A Pilot Randomized Controlled Trial Testing the Feasibility and Acceptability of Helping Ease Anxiety and Depression after Stroke (HEADS: UP): An Online Mindfulness-Based Intervention for Stroke Survivors

Participants were community-dwelling stroke survivors recruited using a UK-focused strategy comprising social media (e.g. Twitter (now X), Facebook) and non-governmental organizations (e.g. Stroke Association; Chest, Heart & Stroke Scotland). Social media adverts included short recruitment videos (1–1.5 min) featuring past participants and members of the project advisory group (https://www.youtube.com/@headsup6765). Potential participants who responded to social media adverts were sent (by mail or email) a project information pack (a project information leaflet (PIL), consent form and data privacy notice). When recruiting through third sector organizations gatekeepers disseminated the project information pack to their members. MBSR is often delivered to groups (15–20 individuals), but to promote adherence the plan was to deliver HEADS: UP to dyads (a participant plus a supportive partner) where possible. The initial plan was to enroll participants (n = 90) into six groups, but COVID-19 restrictions and changes to the research design (moving to online delivery) limited recruitment time (Lawrence et al., 2024). A pragmatic decision was made to reduce the number of participants (n = 60) across four groups in two cohorts (i.e. 15 stroke survivors per group). Cohort 1 was recruited by September 2021. Cohort 2 was recruited by mid-January 2022. Attrition was estimated at 44% by six months (Lawrence et al., 2024) which would leave 10 participants per group by 6-month follow-up. This was in keeping with UK National Institute for Health Research recommendations for pilot and feasibility studies (Julious, 2005) and depending on the standard deviation of the main outcome measures, would fulfill a recommended minimum number of n = 40 (Sim & Lewis, 2012).

The 10-week period of recruitment (October–December 2021) resulted in 120 expressions of interest. Potential participants were eligible if they were: living in the UK, stroke survivors, aged ≥ 18 years, at least 3 months post-stroke, able to speak and understand conversational English, and to follow a two-stage command (Maruya et al., 2018) and scored ≥ 3 on the Patient Health Questionnaire-4 (PHQ-4). The PHQ-4 is a brief screening tool addressing anxiety and depression, selected to reduce participant burden (Lawrence et al., 2023; Löwe et al., 2010); an elevated score indicates that further investigation is warranted (Kroenke et al., 2009). Eighty-three people were screened for eligibility; 64 enrolled (50 as lone participants; 14 (22%) with a supportive partner) and were successfully randomized evenly across the two arms; see Fig. 1 for enrollment flowchart. Sixty-two participants completed baseline questionnaires (T0) and were randomized to HEADS: UP (n = 30) or TAU (n = 32). At baseline mean age, gender split, and time post-stroke of the HEADS: UP and TAU groups were 56.00 and 56.80; 30% and 56% male; and 15 and 24 months, respectively (see Table 1 for full demographic data). Of the n = 30 randomized to HEADS: UP, n = 25 (83.30%) received the intended treatment, compared with n = 27 of n = 32 (84.40%) who received TAU, and were analyzed for the study objectives (see Fig. 1). Unrelated to the trial process one participant died and two participants withdrew because of mental health reasons (one in each arm; one returned at T2, and one withdrew at T3), one participant withdrew because their partner could no longer support them, and another withdrew because of work commitments. The remainder dropped out and were lost to follow-up. Related to the trial process were two participants allocated to TAU (6.30%) who withdrew because they were not randomized to the HEADS: UP arm. Participants analyzed for HEADS: UP versus TAU at T0 numbered n = 30 versus n = 32; at T1 n = 25 (83.30%) versus n = 25 (78.10%); at T2 n = 24 (80%) versus n = 26 (81.30%); and for T3 n = 20 (66.70%) versus n = 25 (78.10%), respectively (Fig. 1). The RCT stopped in September 2022, following collection of 6-month follow-up data.

