Computerized Cognitive Training in Healthy Older Adults : Baseline Cognitive Level and Subjective Cognitive Concerns Predict Training Outcome

Computerized cognitive training (CCT) can improve cognition in older adults. However, specific programs for this target group have rarely been developed, and predictors of training success are largely unclear. In a randomized controlled pilot trial, we examined effects of a six-week CCT, tailored to the cognitive profile of healthy older adults (EG, n = 21) as compared to a nonspecific CCT (CG, n = 18) matched in frequency and intensity. No interaction effects between Group and Time were found. However, within-group analysis showed significant gains in verbal and non-verbal memory, executive and visuospatial functions and subjective cognitive concerns (SCC) in the EG, while the CG only benefitted in non-verbal memory and set-shifting. Low cognitive performance and lower SCC at baseline were the most consistent predictors of cognitive gains in the EG. Thus, CCTs specifically tailored to older adults seem advantageous compared to non-specific CCT. Further, we conclude that SCC may be related to reduced cognitive plasticity.


Introduction
Recently, cognitive interventions to enhance cognitive functions and attenuate cognitive decline in healthy elderly individuals have attracted increasing scientific attention.There is robust evidence that the aging brain is amenable to neuronal and cognitive plasticity [1] [2] [3] and has the potential to enhance cognition and possibly everyday functioning through cognitive training (CT; [4] [5] [6] [7] [8]).However, more research with standardized training protocols and outcome measures is necessary to allow conclusions about the optimal type and dose of cognitive interventions [4].
Computerized cognitive trainings (CCT) have the advantage that they can be administered at home and are usually not cost-intensive.A systematic review of [9] reported that studies with CCTs found benefits in global cognitive performance, reaction time, processing speed, working memory, executive function, attention, memory, and visual spatial ability in older adults.Effect sizes differed according to the type of computerized program used-classic CT tasks, neuropsychological software, and video games.However, in a recent systematic review, which classified studies on CCT in healthy older adults regarding their evidence level, the two studies providing Level III only reported memory improvement with medium to large effects and one of these studies additionally found benefits in processing speed with a large effect size [10].Importantly, the type of CT tasks seems to play an important role [10].Reference [9] concluded that classic CT tasks provided the best results (with the exception of gains in reaction time for which video games were most effective) and that effects of these tasks are most comparable benefits to more traditional CT approaches.Finally, a systematic review and meta-analysis of [11] also emphasized that benefits of CCT in healthy older adults are largely determined by design aspects of the training.
Their findings indicate that supervised training is more effective than home-based CCTs, and that an intensity of 30 minutes per session with not more than three sessions per week is most favorable [11].
Remarkably, CCTs specifically tailored to the typical cognitive profile of healthy elderly individuals, which is characterized by decline in episodic memory, executive functions, attention and processing speed, but sometimes also visuospatial dysfunctions [12] [13] [14], are rare.Furthermore, psychoeducational elements e.g. with topics related to healthy cognitive aging including risk and protective factors, mechanisms of cognitive functions, or cognitive strategies are part of some cognitive group trainings which were developed for elderly people with or without cognitive dysfunctions [15] [16] [17] [18], but are usually not implemented in digital trainings.However, standardized CCT programs tailored for elderly people could be especially useful, as they would constitute a form of "guided" training [11] for this group.Thus, whether CCTs that are specifically tailored to the profile of healthy elderly and that include psychoeducational elements, e.g.delivered with videos, are more effective than untailored CCTs, is a worthy topic which needs research.Predicting training response may be helpful to define which individuals will profit most from which type of CT.In studies examining effects of CT and combined interventions (e.g.CT with physical exercise) in healthy elderly people, cognitive baseline performance [17] [19] [20] [21] and also genetic and neurobiological factors [17] [22] [23] have been identified as possible predictors for training success, although data are inconsistent, as for example both low [17] [20] [21] and high [19] cognitive baseline performance have been associated with training gains.Depressive symptoms seem to be a negative predictor for training success; for example [24] found, in healthy old adults, depressive symptoms are associated with a reduced ability to utilize cognitive resources [see also [25]] and thus, with a reduced ability to profit from memory training.Additionally, a more recent study of [26] identified depressive symptoms as a moderating variable between (low) cognitive scores and cognitive gains induced by a cognitive stimulation program in elderly with dementia.The negative influence of depression on training success may be due to reduced motivation and alertness [27] [28], concentration problems, as well as impaired self-confidence.
Indications that computer familiarity does not seem to be predictive for success of computer-based memory training [29] suggest that CCT has the potential to be a practical and viable method for delivering interventions, even for elderly people without digital familiarity.In patients with cognitive impairment (mild cognitive impairment or dementia), the outcomes of cognitive interventions are related not only to baseline cognitive performance [26] [30] baseline functional abilities and behavioural symptoms [31], but also age [32] [33], sex [17] [32] [34], education [35], and again genetic factors [31] have been related to outcomes of cognitive interventions, although findings rely on only few studies with partly inconsistent results.Remarkably, based on the literature, further predictors of CT success exist, but are so far under investigated; these include self-efficacy [36], motivation [37] [38], and subjective cognitive concerns, the latter of which constitute a risk factor for later dementia [39] [40].Taken together, data suggest that several factors may predict CCT's outcome, but further research is necessary.
On the basis of the noted considerations, the aims of this pilot study were (i) to investigate the effects of an adaptive CCT program specifically developed for aging people including psycho educational elements presented in videos on topics related to healthy cognitive aging, mechanisms of cognitive functions, as well as cognitive strategies, and (ii) to define predictors of intervention outcomes.
For this purpose, we conducted a randomized controlled, single blind study, in which healthy elderly individuals received either a structured six week CCT specifically developed for this target group, or a less structured control CCT not tailored to this group, but comparable in frequency and intensity.Planning and reporting followed the recommendations of the CONSORT Statement for RCT [41] [42] [43]-except for the fact that blinding of neuropsychological assessment pre-and post intervention could not be achieved and only post-hoc power Health calculation was performed.We hypothesized that the tailored CCT program would induce more positive effects on cognitive outcomes compared to a control intervention.Furthermore, we expected that sociodemographic and cognitive performance at baseline as well as self-efficacy, motivation, depression score, and subjective cognitive concerns would predict the intervention's outcome.
Furthermore, in a more explorative way, we also examined whether technology commitment predicted training outcome.

