Review of systematic reviews of non-pharmacological interventions to improve quality of life in cancer survivors

Objectives Over two million people in the UK are living with and beyond cancer. A third report diminished quality of life. Design A review of published systematic reviews to identify effective non-pharmacological interventions to improve the quality of life of cancer survivors. Data sources Databases searched until May 2017 included PubMed, Cochrane Central, EMBASE, MEDLINE, Web of Science, the Cumulative Index to Nursing and Allied Health Literature, and PsycINFO. Study selection Published systematic reviews of randomised trials of non-pharmacological interventions for people living with and beyond cancer were included; included reviews targeted patients aged over 18. All participants had already received a cancer diagnosis. Interventions located in any healthcare setting, home or online were included. Reviews of alternative therapies or those non-English reports were excluded. Two researchers independently assessed titles, abstracts and the full text of papers, and independently extracted the data. Outcomes The primary outcome of interest was any measure of global (overall) quality of life. Analytical methods Quality assessment assessing methdological quality of systematic reviews (AMSTAR) and narrative synthesis, evaluating effectiveness of non-pharmacological interventions and their components. Results Of 14 430 unique titles, 21 were included in the review of reviews. There was little overlap in the primary papers across these reviews. Thirteen reviews covered mixed tumour groups, seven focused on breast cancer and one focused on prostate cancer. Face-to-face interventions were often combined with online, telephone and paper-based reading materials. Interventions included physical, psychological or behavioural, multidimensional rehabilitation and online approaches. Yoga specifically, physical exercise more generally, cognitive behavioural therapy (CBT) and mindfulness-based stress reduction (MBSR) programmes showed benefit in terms of quality of life. Conclusions Exercise-based interventions were effective in the short (less than 3–8 months) and long term. CBT and MBSR also showed benefits, especially in the short term. The evidence for multidisciplinary, online and educational interventions was equivocal.


IntrODuCtIOn
Advances in public awareness, early detection and improved treatments mean that more people are now living with and beyond cancer. For example, Cancer Research UK reports that 50% of people diagnosed with cancer in England and Wales survive 10 years or more, and survival rates have doubled over the last 40 years. 1 This group of survivors includes people at various stages of active treatment, and those in remission, who are gradually restoring their social and occupational roles.
A significant proportion of cancer survivors experience poor quality of life (QoL). 2 The main causes of poor QoL include depression, strengths and limitations of this study ► This is a systematic review of reviews and evidence synthesis of non-pharmacological interventions in cancer survivors. ► Longer term studies are needed and studies of greater methodological quality that adopt similar reporting standards. ► Definitions of survivor varied and more studies are needed for different types of cancer, and specifically for patients who have poor quality of life. ► More studies are needed that investigate educational, online and multidisciplinary teambased interventions. ► This review has some limitations in the methodology.
Studies not in English and grey literature were not included. This was a review of reviews: we did not review individual studies focused on specific cancers or stage, and we did not reassess the quality of the primary studies included in each review.
Open Access anxiety, distress, fear of recurrence and lower levels of social support; impacts on relationships, family and social function; and psychological and social needs, and problems coping. 2 3 The process of diagnosis and treatment is traumatic and disruptive. It is not unusual for patients with cancer to experience distress. Common experiences for those living with and beyond cancer include reduced physical ability, fatigue, changes in sexual activity and developing other medical conditions that affect function for many years. 2 3 If a person is suffering from fatigue, depression or anxiety, they are understandably less motivated to visit friends or engage in social activities; the strain on marital relationships may lead to a loss of support: 25% of people who experience difficulties have broken up with their partner as a result of cancer. 3 4 Thus, the effects of cancer extend beyond the diagnostic and active treatment phases. This review aims to gather the evidence for practitioners, patients and their carers about effective non-pharmacological interventions to improve QoL in cancer survivors. We sought to summarise the effectiveness of non-pharmacological interventions in cancer survivors as part of an (National Institute of Health Research) NIHR-funded programme development grant to inform the design and delivery of a full programme grant.

MethODs
This review of reviews examined existing systematic reviews of non-pharmacological interventions that include information on QoL of those living with and beyond cancer.

Inclusion and exclusion criteria
The study included any systematic reviews that explicitly reported randomised controlled trials (RCTs). Inclusion criteria were organised in accordance with the patient, intervention, comparison, outcome (PICO) reporting structure (see table 1). The population of interest was people living with and beyond cancer, who were aged 18 years or more, and who had received their cancer diagnosis as adults.
We defined non-pharmacological interventions as those that did not involve any drug or medicine, but they could include educational, behavioural, psychosocial approaches or physical activity; we excluded complementary and alternative therapies as defined by the NHS Choices resource. 5 However, we included physical activity and psychological approaches that were part of yoga-based interventions after consulting with patients in the development of the review. Comparators were not specified for the purpose of the inclusion criteria of the review of reviews, but comparators reported in the original reviews were considered in the analysis.
The primary outcome was QoL defined by physical, psychological and social functioning. We reported on studies that used an established and validated measure of global or overall QoL; some of these are cancer-specific. In the literature, the terms 'Quality of Life' and 'Health Related QoL' are used interchangeably; therefore, both are included under the term 'QoL' in this review. The study settings included any healthcare venue, such as hospital inpatient or outpatient services and community services, and also included home and remote e-technology-based interventions.

