Multi-centre cluster randomised trial comparing a community group exercise programme with home based exercise with usual care for people aged 65 and over in primary care

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 06/36/04. The contractual start date was in June 2008. The draft report began editorial review in October 2013 and was accepted for publication in February 2014. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Dawn Skelton and Susie Dinan are directors for Later Life Training, who deliver FaME and OEP training to health and leisure professionals across the UK. The other authors declare that they have no competing interests.

Copyright statement

© Queen’s Printer and Controller of HMSO 2014. This work was produced by Iliffe et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Promoting physical activity

The NHS has attempted to address the problem of inactivity in a variety of ways, including exercise referral schemes in primary care (‘exercise on prescription’), which were provided by approximately 90% of primary care trusts (PCTs) in the 2000s and usually involved referring patients to local leisure centres. 18 Although exercise on prescription has been shown to be feasible and effective in vulnerable older people,19 there appear to be significant barriers to the uptake of exercise classes in leisure centres. For many older people, home exercise or group exercise in non-intimidating environments (e.g. community halls) may be more appealing, and result in higher uptake of exercise programmes and longer continuation of exercise. Peer activity mentors have also been shown to be effective in increasing uptake and adherence to exercise. 20^-23

There are currently two existing exercise programmes designed for use in community settings with people aged ≥ 65 years. The first is a home-based programme, the Otago Exercise Programme (OEP), and the second is a community-based group exercise programme, the Falls Management Exercise (FaME) programme.

The OEP24–30 and FaME programme31,32 are both designed for use in community settings, specifically for people aged ≥ 65 years, to reduce falls. FaME is based on the components of fitness and principles of programming for all older adults (i.e. warm-up, mobility, stretches, strength and balance, endurance and a cool down), while OEP includes brief warm-up and strength and balance exercises appropriate for the age group. Both programmes involve strength and balance training which is tailored to the individual’s ability and health status.

The OEP is a home-based exercise programme for older people which is effective in reducing falls and fall-related injuries, improving balance, strength and confidence in performing everyday activities without falling, and has been shown to be cost-effective for people aged ≥ 80 years. 24–30 It was designed to be delivered by physiotherapists, and nurses trained and supervised by physiotherapists. A 1-year evaluation of the OEP showed considerable improvements in outdoor activities (walking, shopping, gardening and other outside leisure activities) after 6 months (Professor A J Campbell, University of Otago, 2007, personal communication) with participants continuing to exercise after completing the programme. It also showed significant improvements in executive function after 6 months. 30 While the OEP has been evaluated in four controlled trials of older primary care patients in New Zealand and one RCT in Canada, it has not been tested in a primary care setting in the UK for its feasibility, impact, acceptability and cost-effectiveness.

The FaME programme is a group exercise programme which was developed and tested in a controlled trial in the UK,31 but not in a primary care population. It aims to improve balance33 and was designed to be delivered by qualified postural stability instructors (PSIs). 32 It has been shown to be effective in reducing falls, and injuries resulting from falls. 16,31 Good adherence was demonstrated with the FaME programme and nearly two-thirds of people participating in FaME continued in group exercise programmes for over 1 year after trial completion (Professor D A Skelton, Glasgow Caledonian University, 2007, personal communication). The FaME programme remains to be evaluated for its impact, acceptability and cost-effectiveness within primary care.

This trial aimed to fill the gaps in the current evidence base by evaluating the delivery, impact, acceptability and cost-effectiveness of a community-based exercise programme (FaME) and a home-based exercise programme (OEP) supported by similarly aged (peer) mentors (PMs), compared with usual care for primary care patients. The underlying assumption was that the exercises would produce sufficient subjective well-being and improved mobility to encourage continuation of higher levels of PA after the cessation of the intervention. Each exercise programme was compared with usual care for effectiveness in producing sustained change in PA. The two programmes would be compared for cost-effectiveness if both were effective in promoting sustained change in PA. Our primary hypotheses at the start of the study were (1) both exercise programmes would produce sustained changes in PA compared with usual care and (2) OEP would be more cost-effective than FaME.

Objectives

The primary objective of the ProAct65+ study was to determine the effect of two evidence-based exercise programmes designed for older people compared with usual care (i.e. with no special interventions to promote PA), on the achievement of recommended PA targets 12 months after cessation of intervention.

The secondary objectives were to:

1. determine the health benefits of the two programmes to participants starting at various levels of PA – particularly the effects on physical and psychological status, health status, health-related quality of life and quality-adjusted life-years (QALYs)

2. estimate the costs of the exercise interventions, and possible cost offsets, and to assess the cost-effectiveness of community group exercise, and home-supported exercise compared with usual care

3. determine the acceptability of the programmes, adherence rates, enabling factors and barriers to future implementation

4. compare the time course of responses by participants, in terms of exercising at the recommended levels, at 0, 6, 12, 18 and 24 months after cessation of the intervention, between those undergoing the exercise programmes and those receiving usual care

5. determine participants’ perceptions of the value of exercise and the predictors of continued exercise.

Design

The ProAct65+ study was based on a three-arm, parallel design, cluster-controlled trial comparing a community centre-based group exercise programme (FaME), with a home-based exercise programme and walking plan (OEP) and with usual care, and using minimisation for allocation at the level of general practice in two UK centres (London and Nottingham/Derby). We initially planned 2 years’ follow-up post intervention to determine the impact, acceptability and adherence to the programme, longer-term continuation of exercise and cost-effectiveness. The Consolidated Standards of Reporting Trials (CONSORT) diagram35 summarises the design (see Figure 7).

Participants and inclusion/exclusion criteria

Participants were patients aged ≥ 65 years registered with participating general practices who gave informed consent to participate.

Interventions

There were three arms to the trial:

1. home-based exercise programme and walking plan (OEP)

2. community centre-based group exercise programme (FaME)

3. usual care.

Outcome measures

The primary and secondary outcome measures were chosen to reflect the needs of participants (e.g. functional outcomes, falls, confidence, quality of life, participant costs), and of commissioners of exercise services in primary care and policy makers (e.g. PA, falls, NHS costs).

The primary outcome was the proportion of participants reaching the recommended PA target of at least 30 minutes of activity of moderate intensity on at least 5 days each week, measured using the Community Healthy Activities Model Program For Seniors scale (CHAMPS) questionnaires. Although measures were taken at 0, 6, 12, 18 and 24 months after the intervention, our primary analysis was of data collected at 12 months post intervention, as a previous study in New Zealand had suggested that this was the time when the effect of the intervention was maximal. 39

The secondary outcomes included:

1. the direct health benefits, i.e. functional and psychological status, the rate of falls (the major safety outcome measure), the number and nature of falls, and fear of falling

2. self-efficacy for exercise and physical self-perception (self-esteem relative to the physical domain), which includes measurement of perceived importance (the degree to which participants value their physical condition, body image and physical strength) to inform predictors of exercise adherence and continuation, and participants’ judgement of the value or importance of PA

3. health-related quality of life and QALYs40

4. the NHS and private (participant) costs of each exercise programme, and possible cost offsets, identified from a comparison of health and social service utilisation of participants in all groups during the study period.

Economic evaluation

An economic evaluation was conducted alongside the clinical trial. The predefined aims were to:

1. estimate the costs of the exercise interventions, from the NHS and participant perspectives

2. explore the impact of the exercise interventions on participants’ utilisation of health and social services during the 6-month intervention period, and for 12 months post intervention, to assess the extent to which the costs of the interventions are offset by savings elsewhere in the system

3. assess the cost-effectiveness of the interventions, compared with usual care (no exercise intervention), using QALYs as the main measure of effectiveness.

Data collection and analysis related to each aim are described separately, below.