Materials and Method
The present study was conducted in compliance with the World Medical Association Declaration of Helsinki [44].The study was approved by the Ethics Committee of the University of Vechta, Germany and is registered at the WHO ICTRP (ID: DRKS00010096).All participants gave written informed consent before the first neuropsychological assessment.

Participants
Healthy independent home dwelling participants aged between 55

Study Design
The study was designed as a

Experimental and Control Intervention
Both the experimental and the control training were comparable with regard to  S1 and Figures S1-S10 of the Supplementary Material.

Experimental Intervention
The "NeuroVitAALis" Software which has been designed on the basis of the paper and pencil multidomain cognitive group training program "NEUROvitalis" (for a detailed description, see [18] [47]) and was prototypically implemented by the Serious Games group at Technische Universität Darmstadt, using the authoring environment StoryTec (http://www.storytec.de;[48] [49]).The "Neuro-VitAALis" Software is a neuropsychological software application for tablet computers targeting the stabilization and amelioration of age-sensitive functions on the basis of brain plasticity.
The first exercise (20 minutes) of each session targets either memory or executive functions.The second exercise (10 minutes) could be chosen by the user and either trained spatial cognition, attention, language, or, if not already practiced during the session, either memory or executive function.All exercises had different levels of difficulty (with three to twenty-four levels).Each session also contained a psychoeducational video (10 -15 minutes) on topics related to healthy cognitive aging or to cognitive functions and cognitive strategies.Finally, an "advice of the day" was given which aimed at the stimulation of cognitive, mnestic, and social activities as well as structuring the day.One example would be the following: "Do you plan to buy groceries today?Try to memorize your shopping list and do your groceries without using it.Check the list after you have gathered all products in order to review that you did not forget something." The tasks were designed to autonomously adapt to the performance level of the user.However, as the program was still in a pilot phase, the automatic adjustment of difficulty levels was not implemented during the study period yet.In order to approximate comparability to a fully structured automated program, the study protocol included concrete instructions on how to adapt the degree of difficulty of the exercises to participants' personal performance level.
After thirteen lessons, two individuals reported that the exercise for the training of executive functions had an undemanding degree of difficulty on its highest level.After consultation with the research group, a modification of the study protocol was administered in these cases, and the exercise was replaced by the Health exercise for spatial cognition on level ten, which is highly demanding and also requires executive functions in its operation.

