Physical activity and nutrition interventions for older adults with cancer: a systematic review

The conduct and reporting of this review adhere to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) [45]. A data charting/extraction form was adapted from the Johanna Briggs Institute (JBI) Reviewers’ Manual: Methodology for JBI reviews (2015) [46]. A copy of the final form can be found on our open science framework page (https://​osf.​io/​p23jd/​).

Studies were included if they met the following a priori eligibility criteria: (1) delivered a lifestyle intervention for nutrition and/or PA to people with any cancer diagnosis, (2) included a measure of HRQoL, (3) participants over 60 years or at least 50% over 60 years with data analyses by age group and (4) randomized controlled trials. Studies were excluded if (1) we could not determine an age range, (2) the intervention was targeting clinicians or carers rather than older adults with cancer, (3) publication language was not in English or (4) findings were conference abstracts only.

Studies were identified through structured searches of all publication years (final update search performed 30 May 2019) in the following electronic databases: Medline via OVID, Embase via OVID, Cochrane Central Register of Controlled Trials (CENTRAL) and Cinahl via EBSCO. The search strategy was developed in consultation with a specialist librarian at the University of Hull and finalized with the aid of an information specialist. MeSH terms in Medline (see supplemental file 1) were developed to search for all key concepts and modified for other databases. Keyword searches restricted to abstract and title were also completed. Boolean logic was used to combine the terms. The original database searches were conducted by a single author (CF) and updated by an information specialist (SG). For the updated search, search filters for RCTs including Cochrane’s Highly Sensitive Search Strategy were included to retrieve randomized controlled trials.

All identified articles were uploaded into an EndNote X8 database and duplicates removed. Preliminary screening was undertaken by one author (CF) to remove obvious exclusions (e.g. conference abstracts, etc.) after which two authors (CF and SG) independently screened all articles against eligibility criteria taking title, abstract and full-text into account. Disagreements were discussed and resolved by consensus. Any unresolved items were reviewed by a third author (FS) and their decision stood. If criteria were unclear in the manuscript, corresponding authors were emailed and asked for clarification.

A data extraction form was developed and piloted by the research team to extract data about study details and characteristics (e.g. country, setting, sample characteristics, etc.), intervention details (e.g. group descriptions, intervention components and duration, etc.), QoL outcomes and key findings and messages. The form was independently tested using one article by two authors (CF and SG) and revised following discussion. Data were then extracted using the form by a single reviewer (CF). A second (SG) and third (FS) reviewer randomly selected two articles each (i.e. 25%) and reviewed the data extracted. As there were no discrepancies, data extraction by a second reviewer for the remaining articles was considered unnecessary.

Two authors (CF and SG) used the Risk of Bias Assessment Tool version 2 available from the Cochrane handbook (2011) to independently assess quality of life outcomes from all included studies. The articles were judged for bias as either low, high or some concerns for the following: (1) selection (random sequence generation and allocation concealment), (2) performance (blinding of participants and study personnel), (3) detection (blinding of outcome assessment), (4) missing outcome data and (5) reporting (selective outcome reporting). The nature of lifestyle behaviour change studies means double-blinding is very difficult but this tool allows fair judgements despite this fact. The authors discussed any differences and reached consensus; therefore, a third party was not necessary.

To describe the nature of studies currently targeting older adults with cancer, we extracted detailed information related to the intervention groups including (1) type of intervention, (2) intervention delivery methods, (3) all components of intervention, (4) study duration and measurement timing and (5) comparator group information. The primary efficacy outcomes of interest for this review were measures related to QoL or HRQoL. The primary outcomes for each study were identified and noted.

The study selection process is presented in Fig. 1. A total of 6490 records were identified; 5179 remained after de-duplication. After title and abstract screening, 114 articles were identified for full-text review. Of those, 14 studies were deemed eligible and were included in full data extraction for this review.

This review included 14 papers [47‐60]. A further eight related papers were referenced to provide more detail when unable to find information in the primary paper (e.g. protocol papers) [61‐68]. One study was a three-arm randomized controlled trial (RCT) [47] with the remaining 13 studies being two-arm RCTs. One study was described as a pilot [51] and seven as feasibility [47, 48, 50, 54, 55, 58, 60] RCTs. Seven studies took place in the USA [49, 50, 55, 56, 59, 60], two in Canada [47, 51] and one each in China [52], Japan [54], Korea [57], Sweden [58] and UK [48]. Five studies recruited patients with prostate cancer only [47, 48, 55, 57, 60], one each recruited bladder [58], lung [52] and breast cancer only [51], with the remaining recruiting a mixed sample of cancer types [49, 50, 53, 54, 56, 59].