Economic analysis

Standard techniques of economic appraisal were applied. 76 The main measure of cost-effectiveness was the mean difference in QALY scores at 12 months after the end of the intervention, after adjustment for baseline measures in an analysis of covariance (as described in the statistical analysis section). Quality-adjusted life-year scores were obtained by transforming SF-12 health-related quality-of-life scores into EQ-5D utility weights. Transformation of SF-12 version 1 can be conducted using a published algorithm,40 but as version 2 had been used in the study, an amended algorithm was obtained from the authors (Dr Oliver Rivero-Arias, Oxford University, 2013, personal communication). The prepublished protocol specified that, if statistically significant differences in mean-adjusted QALYs were found between groups at the primary end point, comparisons between the usual-care (no exercise) group and each type of exercise programme would be conducted, ICERs would be calculated, and a probabilistic sensitivity analysis undertaken.

Secondary cost-effectiveness analyses were conducted using the primary PA outcome [proportions in each group reaching the recommended PA target of at least 30 minutes of activity of moderate intensity on at least 5 days each week (150 minutes per week), measured using the CHAMPS and Phone-FITT questionnaires] at 12 months after the end of the intervention.

The planned economic evaluation was based on NHS intervention costs only. Service use costs, and those associated with falls, would be added to the analysis if significant differences in these variables were found between groups.

Informed consent

Written informed consent was obtained from all participants. The decision regarding participation in the study was entirely voluntary. The researcher emphasised to potential participants that consent regarding study participation could be withdrawn at any time without penalty and without affecting the quality or quantity of future medical care, or loss of benefits to which the participant was otherwise entitled. No trial-specific interventions were undertaken before informed consent had been obtained.

Ethics committee approval

Ethical approval was granted to the trial from Nottingham Research Ethics Committee 2 (application number 08/H0408/72). National Health Service Research & Development approval were granted by NHS Nottinghamshire County and Westminster, Brent, Harrow, Hounslow and Barnet & Enfield PCTs, and other relevant PCTs as practices were recruited to the study.

Management of the trial

A trial management committee made up of all co-applicants and research staff at each site met regularly, face to face or by teleconference, to review the trial’s progress. Patient and public involvement (PPI) in the study was ensured by involvement in the management group of two lay experts from Nottingham University’s PPI forum. A combined TSC and data management committee met twice yearly to review progress of the trial.

Summary

The ProAct65+ trial was a primary care-based exercise intervention for older people with wide inclusion criteria. The pragmatic trial design replicated the approach taken in successful primary care trials in New Zealand39 and differed from the majority of trials which focus on falls reduction in selected groups by having continuation of PA as its primary outcome. The problems that we anticipated were (1) biases in recruitment, with those already exercising at a relatively high level being more likely to volunteer for this trial; (2) limited retention of recruits to the study, which we hoped to minimise by relatively frequent, but brief, contact with participants after the end of the exercise programmes; (3) variation in ‘doses’ of exercise, which we hoped to avoid through our quality assurance processes; and (4) an increase in falls risk, as in previous studies,39 which we countered through training of staff, risk reduction and risk management programmes.

Because the trial documented the levels of activity of participants (which could then be compared with population norms), the number screened, the number who were ineligible and the number who refused, its findings are generalisable, and can contribute to policy on exercise promotion and falls prevention among older people. They are relevant to older people and to policy-makers working in health, social care and leisure arenas, health and social care commissioners and providers, leisure providers and charities and voluntary organisations working with older people.

Improving the recruitment of general practices and participants

The flow path of participant recruitment to the trial is shown in Figure 1. The trial initially aimed to recruit 30 practices (15 at each site) and 45 patients per practice over a period of 3 weeks, to achieve a sample size of 1200 participants aged ≥ 65 years. The proportion of those who expressed an interest varied between practices, from 8% to 19% in London and from 7% to 21% in Nottingham/Derby, with a mean of 13.4%. In order to achieve the recruitment target, the number of invitations to eligible patients was increased from 450 per practice to 600 to adjust for the lower than anticipated recruitment, and more practices were recruited.

FIGURE 1.

Flow chart of the recruitment and assessment process in the ProAct65+ trial.

Stratified random sampling was planned, whereby eligible patients would be stratified into age groups 65–74 years and ≥ 75 years. To simplify the tasks for the practices and to encourage their co-operation this stratified sampling approach was abandoned and patients were sampled from one list of patients aged ≥ 65 years.

Room availability in practices for baseline assessments was limited and it took up to 6 weeks in some practices to assess and recruit the target number of participants. The recruitment phase of the trial was 9 months longer than anticipated because of the need to recruit more practices at both sites and to allow more time at each practice to undertake recruitment. This extension of the recruitment period altered the time scale of the trial and potentially limited data collection for the 18- and 24-month follow-up. In total, 43 general practices and 1256 participants were finally recruited: 22 practices and 605 participants in London and 21 practices and 651 participants in Nottingham/Derby.

Adding an eligibility screen

Although there were multiple steps for screening eligible patients (including electronic and manual patient searches by the general practices), in the first practices recruited researchers encountered patients at the baseline assessment and consent stage who were ineligible because they met exclusion criteria, particularly falling fewer than three times in the previous 12 months or already exercising at the target level of five sessions of 30 minutes of moderate exercise per week. To limit the number of assessments of participants who would be found to be ineligible, researchers asked questions over the telephone about falls in the last year and current levels of exercise when arranging the baseline assessment appointment, and excluded those who met these criteria.

Peer mentor recruitment and training

Volunteer PMs were recruited to support the participants during the exercise programme. Recruitment was slow (Figure 2) and time-consuming. Despite intense efforts the number of PM who joined the trial did not reach the target. After 8 months of PMs recruitment, the age criterion for PMs was altered to allow the enrolment of adults aged ≥ 50 years. This led to an additional eight PMs being enrolled in London, but no more in Nottingham/Derby.

FIGURE 2.

Number of PMs trained.

Table 1 shows the length of time spent on recruiting PMs, numbers of individuals who expressed an interest in becoming a PM, numbers of individuals trained, the number who subsequently disengaged from the study, and the final number of PMs who volunteered and were allocated participants. There was a large difference in the number of people who expressed an interest in becoming a PM and those that were trained. Feedback from PMs suggests that disengagement was as a result of, in part, the length of time between training and beginning work. This period was long because of the time needed to obtain research management and governance approvals for the PMs, and because of the staggered recruitment and randomisation of the practices. Disengagement was also as a result of, in part, the distance PMs would need to travel to support participants.

Each PM in the trial mentored a mean of three participants (range 1–13) in London, and a mean of three participants (range 1–5) in Nottingham/Derby. Overall, both sites fell short of the target of four to five participants per PM. All participants, regardless of their PM support, received the initial exercise training session and a booklet with tailored exercise instructions. Not all participants received a PM because of the difficulties recruiting them. In London, 123 (53%) participants and in Nottingham/Derby, 21 (12%) participants had a PM.

Despite using the same recruitment methods, recruitment difficulties were greater in Nottingham/Derby, possibly because the trial was competing with existing PM PA programmes for older people in Nottingham/Derby. The TSC advised to keep the intervention true to usual practice in the NHS, i.e. one instruction session plus a manual of exercises. Therefore, where there were insufficient PMs for all participants, they were not supplemented by an alternative person and some participants had no PM support at all.

In another attempt to increase the number of PMs and encourage them to support more participants, the number of their supportive contacts with participants was reduced. Initially PMs were scheduled to visit participants in their home on four occasions and telephone them 12 times during the 24-week intervention. This was reduced to two visits and eight telephone calls. Over both sites, the number of home visits ranged from zero to five (mean 2) and the number of telephone call contacts ranged from 0 to 18 (mean 6). Modification of the number of contacts did not increase PM recruitment or their case load.