Control Intervention
The control training was conducted on computers.Exercises were taken from the German dyslexia software "Tintenklex" (http://www.legasthenie-software.de)and did not specifically aim at the amelioration of age-sensitive functions.Ten "Do you feel like your memory is becoming worse?".Furthermore, for each subjectively impaired cognitive domain, worries concerning this experienced worsening were assessed, i.e. "If yes, does that worry you?" with four answer options "No", "Sometimes", "Yes, that worries me", "I don't know".Total scores for subjectively impaired cognitive domains (0 -5 points) and worries (0 -10 points) were created.

Further Scales Used for the Predictor Analysis of Training Success
Further variables used for the predictor analysis of training success were technology commitment (consisting of technology acceptance, technology competence and technology control) as assessed using the technology commitment scale [58] as well as self-efficacy as measured using the SWE-Scale [59].Individual motivation was assessed by an average grade calculated by school rates (1 to 6 comparable to the A to F system) the participants gave before ("How motivated are you to do the training today?") and after ("How did you like the training today?") each session.Alternatively, in individuals who failed to fill in grades for all sessions, a single average grade for the intervention was obtained retrospectively via telephone interviews.BDI-II depression scores obtained at screening were also included in the analyses.

Statistical Analyses
Statistical analyses were performed using IBM SPSS Statistics 23 for Windows (2015).Post-hoc power analysis was performed using the software G * Power 3.1.7[60].Normal distributions were tested using the Kolmogorov-Smirnov test.
In case of normally distributed measures, parametric analysis was performed.
Otherwise, non-parametric tests were used.
Baseline data of experimental and control group were compared between groups, using t-tests for independent samples to compare the age, years of edu- but not the intensity of worries about the perceived worsening.The assumptions for multiple regression were checked according to the suggestions of [62].

Pre-to Posttest Change Comparisons between Groups
For the analysis of pre-and posttest group comparisons (N = 39, 2-tailed α = 0.05) the power to detect small effects (Cohen's d ≥ 0.2) was 9.2%, the power to detect moderate effects (Cohen's d ≥ 0.5) was 32.3% and the power to detect Health   measure (all p > 0.05).However, note that the power to detect group differences even for strong effects was generally low.

Changes from Pre-to Posttest within Groups
For the detection of effects from pre-to posttest (experimental group n = 21, control group n = 18, 2-tailed α = 0.05), post-hoc power analysis revealed a power of 13.6% in the experimental and 12.2% in the control condition to detect small effects (Cohen's d z ≥ 0.2), a power of 56.6% in the experimental and 49.6% in the control condition to detect moderate effects (Cohen's d z ≥ 0.5) and a Health power of 92.5% in the experimental and 87.6% in the control condition to detect strong effects (Cohen's d z ≥ 0.8).
The results of the outcome measures pre-and post-intervention for both groups are presented in Table 2. On a descriptive level, posttest scores were consistently higher compared to pretest scores in the EG, whereas results in the CG were mixed.While in the EG, significantly higher values post-as compared