Full results from the assessment can be found in Figs. 2 and 3. Based on assessments from two reviewers, no studies received overall risk of bias judgements of low, ten were judged to have some concerns and four had high risk of bias. Twelve studies were low [47‐51, 53, 54, 56‐60] and two some concerns [52, 55] for allocation; ten were low [47, 49, 52‐54, 56‐60], two some concerns [48, 50] and two high [51, 55] for intervention deviations; nine rated low [47‐50, 52‐54, 56, 59], three some concerns [55, 57, 60] and two high [51, 58] for missing data; one ranked low [52] and 13 some concerns [47‐51, 53‐60] for outcome measurement; and finally, 11 rated low [47‐57], two some concerns [58, 59] and one high [60] for selective reporting.

Most interventions focused on PA behaviour only [47, 51‐55, 57‐60] and three on both nutrition and activity behaviour [48, 49, 56]. One study reported improving nutrition as its main objective; however, this was a gardening intervention which could also be considered activity [50]. The studies ranged in duration from 7 days to 1 year, six studies [48‐50, 56, 58] collected follow-up measures beyond post-study though only three studies reported this data in the included articles [48, 49, 58]. Intervention settings included hospital [52‐55, 58], community [57, 59, 60], home [49, 50, 56] or a mixture of settings [47, 48, 51]. Detailed characteristics of the included studies can be found in Table 1.

Eight studies included some specific form of aerobic activity [47, 48, 51, 52, 54‐56, 58], five had traditional-style strength training (e.g. lifting weights or using resistance bands) [47, 48, 56, 58, 60] and five included specific training to build balance and/or functional strength [51, 52, 57‐59]. Seven interventions included at least one supervised individual session [47, 52‐55, 57, 60], seven had home-based activities [47‐51, 53, 56] and four were group-based classes or training sessions [47, 48, 58, 59]. Home-based interventions were delivered using DVD [51], booklets/binders [48‐50, 53, 56] and four included visits or telephone calls to check-in and determine progress [47, 49, 50, 56]. Six home-based studies included personalized advice on activity [47‐49, 51, 55, 56].

One study provided participants with all supplies and guidance on growing their own vegetables at home and had role models in the form of a Master Gardener to teach and assist [50]. One other study provided portion-control tableware to assist eating habits and included tailored nutrition advice as compared to national guidelines [56]. One study had personalized comparisons to general information regarding standard nutrition guidelines [49]. One study had a series of “healthy eating” seminars and a nutrition advice pack [48].

Three studies explicitly stated they used a theory to develop study materials including social cognitive theory [49, 50, 56], transtheoretical model [49] and social ecological model [50]. Other studies included information regarding habit formation [48], autonomy [48], self-efficacy [49] and action/coping planning [47, 50]. Aside from those studies with group-based session, to increase social support, one study formally included spouses or partners in the intervention [60] and one had a private Facebook group for study participants [50]. Detailed descriptions of the interventions included can be found in Table 2.

Study feasibility was stated as the primary outcome for seven studies [47, 48, 50, 54, 55, 58, 60], of which six were deemed feasible based on recruitment, retention, adherence, and compliance rates [47, 48, 50, 54, 55, 60]. Though attendance and compliance rates were high, one study was deemed not feasible due to the large number of dropouts owing to the more severe illness of bladder cancer patients [58]. Change in activity or diet behaviour was a primary outcome for two studies [48, 49] and five focused on physical [52, 56‐58] or cognitive functioning [54]. Seven found significant improvements in the primary outcome [48‐50, 52, 54, 56, 57] while one found no significant group difference [58], though this was a feasibility trial. Further details can be found in Table 3.

