Setting, recruitment and eligibility criteria

Data collection will be conducted at the Interdisciplinary Memory Clinic (IDMC) in Bern and the Cantonal Hospital of Lucerne, Switzerland. Participants will be recruited by local newspaper adverts as well as directly through the respective Memory Clinic. A total of 162 elderly participants (n = 54 per study arm) from a continuum covering cognitively normal individuals at risk for AD with SCD to MCI (diagnosed or cut-off score in the cognitive telephone screening instrument (COGTEL) according to Alexopoulos [42]) and mild AD patients will be recruited. All participants will need to fulfill the following inclusion criteria: (1.) Ability to give their written consent to participate in the study; (2.) Age between 60 and 85 years; (3.) Native or fluent proficiency in German; (4.) Normal or corrected to normal vision and hearing; (5.) Ability to visit the study location repeatedly; (6.) MoCA (Montreal Cognitive Assessment [43] > 11. Exclusion criteria are patients with any contra-indications for MRI-scanning, substance abuse or severe medical conditions (of note: patients with mild depressive symptoms not fulfilling criteria for major depressive disorder are recruited into the sample to better represent clinical reality).

Trial design and study setting

This study will be conducted in a bi-centric, randomized, placebo-controlled, within and partially blinded three arm design to investigate the effect of our CCT. The study design is presented in Fig. 1.

Fig. 1

a Study Design: In the three study arms participants perform a cognitive assessment four times with an interval of 3 months in-between. Neuroimaging is performed at baseline and after 3 months in all study arms. After 6 months neuroimaging is performed in the cognitive training and the waitlist control group and after 9 months in the waitlist control group only. Blood samples are collected once, in the first group appointment. b Assessment: Each assessment has a duration of approximately 3 hours if neuroimaging is included. Abbreviations Figure a: CCT = Computerized cognitive training, CG = Control group, A = Assessment, o = Neuroimaging. Abbreviations Figure b: MoCA = Montreal Cognitive Assessment, AVLT = Auditory verbal learning test, GNT = Graded naming test, fMRI = Functional magnetic resonance Imaging, ASL = Arterial spin labelling

Participants will be randomly assigned to one of three study arms (training, active control, or waitlist control group) based on a stratified Mahalanobis distribution procedure conducted in R [44] balancing arms by level of cognitive impairment, age, and sex through a continuous adjustment of the randomization probabilities.

Participant timeline

Participants will receive information that they will undergo a cognitively activating serious game-based computerized training starting either immediately or after a 3 months’ observation period but without knowing the distinction between the CCT and the active control group. Staff conducting the cognitive testing will be blinded to the study arm. The training will be conducted in small groups of 3–6 participants. The training group immediately starts with the CCT for 10 weeks, including weekly group sessions and will subsequently perform the training for twenty more weeks with monthly booster sessions. The waitlist control group will start the CCT with a delay of 10 weeks. The active control group will perform 10 weeks of unspecific cognitive activation (watching documentaries) before subsequently starting the CCT for 10 weeks. Ten weeks after the fourth assessment (i.e. after 3 × 10 weeks), all participants will be followed-up with a parallel version of the COGTEL.

Intervention description

Based on literature and standardized cogntive tests, we designed four tablet-based training games for each of the three key cognitive domains specifically affected by AD (episodic memory, semantic memory and spatial abilites) in addition to WM-based training. All games are fully developed in-house, programmed in unity (http://unity3d. com/) and provided on 10.2 in. iPads. Individual games are connected by an overarching story-line (a sailboat traveling from one island to the next where each island represents a game) and are performed repeatedly within sessions. A detailed description of each game can be found in the supplementary material (see Additional file 1).

During most on-site group sessions (week 1–4, 6, and 8), two new training games are introduced and practiced with the support of the investigators. The games are distributed in an accumulative pseudo-randomized order to balance between memory components. Each game consists of different levels of difficulty and adapts automatically to the abilities of the participants. Furthermore, cooperative elements like a puzzle which can be solved as a group provides a social experience. According to recommendations of a recent meta-analysis, participants play the games daily for 24 minutes, i.e., three games for 8 minutes each [23]. The study design thus required between 50 and 150 days with training sessions depending on the study arm, resulting in each game being played at least 12 times. Before every at home training session participants will be asked to rate their motivation for the session and how they feel at the moment. After each session, they subjectively will rate their performance.

After 1 week of playing each game, we will assess game perception using the game perception questionnaire [45, 46]. General gaming experience is assessed with the core elements of gaming experience questionnaire (CEGEQ) [47].