Discussion
This study aimed at examining the effects of a CCT especially designed for older adults compared to an unspecific CCT in a group of healthy older adults.The main findings of the study are that (1) no Group × Time interaction effects were found for the CG versus the EG when RCIs and change scores from pre-to posttest were compared, but that (2) in within-group comparisons, the EG showed significant gains in verbal short and long-term memory, non-verbal short and long-term memory, set-shifting, visuospatial functions, and SCC, while the CG only benefitted in non-verbal long-term memory and set-shifting and even worsened in an alternating fluency task.Finally, (3) low cognitive baseline performance as well as lower SCC at baseline were the most consistent predictors of cognitive gains in the EG.
Although we were not able to find Group × Time interaction effects, the fact that substantially more gains were achieved in the CCT target intervention as elicited in the within-group comparisons allow the (tentative) conclusion that it was indeed more effective than the active control intervention.When interpreting these results one has to keep in mind that the effects of our target intervention might be underestimated for two reasons: first, an active control group was used.Although [11] did not find a difference between active and passive control treatments when analyzing CCT effects in cognitively healthy elderly, the meta-analysis conducted by [8] did find differences of effects in studies with active versus passive control treatments in executive-control and working memory training in older adults.Second, it can be assumed that the kind of active control treatment is relevant for the possibilities to find effects.In our study, the control intervention comprised an unspecific CCT which was not tailored to the group of healthy elderly, but which still trained a broad spectrum of functions (e.g. and was cognitively challenging.Thus, the unspecific CCT as an active control intervention was similar to the target CCT, and larger differences might be expected with less similar control conditions that do not include cognitive tasks and which more likely do not yield any cognitive effects; and even stronger effects can be expected for comparisons with passive control groups.However, the usage of an active control intervention challenging the target intervention can be considered a quality criterion of the present study [33].Thus, further studies with tailored programs for older adults and other control groups will have to test our hypothesis that such interventions are especially effective in this group.If effectiveness of (tailored) CCT for memory, visuospatial functions as well as executive functions (the latter of which was not found to be significantly trained with CCT in the meta-analysis conducted by [11] in healthy older adults will be further confirmed-all functions which are vulnerable in aging-this intervention form which is easy to administer and usually not cost-intensive would be a promising approach to support healthy cognitive aging.
The fact that low cognitive baseline performance was predictive for gains in various cognitive domains including verbal and non-verbal memory, executive functions, processing speed and visuospatial functioning is consistent with most, though not all [19], previous studies [17] [20] [21].It has already been discussed that the lower range of a high-functioning sample can be improved more [20], and the neuropsychological scores of the overall group which are considerably below the maximum scores rule out ceiling effects even in the higher range of the sample.Therefore, CT seems to be suitable to strengthen cognitive domains that are lower or "weaker" in healthy individuals.On the other hand, individuals with high performance may need more challenging training programs to even maximise their functions.This aspect merits further investigation, and predictor analysis will be useful here and will help to tailor CT programs to specific target groups.
Our result that also lower SCC is predictive for cognitive gains in relevant domains (verbal memory, verbal fluency, and attention), is intriguing.SCC is regarded as a predictor for dementia development [63] and is associated with Alzheimer typical biomarkers [64].Studies which examine the extent to which non-pharmacological interventions such as CT are effective to enhance cognition in individuals with SCC [65] and whether or not neural plasticity is already reduced compared to healthy older adults without SCC [66] have just begun.
However, to the best of our knowledge, SCC has not been examined as a predictor of CT gains yet.Our results point to the possibility that SCC may be associated with a reduced cognitive plasticity and that individuals with SCC have less gains in various cognitive domains after multidomain CCT.Future studies with larger sample sizes will have to replicate and elaborate this notion.
