Based on the stroke clinicians’ (GM, MM) clinical observation that PSF often coexists with low mood and reduced physical activity and the psychotherapist’s (TC) experience in the management of chronic fatigue, we proposed a psychological intervention to treat PSF by targeting patients’ thoughts and behaviours. We discussed the concept of this intervention with a focus group of service users (10 stroke survivors and five carers, from the Yorkshire Stroke Research Network) for their preliminary opinions on its acceptability. The users shared their experience of PSF and considered that the proposed intervention would be acceptable.

In order to identify the intervention targets in a broader context of stroke illness, we conducted a systematic review to explore the psychological correlates of PSF [9]. Drawing on the evidence from this literature and with input from stroke physicians and clinical psychologists, we developed a stroke-specific model of PSF, which suggested that depressive symptoms, anxiety, lower self-efficacy, passive coping, reduced physical activity, sleeping problems, and inadequate social support were associated with PSF [10].

These interacting psychological, behavioural and environmental factors can be understood within a cognitive behavioural model for neurological functional symptoms [11], which act to perpetuate fatigue symptoms and disability. The premise of cognitive behavioural therapy (CBT) is that changing unhelpful thoughts and behaviours impacts on how people feel physically and emotionally [12]. We hypothesised that a gradual increase in physical activity would challenge any unhelpful beliefs that stop patients with PSF from doing things, by which the vicious cycle of PSF would be interrupted. Patients’ self-efficacy would then be strengthened and a reduction in fatigue and an improvement in physical activity would be achieved. In order to engage the patient in this process, the potential benefits of the approach were explained. Patients were told that PSF was reversible and they were encouraged to overcome the fear of taking physical activity (i.e. the cognitive approach of this intervention), and were encouraged to promote a balance between daily activities, rest and sleep and then gradually increase their level of physical activity (i.e. the behavioural approach of this intervention).

We adapted the intervention from a nurse-delivered behaviourally-oriented intervention for cancer-related fatigue [3] and a self-management cognitive behavioural approach for chronic fatigue syndrome [13]. We drafted a Participant Manual to provide an outline of each session. The provisional programme consisted of three two-hour sessions with two-week intervals between sessions.

We distributed the manual by email to a user group (five stroke survivors, from the Scottish Stroke Research Network) for their opinions on the manual’s content and format. Following the user feedback, we extended the intervention programme to six treatment sessions plus one booster session at one month after session 6, and halved the length of each session to one hour. We modified the Participant Manual to a more user-friendly version and developed a corresponding Therapist Manual.

Consistent with the recommendation for assessing functional recovery after stroke [14], we considered three criteria in selecting outcome measures: a) fitting the framework of the International Classification of Functioning, Disability, and Health [15]; b) good psychometric properties in stroke patients; and c) feasible for postal delivery. We measured the presence of fatigue by a case definition of PSF [16], fatigue severity by Fatigue Assessment Scale [17], depression severity by Patient Health Questionnaire-9 (PHQ-9) [18], independence by Nottingham Extended Activities of Daily Living (NEADL) [19], and stroke-specific quality of life by Stroke Impact Scale (SIS) 3.0 [20].

Inclusion criteria: a) ≥18 years old, b) stroke 3–24 months previously, c) self-reported fatigue, and d) living in the Lothian area, Scotland, UK. Exclusion criteria: a) severe depression (assessed by the screening questionnaire PHQ-9 scored 15 or more), b) significant impairments in cognition or communication (as recorded in medical notes or assessed by responsible stroke physicians or general practitioners who had helped us identify potential participants), c) medically unstable or living in nursing home, or d) currently in other research studies or receiving any treatment for fatigue or depression, or would increase the physical burden of the participants.

Sample size calculation is not normally required for feasibility or pilot studies [21]. Some researchers suggested that a sample of 10 or fewer would be sufficient for studies assessing acceptability of formatting or ease of administration of an instrument [22]. Such a small group of participants had been used in the development and preliminary evaluation of a cognitive behavioural approach to manage fatigue in patients with multiple sclerosis [23], which successfully informed the design of a multi-centre RCT [24]. Thus we pre-specified a sample size of 12 for this feasibility study to allow for two dropouts. This sample size was anticipated to be achievable in the limited time (three to four months) available for the recruitment in this feasibility study.

