Group cognitive behavioural programme to reduce the impact of rheumatoid arthritis fatigue: the RAFT RCT with economic and qualitative evaluations

Inflammation as a possible mechanism for rheumatoid arthritis fatigue

If the inflammatory processes of RA are the major driver of RA fatigue, either directly or through the effects of inflammation on pain and function, then pharmacological interventions for RA inflammation should also significantly improve fatigue. However, systematic review shows that biologic disease-modifying antirheumatic drugs (bDMARDs) to control RA inflammation yield only a small to moderate improvement in RA fatigue (32 studies involving 14,628 patients), although the quality of the studies was variable. 16 Inflammatory activity, not fatigue, is the indication for initiating or changing bDMARD therapy; therefore, these studies did not recruit patients specifically with fatigue. 16 The findings are supported by other studies that show that RA fatigue continues to be problematic even for those patients who achieve disease (i.e. inflammatory) remission. 17 Therefore, it is unlikely that inflammation alone is the single driver for RA fatigue.

Multicausal pathways as a likely mechanism for rheumatoid arthritis fatigue

A systematic review of causal studies relating to RA fatigue could not draw conclusions as a result of the quality of the evidence, which was largely cross-sectional or longitudinal studies that rarely controlled for baseline fatigue. 18 Therefore, with insufficient studies, and often conflicting results on any single causal mechanism for fatigue (e.g. inflammation, disability, depression),18 it is postulated that, conceptually, RA fatigue is likely to have complex and multicausal pathways (see Appendix 1). 19 These pathways are categorised as potentially involving interactions between or within three elements: (1) disease processes (RA elements); (2) thoughts, feelings and behaviours (cognitive/behavioural elements); and (3) life issues (personal elements). 19 Pathways will vary between patients and potentially even within patients for individual fatigue episodes. For example, on one occasion fatigue may be driven by inflammation causing pain and disrupting sleep, interacting with heavy work responsibilities that lead to overactivity (keeping going until exhausted); however, on another occasion fatigue may be driven by low mood, leading to behavioural withdrawal (deactivation) and subsequent deconditioning. In this conceptual model, self-management strategies appear crucial in managing fatigue impact.

Rationale for cognitive–behavioural therapy for fatigue in rheumatoid arthritis

Cognitive–behavioural therapy (CBT) aims to help people work out for themselves the links between their thoughts (cognitions) and feelings and how these may be driving their actions (behaviours), and addresses these underpinning cognitions as a way of prompting behaviour changes. 21 For example, a belief that it is one’s role to do the household chores may prevent someone asking family members to help, which, combined with feeling shame or guilt if chores are left undone, may drive persistence in undertaking chores. This excessive activity leads to episodes of overwhelming fatigue and enforced rest (so-called ‘crashes’), commonly called the ‘boom-and-bust’ cycle. Addressing these previously unchallenged beliefs about what family members would think about a request for help, and whether or not leaving some chores undone would be a catastrophe, might prompt behaviour changes that result in less fatigue. The lack of an existing clear approach to self-management of fatigue from RA professionals may contribute to a sense of powerlessness and passivity among people with RA fatigue who feel their lives are controlled by its perceived unpredictability. CBT provides a framework through which a person can better understand those components of fatigue experiences that are controllable, and, through this, make a shift in core beliefs towards taking control rather than being passive. On the basis of such a shift a person could become progressively less passive and more self-confident through better self-management of activity and fatigue symptoms so that the impact of fatigue is reduced.

A systematic review of group education programmes for general self-management in musculoskeletal conditions found that a CBT approach is more likely to lead to people making behaviour changes than information alone,22 and CBT has been shown to be effective for fatigue in CFS and multiple sclerosis. 23,24 Problem-solving and goal-setting as routes to enhancing self-efficacy (see Underpinning theory: self-efficacy and social cognitive theory) are the core self-management skills that patients learn in CBT.

Underpinning theory: self-efficacy and social cognitive theory

Managing daily life with RA in order to minimise fatigue requires changing behaviours that might have been contributing to fatigue onset, perpetuating it, or limiting its management. Changing habitual behaviours can be hard, and self-efficacy is the belief or confidence that one can successfully execute an action or master a situation and is a strong predictor of behaviour change and persistence in the face of difficulty. 25 Self-efficacy is cultivated through reinterpretation of symptoms (understanding that fatigue has many causes that could be tackled), mastering the new behaviours (through progressive goal-setting), and by modelling and social persuasion (from seeing the successful changes and adaptions of other RA patients). 26 Therefore, learning about fatigue management as part of a group can enhance these self-efficacy processes because other patients in similar situations act as role models: validating the complexity of fatigue and demonstrating more effective ways of mastering it through the efforts they make as fellow participants on the programme [social cognitive theory (SCT)]. 25,26 It has been demonstrated that interventions based on SCT and/or CBT enhance self-efficacy and produce better patient outcomes than simply providing self-management information. 22,27 Learning with others in similar circumstances may be particularly helpful for validating invisible symptoms, such as fatigue.

Evidence for psychosocial interventions on rheumatoid arthritis fatigue

Three RCTs trialled one-to-one CBT, a group programme (based on SCT) and a group CBT programme (SCT/CBT) over 11, 16 and 22 hours, respectively, and reported improved fatigue in RA patients. 28–30 However, interventions were not primarily aimed at improving fatigue severity; thus, patients were not recruited because of fatigue, nor were studies powered to detect fatigue change. Furthermore, participation in the studies was restricted (e.g. to those with early disease, mild disability or psychological distress or who have a partner). However, RA fatigue occurs in the majority of patients;6,7 therefore, self-management interventions should be tested across a wide RA population. Therefore, the original 18-week RCT was conducted to test a fatigue self-management programme in a broad RA population (fatigue score of ≥ 6/10), using group CBT (13 hours SCT/CBT), led by a clinical psychologist and specialist pain management occupational therapist (OT), compared with groups receiving fatigue self-management information alone (n = 127). 31 The group CBT intervention improved fatigue impact (effect size 0.77) alongside fatigue severity and coping with fatigue, as well as disability, sleep and mood (depression, helplessness, self-efficacy). In the qualitative evaluation undertaken through participant interviews, patients spontaneously raised the significance of key elements of the CBT used by the clinical psychologist (working links out for themselves, goal-setting, activity diaries), and SCT components (the value of being in a group). 32

A systematic review identified 15 psychosocial interventions reporting RA fatigue, including the previous study (13 RCTs involving 1556 patients). 20 Interventions included expressive writing, CBT, mindfulness, lifestyle management, self-management, energy conservation and group education, and showed an overall small benefit on RA fatigue severity. The overall quality of trials was rated as being low, largely accounted for by small patient numbers and the difficulty of blinding participants in interventions that require engagement for behaviour change. Within the systematic review, the RCTs detailed above, which were the only interventions in which improving fatigue was an aim or inclusion criterion,28,30,31 had stronger methodological qualities and stronger effects on fatigue.

The RAFT programme delivery: cognitive–behavioural and social cognitive theory approaches

The CB approach used was one of reflective discovery through Socratic questioning, in which tutors ask simple questions to enable patients to identify the links between their thoughts, feelings, behaviours and symptoms (i.e. fatigue). 21 This was in contrast to the more didactic information- or advice-giving by the tutors, in which the patients were passive recipients of information. Self-monitoring (daily activity/rest charts) and goal-setting are the most effective tools in bringing about behavioural change,38 and the use of metaphors helps turn abstract or difficult therapeutic concepts into a form that can be more easily understood and remembered. 39 The value of group CB programmes is that fellow patients are a credible source to legitimise or validate invisible symptoms, such as fatigue, and provide role models and peer support in the goal-setting and problem-solving processes. 25

Inclusion criteria

In order to reflect pragmatic clinical practice, in which the majority of RA patients experience fatigue, inclusion criteria were very broad. Patients aged ≥ 18 years with confirmed RA42 were eligible if they had a fatigue severity score of ≥ 6 (out of 10) on the Bristol Rheumatoid Arthritis Fatigue Numerical Rating Scale (BRAF-NRS),43,44 which they perceived was a recurrent problem.

Exclusion criteria

Patients were excluded if they had recently changed major RA medication (within 16 weeks) or glucocorticoids (within 6 weeks), as that might influence fatigue. Patients were ineligible if they had insufficient English to participate in group discussions or lacked capacity for informed consent.

The RAFT programme intervention

The RAFT programme is a CB fatigue self-management programme delivered to groups of 5–7 RA patients in six 2-hour sessions (weeks 1–6) and a 1-hour consolidation session (week 14) by a pair of local RNSs and/or OTs (see Chapter 1). Taking a pragmatic approach, if one of the tutors was unable to deliver a session, perhaps because of illness, the remaining tutor delivered the session alone. Patients were asked to attend all seven sessions if possible; however, some patients inevitably missed some sessions as RA is a fluctuating condition, in which case tutors offered to explain some of what had been missed, prior to the next session.

