Methods

This chapter is a summary of the full study protocol for the trial as originally funded, except for the paragraph which describes changes to the published protocol. Some of the material, including the tables, has already appeared in publication,70 and is reproduced here under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution and reproduction in any medium, provided the original work is properly cited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article,70 unless otherwise stated. Further funding was later awarded by the National Institute for Health Research (NIHR) Health Technology Assessment programme for a 3-year follow-up of the trial cohort, the methods and results of which are described in Chapter 8.

Study design

The PACE-UP walking intervention trial was a pragmatic, three-arm parallel-cluster trial (randomised by household to allow individuals and couples to participate). It was based in primary care with 45- to 75-year-old inactive adults, with a 12-month follow-up period, and compared the following three groups: control (usual PA); pedometer and written instructions by post (pedometer by post); and pedometer, written instructions and practice nurse individually tailored PA consultations (pedometer plus nurse support).

Study aims and objectives

Study aims

The main hypotheses to be addressed were as follows:

1. Does a 3-month postal pedometer-based walking intervention increase PA in inactive 45- to 75-year-olds at the 12-month follow-up point?
2. Does providing practice nurse support through dedicated PA consultations provide additional benefit?

The study also aimed to assess the cost-effectiveness of both interventions, whether or not any factors modified the intervention effects and the effect of the interventions on patient-reported outcomes, anthropometric measures and primary care-recorded AEs.

Primary objectives (relating to the primary outcome of step counts)

In inactive adults aged 45–75 years, the primary objectives were to:

• confirm that tailored support from practice nurse PA consultations combined with a pedometer-based walking programme can promote an increase in step counts compared with the control group at 12 months (pedometer plus nurse support vs. control)
• determine whether or not the simple provision by post of pedometers plus written instructions for a pedometer-based walking programme can promote an increase in step counts compared with the control group at 12 months (pedometer by post vs. control)
• estimate the effect of tailored support from practice nurse PA consultations combined with a pedometer-based walking programme compared with the postal pedometer-based walking programme, on step counts at 12 months (pedometer plus nurse support vs. pedometer by post).

Secondary objectives (relating to secondary outcomes of time in moderate to vigorous physical activity in bouts, sedentary time and cost-effectiveness)

In inactive adults aged 45–75 years, the secondary objectives were to:

• confirm that tailored support from practice nurse PA consultations combined with a pedometer-based walking programme can promote an increase in steps counts at 3 months and time spent in MVPA in ≥ 10-minute bouts at 3 and 12 months, and a decrease in sedentary time at 3 and 12 months compared with control (pedometer plus nurse support vs. control)
• determine whether or not the simple provision by post of pedometers plus written instructions for a pedometer-based walking programme can promote an increase in step counts at 3 months, an increase in time spent in MVPA in ≥ 10-minute bouts at 3 and 12 months and a decrease in sedentary time at 3 and 12 months compared with the control group (pedometer by post vs. control)
• estimate the effect of tailored support from practice nurse PA consultations in addition to the pedometer-based walking programme alone on step counts at 3 months, and time spent in MVPA in ≥ 10-minute bouts and sedentary time at 3 and 12 months (pedometer plus nurse support vs. pedometer by post)
• determine the cost-effectiveness of these alternative approaches to increasing PA levels at both 12 months and from a lifetime perspective from the viewpoint of the NHS and participants (see Chapter 4).

Other objectives

• To determine the effect of the interventions on anthropometric measures (BMI, waist circumference and body fat) at 12 months.
• To determine the effect of the interventions on patient-reported outcomes (self-reported PA levels, anxiety and depression score, exercise self-efficacy, quality of life, pain, AEs) and on primary care-recorded AEs at 3 months and 12 months.
• To determine whether or not age groups (< 60 years vs. ≥ 60 years), sex, taking part as a couple, socioeconomic group, disability, pain, BMI and exercise self-efficacy modify the effect of the intervention on increasing step count at 3 months and 12 months (ethnic group was originally intended to be included as an effect modifier, but there was inadequate power for this analysis because of the low number of non-white participants; see Changes from the published protocol).
• To compare the age, sex, socioeconomic group and ethnicity of those taking part in the trial with those invited but not participating, and to explore the reasons for not participating (see Chapter 5).
• To assess the fidelity and quality of the intervention implementation over time, by the evaluation of patient diary step count goals and recorded step counts for both intervention groups at the 3-month assessment, and the number and timing of recorded practice nurse contacts for the nurse support group (see Chapter 6).
• To explore the intervention’s acceptability to practice nurses and inactive adults, the reasons for dropout and the durability of effects, by qualitative interviews with participants after the 12-month follow-up, and a focus group with the nurses on study completion (see Chapter 7).