Explanation for the choice of comparators: active control group and waitlist control group

To test whether potential improvements in the intervention group result specifically from the training and not just from e.g. meeting as a group or performing an activating task, we use an active control group. To match the at-home element of the intervention group participants in the active control group will be asked to watch documentaries on the tablet computer matched for time and frequency of the CCT. To ensure that the participants are attentively watching the videos, pop-up windows with easy to solve multiple choice questions (e.g. “What is 2+3?”) randomly appear during the documentaries sessions. This active control condition aims to include the same level of social interaction in a group setting again matched for duration and frequency but without explicit cognitive training. To this end, participants are asked to discuss the documentaries with the other group members. Additionally, participants will play group games that are not cognitively demanding and where success is based on chance rather than skill (e.g., Bingo in slow speed). The active control intervention lasts 10 weeks. Afterwards, participants perform the same CCT as the intervention group.

Before staring the CCT, the wait-list control group will have a 10-week passive waiting period after the first cognitive assessment and MRI scan. The waitlist control group will allow us to quantify test-retest effects as well as habituation to the MRI scanner setting as in previous work by others [48].

Strategies to improve adherence to the intervention and to promote participant retention

As previous CCT studies often faced mal-adherence, we included weekly groups sessions to not solely have a training purpose but to also encourage social interaction among participants. Additionally, based on research on the most useful intervention protocols, we encouraged daily training and followed up with booster sessions over 6 months for long-term benefits in the intervention group. Game characteristics favorable for training success such as a challenging but playful design were implemented based on previous research findings [49, 50]. To promote motivation and to provide direct assistance, participants are called after their first at-home training and a help button within the app triggers an automated e-mail to the investigator. Response data is transmitted to the investigator to track compliance to the training protocol and to provide low-threshold support in case of difficulties or potential non-adherence.

MRI protocol

We will perform several MRI sequences before and after the CCT to assess functional and structural brain changes. Imaging data will help identify elements of restoration and compensation in task related brain activity after the CCT [51, 52]. Two fMRI tasks will be performed in this study. The first task will be an adapted version of a face-occupation task described in a previous study [53] to assess EM, while the second one is an adapted version of a task to assess spatial abilities (SA) [54]. In the face-occupation task, the association between face and occupation pairs has to be learned during the encoding phase. During the control condition, head contours are displayed and participants have to indicate, if the head contour belongs to a male or female. In the cued recall condition, the previously encoded faces are shown again and the question whether the occupation requires a university degree or an apprenticeship has to be answered. During the recognition condition, the faces are presented with two occupations and participants have to indicate which occupation is associated with the face. We expect activity in the medial temporal lobe and occipital brain areas as a previous study with a very similar paradigm showed activity in these brain areas. While the activity in the MTL is expected in EM tasks, the occipital activity might be caused through the task design, where different stimuli were used in the control than in the other conditions (i.e., pictures of faces vs head contours) [55]. In the SA task, participants are asked to mentally translate or rotate two puzzle pieces to indicate if they fit together to build a square. As control condition, two grey squares are presented in similarly shaped puzzle pieces. The participants are asked to indicate if both grey squares have exactly the same color. Based on previous results in the spatial abilities task, we expect activity in the superior parietal cortex bilaterally, the left inferior parietal cortex, bilateral lateral occipital cortex and the medial frontal gyrus. These brain areas are associated with complex visual-spatial construction tasks [54]. Please see the supplementary material for a detailed description of the fMRI tasks (see Additional file 2).

In the analysis of task fMRI, we will predefine regions of interest (i.e., regions with significant activity in the baseline assessment) and investigate how these areas change activity through the training. Additionally, we will also perform whole brain analyses to find new clusters of activity not present in the baseline assessment as a possible sign of compensatory activity [48]. Scanning at both study sites will be performed with 3 Tesla scanners (Bern: Siemens Magnetom Prisma, 32 channel head coil; Lucerne: Siemens Magnetom Vida, 64 channel head coil). From all participants, T1- weighted images (MP2RAGE, TR = 5000 ms, TE = 2.98 ms, TI 1/2 = 700 ms/2500 ms, flip angle 1/2 = 4 degrees/5 degrees, FOV = 256, matrix = 256 × 256, voxel size = 1 × 1 × 1 mm, 176 slices) will be acquired. For the fMRI sequences we will acquire echo-planar imaging (EPI) with 360 (resting state fMRI), 604 (SA Task) and 638 (EM task) volumes (TR = 1000 ms, TE = 37 ms, flip angle = 30 degree, FoV = 230 mm, matrix = 92 × 92, accelerating factor 8, voxel size 2.5 × 2.5 × 2.5 mm, 72 slices). Furthermore, we will perform an ASL sequence (91 volumes, TR = 3300 ms, TE = 13 ms, Flip angle = 90 degree, FoV = 230 mm, matrix = 64 × 64, voxel size 3.6 × 3.6 × 6.0 mm, 22 slices).

Blood analysis

Blood samples will be taken once during the first group session. Genotyping for BDNF ApoE will be performed. Furthermore, plasma pTau181, Amyloid Beta42, Amyloid Beta40, glial fibrillary acidic protein and neurofilament light will be analyzed to estimate AD related brain pathology [38,39,40].

