Perioperative exercise programmes to promote physical activity in the medium to long term: systematic review and qualitative research

Primary outcomes

• Physical activity: amount of PA/exercise conducted at the end of follow-up (such as mean number of steps measured using a step counter).

• Physical activity: number of people who were engaging in PA at the end of follow-up (e.g. as measured in a self-reported questionnaire).

Secondary outcomes

• Physical fitness: mean scores [and standard deviation (SD)] measured using standardised tools at the end of follow-up [cardiopulmonary exercise testing as first choice, but also included data for peak heart rate (beats/minute), peak oxygen consumption (ml/kg/minute), peak power output (W), 6-minute walk (m), bench press (kg), leg press (kg), handgrip (kg), sit and reach (cm), anaerobic or lactic threshold, peak oxygen uptake (VO₂ peak) and minute ventilation to carbon dioxide output (V_E/VCO₂)]. 7

• Quality of life at the end of follow-up: mean scores using a validated tool [e.g. Short Form questionnaire-36 items (SF-36), EuroQol-5 Dimensions (EQ-5D), the Centers for Disease Control and Prevention Health-related Quality of Life measure or condition-specific tools].

• Cancellation of surgery because of improved health: number of participants who no longer needed surgery.

• Participants’ experiences of participation: could include narrative summaries, measures on a continuous scale or number of people who were satisfied with the experience.

• Adherence: may include attendance at PA classes or completion of motivational telephone calls.

• Pain at the end of follow-up: mean scores using a validated tool [e.g. visual analogue scale (VAS)].

• Adverse events: as described by study authors (both related and unrelated to the intervention).

Post operative

Interventions overwhelmingly began post operatively (79.1%).

A number of studies55,59,62,63,66,78,83,90,100 were vague on start point, and, although we suspect that the majority began within a few weeks of surgery, as it was not clear, we have grouped them separately. Some studies indicated commencement a short time after surgery, for example ‘on entering’90 or ‘during’100 cardiac rehabilitation. A number of cancer studies described the intervention beginning before or during adjuvant therapies. 59,63,83 Two studies suggested that participants began interventions over a broader period of time (a mean of 6.3 months since colorectal cancer diagnosis78 and within 9 months of diagnosis66).

Of the other studies, roughly half of the interventions began in the immediate days or early weeks after surgery: 10 interventions began within a few days of, or during the first week after, surgery,72,77,81,82,111 including Kinsey et al. ,88 which involved two interventions, and Losina et al. ,96 which involved three interventions; and nine interventions began in the 2–4 weeks after surgery,58,85,93,109 including Boesch et al. ,56 which involved three interventions, and Johansson et al. ,86 which involved two interventions.

The remaining 23 interventions were delayed post surgery: in 17 studies, there was a delay of > 1 month to 3 months,52,60,68,69,75,76,79,84,101,104,108 including Archer et al. ,51 Jolly et al. 87 and Smith et al. ,105 which each involved two interventions; in six studies, the intervention began > 3 months after surgery,61,92,106,107 including Piva et al. ,102 which involved two interventions.

Perioperatively

Ten interventions (14.9%) were delivered perioperatively: five began between 2 and 4 weeks prior to surgery,70,91,110 including the two intervention arms of Creel et al. ;64 two began 6–8 weeks prior to surgery;94,97 one began 6 months prior to surgery;53 and the intervention arms of Goedendorp et al. ,74 which did not clearly state their initiation points (these were cancer interventions described as beginning at the time, or within days, of diagnosis).

Pre operative

Four interventions (6.0%) were offered pre operatively only: one began up to 18 weeks before surgery;57 and three began between 4 and 9 weeks prior to admission. 54,95,103

Education and advice

Fifty-five (82.1%) interventions described the encouragement of PA adoption through ‘education sessions’ – the provision of information, advice, recommendations, or a formal exercise prescription51–55,57–62,66,68–70,72,75,76,78,82–85,91–94,97,100–103,107–110 – including Boesch et al. ,56 which involved three interventions in this category, and Creel et al. ,64 Goedendorp et al. ,74 Johansson et al. ,86 Jolly et al. ,87 Kinsey et al. ,88 Kraal et al. ,90 Losina et al. 96 and Smith et al. ,105 which all involved two interventions in this category.

The content of these interactions tended to fall into one or more of the following groups:

• Exercise recommendations, in which exercise prescriptions, advice for home exercise, recommendations or goals were described,52,54,55,57–59,62,66,68,69,72,75,78,83–85,87,92–94,97,100–102,108–110 including Creel et al. ,64 Goedendorp et al. ,74 Johansson et al. ,86 Kinsey et al. ,88 Kraal et al. ,90 Losina et al. 96 and Smith et al. ,105 which all involved two interventions in this group. These were sometimes provided in addition to, or following, a period of supervised PA. For example, in Artz et al. ,52 following 6 weeks of supervised classes, ‘participants were provided with a list of exercises, including their individual exercises, to continue with at home on a regular basis’. In others, recommendations might have been given at the start of an intervention: ‘the nurse explained how to break the negative spiral of low physical activity . . . consequently, patients were advised to increase their physical activity level stepwise’. 74 Alternatively, they might have been provided alongside sessions that supported the setting of PA goals. 57,62,69,75,96,109 Recommendations were, however, largely self-managed, although some interventions included a degree of practitioner contact, prompting adherence. For example, one intervention84 provided participants with booster sessions every second month, in which a physiotherapist would review the patient’s experience of the home-based exercise programme, provide instruction for new activities and define a new daily steps target; the option of telephone support from a physiotherapist was also on offer. The use of an exercise diary or activity log,54,59,64,92,94,96,108,109 or an activity device (such as a pedometer or wearable heart rate monitor)55,57,59,62,64,68,69,74,75,78,84,90,96 to monitor progress was also described. Barnason et al. 55 used a telehealth device that provided participants with daily strategies around rest, pain management and progressing PA. Recommendations were often personalised either in response to a baseline assessment (usually fitness) or would be reviewed and set again in response to progress over time. Those that were not personalised were sometimes noted as being based on national clinical guidelines.

• Education or information materials, which comprised 20 interventions,52,59,61,66,68,69,75,76,82,84,91,103,108,109 including both the counselling and pedometer arms of Creel et al. ,64 the nurse and cognitive–behavioural therapy (CBT) arms of Goedendorp et al. ,74 the home-based intervention of Johansson et al. 86 and the home-based intervention of Jolly et al. 87 For these interventions, participants were provided with, for example, ‘a one-page information sheet . . . that also promoted the progressive attainment of 10,000 steps/day’,64 an exercise workbook,69 ‘pictorial and written instructions for [home] exercise’,84 or online role modelling or instruction videos for supporting home exercise adoption. 75,103 We also included here studies in which participants were provided with information about,108 or helped to identify,61 exercise opportunities in the local community.

• Lifestyle education, which comprised 12 interventions,53,60,66,70,82,91 including all three intervention arms of Boesch et al. ;56 the home-based intervention of Jolly et al. ;87 and the financial incentive (FI) and financial incentive plus telephonic health coaching (FI + THC) intervention arms of Losina et al. ,96 in which the focus of contact with practitioners, or of resources, was placed on broader improvements to health, with PA just one strand of an emphasis on lifestyle factors that might also include diet, alcohol or smoking cessation. One study,53 for example, described a voluntary Motivator’s Club that explored PA, nutrition and psychological issues associated with weight management. Several interventions in cardiac and cancer studies framed sessions around risk factors for disease,70,82,87 whereas others touched on PA (goals, benefits, barriers, activity progression or diary-keeping) among a wider programme of sessions covering topics such as nutrition,53,60,70 use of medication,82,87 ‘pain management strategies’84,97,108 or mindfulness. 94 One study59 reported asking participants in both groups not to make significant changes in their dietary habits.

• The value of PA, which comprised eight interventions,54,68 including the education arm of Archer et al. ,51 the nurse and CBT intervention arms of Goedendorp et al. ,74 the home-based arm of Johansson et al. ,86 and both the centre-based and home-based intervention arms of Jolly et al. 87 These studies described interventions highlighting the ‘importance’,53 the ‘benefits’68 or the ‘promotion’54 of exercise, or, as in the CBT intervention in Goedendorp et al. ,74 explained to participants the negative spiral of low PA and fatigue. This focus was often part of direct sessions in person, over the telephone, or, in one case, described as part of computer-tailored PA advice that patients could access from home. 75

Behavioural mechanisms

Therapeutic or behavioural approaches to supporting participants to engage in PA was adopted in over half (59.7%) of all interventions. The majority of these were delivered to the individual,54,55,57–59,61,62,68,69,75,76,78,81,83–85,91,92,95,97,108,109,111 including the CBT arm of Archer et al. ,51 the counselling arm of Creel et al. ,64 the CBT arm of Goedendorp et al. ,74 both the walking and cycling arms of Kinsey et al. ,88 the home-based arm of Kraal et al. ,90 and both the THC and the FI + THC arms of Losina et al. 96

A small number of these approaches was incorporated into group sessions of 6–10 people,94 and between 15 and 25 participants, depending on the site of delivery. 82 Baillot et al. 53 and Sellberg et al. 104 did not report group size. For Engblom et al. 70 and Boesch et al. ,56 which included three intervention arms (all residential models), and the centre-based arm of Jolly et al. ,87 it was not clear whether their residential models were delivered as groups or individually.

Fifteen of these approaches59,61,62,68,69,78,83,84,91,108,109 were delivered by telephone, including the CBT arm of Archer et al. ,51 the home-based arm of Kraal et al. ,90 and the THC and FI + THC arms of Losina et al. ;96 22 approaches53,54,57,58,70,81,82,85,92,94,95,97,104,111 (including all three intervention arms of Boesch et al. ,56 the counselling arm of Creel et al. ,64 the CBT arm of Goedendorp et al. ,74 the centre-based arm of Jolly et al. ,87 and both the walking and the cycling intervention arms of Kinsey et al. 88) were delivered face to face in a clinical, community or inpatient setting. Two interventions were delivered via smartphone55 or computer application,75 and another, which included motivational messaging, was delivered by participants’ method of choice (text message, telephone call, e-mail or post). 76 Therapeutic sessions were sometimes noted as being a ‘one-off’, but were usually reported as taking place over the course of several weeks or months.