Practice and participant inclusion/exclusion criteria

Practice inclusion criteria

General practices were recruited through the Primary Care Research Network – Greater London. Practices were required to be in the south-west London cluster, have a practice list size of > 9000, give a commitment to participate over the duration of the study, have a practice nurse interested and with time to carry out the PA interventions and trial procedures, and have the availability of a room for the research assistant to recruit participants and carry out baseline and follow-up assessments.

Participant inclusion criteria

Participants were patients aged 45–75 years, who were registered with one of the recruited south-west London general practices, were able to walk outside the home and had no contraindications to increasing their MVPA levels.

Participant exclusion criteria

• Physical activity based (by screening question on invitation letter). In order to maximise the benefits of the intervention to individuals and the NHS, the trial focused on less-active adults, using a single-item validated questionnaire measure of self-reported PA as a screening question to identify them.60 Those who reported achieving a minimum of 150 minutes of MVPA weekly1 on their response letter were excluded (participants who, on subsequent baseline accelerometer assessment, were found to be above this PA level were not excluded, as they would be included if this intervention were to be rolled out in primary care).
• Health based [either by the Read code from primary care records or by general practitioner (GP)/practice nurse opinion, or from the telephone or face-to-face baseline assessment with the research assistant] for the following reasons:
  • housebound or living in a residential or nursing home
  • three or more falls in the previous year, or one or more falls in the previous year requiring medical attention
  • terminal illness
  • dementia or significant cognitive impairment
  • registered blind
  • new-onset chest pain, myocardial infarction, a coronary artery bypass graft or an angioplasty within the last 3 months
  • a medical or psychiatric condition that the GP (or practice nurse) considered to exclude the patient (e.g. acute systemic illness such as pneumonia, acute rheumatoid arthritis, unstable/acute heart failure, significant neurological disease/impairment, unable to move about independently, psychotic illness)
  • pregnancy.

Recruitment of practices and participants: informed consent

Practice recruitment

The Primary Care Research Network – Greater London identified practices that fitted the above practice inclusion criteria. Practice recruitment was challenging for a number of reasons, including difficulties in finding practices with sufficient space to accommodate a research assistant on a regular basis, finding practices with nurses willing and with sufficient time to be engaged in delivering the intervention and finding practices that were prepared to provide administrative support. The Primary Care Research Network – Greater London provided us with strong support to recruit practices. Initially, six practices were recruited, with an additional practice added half-way through to boost recruitment. This was necessary, as recruitment at that point was running at just below 10% and we were concerned that we would not achieve the target recruitment from the original six practices within 12 months. The practices were selected to include a range of sociodemographic factors and geographical circumstances based on the practice postcode Index of Multiple Deprivation71 (IMD) scores (at least one practice from each quintile).

Participant recruitment

Practice staff identified patients aged 45–74 years on their primary care electronic patient record system, and, using Read codes and local care home knowledge, excluded ineligible patients (patients were aged 45–74 years when selected, but some were aged 75 years by the time of recruitment or randomisation). A list of potentially eligible patients was produced and ordered by household, with a unique household identifier number. An anonymised list was then used by the research team to create at least four random samples of 400 individuals at each practice. A maximum of two people per household were selected (we were aiming to select couples). If a household had two individuals, one was selected at random, and if the second individual had an age difference of ≤ 15 years, they were also selected; if they fell outside this age range, they were not included. If a household had more than two individuals, one was selected at random, and if there was a second individual aged within ≤ 15 years they were also selected; if there was not a second individual aged within ≤ 15 years, this became an individual household. Each sample list was examined by practice nurses or GPs to ensure trial suitability prior to invitation (see Participant exclusion criteria). Participants were recruited between September 2012 and October 2013, and follow-up was completed by October 2014.

