Efficacy trial

The design of the efficacy trial and adherence sub-study is illustrated in Table 1. The efficacy trial will be a double-blind parallel group multicentric RCT. It will be completed in three sites: Centre Leenaards de la mémoire – Centre hospitalier universitaire Vaudois (CHUV) in Switzerland; Institut universitaire de gériatrie de Montréal (IUGM) of the Centre intégré universitaire de santé et de services sociaux Centre-Sud-de-l’Île-de-Montréal (CIUSSS-CSMTL) in Canada; and Brusano and Centre Public d’Action Sociale (CPAS) of Woluwe-Saint-Lambert in Belgium. Participants will be randomised to one of two home-based computerised intervention conditions, the StayFitLonger training programme (experimental) or the comparator, an active control training programme. Outcome measures (Table 2) will be collected at two timepoints: pre-training (T0; within 6 weeks prior to the start of the intervention) and post-training (T1, within 4 weeks following the end of the 26-week training). Of note, a second exploratory post-training assessment (T2), not part of the RCT, will be performed within 4 weeks following the end of the adherence sub-study. At each timepoint, there will be two assessment visits. Within a month following the T0 assessment, introductory courses in groups of a maximum of six people will take place to introduce the features of the programme and describe the different physical and cognitive exercises. This will mark the beginning of the training that will take place at home for 26 weeks (see Table ​Table11 for details). Participants will be supervised through home visits and monthly phone calls to monitor their use in relation to recommendation and address problems with the use of the programme.

Schedule of enrolment, assessments and interventions

List of outcomes measured for the RCT

^aT2 assessment is listed here but it is not technically part of the RCT

One hundred and twenty-eight French-speaking healthy participants will be recruited, 64 in Switzerland, 32 in Canada and 32 in Belgium. Participants will be community-dwelling older adults. They will be recruited through diverse sources including ads, newsletters, social media, and flyer distribution during various events. Recruitment will be carried out with the help of two community associations, Pro Senectute in Switzerland and Brusano in Belgium, and from the bank of participants of the IUGM research centre in Canada.

Included participants will be fluent French-speaking adults aged 60 years and over, retired and living at home. They will have access to a wireless Internet connection at home and will be open to the use of new technologies including electronic tablets. They will be independent for daily activities based on a normal score on the 4-Instrumental Activities of Daily Living (4-IADL) scale [45]. They will be interested in exercising to stay fit and able to walk at home without a walking aid (e.g., wheelchair, cane, walker, etc.). They will be available to commit themselves for the time period during which the study will take place, with no vision deficits that would prevent them to read information on a tablet and with no current neurological or psychiatric diagnosis (e.g., Parkinson’s disease).

Participants with a MoCA score < 26 [46] or a score ≥ 3 on the Fried’s frailty index [2] will be excluded from the study.

A two-stage screening process will be used to select participants (timepoint T-1; Table 1). Initial contact will be made by phone by research team members with expertise in recruitment. At this stage, only participants who report no major physical, medical, or sensory limitations will be invited to come to the laboratory for further investigation. During the on-site visit, participants will be presented with the information and consent form. In Switzerland, participants will be offered to receive the information and consent form prior to their visit. Once they sign the consent form, inclusion and exclusion entry criteria will be measured for the second-stage screening. The Fried’s phenotype scale [2] will be used to exclude frail individuals and to determine frailty level among other participants who will be identified as either robust (score of 0) or pre-frail (score of 1 or 2). In addition, the participant’s technology (e.g., use of tablet, email, social network) and gaming profile will be established with an ad-hoc questionnaire. Eligible participants will receive an ID number. All data will be anonymized and maintained in REDCap, a secure online database [47]. Access to data will be restricted by type of data (e.g., assessors will only have access to assessment data). Furthermore, data collected directly by the programme will be transmitted and maintained in a secured server located at the Haute École Spécialisée de Suisse Occidentale (HES-SO).

A randomisation list will be generated in Switzerland, independently from the research project and implemented using REDCap. In each site, a team member not involved in assessment or monitoring will assign participants by pressing a “randomisation” button on REDCap. A stratification will be done according to the frailty status. Within each stratum (robust and pre-frail), participants will be assigned to one of the two conditions (StayFitLonger or active control) according to separate randomisation schedules with a 1:1 ratio. Couples (e.g., married individuals) who participate in the study will be assigned to one of the two conditions as a pair: the first member of the couple will be randomised, and the second will be assigned to the same intervention. This has been implemented to avoid contamination in cases where two individuals living in the same household would be randomized to different training programmes.