Consistent with our hypothesis, motivation and self-efficacy were identified as positive predictors of training success in verbal and figural memory, respective-Health ly.Metacognitive and motivational measures were already identified as predictors for gains in objective memory measures in a strategic memory training in older adults [37].The meta-analytic path analysis of [38] further underscores the importance and close interaction between self-efficacy, motivation and training success.In conclusion, individual characteristics and metacognitive attitudes should be taken into consideration when implementing a training program such as our CCT.Therefore, instructors and supervisors would do well to leverage self-efficacy and motivation at the beginning of the training period, for example by persuading trainees that they can succeed and profit from the training and presenting them with vicarious experiences [67].
The other results of our predictor analysis are less clear-cut.Regarding depression, the result that higher depression scores predicted better outcome in verbal memory and speed of processing was unexpected, as depression is regarded to reduce brain and cognitive plasticity [25] and, for example, has also been demonstrated to mediate the relationship between (low) cognitive scores and cognitive gains induced by a cognitive stimulation program in dementia patients [26].However, it should be noted that clinically relevant depression scores and even mild symptoms were an exclusion criterion for our study sample, so that symptoms if present were minimal.One hypothesis for our findings could be that individuals with mild symptoms of depression may have been particularly motivated by the program and the attention they received by the experimenter during the study.This might have resulted in a reduction of depressive symptoms during the training period, leading to better test performance post training.Unfortunately, depression scores were not assessed after the training, so that this notion remains speculative and needs further investigation.
Regarding the influence of sociodemographic factors, the fact that female gender predicts gains in memory corroborates previous findings that women profit more from CT programs especially in the memory domain, although [34] found effects for verbal, not figural memory which fit better to the hypothesis of a "gender-specific cognitive reserve" [68], which assumes larger plasticity in verbal episodic memory in women.Thus, our finding that gains in non-verbal long-term memory, attention and set-shifting was predicted by female gender needs further investigation.The fact that higher age was a negative predictor for gains in (non-verbal) memory and attention is concordant with the notion that higher age is regarded to be associated with less brain and cognitive plasticity [25] and with other findings that younger age predict better training outcome of cognitive or memory training [69] [70] [71] [72], although conflicting data also exist [73] [74].
The inconsistent results regarding technology commitment as a predictor for gains in figural memory, but losses in executive functions and speed of processing are hard to interpret-even more so when the results of [29]  training.This topic needs further investigation.
Some limitations have to be kept in mind when interpreting the results.First, the small sample size in our pilot study limited the power to detect interaction effects between the two training groups and generally limit clear conclusions.
Thus, further studies with larger samples are needed and might be able to dem- older adults [4], which further limit the generalizability of our findings.As the ultimate goal is to prevent cognitive decline long-term in this group of people, this is an important aspect for further studies.Finally, as a more general remark, a recent review on the effectiveness of cognitive trainings [33] challenges the assumption of transfer effects to distantly related tasks and daily life cognitive performance, and even to closely related tasks.With our study, we are not able to make any conclusions whether or not effects on cognitive measures are also related to effects in every life.Therefore, measuring these transfer effects is a central aspect that should be faced in future studies [33].Health The specific strength of our study is that, to our knowledge, this is the first RCT with CCT that was specifically developed for healthy older adults.Such programs which contain both training elements for those domains that are particularly vulnerable in higher age and psychoeducational elements promoting a lifestyle supporting healthy cognitive aging seem very important in an aging society with an increasing prevalence of dementia.A further strength that planning and reporting followed the recommendations of the CONSORT Statement for RCT [41] [42] [43].