Although the detail of sessions was generally limited, we identified several recurring topics or approaches in the content described:

• Most53,54,58,61,62,68–70,75,76,78,81–85,91,92,94,95,97,104,108,109,111 (including the CBT arm of Archer et al. ,51 the counselling arm of Creel et al. ,64 the CBT arm of Goedendorp et al. ,74 the centre-based arm of Jolly et al. ,87 both the walking and the cycling arms of Kinsey et al. ,88 the home-based intervention arm of Kraal et al. ,90 and both the THC and FI + THC arms of Losina et al. 96) talked about behaviour change approaches, counselling or motivational interviewing, describing supporting participants, for example with the ‘restructuring of cognitions and beliefs’ (the CBT arm of Goedendorp et al. 74), or being ‘counselled to establish a home-based exercise programme’ (Lear et al. 92). Some of these interventions emphasised the use of goal-setting,57 or of focusing on participants’ experiences of their exercise programme to consider adaptions and new goals. 84 Details of the THC arm in Losina et al. 96 described coaches using ‘open-ended questions to elicit the participants’ own objectives’ and ‘while expressing empathy . . . helping to resolve any discrepancy between subjects’ PA goals and current behaviour’. Some studies described motivational interviewing techniques, sometimes adopted by PA coaches, and focused on barriers, the discussion of ‘obstacles, fears and solutions’ to PA (the CBT arm of Goendendorp et al. 74) or efforts to ‘allay harmful irrational beliefs and fears regarding activity’. 84

• Improved self-efficacy and promoting self-management were also common foci in 12 interventions55,57,59,61,69,75,78,97 (including the CBT arm of Archer et al. ,51 the counselling arm of Creel et al. ,64 and both the THC and FI + THC arms of Losina et al. 96). This included self-management approaches ‘to reduce pain and disability, and improve PA’ (the CBT arm of Archer et al. 51) or to self-care and management of early recovery symptoms,55 but, more often, studies described a focus on encouraging participants to self-monitor PA,57,61,78 or on participants’ self-efficacy for exercise,59,69,75,108 which included both the THC and FI + THC arms of Losina et al. 96 Other studies used mechanisms such as exploring ‘self-rewards’. 62

• Reference to the following was also common – lifestyle and health coaching68,81,92,94,104 (including the CBT intervention arm of Archer et al. ,51 all three intervention arms of Boesch et al. ,56 the CBT arm of Goedendorp et al. ,74 and both the walking and the cycling intervention arms of Kinsey et al. 88); active lifestyles, such as work-related, leisure-related and daily life activities;81 daily walking;56 or broader counselling approaches, which might include working on addressing fatigue, sleep or ‘risk factor counselling’. 92

Direct physical activity instruction

Thirty interventions (44.8%) provided face-to-face PA instruction,52–54,60,70,72,77,79,82,84,85,92,95,106,107,109,111 including all three intervention arms of Boesch et al. ,56 the aerobic and the resistance arms of Courneya et al. ,63 the centre-based arm of Jolly et al. ,87 the clinic-based arm of Johannson et al. ,86 both intervention arms of Kraal et al. ,90 both the clinic and the community arms of Piva et al. ,102 and the hospital and the home training arms of Smith et al. 105

These were more often one component of a broader approach52–54,60,70,72,82,84,85,92,95,107,109,111 (including all three intervention arms of Boesch et al. ,56 the clinic-based arm of Johansson et al. ,86 the centre-based arm of Jolly et al. ,87 both intervention arms of Kraal et al. ,90 the clinic-based arm of Piva et al. 102 and both arms of Smith et al. 105) that might also include education and advice52–54,60,70,72,82,84,85,92,107,109 (including all three intervention arms of Boesch et al. ,56 the clinic-based arm of Johansson et al. ,86 the centre-based arm of Jolly et al. ,87 both arms of Kraal et al. ,90 the clinic-based arm of Piva et al. 102 and both arms of Smith et al. 105) or therapeutic approaches53,54,70,82,84,85,92,95,109,111 (including all three intervention arms of Boesch et al. ,56 the centre-based arm of Jolly et al. 87 and the home-based arm of Kraal et al. 90). Several studies,52,53,60,77,79,106 (including the centre-based intervention arm of Kraal et al. ,90 the community arm of Piva et al. 102 and the hospital-based arm of Smith et al. 105) indicated that activity sessions took place in groups, although group size was infrequently described. When it was, group size ranged from two people79 to 25 people. 82

The majority of PA sessions52–54,60,72,77,79,82,85,95,106,107,111 (including all three intervention arms of Boesch et al. ,56 both arms of Courneya et al. ,63 the clinic-based arm of Johansson et al. ,86 the centre-based arm of Jolly et al. ,87 both intervention arms of Kraal et al. ,90 the community-based arm of Piva et al. 102 and the hospital training arm of Smith et al. 105) were provided at least weekly, often two or three times per week; all intervention arms of the Boesch et al. 56 study provided five sessions of cycling each week. Participants in this study56 also participated in walking twice daily; it was not clear if these sessions were supervised. In some interventions, the regularity of PA instruction varied over time, for example increasing progressively over 6 months;85 reducing in frequency, as in the clinic-based arm of Piva et al. ;102 or varying as in Baillot et al. ,53 in which participants received instruction (three sessions per week for almost 6 months) in the pre-surgery phase of the intervention only, with post-surgery intervention components focused on counselling only. In Lear et al. ,92 participants received up to eight cardiac rehabilitation exercise sessions over 12 months, and in the home training arm of Smith et al. ,105 participants were offered 2-hour exercise consultations with an exercise specialist at the beginning and after 3 months of home exercise.

Physical activity instruction usually took place in a clinical setting (including a hospital or physiotherapy clinic or gym)52–54,60,72,77,79,82,84,92,95,111 (including the clinic-based intervention arm of Johansson et al. ,86 the centre-based arm of Jolly et al. ,87 both the centre-based and the home-based arms of Kraal et al. ,90 the clinic-based arm of Piva et al. ,102 and both the hospital-based and the home-based arms of Smith et al. 105) but also in community settings (including a leisure centre, community centre or public gym)82,85,106 (including both the aerobic and resistance arms of Courneya et al. 63 and the community arm of Piva et al. 102), within a residential inpatient facility or during a hospital stay70,72 (including all three intervention arms of Boesch et al. 56) or as part of a home visit. 107,109

Six interventions77,79,106 (including both the aerobic and resistance arms of Courneya et al. 63 and the community-based arm of Piva et al. 102) described PA instruction only. In these studies, supervised training or classes took place two or three times per week, in a community or clinical setting, usually for ≥ 3 months.

Other components

Several studies noted additional features of intervention design. Artz et al. ,52 for example, reimbursed participants’ travel and parking costs, or supported access to transport. Losina et al. ,96 in two of their three intervention arms, used FIs, both in relation to the completion of PA logs and again if participants managed to increase their daily step count or minutes of moderate-to-vigorous physical activity (MVPA). Losina et al. ,96 for their THC arm, also reported providing coaching, always at a time most convenient to participants, and coaches were prepared to make several attempts at telephone contact; and Turunen et al. 109 described the use of volunteer students who would support frail participants (who could not manage this alone) to take walks outdoors to meet their activity goals. Additional design features in these studies could be seen as mechanisms to support engagement.

Nutritional counselling54 and a plant-based diet or lycopene supplements76 were noted in some studies; Boesch et al. 56 highlighted the provision of low-fat meals and the rural, high-altitude location of the inpatient facility in which participants spent 4 weeks, shortly after surgery.

Digital technologies

Digital and other technologies were integral components of protocols in just under one-third of interventions (29.9%). Most commonly this was through the use of a pedometer or other step activity monitor,57,62,68,69,75,78,84,90,100,110 including in Brandes et al. ,58 in which participants were blinded to the device, but the data were used by counsellors to encourage increases; in both the counselling and the pedometer arms of Creel et al. ;64 and in both the FI and THC arms of Losina et al. 96 However, also included were the use of heart rate monitors;59,60,90,110 an interactive online application for computer or smartphone (producing tailored advice);75 audio tapes (providing an abridged version of the intervention ‘Heart manual’, spoken in Punjabi for participants with a limited command of English) or online exercise videos, as featured in Santa Mina et al. 103 and the home-based arm of Jolly et al. ;87 and a telehealth device (an interactive device, requiring a telephone line but no internet access, providing individualised intervention sessions). 55

Pedometers in particular, and heart rate monitors, were often referred to as tools to support participants to engage in, or monitor, their activity. Although 22 studies51,53,55,60–62,64,74,75,81,82,85,90,93,96,97,102,104,106,107,109,111 also used movement accelerometers such as ActiGraph GT3X (ActiGraph, LLC, Pensacola, FL, USA) to measure study outcomes (e.g. bouts of MVPA or daily steps), these were generally not available for patients to use during the intervention period; instead, they were provided at specific time points, to be returned either side of measurement periods (usually 1 week). With the exception of five studies55,64,93,96,109 involving eight interventions, two64 of which had the screens on their devices obscured, all such measurement periods fell outside the intervention delivery period (usually prior to intervention commencement, at the immediate end of the intervention or at some point post intervention).

Other resources/equipment

In addition to digital technologies, such as pedometers, used as part of the intervention provision, 32 interventions provided other specific resources or equipment. This included printed materials, such as programme information, educational booklets, exercise workbooks (some of which were personalised) and motivational postcards, in 17 of the studies. 51,52,59,61,64,66,69,74,76,78,82,84,87,108,109

In 13 studies, interventions included the provision of, or the encouragement for participants to keep, exercise diaries, logs or similar; the purpose of these diaries/logs was sometimes described as a method to help participants keep track of progress and to provide additional motivation. 54,57,59–61,64,66,84,92,94,101,105

Participants in both the intervention and control groups in Mundle et al. 100 were asked to self-report daily estimates of their PA via a questionnaire.

In two studies, interventions included open access to fitness centres throughout the intervention period106 or, in both study arms of Courneya et al. ,63 for up to 1 month post intervention. Fitness equipment was provided to participants in five studies: resistance bands,83,84,108,109 and a stability ball and yoga mat, which participants in the study by Santa Mina et al. 103 were able to keep post intervention.

Figure 5 shows the number of interventions per resource used.

FIGURE 5.

Resources to support interventions.

Type of exercise/physical activity encouraged

Studies were frequently vague on the sorts of physical activities or exercises adopted or recommended. However, walking was the most frequently reported activity54,57,58,60–62,68,69,72,76,78,81,83,84,107,109–111 (including in the CBT intervention arm of Archer et al. ,51 all three arms of Boesch et al. ,56 the aerobic arm of Courneya et al. ,63 the counselling and the pedometer arms of Creel et al. ,64 both the nurse and the CBT arms of Goedendorp et al. ,74 the clinic-based and the home-based arms of Johansson et al. ,86 the centre-based and the home-based arms of Jolly et al. ,87 the walking and the cycling arms of Kinsey et al. ,88 the FI and the FI + THC arms of Losina et al. ,96 the clinic-based arm of Piva et al. ,102 and both the hospital-based and the home-based arms of Smith et al. 105), including outdoor walking or trekking; aerobic, brisk walking; running (one study81 only); short or functional walks; or the achievement of daily steps. Cycling, usually on a stationary bike, although also encouraged as a home-outdoor activity, was also described, and sometimes suggested as an alternative to walking54,60,70,81 (including in all three intervention arms of Boesch et al. ,56 the aerobic arm of Courneya et al. ,63 the nurse and the CBT arms of Goedendorp et al. ,74 the centre-based arm of Jolly et al. ,87 the cycling arm of Kinsey et al. 88 and the clinic-based arm of Piva et al. 102). For other interventions52–54,59,69,77,83,85,92,103 (including in the clinic-based arm of Piva et al. ,102 and both the hospital-based and the home-based arms of Smith et al. 105), authors simply reported aerobic training (including endurance activities, moderate-to-vigorous and high-intensity activities, and activities ‘designed to increase general fitness’52). Several studies described other activities such as swimming or pool activities,70,85,111 gymnastics,70 dance,85,111 rowing (in the centre-based intervention)87 and ‘traditional Spanish games’. 85

Resistance or strength training was often included52,61,69,77,83–85,103,106–109,111 (including in the resistance intervention arm of Courneya et al. ,63 the clinic-based and the home-based arms of Johansson et al. ,86 and both the clinic-based and the community-based arms of Piva et al. 102), which, with the exception of the resistance training arm of Courneya et al. ,63 was delivered or encouraged alongside one or more other activities. Ten interventions61,77,92,108,111 (including in both the clinic-based and home-based intervention arms of Johansson et al. ,86 the community-based arm of Piva et al. ,102 and both the hospital-based and the home-based arms of Smith et al. 105) were described as including stretch, flexibility and balance exercises.

