The Survivorship Sleep Program (SSP): A Synchronous, Virtual Cognitive Behavioral Therapy for Insomnia Pilot Program among Cancer Survivors

Abstract

Background:

For cancer survivors, insomnia is prevalent, distressing, and persists for years if unmanaged. Cognitive behavioral therapy for insomnia (CBT-I) is an effective treatment yet can be difficult to access and may require modification to address survivorship-specific barriers to sleep. This two-phase study adapted and assessed the feasibility, acceptability, and preliminary effects of synchronous, virtual CBT-I adapted for cancer survivors (Survivorship Sleep Program; SSP).

Methods:

From April-August 2020, cancer survivors with insomnia (N=10) were interviewed to refine SSP content and delivery. From October 2020-March 2021, 40 survivors were recruited for a randomized controlled trial comparing four, weekly SSP sessions to enhanced usual care (EUC; CBT-I referral plus a sleep hygiene handout). Feasibility and acceptability were assessed by enrollment, retention, attendance, fidelity, survey ratings, and exit interviews. Insomnia severity (secondary outcome), sleep quality, sleep diaries, and fatigue were assessed at baseline, post-intervention, and 1-month follow-up using linear mixed models.

Results:

The SSP included targeted content and clinician-led, virtual delivery to enhance patient-centeredness and access. Benchmarks were met for enrollment (56% enrolled/eligible), retention (SSP: 90%, EUC: 95%), attendance (100%), and fidelity (95%). Compared to EUC, the SSP resulted in large, clinically significant improvements in insomnia severity (d=1.19), sustained at one month follow-up (d=1.27). Improvements were observed for all other sleep metrics, except sleep diary-total sleep time and fatigue.

Conclusions:

Synchronous, virtually delivered CBT-I targeted to cancer survivors is feasible, acceptable, and seemingly efficacious for reducing insomnia severity. Further testing in larger and more diverse samples is warranted.

Precis for use in the Table of Contents:

This two-phase study adapted and assessed the feasibility, acceptability, and preliminary effects of synchronous, virtual cognitive behavioral therapy for insomnia adapted for cancer survivors compared to an enhanced usual care control. The adapted intervention was feasible, acceptable, and associated with large, clinically meaningful reductions in insomnia severity, sustained at one-month follow-up.

1. Introduction

Insomnia affects 30-50% of survivors of various cancers^1,2. Characterized as difficulty falling asleep, staying asleep, or waking, insomnia is associated with worsened mood, social and work functioning, and poorer overall quality of life for cancer survivors^2,3. Insomnia may also result in greater illness burden, increased fatigue, poorer immune functioning, and reduced survival^2-5. Cognitive behavioral therapy for insomnia (CBT-I) is efficacious for insomnia^6-9, and is recommended as a first line treatment¹⁰. Core components of CBT-I include stimulus control and sleep schedule modifications intended to increase sleep ability and reduce conditioned arousal; by reducing excess time in bed and aiming to minimize non-sleep related behaviors from occurring in the bedroom, the bed/bedroom will be more likely to elicit sleep. Supplementary components include sleep hygiene, reframing unhelpful thoughts about sleep, and promoting relaxation to help prepare the body and mind for rest. This therapy is typically delivered in one intake (60-90 min) and 4-10 follow-up sessions (30-60 min each)¹¹. Compared to pharmacotherapy, the effects of CBT-I are more durable because patients learn skills to address current and future insomnia episodes^6,12.

Among cancer survivors, meta-analyses of CBT-I have demonstrated moderate improvements across several aspects of sleep (i.e., insomnia severity, sleep quality, sleep diary metrics), yet challenges with retention and adherence to skills^13,14. Thus, while CBT-I for cancer survivors is promising, further adaptation is needed and may require identifying and targeting population-specific needs¹⁵. Insomnia in cancer survivors is perpetuated and exacerbated by a variety of mechanisms, including discomfort due to treatment-related late effects (e.g., pain, hot flashes), arousal due to physiological (e.g., circadian rhythm) and psychological factors (e.g., fear of recurrence, worry), and changes in lifestyle behaviors (e.g., sleep schedule disruptions, poor sleep hygiene). Several of these factors are modifiable using problem-solving and CBT techniques, and thus suitable targets in interactive interventions¹⁶.

In-person CBT-I is not accessible for many cancer survivors due to time limitations, travel, and illness burden constraints^17,18. Telehealth delivery of CBT-I can address these barriers using a wide variety of formats^19-21. Asynchronous formats, such as self- and video-delivered programs, are efficacious yet produce smaller effect sizes as compared to synchronous modalities, possibly due to less personalization and lower rates of session attendance²²10/28/2021 2:39:00 PM. By contrast, synchronous, virtual CBT-I is a promising option for increasing access and personalized tailoring of skills that has been found to be non-inferior to in-person delivery in non-cancer populations²³. For these reasons, there have been repeated calls for synchronous, virtually-delivered CBT-I for cancer survivors^{24 25}. Yet to our knowledge, no trials have tested synchronous, virtual CBT-I targeted to cancer survivors.

We conducted a two-phase study to (1) adapt CBT-I through interviews with cancer survivor stakeholders and (2) conduct a pilot randomized controlled trial (RCT) testing the feasibility, acceptability, and preliminary effect of the adapted intervention (Survivorship Sleep Program; SSP) compared to enhanced usual care (EUC; referral for CBT-I plus sleep hygiene handout) on insomnia severity at post-intervention and 1-month follow-up. We hypothesized that the SSP would be feasible, acceptable, and lead to significantly lower insomnia severity post-intervention as compared to EUC. Exploratory outcomes assessed changes in subjective sleep quality, sleep diaries (i.e., sleep efficiency, time in bed, total sleep time, sleep onset latency, wake after sleep onset), and fatigue.

