Study findings and key messages
To our knowledge, this is the first large-scale, multicentre, pragmatic, definitive trial to investigate the clinical effectiveness and cost-effectiveness of an early structured exercise programme for women at higher risk of shoulder problems after breast cancer surgery. We delivered the trial in an NHS setting, across 17 breast cancer centres within secondary care services. Treatments were delivered by physiotherapists independent of multidisciplinary oncology teams. We aimed to provide evidence of whether or not early postoperative physiotherapy, currently not offered to women undergoing non-reconstructive breast surgery, is a clinically effective and cost-effective use of NHS resources.
We found evidence that the exercise programme led to greater intermediate and longer-term benefits on upper limb function, postoperative pain, arm symptoms and QoL than information leaflets only. Exercises started within 10 days of surgery were not associated with increased risk of wound-related AEs at 6 weeks or chronic AEs, such as lymphoedema, over 1 year. Overall, our results suggest that physiotherapy-supported exercise is a cost-effective intervention for women undergoing invasive cancer treatments, particularly treatments targeting the axilla, which can increase the risk of shoulder problems.
Key findings from the trial are discussed below. We consider issues relating to recruitment, uptake and retention, and also potential threats to the internal and external validity of the study. We also consider the characteristics of trial participants, the implications of losses to follow-up, and intervention adherence and fidelity. We then discuss our findings in relation to the current literature. Finally, we consider the clinical implications of our findings on recommendations for breast cancer care.
Recruitment uptake
A total of 951 patients were screened in the clinical setting and, of these, 190 (20%) were deemed ineligible for the trial. The main reasons for ineligibility were not being at higher risk of shoulder problems and opting for immediate breast reconstruction surgery. Current UK guidance62 now recommends that all women undergoing mastectomy should be offered either immediate or delayed breast reconstruction and rates of immediate implant-based reconstruction surgery have doubled in England since 2005.138 Of those eligible to participate, a proportion of women were given study materials but had limited time to consider the information (43/761; 6%). Recruitment staff did not always record the specific reason for non-participation (77/761; 10%). Time from diagnostic testing of breast cancer to surgery can be rapid, within weeks, as per recommended NHS cancer referral and treatment guidelines.3 Evidence from the US139 suggests that fewer than 1 in 20 cancer patients approached to take part in clinical trials are actually enrolled, although a recent meta-analysis suggested uptake to cancer clinical trials was nearer 8%.140 Enrolment and uptake into prevention trials and trials testing exercise interventions can present additional challenges, such as the discounting of future perceived benefits,141 particularly when faced with a distressing diagnosis. Other known barriers to participation in cancer clinical trials include structural (e.g. access), clinical and attitudinal (physician and patient) factors.140
Of those eligible and approached, 270 out of 761 (35%) women declined to participate. Before starting PROSPER, we considered the findings of a single-centre RCT67 conducted in the NHS in England, in which two exercise programmes were compared after ANC surgery. The authors reported low recruitment, with 64% of 345 women invited declining participation.67 Timing of invitation was cited as a key issue, with women being distressed or more concerned with their immediate cancer treatment or pending surgery when approached preoperatively.67 These authors recommended approaching women to participate in research after breast cancer surgery.
For PROSPER, we incorporated pre-pilot interviews with patients and BCNs specifically to inform our recruitment procedures and development of trial materials. Based on the literature and interview findings, we opted to screen and recruit women preoperatively. Most breast cancer surgery is undertaken as short-stay or day-case surgery, as per the NHS 23-hour ambulatory care model; thus, postoperative recruitment would be challenging. Consequently, uptake to PROSPER was good, with 59% (392/662) of women actually approached by clinical/research staff about the study agreeing to take part. Our materials explained equipoise, but also that, if allocated to the exercise intervention, participants would then be supported by a trained physiotherapist with flexible appointments arranged around routine clinical follow-ups.