Non-responders and non-participants

See Chapter 5 for more details.

Informed consent

Patients were sent an invitation letter from their own practice, along with a participant information sheet and a screening question on self-reported PA. A reminder invitation was sent if no reply was received after 6 weeks. A log was kept of the response rates from each practice. The decision regarding participation in the study was entirely voluntary. Those interested in participating returned the reply slip, including a response to a single screening question about their usual PA levels. If the participant self-reported as not achieving the PA guidelines,1 the research assistant arranged a baseline appointment for them and ran through the participant information sheet, and handled any questions or concerns that they had. If they were happy to proceed, they signed the study consent form; this form included consent to be contacted for qualitative interviews and consent for their general practice records for the year of the trial to be downloaded after trial follow-up was completed. Participants who had difficulty understanding, speaking or reading English were accompanied by a family member or friend during the research assistant appointment. Participants within a couple could attend together or separately.

Changes from the published protocol

We planned to recruit from six general practices, but to enable target recruitment, a seventh practice was added in December 2012. Changes from protocol-planned analyses70 were approved by the Trial Steering Committee (TSC), prior to analysis. We report MVPA in ≥ 10-minute bouts, as this relates more closely to PA guidelines.1,2 Only 20% of participants were non-white; ethnic group was therefore excluded from the subgroup analyses, as a result of low power.

Interventions

Table 1 shows the components of the intervention provided to the postal and nurse groups. Table 2 shows the content of the patient handbook and the patient diary and the BCTs that were included in each of them, rated according to Michie et al.’s CALO-RE taxonomy.68 Table 3 shows the timing and session content for the three dedicated nurse PA consultations and the BCTs intended to be covered in each session. Figure 1 provides a summary of the 12-week walking programme in terms of steps per day or time spent walking, to be added to each individual’s baseline average daily steps. The training received by the practice nurses in order to deliver the interventions is described in Chapter 6. A figure summarising the trial procedures and complex intervention components is shown in Appendix 1, Figure 17.

FIGURE 1 A summary of the PACE-UP walking programme. Adapted from Harris et al.70 This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. (Figure not reproduced here; see the original article.)

Procedure for the postal intervention group

The participants received (by post) a pedometer (Yamax Digi-Walker, SW-200 model; Yamasa Tokei Keiki Co., Ltd, Tokyo, Japan), instructions and a 12-week step count diary for the 12-week walking intervention (see Figure 1). The research assistant contacted the participant to check that the pedometer had been received and to resolve any difficulties with the equipment. At the end of the 12-month follow-up period, the postal group were offered a single practice nurse PA appointment, if they wanted it.

Procedure for the nurse intervention group

Three dedicated PA consultations (week 1, week 5 and week 9) were arranged with the practice nurse, to individually tailor and support the 12-week pedometer-based walking programme (see Figure 1). At their first appointment, participants were given the same pedometer, diary and handbook that the postal group received. Participants were asked to wear a pedometer and keep a diary record of daily steps for 4 weeks between appointments, in order to review targets and goals at their next appointment. Participants were seen individually or as a couple.

Procedure for the control group

The participants were advised to continue their usual activity levels and were not offered the 12-week walking intervention, but were free to participate in any other PA, just as they would if they were not enrolled in the trial. At the end of the 12-month follow-up period, the control group was offered to receive a pedometer and the PACE-UP 12-week walking programme handbook and diary, either by post or as part of a single PA practice nurse appointment, as preferred.

Outcome measures

Primary and secondary outcome measures

These were selected to reflect the needs of the target population, helping adults and older adults to increase their PA, particularly through walking, and to inform UK public health policy. The primary outcome was the change in average daily step count, measured over 7 days, between baseline and 12 months, assessed objectively by accelerometry (GT3X+; ActiGraph LLC, Pensacola, FL, USA). Secondary outcomes were as follows: changes in step counts between baseline and 3 months; changes in time spent weekly in MVPA in ≥ 10-minute bouts between baseline and 3 months, and between baseline and 12 months; and time spent sedentary between baseline and 3 months, and between baseline and 12 months. All of these secondary outcomes were also assessed objectively by accelerometry. Cost-effectiveness was also a secondary outcome in our protocol [incremental cost per change in step count and per quality-adjusted life-year (QALY)]; this is presented in Chapter 4.