2. Methods

2.1. Participants

Eligible participants for both study phases were recruited from an academic medical center and had a history of non-metastatic, localized, or regional solid or blood malignancy(ies), completed primary cancer treatment (i.e., radiation, surgery, and/or chemotherapy), endorsed chronic insomnia (indicated by DSM-5 criteria and Insomnia Severity Index score ≥ 15 at screening), and were at least 18 years old. Survivors who were on hormonal and/or other long-term maintenance therapy agents were eligible to participate. Survivors were excluded if they were unable to speak and write in English, were unwilling or unable to discontinue night shift work, or had an undertreated non-insomnia sleep disorder (i.e., sleep apnea) or epilepsy, undertreated serious mental illness or suicidality, or had a psychiatric hospitalization in the past year.

2.2. Recruitment and Screening

Survivors were recruited for the Phase 1 qualitative study between April 24, 2020, and August 7, 2020. Survivors were recruited for the Phase 2 RCT between October 1, 2020, and March 24, 2021. Survivors were identified through provider referral, MassGeneralBrigham Rally (a public research recruitment portal), proactive electronic health record screening, and the Massachusetts General Hospital (MGH) Behavioral Sleep Medicine clinic waitlist. If survivors were identified as eligible through electronic health record screening, study staff requested permission from their oncology clinician before pursuing outreach. Identified survivors completed phone screening to determine eligibility. Eligible survivors were asked to complete an electronic informed consent. All procedures were approved and monitored by the Dana Farber Harvard Cancer Center Institutional Review Board and conformed to Declaration of Helsinki ethical standards. Survivors in Phase 1 received $15 for completing a 30-minute interview. Survivors in Phase 2 received $25 per assessment and were offered a list of insomnia treatment resources at the end of the study.

2.3. Procedure

Phase 1: Qualitative Study to Inform Intervention Adaptation

Cancer survivors with insomnia were interviewed to identify the perceived needs and preferences for synchronous, virtual CBT-I. Survivors were asked to (1) describe factors that impact their sleep quality, (2) provide feedback on targeting session content for survivors, and (3) share their preferences for virtual delivery. Interviews were recorded and transcribed.

Phase 2: Randomized controlled trial

Cancer survivors with insomnia (N = 40) were enrolled in a prospective, single site, randomized controlled trial to examine the feasibility, acceptability, and preliminary effects of the SSP compared to EUC. The trial was registered on ClinicalTrials.gov (NCT04566068).

Assessments

Survivors completed surveys (see Measures) at baseline (T0), post-intervention (T1), and one-month follow up (T2). At each assessment, survivors also completed seven days of sleep diaries.

Randomization

After completing the baseline survey, survivors were randomized to either the SSP or EUC through a computer-generated randomization scheme using randomly permuted block sizes of two and four. Survivors were not blinded to their allocation.

Survivorship Sleep Program (SSP)

Table 1 provides a detailed overview of the SSP intervention content. Sessions were modeled after an evidence-based CBT-I protocol¹¹, but adapted with refined content and delivery plans based on findings from the Phase 1 qualitative study (see Results, Phase 1 Qualitative Study). For example, in each session cancer survivors were asked to consider personal experiences with cancer survivorship (i.e., late effects of treatments affecting sleep, sleep-interfering worries about recurrence). To balance survivors’ preferences for a brief number of sessions with opportunities for frequent feedback, we adopted the recommended minimum number of sessions and average session length used previously¹¹ and added brief between-session check-ins that focused on assessing and problem-solving barriers to applying skills and adhering to sleep prescription. The final protocol consisted of four, weekly synchronous sessions (approximately 45 min each) plus three between-session check-ins (approximately 15 min each), delivered via secure videoconferencing technology. Survivors not familiar with telehealth were given the opportunity to complete a brief test call with study staff prior to starting the intervention.

Table 1.

Overview of Session Content: Adapted Cognitive Behavioral Therapy for Insomnia (CBT-I) for Cancer Survivors

Week	Session	Content	Length	Handouts
1	1	Welcome: Building rapport, Tips for virtual sessions, Overview of intervention format, Confidentiality Assessment: Review sleep history and cancer-related factors affecting sleep Education: Sleep diary Sleep drive and circadian rhythm 4 Factor model Skill: Stimulus Control Skill: Efficient Sleep Develop sleep prescription Identify and problem-solve potential barriers to adherence	45min	1.1 Tips for Sleep Diary 1.2 4 Factor Model 1.3 Stimulus Control 1.4 Efficient Sleep 1.5 Things to do if Awake
1	Check in #1	Review sleep diary + problem-solve barriers	15min
2	2	Identify and problem-solve barriers to adherence Review sleep diary Titrate sleep prescription Skill: Sleep Hygiene Skill: Relaxation	45min	2.1 Sleep Hygiene 2.2 Relaxation
2	Check in #2	Review sleep diary + problem-solve barriers	15min
3	3	Identify and problem-solve barriers to adherence Review sleep diary Titrate sleep prescription Skill: Cognitive Techniques	45min	3.1 Thought Record 3.2 Practice Reframing Thoughts 3.3 Worry Time
3	Check in #3	Review sleep diary + problem-solve barriers	15min
4	4	Identify and problem-solve barriers to adherence Review sleep diary Titrate sleep prescription Education: Insomnia Relapse Prevention Next steps in the study	45min	4.1 Personal Plan

Note: Outline adapted from Perlis, Jungquist, Smith, Posner (2005). Cognitive Behavioral Treatment of Insomnia: A Session-by-Session Guide.

