Meta-Analysis of Memory-Focused Training and Multidomain Interventions in Mild Cognitive Impairment

Abstract

Background:

Meta-analysis examining the efficacy of cognitive interventions on neuropsychological outcomes have suggested interventions that focus on memory may actually provide greater benefit against the cognitive declines associated with mild cognitive impairment (MCI). However, it remains unclear if memory-based training would be more effective at addressing the cognitive deficits associated with MCI than multidomain forms of intervention.

Objective:

A meta-analytic review and subgroup analysis was conducted to examine the effects of cognitive training in individuals diagnosed with MCI and to compare the efficacy of memory-based training with multidomain interventions.

Methods:

A total of 32 randomized controlled trials met inclusion criteria for the meta-analysis, which included 9 studies on memory-focused training and 17 studies on multidomain interventions.

Results:

We found significant, large effects for memory-focused training (Hedges’ g observed = 0.947; 95% CI [–1.668, 3.562]; Z = 2.517; p = 0.012) and significant, moderate effects for multidomain interventions (Hedges’ g observed = 0.420; 95% CI [–0.4491, 1.2891]; Z = 3.525; p < 0.001). A subgroup analysis found significant point estimates for memory-based forms of training and multidomain interventions, with memory-based forms of content yielding significantly greater summary effects than multidomain interventions (SMD Z = 2.162; p = 0.031, two-tailed; all outcomes). There was no difference between effect sizes when comparing outcomes limited to its respective domain.

Conclusion:

Overall, these findings suggest that, while both interventions were beneficial, treatment interventions that were strictly memory-based were more effective at improving cognition in individuals diagnosed with MCI than interventions that targeted multiple cognitive domains.

Keywords

Cognitive dysfunction cognitive remediation meta-analysis neuropsychology rehabilitation

INTRODUCTION

Mild cognitive impairment (MCI) is a heterogenous clinical condition that is often conceptualized as the transitional stage between healthy aging and dementia [1 –3]. MCI has also been shown to have a high conversion rate to Alzheimer’s disease [4 –6], making this an ideal stage for potential treatment interventions. Given the limited efficacy of current pharmacological agents as a treatment for MCI (for reviews and discussion, see [7, 8]), there is significant interest in identifying non-pharmacological interventions that could effectively slow down, or even halt, the rate of cognitive decline and prolong functional ability. Non-pharmacological interventions, such as cognitive training, either alone or in combination with pharmacological treatments, may yield greater cognitive benefit and be a more promising avenue for future research (for review, see [9]). To demonstrate its potential efficacy, cognitive training has been associated with increased activation in the hippocampus [10, 11], right inferior parietal lobe [12], frontoparietal network [10], occipito-temporal areas [13], and other various brain regions [14 –18]. However, it remains unclear which cognitive training interventions would be most effective at reducing the rate of cognitive decline and functional impairments.

One challenge in comparing cognitive training programs is the heterogeneity across interventions. For instance, prior research on cognitive training approaches for MCI have included cognitive stimulation, memory-based interventions (e.g., use of mnemonic strategies), multidomain approaches (i.e., targeting multiple cognitive domains such as attention, memory, and executive functioning), compensatory approaches, and multicomponent forms of intervention, which incorporates lifestyle changes in addition to cognitive training [9 , 19–24]. As such, a critical examination of these cognitive trainings is needed to determine which type of intervention may be the most effective and under what circumstances. Furthermore, it is possible that effective cognitive interventions will require greater specificity in order to successfully alter illness trajectory away from more serious cognitive decline [25 –27].

To address some of these questions, a recent meta-analysis investigated the efficacy of cognitive interventions on neuropsychological outcomes and found that both multidomain content and multicomponent approaches were beneficial among individuals with MCI [28]. Interestingly, interventions with memory-based content appeared to be associated with larger effect sizes on neuropsychological outcomes than interventions with multidomain forms of content. These results suggest that, while cognitive trainings aimed at targeting multiple domains may be helpful, interventions that focus exclusively on memory may actually provide greater benefit against the declines associated with MCI. Given that multidomain interventions often include memory-based strategies, it remains unclear why interventions focused exclusively on memory-based training would be more effective at addressing the cognitive deficits associated with MCI than multidomain forms of intervention.

One possible explanation stems from the field of cognitive neuroscience. Using the Scaffolding Theory of Aging and Cognition–revised (STAC-r) as a theoretical framework, it is possible that cognitive training may activate compensatory neural processes or “scaffolding” that prompts neuroplastic reorganization to meet task demands and support primary networks [29 –31]. In healthy aging, this reorganization appears to occur naturally as individuals increase in age. For instance, a longitudinal study found that ‘successful-agers’ showed greater bilateral activation in the hippocampus and prefrontal cortex relative to their peers [32]. These findings support the notion that bilateral activation and hemispheric recruitment can serve as an effective compensatory strategy to meet processing demands [33]. Conversely, in MCI, neurodegenerative processes alter the trajectory of normal structural and functional changes that occur with age and disrupt network activation, which results in neurocognitive failures. For example, MCI has been associated with declines in the integrity of the prefrontal and orbitofrontal cortex, cingulate gyrus, and medial temporal cortex [34, 35]. Moreover, there appears to be a disruption in the topological organization of large-scale neural networks in MCI resulting in an increase in modular function and loss of information transfer ([36]; also see [37]).

Given these changes, one possible approach for treating MCI may be to explicitly target compromised regions in order to promote activation of primary networks and decrease reliance from other areas [31]. For instance, improvements in memory retrieval after memory-focused training may reflect the activation of primary networks and the engagement of memory systems (i.e., direct effects). This specificity may lead to more efficient neurocognitive performance compared to training interventions that indiscriminately activate multiple structures and neural systems. This notion is theoretically consistent with the conceptualization of MCI as an early-onset, predominantly amnestic neurodegenerative disorder in which regions facilitating memory retrieval are affected before the decline of other cognitive and neural systems (i.e., the prodromal stage of Alzheimer’s disease).

On the other hand, cognitive training that targets multiple cognitive domains may provide greater neurocognitive benefits by recruiting support from alternate areas. In this instance, the lack of structural support, loss of white matter integrity, and compromised efficiency of primary network nodes (i.e., reductions in network strength and global efficiency) are unable to provide compensation to meet task demands thereby resulting in suboptimal neurocognitive performance [38]. As such, compensatory neurocognitive processes are needed to recruit other regions [38, 39] From this perspective, MCI is conceptualized as a distinct clinical entity that represents a more advanced stage of neurodegenerative compromise (i.e., an emergence of Alzheimer’s disease). In this instance, there is insufficient activation of primary areas and, while compensatory scaffolding may be present, there is a disruption of local specialization and global integration [37]. Poorer primary network activation relative to general network engagement would suggest that multidomain forms of intervention would offer greater utility in maximizing existing neurocognitive resources and provide greater neurocognitive benefit post training [40].

The present meta-analysis was conducted to investigate the efficacy of both memory-based and multidomain forms of intervention in individuals diagnosed with MCI. We sought to examine these two forms of content to determine: 1) the anticipated cognitive benefits that may occur after individuals with MCI participate in one of these cognitive training interventions, 2) the characteristics of interventions which demonstrate efficacy in MCI, 3) whether the phenotype of MCI (i.e., amnestic versus non-amnestic) has a moderating role on outcome performance, and 4) the insights that may be gained about the possible neural processes involved in MCI. We also sought to examine the effects of training by category (e.g., cognitive stimulation, compensatory, restorative, and multicomponent approaches) to determine what benefits these may have on cognition in MCI.

We hypothesized that domain-targeted training (i.e., memory interventions) would be associated with moderate to large summary effects (direct effects) on outcome measures and this would be significantly greater than multidomain interventions. Provisionally, based on the STAC-r model, we speculated that domain-targeted training would facilitate primary network engagement by creating new primary network paths to meet task demands without recruiting alternate networks [31]. Similarly, in the context of MCI as an early-onset neurocognitive condition, we expected domain-targeted training to have greater specificity than multidomain targeted interventions. Consistent with topological organization of large-scale neural networks reported in MCI, we speculated that neural networks retain sufficient interconnectivity at this prodromal stage to facilitate network interaction and transfer of information [36, 37]. As such, we anticipated memory-focused training to result in moderate to large summary effects on outcome measures.

Conversely, we expected multidomain forms of intervention would be associated with small to moderate effect sizes on cognitive measures. In this instance, we speculated that training would prompt recruitment of alternate networks to support primary network systems as a form of compensatory activation. However, small effect sizes would be anticipated as primary networks would be unable to manage the load due to loss of functionality, decreased efficiency, and decrease in specialization (e.g., dedifferentiation). Similarly, based on theories of structural connectivity, poorer cognitive performance would arise due to an increase in modular function and decrease in information transfer [36, 37]. From this perspective, small summary effects would suggest that MCI is a more advanced neurodegenerative condition with less network connectivity. As such, we anticipated multidomain interventions would be more diffuse and yield small to moderate summary effects on outcome measures.

Finally, the absence of a summary effect would suggest that individuals with MCI have lost sufficient plasticity to benefit from training and are unable to recruit alternative compensatory mechanisms to meet the required task demands.

