Promoting activity, Independence and stability in early dementia (PrAISED): a, multisite, randomised controlled, feasibility trial

We undertook a pragmatic, parallel group, single-blinded, randomised controlled feasibility trial at two study sites, recruiting from Memory Assessment Services (dementia diagnostic services) and the Join Dementia Research register, a United Kingdom National Institute for Health Research (NIHR) web-based initiative to increase participation in dementia research [16]. The full rationale and protocol have been published [17]. There were no significant changes to the research plan during the feasibility study. The CONSORT 2010 checklist of information to include when reporting a pilot or feasibility trial is available in Additional file 1.

Patient participants were identified by clinicians working in Memory Assessment Services and referred to the research team if interested, or by researchers through the Join Dementia Research Register. We included patient participants who were age 65 years or over and had diagnosed mild dementia or mild cognitive impairment (Montreal Cognitive Assessment (MoCA) score of 15–25 [18]; Mini Mental State Examination (MMSE) of 18–26 [19] or Addenbrookes Cognitive Examination III (ACEIII) 60–94 [20], depending on what assessments clinics used). Patient participants had to be able to walk without human assistance and communicate in English, have no co-morbidities that prevented participation in cognitive assessment and capacity to give informed consent (formally assessed by a researcher). A family member or informal carer of the patient participant who could communicate in English was recruited, if willing. Written consent was taken from both patient and carer participants. We chose to study people with mild impairments as our intervention was designed to prevent or slow decline amongst those still well enough to engage and learn new lifestyle activities.

Researchers and NIHR Clinical Research Network clinical support officers (CSO) recruiting patient participants and collecting data had two days training in study procedures. Two researchers or CSOs visited the patient and carer participants in their own home to complete a screening questionnaire, discuss the study and take written informed consent. The commitment needed to complete the study (if randomised to one of the active arms) was emphasised to avoid later withdrawals. Baseline data were collected from the patient and carer participant. Some patient participant data were collected from the carer participant to reduce burden. Visits with two researchers or CSOs allowed the patient and carer participants to be interviewed concurrently in different rooms, reducing the overall time of the interview and allowing the participants to respond freely.

To describe the patient participants, we collected demographic data, past medical history, medications, and falls risk factors from the carer participant and measured cognition by completing a standardised mini mental state examination (sMMSE) [19] with the patient participant.

We tested two versions of the PrAISED intervention, which differed in the amount of professional supervision offered. Both versions aimed for the patient participant to complete three hours of PrAISED exercises each week for the 12 month intervention period. Both versions included assessment, creation of an individualised tailored exercise and activity plan, supervised exercises and activity, and regular reassessments and progression. Patient participants in the high intensity supervision group received up to 50 visits from a therapist or rehabilitation support worker (RSW) over a period of one year. Patient participants in the moderate intensity supervision group received 9 visits from a therapist (or two therapists for initial assessment visits) and three phone calls over a three month period, and encouragement to continue the programme after professional supervision ceased.

The PrAISED physical exercises included balance challenging, strength building, dual-task training and gait re-education. Some exercises could be gained through functional strategies which aimed to maintain or improve independence. Patient participants in the high intensity supervision group received visits that were tapered over the year, starting at twice weekly for three months, then weekly for three months, bi-monthly for three months and monthly for the last three months. Patient participants in the moderate intensity supervision group received tapered visits in the first three months of the twelve month intervention period. The supervision regimens were based on precedents from the literature [6, 21]. Both interventions included motivational strategies to encourage adherence to, and persistence with, the planned exercises and activities. Participants were encouraged to continue their exercise and activities beyond the supervised period and were provided information on appropriate community exercise groups in their area. Therapists and RSWs delivering the intervention received three days training prior to the start of the study and 2 days training during the intervention delivery period. The control group had a single falls prevention assessment and advice with one to two follow-up visits by a therapist if indicated. Standard care controls are common practice in activity and exercise interventions for people with dementia [6, 9, 21].

A feasibility study aims to answer the question “can this study be done?” [22]. They play an important role in the subsequent design of a fully powered trial [13]. Specific trial feasibility questions were [17]:

Outcomes were measured at baseline and 12 month follow-up visits.

Activities of daily living were measured using the Disability Assessment for Dementia scale (DAD) [23], completed by the carer participant and the Nottingham Extended Activities of Daily Living Scale (NEADL) [24], completed by patient participant.