From July 2014 to October 2014, we identified potential participants from patients who had been discharged from a stroke unit (Royal Infirmary of Edinburgh), patients who had visited an outpatient stroke clinic (Western General Hospital), and those who had been visited by nurses from a community stroke service (Chest Heart & Stroke Scotland, CHSS) in Lothian area, Scotland. We reviewed patients’ medical records and contacted their General Practitioners or responsible physicians for their eligibility. Only patients with adequate cognitive and language capabilities would be referred by their doctors for current study. We approached potentially eligible patients by posting two questionnaires: a single question for fatigue and the PHQ-9 for depression, together with an information sheet and a consent form, to each potential participant. Patients who answered ‘yes’ to the fatigue question and had a PHQ-9 score of 14 or less (i.e. no severe depression), were mailed another information sheet for the intervention and invited to meet with the principal investigator (SW) to sign the consent form at the Clinical Research Facility of Royal Infirmary of Edinburgh. For patients with a PHQ-9 score between 10 and 14 (indicating moderate depression), we wrote to them to inform them that they potentially had some symptoms of depression, and suggested that they either take part in the current study or consult their General Practitioners for further advice on depression. Only patients who chose to receive the current intervention and did not receive any other interventions for fatigue or depression were included.

At the meeting in preparation for requesting participants’ consent for inclusion in the study, the researcher (SW) together with a stroke research nurse further assessed the cognitive abilities of potential participants, and patients were included if they were assessed by both the researcher and the nurse as of adequate cognitive capabilities in verbal communication and completing reading and writing tasks.

In the long term, we envisage this intervention to be delivered by stroke nurses. However, we decided a clinical psychologist (KA) should test the feasibility in the current study. If we had tested feasibility with nurses and it hadn’t worked we would not have known whether it was the intervention that hadn’t worked or the person delivering it. KA was supervised by another clinical psychologist (DG). Both psychologists were experienced in the psychological treatment of stroke patients. The intervention was delivered individually to each participant at the Department of Clinical Psychology of a rehabilitation hospital (Astley Ainslie Hospital, Edinburgh). A checklist of key therapeutic components was provided for the therapist to read prior to the delivery of each session and to complete after the session.

Feasibility measures were primary outcomes for this study. We recorded the number of eligible patients at each stage of recruitment and reasons for ineligibility, the number of dropouts, reasons for withdrawal and whether the sessions were attended as planned. After all retained participants completed their sessions and assessments, they were invited to a focus group and were asked to complete an anonymised feedback questionnaire, regarding the usefulness of the intervention, difficulty in completing homework tasks, adequacy of session arrangement, and modes of delivery. Participants who were unable to attend the focus group were invited to give feedback by telephone. The therapist provided written feedback on the acceptability and feasibility of the intervention.

Clinical measures were secondary outcomes for this study. Baseline assessments were completed by all participants at the consenting meeting. Three post-treatment assessments (immediately after Session 6, one month later, and three months later) were completed by participants who completed all sessions (i.e. the retained participants) by postal questionnaires. The three-month follow-up for the final participant was completed in April 2015.

Statistical analyses were performed in IBM SPSS (version 21). We performed descriptive analyses for feasibility outcomes. For clinical outcomes, we tested the assumption of normality of distribution by Shapiro-Wilk test. We used the independent t-test and the Fisher’s exact test for the comparisons between the retained participants and those who dropped out. For the retained participants, we compared their outcomes (mean values or proportions) between baseline assessment and each of the post-treatment assessments, using the paired t-tests and the McNemar test. Considering the correction for multiple comparisons, the critical level of statistical significance would drop from 0.05 to 0.001 (2-tailed). However, there might be interactions between some measures and the results of post-treatment assessments were not independent of each other. Thus the appropriate adjusted critical significance level might be somewhere between 0.001 and 0.05, but the exact value was unknown. Therefore, for each outcome, we reported the mean difference of scores between baseline and each post-treatment assessment, with relevant 95% confidence intervals and p values.

Table 1 summarises the structure and content of the intervention. The duration of each session was one hour. Session 1 had an extra 30 minutes to allow for the development of a collaborative therapeutic relationship. There were two-week intervals between sessions. Homework tasks (e.g. keeping a diary and increasing daily activities) were negotiated during sessions for participants to complete at home. One month after the final treatment session, the therapist delivered a booster session to each participant by telephone.

