Chair-based yoga programme for older adults with multimorbidity: RCT with embedded economic and process evaluations

Eligibility criteria for participants

Patients were eligible to join the study if they were aged 65 years or older (both male and female), community-dwelling (including sheltered housing living with support) and had two or more of the following chronic conditions, as derived from a list included in the NHS Quality and Outcomes Framework (QOF version 31.0) and following discussions with the Trial Management Group clinical oversight and yoga consultants:

• arthritis: including osteoarthritis, rheumatoid arthritis and history of shoulder, hip or knee arthroplasty for arthritis

• asthma or chronic obstructive pulmonary disease

• atrial fibrillation

• bowel problems: including irritable bowel syndrome, diverticulitis and inflammatory bowel disease

• cancer, diagnosed within the last 5 years

• cardiovascular disease: including coronary heart disease (includes angina and history of heart attack, bypass surgery or angioplasty), hypertension, heart failure and peripheral arterial disease

• chronic kidney disease

• dementia (only if patients have the capacity to provide written informed consent)

• depression or anxiety

• diabetes

• epilepsy

• fibromyalgia

• multiple sclerosis

• osteoporosis or osteopenia

• Parkinson’s disease

• sensory conditions: including hearing loss, macular degeneration, cataracts and glaucoma

• stroke, within the last 5 years.

Patients were ineligible for the study if they met one or more of the following exclusion criteria:

• inability to attend one of the yoga courses on offer [participants needed to indicate that they would be available to attend at least 9 of the 12 classes on offer for a particular course. In relation to online classes, additional factors that made someone ineligible included: no internet access, unfamiliarity with or inability to use the internet, no suitable device for accessing the online classes (e.g. tablet-size screen or larger; device with camera and microphone), insufficient space at home and no sturdy chair for use in the classes]

• attended yoga classes twice a month or more in the previous 6 months

• contraindications to yoga participation (as identified by the patient’s GP)

• severe mental health problem: Schizophrenia, bipolar affective disorder or other psychotic illness (on advice from the Trial Management Group yoga consultants, potential participants with severe mental health problems and learning disabilities were listed as exclusions on the basis that the GYY teacher training programme does not cover how to accommodate these people, who may have specific support needs, and it was thought that this might make the yoga class size of 12–15 difficult to manage)

• learning disability (on advice from the Trial Management Group yoga consultants, potential participants with severe mental health problems and learning disabilities were listed as exclusions on the basis that the GYY teacher training programme does not cover how to accommodate these people, who may have specific support needs, and it was thought that this might make the yoga class size of 12–15 difficult to manage)

• unable to read or speak English (potential participants who were unable to read or speak English were listed as exclusions due to the uncertainty of being able to adapt the course delivery and questionnaires to accommodate different languages)

• unable to provide consent

• unable to complete and return a valid baseline questionnaire

• no more than one trial participant per household [no more than one trial participant per household could take part to avoid contamination effects (i.e. if one was allocated to the intervention group and the other to the usual care group)]

• currently enrolled in another research study for which concurrent participation is deemed inappropriate by their GP or a clinician co-investigator.

All eligibility criteria were assessed by reviewing responses to specific questions posed in a screening questionnaire [see Trial Participant project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)] that was self-reported by potential participants either in written format or over the telephone with a researcher. Patients who did not meet the eligibility criteria were notified in writing or via a phone call that they were ineligible, and no further correspondence was sent. Patients who were deemed eligible were required to complete a baseline questionnaire [see Trial Participant project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)] either in written format or over the telephone with a researcher. Participants were eligible to be randomised once the study team had received their completed consent form and baseline questionnaire. If a completed baseline questionnaire was not received by the end of the recruitment period, the patient was sent a letter to inform them that recruitment to the trial had closed and that they were unable to participate.

Recruitment of participants

Potential participants were identified by searching the electronic patient databases (SystmOne or EMIS) of 15 general practices. Participating practices ran a custom-built search based on pre-defined read codes, which identified patients with health conditions within the eligibility criteria. A GP then reviewed the resulting list to rule out patients who did not meet the eligibility criteria. Where there were more potentially eligible patients identified than required, practices were asked to order the list of patients by NHS number and select the top required number. Docmail (a third-party information handler) then mailed out a recruitment pack to each of the remaining potentially eligible patients, which included a covering invitation letter, a participant information sheet (PIS) which included a link to an audio version on the trial website, a consent form and a screening questionnaire [see Trial Participant project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)] and two prepaid envelopes for returning the completed forms. If the patient was deemed eligible (based on the information provided in the screening questionnaire), a researcher notified the patient’s GP and asked them to confirm the patient’s suitability for participation. Eligible patients were then sent a baseline questionnaire [see Trial Participant project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)] to complete and return by a specified date. This questionnaire collected data on sociodemographic measures, primary and secondary outcome measures and preferences/beliefs for the treatments on offer in the trial.

Implementation of an alternative process was required during the COVID-19 pandemic as it was difficult to collect study forms by post, particularly when the trial team was required to work from home. Therefore, participants gave consent electronically by completing an online form and provided screening and baseline questionnaire data via a telephone call with a researcher, who entered these data into an online form during the call.

Identification of trial groups

Participants had to be recruited in groups for the yoga courses, and the courses were run over a period of 29 months. To facilitate identification, each recruitment drive was called a ‘wave’. There were two waves in the internal pilot phase: pilot wave 1 (PPW1) and pilot wave 2 (PPW2) and two waves in the main phase of the trial: main wave 1 (MPW1) and main wave 2 (MPW2). The number of courses run within a wave varied depending on GP practice capacity to recruit and yoga teacher availability.

Consenting participants

As detailed above, potentially eligible patients were posted a recruitment pack, which included a covering invitation letter, an information sheet, a screening questionnaire and a hard copy of (or electronic link to) the consent form.

The information sheet provided a balanced written account of the purpose and design of the trial and also included details of who to contact to ask any questions and how they could access an audio recording of the information sheet on the trial website. Part way through recruitment, we introduced a simple diagram of the trial design to be provided along with the information sheet to facilitate participants’ understanding of the randomisation process and what each group was to receive [see Trial Participant project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)].

Participants indicated on the consent form if they also wanted to be considered to take part in the process evaluation interviews. If the participant indicated ‘yes’ and was selected by the process evaluation researcher to take part in the interviews, they were then provided with an information sheet regarding the interviews [see Process evaluation project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)]. Additional consent was sought, either via hard copy or electronically, in line with alternative arrangements, from those who agreed to take part [see Process evaluation project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)]. Individuals who declined to participate in the trial were able to indicate their willingness to take part in the interviews as a ‘trial decliner’, and additional consent was sought from them, as above.

Eligibility criteria for primary care general practices

The selection of general practices was informed by location (i.e. located close, generally within five miles, to the yoga class venue); local transport routes and teacher recommendations for face-to-face classes; computer system used; patient list size and practice staff availability for conducting recruitment activities.

Recruitment of primary care general practices

General practices that were potentially interested in taking part in the trial were identified with help from the NHS Clinical Research Networks (CRN) in England and the Health and Care Research Wales (HCRW) Support and Delivery Centre. We also worked with NHS Research Scotland (NRS) Primary Care Research Network but did not receive any expressions of interest from invited general practices via this network. Trial coordinators then liaised with key stakeholders at each practice (e.g. practice managers, GPs) to explain the requirements of the study. If a practice agreed to take part, a practice-level agreement form was signed to confirm capacity and capability before any recruitment activity commenced. This initially was the Health Research Authority (HRA) and HCRW Statement of Activities document [see GP practice project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)] which was later replaced by the Integrated Research Application System (IRAS) Organisations Information Document [see GP practice project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)]. Practice recruitment is represented in Figure 1.

FIGURE 1.

General practice recruitment flowchart.

Eligibility criteria for yoga teachers

To be eligible for inclusion to deliver the intervention programme, yoga teachers needed to have completed the British Wheel of Yoga Qualifications (BWYQ) Level 4 Teaching GYY qualification and have valid BWY membership and insurance. For online courses, teachers also needed to be proficient in remote teaching. The selection of yoga teachers was informed by observations of them leading online non-trial yoga classes, as conducted by the trial’s yoga consultants (LB and JH).

Recruitment of yoga teachers

Yoga teachers who were potentially interested in taking part in the trial were identified by the trial’s yoga consultants (LB and JH). The consultants were aware of who had completed the Level 4 GYY teacher training course (because of their work at the awarding organisation BWYQ) and where they were based. The consultants and trial coordinators liaised with potential teachers to explain the requirements of the study. This included both delivery of the courses and provision of backup yoga teachers to cover absence. If a teacher agreed to take part, a contract with the University of York was signed before any trial classes were delivered.

Comparator description

The comparator was usual care alone. Usual care was defined as ‘The wide range of care that is provided in a community, whether it is adequate or not, without a normative judgment’. 20 Throughout the trial, both trial arms continued with any usual care provided by primary care, secondary care, community and social services. This approach reflects the main aim of this pragmatic trial: to determine the clinical effectiveness and cost-effectiveness of offering the GYY programme in addition to usual care.

To characterise and quantify usual care, self-reported healthcare resource use (NHS and private care) was collected at baseline and at each follow-up assessment for all participants in both intervention and usual care groups. Prescription data from general practices were also collected for the period between 3 months prior to baseline and up until 12 months after, for a sub-sample of participants in both groups.

The protocol did not restrict access to yoga or any other intervention during the follow-up period. However, to reduce contamination, the trial yoga teachers were asked to only deliver trial classes to the participants who had been randomly allocated to their courses. In the follow-up questionnaires, all participants were asked to report any trial and non-trial, supervised yoga classes and self-managed yoga practice that they had done since the previous study time point.

Intervention description

Gentle Years Yoga is a yoga programme for older adults, including those with chronic health conditions. It is based on standard Hatha Yoga and incorporates traditional physical postures and transitions as well as breathing, concentration and relaxation activities. Each class also includes optional, additional time immediately following the class for participants to stay on and socialise. The main aims of the programme are to improve muscle strength, flexibility, balance, mobility and mental and social well-being. Chairs are used for seated exercise and can be used to provide support when standing, although the whole session can be carried out on a chair. Figure 2 shows examples of seated postures that are commonly used. The yoga practices are modified to allow individuals with varying medical conditions and functional abilities to participate safely. Props are also sometimes used to modify some of the postures and concentration activities. A list of props used in the trial is detailed in Table 1. The physical challenge of each posture can be progressed throughout the course as participants become more able and confident. The following summarises how the GYY classes differ from standard Hatha Yoga classes:

FIGURE 2.

Chair-based postures that are commonly used in GYY. Reproduced from Tew et al. 19

• For the most part, participants are seated on chairs, and when standing, they use the chair or other aids for support.

• The classes do not use supine, semi-supine or prone postures; instead, the key elements of traditional supine and prone postures are integrated into seated or standing postures.

• The classes hold static postures for a shorter length of time, especially those that could cause more pronounced acute increases in blood pressure.

• The physical set-up of classes has been adapted to suit people with sensory impairments; specifically, participants being relatively close to the teacher; lighting levels being higher; the colour of equipment being in contrast to that of the walls, the floor and the teacher, and no music played during verbal instructions.

• The pace and overall structure of the class allow greater time for recovery from the more intense activities (e.g. by having a simple breathing practice follow a more-challenging physical posture).

• If there are individuals with mild cognitive impairment in the group, the teacher will use short, single-subject phrases and pace the instructions to allow time for processing each element of the instructions.

• The classes have a longer warm-up period and an overall slower pace, making it safer for older adults and at a level where they can work without feeling ‘left behind’ or ‘too old for yoga’ or having their self-confidence eroded.

• Breathing practices avoid retention, as this is contraindicated for individuals with hypertension.

• Mobilisation, postures and concentration activities are incorporated that specifically focus on balance and co-ordination.

