Randomized control trial evidence for the benefits of massage and relaxation therapy on sleep in cancer survivors—a systematic review

A comprehensive data search was performed on PubMed, Scopus, Web of Science, PEDro, and CINAHL from inception to September 2020. The search terms used for cancer were the following: Carcinoma (MeSH), Cancer, Neoplasms (MeSH), Malignancy and Tumor. Sleep (MeSH), sleep quality, sleep disturbances, Insomnia, Sleep wake disorders (MeSH), Sleep initiation and maintenance disorders (MeSH) and Sleep disorders were the search terms related to sleep disturbance. For the interventions, Relaxation techniques, Relaxation therapy (MeSH) and Massage (MeSH) were the terms used. The search terms were combined with Boolean operator ‘AND’ or ‘OR’ wherever relevant. The reference list of the included articles was also screened for potentially relevant studies. The search was limited to RCTs involving human participants that were published in English.

After the results from each database were compiled and duplicates removed, title and abstracts were screened independently by 2 reviewers (RG and PV). Any disagreements were sorted by discussion with SRS. The screening for eligible articles was done based on the pre-set criteria. To be included in the systematic review, studies had to (1) include cancer survivors of any ages, type of cancer, or treatment status; (2) have massage or relaxation therapy as the primary intervention; (3) include comparisons to usual care or active/sham treatments; (4) report self-reported or objectively measured sleep outcomes; and (5) involve RCT designs, with the results reported in peer-reviewed journals. Cancer survivors were defined according to the National Cancer Institute definition that states “a person is considered to be a survivor in cancer from the time of diagnosis until the end of life” [23]. Studies reporting any form of massage such as deep friction massage, Swedish massage, and its components used for therapeutic benefit were included in the review. Studies evaluating relaxation therapy techniques used independently, consisting of breathing exercises, somatic relaxation, and progressive relaxation techniques were included in the review. However, other forms of complimentary treatments that included relaxation therapy as a part of the treatment were excluded.

All the eligible studies were screened further based on full text independently by RG and PV. Any disagreements were sorted after discussion with SRS, VK, and JK. A data extraction template was developed for the eligible studies which included information on the demographic characteristics of the participants, description of the intervention, comparator, treatment status, and key findings. Data extraction was performed by RG and PV and disagreements were sorted by discussions with SRS, VK, and JK.

The assessment of risk of bias for the included studies was performed independently by RG and PV using the Cochrane risk of bias tool [24]. The studies were scored under the domains of random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias. The judgment of yes was marked as low risk and no as high risk. Unclear was marked if the details were missing or unknown bias. Any disagreements in the markings between the 2 reviewers were sorted by discussions with SRS.

As seen in the PRISMA flowchart (Fig. 1), a total of 371 articles were retrieved from the data search (PubMed-77, Scopus-159, Web of Science-78, CINAHL-2, and PEDro-55). A total of 278 articles were screened based on title and abstract after removing the duplicates. Of the 22 articles that were found to be potentially eligible and underwent full-text screening, 7 articles which met the inclusion criteria were included in the review. The most common reasons for exclusion of the articles included non-randomized controlled trial methodologies and a lack of reported sleep outcomes.

The summary of the risk of bias assessments of the included studies has been represented in Fig. 2a and b. Of the 4 massage therapy studies, a high risk of bias was reported for Incomplete Outcome Data [25] and Other Biases [26]; with unclear risk of bias reported for Random Sequence Generation [25, 27, 28], Allocation Concealment [23, 26‐28], Blinding of Participants [25, 27], and Blinding of Outcome Assessment [25]. Of the 3 relaxation therapy studies, a high risk of bias was reported for Incomplete Outcome Data, Selective Reporting [29, 30], and Other Biases [29‐31]; with unclear risk of bias reported for Allocation Concealment [29], Blinding of Participants [29], and Blinding of Outcome Assessment [29].

Among the 7 included studies, 6 studies involved adults from 18 to 78 years [25, 27‐30] and 1 involved pediatric population with an age range of 4–8 years [31]. Four studies included participants with mixed type of cancers [25, 27, 29, 30], 3 studies included participants with metastatic cancer [25, 28, 29], and 2 studies included participants with leukemia [26, 31].