Cognition was measured using Cambridge Neuropsychological Test Automated Battery (CANTAB) tests [25, 26]: Paired Associated Learning (PAL), Attention Switching Task (AST), Spatial Working Memory (SWM). We also measured a verbal fluency test (from MoCA) [18]. We used these tests as previous studies had shown that it was unlikely that exercise would improve global cognition scores, but may influence specific elements such as executive function. The clinical dementia rating [27] was completed by the researcher.

Sitting and standing blood pressure and pulse rate were measured using an automated sphygmomanometer (Omron, Milton Keynes, UK). The Berg Balance Scale [28], leg and hand strength (Lafayette dynamometer, Loughborough, UK), timed up and go (TUG) test and dual task TUG (whilst counting backwards in 3’s) were measured directly with the patient participant. The SHARE Frailty instrument [29], and International Physical Activity Questionnaire (IPAQ) [30] were completed by carer participant.

Quality of life was measured from the patient participant using EQ-5D-3L [31]; and Dementia Quality of Life Scale (Demqol) [32]. Demqol proxy was completed by the carer participant. Anxiety and depression (Hospital Anxiety and Depression Scale, HADS) [33] and Falls Efficacy Scale – International (FES-I) [34] were completed by the patient participant. Carer participants completed the carer strain index (CSI) [35].

Throughout the year of follow-up, patient participants were asked to complete and return monthly falls and exercise calendars, by post, in a stamped addressed envelope provided by the researchers or CSOs, prompted by monthly telephone reminders. At baseline, six and 12 months patient participants were asked to wear and then return, by post, a tri-axial accelerometer for one week to objectively measure activity. A range of accelerometers were provided to participants. See Table 1 for data collection time points.

Adverse events were monitored and reported to the principal investigator at each site by therapists, RSWs and researchers during telephone contacts to prompt diary return. Adverse events were defined as hospitalisation, or incidents which were life threatening, causing persistent or significant disability or incapacity, or an incident, injury or symptom related to therapy sessions or exercise undertaken independently.

Participants were randomly allocated to one of the three arms, stratified by site, presence of a co-resident carer and history of previous falls (to control for confounding variables) using a dynamic adaptive allocation algorithm [36] accessed by a secure web portal held at the NWORTH clinical trials unit (CTU), Bangor University. The randomisation system was maintained by a statistician independent of the analysis and research teams to ensure blinding of analysis. The researchers and CSOs entered the patient participant’s details into the web portal. Notification of allocation was sent to ‘unblinded’ members of the research team who notified the therapists of allocation. The therapists then informed the patient participant. Patient participants were randomised 1:1:1 to high intensity supervision, moderate intensity supervision or standard falls prevention assessment and advice (the control arm).

We could not blind the patient and carer participants to the intervention. We took measures to ensure the researchers collecting primary outcome data were blind to allocation. For example, the ‘blinded’ researchers sat in a different office, spreadsheets identifying allocation were password protected to avoid accidentally seeing allocations and we took care in meetings not to reveal allocation. Randomisation emails giving details on allocation were not sent to ‘blinded’ researchers.

Data were entered into a MACRO database. Data which could ‘unblind’ researchers was kept separately from ‘blinded’ data on MACRO with access given as appropriate. Database management and data cleaning were managed by the CTU. Source data verification was undertaken for 5% of data from both carers and patient participants at baseline and follow-up at each site. Source data to be verified were randomly selected by the CTU and researchers completed the verification checks.

A sample size of 60 was chosen to be sufficient to answer feasibility questions. A sample of this size would give adequate precision using a confidence interval approach [37] which considers the likelihood of a future definitive study finding a relevant effect size.

A statistical analysis plan was drawn up by the trial statistician prior to completion of recruitment. Statistical analyses provided descriptive statistics on recruitment and retention rates, participation, missing data, adverse events, follow-up, and falls ascertainment rates, and the distributions of key variables required for main trial sample size calculation. Mean health status variables were calculated by group, at baseline and follow-up, with mean differences. Standardised effect size estimates from analyses of covariance (ANCOVA), equivalent to Cohen’s d [38], were calculated. A Cohen’s d of 0.2 to 0.5 is a small effect size; 0.5 to 0.8 is a medium effect size and greater than 0.8 is a large effect size [38].

We set discontinuation rules of recruitment (> 2.5 participants per week), retention (> 77%), intervention set up at sites, data collection (< 20% missing primary outcome).

