Interventions for insomnia in cancer patients and survivors—a comprehensive systematic review and meta-analysis

Abstract

Background

Considering the persistent nature and higher prevalence of insomnia in cancer patients and survivors compared with the general population, there is a need for effective management strategies. This systematic review and meta-analysis aimed to comprehensively evaluate the available evidence for the efficacy of pharmacological and nonpharmacological interventions for insomnia in adult cancer patients and survivors.

Methods

Following the PRISMA guidelines, we analyzed data from 61 randomized controlled trials involving 6528 participants. Interventions included pharmacological, physical, and psychological treatments, with a focus on insomnia severity and secondary sleep and non-sleep outcomes. Frequentist and Bayesian analytical strategies were employed for data synthesis and interpretation.

Results

Cognitive-Behavioral Therapy for Insomnia (CBT-I) emerged as the most efficacious intervention for reducing insomnia severity in cancer survivors and further demonstrated significant improvements in fatigue, depressive symptoms, and anxiety. CBT-I showed a large postintervention effect (g = 0.86; 95% confidence interval [CI] = 0.57 to 1.15) and a medium effect at follow-up (g = 0.55; 95% CI = 0.18 to 0.92). Other interventions such as bright white light therapy, sleep medication, melatonin, exercise, mind-body therapies, and mindfulness-based therapies showed benefits, but the evidence for their efficacy was less convincing compared with CBT-I. Brief Behavioral Therapy for Insomnia showed promise as a less burdensome alternative for patients in active cancer treatment.

Conclusions

CBT-I is supported as a first-line treatment for insomnia in cancer survivors, with significant benefits observed across sleep and non-sleep outcomes. The findings also highlight the potential of less intensive alternatives. The research contributes valuable insights for clinical practice and underscores the need for further exploration into the complexities of sleep disturbances in cancer patients and survivors.

Insomnia, a prevalent and debilitating sleep disturbance, is characterized by dissatisfaction with sleep quality or duration accompanied by difficulties in initiating or maintaining sleep, leading to significant distress or impairment in daily functioning (1). In the general population, approximately 30%-35% experience insomnia symptoms, and the average prevalence rate of insomnia disorder has been reported to be approximately 10% (2). The course of insomnia is often chronic and not only impairs quality of life but also increases the risk for several physical and mental health issues, such as cardiovascular diseases, diabetes, dementia, depression, and anxiety (3-5). Impaired sleep quality, which is highly correlated with insomnia, is a closely related but broader concept of sleep disturbance (6).

In cancer patients and survivors, the prevalence of such sleep disturbances is markedly higher than in the general population, with a meta-analysis of 160 studies reporting that, on average, 67% of individuals with a cancer diagnosis experience sleep disturbances (7). The sample characteristics generally associated with a higher prevalence of sleep disturbances include younger age and type of cancer, with higher rates found in lung, breast, and gynecological cancers. In contrast, prevalence rates do not appear to differ according to cancer stage or treatment status (ie, active vs completed treatment) (7).

The heightened prevalence in cancer populations is likely attributed to a range of disease and treatment-related factors, including physical and psychological side and late effects of cancer treatments such as pain, menopausal symptoms such as hot flashes, anxiety, depression, and worries (8,9). As in the general population, once manifested, sleep disturbances in cancer patients and survivors tend to become chronic. For example, 58% of a large nationwide cohort of women treated for breast cancer experienced clinically significant sleep disturbances 3 to 4 months after surgery (10), and almost 52% reported sleep disturbance 7 to 9 years later (11).

The consequences of sleep disturbance in cancer patients and survivors are profound and multifaceted. In addition to being associated with reduced quality of life, sleep disturbances in this population have been linked not only to increased symptom burden and higher rates of depression and anxiety (12) but also to poorer prognostic outcomes (12-14). The role of sleep in cancer outcomes could partly be explained by the known bidirectional associations between sleep and the immune system (15). Sleep disturbance has been found to be associated with increased levels of inflammation in cancer patients (16), and the role of inflammation in cancer progression is well established (17). Inflammation may, in turn, play a central role in the cluster of sleep disturbance, fatigue, depression, and pain symptoms commonly observed in cancer patients and survivors (18,19). The interplay between sleep, cancer progression, and treatment response underscores the critical need for effective management strategies for insomnia in this vulnerable population.

In the general population, treatment guidelines for insomnia typically include cognitive-behavioral therapy for insomnia (CBT-I), pharmacological interventions, and lifestyle modifications (20,21). CBT-I, in particular, has been established as the first-line treatment because of its efficacy and minimal adverse effects compared with those associated with medication (22). However, the applicability and effectiveness of these interventions in cancer patients and survivors remain an area of active research.

The management of insomnia in oncology settings is complicated by the unique psychological and physiological challenges faced by cancer patients and survivors. Although some studies have explored the effectiveness of standard insomnia treatments in this population, there is a growing interest in identifying the interventions that address their specific needs and circumstances. Previous studies and reviews have begun to shed light on this area, suggesting potential benefits of both pharmacological (eg, melatonin) (23) and nonpharmacological (eg, CBT-I) approaches (24), mindfulness-based interventions (25), and exercise (26). Yet the evidence remains fragmented and incomplete.

Therefore, the present systematic review and meta-analysis aims to comprehensively evaluate the efficacy of interventions for insomnia in cancer patients and survivors, seeking to provide a clearer understanding of the most efficacious strategies for managing this challenging condition in the context of cancer care. In contrast to previous reviews, our analysis will encompass a wide range of interventions, including pharmacological, physical, and psychological interventions, examining their impact on the primary outcome of insomnia severity as well as on a number of secondary sleep outcomes, both self-reported and objectively assessed, and secondary non-sleep outcomes, including fatigue, anxiety, and depressive symptoms. In doing so, this review will not only contribute to the growing body of literature on insomnia in cancer patients but also offer valuable insights for clinicians and researchers. By identifying effective interventions, we aim to improve the management of insomnia in cancer patients and survivors, ultimately enhancing their quality of life and potentially influencing their overall cancer journey.

Methods

The present review was conducted and reported following the PRISMA recommendations for reporting systematic reviews and meta-analyses (27) and the protocol outline preregistered with PROSPERO (#CRD42021281153) (28).

Search strategy

Keyword-based searches were conducted in the databases of PubMed, PsycINFO, Cochrane, CINAHL, and Embase. Keywords related to cancer (eg, neoplasm OR oncology) were combined with keywords related to insomnia (eg, sleep OR sleep problem OR sleep disturbance) and terms related to study type (randomized OR RCT). The full search string is shown in Supplementary Methods (available online). Searches were conducted for the period from the earliest time available until November 22, 2023, together with backward searches (snowballing) of reference lists of identified articles and earlier systematic reviews and forward searches (citation tracking). Regular database search updates and de-duplication were performed following previously described procedures (29,30).

Selection procedure and data extraction

Only English-language reports published in peer-reviewed sources were included. Study eligibility was assessed using the PICO (Population, Intervention, Comparison, Outcome) approach (31).

Population: adult cancer patients or survivors (age ≥ 18 years) with symptoms of insomnia/sleep disturbance (ie, difficulties falling asleep and/or maintaining sleep). Insomnia diagnosis or symptom scores above prespecified cutoffs were not required. Studies of children and adolescents with cancer, patients without current cancer or a cancer history, or cancer patient caregivers were excluded, together with studies focusing on sleep problems other than insomnia (eg, sleep apnea, restless legs syndrome, and narcolepsy).

Intervention: any psychological, physical, or pharmacological treatment with the explicit aim of treating sleep disturbance or improving sleep quality as the primary outcome. Excluded were studies of complementary and alternative therapies with no clear scientifically accepted theory about their working mechanisms and not presently considered part of conventional medicine (eg, homeopathy, acupuncture, reflexology, herbal medicines, qigong).

Comparison: Eligible studies were required to be randomized controlled trials (RCTs) employing at least 1 passive, active, or competing control condition. Case studies, open trials, and other studies without control groups were excluded.