Participants randomised to the intervention were invited to take part in 12 free-of-charge, 75-minute group-based GYY classes. Each class was immediately followed by an optional 15- to 30-minute period of socialising. All courses commenced within 3 weeks of randomisation. Each course involved one class per week for 12 (mostly) consecutive weeks; however, there was allowance for a gap in delivery for public holidays and unforeseen circumstances. A summary of actual time between randomisation and commencement of a yoga course and between each class, indicating any gaps in delivery, is provided in Report Supplementary Material 3. Before the first class, participants were required to complete and submit a BWY Health Questionnaire to their yoga teacher [see Yoga participant project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)]. This questionnaire was needed for BWY teaching insurance purposes and so that the teachers were aware of each participant’s medical conditions and physical activity status.

Each class included: (1) ‘housekeeping’ activities, such as completing the class register and discussing any home practice or health issues (5 minutes); (2) an introduction to the theme and practices of the class, basic breathing and focusing activities (5 minutes); (3) an extended warmup/mobilisation and preparatory postures (30–35 minutes); (4) focused postures and restorative activities (10–15 minutes); (5) breathing exercises (5–10 minutes) and (6) relaxation and concentration activities (5–10 minutes), followed by optional after-class social time (15–30 minutes).

The after-class social time provided an opportunity for participant interaction and building of social networks. The yoga teachers encouraged participants to stay on for this component, but it was not mandatory. The focus of conversations was not standardised; however, the teachers were advised that there was a preference for non-yoga-based discussions. The teacher could still provide general yoga advice to participants if directly asked.

The classes also included instruction on yoga activities that the participants could do at home. Yoga teachers provided participants with a home practice sheet that detailed these activities, which were to be completed at home on non-class days. As the class-based activities became more challenging, students were given new home practice sheets to allow progression of their home yoga routine. Each yoga participant received four home practice sheets in total, covering weeks 1–3, 4–6, 7–9 and 10–12, respectively [see Yoga participant project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)]. Each sheet included at least five practices, providing an expected practice time of 10–20 minutes per session.

Towards the end of the course (i.e. classes 11 or 12), the yoga teachers provided participants with general verbal advice about continuing yoga practice and a written or electronic handout [see Yoga participant project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)] sign-posting participants to suitable yoga classes (i.e. GYY or similar) in their local community or online, which they could attend on a self-pay basis.

The delivery mode of the trial yoga classes was originally face to face. Online classes were implemented during Pilot Phase Wave 2 (autumn 2020) and addressed the closure of venues during the first national lockdown and, later, the need for social distancing during the COVID-19 pandemic. The trial classes included intervention participants only (i.e. members of the public could not participate), with each 12-week course having 12 participants allocated if delivered online and up to 15 participants allocated if delivered face to face (see Report Supplementary Material 2). Face-to-face classes were conducted in non-medical community-based facilities (e.g. yoga studio, community hall, leisure centre) following checks for venue suitability by the yoga consultants (LB and JH). Accessibility factors that were considered included close proximity to public transport links, parking facilities and disability access. Online classes were conducted via Zoom, an online livestream platform freely accessible to participants using a computer or tablet. Pre-course contact by the yoga teachers with participants was introduced from the start of Pilot Phase Wave 2. Prior to online classes commencing, the yoga teacher conducted one-to-one Zoom meetings with their participants to optimise the set-up of their equipment and environment and to discuss any health issues or course queries. Prior to face-to-face classes commencing, the yoga teachers conducted one-to-one telephone meetings with their participants to discuss any health issues or course queries.

Criteria for discontinuing or modifying allocated interventions

There were no specific criteria for discontinuing or modifying allocated interventions. Participants could decide to stop doing the yoga programme at any point and for any reason.

Strategies to improve adherence to interventions

To optimise and encourage attendance, the teachers were asked to contact participants who missed two consecutive classes without prior notification.

Relevant concomitant care permitted or prohibited during the trial

Participants continued to receive independent, usual care throughout the trial, and this was not prohibited in any way.

Provisions for post-trial care

Towards the end of the course (i.e. classes 11 or 12), the yoga teachers provided intervention participants with general verbal advice about continuing yoga practice and a paper or electronic handout [see Yoga participant project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)] sign-posting participants to suitable yoga classes (i.e. GYY or similar) in their local community or online, which they could attend on a self-pay basis. The usual care participants received the same information after completing the final (12-month) follow-up questionnaire. Usual care participants were also randomised into a methodological substudy in which half were offered a GYY class at the end of their 12-month follow-up. They were informed of this offer shortly after randomisation in order to determine whether this improved their retention in the trial. Methods and results of this substudy are detailed in Chapter 6.

Primary outcome measure

The primary outcome was the EQ-5D-5L utility index score. 21 The EQ-5D-5L was self-reported by the participant and collected using questionnaires at baseline and at 3, 6 and 12 months post randomisation. The primary end point was the overall difference over the 12-month follow-up period. There is currently work ongoing to develop a valuation set for the EQ-5D-5L for England; in the meantime, the UK’s National Institute for Health and Care Excellence (NICE) recommends22 that utility index scores should be calculated using the crosswalk developed by van Hout et al. ;23 hence, they were calculated on this basis.

The EQ-5D^TM is a widely used self-reported health utility measure that comprises two parts: the classification of five dimensions of health (mobility, self-care, usual activities, pain/discomfort and anxiety/depression) and a visual analogue scale (VAS), which records participants’ overall evaluation of their health on a scale from 100 (best imaginable health) to 0 (worst imaginable health). The EQ-5D has been validated in many different patient populations, including those with diabetes, cardiovascular problems, chronic obstructive pulmonary disease, cancer, chronic pain and rheumatoid arthritis.

There are currently two versions of the instrument that can be used for adults: the original EuroQol-5 Dimensions, three-level version (EQ-5D-3L) with five dimensions of health and three response levels of problems and the more recent EQ-5D-5L that has the same five dimensions of health but has five response levels of problems (1 = no problems, 2 = slight problems, 3 = moderate problems, 4 = severe problems and 5 = unable/extreme problems). The EQ-5D-5L helps overcome problems with ceiling effects and has greater sensitivity. 21 It showed evidence of good sensitivity in the pilot trial of the GYY programme19 and has been the primary outcome measure in other primary care-based multimorbidity trials. 24 Responses to the EQ-5D-5L lead to 3125 unique possible combinations of health states where each health state is mapped to a utility index score (on a scale where negative values correspond to a state worse than death, 0 corresponds to a health state equivalent to being dead and 1 corresponds to perfect health) by making use of a valuation set. Participants who die can be given a score of 0 (for both the utility index score and VAS) for any assessment time point following their date of death.

Besides being used as the primary outcome measure in the analysis, the EQ-5D-5L was also used to estimate QALYs for the economic evaluation (see Chapter 4).

Secondary outcomes measures

All secondary outcomes were self-reported by the participant and collected using questionnaires at baseline and during the 12-month follow-up [see Trial Participant project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)]. The secondary outcomes were:

• EQ-5D-5L utility index score at 3, 6 and 12 months after randomisation

• EQ-5D-5L VAS score at 3, 6 and 12 months and overall

• HRQoL at 3, 6 and 12 months and overall using the PROMIS-2925

• depression severity at 3, 6 and 12 months and overall using the Patient Health Questionnaire-8 (PHQ-8)26

• anxiety severity at 3, 6 and 12 months and overall using the Generalised Anxiety Disorder-7 (GAD-7)27

• loneliness at 3, 6 and 12 months and overall. Four questions were used to capture different aspects of loneliness. The first three questions were taken from the University of California, Los Angeles 3-item (ULCA-3) loneliness scale. 28 The wording of the UCLA questions and response options was taken from the English Longitudinal Study of Ageing (ELSA). 29 The last was a direct question about how often the respondent feels lonely

• incidence of falls over 12 months

• AEs over 12 months

• healthcare resource use over 12 months (see Chapter 4, Economic Evaluation).

Process evaluation

A qualitative process evaluation was informed by qualitative interviews with trial participants, trial decliners, trial yoga teachers and stakeholders, as well as by observations of standardisation training sessions and yoga classes (see Chapter 5, Process Evaluation).

Intervention fidelity

To ensure the trial yoga courses were delivered in accordance with the GYY teacher training programme and GYY trial guidelines, each yoga teacher underwent an observation of one of their trial classes by one of the yoga consultants (LB and JH). A checklist was completed for each observation [see Yoga teacher project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024) for the yoga teacher fidelity check assessment template.]. Intervention fidelity was also assessed via class observations and interviews with each yoga teacher as part of the process evaluation.

Other data collected

Sociodemographic measures (age, gender, ethnicity, residential status, employment status, smoking status) and details of health conditions were collected at baseline via the screening and baseline questionnaires. An adapted Bayliss measure of illness burden31 was calculated by summing the amount that each self-reported health condition limits a participant’s daily activities from 1=‘Not at all’ to 5=‘A lot’. Participant beliefs and preferences for the GYY programme and usual care were assessed in the questionnaires at baseline and 12 months. Adherence by participants to the supervised GYY classes was recorded by the yoga teachers using class attendance registers [see Yoga teacher project documents, https://fundingawards.nihr.ac.uk/award/17/94/36 (accessed May 2024)]. Participants also self-reported in the follow-up questionnaires any other supervised or self-managed yoga practice that they had done since the previous study time point. Adverse events were recorded (see ‘Adverse event reporting and harms’ below).

Original

Walters and Brazier, in a review paper of the EQ-5D-3L,32 found a difference of 0.074 (mean) or 0.081 (median) to be a minimum clinically important difference among a variety of patients, while McClure and colleagues found a minimum clinically important difference of 0.063 (mean) or 0.064 (median) for the EQ-5D-5L using simulated data. 33 To be conservative, we powered the trial to detect the lowest estimate (0.06), assuming a standard deviation (SD) of 0.20. 19 Accounting for loss to follow-up of 20%, we needed to randomise 586 participants to have 90% power with 5% significance (two sided).

Although this was an individually randomised trial, there was the potential for clustering within the intervention arm by yoga class. Assuming an intraclass correlation coefficient of 0.04 and average class size of 12 in the intervention arm, with the proposed sample size of 586, we would have still retained 84% power to detect the same magnitude of effect (ceteris paribus). In this calculation, we considered the level of clustering at the yoga class level rather than at the level of the yoga teacher, since we believed this to be the most influential level of clustering. Accounting for potential clustering within the intervention arm only leads to small reductions in power, which could potentially be recovered in the analysis of the repeated measures, adjusting for baseline value, which was not accounted for in the original sample size calculation.

Revised

Trial recruitment commenced prior to the COVID-19 pandemic. The onset of the pandemic caused challenges to recruitment. Considering these challenges, in October 2021, the sample size calculation was revisited. At this point, 454 participants had been randomised (240 intervention; 214 usual care) and the trial team were considering whether to apply for a funded 7-month extension to give extra time to pursue the original recruitment target of 586.

The original calculation did not account for the correlation of outcome with baseline, as this provides the most conservative target. At the start of the trial, there was little to base the estimate of the correlation on, and assuming a correlation that is larger than is eventually observed can lead to an underpowered trial. However, in practice, it was always planned to include the baseline EQ-5D-5L utility index score as a covariate in the primary analysis model, and this provides gains in power.

Assuming all other parameters remained the same, with 454 participants, we would have had 81% power to detect the same effect size. With a higher rate of follow-up that was consistent with what had been observed up until this time (90%), this would have increased the power to 86%. Accounting for a higher rate of follow-up and also correlation of the outcome with baseline of at least 0.35, then we would have retained 90% power. An interim calculation of the correlation between baseline and 12-month EQ-5D-5L utility index score based on data received and processed up to 15 October 2021 (n = 86 observations) was 0.67 (95% CI 0.54 to 0.77), and at that time we had observed a response rate of 90% at 3 months. Therefore, we were confident that we would be able to detect a clinically important difference with close to or greater than 90% power with 454 participants randomised. Hence, it was agreed to cease recruitment at 454 participants and not apply for an extension.

Recruitment

The flow of participants through the trial is detailed in three CONSORT diagrams depicting (1) recruitment of participants in the two pilot phase waves; (2) recruitment of participants in the two main phase waves; and (3) the retention and follow-up of participants post randomisation. The total number of participants approached and randomised are reported, with reasons for non-participation (ineligibility or non-consenting) provided where available.