Participant retention
Although uptake was good, eight participants were randomised in error and two allocated to the exercise group declined when notified of their treatment allocation. In all cases, errors in randomisation or declining to participate occurred on the day of randomisation and no further data were collected. An additional 8% of randomised participants were lost because they did not return their baseline questionnaires, and thus the baseline DASH scores required for treatment group comparison over time could not be calculated. These losses mostly occurred during the internal pilot study, when procedures were being refined. We extended recruitment by 1 month to achieve the required sample size of 350. We analysed data according to the ITT principle of allocated treatment, irrespective of subsequent non-compliance, and without imputation. Imputation for data missingness did not change the strength or direction of our estimates. Unlike other surgical trials, we did not undertake a modified ITT142 (excluding the 10 randomisations in error and those withdrawn at point of randomisation) or a per-protocol analysis, although the compliance-based approach does provide a realistic estimate of those who complied with treatment.
Response rates to follow-up postal questionnaires were lower than predicted, increasing the risk of attrition bias. We predicted a 25% loss to follow-up at 1 year, based on other clinical trials investigating exercise interventions.67 However, we observed a slightly higher loss to follow-up, with 70% (274/392) of our randomised sample returning final 12-month questionnaires (although this equated to 78% of those returning baseline data). Despite losses, the trial was sufficiently powered, as the required sample size was calculated to be 256 participants at 12 months. Rates of withdrawal were comparable by treatment group, and reasons for withdrawing were mostly cited as treatment burden, including being informed of a cancer recurrence, or that women did not want to complete questionnaires during adjuvant treatment.
Risk screening criteria
Only women at high risk of developing shoulder problems after breast cancer treatment were eligible to take part in the study. We developed our own criteria, based on the published literature, to determine future risk of developing shoulder problems. Screening criteria were applied preoperatively, based on planned cancer treatment pathways. Over half of participants (59%) were booked for ANC surgery when recruited, although actual cancer treatments delivered after randomisation differed from preoperative treatment plans. A total of 327 out of 392 (83%) of women underwent ANC, thus we are confident that we recruited those who were, indeed, at higher risk of developing shoulder problems. One-third of women had both ANC and radiotherapy to the axilla/supraclavicular area, regardless of other entry criteria. A high proportion (75%) of women were screened as being overweight or obese, having a BMI of ≥ 25 kg/m2. Preoperative height and weight were recorded at preoperative clinics and used by clinical teams for risk screening. We also captured patient-reported height and weight from baseline questionnaires. Patient-reported BMI was marginally lower than clinically recorded data (72% overweight/obese). Cancer treatment covered a wide spectrum and different combinations of modalities, including repeated surgeries, adjuvant chemotherapy, radiotherapy and hormonal therapy. Although treatment modality and sociodemographic variables were well balanced across treatment groups at baseline, we noted differences between planned and actual cancer treatments delivered over the 12-month follow-up. For example, a higher proportion of women in the exercise group underwent mastectomy (44%) than those randomised to usual care (38%), and thus those in the exercise group were more likely to have more extensive breast surgery and less likely to have breast-conserving procedures with adjuvant therapy. We adjusted for cancer treatment in all analyses. We also noted differences in proportions of participants reporting a history of shoulder problems, although this was not reflected in baseline DASH scores. We did not stratify allocation by risk criteria; therefore, some random variation is expected. Nevertheless, we undertook sensitivity analyses to explore this further, but adjustment for self-reported shoulder problems did not change effect estimates.
Almost one-quarter of women (22%) underwent repeat surgery; this reoperation rate is similar to that reported across 156 NHS trusts for women having breast conserving procedures over a 3-year period (20%; 95% CI 19.6% to 20.3%), with higher repeat surgery rates (30%) noted after invasive cancer surgery.138 Within PROSPER, reoperations were mostly revisions of tumour resection margins undertaken as an additional day-case procedure. Five women opted for a delayed breast reconstruction. No adjustment was undertaken for revisional surgeries, as these were equally distributed by treatment group and all statistical analyses were adjusted based on the most invasive breast and axillary procedure over follow-up. For intervention participants already following the exercise programme, physiotherapists advised them to restart exercises from the 7th postoperative day after revisional surgery.