Outcome: Eligible studies had to include pre- and postintervention data or pre-post change score data on one or more quantitative sleep-relevant constructs assessed with validated instruments. Insomnia severity was chosen as the primary outcome as assessed with a validated measure of insomnia, for example, the insomnia severity index (ISI) (32). Secondary sleep outcomes included sleep quality assessed with validated scales, for example, the Pittsburgh Sleep Quality Index (PSQI) (33), and other standard sleep outcome measures such as sleep onset latency (SOL), nocturnal awakenings, wake after sleep onset (WASO), early morning awakenings (EMA), time in bed (TIB), total sleep time (TST), and sleep efficiency (SE) assessed either subjectively with sleep diaries or objectively with actigraphy or polysomnography (PSG). We also explored the effects on the secondary non-sleep outcomes of fatigue, anxiety, and depression. In addition, studies needed to report results as either pre-post means and SD/SE in all groups, change-scores in all groups, effects sizes (ESs) (eg, Cohen’s d, η²), or provide other data that could be converted into an ES.

The final search and removal of duplicates was conducted by 1 author (ERN), and 4 authors (ERN, SMK, HN, and EN) screened titles and abstracts in pairs, ensuring that all records were independently evaluated by 2 authors. Full texts were assessed for eligibility and reasons for exclusion registered using the Covidence software (34). Disagreements were discussed with a third author (RZ) until a negotiated conclusion was reached. Data were extracted by 1 author and checked by a second author (ERN, SMK, HN, or RZ). Studies were coded according to a priori specified study, intervention, and participant characteristics. If needed, study authors were contacted for additional data or clarifying details.

Risk of bias assessment

The revised Cochrane Risk of Bias tool (RoB 2) (35) was used to evaluate possible bias in the included studies. As RoB 2 is outcome specific, self-reported and objective outcomes were evaluated separately. Studies investigating more than 1 intervention were evaluated for each comparison included in the meta-analysis. Studies that included both an active and passive control condition were assessed on the basis of the active control. Five sources of bias were assessed: 1) bias arising from the randomization process, 2) bias because of deviations from intended interventions, 3) bias because of missing outcome data, 4) bias in the measurement of the outcome, and 5) bias in the selection of the reported result. All studies were rated as “low risk,” “high risk,” or “some concerns” for each source of bias, and an overall assessment of the risk of bias was conducted for each study. The assessments were performed independently by 3 authors (SMK, RN, ERN). Disagreements were solved by negotiation.

Computing effect sizes

Hedges’s g, a variation of Cohen’s d (36), correcting for possible bias because of small sample sizes (37), was used as the standardized between-group effect size (ES). Whenever possible, ESs were computed using means and their standard deviations for preintervention, postintervention, or change scores. If unavailable, ESs were estimated on the basis of other reported statistics (eg, P-values, F-values, or B-values) and, in some instances, medians and interquartile ranges.

Frequentist analytical strategy

When the number of studies (K) was 3 or more for an outcome, ESs were pooled and weighted by the inverse standard error, taking the precision of each study into account. A random-effects model was chosen for all analyses, with positive values indicating ESs in the hypothesized direction. If studies reported results for more than 1 measure per outcome, independence of results was ensured by averaging ESs across all outcomes so that only 1 result per study was used for each quantitative data synthesis. Heterogeneity was explored by calculating the I² statistic, which estimates the variance in a pooled ES that is accounted for by heterogeneity, that is, true differences between effect sizes rather than variation because of sampling error (38). If the results indicated heterogeneity (I² > 0.0), we calculated the 95% prediction interval, which quantifies the distribution of the ESs, indicating that in 95% of cases, the true effect of a new and unique study (from the same family of studies) will fall within this range (39).

Possible sources of heterogeneity were explored by comparing the ESs of studies according to the following study characteristics: participant characteristics (ie, mean sample age, percentage women, percentage White participants); cancer stage (early-stage vs advanced), and treatment status (survivors vs patients in active primary cancer treatment); intervention characteristics, that is, delivery type (face-to-face vs remote), therapist contact (yes, no), delivery format (individual vs group), type and number of CBT-I components (sleep restriction, relaxation, stimulus-control, cognitive therapy, and sleep hygiene education), and control condition characteristics (passive vs active and competing). These moderators were explored for the primary outcome of insomnia severity with subgroup analyses when K ≥ 3 and meta-regression when K ≥ 10. Likewise, when K ≥ 10, the possibility of publication bias was evaluated with funnel plots and Egger’s method (40). If the results were suggestive of publication bias, we planned to calculate an adjusted ES using the Duval and Tweedie trim and fill method (41). The calculations were conducted using Comprehensive Meta-Analysis, version 4 (42), and various formulas in Microsoft Excel.

Supplementary Bayesian analyses

To aid the interpretation of the results, we conducted, as a supplement to the conventional frequentist meta-analysis, a Bayesian Model-Averaged meta-analysis (43) of the self-reported and objective sleep outcomes. The procedure examines the results of 4 models: 1) the fixed-effect null hypothesis (fH₀), 2) the fixed-effect alternative hypothesis (fH₁), 3) the random-effects null hypothesis (rH₀), and 4) the random-effects alternative hypothesis (rH₁). Bayesian Model-Averaged analysis thus avoids selecting either a fixed- or random-effects model and addresses 2 questions in light of the observed data: What is the plausibility that the overall effect is nonzero and the effect sizes are heterogeneous? We chose an uninformed prior probability (ie, 25%) of each of the 4 models and used 2000 iterations. Concerning parameter distributions, we chose previously recommended defaults (43). We thus used a zero-centered Cauchy prior with a scale of 0.707 for the ES. For the between-study variation, we used an empirically informed prior distribution on nonzero between-study deviation estimates based on standardized mean difference ESs from 705 meta-analyses published in Psychological Bulletin between 1990 and 2013 (44). This distribution has been approximated by an inverse-gamma (1, 0.15) prior on the standard deviation (Tau) (26). The results are presented as Bayes factors (BFs) (45), which represent the posterior probability of the alternative hypothesis (H₁) relative to the null hypothesis (H₀). The levels of evidence in favor of nonzero effects (H₁) are categorized as weak (BF = 1-3), moderate (BF = 3-10), strong (BF = 10-30), very strong (BF = 30-100), and decisive (BF > 100) (46). The Bayesian analyses were conducted with the computer software JASP (version 0.17) (47).

Results

The literature search yielded 15 823 hits. After removing duplicates and title, abstract, and full text screening, 61 independent randomized controlled trials (RCTs) reported in 63 publications were eligible to be included in the meta-analysis. The study selection process with reasons for exclusion is described in Figure 1. A reference list of all included studies is included in Supplementary Materials (available online).

Figure 1.

Study selection process.

Study and participant characteristics

The study characteristics are summarized in Table 1. The 61 randomized controlled trials (RCTs) had recruited a total of 6528 cancer patients and survivors. Sample sizes ranged from 18 (61) to 709 (97), with a median of 72. The majority of RCTs were conducted in the United States (K = 27), followed by Canada (K = 7), China (K = 5), Iran (K = 4), Denmark (K = 3), Australia (K = 2), Taiwan (K = 2), and India (K = 2). Nine additional studies were conducted in 9 other various countries. The average age of participants was 54.8 years, and 78.9% were women. Only half of the studies (K = 29) reported on the racial composition of the sample. Those that did were almost exclusively from the United States. The majority of participants in these studies were White, with an average proportion of 85% White participants across samples. Most studies focused on breast cancer (K = 27), mixed cancers (K = 23), lung (K = 4), and hematological cancers (K = 3). Four studies explored lymphoma, colorectal, gynecological, and prostate cancer. Cancer severity was primarily early-stage (K = 26) or mixed-stage (K = 19), whereas 4 studied advanced-stage patients and 12 did not report disease severity. More than half of the studies included cancer survivors (K = 33), whereas 24 focused on patients undergoing active primary cancer treatment. Inclusion criteria for more than half of the studies (K = 36) involved insomnia severity or sleep quality instrument cutoffs.

Table 1.