METHODS

Search strategy

The search process and meta-analyses performed followed guidelines outlined by the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) [41] using a PICOS approach (Participants, Interventions, Comparators, Outcomes, and Study Design). Specifically, we used the same procedure outlined by Sherman and colleagues [28], except we expanded our search to include more recent studies. The current review focused on studies published from January 1, 1995 through June 1, 2019 that: 1) selected subjects based on established MCI criteria [2 , 42], 2) performed a randomized controlled trial (RCT) in an outpatient setting, 3) compared cognitive training versus controls (active or passive), and 4) reported outcomes based on objective neuropsychological measures. The definition used for cognitive intervention was any skill or strategy that aimed at improving mental processes of attention and concentration, speed of information processing, memory, or executive function, similar to the guidelines offered by Gates and Valenzuela [43]. We adopted the following terminology: 1) intervention as a broad-based idiom to refer to any effort that was employed, 2) cognitive stimulation, defined as nonspecific and leisurely forms of activities, 3) cognitive training to denote either compensatory or restorative forms of training, and 4) multicomponent forms of intervention, which refers to a combination of several cognitive-based approaches and lifestyle changes.

Inclusion criteria

Studies selected for inclusion in the meta-analysis were RCTs with a clearly defined sample of individuals that were diagnosed with MCI using well-established criteria [1–3 , 44] or an analogous definition using an algorithm of 1.5 standard deviations below the mean [45] on established neuropsychological instruments, such as the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) [46]. The start date of January 1, 1995 was chosen as a starting point since a cursory search for studies prior to this date did not yield any relevant research and it was believed that studies published prior to 1995 would not have recruited subjects according to the current MCI diagnostic criteria. Our review was conducted through OVID-MEDLINE search engine using a collection of source databases: MEDLINE-R, PubMed, Healthstar, Global Health, PSYCH-INFO, and Health and Psychological Instruments. The primary search parameters included 1) terms representative of an MCI diagnosis (mild cognitive impairment, MCI, pre-Alzheimer’s disease, early cognitive decline, early-onset Alzheimer’s disease, or preclinical Alzheimer’s disease), 2) a descriptor of the intervention or training conducted (intervention, training, stimulation, rehabilitation, or treatment), 3) terms that refer to cognitive functioning (cognition, thinking, or neuropsychology), 4) RCT, 5) limit to “1995-Current”, and 6) limit to human (see Fig. 1 for a flowchart of the search criteria). Each of the three authors (DSS, KD, and DR) conducted an independent search of the literature for the period from June 1, 2017 to June 1, 2019 with periodic confirmation of the eligibility criteria and progress made in study selection.

Fig. 1

Literature Review Flow Diagram.

Data extraction

The effects of training were evaluated through subjects’ performance on outcome measures, defined as scores obtained on standardized neuropsychological tests. Cognitive domains included mental status and general cognition, working memory and attention, speed of information processing, language, visual-spatial ability, memory (verbal and non-verbal), and executive functioning [47]. Means and standard deviations for participants’ performance on these neuropsychological tests were extracted from each study to calculate summary statistics, effect sizes (weighted and un-weighted), and 95% confidence intervals (CI). When means and standard deviations were not reported directly, p-values, t-values, F-values, or confidence interval data were extracted to calculate the mean and standard deviation statistics for intervention and control groups. If data was collected post-training, outcome data was extracted from the timepoint closest to the conclusion of training.

Statistical analysis

The overall summary effect size, forest plots, and individual effect sizes within specific cognitive domains were examined. Differences between means were calculated according to Hedges’ g metric. Interpretations of effect sizes used the established guidelines that a small effect size is 0.20 or smaller (range 0.00–0.20); moderate = 0.50 (range 0.30–0.70); and large = 0.80 (range ≥0.80; [48, 49]. Given the likely heterogeneity resulting from the variability of training approaches, range of outcome measures, and differing methodological procedures across studies, a random-effects model was assumed for all analyses [50]. For studies with more than one outcome measure, a combined outcome or ‘synthetic variable’ was computed by combining all test results reported from the study to produce a single mean difference, in accordance with the procedure recommended by Borenstein, Hedges, Higgins, and Rothstein [51]. A minimum of five studies were used as the criterion for analysis due to concerns of over-interpretation given the range of outcome measures administered and due to cautions performing meta-analyses with less than five studies [51]. Meta-analyses and subgroup comparisons were performed using Comprehensive Meta-Analysis software (CMA, Software Version 3.3.070). Additional comparisons between subgroups were conducted with a Z-test using standardized summary effect scores derived from CMA software. We also tested significance and calculated confidence intervals for Hedges’ g overall effect for meta-analytic trials that included ≤20 studies using the Hartung-Knapp-Sidik-Jonkman (HKSJ) method [52].

Several sensitivity analyses were performed to determine error and possible artificial influence introduced by variance estimates on Hedges’ g [53]. We examined the effects of between-study variance within the dataset calculating Hedges’ g at various levels of τ2 (0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 2.5, 5.0, 7.5, 10.0). We also computed between-study variance and values of Hedges’ g using the DerSimonian-Laird (DL) estimator [50] and sample size-based methods [54]. While we expected adequate performance of random effects model with the DL method [51 , 56], we examined any possible bias introduced by these weighting methods [53].

RESULTS

Study selection

The search strategy and selection process resulted in a total of 681 studies for consideration. From this group of publications, a total of 649 studies were excluded for various reasons, including a failure to use well-established diagnostic criteria for MCI, not meeting the criteria for a RCT, unclear or inadequate study design, non-cognitive interventions (e.g., diet, physical activity, social skills, etc.), non-standardized outcome variables (i.e., neuropsychological tests that have not been standardized), and incomplete/preliminary datasets (see Fig. 1 for a flowchart of the screening process and reasons for exclusion).

Study characteristics

A total of 32 studies met the criteria for inclusion in the meta-analysis (see Table 1 for a detailed overview of the included studies). A summary of age, education, mental status, number of participants, treatment duration, and span of involvement is presented in Table 2. The interventions employed across studies included cognitive stimulation = 1/32 (3.13%), restorative training = 9/32 (28.13%), compensatory training = 5/32 (15.63%), and multicomponent approaches = 17/32 (53.13%). The domains targeted by these interventions primarily included attention and working memory (6.25%), speed of information processing (6.25%), memory (31.25%), executive functioning (3.13%), and multidomain training (cognitive-based approaches that included interventions related to lifestyle and socialization, 53.13%). Trainings were conducted in group format (43.75%), one-on-one/dyad training (12.5%), and computer-based programs (43.75%). Memory-based interventions included errorless learning, spaced retrieval, method of loci, face-name associations, spaced-retrieval/lags, journaling, cue/cue-generation, and categorization. Multicomponent interventions trained participants in several cognitive domains, such as sustained attention, speed of information processing, visual-spatial functions, language, memory, logical reasoning, hierarchical organization, psychoeducation, and progressive muscle relaxation.

Table 1

Summary of findings: study characteristics, types of interventions, and targeted domains