Quality control of the Falls Management Exercise programme

The FaME intervention was a weekly group-based exercise session, supplemented with additional home exercises (modified from the OEP) described in a booklet. Postural stability instructors were recruited to lead the classes. The trial aimed to recruit 12 PSIs per site. In London, 16 PSIs were recruited, with a total of seven working on the trial. As there were few qualified PSIs available to recruit in Nottingham/Derby, the trial recruited and trained physiotherapists and exercise professionals who were interested in becoming a PSI and working on the trial. Sixteen individuals embarked on the PSI training course (15 completed the training) and seven of them worked on the trial. Some PSIs were not employed on the trial because of their limited availability. Additionally the complex and lengthy process of completing research governance approvals resulted in losing some available PSIs. The recruitment target was reached with 32 PSIs recruited and trained over both sites. Of these, 14 (44%) delivered the intervention, enabling the intervention to be fully staffed.

In order to quality assure and standardise the FaME intervention, two quality assurance members of the trial oversaw the intervention delivery by attending four exercise sessions over the 24-week intervention period for each PSI in all of the FaME practices. The quality assurers went to the sessions individually, except the first two sessions when they attended together to standardise their method. Overall, 45 FaME classes in London and 38 in Nottingham/Derby were quality assured. Using a standard checklist (Figure 3), the quality assurers observed the PSI leading the exercise class and then gave them feedback and an action plan in order to improve intervention delivery, optimise participants’ ability to undertake progressively demanding exercises and standardise the exercise intervention as much as possible.

FIGURE 3.

Postural stability instructors quality assurance checklist.

Measuring falls, service use and physical activity

During the intervention self-completion diaries were posted to participants every 4 weeks. During the follow-up period, participants were posted self-completion diaries every 3 months, larger self-completion questionnaires every 6 months and telephoned for a short questionnaire every 6 months. See Table 2 for the schedule of questionnaires at different time points and Chapter 2 for full details of questionnaires.

Because non-monetary incentives are known to assist retention in trials,17 small incentives were sent to participants to encourage completion of postal questionnaires. With diary 6 and 12, participants received a ProAct65+ pen and with diary 8 and 10, they received a ProAct65+ cotton shopping bag. Participants were also sent an annual Christmas card and brief newsletters with each diary they received.

Research staff at both sites telephoned participants every 3 months to remind them to return questionnaires. Up to three contacts with participants were made to undertake each telephone interview. Some participants did not return self-completion diaries and/or questionnaires and some were not available for a telephone interview as a result of a variety of reasons, including being on holiday, at work, too busy, or forgetting or losing the questionnaires.

The self-completion diaries requested information on participants’ health and social service use, falls and current exercise levels. 79 Initially, it was planned for participants to receive monthly prospective diaries to complete throughout the full length of the trial. When participants said that they wished to withdraw from the trial because of the quantity and frequency of questionnaires they were offered the opportunity to remain in the trial but complete only the 6-monthly questionnaires and not receive further diaries. By doing this, the trial retained 52 participants in London and 28 in Nottingham/Derby (6% of total trial participants) who would otherwise have withdrawn from the trial. To further limit the number of participants who withdrew from the trial because of the burden of the questionnaires and diaries the frequency of the diaries sent during the 2-year follow-up phase was reduced from monthly to quarterly. The diaries sent during the follow-up phase required the participants to recall their service use and falls from the last 3 months and record a 1-week prospective snap-shot of their exercise activities.

Capturing adverse events

Adverse events were monitored throughout the trial to assess the trial’s safety and manage participant risks. This is especially important as exercise within this age group may be associated with an increased risk of falls. 39,80 The ProAct65+ trial used a risk management pathway for capturing, classifying and dealing with participant AEs (Figure 4), which initially categorised all occurrences as SAEs, AEs, adverse reactions (ARs) or adverse incidents (AIs). All data were logged and any SAEs were reported to the TSC. The original risk management pathway and the definitions of events, reactions and incidents are reported in the trial protocol. 34

FIGURE 4.

The ProAct65+ risk management pathway. CI, chief investigator.

A comparison of all events between trial sites was carried out towards the end of the trial’s intervention phase. There were noticeable differences in the numbers of ARs recorded between sites with London categorising 5%, and Nottingham/Derby categorising 16% of their total events as ARs. A cross-checking system was therefore implemented between sites in an attempt to standardise categorisation. All events from each site, except AIs, were checked by the other site. If the other site’s categorisation was different to the original categorisation, this was deemed a mismatch. Mismatches between sites were identified, and blinded forms then passed to the principal investigators who discussed and agreed a final categorisation. The initial calculation of mismatches was performed towards the end of the intervention phase, when there were 51 mismatches, giving a mismatch rate between sites of 19%.

The decision whether or not an event is ‘possibly related’ to the trial is open to subjective interpretation. Consequently, 45 of the 51 (88%) discrepancies in the categorisation of events recorded at each site were between AEs and ARs. The category ‘possible adverse reaction’ (possible AR) was therefore added. After the introduction of the possible AR category, the mismatch rate (prior to discussion between principal investigators) fell to 2.6%.

After advice from the TSC, the categorisation was further modified to enable unrelated SAEs to be distinguished from non-SAEs. The final categories applied to the trial’s events were, therefore, SAEs, unrelated SAEs, AEs, ARs, possible ARs and AIs (see Figure 4).

Recruitment of general practices

Forty-three practices were recruited to the trial, to ensure that the target study population could be reached (see Chapter 3 for details). The characteristics of practices that joined the trial are shown in Table 3.

Postural stability instructors

The target of recruiting 12 PSIs per site was achieved and FaME-arm classes were fully staffed. In London, 16 PSIs were recruited with a total of seven working on the trial, whilst in Nottingham/Derby 15 completed the training, and seven of them worked on the trial. The mean class size was less than planned, at five not nine. The quality assurance reviewers noted that PSIs largely achieved standardisation of the intervention, although they varied most in progression of the exercise programme, and needed reminding about collecting data for the trial.

Peer mentors

Thirty-eight PMs were recruited, trained and deployed in the trial, 31 in London and seven in the Nottingham/Derby practices (details of the recruitment processes can be found in Chapter 3). The planned and actual engagement by PMs with participants is shown in Table 4. Research staff carrying out quality assurance through discussions with PMs concluded that, as a whole, PMs made only a limited attempt to standardise the intervention (i.e. to implement the individualised plan given to the participant at the first encounter) and the participants’ progression was also limited, even though the PMs tailored their advice in other aspects of exercise. They returned trial paperwork (follow-up sheets detailing call and visit information, and time and travel log for the economic analysis) promptly.

Recruitment of participants

Steps that were taken to ensure recruitment to the trial are described in Chapter 3. Figure 5 shows the recruitment of 1256 participants to the study. In total, 20,507 patients were invited to participate (Nottingham/Derby, 10,738; London, 9769). Expressions of interest were received from 2752 (13%) (Nottingham/Derby, 1481; London, 1271) and 1256 (6% of those approached) consented (Nottingham/Derby, 651; London, 605).

FIGURE 5.

Recruitment of participants to the trial.

Baseline characteristics of the study population

The average age of participants was 73 years (range 65–94 years), with 84% of participants aged less than 80 years, and 62% of participants were female. Thirty-four languages were spoken (33 in London and 12 in Nottingham/Derby) and 14% of participants were non-white, with greater ethnic diversity among the London participants. A total of 44% of participants had completed some form of further education, as shown in Figure 6. On average, each participant had 1.7 comorbidities [range 0–7, standard deviation (SD) 1.4 comorbidities] and was taking 3.7 medications on repeat prescription (range 0–18, SD 3.7 medications).