Between September 2016 and March 2017 371 people with mild dementia or MCI were referred to the study and 201 were pre-screened. Of these, 52 were ineligible and 86 either not interested or lost. We screened the remaining 63, one was ineligible and two withdrew. We recruited 60 patient participants and 54 carer participants, on schedule (consort diagram, Fig. 1). Seventeen were recruited via Join Dementia Research. Patient participants had a mean age of 76 years (range 65–91), 34/60 (57%) were male and 45/60 (75%) married. Most were white ethnicity (58/60, 97%); lived with another person (48/60, 80%); had received secondary education 32/60 (53%) and many had received further education 26/60 (43%). Mean sMMSE was 25.6/30 (range 18–30), 55/59 (93%) had a diagnosis of dementia. Carer participants were younger with a mean age of 68 years (range 33–87), 45/60 (75%) were spouses and 39/60, (65%) were female (Table 2). Twenty one patient participants were randomised to standard care, nineteen to the moderate intensity supervision intervention and twenty to the high intensity supervision intervention. There was a gender imbalance between groups with 62% of the standard care group being female compared to 37 and 30% in the moderate and high intensity supervision groups; groups were otherwise well-matched at baseline.

The intervention was successfully delivered in participant’s homes, across the two sites with the control group receiving a mean of 2 sessions, the moderate intensity supervision group receiving 12/12 (100%) sessions (including telephone calls) and the high intensity supervision group receiving 33/50 (66%) sessions, although some sessions had more than one therapist attending.

Three hundred and ninety one out of five hundred and twenty eight (74%) of the exercise and falls calendars were returned. From these calendars, mean PrAISED related physical activity minutes per week for both intervention groups was 72 min (Standard deviation (SD) = 63) when calculated for the time between baseline and follow up (approximately 11 months). In the moderate intensity supervision group, participants completed a mean of 77 min (SD = 71; range 15 to 228 min) per week and 71 min (SD = 56; range 11 to 246 min) in the high intensity supervision group. Those in the control group had not been given PrAISED exercises to complete.

Seven participants withdrew from therapy before the 12 month outcome visit (three from the high intensity supervision group, two from the moderate intensity supervision group and two from the control group). Two participants gave depression as the reason for withdrawing, one was feeling overwhelmed, one felt he wasn’t eligible because he was physically well, one thought it was the wrong time and that the visits stopped her going out and doing what she wanted to do, one withdrew due to physical and mental deterioration and one withdrew due to bereavement and feeling the exercises were not appropriate to her age due to risk of falls.

Between September 2017 and March 2018, 49/60 (82%) participants completed the follow-up visit with nine withdrawals and two participants lost to follow-up. 16/21 (76%) participants from the control group completed the follow up visit, 15/19 (79%) from the moderate intensity supervision group and 18/20 (90%) from the high intensity supervision group. Assessment interviews proved lengthy with a median duration of 115 min. 32/52 (62%) patient participants reported that the time taken was acceptable 16/52 (31%) considering that it was too long. Missing data at baseline and follow-up varied across the different measures and questionnaires (Table 3). The primary outcome measure, the DAD, was completed for 52/60 (87%) participants at baseline and 46/49 (94%) participants at follow-up. DAD is an informant rated scale. Six participants were recruited without a carer, so did not provide DAD data. A ceiling effect was seen with 11/52 participants at baseline and 4/46 at follow-up gaining full scores on the DAD. CANTAB assessments and accelerometer data showed considerable missing data due to technical problems. The carer participants reported finding the IPAQ difficult to complete.

Despite efforts to blind researchers collecting follow-up data the researchers were able to correctly identify allocation (intervention or control) for 42/49 (86%) outcome visits.

There was a trend towards better performance on the primary outcome measure, DAD, and less deterioration from baseline, with higher intensity supervision. Mean DAD scores were 76/100 at baseline versus 58/100 at follow-up for the control group, 83 versus 66 for the moderate intensity supervision group and 76 versus 68 for the high intensity group. Cohen’s d was 0.03 (− 0.7 to 0.8) for moderate (no effect) and 0.43 (− 0.3 to 1.2) for high intensity supervision (a small effect size), compared with control (Tables 3 and 4) [38].

The Berg Balance Scale showed a large and statistically significant positive effect size for the moderate intensity d = 0.9, 95% CI [0.1, 1.8] and high intensity d = 1.0, 95% CI [0.1, 1.8] groups compared with control. Fear of falling improved for the moderate intensity supervision group d = − 0.8, 95% CI [− 1.7, − 0.03] (Table 4).

There were 19 recorded adverse events. Five were related to the intervention but not serious, 12 were serious but not related, 2 were neither serious nor related to the intervention. They were all recorded in the active intervention groups, but were subject to ascertainment bias as these groups had much more contact with therapists.