Four out of 7 RCTs studied the effect of massage therapy [25‐28], with the other 3 examining relaxation therapy [29‐31] as their intervention. The total number of participants recruited for the 4 RCTs studying the effect of massage therapy was 205, out of which 187 completed the entire study. There were 50 participants recruited across the 3 RCTs studying the effect of relaxation therapy, with 33 participants completing the study.

The 4 included RCTs that studied the effect of massage therapy, delivered either by massage therapists or nurses trained in massage, are summarized in Table 1. The duration of the interventions ranged from 2 days to 4 weeks, involving 2 to 3 massage sessions per week, with each massage therapy session ranging from 10 to 45 min in duration. Two studies compared the effect of massage with usual care [26, 27], 1 study compared the intervention to a social attention control [28], and the remaining study compared massage with usual care and no-touch control interventions [25]. One study used a MiniMotionlogger actigraph with sleep diary to objectively measure sleep [27], whereas the remaining 3 studies used self-reported questionnaires including the Richards-Campbell Sleep Questionnaire (RCSQ) and Pittsburg Sleep Quality Index (PSQI) to assess sleep quality [25, 26, 28].

Significant improvements in self-reported sleep quality were reported in 2 studies, with this including a significant between-group improvement in PSQI [26] and a trend for a significant between-group improvement in RCSQ (p = 0.07) after the first day of the massage intervention [28]. The only significant benefit observed for any objective sleep outcomes was reported in a study utilizing the Motionlogger device, whereby Jacobs et al. [27] observed a significant between-group increase in the number of nighttime long sleep episodes for the massage compared to the control group.

Feasibility of the massage interventions were reported in the studies conducted by Toth et al. and Jacobs et al. [25, 27]. Toth et al. [25] reported that it was feasible to provide in-home massage therapy to the participants by professional massage therapists, with each participant averaging 2.8 sessions of the required 3 sessions during the 1-week trial. In the study of Jacobs et al. [27], the goal was for massage therapists to provide 2–3 massages to adolescent patients with cancers who were staying in the hospital for a minimum of 4 days. Results indicated reasonable feasibility, in that 94% of the participants received 1 massage and 69% received 2 massages during their hospital stay. While feasibility outcomes were not directly assessed in the other studies, some indication of the feasibility for the other 2 studies may be provided by comparing the number of participants initially recruited with the number of participants who completed the study. Of the 72 participants recruited in the study by Jane et al., 2 and 3 dropped out by the second session (T2) in the massage and social attention group, respectively [28]. In the study conducted by Miladinia et al., out of the 64 recruited participants, only 3 participants in the massage therapy group and 1 in the usual care group were lost to follow-up [26].

The safety of these massage interventions was examined by monitoring the number of adverse events in 2 of the 4 studies. No adverse events related to the massage therapy were reported in either of these studies [25, 28].

As summarized in Table 2, there were 3 RCTs that studied the effect of relaxation therapy on sleep disturbances in cancer populations [29‐31]. All 3 RCTs were pilot studies with the duration of intervention lasting from 3 to 60 days that examined if a variety of sleep outcomes could be improved compared to usual care. Two out of the 3 studies provided some degree of relaxation guidance by the therapist, with 1 study appearing to only provide relaxation resources and written instructions [30]. The outcome tools used 3 self-reported questionnaires including the Numerical Rating Scale of Satisfaction of Sleep (NRSSS), Patient-Reported Outcomes Measurement Information System (PROMIS), and the Children Sleep Habits Questionnaire (CSHQ) as well as objective sleep data from the accelerometers including the Minimotionlogger actigraph.

No statistically significant improvements were observed in any sleep outcomes as a result of relaxation therapy, although there were some potential trends observed. Specifically, a trend for significant improvements in self-reported sleep quality was observed after the second day of relaxation training as assessed by the NRSSS [29], although no exact p value was provided. Sleep disturbances, as assessed by the PROMIS, were also concluded to have been improved by the authors in 2 of the 3 participants in the relaxation group and in 1 of the 3 participants in the relaxation plus guided imagery groups in the case series by Nooner et al. [30]. The authors also stated that all but 1 of these participants reported more refreshing sleep and fewer sleep symptoms after 30 days of relaxation therapy [30]. However, it must be acknowledged that none of these improvements in the Nooner et al. [30] study appears to have undergone statistical comparisons. Of the 7 objective sleep outcomes assessed using an actigraph by Zupanec et al. [31], there was a trend for an improvement in the wake time after sleep onset (p = 0.08) for the relaxation compared to the usual care group.

