia severity was significantly different between groups (P < .001).

Figure 3.

Changes in diary-based sleep onset latency and wake after sleep onset following either cognitive behavioral therapy for insomnia (CBT-I) or waitlist control (WLC). Change in sleep onset latency trended toward differing between groups (P = .09) and change in wake after sleep onset was significantly different between groups (P = .02).

Unsurprisingly, insomnia severity, sleep onset latency, and wake after sleep onset also improved significantly when evaluated in the combined pre– to post–CBT-I analyses. Effect sizes for improvement in insomnia symptom severity immediately post-treatment and at follow-up were large (0.99 and 1.05, respectively), and effect sizes for diary values of sleep onset latency and wake after sleep onset were medium to large immediately and at follow-up (0.63-0.89). While models were adjusted by sex, age, and presence or absence of active inflammation at baseline, none of these covariates were significant (Ps ranged from .23-.84).

Secondary outcomes

All secondary outcomes were investigated using the combined group to assess for changes pre– to post–CBT-I. Diary-based sleep efficiency and total sleep time both improved significantly (P < .001 and P = .02, respectively), while most objective measures did not significantly change; only polysomnography-based wake after sleep onset neared significance (P = .09). Age was a significant covariate in the polysomnography-based wake after sleep onset model (P < .001) and actigraphy-based total sleep time model (P = .02). Sex was significant in the polysomnography-based models for sleep efficiency (P = .09) and total sleep time (P < .001).

Self-reported pain intensity, pain interference, and Crohn’s disease symptoms all improved significantly over the course of CBT-I (Ps ranged from .01 to .02). Effect sizes were in the small-to-medium range immediately after treatment (0.30-0.56) and 1 month later (0.37-0.61). Tolerance of pain or discomfort as measured by the cold pressor test did not change significantly; maximum pain rating increased nonsignificantly (P = .06), and submersion time was nearly identical pre- to post-treatment (P = .51). Time spent in slow wave sleep increased significantly with intervention as measured by percent time in slow wave sleep (P = .007) and trended toward significance in number of minutes spent in slow wave sleep (P = .09; increase of 16.33 minutes). CRP did not change significantly, though it trended toward significance (P = .13), and fecal calprotectin did not change significantly (P = .30). Finally, reductions in CRP were partially explained by increased percent time spent in slow wave sleep (P = .02); of note, sleep was modeled nonlinearly; moderate increases in slow wave sleep were associated with the largest CRP reductions.

Exploratory outcomes

As is also included in Table 2, awakenings during the night improved significantly during CBT-I (P < .001), and there was a trend for reduced nighttime bowel movements (P = .10). With regard to psychosocial factors, fatigue (P < .001), pain catastrophizing (P = .03), anxiety (P = .01), and depression (P < .001) all improved significantly following treatment. In all cases, effect sizes increased between the immediate and 1-month follow-up assessments.

Discussion

This pilot trial sought to investigate the impact of CBT-I on insomnia symptoms, pain, Crohn’s disease symptoms, psychosocial concerns, and objective measures of sleep and inflammation. Primarily, over the course of 1 year, we were able to successfully enroll 30 participants. Of the enrolled participants, 2 withdrew prior to their assessment due to factors unrelated to the study, 28 completed a baseline assessment, and 26 of 28 completed the full study. This high retention rate (93%) is consistent with research indicating people with Crohn’s disease are interested in insomnia treatment.⁵ Further, not only did participants complete most study measures, but also 25 of 28 participants completed all 5 treatment sessions, supporting the ability of people with Crohn’s disease to fully engage in insomnia treatment despite ongoing symptoms.

Results indicated that in people with self-reported active Crohn’s disease symptoms, CBT-I resulted in significantly better improvements in self-reported sleep continuity than a waitlist control. Specifically, participants in the CBT-I group reduced their time to fall asleep (sleep onset latency) by about 22 minutes, compared with 6 minutes in the waitlist group, and time awake during the night (wake after sleep onset) by about 30 minutes, compared with 3 minutes in the waitlist control group. The overall perceived insomnia severity also reduced significantly with CBT-I, while it was nearly unchanged during the waitlist period. In analyses comparing pre- and post-treatment, participants reported significant changes in sleep continuity, pain intensity, pain interference, Crohn’s disease symptoms, and psychosocial outcomes, including anxiety, depression, pain catastrophizing, and fatigue. Percent time spent in slow wave sleep increased, and there was a trend toward improvement in CRP, though other objective measures of sleep and inflammation did not evidence significant change. All of these changes were sustained 1 month after treatment ended. In addition, over the course of the trial 7.7% of participants made changes to Crohn’s disease medications and 19.2% made changes to sleep/mental health medications. As some participants discontinued medications and others added medication, it is unlikely that these changes significantly affected outcomes.

Broadly, our results are consistent with the extant literature. Meta-analytic evidence demonstrates that both psychiatric and chronic medical illness populations experience significant improvements in insomnia following CBT-I⁴⁷; while effects of CBT-I are sustained at least a year post-treatment, they may decline in potency over time.⁴⁸ In addition, research in people with other chronic, painful conditions has demonstrated that improvements in insomnia may result in improved pain severity, pain interference, and pain catastrophizing.^49-52 Importantly, while CBT-I does not often result in improvements in total sleep time in the short term, in this study and in our prior feasibility study in IBD,²⁰ total sleep time did increase significantly. This particular change may contribute to the self-­reported improvements in pain, as we have previously demonstrated that short-term improvements in total sleep time during CBT-I are particularly important for experiencing long-term reductions in chronic pain.⁵³ However, as we only analyzed pain pre- to post-treatment, it is also important to consider the possibility that changes in pain-related variables were a product of time and not the intervention itself.