Study characteristics^a

Study (country)a	Cancer type; Treatment status; Stage	Demographic characteristics (mean age; % women; race)	Intervention arms; Intervention type(s)	Control type Control condition	Insomnia as inclusion criteria (Yes/No)	Sleep outcomes	Secondary outcomes	Treatment format; Therapist contact (Yes/No); Delivery mode (specified) b	No. of sessions; (Time to post); [Time to FU]	Initial N; Analyzed N (post, FU)	ROB-2: self-report; objective outcome
Barsevick et al. 2010 (USA) (48)	Mixed; Active; Stage I-IV	Mean age = 54.0; 83% women; 0.7% Asian, 5% Black, 2.1% Native American, 90.7% White	BBT-I	Active: Healthy eating control	No	SQ; Actigraphy (SOL, WASO, NA, TST, SE)	Fatigue; Depression; Anxiety	Individual; YES; Remote (phone call)	3 sessions; (3 wks)	292 (276)	Some concerns; Some concerns
Bean et al. 2022 (Australia) (49)	Breast; Active; Stage I-IV	Mean age = 39.6; 100% women; 4.3% Asian, 87.2% White	CBT-I + BWL	Passive: CAU	No	Insomnia severity	Fatigue; Depression; Anxiety	Individual; YES; Remote digital + phone	(6 wks) [12 wks]	101 (70, 55)	Some concerns; NA
Berger et al. 2009 (USA) (50)	Breast; Active; Stage I-III	Mean age = NR; 100% women; 3% Hispanic, 97% White	CBT-I	Active: Healthy eating control	No	SQ; Sleep diary (SOL, WASO, TST, SE); Actigraphy (SOL, WASO, TST, SE)	Fatigue	Individual; YES; Face-to-face	NR	219 (182)	Some concerns; Some concerns
Casault et al. 2015 (Canada) (51)	Mixed; Mixed; Stage I-III	Mean age = 56.9; 92.1% women; race: NR	CBT-I	Passive: CAU	Yes	Insomnia severity; Sleep diary (SOL, WASO, TST, SE)	Fatigue; Depression; Anxiety	Individual; YES; Remote (6 booklets + 3 phone calls)	6 + 3 sessions; (6 wks) [26 wks]	38 (32, 30)	High risk; NA
Celik et al. 2022 (Turkey) (52)	Mixed; Active (palliative care); Stage II-IV	Mean age = NR; 50% women; race: NR	BWL	Active: DRL	Yes	SQ	Fatigue	–	14 sessions; (2 wks) [2 wks]	52 (52, 52)	Some concerns; NA
Chaoul et al. 2018 (USA) (53)	Breast; Active; Stage I-III	Mean age = 49.2; 100% women; 5.4% Asian, 14.0% Black, 13.5% Hispanic, 64.0% White	a) Exercise (stretching) b) Mind-body (yoga)	Passive: CAU	No	SQ; Actigraphy (SOL, WASO, TST, SE)	Fatigue	Individual; YES; Face-to-face	4 sessions + 3 booster sessions; (NR) [52 wks]	a) 153 (138, 99) b) 159 (143, 107) c) 142 (123, 86)	a) High risk; b) High risk
Chen et al. 2014 (USA) (54)	Breast; Survivors; Stage I-III	Mean age = 59; 100 % women; 1.1% Asian, 2.1% Black, 96% White	Pharmacological: Melatonin	Active: Placebo	No	SQ	Depression	–	3 mg daily (16 wks)	95 (86)	High risk; NA
Chen et al. 2016 (Taiwan) (55)	Lung; Mixed; Stage I-IV	Mean age = 63.6; 55.86% women; race: NR	Exercise: walking	Passive: CAU	No	SQ; Actigraphy (SOL, WASO, TST, SE)	–	Individual; YES; Remote (weekly phone calls)	36 sessions; (12 wks) [13 wks]	111 (95, 89)	High risk; Some concerns
Dean et al. 2019 (USA) (56)	Lung; Mixed; Stage I-II	Mean age = 66; 60% women; 10% Black, 90% White	BBT-I	Active: Healthy eating control	Yes	SQ; Sleep diary (SOL, TST, TiB, SE)	–	Individual; YES; Face-to-face (plus 2 phone calls)	4 sessions + 2 follow-up phone calls; (4 wks) [2 wks]	40 (25)	Some concerns; NA
Dimsdale et al. 2011 (USA) (57)	Hematologic; Active; Stage NR	Mean age: NR; % women: NR; race: NR	Pharmacological: Eszopiclone	Active: Placebo	No	Sleep diary (SOL, NA, TST)	Fatigue	–	2-3 mg.; (2 days)	45 (45)	Some concerns; NA
Epstein & Dirksen 2007 (USA) (58)	Breast; Survivors; Stage I-III	Mean age = 58.2; 100% women; 2% Black, 2% Native American, 96% White	CBT-I	Active: Sleep hygiene	Yes	SQ; Sleep diary (SOL, WASO, TST, TiB, SE); Actigraphy (SOL, WASO, TST, TiB, SE)	–	Group (+ individual); YES; Face-to-face (+ Remote phone calls)	4 group sessions + 2 weekly phone calls; (4 wks)	81 (61-72)	High risk; High risk
Espie et al. 2008 (UK) (59)	Mixed; Survivors; Stage NR	Mean age = NR; 68.7 % women; race: NR	CBT-I	Passive: CAU	Yes	Sleep diary (SOL, WASO, TST, SE); Actigraphy (SOL, WASO, TST, SE)	Fatigue; Depression; Anxiety	Group; YES; Face-to-face	5 sessions (5 wks) [26 wks]	150 (121, 106)	High risk; Some concerns
Etedali et al. 2022 (Iran) (60)	Prostate; Active; Stage I-III	Mean age = 69.3; 0 % women; race: NR	Pharmacological: Melatonin	Active: Placebo	Yes	SQ	Depression; Anxiety	–	6 mg daily (4 wks)	49 (43)	High risk; NA
Fox et al. 2021 (USA) (61)	Ovarian, peritoneal, cancer of the fallopian tube; Survivors; Stage I-III	Mean age = 57.1; 100% women; 100% White	BWL	Active: DRL	Yes	SQ; Actigraphy (SOL, WASO, NA, TST, TiB, SE)	Fatigue; Depression	–	28 sessions; (4 wks) [13 wks]	18 (18, 18)	High risk; High risk
Garland et al. 2014 (Canada) (62)	Mixed; Survivors; Stage I-III	Mean age = 58.9; 72% women; 3% Aboriginal, 7% Asian, 1% Black, 90% White	CBT-I d	Competing: MBSRd	Yes	Insomnia severity; SQ; Sleep diary (SOL, WASO, TST, SE); Actigraphy (SOL, WASO, TST, SE)	Depression	Group; YES; Face-to-face	8 sessions; (8 wks) [13 wks]	111 (111, 111)	High risk; High risk
Garland et al. 2019 (Canada) (63)	Mixed; Survivors; Stage 0-IV	Mean age = 61.5; 56.9 % women; 17% Black, 71% White	CBT-I	Active: Acupunc-ture	Yes	Insomnia severity; SQ; Sleep diary (SOL, WASO, TST, SE)	Fatigue; Depression; Anxiety	Individual; YES; Face-to-face	7 sessions (8 wks) [12 wks]	160 (148, 147)	Some concerns; NA
Grégoire et al. 2022 (Belgium) (64)	Mixed; Survivors; Stage I-III	Mean age = 53.9; 100 % women; race: NR	Mind-body: Self-hypnosis	Passive: WL	No	Insomnia severity	Fatigue; Depression; Anxiety	Group; YES; Face-to-face	8 sessions (8 wks)	95 (95)	High risk; NA
Hall et al. 2022 (65); Wieman et al. 2023 (USA) (66)	Mixed; Survivors; Stage I-III	Mean age = 53.2; 85 % women; 5% Asian, 2.5% Black, 10% Latinx, 85% White	CBT-I	Active: Enhanced CAU	Yes	Insomnia severity; SQ; Sleep diary (SOL, WASO, TST, TiB, SE)	Fatigue; Depression	Individual; YES; Remote (video conference)	7 sessions (4 wks) [8 wks]	40 (38, 37)	Some concerns; NA