Baseline characteristics of randomised and analysed participants

All participant baseline data are summarised descriptively by the trial arm, both as randomised and as included in the primary analysis. No formal statistical comparisons were undertaken on baseline data between the arms. Continuous measures are reported as means and SDs (and/or median, interquartile range and range), and categorical data are reported as counts and percentages.

Withdrawals and follow-up

Response rates to the participant questionnaires at 3, 6 and 12 months are presented by the trial arm, and overall, as number expected (i.e. not withdrawn before the time point), they are presented as a percentage of number randomised, number returned (% of expected, % of randomised) and median (interquartile range) time to completion in days from due date. Mode of completion (postal or over the telephone) is summarised for the 6-month time point. Reasons for non-completion are provided where known. Type and timing of withdrawals are presented overall and by the randomised group.

All outcome data are summarised descriptively by the randomised group and overall at each time point.

Interim pilot phase

Descriptive statistics only were used to evaluate the progression criteria for the internal pilot sites. Data from participants in the internal pilot were included in the main study analysis.

Primary outcome analysis

The EQ-5D-5L utility index score was modelled using a linear mixed-effects model to incorporate the outcome at all post-randomisation time points. The model adjusted for baseline EQ-5D-5L utility index score, time point, trial arm and trial arm by time interaction as fixed effects. Participant and site were included as random effects, to account for the repeated measures of scores by participants over time, and the clustering of participants within sites. The different covariance structures for repeated measurements that are available as part of the analysis software were applied to the model. The most appropriate pattern was used for the final model based on diagnostics, including Akaike’s information criterion (smaller values are preferred). The treatment effect in the form of the adjusted MD in EQ-5D-5L utility index score was extracted with its 95% CI and p-value for each time point and overall. The overall difference between the two groups over the 12 months from randomisation was the primary end point, but differences at each time point were extracted for secondary investigations aimed at determining the potential pattern of improvement. Model coefficients for the covariates with 95% CIs are also presented to aid understanding of the fitted model. Participants were only included in the model if they had full data for the baseline covariates and outcome data for at least one post-randomisation time point; the model assumes any missing data were missing at random (MAR). Model assumptions were checked as follows: the normality of the standardised residuals was checked using Q–Q plot, and homoscedasticity was assessed by means of a scatter plot of the standardised residuals against fitted values.

Sensitivity analyses

Subgroup analysis

Secondary outcome analyses

Each of the following secondary outcomes was analysed using the same methods as described for the primary outcome, with baseline EQ-5D-5L utility index score swapped as a covariate for baseline value of the outcome:

• EQ-5D VAS

• GAD-7 score

• PHQ-8 score

• T-scores from each of the seven subscales of the PROMIS-29 v2.1, the PCS and MCS and the global item score

• UCLA-3 score

• ELSA single-item direct loneliness question.

The incidence of falls during the 12-month follow-up period was analysed by modelling the number of participant falls over the 12-month follow-up period with a mixed-effect negative binomial regression model, adjusting for the number of falls in the 3 months prior to baseline, and site as a random effect. The model includes an exposure variable for the number of months for which the participant provided falls data. The point estimate for the treatment effect in the form of an incidence rate ratio and its associated 95% CI and p-value is provided.

Adverse events

Adverse events for this trial are summarised descriptively, stratified by whether they are classified as serious or non-serious. Only events deemed at least possibly related to the study are reported. The number, type, outcome, action and relatedness of the events are summarised.

Design and processing of participant questionnaires

The participant postal questionnaires were designed using TeleForm software (version 11.2; Cardiff Software, Cambridge, UK). Specification questionnaires were populated with variable names and appropriate scoring. To maximise data quality, when questionnaires were returned to YTU (by post), key variables required for the statistical analysis were reviewed for completion and accuracy by a data administrator who resolved any queries by contacting the participant via telephone. After this initial check, all questionnaires were prepared for scanning by a data administrator using the TeleForm software. When a form would not scan, the data were manually entered. When a form was scanned, the data were then verified depending on what TeleForm identified as requiring correction. The verified data were then temporarily held in the Download Database and available for second checking. This involved each hard copy of all forms being compared against the entry stored in the Download Database and correcting the electronic data as necessary. All data were scanned, downloaded and second-checked in the Validate Database. The automated data validation was undertaken by the Data Manager who applied predetermined rules to check for agreed variables and whether the data were recorded correctly.

The participant electronic questionnaires were designed using Qualtrics. To maximise data quality, response requirements and validations were used for key variables required for the statistical analysis (e.g. to alert when a question had been missed or entered in the incorrect format). The data from each form submitted was regularly downloaded in batches from Qualtrics and imported into a Structured Query Language (SQL) database. Images of each form submitted were also downloaded in PDF format from Qualtrics. The data verification process was similar to that for the postal questionnaires in that the data were checked against the images and a code was run to apply the data rules and identify any anomalies.

Data that had been validated were made available, upon a formal request, to the trial statistician and health economist.

Plans to promote participant retention and complete follow-up

The following processes were implemented to promote participant retention and follow-up:

• Yoga teachers were asked to encourage regular class attendance and to contact participants who missed two consecutive classes without prior notification.

• Participants who withdrew from the yoga programme were asked if they would be happy to remain in the trial to provide follow-up data.

• Participants received a £5 shopping voucher or £5 cash with every follow-up questionnaire as a thank you for their continuing participation in the trial.

• A text message was sent to participants who consented to receive this form of communication 1–7 days before each follow-up questionnaire was sent, saying that they would soon receive a follow-up questionnaire and to complete and return it as soon as possible. The text message also acted as a prompt for the participant to inform the research team if they had moved address.

• A reminder letter was sent if the follow-up questionnaire had not been returned within 14–21 days.

• A researcher called the participant to complete the follow-up questionnaire over the telephone if it had not been returned within 28 days.

• If a questionnaire was returned incomplete or with errors, a researcher called the participant for clarification or completion of missing or invalid data.

• Participants were advised that they were able to call a member of the research team if they required assistance with completing a questionnaire.

• A newsletter containing information about trial progress and any relevant updates was sent out to participants via post or e-mail every 3–6 months.

• After completing the final follow-up questionnaire (12 months), all usual care participants received details of GYY or other suitable yoga classes that they could join on a self-pay basis. That they would do so was specified in the PIS.

• After randomisation to the main trial, participants allocated to the usual care group were randomised again to receive either the offer of a one-off yoga class, which would take place after the final follow-up or no offer (see substudy 4 in Chapter 6: Does offering a free yoga session to control participants after the 12-month follow-up assessment enhance retention and reduce contamination?).

• After randomisation to the main trial, participants allocated to the intervention group were randomised again to receive either a pen with their 3-month follow-up questionnaire or no pen (see substudy 3 in Chapter 6: Does sending a pen with a follow-up questionnaire enhance return rates?).

Participant withdrawal

Participants could withdraw from the trial at any point during the study. If a participant indicated that they wanted to withdraw from the study, they were asked whether they wished to withdraw from the intervention only (i.e. withdraw from the yoga programme if allocated to the intervention group) or withdraw fully from the study. When withdrawal was only from the intervention, follow-up data continued to be collected. Data provided by participants who fully withdrew were retained for analysis up to the point at which they withdrew. A member of the research team completed a change of status form for any participant who changed status (died, withdrew, lost to follow-up) during the trial.

Trial completion and exit

Participants completed the trial once they had returned a complete 12-month follow-up questionnaire. Participants exited the trial prematurely if they had fully withdrawn, were lost to follow-up or died.

Trial registration

The study was registered to the ISRCTN Registry on 18 March 2019 (reference number ISRCTN13567538).

Summary of changes to the protocol

Amendments to the trial protocol are summarised in Appendix 1.

Trial Management Group

The day-to-day management of the trial was overseen by the Trial Management Group, which met every 2 months. This group comprised the Chief Investigator, co-applicant co-investigators, Trial Manager, Trial Coordinators, Trial Support Officers, Data Manager, Data Administrator, Statistician, Health Economist, Process Evaluation Researcher, Clinician, Patient Representative and a BWY Representative (see Acknowledgements section for member details).

Trial Steering Committee

A TSC was appointed by the funding body to provide overall supervision for the trial and to advise on its continuation. It met every 6–9 months and included an independent chairperson and members as well as representatives of the Trial Management Group (the Chief Investigator, a Statistician and a Trial Coordinator). Membership is listed in the Acknowledgements section.

The TSC agreed to additionally take on the role of a data monitoring committee. A separate Data Monitoring Committee was considered unnecessary due to the low risk and open nature of the trial. Therefore, the TSC reviewed and discussed data pertaining to both participant safety and trial progress.

Baseline characteristics of randomised participants

The mean age of randomised participants was 73.5 years (range 65–99); 60.6% were female, and participants had a median of three chronic conditions (see Table 5). The most commonly reported conditions were cardiovascular diseases (n = 307 participants, 67.6%), which included participants who reported at least one of coronary heart disease, hypertension, heart failure or peripheral arterial disease (of which hypertension was the most prevalent) and arthritis (osteo or rheumatoid arthritis, n = 242, 53.3%) (see Table 6). The intervention and usual care groups were reasonably comparable in terms of baseline characteristics, except that there was a slightly higher proportion of females in the intervention group (64.2%) than usual care (56.5%).

Participants were asked at baseline about their expectations and preferences in relation to the health care offered in the GYY trial (see Table 7). Half of the respondents (n = 235, 52.3%) thought that usual care would be fairly or very effective at improving their quality of life, and a slightly higher proportion (n = 277, 61.0%) thought the GYY programme would be fairly or very effective at improving their quality of life. Given the choice, three-quarters (n = 339, 74.7%) said they would prefer to be allocated to the intervention group rather than usual care alone. Most of the rest had no preference (n = 103, 22.7%), and only a small number preferred usual care (n = 12, 2.6%).

The baseline values of the primary and secondary outcome measures are summarised in Table 8 and are reasonably well-balanced between groups.

Withdrawals and follow-up

Participant follow-up was completed in October 2022.

In total, we became aware of seven deaths [2 (0.8%) in the intervention group and 5 (2.3%) in the usual care group]. Six of these occurred within the 12 months from randomisation, and one just beyond this. For one death that occurred within 12 months, we only became aware of the event after their 12-month questionnaire was sent out.

A further 36 participants [14 (5.9%) intervention participants and 22 (10.5%) usual care participants] withdrew from follow-up data collection during the trial; 15 before month 3, 12 between months 3 and 6, 8 between months 6 and 12 and 1 just beyond 12 months (participants contacted the research team upon receipt of their 12-month postal questionnaire to say they were unable to complete the questionnaire due to ill health).

The overall follow-up rates for the 454 randomised participants were 91.0% at month 3, 87.4% at month 6 and 85.9% at month 12 (see Table 9). At all time points, the return rate was slightly higher (by approximately 5 percentage points) in the intervention group than in the usual care group. Median time to completion was 7 days from the due date at both months 3 and 6 and 10 days at month 12 and was similar for the two groups at each time point.

At the 6-month time point, 70 (17.6%) of the questionnaires were completed over the phone with a researcher rather than on paper and returned by post. This was when COVID-19 restrictions prevented researchers from being in the office to facilitate the mailing and return of postal questionnaires.

Raw scores

Summaries of the EQ-5D-5L utility index score by trial arm and time point are presented in Table 10. At each time point, mean scores are slightly higher in the intervention arm than in the usual care arm. Overall, at baseline, the mean score was 0.739 (SD 0.169) and decreased over time to 0.707 (SD 0.214) at 12 months.

The correlation between baseline EQ-5D-5L utility index score and scores at the follow-up time points is: 3 months 0.72 (95% CI 0.67 to 0.77), 6 months 0.63 (95% CI 0.57 to 0.69) and 12 months 0.59 (95% CI 0.52 to 0.65).

Baseline characteristics of participants included in primary analysis

The primary analysis included participants with a valid EQ-5D-5L utility index score at baseline and at least one post-randomisation time point (n = 422, 93.0%; intervention n = 227, 94.6%; usual care n = 195, 91.1%). The baseline characteristics of these participants are included in Tables 11–14; these are very similar to the randomised population, which indicates that there is little evidence that loss to follow-up has introduced attrition or selection bias.