Author	Study	Sample	Age	Duration	Format	Domain	Type	Interventions	Outcome
design	size	mean (SD)	targeted	measures
1.	Bahar-Fuchs et al. (2017) [78]	RCT, double blind, active control	32	74.8 (6.80)	2 sessions/day, 20–30 min per session (3 days/week, for 8–12 weeks)	Computer	Multidomain	Multicomponent/Lifestyle	Repeated practice on standardized, game-like computer tasks, psychoeducation, and a range of behavior-change techniques used to optimize engagement, adherence, and perseverance. Training program consisted of 33 tasks designed to train a broad range of cognitive abilities.	ACE-III, SydBat (Naming, Word Repetition, Word Comprehension, Semantic Association), L-Hermitte Board (Learning Trials, Delayed Recall), Logical Memory, RAVLT, RCFT, Verbal Fluency (FAS, Animals, Category Switching), DS-FWD, DS-BWD, Coding, TMT A &B, Olfaction
2.	Balietti et al. (2016) [57]	RCT, inactive control	70	75.7 (0.72)	10 sessions, 60 min per session (1×/week for 10 weeks)	Group	Memory	Multicomponent/Lifestyle	Errorless learning, spaced retrieval, organization, categorization, clustering, method of loci, visual imagery, and face-name associations.	DS-FWD, CSST, AMT, phonemic fluency, semantic fluency, prose recall, word-pair learning
3.	Barban et al. (2016) [14]	RCT, single blind, inactive control	106	74.4 (5.70)	24 sessions, 60 min per session (2×/week for 3 months)	Computer	Multidomain	Multicomponent/Lifestyle	Software designed in multicomponent cognitive exercises/training in memory, logical reasoning, orientation, language, constructional praxis, questions concerning autobiographical data (SOCIABLE).	RAVLT, ROCT, TMT A &B, Phonemic Fluency, MMSE
4.	Barnes et al. (2009) [79]	RCT, single blind, active control	47	74.0 (—) Range = 54–91	100 min/day, 5 days/week	Computer	Speed	Restorative	Restorative training strategies targeting auditory processing speed and working memory.	1° - RBANS
										2° - CVLT-II, COWAT, BNT, California TMT, Design Fluency, Spatial Span, GDS
5.	Buschert et al. (2011) [80]	RCT, single blind, active control	43	71.2 (7.00)	11 sessions minimum, 20 “units”, 120 min/week	Group	Multidomain	Multicomponent/Lifestyle	Multicomponent cognitive intervention including memory, cognitive stimulation, face-name associations, errorless learning, principles of meta-cognition, and education.	ADAS-cog, MMSE, TMT B, RBANS, story recall, MADRS, QoL-AD
6.	Carretti et al. (2013) [81]	RCT, active control	20	71.8 (2.20)	5 sessions, 90 min per session	Individual	Working Memory	Restorative	Individual-based, one-on-one sessions using restorative strategies focused on working memory techniques.	NPE, Vocab, CWMS, DS-FWD, DS-BWD, Dot matrix, List recall, Pattern comparison, Cattell test
7.	Djabelkhir et al. (2017) [82]	RCT, single blind, active control	19	75.2 (6.40)	12 sessions, 90 min per session (1×/week for 3 months)	Group &Computer	Multidomain	Multicomponent/Lifestyle	Computerized cognitive stimulation (CCS) program designed to stimulate several cognitive domains with computerized cognitive exercises (described as “playful and ecological”) and social interactions among participants.	MMSE, VST, RL/RI-16 Free/Cued Recall, TMT A &B, DS-BWD, Verbal Fluency (Phonemic, Categorical)
8.	Donnezan et al. (2018) [83]	RCT, inactive control	69	76.3 (1.50)	24 sessions, 60 min per session (2×/week for 12 weeks)	Computer	Multidomain	Cognitive Stimulation	Thirty-three preselected games aimed to stimulate attention and executive functions; namely, working memory, mental flexibility, inhibition, reasoning and updating by using words, numbers, colors, shapes, and logical exercises.	Matrix Reasoning, Stroop Color-Word, DS-FWD, DS-BWD
9.	Emsaki et al. (2017) [84]	RCT, inactive control	20	63.1 (5.56)	5 sessions, 80 min per session (1×/week)	Group	Memory	Compensatory	Memory Specificity Training (MEST): Consists of a psychoeducational overview of memory functioning followed by training including in-session practice, homework, and journaling on differentiating between autobiographical memory recall with positive, negative, and neutral words with special attention paid to spatio-temporal and contextual details, and sensory-perceptual details for the memories.	AMT (Autobiographical Memory Test), PRMQ, WMS (Spatial Span, Logical Memory, Verbal Paired Associates I &II, Designs I &II, Visual Reproduction I &II)
10.	Fiatarone Singh et al. (2014) [85]	RCT, double blind, active control	100	70.1 (6.70)	4 45-minute exercises (initial/group setting); then 45 min per session (2×/week for 24 weeks)	Computer	Multidomain	Compensatory	Computer-based training in memory, executive functions, attention, and speed of information processing (COGPACK).	ADAS-Cog, MMSE, GP-Cog, CDR, Matrices, Similarities, TMT A &B, LM I &II, BVMT-R; SDMT; Semantic Fluency, COWAT, MARS-MF
11.	Finn and McDonald (2011) [86]	RCT, single blind, inactive control	16	69.0 (7.69)	30 sessions, on average 11.43 weeks/completion	Computer	Multidomain	Restorative	Computer-based training program involving attention, processing speed, visual memory, cognitive control (LUMOSITY).	1° - CANTAB, RVP A, PAL, IED2° - MFQ, DASS-21
12.	Finn and McDonald (2015) [87]	RCT, inactive control	24	72.8 (5.70)	6 sessions over several weeks (plus one practice session).	Computer	Memory	Restorative	Computer-based training program using repetition-lag training.	1° - VPA I &II2° - WMS-IV –SS, D-KEFS N-S, D-KEFS N-LS, CFQ, DASS-21
13.	Förster et al. (2011) [88]	RCT, single blind, active control	24	74.5 (8.60)	26 sessions, ∼120 min per session (1×/week for 6 months)	Group	Multidomain	Multicomponent/Lifestyle	Multicomponent cognitive interventions focused on global cognition, mood, and quality of life.	ADAS-cog, MMSE, FDG-PET scan
14.	Gagnon and Belleville (2012) [89]	RCT, double blind, active control	24	68.42 (6.04)	6 sessions, 60 min per session (3×/week for 2 weeks)	Computer	Divided Attention	Restorative	Restorative training to increase attentional control and meta-cognition.	1° - (modified) Dual task visual detection/ classical digit span2° - TEA, TMT A &B, DAQ, WBS
15.	Giuli et al. (2016) [90]	RCT, inactive control	97	76.0 (6.30)	10 sessions, 45 min per session (1×/week)	Individual	Multidomain	Multicomponent/ Lifestyle	Strategies in orientation, memory, categorization and clustering, psychological support, psychoeducation, as well as education regarding healthy lifestyles.	MMSE, CSST, DS-FWD, DS-BWD, Prose memory, VPA, AMT, Semantic Fluency, Phonemic Fluency, CDR, GDS-30, PSS, ADL, IADL, MAC-Q, Questionnaire of Confidence
16.	Greenaway et al. (2012) [59]	RCT, inactive control	40	72.7 (6.90)	12 sessions, 60 min per session (2×/week for 6 weeks)	Dyad-based (participant and partner)	Memory	Compensatory	Compensatory training memory strategies (Memory Support System).	DRS-2, MMSE, ECog, QoL-AD, CBQ, Chronic Disease Self-Efficacy Scale, Adherence Assessment
17.	Hampstead et al. (2012) [91]	RCT, single blind, inactive control	28	73.5 (10.10)	3 sessions, 60–90 min per session, over a 2-week period	Individual	Memory	Compensatory	Compensatory training in object-location built on face-name associations and mnemonic strategies.	MMSE, RBANS, TMT A &B, GDS, FAQ, ILV, F-N, fMRI imaging
18.	Herrera et al. (2012) [58]	RCT, single blind, active control	22	75.1 (1.97)	24 sessions, 60 min per session (2×/week for 12 weeks)	Computer	Memory	Restorative	Memory &attention.	DS-FWD, DS-BWD, DMS48, 12-word recall (BEM-144), 16-FR/CR test, MMSE, Doors/ People memory
19.	Jean et al. (2010) [92]	RCT, single blind, active control	22	68.6 (9.16)	6 sessions, 45 min per session (2×/week for 3 weeks)	Individual	Memory	Restorative	Face-name associations using errorless learning and spaced retrieval.	1° - Training measure (free recall and cued recall)
										2° - CVLT-II, DRS-2, F-N, MMSE, RBMT, MMQ, SES
20.	Jeong et al. (2016) [60]	RCT, double blind, inactive control	147	70.8 (6.90)	24 sessions, 90 min per session (2×/week for 12 weeks)	Group	Multidomain	Multicomponent/ Lifestyle	Multicomponent cognitive training (memory, attention, executive functions, language, reality orientation, visual-spatial functions), activities to improve ADLs, knowledge for health and daily life, reminiscence therapy and discussion.	1° - ADAS-Cog2° - MMSE, Digit Symbol Coding, Stroop, Animal Fluency, COWAT, SRT, DS-FWD, DS-BWD, CDR, GDS-15, NPI, Bayer ADL, PRMQ, AD8, PMT, MMT-Strategy, Composite scores
21.	Lam et al. (2015) [93]	RCT, double blind, inactive control	276	74.4 (6.40)	48 sessions, 60 min per session (3×/week for 4 months [Time 1], 12 months total)	Group, Homework assignments	Multidomain	Multicomponent/ Lifestyle	Multicomponent or lifestyle activities categorized into cognitive, physical, social, and recreational activities (33 in total). Cognitive group attended cognitively-demanding activities (i.e., reading, discussing newspapers, etc.).	1° - CDR2° - ADAS-Cog, CMMSE, List Learning (delayed recall), Digit Span, Visual Span, CVFT, CTMT, MIC, CSDD, CDAD
22.	Mowszowski et al. (2014) [94]	RCT, double blind, inactive control	40	74.4 (6.40)	14 sessions, 120 min per session (2×/week for 7 weeks)	Computer	Multidomain	Multicomponent/Lifestyle	Multicomponent, computer-based cognition training (not specified) and psychoeducation.	1° - EEG2° - WTAR, MMSE, RAVLT, FAS, Semantic Fluency (Animals), WAIS-III-DS (total), TMT-B
23.	Mudar et al. (2017) [95]	RCT, active control	50	75.65 (8.51)	8 sessions, 60 min per session (2×/week for 4 weeks)	Group	Executive Functions (Reasoning &Problem Solving)	Restorative	Gist reasoning training focused on the ability to synthesize and derive diverse interpretations from complex information (e.g., written documents and verbal information). The primary objectives were to promote utilization of three core strategies: strategic attention, integrated reasoning, and innovation.	WAIS-III Similarities, List Learning, Text Processing, Complex Abstraction, WMS-III LM, Meta-Memory Questionnaire
24.	Olchik et al. (2013) [96]	RCT, single blind, inactive control	30	70.3 (4.30)	8 sessions, 90 min per session (2×/week for 4 weeks)	Group	Memory	Multicomponent/ Lifestyle	Multicomponent training focused on memory with each session beginning with explanations/ education followed by training in a memorization target task/ strategy (active attention, categorization, association, or visual imagery), and exercises to practice the strategy.	MMSE, Lawton IADL, CRD, Semantic Fluency (Animals), COWAT (FAS), RAVLT, RBMT
25.	Park et al. (2018) [97]	RCT, single blind, active control	78	67.64 (4.55)	30 sessions, 30 min per session (3×/week for 10 weeks)	Computer	Multidomain	Compensatory	Cognition-specific computer training consisting of various training programs including attention, memory, and visual spatial abilities. The difficulty level was adjusted during the sessions depending on each subject’s performance. Training targeted attention (sustained and selective attention), memory (working memory), and visual spatial ability (visual discrimination and visual constancy).	DS-FWD, DS-BWD, RAVLT, TMT-B, RCFT, MTCF
26.	Polito et al. (2015) [98]	RCT, single blind, inactive control	44	74.0 (1.40)	10 sessions, 100 min per session (2×/week for 5 weeks)	Group	Multidomain	Multicomponent/Lifestyle	Multicomponent training using body awakening, reality orientation, and multiple compensatory cognitive exercises. Exercises were designed to stimulate attention (auditory and visual), executive reasoning, language (fluency), semantic memory, visual perception, encoding, information storage, nonverbal learning and executive problem solving.	MMSE, MOCA, CSST
27.	Rapp et al. (2002) [99]	RCT, single blind, inactive control	19	73.3 (6.61)	6 sessions, 120 min per session (1×/week for 6 weeks)	Group	Memory	Multicomponent/ Lifestyle	Multicomponent training using education, relaxation training, memory skills training, and cognitive-restructuring.	CERAD neuropsychological battery, MMSE, Face-Name, MFQ, Memory Controllability Inventory, POMS
28.	Rojas et al. (2013) [100]	RCT, inactive control	30	72.0 (14.29)	52 sessions, 120 min per session (2×/week for 6 months)	Group	Multidomain	Multicomponent/Lifestyle	Multicomponent intervention program including cognitive-stimulation, cognitive training, and education.	1° - MMSE, CDR2° - SMB, SF, BNT, PhF, Verbal Fluency, WASI (Similarities, Matrix Reasoning, Block Design), TMT A &B, WAIS-III DS-FWD, WAIS-III DS-BWD, QoLQ, NPI, ADL Scale
29.	Schmitter-Edgecombe &Dyck (2014) [101]	RCT, single blind, inactive control	46	72.96 (7.05)	20 sessions, 120 min per session, (2×/week for 10 weeks)	Group	Memory	Multicomponent/ Lifestyle	Multicomponent training including workbook lessons and education workshop.	WTAR, TICS, MMAA, EFPT, ADL-PI, RBANS, QOL-AD, CSE, GDS, RBMT-II
30.	Tsolaki et al. (2011) [102]	RCT, inactive control	201	68.5 (6.99)	60 sessions, 90 min per session, (3×/week for 5 months)	Group	Multidomain	Multicomponent/Lifestyle	Multicomponent training in cognitive strategies, cognitive stimulation, and psycho-therapeutic techniques.	HVLT, RAVLT, RBMT, Digit Symbol, FUCAS, BNT, FRSSD, MMSE
31.	Valdes et al. (2012) [103]	RCT, single blind, inactive control	195	76.95 (6.53)	10 sessions, 60 min per session (5-week duration)	Computer	Speed	Restorative	Restorative training using a standardized set of visual attention tasks designed to improve speed of information processing (SOPT).	RBMT, RAVLT, LS, WS, Computerized UFOV
32.	Vidovich et al. (2015) [45]	RCT, single blind, active control	160	75.1 (6.10)	10 sessions, 90 min per session (2×/week for 5 weeks)	Group	Multidomain	Multicomponent/Lifestyle	Multicomponent training in cognition, activities related to enhance attention, memory, and executive functions as well as methods to adapt these to everyday life.	CAMCOG-R, CVLT-II, WAIS-III DS, WAIS-III DS-BWD, WAIS-III DS Total, WAIS-III SS, TMT A &B, COWAT