Hansen et al. 2014 (Denmark) (67)	Breast; Active; Stages I-III	Mean age = 55.3; 100% women; race: NR	Pharmaco-logical: Melatonin	Active: Placebo	No	Sleep diary (SOL, NA, TST SE)	–	–	6 mg (1 wks) [12 wks]	54 (45)	High risk; NA
Harorani et al. 2020 (Iran) (68)	Mixed; Active; Stage NR	Mean age = 45.4; 50% women; race: NR	Relaxation	Passive: CAU	No	SQ	–	Individual; NO; audio records	10 sessions (5 days)	84 (80)	High risk; NA
He et al. 2022 (China) (69)	Breast; Active; Stages I-III	Mean age = 48.2; 100% women; race: NR	Exercise: dancing	Active: Health consultation	No	SQ	Fatigue; Depression	Group/Individual; YES; Face-to-face and remote (self-administered)	6 group sessions + 16 weeks of home practice (16 wks)	176 (159)	Some concerns; NA
Hrushesky et al. 2022 (USA) (70)	Lung; Active; Stages III-IV	Mean age = 61.33; 22.62 % women; race: NR	Pharmacological: Melatonin	Active: Placebo	No	SQ	–	–	20 mg (12 wks)	50 (36)	High risk; NA
Irwin et al. 2017 (USA) (71)	Breast; Survivors; Stage NR	Mean age = 59.8; 100% women; 86% White, Other: NR	CBT-Ic	Competing: Tai Chi meditationc	Yes	SQ; Sleep diary (SOL, WASO, TST, SE); Actigraphy (SOL, WASO, TST, SE); PSG (SOL, WASO, TST, SE)	Fatigue; Depression	Group; YES; Face-to-face	12 sessions (12 wks) [52 wks]	90 (80, 73)	Some concerns; Some concerns
Jakobsen et al. 2023 (Norway) (72)	Mixed; Active; Stage IV	Age = median: 66; 44% women; race: NR	Pharmacological: Zopiclone	Active: Placebo	Yes	SQ	–	–	3,75-7,5 mg daily (6 days)	41 (39)	Some concerns; NA
Kurdi & Muthukalai 2016 (India) (73)	Mixed; NR; Stage I-IV	Mean age = 52.4; 48% women; race: NR	Pharmacological: Melatonin	Active: Placebo	Yes	Insomnia severity	–	–	3 mg daily (2 wks)	50 (48)	Some concerns; NA
Lengacher et al 2015 (USA) (74)	Breast; Survivors; Stage 0-III	Mean age = 57; 100 % women; 10.1% Black, 84.8% White	MBSR	Passive: CAU	No	SQ; Sleep diary (SOL, WASO, NA, TST, SE); Actigraphy (TST, SOL)	–	Group; YES; Face-to-face	6 sessions (6 wks) [6 wks]	79 (76, 77)	High risk; Some concerns
Liu et al. 2022 (China) (75)	Breast; Active; Stage 0-IV	Mean age = 51.0; 98.15% women; race: NR	MBSR	a) Passive: CAU b) Active: acupressure	Yes	SQ; Actigraphy (SOL, WASO, NA, TST, SE)	Fatigue; Depression; Anxiety	Group; YES; Face-to-face	8 sessions (8 wks)	108 (a)62; b)64)	High risk; High risk
Lubas et al. 2022 (USA) (76)	Mixed; Survivors; Stage NR	Mean age = 33.5; 53.45% women; 87% White	Pharmacological: Melatonin	Active: Placebo	Yes	SQ; Sleep diary (TST; SOL; SE); Actigraphy (WASO; TST; SOL; SE)	–	–	3 mg daily (26 wks)	580 (298-374)	High risk; High risk
Madsen et al. 2016 (Denmark) (77)	Breast; Active; Stage I-III	Mean age = 54.5; 100% women; race: NR	Pharmacological: Melatonin	Active: Placebo	No	SQ; Actigraphy (SOL, WASO, TST, TiB, NA, SE)	–	–	6 mg daily (3 days) [1 wk]	54 (48, 48)	High risk; High risk
Matthews et al. 2014 (USA) (78)	Breast; Survivors; Stage I-III	Mean age = 52.49; 100% women; race: NR	CBT-I	Active: Desensiti-zation	Yes	Sleep diary (TST)	Fatigue; Depression; Anxiety	Individual; YES; Face-to-face and remote (phone calls)	6 sessions (6 wks) [26 wks]	60 (60, 60)	Some concerns; NA
Mercier et al. 2018 (Canada) (79)	Mixed; Survivors; Stage 0-III	Mean age = 57.1; 78.1% women; 100% White	CBT-I	Competing: Exercise	Yes	Insomnia severity; SQ; Sleep diary (SOL, WASO, TST, SE); Actigraphy (SOL, WASO, TST, SE)	–	Individual; YES; Remote digital (+ weekly phone calls)	(6wks) [26 wks]	41 (41, 41)	High risk; High risk
Moon et al. 2020 (Korea) (80)	Mixed; Survivors; Stage I-IV	Mean age = 63; 50% women; race: NR	CBT-Id	Competing: Herbal medicine + sleep hygiened	Yes	Insomnia severity; SQ	Fatigue; Anxiety	Individual; YES; Face-to-face	4 sessions (4 wks)	22 (20)	Some concerns; NA
Mustian et al. 2013 (USA) (81)	Mixed; Survivors; Stages I-IV	Mean age = 54.1; 96% women; 6% Black, 93% White	Mind-body: Yoga	Passive: CAU	Yes	SQ; Actigraphy (SOL, WASO, SE)	–	Group; YES; Face-to-face	8 sessions (4 wks)	410 (321)	High risk; Some concerns
Nakamura et al. 2013 (USA) (82)	Mixed; Survivors; Stage NR	Mean age = 52.59; 75.44% women; 5% Hispanic, 95% White	a) Mind-Body Bridging b) Mindfulness Meditation	Active: Sleep hygiene education	Yes	SQ	Depression	Group; YES; Face-to-face	3 sessions (3 wks) [9 wks]	57 (55, 44)	a) Some concerns; NA b) Some concerns; NA
Nguyen et al. 2020 (Australia)	Breast; Survivors; Stage I-III	Mean age = 62; 100% women; race: NR	Exercise: Physical activity device	Passive: WL	No	SQ; Actigraphy (SOL, WASO, TST, SE, NA)	–	Individual; YES; Face-to-face (1 session) + remote (5 phone calls) + Remote Digital	6 sessions (12 wks) [26 wks]	83 (80, 70)	High risk; Low risk
Nourizadeh et al. 2022 (Iran) (83)	Breast; Survivors; Stage II-III	Mean age = 38.4; 100% women; race: NR	Exercise: Aerobicc	a) Passive: CAU b) Active: acupressurec	Yes	SQ	–	Group; YES; Face-to-face	24 sessions (8 wks)	99 (99)	High risk; NA
Oswald et al. 2022 (Puerto Rico) (84)	Breast; Survivors; NR	Mean age = 58.44; 100% women; race: NR	CBT-I	Passive: WL	Yes	Insomnia severity; SQ; Sleep diary (SE)	–	Group; YES; Remote (video conference)	6 sessions (6 wks)	30 (29)	High risk; NA
Padron et al. 2021 (USA) (85)	Gynecological; Survivor; Stage I-IV	Mean age = 59.39; 100% women; 89% White, Other: NR	CBT-I	Active: Psychoeducation	Yes	SQ; Sleep diary (SOL, WASO, SE); Actigraphy (SOL, WASO, SE); PSG (SOL, WASO, SE)	–	Individual; YES; Face-to-face	6 sessions (6 wks) [9 wks]	35 (27, 25)	Some concerns; High risk
Palesh et al. 2018 (USA) (86)	Breast; Active; Stage I-III	Mean age = 52.5; 100% women; 1% Asian, 3% Black, 96% White	BBT-I	Active: Healthy eating	Yes	Insomnia severity	–	Individual; YES; Face-to-face (session 1 + 6) + Remote (phone calls sessions 2-5)	6 sessions (3 wks)	71 (45)	Some concerns; NA