Fidelity to the RAFT programme during delivery

Quality and homogeneity of the RAFT programme were facilitated by standardised training and a standardised manual, with programme delivery by the same tutor pairs in each centre. However, unlike a pharmacological RCT, in a group CB intervention it is not possible to adhere to a rigid protocol in every session, as tutors must respond to the individual issues raised by patients. Therefore, fidelity evaluation was pragmatic and involved monitoring in one randomly selected session of each of the four programmes run in each centre. An independent clinical psychologist observer (Richard Cheston) used a template to record evidence of CB approaches (e.g. reflective questioning, group management, goal-setting), delivery of the RAFT programme as planned (adherence to session plans), use of the RAFT programme materials (handouts, metaphors) and any unhelpful delivery styles (e.g. didactic teaching) (see Report Supplementary Material 1). 41

Control intervention (usual care alone)

Unlike pain, RA fatigue is not routinely addressed in usual care in great detail and patients are generally given written fatigue information by the RNS. Most rheumatology units supply the free Versus Arthritis (formerly Arthritis Research UK) fatigue booklet,45 a comprehensive fatigue self-management booklet for arthritis authored by Hewlett et al. and based on their original RCT. 31 It contains information on all the topics in the RAFT programme, includes a pull-out sample activity and rest chart to complete and could be considered ‘usual care’. Usual care (i.e. provision of the Versus Arthritis fatigue booklet) was given to patients in both the control and intervention arms of the trial at the baseline visit, after consent and assessment but before randomisation. The research nurse spent approximately 5 minutes showing their patient the sections in the booklet, using a brief standardised guideline and pointing out that the booklet suggests patients might wish to request specific support from the rheumatology team to help them try the fatigue self-management activities described (generally requested via the local RNS helpline). Seeking help is thus an intended outcome of the booklet, which was captured within health-care use data collected in the trial. In order to minimise the risk of contamination between the trial arms, the helpline RNS were asked not to book in a control patient requesting fatigue support to see a clinician who was also a RAFT programme tutor if there was an alternative health professional available, and tutors recorded any control patients seen for fatigue support.

Primary outcome measure (26 weeks)

As the impacts of fatigue are wide-ranging and personal to individuals’ lives, the perceived impact of fatigue was the primary outcome measure, rather than fatigue severity. 8–11 Fatigue impact was measured by a single NRS item, ‘Please circle the number that describes the effect fatigue has had on your life in the last 7 days’, with anchors of ‘No effect’ to ‘A great deal of effect’ (0–10), using the BRAF-NRS for impact (see Report Supplementary Material 2). 43,44

Secondary outcome measures (clinical)

The secondary fatigue outcomes were fatigue severity and coping (as measured using the BRAF-NRS severity, BRAF-NRS coping questionnaires) (see Report Supplementary Material 2) and a RA fatigue multidimensional questionnaire [i.e. Bristol Rheumatoid Arthritis Fatigue Multidimensional Questionnaire (BRAF-MDQ)] (see Report Supplementary Material 2), which measures overall fatigue impact and includes subscales for physical fatigue, living with fatigue, emotional fatigue and cognitive fatigue. 43,44 Each of these fatigue variables might respond differently to interventions. Other components of the RA fatigue conceptual model19 evaluated were mood [via the Hospital Anxiety and Depression Scale (HADS)],46 pain (via the NRS), disability [via the modified Health Assessment Questionnaire (MHAQ)],47 quality of life [global arthritis impact question from the Arthritis Impact Measurement Scale (AIMS)],48 sleep quality (via a single question from the Pittsburgh Sleep Quality Index),49 and returning to important leisure activities previously lost to fatigue [via the discretionary activity subscale of the Valued Life Activities scale for rheumatoid arthritis (VLA)]. 50 Beyond pain, disability and fatigue, remaining items from the RA core set51 were an objective assessment of painful joints and swollen joints, an inflammatory marker (i.e. C-reactive protein) and patient global opinion of disease activity [visual analogue scale (VAS)], forming a single disease activity score [Disease Activity Score (DAS28)] through a weighted algorithm. 52 As the DAS28 requires a hospital visit to the research nurse, it was measured at weeks 0 and 26 only, and replaced at all other time points by a patient self-reported version [i.e. the simplified Patient-derived Disease Activity Score (sPDAS2)]. 53,54 Social contact questions (unvalidated) were added for weeks 52 and 104.

Secondary outcome measures (processes)

To understand what key processes may relate to behavioural change, helplessness [via the Arthritis Helplessness Index (AHI)]55 and self-efficacy [via the Rheumatoid Arthritis Self-Efficacy (RASE) scale]56 were measured. The RASE scale contains questions on self-efficacy beliefs for managing RA covered in both the RAFT programme and the usual-care Versus Arthritis fatigue booklet. 45

Secondary outcome measures (acceptability and feasibility)

Acceptability of the RAFT programme and the usual-care booklet45 was assessed at week 26 by patient satisfaction and the likelihood that patients would recommend the programme or booklet to others, and by programme attendance. Feasibility of delivery in the NHS was captured through monitoring of programme scheduling and delivery, and the tutor qualitative evaluation.

Minimising attrition

In order to reduce attrition prior to starting a RAFT programme, patients were consented and randomised as close to the dates of the CB programmes as possible and the local research nurse kept patients informed of the potential programme dates. Three-quarters of participants in the original RCT attended most of their seven CBT sessions, suggesting that the RAFT programme would be acceptable. 31 In this RCT a 2-year follow-up potentially risked significant loss to questionnaire completion over time; therefore, to maximise questionnaire returns, the primary outcome data (i.e. fatigue impact NRS scores) were collected by the central trial team by telephone each time (see Follow-up time points, data collection and management). To enhance feelings of community, engagement and responsibility, all RAFT programme study materials included the study logo; questionnaire packs were accompanied by a short newsletter highlighting overall questionnaire return rates, and thank-you notelets were sent at major time points.

Primary outcome measure

Secondary analyses of the primary outcome included additional adjustment of the primary analysis for any prognostic variables demonstrating a marked imbalance at baseline (ascertained using descriptive statistics). There was a gap in time between eligibility screening and baseline measurement. Owing to this time gap, some individuals who had a BRAF-NRS severity score of ≥ 6 points (out of 10 points) at screening (an eligibility criterion) subsequently had a lower BRAF-NRS severity score of < 6 points at the time of randomisation. For this reason, the primary analysis was repeated and restricted to those individuals who had a BRAF-NRS severity score of ≥ 6 points at randomisation (analysis not prespecified).

A repeated measures mixed-effects model was used to examine the effect of the intervention over time by including up to four follow-up BRAF-NRS impact scores (at weeks 26, 52, 78 and 104), adjusted for baseline values of the outcome and for recruitment centre. This analysis provided an estimate of the average effect size over the duration of follow-up. A potential convergence or divergence between trial arms was investigated by repeating the analysis with the inclusion of a time-by-treatment group interaction term. The repeated measures analysis was repeated on those individuals with a BRAF-NRS severity score of ≥ 6 points at randomisation.

Secondary outcome measures

The effect of the intervention on the secondary outcomes collected at 26 weeks was also examined using appropriate regression models adjusted for baseline values of the outcome being investigated and for recruitment centre. The secondary outcomes were also subject to repeated measures analysis using data collected at the 26-, 52-, 78- and 104-week follow-ups.

A potential difference between trial arms in the changes in major RA medications that might influence the primary outcome was investigated using a chi-squared test. A change in major RA medications was defined as stopping, starting or changing the dose of one or more disease-modifying antirheumatic drugs (DMARDs), bDMARDs or glucocorticoid medications during a single time period. Acceptability of the RAFT programme or the usual-care booklet to patients was measured by the RAFT programme attendance rates, satisfaction and recommendation to others, and is summarised using descriptive statistics.

Number of sessions attended

It was hypothesised a priori that those individuals attending a greater number of the RAFT programme sessions would show a greater improvement than those who attended fewer sessions. The theoretical justification was that greater exposure to the RAFT programme intervention should improve self-management skills. A structural mean model 2SLS instrumental variable approach, which respects randomisation,64 was used to investigate the impact of number of sessions of the RAFT programme attended on BRAF-NRS impact score. As with the treatment efficacy CACE analysis, the non-follow-up of those individuals randomised to the RAFT programme intervention arm but who did not attend the first session of the RAFT programme is problematic. The same approach (i.e. a separate two-stage approach) was used to address this issue. A further assumption of this model was that of a linear relationship between session attendance and BRAF-NRS impact score, that is, that the impact of an increase in attendance of one session was the same across all the different sessions (i.e. all sessions had a similar effect on patient outcome).