Please see Supplementary Table 1 for a list of abbreviations.

Table 2

Demographic Information (Total Included Studies), N = 32

Variable	Mean (SD)
Age	72.36 (6.32)
Education (y)	11.91 (3.21)
Mental status (mean score on MoCA/MMSE)	27.12 (1.67)
Number of participants per study	30.75 (32.43)
Treatment duration (total hours)	24.94 (22.83)
Span of involvement (weeks)	10.11 (6.87)

Meta-analyses: Cognitive interventions –benefits of training

Several meta-analytic procedures were performed to examine the effects of training on outcome measures (overall, by domain, by training type, and by domain targeted). Subgroup analyses were also conducted to examine the potential influence of moderator variables, such as intervention content and MCI subtype, on cognition. The overall summary effect of cognitive interventions on outcome measures demonstrated a significant, moderate effect (Hedges’ g observed = 0.510; 95% CI [–0.9206, 1.9406]; Z = 3.734; p < 0.001). As anticipated, heterogeneity was large (Q = 230.897; df = 31; p < 0.001; I ² = 86.574%; τ ² = 0.472). Visual inspection of the funnel plot was somewhat asymmetrical with five outliers, although there were no adjustments for publication bias after calculation of Duval and Tweedie’s trim-and-fill method. Egger’s regression intercept was suggestive of small-study effects (Intercept = 3.060; t = 3.046; p = 0.005; two-tailed). Examining memory outcomes specifically, the summary effect of all cognitive interventions on memory outcomes also generated a significant, large effect (Hedges’ g observed = 0.748; 95% CI [–0.5681, 2.0641]; Z = 4.991; p < 0.001). Heterogeneity was significant and large (Q = 117.489; df = 22; p < 0.001; I ² = 81.275%; τ ² = 0.378). Visual inspection of the funnel plot was asymmetrical with five outliers and there were no adjustments for publication bias after calculation of Duval and Tweedie’s trim-and-fill method. Egger’s regression intercept was significant (Intercept = 2.531; t = 2.601; p = 0.017; two-tailed). A forest plot of the effects and overall summary on neurocognitive outcomes is presented in Fig. 2.

Fig. 2

Meta-analysis of interventions on outcome measures. Test for heterogeneity Q = 230.897; df = 31; p < 0.001; I² ⁼86.574; τ ² = 0.472.

In terms of intervention type, nine studies reported outcome data for restorative forms of training, yielding a moderate, non-significant effect (Hedges’ g observed = 0.663; 95% CI [–2.9920, 4.3180]; Z = 1.307; p = 0.191). Calculating the HKSJ adjustment to account for a small number of studies was also non-significant (HKSJ point estimate adjustment SMD = 0.423; t = 1.63; df = 8; p = 0.142). Indicators of heterogeneity were significant (Q = 142.217; df = 8; p < 0.001; I ² = 94.375%; τ ² = 2.132) and visual inspection of the funnel plot was symmetrical with multiple outliers. No studies were trimmed with Duval and Tweedie’s trim-and-fill method. Egger’s regression intercept was significant (Intercept = 8.600; t = 5.229; p = 0.001; two-tailed). These results suggest that restorative forms of training are not associated with improvements on neuropsychological outcome measures.

Five studies reported outcome data for compensatory forms of intervention, yielding a moderate, significant effect (Hedges’ g observed = 0.554; 95% CI [–0.0573, 1.1653]; Z = 3.589; p < 0.001). Calculating the HKSJ adjustment to account for the small number of studies was also significant (HKSJ point estimate adjustment SMD = 0.566; t = 3.468; df = 4; p = 0.026). Indicators of heterogeneity were unremarkable (Q = 4.468; df = 4; p = 0.346; I ² = 10.472%; τ ² = 0.013) and visual inspection of the funnel plot was somewhat asymmetrical. Duval and Tweedie’s trim-and-fill method trimmed two studies with a moderate adjusted point estimate (Adjusted point estimate = 0.428; Q = 9.010). Egger’s regression intercept was not significant (Intercept = 2.220; t = 1.380; p = 0.261; two-tailed). A significant, moderate adjusted point estimate with nominal markers of heterogeneity would indicate that compensatory forms of training are associated with improvements in neurocognitive performance.

Seventeen studies reported outcome data for multicomponent forms of intervention yielding a moderate, significant effect (Hedges’ g observed = 0.420; 95% CI [–0.4491, 1.2891]; Z = 3.525; p < 0.001). The HKSJ calculation was also significant (HKSJ point estimate adjustment SMD = 0.428; t = 3.026; df = 17; p = 0.008). Heterogeneity indicators were significant (Q = 58.363; df = 16; p < 0.001; I ² = 72.585%; τ ² = 0.152). With respect to possible publication bias, visual inspection of the funnel plot was asymmetrical with one prominent outlier [57]. No studies were trimmed with Duval and Tweedie’s trim-and-fill method and Egger’s regression intercept was not significant (Intercept = 1.773; t = 1.708; p = 0.108; two-tailed). Significant point estimates and relatively low values on bias indicators would indicate that multicomponent training is likely to have a positive benefit on cognition and yield improved performance on outcome measures, in the context of some heterogeneity.

There was an insufficient number of studies to perform a meta-analysis on studies that applied cognitive stimulation forms of training (k = 1).

