Lifestyle coaching is feasible in fatigued brain tumor patients: A phase I/feasibility, multi-center, mixed-methods randomized controlled trial

Abstract Background There are no effective treatments for brain tumor-related fatigue. We studied the feasibility of two novel lifestyle coaching interventions in fatigued brain tumor patients. Methods This phase I/feasibility multi-center RCT recruited patients with a clinically stable primary brain tumor and significant fatigue (mean Brief Fatigue Inventory [BFI] score ≥ 4/10). Participants were randomized in a 1–1–1 allocation ratio to: Control (usual care); Health Coaching (“HC”, an eight-week program targeting lifestyle behaviors); or HC plus Activation Coaching (“HC + AC”, further targeting self-efficacy). The primary outcome was feasibility of recruitment and retention. Secondary outcomes were intervention acceptability, which was evaluated via qualitative interview, and safety. Exploratory quantitative outcomes were measured at baseline (T0), post-interventions (T1, 10 weeks), and endpoint (T2, 16 weeks). Results n = 46 fatigued brain tumor patients (T0 BFI mean = 6.8/10) were recruited and 34 were retained to endpoint, establishing feasibility. Engagement with interventions was sustained over time. Qualitative interviews (n = 21) suggested that coaching interventions were broadly acceptable, although mediated by participant outlook and prior lifestyle. Coaching led to significant improvements in fatigue (improvement in BFI versus control at T1: HC=2.2 points [95% CI 0.6, 3.8], HC + AC = 1.8 [0.1, 3.4], Cohen’s d [HC] = 1.9; improvement in FACIT-Fatigue: HC = 4.8 points [−3.7, 13.3]; HC + AC = 12 [3.5, 20.5], d [HC and AC] = 0.9). Coaching also improved depressive and mental health outcomes. Modeling suggested a potential limiting effect of higher baseline depressive symptoms. Conclusions Lifestyle coaching interventions are feasible to deliver to fatigued brain tumor patients. They were manageable, acceptable, and safe, with preliminary evidence of benefit on fatigue and mental health outcomes. Larger trials of efficacy are justified.

Brain tumor-related fatigue is a common and pervasive problem. Frequency estimates for fatigue range from 39 to 96% 1,2 seemingly irrespective of brain tumor grade, 3 histology, 4 or clinical time-point, 5 and often comorbid with cognitive impairment or depressive symptoms. 2,6 The profound impact of fatigue 4,7 is widely recognized. For example, The Brain Tumour Charity highlights its devastating impact on a person's capacity to cope with daily life, 8 while the UK James Lind Alliance considers the management of fatigue to be a top research priority for the neuro-oncology community. 9 There is a lack of evidence to underpin effective management of brain tumor-related fatigue, 10 despite several psychostimulant drug trials over the past decade. Trials of Methylphenidate and Armodafinil included non-fatigued patients, [11][12][13] while studies of Modafinil, Dexamphetamine, and Armodafinil recruited only highly fatigued patients but found no evidence of benefit over placebo on their primary outcome of fatigue. [14][15][16] Non-pharmacological treatments are a proposed alternative to medication for many cancer-related symptoms. [17][18][19] Some of these treatments incorporate more complex "lifestyle coaching" approaches that aim to reinforce healthy behavior, 20 social activity, 21 stress reduction, 22 exercise, 23 and/or better dietary habits. 24 Lifestyle coaching interventions have shown promise in alleviating fatigue in patients with cancers arising out-with the CNS. 25 Within neurooncology it was established recently that exercise 23 and yoga 26 are feasible to deliver to non-fatigued patients. However, lifestyle coaching interventions have not been studied in highly fatigued patients. These patients are often quite impaired, and it is far from clear whether they would find such interventions feasible and acceptable.
We therefore developed BT-LIFE (Brain Tumors, Lifestyle Interventions, and Fatigue Evaluation), a pilot (Phase I/feasibility) randomized controlled trial of two lifestyle coaching interventions for clinically significant brain tumor-related fatigue. The first intervention, "Health Coaching" , promoted a healthy lifestyle. The second, "Activation Coaching" , promoted self-efficacy. Both interventions are described in detail below. Our primary aim was to determine the feasibility of recruiting and retaining brain tumor patients with moderate or severe fatigue to HC and AC interventions. Secondary aims were to determine interventional acceptability, engagement, and safety; and to gather exploratory outcome data to inform a larger trial.