Participant characteristics
Our recruited sample had a mean age of 58 years and were, thus, representative of newly diagnosed breast cancer patients.1 Over one-quarter (28%) of women were < 50 years. Among our sample returning baseline questionnaires, participants were predominantly white (92%), although we had fair representation from the black and ethnic minority population (8%) relative to national English statistics (UK Census data: 86% white).143 As described, most women were overweight/obese and were relatively inactive in the week prior to recruitment. Three-quarters (76%) reported not taking part in any physical activities or sport, such as dancing, swimming, jogging, cycling or tennis. We are confident, therefore, that we did not recruit a sample of highly active women who exercised regularly. This may be important when considering generalisability and wider implementation of the intervention, as well as the longer-term potential health benefits from promoting physical activity. Other studies have shown that physical activity declines after a breast cancer diagnosis and during treatment.144 Women receiving adjuvant treatment are least likely to remain physically active; one large cohort study found that physical activity halved in those receiving both radiotherapy and chemotherapy.145
Despite being relatively inactive in the week before recruitment, our participant sample were confident that they would return to usual activities and/or regular physical activity in the future, after cancer treatment. Confidence scores were higher in the exercise group than in the usual-care group across all postoperative time points. Some argue that the time of diagnosis is the window of opportunity to identify and motivate sedentary patients with breast (and colon) cancer, to target motivational support and sustain lifestyle changes.146 Having confidence in ability to return to future activity is therefore an important consideration for encouraging behaviour change.
We found differences in rates of outdoor walking between groups over time, with higher rates of activity in the exercise group than in the usual-care group at 6 weeks and 6 months, but rates were broadly similar at 1 year. Light and moderate physical activity, as encouraged within the PROSPER exercise programme, has been associated with improved QoL.147 Interestingly, at the 1-year follow-up, the majority of women reported that the arm and shoulder exercises had helped their recovery, suggesting that participants in both treatment groups attributed at least some of their recovery to exercise. We can assume that those in the exercise programme undertook their prescribed exercises, but we cannot know what exercises those in the usual-care group were following. Although women in the usual-care group perceived their exercises as being helpful to their recovery, these did not have the same impact on their upper limb disability (as measured by the DASH scale) as the structured, supported programme.
Health-related quality of life
Our sample had lower QoL scores on recruitment than population normative scores for UK females aged 45–64 years [SF-12 PCS 49.1 (SD 10.6); MCS 51.4 (SD 9.8)].148 We observed lower scores in our sample for mental health at recruitment than at 6 and 12 months and also when compared with population normative scores for women < 45 years and > 64 years.148 These lower scores at baseline may reflect diagnosis-related distress and pending cancer surgery.149 QoL scores were lower at the 6-month follow-up than at 1 year, reflecting the impact of ongoing adjuvant treatment. Among postal responders, item missingness was low for SF-12 data at baseline (4%), 6 months (5%) and 12 months (6%). We followed validated scoring guidelines for all standardised measures. Comparison of complete-case and imputed data for EQ-5D-5L scores used in the cost-effectiveness analysis did not change estimates. At 12 months, the exercise group had significantly better physical QoL than women receiving usual care.
Exercise intervention: uptake and adherence
Usual NHS postoperative care for non-reconstructive breast surgery is written information about exercise, and referrals are made to physiotherapy only when a problem has been identified. We designed an exercise programme cognisant of busy physiotherapy clinics and what could reasonably be offered within the NHS. The exercise programme was planned to start after the first postoperative week and continue throughout cancer treatment. Uptake of the intervention was excellent, with 181 out of 196 (92%) of those randomised attending at least one appointment with their physiotherapist (95% of referred). Despite the known barriers to exercise during cancer treatment, adherence to the programme was good. Twenty per cent of participants either withdrew or were discharged after two appointments, and the remainder completed three or more physiotherapy sessions (143/191; 75%). The adherence rate was higher than those reported in other trials testing exercise intervention with patients with breast cancer.150,151 We recommended that women continue with shoulder-specific exercises and physical activity for up to 12 months, despite the fact that symptom burden, such as fatigue and other side effects caused by chemotherapy, may overwhelm the motivation to exercise.146
All participants were advised to restrict arm movements for the first postoperative week to avoid the risk of increasing wound drainage.12 There is debate around the optimal timing to start exercise in relation to cancer surgery.152 Our findings show that unrestricted ROM shoulder exercises started at 7–10 days postoperatively did not increase the risk of self-reported wound-related AEs at 6 weeks. Physiotherapists notified the study team of six adverse events (6/191; 3%), although four of the six women who experienced postoperative events continued with the exercise programme. We did not objectively measure postoperative wound drainage (volume or duration), as this would have required additional hospital clinic visits. For women requiring additional surgery after starting the exercise programme, physiotherapists advised that they restart the programme from the 7th postoperative day. The intervention was designed such that face-to-face appointments were strategically planned to allow controlled, supervised progression of exercise difficulty while dovetailing with routine oncological follow-up. We used the frequency, intensity, time and type (FITT) principle to prescribe and tailor the exercise programme. The programme was adaptable and flexible, and thus could be adjusted in accordance with individual needs, preferences and cancer treatments.
