Summary of findings
Forty-seven participants were recruited (78% target), thirty-seven were retained at three months and 29 of these provided at least four days accelerometry data (64% target). The number receiving the Booster interventions per protocol surpassed the feasibility target and data error rates were within acceptable limits, with the exception of the EXERT questionnaire.
From the data recorded, the standard deviation is estimated to be 103,394 PAC per week or 305 kcal per day for TEE and the sample size has been recalculated. With 600 randomised participants and 30% loss to follow-up the study will have approximately 90% power to detect a mean difference of 102 kcal per day in TEE between the groups as statistically significant at the 5% two-sided level. The observed attrition rate in the pilot study was, however, higher than this at 38%. With an initial sample size of 600 and assuming an attrition rate of 40%, and hence 360 participants with valid accelerometry data at three months then the study would have 85% power to detect a mean difference of 102 kcal in TEE between the groups.
Improving recruitment and retention
The fact that 82 out of the 104 DVD recipients who were screened after three months had increased their physical activity and were eligible for the study is promising. But, as only 47 were consented and randomised, strategies to improve recruitment are clearly required. The initial invitation to use the brief intervention (DVD) is being extended to all 40-64 year olds resident in EPHP areas giving a pool of 53,000 rather than only 30,000 residents (as originally intended) from which to draw (Figure 1 Box A). Extrapolating from the pilot data and with other things being equal, we would now expect the proposed mailout to 30,000 residents to result in 420 (95% CI: 380 to 570) participants randomised to the main trial (Figure 1 Box M), somewhat short of the proposed 600. With a mailout of 50,000, however, we would expect to be able to recruit 700 (95% CI: 550 to 950) participants. We have also increased the intensity with which the study is promoted. In the pilot area, the study was implemented without NHS or community-based support, awareness raising or publicity. Knowledge of the brief intervention's availability was communicated purely by one-to-one postal contact. In future areas, the invitation letters are being supported by local awareness raising involving EPHP area leads and other NHS links, community activists and wider media promotion.
A large minority (30%) of those who responded to the initial invitation and were contactable (Figure 1 Box F) were ineligible for the brief intervention almost exclusively because they were already too active. While it is too late to alter the recruitment materials for this study, any future trial attempting to identify people with sedentary lifestyles might consider incorporating a single item measure of current physical activity into the reply card, to avoid unnecessary telephone screening.
Further down the CONSORT diagram, many of those who were sent the brief intervention were not contactable for pre-trial screening after three months (Figure 1 Box H). To address this, during initial screening preferred method of contact is now confirmed and SMS messaging or e-mail offered, as well as post and telephone.
The refusal of consent by 43% of those eligible to be randomised is unsurprising considering the recruitment yield of prevention trials generally, e.g. Spilker and Kramer cite typical recruitment yields of only 1% to 6% . Putting recruitment figures in context is not easy as many trial reports do not record how many people were invited to participate, only how many were screened and randomised. For instance, the EXERT trial, which evaluated GP referral for different exercise programmes, report randomising 949 out of 1105 people who contacted the team, but did not record the number of patients who were advised by their GPs to take part in the programmes, but did not wish to do so . Korde eventually randomised 29 out of 352 women (8.2%) at risk of breast cancer who initially expressed interest in the study; it is not known how many women the team initially contacted . Similarly, the BRUM-CHF trial evaluating home-based exercise rehabilitation for patients with congestive heart failure screened 1639 patients, invited 642 (39.2%) to participate and randomised 169 (10.3%) . The SMART trial, evaluating different modes of self-monitoring on short- and long-term weight loss, finally randomised 210/704 (29.8%) originally screened for eligibility by phone but again there is no information about the numbers contacted in the original mailout . The numbers in these studies further suggests that converting eligible to randomised participants is unlikely to be easy. There is little difference between the recruitment rates of individual research assistants working on Booster, so training is unlikely to be an issue. Where reasons have been given for non-consent, the most frequently cited is lack of interest in further support to stay active.