Ancillary outcomes

• Change in self-reported PA, measured over the same 7 days as accelerometry using the short International Physical Activity Questionnaire (IPAQ)72 and the General Practice Physical Activity Questionnaire [(GPPAQ)73 as part of the 7-day PA questionnaire; see Appendix 1].
• Change in other patient-reported outcomes (from the health and lifestyle questionnaires at baseline, 3 months and 12 months; see Appendix 1): confidence in ability to do PA, as measured by exercise self-efficacy;74 anxiety and depression, as measured by the Hospital Anxiety and Depression Scale (HADS);75 perceived health status (health-related quality of life), as measured by the EuroQol-5 Dimensions, five-level version (EQ-5D-5L);76 and self-reported pain, measured by two items from the Medical Outcomes Study 36-item short-form health survey.77
• Change in anthropometric measurements (weight, BMI, waist circumference, body fat; see Ascertainment of outcomes, Anthropometry, for measurement details).
• Adverse outcomes [falls, fractures, injuries, exacerbation of pre-existing conditions, major cardiovascular events and deaths from serious AEs (SAEs)] were collected as part of safety monitoring for the trial, by questionnaire self-report items designed by us at 3 and 12 months, and from primary care records after the 12-month follow-up period, for those giving consent.
• Health service use for those giving consent to primary care record access for the 12-month trial period, numbers of the following occurrences were collected for health economic evaluations (see Chapter 4): primary care consultations, accident and emergency (A&E) attendances, emergency and elective hospital admissions and outpatient referrals.

Ascertainment of outcomes

See Figure 2 for the schedule for outcome assessments and measures.

Accelerometry

Participants were asked to carry on with their usual PA levels and to wear an accelerometer (GT3X+, ActiGraph LLC) on a belt over one hip, during waking hours (from rising until going to bed) for 7 days, only removing it for bathing, at baseline, 3 months and 12 months. Participants were offered the option of text messaging to remind them to wear the accelerometer each day and to return it after the 7 days. A diary was provided to record what activities were done and for how long. The monitor, belt and diary were posted back on completion. Once returned, the participants received a £10 gift voucher.

Anthropometry

At the baseline and 12-month face-to-face assessments, the following measurements were taken: height (measured in bare feet to the nearest 0.5 cm using a stadiometer), weight (measured to the nearest 0.1 kg), body fat, bioimpedence [using the Tanita body composition analyser BC-418 MA (Tanita Corporation, Tokyo, Japan)] and waist and hip circumference (using a standard technique and tape measure with a clear plastic slider).

Questionnaire measures

Questionnaire measures were collected using validated tools (detailed under Outcome measures, Ancillary outcomes), as part of self-completed questionnaires at 3 months and 12 months.

In terms of the self-reported PA in the previous week, for the IPAQ72 we used the measure of MVPA (total minutes of vigorous and moderate PA weekly) and the measure of walking (total minutes of walking weekly). The GPPAQ73 provides a PA index (PAI), which is calculated from a combination of PA from both work and leisure activities. Active individuals are those who self-report ≥ 3 hours of MVPA per week on the PAI. However, walking is not included in the calculation of the PAI, although it is asked about in the questionnaire. Analysis of similar GPPAQ data compared with objective accelerometry in our earlier PACE-LIFT trial21 demonstrated that a modified PAI, which also included walking at a brisk pace for at least 3 hours per week, improved validity and repeatability compared with the standard GPPAQ.78 A modified index, GPPAQ-Walk, was therefore generated.

In addition, the following were also recorded at baseline: demographic information, based on 2011 census questions79 (e.g. marital status, ethnic group, occupation, employment, household composition, home ownership); a list of common self-reported chronic conditions (e.g. heart disease, lung disease, arthritis, stroke, diabetes mellitus, depression); disability, as measured by the Townsend score;80 limiting long-standing illness;79 current medications; smoking; and alcohol consumption.