Afbeelding vergroten

After obtaining informed consent, confirmed by email, a researcher conducted screening by phone or on Zoom^© (an online video platform) using a bespoke screening and enrolment questionnaire, which incorporated the Modified Telephone Interview for Cognitive Status (TICSm; Brandt et al., 1988) and the PHQ-4. Stroke survivors who were eligible to participate were asked whether they wanted to nominate another person (e.g. a family member) to take part with them (Lawrence et al., 2023; Morris et al., 2023). This was not an inclusion criterion, and the role varied, according to participant preference (e.g. some family members simply assisted the participant to get online, others attended sessions with the participant). Participants were also asked for ''logistics'' data: availability on pre-set dates, internet access issues, help (if any) to complete questionnaires; accessibility requirements e.g. paper study materials. Finally, the researchers reminded participants of the study requirements, including completing repeated questionnaires and taking part in focus groups, irrespective of group allocation. Screening took approximately 45 min; breaks were offered.

Following collection of baseline data, individual participants were randomized (1:1 ratio) to HEADS: UP or TAU. Randomization was conducted by an independent statistician who used a randomization generator (Microsoft Excel) to assign participants, stratified by participant status i.e. lone participant or participant with a partner. This aimed to achieve between-group equivalence for lone participants vs. those with a partner; important because of the anticipated non-specific effects of sharing the experience with another. To minimize bias, participants self-completed the patient reported outcomes measures (PROMs) unless they requested assistance; the researchers recorded details of any assistance given. One researcher (NC) entered all PROMs data, which were independently checked by a second researcher (BD or ML). The statisticians who conducted the analysis were blind to allocation (ND; MJ). HEADS: UP participants received a course manual and had access to downloadable audio practices. Participants (HEADS: UP and TAU) did not receive any financial or other incentive to participate.

HEADS: UP was delivered online, to groups, over nine weeks using Zoom. The first week was an introductory session which was followed (Weeks 2–9) by the eight core MBSR sessions (Santorelli et al., 2017). The core sessions cover a range of mindfulness techniques including body-focused meditation, gentle movement (yoga), and breathing meditations. Students learn about, practice, and then discuss each technique within the group. Weekly sessions were 2.5 hr and include two 15-min breaks. In week seven the session is longer (6 hr) to accommodate a silent retreat. Full details of course content, development and optimization have been reported previously (Lawrence et al., 2023, 2024). The course was delivered by experienced MBSR trainers who fulfilled the British Association of Mindfulness-Based Approaches (2011) good practice requirements and had completed bespoke HEADS: UP Train the Trainer training. Trainers' audio-recorded the individual weekly practices and uploaded the recordings to an online repository (Padlet). The TAU treatment was not structurally equivalent to HEADS: UP and was not intended to augment any therapeutic intervention typically received by individual participants during the course of the study. TAU participants remained under the care of their GP (General Practitioner) with no specific recommendations on prescription of medication, psychological interventions, or referral to a mental health professional.

Each person expressing interest in the study was assigned an identity number (ID), so they could be tracked throughout the various stages of the study. Demographic data (gender, age, time post-stroke, ethnicity, marital status, living arrangements, educational attainment, employment status, dyadic versus lone participation, recruitment origin) was used to assess whether randomization had been successful. Recruitment origin (e.g. X; Stroke Association) data allowed assessment of the various strands of our recruitment strategy. Most studies have defined MBSR completion as attending ≥ 4 of the eight core sessions; most have not achieved this with reported dropout rates of 15–30% (e.g. Cash et al., 2016; Marjani, 2017; Wang et al., 2019). In this study, intervention feasibility was defined as ≥ 70% of participants attending ≥ 4 core sessions. Of n = 26 HEADS: UP participants, n = 24 (92.30%) attended at least four core sessions. Participants with a supportive partner attended an average (median) of seven sessions (median; IQR 6–8); lone participants attended eight (IQR 7–8). Participants practiced on 6 days of the week (IQR 5–7) for 22.70 (SD 14.60) mins per practice day. Across 16 core sessions, the Cohort 2 trainer recorded nine incidences of deviation from session plans (minor overruns n = 5; breakout rooms not used n = 3; one practice omitted n = 1). HEADS: UP participants were invited to keep personal practice logs (PPLs, simple check-box forms submitted weekly; Online Resource 2) to record their home practice. Reminder SMS (Short Message Service) prompts were sent, as required. Weekly session plans and fidelity logs, completed by the trainers, allowed assessment of fidelity.

