van de Ven 2017.

Study characteristics
Methods	Design: parallel RCT Duration of trial: April 2013 to March 2015; last follow‐up measurement in November 2015 Unit of randomisation: people with stroke 3 months to 5 years prior, aged 30–80 years, and had received inpatient or outpatient rehabilitation therapy Recruitment and allocation: participants were recruited from a patient database of rehabilitation centres and 1 hospital and from advertisements in newsletters and forums of national stroke patients associations. Of 223 assessed for eligibility, 97 were randomised to 1 of 3 groups: IG (n = 38), active CG (n = 35), and waiting list CG (n = 24). Data for the IG and waiting list CG were used in this review.
Participants	Setting: participants' homes Country: the Netherlands Sample size: 97 adults, 69% men; IG: 38; waiting list CG: 24 Inclusion criteria: stroke 3 months to 5 years prior, aged 30–80 years, and had received rehabilitation therapy as inpatient or outpatient, cognitive impairments after stroke (as testified by medical records), with cognitive complaints still present at study entry, able to work with a computer and have daily access to a computer with Internet connection Exclusion criteria: presence of neurodegenerative disease; epilepsy; serious psychiatric illness; any disease other than stroke that resulted in severe cognitive impairments; drug or alcohol dependency; severe colour blindness, aphasia, neglect, or computer fear; disabling vision or auditory problems; diagnosed learning disability, not mentally or physically fit enough to be able to complete 12 weeks of training, unable to understand the training instructions or who could not execute the training due to any other unforeseen reason, after instructions or after the first training week (van de Ven 2017a) Age: mean: IG: 57 (SD 9.1) years; waiting list CG: 61.2 (SD 9.0) years Time since stroke onset: mean: IG: 28.3 (SD 16.4) months; waiting list CG: 29.1 (SD 17.0) months Types of stroke: not reported Hemisphere: not reported
Interventions	Intervention group Brief name: cognitive flexibility training (BrainGymmer) Recipients: adults with stroke within last 5 years with cognitive impairments Why: to improve cognitive flexibility and executive functioning through cognitive computer‐based training that included frequent switching between various training tasks. What (materials): online platform BrainGymmer (in Dutch); English version: www.braingymmer.com/en/; daily access to a computer with Internet connection and sound (either through headset or speakers); 9 tasks consisting of 20 levels selected to train 3 cognitive domains: attention, reasoning, and WM (see Additional file 2 van de Ven 2015 for brief descriptions and pictures of tasks; www.ncbi.nlm.nih.gov/pmc/articles/PMC4545547/: attention tasks: Pattern Matrix, Birds of a Feather, Mind the Mole reasoning tasks: Square Logic, Out of Order, Patterned Logic memory tasks: Toy Shop, Multi Memory, Moving Memory Training videos of each task on individual computers Instruction booklet What (procedures): training instructions were provided at baseline using videos on computers followed by supervised practice by participants and provision of training booklet. Participants trained daily on several tasks within 1 session and frequently switching between the tasks. Participants were asked to train when they had ≥ 50 minutes available and when not mentally fatigued (e.g. late at night). Personal feedback was provided based on individuals' scores after each task, using a 3‐star rating scale, with more stars for better performance and at the end of each session with more detailed feedback of their scores on each task. Email availability for questions or to address any problems. Weekly or fortnightly contact by telephone for participants to ask questions and for discussion about training adherence. Reminder email sent if no training for 2 days and personal contact if 3 days without training. Daily log by participants. Exit questionnaire administered. Who provided: trained Master's degree‐level students who were familiar with all training tasks and login procedures for remote support and training instruction. Neuropsychologist for weekly or fortnightly contact. How: individually and independently without face‐to‐face supervision at the time but with remote assistance if required; individual and personal telephone and email contact with trainers/neuropsychologist Where: at home for computer tasks; training at University of Amsterdam When and how much: 10 tasks daily for 3 minutes each (30 minutes), 5 times per week for 12 weeks; 58 sessions (29 hours in total); In week 1, participants trained for 10 minutes each day on 3 tasks then from week 2, the number of trials per task was reduced to promote frequent switching between task. Trainer contact weeks 1, 2, 3, 5, 6, 8, and 10. Email contact if no training for 2 days Tailoring: difficulty of tasks was adapted individually to the performance of participants. Participants were instructed to go to the next level when obtaining 2 or 3 stars. However, participants could choose to stay at the same level when receiving 2 stars, whereas they were obliged to stay at the same level when obtaining no or 1 star. Tasks were set up each session for each participant. Order of tasks ensured that tasks from the same cognitive domain (attention, reasoning, and WM) were not presented immediately after each other (van de Ven 2015). Emails were sent as soon as participants did not train for 2 days. Modifications: none reported How well (planned): weekly or fortnightly discussion by telephone about training adherence. Number of sessions was recorded. Degree of adaptiveness of the CT was assessed. Amount of extra personal contact (telephone or email) due to questions or technical issues during training and level of engagement (i.e. how often a reminder to train was needed) was recorded. Participant daily log of their level of motivation during training, amount of PE at the day of training, how interesting and difficult the tasks of that day were, fatigue level before and after training. Exit questionnaire about subjective training effectiveness; change of strategies during training; check of blinding to experimental condition; changes in cognitive stimulation in daily life besides study related training; and major changes during training. How well (actual): the number of sessions did not differ significantly between the intervention training group and the active control group. The degree of adaptiveness was compromised in 17% of the intervention group participants. 