FIGURE 6.

Educational attainment among trial participants. FE, further education.

Baseline characteristics of participants in the trial are compared with normative data in Table 5. Trial participants performed below normative levels on most scales, except for Phone-FITT, PASE, ConfBal and OPQoL, but similarly to normative values on the AFRIS questionnaire. The normative values for Phone-FITT apply to an older (mean age 81 years) male population, so the higher level of household PA and the lower level of recreational activity in the ProAct65+ population may reflect its lower median age and the predominance of female participants. The normative values for ConfBal were calculated from the published data, which were derived from a population attending day centres, so the better performance of trial participants is not surprising.

Retention of participants

Of the 1256 randomised study participants, 426 (33.9%) did not reach 12 months’ follow-up after the end of the intervention period. Of these, 69 were excluded by their GP, 12 died, three withdrew at an unknown time point, one withdrew before providing any baseline data and one withdrew but asked for their data to be destroyed. Overall, 340 participants were defined as lost to attrition. Almost half of these dropouts withdrew within the first 3 months of the intervention (49.7%). A total of 830 participants were retained in the trial at 12 months’ follow-up. Figure 7 summarises the pattern of attrition over time, for all arms and sites.

FIGURE 7.

The CONSORT diagram.

Illness events were common in this study population and 30% of those who dropped out cited illness (their own or others’) as their reason for discontinuing with the study. Disappointment at allocation and research burden (principally related to the number of questionnaires and diaries to complete) were responsible for at least 18% and 11% of dropouts, respectively.

Tables 6 and 7 describe participants’ characteristics by withdrawal status. Associations of baseline measures with later withdrawal were investigated using logistic regression, with robust standard errors calculated to allow for clustering of participants within practices.

TABLE 6 - Univariate associations between attrition and sociodemographic dataa

CI, confidence interval; IQR, interquartile range; NS-SEC, National Statistics Socio-Economic Classification; OR, odds ratio.

Data are n (%) [n missing], unless otherwise stated.

Current or previous job coded to the National Statistics Socio-economic Classification 2000.

Note

A total of 69 participants were withdrawn by their GP after allocation, two withdrew their data from the study and 15 had incomplete demographic data.

Characteristics	Retained	Withdrew	OR (95% CI)	p-value
Number of participants (n = 1170)	830 (70.9)	340 (29.1)
Sex (% female)	518 (62.4)	219 (64.4)	1.09 (0.90 to 1.32)	0.380
Age (years) median (IQR)	71 (68–76)	74 (69–79)	1.05 (1.03 to 1.08)	< 0.001
Group allocation
FaME	256 (30.8)	95 (27.9)	1.00	0.750
OEP	278 (33.5)	114 (33.5)	1.11 (0.77 to 1.58)
Usual care	296 (35.7)	131 (38.5)	1.19 (0.74 to 1.92)
BMI (kg/m2) mean (SD)	26.7 (4.9) [20]	27.2 (5.1) [26]	1.02 (0.99 to 1.05)	0.140
Number of comorbidities	[1]
0 comorbidities	157 (18.9)	46 (13.5)	1.00	0.007
1 or 2 comorbidities	404 (48.7)	157 (46.2)	1.33 (0.90 to1.96)
≥ 3 comorbidities	268 (32.3)	137 (40.3)	1.74 (1.22 to 2.49)
Number of current medications median (IQR)	3 (1–6) [5]	4 (2–6) [3]	1.06 (1.02 to 1.10)	0.004
English main language	738 (89.9) [9]	290 (86.3) [4]	0.71 (0.50 to 1.01)	0.060
White self-reported ethnicity	707 (87.2) [19]	291 (86.6) [4]	0.95 (0.67 to 1.34)	0.780
Living alone	285 (34.6) [5]	122 (35.9)	1.06 (0.78 to 1.44)	0.710
Children < 18 years living at home	8 (1.0) [2]	6 (1.8)	1.84 (0.68 to 4.98)	0.230
Living with dependent adults	40 (5.0) [35]	27 (8.2) [9]	1.68 (0.95 to 2.95)	0.073
Education	[13]	[3]
Primary/secondary school	436 (53.4)	206 (61.1)	1.00	0.030
College/university	381 (46.6)	131 (38.9)	0.73 (0.55 to 0.97)
Employed full- or part-time	69 (8.4) [8]	33 (9.7) [1]	1.18 (0.76 to 1.82)	0.470
NS-SEC job gradeb	[35]	[15]
1–2: managerial and professional occupations	354 (44.5)	125 (38.5)	1.00	0.047
3–4: intermediate occupations	229 (28.8)	84 (25.9)	1.04 (0.76 to 1.41)
5–7: routine and manual occupations	205 (25.8)	109 (33.5)	1.51 (1.08 to 2.11)
8–9: never worked and long-term unemployed	7 (0.9)	7 (2.2)	2.83 (1.01 to 7.95)
Annual household income ≥ £20,000	291 (40.1) [104]	110 (40.0) [65]	1.00 (0.75 to 1.32)	0.980
Smoking status	[2]
Non smokers	444 (53.6)	148 (43.5)	1.00	0.011
Ex-smokers	341 (41.2)	172 (50.6)	1.51 (1.15 to 1.99)
Current smokers	43 (5.2)	20 (5.9)	1.40 (0.70 to 2.76)

TABLE 7 - Univariate associations between attrition and risk factors for falling, exercise, psychosocial and functional measuresa

CI, confidence interval; LSNS-6, Lubben Social Network Scale-6; MCS, mental component summary; OR, odds ratio; PCS, physical component summary.

Data are n (%) [n missing], unless otherwise stated.

FRAT is a five-item tool used to assess falls risk. The presence of ≥ 3 risk factors indicates a higher risk of falling.

ConfBal scale: total score provided as a measure of confidence, range from confident ‘10’ to ‘30’ not confident.

The short FES-I dichotomised into low (score 7 to 10) or high (score ≥11) concern about falling.

CHAMPS is a 40-item scale measuring PA duration and frequency in older people.

SF-12 PCS scores range from 0 to 100, where a zero score indicates the lowest level of health measured by the scale and 100 indicates the highest level of health.

SF-12 MCS scores range from 0 to 100, where a zero score indicates the lowest level of health measured by the scale and 100 indicates the highest level of health.

LSNS-6 where a score of ≤ 11 indicates the participant is isolated.

MSPSS is a 12-item scale measuring perceived availability of support. Scores range from 12 (weak social support) to 84 (strong social support).

OPQoL range 33–165 (higher score implies a higher subjective quality of life).

OEE scores range from 1 to 5, with 1 indicative of low outcome expectations for exercise, and 5 strong outcome expectations for exercise.