A facilitator’s guide and participant workbook were developed to promote delivery fidelity and participant engagement. Facilitators (TJC, MG) were PhD-level psychologists with three or more years of experience delivering CBT-I. Facilitators received two days of training on the adapted content and study procedures and participated in weekly supervision meetings.

Enhanced usual care (EUC)

Survivors randomized to EUC were emailed an electronic handout with sleep hygiene tips and referral information for CBT-I through the MGH Behavioral Sleep Medicine Clinic.

Exit Interviews

Individual exit interviews (~30min) were conducted with survivors who participated in the SSP. Survivors rated the intervention acceptability in terms of enjoyability, convenience, helpfulness, likelihood of continued use of skills learned, and overall satisfaction on a scale from 1 (Not at all) to 5 (Very high). Open-ended response probes asked survivors to identify aspects that were most/least acceptable. Interviews were recorded and transcribed.

Measures

Sociodemographic and medical characteristics

Sociodemographic characteristics were self-reported. Medical characteristics were self-reported and verified by medical record, including physical and psychiatric co-morbidities, medication use, date(s) of cancer diagnoses, cancer treatment history, and date(s) of primary cancer treatment completion. Risk for obstructive sleep apnea was assessed using the 10-item Berlin questionnaire.²⁶

Primary outcomes

Feasibility and acceptability:

Feasibility was measured using a combination of factors and benchmarks determined a priori, including: the proportion of survivors screened and determined to be eligible who enrolled in the trial (≥ 50%), retention at each assessment timepoint (≥ 70%), and session attendance (≥ 70%). Reasons for ineligibility, refusal, or drop out were measured. Facilitator fidelity was assessed for 20% of sessions (randomly selected after stratifying by facilitator and session number) using a checklist. Acceptability was assessed among survivors in the SSP arm using quantitative ratings [overall satisfaction, enjoyableness, convenience, helpfulness, and odds of future use of skills, from a scale from 1 (very low) to 5 (very high)] and in qualitative exit interviews probing for aspects that were most/least acceptable.

Secondary outcome

Insomnia severity:

Insomnia severity was assessed at all timepoints using the validated Insomnia Severity Index questionnaire (ISI; clinical cutoff = 15)^13,27. ISI reductions of > 8 points are considered clinically meaningful^28,29.

Exploratory outcomes

Sleep quality:

Sleep quality over the past month was assessed using the “Sleep Quality” item from the Pittsburgh Sleep Quality Index³⁰ (PSQI; item 6), scored from 0 (Very good) to 3 (Very bad).

Sleep diary:

Seven nights of sleep diary data were collected at each timepoint. Data were collected using a REDCap electronic sleep diary adapted from the Consensus Sleep Diary^31-33. Diaries captured several sleep metrics, including time in bed (TIB), total sleep time (TST), sleep efficiency (SE; amount of time spent sleeping divided by amount of time spent in bed), sleep onset latency (SOL), and wake after sleep onset (WASO). For each metric, the mean score at each timepoint was calculated and used in analyses.

Fatigue:

Fatigue was assessed by summing 13 items (scored 0 to 10) on the Fatigue Symptom Inventory³⁴ (FSI) to create a total score, with higher scores reflecting more fatigue severity and interference.

2.5. Statistical analyses

Quantitative analyses were conducted using SPSS (v25), and qualitative analyses were conducted using NVivo 12. For Phase 1, transcriptions underwent directed content analysis to identify preferences and concerns related to each CBT-I skill and plans for virtual delivery. Codes were initially established and then derived independently by two coders (BJ, HM) with a third coder (DH) for resolving discrepancies.

For the Phase 2 RCT, feasibility calculations included: the ratio of screened, eligible survivors who enrolled, rates of retention at each timepoint, and percentage of intervention sessions attended. Treatment fidelity was computed as a percentage of topics covered and an inter-rater reliability kappa (% agreement between independent rater and facilitators’ post-session checklists). Durations of the sessions and check-ins were computed. Acceptability ratings were analyzed with descriptive statistics. To identify aspects of the intervention that were most and least acceptable, transcriptions of exit interviews underwent content analysis by two coders (RL, HM) with a third coder (DH) for resolving discrepancies.

Sleep outcomes were analyzed using intent-to-treat with separate general linear mixed models to identify preliminary effects of the intervention across all three timepoints (T0 - T2). Within linear mixed models, missing data were handled using maximum likelihood (ML), which incorporates information from all randomized participants (N=40). Sleep diaries with impossible values (i.e., TST > TIB) were flagged and analyzed as missing (3.5%), prior to computing mean scores. To examine the effect of intervention condition on change in sleep outcomes over time, two random intercept models were conducted. Across models, time was treated as a 3-level categorical variable (baseline = T0, post-intervention = T1, 1-month follow-up = T2) to obtain estimated marginal means and standard errors of sleep variables by condition at each timepoint. Model 1 included T0 as the timepoint reference and EUC as the treatment group reference to produce separate contrasts between T0 - T1 and T0 - T2 for SSP vs. EUC participants. Model 2 was identical to Model 1 but included T2 as the timepoint reference to evaluate changes from T1 to T2. Time, treatment condition, and time x treatment condition were included as fixed effects, with magnitudes calculated as Partial r² values. Cohen’s d values of between-group effects were calculated using estimated marginal means at T1 and T2.