From July 2014 to October 2014, we screened 421 patients (who had had a stroke in the past three months to two years) and sent invitation letters to 120 potentially eligible patients. Forty-nine eligible patients completed questionnaires, of whom 31 reported fatigue by the single question and the other 18 did not. Of the 31 fatigued patients, 11 had a PHQ-9 score of 15 or more (indicating severe depression) and so were excluded; the remaining 20 patients were eligible for the intervention, of whom eight declined to receive it (reasons for decline see Fig 1).

Twelve eligible patients (five women and seven men, all ischaemic stroke) consented to receive the intervention, with a mean age of 63 years (range 47 to 85). Eight participants had had their first-ever stroke and the other four recurrent stroke in the past two years. The mean duration from the recent stroke to recruitment was 16 months (range 5 to 23 months). All 12 participants were allocated to receive the intervention.

Four participants (all women) dropped out: one participant withdrew after the first session as, following discussion with the therapist, she explained that she did not desire an intervention for her fatigue but that her chief purpose in participation had been to contribute to the research; a second participant withdrew after the initial session and the third participant after the second session because of unrelated physical ill-health problems; the fourth participant failed to attend any of the sessions despite reminders and gave no reasons.

The remaining eight participants (one woman and seven men) completed all sessions and all post-treatment assessments. The three-month follow-up assessment for the final participant was completed in April 2015. Women were more likely to drop out (p = 0.01) as were those with lower scores of physical strength in SIS (p = 0.03). There were no differences between groups in age, time since stroke or other baseline outcomes (Table 2).

Of the eight retained participants, all attended their face-to-face sessions as planned. For the telephone-delivered booster session, only one participant was available as planned, whilst the other seven participants requested that their sessions be rearranged as they had forgotten about the appointment or were too busy to be available at the designated time.

Feedback was obtained from the eight retained participants (five attended the feedback meeting and three gave feedback by telephone). Seven participants (87.5%) rated the intervention as ‘very useful’ or ‘somewhat useful’ (Table 3). Three participants reported that they had not used the strategies to regulate sleep or challenge thoughts, as these problems were not relevant to them. The rating of the usefulness of regulating sleep varied among participants. Regarding the other strategies, each was rated as ‘very useful’ by more than 70% of the participants who had used it. One participant rated the intervention as ‘not useful at all’. This included strategies for regulating sleep and the booster session, though interestingly the fatigue score of this participant was improved from baseline to after treatment.

More than 60% of participants rated homework tasks as ‘not difficult at all’ or ‘a little difficult’ (Table 4). Some participants reported difficulties in keeping diaries, making plans, increasing activities or regulating sleep, though they commented that these tasks were useful in helping them overcome fatigue. Five participants rated the content of sessions as being of ‘the right amount’, whilst two participants reported Session 1 was ‘too much’ and the other thought Session 6 was ‘too little’.

All participants had concerns about our future plans for stroke nurses to deliver this intervention. Nevertheless, they conceded that the nurse-delivery approach could be acceptable if stroke nurses were to be trained and supervised by experienced CBT psychotherapists. All participants suggested that treatment sessions and booster sessions should be delivered in person. They felt that they would be taken more seriously when meeting the therapist in person and that this would result in their greater commitment to the therapeutic process. Based on their experience of telephone-delivered booster sessions, they had had difficulty in being available at the agreed times and were easily distracted by their home environment during the telephone session. Participants also intimated that they would not wish to undertake such a therapeutic process online as it was unlikely that they would regard it as personal or as relevant as a person-to-person approach.

The therapist (KA), a clinical psychologist experienced in delivering CBT, commented that the intervention ‘had been carefully designed on the basis of standard CBT literature’ and that the manuals ‘had been clearly laid out’. However, she suggested a more even distribution of materials between sessions, as there was too much new information in Session 1 and insufficient material in Session 6 for some participants.

Clinical outcomes of the eight retained participants are summarised in Table 5. Compared to the baseline scores, there was improvement in fatigue severity, self-reported general recovery, memory and thinking, emotion, mobility, and social activity at three months after the end of treatment (all p<0.05).