Trial Design
BT-LIFE was a Phase I/feasibility, multi-center, mixedmethods, three-arm randomized controlled trial with 1:1:1 allocation stratifying by center (ISRCTN17883425, Supplementary Figure S1). The study was coordinated by the Scottish Clinical Trials Research Unit. The final study protocol is presented in Supplementary Methods.

Participants
Eligible participants were: aged 18+; diagnosed with a primary brain tumor of any grade or histological subtype; ≥ 3 months after completion of primary treatment (any combination of surgery, radiotherapy and/or chemotherapy) with no evidence of disease progression; and clinically significantly fatigued (mean Brief Fatigue Inventory [BFI] score of at least 4/10). 27 Exclusion criteria were: low fatigue (mean BFI < 4/10); clinical concern about disease progression; clinically adjudged severe cognitive, language, or visual impairment; or inability to give informed consent.
Participants were recruited from neuro-oncology outpatient clinics in three UK centers (Edinburgh, Glasgow, and Manchester). Patients describing fatigue in their usual clinical appointment and who showed interest in the trial were screened for eligibility. WHO tumor grade was based on WHO 2016 diagnostic criteria. Baseline and follow-up trial outcome data were gathered face-to-face. These outcome assessments usually occurred in clinical settings with reimbursement of travel costs. Rarely (< 5% of cases) other locations were used such as the participant's home.

Interventions
Control.-Control arm participants received the Brain Tumour Charity's information leaflet about fatigue and "Usual Care" . The information leaflet included written advice on managing fatigue. In our centers Usual Care included ongoing access to the neuro-oncology clinical team with routine scheduled follow-up appointments.
Health Coaching (HC).-Participants in this arm received the information leaflet plus Health Coaching (hereafter HC). HC was a multimodal lifestyle coaching intervention which promoted incremental patient-led improvements in: fluid intake (reducing caffeine and alcohol and drinking more water); sleep (promoting rest); diet (encouraging healthy eating); exercise and movement (increasing number of steps per day); and reducing stress. Our HC model was developed in private practice by an accredited personal trainer. This individual (GA) delivered HC to patients recruited in Edinburgh and Glasgow. In Manchester HC was delivered by an NHS physiotherapist (AD) working to the same format.
Participants were offered eight weekly HC sessions. In this pilot study none of the sessions were formally manualized. However, efforts were made to standardize them between sites and patients. The initial session was a standardized 45-min face-to-face assessment with the Health Coach (Supplementary Methods). Each participant was provided with an infographic reminder card (Supplementary Figure S2), a wearable step counter (Omnicron), and a daily home diary (Supplementary Figure  S3). Up to seven 30-min follow-up sessions were then delivered weekly by telephone or in person, informed by the diary and tailored to individual goals. These follow-up sessions followed a set structure (Supplementary Methods) with their content reflecting the goals and progress of individual patients.

Health Coaching plus Activation Coaching (HC + AC).-
Participants in this arm received the above, plus Activation Coaching (AC). AC was developed by the UK brain tumor charity brainstrust. AC was targeted at coaching Neuro-Oncology Practice improvement in participants' self-efficacy for managing fatigue. It was delivered by trained life coaches in two 1-hour sessions for each patient, with sessions separated by 4 weeks. Before each AC session participants completed the Patient Activation Measure (PAM) 28 to inform the coach about the individual's current level of knowledge, skills, and confidence about fatigue. AC then focused on promoting self-efficacy using Dilts' Logical Levels 29 in session one, and the FRAME 30 and GROW 31 models in sessions one and two (Supplementary Methods). During the trial this arm had been called simply "Patient Activation" . We have since re-named the intervention "Activation Coaching" to distinguish it more clearly from the original Patient Activation Measure to which it is unrelated.

Primary Outcome
The primary outcome was feasibility of recruitment and retention. Recruitment feasibility required a recruitment rate "equivalent to 60 fatigued patients over 12 months" . This rate would permit a larger trial to recruit in a timely manner. Retention feasibility required all-cause attrition of under 40% at the study endpoint. There is no agreed critical threshold for attrition in QOL outcomes from cancer RCTs. 32 In neuro-oncology attrition is often high: leading studies of QOL outcomes in highly impaired patients may report missing data frequency (even among surviving patients) of over 30% on outcome measures at four months. 33 In such an understudied group we also anticipated a process of learning and systems development to iteratively minimize both avoidable attrition, and the impact of unavoidable attrition, which might inform future trials of interventions for brain tumor-related fatigue.