Those studies that were less specific55,66,68,72,75,79,82,91,93–95,97,100,101,104,106 (including in both the CBT and the education intervention arms of Archer et al. ,51 the counselling arm of Creel et al. ,64 the home-based arm of Jolly et al. ,87 both the centre-based and the home-based arms of Kraal et al. ,90 and the THC arm of Losina et al. 96) tended to include reference to interventions supporting ‘activities of choice’, as in the home-based intervention for Jolly et al. ;87 ‘progressing physical activity’;55 participants being encouraged to engage in, for example, 30 minutes of daily exercise;94 or an exercise prescription (these, as in Painter et al. ,101 might include goal intensity markers, but there were no details on the activity or activities prescribed).

How much physical activity and at what sort of intensity?

Studies were also not always clear on the frequency or intensity of PA that was being encouraged or delivered. Sometimes this was explicitly undefined: when participants were encouraged to focus on simply being more active95 or on the ‘quality of movement rather than quantity’. 52 Other studies focused on incremental progression, such as a 5% increase in daily steps58 or increasing ‘speed and duration . . . as tolerated up to a maximum 30 minutes’. 72 Some studies focused on perceived exertion, describing exercise classes ‘designed to be challenging for active older adults and safe for more frail individuals’, as in the community-based intervention for Piva et al. ;102 or the use of the Borg Scale to estimate participants’ training intensity, gradually increasing, for example from a score of 15 in the initial phase of the intervention to a Borg Scale score of 17. 106 Boesch et al. ,56 for both the self-regulation and the objective/subjective intervention arms, focused on a lower score of 12–14. A number of studies described activities of moderate to vigorous intensity, often also inclusive of bout-related goals, such as walking a ‘brisk pace for 30 minutes on at least five days a week’,76 or a goal of meeting a PA guideline of at least 150 minutes of MVPA each week, as in all three intervention arms in Losina et al. 96

A number of interventions described being guided by exercise testing, such as working (cycle ergometer, treadmill or elliptical trainer) at 60% of VO₂ peak for 15 minutes and progressing to 80% of VO₂ peak for 45 minutes, as found in the aerobic training arm of Courneya et al. 63 Similarly, the hospital training intervention arm in Smith et al. 105 used an exercise prescription based on a target intensity of 60% of VO₂ peak for 3 months, revised, when appropriate, to 70% of VO₂ peak for a further 3 months. Participants were advised to follow this regime using a cycle ergometer, arm cycle ergometer, treadmill and track-walking for 40 minutes, 5 days per week. Sessions would also include additional time doing warm-up and stretching exercises. Kinsey et al. ,88 in their walking programme, used metabolic equivalent of task (MET), encouraging short walks at a level of 3.3 METs three to five times weekly or, in their cycling programme, progressive cycling goals (also three to five times weekly) from 1.5 METs at the start of the intervention to 5.3 METs over 12 weeks. Yates et al. 111 encouraged a PA minimum goal of 150 minutes per week (but preferably activity all days of the week) at 3 METs, whereas Painter et al. 101 used an exercise prescription based on an intensity goal of 60–65% of maximal heart rate, increasing gradually to 75–80% of maximal heart rate, four times per week, over a 10-month period.

The total amount of PA either delivered or encouraged was sometimes hard to gauge from the limited details provided. However, we considered these as falling into one of three groups:

1. low – those encouraging PA once or twice per week (17.9%)52,61,62,79,82,106,107 (including in the centre-based intervention arm of Jolly et al. ,87 both the centre-based and the home-based arms of Kraal et al. ,90 and the clinic-based and the community-based arms of Piva et al. 102)

2. regular – those largely in line with typical PA recommendations of three (or four, as with Santa Mina et al. 103) times per week (11.9%)53,77,84,85,103,108 (including in the aerobic and the resistance intervention arms of Courneya et al. 63)

3. frequent – those encouraging PA on ≥ 5 days per week (46.3%)54,55,57,59,60,66,70,72,76,78,83,93,94,109–111 (including in all three intervention arms of Boesch et al. ,56 the pedometer arm of Creel et al. ,64 the CBT arm of Goedendorp et al. ,74 the clinic-based and the home-based arms of Johansson et al. ,86 the home-based arm of Jolly et al. ,87 both the walking and the cycling arms of Kinsey et al. ,88 all three intervention arms of Losina et al. ,96 and both the hospital-based and the home-based arms of Smith et al. 105).

Just under one-quarter of interventions (23.9%) did not provide this information58,68,69,75,81,91,92,95,97,100,101,104 (including in both the CBT and the education arms of Archer et al. ,51 the counselling arm of Creel et al. 64 and the nurse intervention arm of Goedendorp et al. 74).

Tailoring activity

Forty-one (61.2%) interventions took a personalised or tailored approach to PA. This was usually described in respect to participants’ capacities or fitness (such as tailored goals or activity intensity)54,57,58,60–62,66,68,69,72,75,77,79,81,82,84,92,95,101,103,107–109,111 (including in the CBT intervention arm of Archer et al. ,51 all three intervention arms of Boesch et al. ,56 the counselling arm of Creel et al. ,64 the CBT arm of Goedendorp et al. ,74 both the walking and the cycling arms of Kinsey et al. ,88 the centre-based and the home-based arms of Kraal et al. ,90 the clinic-based arm of Piva et al. ,102 and the hospital-based and the home-based arms of Smith et al. 105) and/or participant preferences (such as activity of choice)52,59,61,66,69,75,79,81,82,84,92,95,101,109,111 (including in the CBT intervention arm of Archer et al. ,51 the counselling arm of Creel et al. ,64 both the walking and the cycling arms of Kinsey et al. ,88 the centre-based and home-based arms of Kraal et al. ,90 all three intervention arms of Losina et al. ,96 the clinic-based arm of Piva et al. 102 and the home-based arm of Smith et al. 105). This included, for example, participants being asked to identify activities that they would like to return to following their surgery;52 activities that they wished to become better at;79 or PA plans developed around individuals’ interests, found in both intervention arms of Kinsey et al. 88 Other studies described interventions tailoring the number of sessions, intensity and progress of training (Barberan Garcia et al. 54); ‘the frequency and duration’ of rehabilitation components being ‘individualised to patient characteristics’;58 and ‘exercise parameters’ and ‘rate of progression’ that accounted for patients functional capacity, and treatment side effects (as well as their exercise preferences or previous exercise history). 69

Home

Thirty-nine interventions took place in participants’ homes (or neighbourhood environment): 21 exclusively55,59,66,69,75,76,78,83,93,101,103,107–110 (including in both the CBT and the education arms of Archer et al. ,51 the home-based arm of Jolly et al. 87 and all three intervention arms of Losina et al. 96), whereas 19 included home-based components57,60–62,68,72,81,84,85,91,92,111 (including in both the clinic-based and the home-based intervention arms of Johansson et al. ,86 the walking and the cycling arms of Kinsey et al. ,88 the home-based arm of Kraal et al. ,90 the clinic-based arm of Piva et al. 102 and the home-based arm of Smith et al. 105). Of those taking place exclusively at home, the majority included some contact (usually weekly, but it was variable) with their clinical or intervention team via telephone, post or the participant’s method of choice, be that text message, telephone, post or e-mail, as in Hackshaw-McGeagh et al. 76 A small number had no ‘human’ contact, relying instead on support/guidance delivered through the use of a telehealth device,55 computer-generated messaging75 or a smartphone application. 110

Three of the exclusively home-based interventions included, in addition to telephone support, home visits from physiotherapists (delivering motivational interviewing with musculoskeletal patients108,109) or a nurse (supporting exercise goals among cardiac patients87). The addition of volunteer health sciences students to support frail participants, who were unable to go outdoors alone, to take outdoor walks was also offered to participants in one of these studies. 109

Clinical settings

The use of clinical settings, including hospitals and hospital gyms, outpatient and physiotherapy clinics, and health centres, was also common (49.3%), but was more usually accompanied by expectations for home exercise or further telephone support provided to participants at home52–54,57,60,62,68,72,77,79,81,82,84,91,92,95,97,104,111 (including in both the counselling and the pedometer arms of Creel et al. ,64 the nurse and the CBT arms of Goedendorp et al. ,74 the clinic-based and the home-based arms of Johansson et al. ,86 the centre-based arm of Jolly et al. ,87 the walking and the cycling arms of Kinsey et al. ,88 the centre-based and the home-based arms of Kraal et al. ,90 the clinic-based arm of Piva et al. ,102 and both the hospital-based and the home-based arms of Smith et al. 105).

Community

Community spaces (public or private leisure centres or gyms; local activity groups, such as walking groups; or community centres) were described less frequently (10.4%)61,82,85,106 (including in both the aerobic and the resistance intervention arms of Courneya et al. 63 and the community-based arm of Piva et al. 102). Some of those delivered in community settings were delivered as group activities, whereas some also provided access to leisure facilities outside (and, for a short period, beyond) the intervention delivery. One intervention61 that did not directly deliver PA described taking a therapeutic and goal-setting approach to behaviour change, and supported participants to identify walking opportunities within their local community.

Inpatient care

Six (9.0%) interventions had components set in inpatient care: one72 in a hospital setting and five in a residential rehabilitation centre,58,70 including in all three intervention arms of Boesch et al. 56 With the exception of Brandes et al. 58 (hip or knee replacement patients), these were all interventions aimed at cardiac patients. Stays lasted 2 weeks,72 3 weeks58 and 1 month,56 with one intervention involving three phases of admission that included a 2-day stay pre surgery, a 3-week stay 6–8 weeks post surgery and a further 2-day stay at 8 months post surgery. 70 The intervention in Foster et al. ,72 the only study to describe an intervention as part of their hospital stay, began within 48 hours of surgery, but continued in an outpatient clinic and as home exercise on discharge, after approximately 5 days.

Two further studies94,100 included no details on intervention setting. Figure 6 shows the number of interventions per setting.

FIGURE 6.

Settings used to deliver interventions. Two interventions are not included in this figure because the delivery setting is not described in the study reports. a, Participants also used other settings to complete PA.

Remote delivery

Twenty-six interventions55,59,61,62,66,68,69,75,76,78,83,84,91,93,101,103,108–110 (including in the CBT and the education arms of Archer et al. ,51 the home-based arm of Jolly et al. ,87 the home-based arm of Kraal et al. 90 and all three intervention arms of Losina et al. 96) included some form of remote delivery. In 16 interventions55,59,66,69,75,76,83,93,101,103,110 (including in the CBT and education arms of Archer et al. 51 and all three intervention arms of Losina et al. 96) it was exclusively remote, reliant on contact with practitioners or study staff via telephone or post and/or supported by technology. Several studies,61,62,68,84,91 including in the home-based intervention arm of Kraal et al. ,90 took a blended approach, combining in-person/clinic-based support with remote (usually telephone) contact; three of these108,109 (including in the home-based arm of Jolly et al. 87) also included home visits in addition to telephone support.