Meta-analyses: Cognitive interventions –memory training and multidomain interventions

With respect to targeted domain and intervention content, nine studies reported outcomes on memory training yielding a large and significant summary effect on memory measures (Hedges’ g observed = 0.947; 95% CI [–1.668, 3.562]; Z = 2.517; p = 0.012). HKSJ adjustment was also significant (HKSJ point estimate adjustment SMD = 0.975; t = 2.443; df = 8; p = 0.040). Heterogeneity across studies was significant (Q = 60.936; df = 8; p < 0.001; I² = 86.871%; τ ² = 1.082). Visual inspection of the funnel plot was symmetrical with two outliers [57, 58]. There were no adjustments after calculation of Duval and Tweedie’s trim-and-fill method and Egger’s regression intercept was not indicative of small-study effects (Intercept = 2.129; t = 0.477; p = 0.648; two-tailed). These results suggest that, in the context of some variability, memory training is likely to have a positive benefit on memory performance among individuals with MCI. A forest plot of the effects and overall summary of memory strategies on memory outcomes is shown in Fig. 3. It should be noted that while there were eleven studies using memory-focused interventions, two were removed from the analysis. One study did not administer outcomes specifically assessing memory [59] and another study used a memory composite score which would introduce unrelated, non-delay data into the analysis [60].

Fig. 3

Meta-analysis of memory-focused interventions. HKSJ point estimate adjustment SMD = 0.975; t = 2.443; p = 0.040. Test for heterogeneity Q = 60.936; df = 8; p < 0.001; I² ⁼86.871; τ ² = 1.082.

There were seventeen studies that applied multidomain forms of intervention (by definition, these were the same studies that used a multicomponent approach). As noted above, this yielded a moderate, significant effect (Hedges’ g observed = 0.420; 95% CI [–0.4491, 1.2891]; Z = 3.525; p < 0.001). Significant point estimates and relatively low values on bias indicators suggest that multidomain training has a positive benefit on cognition and is likely to improve performance on outcome measures, in the context of some heterogeneity. A forest plot of the effects and overall summary of memory strategies on memory outcomes is illustrated in Fig. 4.

Fig. 4

Meta-analysis of multicomponent interventions. HKSJ point estimate adjustment SMD = 0.428; t = 3.026; p = 0.008. Test for heterogeneity Q = 58.363; df = 16; p < 0.001; I² ⁼72.585; τ ² = 0.152.

Subgroup analysis: Comparing memory and multidomain interventions

With regards to the possible influence of moderator variables on training outcomes, a subgroup analysis examining intervention content was significant (Total Between Q = 21.896; df = 4; p < 0.001; all outcomes). Individual point estimates were significant for memory-based forms of training (Hedges’ g = 0.950; SE = 0.300; Z = 3.162; p = 0.002) and multidomain content (Hedges’ g = 0.285; SE = 0.065; Z = 4.359; p < 0.001). Further analysis determined that memory-focused interventions were more effective than training programs that targeted multiple domains. Specifically, summary effects associated with memory-based training was significantly higher than summary effects associated with multidomain content (SMD Z = 2.162; p = 0.031; all outcomes). Additional analyses comparing memory-focused interventions with multidomain training for domain-related outcomes were unremarkable (SMD Z = 1.762; p = 0.078; memory training –memory measures versus multiple domain –all outcomes; SMD Z = 1.383; p = 0.167; memory –verbal memory only versus multiple domain –all outcomes). There was also no difference between training types when limiting outcomes solely to memory measures (SMD Z = 1.338; p = 0.181; memory –memory only versus multiple domain –memory only).

Sensitivity analyses of between-study variance were suggestive of low inflation on Hedges’ g using the DL method at the effect size observed, consistent with nominal bias [53, 56]. Analyses limited to studies applying memory interventions alone (all outcomes), revealed a similar pattern of between-study variance for inverse-variance compared with sample-size methods, although the magnitude of effect was slightly smaller for sample-size methods. Sensitivity analyses would suggest that, while there may be some artificial inflation on Hedges’ g, we anticipate this to be small given the dataset and effects observed, although the dataset would be improved by greater uniformity of outcomes and increased sample sizes.

These findings suggest that, while both forms of training interventions are beneficial, interventions that are memory-based appear to be more effective than trainings that target multiple domains. Moreover, given the summary effects for memory-based training included all cognitive outcomes, it is possible that memory-based training could have transfer effects and generalize to other domains. Non-significant findings with domain-specific outcomes are apt to reflect the limited number of studies in each cell. It should be noted that results regarding subgroup analysis would be regarded as observational only and not the equivalent of conducting a well-designed comparison of each training approach. In addition, the number of studies in other cognitive areas was insufficient to perform an examination across all cognitive domains.

Qualitatively, the memory-based interventions used in these studies included errorless learning, spaced retrieval, categorization and clustering, chunking, method of loci, associations, enhanced encoding (i.e., cue generation, mnemonics, instrumental aids [memory notebook]), and various methods to facilitate metacognitive awareness (e.g., psychoeducation, memory-related beliefs). Multidomain interventions included training in multiple cognitive domains (e.g., orientation, attention, processing speed, language, fluency, visual-spatial skills, memory, reasoning, executive functions) as well as psychoeducation, metacognition, reminiscence therapy, psychomotor/recreational tasks, social activity, lifestyle changes, affect, and other non-cognitive aspects (e.g., mood, stress reduction, quality of life, engagement).

Other moderator variables: MCI subtype, intervention type

In terms of MCI diagnostic category (aMCI, MCI-MD, MCI [all subtypes]), a subgroup analysis was not significant (Total Between Q = 1.623; df = 2; p = 0.444), suggesting that cognitive training is apt to have similar benefits across MCI type. Comparing memory-based training for aMCI (memory outcomes) versus multidomain training for all-MCI studies (all outcomes) was not significant (SMD Z = 1.181; p = 0.238; aMCI memory training –memory measures; all-MCI multiple domain training –all outcomes). In addition, subgroup analysis of intervention type (restorative, compensatory, multiple domain) was also not significant (Total Between Q = 0.836; df = 3; p = 0.841).

DISCUSSION

Examining the effects of training in MCI, the present meta-analysis found that two broad categories of training, compensatory forms of intervention and multicomponent approaches, were associated with moderate effect sizes on outcome measures. This suggests that, generally, individuals with MCI who are taught skills to supplement existing abilities (compensatory strategies) or provided training with multiple components (e.g., training in various domains of cognition, lifestyle changes, nutrition, etc.) are apt to demonstrate an improvement in cognition. In addition, with respect to intervention content, we found that both memory-based training and multidomain interventions were associated with improved cognitive performance on neuropsychological outcome measures post-intervention. We found significant, large effects for memory-focused training (Hedges’ g observed = 0.947; 95% CI [–1.668, 3.562]; Z = 2.517; p = 0.012) and significant, moderate effects for multidomain-based strategies (Hedges’ g observed = 0.420; 95% CI [–0.4491, 1.2891]; Z = 3.525; p < 0.001). These findings suggest that individuals with MCI who are taught memory strategies or multidomain forms of intervention are apt to do better on cognitive measures than those with MCI who did not receive training.

Furthermore, a subgroup analysis of intervention content found significant point estimates for memory-based forms of training and multidomain interventions, with memory-based forms of content yielding significantly greater summary effects than multidomain interventions (SMD Z = 2.162; p = 0.031, two-tailed; all outcomes). This suggests that, while both are beneficial, treatment interventions that are strictly memory-based may be more effective at improving cognition in individuals diagnosed with MCI than interventions that target multiple cognitive domains. Strategies such as errorless learning, method of loci, spaced retrieval, categorization, and clustering, are apt to have greater cognitive benefits than interventions that target multiple domains simultaneously. There was, however, no significant difference between effect sizes when comparing outcomes limited to its respective domain (SMD Z = 1.762; p = 0.078; memory training –memory measures versus multidomain training –all outcomes). While the summary effects for each intervention appeared markedly discrepant (Memory SMD = 0.975; Multidomain SMD = 0.285), this difference did not reach statistical significance. This may reflect a limitation of memory training or methodological constraints of the studies examined (e.g., diversity of training strategies, array of outcome measures, and other differences in study design).

A subgroup analysis of MCI diagnostic category (aMCI, MCI-MD, MCI [all subtypes]), found that there was no difference between MCI type on cognitive outcomes, although this may be indeterminant given the lack of uniformity across studies in how the subtype was diagnosed, questionable discriminability between subtypes, a lack of data regarding illness onset/duration, and potentially different disease etiologies [61 –65]. It should be noted that results from the subgroup analysis should be regarded as observational and not the equivalent of conducting a well-designed head-to-head comparison of each training approach.

Taken together, these data provide additional direction regarding the areas to target in cognitive training and, while strictly speculative, may shed insight on the possible cognitive processes operating in MCI. Provisionally, based on the STAC-r model [31], the large effect size observed with memory-based trainings suggests that neoplastic processes remain intact in MCI and, possibly, that memory-focused interventions prompt hippocampal activation and the primary networks that are associated with memory, such as the networks involved in memory consolidation and retrieval [66, 67]. In addition, memory-based trainings may also have transfer effects that benefit other brain regions, such as prefrontal areas, premotor cortex, fusiform cortex, and parietal cortex [68 –72] which facilitate the engagement of compensatory processes and neural reorganization [31]. As such, these findings are consistent with the theory that MCI is an early-onset neurodegenerative disorder where primary networks retain some facility for activation and compensatory neurocognitive processes can have indirect effects by recruiting other regions to provide “compensatory scaffolding” to meet task demands.