3. Results

3.1. Phase 1 Qualitative Study

Ten cancer survivors with insomnia participated in Phase 1 (demographic and clinical characteristics summarized in Appendix Table 1). Twelve out of 45 (27%) survivors screened were eligible, and 10/12 (83%) eligible survivors participated in interviews. As summarized in Appendix Figure 1, survivors voiced multiple preferences for making CBT-I skills more salient to cancer survivors (e.g., providing education about sleep and cancer; problem-solving tips for managing late effects of treatments interfering with sleep, such as hot flashes and incontinence; tailoring cognitive restructuring to address fears of cancer recurrence and scheduling “Worry Time” to limit interference of cancer-related worries on sleep), as well as considerations for remote CBT-I delivery (e.g., synchronous sessions, brief number of sessions with check-ins). Feedback was reviewed with the study team and integrated into final intervention delivery plans and materials (see Table 1).

3.2. Phase 2 RCT Primary Outcomes

Sociodemographic and medical characteristics of enrolled survivors are summarized in Table 2. Survivors were on average 53.2 years old (SD = 14.1 years) and were predominantly female, non-Hispanic, white, married or living as married, highly educated, employed, and had access to private insurance. On average, survivors were 93.5 months (7.8 years) since their initial cancer diagnosis and 61.6 months (5.1 years) since completing cancer treatment. Survivors had been treated for a variety of cancers, most commonly breast, hematological, thyroid, prostate, and skin. The intervention conditions did not differ on the majority of sociodemographic or medical variables. However, compared to survivors randomized to the SSP, survivors randomized to EUC were more likely to be employed (p < .001), have private insurance (p < .01), and be married/living as married (p = .08). Thus, preliminary efficacy analyses were conducted unadjusted and adjusting for these factors.

Table 2.

Characteristics of Participants in the Pilot Randomized Controlled Trial (N=40)

Characteristic	Total	CBT-I	EUC	Test Statistic	P
Sociodemographic Characteristics
Age (years), M (SD)	53.2 (14.1)	51.2 (17.1)	55.3 (10.3)	t = −.91	.37
Gender (female)	34 (85.0%)	18 (90.0%)	16 (80.0%)	X2 = .78	.38
Ethnicity (Hispanic)	5 (12.5%)	3 (15.0%)	2 (10.0%)	X2 = .23	.63
Race*				X2 = .29	.86
White	34 (85.0%)	18 (90.0%)	16 (80.0%)
Latinx	4 (10.0%)	2 (10.0%)	2 (10.0%)
Asian	2 (5.0%)	1 (5.0%)	1 (5.0%)
Africana	1 (2.5%)	0 (0.0%)	1 (5.0%)
Married/Living as married	29 (72.5%)	12 (60.0%)	17 (85.0%)	X2 = 3.13	.08
Number in household, M (SD)	2.7 (1.2)	2.6 (1.2)	2.9 (1.1)	t = .82	.42
Completed college	33 (82.5%)	16 (80.0%)	17 (85.0%)	X2 = .17	.68
Employed (at least part-time)	21 (52.5%)	5 (25.0%)	16 (80.0%)	X2 = 12.13	< .001
Private insurance	30 (75.0%)	11 (55.0%)	19 (95.0%)	X2 = 8.53	< .01
Medical Characteristics
Cancer type*				X2 = 5.85	.83
Breast	26 (65.0%)	13 (65.0%)	13 (65.0%)
Hematological	5 (12.5%)	2 (10.0%)	3 (15.0%)
Thyroid	4 (10.0%)	3 (15.0%)	1 (5.0%)
Prostate	3 (7.5%)	1 (5.0%)	2 (10.0%)
Skin	3 (7.5%)	1 (5.0%)	2 (10.0%)
Bladder	2 (5.0%)	1 (5.0%)	1 (5.0%)
Ovary	2 (5.0%)	1 (5.0%)	1 (5.0%)
Lung	1 (2.5%)	1 (5.0%)	0 (0.0%)
Testicular	1 (2.5%)	1 (5.0%)	0 (0.0%)
Urethra	1 (2.5%)	0 (0.0%)	1 (5.0%)
Uterus	1 (2.5%)	0 (0.0%)	1 (5.0%)
Time since diagnosis (months), M (SD)	93.5 (91.7)	92.5 (112.0)	94.5 (68.7)	t = .07	.94
Time since treatment (months), M (SD)	61.6 (61.5)	53.0 (66.4)	70.1 (56.5)	t = .88	.39
Cancer treatment type*				X2 = 2.80	.83
Surgery	37 (92.5%)	18 (90.0%)	19 (95.0%)
Radiation therapy	23 (57.5%)	13 (65.0%)	10 (50.0%)
IV chemotherapy/other IV cancer therapy	22 (55.0%)	11 (55.0%)	11 (55.0%)
Hormonal or endocrine therapy	18 (45.0%)	10 (50.0%)	8 (40.0%)
Oral chemotherapy/other oral therapy	7 (17.5%)	3 (15.0%)	4 (20.0%)
Complementary or alternative therapy	3 (7.5%)	2 (10.0%)	1 (5.0%)
Other	2 (5.0%)	2 (10.0%)	0 (0.0%)
Comorbid medical illness (1+)	22 (55.0%)	9 (45.0%)	13 (65.0%)	X2 = 1.62	.20
Comorbid psychiatric illness (1+)	16 (40.0%)	7 (35.0%)	9 (45.0%)	X2 = .42	.52
COVID-19 diagnosis	2 (5.0%)	0 (0.0%)	2 (10.0%)	X2 = 2.11	.15
Sleep Characteristics
Insomnia severity (ISI), M (SD)	18.0 (4.2)	18.1 (4.6)	17.9 (3.8)	t = −.19	.85
Sleep quality (PSQI Item 6), M (SD)	2.1 (0.7)	2.3 (0.6)	2.0 (0.8)	t = −1.35	.18
Fatigue (FSI), M (SD)	56.8 (28.4)	55.7 (30.4)	58.0 (26.9)	t = .26	.80

Note.

^*Participants could select more than one response.