Assessors will be blind to the hypotheses and to participants’ assignment, as they will only have access to the testing sessions. Participants will be asked not to discuss their training programme with assessors. If such a circumstance were to occur, it will be reported but should have minimal effect on data integrity, as assessors will be blinded to the hypotheses. Research team members responsible for the statistical analyses will be blind to the training conditions. Study coordinators and instructors involved in the introductory courses and supervision of home-based training will not be blind. Participants will be aware that the trial has two different training conditions and that they will be randomly allocated to one of them. However, they will not be informed of the study hypotheses and therefore will not know which one is the experimental condition. Both programmes will have a similar main screen layout and name, and the wording of the recruitment documents and consent forms will not convey the notion that one condition is hypothesised as inferior in terms of its effects on physical capacities and cognition [48].

Four face-to-face introductory sessions will be provided to familiarise participants with the material and the assigned application (Table ​(Table1).1). Two sessions of 3 h will present how to use the tablet and accessories (e.g., handling, charging), how to navigate in the application and how to complete the cognitive exercises. Two sessions of 2 h will present the physical exercises and teach participants how to place the motion sensor (Physilog®5, GaitUp, Switzerland) that will be used to record bodily measurements. In both programmes, physical activity instructors will ask participants to practice physical exercises for a total of 30 to 45 min distributed over the day. They will be recommended to train using the same physical exercises for at least 3 weeks with three sessions per week and a day of rest between each session. Cognitive training instructors will encourage participants to practice the cognitive exercises at least 3 times per week for 15 min each time. Participants will be made aware that during an ideal training session, activity should be perceived as of moderate difficulty. Instructions on physical and cognitive activities will be provided by a different instructor, the same for both programmes. Instructors will specifically be asked to present and explain the two programmes in similar ways.

The StayFitLonger programme will be accessible through the application RestonsEnForme, which will be available on a tablet (Galaxy Tab S2, Samsung) that will be provided to each participant. When launched, the main screen of the application provides access to different physical and cognitive activities (Fig. 1a) as well as other features listed below.

Illustration of the different activities of the StayFitLonger training programme

The physical exercise activity (Exercises) will be based on the T&E home-based programme using the concept of self-efficacy and empowerment [39]. Participants will be invited to create a personalised 8-exercise programme (Fig. ​(Fig.1b).1b). Those will be selected from 50 available exercises, which vary as function of themes (e.g., on a chair, with a pillow) and difficulty level (e.g., different body position or workload). Participants will try the exercises before including them in their programme. Exercices will only be included if judged as not too difficult by the participant. More details on the T&E programme can be found in [39]. During the intervention, participants will be allowed to add new exercises to their 8-exercise programme after a period of at least 3 weeks to introduce variety and increase challenge.