Several other questionnaire variables were collected at all three time periods, but were not considered to be ancillary trial outcomes in the trial protocol:70 loneliness, measured by a single item;81 risk of falls, measured using the Falls Risk Assessment Tool,82 was assessed using a combination of both self-reported items and direct observation of the ability to rise from a chair without using arms; and self-reported usual PA, as measured by the modified Zutphen Physical Activity Questionnaire.83 Data from these variables are not presented in this report.

All study groups were asked about falls, injuries, fractures, exacerbation of any pre-existing conditions and the costs of any treatments in the 3- and 12-month questionnaires. Questions on the financial costs of participating in walking and other PAs were asked in the 3- and 12-month questionnaires.

Primary care computerised record measures

For participants who gave written consent, the following data were collected from their electronic primary care records, for the 12-month duration of the trial, after the 12-month follow-up:

• adverse events potentially relating to trial participation [Read codes relating to falls, fractures, injuries, cardiovascular events (myocardial infarction, coronary artery bypass graft, coronary angioplasty, transient ischaemic attack, stroke) and death]
• health service use – GP consultations, practice nurse consultations (excluding those for the trial), A&E attendances, emergency and elective hospital admissions and outpatient referrals.

These data were downloaded and pseudoanonymised before removal from the practice.

Baseline and follow-up data collection

Baseline data collection

At baseline, a face-to-face assessment with the research assistant occurred at the participant’s general practice, and questionnaire and anthropometric data were collected (see Ascertainment of outcomes). Participants were then given a belt with an accelerometer (GT3X+, ActiGraph LLC) and a blinded pedometer (Yamax Digiwalker CW200) on it, and asked to wear this for 7 days. The CW200 pedometer model was used to enable the baseline target-setting of the pedometer step count, because of its 7-day memory of consecutive daily steps. However, it is bulky to wear and complicated to use, so this model was not used for the intervention.

Follow-up data collection

Follow-up data collection was conducted in the same way for all trial groups (Figure 2): (1) 3 months (postal) after randomisation (questionnaires and accelerometry) and (2) 12 months (face to face) after the baseline assessment (questionnaires, accelerometry and anthropometry). Participants were also contacted by the research assistant at 6 and 9 months after randomisation by telephone or e-mail to check on falls for trial safety reporting and contact details. For those in the intervention groups, a replacement pedometer or batteries were offered at each contact point, if required. The intervention groups were asked to return their 12-week step count diary following the intervention at 3 months. This was then photocopied and sent back to participants.

FIGURE 2 Schedule of outcome assessment measures used in the PACE-UP trial. (Figure not reproduced here; see the original article.)

Accelerometer data reduction

The accelerometer measured vertical accelerations in magnitudes from 0.05 to 2.0 g sampled at 30 Hz, then summed over a 5-second epoch time period. ActiGraph data were reduced using Actilife software (v 6.6.0; ActiGraph LLC, Pensacola, FL, USA), ignoring runs of ≥ 60 minutes of zero counts.70 Vertical counts were used, as these are the basis of the validated step count and MVPA algorithms. The analysis summary variables used were step counts, accelerometer wear time, time spent in MVPA (≥ 1952 counts per minute, equivalent to ≥ 3 metabolic equivalents),84 time spent in ≥ 10-minute MVPA bouts and time spent being sedentary (≤ 100 counts per minute, equivalent to ≤ 1.5 metabolic equivalents).85

Adverse events and serious adverse events

An AE was defined as any unfavourable and unintended sign, symptom, syndrome or illness that developed or worsened during the observation period of the trial. This included:

• exacerbation of a pre-existing illness
• an increase in frequency or intensity of a pre-existing condition
• a condition detected or diagnosed after the trial started (but which might have been present at baseline)
• a persistent disease or symptoms present at baseline that worsened following the start of the trial.

A SAE was defined as any AE occurring during the trial for any of the three groups that resulted in any of the following outcomes:

• death
• a life-threatening AE
• inpatient hospitalisation or prolongation of existing hospitalisation
• a new disability/incapacity.