However, the benefits of memory-based interventions appear to be limited. When comparing summary effects of training content matched with domain-related outcomes, there was a non-significant difference between memory-based training and multidomain forms of interventions. We found no difference between outcomes from memory measures versus outcomes from all domains, after memory and multidomain training, respectively (p = 0.0782, two-tailed). The marginal difference between summary effects on domain-related outcomes may reflect compromise of primary memory networks and loss of resilience in memory. As such, memory training may generalize to other domains and engage alternate processing mechanisms [73]. Hence, while memory-focused training is apt to yield greater benefits than multidomain interventions (all outcomes), the non-significant difference observed with domain-related outcomes may indicate that memory systems are strained and, possibly, in a transitional state.

Finally, the moderate effects associated with multidomain forms of training are consistent with the notion that there is activation of compensatory mechanisms and modest engagement of primary cognitive systems. In this instance, multidomain interventions may provide neurocomputational support to process load demands by activating alternate networks. For instance, the loss of detail in encoding and retrieval may result in greater reliance on the recognition of target items or the use of associations to aid in recall. Additionally, frontal systems may provide support through determining the ‘gist’ of what needs to be recalled, narrowing plausible responses, problem-solving, etc. However, while beneficial, compensatory activation may be ineffective at high levels of demand and reach a ceiling, resulting in poorer cognitive performance [29]. Multidomain forms of training may also activate multiple networks indiscriminately, compared to more specific, high-demand interventions. As such, the benefits of multidomain forms of training may be to boost cognitive performance by combining strategies such as focusing attention, identifying key ideas to encode, and providing education about cognitive processes, collectively. It should be noted that discussion about the underlying neural mechanisms and network activation is only speculative in nature. Although the findings reported in this meta-analysis provide evidence to support these inferences, future research using neuroimaging techniques is needed to demonstrate that memory-based interventions do in fact provide primary network activation and the activation of compensatory mechanisms in individuals with MCI.

There are several limitations of this study worth noting. Specifically, some of the analyses only included a small number of studies, multiple outcomes were used to measure cognition, and there was a diverse range of experimental conditions reported across studies. The technical challenges associated with a large range of intervention types and outcome measures may have contributed to some of the heterogeneity and publication bias observed. Additionally, some studies that were included in this meta-analysis had missing data in some cognitive domains and had limited methodology [74]. This prevented a more thorough analysis of domain-based interventions and domain-related outcomes. Specifically, the non-significant difference between training-content matched with domain-associated outcomes may indicate there is an insufficient amount of data to make meaningful comparisons between these two forms of interventions. In a broader context, the neurocognitive processes involved in cognitive training are likely to be more complex, particularly in an illness such as MCI [75]. Moreover, hippocampal activation in MCI has been described as a dysfunctional condition so decreasing hippocampal activation may be more beneficial than increasing hippocampal activation [76]. Given this, training interventions that include the activation of inhibitory processes and incorporate wider network regions may be of greater benefit to individuals with MCI [77]. Future research is needed to explore this possibility.

This review conducted a meta-analysis to compare RCTs that use cognitive-based interventions to slow down the rate of cognitive decline among individuals diagnosed with MCI. Results indicated that both memory-based trainings and multidomain interventions are likely to improve cognitive performance in individuals diagnosed with MCI based on their performance post-intervention on neuropsychological measures. However, subgroup comparisons showed that interventions that focus exclusively on memory-enhancing strategies are associated with greater improvements than multidomain interventions, which provide training strategies that target multiple cognitive domains. Consistent with the STAC-r model, these findings suggest that memory-based interventions can have direct effects that facilitate primary activation of brain regions associated with memory as well as indirect effects on other brain regions (e.g., prefrontal cortex, parietal cortex) that facilitates neural reorganization and the engagement of compensatory processes.

Footnotes

ACKNOWLEDGMENTS

This work was supported by the National Institute on Aging under Grant F31AG057174.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Petersen

, Doody

, Kurz

, Mohs

, Morris

, Rabins

, Ritchie

, Rossor

, Thal

, Winblad

(2001) Current concepts in mild cognitive impairment. Arch Neurol 58, 1985–92.

Petersen

, Roberts

, Knopman

, Boeve

, Geda

, Ivnik

, Smith

, Jack

(2009) Mild cognitive impairment. Arch Neurol 66, 1447–55.

Winblad

, Palmer

, Kivipelto

, Jelic

, Fratiglioni

, Wahlund

, Nordberg

, Bäckman

, Albert

, Almkvist

, Arai

(2004) Mild cognitive impairment–beyond controversies, towards a consensus: Report of the International Working Group on Mild Cognitive Impairment. J Intern Med 256, 240–246.

Mitchell

, Shiri-Feshki

(2009) . Rate of progression of mild cognitive impairment to dementia –meta-analysis of 41 robust inception cohort studies. Acta Psychiatr Scand 119, 252–265.

Roberts

, Knopman

, Mielke

, Cha

, Pankratz

, Christianson

, Geda

, Boeve

, Ivnik

, Tangalos

, Rocca

(2014) Higher risk of progression to dementia in mild cognitive impairment cases who revert to normal. Neurology 82, 317–25.

Petersen

, Lopez

, Armstrong

, Getchius

, Ganguli

, Gloss

, Gronseth

, Marson

, Pringsheim

, Day

, Sager

(2018) Practice guideline update summary: Mild cognitive impairment: Report of the Guideline Development, Dissemination, and Implementation Subcommittee of the American Academy of Neurology. Neurology 90, 126–35.

Diniz

, Pinto

, Gonzaga

MLC

, Guimarães

, Gattaz

, Forlenza

(2009) To treat or not to treat? A meta-analysis of the use of cholinesterase inhibitors in mild cognitive impairment for delaying progression to Alzheimer’s disease. Eur Arch Psychiatry Clin Neurosci 259, 248–256.

Raschetti

, Albanese

, Vanacore

, Maggini

(2007) Cholinesterase inhibitors in mild cognitive impairment: A systematic review of randomised trials. PLoS Med 4, e338.

Simon

, Yokomizo

, Bottino

CMC

(2012) Cognitive intervention in amnestic mild cognitive impairment: A systematic review. Neurosci Biobehav Rev 36, 1163–1178.

10.

Hampstead

, Stringer

, Stilla

, Deshpande

, Hu

, Moore

, Sathian

(2011) Activation and effective connectivity changes following explicit-memory training for face-name pairs in patients with mild cognitive impairment: A pilot study. Neurorehabil Neural Repair 25, 210–222.

11.

Rosen

, Sugiura

, Kramer

, Whitfield-Gabrieli

, Gabrieli

(2011) Cognitive training changes hippocampal function in mild cognitive impairment: A pilot study. J Alzheimers Dis 26, 349–357.

12.

Belleville

, Clément

, Mellah

, Gilbert

, Fontaine

, Gauthier

(2011) . Training- related brain plasticity in subjects at risk of developing Alzheimer’s disease. Brain 134, 1623–1634.

13.

Onur

, Kukolja

, Nolfo

, Schlegel

, Kaesberg

, Kessler

, Fink

, Kalbe

(2016) Cognitive training fosters compensatory mechanisms in MCI. Alzheimers Dement 12, P420.

14.

Barban

, Annicchiarico

, Pantelopoulos

, Federici

, Perri

, Fadda

, Carlesimo

, Ricci

, Giuli

, Scalici

, Turchetta

(2016) Protecting cognition from aging and Alzheimer’s disease: A computerized cognitive training combined with reminiscence therapy. Int J Geriatr Psychiatry 31, 340–348.

15.

Chapman

, Aslan

, Spence

, Keebler

, DeFina

, Didehbani

, Perez

, Lu

, D’Esposito

(2016) Distinct brain and behavioral benefits from cognitive vs. physical training: A randomized trial in aging adults. Front Hum Neurosci 10, 338.

16.

Ciarmiello

, Chiara Gaeta

, Benso

, Del Sette

(2015) FDG-PET in the evaluation of brain metabolic changes induced by cognitive stimulation in aMCI subjects. Curr Radiopharm 8, 69–75.

17.

Maffei

, Picano

, Andreassi

, Angelucci

, Baldacci

, Baroncelli

, Begenisic

, Bellinvia

, Berardi

, Biagi

, Bonaccorsi

(2017) Randomized trial on the effects of a combined physical/cognitive training in aged MCI subjects: The Train the Brain study. Sci Rep 7, 39471.

18.

McPhee

, Downey

, Stough

(2019) Effects of sustained cognitive activity on white matter microstructure and cognitive outcomes in healthy middle-aged adults: A systematic review. Ageing Res Rev 51, 35–47.

19.

Bahar-Fuchs

, Clare

, Woods

(2013) Cognitive training and cognitive rehabilitation for mild to moderate Alzheimer’s disease and vascular dementia. Cochrane Database Syst Rev 6, CD003260.

20.

Huckans

, Hutson

, Twamley

, Jak

, Kaye

, Storzbach

(2013) Efficacy of cognitive rehabilitation therapies for mild cognitive impairment (MCI) in older adults: Working toward a theoretical model and evidence-based interventions. Neuropsychol Rev 23, 63–80.

21.

Jean

, Bergeron

, Thivierge

, Simard

(2010) Cognitive intervention programs for individuals with mild cognitive impairment: Systematic review of the literature. Am J Geriatr Psychiatry 18, 281–296.

22.

, Li

, Wang

, Zhou

(2011) Cognitive intervention for persons with mild cognitive impairment: A meta-analysis. Ageing Res Rev 10, 285–296.

23.