Only 12 falls were reported by participants over the 12 months, fewer than anticipated [2, 17]. The incident rate ratios (IRR) were 0.68, (95% CI 0.16, 2.8) and 1.20, (95% CI 0.32, 4.5) for moderate and high intensity supervision, neither being statistically significant.

For the main trial primary outcome is disability at 12 months. We calculated that a sample size of 184 participants per group, including 23% attrition (based on previous studies [6, 39], has 80% power to detect changes in the disability outcome, DAD, with effect size 0.5 (11 points on a baseline of 70, standard deviation 22, data from [39, 40]).

The results from this feasibility study broadly supported our original sample size calculation for the high intensity supervision group and we have not revised the original sample size calculation based on these results.

Our study was conducted at two sites with different ways of organising staff to deliver the intervention. The ‘Join Dementia Research’ register proved useful for recruitment, but may have skewed the population to more educated and younger participants.

The adherence rate for the supervised activity and exercise sessions of 66% for the high intensity supervision version and 100% for the moderate intensity supervision version is comparable to other studies of this kind which show a mean adherence rate of 71% [unpublished observation from Di Lorito C, Bosco A, Booth V, Goldberg S, Harwood RH, van der Wardt V]. However, mean activity per week was only 72 min of PrAISED exercises a week, falling considerably short of the required ‘dose’ of 180 min and WHO recommendations [10]. Therapist training has been changed for the main trial with a greater focus on motivation and adherence. Our attrition rate of 18% was acceptable compared to similar studies (23% for FiNALEX [6]; 9% for Wesson [21]).

The standardised effect size estimates for the primary and some intermediate outcome measures (whilst difficult to interpret due to small sample size) suggest the intervention, delivered with a high level of supervision, could feasibly show a positive effect at 12 month outcomes in a fully powered comparison (in other words, the intervention is not shown to be futile). The PrAISED intervention required patient participants to complete 3 h of PrAISED exercises a week for 12 months, but differed according to the amount of professional contact they received. Of the two intervention groups, the greater effect size was seen with the high intensity supervision group. Measured adherence was not different between these groups, but these data are difficult to collect, not least because the difference between an ‘exercise’ and an ‘activity’ can be difficult to determine, and therapists reported that some participants chose to do ‘activities’ over ‘exercises’ as time went by. Care may be enhanced in the therapy arms through the interactions between participant and therapist (such as if the therapist referred the patient participant to services for a problem not related to the intervention [41]). We cannot be sure whether it was the ‘social’ contact or the therapy programme that changed outcomes, however, in complex intervention trials all elements are typically considered together. It was not feasibly to have a sham control and it is common practice to have a ‘usual care’ control arm [6, 9, 21] for activity and exercise interventions for this patient group. Our research suggests participants are less likely to do the exercise without human contact (through supervision or support from carers) [42].

Unblinded trials can result in bias [43], however, we found blinding was impossible to maintain in practice. Issues of bias in outcome data collection need to be addressed through specific anti research bias training, which we have put in place for the main trial.

Many people with dementia live alone in the community and some do not have any family carers. We wanted to be inclusive in our recruitment, but found we could not complete primary outcome data with this policy and will restrict our inclusion criteria for the main trial to those with a carer willing to take part.

The DAD as a primary outcome measure has been used before in similar trials; is recommended [44] and was well completed when a carer was available. However, the DAD did show a ceiling effect at baseline and outcome; potentially reducing the chances of demonstrating effectiveness. Better ADL outcome measures for people with early dementia are needed. Measurement of physical activity using accelerometers was inaccurate due to technical difficulties, which were resolved during the feasibility trial and which has informed researcher training. We selected the best performing accelerometer for the main trial (Misfit Shine 2). Carer participants found the IPAQ difficult to complete and we have changed it to the Longitudinal Aging Study Amsterdam physical activity questionnaire for the main trial [45]. An independent CTU managed the data and undertook statistical analysis blind to intervention group. Some scales had high levels of missing data resulting in some being dropped or changed and instructions on completion being clarified for the planned main trial.

There were no serious adverse events related to the intervention, suggesting the intervention does not cause harm. In a trial where one group has considerably more contact than the other, there is the problem of differential reporting of adverse events, and comparing rates of adverse events in different groups is impossible to interpret. We have embedded safety parameter variables (mortality, hospitalisation, falls, injury) in the outcome dataset, which is collected independently of contact with therapists delivering the clinical intervention. This approach was agreed by our Data Monitoring Committee.

Our sample size assumptions are shown to be reasonable.