All AEs were assessed for seriousness, expectedness and causality. All AEs were recorded and closely monitored until resolution or stabilisation, or until it had been shown that the study intervention was not the cause. Participants were asked to contact the trial site immediately in the event of any SAE. The chief investigator was informed immediately and determined seriousness and causality in conjunction with two other medically trained trial investigators. A SAE that was determined to be directly or possibly trial related was reported, within agreed time frames, to the TSC and the ethics committee. All SAEs were reported annually to the TSC, ethics committee and the trial sponsor.

Although it was important to record AEs contemporaneously for trial safety monitoring during the trial, there was a risk of bias in their reporting, with those having nurse contact having more opportunities for reporting falls, injuries and illnesses. For analyses and reporting, we therefore concentrated on measures for which there were fewer risks of bias between groups: (1) spontaneously reported SAEs, (2) falls, fractures and injuries from questionnaire self-report at 3 and 12 months and (3) falls, fractures, injuries, cardiovascular events (new episode of any of the following: myocardial infarction, angioplasty, coronary artery bypass, stroke, transient ischaemic attack, new-onset angina, ischaemic heart disease) and deaths from primary care records after the 12-month follow-up point.

Sample size

A total of 217 patients in each of the three trial arms would allow a difference of 1000 steps per day to be detected between any two arms of the trial, with 90% power at the 1% significance level. However, we planned to randomise households. Assuming an intracluster correlation of 0.5 and an average household size of 1.6 eligible patients, we needed to analyse 282 patients per trial arm. Allowing for approximately 15% attrition, we needed to randomise a total of 993 patients (331 control participants, 331 participants receiving a pedometer by post and 331 participants receiving a pedometer plus nurse support). We initially planned on six practices to each recruit approximately 166 patients (approximately 55 participants to each of the three groups), but to enable target recruitment, a seventh practice was added.

We anticipated a 20% recruitment rate among eligible participants, based on other PA interventions (including with pedometers) among middle-aged and older adults in primary care, where the recruitment rate was between 17% and 35%.21,33,86–89 We estimated that, even if our recruitment rate was as low as 10%, we would have enough eligible participants at practices. In fact, the recruitment rate dipped to below 10%, so a seventh practice was added.

Randomisation, concealment of allocation, contamination and treatment masking

Randomisation and concealment of allocation

Following completion of the baseline assessment (including providing accelerometry data on ≥ 5 complete days of ≥ 9 hours/540 minutes), each participant was allocated to a trial group using the King’s College London clinical trials unit internet randomisation service, to ensure independence of the allocation. If participants were unable to provide at least 5 days of ≥ 540 minutes wear time on accelerometry, they were asked to wear the accelerometer for a further 7 days, or they were excluded if this was not possible. Randomisation was at a household level. Randomisation of a group household took place only after both members of the household had completed the baseline assessment. Block randomisation was used within the practice with randomly sized blocks (2, 4 or 6) to ensure balance in the groups and an even nurse workload. Participants were informed by telephone which group they had been allocated to.

Contamination

Contamination could occur between partners in a household; we minimised this by ensuring that, if two household members were recruited, they were allocated to the same group (i.e. randomisation was at a household level). Contamination would have occurred if the control group used a pedometer to increase their walking during the 12-month trial follow-up. We tried to discourage participants in the control group from buying a pedometer, by ensuring that they knew that they would receive one at the end of follow-up. A question was included in the 12-month questionnaire to ask if they had used a pedometer during the course of the trial.

Treatment masking

Participants were randomised only after the successful return of accelerometers with 5 days’ recording. It was not possible to mask participants to their intervention group. The research assistants who carried out follow-up assessments were not masked to group allocation for pragmatic reasons alone: the study was funded to support only enough researchers to carry out recruitment and follow-up simultaneously. However, the main outcome was assessed objectively through accelerometry, and the assessment of the quality of the outcome data was done blind to intervention group: days with < 9 hours of data were excluded. Weight and body fat were also assessed objectively using the Tanita scales, which provided electronic printouts of results, and other outcomes were assessed using standardised measures (e.g. patient-reported outcomes from questionnaires). The statistician carrying out the primary analyses was masked to group allocation as far as possible.