Consolidation session (week 14)

The RAFT programme intervention was delivered in seven sessions: sessions 1–6 delivered weekly over weeks 1–6, followed by session 7 at week 14 to consolidate skills and behaviours. Separate linear regression models were used to examine the effect of the RAFT programme on BRAF-NRS impact score at 6 weeks (immediately after the final RAFT programme session), at 10 weeks (4 weeks before the consolidation session) and at the 18-week follow-up (4 weeks after the consolidation session).

Predictors of fatigue impact change

Certain clinical and demographic characteristics were examined as potential predictors of fatigue impact change. All of these variables of interest were first added as covariates to the primary analysis model and then removed from the model one by one in a process of stepwise deletion until only covariates with a p-value of < 0.05 were retained (i.e. the covariate with the highest p-value from the initial model was removed, the model was then repeated without this covariate and the variable with the highest p-value in this second model was then removed, etc.). The starting model comprised the primary analysis model additionally adjusted for baseline pain (i.e. NRS score), disease activity (i.e. DAS28 score), BRAF-NRS coping score, age, comorbidity, work status, disease duration and sex.

Self-efficacy

The RASE scale has 28 items that ask about beliefs regarding self-management tasks, many of which are discussed during the RAFT programme. To examine the impact of the RAFT programme intervention on particular components of self-efficacy, the change in RASE scale item scores from 0 to 26 weeks in the control and the RAFT programme intervention trial arms was compared using t-tests.

Tutor delivery over time

It was hypothesised that the size of the effect of the RAFT programme intervention may differ between cohorts as tutors within each centre may have improved in their delivery of the intervention over time. To explore this, the mean change in fatigue, impact from 0 to 26 weeks, is presented by intervention arm and cohort. The difference in mean change (between trial interventions) by cohort is also presented (this was not in the SAP but was prespecified before any analysis commenced).

Social contact

Group work may enhance the seeking of social support (see Chapter 2, Protocol amendments during the trial). The social contact data are analysed using simple percentages.

Participant flow

Recruitment commenced in December 2013, with the first patient consent received in February 2014 and the last in October 2015, and the trial completed as planned with the final follow-up questionnaires returned in January 2018. The flow of eligible patients with RA and a fatigue severity score of ≥ 6 points from invitation to participation through to week 104 is detailed in Figure 1. Over the 2 years a total of 333 patients were randomised, comprising four consecutive cohorts in each of the seven centres (see Chapter 2, Recruitment procedures), with similar numbers across centres (see Appendix 3). All patients accepted their randomised allocation.

FIGURE 1.

Participant flow chart. Reproduced from Hewlett et al. 65 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text.

Reproduced from Hewlett et al. 65 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text.

After consent and baseline assessment, all patients received usual care before randomisation, i.e. the Versus Arthritis fatigue booklet. 45 Of the 175 patients randomised to additionally receive the RAFT programme, 14 did not receive the intervention because of an inability to find suitable programme dates. Of the 161 patients who received the RAFT programme, five patients withdrew for personal/health reasons (four patients during programme delivery, having received up to three of the seven sessions, and one patient withdrew before the week 10 assessment). Therefore, 156 of the 175 RAFT programme patients provided primary outcome data at week 26, and 153 patients provided data at week 104 (89.9% and 87.4% of those patients randomised, respectively). Of the 158 patients randomised to usual care only, six patients withdrew before week 26, leaving 152 patients providing primary outcome data at week 26 and 147 patients providing data at week 104 (96% and 93.0%, respectively).

Primary outcome measure

Analysis adjusting for variables imbalanced at baseline was not required as no variables were imbalanced.

Primary outcome at 26–104 weeks

The BRAF-NRS impact score differed between trial arms over weeks 0–104 (Figure 3). Based on repeated measures analysis and using data from the 26-, 52-, 78- and 104-week follow-up time points, the group receiving the RAFT programme intervention had a BRAF-NRS impact score that was, on average, 0.49 units lower (i.e. better) than in the control arm over the 2 years, on a scale of 0–10 (adjusted mean difference –0.49) (Table 8).

FIGURE 3.

The BRAF-NRS impact scores over weeks 0–104, by trial arm. Reproduced from Hewlett et al. 65 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text.

Reproduced from Hewlett et al. 65 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text.

TABLE 8 - Repeated measures analysis of BRAF-NRS impact score at 26, 52, 78 and 104 weeks

Adjusted for baseline score and centre.

Note

Reproduced from Hewlett et al. 65 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The table includes minor additions and formatting changes to the original text.

Trial arm	Time point								Repeated measures (N = 308)
	26 weeks (N = 308)		52 weeks (N = 305)		78 weeks (N = 300)		104 weeks (N = 296)
	n	BRAF-NRS impact score, mean (SD)	n	BRAF-NRS impact score, mean (SD)	n	BRAF-NRS impact score, mean (SD)	n	BRAF-NRS impact score, mean (SD)	Adjusted mean differencea (95% CI)	p-value
Control	152	6.36 (2.42)	151	6.38 (2.19)	146	6.38 (2.23)	145	6.05 (2.14)	–0.49 (–0.83 to –0.14)	0.01
RAFT	156	5.74 (2.41)	154	5.72 (2.23)	154	6.17 (2.24)	151	5.54 (2.28)

When a treatment arm-by-time interaction term was added to the model there was no evidence that the difference between the RAFT programme and the control arm varied over time (interaction between treatment and time, p = 0.52). The estimated treatment effects at each of the follow-up time points were similar and in favour of the RAFT programme, with the exception of that at the 78-week follow-up time point, which was non-significant (see Appendix 7).

Secondary outcome measures

Number of sessions attended

As with the CACE analysis (see Appendix 11), a structural mean model 2SLS instrumental variable approach62 was used to investigate the effect of the total number of RAFT programme sessions attended on the BRAF-NRS impact score. The results suggest that the more RAFT sessions attended, the greater the reduction in fatigue impact at 26 weeks, with each additional RAFT session attended giving a decrease of approximately 0.1 BRAF-NRS impact units (95% CI –0.21 to –0.01 BRAF-NRS impact units; p = 0.03). However, the model assumes a linear relationship between session attendance and BRAF-NRS impact score and that the impact of an increase in attendance at any one of the seven sessions is the same across all sessions (i.e. each of sessions 1–7 has a similar effect on patient outcome, even though each session covers a different topic that may or may not hold personal relevance for individual patients). This analysis is exploratory and assumptions of linearity may not hold.

Consolidation session (week 14)

Additional data were collected at the end of the intensive six sessions of the RAFT programme (week 6) and 4 weeks before and after the week 14 consolidation session (i.e. weeks 10 and 18), to provide preliminary data on the possible impact of the consolidation session. Week 6 data are shown in Appendices 12 and 13 and fatigue data collected at weeks 10 and 18 are provided in Appendices 14–17. After adjusting for baseline score and centre, there was evidence of a difference between trial arms at week 6 in favour of the RAFT programme for the primary outcome (BRAF-NRS impact, –0.65), as well as coping with fatigue (BRAF-NRS coping), overall fatigue impact (BRAF-MDQ overall) and living with fatigue, cognitive fatigue and emotional fatigue, plus sleep quality and the process measure self-efficacy (RASE scale) (Table 14). Effect sizes for these variables ranged from 0.21 (cognitive fatigue) to 0.6 (coping with fatigue).

At 10 weeks, the adjusted mean difference between the arms for the primary outcome was maintained in favour of the RAFT programme, but was lower than that at 6 weeks (BRAF-NRS impact score –0.51 vs. –0.65). In addition, there was no longer evidence of a difference between arms for coping with fatigue. At 18 weeks, 4 weeks after the consolidation session 7, the adjusted mean difference in BRAF-NRS impact score was again similar to that at week 6, at –0.64, and coping with fatigue was once again significantly different between arms (Table 15). However, these were exploratory analyses and any differences between effect sizes at these time points may reflect random variation in treatment effect as the trial progressed.

Predictors of fatigue impact change

Certain clinical and demographic characteristics were examined as potential predictors of fatigue impact. The starting model comprised the primary analysis model additionally adjusted for baseline pain (NRS), disease activity (DAS28), BRAF-NRS coping score, age, comorbidity, work status, disease duration and sex. After the process of stepwise deletion, the only additional variables retained (i.e. variables other than those in the primary analysis model) were sex and disease activity (DAS28) score. The final linear regression model (primary analysis model plus additional adjustment for sex and baseline DAS28) suggested that a worse outcome was predicted by being female and having a higher baseline DAS28 score: being female was associated with a 0.75 unit of scale increase in BRAF-NRS impact score at 26 weeks (95% CI 0.08 to 1.41; p = 0.03), whereas a 1 unit of scale increase in baseline disease activity (DAS28) was associated with a 0.25 unit of scale increase in BRAF-NRS impact score at 26 weeks (95% CI 0.02 to 0.49; p = 0.04).