3.2.1. Feasibility

As shown in Figure 1, 71/127 (56%) of survivors screened were eligible, and 40/71 (56%) of those who were eligible enrolled in the trial. The chief reason for ineligibility (44/56) was having a non-elevated insomnia severity score. Retention across study assessments was high at T1 (95% in SSP and EUC arms) and T2 (90% in SSP, 95% in EUC). Missingness on sleep outcomes varied by timepoint, ranging from 0% to 15% (wake after sleep onset at T2). One survivor in the SSP arm withdrew after completing T0 assessments but prior to initiating sessions, citing time constraints. One survivor in the EUC arm withdrew from the study after completing the T0 assessments, citing dissatisfaction with their study allocation.

Figure 1.

CONSORT diagram of cancer survivors’ participation in the trial and completion of study assessments.

Note. T0 = Baseline, T1 = Post-Intervention, T2 = 1-month follow-up.

Among the 19 survivors who initiated the SSP, 100% of sessions and check-ins were attended. Due to scheduling challenges, one survivor received a combined Session 3 and 4. When survivors anticipated a scheduling conflict, facilitators offered survivors the possibility of rescheduling. Chief reasons for rescheduled sessions were conflicts with medical appointments, caregiving roles (e.g., childcare, caring for family members), and interference from late effects of cancer treatments (e.g., fatigue). Fidelity was high, with 95% of reviewed sessions covering all required content (k = .95; proportion of observed agreement between coders). On average, session durations were 57:02min (SD = 6:16min) for Session 1 and 51:41min (SD = 8:22min) for Sessions 2-4. Check-ins lasted on average 15:39min (SD = 4:40min).

3.2.2. Acceptability

All survivors who initiated the SSP (N = 19) provided intervention acceptability ratings. As summarized in Table 3, the program was rated “very high” in terms of overall satisfaction, enjoyability, convenience, helpfulness, and likelihood of continued use of skills learned. The most acceptable components of the program were the improvements survivors observed in a relatively short time and the patient-centeredness of facilitators and session content. The least acceptable components included the brief number of sessions and adhering to prescribed time into and out of bed. Virtual delivery was identified as being convenient for most survivors, although several noted that they would have preferred to meet in-person.

Table 3:

Acceptability of the Adapted CBT-I Intervention for Cancer Survivors

Acceptability Item	Rating (1=Very low, 5=Very high)	Most/Least Acceptable Aspects	Exemplary Quote
Overall Satisfaction	Median=5, Mean=4.5, SD=0.7	Most - Amount of improvement	"I was satisfied because I saw the results at the end… I got results fairly quickly um I think that…was very satisfying. My sleep pattern is a lot better, and my mood is better"
		Least - Brief duration	"I felt I would've liked something that went beyond [4 weeks]. It didn't necessarily have to be with [the facilitator] who could be busy, but something added to what I was doing to keep moving me along"
Enjoyableness	Median=5, Mean=4.3, SD=1.0	Most - Patient-centeredness	“I thought it was great like having like somebody actually teaching me what to do and kind of like what things mean in regards to your sleep and not just you know giving me something to follow”
		Least - Sleep prescription	"The biggest problem I had was that I can't fall asleep in 15 minutes and so I would wake up in the [prescribed] time frame… the least enjoyable [aspect] was not getting enough sleep based on the parameters of the study"
Convenience	Median=5, Mean=5.0, SD=0.0	Most - Virtual delivery	"I couldn't be able to just go into the office just on a willynilly… so [virtual delivery] just worked out really well… just really makes it possible to do anything anywhere meeting wise"
		Least - Virtual delivery	"I normally prefer things in person, nothing specific about this program, but I would say if I if I lived close to the hospital, I think I would prefer in person"
Helpfulness	Median=5, Mean=4.4, SD=0.8	Most - Content of program	"I liked learning, you know, what actually helps you get better sleep and can fix the insomnia… was really helpful"
		Least - Brief duration	"A lot kind of jammed into those sessions…I just think if it was a little bit longer…maybe a couple of 45 minute sessions longer it would've been more helpful because I find that it might be easier to fall out of the pattern… once I started to look better I was like oh it's over now"
Odds of Future Use	Median=5, Mean=4.5, SD=0.7	Most - Relaxation training	"The deep breathing exercises like one hand on my stomach and one on my heart… just taking breaks throughout the day to do that"
		Least - Sleep prescription	"I think all the strategies that were taught I will use except for increasing my sleep drive… staying up later and like only getting 6 hours of sleep… that was really difficult for me"

3.3. Preliminary Efficacy of SSP vs. EUC

Intervention condition (SSP vs. EUC) was examined as a predictor of change over time for each sleep outcome. Plots of estimated marginal means are presented in Figure 2.

Figure 2.

Compared to enhanced usual care, the Survivorship Sleep Program resulted in significant reductions in insomnia severity, sleep onset latency, and wake after sleep onset and significant increases in sleep quality, sleep efficiency, and total time in bed. Total sleep time and daytime fatigue were not affected by intervention assignment.

Note.T0 = Baseline, T1 = Post-Intervention, T2 = 1-month follow-up. ISI = Insomnia Severity Index (clinical cutoff = 15); PSQI = Pittsburgh Sleep Quality Index; FSI = Fatigue Symptom Inventor

As depicted in Table 4, the reduction in insomnia severity from T0 - T1 was significantly greater for survivors in the SSP (vs. EUC). The difference in insomnia severity between conditions at T1 was large (d = 1.19) and was greater at T2 (d = 1.27). The change in insomnia severity from T1 - T2 was small and did not differ by condition. Similarly, sleep quality improvement from T0 - T1 was significantly greater for survivors in the SSP (vs. EUC). The difference in sleep quality between conditions at T1 was moderately large (d = .74) and was greater at T2 (d = .83). Across conditions, sleep quality did not change significantly from T1 - T2. Changes in fatigue were not found to be moderated by treatment condition. For all models, controlling for differences in participants’ employment status, insurance type, and marital status did not alter the pattern of results. The magnitudes of between-condition effects obtained from the adjusted ISI model were d = 1.10 and d = 1.15 for T1 and T2, respectively.