Characteristics	Retained	Withdrew	OR (95% CI)	Unadjusted p-value
Number of participants (n=1170)	830 (70.9)	340 (29.1)
Use public transport easily	788 (95.6) [6]	312 (92.3) [2]	0.55 (0.30 to 1.01)	0.052
Use a walking aid	112 (13.5) [2]	51 (15.0) [1]	1.13 (0.78 to 1.64)	0.520
FRATb
History of any fall in the previous year	134 (16.2) [1]	67 (19.8) [1]	1.28 (0.94 to 1.75)	0.120
On ≥ 4 medications per day	361 (43.6) [1]	174 (51.3) [1]	1.37 (1.07 to1.75)	0.014
Diagnosis of stroke or Parkinson’s disease	14 (1.7) [1]	12 (3.5) [1]	2.14 (0.91 to 5.03)	0.082
Any self-reported problems with their balance	197 (23.9) [5]	83 (24.8) [5]	1.05 (0.80 to 1.38)	0.730
Unable to rise from a chair of knee height	25 (3.0) [3]	19 (5.6) [3]	1.92 (1.01 to 3.63)	0.046
ConfBal score median (IQR)c	10 (10.0–13.0) [81]	12 [10–15) [82]	1.08 (1.05 to 1.12)	<0.001
High concern about falling (measured by short FES-I)d	123 (16.3) [74]	64 (24.2) [75]	1.64 (1.15 to 2.34)	0.007
CHAMPS (minutes/week MVPA)e	[64]	[70]
0 minutes	163 (21.3)	99 (36.7)	1.00	<0.001
1–149 minutes	274 (35.8)	82 (30.4)	0.49 (0.35 to 0.70)
≥ 150 minutes	329 (43.0)	89 (33.0)	0.45 (0.32 to 0.62)
SF-12 PCS mean (SD)f	37.6 (6.2) [4]	36.3 (6.9) [2]	0.97 (0.95 to 0.99)	<0.001
SF-12 MCS mean (SD)g	48.7 (6.1) [3]	48.4 (6.7) [2]	0.99 (0.97 to1.02)	0.540
Socially isolated (based on LSNS-6)h	149 (20.0) [85]	72 (27.4) [77]	1.51 (1.06 to 2.15)	0.023
Perceived social support (MSPSS) median (IQR)i	70 (58.0–79.0) [130]	70 (55.0–78.0) [95]	0.99 (0.99 to1.00)	0.250
OPQoL mean (SD)j	131.0 (13.0) [169]	126.9 (13.8) [126]	0.98 (0.96 to 0.99)	<0.001
OEE-positive subscale median (IQR)k	3.9 (3.6–4.2) [107]	3.8 (3.4–4.2) [93]	0.78 (0.62 to 0.98)	0.035

Those participants who dropped out were significantly more likely to be older, have three or more comorbidities, have more medications, have a lower level of education, have worked or currently work in a routine or manual occupation, be an ex-smoker, be unable to rise from a chair of knee height, be less confident about their balance, have a high concern about falling, be inactive, perceive their physical health as poor, be at risk of social isolation, have a lower subjective quality of life, have lower outcome expectations for exercise, take longer than 13.5 seconds to complete the TUG test, have a reduced functional reach, score lower on the Romberg test and perform fewer sit to stands in 30 seconds.

Sensitivity analyses

Analysis of the log(CHAMPS score + 1) outcome was repeated (1) after excluding an outlying value for a participant in the OEP arm (see Figure 10), (2) after excluding an individual whose reported number of falls was extremely high post entry and (3) after excluding 18 participants who were considered ineligible by GPs after randomisation had taken place. In all three analyses, the results concerning estimates of intervention effects were essentially unaltered.

The ICCs for the primary outcome (CHAMPS minutes per week of moderate or greater intensity activity at the 12-month post-intervention follow-up) were for the untransformed outcome 0.000 (95% CI 0.000 to 0.032) and the logarithmic transformation 0.009 (95% CI 0.000 to 0.044).

Adherence analysis

Different definitions were applied to participants in the FaME and OEP arms. In each case, analysis comparing adherent and non-adherent participants was carried out. Comparisons of log_e(CHAMPS score + 1) were made, with multilevel modelling as described above.

Safety: adverse events in the trial

There was only one documented SAE that the chief investigator thought could have been because of the trial, and this was reported to the TSC’s chairperson. On further enquiry, this potential SAE was judged by the TSC chairperson to be unrelated to the trial.

Table 12 summarises the categorisation process for harms arising in the participant population that could potentially be attributed to the trial (see Chapter 3 for the categorisation algorithm).

Table 13 shows the numbers of events documented during the trial and Table 14 shows AEs, reactions and incidents by arm, during the intervention period and in the 12 months post intervention, per person-month.

Table 15 shows the clinical conditions categorised as AEs.

The number of falls recorded during the trial in diaries and followed up by telephone contact are shown in Table 17 in Chapter 6. The different types of ARs are shown in Appendix 3.

Other physical activity measures

The CHAMPS questionnaire allowed calculation of weekly calories expended (Table 16). No significant differences were observed whether the measure was analysed as its original value or after logarithmic transformation. The full-time profiles are shown in Appendix 4 (Table 42 and Figure 12). Other questionnaire batteries (PASE and Phone-FITT) measuring reported activity were also analysed. The PASE scores showed a small, but statistically significant, benefit for FaME compared with usual care (difference in means 11.2, 95% CI 0.2 to 20.2; p = 0.046), but no statistically significant benefit for OEP (difference in means 7.5, 95% CI –3.8 to 18.8; p = 0.20). No significant differences were observed according to Phone-FITT (reported through telephone interviews) [Table 17 and Appendix 4 (Table 42 and Figures 13 and 14)].

Falls and falls risk

The number of falls was analysed (1) during the intervention period and (2) in the 12 months following the intervention (see Table 18). One very frequent faller (> 100 falls reported during the intervention period) was excluded from analysis, since his rate of falling should have excluded him from the trial, even though his reported fall rate prior to baseline was within inclusion criteria. There was no statistically significant difference in the number of falls among the FaME, OEP and the control arms during the intervention period [adjusted incidence rate ratio (IRR) 0.91 (95% CI 0.54 to 1.52)] for FaME compared with usual care, and IRR 0.93 (95% CI 0.64 to 1.37) for OEP compared with usual care. In the 12 months post intervention there was a statistically significant reduction in falls in the FaME arm compared with the usual-care arm (IRR 0.74, 95% CI 0.55 to 0.99; p = 0.009) and a non-significant reduction in the OEP arm (IRR 0.76, 95% CI 0.53 to 1.09; p = 0.14).

During the intervention period, participants in the FaME arm reported 39 falls with no injury, 31 falls with a bruise or cut, 13 falls with muscle or ligament damage and one fall resulting in a broken bone. In the OEP arm, participants reported 59 falls with no injury, 45 falls with a bruise or cut, 19 falls with muscle or ligament damage, and two falls resulting in a broken bone. In the usual-care arm, participants reported 34 falls with no injury, 59 falls with a bruise or cut, 23 falls with muscle or ligament damag, and six falls resulting in a broken bone.

The FES-I index, which measures participants’ fear of falling, showed no significant differences according to intervention arm (Table 18 and Appendix 4).

Quality-of-life measures

No significant differences were apparent at 12 months for either component of the SF-12 (mental or physical), the EQ-5D scores or the OPQoL (Tables 19 and 20). Details of profiles over time are shown in Appendix 4 (Table 44 and Figure 16).

Balance confidence and social networks

Table 21 shows differences between arms on balance confidence and social network support. Significant improvements in balance confidence were seen in both intervention arms at 12 months post intervention. The mean difference for FaME compared with usual care was –0.529 (95% CI –0.998 to –0.061; p = 0.027) while the mean difference for OEP compared with usual care was –0.545 (95% CI –1.033 to –0.057; p = 0.029). No significant difference in either social network scale (MSPSS or LSNS) was observed when comparing FaME and OEP to the usual-care arm. Further information concerning changes over time is shown in Appendix 4 (Table 45 and Figure 17).

Other secondary outcomes (measures taken only at baseline and immediately post intervention)

Table 22 shows no evidence for effect of either intervention on the FRAT. However, clear benefits were seen on the OEE scale (Table 22). Those in the FaME arm whose expectations of exercise were positive at baseline had significantly increased expectations of exercise at follow-up compared with those in the usual-care arm. Those who were negative about exercise at baseline improved their expectations at follow-up, in both FaME and OEP arms, compared with usual care. Changes from baseline to post intervention are shown in Appendix 4 (Table 46). Results are shown for three functional measures in Table 23. No significant effects of intervention were observed. Table 24 shows outcomes for PASE, Phone-FITT and the mental and physical components of the SF-12 scale. Table 25 shows results for quality of life (OPQoL) and falls risk assessment (FRAT score). Table 26 shows results for the FRAT binary score (0 or ≥ 1). Table 27 shows the results from multilevel modelling of post intervention and 6- and 12-month post-intervention scores on EQ-5D.