Self-efficacy

The RASE scale addresses beliefs about RA self-management behaviours56 and most of the 28 items are touched on during the RAFT programme, except the final three medication items (see Appendix 18). At 26 weeks, six RASE scale items demonstrated evidence of a difference between trial arms for change over weeks 0–26 in favour of the RAFT programme (Table 16). Avoiding doing things that cause pain, relaxation, pacing, acceptance of fatigue, asking for help and using relaxation tapes were all covered during RAFT programme sessions.

Tutor delivery over time

It was hypothesised that tutors might increase in skills and confidence to deliver the RAFT programme over time, which might be reflected in improved outcomes in the later RAFT programmes (i.e. from cohort 1 through to cohort 4). Although it appears that the difference between trial arms in mean change in BRAF-NRS impact score was greatest in cohort 4, the study was not powered to formally test this (Table 17).

Social contact

Both the usual-care booklet and the RAFT programme highlight the value of obtaining support. At 52 and 104 weeks, 7–13% of patients across both trial arms reported making contact with other patient(s) for support with fatigue or arthritis (see Appendix 19). In addition, at 52 weeks, 20% of RAFT programme patients (n = 35) had made contact with patient(s) in their RAFT programme group after their programme had finished and 12.6% (n = 22) had maintained contact during the year. At 104 weeks, 9% of RAFT programme patients (n = 16, including 12 of the original 35 patients) had made contact and 5.6% of RAFT programme patients (n = 10) maintained contact during this second year.

Identification of relevant resource use

Data were collected for resource use associated with RA or RA-related fatigue only. Routine monthly blood monitoring visits were excluded, whether performed in primary or secondary care, as these were considered to be unrelated to the intervention. For the NHS and PSS perspectives, data were collected on arthritis medication, primary and community medical care, physiotherapy, occupational therapy, podiatry or nursing care, secondary care, use of a rheumatology telephone helpline and use of social care by the patient. Staff time and other expenses incurred during the training and delivery of the RAFT programme were recorded.

For the primary analysis from the societal perspective, we additionally collected data on work productivity (presenteeism and absenteeism), and patient-incurred expenses for travel to a general practitioner (GP) and hospital appointments, including the RAFT programme itself.

Measurement of resource use

Time logs and expense forms were used to track all resources used in the delivery of the 4-day RAFT training programme, including trainee and trainer time (preparation time and observation of the practice programme), travel costs and programme materials, to calculate the fixed cost of training (Table 18). For the delivery of the RAFT programme CB sessions, NHS resources were captured in staff time logs recording session preparation, delivery, debriefing, supervision time and materials.

NHS community care and emergency care, and patient personal resource use, during the 2-year follow-up period were captured using patient-reported questionnaires at baseline and at weeks 6, 26, 52, 78 and 104, with patients completing the questionnaires at their research nurse visits at baseline and week 26, and by post at all other time points. Medications were recorded by the research nurse at the baseline visit and patients were asked to indicate changes from the previous record at each follow-up questionnaire. Secondary care data on outpatient appointments, inpatient stays and day-case attendances were extracted from hospital computer records by research nurses.

The research nurse documented the patient’s normal transport method for hospital and GP appointments and the cost (for public transport) or mileage (for private or hospital transport) at the baseline visit, to use as a multiplier for calculating costs. Work disability, presenteeism and absenteeism were captured using the Work Productivity and Activity Impairment (WPAI) scale, which measures hours worked or lost over the past 7 days. 70

Valuation of resource use

All costs are reported in pounds sterling at 2015/16 rates. Unit costs for community care, primary care and NHS staff time to train for and deliver the intervention were based on the Unit Costs of Health and Social Care 2016. 71 Actual expenses incurred for training materials, refreshments and staff travel were recorded. The costs of medications used during follow-up were based on data in the prescriptions cost analysis for England. 72 Secondary care (including routine RA care) costs were derived from the NHS Reference Costs 2015 to 2016. 73 Private care costs were estimated from the median of three quotes or by using data from national bodies. The WPAI scale responses were combined to derive an estimate of the overall work impairment due to RA or RA-related fatigue. 70 The impairment factor was applied to average weekly earnings stratified by age. 74 Productivity losses were estimated as changes in productivity from baseline. Unit costs used in the analysis are given in Table 19. Costs occurring during the second year of follow-up were discounted in line with NICE guidance (currently 3.5%). 66

Primary cost-effectiveness analysis

Cost and QALY data were combined to calculate the incremental cost-effectiveness ratio (ICER) and incremental net monetary benefit (INMB) statistics:

The INMB statistic is given as the societal willingness to pay (WTP) for a QALY, λ, multiplied by the difference in patient outcomes, E (i.e. QALYs), from which the difference in total costs, C, is subtracted. In the primary analysis, INMB was estimated to see whether or not the RAFT programme intervention is cost-effective at the established NICE thresholds of £20,000 and £30,000 per QALY gained.

The probability that the RAFT programme intervention is cost-effective at various ‘WTP-for-a-QALY’ thresholds was explored using cost-effectiveness acceptability curves. 79 All cost-effectiveness statistics (i.e. ICERs, cost-effectiveness acceptability curves and INMB) and CIs were estimated parametrically from the output of seemingly unrelated regression analysis, using the Stata ‘sureg’ command to allow for the correlation between costs and outcomes. Differences in baseline utility (or BRAF-NRS impact score) between the arms were controlled for. Study centre was included as an independent variable in the regression analysis.

Sensitivity analyses

One-way sensitivity analyses were used to judge the potential impact of sources of uncertainty. A sensitivity analysis was conducted in which rare high-cost events thought unlikely to be related to fatigue by two clinicians (unaware of randomised treatment allocation) were excluded. A sensitivity analysis dropping training costs was also conducted to estimate the impact of training becoming routine in a roll-out situation. Although it was possible to restrict the primary statistical (clinical) analysis to complete cases, the number of missing data was higher for the economic analysis and so data were imputed and a complete-case analysis was conducted as a sensitivity analysis. There was an inevitable delay between screening assessment and formal baseline assessment as the cohorts (i.e. 10–14 patients sufficient for randomisation) took time to accrue (see Chapter 2, Recruitment procedures). This meant that 24 out of the 158 randomised control patients and 28 out of the 175 randomised RAFT programme patients had a baseline BRAF-NRS fatigue severity score that fell below the eligibility criterion of ≥ 6 points. A further post hoc subgroup analysis therefore investigated only those patients who were eligible to participate in the trial at baseline, as was performed for the primary clinical outcome (see Chapter 3).

Missing data

Complete sets of data were available for the intervention costs (which were recorded by staff) and secondary care costs (which were extracted from medical records). Participants did not always fully complete questionnaires, leading to missing data at some time points in the medication, emergency care, social care, community care, work productivity, travel and private care cost categories, and in the utility measurements. For the primary 26-week analysis, one or more data points were imputed for 59 participants (17.7%), comprising 24 patients in the control arm (15.2%) and 35 patients in the RAFT programme arm (20.0%) (p = 0.25). At the 2-year follow-up, one or more data points were imputed for 107 participants (32.1%), comprising 51 patients in the control arm (32.3%) and 56 patients in the RAFT programme arm (32.0%) (p = 0.96). Data were imputed for a mean of 9.9 out of 18 items in the RAFT programme arm and a mean of 8.0 out of 18 items in the control arm over weeks 0–26 (p = 0.03); and a mean of 15.8 out of 48 items in the control arm and a mean of 20.8 out of 48 items in the RAFT programme arm over weeks 0–104 (p = 0.07).

Protocol amendment to data collection

In the original protocol [see URL: www.journalslibrary.nihr.ac.uk/programmes/hta/1111201/#/documentation (accessed 12 April 2019)], the study design included two focus groups with RAFT programme tutors to evaluate the impact of learning CB approaches and delivering the RAFT programme intervention. The first focus group with approximately half of the RAFT programme tutors (as available) would then be followed by a focus group with the remaining tutors to confirm, challenge or elaborate those findings. Focus groups allow for discussion and reflection of a common experience, which can help participants to clarify their own thoughts and so provide a collective consideration of the topic under investigation. 83 However, the RAFT programme tutor feedback immediately after the training and clinical supervision during delivery indicated great variation in individual tutor experiences, even though the RAFT programme tutors participated in the training together and all delivered the RAFT programme from a standard manual. The RAFT programme study team concluded that individual face-to-face interviews would be the most appropriate method of data collection to capture this diversity, as these types of interviews allow for in-depth exploration of a single viewpoint and facilitate discussion of sensitive topics. 84 Therefore, the final qualitative study design was individual tutor interviews followed by a single focus group (see Chapter 2, Protocol amendments during the trial).

Recruitment and sampling

All RAFT programme tutors, comprising nurses and OTs who delivered RAFT programme sessions at the seven participating sites (see Chapter 3, RAFT programme delivery and tutor adherence), were invited to take part in a one-to-one, face-to-face semistructured interview. This included a locum tutor who had not attended RAFT programme training, but had co-delivered one RAFT programme after observing the previous programme delivered at that site. Fourteen tutors accepted and the one tutor who declined did not offer a reason. At least one RAFT programme tutor from each RAFT programme site participated in an interview (nine nurses and five OTs).