Conclusion
In conclusion, although our hypothesis that a CCT especially tailored to the profile of healthy older adults would lead to better cognitive outcome compared to a non-tailored CCT in terms of significant interaction effects could not be verified, benefits in more cognitive domains in within-comparisons for the CCT indicate that this type of intervention may be a fruitful approach to support older adults in sustaining their cognitive level.More studies are needed to corroborate this notion and also to further specify which individuals have a high probability to benefit from such programs.SCC as a possible predictor for future cognitive decline will have to be considered in this research.Exercise "Daily plan" (Example in Figure S2) Mainly trained domain: executive functions At the beginning of the task a daily calendar with time categories is presented.
On the right side, there are different activities displayed that should be classified into the daily schedule following specific rules.The classification of activities not only depends on a predefined time schedule but also on an order of succession of activities that needs to be considered.Difficulty variations result from the following conditions: Individual timeslots in the schedule are blocked, while others may be occupied twice, appointments are subject to a fixed duration, appointments are only shown once and then must be kept in mind/remembered until they are processed.There are 14 difficulty variations/levels for this exercise.Health Exercise "Lateral thinking" (Example in Figure S3) Mainly trained domain: Complex attention The exercise is composed of varying stimuli in different colors and shapes moving from the left to the right side of the screen.If a stimuli fits the displayed rule/criteria (color and shape) it has to be selected before reaching a red marked area on the right side of the screen.At higher levels, attention must also be given to the patterns of the stimuli.Difficulty variations result from the following conditions: Density of stimuli, number (and combination) of features that have to be considered before selection, number and speed of the change of rules/criteria, presentation speed of the stimuli.There are 14 difficulty variations for this exercise.
Exercise "City map" (Example in Figure S4) Mainly trained domain: Spatial cognition The task is composed of two parts: The first part of the task consists in the completion of a map by inserting matching parts.The cards have to be selected and aligned in order to fit the layout of a street map.In the second part it is required to determine the shortest possible distance from a certain starting point, Figure S3.Exercise "Lateral thinking".In the left part of the screen, criteria are displayed in German: "Muster gleich" for "same pattern" and "Farbegleich" for "same color".Exercise "Word fluency".First example: sports ("Sportarten" in German) and the first three letters of the German word for diving ("tauchen").Second example: Blank crossword puzzle including the first word "Erde", which is German for "earth".Health through specification of the categories (e.g.insect instead of animal).During the next task, the user has to contrive words matching a specific category, the number of clues given decreases alongside the increase of task difficulty.The third and most complex exercise is a blank field of a crossword puzzle in which the user has to fill in all cases with a term without contradictions.Each task has a specific amount of letters that the participant is required to use.Task difficulty increases by number of the blank fields to fill and number of compulsory letters.
Exercises: Control intervention "Tintenklex" Example: Exercise "Word rain" (Example in Figure S6) Mainly trained domain: Spelling, speed of processing The word that is initially displayed is to be composed with letters, moving from the top of the screen to the bottom.The user solves the exercise by moving each letter into the right column before it reaches the bottom.The user has the possibility to adapt the speed.
Exercise: "Griddle" (Example in Figure S7) Mainly trained domain: Reading/attention Words which can be displayed diagonally, vertically and horizontally, forwards and backwards have to be found.
Exercise: "Klexsklick" (Example in Figure S8) Mainly trained domain: Visuo-spatial processing, reading At the right eight words are displayed.In the middle of the screen is a grid of 5 × 6 fields, and one field after the other is revealed until the user recognizes the symbol and clicks the field with the matching word.
Exercise: "Labyrinth" (Example in Figure S9) Mainly trained domain: Visual-motor processing The task is part of three exercise-units with a duration of ten minutes.The user has to solve a labyrinth by moving the red dot via the exit.
Exercise: "Different pictures" (Example in Figure S10) Mainly trained domain: visuospatial processing, attention The task is part of three exercise-units with a duration of ten minutes.The user has to compare the two images, find 10 differences and click on the appropriate places on one of the fields.
Figure S7.Exercise "Griddle".Examples: the German word "Elch" for "moose" in the upper left corner (forwards) and the German word "Giraffe" for "giraffe" in the middle of the word grid (vertically backwards).