PROMs data were collected at four time points: baseline (T0); post-intervention (T1); 3-month follow-up (T2); and 6-month follow-up (T3). Changes to study design made as a consequence of COVID-19 restrictions meant that the planned 12-month follow-up was no longer possible (Lawrence et al., 2024). Reminders (phone calls; SMS texts) were sent after three days and a member of the research team assisted participants with completing the measures via phone. Sixty-two (96.90%) participants completed all baseline PROMs (completion time 40 min approx.). Eleven (17.20%) participants requested paper copies of all research materials including PROMs. Eleven participants (17.20%) required assistance to complete the PROMs, whether online or on paper: reading the questions n = 5 (8.10%); explaining the questions n = 1 (1.60%); clicking on/writing the answer n = 5 (8.10%). At 3-month follow-up two participants returned only partially complete PROMs. Twenty-four (92.30%) HEADS: UP participants returned PPLs; 23 (88.50%) returned ≤ 4 PPLs. Some paper copies returned included omissions (e.g. ''practice duration'' was incomplete: n = 15 logs, at least 1 missing item), illegible handwriting, and contradictory information (n = 2 logs). Long-term engagement with data collection processes was maintained through intermittent contact with all participants (e.g. sending seasonal greetings cards).

A statistical analysis plan was developed and approved before commencing analysis of the PROMs data and the unit of analysis was the participant—i.e., the stroke survivor. Descriptive statistics and R statistical software (RStudio Team 2020.03.0 + 454, "Chocolate Cosmos", running R 4.3.2) were used to summarize participant characteristics, their outcome measures, and any adverse events. There were no formal tests of statistical significance as this pilot trial was not powered to detect effectiveness. Analysis was conducted using intention to treat. CONSORT guideline extension for pilot trials informed the data analysis (Eldridge et al., 2016). There was no sub-group analysis. EQ-5D-5L was scored using the EuroQol crosswalk (van Hout et al., 2012). Mean change in utility scores and EQ-VAS scores were reported with estimates of precision (standard deviation (SD), min and max scores). Self-reported healthcare resource use during the study period was summarized and presented by trial arm. Components of intervention delivery were itemized and unit costs from the Personal Social Services Research Unit were attached to each item of resource (Jones & Burns, 2021).

The economic evaluation feasibility study findings are summarized here. For full details of health economic assessment methods and results see Fenocchi et al., (in peer review). Most participants (> 95% at each timepoint) provided data about health outcomes measured using EQ-5D-5L and about health care resource use. Completion of RUQ and EQ-5D-5L are presented in Table 3. From the EQ-5D-5L, no statistically significant difference in health-related QoL was observed between groups cross-sectionally at any timepoint (Online Resource 9). Both groups indicated improvement over time (utility scores for the HEADS: UP group ranged from −0.04 to 1.00 at T0 (n = 25), and 0.32 to 1.00 at T3 (n = 20), compared with TAU from 0.08 to 0.85 at T0 (n = 25), and 0.16 to 1.00 at T3 (n = 25)) although small numbers preclude conclusions. The highest reported categories of health care resource use were for hospital outpatient appointments and GP appointments (telephone, online and in person combined). A few participants reported appointments with a psychologist about their anxiety or depression symptoms, or both. Healthcare resource use at each time point is presented in Online Resources 10, 11, 12. Estimates of the costs of delivering HEADS: UP indicated a time commitment of 70.5 hr from a single practitioner for training specifically in HEADS: UP course delivery (10.5 hr) and online course delivery (60 hr, ranging 22.5 hr to 26 hr), plus central resources (time of trial team at university to train-the-trainer, and provision of materials) necessary for ensuring fidelity to the HEADS: UP delivery model.