Comparison of findings with other studies
We found evidence of improved upper limb function, lower intensity acute pain, reduced chronic pain, fewer arm symptoms and improved QoL in the exercise group over 1 year. We found statistically significant differences between groups for the DASH score at 6 and 12 months; the mean difference at 12 months was greater than the estimate specified a priori as the MCID. We defined MCID as a 7-point difference and our primary and sensitivity analyses confirmed this difference. Other studies38 suggest that slightly larger differences are clinically meaningful but in relation to those with other chronic conditions receiving different interventions with shorter follow-up time periods. We specified 7 points to account for the pragmatic nature of the trial and the longer duration of follow-up of over 1 year. This was also to account for the prolonged and complex cancer treatment pathways after primary surgery. We are confident of some benefit from our programme, with CIs suggesting evidence of benefit in total DASH score, and also for impairment, activity limitations and participant restriction subscores, although CIs were wider for subscores.
The primary source of evidence for exercise after breast cancer treatment remains the Cochrane review,12 which is now out of date given that at least 10 RCTs have since been registered or completed, as per our literature review update. Our findings are similar to findings from some small trials comparing exercise programmes to either non-active or information-only control groups.14 For example, the small Dutch trial41 that we used for our sample size calculation compared a shoulder exercise programme with information leaflets on outcomes at 3 and 6 months in 30 women. The authors reported improved ROM and strength and lower pain scores in the exercise group, with a similar difference in DASH scores (unadjusted MD –9.0; p = 0.03; n = 30 participants) at 6 months after exercise41 to that observed in PROSPER (adjusted MD –8.74; p = 0.001; n = 235 CACE analysis).
Other trials have recruited patients to exercise postoperatively, either during or months after completion of adjuvant therapy.12 ROM exercises are important to maintain shoulder mobility and to avoid common problems related to axillary surgery, such as axillary web syndrome. We found evidence of progression in ROM exercise difficulty, but this was not captured in the work capacity outcome. Participants increased the difficulty of exercises over subsequent appointments but maintained the same number of repetitions and sets. We observed improvements in strength over time using our composite measure of work capacity (incorporating band resistance) among intervention compliers. Strength exercises are important to maintain muscle strength and mass during cancer treatments; chemotherapy can reduce upper limb strength by as much as 16%.80 A recent systematic review,153 based on several small trials with short-term follow-up, concluded that there was low-level evidence regarding the effectiveness of range of motion and muscle strength exercises to improve arm function. Our findings suggest that an early, structured, progressive exercise programme is beneficial for improving upper arm function in those who are at highest risk of shoulder problems and these findings are clinically relevant.
Intervention fidelity
Intervention fidelity is an important consideration for complex intervention trials. Multicentre trials delivering complex interventions are more vulnerable than studies testing simple interventions, being at greater risk of not being implemented as intended.154 We developed clear treatment pathways and protocols throughout the first year, thus developing and testing the exercise intervention, incorporating multidisciplinary expert and patient input to co-design a programme suitable for delivery within the NHS. Some intervention elements were novel for physiotherapy staff, for example behavioural change strategies and co-decision of exercise prescription to encourage adherence. Although participating physiotherapists were experts in the management of musculoskeletal conditions, none was formally embedded within oncology services. Each physiotherapist completed training and achieved the required competence before training certificates were signed. Quality of training delivered was assessed and adapted during the internal pilot study. We provided clear supportive materials (e.g. laminated prompt sheets for motivational interviewing, ROM, strength assessment etc.). We did not undertake video recordings of participant consultations, as this may have affected adherence, particularly given the sensitive nature of consultations. Qualitative interviews revealed that some women became emotional during appointments, particularly in the earlier postoperative assessments, when adjusting to their new body image. We are confident, through our quality assurance procedures, that physiotherapists adhered to and prescribed the recommended programme. We are also confident that our intervention is ‘futureproof’, given that women underwent contemporaneous cancer treatment for those requiring invasive axillary treatment.