Martin

, Clare

, Altgassen

, Cameron

, Zehnder

(2011) Cognition-based interventions for healthy people and people with mild cognitive impairment. Intervention Review. Cochrane Database Syst Rev 1, CD006220.

24.

Reijnders

, van Heugten

, van Boxtel

(2013) Cognitive interventions in healthy older adults and people with mild cognitive impairment: A systematic review. Ageing Res Rev 12, 263–275.

25.

Kim

, Kim

(2014) . A theoretical framework for cognitive and non-cognitive interventions for older adults: Stimulation versus compensation. Aging Ment Health 18, 304–315.

26.

Shatenstein

, Barberger-Gateau

(2015) Prevention of age-related cognitive decline: Which strategies, when, and for whom? J Alzheimers Dis 48, 35–53.

27.

Sitzer

, Twamley

, Jeste

(2006) Cognitive training in Alzheimer’s disease: A meta-analysis of the literature. Acta Psychiatr Scand 114, 75–90.

28.

Sherman

, Mauser

, Nuno

, Sherzai

(2017) The efficacy of cognitive intervention in mild cognitive impairment (MCI): A meta-analysis of outcomes on neuropsychological measures. Neuropsychol Rev 27, 440–484.

29.

Reuter-Lorenz

, Lustig

(2017) Working memory and executive functions in the aging brain. In Cognitive Neuroscience of Aging: Linking Cognitive and Cerebral Aging, Cabeza R, Nyberg L, Park DC, eds. Oxford University Press, New York, pp. 235-258.

30.

Reuter-Lorenz

, Park

(2010) Human neuroscience and the aging mind: A new look at old problems. J Gerontol B Psychol Sci Soc Sci 65, 405–415.

31.

Reuter-Lorenz

, Park

(2014) How does it STAC up? Revisiting the scaffolding theory of aging and cognition. Neuropsychol Rev 24, 355–370.

32.

Pudas

, Persson

, Josefsson

, de Luna

, Nilsson

L-G

, Nyberg

(2013) Brain characteristics of individuals resisting age-related cognitive decline over two decades. J Neurosci 33, 8668–8677.

33.

Cabeza

(2002) Hemispheric asymmetry reduction in older adults: The HAROLD model. Psychol Aging 17, 85–100.

34.

Bubb

, Metzler-Baddeley

, Aggleton

(2018) The cingulum bundle: Anatomy, function, and dysfunction. Neurosci Biobehav Rev 92, 104–127.

35.

Wee

C-Y

, Yap

, Zhang

, Denny

, Browndyke

, Potter

, Welsh-Bohmer

, Wang

, Shen

(2012) Identification of MCI individuals using structural and functional connectivity networks. Neuroimage 59, 2045–2056.

36.

Dai

, He

(2014) Disrupted structural and functional brain connectomes in mild cognitive impairment and Alzheimer’s disease. Neurosci Bull 30, 217–232.

37.

Sun

, Yin

, Fang

, Yan

, Wang

, Bezerianos

, Tang

, Miao

, Sun

(2014) Disrupted functional brain connectivity and its association to structural connectivity in amnestic mild cognitive impairment and Alzheimer’s disease. PLoS One 9, e96505.

38.

Bai

, Shu

, Yuan

, Shi

, Yu

, Wu

, Wang

, Xia

, He

, Zhang

(2012) Topologically convergent and divergent structural connectivity patterns between patients with remitted geriatric depression and amnestic mild cognitive impairment. J Neurosci 32, 4307–4318.

39.

Perry

, Wen

, Lord

, Thalamuthu

, Roberts

, Mitchell

, Sachdev

, Breakspear

(2015) The organisation of the elderly connectome. Neuroimage 114, 414–426.

40.

Cespón

, Miniussi

, Pellicciari

(2018) Interventional programmes to improve cognition during healthy and pathological ageing: Cortical modulations and evidence for brain plasticity. Ageing Res Rev 43, 81–98.

41.

Moher

, Liberati

, Tetzlaff

, Altman

(2009) Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. Ann Intern Med 151, 264–269.

42.

Albert

, DeKosky

, Dickson

, Dubois

, Feldman

, Fox

, Gamst

, Holtzman

, Jagust

, Petersen

, Snyder

(2011) The diagnosis of mild cognitive impairment due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 270–279.

43.

Gates

, Valenzuela

(2010) Cognitive exercise and its role in cognitive function in older adults. Curr Psychiatry Rep 12, 20–27.

44.

Petersen

(2004) Mild cognitive impairment as a diagnostic entity. J Intern Med 256, 183–194.

45.

Vidovich

, Lautenschlager

, Flicker

, Clare

, McCaul

, Almeida

(2015) The PACE study: A randomized clinical trial of cognitive activity strategy training for older people with mild cognitive impairment. Am J Geriatr Psychiatry 23, 360–372.

46.

Fillenbaum

, van Belle

, Morris

, Mohs

, Mirra

, Davis

, Tariot

, Silverman

, Clark

, Welsh-Bohmer

, Heyman

(2008) Consortium to Establish a Registry for Alzheimer’s Disease (CERAD): The first twenty years. Alzheimers Dement 4, 96–109.

47.

Lezak

, Howieson

, Loring

, Fischer

(2004) Neuropsychological assessment. Oxford University Press, New York.

48.

Cohen

(1988) Statistical power analysis for the behavioral sciences. Erlbaum, Hillsdale, NJ.

49.

Durlak

(2009) . How to select, calculate, and interpret effect sizes. J Pediatr Psychol 34, 917–928.

50.

DerSimonian

, Laird

(1986) Meta-analysis in clinical trials. Control Clin Trials 7, 177–188.

51.

Borenstein

, Hedges

, Higgins

JPT

, Rothstein

(2009) Introduction to meta-analysis. JohnWiley & Sons, West Sussex.

52.

IntHout

, Ioannidis

JPA

, Borm

(2014) The Hartung-Knapp-Sidik-Jonkman method for random effects meta-analysis is straightforward and considerably outperforms the standard DerSimonian-Laird method. BMC Med Res Methodol 14, 25.

53.

Hamman

, Pappalardo

, Bence

, Peacor

, Osenberg

(2018) Bias in meta-analyses using Hedges’d. Ecosphere 9, e02419.

54.

Hedges

, Olkin

(1985) Statistical methods for meta-analysis. Academic Press.

55.

Borenstein

(2019) Common mistakes in meta-analysis and how to avoid them. Biostat, Incorporated.

56.

Doncaster

, Spake

(2018) Correction for bias in meta-analysis of little-replicated studies. Methods Ecol Evol 9, 634–644.

57.

Balietti

, Giuli

, Fattoretti

, Fabbietti

, Postacchini

, Conti

(2016) Cognitive stimulation modulates platelet total phospholipases A2 activity in subjects with mild cognitive impairment. J Alzheimers Dis 50, 957–962.

58.

Herrera

, Chambon

, Michel

, Paban

, Alescio-Lautier

(2012) Positive effects of computer-based cognitive training in adults with mild cognitive impairment. Neuropsychologia 50, 1871–1881.

59.

Greenaway

, Duncan

, Smith

(2012) The memory support system for mild cognitive impairment: Randomized trial of a cognitive rehabilitation intervention. Int J Geriatr Psychiatry 28, 402–409.

60.

Jeong

, Na

, Choi

, Kim

, Na

, Seo

, Chin

, Park

, Kim

, Han

(2016) Group-and home-based cognitive intervention for patients with mild cognitive impairment: A randomized controlled trial. Psychother Psychosom 85, 198–207.

61.

Csukly

, Siraly

, Fodor

, Horvath

, Salacz

, Csukly

, Sirály

, Fodor

, Horváth

, Salacz

, Hidasi

, Csibri

, Rudas

, Szabo

(2016) The differentiation of amnestic type MCI from the non-amnestic types by structural MRI. Front Aging Neurosci 8, 1–10.

62.

Guan

, Liu

, Jiang

, Tao

, Zhang

, Niu

, Zhu

, Wang

, Cheng

, Kochan

, Brodaty

(2017) Classifying MCI subtypes in community-dwelling elderly using cross-sectional and longitudinal MRI-based biomarkers. Front Aging Neurosci 9, 309.

63.

, Farias

, Martinez

, Reed

, Mungas

, DeCarli

(2009) Differences in brain volume, hippocampal volume, cerebrovascular risk factors, and apolipoprotein E4 among mild cognitive impairment subtypes. Arch Neurol 66, 1393–1399.

64.

Kantarci

, Petersen

, Przybelski

, Weigand

, Shiung

, Whitwell

, Negash

, Ivnik

, Boeve

, Knopman

, Smith

(2008) Hippocampal volumes, proton magnetic resonance spectroscopy metabolites, and cerebrovascular disease in mild cognitive impairment subtypes. Arch Neurol 65, 1621–1628.

65.

Van De Pol

, Verhey

, Frisoni

, Tsolaki

, Papapostolou

, Nobili

, Wahlund

, Minthon

, Frölich

, Hampel

, Soininen

(2009) White matter hyperintensities and medial temporal lobe atrophy in clinical subtypes of mild cognitive impairment: The DESCRIPA study. J Neurol Neurosurg Psychiatry 80, 1069–1074.

66.

Gutchess

, Welsh

, Hedden

, Bangert

(2005) Aging and the neural correlates of successful picture encoding: Frontal activations compensate for decreased medial-temporal activity. J Cogn Neurosci 17, 84–96.