There will be four cognitive training activities which target problem solving, semantic memory, prospective memory, and divided attention. The Quiz activity will teach different strategies [49, 50] to learn new vocabulary and semantic repertoires (e.g., mushrooms, trees, flowers, dogs, etc.; Fig. ​Fig.1c).1c). Participants will choose first a repertoire of interest and will be asked to perform word-image associations related to the repertoire. Based on their level of proficiency in the selected repertoire, participants will then be offered different learning techniques: completing (relying on cues for help) or copying/completing (copying the word while using pure errorless learning and then completing while using encoding cues). Participants will continue to explore the repertoire through practice using an optimal number of cues to obtain the best performance while limiting the production of errors. This practice will be completed once participants reach at least 60% of correct responses without cues. Then, participants will be invited to a final evaluation without any help. Feedback will be provided with the option of continuing training using the repertoire or choosing another one. The 4 Images/1 Word activity will train cognitive flexibility (Fig. ​(Fig.1d).1d). Participants will be shown four images that are associated with an overarching concept and will be asked to find and write down the associated concept. Two types of cues will be provided to help them solve the task: number of letters in the target and some of the target letters mixed with distractors. The Attention! activity will train participants to vary their attentional priority in dual-tasks [40] while exploring a city on a two-wheel vehicle (Fig. ​(Fig.1e).1e). The dual-task will involve detecting different targets in the environment (i.e., people, 4-wheel vehicles or buildings) by pressing a button on the screen (task A), and at the same time detecting sewer covers with foot taping (task B). The foot response will be recorded by a motion sensor attached to the waist or shoe. The activity will comprise 30 levels with a progressive increase in the degree of difficulty. Difficulty will be increased by manipulating the number of targets, the number of distractors and the speed of the vehicle driven by the player (i.e., bicycle, scooter, motorcycle) and by introducing a response contingency condition (if/then). Participants will complete first each task (detection of targets in the environment and detection of sewer covers) in focused attention. They will then be asked to combine the tasks with different priority levels during a series of trials: one trial in which they will devote 80% of their attention to task A and 20% of attention to task B, one with 20% on task A and 80% on task B, and one with 50% of their attention on each task. Each priority trial will last about 1 min and will be repeated twice in random order. The Recall you activity will be embedded into the physical exercises to train prospective memory [51] (Fig. ​(Fig.1f).1f). On every 3 to 4 sessions, the Exercises activity will start with an instruction asking participants to complete a casual task (e.g., to get and drink a glass of water or to open a window, etc.) after a certain amount of time in the physical training. A timer will appear on the top left corner so that participants can track time while doing their exercises. For safety reasons, participants will be instructed to complete the exercise they are engaged in before performing the cognitive task.

In addition to the physical and cognitive activities, the StayFitLonger programme will include the following features:

A Chat room will provide a venue for participants to share views about topics of interest and tips for common real-life problems (Fig. 2a). Pre-established themes will be available (e.g., cooking, gardening, handiwork, etc.) and participants will have the opportunity to enrich this setting and create their own themes. When entering the chat room, a moderator message will inform participants to be respectful while chatting and to avoid revealing sensitive information (e.g., address, name, credit card information).

Illustration of the unique features of the StayFitLonger training programme

Creation of material. Participants will be invited to create material for the 4 Images/1 Word and Quiz activities. Once validated by the research team through a moderation platform, the material will be shared with all participants who will have the opportunity to use it for their own training and to rate the material created by other participants. This feature has been implemented to foster social interactions across participants.

Psychoeducation. From the application homepage, participants will have access to psychoeducational content (Fig. ​(Fig.2b)2b) on different topics related to physical, psychological and cognitive health. Twenty-two topics will be available (e.g., divided attention improvement; stress regulation; fatigue management, etc.).

Virtual coach. A customisable virtual coach using verbal (but written) and non-verbal communication (Fig. ​(Fig.2c,2c, d) will guide participants along the proposed exercises by giving them instructions, reminding them to practice a variety of available activities repeatedly, providing appropriate and timely feedback (through congruent facial expressions) on participant’s performances (e.g., encouraging messages) and rewarding assiduity, perseverance and performance with achievements and virtual credits (“physio-coins” and “cogni-coins”). Some achievements will unlock new icons, backgrounds and frames to modify the user interface, and by spending the coins obtained, it will be possible to get additional icons, background, frames and equipment to customise the virtual coach appearance (e.g., hat, glasses, etc.). These different functions of the virtual coach have been implemented to improve adherence by helping participants through a direct interlocutor (rather than neutral messages) and to keep them motivated [43].

The active control programme will be similar in structure and layout to the StayFitLonger programme (Fig. 3a) and will include physical and cognitive exercises.

Illustration of the different activities of the active control training programme

The physical exercise activity (Exercises) will be a computerised version of Helsana’s physical training programme (Fig. ​(Fig.3b).3b). Helsana, a Swiss health insurance company, offers this programme in a booklet. The computerised version will include advice and tips to stay physically active (e.g., to go shopping by foot) and 12 exercises to train upper and lower extremity strength, mobility and balance. It will also provide information about which exercises to choose, the training frequency and precautionary measures to follow. This programme has been judged close to “standard care”, as it is similar to a large range of programmes and recommendations available to the general public. It will differ from the Exercise activity available on the StayFitLonger programme, as it only contains a limited number of exercises and does not benefit from interactive content (e.g., videos of exercises), self-management, personalization features, and rewards from the virtual coach.