Data sources
We searched the following databases: PubMed, Cochrane Central, EMBASE, MEDLINE, Web of Science, the Cumulative Index to Nursing and Allied Health Literature, and PsycINFO. The final search was from inception to May 2017 and is shown in online supplementary annex 1. We consulted experts in the field to assess completeness of the list of identified reviews, and where necessary contacted authors to secure the full-text versions. study selection Two authors (MD, JD) independently screened all titles and abstracts of studies identified by the search strategy against inclusion and exclusion criteria, and when eligibility was determined the full text was read. Discrepancies around inclusion were resolved by discussion or in consultation with Open Access a third author when required (KB). We searched the reference lists of all included reviews to identify any further relevant reviews. The research team was not blinded to authors. Citations were downloaded and managed in an EndNote library.

Data extraction
Two authors (EM, EH) independently extracted data from each of the eligible reviews into a purpose-built, predesigned, structured template. The data extraction forms were then summarised in a table and  Assessment of methodological quality of included reviews The methodological quality of the systematic reviews was evaluated using Assessing Methdological Quality of Systematic Reviews (AMSTAR), 6 a measurement tool for the assessment of multiple systematic reviews that has good reliability and validity (table 2). The AMSTAR checklist used can be found at https:// amstar. ca/ Amstar_ Checklist. php.

Data analysis and narrative synthesis
The intervention components were listed, followed by a narrative synthesis. 7 This included understanding components of the interventions, exploring patterns of findings across studies and within primary reviews, and giving greater weight to studies of higher quality in the interpretation of the findings, especially if there were contradictions between the findings of reviews. Ultimately, the purpose was to put into text format the key findings from the most robust evidence available, to guide treatment and future research recommendations. The synthesis set out reported effect sizes across studies, means and SD. Meta-analysis was not undertaken, due to heterogeneity of methods, outcomes and absence of reported effect sizes (10 reviews did not provide effect sizes). The publications were segmented into those reporting meta-analyses to which the greatest weighting was given in the synthesis; some reviews did not undertake or report meta-analyses but rather reported each study, trends and the range of effect sizes; a third group reported no effect sizes but provided narrative statements.

Patient and public involvement
Patients and carers (and respective organisations) were involved in the design and development of the programme development grant application (from which this review is one output). Patients and carers attended all the steering group meetings and were an integral part of the research team, commenting on and critiquing the inclusion and exclusion criteria, outcome selection, and the acceptability and likely value of interventions. As part of the steering group, they received and commented on study progression, emergent findings and reports. They are integral to the dissemination plans, including sharing the publication, but also helping craft lay summaries of the overall research project and key findings. A public -patient representative (EH) performed the data extraction together with research and clinical colleagues,  26 6 Moderate *The maximum score on AMSTAR is 11 and scores of 0-3 indicate that the review is of low quality, 4-7 of moderate quality and of 8-11 as high quality.

Open Access
and coauthored and edited the review. Public-patient representatives were also part of the steering group and informed the design and delivery of the review.

Participants
The number of patients included in the reviews ranged from 262 8 to 7164. 9 Thirteen reviews covered mixed tumour groups, [10][11][12][13][14][15][16][17][18][19][20][21][22] seven specifically focused on breast cancer 8 9 23-27 and one on prostate cancer. 28 Intervention type and components Face-to-face delivery of interventions was often combined with online delivery (three reviews) 9 24 28 ; others included telephone communication (five reviews) 9 11 23 25 26 and printed information (two reviews). 11 25 Four reviews included interventions that provided supplementary compact discs, manuals or video tools. 11 23 24 28 Two reviews were from inpatient rehabilitation. 8 18 None of the reviews reported the use of structured manuals, and interventions were often not fully described or broken down into different components, nor was there attention to a mechanism or theory of change.
Ten of the reviews focused on physical interventions, 10             Open Access yoga 11 14 23 ; four reviews were of psychosocial or behavioural interventions 9 17 24 27 ; and one review focused on online interventions including connecting patients and online education (see tables 3 and 4). 15 One review compared multidimensional versus monodimensional interventions, 18 and one tested multidisciplinary rehabilitation models. 8 Finally, one review focused on the effects of expressive writing. 22 The duration and frequency of the interventions varied greatly from a single 20 min session 17 to 60 weekly sessions. 10 The most common components of physical interventions were aerobic exercise 9 10 12 13 16 19 26 and resistance/strength training. 9 10 12 13 16 26 Psychological education 8 9 17 18 24 and cognitive behavioural therapy (CBT) 9 17 18 24 were the most commonly used psychological and educational interventions. Peer support was often used as a psychological and a behavioural intervention. 8 9 15 Components of the interventions were thematically organised into two groups (see table 4 for a more detailed itemisation): biological or physical actions (19 types of activity or diet change), and psychological, behavioural or educational (24 types of intervention about mind and body, including CBT, mindfulness-based stress reduction, psychosexual therapy, supporting existing coping methods, emotional support, relaxation, psychotherapy and psychosocial therapy, and interventions focusing on social support, guided imagery, self-management, use of peer support, bibliotherapy, telephone and web-based interventions, and return to work interventions).