In addition to the individual interviews, the same 15 RAFT programme tutors were invited to take part in a focus group, of whom eight agreed to participate. Reasons cited for declining included clinical commitments, the focus group taking place on a non-work day and the time needed to travel to Bristol. At least one RAFT programme tutor from five of the seven RAFT programme sites participated in the focus group (three nurses and five OTs).

Data collection

Data collection occurred after the completion of the last of the four RAFT programmes in a centre. On average the tutor interview or focus group occurred 9 weeks later (range 4–13 weeks), apart from the interview with the locum tutor who had delivered an early RAFT programme, resulting in a longer time period of 39 weeks.

Data analysis

Interview and focus group audio-recordings were transcribed by a professional service and then transcripts checked for accuracy against the original audio-recordings and anonymised (ED). The interview and focus group transcripts were analysed using an inductive thematic approach. 85 Two main characteristics of thematic analysis made it appropriate for the RAFT programme qualitative study:

1. It is a useful method when researchers want to stay close to the data and ensure that the findings are recognisable to participants.

2. It can summarise the key features of a large body of data, highlighting both similarities and differences across participants. 86

First, ED read through all the interview transcripts and coded items of interest and chunks of text that related to the research topic. These initial codes were made in the margins of the transcript itself. ED then re-read the transcripts to identify patterns of codes. Codes and their supporting text that appeared to be addressing similar topics or ideas were copied from the transcript and placed together in a Microsoft Word document (Microsoft Corporation, Redmond, WA, USA). This sorting process formed the basis of the analysis. Related clusters of coded text formed subthemes, which were then grouped together to form a smaller number of higher-order themes that were more generic and described broader, and often more abstract, elements in the data set. Two transcripts were independently coded by SH and AH, and a single transcript by CR. These independent analyses were incorporated into the final analysis by ED, as there were no substantial differences or inconsistencies between the interpretations of the data. The focus group transcript was analysed by ED and AH. Owing to the overlap in the topics discussed in the interviews and the focus group, the RAFT trial team decided to use the focus group data as a form of triangulation (i.e. to confirm, challenge or elaborate the themes identified in the interviews, and to present a single, integrated analysis). In this final analysis presented here, each theme has three parts: the label, the summary and the supporting subthemes. These subthemes are supported by data excerpts from the RAFT programme tutors to ensure that data interpretation remains directly linked to their words. 87 Tutor excerpts are identified using arbitrarily allocated numbers not linked to the RAFT programme site or tutor pairing (INT 1–14 for interview participants, FG 1–8 for focus group participants).

A different way of working

To differing degrees, participants identified elements of the RAFT programme intervention that required a different way of working and for some it was ‘completely unfamiliar territory’ (INT4). In particular, the use of CB approaches and the ‘ask don’t tell’ approach contrasted with participants’ usual clinical practice, which involved giving advice and problem-solving for patients. Group work was also a new way of working for several participants:

I think it was really interesting and of course, I guess, the complexity of it was, there was all of those things to go through as somebody that was learning.

INT2

I wasn’t used to that kind of role, the cognitive–behavioural role rather than, just, as a nurse you just want to help them and say ‘Yes, I’ll do that for you’ so it was changing my kind of way of thinking.

INT3

I found it quite hard to change my way of thinking to follow it as the ‘ask don’t tell’ sort of process I found very difficult. Just purely from the way I’d worked over the years really.

INT7

. . . it’s the techniques because we’re used to giving information aren’t we and obviously we have to deliberately not do that and reflect things all the time and that wouldn’t come naturally to most practitioners.

FG1

. . . the fact that it could be delivered by an OT and somebody within rheumatology, it was quite exciting to be able to do that.

INT11

Requiring time and effort

At the start of the study, participants had not anticipated the high levels of time and effort that would be expended during training and delivery of the RAFT programme intervention. The amount of hard work required was due to the challenge of becoming familiar and confident with a large amount of material, delivered using a new approach, to patients in a group setting. Participants described a sense of responsibility and commitment to delivering the RAFT programme to the best of their ability:

It wasn’t something you were just going to be able to go away with and think ‘I could just do this’ . . . you really did need to know your material. You needed to be well read . . . before you delivered . . . and it made you realise that it, there was quite a lot of preparation to be done and it wasn’t just something you could run off on a whim.

INT6

I remember just taking it very seriously, that it wasn’t a practice run at all and putting as much effort and time as I could just to try and understand the manual and get it across.

INT9

We spent hours, hours and hours sitting and going through it and going through it again and writing it out on to cards.

INT12

Feeling challenged

Along with being enthusiastic about the RAFT programme, participants reported feeling challenged as they started to engage with the training and intervention:

. . . interested, enthusiastic, I just wanted to try and do it.

INT12

. . . it was a bit scary actually and I felt quite challenged.

INT1

I did feel very daunted starting it. I was very nervous delivering the groups as it first began.

INT7

. . . it was the managing the group and actually when to try to stop somebody or try and include somebody else and the actual feeling at one point . . . you know the group can change and there are dynamics and sometimes it can feel quite . . . certainly, intimidating sometimes.

INT14

Not as easy as it looks

When the RAFT programme trainers demonstrated sessions during training, the delivery flowed and the interaction appeared natural and almost effortless. When participants practised, they found it difficult to emulate this sense of being at ease with the material and with the group. Although, initially, this diminished participants’ confidence as they compared their performance to that of the RAFT programme trainers, it also helped to get a sense of what participants were aiming for with the intervention:

I didn’t feel particularly confident in delivering it as well because you’ve got professionals who have shown us how to do it, they’re so good at what they do you know it’s quite daunting to think ‘Oh crikey how can you do that?’

INT3

I mean they made it look so easy and then when you try and do it yourself it is like ‘Let’s have a look at my card again and I’ll just have a look at the manual again’. Whereas they were just able to do it.

INT7

[RAFT programme trainers] made it look so simple, so easy. I just think ‘why can’t I do it like that?’ they’ve made it look so easy.

FG2

Being new to the RAFT programme together

Participants felt sufficiently comfortable with each other and with the training team to engage and take part in the training. They found it helpful to realise that they were all new to the RAFT programme and to be able to share their thoughts and concerns:

. . . you definitely have that camaraderie and that you can continue to bounce things off each other a little bit or whatever, I personally like that environment, and I like to learn in that way.

INT2

It’s reassuring in a way to know that you have people with the same mind-set and obviously finding that opportunity to explore this territory further helps us because it’s like putting a blank canvas and at the end of the day it’s like painting it together.

INT4

. . . having the group dynamics of different nurses, allied health professionals, I think we fed off each other.

INT6

Learning from expert demonstrations

Observing the RAFT programme trainers demonstrate sessions brought the RAFT programme intervention to life, adding depth and detail to the manual. Demonstrations were very helpful for facilitating participants’ learning and understanding of what the intervention should look like in action:

I thought it was quite good when they demonstrated, they demonstrated how to use the CB [cognitive–behavioural] approach in the group setting using the manual. That was good because I work best by being able to observe and sort of see how you do things rather than reading it off a page.

INT7

We sat back and observed didn’t we and we actually watched them do a session. That was really helpful when we didn’t know how it was going to do anyway, to actually see it happen that was really helpful.

FG6

Role play: invaluable despite the discomfort

Participants acknowledged role play as one of the most useful aspects of the central training. Practising parts of the sessions out loud then getting feedback from the RAFT programme trainers was invaluable in the process of becoming familiar with the intervention, although it could be uncomfortable at times:

I mean I hate role play; I absolutely detest it. It is just not in my nature to do it. I just think ‘God’, but it was good because it gave us the chance to practise it, and to get feedback.

INT7

For me to actually have a go and to practise . . . I found that was very useful even though I didn’t like it, I would say one of the most valuable things of the course. I think we should have more of it, although people will hate me for saying that.

INT8

I think you need to have that practice, that . . . it’s not until you realise the benefit, I said this before, about the role play and that and actually practising saying things and scenarios.

FG2

I didn’t enjoy the role plays at all but if we had to do some more training that is what I would want is what I’m saying and I would know the benefits of it and that’s probably . . . for me it was the only way that I could actually practise and be semi-confident going out for the pilot [practice run].

FG4

Personalising the RAFT programme manual

Participants described consolidating and deepening their understanding of the RAFT programme intervention through adapting their manual, including writing summaries and rephrasing text. This was done with great care to ensure that delivery stayed true to the underpinning CB approach. Participants explained that this was a difficult but essential part of the process of becoming a competent RAFT programme tutor:

I was worried about if I changed it in certain ways that actually the fundamental way of interacting with the patients wouldn’t have been right, do you see what I mean? So because you’re trying to use that cognitive–behavioural approach, I was conscious that maybe if I alter it too much, I wouldn’t be doing that and therefore you wouldn’t get where you wanted to with it, but at the same time, the language, for me, didn’t sound familiar for me, so it’s trying to get that balance of extracting the essence of what you’re getting from it and putting it across in a way that feels comfortable for you as a practitioner.