Palesh et al. 2020 (USA) (87)	Breast; Active; Stage I-IV	Mean age = 50.13; 100% women; 12.9% Asian, 4.3% Black, 2.9% Native American, 2.9% Pacific Islander, 65.7% White	BBT-I	Active: Sleep hygiene	Yes	Insomnia severity	Depression; Anxiety	Individual; YES; Face-to-face (session 1 + 6) + Remote (phone calls sessions 2-5)	6 sessions (6 wks) [26 wks]	74 (70, 70)	Some concerns; NA
Palmer et al. 2020 (Brazil) (88)	Breast; Active; Stage NR	Mean age = 54.18; 100% women; race: NR	Pharmacological: Melatonin	Active: Placebo	No	SQ	Depression	–	10 days	36 (35)	Low risk; NA
Pang et al. 2023 (China) (89)	Breast; Active; Stage NR	Mean age = 52; 100% women; race: NR	Mind-body: Relaxation	Passive: CAU	No	SQ	–	Individual; YES; Remote (phone)	6 sessions (12 wks)	60 (60)	High risk; NA
Rao et al. 2017 (India) (90)	Breast; Survivor; Stage IV	Mean age = 49.6; 100% women; race: NR	Mind-body: Yoga	Active: Supportive therapy	No	Insomnia severity	–	Individual; YES; Face-to-face	24 sessions (12 wks)	91 (64)	Some concerns; NA
Rissling et al. 2022 (USA) (91)	Breast; Active; Stages I-III and unknown	Mean age = 53.95; 100% women; 7.7% Asian, 15.4% Black, 71.8% White	BWL	Active: DRL	No	SQ; Actigraphy (WASO, TST, SE)	–	–	56-84 sessions (8-12 wks)	41 (29)	Some concerns; Some concerns
Ritterband et al. 2012 (USA) (92)	Mixed; Survivors; Mixed stages	Mean age = 56.7; 85.71% women; 4% Black, 93% White	CBT-I	Passive: WL	Yes	Sleep diary (SOL, WASO, TST, TiB, NA, SE)	Fatigue; Depression; Anxiety	Individual; NO; Remote digital	(9 wks)	28 (26)	High risk; NA
Roscoe et al. 2015 (USA) (93)	Mixed; Survivors; Stage I-III	Mean age = 56; 87.5% women; 8.3% Black, 89,6% White	a) CBT-I b) Pharmacological: Armodafinil	Active: Placebo	Yes	SQ	–	a) Individual; YES; Face-to-face and remote (phone) b) b	a) 7 sessions (7 wks) b) (7 wks)	a) 49 (35) b) 49 (32)	a) Some concerns; NA b) Low risk; NA
Roveda et al. 2017 (Italy) (94)	Breast; Survivors; Stage I-III	Mean age = 56.78; 100% women; race: NR	Exercise: Walking + stretching	Passive: CAU	No	Actigraphy (SOL, WASO, TST, SE)	–	Group; YES; Face-to-face	24 sessions (12 wks)	42 (40)	NA; Some concerns
Savard et al. 2005 (Canada) (95)	Breast; Survivors; Stage I-III	Mean age = 54.05; 100% women; 100% White	CBT-I	Passive: WL	Yes	Insomnia severity; Sleep diary (SOL, WASO, TST, SE); PSG (SOL, WASO, TST, SE)	Fatigue; Depression; Anxiety	Group; YES; Face-to-face	8 sessions (8 wks)	58 (57)	High risk; Some concerns
Savard et al. 2014 (Canada) (96)	Breast; Survivors; Stage I-III	Mean age = 54.4; 100% women; race: NR	CBT-I a) Face-to-face b) Video-based	Passive	Yes	Insomnia severity; Sleep diary (SOL, WASO, TST, SE)	Fatigue; Depression; Anxiety	Individual; a + b) YES; Face-to-face c) NO; Remote digital	6 sessions (6 wks)	242 (204)	a) High risk; NA b) High risk; NA
Seely et al. 2021 (Canada) (97)	Lung; Active; Stage I-IV	Mean age = 67.2; 56.14% women; race: NR	Pharmacological: Melatonin	Active: Placebo	No	SQ	Fatigue	–	20 mg (52 wks)	709 (499)	Low risk; NA
Shahrokhi et al. 2021 (Iran) (98)	Colorectal; Active; NR	Mean age = 63.87; 46.67% women; race: NR	Pharmacological: Zolpidem	Competing: Melatonin	Yes	SQ	Depression; Anxiety	–	(4 wks) [9 wks]	101 (90, 90)	High risk; NA
Starreveld et al. 2021 (Netherlands) (99)	Lymphoma; Survivors; stages I-IV	Mean age = 45.7; 59.6% women; race: NR	BWL	Active: DRL	No	SQ; Actigraphy (TST, SE)	Fatigue; Depression; Anxiety	–	25 sessions (4 wks) [39 wks]	166 (157, 144)	Some concerns; Some concerns
Tang et al. 2010 (Taiwan) (100)	Mixed; Mixed; NR	Mean age = 51.80; 76.1% women; race: NR	Exercise: walking	Passive: CAU	Yes	SQ	–	Individual; NO; Face-to-face	24 sessions (8 wks)	72 (59)	High risk; NA
Wells di- Gregorio et al. 2018 (USA) (101)	Mixed; Active; Stage IV	Mean age = 56.54; 0.82% women; 7% Black, 93% White	ACT	Passive: WL	Yes	Insomnia severity; Sleep diary (SOL, SE)	Fatigue; Depression; Anxiety	Individual; YES; Face-to-face and remote (DVD)	3 sessions (6 wks)	30 (25)	High risk; NA
Wu et al. 2018 (USA) (102)	Mixed; Survivors; Stage NR	Mean age = 53.6; 75% women; 36.4% Black, 43.2% White	BWL	Active: DRL	No	SQ; Actigraphy (WASO, TST, SE)	Fatigue	–	28 sessions (4 wks) [13 wks]	54 (34, 31)	Some concerns; Some concerns
Wu, Gao et al.2022; Wu, Gao et al. 2023 (USA)	Breast; Survivors; Stage I-III	Mean age = 52.86; 100 % women; 9.5% Black, 90.5% White	BWL	Active: DRL	Yes	SQ; Sleep diary (SOL, TST, SE) PSG (SOL, WASO, TST, SE)	Fatigue; Depression	–	14 sessions (2 wks)	21 (19)	Some concerns; Some concerns
Wu, Valdemars-dottir et al. 2022 (USA) (103)	Hematological; Survivors; Stage NR	Mean age = 56.8; 63.8% women; 6.4% Black, 80.9% White	BWL	Active: DRL	No	SQ; Actigraphy (SOL, WASO, TST, TiB, NA)	Fatigue	–	28 sessions (4 wks) [9 wks]	47 (42, 41)	Some concerns; Some concerns
Yennurajalingam et al. 2020 (USA) (104)	Mixed; Active; Stage IV	Mean age: NR; % women: NR; race: NR	a) BWL b) Pharmacological: Melatonin	a) DRL b) Active: Placebo	Yes	Insomnia severity; SQ;	Fatigue; Depression; Anxiety	–	a) 15 sessions b) 20 mg	26 (23)	a) High risk; NA b) High risk; NA
Zachariae et al. 2018 (Denmark) (105)	Breast; Survivors; Stage 0-III	Mean age = 53.1; 100% women; race: NR	CBT-I	Passive: WL	Yes	Insomnia severity; SQ; Sleep diary (SOL, WASO, TST, TiB, NA, SE)	Fatigue	Individual; NO; Remote digital	6 sessions (9 wks) [15 wks]	255 (203, 198)	High risk; NA
Zhang et al. 2017 (China) (106)	Hematological; Active; Stage NR	Mean age = 39.03; 48.68% women; race: NR	Mindfulness	Passive: CAU	No	SQ	Depression; Anxiety	Individual; YES; Face-to-face	5 sessions (5 wks)	76 (65)	High risk; NA
Zhao et al. 2020 (China) (107)	Breast; Survivors; Stage I-III	Mean age = 53.04; 100% women; race: NR	Mindfulness	Passive: WL	Yes	Insomnia severity; Actigraphy (SOL, WASO, TST, SE)	–	Group; YES; Face-to-face	6 sessions (6 wks) [26 wks]	136 (132, 127)	High risk; Some concerns