Overall effectiveness of interventions: meta-analysis findings
Meta-analyses were reported in 11 reviews and the effect sizes (as reported in the original reviews) are tabulated (table 5). Of six publications providing meta-analyses of physical activity (not including yoga), all found convincing positive associations for studies testing response between 1 and 26 weeks post-treatment. Long-term effects were not tested by all, although Fong et al and Zeng et al did show persistent effects at 6 months and a year, respectively. 10 26 One review 19 showed uncertain outcomes at 3-6 months, although shorter and longer term outcomes were favourable. This review showed equivocal effects when the intervention group was compared with the control group, once adjusted for QoL and covariates at baseline. The two meta-analyses of yoga interventions showed positive effects, 11 23 as did a review of CBT. 17 There was no evidence of benefit in QoL following patient education 17 and behavioural interventions. 9 Two reviews reported effect sizes from individual studies but did not undertake meta-analyses. 18 24 Mewes et al's 18 review of multidimensional rehabilitation included 10 studies, 9 of which had global QoL outcomes; of these, 7 showed benefit, with effect sizes ranging from 0.04 to 0.99 (no CIs reported). Fors et al's 24 review included six RCTs only, four of which included a QoL measure; two of these showed positive effect sizes (ranging from 0.56, 95% CI 0.09 to 1.03; 0.63, 95% CI: 0.11 to 1.18); one showed improved and one a worsening of QoL as a non-standardised mean score. Five reviews 8 13 15 16 25 did not report meta-analyses or effect sizes; mostly these provided mean change scores or narrative statements. On the whole these gave a mixed picture, often resorting to subgroup analysis by cancer type or different dimensions of QoL.
Physical activity: summary findings Cramer et al's 23 high-quality review of 6-12 weeks of yoga in patients with breast cancer showed a large increase in general QoL, a finding that was consistent with reviews by Buffart et al 11  There was inconsistency across the reviews with regard to the types of exercise interventions that were most effective. Fong et al 10 found aerobic plus resistance training to be significantly more effective than aerobic training alone on many aspects of QoL. However, Zeng et al's 26 moderate-quality review suggested that single types of exercise interventions (general aerobic, yoga or tai chi) were more effective at increasing QoL at 4-52 weeks after intervention; half of the studies assessed interventions between 8 and 12 weeks. Duijts et al's 9 study of patients with breast cancer found only small effects of physical activity on QoL (at 8-26 weeks after intervention), and Spence et al's 16 study of mixed but mostly patients with breast cancer reported evidence that physical activity improved overall QoL, but only four of ten trials maintained the intervention and only a fifth of trials seemed to assess outcome at 3 months and beyond. Zeng et al's 26 review of patients with breast cancer found small but positive benefits of physical activity on overall QoL. Galvão and Newton's 13 review of mixed cancers gave preliminary evidence of positive benefits on a Modified Rotterdam QoL measure, but no overall effects were reported. However, Spark et al's 25 review of patients with breast cancer showed that the impact of physical activity on QoL was not convincing. Although Spark et al did not report effect sizes, two of the studies in that review included QoL measures, both of which reported effect sizes in the original papers: one showed positive benefits on Functional Assessment of Cancer Therapy-General (FACT-G) and Functional Assessment of Cancer Therapy-Breast Cancer at 8 months (effect sizes 9.8-13.4), but not at 24 months of follow-up; the other showed no significant effects on FACT-G overall, but when the cancer-specific FACT-G Open Access Table 4 Components of the interventions by study  individual interventions were more effective than groupbased treatment. CBT showed both short-term 24 and longterm improvements in QoL. 17 Five primary papers in one review assessed the effect of social and emotional support as an intervention, four of them finding no effect, and one reporting a significant improvement in QoL on one measure. 24 There was no evidence that psychosocial education increased QoL. 17 24 Multidimensional and multidisciplinary rehabilitation Khan et al's 8   Open Access writing interventions, but this found no benefit on QoL, although small effects would be undetected. 22 Individuals with low levels of emotional support appeared to benefit more than others.
Adverse effects Five reviews 11  Three participants in one study reported thrombosis and infection following exercise interventions. Another study found hip pain, sciatica, arm discomfort (n=4), knee discomfort (n=10), ankle discomfort (n=3), and foot discomfort (n=8) with asymptomatic ischaemia and conduction problems on ECG. A further study reported lung metastases, pulmonary embolism and palpitations.
Another study reported soft tissue injury following exercise, and cholecystitis following stroke. Cancer recurrence, although not a direct effect of interventions, was common and another reason to stop participation in the research.