The exclusion of people with aphasia is a limitation frequently noted in stroke intervention research (Shiggins et al., 2024) and may, at least in part, be related to the additional time and resources required to ensure meaningful involvement of people with moderate to severe communication impairment. An additional concern for people with aphasia in the context of a standardized MBSR course is that course content increases in difficulty, both in terms of the concepts being explored and the language used to describe them as the course progresses (Pieri et al., 2022a). Therefore, we determined that further adaptation of HEADS: UP would be required to ensure its accessibility for people with moderate-severe aphasia, and a PhD study commenced concurrently with this RCT, to develop HEADS: UP Aphasia (Pieri et al., 2022a, 2022b).

The study aimed to recruit a representative sample of the stroke population of the UK across a range of characteristics, including geographic location, socioeconomic profile, and ethnic diversity. Whilst the sample included some island-dwelling participants, most lived in or close to major conurbations. Whilst sample characteristics did reflect those of populations typically associated with the uptake of mindfulness, yoga, and other complementary therapies in terms of age and educational attainment, they did not reflect the demographic profile of the wider UK stroke population. It is acknowledged that in Western societies MBIs are typically accessed by well-educated, White populations, so future research will need to develop recruitment strategies to enhance inclusivity across social groups and cultures (Oman, 2023). Such strategies include expanding networks to include community-based faith groups and partnerships and working collaboratively with such groups to consider how best to ''brand'' HEADS: UP to enhance perceptions of relevance across diverse populations, ahead of and during a future definitive trial (Zimmerman et al., 2022).

The observed attrition rate in the HEADS: UP group at six-month follow-up was lower than the anticipated 44% benchmark, aligning with UK NIHR recommendations for pilot and feasibility studies. However, attrition was higher in the HEADS: UP group (33.3%) than in the TAU group (21.9%). Analysis of participant characteristics revealed that fewer males (9) than females (21) were randomized to the HEADS: UP group, whereas the TAU group had a higher proportion of males (18) than females (14). This imbalance may have influenced the attrition patterns observed, although the extent to which this was directly related to the intervention remains unclear. Additionally, participants further post-stroke were more likely to discontinue. A large, definitive RCT would allow for a more robust randomization to better account for time-post stroke differences and ensure balanced group characteristics. Conversely, the relatively low attrition rate in the TAU group, often harder to retain due to the absence of an active intervention, may reflect the effectiveness of recruitment and retention strategies, such as regular participant contact and emphasizing the altruistic value of participation.

Whilst the independent statistician conducting the analysis was blinded to allocation, the research team and participants were not—a potential source of bias. In a future definitive RCT, independent researchers will manage data collection, analysis, and reporting processes. Randomization was effective in ensuring equal distribution of participants attending with a supportive partner between the two trial arms. However, between group differences were evident in relation to gender, time post-stroke, employment status and educational attainment, reflecting the pilot trial’s smaller sample size and the associated increased likelihood of random imbalances in baseline characteristics (Netz et al., 2019). In a future definitive RCT a larger sample size will help eliminate these imbalances, and a more robust randomization strategy will be used. Specifically, variable-sized permuted blocks could be used to maintain allocation concealment and flexibility, while stratification could incorporate key variables such as gender, time-post stroke, and baseline anxiety and depression severity to maintain group balance. These measures will support more equitable group characteristics and strengthen validity of trial outcome data.

In conclusion, HEADS: UP Online intervention and research processes are feasible and acceptable. Results indicate a positive impact on mood and quality of life outcomes. Future large-scale research is warranted to assess effectiveness and cost-effectiveness and to identify any implementation and scalability issues.