Primary end-point analysis

There was no evidence of a statistically or clinically significant difference in EQ-5D-5L utility index score between the intervention and usual care arms over 12 months, with an adjusted MD of 0.02 in favour of the intervention group (95% CI −0.006 to 0.045, p = 0.14). The predicted means and associated 95% CIs over time are presented in Table 15 and displayed in Figure 6, by group.

FIGURE 6.

Adjusted mean EQ-5D-5L utility index scores (with 95% CIs) for primary analysis over time by randomised group.

Different covariance structures were applied to the model, and the Akaike information criterions (AICs) were compared. An unstructured pattern that models all variances and covariances separately was used in the final model, as this resulted in the lowest AIC.

Model fit diagnostics indicated that the standardised residuals demonstrated only a minor deviation from normality and were uniform against fitted values; therefore untransformed values were used in analyses.

Model coefficients for the covariates with 95% CIs are provided as software output in Appendix 2 to aid understanding of the fitted model, along with summaries of the EQ-5D-5L index value by trial site and time point to assess variation between sites (see Table 45).

Adjusting for other covariates (sensitivity 1)

Results were very similar when the primary analysis was repeated with age, gender and adapted Bayliss score additionally adjusted for as fixed effects (see Table 15).

Clustering by yoga teacher (sensitivity 2)

Nineteen yoga courses were delivered within the trial by 12 yoga teachers (1 teacher delivered 3 courses, 5 teachers delivered 2 courses each and 6 teachers delivered 1 course each). Analyses to account for possible clustering by yoga teacher were undertaken by including the intended yoga teacher as a random effect instead of site in the primary analysis model; results were virtually unchanged (see Table 15).

Compliance with random allocation and treatment received

One participant in the usual care group was invited to attend classes in error; they attended eight sessions, including five of the first six sessions.

A summary of attendance at weekly GYY sessions for intervention participants is presented in Table 16. Among the intervention group, 222 (92.5%) participants attended at least 1 GYY class, while 53 (22.1%) attended all 12 (see Figure 7). The mean number of sessions attended among all randomised intervention participants was 8.8 (SD 3.7, median 10) and 9.6 (SD 2.8, median 11) among those who attended at least one. Eighty per cent (n = 192) of participants attended at least six sessions, including at least three of the first six (see Table 17).

FIGURE 7.

Number of sessions attended by GYY intervention group participants.

On average, the first class in a course took place 18.2 days (SD 2.7, median 19, range 13–21) after the participant was randomised, and classes were scheduled a median of 7 days apart (range 7–28; longer intervals tended to be due to the Christmas period) (see Report Supplementary Material 3).

Three CACE analyses for the primary outcome were undertaken to explore the impact of non-compliance on treatment effect estimates, with compliance defined as:

• Attendance at one yoga session or more (n = 222 intervention participants, 92.5%; n = 1 usual care participant, 0.5%). The CACE estimate of the treatment effect is a difference of 0.025 at 12 months in favour of the intervention group (95% CI −0.002 to 0.052, p = 0.07). This difference is larger than the ITT estimate [The CACE analysis is not directly comparable with the primary ITT analysis as the CACE analysis cannot take account of the repeated measures for the EQ-5D-5L utility index score at 3, 6 and 12 months; it simply considers the difference at 12 months. Therefore, we conducted a linear regression with 12-month EQ-5D-5L utility index score as the outcome variable, adjusting for baseline score and gender with robust standard errors to account for clustering within site], but neither the treatment effect nor the upper 95% CI limit exceeds the clinically meaningful difference of 0.06.

• Attendance of at least three of the first six sessions and at least three other sessions (n = 192 intervention participants, 80.0%; n = 1 usual care participant, 0.5%). The CACE estimate of the treatment effect is a difference of 0.029 at 12 months in favour of the intervention group (95% CI −0.002 to 0.059, p = 0.06). This difference is larger than the ITT estimate [The CACE analysis is not directly comparable with the primary ITT analysis as the CACE analysis cannot take account of the repeated measures for the EQ-5D-5L utility index score at 3, 6 and 12 months; it simply considers the difference at 12 months. Therefore, we conducted a linear regression with 12-month EQ-5D-5L utility index score as the outcome variable, adjusting for baseline score and gender with robust standard errors to account for clustering within site.], but neither the treatment effect nor the upper 95% CI limit exceeds the clinically meaningful difference of 0.06.

• Number of sessions attended in its continuous form (intervention: mean 8.8, SD 3.7, one usual care participant attended eight sessions). The CACE estimate was a difference of 0.003 per session (95% CI −0.000 to 0.005, p = 0.07), indicating a very small additional benefit of the intervention for each session attended.

Yoga practices

Participant self-reported data on attendance at yoga classes and home yoga practice at 3-, 6- and 12-month follow-ups are presented in Tables 18–20, respectively.

The self-reported data relating to attendance at GYY classes at the 3-month follow-up match well with the attendance register data (see Table 18). All 214 participants in the intervention group who self-reported as having attended a GYY session were recorded as having attended at least one session (8 participants were recorded on the attendance registers but did not return a 3-month questionnaire). The one usual care participant who self-reported as attending was the person we expected. Only a small number of participants in both groups reported attending other (non-trial) yoga sessions. Most of the intervention group reported practising yoga at home, which could include as part of the GYY programme (n = 185, 82.6%), but only a small number of usual care participants (n = 6, 3.2%). Where undertaken, participants in the intervention group did twice as many home yoga sessions as usual care participants (median 4 vs. 2), though these sessions tended to last for a similar length of time (median 15 minutes).

At 6 months, 72 (33.3%) intervention participants reported that they had attended a GYY session in the previous 3 months, all of whom were confirmed to have attended at least one session according to the class registers (see Table 19). It is likely the other attendees had completed their sessions more than 3 months prior to completing this follow-up questionnaire. Three usual care participants reported having attended a GYY session, though none of these were present on the class registers. GYY classes are available to the public through the BWY, so it is possible these participants had sought out and attended a GYY session not delivered as part of the trial, and therefore this would not be captured as part of our evaluation. The proportion of participants reporting home yoga practices decreased relative to month 3 (in accordance with the cessation of the GYY course) in the intervention group (as did the median number of home yoga sessions from 4 to 3) but increased very slightly in the usual care group.

At 12 months, only a fifth of participants reported having attended GYY classes in the previous 6 months (n = 41, 19.3%) and two usual care participants (again, not participants who were on a class register) (see Table 20). As at 6 months, the proportion reporting home yoga practice decreased in the intervention group relative to the previous follow-up time point (to just less than half) but increased in the usual care group (9.6%). Participants in the intervention group reported doing a median of three home yoga sessions a week lasting a median of 15 minutes, while for the usual care group, this was a median of two sessions a week for 10 minutes.

Intervention fidelity

All yoga teachers submitted a course plan to the yoga consultant for pre-approval ahead of their first class. Each teacher received timely feedback on their plan. The feedback was mostly positive, and all plans met the assessment criteria and were therefore deemed appropriate for delivery.

Each yoga teacher underwent an observation of one of their trial classes by one of the yoga consultants. A fidelity check assessment form was completed for each observation. All yoga teachers passed all aspects of the fidelity check assessment criteria.

Intended mode of delivery of Gentle Years Yoga

More participants were randomised in a site intended for online GYY delivery (61.9% across 12 sites; intervention group n = 144, 60.0%; usual care group n = 137, 64.0%) than face to face (38.1% across 7 sites; intervention group n = 96, 40.0%; usual care group n = 77, 36.0%). A subgroup analysis was conducted in which the primary analysis was repeated, including, as a fixed effect, an indicator for this factor plus an interaction with trial arm. There was no evidence of an interaction between trial arm and intended mode of delivery (interaction effect 0.007, 95% CI −0.042 to 0.057, p = 0.77).

EuroQol-5 Dimensions, five-level version utility index scores at the secondary time points

Adjusted EQ-5D-5L utility index score means and group differences from the primary analysis model are presented in Table 22 and displayed in Figure 6. There was no evidence of a statistically significant difference at 3, 6 or 12 months, and none of the CIs for the differences contained the clinically meaningful difference of 0.06.

EuroQol-5 Dimensions, five-level version visual analogue scale

Raw EQ-5D-5L VAS scores are summarised in Table 21. Adjusted means and group differences are presented in Table 22. The analysis included data from 423 participants (intervention n = 227, 94.6%; usual care n = 196, 91.6%). There was no evidence of a statistically significant difference at any time point.

Generalised Anxiety Disorder-7

Raw GAD-7 scores are summarised in Table 21. Adjusted means and group differences are presented in Table 22. The analysis included data from 420 participants (intervention n = 227, 94.6%; usual care n = 193, 90.2%). There was no evidence of a statistically significant difference at any time point.

Patient Health Questionnaire-8

Raw PHQ-8 scores are summarised in Table 21. Adjusted means and group differences are presented in Table 22. The analysis included data from 419 participants (intervention n = 227, 94.6%; usual care n = 192, 89.7%). There was no evidence of a statistically significant difference at any time point.

University of California, Los Angeles-3 loneliness

Raw UCLA-3 scores are summarised in Table 21. Adjusted means and group differences are presented in Table 22. The analysis included data from 419 participants (intervention n = 227, 94.6%; usual care n = 192, 89.7%). There was no evidence of a statistically significant difference at any time point.

English Longitudinal Study of Ageing single-item direct loneliness question

Raw ELSA single-item direct loneliness question scores are summarised in Table 21. Adjusted means and group differences are presented in Table 22. The analysis included data from 421 participants (intervention n = 227, 94.6%; usual care n = 194, 90.7%). There was no evidence of a statistically significant difference at any time point.

Patient-Reported Outcomes Measurement Information System-29

Raw PROMIS-29 scores are summarised in Table 23. Adjusted means and group differences are presented in Table 24. These analyses included data from between 419 and 421 participants. There was evidence of a statistically significant difference in the T-score for the pain interference subscale of the PROMIS-29 at 3 months (−1.44, 95% CI −2.63 to −0.26; p = 0.02) and over the 12 months (−1.14, 95% CI −2.24 to −0.04; p = 0.04) and in the global (pain intensity) PROMIS-29 item at 12 months (−0.45, 95% CI −0.83 to −0.08; p = 0.02) and over the 12 months (−0.32, 95% CI −0.61 to −0.04; p = 0.03). Differences favoured the intervention. Otherwise, no statistically significant differences were observed.

Falls

A total of 421 participants responded to the question asking whether they had had a fall in the previous 3 or 6 months on at least one of the post-randomisation questionnaires, of which 112 (26.6%) said they had [60/227 (26.4%) in the intervention group and 52/194 (26.8%) in usual care].

A mean of 0.82 (SD 2.0, median 0, range 0–21) falls per person was reported over an average of 10.5 (SD 3.6, median 12) months [intervention 0.91 (SD 2.1, median 0, range 0–21) falls over 10.8 (SD 3.2, median 12) months; usual care 0.71 (SD 1.9, median 0, range 0–15) falls over 10.2 (SD 3.9, median 12) months]. There was no evidence of a statistically significant difference in the rate of falls reported over the 12 months of follow-up (incidence rate ratio 1.38, 95% CI 0.95 to 2.01, p = 0.09).

Within-trial economic evaluation

A within-trial economic evaluation was undertaken using a cost–utility analysis to evaluate the intervention relative to usual care in terms of the incremental cost per QALY. The perspective of the UK NHS and personal social services (PSS) was taken for the evaluation on an ITT basis. Analyses used a 12-month time horizon; therefore discounting of health benefits and costs was not necessary. Multiple imputation was undertaken in the base-case analysis, with a complete-case analysis (CCA) explored also as a SA. Further sensitivity analyses were undertaken to investigate the effect of varying assumptions made in the analysis in terms of the cost-effectiveness findings. Personal costs paid by participants regarding private health care were also explored as a SA. All analyses were conducted in line with NICE recommendations as far as possible35 and taking into consideration the reporting requirements of the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) www.ispor.org/TaskForces/EconomicPubGuidelines.asp.