Of those taking place exclusively at home, three interventions55,75,110 relied on no practitioner (or study) contact throughout the intervention period at all. Barnason et al. 55 was, for example, designed around a telehealth system that provided participants with daily strategies around rest, pain management and incremental progression of PA over the course of 6 weeks, but no direct contact with practitioners or the study team throughout that period. Similarly, Van der Walt et al. ,110 also a 6-week intervention, relied on predefined daily steps goals supported by the use of a Garmin Vívofit^® (Olathe, KS, USA) activity tracker, but again included no interaction with practitioners throughout. Golsteijn et al. 75 was an application-driven intervention (although apparently a more enhanced package) providing computer-generated PA advice to participants at three time points (monthly intervals for 3 months) across the intervention period. The advice was generated automatically using a message library, questionnaire data and computer-based data-driven decision rules. The advice was also made available on paper and mailed to participants. This advice was provided alongside interactive content on the intervention website including role-modelling videos, home exercise instruction videos and a module for goal-setting using a pedometer. There was, however, an option to consult a physical therapist for additional information if needed (no details were provided as to whether or not participants selected this option).

Twenty-one59,61,62,68,69,78,83,84,91,93,101,103,108,109 (including in the CBT and the education arms of Archer et al. ,51 the home-based arm of Jolly et al. ,87 the home-based arm of Kraal et al. 90 and all three intervention arms of Losina et al. 96) of these remote delivery models predominantly relied on frequent (e.g. weekly) or irregular or less frequent (e.g. monthly, or reducing or variable over time) telephone support or coaching. Others were reliant on a smartphone, device or computer application guiding the intervention55,75,110 (including the home-based training intervention arm of Kraal et al. 90), or the exchange of workbooks and tailored newsletters via post (containing activity/lifestyle goals);66 just workbooks via post was a further mechanism for engaging with participants. 101 In Hackshaw-McGeagh et al. ,76 participants could choose the method of contact (text message, telephone, post or e-mail) through which they received motivational messages to encourage continuation of the intervention.

When remote and face to face was blended, some interventions appeared to begin with face-to-face contact and then move to a less frequent telephone model. For example, Christiansen et al. 62 provided weekly face-to-face therapeutic support (as an add-on to usual care rehabilitation) for the first 6–8 weeks, after which participants received monthly telephone contact with a physical therapist. In Ilves et al. ,84 participants received clinic booster sessions every second month, but otherwise exercised independently and, only if they needed it, had contact via telephone with a physiotherapist. However, Kraal et al. 90 was based on short but regular telephone-delivered motivational interviewing (of 10–20 minutes per session) each week in addition to in-person supervised PA training.

Practitioners and those involved in delivery

Interventions tended to be delivered by a single practitioner (62.7%): 12 by a physiotherapist52,54,60,69,79,84,95,106–108 (including in both the clinic-based and the home-based intervention arms of Johansson et al. 86), nine by study staff83,101,103,104,110 (including in both the aerobic and the resistance intervention arms of Courneya et al. ,63 the pedometer arm of Creel et al. 64 and the centre-based arm of Jolly et al. 87), nine by a physical therapist61,62,77,97 (including in the CBT and the education arms of Archer et al. ,51 the centre-based and the home-based arms of Kraal et al. ,90 and the clinic-based arm of Piva et al. 102), four by a health coach78 (including in all three intervention arms of Losina et al. 96), three by a nurse72,76 (including the nurse intervention arm of Goedendorp et al. 74), two by a PA specialist85 (including in the community-based arm of Piva et al. 102), one by a psychologist,74 one by a psychiatrist57 and one by a student. 58 Others (23.9%)53,70,81,82,91,92,94,109,111 (including in all three intervention arms of Boesch et al. ,56 the counselling arm of Creel et al. ,64 the home-based arm of Jolly et al. ,87 and both the hospital-based and the home-based arms of Smith et al. 105) were delivered by a MDT or provided no details55,59,66,68,75,93,100 (13.4%) (including in both the walking and the cycling intervention arms of Kinsey et al. 88) (Figure 7).

FIGURE 7.

Practitioners involved in the delivery of interventions.

Multidisciplinary teams included staff from between two and six different specialties. Physiotherapists were involved with nine interventions81,82,94,109 (including in all three intervention arms of Boesch et al. ,56 and the home-based and the hospital-based arms of Smith et al. 105), PA specialists were involved with seven53,81,92,111 (including in the counselling arm of Creel et al. ,64 and both the home-based and the hospital-based arms of Smith et al. 105), physicians were involved with five70,111 (including all three intervention arms of Boesch et al. 56) and dietitians with four. 53,70,92,111 The following were also featured in interventions as part of these multidisciplinary approaches: psychologists in two interventions,70,94 physiotherapist assistants in one,82 students in one,109 a pharmacist in one111 and an occupational therapist in one. 81 The involvement of family or another key support person whom participants would invite to engage in activities111 or education sessions91 was mentioned in two studies.

Studies sometimes made reference to specific training or qualifications held by practitioners, including accredited qualifications, such as physical therapists with graduate diplomas in CBT,97 or exercise and sports science professionals with post-graduate qualifications in PA and health. 85 One study noted peer training, whereby physiotherapists and nurses had received training relating to cancer and exercise from a cancer exercise specialist,82 and several studies mentioned that practitioners had received some form of training from the study team69,78,85,87 or were following an accredited (intervention) course. 81 One study74 described regular supervision and therapists experienced in treating chronically fatigued cancer survivors, and two studies97,107 referenced staff with substantial clinical experience.

Interventions varied considerably in terms of the amount of practitioner contact or support they offered. This sometimes varied over time, or might be quite intense over a short period of time, for some interventions, or much more infrequent but over a much longer period of time in others. Therefore, when data were available, we considered the interventions as falling into one of three groups: 24 low contact (one to nine total contacts)52,54,57,58,66,68,76,81,83,84,91,93,94,97,103,104 (including in both the CBT and the education intervention arms of Archer et al. ,51 the counselling and the pedometer arms of Creel et al. ,64 the nurse arm of Goedendorp et al. ,74 both the home-based and the clinic-based arms of Johansson et al. ,86 and the home-based arm of Jolly et al. 87), 19 intermediate contact (10–20 total contacts)61,62,69,78,79,82,92,95,107–109 (including in the CBT arm of Goedendorp et al. ,74 the centre-based arm of Jolly et al. ,87 the home-based arm of Kraal et al. ,90 all three intervention arms of Losina et al. ,96 the clinic-based arm of Piva et al. 102 and the home-based arm of Smith et al. 105), and 17 high contact (≥ 21 total contacts)53,59,60,70,72,77,101,106,111 (including in all three intervention arms of Boesch et al. ,56 both the aerobic and the resistance arms of Courneya et al. ,63 the centre-based arm of Kraal et al. ,90 the community-based arm of Piva et al. 102 and the hospital-based arm of Smith et al. 105). We also looked at average weekly contacts. This ranged from 0.03 contacts (one contact across a 7-month pedometer-use intervention in the pedometer arm of Creel et al. 64) to 19 contacts (in which participants were residents at a residential facility providing several daily sessions, for 1 month, as in all arms of Boesch et al. 56). The mean number of contacts was 1.89 (median 0.97, mode 1.0). Excluding the 1-month residential intervention, which was an outlier, average weekly contacts ranged from 0.03 to 3.35 (mean 1.02, median 0.75).

Four studies were not included in the previous calculations; participants in Barnason et al. 55 received daily self-management messaging via the Health Buddy Telehealth device, but it was not clear whether this was entirely automated or whether physicians inputted messaging; Golsteijn et al. 75 provided automated computer-tailored advice at monthly intervals (3 months) across the intervention period (participants had the option to consult a physical therapist if they wished, although it was not clear whether or not participants took up this offer); and Van der Walt et al. ,110 an intervention based on daily steps goals, involved use of an activity tracker, but received no practitioner support. Mundle et al. ,100 an add-on to usual care cardiac rehabilitation, provided participants with accelerometers, but provided no details of the rehabilitation programme or any practitioner contact.

Duration of intervention

Interventions varied in duration from a mean of 19.4 (SD 1.4) days58 to 26 months. 94 Thirty-six interventions52,54,55,57,58,61,75,77,79,82,95,97,100,103,104,107,110,111 (including in both the CBT and the education intervention arms of Archer et al. ,51 all three arms of Boesch et al. ,56 the aerobic and the resistance arms of Courneya et al. ,63 the nurse arm of Goedendorp et al. ,74 the clinic-based and the home-based arms of Johansson et al. ,86 the centre-based and the home-based arms of Jolly et al. ,87 the walking and the cycling arms of Kinsey et al. ,88 the centre-based and the home-based arms of Kraal et al. ,90 and the clinic-based and the community-based arms of Piva et al. 102) lasted < 6 months, whereas 15 interventions72,76,78,85,93,106,109 (including in the CBT arm of Goedendorp et al. ,74 all three intervention arms of Losina et al. ,96 and the hospital-based and the home-based arms of Smith et al. 105) lasted 6 months. Seven studies108 took place over the course of 1 year or up to 2 years. One intervention83 was delivered throughout adjuvant therapy, and so varied in relation to patients’ treatment (Figure 8).

FIGURE 8.

Duration of delivery of interventions.

Education and advice

Seven interventions65,67,73,80,89,99,112 described the encouragement of PA adoption through ‘education sessions’: the provision of information, advice, recommendations or a formal exercise prescription.

The content of these interactions tended to fall into one or more of the following groups:

• Exercise recommendations, including studies in which exercise prescriptions, advice for home exercise, recommendations or goals were described. 65,67,73,80,89,99,112 This sometimes involved the recommendation of, for example, 10,000 daily steps,73 the self-management of a personalised exercise programme67 or activity goals. 99 Sometimes advice was provided in addition to supervised PA; for example, in Zopf et al. ,112 participants were recommended to undertake an additional 60 minutes of weekly exercise on top of weekly supervised classes. Studies provided variable details on how PA recommendations were delivered or managed; in some cases, participants self-managed these expectations, sometimes through the use of a PA diary73,89 or an activity device that monitored or measured activity. 73,99 Doganay et al. 67 describes the weekly monitoring and modification of participants’ adherence to the exercise programme (recommendations) by the participants’ clinical team, and Macleod et al. 99 gave weekly feedback during the first weeks of the intervention from their lifestyle coach. Most recommendations were personalised in response to a baseline assessment (usually fitness) or to the participants’ personal goals, or would be reviewed and set again in response to progress over time.

• Education or information materials, including studies in which participants were provided with exercise sheets;73 written resources with a focus on self-monitoring activity and goal-setting;99 or, as in Dantas et al. ,65 an ‘educational folder’ at hospital discharge, which detailed common physical and emotional experiences of patients following a coronary artery bypass graft (CABG), as well as information around the use of medication, activity progression and benefits, healthy diet, smoking cessation, sexual activity and managing weight.

• The value of PA,65,67,73 including studies that reported participants being encouraged to recognise the importance of ‘exercise for well-being’;73 incorporating ‘routine exercise’ into daily life for ‘lasting benefit’67 was also described.

• Lifestyle education – one study65 noted that the focus of education or advice was placed on broader improvements to health. This was delivered in the form of an ‘educational folder’, provided to patients on hospital discharge following CABG surgery (see previous list item ‘educational or information materials’).

Behavioural mechanisms

A therapeutic or behavioural approach to supporting participants to engage in PA was adopted in over half (66.7%) of interventions. 65,67,73,89,98,99 These sessions were delivered one to one, although in both Dantas et al. 65 and Macleod et al. 99 the participation of the patient’s family or a support person was encouraged.