Intervention costs – NHS perspective

Discussion

The reasons for the low PM contact with mentees in the OEP arm are not fully known. The lower PM input, or lack of PM input for a significant proportion of participants, may have impacted effectiveness to an unknown extent.

Although PMs were volunteers, a cost was applied to PM time, inputted using a replacement cost method (based on a clinical support worker). Use of the opportunity cost method would have required further information from PMs about the activities that they were not doing in order to carry out PM responsibilities. If PMs were retired, the opportunity cost of their time may have been less than the replacement cost used and the per-participant cost in the OEP arm would have been lower. Moreover, if volunteer PMs had obtained positive utility from their contribution, their cost should have been reduced accordingly. 75 Problems with recruiting PMs might have been mitigated if they had been remunerated and this may have also resulted in more active support of mentees.

Per-participant costs in delivery of the FaME programme would be inversely affected by group size, but group size might also affect effectiveness to an unknown extent. In the trial, the mean group size was similar in both sites at allocation (9 or 10 participants, as planned per protocol). However, poor attendance at some groups [mean group attendance rates were 50.5% (range 35–68%) in London and 56.9% (range 28–68%) in Nottingham] indicates that actual costs per attendee were higher and that a larger group size at the outset may be possible so that there is still a core group of active participants after drop-out. Attendance rates of participants assigned to the FaME programme ranged from 0% to 100% in both sites.

Service use

Cost-effectiveness analysis

The primary outcome for the economic evaluation was to be QALYs derived from transformation of SF-12, as described in the methods. The main analysis failed to find a significant difference between groups in this outcome, with or without imputation, and after adjusting for baseline values, cluster and other confounders. As a result, the economic analysis focused on cost-effectiveness using the primary clinical outcome, i.e. the proportion of people reaching or exceeding 150 minutes of MVPA per week at 12 months after the end of the intervention.

A significant effect in favour of FaME, compared with usual care, was identified by the main statistical analysis (see Chapter 5). Against 38% of patients receiving usual care who at least met the exercise target at 12 months post intervention, an OR of 1.78 was associated with FaME, meaning that 52% of participants assigned to that group would be meeting the target, an absolute increase of 14%. This benefit was achieved through NHS expenditure on delivering the FaME interventions in London and Nottingham of £268.75 and £218.43 per participant, respectively (mean £243.59).

A cohort of 100 people assigned to the FaME intervention would therefore incur a total cost to the NHS of £26,875 in London and £21,843 in Nottingham (average £24,359), compared with no cost for usual care. As the FaME programme, compared with usual care, results in 14% more people achieving or exceeding the 150-minute per week moderate or vigorous exercise target at the 12-month post-intervention end point, the cost per extra person exercising can be calculated as the total cost for 100 people divided by 14, i.e. £1919.64 in London and £1560.21 in Nottingham (mean £1739.93).

Discussion

These findings need to be interpreted with caution. The per-participant costs for FaME are based on those recorded in the trial and would be affected by class size, with smaller groups increasing the average costs. Class size (and instructor) might also affect compliance and outcomes, but the impact of these factors is not known. The difference in costs between sites is largely a reflection of the higher costs of facilities hired for group classes in London, but considerable variability was observed within both sites.

The calculations reflect the NHS perspective and do not take account of the expenditures incurred by individuals in travelling to the exercise class venues or other out-of pocket expenses associated with exercise. No allowances are made for offsets against the costs of the interventions because no differences were found between groups in the costs of primary care contacts (used as an indicator or general health effects of exercise) or injurious falls in the 18 months post recruitment. The lack of difference between groups in use of primary care utilisation is consistent with the finding of no difference in health-related quality of life between groups.

Data from participants returning diaries indicated that those in the FaME group reported fewer falls than those in the usual-care group in the 12 months post intervention, but this was not reflected in the GP data (which covered all participants in GP practices that permitted access for data gathering, i.e. all practices except for three in London). Owing to lack of a statistically significant difference in QALYs between groups, a probabilistic sensitivity analysis was not undertaken, so no interpretation of the findings against a cost/QALY gained benchmark is possible. The study recruited volunteers, some of whom were already achieving the 150 minutes of MVPA at baseline. Further analysis is needed to explore the differential impact of the interventions on sustaining exercise among those already at target, and encouraging people not exercising to start, since this may impact on the cost-effectiveness ratios.

What this study shows

Exercise classes using the FaME programme significantly increased PA in older people. The proportions reporting at least 150 minutes of MVPA per week rose from 40% at baseline to 49% at 12 months post intervention in the FaME arm, from 41% to 43% in the OEP arm and from 37.5% to 38.0% in the usual-care arm. The odds of reporting at least 150 minutes of MVPA were 78% higher in the FaME arm than in the usual-care arm, equating to an absolute increase of 14% in the number of participants reaching or exceeding the PA target. In terms of minutes of MVPA, the FaME arm reported an additional 13–15 minutes of MVPA per day (91–105 minutes per week) compared with the usual-care arm, depending on the imputational model used. There was no statistically significant increase in MVPA in the OEP arm compared with the usual-care arm.

In the 12 months post intervention there was a statistically significant reduction in the rate of falls in the FaME arm compared with usual care (IRR 0.74, 95% CI 0.55 to 0.99; p = 0.042). Although the falls rate was lower in the OEP arm than in the usual-care arm, there was no statistically significant difference between these two arms. The PASE scores showed a small, but statistically significant, benefit for FaME compared with usual care (difference in means 11.2, 95% CI 0.2 to 20.2; p = 0.046), but no statistically significant benefit for OEP (difference in means 7.5, 95% CI –3.8 to 18.8; p = 0.20). Significant improvements were seen in balance confidence for both intervention arms at 12 months post intervention. The mean difference for FaME compared with usual care was –0.529 (95% CI –0.998 to –0.061; p = 0.027), while the mean difference for OEP compared with usual care was –0.545 (95% CI –1.033 to –0.057; p = 0.029). Participants in the FaME and OEP arms were significantly less likely to dismiss exercise as not beneficial, and in the FaME arm were more likely to be positive about exercise, 12 months after the end of the interventions.

There were no statistically significant differences between intervention arms and the usual-care arm in self-efficacy, mental and physical well-being, quality of life, social networks, falls risk or functional abilities. The lack of change in quality of life is perhaps not surprising, given the high baseline level of OPQoL scores and the limited likelihood that an extra 15 minutes of PA in relatively active people would change perceptions of quality of life. The interventions were not associated with increased risk of AEs or ARs, during or, after the intervention period.

FaME is more expensive than OEP delivered with PMs (£269 vs. £88 per participant in London; £218 vs. £117 per participant in Nottingham), because of more direct participant contact from PSIs and hire of halls for the exercise classes. The cost per additional person meeting the target of 150 minutes MVPA per week at 12 months post intervention in FaME, compared with usual care, is £1920 in London and £1560 in Nottingham.