Economic data collection

Health outcome and healthcare resource use data were collected during the 12-month study follow-up period. Specifically, healthcare resource use and EQ-5D-5L data were obtained at baseline, 3, 6 and 12 months. Medication prescription data were collected via GP practices for a sample of participants over a period of 3 months prior to randomisation and for the 12 months following randomisation. Information regarding the intervention cost was obtained from the study team (i.e. invoices, etc.).

Health-related quality of life

The health-related quality-of-life utility data for the economic evaluation comprised the EQ-5D-5L, as described previously in Chapter 2, which was used to elicit patient utility index scores at the different time points in order to estimate QALYs. QALYs are able to combine the effect of a treatment on both the quantity and quality of life for an individual, where one QALY is equivalent to a year spent in full health. Participants’ responses to the EQ-5D-5L questions were used to derive utilities, with the EQ-5D-5L scored according to the User Guide36 and the mapping function developed by van Hout et al. 23 used to calculate utility index scores, as recommended by NICE. 35 QALYs were estimated for each participant based on utility index scores at the different time points (i.e. at baseline, 3, 6 and 12 months) using the area under the curve (AUC) approach; this assumes linear interpolation between measurement points, following the trapezium rule. 37 In the primary analysis, the MD in QALYs gained over the 12-month period between the groups was adjusted for baseline utility only. An additional SA was further adjusted for baseline utility, age, gender and study site.

The EQ-5D-5L utility index scores are provided for the two groups, with movements between levels examined for each domain by trial arm. Descriptive statistics of the utilities for both groups and the difference in utilities between the groups at each time point have been presented. The overall difference in EQ-5D-5L utility index scores between the two groups has been examined through regression methods, consistent with the statistical analysis.

Healthcare resource use

Healthcare resource use data were collected for each participant, specifically resource use within primary care and the community (i.e. appointments with a GP, nurse, physiotherapist, mental health services) and the hospital setting [i.e. hospital outpatient attendances, day-case visits, inpatient stays and accident and emergency attendances). Information on private treatments was also collected (i.e. private GP visits, physiotherapists or mental health service specialists) to feed into a SA. The resource use data were collected during the trial via self-reported participant questionnaires at baseline, 3, 6 and 12 months. Healthcare resource use results are presented for both arms in terms of mean value, SD and MD (with 95% CIs) between the groups.

Costs

Unit costs for the resources used were obtained from established costing sources such as NHS Reference Costs38 and Personal Social Services Research Unit (PSSRU) Unit Costs of Health and Social Care. 39 No data were collected as to how privately funded health care was paid for, for example, private plan, employer plan or pay as you go. For the purposes of costing, we assumed participants paid out of pocket on a pay-as-you-go basis. Private healthcare costs were sourced from published online sources. The unit cost for all items used in the analysis is summarised in Table 26. All costs were evaluated in Great British pounds (£) for 2020–1. In cases where costs were sourced from previously published data, costs were inflated to the appropriate year figures. The mean cost per participant for each resource use item is provided by trial arm, with costs calculated by multiplying the resources used by their corresponding unit costs. For participants’ responses to accident and emergency visits that involved overnight stays, the corresponding cost was calculated by multiplying the number of visits to accident and emergency by the cost of a visit to accident and emergency plus multiplying the number of nights they spent in hospital (as part of this visit) by the cost of an inpatient night in hospital.

Medications data were collected for a sample of participants across three sites using prescription data collected from GP practices. This sample comprised all participants in the first pilot phase who remained within the trial at 12 months post randomisation. These data were used to generate an estimate of the average medication cost per participant; all prescribed medications were recorded for the 3-month period prior to randomisation and for the 12-month period post randomisation. For each participant in the sample, information was collected regarding the medication name, dose, quantity taken, frequency (i.e. once daily, twice weekly, etc.), number of days taken and whether it was a repeat prescription. Unit costs were attached to the medications using costs sourced from the British National Formulary (BNF)40 and NHS Drug Tariff. 41 Costs were applied using the drug that matched as closely as was available from BNF, and where there was a discrepancy between the information reported in the frequency and quantity, the more conservative option was selected. For participants who had not had medication data collected for them (i.e. they had missing medication data), the average medication cost for the relevant group (i.e. GYY or usual care) was applied.

The total cost for each participant comprised the following: healthcare resource use, medication cost and intervention cost (for those who received the intervention). Additional information regarding the GYY intervention is provided in the ‘Costing the intervention’ section below.

Costing the intervention

The cost of the intervention comprised (1) the cost of training the yoga teachers to undertake the GYY classes and (2) the cost of running the GYY classes, which included the cost of providing participants with the equipment and home practice sheets provided as part of the intervention. The intervention cost is described further below.

Missing data

Missing data are likely to arise in trial-based economic analyses involving participant-level data. This is due to the summative nature of the data; the total cost is comprised of several cost items over different time points, whereby if one item is missing, the total cost is then considered missing. Similarly, QALYs can be calculated only when all EQ-5D-5L questionnaires have been completed fully at all time points. The use of CCA, that is, including only participants with complete data on costs and EQ-5D-5L, can introduce bias unless the data are missing completely at random. An approach used to deal with missing data in economic evaluations in clinical trials is multiple imputation48 which estimates plausible, unbiased values for the missing data. This is based on the assumption of the data being MAR, that is, the observed data can be used to predict the missing values, but the missingness of the data is not associated with its value. We explored the validity of this assumption through the exploration of missing economic data at all follow-up points for both study groups and missing data patterns. Logistic regression was used to investigate which factors were associated with the missingness of data. Our findings indicated the MAR assumption to be reasonable; hence, multiple imputation with chained equations was used for the base-case analysis, with predicted mean matching on QALY and cost estimates, thereby generating plausible values from the imputation.

The imputation model included age, gender, study site, utilities (at baseline, 3, 6 and 12 months) and total costs at the resource use level. The base-case analysis undertook 60 imputations due to the finding of 58% missing data and following guidance that the number of imputed data sets should be similar to the proportion of incomplete cases. 49 All sensitivity analyses also followed this guidance. Graphical plots compared the distributions of the observed and imputed data to check whether the imputed data resembled the observed data. The imputed data sets were combined using Rubin’s rules50 to generate mean QALY and cost estimates. Usual imputation methods were followed for participants who died during the trial, where they had any missing data prior to their death. For data that would have been obtained after their death, zero QALYs and costs were assumed.

The base-case analysis used the multiply imputed data set with a CCA (i.e. includes only participants with observed data for all utilities and costs) undertaken as a SA. This enabled an alternative missing data assumption to be explored, as CCA relies on the data being not MAR. An initial exploration of the data was conducted using available case analysis.

Incremental analysis

Costs and QALYs are presented for the two groups in terms of mean value and SD and MD (with 95% CIs) between the groups. Findings are presented using ICERs, where appropriate, with standard decision rules followed when evaluating the two options. The incremental cost per QALY gained was determined based on the following:

where incremental costs and incremental effects are shown by ΔC and ΔE, respectively.

The ICER is compared to a willingness-to–pay (WTP) threshold, that is, the amount that a decision-maker is willing to pay for an additional QALY. The threshold of £20,000, following NICE recommendations, has been used for the analyses;35 where interventions have an ICER below £20,000, they would generally be deemed to be cost-effective. The results are also presented in terms of net monetary benefit (NMB),51 thereby representing the intervention in monetary terms based on a WTP threshold λ. Where the mean incremental NMB (i.e. the difference in NMB between the intervention and usual care) is positive, the intervention is found to be cost-effective at the given threshold.

The cost–utility analysis produced estimates of the MDs in costs and effects using seemingly unrelated regression equations, with 95% CIs estimated using bias-corrected and accelerated (BCA) bootstrap methods. Bootstrapping was conducted for 10,000 replications, producing 10,000 estimates of incremental costs and incremental effects. The 10,000 point estimates generated from the analyses are plotted as a scatterplot on the cost-effectiveness plane, across four quadrants, as summarised below:

1. North-east quadrant: the intervention is more costly and more effective than usual care; an ICER is estimated to evaluate whether the incremental effects (i.e. QALYs here) are worth the incremental costs.

2. South-east quadrant: the intervention is cost saving and more effective than usual care; it therefore dominates usual care.

3. South-west quadrant: the intervention is cost saving but less effective than usual care; an ICER is estimated to determine the cost per QALY lost.

4. North-west quadrant: the intervention is more costly and less effective than usual care; it is therefore dominated by usual care.

Cost-effectiveness acceptability curves (CEACs) were used to investigate the probability of the intervention being cost-effective52 compared with usual care at different WTP threshold possibilities faced by decision-makers.

Regression methods have been used to account for differences in stratification or prognostic variables and other sources of heterogeneity. QALY data were adjusted for baseline EQ-5D-5L scores, age, gender and trial site. Differences between the groups were found to be statistically significant if p < 0.05 and are presented alongside CIs around the differences in costs and outcomes. Analyses were undertaken using Stata release 17. 34

Sensitivity analysis

The uncertainty around the cost-effectiveness findings was explored by sensitivity analyses to investigate the robustness of the findings to assumptions made in the base-case analysis. Specifically, the following sensitivity analyses have been undertaken:

1. CCA, as an alternative to the use of multiple imputation which dealt with missing data

2. inclusion of personal expenses for private health care

3. face-to-face yoga classes only (i.e. no online classes included)

4. online yoga classes only (i.e. no face-to-face classes included)

5. medication costs removed

6. age and gender removed as covariates.

Cost–consequences analysis

A cost–consequences analysis was also undertaken to present disaggregated costs alongside the full range of outcomes. The analysis summarises resource use captured during the trial and the costs associated with these for each of the trial arms, as well as the consequences of the trial. As per the cost–utility analysis, the cost–consequence analysis was conducted on an ITT basis, following the 12-month time horizon for the trial and using Stata release 17. 34

For each resource use item, the mean resource use was calculated according to the appropriate unit (e.g. mean number of visits/appointments). To calculate the incremental difference for each individual item of resource use between the two trial arms, the values for participants allocated to the usual care group were subtracted from the values of participants allocated to the intervention (GYY) group. 53 Mean unit costs were calculated by resource item for each of the trial arms by multiplying the resource by an appropriate unit cost, with resource use costs summed by type, for example, primary care. Multiple imputation was not undertaken for the purposes of the cost–consequence analysis, as the model serves only to highlight the likely resource use and the average costs per resource, alongside the consequences (outcomes) of the trial. The consequences presented in Table 35 reflect the outcomes of the GYY trial, and full details as to how these were calculated are presented in Chapter 2.

Participant population and missing data

As previously reported, overall, seven participants died during the study period: 2/240 (1%) in the intervention group and 5/214 (2%) in the usual care group. A CCA was undertaken (SA1) that included participants with complete data on all EQ-5D-5L and cost items over the 12-month study period. In total, 192 (42.3%) participants had complete economic data: 105 in the intervention group and 87 in the usual care group. Therefore, overall, 262 (57.7%) participants had missing economic data. Details of the completion of data at the data collection points in the trial can be seen in Appendix 3 (see Table 46). At baseline, 72.9% participants in the intervention group and 77.1% in the usual care group had complete data, which decreased at 3 months and further by 6 months for the two groups. Completion levels increased by 12 months, however, to 70.8% in the intervention group and 67.8% in the usual care group (see Appendix 6, Table 49). The pattern of missing data for both cost and EQ-5D-5L data was not monotonic, that is, there were some participants who had missing data at 6 months but who had observed data at 12 months. The odds ratios (ORs) from a logistic regression for indicators of missing data, for QALYs and costs, on study group allocation and several baseline variables are presented in Appendix 4 (see Table 47). Age, site and allocation were found to be significantly associated with both missing cost data at 12 months and missing QALY data at 12 months. In addition, gender was significantly associated with missing QALY data at 12 months. When investigating observed values, observed QALYs at 3 and 6 months were found to be significant predictors of missing QALY data at 12 months and cost data at 12 months. Our findings therefore support the assumption of the data being MAR rather than missing completely at random. There is no way of definitively distinguishing whether the data are MAR rather than missing not at random (MNAR), so the best we can do is assess how reasonable the assumption is based on how likely it is that people would not want to provide data depending on what their response was. Data that are particularly sensitive or personal or that reflect an unfavourable opinion may be more likely to be MNAR, but arguably resource use data are not particularly sensitive. Missingness was therefore assumed to depend on baseline covariates (EQ-5D-5L at baseline, age and gender) and observed QALYs and costs. The fit of the imputation model was checked by comparing the distributions of the observed data with the imputed data for costs and utilities.