Three interventions65,73,99 combined an initial period of face-to-face clinic contact with subsequent home telephone contact. For example, in Frawley et al. ,73 participants attended 8 weeks (two sessions per week) of in-person education sessions that used motivational interviewing and CBT techniques to discuss emotional management, diet and PA. These were followed by a series of six telephone motivational coaching sessions over the remaining intervention period (4 months). In Dantas et al. ,65 it appeared to be optional whether the latter counselling sessions, delivered at increasingly extended periods across the 6 months following hospital discharge, were provided at the clinic or on the telephone. Dantas et al. 65 also reported an invitation to patients or their families to call with any questions or problems, at any time.

The Komatsu et al. 89 intervention was delivered in a clinical setting; the Macchi et al. 98 intervention was delivered as part of a residential stay.

In Doganay et al. ,67 it was not clear from authors’ reporting where the intervention was delivered.

The finer detail of these sessions was limited. However, we identified some recurring topics or approaches in the content described. Five interventions67,73,89,98,99 talked about behaviour change approaches, counselling or motivational interviewing, such as coaching focused on removing bad habits or incorporating PA into participants’ daily routines,67 or motivational interviewing and CBT around emotional management, diet and PA. 73 As we saw in Frawley et al,73 these tended to overlap with references to lifestyle and health coaching,67,73,99 in which the focus might be on supporting participants to make broader behavioural changes, such as support from a lifestyle coach to reach ‘collaborative agreements’ towards achieving and maintaining, for example, smoking cessation, increased PA, caloric intake appropriate to weight status and a nutrient-dense diet. 99

Direct physical instruction

Five interventions71,73,80,98,112 (55.6%) provided face-to-face PA instruction. With the exception of Fontana et al. ,71 who described their intervention solely as medically supervised treadmill walking (three times per week), these sessions ran as one component of a broader approach that included the provision of education and advice73,80,112 and/or therapeutic behavioural mechanisms. 73,98 Activity sessions took place either individually71,98 or as group classes. 73,112 Among other activities (such as aerobic and resistance training), Zopf et al. 112 described ‘games and exercise’ ‘to promote interaction and communication’.

Physical activity sessions were provided at least weekly (although no details on the frequency of supervised training were provided in Heitkamp et al. 80). This ranged from once per week112 (over 15 months), to two or three sessions per week (for 8 weeks73 or 12 weeks71), to twice-daily sessions during a 3-week residential stay. 98

Physical activity instruction took place in a clinical setting71,73 or during a residential rehabilitation centre stay,98 or, as in Zopf et al. ,112 participants attended weekly ‘exercise classes’ in a local leisure centre. Again, no details were provided for Heitkamp et al. 80

Other components

Several studies65,67,99 reported additional features of intervention design. Dantas et al. ,65 for example, described efforts to improve communication among health teams and the patient and/or family and encouraged family participation in the patient’s health management. Macleod et al. ,99 too, noted the invitation for participants to engage a support person, such as a spouse, to assist in their adherence to the programme. Doganay et al. 67 described the reiteration of PA messaging by patients’ broader multidisciplinary clinical team at clinical follow-up appointments.

Digital technologies

There was limited utilisation of digital and other technologies to support delivery of interventions, or to provide objective outcome measures of participants’ PA. Two studies73,99 (22.2%) included the use of pedometers, provided to participants to support self-monitoring and goal-setting of PA99 and step count;73 this is a similar proportion of studies as in the RCTs.

Other resources/equipment

In addition to pedometers, the provision of specific resources or equipment was described in four studies. 65,73,99,112 This included the provision of an ‘educational folder’ containing information relating to the physical and emotional experiences often experienced by patients following CABG surgery, as well as guidance around the use of medication, activity progression and benefits, healthy diet, smoking cessation, sexual activity and body weight control;65 an exercise diary and home exercise sheets, provided to patients alongside a pedometer to support self-management of activity;73 and an ‘equipment toolkit’ for participants to use at home included a pedometer, resistance bands, digital versatile discs (DVDs) and booklets. 99 It is also worth noting that participants in Frawley et al. 73 had access to a well-equipped gym (aerobic and resistance training equipment including a treadmill, a stationary bike, a Pilates reformer, resistance bands and dumbbells).

Type of exercise/physical activity encouraged

Studies were often vague on the sorts of physical activities or exercises adopted by or recommended to participants. Some referred to aerobic training, including supervised aerobic training,73 aerobic exercise98,112 and counselling that aimed to facilitate at least 150 minutes of moderate-intensity activity per week. 99 Walking or the measurement of steps was described in three studies, including participants following a progressive programme of daily walking,65 treadmill walking71 and the encouragement of 10,000 daily steps. 73 Resistance or strength training,73,98,112 such as gentle calisthenic exercises to improve muscle strength or exercises for balance, was also reported,98 as were games and exercises focused on flexibility and relaxation,98,112 such as ‘gentle passive stretching’. 98 Both were in conjunction with other activities.

Doganay et al. 67 described a ‘personalised exercise programme’ with no details on the sorts of activities adopted by participants, except that the aim of the programme was to focus on PA that supported recovery and could be incorporated into daily routines. Similarly, Komatsu et al. ,89 noted counselling that encouraged PA, and that participants were to maintain a PA diary, but did not describe the sorts of activities participants undertook. Heitkamp et al. 80 reported supervised and home-based training, but provided no further details.

How much physical activity and at what sort of intensity

Studies reported varying approaches to the specification of the intensity at which participants should exercise or the frequency with which they should do it. In several studies65,67,99 this was largely undefined; rather, participants were encouraged to increase their activity such as by walking an additional 5 minutes each day up to a specified goal of 60 minutes,65 or to progress PA for lasting benefit. 67 Others used standard measures to focus intensity; Frawley et al. ,73 for example, stated participants were required to do at least 150 minutes of moderate-intensity activity at a Borg Scale score of ≥ 13. In Zopf et al. ,112 intensity was adjusted individually based on each patient’s condition and experience, but was aimed at moderate-intensity exercises of approximately 3.84 to 4.84 METs/hour/week. Similarly, for Fontana et al. ,71 target training intensity (for 35 minutes of treadmill walking each week) was tailored to each individual’s condition, to within 85% of maximal heart rate, determined by exercise testing at the start of, and 6 weeks into, the intervention period.

Although Macchi et al. 98 provided no details in relation to the aerobic activities, they referred to the adoption of ‘gentle passive’ stretching and ‘gentle’ strengthening exercises. Two studies80,89 provided no details.

The total amount of PA either delivered or encouraged tended to fall into one of three groups. The low-PA group comprised those studies encouraging PA once or twice per week, for example the recommendation of one supervised session and 60 minutes of home exercise each week. 112 The regular-PA group comprised studies largely in line with typical PA recommendations of three or four times per week. 71 Studies in the frequent group encouraged PA on at least five days per week. 65,67,73,99 Three studies80,89,98 did not provide this information.

Tailoring activity

The majority (88.9%) of interventions65,67,71,73,89,98,99,112 took some form of personalised or tailored approach to PA. This was usually65,71,73,98,112 described with respect to participants’ capacities or fitness, such as target heart rates determined by participants’ fitness,71 or the adjustment of intensity based on a patient’s condition and activity experience. 112 Two interventions, however, focused on participants’ preferences, such as determining specific action goals,99 or the collaborative development of an exercise programme between participants and their clinical team. 67

Clinical settings

Over half of interventions65,71,73,89,99 took place in a clinical setting (outpatient clinic or rehabilitation centre), either exclusively71,89 or in conjunction with additional telephone contact with participants. Komatsu et al. 89 included expectations for self-directed home exercise.

Home

Four interventions65,73,80,99 took place in participants’ homes (or neighbourhood environment): coaching or counselling took place via telephone, in addition to some expectation of self-directed PA, and some face-to-face clinic contacts,73,99 including Dantas et al. 65 in which participants had the option to decide whether to meet in person or to receive telephone calls. Telephone contacts between participants and their clinical team (lifestyle coach, physiotherapists, research nurse) took place roughly every 3 or 4 weeks,73,99 or at months 1, 3 and 6 after hospital discharge. 65 Heitkamp et al. 80 noted home-based exercise training, but provided no further details.

Inpatient care

One intervention,98 for cardiac surgery patients, was delivered in a residential rehabilitation centre, across a 3-week stay.

Community

In Zopf et al. ,112 patients attended a weekly exercise class at a community sports centre. There was also an expectation for self-directed home exercise.

Although Heitkamp et al. 80 included home-based training, the authors provide no details as to where the supervised sessions took place. One further study67 provided no details on intervention setting.

Remote delivery

Three studies65,73,99 described some form of remote delivery. None was exclusively remote and instead combined either an initial or early face-to-face contact along with home resources, such as an education folder;65 exercise diaries, exercise sheets and pedometers;73 and subsequent telephone support (optional in-person in Dantas et al. 65). In Macleod et al. ,99 colorectal patients received a face-to-face session at the beginning of each phase of the three-phase, perioperative programme, followed by ongoing telephone contact to support participants’ progress. Each participant received an ‘equipment toolkit’, which included a pedometer, resistance bands, DVDs and booklets. Participants were also encouraged to engage a support person, such as a spouse, to assist in their adherence to the programme.

Practitioners and those involved in delivery

Interventions were often delivered by a MDT. 65,67,73,99 Specialisms within these teams varied between interventions; several included physiotherapists,65,73,99 with a lifestyle coach,99 dietitian,73 research nurse65 and psychologist73 also involved. In addition, Dantas et al. 65 and Macleod et al. 99 encouraged a family member or support person to actively support patients’ health management and adherence to programmes.

Doganay et al. 67 described a multidisciplinary clinical team, but provided no further details.

Two interventions were delivered by a single practitioner: one described as a nurse89 and one as a ‘qualified trainer’. 112 Fontana et al. 71 described the intervention being delivered under ‘medical supervision’. Heitkamp et al. 80 and Macchi et al. 98 provided no details.

Two studies mentioned specific training, qualifications or experience held by practitioners. Frawley et al. 73 described exercise physiologists who were experts in exercise prescription, with at least 2 years’ clinical experience, and Macleod et al. ,99 described lifestyle coaches with a ‘nursing background’; experience with cancer patient management; and having undertaken a 3-day bespoke training programme covering smoking cessation, increasing moderate PA, and brief interventions on alcohol and weight management.

Interventions varied considerably in terms of the amount of practitioner contact or support they offered. This sometimes varied over time, or might be quite intense over a short period of time for some interventions and much more infrequent over a much longer period of time in others. As a result, where data were available, we considered the interventions as falling into one of three groups:

1. low contact89 (25%) (one to nine total contacts), including Dantas et al. ,65 in which patients or family members were encouraged to call the MDT at any time if questions or problems arose

2. intermediate contact67 (25%) (10–20 contacts), including Macleod et al. ,99 in which a minimum of 12 contacts with the lifestyle coach was offered; they also reported that this might vary, depending on a patient’s individual treatment regime

3. high contact71,73,98,112 (50%) (≥ 21 contacts).

As the duration of interventions varied considerably, we also looked at the average number of weekly contacts. This ranged from 0.15 contacts (one contact at hospital discharge with the provision of educational materials, followed by three contacts in person or on the telephone over a 6-month period, as in Dantas et al. 65) to 14 contacts (two sessions per day during a 3-week residential rehabilitation centre stay, as in Macchi et al. 98). The mean number of contacts was 2.98, the median was 0.92 and the mode was 1. Excluding the 3-week residential intervention, which was something of an outlier, average weekly contacts ranged from 0.15 to 3.0 (mean 0.95, median 0.85). This contact time was comparable to the interventions in the RCTs.