INT2

That was one of the hardest things I think we’ve found. Is actually putting it in our words because . . .

We really struggled to start with because we were trying to learn it as a script and it wasn’t how we would say things. So, yeah, we spent hours didn’t we . . . Hours taking it all apart.

Sort of changing it. Changing bits to how we would speak.

And we had to have crib sheets for every session.

Delivery improved through clinical supervision

In addition to the 4 days of central training, participants reported that ‘you need the feedback on your performance’ (INT13) when delivering to patients. The RAFT programme trainers’ ongoing support, comprising clinical supervision and feedback on observed sessions, was described as very helpful. Feedback was perceived as constructive and a mechanism for enhancing participants’ confidence and skills. Occasionally, participants sought specific guidance, aware of the RAFT programme trainers’ high level of skills and familiarity with the intervention:

[RAFT programme trainer] said ‘You really need to get up and get the people engaged more, and you were struggling to get the information from them’ and that was really, although it was a bit of criticism it was really useful because the next time we did that session we did it differently and it was much better.

INT1

I think it gave us more confidence that they are there, we can actually ask questions afterwards and then they can give us feedback on how we could have delivered it better or what are we lacking in terms of explanations or something.

INT4

. . . just please give me an answer and then I can decide whether that’s what I would do or not, but I want to know what you think I should do.

INT8

I think the feedback after the pilots [practice run] was essential, as much as it was really scary to have someone observe you the feedback and the debrief afterwards was really good to just give you focus, direction and just reassurance that you were doing something right.

FG5

Using the dynamics of pair work

In addition to the work that participants did to consolidate their individual understanding of the intervention, the participants developed ways of working with their co-tutor that were mutually supportive and allowed them to each play to their strengths:

. . . you need two people, if you’re delivering stuff and you can’t, you listen to what is going on but sometimes you can miss other things that are going on in the room and having the other person there just reminds you to pick it up as well and should you get stuck or lost, it’s good to have the other person there because they could always pick it up for you, it was back up, it was reassurance to have two of them there and I think that worked really well.

INT8

. . . you’re helping each other out, aren’t you, you’re writing things up on the board and even if the one is struggling a bit, then we would, we tried, you have to be careful not to take over each other’s little roles, but we did just support each other.

INT10

I think it’s really important that you do it in pairs, really important because we helped each other didn’t we?

FG3

Working with the whole person

During delivery of the intervention, participants explored with patients ways in which fatigue can permeate all spheres of patients’ lives and well-being. This heightened the tutors’ awareness of the importance of looking at all of the individual and contextual factors that might be impacting on patients’ health in their wider clinical practice, and not focusing solely on a set of symptoms:

. . . it’s like you become more sensitive of the person that you’re seeing all the time and you sort of separate it slightly different in a way that am I just going to be focusing on treatments or am I going to prompt them to say something about, you know more on how they feel, how they live their life and how they do their day to day, so it helps in a way.

INT4

. . . that’s been so helpful just in day-to-day treatment with patients, just being able to get a bit more out of them and . . . yeah, it’s a really, really useful skill.

INT11

I feel I’m a better practitioner, I feel I’m more compassionate and empathetic. I think I look at them much more holistically as opposed to ‘Right what drug can I throw at them now?’

INT13

Knowing how to draw things out, sit back and listen

Participants contrasted using new approaches to support self-management in their clinical practice, such as ‘ask don’t tell’, with their previous ways of working, such as advice-giving. In particular, they identified that eliciting patients’ thoughts and feelings and letting them come up with answers to challenges was beneficial:

I think as nurses you tend to often want to give the answer all the time and give advice and it’s very nursey to do that, but it’s learning when to listen and stand back and try and get the patients to find the answers more rather than you delivering the answers to them and the solutions all the time.

INT6

Confidence that you have the knowledge, but you question them to bring up the answer, I do that a lot more.

INT10

It’s so easy and quick to see, say a hand deformity and think that’s clearly the problem, but it’s actually establishing ‘Why is that causing the problem, how is that affecting you functionally’ and it’s, yeah it’s just drawing out the, yeah a better insight really and getting that more out of a patient . . . getting them to open up a bit more which still does often require a box of tissues sometimes.

INT11

More confident talking about fatigue

Participants noted an increased confidence to discuss fatigue in their everyday clinical practice, in the knowledge that they had some ideas, skills and tools for supporting patients. Often these discussions involved adapting or distilling parts of the RAFT programme session:

I got used to using the [activity] diaries quite happily with those and trying to set goals from those was quite useful as well. So, used them a lot more in practice now with people which is good. I feel a lot more happy to talk about fatigue and that with patients on a one-to-one session now rather than before.

INT7

I liked the communication one [session] and I’ve used some of those concepts a lot in clinic because I feel that it’s quite useful even in, a few sentences in the clinic we have got somebody who has got fatigue and they’re isolating themselves, you can talk to them about communicating with family and friends.

INT12

Buy-in from managers and clinical colleagues

At an individual level, participants would be delighted if they could deliver the RAFT programme intervention in their clinical practice. This would typically require a business case from managers and the support of colleagues and clinical leads. Participants acknowledged that there might need to be some flexibility in relation to who delivers the intervention. There was also enthusiasm about the possibility of delivering it to other patient groups. Although participants could envisage some changes to delivery, they did not think that the intervention could be adapted in relation to the number of sessions and the topics covered:

We were the same, I think it’d be a business case.

Yes, because [RAFT programme local principal investigator] is looking at a business case for us to do it. [RAFT programme local principal investigator] bought into it but that’s as far as it’s got. I don’t know how far [RAFT programme local principal investigator] has got with the business case.

Is that an OT manager or a . . . who is that?

[RAFT programme local principal investigator] our consultant.

And certainly, I don’t think they’d put our banding in either, because we’re band 7s, they’d probably put lower bands in.

INT10

I just want to roll it out for everybody now, and I want to have CTD [connective tissue disease] groups as well as the RA groups and general arthritis groups that would be good. And also fibromyalgia patients as well, it would be interesting for everybody really.

INT12

. . . it’s difficult to think about things that could maybe be cut out to make it quicker because I think actually the benefit of it has been that all of the areas have been covered.

INT11

Models of training and support

Although participants valued their 4 days of central training, they acknowledged that this is unlikely to be a feasible model to roll out. One option was new RAFT programme tutors observing the intervention being delivered by experienced RAFT programme tutors, either in a live setting or on a digital versatile disc (DVD). However, there was a strong sense that some face-to-face training with skills practice would be important. Participants also stressed the need for clinical supervision to ensure a level of fidelity to the process:

I wondered if new people could maybe come and observe people who are already doing the course perhaps, after the training?

FG3

. . . you’re reading through the manual and you think that you’ve got your head around what it is that you’re doing and then when it comes to actually delivering that again which is why the practice is so good because at least you get that practical element of it but trying to put those words into something that made sense to me as a non-psychologist so having something like that [DVD], a kind of more visual thing to use alongside it I think would be really useful.

FG7

I think you do need some supervision as well; you need somebody to come down, every now and again to go through because you can get into bad habits.

INT8

I think you’d want to be supervised by people who’ve got experience in it and if there aren’t people with experience in it, then a support network amongst each other, I think.

INT10

. . . it’s good to have some kind of structure and observation just to check that people are doing it right because you can just assume can’t you and then it not being delivered . . . as required.

INT11

Wonderful seeing how patients can benefit

It was rewarding for participants to see patients benefit from taking part in the RAFT programme intervention. They described how patients’ engagement with the intervention evolved during the seven sessions and the satisfaction of seeing positive changes over time. There was the sense that some patients, who had taken part in RAFT programme groups, might stay in contact and organise to meet up outside the hospital and, subsequently, be an ongoing source of peer support:

. . . it was really lovely to see that light bulb moment for some patients, to click, I love that bit, that’s really satisfying.

INT3

I can think of, well, quite a few patients said it’s been life-changing.

INT10

. . . after the last session they [patients] had already met up for coffee, they went shopping, they’re giving each other lifts so it’s good to know that they had managed to find a group that they think they can get support as well if they need to which is reassuring.

INT4

. . . it just gives you such a lift just seeing how well they look and how well . . . the groups interacted and how they’re staying in touch and I think as a therapist that’s what you want to see, that’s why you do your job really.