Secondary and Exploratory Outcomes
Secondary outcomes were the acceptability of, and engagement with, the interventions among participants (evaluated by sessional attendance, diary inspection, and qualitative interviews). A further secondary outcome was to explore potential benefit via a quantitative analysis of outcome measures. In these exploratory outcomes participants completed the Brief Fatigue Inventory (BFI); 27 the FACIT-Fatigue scale; 34 the Hospital Anxiety and Depression Scale (HADS); 35 the Psychological Outcome Profiles Questionnaire (PSYCHLOPS); 36 the EQ-5D-5L; 37 and the Addenbrooke's Cognitive Examination III (ACE-III) 38 at T0, T1, and T2. More detail on these measures, including comments on Minimal Important Clinical Differences for the two fatigue scales, is provided in Supplementary Methods.

Qualitative Interviews
All those completing HC (+/− AC) were invited to participate in a semi-structured qualitative interview about their experiences and the perceived acceptability of the interventions. Interviews were guided by a template (Supplementary Methods), conducted by a qualitative researcher or allied health professional blinded to treatment allocation, audio-recorded, and transcribed. Interviews were audio-recorded, transcripts were analyzed using the framework method 39 : (1) Familiarization; (2) Construction of initial themes; (3) Indexing and sorting; (4) Review; and (5) Data summary. 39,40 A realist approach 41 provided a thematic framework (Context-Mechanism-Outcome) to analyze transcripts.

Sample Size Calculation
Sample sizes were not calculated for this feasibility study. 42 Results were used instead to calculate sample sizes for a future Phase II trial of efficacy (see Supplementary Methods).

Randomization and Follow-Up
After obtaining informed consent a Research Assistant administered baseline measures. They accessed www. sealedenvelope.com to randomize with stratification by study site. It was not possible to blind participants or Research Assistants to allocation. Follow-up assessments were conducted 10 weeks (T1) and 16 weeks (T2) after randomization. The T1 time-point was chosen to measure symptoms shortly after the end of the 8-week interventionassumed to represent the point of maximal signal, if any. The T2 time-point was chosen to allow a modest period of follow-up extending to a total of four months after baseline, to examine for early attenuation of any signal in this pilot study.

Statistical Analyses: Primary Outcomes
For recruitment feasibility, the number of patients recruited was compared to the target rate of 60 patients per year. The number and proportion of patients retained to the T2 endpoint was compared to the target that at least 60% of patients should be retained.

Statistical Analyses: Exploratory Outcomes
Box-and-whisker plots were produced for each outcome scale, broken down by treatment arm and separated by time-point. Changes to outcome scale scores from T0 (baseline) to T2 (endpoint) were compared among the study arms using a one-way ANOVA. Ninety-five percent confidence intervals were produced and two sided posthoc comparisons made using Bonferroni (equal variances assumed) and Dunnett's C (equal variances not assumed) statistics. A significance level of .05 was used. Waterfall plots were constructed to examine whether intervention arm effects clustered separately from control arm effects. Further statistical analysis detail is given in Supplementary Methods and below.

Data Protection and Ethics
All participants gave written informed consent and GDPRspecific consent. The trial team used encrypted nhs.net email for all written communication. The trial was ethically approved by the South-East Scotland Research Ethics Committee 2 (18/SS/0025) and the local Caldicott Guardian.

Primary Outcome: It was Feasible to Recruit and Retain Fatigued Brain Tumor Patients
During a 9-month recruitment period (16th August 2018-15th May 2019) n = 103 primary brain tumor patients were screened from whom n = 46 fatigued patients were recruited ( Figure 1, Table 1 and Supplementary Figure  S4). This recruitment rate equated to 61 participants in 12 months, establishing recruitment feasibility ( Figure  2A). At T2 endpoint 34/46 participants remained (retention = 74% [95% CI = 59, 86%], Figure 2B) establishing retention feasibility. Recruitment and retention were generally acceptable across sites and arms (Supplementary Table  S1). The most frequent reason for drop-out was withdrawal of consent due for example to interventional burden, difficulty traveling, family commitments, and medical illness (for full details see Figure 1).
In this small study the intervention arms were balanced on most characteristics, but participants randomized to HC were significantly younger. At baseline the groups also differed on the PSYCHLOPS outcome measure: participants in the HC arm reported better function and participants in the HC+AC arm reported poorer function on this scale, relative to controls (Table 1).