One pathway by which insomnia associated sleep disturbances can augment clinical pain is by reducing inflammation.^10,54-56 Meta-analytic research suggests broadly that CBT-I may reduce CRP,57 and in IBD, one longitudinal study demonstrated that insomnia symptoms were predictive of CRP 1 year later for women, though not for men.⁵⁸ Herein, while CRP improved, the overall changes were nonsignificant. However, increased time spent in slow wave sleep significantly predicted decreases in CRP. Prior insomnia intervention trials have demonstrated improvements in slow wave sleep,^59,60 while experimental and observational research supports the relationship between slow wave sleep and CRP.^61,62 Thus, it is likely that CRP changes were contingent upon CBT-I–specific mechanisms. As our population had a low mean CRP at baseline, future research in people with higher CRP or other biochemical inflammatory markers at baseline is needed to better understand this phenomenon.

We also did not observe changes in objective measures of sleep continuity, including actigraphy and most electroencephalography-based sleep architecture assessments. This pattern is consistent with the literature; a recent meta-analysis indicated that CBT-I does not reliably improve actigraphic or polysomnography-based sleep parameters,⁶³ We have previously demonstrated that sleep was most closely related to abdominal pain in diary data, while disease activity was the most consistent predictor of abdominal pain when using actigraphic data.¹³ As actigraphy-measured sleep continuity problems may increase with disease activity,⁶⁴ it is also possible that the results would be different in a population with more severely active disease. However, as insomnia is a diagnosis made based on patients’ subjective experiences of sleep, and not objective criteria, improvement in self-report measures can be considered more important than objective measures.

Taken together, our findings support continued investigation into and treatment of insomnia in IBD. We have previously demonstrated that 70% of patients with IBD want their IBD providers to ask about sleep, and most patients with insomnia symptoms are interested in seeking treatment.⁸ Thus, it is likely that sleep-related questions and recommendations will be well received. In addition, results of this study suggest that insomnia treatment may also reduce pain and fatigue in Crohn’s disease. As people with IBD report that pain and fatigue impact their quality of life more than bowel incontinence,⁶⁵ and want help with these symptoms,⁶⁶ it is important to capitalize on treatments that can improve these aspects of IBD.

Broadly, this pilot study was an important step toward better understanding the ways in which insomnia treatment can impact Crohn’s disease. Strengths of this study include a well-characterized population, comprehensive, multinight assessment of objective and self-reported sleep parameters and objective assessment of disease activity, and engagement of participants from a wide area outside of the medical center. However, this research is not without limitations. First, while our population was diverse from a social determinants of health perspective and consistent with the greater population that our medical center serves, it was not racially or ethnically diverse. Given the possible race and ethnicity related differences in sleep patterns,⁶⁷ it is important that subsequent studies enroll a more diverse population to support generalizability. Second, while CBT-I is the gold-standard treatment for insomnia, it can be resource and time intensive. While virtual delivery and expanding interstate psychology practice authorizations (ie, PSYPACT) reduce barriers, many people with IBD still may not have access to a provider trained in this intervention. Research that investigates the most potent components of this intervention so that fewer sessions are needed, or which explores novel delivery mechanisms (eg, nurse led, group based) may be the next step in disseminating this treatment more widely. Finally, due to the pilot nature of this work, CBT-I was compared with a waitlist as opposed to an attentional control that included interventionist contact. In other areas of treatment (eg, depression), waitlists have performed worse than care as usual.⁶⁸ Thus, future research comparing CBT-I to a nonspecific control intervention is warranted both to validate our insomnia findings and further investigate the impact of CBT-I on pain and inflammation.

Conclusions

This pilot study demonstrated that CBT-I improves insomnia severity, sleep onset latency, and wake after sleep onset significantly better than a waitlist control. Analyses comparing pre-treatment and post-treatment also suggested that Crohn’s disease symptoms, pain intensity, pain severity, pain catastrophizing, fatigue, anxiety, and depression may improve during treatment. Future research investigating these relationships in a larger, more diverse sample is warranted to fully realize the impacts of sleep treatment in Crohn’s disease.

Author Contributions

J.K.S.-D. designed this trial with input from C.A.S., M.T.S, and K.A.A. J.K.S.-D. and S.J.W. were responsible for data acquisition and S.J.W. analyzed the data. All authors contributed to interpretation of the data. J.K.S.-D. and S.J.W. drafted the initial manuscript and C.A.S., M.T.S., and K.A. revised it for important intellectual content. All authors have approved the final version of the manuscript, including the authorship list.

Funding

This work was supported by the Crohn’s & Colitis Foundation (Litwin IBD Pioneers Award 828830 to J.K.S-D.) and the National Institute of Diabetes and Digestive and Kidney Diseases (K23DK134814 to J.K.S-D.). J.K.S.D. received support for this research from Buhlmann Diagnostics.

Conflicts of Interest

None declared.

Data Availability

The data underlying this article cannot be shared publicly due to the privacy of individuals that participated in the study. The data will be shared on reasonable request to the corresponding author.