Withdrawals, losses to follow-up and missing data

Withdrawals and losses to follow-up

Participants could withdraw from the trial at any point. Participants who withdrew following informed consent, and prior to randomisation, were replaced with another participant. Participants who withdrew after randomisation were not replaced and were asked if they were prepared to contribute to further data collection on outcomes at 3 and 12 months. Participants were made aware that withdrawal from the trial would not affect future care and that information on those who withdrew or were lost to follow-up that had already been collected would still be used, unless consent for this was withdrawn.

Procedure for accounting for missing data

Only days with at least 540 minutes of registered time on an accelerometer on a given day were used, which was consistent with previous work (the PACE-LIFT trial,21 Trost et al.90 and Miller et al.91). Participants were randomised only if they provided at least 5 such days of accelerometer data at baseline. A multilevel linear regression model was used, taking account of repeated days within individuals to estimate the baseline average daily step count for each subject, adjusted for the day of the week and the day order of wearing the accelerometer. The same approach was used to estimate the average daily step count at 3 months and 12 months. The main covariates – age, sex, practice, month of baseline accelerometry and whether or not participants were taking part as a couple – were known for all participants, and most patients had complete data for other measures. To lessen attrition bias, the primary analysis included all participants with at least 1 satisfactory day of accelerometry recording at 12 months (i.e. a wear time of ≥ 540 minutes). The main analysis assumed that, depending on the model covariates, outcome data were missing at random. This was likely to be true for missing data as a result of accelerometer failure, and was plausible for missing days and participants who did not return accelerometers. However, an alternative plausible assumption is that participants who failed to provide outcome data were less active. Multiple imputation was used to impute values for those with no accelerometer data at 12 months (see Statistical methods, Sensitivity analyses). Further sensitivity analyses examined the impact of assuming that missing step counts at 12 months in the control group were equal to their baseline values and, in the two intervention groups, varied between 1500 steps lower and 1500 steps higher than their baseline values.

Statistical methods

The analysis and reporting was in line with the Consolidated Standards of Reporting Trials (CONSORT) guidelines, with the primary analysis on an intention-to-treat basis. That is, all participants with outcome data were included, regardless of their adherence to the interventions. All participants were included in the primary analysis if they had at least 1 satisfactory day of accelerometer recording (≥ 540 minutes) of registered time during a day, out of 7 days, at 12 months. The adequacy of the randomisation process to achieve balanced groups was checked by comparing participant characteristics in the three arms (e.g. sex, age, socioeconomic group, baseline PA level or BMI). Stata^®, version 12 (StataCorp LP, College Station, TX, USA) was used for all analyses.

Primary analysis

The primary outcome measure was change in average daily step count from baseline to the 12-month follow-up point measured over 7 days. However, to overcome Lord’s paradox,92 the analytic approach regressed 12-month outcomes on baseline measures, thus allowing for regression to the mean. Eligibility was defined on the basis of ≥ 5 days with ≥ 540 minutes’ activity at baseline. If the participant was asked to wear the accelerometer for a second time, the second 7 days was used in the analysis. If there were > 7 days’ wear on the accelerometer, then the first 7 days were used and later readings were discarded. The primary analysis used all participants providing at least 1 day of ≥ 540 minutes accelerometry wear time at 12 months (i.e. a complete-case analysis).

All analyses were carried out using Stata, version 12. The xtmixed procedure was used for regression models. A two-stage process was used for accelerometry data. Stage 1 estimated the average daily step count at both baseline and 12 months, using a multilevel model in which daily step counts were regressed on day of the week and day order of wearing the accelerometer (from day 1 to day 7) as fixed effects, and with day within individual as the random effect (i.e. level 1 was the day within individual and level 2 was individual). In stage 2, average daily step count at 12 months was regressed on baseline average daily step count, sex, age, general practice, month of baseline accelerometry and treatment group as fixed effects, and household as the random effect, to allow for clustering at a household level (i.e. level 1 was individual and level 2 was household). This method effectively measured the change in step count from baseline to 12 months, minimising bias and maintaining power. Adjusting for baseline steps controlled for many factors that predict the number of steps in cross-sectional analyses (e.g. BMI, socioeconomic group, health status). The reference group for the intervention group comparisons was the control group. The post-estimation command pwcompare was used to obtain the estimates of change with 95% CIs and p-values for the difference in change in steps for the postal group versus the control group, the nurse-support group versus the control group and the nurse-support group versus the postal group. This last comparison provided information on whether or not the nurse intervention promoted a worthwhile increase in activity compared with a pedometer alone. It should be noted that, although this estimate can be obtained from the difference of the first two estimates, pwcompare also provided 95% CIs for this comparison. Checks were carried out to confirm that the distribution of residuals from the regression model for change in steps was normally distributed.