Secondary Outcome: Lifestyle Coaching was Acceptable, Manageable, and Safe
Acceptability and engagement.-Fifteen participants received Control and 31 patients were randomized to an intervention ( Figure 1, HC n = 16, HC + AC n = 15). Two participants randomized to HC + AC dropped out for unrelated medical reasons before starting any intervention. Those remaining started HC on average 17 days (SD = 9, range = 2-36) after randomization, completing a median of seven sessions (IQR = 6-8, mode = 8, range = 1-8, Figure  2C). The first AC session occurred after a median of four HC sessions and the second after seven. All patients who received the first AC session also completed the second.

Allocated to HC & AC (n = 15)
Received HC & AC (13) T1: Lost to follow-up (n = 1) No reason given (1) No reason given (1) Withdrew due to medical illness (1) Tumor progression (1) T2: Further loss to follow-up (n = 1) Retained at T2 endpoint (n = 11) Figure 1. CONSORT Flow diagram. Patients who declined to participate (n = 29) were typically either unable to commit to regular coaching sessions (eg, due to traveling distance, family duties, or holidays), or unwilling to receive the interventions. The patients listed as "other reasons" (n = 9) gave no specific reason for not taking part. .233 a range = 35-90%). Cell completion was similar between intervention groups and remained stable throughout the study ( Figure 2D 1 ) with no evidence of "diary fatigue" over time.
We examined step count data as an objective test of engagement with HC. In 23/31 participants returning step data, the mean weekly count was 6674 steps (SD = 2815, range = 2569-12 099), with no significant difference in frequency between intervention arms. In an exploratory post-hoc analysis there was a positive effect of coaching over time (mean steps in the first half of coaching = 6273/ week [SD = 2834]; second half of coaching = 7031/week [SD = 2798], paired-samples t-test P = .027). Normalized to their performance in the first half of coaching, participants showed a 17% increase in physical activity in the second half (SD = 29%, range −21 to +83%, Figure 2D 2 ). These modest improvements suggested that coaching could improve outcomes in fatigued patients.
Qualitative feedback.-Among patients who completed HC or HC + AC, n = 21 were interviewed in detail about their experience of the interventions. This sub-sample included 10 female and 11 male participants aged between 25 and 63 (mean age 44). Participants came from both the HC (n = 9) and HC + AC (n = 12) arms of the trial, and from all three sites: NHS GGC (n = 10), NHS Lothian (n = 8) and The Christie Hospital, Manchester (n = 3). They represented a spectrum of fatigue with moderate (n = 13) and severe (n = 8) levels of baseline BFI fatigue.
Overall acceptability was good although influenced by contextual factors, including patient expectations and attitudes to life, prior experience of making lifestyle changes, and the emotional and physical constraints of living with a brain tumor. The degree to which the coaching approach was perceived to "fit" with the individual's beliefs and way of life was important to overall engagement with the interventions.
Some patients found the interventions compatible with their needs whereas others expressed needs that were unmet by the coaching approach. Potential mechanisms of change included pragmatic factors such as goal-setting and monitoring, and supportive factors such as the motivational "push" from coaches and family or friends. Many participants also reported improvements in general wellbeing and awareness of health behaviors. Some individuals reported specific improvements in fatigue including being "better able to cope" with fatigue. Full qualitative results will be reported elsewhere.

Neuro-Oncology Practice
Severe depressive symptoms may limit potential coaching benefits.-Correlation of baseline measures with BFI improvement at T1 were negative ( Figure 3C 1 T0  T1  T2  T0  T1  T2  T0  T1  T2   4004   HC  Control  H  that coaching had greater impact in participants who were relatively less impaired at baseline. As there were no clear differences of effect between the two intervention arms, we combined them to create one "grouped interventions" arm in exploratory modeling analyses. Linear modeling showed a mean T1 BFI improvement of 2.5 points (95% CI 1.8, 3.2) in the "grouped interventions" arm compared to the control arm (Supplementary Table S5). This model explained 28.6% of BFI score variation at T1 (adjusted R 2 = 26.3%).
To explore whether depressive symptoms mediated this improvement, a second linear model considered baseline HADSdepression (HADS-D T0 ) scores in addition to allocation. In the "grouped interventions" arm, baseline HADS-depression was a significant explanatory variable for the change in BFI at T1 (Supplementary Table S5, Figure 3C 2 ). As baseline HADSdepression increased there was less improvement on the BFI following an intervention. This improved model explained 42.5% (adjusted R 2 = 36.8%) of the variance, suggesting that patients with lower depression scores at baseline may benefit more from lifestyle coaching for fatigue.