The four cognitive activities provided in the active control programme will be commercially available leisure activities that do not target specific cognitive processes and do not teach cognitive strategies [52–56]. The Crosswords activity will include 219 puzzles with five different sizes (Fig. ​(Fig.3c).3c). The Sudoku activity will include around 5000 puzzles with four levels of difficulty (Fig. ​(Fig.3d).3d). The Attention! activity will be a maze arcade game inspired from Pac-Man in which participants eat dots in a maze while trying to avoid coloured ghosts (Fig. ​(Fig.3e).3e). The Countdown activity will be embedded into the Exercises activity and triggered randomly every 3 to 4 days. It will require that participants count backward from 100 to 1 or recite the alphabet from Z to A while doing their exercise.

There will be no chat room, psycho-educational content or virtual coach included in the active control training programme.

Participants will receive a phone call and a home visit on week four and on week eight. Then, they will receive two phone calls (one from the physical activity instructor and one from the cognitive activity instructor) every four weeks. These will serve to identify and help participants to resolve difficulties with the programme, devices or exercises, and to obtain information about their health.

The primary outcome will be the performance on the TUG test [57]. In this test, the person will be sitting on a chair and will be asked to stand-up, walk three meters, turn around, walk back to the chair, and sit down. Time will be measured from the moment the person stands up until s/he sits down. Participants will perform the TUG twice providing two measures that will be averaged.

Measures in four domains will be used as secondary outcomes (Table ​(Table22).

Physical domain. 1) Walking speed will be measured over a 20-m distance. Participants will be instructed to walk as quickly as possible without running and in a safe manner. Time will be measured in seconds using a smartwatch. The task will be carried out twice, and measures will be averaged. 2) Lower extremity strength will be measured with the Five Time Sit to Stand Test (FTSST) [58]. Participants will sit on a chair with arms folded across their chest and will be asked to stand up and sit as quickly as possible five times while keeping their arms folded. The task will be administered twice, and the two measures averaged. 3) In the Four Stage Balance Test (FSBT) [59], participants will be asked to perform four progressively more challenging positions and to hold each of them as long as they can for a maximum of 10 s (parallel, semi-tandem and tandem positions) or as long as possible (one-leg stance position). The test will be stopped if a participant fails at holding a given position. 4) A smartwatch (Huawei Watch 2) connected to two motion Physilog®5 sensors worn by participants will be used during the TUG and the 20-m Walking task to collect additional specific gait movement parameters (Table ​(Table2).2). These sensors are a standalone 7 degree-of-freedom MEMS inertial measurement unit with wireless synchronisation, including 3D accelerometer, 3D gyroscope, and a barometric pressure sensor. The system is non-invasive, as sensors will be directly strapped on right and left shoe/foot.

Cognitive domain. 1) Global cognition will be measured with the ZAVEN composite score [60, 61] computed by averaging z-scores from the following tests: delayed free recall of the California Verbal Learning Test (CVLT); delayed recall of the WMS-IV logical memory subtest [62]; number of correct symbols reported in the WAIS-IV digit symbol substitution test (DSST) [63]; and letter fluency of the verbal fluency task [64]. 2) An executive composite score will be computed by averaging z-scores from the following tests: letter fluency of the verbal fluency task; time to complete the Trail Making Test part B-A (TMT) [65]; interference index of the Victoria Stroop Test (VST) [66]; number of total visual and auditory omissions of the divided attention subtest (Test of Attention Performance 2.3.1; TAP [67];). 3) A memory composite score will be obtained by averaging z-scores from the delayed free recall score of the CVLT [68, 69] and the delayed recall of the logical memory task. 4) A processing speed composite score will be obtained by averaging z-scores from the following tests: time to complete the TMT part A; number of correct symbols reported in the DSST; time to complete the “naming condition” of the VST [70]. 5) Divided attention will be measured with a customized computerised task performed on a tablet [40]. Participants will be asked to deliver newspapers by pressing on a screen button while on a bicycle that moves forward automatically. At the same time, they will have to follow the road traffic regulation to ensure their safety (e.g., stopping when traffic lights go from green to red and avoiding animals crossing the road). The tasks will involve different distractors to vary participants’ attentional demand. Participants will be made aware that they should prioritize their safety as they would in real life. Each task will be done first in focused attention and then both tasks will be combined and performed using three levels of speed. The number of delivered items, reaction time, and errors will be recorded. 6) Prospective memory will be measured with two subtests of the Rivermead Behavioural Memory Test (RBMT-3) [71]. In the “belonging” subtest, participants will be instructed to remember asking for two personal belongings at the end of the session. In the “appointment” subtest, participants will be asked to remember asking two questions when an alarm rings 25 min later. 7) Concept elaboration will be assessed with the TAP flexibility sub-test, a “set shifting” computerised task [67] and the WAIS-IV Similarities subtest [63]. In the Similarities subtest, participants will be presented with pairs of words (e.g.: apple and peach) and will be asked how the two words are alike.