Health-related quality-of-life utility

The EQ-5D-5L was very well completed at baseline, with completion rates being 100% for intervention participants and 99.5% for usual care participants (only one participant missed one item) (see Appendix 5, Table 48). Completion rates decreased over time for the two groups, though they remained high at 12 months, at 87.9% and 83.2% for the intervention and usual care groups, respectively. Further information on the missing dimensions and completion of the EQ-5D-5L questionnaires is provided in Appendix 6 (see Table 49). A breakdown of the different EQ-5D-5L levels reported by participants over the 12-month follow-up period, according to dimension, group and time point is shown in Appendix 7 (see Table 50), for all available cases (i.e. not imputed). Mean (SD) VAS scores were similar at baseline, though slightly higher in the intervention group: 75.1 (18.2) for the intervention group versus 73.4 (17.6) in the usual care group. As presented previously (see Chapter 3, Table 21), mean VAS scores increased at 3 months, then decreased over time for both groups. The mean EQ-5D-5L utility index scores are presented in Table 27 for all time points, based on the available cases, alongside the MD between the groups (both unadjusted and adjusted for baseline utility). Utilities were slightly higher at baseline for participants in the intervention group (0.742) compared with those in the usual care group (0.736). The MD was not found to be statistically significant at any time point, irrespective of whether it was adjusted for baseline utility or not. The total QALYs gained over the duration of the trial are shown in Table 28 for all available cases, indicating a small gain in QALYs for the intervention group over the 12-month period. This increase in QALYs was not found to be statistically significant based on the non-imputed (all available cases) data set when adjusting for baseline utility only or when adjusting for baseline utility, age, gender and study site.

Healthcare resource use and costs

The healthcare resource use and associated costs for all available cases are summarised in Tables 29 and 30, respectively, according to item of resource and trial arm.

In terms of resource use outside of the hospital, although participants in the intervention group had, on average, fewer nurse visits at the GP practice, they had, on average, additional GP visits at the GP practice/home/phone, nurse visits by phone/home, physiotherapist visits (community-based) and mental health services (community-based). For hospital-based resource use, intervention group participants had, on average, fewer physiotherapist visits (hospital-based), hospital outpatient visits, inpatient nights in hospital and accident and emergency visits, though they had, on average, more day-case hospital visits and attendances with mental health services (hospital-based) than usual care participants over the study period.

Hospital-based services tended to be the major cost drivers, along with the cost of medications and the intervention, for participants in the intervention group. Based on the multiply imputed data set (see Table 31), the largest (mean; 95% CI) reductions in costs were seen for the intervention group versus usual care group for inpatient nights spent in hospital (−£139.55; −£80.22 to −£98.89), accident and emergency visits (−£25.89; −£73.32 to £21.54) and hospital outpatient attendances (−£11.30; −£46.32 to £23.73). In addition, medication costs were lower for the intervention group (−£68.90; −£77.19 to −£60.62). The most notable increases in costs for participants in the intervention group over the 12-month period were found for day-case visits (£85.03; £15.22 to £154.85) and mental health services (£32.88; £19.66 to £46.11).

Intervention cost

A total of 240 participants were randomised to receive the intervention, with the mean cost of the intervention estimated to be £187.49 per participant. The different cost components of the intervention can be seen in Table 32.

For the training of yoga teachers for intervention delivery, in PPW1, face-to-face training took place (six trainees), with all remaining training taking place online in the subsequent training workshops for PPW2 (six trainees), MPW1 (three trainees) and MPW2 (nine trainees, involving three half-day sessions and one 1-hour additional meeting). The total cost of training that took place over the study period was divided by the number of participants in the intervention group to generate a mean training cost per participant of £31.92.

The cost of running the GYY classes was estimated to be £142.25 for a course of 12 classes; some courses were delivered face to face, and some were delivered online; hence both delivery methods were considered in the cost estimations. Four face-to-face courses took place in PPW1, four online courses in PPW2 and three online courses in MPW1. In MPW2, there was a combination of (three) face to face and (five) online courses. In terms of the cost of liaising with participants prior to the course starting, in PPW2 a cost of £135.00 (i.e. by phone) was applied for four courses, in MPW1 £270.00 (i.e. by Zoom) was applied for three courses, and during MPW2 three phone (£135.00) and five Zoom (£270.00) costs were applied.

The cost of equipment used by participants was estimated to be £12.50 on average per participant, which considered that an additional charge was incurred for the participants who received their equipment via post, for those who undertook classes online (those who undertook face-to-face classes received their equipment at the classes, hence no postage charge). The cost of home practice sheets was £0.82 per participant.

The intervention cost was explored as part of the sensitivity analyses to investigate whether the mode of delivery (i.e. face to face vs. online) affected the cost-effectiveness findings. When only face-to-face classes were considered, the intervention cost became £175.44, and for online classes only, it became £195.52. The online classes had higher costs associated with them due to the costs of liaising beforehand with participants via the additional one-to-one session and also the video-conferencing subscription cost, plus the cost of equipment was slightly higher due to sending out via post; though the face-to-face classes did involve the room hire cost, refreshment expenses and some travel expenses, overall there was an increase in cost for the online classes when compared with face-to-face classes.

Cost–utility analysis and uncertainty

In the base-case analysis, participants in the intervention group incurred higher costs than the usual care group, at a mean (95% CI) cost of £1964.96 (£1882.38 to £2047.55) per participant in the intervention group and £1885.69 (£1795.53 to £1975.85) per participant in the usual care group (see Table 33). Additional QALYs were experienced by participants in the intervention group; mean (95% CI) QALYs were found to be 0.731 (0.724 to 0.738) in the intervention group versus 0.708 (0.700 to 0.716) in the usual care group (see Table 33).

The incremental analysis found that the intervention was associated with additional costs of £80.85 per participant, and generated an additional 0.0178 QALYs per participant, on average over the 12 months (see Table 34) in the base case. The resulting ICER was found to be £4546 per QALY, which falls under the WTP threshold used by NICE, of £20,000 per QALY. The intervention is therefore considered to be cost-effective. The incremental NMB, at the £20,000 per QALY threshold, was £274.85 (£268.29 to £281.41), also indicating that the intervention falls in the cost-effective region in comparison to usual care. This is shown by the incremental NMB associated with the intervention being positive; the NMB is positive if the ICER is below the WTP threshold, that is, indicating here that the benefit in QALYs gained exceeds the resources displaced if the intervention were to be adopted.

The point estimates of the bootstrap estimates presented in the cost-effectiveness plane in Figure 8 indicate there is uncertainty associated with the cost-effectiveness findings, as estimates populate all four quadrants of the plane. The estimates are more heavily populated in the north-east and south-east quadrants, that is, showing an improvement in QALYs from the intervention, but indicate quite a variation in findings in terms of the costs, with points below the x-axis showing a reduction in cost for the intervention, so a cost saving, and points above showing an increase in cost. The associated CEAC is shown in Figure 9, illustrating the probability of the intervention being cost-effective for different WTP thresholds. The probability of the intervention being cost-effective at a WTP threshold of £20,000 per QALY gained threshold is 79% (base case) and 85% at a £30,000 per QALY gained WTP threshold.

FIGURE 8.

Cost-effectiveness plane for the GYY intervention relative to usual care: base-case analysis (MI adjusted for all covariates).

FIGURE 9.

Cost-effectiveness acceptability curve for the GYY intervention relative to usual care: for the base-case analysis (MI adjusted for all covariates).

Sensitivity analysis

The results of the sensitivity analyses are presented in terms of the incremental mean costs, incremental mean QALYs, the cost-effectiveness finding (i.e. the ICER) and the probability of the intervention being cost-effective (based on a WTP threshold of £20,000 per QALY) in Table 34. The QALY findings remained relatively similar across the sensitivity analyses, with most notable differences seen for the CCA in SA1 for the complete analysis and in SA6 (where gender and age were no longer adjusted for in the model). The intervention group experienced additional QALYs, on average, with the difference found to be statistically significant for all analyses with the exception of SA1. In terms of the cost findings, all sensitivity analyses found there to be higher costs incurred by the intervention group, though these were of varying magnitude ranging between £28.31 (£24.35 to £32.27) for SA6 and £149.23 (£145.14 to £153.32) for SA5 (see Table 34). The cost-effectiveness results remained robust to the sensitivity analyses that were conducted; the ICER remained below the WTP threshold of £20,000 in all instances. The cost-effectiveness planes and CEACs for all sensitivity analyses can be seen in Appendix 8 (see Figures 12–17) and Appendix 9 (see Figures 18–23).

FIGURE 10.

Receiver operating characteristic curves for the PROMIS-29 physical and mental health scores against the external criterion of an improvement equating to the minimum clinically important difference of 0.06 in the EQ-5D-5L index value between baseline and month 12.

FIGURE 11.

Location of GYY-trained yoga teachers.

Cost–consequence analysis

Unit costs of all NHS and private healthcare resource items are presented in Table 35. The cost–consequence balance sheet summarises the results of the cost–consequences analysis (see Table 35). For each item of cost, the mean per participant resource use for both arms is displayed, alongside the incremental resource use. The mean cost per participant is also shown for each item according to trial arm, along with the incremental cost. In terms of the consequences, the outcomes of the GYY trial, as per Chapter 3, are listed. A descriptive summary of the costs and consequences is provided, and items are discussed in a disaggregated format.

Study design

The qualitative process evaluation involved interviews and observations with a diverse range of trial participants and yoga teachers at all sites, as well as additional stakeholders.

Sampling strategy

Our sampling strategy aimed to achieve a balance between spread of data and depth. We were also responsive to the study context, with additional data collection in response to our emerging analysis and study events in order to support meaning saturation. 55 Our sample of people recruited to interview was purposive,56 using the following criteria:

• Trial participants: a maximum variation sample of people in the yoga arm to include the breadth of demographic data relating to age, gender, ethnicity, Index of Multiple Deprivation (IMD) and number, severity and type of health condition, as well as site and mode of delivery. A convenience sample of people in usual care arm and trial decliners.

• Trial yoga teachers: a total population sample of all yoga teachers in the trial.

• Additional stakeholders: a snowball, emergent and theoretical sample of representatives of healthcare, academic and community-based groups involved in the development, delivery, evaluation, implementation and/or commissioning of yoga and related UK health programmes.

One trial class was observed for all but one of the yoga teachers. Sampling was monitored and discussed regularly at each team meeting.

Recruitment, consent and data collection

All interview participants were sent interview-specific study information. In relation to observations, all the trial participants attending a class and the yoga teacher were sent observation-specific study information. Interviews and observations took place throughout the trial, from January 2020 to April 2022. All process evaluation participants provided written informed consent for the interviews and/or observations, in addition to main trial consent.

The majority of interviews and observations were conducted remotely, with the exception of PPW1, which was undertaken face to face prior to COVID-19 restrictions. Some yoga participants and yoga teachers took part in a follow-up interview.

Interview topic guides were initially developed and agreed within the team and included the following:

• Trial participants: trial processes (e.g. ideas and/or concerns about randomisation and consent); the yoga intervention (e.g. expectations about yoga and/or concerns about impact on health and acceptability, delivery format, the timing, location, content, home practice); general health and well-being.

• Trial yoga teachers: views and experiences of yoga, trial processes, trial delivery and participant engagement.

• Additional stakeholders: views and experiences of yoga research, yoga provision for health and social prescribing.

The content of the interview was flexible to accommodate additional unanticipated areas, the developing analysis and in the case of follow-up interviews, what was known from the prior interview. Interviews and observations were conducted by an experienced qualitative researcher (LW), who, as necessitated by the purposive sampling strategy, was unblinded to allocation.