INT11

Summary

The trial demonstrated that improvements in fatigue impact were made in both the control and RAFT programme trial arms. However, the RAFT programme resulted in a greater improvement in fatigue impact at 26 weeks than usual care alone (BRAF-NRS impact score adjusted mean difference –0.59; p = 0.03), with treatment differences lasting for the 2 years of follow-up (BRAF-NRS impact score adjusted mean difference –0.49; p = 0.01). A stronger effect on the primary outcome of fatigue impact was seen in those patients who met the eligibility criteria of BRAF-NRS fatigue impact ≥ 6 at baseline, at both 26 weeks (BRAF-NRS impact score adjusted mean difference –0.82 units; p = 0.01) and 2 years (BRAF-NRS impact score adjusted mean difference –0.58 units; p = 0.02). Similar treatment differences lasting over the 2 years were seen for overall impact of fatigue (BRAF-MDQ) and subscales emotional fatigue and living with fatigue; in addition, self-efficacy and coping with fatigue also showed treatment differences at 26 weeks and over the 2 years, respectively. There was no evidence of a treatment effect on fatigue severity, cognitive fatigue or other clinical variables. Patient attendance at the RAFT programme sessions was high, as were satisfaction with the programme and willingness to recommend it to other patients.

The RAFT programme targeted the impact of RA fatigue on patients’ lives and there was a treatment difference between trial arms in favour of the RAFT programme for reducing fatigue impact as measured by both the BRAF-NRS impact and BRAF-MDQ overall impact scales, as well as BRAF-MDQ subscales of emotional fatigue and living with fatigue. Patients in the control arm also reported improvements in fatigue impact, which was not seen in the previous RCT of this intervention in which the control was a group session comprising 1 hour of didactic information on general self-management of RA. 31 The usual-care booklet, written after the first RCT and based on the results of that intervention,45 has been available in rheumatology units since 2011 and was therefore used as a ‘usual-care’ active control in the present study. Although the booklet has never been subjected to rigorous effectiveness testing, a recent qualitative study88 suggested that patients felt that it made them think differently about their fatigue. Patients in the study88 reported that having a discussion of the booklet with a researcher made them consider taking responsibility for managing some aspects of their fatigue and try the suggestions the booklet contained. Although the focus of the qualitative study was for patients to review the booklet for research, it raises the possibility that the brief booklet discussion with the health professional (usual care in this RCT) may have delivered an effect over and above a patient simply picking the booklet up from the clinic display.

The trial demonstrated an effect size of 0.36 on fatigue impact. As recommended,62 in interpreting these findings against the traditional 0.5 guideline for clinically meaningful effect sizes, the broader evidence relating to the specific issue of RA fatigue (see Chapter 3, Interpretation of findings) was examined. The level of change in fatigue impact that might be clinically important for people with RA is unknown, as measurement of fatigue impact is relatively new. However, in terms of fatigue severity, a study to examine the level at which patients reported a noticeable improvement in their RA-related fatigue demonstrated that this occurred at effect sizes of 0.27–0.39 (three different anchors used for analysis). 63 Non-pharmacological interventions in RA are normally given in addition to pharmacological management, meaning that findings must be placed in this context. For example, meta-analysis shows RA patient education to have an effect size of 0.16 on painful joints compared with 0.66 seen in RCTs of non-steroidal anti-inflammatory drugs. However, all patients in the patient education trials were already taking these; thus, the 0.16 effect size represented additional benefits. 89 Similarly, the RAFT programme study was conducted with patients who were already receiving routine clinical management of their RA, that is pharmacological therapies to address inflammatory activity, which meta-analysis shows to have an effect size of 0.43 on RA fatigue;16 therefore, benefits in both intervention arms were over and above existing substantial improvements. Three studies analysing the minimal clinically important difference (MCID) on a number of fatigue severity scales in RA suggest that a change of between 7% and 11% would be clinically important. 63,90,91 This approximate 10% RA fatigue severity MCID might be a guide for meaningful change in RA fatigue impact, and comparable to multiple sclerosis, for which results from two studies average a fatigue impact MCID of around 15%. 92,93 The RAFT programme demonstrated an adjusted mean difference of 0.59 in BRAF-NRS fatigue impact beyond the improvement seen in the usual-care arm, with the RAFT programme showing a 19% change from baseline mean and the control arm showing a 12% change from baseline mean (fatigue impact improving by 1.36 and 0.88 units, respectively).

The improvement of 0.59 in BRAF-NRS fatigue impact beyond the improvement seen in usual care is smaller than the RCT was powered to detect, which was an improvement of 1.45 units. This had been based on an assumption that the clinical team delivering CBT might achieve around 75% of the 1.95 fatigue impact change seen in the earlier trial, in which delivery was by a clinical psychologist and in which the usual-care arm experienced no change at all (usual care comprised a 1-hour group session of RA self-management information delivered didactically by a nurse). 31 The statistical significance of the 0.59 improvement would have been enhanced by the lower than expected attrition rate. Therefore, taking the wider evidence from the trial and the literature presented above, it seems likely that in the current RCT the improvement in fatigue impact achieved by the RAFT programme beyond that of usual care may be clinically meaningful. This is further supported by the high levels of patient satisfaction demonstrated in attendance rates, and satisfaction and recommendation scores. However, further work to establish the MCID for the BRAF-NRS impact would be beneficial.

The treatment differences for the primary outcome of fatigue impact were greater in those patients who were eligible for the RAFT programme (i.e. had a fatigue severity score of ≥ 6 out of 10 at baseline). As with any symptom, RA-related fatigue severity will naturally fluctuate slightly from week to week; therefore, it is not surprising that, during the time it took to build cohorts in each centre, some patients’ fatigue severity over the previous 7 days was ≥ 6 at screening, but was reported as < 6 at their later baseline assessment. However, the screening process also required that patients considered their fatigue to be a recurrent problem, thereby allowing for normal short-term fluctuations in severity. The RAFT programme demonstrated significant treatment effects even including those patients with less severe fatigue (i.e. using complete-case analysis). This suggests that the RAFT programme is a pragmatic intervention to deliver clinically, as its timing in relation to symptom severity may not be crucial; it still had an effect on a recurrent symptom even when intervention delivery was delayed by the need to build an adequately sized group, or to accommodate patients’ personal availability.

There was no treatment difference between trial arms for change in fatigue severity at 26 weeks, although both trial arms had small reductions, which for the RAFT programme patients reached the proposed MCID63 for people with RA-related fatigue (MCID –0.82 to –1.12: control, –0.72; RAFT, –0.97). It is possible that the RAFT programme participants made a shift in how they view and manage a persistent symptom, finding a way to reduce the impact of fatigue on their lives. 32 Large longitudinal observational studies of current best practice pharmacological management aiming to control RA disease activity have demonstrated that fatigue severity remains largely stable over the years,94 with the majority following either a persistent moderately high or a persistently high fatigue trajectory and only a minority (one-third) having an improving fatigue trajectory. 95

Symptom severity and symptom impact are different concepts. For example, two studies have found that RA patients can report low impact from high levels of disability, whereas others report that their low level of disability has a high impact. 96,97 Personal impact is likely to be a result of the interactions between symptom severity, symptom importance in the individual context of a person’s life and perceived ability to manage or cope with that symptom: the ‘impact triad’. 98 There was a treatment difference in favour of the RAFT programme in coping with fatigue over the 2 years (as measured on the BRAF-NRS coping). The perceived ability to cope or manage fatigue better may reduce the importance of fatigue in an individual’s life; thus, it is understandable that fatigue impact could still be changed by the RAFT programme even without significant improvements in fatigue severity (the third side of the impact triad).

The effects of the RAFT programme lasted over 2 years without further RAFT programme input after the 1-hour group session at week 14, suggesting that patients’ newly acquired fatigue self-management skills became embedded. This is supported by evidence of a treatment difference between intervention arms for improvements in self-efficacy (the belief or confidence that one can successfully make behavioural changes), perhaps translating into the improvements in perceived coping over the 2 years. Self-efficacy, a key process underpinning behaviour change, was measured with the RASE56 scale (assessing beliefs about RA self-management behaviours), and six of the RASE scale component items demonstrated differences between trial arms for change over weeks 0–26 in favour of the RAFT programme, including key fatigue self-management activities such as pacing and acceptance of fatigue. Problem-solving and goal-setting, as a route to enhancing self-efficacy,25 are core self-management skills that patients learn in CBT and thus these changes in coping and self-efficacy may reflect the CB approaches that the clinical rheumatology tutors were trained to use in the RAFT programme.

The CB approaches were delivered by clinical rheumatology nurses and OTs after a brief training, rather than clinical psychologists. This novel approach aimed to circumnavigate the lack of psychology support available in most rheumatology units33 and shows similar results to a RAFT programme intervention delivered by the usual clinical team for multiple sclerosis fatigue. 99 A systematic review of nurse-led psychological interventions for depression in people with long-term physical conditions demonstrated their effectiveness. 100 In addition, a systematic review of psychological interventions for people with diabetes showed generalist nurses/physicians to be equally as effective as psychologists. 101 Delivery of the RAFT programme intervention by the clinical team was feasible, in that all RAFT programmes were delivered as planned and tutor sickness/absence could be managed. Furthermore, the RAFT programme feasibility was supported by high patient attendance and satisfaction rates; these suggest that the RAFT programme could be implemented in practice.