^a“Country” refers to the place of residence of the participants. CBT-I = cognitive-behavioral therapy for insomnia; BBT-I = brief behavioral therapy for insomnia; ACT = acceptance and commitment therapy; NR = not reported; CAU = care as usual; BWL = bright white light; DRL = dim red light; SQ = sleep quality; SE = sleep efficiency percentage; TST = total sleep time; TiB = time in bed; WASO = wake after sleep onset; NA = nocturnal awakenings; SOL = sleep onset latency; MBSR = mindfulness-based stress reduction; WL = waitlist; PSQ = polysomnography; ACT = acceptance and commitment therapy; NA = not applicable.

^bDelivery not specified for pharmacological and light interventions.

^cIntervention and control groups are reversed according to original study to match grouping of remaining studies in the meta-analysis.

^dData from selected study groups only.

Several interventions were studied, including cognitive-behavioral therapy for insomnia (CBT-I) (K = 18), melatonin (K = 10), bright white light (BWL) therapy (K = 8), mind-body therapies (K = 7), exercise (K = 7), mindfulness-based and other contemporary CBTs (K = 5), sleep medication (K = 4), and brief behavioral therapy for insomnia (BBT-I) (K = 4). One study examined a combination of CBT-I and BWL. Control conditions included passive control (K = 18), placebo (K = 10), active control such as sleep hygiene education (SHE) (K = 25), and comparisons between competing interventions (K = 6). Half of the studies (K = 30) also compared conditions at follow-up with follow-up times ranging from 1 to 52 weeks, averaging 18.5 weeks.

The sleep-related outcomes assessed were insomnia severity (K = 24), with most studies (K = 20) using the Insomnia Severity Index (ISI) (32), and sleep quality (K = 44), with the majority (K = 35) using the Pittsburgh Sleep Quality Index (PSQI) (33). Both measures have been validated in cancer populations (108,109). Additionally, 24 studies used sleep diaries (eg, the Consensus Sleep Diary) (110), monitoring parameters such as SOL, WASO, NA, TIB, TST, and SE (ie, the percentage of TIB spent asleep). Actigraphy (111) was employed in 21 studies, whereas 4 studies used polysomnography (PSG) (112). The most frequently reported subjective and objective outcomes included SE, SOL, WASO, and TST. The secondary non-sleep outcomes of fatigue, depression, and anxiety were included in 29, 19, and 14 studies, respectively.

Insomnia severity

As seen in Table 2 and Figure 2, a large (Hedges’s g = 0.86) statistically significant effect of CBT-I was observed for insomnia severity, the primary outcome, which had been assessed in 13 independent trials. The effect corresponded to a larger mean improvement of 4.6 points compared with controls on the ISI (K = 11), with the scoring range of the ISI at 0 to 28. In the CBT-I condition, participants improved, on average, by 7.8 points from pre- to postintervention (data not shown). Eight studies had included follow-up assessment, yielding a medium (0.55) effect (Figure 3). A statistically significant, medium effect (g = 0.71) was also seen for BBT-I, which had been examined in 3 studies. On the basis of the available data, the supplementary Bayesian analyses indicated that a nonzero postintervention effect of CBT-I was 519 times more likely than the null hypothesis, which corresponds to “decisive evidence” (46). Although the available evidence was very strong for BBT-I at postintervention, with a nonzero difference of 43.5 times more likely than the null hypothesis, the evidence for the efficacy of CBT-I at follow-up was moderate. The number of studies examining insomnia severity did not meet our requirement for meta-analysis (K ≥ 3) for the remaining interventions examined. There was no indication of publication bias (Egger’s test, P = .99).

Table 2.

Results of meta-analyses of randomized controlled trials of interventions for insomnia in cancer patients and survivors—self-reported sleep outcomes

				Heterogeneity				Pooled effect size				Bayesian analysis
Outcome (intervention/subtype)	Time-pointa	Kb	Nc	I 2	T 2	Mean diff.d	95% CI	Hedges’s ge	95% CI	P f	95%PI g	BF h	Level of evidencei
Insomnia Severityj
CBT-I	Post	13	1000	75.0	0.19	4.6	3.2 to 6.0	0.86	0.57 to 1.15	<.001	−0.15 to 1.87	519	★★★★★
-	FU	8	713	81.0	0.22	2.3	−0.1 to 4.8	0.55	0.18 to 0.92	.004	−0.69 to 1.79	4.1	★★
Brief behavioral therapy	Post	3	171	00.0	0.00	4.1	2.4 to 5.9	0.71	0.41 to 1.02	<.001	NA	43.5	★★★★
Sleep qualityk
CBT-I	Post	11	957	80.8	0.22	2.2	1.2 to 3.2	0.52	0.19 to 0.84	.002	−0.60 to 164	4.2	★★
-	FU	6	618	82.3	0.20	0.4	−2.5 to 3.2	0.34	−0.06 to 0.75	.095	−1.03 to 1.71	0.8	○
MBT and contemp. CBTs	Post	4	323	88.0	0.45	2.9	0.5 to 5.3	0.72	0.02 to 1.42	.045	−2.55 to 3.99	1.9	★
Mind-body therapies	Post	5	641	91.9	0.47	0.8	0.2 to 1.4	0.94	0.30 to 1.58	.004	−1.48 to 3.36	3.8	★★
Exercise	Post	6	593	81.2	0.19	1.4	−0.3 to 3.0	0.45	0.06 to 0.84	.023	−0.88 to 1.78	1.8	★
-	FU	3	259	35.3	0.03	1.1	−0.2 to 2.4	0.26	−0.04 to 0.57	.094	−2.72 to 3.24	0.8	○
Bright white light (BWL) therapy	Post	8	368	82.7	0.52	1.5	−0.3 to 3.4	0.58	0.01 to 1.14	.045	−1.32 to 2.48	1.3	★
-	FU	6	302	84.9	0.56	0.4	−2.5 to 3.2	0.09	−0.58 to 0.76	.797	−2.19 to 2.37	0.3	○○
Sleep medication	Post	5	288	59.6	0.11	1.8	1.3 to 2.4	0.56	0.18 to 0.95	.004	−0.67 to 1.79	9.3	★★
Melatonin	Post	6	1152	76.3	0.11	2.5	−0.2 to 5.2	0.35	0.01 to 0.69	.042	−0.69 to 1.39	0.7	○
Sleep efficiency (SE) (diary) (%)
CBT-I	Post	15	1328	65.5	0.10	6.1	3.6 to 8.5	0.63	0.43 to 0.84	<.001	−0.09 to 1.35	4422.9	★★★★★
-	FU	9	657	37.7	0.04	3.5	1.4 to 5.5	0.45	0.25 to 0.66	<.001	−0.08 to 0.99	95.0	★★★★
Sleep onset latency (SOL) (diary) (min)
CBT-I	Post	9	657	44.8	0.05	−8.6	−13.8 to −3.4	0.43	0.21 to 0.65	<.001	−0.16 to 1.02	36.2	★★★★
-	FU	5	290	49.7	0.07	−6.8	−13.3 to −3.2	0.30	−0.04 to 0.64	.083	−0.71 to 1.31	1.6	★
Nocturnal awakenings (NA) (diary) (number)
CBT-I	Post	3	411	73.8	0.10	−0.4	−0.9 to 0.0	0.37	−0.06 to 0.80	.095	−4.52 to 5.26	1.4	★
Wake After Sleep Onset (WASO) (Diary) (min)
CBT-I	Post	14	1302	53.9	0.05	−10.2	−16.1 to −4.3	0.42	0.25 to 0.60	<.001	−0.11 to 0.95	223.4	★★★★★
-	FU	9	657	41.8	0.04	−3.5	−9.2 to −2.2	0.22	0.00 to 0.43	.048	−0.32 to 0.76	1.4	★
Time in bed (TiB) (diary) (min)
CBT-I	Post	3	301	00.0	0.00	−36.0	−46.5 to −24.9	0.73	0.50 to 0.96	<.001	NA	32.7	★★★★
Total Sleep Time (TST) (diary) (min)
CBT-I	Post	14	1325	46.1	0.04	17.2	−3.0 to 37.4	0.05	−0.11 to 0.20	.536	−0.42 to 0.52	0.1	○
-	FU	8	622	15.3	0.01	−6.4	−17.7 to 4.9	−0.01	−0.18 to 0.17	.928	−0.34 to 0.32	0.1	○

^aPost = postintervention; FU = follow-up; Rel = relapse.

^bK = number of studies.

^cN = total number of participants.