There are a number of methodological lessons from this trial. We have demonstrated that it is possible to recruit older people who would benefit from increasing their PA (as shown by their performance on a range of functional and psychological measures) to exercise promotion trials in general practice. As we outlined in Chapter 3, organisational factors in practices (such as room availability and space to carry out functional assessments) mean that planned recruitment rates may overestimate the speed of recruitment. Given that participation in an exercise trial attracts some who are already physically active at or above the recommended target level, telephone prescreening is useful to minimise the conduct of baseline assessments on individuals who are subsequently found to be ineligible. Quality assurance of interventions is necessary to optimise the fidelity of their application. The quality assessment process developed for the FaME intervention proved workable and may be of use in other similar studies. Recruitment of PMs was difficult despite relaxing the eligibility rules and the OEP arm was disadvantaged by this. Finally, frequent request of participants to complete exercise diaries challenged retention, and we lessened this research burden by reducing the frequency of requests. We will return to these methodological issues later in this chapter.

As we described in Chapter 4, those recruited to the trial were more active than their peers (with a median weekly MVPA level at baseline of 105 minutes), but on other characteristics functioned below population norms. Although not inactive, in other respects the participant population was an ideal one for testing PA interventions, as noted above. Retention of trial participants in the study remained problematic, despite the efforts made to increase it described in Chapter 3. However, this may in part be an unavoidable consequence of targeting an older age group, as illness events are common in the older population; 30% of those who dropped out cited illness as their reason for discontinuing with the study. Disappointment at allocation and research burden (the number of questionnaires and diaries to complete – see above and Chapter 3) were responsible for at least 18% and 11% of dropouts, respectively. As Chapter 4 shows, those who dropped out were older and more disabled than those who remained in the study. Those lost to follow-up were the subgroup which would probably have benefited most from increasing PA and may have been those least likely to increase their MVPA as a result of the intervention. This suggests that post-intervention levels of MVPA may overestimate activity levels that could be achieved in the general population of older people with delivery of FaME and OEP programmes. However, as losses to follow-up were similar across treatment arms, it is unlikely that this will have biased our estimates of the difference in MVPA between treatment arms.

Comparison with other studies

Our systematic review of the effectiveness of PA interventions for adults aged ≥ 50 years delivered through general practice84 identified six studies published between 1998 and 2011, with a total of 1522 participants. Four interventions were delivered by GPs or nurses and exercise specialists. 85–88 Three used only exercise specialists86,89 or an exercise counsellor. 90 Two used specific PA measures, such as the PASE and the Auckland Heart Exercise Questionnaire. 89,90 Four studies had 12 months’ follow-up. 86–89 Only two of these studies reported a statistically significant increase in PA levels. Kolt et al. 90 report that moderate-leisure PA increased by 86.8 minutes per week in the intervention participants compared with controls (p = 0.007). More intervention participants than controls reached 2.5 hours per week of moderate/vigorous leisure physical activity at 12 months (42% vs. 23%; OR 2.9, 95% CI 1.33 to 6.32; p = 0.007). Halbert et al. 86 reported that PA increased in both groups (p < 0.05), but more participants in the intervention group than in the control group increased their intention to do PA (p < 0.001). The increase was greater in the intervention than in the control group for all measures, except time spent walking. Two studies showed no significant increase in activity. 85,88

The ProAct65+ trial almost doubles the number of participants in such studies and shows the impact of interventions using standard scales for assessing PA for up to 2 years post randomisation. In contrast to the methodological variability of the other six studies, the ProAct65+ trial reported the method for generating the randomisation sequence, concealment of allocation, blind assessment of outcomes, an intention-to-treat analysis controlled for confounding variables and differences between treatment groups at baseline. In the six other studies, all interventions left participants to motivate and organise their own PA and the quantity of PA undertaken was not monitored, making it difficult to know whether or not the dose of the intervention affected the results. The effectiveness of the FaME arm in increasing self-reported PA may reflect the direction and encouragement provided by PSIs to participants.

Strengths and limitations of the study

To the best of our knowledge this study is the largest general practice-based trial of exercise interventions for older people in the UK, to date, and the first to deploy PMs to augment an exercise programme. Both exercise interventions were evidence based, but also pragmatic (i.e. feasible to use in general practice), and tailored to individual participants’ capabilities.

The ProAct65+ trial largely fulfils the RE-AIM criteria for evaluation of the public health impact of health promotion studies, using five dimensions:91

1. Reach (the proportion of the target population reached and the characteristics of participants compared with the target population). The ProAct65+ trial recruited a large number of people aged ≥ 65 years, whose performance on most of the measures used fell below population norms, despite their relatively high level of PA at baseline. The trial attracted participants who would be likely to benefit from increasing their PA level.

2. Efficacy (how the intervention benefited the participants). Participants in the FaME arm reported increased physical activity (on two measures), had a lower risk of falls and improved balance confidence, as well as becoming more positive about the beneficial effects of exercise. These findings are consistent with the conclusions of the Cochrane review of falls prevention. 5

3. Adoption (engagement of the settings participating in the study). Recruitment through general practice was feasible and 70% of participants remained in the study for 1 year after the intervention period.

4. Implementation (the extent to which the intervention was delivered as intended; including the adherence to the intervention, and the involvement of staff in the setting). Adherence to the intervention was easier to maintain in the FaME arm than in the OEP arm. As reported in Chapter 5, we were unable to show any difference in outcome in either intervention arm attributable to adherence.

5. Maintenance (long-term maintenance of behaviour change, defined as ≥ 2 years). We have reported findings at 12 months post intervention, our predetermined analysis point, but have collected data for up to 24 months post intervention. Significant increases in self-reported MVPA were found in the FaME arm of this study at 12 months post intervention (18 months after allocation); as Chapter 5 shows, this increase persisted, although slightly attenuated, at 18 months post intervention (2 years post allocation). Further studies are needed to measure attenuation of effects and to test the impact of reinforcement of the intervention.

Because of the difficulties of recruiting sufficient PMs we were unable to ensure a consistent dose of peer mentoring, which means that we have not measured the true impact of the OEP intervention.

The trial was reliant on self-report of PA, which is criticised for overestimating actual levels of activity. 92 However, this is less of a limitation than some suggest, for several reasons.

First, associations between self-reported PA and health outcomes93 are the basis of guidelines on 150 minutes of MVPA94 and, therefore, self-report of PA is also an appropriate measure of change in behaviour. 95 Using objective measures to assess compliance with guidelines that are based on evidence from self-reported activity could give an inaccurate picture of the proportion of the population that is insufficiently active. 94

Second, self-reported engagement in activities predicts both self-reported and measured functional ability 3–5 years later96 and all-cause mortality in middle-aged men 21 years later. 97 Self-reported PA scales can have acceptable validity98 and a single question can reflect a physiological measure like VO_2max. 99

Third, it appears that social desirability may influence self-reporting of PA, but this bias may also be determined by the type of questions asked100 and the characteristics of respondents. For example, misperception of activity level in one study was associated with older age, female sex, poorer walking performance, lower social support and lower self-efficacy,101 while another found that the difference between self-report and objectively measured MVPA was greatest among older men with lower educational level, at higher activity and intensity levels. 101 In the ProAct65+ trial the treatment arms were well balanced, and factors known to be associated with reporting PA were similar across treatment arms, suggesting social desirability to report PA might be expected to be similar across treatment arms. In addition, although those who were less active at baseline were more likely to withdraw, attrition did not vary significantly between treatment arms, again, suggesting this should not have resulted in differential reporting of PA between treatment arms.

Finally, although self-reported PA may overestimate objectively measured PA, this finding has not been consistent across studies. Some of the discrepancies between self-reported and objectively measured PA are because subjective and objective measures of PA are actually measuring different aspects of activity that are independently associated with biomarkers. 102 For example, PA monitors cannot accurately assess upper-body activities or account for movements that require extra effort, such as walking uphill or carrying loads. 103 Overall, it is not yet possible to draw definitive conclusions about the validity of self-reporting of PA compared with objective measurement,104 and this is an important research topic needing further investigation.