Discussion
It is feasible to deliver lifestyle coaching interventions to fatigued brain tumor patients. Willingness to take part and

Neuro-Oncology Practice
remain in BT-LIFE was high and most participants engaged meaningfully with the coaching procedures. Qualitative analysis suggested that the compatibility of coaching with an individual's prior experiences was a key mediator of engagement. There were favorable but preliminary signals of beneficial effects on fatigue, psychological outcomes, and depressive symptoms. Exploratory modeling suggested a limiting effect of higher depressive symptoms on the effect of coaching on fatigue. Our data suggest that patients who were moderately rather than extremely impaired, and who "believe" in a coaching approach, may be more likely to benefit-but much more study is needed.

Limitations and Strengths
In this pilot RCT we did not record the number of patients who declined to be screened, nor did we measure ethnicity or tumor histological subtype. Follow-up duration was relatively short meaning that sustainability of the interventions is not known. Analyses were not powered or intended to test efficacy but to the extent that quantitative data can be weighed, outcome scales conflicted on whether improvements were seen in both intervention arms or sustained until the study endpoint; reasons for these conflicts are not clear and will require more data to unpick. Whether the interventions can be scaled beyond the study authors also remains open: it will benefit future studies to manualize each separate coaching session. The optimal timing for coaching, and mechanisms of action, if any, are not known and likewise questions for future studies. Therefore apart from the primary outcome of feasibility, much of the current output is preliminary.
In conducting the study, we became aware that "coaching" itself remained poorly defined. We established that it is feasible for fatigued patients to engage with people who call themselves coaches but did not examine which elements distinguish a coaching intervention from a more typical clinical intervention. What aspects of coaching are irreducible? What makes a good coach? Who is best placed to coach cancer patients in healthcare settings? Others may know the answers to these questions, but we do not. Exploring them may be helpful for future studies.
Alongside these limitations BT-LIFE had several novel or notable features. To our knowledge it is the first study to establish the feasibility of targeting highly fatigued brain tumor patients with complex interpersonal interventions. It is one of a handful of trials to report a signal of clinical and statistical effect on fatigue in these patients. The study blended private, charity, and health sectors to present a successful model of "multi-sectoral research" . Multisectoral research partnerships in which each sector has something valuable to the others have great potential for synergy, innovation, and new ways of delivering new interventions that may benefit patients.

Results in Context
In neuro-oncology, there is mounting evidence from RCTs that psychostimulant medication is ineffective for fatigue and most secondary QOL outcomes, [11][12][13][14][15][16] while two pilot RCTs of non-pharmacological interventions have reported preliminary improvements in fatigue and neurocognitive functioning. 23,26 Although the effectiveness of any nonpharmacological strategy is yet to be proven for brain tumor patients, the wider oncology literature supports a shift in this direction: exercise, resistance training, core stability, psychological therapies, yoga, mindfulness, and hypnosis are listed among the therapies showing varying degrees of promise for fatigue in other cancer populations. 43 Engagement with HC overall was good. Every patient who started AC continued to the second session. Most patients who were interviewed reported benefits on symptoms, lifestyle behaviors, or ability to manage ongoing fatigue. These insights suggest that fatigue is not an intractable or treatment-refractory problem: it may simply be that such a complex symptom is better suited to management by complex interventions.

Next Steps
The next steps will study the essence, scalability, and efficacy of the coaching interventions piloted in BT-LIFE. A Phase II trial could formally manualize interventions, develop mechanisms to ensure blinding of outcome assessors, and perhaps stratify by symptom severity.

Conclusion
BT-LIFE established that lifestyle coaching interventions are feasible to deliver to fatigued brain tumor patients. Coaching interventions were manageable, acceptable, and safe, with preliminary evidence of benefit on fatigue and mental health outcomes. Further trials are justified to test the efficacy of lifestyle coaching interventions in this highly impaired patient group.