No details were provided in Heitkamp et al. 80

Duration of intervention

Interventions varied in duration from 3 weeks98 to 64 weeks. 112 With the exception of Macchi et al. 98 (3 weeks), all interventions lasted at least 3 months, although most lasted ≤ 6 months. 65,67,71,73,89,98 One intervention,67 delivered to oesophagogastric cancer patients, varied in duration, typically between 8 and 18 weeks, depending on a patient’s treatment plan; we have used a median point for the purpose of calculations.

Amount of physical activity

Amount of physical activity at the end of follow-up measured as minutes per day or per week

Twenty-one studies57,59,60,62,64,66,74,75,78,81,82,85,94,96,97,100,104,106,107,109,111 used instruments to measure minutes of PA per day or per week. A combination of objective methods (such as using accelerometers) and subjective methods (such as using questionnaires) was used to record PA data in these studies.

Twelve studies59,66,74,75,78,82,85,94,96,100,104,109 reported PA as post-intervention values. Of these studies, Demark-Wahnefried et al. ,66 Hawkes et al. ,78 Lier et al. 94 and Sellberg et al. 104 measured the outcome at 12 months; Jiménez-Loaisa et al. 85 measured the outcome at 13 months; and the remaining studies measured the outcome at 6 months. When combining these studies, we found that people participated in more PA at the end of follow-up when they had received the intervention (SMD 0.15, 95% CI 0.04 to 0.27; participants, n = 1947; studies, n = 12; I² = 24%; favours intervention) (Figure 9). This effect was likely to be small. 49

FIGURE 9.

Combined data: amount of PA at end of follow-up measured as minutes per day or week. a, Multiarm study: in this analysis, we combined CBT vs. usual care; b, multiarm study: in this analysis, we combined FI + THC vs. attention control. df, degrees of freedom; IV, inverse variance.

We conducted the following subgroup analyses on these data:

• Duration of intervention. We subgrouped the studies according to whether the intervention was undertaken for < 6 months75,82,100,104 or for at least 6 months. 59,66,74,78,85,94,96,109 From our visual inspection of the data, we noted that, when the intervention was given for at least 6 months, people receiving the intervention appeared to participate in more PA at the end of follow-up (SMD 0.23, 95% CI 0.12 to 0.33; participants, n = 1303; studies, n = 8; I² = 0%; favours intervention) (see Appendix 2, Figure 16). However, this difference was not supported by a formal test for subgroup interactions (p = 0.10).

• Time of intervention commencement. We subgrouped the studies according to whether the intervention was given pre surgery74,94 or post surgery. 59,66,75,78,82,85,96,100,104,109 From our visual inspection of the data, we noted that, when the intervention was given post surgery, people receiving the intervention appeared to participate in more PA at the end of follow-up (SMD 0.14, 95% CI 0.02 to 0.25; participants, n = 1735; studies, n = 10; I² = 18%; favours intervention (see Appendix 2, Figure 17). However, this difference was not supported by a formal test for subgroup interactions (p = 0.65).

• Type of surgery. We subgrouped the studies according to whether the participants underwent surgery for types of cancer59,66,74,75,78,82 or for other conditions. 85,94,96,100,104,109 During our visual inspection of the data, we noted that, when the intervention was given to participants undergoing surgery for cancer, people receiving the intervention appeared to participate in more PA at the end of follow-up (SMD 0.22, 95% CI 0.07 to 0.38; participants, n = 1472; studies, n = 6; I² = 45%; favours intervention) (see Appendix 2, Figure 18); we noted a moderate level of statistical heterogeneity in this effect. However, this difference was not supported by a formal test for subgroup interactions (p = 0.95).

• Age. We subgrouped the studies according to whether the participants had a mean age of < 60 years59,66,74,85,94,104 or at least 60 years;75,78,82,96,109 one study100 did not report baseline characteristics for age. There was no statistically significant difference between the two subgroups (favours intervention; see Appendix 2, Figure 19).

• BMI. We were unable to perform a subgroup analysis on the basis of BMI because of insufficient data.

We also conducted sensitivity analyses on this meta-analysis. When we excluded studies that measured this outcome immediately after the end of the intervention period, we found evidence of a possible increase, as well as a decrease, in amount of PA at the end of follow-up (SMD 0.11, 95% CI –0.03 to 0.26; participants, n = 1517; studies, n = 7; I² = 35%); we noted a moderate level of statistical heterogeneity in this effect. For all sensitivity analyses for this outcome, see Appendix 4, Table 22.

The following additional data also provided evidence for this measure:

• Two studies60,97 that used instruments to measure minutes of PA per day or per week reported their data based on changes from baseline; both measured the outcome at 6 months, and both used accelerometers to collect their data. We found little or no difference in the amount of PA according to whether an intervention or usual care was provided (MD –1.56, 95% CI –15.71 to 12.60; participants, n = 214; studies, n = 2; I² = 76%; favours usual care) (see Appendix 2, Figure 20); we noted a substantial to considerable level of statistical heterogeneity in this effect.

• Bond et al. 57 measured the amount of PA as mean bout-related MVPA minutes per day, but did not report a SD for these data, and did not clearly report the number of participants analysed at follow-up. The study authors reported little or no difference in the amount of PA according to whether an intervention or usual care was provided (p = 0.15) (see Appendix 3, Table 10).

• Christiansen et al. 62 measured the amount of PA as total MVPA minutes per day, and reported their data as median [interquartile range (IQR)]. The study authors reported little or no difference in the amount of PA according to whether an intervention or usual care was provided (p = 0.55; participants, n = 22) (see Appendix 3, Table 10).

• Creel et al. 64 reported their PA data as mean bout-related minutes per week, but did not include a SD with these data; the outcome was measured at 6.5 months. The study authors reported that people participated in more PA at the end of follow-up when they had received the intervention (p < 0.05; participants, n = 58) (see Appendix 3, Table 10).

• Stolberg et al. 106 reported their data as the difference between groups at 24 months post surgery. The study authors reported little or no difference in the amount of PA according to whether an intervention or usual care was provided (p-value not stated; participants, n = 42) (see Appendix 3, Table 10).

• Taraldsen et al. 107 reported their PA data as mean minutes per day of upright time, and based on between-group differences at 6 months post intervention. The study authors reported little or no difference in the amount of PA according to whether an intervention or usual care was provided (p = 0.346; participants, n = 113) (see Appendix 3, Table 10).

• Yates et al. 111 reported their PA data based on changes from 3 to 6 months post surgery, and gave their data as a median (range). The study authors reported little or no difference in the amount of PA according to whether an intervention or usual care was provided (p = 0.79; participants, n = 34) (see Appendix 3, Table 10).

• Hoorntje et al. 81 reported their data based on changes in the percentage of time spent active from baseline to 6 months post surgery. The study authors reported little or no difference in the amount of PA according to whether an intervention or usual care was provided (p = 0.59; participants, n = 97) (see Appendix 3, Table 10).

Engagement in physical activity at the end of follow-up

Ten studies63,69,70,72,91,94,95,101,108,109 measured engagement in PA.

Seven of these studies69,70,72,91,94,101,109 measured the outcome at 12 months, one study63 measured the outcome at 6 months and one study108 measured the outcome at 24 months. When analysing these studies, we found little or no difference in the number of people engaged in PA at the end of follow-up when they had received the intervention (RR 1.19, 95% CI 0.96 to 1.47; participants, n = 882; studies, n = 9; I² = 25%) (Figure 10). We had an insufficient number of studies to conduct a subgroup analysis. Exploration of the data in sensitivity analyses identified no differences in our interpretation of this effect (see Appendix 4, Table 28).

FIGURE 10.

Combined data: engagement in PA at the end of follow-up. a, Multiarm study: in this analysis, we compared aerobic exercise training group with usual care. df, degrees of freedom; M–H, Mantel–Haenszel.

In addition, one study95 reported engagement in PA as analysed data at 12 months. The study authors reported that more people were engaged in PA when they had received the intervention (p = 0.020; participants, n = 197) (see Appendix 3, Table 14).

Physical fitness

Health-related quality of life

Twenty-two studies52–54,58,59,63,66,69,75,78,79,82,85,95,97,101–104,106,107,110 measured HRQoL.

Seventeen studies52–54,58,59,63,66,75,78,79,82,85,101–104,110 reported the outcome as a set of post-intervention values. Three of these studies measured the outcome using components of the Functional Assessment of Cancer Therapy (FACT): two studies66,103 used FACT-General and one study63 used FACT-Anaemia. Two of these studies63,103 measured the outcome at 6 months and one66 measured the outcome at 12 months. Eight studies54,58,59,78,85,101,102,104 measured the outcome using components of the SF-36; we extracted the data from the global score or general score for Brandes et al. ,58 Cadmus et al. ,59 Jiménez-Loaisa et al. 85 and Sellberg et al. ;104 for the remaining studies, we extracted the physical component summary (PCS). Of these studies, four54,58,59,102 measured the outcome at 6 months, three78,101,104 measured the outcome at 12 months and one85 measured the outcome at 13 months. One study52 measured the outcome, at 6 months, using the Knee injury and Osteoarthritis Outcome Score (KOOS). One study79 measured the outcome, at 5 years, using the Hip disability and Osteoarthritis Outcome Score (HOOS). One study53 measured the outcome using the Laval Questionnaire and reported their data as a weight-related quality-of-life score, measuring the outcome at 12 months. One study75 measured the outcome using the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire-C30 Global (EORTC QLQ-C30 Global), and measured the outcome at 6 months. One study110 measured the outcome, at 6 months, using the EQ-5D. The remaining study82 measured the outcome using the EuroQol-5 Dimensions, five-level version (EQ-5D-5L), and measured the outcome at 6 months. When analysing these studies, we found a possible improvement in HRQoL at the end of follow-up with an intervention, as well as a possible reduction (SMD 0.12, 95% CI –0.03 to 0.26, favours intervention; participants, n = 2455; studies, n = 17; I² = 60%; favours intervention) (Figure 11); we noted a moderate to substantial level of statistical heterogeneity in this effect. The effect size is likely to be small. 49

FIGURE 11.

Combined data: HRQoL at the end of follow-up. Measured using a range of measurement tools. a, Multiarm study: in this analysis, we compared aerobic exercise group with usual care; b, multiarm study: in this analysis, we compared clinic-based group with usual care. df, degrees of freedom; IV, inverse variance.

We conducted the following subgroup analyses on these data:

• Duration of intervention. We subgrouped the studies according to whether the intervention was undertaken for < 6 months52,54,58,63,75,79,82,102–104,110 or for at least 6 months. 53,59,66,78,85,101 There was no statistically significant difference between the two subgroups. During our visual inspection of the forest plots, we noted that people receiving the intervention appeared to have improved HRQoL at the end of follow-up when the intervention was given for at least 6 months (SMD 0.32, 95% CI 0.01 to 0.62; participants, n = 1045; studies, n = 6; I² = 75%; favours intervention); we noted a substantial to considerable level of statistical heterogeneity in this effect. However, this effect was not supported by a formal test for subgroup interactions (p = 0.11) (see Appendix 2, Figure 40).

• Time of intervention commencement. We subgrouped the studies according to whether the intervention was given pre surgery53,54,103,110 or post surgery. 52,58,59,63,66,75,78,79,82,85,101,102,104 There was no statistically significant difference between the two subgroups (see Appendix 2, Figure 41).

• Type of surgery. We subgrouped the studies according to whether the participants underwent surgery for types of cancer59,63,66,75,78,82,103 or for other conditions. 52–54,58,79,85,101,102,104,110 There was no statistically significant difference between the two subgroups (see Appendix 2, Figure 42).