Table 4.

Estimates of fixed effects from linear mixed models examining treatment condition as a predictor of change in sleep outcomes over time.

Insomnia Severity Index (ISI)
Model 1	B	SE	df	T	p	r p 2	Model 2	B	SE	df	t	p	r p 2
Intercept	17.85	.97	72.91	18.38	<.001	.82	Intercept	13.05	1.01	78.28	12.93	<.001	.68
t0 – t1	−3.48	1.00	74.70	−3.48	.001	.14	t2 – t0	4.80	.98	74.19	4.89	<.001	.24
t0 – t2	−4.80	.98	74.19	−4.89	<.001	.24	t2 – t1	1.32	1.01	72.28	1.30	.20	.02
Condition (t0)	.25	1.37	72.91	.18	.86	.0004	Condition (t2)	−5.74	1.43	78.41	−4.03	<.001	.17
t0 – t1 * Condition	−5.66	1.39	73.84	−4.08	<.001	.18	t2 – t0* Condition	5.99	1.39	73.81	4.32	<.001	.20
t0 – t2 * Condition	−5.99	1.39	73.81	−4.32	<.001	.20	t2 – t1* Condition	.33	1.41	72.27	.23	.82	.0007
Pittsburgh Sleep Quality Index (PSQI) Item 6 (Sleep Quality)
Model 1	B	SE	df	T	p	r p 2	Model 2	B	SE	df	t	p	r p 2
Intercept	1.95	.15	66.51	19.73	<.001	.85	Intercept	1.45	.16	74.54	15.54	<.001	.76
t0 – t1	−.25	.15	74.06	−1.72	.09	.04	t2 – t0	.50	.14	73.68	3.48	.001	.14
t0 – t2	−.50	.14	73.68	−3.48	.001	.14	t2 – t1	.25	.15	70.86	1.65	.10	.04
Condition (t0)	.30	.21	66.51	1.42	.16	.03	Condition (t2)	−.58	.22	73.18	−2.64	.01	.09
t0 – t1 * Condition	−.82	.20	72.83	−4.07	<.001	.19	t2 – t0* Condition	.88	.20	72.79	4.41	<.001	.21
t0 – t2 * Condition	−.88	.20	72.79	−4.41	<.001	.21	t2 – t1* Condition	.07	.20	70.84	.32	.75	.001
Sleep Diary: Sleep Efficiency (SE)
Model 1	B	SE	df	T	p	r p 2	Model 2	B	SE	df	t	p	r p 2
Intercept	79.81	2.12	56.61	37.62	< .001	.96	Intercept	80.01	2.15	59.06	37.17	< .001	.95
t0 – t1	2.04	1.73	69.01	1.18	.24	.02	t2 – t0	−.20	1.77	69.40	−.11	.91	.0002
t0 – t2	.20	1.77	69.40	.11	.91	.0002	t2 – t1	1.84	1.77	69.40	1.04	.30	.02
Condition (t0)	−6.77	2.94	57.64	−2.30	.03	.08	Condition (t2)	6.13	2.96	58.94	2.07	.04	.07
t0 – t1 * Condition	11.67	2.41	69.18	4.85	< .001	.25	t2 – t0* Condition	−12.90	2.46	69.57	−5.25	< .001	.28
t0 – t2 * Condition	12.90	2.46	69.57	5.25	< .001	.28	t2 – t1* Condition	−1.23	2.44	69.38	−.50	.62	.004
Sleep Diary: Time in Bed (TIB)
Model 1	B	SE	df	t	p	r p 2	Model 2	B	SE	df	t	p	r p 2
Intercept	520.97	18.11	49.34	28.76	< .001	.94	Intercept	525.35	18.30	51.08	28.71	< .001	.94
t0 – t1	3.40	12.38	69.04	.28	.78	.001	t2 – t0	−4.39	12.65	69.30	−.35	.73	.002
t0 – t2	4.39	12.65	69.30	.35	.73	.002	t2 – t1	−.98	12.65	69.30	−.08	.94	.0001
Condition (t0)	1.31	25.04	50.07	.05	.96	<.0001	Condition (t2)	−67.81	25.18	50.99	−2.69	.01	.12
t0 – t1 * Condition	−80.64	17.20	69.15	−4.69	< .001	.24	t2 – t0* Condition	69.11	17.56	69.41	3.94	< .001	.18
t0 – t2 * Condition	−69.11	17.56	69.41	−3.94	< .001	.18	t2 – t1* Condition	−11.53	17.40	69.29	−.66	.51	.01
Sleep Diary: Total Sleep Time (TST)
Model 1	B	SE	df	t	p	r p 2	Model 2	B	SE	df	t	p	r p 2
Intercept	413.49	15.97	46.18	25.89	< .001	.94	Intercept	419.95	16.11	47.56	26.08	< .001	.93
t0 – t1	12.00	9.78	68.90	1.23	.22	.02	t2 – t0	−6.46	10.00	69.11	−.65	.52	.01
t0 – t2	6.46	10.00	69.11	.65	.52	.01	t2 – t1	5.53	10.00	69.11	.55	.58	.004
Condition (t0)	−33.05	22.06	46.76	−1.50	.14	.05	Condition (t2)	−27.39	22.16	47.49	−1.24	.22	.03
t0 – t1 * Condition	−6.08	13.59	68.99	−.45	.66	.003	t2 – t0* Condition	−5.66	13.88	69.20	−.41	.69	.002
t0 – t2 * Condition	5.66	13.88	69.20	.41	.69	.002	t2 – t1* Condition	−11.74	13.75	69.10	−.85	.40	.01
Sleep Diary: Sleep Onset Latency (SOL)
Model 1	B	SE	df	t	p	r p 2	Model 2	B	SE	df	t	p	r p 2
Intercept	24.20	3.51	79.55	6.89	< .001	.37	Intercept	23.92	3.59	82.60	6.66	< .001	.35
t0 – t1	−2.97	3.82	69.35	−.78	.44	.01	t2 – t0	.28	3.89	70.10	.07	.94	.0001
t0 – t2	−.28	3.89	70.10	−.07	.94	.0001	t2 – t1	−2.69	3.89	70.10	−.69	.49	.01
Condition (t0)	8.63	4.88	80.87	1.77	.08	.04	Condition (t2)	−9.51	4.94	82.45	−1.92	.06	.04
t0 – t1 * Condition	−13.85	5.30	69.67	−2.61	.01	.09	t2 – t0* Condition	18.14	5.40	70.42	3.36	.001	.14
t0 – t2 * Condition	−18.14	5.40	70.42	−3.36	.001	.14	t2 – t1* Condition	4.30	5.35	70.06	.80	.43	.01
Sleep Diary: Wake After Sleep Onset (WASO)
Model 1	B	SE	df	t	p	r p 2	Model 2	B	SE	df	t	p	r p 2
Intercept	37.90	7.28	53.60	5.21	< .001	.34	Intercept	31.36	7.28	53.60	4.31	< .001	.26
t0 – t1	−9.58	5.51	65.93	−1.74	.09	.04	t2 – t0	6.54	5.63	66.28	1.16	.25	.02
t0 – t2	−6.54	5.63	66.28	−1.16	.25	.02	t2 – t1	−3.04	5.51	65.93	−.55	.58	.005
Condition (t0)	16.24	10.01	53.50	1.62	.11	.05	Condition (t2)	−1.23	10.08	54.53	−.12	.90	.0003
t0 – t1 * Condition	−15.61	7.66	66.36	−2.04	.05	.06	t2 – t0* Condition	17.48	7.83	66.69	2.23	.03	.07
t0 – t2 * Condition	−17.48	7.83	66.69	−2.23	.03	.07	t2 – t1* Condition	1.87	7.68	65.92	.24	.81	.001
Fatigue Symptom Inventory (FSI)
Model 1	B	SE	df	t	p	r p 2	Model 2	B	SE	df	t	p	r p 2
Intercept	58.00	5.92	57.81	9.80	<.001	.62	Intercept	44.07	6.19	65.06	7.12	<.001	.44
t0 – t1	−13.19	5.12	72.06	−2.58	.01	.08	t2 – t0	13.93	5.00	71.78	2.78	.007	.10
t0 – t2	−13.93	5.00	71.78	−2.78	.007	.10	t2 – t1	.74	5.17	69.38	.14	.89	.0003
Condition (t0)	−2.35	8.37	57.81	−.28	.78	.001	Condition (t2)	−14.33	8.67	63.70	−1.65	.10	.04
t0 – t1 * Condition	−2.37	6.99	71.05	−.34	.74	.002	t2 – t0* Condition	11.98	6.98	71.02	1.72	.09	.04
t0 – t2 * Condition	−11.98	6.98	71.02	−1.72	.09	.04	t2 – t1* Condition	9.61	7.11	69.36	1.35	.18	.03