E.
Kalbe et al.DOI: 10.4236/health.2018.10100322 Health and 100 years were recruited by the two study German centers Cologne and Vechta in the cities Cologne and Düsseldorf as well as Vechta, Bremen and Osnabrück, respectively.Individuals were recruited with flyers distributed via local pharmacies and through presentations at a Senior Service Center in Bremen.Participants did not receive monetary compensation.At first, individuals interested in the study were given standardized information either by telephone or, for those attending the presentations, on site.This information included the aims and the course of the study, inclusion and exclusion criteria, as well as a short description of the neuropsychological assessment.In case people were interested in participating in the study, personal information was collected and individuals were contacted later for the study.The inclusion criteria were: male and female individuals aged 55 or older, native German language or very good German language, normal or sufficiently corrected vision and hearing ability, motor capacities that ensured an unimpaired work on a computer and written informed consent to participate in the study.Exclusion criteria were: psychiatric or neurological disorder, cognitive impairment operationalized by the Montreal Cognitive Assessment (MoCA; cut-off score < 26 points; [45]), mild or clinically relevant depressive symptoms, operationalized by the German version of the Beck Depression Inventory 2 (BDI-II; cut-off score > 12; [46]), current drug abuse, and life-threatening illness.
frequency and intensity in order to warrant comparability.Both trainings consisted of eighteen lessons, with an average training period of 40 to 45 minutes each.Training sessions were performed three times per week over a period of six weeks.Every session contained two exercises (one 20 and one 10 minutes), a video (10 -15 minutes), and an "advice of the day".An overview on the structure of both interventions with examples of the exercises is given in Table different tasks were chosen training reading abilities, orthography, perception, concentration, visual-spatial perception as well as visual-motor skills.In each session, two tasks defined in the study protocol were trained 20 and 10 minutes, respectively.Videos of 10 -15 minutes that do not belong to the software were chosen from freely available sources from the internet and included topics like doing sports and exercise, nutrition, relaxation, stress and sleep.Participants watched one video per training lesson.The "advices of the day" (one per training lesson) focused on healthy aging and achievement of well-being in a more general way than in the experimental intervention.They did not focus on cognition.Verbal memory was assessed with the German verbal learning and memory test "Verbaler Lern-und Merkfähigkeitstest" (VLMT;[50]).VLMT 1-5 as the sum of words recalled in the five presentation runs was used as an indicator of intermediate verbal memory, and VLMT 7 as the amount of words recalled in the delayed condition and an indicator of verbal long term memory.For the assessment of figural memory, the Wechsler Memory Scale (German version) subtest Visual Reproduction (WMS VR;[51]) with a direct and delayed recall condition was used.The "Brief Test of Attention" (BTA;[52]) was performed to assess attention.As measures of speed of processing and set shifting as an executive function, the "Trail Making Tests A and B" (TMT A & B;[53] [54]) was performed.Executive functions were also operationalized with verbal letter and alternating semantic fluency tasks using the "Regensburger Wortflüssigkeits Test" (RWT;[55]).As recommended in the manual ([55], p.15), two different letter fluency tasks (P and M) and two alternating semantic fluency tasks (Sports-Fruits and Clothing-Flowers) were used for the two points of measurement at pre-and posttest.Furthermore, planning was assessed with the Key Search Task a subtest of the "Behavioural Assessment of Dysexecutive Syndrome" (BADS;[56]) test battery.Finally, for the assessment of visuospatial functioning, the subtest nine (visuospatial imagination) from the "Leistungsprüfsystem für 50 -90 jährige" (LPS 50+;[57]) was used.To assess Subjective Cognitive Concerns (SCC) a modified and extended version of the Subjective Memory Impairement Questionnaie (SMI-Q) proposed by[40] was used.The extended version of the SMI-Q contains "yes or no" questions concerning subjective cognitive impairment in five cognitive functions (memory; attention; language; executive functions; and visual-spatial skills), e.g.E. Kalbe et al.DOI: 10.4236/health.2018.10100327 Health Data assessment took place between March 2015 and October 2015.N = 54 persons were assessed for eligibility and randomized, of which a total of N = 49 engaged in the training conditions EG (n = 28) and CG (n = 21).During the study, n = 10 (20.4%) persons dropped out due to technical issues with the software and/or hardware issues and, consequently, deviations from the training protocol (EG: n = 3; CG: n = 1), discontinuation of the training (EG: n = 4; CG: n = 1), and non-compliance with the training protocol (CG: n = 1).Thus, a total of N = 39 datasets (EG: n = 21; CG: n = 18) were included for statistical analysis.The flow of participants during the study including information for the flow of participants per study center is presented in Figure 1.