Qualitative data management and analysis

Interviews were, with consent, audio-recorded, transcribed verbatim and edited to ensure anonymity of respondents. Contemporaneous field notes from non-participant observation of yoga classes were edited to ensure anonymity of participants. The analysis was conducted according to the standard procedures of rigorous qualitative analysis,57 including pre-coding,58 open and focused coding, constant comparison, memoing,59,60 deviant case analysis61 and within- and across-participant thematic charting. 62 A sample of data was independently coded and cross-checked, to identify and reflect on similarities and differences. The majority of data were also analysed collectively in weekly ‘data clinics’ where the research team shared and exchanged interpretations of key issues emerging from the data. Analysis was conducted by LW/TR (both with social science backgrounds) who were joined for some discussions by GT. All names and other potentially identifying details have been changed to ensure the anonymity of participants.

Relationship between process evaluation and trial

The emergent analysis was regularly discussed at monthly TMGs and, where appropriate, informed changes to study processes. We produced two overviews with large numbers of detailed (minor) recommendations during the PPW1, in relation to the provision of face-to-face classes in March 2020 for initial pilot work, then following the shift to online provision, PPW2, in February 2021. Examples of recommendations ranged from supporting more effective class register practices to clarifying the concept, purpose and meaning of ‘usual care’ in participant trial information to enhancing class engagement during online class delivery. Nearly all recommendations were acted upon. This chapter will not focus on exploring the trial optimisation and fidelity work; instead it will offer an overview of some core issues that help enable reflection on the effectiveness outcomes of the trial.

Enrolment: clear willingness to engage with Gentle Years Yoga

Sixteen of the interviewed trial participants had not tried yoga before. Of these, some were familiar with what yoga involved through family members who practised, while others were unaware of the variety of yoga styles available. Trial participants’ preconceptions of yoga, based on media representation, experience or word of mouth, ranged from a physically challenging form of exercise targeting younger adults, women and the financially comfortable, through to a ‘soft’ form of stretching targeting physical and mental well-being. Several participants noted a generational knowledge of yoga among older adults of the 1960’ culture, associating its practice with Eastern spirituality. Common reasons for lack of engagement in yoga prior to the trial included barriers of cost, transport and class availability. Those participants with prior engagement mostly described practising yoga in their 30s or 50s, with most stopping due to the physical content or ability to self-mobilise to and from the floor precluding ongoing engagement with their community- or gym-based classes. Only one participant described a long-term, mat-based yoga practice.

Participants outlined four main reasons for joining the GYY trial: potential benefit to health, invitation-enabled opportunity for exercise, curiosity about yoga and, as with all studies, discussions of altruism (see Table 38). Several participants noted the encouragement of family members and their GPs to try the yoga classes. Notably, the majority of participant interviewees, those with prior experience of yoga and the yoga naïve, did not know yoga could be adapted either to a seated practice or to older adults prior to receiving their trial invitation. Despite some concerns about the potential physical challenges of yoga, the trial participants considered the GYY intervention would be within their physical capabilities, due to it being chair-based, and delivered by yoga instructors who had undertaken specialist training in yoga for older adults with health issues. As such, there were minimal concerns regarding the physical content of the trial classes. However, several noted social anxiety around meeting new people and potentially feeling embarrassed at learning a new activity in front of others.

Unsurprisingly for individuals joining a yoga trial, many held a belief, or were interested in exploring, the use of positive health behaviours as an effective and proactive means of maintaining independent living and improving health-related well-being. Self-management was regularly mentioned in preference to pharmaceutical management of chronic health conditions – ‘I hate taking tablets, I really do … I would choose an activity, you know, rather than [medication]’ (Joy, YPart-Int1) – with participants concerned regarding the side effects and potential dependency issues of certain pharmaceuticals. However, trial participants’ advocacy of proactive self-management came with the caveat of ensuring management choices were age-appropriate.

Integration: many already felt (relatively) active and healthy

The number of self-reported chronic health conditions ranged from two to eight per participant (mean 4.5). Fifteen interview participants reported physical health issues, predominantly related to mobility and pain, and eight participants reported both physical and mental health issues. Conditions were often co-related, for example, pain was commonly associated with sleep disturbance, fatigue and low mood. Six of the fifteen longitudinal interview yoga participants reported decreased health status over time, two due to accidents and one related to mental health. However, despite the high level of multimorbidity, most participants viewed their health as good. As Derek notes:

I’ve got high blood pressure too. That’s well controlled on medication. What else? I’ve been on antidepressants for about ten years now, and again that’s reasonably well controlled. [...] I’ve got some gastric reflux, which I’m on treatment for as well, but that’s not very severe.

YPart-Int 1

In his follow-up interview, Derek went on to note that ‘I’d describe myself as being fairly healthy despite all of the issues, high blood pressure, arthritis, reflux and that’ (YPart-Int2). Many did not consider multimorbidity indicative of poor or ill health. Certain health conditions, such as osteoarthritis, were expected among this older adult age group. The presence of multiple conditions was not seen as cumulative, with more conditions not necessarily equating to poorer health. Additionally, the diagnosis of a chronic health issue has often instigated adopting healthier lifestyle behaviours to minimise disease progression and maximise quality of life.

The impact of their multimorbidity was related to acute rather than chronic health status, and pharmaceutical management meant many health conditions had little to no impact on their activities of daily living or general well-being. As such, it was the day-to-day presentation, rather than the presence, of a health condition that determined its impact on the person. For example, when asked, Heather noted that her health is ‘not bad really, because, although I’ve got these conditions, they’re controlled’ (YPart-Int1). Several chronic health conditions, such as hypertension and coronary heart disease, were experienced as stable and/or asymptomatic conditions, often amendable to pharmaceutical management. These had little impact on participants’ daily life, to the point that many participants did not mention them in the interviews when prompted, even though they had indicated them on their screening questionnaires. Similarly, many participants had adjusted to living with chronic but stable pain, viewing it as a protective signal for managing their health condition. However, conditions with variable and/or intense symptom presentation, such as rheumatoid arthritis or clinical anxiety, were highly impactful, requiring adaptive and tailored management of both the acute and chronic phases of the condition. Notably, seven of the participants deemed their health to be above what they considered the level required to qualify for this trial. For example, Anne noted, while filling in the initial trial-related forms ‘I thought, “oh, they won’t want me, because there’s nothing wrong with me” sort of thing’ (YPart-Int1). All the trial participants interviewed queried their inclusion in the trial.

Workability: engaged well with online and face-to-face delivery

There was no clear delivery preference among participants. Some participants favoured online delivery, citing pragmatic reasons including the removal of geographical barriers around choosing a yoga teacher, the removal of travel barriers related to transport logistics, adverse weather and acute health status. This in turn enabled consistency of practice, independent of external or health factors. Several participants noted that the ability to look at the yoga teacher without being distracted by other participants enabled them to focus more on their practice. It also provided a level of anonymity and emotional security for those with emotional and mental health issues. Conversely, those preferring face-to-face delivery favoured the physical presence of class peers for social support and interaction and of the yoga teachers for receiving positional information and adaptations. On a pragmatic level, face-to-face classes were also preferred for those with poor internet connections impacting online classes.

Most yoga teachers indicated they adapted quickly to the onset of COVID-19 restrictions in March 2020, quickly moving their community yoga classes online where possible. Albeit that participants’ IT skills were variable, ranging from basic computer literacy to highly competent business users, the majority adapted quickly to the computer-based classes. Despite the lifting of restrictions, which allowed a return to face-to-face teaching, many yoga teachers took a pragmatic approach to class delivery and continued to offer online class options. They have continued online delivery of GYY and their general yoga provision in response to a stable client base who prefer the convenience and accessibility of being able to engage in yoga without having to travel outside their home. Centrally, whether delivered online or face to face, the GYY style of yoga was viewed as a suitable and safe form of physical movement for older adults with varying health issues, independent of delivery format. Despite restricted participant visibility and communication, yoga teachers had no safety concerns for online yoga beyond those inherent to face-to-face-delivered classes. The only potential safety concerns Kirsty had were the times when:

participants will want to do a little bit more than I am ready for them to do. But would that make much difference between in person and online? Not really, no. Because if they’re going to push themselves a bit more, they’re going to do it anyway, in person or online. I think that the responsiveness, when it’s in-person, it’s the two-way verbal. When it’s online, it’s one-way verbal, and I’m observing so much more.

YTeach-Int1

Safety reminders are a consistent and integral element of all yoga teaching. Both participants and yoga teachers identified two key safety messages of self-responsibility and non-harm, working within your personal limits, within a safe practice environment. The most common adverse impacts of a GYY class reported were those typically associated with exercise classes in general, including delayed-onset muscle soreness and upper body pain.

Finally, nearly all yoga teachers reported delivering class content which contained elements that differed from their 12-week plan submitted to and authorised by the yoga consultants prior to intervention delivery. The consistent reason given for this was that the class plans, written prior to meeting their participants, needed to be adapted once their health issues and physical abilities were known. Most notably, we also observed a variable focus on home practice content and the timely introduction of home practice sheets. Home practice should have been covered in the first segment (i.e. ‘housekeeping’) of each yoga class as per yoga teacher research training (Report Supplementary Material 4). Home practice was, in some cases, however only referred to in passing during social time after class, meaning only those who stayed on for this optional time received any home practice guidance.

Appraisal: experienced Gentle Years Yoga as gentle exercise (with mindful breathing)

The majority of participant interviewees viewed GYY as a form of gentle exercise. Only two participants found the class content physically challenging, with one further describing the level of physicality as appropriate both for beginners and for online delivery. However, while lacking overt physical challenge, participants commonly mentioned the challenge presented by coordination of the breath with movement. So, Michael, who did not find GYY ‘at all physically challenging’ did however find ‘the complexity of the movements’ challenging:

making certain you replicated what [trial YT] was doing, so there was a lot of thinking: ‘hang on, that’s out, then up and breathe out and breathe in’, you know [laughs].

YPart-Int1

This was noted as a unique feature differentiating yoga from exercise and provided a mental focus commonly described as inducing a calmness not found in other forms of physical activity. General participant feedback noted a lack of background information on defining what ‘yoga’ was in terms of basic concepts and the role of the breath within its practice.

Overall, the participant interviewees had mixed views on the role, place and target audience of the GYY form of yoga. Those in favour of the practice viewed GYY as filling a gap in the market for age-appropriate exercise, particularly for those with no previous experience of physical activity or those seeking an alternative to previous aerobic-based activities which had become too physically challenging. For these participants, GYY was viewed as both a physical and mental health tool, with the use of chairs and adaptability of the physical postures key to making the practice inclusive to all regardless of chronic and acute health status and physical ability. However, those with physical health issues found the repetitive movements and upper body focus of some GYY classes induced or increased musculoskeletal pain. Conversely, those with minimal health issues found the GYY classes lacked physical challenge beyond gentle stretching.

Dave: ‘I thought it complemented what I’m doing with my [other diagnosis-related exercise] class really. Although a little different I have to say. I find the yoga gentler; the [other exercise] class is more challenging physically’.

Interviewer: ‘Which do you prefer?’

Dave: ‘I prefer the challenging one. A lot of the stuff so far in terms of the yoga we’ve been sat down. Well I never sit down when I have my exercise class with [Diagnosis-group]. … No, it’s not physically demanding enough. For me’.

YPart-Int1

They often preferred their pre-existing physical routines to this new form of physical activity. Some questioned its potential to be beneficial as a form of exercise – with Walter noting that: ‘but it was very easy, I thought, I didn’t think that I was doing anything that would exercise me in any way … to me, exercise should be getting the heart rate up and sweating’ (YPart-Int1).

Several yoga teachers also noted that the GYY style of yoga may not be sufficiently challenging for more functionally able individuals they worked with. As Faye notes:

But our bodies age, and our bodies change, and it’s not a number. And there are some people who are very, very able. In my classes, in this trial, they’re in their 70s and 80s but they could be doing so much more. And so I feel sometimes I’m holding them back a little bit.

YTeach-Int1

Another yoga teacher, Jackie, noted that she felt ‘like I’ve got people that are more able than this course is designed for’ (YTeach-Int1). She describes her surprise at how her participants were ‘really keen, really able’ from the first class onwards. Some yoga teachers thought some participants would be better suited to attending general mat-based yoga classes. Several trial participants also queried the inclusion criteria of the trial, feeling that they should have been recruited based on health status rather than age. Others were uncomfortable being classified as an ‘older’ adult with multiple health issues. Participants in the younger end of the age range, and those who felt in good health, stated feeling out of place among participants older and in poorer health than them.