Affective domain. Mood will be assessed using the Hospital Anxiety and Depression Scale (HADS) [72]. Fear of falling will be measured with the Falls Efficacy Scale-International (FES-I) [73].

Psychosocial domain. Quality of life will be assessed with the 35-item Older People Quality of Life questionnaire (OPQOL 35) [74]. Cognition in everyday life will be measured with the self-reported Cognitive Function Instrument (CFI) [75] and Everyday Cognition scale (E-Cog) [76]. The CFI will include 14 questions to measure subjective concerns regarding cognition and activities of daily living over the last year. The E-Cog will measure how cognitive functions in different domains (everyday memory, language, visuospatial abilities, planning, organisation, divided attention) impact activities of daily living compared to 10 years ago. Participant’s expectation toward the efficacy of the training programme will be assessed with an ad-hoc 17-item questionnaire.

Given that our secondary analyses will stratify participants into two categories (robust and pre-frail), we determined our sample size to ensure that we have the capacity to test the hypotheses related to this stratification. This was done with a Marker Stratified Design using the following plan: marker-by-treatment interaction using separate test (see: http://www.bigted.org/NonAdaptiveDesigns/MarkerStratifiedDesigns.html). For pre-frail participants, it was estimated that 16 participants per group (StayFitLonger vs. active control) would be required to detect a significant difference of 3.22 s in the TUG test using a two-sided t-test (alpha = 0.05) based on the T&E pilot study. As data might not be normally distributed, a non-parametric test was required resulting in a sample size of about 18 participants per group. Considering a dropout rate of about 25% based on prior studies, a sample size of 24.5 pre-frail participants should be enrolled for each group. For robust older adults, a sample size of 23.5 participants per group would allow to detect a difference of 0.82 s on the TUG test using a two-sided t-test (alpha = 0.05) with a power of 80% based on the study by Uemura et al. [77]. By accounting for the non-normality of data (using a non-parametric test) and the dropout rate, we targeted recruiting 36 robust participants per group. Thus, a total sample of 122 participants was determined as sufficient to have the appropriate power based on sample size calculation. To have a balanced distribution in the three countries, the total N targeted for recruitment was set at 128 participants.

All statistical tests will be two-tailed and a p value < .05 will indicate statistical significance. Effect sizes will also be assessed. Standard descriptive statistics will be provided with means and standard deviation for demographics and baseline characteristics. Group comparisons will be made using t-tests for continuous variables and chi-square analyses for discrete variables.

The primary efficacy analysis will be done with a modified intention-to treat (mITT) approach. All participants will be included in the analyses and the characteristics of those who withdrew will be analysed, as well as the causes leading them to leave the study. A linear mixed model will be used to analyse the data, as it handles correlated data and unbalanced designs and are robust against missing values. The fixed effects will be Intervention (StayFitLonger vs. active control), Time (T0, T1) and their interaction. If the StayFitLonger training is more beneficial than the active control training, a significant interaction will be expected. In such case, the presence of a significant difference between T0 and T1 in each group will be evaluated, as well as group difference on change scores at post-training using the pre-training and control group as reference points. The same analysis will be used with primary and secondary outcomes. To examine the effect of frailty status on efficacy, participants will be stratified into robust and pre-frail seniors and data will be analyzed separately in these two populations using the same method described above.

Age, sex, education and score on MoCA, four variables considered as time-invariant for the duration of the study, will be assessed as potential moderators of the impact of training on primary and secondary outcomes. Prior to their use, we will verify that they are independent from each other with a chi-squared test or correlations.