Lessons learned

The challenges faced during the ProAct65+ trial and solutions to these challenges are summarised in Chapter 3. Other research which has faced similar challenges is discussed here, and implications for future research and public health practice are suggested.

Although the trial exceeded its recruitment target, the recruitment process was more difficult and slower than anticipated. The time needed to recruit participants was underestimated and an extension in recruitment time was needed. Other trials recruiting from general practice have found similar slow and difficult recruitment, with lower than anticipated numbers recruited and required time extensions. 105–107 In ProAct65+, the recruitment phase was extended, more general practices were recruited and more patients at each practice were invited to participate to achieve the target numbers. We learned that it is advisable to keep recruitment as straight forward as possible and to minimise the work demanded of general practices. 108

Expressions of interest were received from patients already exercising at the target level of 150 minutes of moderate activity per week, and from frequent fallers. Others have reported that exercise trials can attract the more active part of the population. 109 Telephone prescreening was introduced to exclude such patients before they reached the baseline assessment appointment, but further studies are needed of ways to recruit the less active population.

The use of volunteers to act as PMs proved complicated. As others have found with interventions using volunteers, recruitment can be slow and the numbers deployed may be low. 110 Our PMs had a case load lower than we planned and, because we also had fewer PMs than intended, some participants received little or no PM support. Other studies have also encountered these problems. 111–113 The lower age limit for PMs was reduced, as was the frequency of their contacts with participants, but with only limited benefit to PM recruitment. It will be important for future interventions testing peer mentoring to allow enough time and resources (human and financial) when planning recruitment and training programmes. In order to minimise the time from training a PM to deployment, and to retain interested volunteers, attention needs to be focused on speeding up the process of gaining Criminal Records Bureau checks and Research Management and Governance approvals. Strategies to optimise PM motivation and involvement need further investigation.

In addition, the number of supportive contacts between PMs and participants varied and often differed from the number of contacts advised by the research team, which may reflect the needs of the individual participants. Future projects implementing PM support should be aware of participants’ needs for more or less support, which may lead to varied numbers of contacts with PMs. Overall, our experience of recruiting and retaining PMs within a trial raises questions about the feasibility of doing this in routine provision of exercise programmes in the community. Community-based exercise programmes proposing to use PMs should explore the feasibility of this prior to embarking on the programme.

Failure to ensure the fidelity of interventions is an important source of variation affecting the credibility and utility of research. 114 Quality assurance observation visits to classes were carried out by expert instructors, with verbal and written feedback on performance. Exercise instructors may not always achieve a balance between tailoring exercise and providing a standardised programme, and observations of intervention delivery are recommended. 115

Participants can be burdened by frequent data collection, which can impact on response rates to self-completion questionnaires and falls diaries. Response biases may occur, and we found that those with lower educational attainment and those whose first language is not English were less likely to complete falls diaries. 78 Compromises in the frequency of data collection were made, and the frequency of the self-completion questionnaires and diaries was reduced. However, maintaining between-assessment contacts is important to reduce attrition. 116 Personal contact with the research team improves response rates,117 as do reminders, incentives and printed educational materials. 118,119 Home visits to collect follow-up data are useful and can reduce attrition bias in longitudinal studies. 120 Alternatively, higher response rates to postal questionnaires have been found when they are sent by the general practices rather than by the research team; this may also be a method to aid retention of participants during a trial. 121

The classification of safety events between sites was variable, so a method of cross-checking and standardisation was developed. Both site principal investigators reviewed and discussed discrepancies in categorisation and a new possible AR category was introduced to reduce variability. This method of cross-checking and the classifications of safety events used in ProAct65+ could be applied to future exercise or indeed any multisite trials.

The ProAct65+ trial was a large pragmatic RCT, which, despite difficulties, reached its recruitment target, making it the largest exercise trial in UK general practice to date. The research team’s flexibility in being able to adapt to unexpected problems may have led to the successful implementation of the trial. 106 The lessons learnt during the ProAct65+ trial have been valuable and have potential implications for similar trials in general practice.

Conclusions

Our first hypothesis, that both exercise interventions would increase self-reported PA, has been refuted in this study, as has the second, that the OEP intervention would be more cost-effective. The FaME intervention increased self-reported PA, adding almost 15 minutes per day of MVPA. This effect persisted for 12 months after cessation of classes. The cost of getting one person to achieve or exceed the target level of PA was between £1560 (Nottingham) and £1920 (London). The OEP arm participants did not show any statistically significant increase in self-reported PA 12 months post intervention. This may be as a result of the limited support from PMs experienced by many participants in the OEP arm, and needs further investigation.

Contributions of authors

Steve Iliffe conceived and designed the study, submitted it for funding, was chief investigator for the study and drafted this report.

Denise Kendrick conceived and designed the study, submitted it for funding, was the principal investigator for the study in Nottingham and Derby and helped draft this report.

Richard Morris was senior statistician for the trial, led the analyses and helped draft this report.

Tahir Masud conceived and designed the study, submitted it for funding, supported implementation of the study, and contributed to this report.

Heather Gage supported implementation of the study, led the economic analysis and contributed to this report.

Dawn Skelton conceived and designed the study, guided the development of the exercise interventions, and contributed to this report.

Susie Dinan guided the development of the exercise intervention and contributed to this report.

Ann Bowling supported implementation of the study, led on quality-of-life measurement and contributed to this report.

Mark Griffin was trial statistician, supported implementation of the study and contributed to this report.

Deborah Haworth was trial manager, drove implementation of the trial at both sites and contributed to this report.

Glen Swanwick was PPI representative on the management board, supported implementation of the study and contributed to this report.

Hannah Carpenter co-ordinated the Nottingham research team, supported implementation of the study and contributed to this report.

Arun Kumar supported implementation of the study, led the analysis of fear of falling and contributed to this report.

Zoe Stevens was the trial administrator, supported implementation of the study and contributed to this report.

Sheena Gawler was research associate in the London research team, supported implementation of the study and contributed to this report.

Cate Barlow was research associate in the London research team, supported implementation of the study and contributed to this report.

Juliette Cook was research associate in the Nottingham research team, supported implementation of the study and contributed to this report.

Carolyn Belcher was research associate in the Nottingham research team, supported implementation of the study and contributed to this report.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.

Publications

Perry L, Kendrick D, Morris R, Dinan S, Masud T, Skelton D, et al. Completion and return of fall diaries varies with participants' level of education, first language, and baseline fall risk. J Gerontol A Biol Sci Med Sci 2012;67:210–14. doi: 10.1093/gerona/glr175

Iliffe S, Kendrick D, Morris R, Skelton D, Gage H, Dinan S, et al. Multicentre cluster randomised trial comparing a community group exercise programme with home based exercise with usual care for people aged 65 and over in primary care: protocol of the ProAct 65+ trial. Trials 2010;11:6. doi: 10.1186/1745-6215-11-6

Stevens Z, Barlow C, Kendrick D, Masud T, Skelton D, Dinan-Young S, et al. Effectiveness of general practice-based exercise promotion for older adults: a systematic review. Prim Health Care Res Dev 2013:15;190–201. doi: 10.1017/S1463423613000017

Kumar A, Carpenter H, Morris R, Iliffe S, Kendrick D. Which factors are associated with fear of falling in community dwelling older people? Age Ageing 2014;43:76–84. doi: 10.1093/ageing/aft154

Stevens Z, Carpenter H, Gawler S, Belcher C, Haworth D, Kendrick D, et al. Lessons learnt during a complex, multi-centre cluster randomised controlled trial: the ProAct65+ trial. Trials 2013:14;192. doi: 10.1186/1745-6215-14-192