• Body mass index: we subgrouped the studies according to whether the participants had a mean BMI of < 30 kg/m² (10 studies58,59,63,66,75,78,79,101,103,110) or of at least 30 kg/m² (three studies85,102,104). Four studies52–54,82 did not report baseline characteristics for BMI. There was no statistically significant difference between the two subgroups (see Appendix 2, Figure 43).

• Age. We subgrouped the studies according to whether the participants had a mean age of < 60 years53,59,63,66,85,101,104 or of at least 60 years. 52,54,58,75,78,79,82,102,103,110 There was no statistically significant difference between the two subgroups (see Appendix 2, Figure 44).

We also conducted sensitivity analyses on this meta-analysis (see Appendix 4, Table 34). We noted that the estimate clearly favoured the intervention group when we excluded studies that were deemed to be at a high or an unclear risk of selection bias for random sequence generation (SMD 0.17, 95% CI 0.01 to 0.33; participants, n = 2167; studies, n = 12; I² = 63%), or we excluded studies at a high risk of attrition bias (SMD 0.34, 95% CI 0.00 to 0.68; participants, n = 913; studies, n = 6; I² = 77%). We noted moderate to considerable levels of statistical heterogeneity in these estimates.

The following additional data also provided evidence for this measure:

• Two studies95,97 measured the outcome using the EQ-5D and reported the outcome based on change from baseline. Lindbäck et al. 95 reported the outcome based on change from baseline to 12 months, whereas Lotzke et al. 97 reported the outcome based on change from baseline to 6 months. When analysing these studies, we found little or no difference between groups in this outcome according to whether an intervention or usual care was provided (MD –0.05, 95% CI –0.11 to 0.01; participants, n = 315; studies, n = 2; I² = 0%; favours usual care) (see Appendix 2, Figure 45).

• One study69 measured the outcome using the Functional Assessment of Cancer Therapy – Breast cancer plus arm subscale (FACT-B+4), and reported the outcome based on change from baseline to 12 months. For this study we found little or no difference between groups in this outcome according to whether an intervention or usual care was provided (MD 3.70, 95% CI –1.48 to 8.88; participants, n = 126; studies, n = 1; favours intervention) (see Appendix 2, Figure 46).

• One study106 measured the outcome using the PCS of the SF-36; the outcome was reported as the difference between groups, and measured at 24 months post surgery. The study authors reported little or no difference in the outcome according to whether an intervention or usual care was provided (the study authors reported the p-value as not significant; participants, n = 42) (see Appendix 3, Table 17).

• Taraldsen et al. 107 reported the outcome data based on between-group differences at 6 months post intervention. The study authors reported little or no difference in HRQoL according to whether an intervention or usual care was provided (p = 0.965; participants, n = 113) (see Appendix 3, Table 17).

Pain

Eleven studies52,58,59,79,84,95,97,103,104,106,110 measured pain using a range of measurement scales.

Seven studies52,58,59,79,103,104,110 reported pain as a set of post-intervention values. Three studies58,59,104 measured pain using the relevant subscore of the SF-36 HRQoL tool. Of these, Brandes et al. 58 and Cadmus et al. 59 measured the outcome at 6 months, whereas Sellberg et al. 104 measured the outcome at 1 year. Two studies52,110 measured pain using the relevant subscore of the KOOS HRQoL tool; both of these studies measured the outcome at 6 months. Heiberg et al. 79 measured pain using the relevant subscore of the HOOS HRQoL tool and measured the outcome at 5 years. Santa Mina et al. 103 measured pain using the Pain Disability Index, and measured the outcome at 6 months. Whereas all previously mentioned measurements report pain using a scale whereby a higher score indicates a better outcome, the Pain Disability Index reports pain using a scale whereby a higher score indicates a worse outcome. To avoid a unit-of-analysis error, we inverted the data for Santa Mina et al. 103 When analysing these studies, we found little or no difference between groups in pain according to whether an intervention or usual care was provided (SMD 0.19, 95% CI –0.14 to 0.53; participants, n = 602; studies, n = 7; I² = 71%) (Figure 12); we noted a substantial level of statistical heterogeneity in this effect. Exploration of the data in sensitivity analyses identified no differences in our interpretation of this effect (see Appendix 4, Table 35).

FIGURE 12.

Combined data: pain measured at the end of follow-up using a range of pain scores. df, degrees of freedom; IV, inverse variance.

In addition, two studies95,97 measured back pain and leg pain using the region-specific VAS, and reported their data based on changes from baseline; in this scale, a lower score indicates less pain. Lindbäck et al. 95 measured the outcome at 12 months, whereas Lotzke et al. 97 measured the outcome at 6 months. When analysing the data for back pain, we found little or no difference between groups in this outcome according to whether an intervention or usual care was provided (MD 5.45, 95% CI –1.03 to 11.92; participants, n = 315; studies, n = 2; I² = 0%; favours intervention). Similarly, when analysing the data for leg pain, we found little or no difference between groups in this outcome according to whether an intervention or usual care was provided (MD 2.00, 95% CI –6.19 to 10.18; participants, n = 315; studies, n = 2; I² = 0%; favours intervention) (see Appendix 2, Figure 47). Exploration of the data in sensitivity analyses identified no differences in our interpretation of this effect (see Appendix 4, Table 35).

The following additional data also provided evidence for this measure:

• Ilves et al. 84 measured back and leg pain using the region-specific VAS, and reported their data based on changes from baseline to 12 months post intervention; in this scale, a lower score indicates less pain. The study authors reported the data using median (IQR). They reported little or no difference in pain according to whether an intervention or usual care was provided (back pain, p = 0.76; leg pain, p = 0.40) (see Appendix 3, Table 18).

• Stolberg et al. 106 measured bodily pain as a subscore of the SF-36 scale, and reported their data as a difference between groups 24 months post surgery. The study authors reported a non-significant difference (see Appendix 3, Table 18).

Adverse events

Seventeen studies51,56,59,61,66,69,76,77,79,82,83,97,102–104,107,109 reported information regarding adverse events. Of these, seven51,56,59,79,97,104,109 reported that no participants experienced adverse events that were related to the intervention.

Ten studies61,66,69,76,77,82,83,102,103,107 reported numbers of adverse events that were related or unrelated to the intervention, or were serious or not serious. Table 1 presents a summary of these events.

Adherence

Fifteen studies52,53,59,63,69,76,77,82–84,92,96,103,106,108,109 reported adherence data (Table 2).

Participant experiences of the interventions

Four studies52,69,76,82 reported data on participants’ experiences of the interventions. Table 3 presents a summary of these data.

Amount of physical activity (37 studies, 4969 participants)

We found moderate-certainty evidence that PA interventions may increase the amount of PA in the ‘medium term’ after surgery. Most studies measured this outcome at either 6 or 12 months after surgery.

This outcome was evaluated using a range of measurement values that we could not combine in a single analysis: minutes of PA per day or week, energy expenditure measures such as kcal/kg/day, METs/minutes/week, steps/day, daily activity scores or results from activity-related questionnaires.

Physical activity was most commonly recorded as minutes per day or week (12 studies, 1947 participants), and analysis revealed a small increase in PA among participants who received the intervention. We similarly found a small increase in PA when reported as steps/day or from questionnaire data. Although energy expenditure measures, METs/minutes/week and the results of daily activity scores showed both a reduction and increase in PA when participants received the intervention, we noted that the effect size in all these analyses favoured the intervention. We believe that there was a consistent finding across all measures that the intervention may increase the likelihood that people would do more PA.

Engagement in physical activity (10 studies, 1097 participants)

We found moderate-certainty evidence that PA interventions probably slightly increase people’s engagement in PA, compared with usual care. For this outcome, we found that 60 more participants per 1000 would still be engaging in PA at the end of follow-up after receiving the intervention. However, we noted a wide CI for the data from this outcome, indicating that some people receiving the intervention will do less PA.

Physical fitness (15 studies, 1031 participants)

Although we noted that effects generally favoured the PA intervention, we found low-certainty evidence that PA interventions may improve physical fitness in the ‘medium term’ after surgery. Most studies measured this outcome at either 6 or 12 months.

This outcome was evaluated using various measures, and the differences in these measures of fitness may reflect the age of participants or the reason for surgery, or both. This outcome was measured using walking tests (such as the 6MWT), exercise tolerance tests (measured as METs maximum), sit-to-stand tests (such as the TUG test), handgrip strength tests, leg-press tests and VO₂ peak tests. It was not feasible for us to combine the data when the measurement tools were so different, and most of our effects were derived from single studies or from two studies. In general, we noted a similar trend that suggested an improvement in fitness when people had received a PA intervention.

Health-related quality of life (22 studies, 3051 participants)

We found moderate-certainty evidence that PA interventions probably slightly increase HRQoL at the end of follow-up.

This outcome was evaluated using a range of measures that allowed for the combining of data using a SMD. The tools used were appropriate measures of quality of life, with some of these measures specific to the surgery within the study criteria (e.g. KOOS and FACT-Anaemia). Although the primary analysis showed a slight reduction in HRQoL as well as an increase, the findings more clearly favoured the PA interventions once we removed studies rated as having a high or an unclear risks of bias from the analysis.

Pain (11 studies, 1057 participants)

Again, the findings for pain tended to favour the intervention. However, the estimates were all imprecise and included possible benefits as well as harms; we judged the certainty of this evidence to be low.

Adherence (15 studies, 766 participants randomised to intervention groups)

Among studies clearly reporting this outcome, the level of adherence ranged from 47% to 93%. We noted that 93% of studies reported that adherence to the intervention was > 53% and that 20% reported that adherence was > 79%. However, the definitions of adherence varied between studies, and, because the designs of the interventions differed significantly, it was not reasonable to draw confident conclusions about adherence to all PA interventions; we judged the certainty of this evidence to be very low.

Adverse events (10 studies, 1410 participants)

We could not be certain of the risk of adverse events for interventions promoting PA because few studies reported adverse events data; we judged the certainty of this evidence to be very low. We noted that most events were described as not serious, and that most events were unrelated to the intervention. The few events described as possibly related to the intervention included muscle soreness, musculoskeletal injury, indigestion, abdominal bloating, knee pain/discomfort, syncope, arthralgia, back pain, myalgia, falls, musculoskeletal/connective tissue effects and skin/subcutaneous tissue effects; these events were reported for 30 participants.

Participant experience of intervention (four studies, 159 participants randomised to intervention groups)

Very few studies reported details of participants’ experiences. In the four studies that clearly reported this information, feedback was generally positive and participants were satisfied and/or felt that they had benefited from being able to engage with the intervention. We did not downgrade the certainty of this narrative evidence.

Amount of physical activity

Engagement in physical activity at the end of follow-up

This outcome was measured in three studies,86,88,105 which we did not pool because of differences in intervention designs.

Johansson et al. 86 compared a clinic-based intervention with a home-based intervention, and measured the outcome at 12 months. For this study, we found that more people were engaged in PA at the end of follow-up after receiving the clinic-based intervention (RR 1.25, 95% CI 1.03 to 1.52; participants, n = 57; studies, n = 1; favours clinic-based intervention) (see Appendix 2, Figure 52).

Smith et al. 105 compared a home-based intervention with a hospital-based intervention, and measured the outcome at 6 years. For this study, we found little or no difference in engagement in PA at the end of follow-up according to which type of intervention was provided (RR 1.30, 95% CI 0.90 to 1.87; participants, n = 108; studies, n = 1; favours home-based intervention) (see Appendix 2, Figure 53).