Note. In Model 1, reference time point is baseline (t0) and reference condition is enhanced usual care (EUC). In Model 2, reference time point is 1-month follow-up (t2) and reference condition is EUC.

Improvements in sleep efficiency from T0 - T1 and T0 - T2 were significantly greater for survivors in the SSP (vs. EUC). Differences observed at T1 (d = .53) and T2 (d = .65) were medium in magnitude. Notably, these improvements in sleep efficiency appeared to be a function of reduced time in bed, rather than an increase in total sleep time. Deceases in time in bed from T0 - T1 and T0 - T2 were significantly greater in the SSP (vs. EUC). Large differences between conditions were observed at T1 (d = 1.01) and T2 (d = .85). In contrast, no significant time x treatment condition effects emerged from the total sleep time analyses. The SSP (vs. EUC) also demonstrated significantly greater reductions in sleep onset latency from T0 - T1 and T0 - T2. At T1, this difference was small (d = .34), but at T2, the difference was moderate in size (d = .61). Finally, the SSP (vs. EUC) demonstrated significantly greater reductions in wake after sleep onset from T0 - T1 and T1 - T2, although the magnitudes of effects were small (ds = .02 and .04 for T1 and T2, respectively) and not statistically significant. Controlling for employment status, insurance type, and marital status did not alter the pattern of results for total sleep time, sleep onset latency, or wake after sleep onset. The difference in initial sleep efficiency between conditions at T0 was no longer significant (p = .32), and the difference in time in bed at T2 was attenuated to a non-significant trend (p = .08).

4. Discussion

Using a two-phase approach, this study developed and preliminarily tested the Survivorship Sleep Program (SSP) comprised of CBT-I delivered in virtual, synchronous sessions with content targeted to cancer survivorship. We found that delivering the SSP was feasible, with high acceptability and large, statistically- and clinically- meaningful reductions in insomnia severity post-intervention and 1-month follow-up. Collectively, this study has several implications for the treatment of insomnia in cancer survivors and the development of scalable behavioral interventions for reaching this growing population.