5 participants (17%) were slightly less challenged in the last weeks of the training because they reached the highest level and score possible on 1 of the 9 tasks. Comparator group Brief name: waiting list CG or mock training (active CG) Recipients: adults with stroke within last 5 years with cognitive impairments Why: active CG: to provide comparable computer‐based training selected to be sufficiently challenging but not too difficult and thought to be to unlikely to improve executive functioning; waiting list CG: no intervention What (materials): active CG: non‐adaptive mock training consisting of 4 computer tasks using same online platform braingymmer with 9 levels each (see Additional file 2 van de Ven 2015 for brief descriptions and pictures of tasks); Fuzzle, Sliding Search, Pay Attention, Grid Tracks; daily access to a computer with Internet connection and sound (either through headset or speakers) Training videos of each task on individual computers Instruction booklet; waiting list CG: none What (procedures): active CR completed computer‐based tasks with the same amount of feedback, motivational instructions, visual stimulation, and use of mouse as the intervention training. Email availability for questions or to address any problems. Weekly or fortnightly contact by telephone for participants to ask questions and for discussion about training adherence. Reminder email sent if no training for 2 days. Daily log by participants. Exit questionnaire administered; waiting list AC did not receive training and were not contacted by telephone during the first 12 weeks. Who provided: active CG: trained Master's degree level students who were familiar with all training tasks and login procedures for remote support and training instruction. Neuropsychologist for weekly or fortnightly contact; waiting list CG: no‐one How: active CG: individually and independently without face‐to‐face supervision at the time but with remote assistance if required; individual and personal telephone and email contact with trainers/neuropsychologist; waiting list CG: none Where: active CG: at home; training at University of Amsterdam; waiting list CG: none When and how much: active CG: 3 tasks per day for 10 minutes (30 minutes) switching between tasks every 10 minutes; waiting list CG: none Tailoring: active CG: level increased each week during first 5 weeks, then every 2 weeks during weeks 6–12 unless participant did not master a level (at least 1 star). Emails were sent as soon as participants did not train for 2 days; waiting list CG: none Modifications: none How well (planned): active CG: weekly or fortnightly discussion by telephone about training adherence. Number of sessions was recorded. Degree of adaptiveness of the cognitive training assessed. Amount of extra personal contact (telephone or email) due to questions or technical issues during training and level of engagement (i.e. how often a reminder to train was needed) recorded. Participant daily log of their level of motivation during training, amount of PE at the day of training, how interesting and difficult the tasks of that day were, and fatigue level before and after training. Exit questionnaire about subjective training effectiveness; change of strategies during training; check of blinding to experimental condition; changes in cognitive stimulation in daily life besides study related training; and major changes during training; waiting list CG: none How well (actual): active CG: participants were asked not to train beyond level 9. However, some participants disregarded this and trained at higher levels. The degree of adaptiveness was compromised in 83% of the active control group; waiting list CG: none
Outcomes	Primary: none Secondary Lawton & Brody Instrumental Activity of Daily Living scale USER‐P CFQ DEX TMT‐A TMT‐B D‐Kefs TMT number‐letter switching condition Tower of London Letter‐Number Sequencing PASAT Digit‐Symbol‐Coding Rey's Auditory Verbal Learning Test Blokkenreeksen (NeuroTask BV); online modified version of Corsi task Raven's Progressive Matrices Shipley Institute of Living Scale‐2 Other Category Fluency Letter Fluency Switch Task Dual Task Category Fluency switch condition N‐back D‐Kefs TMT motor speed condition Mouse skills tasks Stop‐signal task Methods of data collection: neuropsychological outcomes assessed by assessor blinded to groups. Some outcomes were measured online without an assessor and remainder in person at the University Data collection time points: baseline (week 1), after 12 weeks' intervention (week 13), 4 weeks after completion of intervention (week 17)
Notes	Funding: yes Conflict of interest: none Published trial protocol: yes (van de Ven 2015) Trial registration: yes: Central Committee on Research Involving Human Subjects Register NL4468502913; Netherlands National Trial Register NTR5174 Ethics approval: yes Data: reported results included some for participants randomised to the IG and CG and some for participants who completed the training according to the protocol (e.g. completed ≥ 50 sessions) and who completed outcome measures after the intervention and on follow‐up. Where possible we used the data reported for those participants who completed the protocol and did not drop out.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were randomly assigned by randomisation software (Minimpy).
Allocation concealment (selection bias)	Unclear risk	Process of allocation not reported.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Participants of the IG and active CGs were blinded, the participants of the waiting list CG would not have been blinded. The person administering the computer tasks and training instructions was not blind to training allocation.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Groups were coded by the research co‐ordinator such that the assessors were blind to which training condition participants were in.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Although the paper indicates loss to follow‐up, the reasons for losses were reported and these were balanced between groups.
Selective reporting (reporting bias)	Low risk	All planned outcomes reported as per protocol.
Other bias	Low risk	No other identifiable bias.