To measure QoL, five studies used the Medical Outcomes Study’s Short-Form 36 (SF-36) [50, 56‐58, 60], five used all or part of the general Functional Assessment of Cancer Therapy (FACT-G) [47, 48, 51, 53, 54] and one used the European Organization for Research and Treatment of Cancer QoL Questionnaire 30 (EORTC-QLQ-30) [52]. Six used cancer-specific measures including the FACT-Prostate [47‐49, 55], FACT-Breast [49, 51] and the EORTC Lung Cancer supplement [52]. Six studies assessed symptom specific QoL including fatigue [47, 48, 51, 55, 59], anxiety and mood [53] and side effect burden [59]. Six studies reported significant intervention group improvements [48, 55‐59] and six reported no significant group difference in one or more QoL measure [48, 49, 51, 52, 54, 60]. Two studies reported significant negative intervention difference in QoL [47, 50]. There was no difference in effectiveness between studies that reported using theory/psychosocial components to guide the interventions versus those reporting no theory. Three studies using theory reported significant group differences in QoL measures [47, 50, 56]; however, two of these were favouring the non-intervention group [47, 50]. Both theory-based and non-theory-based studies had within group improvements in both quality of life and physical outcomes. Select QoL results are described below; full details can be found in Tables 3 and 4.

Participants in a home- or group-based activity programme reported poorer general and cancer-specific QoL than those in a 1:1 personal training group [47]. Participants that received a home-based personalized activity and nutrition intervention reported similar improvements in general QoL to an attention control group receiving general information at study end which was maintained at follow-up [49]. A wait-list control group showed more positive improvement in scores for pain, physical role and overall mental health compared to those receiving a gardening intervention [50]. Both the home-based and group session-based participants in an exercise intervention for women diagnosed with breast cancer improved their overall and breast cancer-specific QoL [51]. No differences were observed between groups on global or lung cancer-specific QoL in those participating in a prehabilitation intervention versus usual care [52]. In a group of mixed cancer patients undergoing chemotherapy, between group differences favouring the intervention were noted among those having poorer social and emotional well-being at baseline measures compared to the wait-list control [53]. No between-group or within-group differences were found for QoL in speed feedback therapy group versus usual care [54]. Those receiving a supervised activity programme had significant between-group improvements in overall and prostate cancer-specific QoL compared to those receiving usual care [55]. While both groups in a home-based activity and nutrition intervention versus wait-list control had declines in overall QoL throughout the study period, the intervention group had significantly smaller declines than the control group [56]. Physical and mental composite scores of the SF-36 returned to preoperative levels in participants in a functional exercise intervention when compared to those in a usual care group [57]. In an exercise study among people with bladder cancer, only the role-physical domain scores improved significantly in the intervention group compared to usual care; all other scores had no differences [58]. No differences were found between exercise and control groups in men with prostate cancer, but among spouses also participating, there was a non-significant increase in partners’ mental health scores [60].

Men starting or currently on androgen deprivation therapy (ADT) reported a similar change in fatigue among three groups receiving an activity programme [47]. Participants that received a combination of supervised and home-based activity reported more improvement in fatigue than control groups after the study period which was maintained at follow-up [48]. An activity and nutrition intervention found improvement in depression scores in both the tailored versus non-tailored groups [49].

In a study comparing the same programme either home or group based, there were non-significant improvements in fatigue in both groups [51]. Participants receiving a low- to moderate-intensity home-based activity programme, with poorer anxiety and mood at baseline, had significant improvements compared to the control group [53]. Men with prostate cancer participating in aerobic exercise before radiotherapy reported significantly better fatigue scores than those in usual care [55]. Those participating in a yoga intervention reported significantly lower cancer-related fatigue and global side effect burden than the wait-list control group [59].

Strengths include a broad search method and the use of independent researchers. However, as with any review, important papers may have been missed. As with much of cancer research, most patient groups in this review were either breast or prostate cancer reducing the generalizability of the findings. Over half of the articles included were identified as feasibility studies. Though the majority of studies indicated the interventions were feasible, they were underpowered for effectiveness. Finally, few studies indicated using any theoretical base or specific BCTs. Future research should incorporate appropriate techniques to assist self-management and help encourage higher completion rates for older adults, especially those with frailty and sarcopenia, learning from research in other conditions as relevant [72].

This work identified key gaps in the evidence supporting rehabilitation-based programmes for older adults with cancer, and a paucity of work including nutrition interventions alongside those aiming to improve PA. Development of acceptable and relevant interventions, flexible across the cancer continuum and cancer type and stage are needed. One size is unlikely to fit all. Future research should be underpinned by behaviour change theory and include studies to explore how best to support attendance and completion by those with frailty and sarcopenia. There is likely to be overlap with research in other areas of older adults’ health and rehabilitation but although there is interest in generic rehabilitation programmes, there is little evidence to date to confirm benefits in people with cancer, or in older adults [86, 87].