The adapted protocol was designed to address delivery preferences and target CBT-I skill content to survivorship-specific barriers to sleep (e.g., fear of recurrence, late effects of treatments). Consistent with recent literature, survivors were willing to access sessions remotely, citing time and physical barriers to in-person care^17,18. Remotely delivered CBT-I interventions have largely relied on technologies that are asynchronous (e.g., video recordings³⁵), lack a facilitator (e.g., an app or website^16,19-22,36), or have hybrid in-person and remote sessions (e.g., brief behavioral therapy for insomnia; BBT-I³⁷). As noted previously^35,38, survivors expressed that professional support is necessary to stay engaged in CBT-I sessions and personalize treatment based on cancer-related experiences. These preferences were substantiated by our RCT findings demonstrating rates of session attendance (100%) and retention across assessments (90% by one-month follow-up) that were higher than in other CBT-I trials with cancer survivors^39,40.

Although the remote, synchronous format offered advantages, using live facilitators did require initial training, development of a facilitator’s guide, fidelity checks, and weekly supervision. Additionally, to cover all content, sessions lasted an average of 6-12min longer than anticipated. These findings have implications for guiding future implementation, as the average durations for Session 1 (57min), Sessions 2-4 (51min), and check-ins (16 min) overlap with existing CPT clinical billing codes for an intake session (90791), follow-up session (90834), and check-in (90832), respectively. In the clinical setting, it is possible the brief format may not be optimal for some survivors (i.e., who prefer to have additional sessions), and stepped care models have emerged as one approach to offering progressively more intensive treatment as needed to reduced burden and costs^41,42. Subclinical insomnia severity was a chief reason why cancer survivors were ineligible for this trial, suggesting this subgroup may have unmet needs for a low-intensity sleep intervention. Alternatively, in exit interviews, some survivors expressed a desire for extending the number of sessions. Future research could evaluate the implementation of SSP within hospitals or outpatient clinics and the optimal timing and dose of CBT-I with respect to the severity of survivors’ sleep concerns.

Compared to EUC, the SSP resulted in significant, large, clinically meaningful reductions in insomnia severity (i.e., below clinical cutoff = 15 and > 8-point decrease) at post-intervention, sustained by one month follow-up. Similar to other CBT-I trials with cancer survivors, improvements were also found in other sleep metrics, including sleep quality, sleep efficiency, total time in bed, sleep onset latency, and wake after sleep onset^13,14. Two outcomes, total sleep time and daytime fatigue, were not affected by intervention assignment. It is possible that total sleep time did not increase due to the brevity of the intervention. This finding suggests that SSP-related improvements in insomnia and sleep quality were driven by survivors falling asleep faster, having less fragmented, more efficient sleep, and spending less time awake in bed. The lack of significant effects for fatigue is consistent with literature that cancer-related fatigue is related yet distinct from insomnia, and that biophysiological mechanisms underlying daytime fatigue (e.g., history of radiation therapy) may not be modifiable through targeting sleep behaviors alone. Survivors in the EUC arm did report nominal benefits. Notably, none had initiated CBT-I or other sleep treatments while participating in the study, so these benefits may have resulted from practicing sleep hygiene or expectation effects.

This pilot study had several limitations worth noting. The modest sample size precluded our ability to test moderators and mediators of group-by-time effects. Baseline imbalances on employment and insurance status required both unadjusted and adjusted models. Perhaps because sleep diaries were collected after randomization, SSP participants reported somewhat poorer sleep on T0 diaries, which may have prevented detection of significant between-group differences at T1 and T2. Although the sample was diverse with respect to cancer type, there was limited diversity in terms of race, (15% non-White), ethnicity (12.5% Hispanic) and education (17% not college-educated), reflecting the demographics of the recruitment site. These subgroups may require additional considerations for remote, synchronous CBT-I, which is especially important considering the growing prevalence of cancer survivors across diverse backgrounds. Similarly, future studies should assess cancer survivors’ willingness to pay for CBT-I to contextualize preferences for brief sessions, inform future implementation, and guide policy about insurance coverage and reimbursement for synchronous, provider-delivered behavioral healthcare.

Finally, this trial was conducted at a single site, and testing across multiple sites is required to establish effectiveness of the adapted intervention.

Collectively, the SSP was feasible, acceptable, and evidenced preliminary efficacy as compared to EUC. Synchronous, virtual delivery of CBT-I with targeted content has the promise of reaching and improving sleep outcomes among the growing population of cancer survivors.

5.0. Appendix

Table A1.

Phase 1 Qualitative Interview Participant Characteristics

	N (%) / M(SD) [range]
Age (years	52.4 (15.5) [31-71]
Time Since Treatment (months)	45.5 (29.8) [3-101]
Gender (female)	8 (80%)
Ethnicity
Non-Hispanic	8 (80%)
Race
White	8 (80%)
Latinx	1 (10%)
Asian	1 (10%)
Cancer Type
Breast	3 (30%)
Colon	2 (20%)
Endometrial	1 (10%)
Chondrosarcoma	1 (10%)
Non-Hodgkin’s Lymphoma	1 (10%)
Retinoblastoma	1 (10%)
Prostate	1 (10%)
Thyroid	1 (10%)
Bladder	1 (10%)
Treatment History
Surgery	8 (80%)
Chemotherapy	7 (70%)
Radiation	5 (50%)
Hormonal/Endocrine Therapy	3 (30%)
Immunotherapy	1 (10%)
Radioactive Iodine	1 (10%)
Stem Cell Transplant	1 (10%)

Figure A1.

Phase 1 Qualitative Interview Findings and CBT-I Considerations for Cancer Survivors

Data Availability Statement:

The data that support the findings of this study and intervention materials are available from the corresponding author upon reasonable request.