^dMean weighted differences only calculated for studies using the Insomnia Severity Index (ISI) (scoring range = 0-28) [32] for insomnia severity and the Pittsburgh Sleep Quality Index (PSQI) (scoring range = 0-21) [33] for sleep quality.

^eHedges’s g: Standardized mean difference adjusted for small sample bias [37], pooled with random-effects models when K ≥ 3. When K ≥ 10, the possibility of publication bias was analyzed with funnel plots and Egger’s test [40]. If the results indicated publication bias, it was planned to impute “missing studies” and calculate an adjusted effect size using the trim-and-fill procedure [41]. None of the analyses showed indications of publication bias.

^f P-values (two-tailed): Statistically significant (P < .05) in bold.

^g95% prediction interval, ie, the interval in which 95% of future observations from the same family of studies will fall. Only calculated when I² > 0.0 [39].

^hThe Bayes factor (BF) [45] represents the posterior probability of the alternative hypothesis (H₁) relative to the probability of the null hypothesis.

ⁱValues <1 indicate evidence in favor of the null hypotheses: weak (0.33-1.00): ○, moderate (0.10-0.33): ○○, strong (0.03-0.10): ○○○, very strong (0.01-0.03): ○○○○, decisive (< 0.01): ○○○○○. Values >1 correspond to evidence in favor of the alternative hypothesis: ^★weak (BF = 1-3), ^★★moderate (3-10), ^★★★strong (10-30), ^★★★★very strong (30-100), ^{★★★★★}decisive evidence (>100) [46].

^jInsomnia severity measures, primarily the ISI (19 out of 22 studies); l) Sleep quality measures, primarily the PSQI (37 of 43 studies).

Figure 2.

Postintervention effects of cognitive behavioral therapy for insomnia on insomnia severity in cancer survivors.

Figure 3.

Effects at follow-up of cognitive behavioral therapy for insomnia on insomnia severity in cancer survivors.

As indicated by the large I² values, the ESs found for CBT-I and insomnia severity were characterized by considerable heterogeneity. When categorical moderators showed sufficient variation, the possible sources of heterogeneity were explored with subgroup analyses and meta-regression. The results are shown in Supplementary Tables 1 and 2 (available online). Older sample age was statistically significantly associated with smaller effects of CBT-I on insomnia severity, whereas a higher percentage of women in the sample was associated with larger effects. When adjusting for cancer type (breast cancer vs other cancers), both moderators remained significant. In contrast, when including both age and gender simultaneously, only sample age remained statistically significant (data not shown). Studies that had included a larger number of CBT-I components, specifically the CBT-I component of relaxation, had statistically significantly smaller effects than the remaining studies. The pooled effect in studies that had included a passive control condition was more than twice as large (g = 1.29) than in studies using active or competing control conditions (g = 0.52). Studies at “high risk of bias” also had larger effects (g = 1.10) than studies with “some concerns” (g = 0.49). When adjusting for the type of control condition, the 3 moderators—number of components, having included the relaxation component, and high risk of bias—no longer reached statistical significance (data not shown).

Secondary sleep outcomes

As seen in Table 2, 45 studies had tested the efficacy in improving sleep quality of 7 different interventions. Statistically significant small (g = 0.35, melatonin) to large posttreatment effects (g = 0.94, mind-body therapies) were found for all interventions (see also Supplementary Figures 1 and 2, available online). The Bayesian analyses indicated weak to moderate levels of evidence, with BFs ranging from 0.7 (melatonin) to 4.2 (CBT-I). No effects at follow-up reached statistical significance, and all Bayesian analyses indicated weak to moderate evidence in favor of the null hypothesis.

In contrast, strong to decisive evidence was found for the efficacy of CBT-I in improving the posttreatment diary-based sleep parameters of SE, SOL, WASO, and TIB. Compared with control conditions, CBT-I, on average, improved SE by 6.1%, SOL by 8.6 minutes, WASO by 10.2 minutes, and reduced TIB by 36 minutes. TST, on average, increased 17.2 minutes more after CBT-I than in controls, but the difference did not reach statistical significance, and the Bayesian analysis indicated weak evidence in favor of the null hypothesis. The effects found at follow-up were generally smaller and reached statistical significance only for SE and WASO. Except for SE, where the level of evidence was very strong, the Bayesian analyses generally suggested only weak evidence for effects at follow-up. Effects of CBT-I on insomnia severity were statistically significantly correlated with effects on sleep quality (Pearson r = 0.86; P = .006; K = 8), diary-based SE (r = 0.65; P = .03; K = 11), and TST (r = 0.67; P = .03; K = 10), but not with SOL (r = 044; P = .32; K = 7) or WASO (0.55; P = .10; K = 10).

The actigraphy- and PSG-assessed sleep parameter results are shown in Supplementary Table 3 (available online). Sufficient data to conduct a meta-analysis (K ≥ 3) were available only for CBT-I, exercise, and BWL. All the effects were small (-0.24 to 0.26) and statistically nonsignificant. The Bayesian analyses indicated weak to moderate evidence in favor of the null hypothesis. There were no indications of publication bias (Egger’s test: P = .61 (sleep quality) to 0.93 (diary-based SE)).

Secondary non-sleep outcomes

As seen in Supplementary Table 4 (available online), small (g = 0.23-0.36) but statistically significant effects of CBT-I were found for all 3 non-sleep outcomes of fatigue, depression, and anxiety. Statistically significant effects were also seen of CBT-I for fatigue at follow-up and for melatonin on depression at postintervention. The effects of the remaining interventions examined (ie, BWL and mindfulness-based therapies and other contemporary CBTs) did not reach statistical significance. The effects of CBT-I on insomnia severity were significantly correlated with the effects on fatigue (Pearson r = 0.89; P = .003; K = 8) and anxiety (r = 0.98; P = .02; K = 4) but not with the effect on depressive symptoms (r = 0.54; P = .27; K = 6).

Risk of bias

The risk of bias assessments are shown in Supplementary Figures 3-6 (available online). For self-reported sleep outcomes, 36 studies (55%) were characterized as having an overall high risk of bias, 26 (40%) some concerns, and 3 studies (5%) had a low risk of bias. Bias stemmed primarily from “bias in selection of the reported result” domain, here mainly because of insufficient preregistered specification of the analyses, and “bias in the measurement of outcome” domain, here because of combinations of nonblinding and a self-reported outcome, and using a nonactive control condition. For the objective sleep outcomes, 9 studies (36%) were characterized as having a high risk of bias, 15 studies (60%) had some concerns, and 1 study (4%) had a low risk of bias. Here, potential bias arose primarily from the “bias in the selection of the reported result” domain, again mainly because of insufficient preregistered specification of the analyses, and from the “bias because of deviations from intended interventions” domain, and here more specifically from nonmasking of the condition to participants and researchers, combined with changes from assigned interventions and the use of non-intention-to-treat analyses.

Discussion

Insomnia severity

The primary aim of the present study was to provide a comprehensive quantitative review of the current evidence for pharmacological, physical, and psychological interventions for insomnia in cancer patients and survivors. The primary outcome measure of insomnia severity was assessed primarily in studies of CBT-I with cancer survivors, with the 13 currently available studies providing decisive evidence in favor of this intervention. A nonzero positive effect was thus 519 times more likely than the null hypothesis. The pooled postintervention effect corresponded to a large effect size and a mean reduction of 4.6 points on the ISI, the most commonly used measure of insomnia. Although the medium effect found at follow-up was less convincing, the available evidence unequivocally supports CBT-I as the first-line treatment for insomnia in cancer survivors.

The heterogeneous results for CBT-I and the relatively wide prediction interval suggest that a considerable proportion of the variation in effect sizes is due to systematic differences in the true effect rather than sampling error (38). Although the moderator analyses pointed toward several potential sources of heterogeneity, after having adjusted for possible confounding factors, the most consistent moderators of the effect of CBT-I on insomnia severity were the type of control condition employed in the trial and the average age of the study sample. Including a passive control condition (eg, care as usual or a waitlist control) was associated with more than twice as large an effect than having included an active or competing control condition (eg, sleep hygiene education, acupuncture, or mindfulness-based stress reduction). Although this result may not be too surprising, it is worth noting that even when compared with active or competing control conditions, CBT-I yielded a statistically significant medium effect. The results also suggested a robust association with age, regardless of cancer type, with younger samples reporting twice as large effects on insomnia severity compared with older samples. This inverse association with age has also been found in studies on CBT-I with noncancer samples (113) and in studies on psychotherapy for depression (114). Although the available data in the included trials do not provide a ready explanation, younger individuals may experience larger effects because of several factors, including greater neuroplasticity yielding more adaptable brains; better cognitive functions (eg, memory and learning capabilities); greater lifestyle flexibility making it easier to implement the necessary behavioral changes; and, perhaps, fewer comorbidities such as chronic pain making insomnia treatment more challenging.