Kinsey et al. 88 compared a home walking intervention with a home cycling intervention, and measured the outcome at 4 years. The study authors did not report the data by group, and so we could not determine any differences between groups (see Appendix 3, Table 19).

Physical fitness

Health-related quality of life

Five studies51,86,87,90,105 measured HRQoL:

1. Smith et al. 105 measured the outcome using the PCS of the SF-36, and compared a group receiving a home training intervention with a group receiving a hospital training intervention; the study investigators measured the outcome at 6 years. In this study, we found little or no difference in HRQoL between the home and the hospital training groups at the end of follow-up (MD –2.30, 95% CI –5.70 to 1.10; participants, n = 144; studies, n = 1; favours centre-based intervention) (see Appendix 2, Figure 57).

2. Jolly et al. 87 compared a centre-based intervention with a home-based intervention and measured the outcome at 24 months using the EQ-5D. In this study, we found little or no difference in HRQoL between the centre- and the home-based intervention groups at the end of follow-up (MD 0.02, 95% CI –0.03 to 0.07; participants, n = 454; studies, n = 1; favours centre-based intervention) (see Appendix 2, Figure 58).

3. Kraal et al. 90 measured the outcome using the total score from the MacNew Heart Disease Health-Related Quality Of Life Questionnaire, which uses a scale from 1 to 7, whereby a higher score indicates a higher quality of life. The study investigators compared a group receiving a home-based intervention with a group receiving a centre-based intervention, and measured the outcome at 12 months. In this study, we found a small improvement in HRQoL at the end of follow-up when participants had received the home-based intervention (MD 0.32, 95% CI 0.27 to 0.37; participants, n = 78; studies, n = 1; favours home-based) (see Appendix 2, Figure 59).

4. Johansson et al. 86 measured the outcome using the EQ-5D and compared a clinic-based intervention with a home-based intervention and measured the outcome at 12 months. We did not calculate an effect estimate because the data were reported with median values. The study authors reported little or no difference between groups in HRQoL (p = 0.35; participants, n = 57) (see Appendix 3, Table 20).

5. Archer et al. 51 measured HRQoL using the PCS of the Short Form questionnaire-12 items compared a group receiving a CBPT intervention with a group receiving an education intervention and measured the outcome at 12 months. For this study we found little or no difference in the outcome according to whether a CBPT intervention or an education intervention was provided (MD 1.82, 95% CI –1.44 to 5.08; participants, n = 229; studies, n = 1; favours CBPT) (see Appendix 2, Figure 60).

Pain

Three studies51,86,87 measured pain using various components:

1. Archer et al. 51 measured back and leg pain using the region-specific Brief Pain Inventory, in which a higher score indicated a worse outcome. The study investigators compared a group receiving a CBPT intervention with a group receiving an education intervention, and reported the outcomes at 12 months. When analysing the data for back pain, we found little or no difference between CBPT and education intervention groups (MD 0.12, 95% CI –0.54 to 0.78; participants, n = 229; studies, n = 1; favours education). Similarly, when analysing the data for leg pain, we found little or no difference between the CBPT and the education intervention groups (MD –0.46, 95% CI –1.17 to 0.25; participants, n = 229; studies, n = 1; favours CBPT) (see Appendix 2, Figure 61).

2. Jolly et al. 87 measured pain using self-reported chest pain on movement; a higher score indicated a worse outcome. The study investigators reported the outcome at 24 months, and compared a group receiving a centre-based intervention with a group receiving a home-based intervention. When analysing these data, we found little or no difference between the centre- and the home-based intervention groups (MD 0.09, 95% CI –0.09 to 0.27; participants, n = 342; studies, n = 1; favours home-based intervention) (see Appendix 2, Figure 62).

3. Johansson et al. 86 measured the outcome using the region-specific VAS in which a lower score indicates a better outcome. The study investigators measured back pain and leg pain, and reported the outcomes at 12 months. They also compared a group receiving a clinic-based intervention with a group receiving a home-based intervention. We did not calculate an effect estimate for these data, which were reported as median values. The study authors reported a reduction in back pain for participants receiving the home-based intervention (p = 0.04; participants, n = 57) (see Appendix 3, Table 21).

Adverse events

Two studies87,90 reported information regarding adverse events. Both studies reported numbers of adverse events. These events are summarised in Table 4.

Adherence

One study90 reported adherence data; Table 5 shows the data for this study.

Participant experience of intervention

One study87 reported data on participants’ experiences of the interventions; Table 6 presents this data.

Amount of physical activity (four studies, 785 participants)

We were uncertain whether or not there was any difference between the interventions in the amount of PA at the end of follow-up; we judged the certainty of the evidence to be very low.

In each study, the estimates indicated little or no difference between intervention groups when measured at 12 months or 24 months. The outcome was evaluated using a different measurement tool in each study: energy expenditure measures, activity counts per minute, the GSLTPAQ and the PAL score.

Engagement in physical activity at the end of follow-up (three studies, 204 participants)

We were uncertain whether or not there was any difference between the interventions in engagement in PA; we judged the certainty of the evidence to be very low.

In one study,86 the evidence indicated that more participants receiving a clinic-based intervention engaged in PA at 12 months than participants who had received a home-based intervention. However, this finding was not demonstrated in another study105 that compared a hospital-based intervention with a home-based intervention and found little or no difference between groups. Differences between the interventions in these studies meant that comparing the findings in these two studies was not appropriate.

Physical fitness (three studies, 564 participants)

We were uncertain whether or not there was any difference between the interventions in physical fitness; we judged the certainty of the evidence to be very low.

We pooled the data from two studies90,105 and found little or no difference between a home-based and a clinic-based intervention in VO₂ peak. We similarly found little or no difference between interventions in the distance walked in a walking test. In one study,105 a home-based intervention showed improved results in exercise tolerance tests, compared with a clinic-based intervention.

Health-related quality of life (five studies, 962 participants)

We were uncertain whether or not there was any difference between the interventions in HRQoL; we judged the certainty of the evidence to be very low.

We did not pool the data from any of these studies because of the differences in the study designs. Only one small study90 (78 participants) noted an improvement in HRQoL when a home-based intervention was provided, rather than a centre-based intervention. Other studies, however, noted little or no difference in HRQoL between their intervention groups, and effects were often imprecise.

Pain (three studies, 628 participants)

We were uncertain whether or not there was any difference between the interventions in pain; we judged the certainty of the evidence to be very low.

Again, we did not pool the data from these studies. Only one small study86 (57 participants) noted an improvement in pain when a home-based intervention was provided, rather than a clinic-based intervention. However, this finding was not consistent with the other two studies, for which there appeared to be little or no difference between interventions in pain.

Adverse events (two studies, 615 participants)

We were uncertain whether or not any of interventions led to adverse events, and the certainty of this evidence was very low.

There were few adverse events reported in only two studies,87,90 and, because of limits in the reporting of these data, we could not ascertain whether these adverse events were related or unrelated to the interventions.

Adherence (one study, 90 participants)

It was not reasonable to draw confident conclusions about adherence to all PA interventions, as this was reported in only one study. 90 There appeared to be a comparable number of exercise sessions completed in both intervention groups in this study. We assessed the certainty of this evidence to be very low.

Participants’ experiences of the intervention (one study, 525 participants)

Only one study87 reported feedback on participants’ experiences. Participants in the centre-based intervention had enjoyed exercising in a group, and those who exercised at home were motivated by regular contact with a study nurse. We did not downgrade the certainty of this narrative evidence.

Supporting a holistic, well-being approach

A ‘holistic’ approach, focused on a patient’s overall well-being, was frequently described, with many participants commenting on the link between PA and a patient’s psychological profile and ‘sense of self’. Several participants indicated that they frame the impact of their programmes to patients using the same narrative, encouraging patients to understand how activity might support them through their treatment, and help them manage stress and anxiety. This appeared to be seen as central to the efficacy of their approaches and an acceptable narrative with which patients could connect. This focus often included broader lifestyle aspects such as smoking and alcohol reduction; efforts to keep patients nutritionally ‘stable’; or working closely with dietitians, or with nurses around pain management, optimising medication and any ‘other elements of supporting patients’ that might become apparent. One participant described current efforts to find a way (perhaps using technology) to more systematically stratify the different aspects of patients’ physical, psychological and nutritional needs when they first meet a person, to ensure that every patient gets the right amount of support for each of their varying needs.

Well-being was often aligned with thinking about helping patients to feel supported, and this was often talked about in terms of making sure people ‘feel like they’re in a safe space’; one participant described feedback from patients that they had had ‘enough clinical appointments and don’t want this to be seen as clinical as well . . . so every effort is taken to create a more informal and relaxed service’. This was described as helping patients ‘open up’ much more than they might do with their clinical team.

Motivating and inspiring

Another common frame participants spoke about was driving ‘self-empowerment’ and motivating or inspiring people to make permanent changes; one participant explained that they focus on ‘self-efficacy for exercise because we know somebody with a higher self-efficacy for exercise are more likely to be independent once they’re discharged from our service’. Another talked about helping patients to understand ‘how to monitor their own body, [and] how to put together their own gym programme’. Attempts to motivate change sometimes included an element of education: group sessions, for example, focused around ‘physical activity and why it’s important and the kind of benefits of it’ or conversations, for example in which a patient’s anaesthetist might be talking to a patient about how their aerobic fitness is going to affect their care. This was seen as incentivising patients to concentrate on how they may be able to influence treatment outcomes and reduce side effects (e.g. of cytotoxic chemotherapy).

Several participants talked about providing information or guidance on how to find support and ‘the right exercise classes or activities’ available in patients’ local areas. This was described as ‘key for the long term’ and ‘massively important’ in making sure ‘they are able to motivate themselves, have the know-how and the confidence to keep active and address their activity needs’.

Several participants also commented on leading by example. One service, in working on becoming an active practice, described focusing first on getting their staff and GPs within the practice moving more. This included walking challenges and signing up to be a Park Run Practice (advertising local runs and a couple of staff going along). The participant felt that, on the whole, everyone now ‘feels more active’, more ‘willing’ and better placed to discuss PA more with their patients. Others too talked about role modelling, with one participant suggesting that it might be extremely helpful if activity was ‘something that is habitual’ and ‘health-care professionals themselves [are] physically active’.

Participants generally echoed a perception that their provision was ‘one area that [patients] do actually have control over . . . they can control what they come to and what they don’t come to . . . it’s a choice at the end of the day’. There was some understanding that ‘once you get into a hospital system, you don’t get a lot of control over many things . . .’ and that the programmes provided an opportunity to ‘actively participate in something’, regain an element of control over part of their lives among unfamiliar and unwelcome aspects of their illness, and that this feeling of choice can itself be enabling and ‘quite a kind of powerful thing for patients’.

Embedding physical activity in usual care

A further frame was that PA was effectively part of a patient’s treatment options, it was usual care or a prescription, and the ultimate aim of services was that their programmes were embedded in standard care. One participant explained that clinical teams at patients’ MDT appointments would ‘present a treatment plan’, and ‘this plan includes . . . “we’re going to send you to the exercise team” . . . it’s part of patients’ prescription . . . it’s just like them going to see the radiotherapist, they come and see us’. The message from participants was that PA must form an integral part of a patient’s treatment, and, whether implicit or overtly stated by those involved in patients’ care, having this reinforced was very helpful.