Feasibility data for relaxation therapy was provided by Zupanec et al. [32]. The 8 participant families (childhood cancer survivor and their parents) who completed the study rated the education session to be somewhat or very useful, with the sleep tips as well as the relaxation books and associated techniques reported to be the major resources used [31]. The major barriers faced related to the children’s (1) fatigue levels being so high that daytime naps were still required, with such naps potentially negatively affecting the time required for implementing the relaxation intervention; late-night medication interfering with early bedtimes; and (3) resistance to reading the same relaxation book every day [31]. Although feasibility was not directly assessed in the remaining studies, some indication of the feasibility of these interventions may also be provided by examining the number of participants who started and completed each of the relaxation therapy programs. For example, Ducloux et al. [29] recruited 9 participants into both their immediate and delayed intervention groups, with 7 participants completing the immediate intervention but only 4 completing the delayed intervention. Twenty participants were randomized to the Nooner et al. study. Nine of the 11 relaxation therapy participants and all 9 usual care participants completed the study [30]. In contrast, of the 3 participants allocated to the each of the 4 groups (relaxation, guided imagery, relaxation and guided imagery, and usual care), the relaxation group maintained only 2 participants at 30 days and 1 participant at 60 days, compared to the combined relaxation and guided imagery group who maintained 2 participants at 30 and 60 days. The dropouts in the study of Nooner et al. [30] were attributed to death, loss to follow-up due to complications of treatment, and patient request. The reasons that contributed to some participants discontinuing the Nooner et al. [30] study were described in a participant evaluation of the study. Specifically, the negative perceptions of the intervention that may have increased dropout rates were that the relaxation program may have been “too much to keep up” during complicated periods of cancer treatment and that completing the symptom log was too much work.

There appeared to be no monitoring of adverse events in any of the 3 relaxation therapy studies [29‐31].

To the best of our knowledge, this is the first systematic review summarizing the effect of massage or relaxation therapy for sleep disturbances in cancer survivors. We used a list of comprehensive search terms developed by in-depth search and discussions among the review team, and the search was run on a wide range of databases including PEDro which is specifically for physiotherapy interventions. The observation that some of the reviewed studies consisting of participants with metastatic cancer obtained some significant benefits in sleep outcomes was of interest, as there appears to be no current RCT-level evidence regarding the effectiveness of other non-pharmacological interventions for improving sleep outcomes in cancer survivors with metastatic cancer. Furthermore, the inclusion of self-reported sleep outcomes that provided insight into the patients’ perspective was also considered a strength of the studies included in this review.

The limitations of this review process include restriction of articles to English and also the exclusion of gray literature. Unfortunately, there are also many limitations of the literature that were included in the review. Several of the included studies, especially those examining relaxation therapy, were unable to be described as low risk of bias due to the missing information reported regarding selection, performance, and detection bias. Many of the studies also did not provide detailed information regarding the actual massage or relaxation therapy intervention, which then may make replication of the intervention by practitioners or other researchers difficult. We also acknowledge that the included studies typically involved a small number of participants, cancer types, stages of cancer, and current treatments, with only one of the studies having inclusion criteria specifying that all participants needed to have been diagnosed with a sleep disorder by the International Classification of Sleep Disorders definition. With respect to the sleep outcomes, while there were some objective outcomes provided via accelerometry-derived data, such measures are indirect and not an example of outcomes recommended to be used in insomnia research and clinical practice [38]. Some relaxation therapy studies had a high dropout rate, which suggests there may be some issues in the widespread uptake of this therapy in cancer survivors. Furthermore, as many studies utilized a relatively large number of sleep outcomes but at best only found significant improvements in a small subset of these outcomes, there exists the potential for type I errors to have occurred. Finally, none of these studies utilized a follow-up period to assess the length of time the treatment effects may last post-therapy.