Of the 47 participants randomised into the trial, ten did not complete three month follow up assessments. Eight of these participants had withdrawn prior to the three month assessment, with the most common reason cited as being too busy to continue (n = 5). Strategies to minimise numbers withdrawing are needed. During the feasibility phase, both intervention and assessment sessions were being delivered at venues local to the participant's residence with appointments scheduled at the participant's convenience, including in the evening and at weekends. During the main trial, however, sessions will also be offered at venues which are close to large employment hubs. It is hoped that this wider choice of venue will make the sessions more readily accessible to participants. Although it is not possible to abbreviate the intervention, it may be possible to shorten the assessment visits for those participants who express concerns about time constraints, e.g. by mailing questionnaires out in advance, to be returned at a shortened assessment visit.
Three Actiheart devices were returned with no usable data Two of these cases were due to set-up error resulting in the device not recording. This is a training issue and all research assistants will be given refresher training on programming the devices prior to the main trial to minimise the risk of similar set-up errors occurring in future. The remaining device with no data was due to removal soon after fitting due to skin irritation. This is an acknowledged issue in research using Actiheart devices: during fitting the participant's skin is cleaned and abraded to remove the top layer of skin prior to the application of the two self-adhesive ECG pads which the Actiheart device attaches to. The process of cleaning and abrasion combined with the presence of the ECG pads can cause skin irritation in some participants. Previous studies using Actiheart devices have reported skin irritation rates in up to 12% participants for protocols requiring seven days continuous wear . Five devices had less than four full days data; there was no specific reasons were identifiable for this. As some initial data was recorded in all cases and the Actihearts all recorded for the full seven days, it is likely that participants removed the monitors before the end of the monitoring period but did not report that fact. Examination of the assessment records did not reveal any systematic failure to provide full data sets, e.g. due to a particular research assistant delivering instructions. During the main trial, however, additional emphasis will be placed on ensuring that participants understand the importance of wearing the monitor for the full monitoring period. Spare ECG pads will also be provided in case shorter periods of wear were due to the original set of pads becoming detached from the skin.
Reach and representativeness
Glasgow and colleagues define 'reach' as an individual-level measure of participation, which refers to the percentage and risk characteristics of persons who receive or are affected by a policy or program . A casual glance at the CONSORT flow diagram (Figure 1), may lead to the conclusion that this intervention is lacking the reach necessary for an effective health behaviour intervention. Study eligibility criteria, however, means that the recruitment yield should be understood as 57% of those having received and benefited from a brief intervention (Figure 1 Box K) rather than 25% of those eligible for the brief intervention (Box G), or.4% of those originally contacted (Box A).
The other concern is whether the study population is representative of target populations. The high number of women (68.1%) in the trial population reflects those who were eligible for the brief intervention (Figure 1 Box G, 68.5% female) and those who came forward for the brief intervention (Figure 1 Box E: 63.5% female) rather than the target population of the pilot area, High Green (50.5% female). The age distribution of the pre-trial population (Figure 1 Box E: n = 277) was not only skewed but differed markedly between genders. Women showed a linear increase in the response to the initial mailout with increasing age, such that almost twice as many women in the 60-64 years bracket responded as those aged 40-44 years. On the other hand, the response to the mailout among men was bimodal with peaks in the 40-44 years and 60-64 years categories and the fewest respondents from the 50-54 years category. High Green residents are predominantly White-British (97.4%) and this seemed to be reflected in both our contactable respondents and those eligible for the DVD, although no ethnicity data was collected during screening as it was not relevant to eligibility. Similarly, no income data was collected on the pre-trial population. Concerns expressed by local public health leads, however, that a mass mailout would attract people from the relatively wealthy periphery rather than those in greater need living in the more deprived core of the neighbourhood appear unwarranted; an informal mapping exercise revealed a relatively even spread over the pilot area.