Summary

The primary economic evaluation found no difference between intervention arms in either overall costs (which incorporated intervention training and delivery costs) or QALY gains at 26 weeks or 2 years, and there was no evidence that the RAFT programme reduced other health-care costs. However, there was substantial uncertainty around the cost and QALY estimates. The primary cost–utility analyses suggested that the RAFT programme is not very likely to be cost-effective at the conventional societal WTP values, with probabilities below the 50% level (probability of 28% at the £20,000 NICE threshold and a probability of 35% at the £30,000 NICE threshold) from the societal perspective. A subgroup analysis of those patients who were eligible at baseline (i.e. BRAF-NRS impact ≥ 6/10) gave a 52% probability that the RAFT programme was cost-effective at £20,000 per QALY, or 60% probability that the RAFT programme was cost-effective at £30,000 per QALY.

The sustained improvement in fatigue impact seen clinically over 2 years was not reflected by increased QALYs measured using the EQ-5D-5L. This may be indicative that the EQ-5D-5L is not responsive enough to capture improvements in RA-related fatigue. Of the five EQ-5D domains (mobility, self-care, usual activities, pain and mood), ‘usual activities’ is most likely be the domain to be affected by a fatigue self-management intervention, a theory supported by the correlation demonstrated here between change in EQ-5D-5L domains and change in BRAF-NRS impact score. However, ‘usual activities’ is the EQ-5D domain that contributes least to the overall utility weights. 102 The EQ-5D ‘usual activities’ domain includes the ability to perform work, housework and family and leisure activities, and people with RA value doing such ‘daily activities’ as the second most important of seven health domains, with fatigue as third, giving highest weights to these two. 103 This contrasts with the low weight for ‘usual activities’ and absence of fatigue assessment in the EQ-5D, raising questions about the relevance of the EQ-5D for evaluating quality of life in RA. Energy (or drive) and sleep have been identified as key domains for people with RA, and these may be missed by the EQ-5D, leading to recommendations for further research to identify relevant domains and weightings for the RA population for use in the EQ-5D. 104

There is robust evidence to support the satisfactory psychometric performance of EQ-5D in RA. 105 However, it has also been shown that several commonly used quality-of-life measures for calculating QALYs in RA give a wide range of results on the same data set, demonstrating cost–utility in favour of a bDMARD that varies from 12% to 91% from four instruments (EQ-5D, 63%). 106 Previous research has also suggested that the EQ-5D is more responsive to worsening health states than to improving health states in RA patients. 107 However, much of this research has not evaluated the more recent five-level EQ-5D (i.e. EQ-5D-5L) used in this RCT, which might be more responsive to changes in health.

Economic evaluations of interventions designed to improve fatigue in RA are lacking, although cost-effectiveness has been assessed in other long-term conditions. A short-term economic evaluation of a fatigue management programme for patients with multiple sclerosis found similarly insignificant between-arm differences in EuroQol-5 Dimensions, three-level version (EQ-5D-3L), derived QALYs, despite significant improvement in fatigue in the intervention arm. 99 A cost-effectiveness study comparing a rehabilitation programme with a CBT intervention for patients with CFS over 12 months also found no statistical differences in EQ-5D-3L-derived QALYs, despite a significant difference between the intervention arms for fatigue severity. 108 These studies support the view that the EQ-5D may not be responsive to changes in fatigue in long-term conditions.

Work productivity was improved for patients in the RAFT programme intervention arm at 26 weeks and 2 years, although the difference between intervention arms was not statistically significant. However, the number of patients in paid employment at baseline was small (n = 76, 23% of randomised patients), suggesting that the power to capture a significant difference was low. There has been little research on the effect of RA-related fatigue severity on work productivity, but a study of > 8000 RA patients in 32 countries demonstrates greater fatigue severity in patients who are work disabled than in those staying in work. 109 The BRAF-NRS impact score is a relatively new measure and there have been no reported quantitative studies examining the relationship between fatigue impact and work productivity. However, qualitative studies consistently report that people with RA consider the impact of fatigue on work to be significant;8,12 thus, improvement in work productivity could be consistent with the reduction seen in fatigue impact with the RAFT programme. The EQ-5D scores are derived from surveys of the general population. Recent work comparing people with RA with the general population found that both groups placed a similar value on pain and function limitations, but the general population put a lower value on the ability to work. 110 This suggests that the EQ-5D scores may have been different had they been based on values from people with RA rather than the general population, which is part of a wider debate on obtaining better-informed population preferences for estimating QALYs. 111

The algorithm for work productivity was based largely on items relating to hours lost (absenteeism) and included just one item on presenteeism. Fatigue causes major difficulties for people while at work and may mean that they are present but not functioning at full capacity (presenteeism). 8,12 An uncontrolled pilot study of a self-management programme specifically for fatigue at work for patients with inflammatory rheumatic diseases examined presenteeism and demonstrated improvements in four domains of functioning while at work: physical and mental demands, work scheduling and work outputs. 112 It is possible that the weekly review of daily activity diaries in the RAFT programme, accompanied by encouragement to break up large periods of intense activity through factoring in short breaks, may help presenteeism.

Summary

Five themes were identified in the tutors’ data. The rheumatology nurses and OTs (tutors) were interested and enthusiastic about being involved in the trial of a novel fatigue intervention. The tutors described delivering the RAFT programme as a very rewarding experience, seeing patients make big changes to their lives, although the process of becoming familiar with the material and gaining confidence in delivering it proved hard work (theme 1). The tutors’ perceived personal benefits included a sense of professional development and fulfilment, especially when they saw patients making life changes. Learning together and rehearsing the RAFT programme elements as a group was helpful (theme 2). Challenges included using an ‘ask don’t tell’ approach to prompt patients to generate their own ideas for behaviour change, as it contrasted with tutors’ usual advice-giving and problem-solving approach. Tutors’ understanding of the intervention was consolidated through re-reading and individualising their manuals (e.g. paraphrasing sections), and feedback from the RAFT programme trainers during clinical supervision (theme 3). Over time, tutors deepened their understanding of how a CB approach and the group setting could be used to support patients in self-managing the impact of their fatigue. In addition, the usefulness of the skills gained by tutors extended beyond the RAFT programmes and into their everyday clinical practice (theme 4). There was widespread support for delivering the RAFT programme in clinical practice beyond the trial, as tutors perceived that patients were benefiting (theme 5). Tutors valued their 4 days of central training, but believed that it would not be a feasible future clinical model. Some face-to-face training with the opportunity to role play key aspects of the intervention, along with some clinical supervision and quality control, were deemed necessary in any future training of new tutors.

The qualitative findings support two key ideas about skills training and its subsequent transfer into clinical practice. First is the importance of role play, a method of simulation commonly used to teach communication skills. 113 In the RAFT programme central training, role play was employed through both ‘role reversal’ (i.e. participants taking on the role of patients to develop insight into what the group dynamics might feel like from a patient’s perspective) and ‘role training’ (i.e. participants practising the skills that can help them become more expert in their professional role). Tutors identified both forms of role play as helpful for learning. Second is the importance of clinical supervision, as although communication skills training enhances skills, it has limited effect on clinical practice without subsequent clinical supervision. 114–116

Tutors were aware that they were delivering a standardised intervention and made great efforts to adhere to the manual content. However, tutors also expressed their need to individualise the manual and to put material into their own words. The RAFT programme trainers and fidelity observer often helped in this process by feeding back on if and how tutors were incorporating essential techniques and approaches into the sessions. Fidelity was recorded in the observer template (see Report Supplementary Material 1), and confirmed that tutors were delivering the RAFT programme as intended. Fidelity to the RAFT programme is an important consideration in any future roll-out. Inevitably, interventions are adapted to individual settings, resources available, or practitioners’ beliefs and interpretations, once they are introduced into clinical practice. Such changes may reduce or enhance outcomes seen in clinical practice and thus there is an argument for capturing not only outcomes but implementation fidelity and adherence to delivery format, content and underpinning CB theory in the roll-out of such interventions into clinical practice. 117

Tutors’ perceptions that the RAFT programme training and subsequent delivery impacted and benefited their wider clinical practice highlights the potential of up-skilling existing members of clinical teams to provide low-level psychological support. Evidence from studies of other long-term conditions, such as diabetes, cardiac rehabilitation and depression in long-term physical health conditions, shows that the incorporation of psychological skills into the nursing role is viewed positively by both nurses and patients. 100,118,119 However, the research evidence reiterates the finding that nurses and allied health professionals need time and support to acquire and use these skills.