67.

Wais

(2008) fMRI signals associated with memory strength in the medial temporal lobes: A meta-analysis. Neuropsychologia 46, 3185–3196.

68.

Kim

(2011) Neural activity that predicts subsequent memory and forgetting: A meta-analysis of 74 fMRI studies. Neuroimage 54, 2446–2461.

69.

Maillet

, Rajah

(2011) Age-related changes in the three-way correlation between anterior hippocampus volume, whole-brain patterns of encoding activity and subsequent context retrieval. Brain Res 1420, 68–79.

70.

Persson

, Nyberg

(2006) Altered brain activity in healthy seniors: What does it mean? Prog Brain Res 157, 45–56.

71.

Wang

, Yu

, Zhao

, Li

, Yin

, Han

(2018) Altered whole brain structural covariance of the hippocampal subfields in subcortical vascular mild cognitive impairment and mild cognitive impairment patients. Front Neurol 9, 1–12.

72.

Yang

, Wen

, Jiang

, D Crawford

, Reppermund

, Levitan

, J Slavin

, A Kochan

, L Richmond

, Brodaty

, N Trollor

(2016) Structural MRI biomarkers of mild cognitive impairment from young elders to centenarians. Curr Alzheimer Res 13, 256–267.

73.

Davis

, Dennis

, Daselaar

, Fleck

, Cabeza

(2008) Qué PASA? The posterior-anterior shift in aging. Cereb Cortex 18, 1201–1209.

74.

Stott

, Spector

(2011) A review of the effectiveness of memory interventions in mild cognitive impairment (MCI). Int Psychogeriatr 23, 526–538.

75.

Farràs-Permanyer

, Guàrdia-Olmos

, Peró-Cebollero

(2015) Mild cognitive impairment and fMRI studies of brain functional connectivity: The state of the art. Front Psychol 6, 1095.

76.

Bakker

, Krauss

, Albert

, Speck

, Jones

, Stark

, Yassa

, Bassett

, Shelton

, Gallagher

(2012) Reduction of hippocampal hyperactivity improves cognition in amnestic mild cognitive impairment. Neuron 74, 467–474.

77.

Jurick

, Weissgerber

, Clark

, Wierenga

, Chang

, Schieher

, Han

, Jak

, Dev

, Bondi

(2018) Faulty adaptation to repeated face-name associative pairs in mild cognitive impairment is predictive of cognitive decline. Arch Clin Neuropsychol 33, 168–183.

78.

Bahar-Fuchs

, Webb

, Bartsch

, Clare

, Rebok

, Cherbuin

, Anstey

(2017) Tailored and adaptive computerized cognitive training in older adults at risk for dementia: A randomized controlled trial. J Alzheimers Dis 60, 889–911.

79.

Barnes

, Yaffe

, Belfor

, Jagust

, DeCarli

, Reed

, Kramer

(2009) Computer-based cognitive training for mild cognitive impairment: Results from a pilot randomized, controlled trial. Alzheimer Dis Assoc Disord 23, 205–210.

80.

Buschert

, Friese

, Teipel

, Schneider

, Merensky

, Rujescu

, Möller

H-J

, Hampel

, Buerger

(2011) Effects of a newly developed cognitive intervention in amnestic mild cognitive impairment and mild Alzheimer’s disease: A pilot study. J Alzheimers Dis 25, 679–694.

81.

Carretti

, Borella

, Fostinelli

, Zavagnin

(2013) Benefits of training working memory in amnestic mild cognitive impairment: Specific and transfer effects. Int Psychogeriatr 25, 617–626.

82.

Djabelkhir

, Wu

Y-H

, Vidal

J-S

, Cristancho-Lacroix

, Marlats

, Lenoir

, Carno

, Rigaud

A-S

(2017) Computerized cognitive stimulation and engagement programs in older adults with mild cognitive impairment: Comparing feasibility, acceptability, and cognitive and psychosocial effects. Clin Interv Aging 12, 1967–1975.

83.

Donnezan

, Perrot

, Belleville

, Bloch

, Kemoun

(2018) Effects of simultaneous aerobic and cognitive training on executive functions, cardiovascular fitness and functional abilities in older adults with mild cognitive impairment. Ment Health Phys Act 15, 78–87.

84.

Emsaki

, NeshatDoost

, Tavakoli

, Barekatain

(2017) Memory specificity training can improve working and prospective memory in amnestic mild cognitive impairment. Dement Neuropsychol 11, 255–261.

85.

Fiatarone Singh

, Gates

, Saigal

, Wilson

, Meiklejohn

, Brodaty

, Wen

, Singh

, Baune

, Suo

, Baker

(2014) The Study of Mental and Resistance Training (SMART) study—resistance training and/or cognitive training in mild cognitive impairment: A randomized, double-blind, double-sham controlled trial. J Am Med Dir Assoc 15, 873–880.

86.

Finn

, McDonald

(2011) Computerised cognitive training for older persons with mild cognitive impairment: A pilot study using a randomised controlled trial design. Brain Impair 12, 187–199.

87.

Finn

, McDonald

(2015) Repetition-lag training to improve recollection memory in older people with amnestic mild cognitive impairment. A randomized controlled trial. Neuropsychol Dev Cogn B Aging Neuropsychol Cogn 22, 244–258.

88.

Förster

, Buschert

, Teipel

, Friese

, Buchholz

, Drzezga

, Hampel

, Bartenstein

, Buerger

(2011) Effects of a 6-month cognitive intervention on brain metabolism in patients with amnestic MCI and mild Alzheimer’s disease. J Alzheimers Dis 26, 337–348.

89.

Gagnon

, Belleville

(2012) Training of attentional control in mild cognitive impairment with executive deficits: Results from a double-blind randomised controlled study. Neuropsychol Rehabil 22, 809–835.

90.

Giuli

, Papa

, Lattanzio

, Postacchini

(2016) The effects of cognitive training for elderly: Results from My Mind Project. Rejuvenation Res 19, 485–494.

91.

Hampstead

, Stringer

, Stilla

, Giddens

, Sathian

(2012) Mnemonic strategy training partially restores hippocampal activity in patients with mild cognitive impairment. Hippocampus 22, 1652–1658.

92.

Jean

, Simard

, Wiederkehr

, Bergeron

M-E

, Turgeon

, Hudon

, Tremblay

, van Reekum

(2010) Efficacy of a cognitive training programme for mild cognitive impairment: Results of a randomised controlled study. Neuropsychol Rehabil 20, 377–405.

93.

Lam

, Chan

, Leung

, Fung

, Leung

(2015) Would older adults with mild cognitive impairment adhere to and benefit from a structured lifestyle activity invention to enhance cognition?: A cluster randomized controlled trial. PLoS One 10, e0118173.

94.

Mowszowski

, Hermens

, Diamond

, Norrie

, Cockayne

, Ward

, Hickie

, Lewis

, Batchelor

, Naismith

(2014) Cognitive training enhances pre-attentive neuropsychological responses in older adults ‘at risk’ of dementia. J Alzheimers Dis 41, 1095–1108.

95.

Mudar

, Chapman

, Rackley

, Eroh

, Chiang

H-S

, Perez

, Venza

, Spence

(2017) Enhancing latent cognitive capacity in mild cognitive impairment with gist reasoning training: A pilot study. Int J Geriatr Psychiatry 32, 548–555.

96.

Olchik

, Farina

, Steibel

, Teixeira

, Yassuda

(2013) Memory training (MT) in mild cognitive impairment (MCI) generates change in cognitive performance. Arch Gerontol Geriatr 56, 442–447.

97.

Park

, Park

(2018) Does cognition-specific computer training have better clinical outcomes than non-specific computer training? A single-blind, randomized controlled trial. Clin Rehabil 32, 213–222.

98.

Polito

, Abbondanza

, Vaccaro

, Valle

, Davin

, Degrate

, Villani

, Guaita

(2015) Cognitive stimulation in cognitively impaired individuals and cognitively healthy individuals with a family history of dementia: Short-term results from the “Allena-Mente” randomized controlled trial. Int J Geriatr Psychiatry 30, 631–638.

99.

Rapp

, Brenes

, Marsh

(2002) Memory enhancement training for older adults with mild cognitive impairment: A preliminary study. Aging Ment Health 6, 5–11.

100.

Rojas

, Villar

, Iturry

, Harris

, Serrano

, Herrera

, Allegri

(2013) Efficacy of a cognitive intervention program in patients with mild cognitive impairment. Int Psychogeriatr 25, 825–831.

101.

Schmitter-Edgecombe

, Dyck

(2014) Cognitive rehabilitation multi-family group intervention for individuals with mild cognitive impairment and their care-partners. J Int Neuropsychol Soc 20, 897–908.

102.

Tsolaki

, Kounti

, Agogiatou

, Poptsi

, Bakoglidou

, Zafeiropoulou

, Soumbourou

, Nikolaidou

, Batsila

, Siambani

, Nakou

(2011) Effectiveness of nonpharmacological approaches in patients with mild cognitive impairment. Neurodegener Dis 8, 138–145.

103.

Valdes

, O’Connor

, Edwards

(2012) The effects of cognitive speed of processing training among older adults with psychometrically-defined mild cognitive impairment. Curr Alzheimer Res 9, 999–1009.