Integration: variable engagement with Gentle Years Yoga, social time and home practice

Participant interviewees held diverse views on the role of social time. Some described it as potentially beneficial, others as non-essential and independent of a GYY class. Irrespective of whether the classes were face to face or online, those in favour cited both social and educational benefits of social time as an opportunity to meet new people, gain peer support for those experiencing isolation and low mood and learn more about the practice of yoga. Notably, the value of social time was tied to the facilitation skill of inclusivity, independent of delivery format. Yoga teachers more adept at facilitation promoted inclusivity through moderating the group dynamic to ensure everyone had the opportunity to contribute to conversation if desired. However, a majority of participant interviewees had no interest in the social aspect of classes, and yoga teachers and participants noted that attendance was low across several classes. These participants commonly stated a lack of need or interest in socialising with a new group of people, noting they had joined the trial due to an interest in yoga not meeting new people. For example, Dot explained that ‘I don’t go for the social business’ (YPart-Int1). Some also noted the communication issues inherent to group conversation via an online platform, outlining that there could be limited opportunity to verbally socialise for those of a quieter nature.

Reported adherence to home practice during the 12-week intervention and into the follow-up period was also variable, ranging from zero through to daily practice. One of the factors in adherence was the level of integration of the home practice content into the group classes by the yoga teachers. Some participants expressed disappointment in the lack of focus on home practice by their yoga teachers. Some suggested more dynamic and progressive home practice options for those more physically capable, synchronising home practice with the progressive class content, including props and relaxation practices.

The primary facilitator of sustained self-practice was the perceived biopsychosocial benefit gained from practice. Those who continued to practice yoga at home into the follow-up period described a practice which had evolved over time. Initially basing their home practice on the content of the trial-provided home practice sheets, they gained a skillset from the 12-week intervention, enabling them to tailor the exercises, breathing and relaxation practices to their acute health needs, time commitments and/or personal preferences. For several participants, this skillset precluded the need for continued attendance at group-based classes, with a preference for self-directed practice at home. Conversely, the primary reason for a lack of home practice during or following the trial was a noted lack of benefit from, or interest in yoga, or a preference for pre-existing physical exercise routines. So Simon, who was ‘a bit lukewarm with the exercises’ initially said that he did not find home practice ‘motivating’ (YPart-Int1) and in the follow-up interview noted that ‘Well, I didn’t do very much home practice in the trial, to be honest’ (YPart-Int2). Many non-adherent participants also noted a lack of extrinsic motivation for self-practice, particularly among those experiencing low mood or mental health challenges. Finally, several participants with work or social demands cited the time cost of trying to fit home practice into an already busy lifestyle.

Appraisal: embodied impact (and future engagement) varied

The reported impact of GYY practice focuses around two main areas of improvements: physical health and emotional well-being (see Table 39). Tangible improvements in functional activities of daily living, such as dressing and housework, were also noted. Most notably, the impact of the GYY classes ranged from minimal to transformative. Some participants noted no impact of yoga on their health or lifestyle. This was primarily associated with describing a state of good health and physical activity when entering the trial, with the physical yoga content not being at a level capable of providing additional functional or sustained benefits. Irrespective of impact, some noted that taking part in the trial had stimulated an interest in engaging in other forms of yoga or physical activity, such as mat-based yoga, Tai Chi or more physically challenging forms of exercise.

In contrast, some participants described the transformative impact of the intervention had placed yoga as integral to their lives. For example, GYY has impacted Audrey at a biopsychosocial level, in a way she describes as ‘lifechanging’:

It is the holistic bit. It’s the mental, the physical, the intellectual, the social. […] It closes down negative thoughts but opens up possibilities, and that very much resonates with me because that’s the sort of person I am. It puts me in control. But it’s something about letting go of control as well, if that makes sense. It’s just okay to be me. And, yes it can be one of many things, but it is – it is lifechanging. That is without a doubt.

YPart-Int2

Continued engagement was aligned with initial and sustained biopsychosocial improvements in physical health, emotional well-being and independence over the 12-week intervention, with associated decreases in analgesic, opiate-based medication.

Some participants noticed physical and psychosocial improvements from within the first few classes. For example, after one class, Rose felt a notable impact:

Rose: ‘And my first thing that I felt came from it really, was that first night of doing yoga, after I’d done that first class, and it was very gentle, I actually didn’t have any pain at all going to bed for the first time in years’.

Interviewer: ‘After just one class?’

Rose: ‘After just one class, I couldn’t believe it. I was a little tiny bit stiff, in the more sensitive areas of the knee and things like that, but that wasn’t unbearable by any stretch of the imagination. So, I just thought, well, you know, this has got real good benefits. And it was just like flowing in water, really, because I really felt into it. Once I’d started, it was just like, you know, swimming. You just do it [laughter]’.

YPart-Int1

For others, the improvements occurred more gradually over the 3-month trial period and/or beyond. Tom’s awareness of the impact of yoga on his well-being increased once the classes had stopped:

It opened up something of which I was unaware. Because I was unaware, I could see no value. I could put no value on it and I could see no value in it beforehand. And now looking back, I can. [...] I can’t say there’s like this little sort of magic moment of revelation or something like that, but I suppose that glimpse or that awareness of that sort of connection of mind, body, brain, movement, you know, just going from a flash to being in that world for a little bit of time, into being in that world for a slightly longer period of time. That awareness grew during the course.

Ypart-Int2

Some had engaged with (or intended to) post-trial GYY classes, as well as different types of local yoga classes (some with different yoga teachers) and different styles of yoga, including mat-based for those who found the GYY classes lacking in physical challenge. Others had integrated self-directed yoga practice as part of their home routine, having a set time and place in the home for a daily or multiweekly practice.

Participants who had not continued with yoga post trial described two main reasons: a lack of benefit from the 12-week intervention and a lack of self-motivation to find a new class or engage with self-practice. Tom reflected on his lack of continuation:

And now it’s [GYY classes] gone out of my life and I’ve been talking to you about how beneficial it’s been and things like that, and yet I’ve not made any efforts to go back to it, or yet if I had an email saying, ‘There’s a follow-up class,’ I’d say, ‘Yeah, ding.’ That’s my experience. That’s a very strange combination.

Ypart-Int2

Decreased self-motivation was often associated with low mood rather than lack of interest in yoga, with these participants suggesting some form of extrinsic motivation may be enough to re-engage them with a regular practice. Additionally, some of the more physically able participants noted that while GYY was not something they were interested in pursuing further at the present time, it presented as a suitable option for when they were physically unable to continue with their current physical activities.

Reconfiguration: introducing and embedding gentle yoga

Yoga teachers, trial participants and other stakeholders were generally supportive of the integration of yoga within the NHS as a means of providing a more holistic approach to health care than the predominant biomedical model. The potential biopsychosocial health benefits of yoga were viewed as aligning with the initial NHS vision of preventative medicine, focusing on maintaining health rather than treating illness, as well as new agendas like social prescribing. They outlined a range of factors that would be central to consider in any future roll-out of GYY in the UK context (see Table 40).

Many favoured the representation of yoga as a proactive health behaviour aimed at health maintenance, sustained independence and illness prevention independent of age. The impact of a health condition was deemed more relevant than its presence. Several participants noted that the option of a gentle seated form of yoga, such as GYY, would have been useful when many of their chronic health conditions presented in their 30s–50s. Age was not viewed as indicative of health status or ability, with people noting little commonality or homogeneity among ‘older adults’ as a group beyond their similar ages. Notably, several yoga teachers reported not putting age restrictions on their current community-based GYY and related classes. Alison, a yoga teacher, noted that:

So, in my chair-based classes, I have somebody with rheumatoid arthritis who is in her fifties. I think that the age cut-off is – I wouldn’t say it’s irrelevant, but I suppose you’ve got to … Do we need it? I don’t know if we do need it, you know, a chair-based class … I don’t enforce it, in my chair-based class. It’s just anybody who would prefer to do something – you know, whatever we do on a mat, we do on a chair. So, if people can’t get down onto the floor and up off the floor easily then they come to my chair-based class.

Yteach-Int1

As such, GYY could focus on inclusivity rather than age, promoting GYY as a beginner’s form of gentle yoga, suitable for those who are currently non-active or those unable or uninterested to engage with a mat-based class.

Moving the emphasis of GYY from age to inclusivity may facilitate additional engagement of GYY classes among health professionals and patients. The potential target demographic could include those with mobility, mental health and rehabilitation issues. Gentle Years Yoga was seen as suitable for those who are wheelchair dependent and for a younger demographic with health-related mobility restrictions who require a gentler style of yoga. As one participant noted, ‘People I think with chronic [illnesses] could easily, you know, keep up with it. … I think it’s open for all, open for all’ (Joy, Ypart-Int1). Broadening the GYY remit to include people based on their physical ability would provide accessibility to a younger demographic with chronic health needs, to individuals in transition from injury or surgery and to experienced yoga practitioners no longer able to meet the physical requirements of a general yoga class due to onset or progression of health issues.

Introduction

Health-related quality of life has been defined as ‘how well a person functions in their life and his or her perceived well-being in physical, mental, and social domains of health’. 80 A common mode of assessing HRQoL is through patient-reported outcome measures (PROMs), via self-administered surveys (e.g. postal) or interviewer administration (e.g. telephone). Measuring HRQoL accurately is important in a range of fields, including research, policy-making and informing clinical care. Two HRQoL PROMs were collected in this trial, the PROMIS-29 and the EQ-5D-5L.

The PROMIS was developed by the US National Institute of Health (NIH) using item response theory to calibrate a large number of questions for each health domain to create an item bank. 81 PROMIS has over 100 health domains with a calibrated ‘item bank’ of questions that can be administered via standard short forms, custom short forms or computer adaptive testing. 82 PROMIS aims to address the lack of generalisable and universal PROMs to assess many HRQoL domains. 82 It has undergone extensive psychometric testing for validity and reliability in the general adult population, but it has not been widely used in older adults with multimorbidity83 and has received limited attention in the UK.

The EQ-5D has been validated in many different patient populations including diabetes, cardiovascular problems, chronic obstructive pulmonary disease, cancer, chronic pain and rheumatoid arthritis. The EQ-5D has some limitations arising from the descriptions of health used in the measure, which include (1) large proportions of the respondents scoring at the very top or very bottom of the scale (i.e. ceiling effects in the very healthy or floor effects in the very ill), (2) imprecise measurement for individuals, (3) poorly worded questions such as those that combine concepts (double-barrelled items) and (4) lack of coverage of the full range of health. 84 PROMIS addresses several of these limitations, including: (1) capturing a wider range of each health domain, (2) measuring individual health status with greater precision and (3) using rigorously designed and tested questions. 84

Although EQ-5D-5L and PROMIS-29 are both concise generic measures that allow the estimation of HRQoL, a recent systematic review of studies comparing EQ-5D and PROMIS-29 instruments has noted the lack of research comparing their measurement properties and utilities. 85 The impact of choosing either instrument for research, policy and clinical care decision-making is therefore unclear, and more research is needed to understand the relationship between the two instruments.

Objectives

To assess the concurrent validity of the PROMIS-29, a short-form version of the whole PROMIS system, against the EQ-5D-5L in adults aged 65 years or older with multimorbidity.

Methods

Results

Summary scores for the PROMIS-29 and EQ-5D-5L outcomes at each time point are presented in Chapter 3.

Correlations between the PROMIS-29 and EQ-5D-5L scores are presented in Table 41. For the EQ-5D-5L index value and VAS score, and the PROMIS-29 physical function and social participation subscales and both the PCS and MCS, a higher score indicates a more favourable outcome, whereas for all other scores a lower score indicates a better outcome. Hence, some correlations are negative. At all time points, the strongest correlation was between the EQ-5D-5L index value and the PROMIS-29 PCS (ρ ≥ 0.74). The EQ-5D-5L utility index score also correlated well with the PROMIS-29 MCS (ρ ≥ 0.66).