Secondary and ancillary outcome analyses

Secondary PA outcome measures from accelerometry were total weekly minutes of MVPA in ≥ 10-minute bouts, average daily sedentary time at 12 months and steps, MVPA in bouts and sedentary time at 3 months. These data were processed and analysed in the same way as described for the step counts (see Primary analysis). MVPA was highly positively correlated with step counts and sedentary time was negatively correlated with step counts.

Other ancillary outcomes were changes in exercise self-efficacy, anxiety, depression, perceived health status (health-related quality of life, as measured using the EQ-5D-5L), self-reported pain, anthropometric measures (weight, BMI, waist circumference, body fat) and self-reported PA from the IPPAQ and GPAQ questionnaires. Changes in these outcomes from baseline to 3 and 12 months were analysed using identical models to stage 2, as described in Primary analysis (i.e. level 1 was individual and level 2 was household).

Adverse event analyses

The number of participants who suffered an AE between 0 and 3 months or between 0 and 12 months (a spontaneously reported SAE or a systematically reported SAE from the 3- or 12-month questionnaire, or a SAE collected from the primary care record data) was compared between groups using exact tests for categorical tables.

Subgroup analyses

Sex, age groups (< 60 years or ≥ 60 years), taking part as a couple, socioeconomic subgroups, BMI, disability, pain and exercise self-efficacy were examined as potential effect modifiers by adding interaction terms to the regression model for the primary outcome, which was changes in step counts at 12 months.

Sensitivity analyses

Sensitivity analyses were carried out for the primary outcome (change in step counts from baseline to 12 months). The effects of using different criteria for defining satisfactory wear at 12 months were examined as follows: (1) at least 5 days of ≥ 540 minutes’ wear time, (2) ≥ 1 day of ≥ 600 minutes’ wear time and (3) ≥ 5 days of ≥ 600 minutes’ wear time. The effect of adjusting for change in wear time between baseline and 12 months was also examined.

Additional sensitivity analyses assessed whether participants lost to follow-up or who failed to provide a single adequate day’s recording might have introduced bias. This was first done by assuming that outcome data were missing at random, depending on the model covariates, using the Stata procedure mi impute. The first model used the standard model covariates to impute missing step counts at 12 months (treatment group, baseline steps, sex, age, general practice, month of baseline accelerometry and household as a random effect) and the second model added in the National Statistics Socio-economic Classification (NS-SEC), self-reported pain and fat mass as additional covariates. Further analyses explored the possible impact of outcomes not being missing at random, using the following assumptions: among those with missing data in the control group, the change in mean steps from baseline to 12 months was 0, and among those with missing data in each of the intervention groups, the change in mean steps from baseline to 12 months was –1500, 0 or +1500.

Ethics approval and research governance

Ethics approval was granted for the trial from London, Hampstead Research Ethics Committee (reference number 12/LO/0219). The NHS Research and Development approval was granted by the Clinical Commissioning Groups in south-west London, through the Primary Care Research Network, to cover all the practice sites.

Management of the trial

The trial progress, including recruitment, safety, finance and data management, was reviewed regularly by the trial management group (TMG). This was made up of the chief investigator, two trial investigators, the trial statistician and the trial manager. The TMG met on a monthly basis. All of the trial investigators met as a group (the trial investigator group) on a biannual basis, and the TSC met prior to participant recruitment, and then annually or biannually as necessary. Minutes were kept of all TMG, trial investigator group and TSC meetings. The TSC included a patient advisor and more details of their role in terms of patient and public involvement are given in Appendix 1.

Further trial follow-up at 3 years

After the initial trial results were analysed, funding was obtained to follow up the trial cohort at 3 years. Details of the methods and results for this further follow-up are given in Chapter 8.