Figure 1 .
Figure 1.Flow of participants trough the study.EG, Experimental condition; CG, Control condition.
onstrate the possible favour of cognitive training programs that are tailored to the specific needs of healthy older adults.Second, the fact that a pilot version of the training program was used in which the automatic adjustment of difficulty levels had not been implemented yet and in which individuals had to follow concrete instructions to choose the adequate training levels, might have influenced the results.Thus, future studies will have to show whether the expectation that fully adaptive programs which lead the individual smoothly through the program will be more effective.Third, our study was only single-blind.While individuals who performed the training were blind to the program, the person who administered the neuropsychological test battery was not.Thus, although test battery was fully standardized, a bias is possible and should clearly be avoided in further studies.Fourth, it is not possible to define the specific efficacious elements in a multidomain training program and, in our case, differentiate the impact of the direct training from the psychoeducational elements.One important difference between the groups was that the experimental group included a video that explained effects of cognitive training, which might have promoted expectations and corresponding effects.Therefore, for future studies and to extract the effects of the direct training, the psychoeducational elements should be designed more similar between groups, especially regarding contents that could directly influence the training outcome via a placebo effect as for example information on the effectiveness of cognitive trainings.Another control group that only differs in the nature of the psychoeducational elements could further disentangle effects of the specifically targeted cognitive training intervention from psychoeducational effects.With regard to the fact that effects of specific cognitive training tasks in the programs cannot be derived, it is important to note that multidomain trainings rather than interventions focused on only one domain have been recommended for inducing lasting improvement in cognition in healthy older adults[5] [75][76].Fifth, no long term follow-up was performed in this study, as in many other cognitive training studies conducted with healthy
Figure S5.Exercise "Word fluency".First example: sports ("Sportarten" in German) and the first three letters of the German word for diving ("tauchen").Second example: Blank crossword puzzle including the first word "Erde", which is German for "earth".
to ensure achievement of best model fit while taking into account each relevant predictor.The significance level was set at α = 0.10.Based on the current literature outlined above, the following predictors were integrated in the regression models: baseline level of the cognitive tests, depression score, cation (including both school and professional education), cognitive state (Mo-CA), depression score (BDI-II), technology commitment, self-efficacy (SWE) and motivation assessed via an average grade, and chi-square test for the comparison of sex distribution.As Kolmogorov-Smirnov test for normality indicated that neuropsychological data could not be assumed to be normally distributed (0.086 ≥ D ≤ 0.275 with pre-and/or posttest scores at p < 0.05 in 69.23% of the tests), non-parametric analyses were performed.To quantify the gains achieved within the experimental and control training, we calculated reliable change inposttest within the intervention groups were analyzed with Wilcoxon signed-rank tests.For all comparisons, the significance level was set at α = 0.05.Predictors of training success within the EG were estimated using backwards multiple regressions

Table 1 .
Baseline characteristics of the study sample.Cohen's d ≥ 0.8) was 67.1%.Outcome data for pre-and posttest are presented in Table2; RCIs and change scores of outcome variables are shown in Table3.The analysis of differences, based on RCIs or percentage change scores, between groups did not reveal significant results in any outcome Note.BDI-II = Beck Depression Inventory.MoCA = Montreal Cognitive Assessment.SWE = Self-Efficacy Expectation.E. Kalbe et al.DOI: 10.4236/health.2018.10100330 Health strong effects (

Table 2 .
Pre-/Post test comparison of test performance in both intervention groups.

Table 3 .
Comparison of pre-to posttest changes in outcome variables between both intervention groups.

Table S2 .
Backward multiple regression predicting cognitive improvement in experimental condition.