Qualitative findings
We included the qualitative study to explore insights and perspectives from patients and health-care providers taking part in the trial. Women described feeling motivated to comply with the exercise programme because they felt that it contributed to their overall well-being and recovery. Women also reported that it was something that they could achieve themselves during cancer treatment, thus having control over their own body rather than passively receiving cancer treatments. Physiotherapists described the exercise programme as rewarding; they enjoyed providing support and encouragement to women and reassuring them that it was safe to move their arm after surgery. Other positives described by physiotherapists included the longer appointment sessions and emphasis on patient choice and joint decision-making. Several issues for future implementation within the NHS setting were identified: the need for a private space for appointments, access to emotional support for physiotherapists and integration of physiotherapy expertise within the multidisciplinary oncology team.
Strengths of the study
The strengths of the study included the methodological rigour, excellent uptake of exercise and good adherence to interventions over time. Clinical data confirmed that most of our recruited sample had axillary surgery or radiotherapy treatment; only 14% were recruited for other risk factors (obesity or shoulder problems only). Although we cannot fully eliminate the risk of selection bias, we believe that the risk of bias was low, as randomisation was stratified by site and we avoided the use of permuted blocks. A major strength was the effort invested in the production of a well-designed programme with high-quality intervention materials packaged as a deliverable, stand-alone treatment programme. The intervention was delivered by NHS staff with the aim of testing the programme in an everyday clinical setting. We carefully tracked treatment referrals and monitored compliance. We incorporated an embedded processes valuation with qualitative interviews to better understand the challenges and acceptability of our interventions from the perspectives of both patients and physiotherapists. For the health economic analyses, data were triangulated from multiple sources: self-report and clinical records and routine national statistics via HES data. We undertook sensitivity analyses and also explored the impact of imputation for missingness, but these analyses did not change our findings.
Limitations
The main limitation of the trial was the loss to follow-up over study duration, being 5% higher than predicted. Although disappointing, losses to follow-up were equally distributed by treatment allocation; therefore, we are confident that systematic error (attrition bias) from disproportionate losses did not occur. Drop-out mostly occurred between 6 weeks and 6 months during ongoing cancer treatments, with minimal loss thereafter (2%). We compared the characteristics of the randomised sample with those of participants remaining in the trial at 12 months and found no differences in marital status, comorbidity, body weight or cancer treatments. However, younger women were more likely to withdraw, possibly because of family or work commitments. We could postulate that younger women are possibly more physically active in general or less in need of a supportive exercise intervention throughout cancer treatment. However, evidence from many studies suggests that younger women are more likely to experience more problems with functional limitations,155 delayed return to work, declines in physical activity156 and chronic post-surgical pain.13,15 Nevertheless, strategies to encourage adherence to exercise during periods of intensive treatment, chemotherapy in particular, should be explored. Despite losses, the trial study was adequately powered to be definitive and we completed all prespecified statistical analyses. We made one amendment to the statistical analysis plan after protocol publication to adjust for baseline values. We observed consistency in findings across a range of patient-reported outcomes, all of which support rejection of the null hypothesis. Another limitation was that we used self-report indicators for the secondary outcome of lymphoedema, rather than objective perometry, although our primary focus was to assess patient-reported upper limb function as per the commissioned brief.
Cost-effectiveness findings
Exercise was found to be more cost-effective than usual care, with the exercise intervention having a 78% chance of being the more cost-effective option at the NICE-recommended cost-effectiveness threshold of £20,000 per QALY, and an 84% chance at £30,000 per QALY. These results were robust to a range of sensitivity analyses. The intervention itself was relatively cheap to implement, at an additional £129 per person, and was associated with lower health-care costs and improved HRQoL. As QoL utility scores were diverging at the 12-month follow-up, we conclude that these estimates are conservative as benefits may accrue beyond the end of the trial.
Patient and public involvement
We included lay members in our external committees and at all stages throughout the trial. We included PPI in the design stage and during early intervention development. A patient dissemination event was planned at the University of Warwick to feed back study findings to trial participants. This was postponed because of coronavirus but will be rearranged in due course.