Although our results show CBT-I to be highly efficacious, the intervention involves multiple components, each requiring significant commitment from the patient (115). For patients undergoing active cancer treatment, adding the demands of CBT-I could exacerbate stress and fatigue, making it an impractical or less suitable option. Brief behavioral therapy for insomnia, a less intensive alternative, was investigated in 3 studies with this patient group, demonstrating a statistically significant medium effect size. According to the Bayesian analysis, the evidence very strongly favors a positive effect, suggesting that the scaled-down approach holds promise for patients currently in cancer treatment. However, further research is needed—for example, regarding its long-term impacts as patients transition into survivorship.

Secondary sleep outcomes

We also examined the impact on several other self-reported and objectively measured sleep outcomes. These included global sleep quality, typically assessed using the PSQI (33). Seven interventions were evaluated, with CBT-I being the most studied, followed by BWL therapy, melatonin, exercise, mind-body therapies (eg, relaxation), and contemporary CBTs (eg, mindfulness-based therapies). All interventions showed statistically significant postintervention effects, ranging from small (melatonin) to large (mind-body therapies), but Bayesian analyses provided only weak to moderate evidence of effectiveness. Follow-up studies did not confirm sustained improvements in sleep quality. Surprisingly, CBT-I showed less convincing effects on sleep quality despite its strong impact on insomnia severity. This result could be due to negative or smaller effects in studies assessing sleep quality without focusing on insomnia severity, including studies with active controls (50, 58) or studies not using insomnia as an inclusion criterion (85).

Taken together, the available studies provide robust evidence for the efficacy of CBT-I on several sleep diary-based outcomes. Effect sizes ranged from small to medium, and compared with controls, SOL improved by, on average, 8.6 minutes, WASO improved by 10.2 minutes, TIB reduced by 36 minutes, and SE by 6.1%. The few trials of other interventions that had included sleep diary-based outcomes did not allow for meta-analysis. In contrast to diary-based outcomes, the effects of CBT-I, exercise, and BWL therapy on actigraphy and polysomnography-based sleep outcomes generally approached zero, with Bayesian analyses indicating moderate evidence in favor of the null hypothesis. The observed discrepancies between self-reported and objective sleep outcomes could be related to the commonly reported phenomenon of paradoxical insomnia, a subtype of insomnia where individuals perceive that they sleep considerably less than when measured objectively (116). The available research suggests that individuals with this insomnia subtype appear to have increased brain activity during sleep, making them more aware of their environment and leading them to believe that they are awake when they are actually asleep (117). Although the average sleep duration at baseline across interventions generally differed, depending on whether it was assessed with a sleep diary (6 hours 17 minutes) or actigraphy and polysomnography (6 hours 54 minutes), it is unknown to which degree paradoxical insomnia is present in the included studies of cancer patients and survivors.

Secondary non-sleep outcomes

Finally, we explored the possible effects of improved sleep on secondary non-sleep outcomes. Except for a large pooled effect of melatonin on depressive symptoms in three studies, the results again favored CBT-I, which showed statistically significant effects on fatigue, depressive symptoms, and anxiety, with the effects on fatigue being maintained at follow-up. The effects of BWL therapy and mindfulness-based therapy and contemporary CBTs, on the other hand, did not reach statistical significance. The improvements likely stemmed from better sleep, as indicated by correlations between reduced insomnia severity and fewer comorbid symptoms. Commonly co-occurring across the cancer treatment and survivorship trajectory (118), symptoms such as fatigue, depression, anxiety, and insomnia may have shared roots, such as inflammation (18). These symptom clusters tend to persist, significantly burdening patients and survivors (119). Our findings suggest that addressing insomnia and enhancing sleep quality could be a novel strategy to mitigate fatigue and other challenging cancer-related symptoms and late effects.

Limitations and need for additional research

Although the present study provides a comprehensive overview of the available evidence concerning treatments for insomnia in cancer patients and survivors, some limitations should be noted. First, CBT-I is the most well-studied intervention, and only very few studies focusing on the primary outcome of insomnia severity have investigated other interventions. Although the effects on the broader concept of sleep quality had also been investigated for a wider range of behavioral, physical, and pharmacological therapies, the number of studies on each of these interventions was relatively small, challenging the interpretability of the results. Second, our ability to generalize the findings is limited to some degree by the over-representation of certain demographic and clinical characteristics, including women and survivors of early-stage breast cancer. Third, the lack of data on race and the relatively few studies of patients with other cancers, patients with advanced-stage cancers, and patients in active treatment limits our ability to explore how the efficacy of interventions for insomnia may vary between different patient groups. Finally, one-third of the studies included fewer than 50 participants and may thus be underpowered.

Although the available evidence clearly extends the support for CBT-I as the first-line treatment of insomnia to comorbid insomnia in cancer patients and survivors, the limitations of the current body of research indicate a need for additional research in this population. More adequately powered studies focusing on other interventions and other cancers, including studies comparing the relative efficacy of different approaches, are needed to explore the possibility of providing alternative treatments to patients for whom the full CBT-I program may prove excessively burdensome. Given the challenges of meeting the needs of the many cancer patients and survivors with insomnia, there is also a need to investigate alternative delivery formats. Although they should be regarded as preliminary, the results of our moderator and subgroup analyses indicate that remote delivery formats, including fully automated CBT-I programs, hold considerable promise. Additional research is needed to explore for whom these delivery formats are most efficacious. Finally, the promising results indicating that treating insomnia in this group may also reduce common physical and psychological sequelae such as fatigue, depression, and anxiety, together with the increasing evidence suggesting that insomnia may be a significant factor contributing to poor mental and physical health outcomes in cancer patients and survivors, call for intervention studies examining the possible beneficial effects on other cancer-related mechanisms and outcomes, including prognosis (13).

In conclusion, our comprehensive review provides robust support for CBT-I as the most efficacious intervention for reducing insomnia severity in cancer patients and survivors, with substantial benefits observed across various sleep and non-sleep outcomes. Although other behavioral, physical, and pharmacological treatments showed benefits, the evidence for their effects was considerably less convincing compared with CBT-I, supporting CBT-I as a first-line treatment for insomnia not only in the general population but also in cancer survivors. The results also suggest the potential of less intensive alternatives such as brief behavioral therapy for insomnia, especially for patients undergoing active cancer treatment. The research underscores the intricate connection between insomnia and secondary outcomes such as fatigue, depressive symptoms, and anxiety, suggesting that improving sleep could play a crucial role in alleviating these common, persistent symptom clusters in cancer care. However, the discrepancies between self-reported and objective sleep outcomes indicate the need for further exploration into the complexities of sleep perception and experience in this population, and additional research is needed to explore the relative efficacy of different approaches, to examine different delivery formats, and to investigate the possible beneficial effects on cancer-related mechanisms and outcomes.

Data availability

The data underlying this article are available in the article and in its online supplementary material.

Author contributions

Eva Rames Nissen, PhD (Conceptualization; Data curation; Validation; Writing—review & editing), Henrike Neumann, MSc (Data curation; Validation; Writing—review & editing), Sofie Knutzen, MSc (Data curation; Validation; Writing—review & editing), Emilie Henriksen, MSc (Conceptualization; Data curation; Validation; Writing—review & editing), Ali Amidi, PhD (Writing—review & editing), Christoffer Johansen, DMSc (Writing—review & editing), Annika von Heymann, PhD (Writing—review & editing), Peer Christiansen, DMSc (Writing—review & editing), Robert Zachariae, DMSc (Conceptualization; Formal analysis; Methodology; Supervision; Writing—original draft; Writing—review & editing).

Funding

Not applicable.

Conflicts of interest

None of the authors have conflicts of interest in relation to the research presented in this article.