Outcomes

Each participant will perform the cognitive assessments four times in intervals of approximately 10 weeks. The individual cognitive profile will be assessed through a set of paper-pencil and tablet-based cognitive tests designed to evaluate typical AD components. More precisely, the general cognitive performance will be assessed in a paper-pencil version of the MoCA. The following tests will be administered as tablet-based versions: the auditory verbal learning test (AVLT) [56] to assess episodic memory, the Rey-Osterrieth Complex Figure Test (ROCF) [57] for spatial abilities/memory, and verbal fluency as well as the Graded Naming Task (GNT-30) [58] will be used to probe semantic memory. Additionally, participants will performed the Flanker Test [59] to assess attention/response inhibition abilities and the digit span test (forward and backwards) as measure of short-term memory and working memory respectively. Parallel versions of the tests will be used whenever possible. Based on this comprehensive test battery, a cognitive composite score will be calculated as primary outcome of cognitive performance.

Moreover, participants will be asked to complete questionnaires to assess their form of the day, their situational motivation [60], quality of life [61] and depressive symptoms trough the geriatric depression scale (GDS) [62]. Through in-house designed or adapted questionnaires, participants’ expectancy towards the cognitive training (Additional file 3) [63], and subjective cognitive performance (Additional file 4) will be assessed. Subjective cognitive performance will additionally be assessed in an informant rated version of the questionnaire (filled out by a close friend or relative) (Additional file 5). Unexperienced participants will have the opportunity to get used to the tablet and to have one on one support from the investigator before the training starts.

The mentioned tests will be supplemented by a set of paper-pencil based questionnaires that will be sent to the participants before the first on-site meeting. These preliminary questionnaires asses activities of daily living [64], handedness [65], and the informant rated subjective memory performance questionnaire.

Statistical methodsSample size

For our primary analysis (“Does change in performance on the three typical AD domains 3 months after the start of the training differ between groups?”), a repeated- measures ANOVA is planned (3 groups with 2 time points). We computed the necessary sample size using G*Power [66] to ensure a power of 80% to detect an assumed moderate effect (i.e. effect size: η2 = 0.06, f = 0.25) with α of 5%. The power analysis revealed a total sample size of 162 (54 participants per group). Moreover, we performed a power analysis for testing the interaction between the groups and all four time points within a repeated-measures ANOVA using the same parameters as before (α = 5%, power = 80%, η2 = 0.06) as well as a non-sphericity correction factor of 0.75. According to this power analysis a total sample size of 96 (32 participants per group) is needed.

Statistical methods for primary and secondary outcomes

Primary outcomes are three composite scores corresponding to the domains mostly affected by AD. One advantage of composite scores is a reduced number of statistical tests. At the same time, separate composite scores for the three domains allow to assess and compare changes in the scores individually. For each of the domains, three variables will be included in the composite score (EM: AVLT learning sum, immediate and delayed recall; SA: ROCF encoding, immediate and delayed recall; semantics: semantic fluency from the MoCA, phonemic fluency and GNT). We will perform a principal component analysis (PCA) for each of the respective domains.

All scores will be scaled and centralized on the scores from the baseline assessment and weighted if appropriate according to their loadings in the PCA. The main hypothesis will be tested with a repeated-measures ANOVA for which an adequate level of power should be guaranteed (according to the power analysis described before). Further analyses (using all time points and additional numeric subject characteristics as predictor variables) will be performed with linear mixed models as they allow for more sophisticated model design and a better handling of missing data. As covariates, the factors study site and study arm will be included in all analyses. Neuroimaging data will be analyzed with SPM12 and related toolboxes (e.g., Functional Connectivity (CONN) toolbox [67]). Apart from task related fMRI, resting state fMRI, volumetrical images and ASL will be performed.

To assure meaningful data, non-adherence to the training regimen leads to exclusion of further analysis. Non-adherence is defined as completing less than 50% of the required training sessions and attendance of less than 70% of the group sessions.

Given that a relatively high number of games are purposely built for this study, we expect the need for continuous adjustments based on feedback from participants. Should substantial adjustments be necessary, we will introduce release number as a covariate in the analysis.

Dissemination of outcomes

We will disseminate findings from this research as widely as possible to reach academic, clinical and public audiences. All academic publications will be peer-reviewed, open access and we will share findings through a range of other channels, e.g. newsletters, study websites, social media, presentations at conferences and events.

Adverse event reporting and study integrity

Health hazards will be reported to the Sponsor-Investigator within 24 hours upon becoming aware of the event. Health hazards will be reported to the local Ethics Committee within 2 days.

Substantial changes to the study setup and study organization, the protocol and relevant study documents are submitted to the local Ethics Committee for approval before implementation.