DIsCussIOn
Main findings Twenty-one reviews were included and showed a lack of definitive and consistent evidence across 465 primary studies, of which 362 were RCTs. In part this is explained by substantial variation in study designs and outcome measures used to indicate QoL. All systematic reviews of physical activity demonstrated improved overall QoL, but few studies assessed long-term outcomes beyond 3 months, and even fewer assessed outcomes beyond a year after the intervention. More focused research and a consistent approach are required to explore the effect on the subdomains of QoL. 12 A higher quality review suggests that aerobic plus resistance training provides maximum improvements in QoL. 10 There was more evidence of physical rather than psychological or other types of interventions.
One of the included reviews for psychological or behavioural interventions was of high quality. 27 CBT is effective for improving QoL in the short and long term, 17 24 especially when provided as an individual intervention. 17 There is not much evidence to support †Random effects assumption. ‡Findings sustained for random or fixed effects, random effects reported.
CBT, cognitive behavioural therapy; FU, follow up; QoL, quality of life; SMD, Standardised Mean Difference.

Table 5 Continued
Open Access comparative effectiveness of intervention modalities such as group versus individual, monodimensional versus multidimensional or multidisciplinary; further work is needed to examine these different approaches. Given the accessibility of social media and its popularity, the findings that email contact was related to poorer QoL need further investigation; although interactive websites were beneficial, the overall findings about digital interventions were equivocal.

limitations
The current review has some limitations in the methodology. Studies not in English and grey literature were not included due to time constraints as the review was undertaken as part of a programme development grant to inform the design of a future research programme application.
We encountered some methodological limitations in included reviews. Some used multiple outcomes and often had a very broad understanding of QoL and used diverse measures of QoL. There was no consistent reporting standard.
We did not consider outcomes such as well-being or the multiple subdomains of QoL to avoid the risk of generating findings due to multiple testing in smaller subsamples in underpowered analyses. Some reviews included few primary papers. We examined the sample sizes of RCTs included in reviews and whether there seemed to be any relationship with AMSTAR ratings. We found no obvious relationship, given AMSTAR scores refer to review quality rather than the quality of or sample size of individual RCTs. A review of primary RCTS might help to better understand and report robust findings from RCTs with large and adequate sample sizes, findings which may otherwise be less visible in a review of reviews.
We found little overlap between reviews (tabulation available on request), reflecting their specific inclusion and exclusion criteria and interest in very specific interventions and cancer types. We did not evaluate the methodological quality or bias of the original studies within each systematic review. Ten reviews planned to assess publication bias; three of these could not perform any specific tests of bias due to small samples. 8 23 27 Consequently seven studies tested for publication bias. 9 10 12 17 19 20 22 Three of these reported that publication bias was not significant. 10 20 22 Four reviews 9 12 17 19 reported significant publication bias suggesting caution in assuming there is definitive evidence for exercise and CBT.
The physical and psychosocial concerns of patients at different time periods of the cancer experience will vary greatly, and interventions effective at one stage may not be suitable for another. Most reviews defined 'survivors' as those who had completed active treatment before the onset of the study. 10 13 14 16 18 19 23 24 26 Some specified a time frame, from immediately after surgery to 15 years after active treatment. 12 One review defined survival as being from diagnosis onwards. 17 Another included terminal stages of cancer. 15 The majority of the reviews incorporated studies combining patients during and post-treatment. 9 11-15 23-25 These differing definitions of living with and beyond cancer make comparison difficult, and a standardised approach to trials and reporting of studies is needed.
Interventions were offered to patients based on their diagnosis of cancer, rather than low QoL, which may have led to underestimation of potential beneficial effects. Future research should consider the effectiveness of interventions targeting people living beyond all types of cancer and with poor overall QoL.

COnClusIOns
Systematic reviews of patients with cancer and their QoL showed that effective interventions included physical activity, CBT and mindfulness-based stress reduction training. Personalised lifestyle interventions showed promise, as did social and emotional support. Educational and information provision appears ineffective, and there were few studies of electronic interventions. Currently, there is no standard study design, outcome selection or reporting convention adopted across these reviews. No single intervention can be recommended to those patients with a poor QoL following cancer treatment as interventions were not targeting poorer QoL, but cancer survivors in general.