Background
Methods
Search strategy
Study selection
Data extraction and quality assessment
Results
Overview of studies included
Reference | Study Samplea
| Study Design Length of follow-up | MRI Magnet | Neuroimaging Outcome Measures | Cognition measured (test name) |
---|---|---|---|---|---|
Suo et al. [19] 2016 | Older adults with MCI
N = 100 70.1 ± 6.7 years Completed MRI: N = 79 | RCT Assessments at baseline and 6 months | 3 T | • Volumetric Structural MRI • Resting-state fMRI | Global Cognition (ADAS-Cog) [41] ○ Memory Domain ○ Executive Function ○ Attention-Speed |
Rosen et al. [18] 2011 | Older adults with MCI
N = 12 74.34 ± 9.25 years Training
N = 6 70.67 ± 10.58 years Control
N = 6 78 ± 7.92 years | RCT Assessments conducted on average 72 ± 26 days apart | 3 T | • Task-based fMRI • Incidental Auditory-Verbal Repetition paradigm
| Memory (Repeatable Battery for the Assessment of Neuropsychological Status: RBANS) [42] |
Lampit et al. [16] 2015 | Healthy older adults: Subsample from Timecourse Trial
N = 12 71.43 ± 7.48 years Training
N = 7 72.3 ± 8 years Control
N = 5 70.2 ± 6.7 years | RCT Assessments at baseline, 3 weeks: Follow Up 1(FU1), 3 months: Follow Up 2 (FU2) Secondary analysis | 3 T | • Volumetric Structural MRI • Resting-state fMRI • Proton Magnetic Resonance Spectroscopy • DTI | Global Cognition: |
Belleville et al. [21] 2014 | Healthy community-dwelling older adults
N = 40 69 ± 6.27 years Training group 1
N = 12 68.58 ± 8.16 years Training group 2
N = 14 69.57 ± 5.81 years Training group 3
N = 14 68.79 ± 5.13 years | Quasi-experimental Pre-post Assessments 1 week before and 1 week after training | 3 T | • Task-based fMRI • Alphanumeric equation task
• Visual detection task
Tasks performed as single-task and dual-task | Reaction Time (Alphanumeric equation task and visual detection task) Accuracy (Alphanumeric equation task and visual detection task) |
Lin et al. [17] 2014 | Older adults with a history of a stroke
N = 34 69.21 ± 4.93 years Training
N = 16 62.4 ± 6.0 years Control
N = 18 63.2 ± 5.7 years | RCT Assessments at baseline and 10 weeks | 3 T | • Resting state fMRI | Memory (Wechsler Memory Scale) [46] Executive Function (Trail Making Test) [47] |
Strenziok et al. [24] 2014 | Healthy older adults
N = 42 69.21 ± 4.93 years Training group 1
N = 14 69.70 ± 6.9 years Training group 2
N = 14 68.52 ± 5.6 years Training group 3
N = 14 69.41 ± 2.3 years | Quasi-experimental Pre-post Length of follow up: Not stated | Not stated | • Resting-state fMRI • DTI | Reasoning/Problem Solving (WAIS III Matrix Reasoning subtest, [48] Everyday problems Test, [49] Word Series and Letter Series Tests) [50] Auditory Working Memory (Letter Number Sequencing subtest of WAIS III) [48] |
Lövden et al. [23] 2010 | Healthy older adultsb: Subsample COGITO study
N = 25 69.32 ± 3.12 years Training
N = 12 68.9 ± 2.7 years Control
N = 13 69.7 ± 3.5 years | Quasi-experimental Pre-post Training: Pre-post MRI on average 179 ± 25.2 days apart Control: Pre-post MRI on average 184 ± 15.0 days apart | 1.5 T | • DTI | Spatial Working Memory (3-Back)c
Numerical Working Memory (Memory Updating)c
Figural-Spatial Episodic Memory (Object-Position Memory)c
Numerical Episodic Memory (Number-noun pairs)c
Verbal Episodic Memory (Wordlist)c
Perceptual Speed (Choice Reaction Task, Comparison tasks)c
|
Antonenko et al. [20] 2016 | Healthy older adults
N = 25 69 ± 6 years | Quasi-experimental Pre-post Assessments 1 day before (pre), 1 day after (post) and 1 month after (follow-up) trainingd
| 3 T | • Volumetric Structural MRI • DTI | Cued recall (3-alternative-forced-choice recall task (AFC); main outcome) [55] and recognition Episodic Memory control task (German Rey Auditory Verbal Learning Test) [56] |
Heinzel et al. [22] 2014 | Healthy older adultsc
N = 19 65.95 ± 3.73 years | Quasi-experimental Pre-post Subset of 15 older individuals performed pre-post MRI Length of follow up: Not stated | 3 T | • Volumetric Structural MRI • Task-based fMRI • N-back [57]: two runs (16 blocks/run) with 4 working memory loads (0, 1, 2, 3) • Functional Connectivity (PPI) | Relative Working Memory Training gain (n-Back) [57] Short-term memory (Digit span Fwd and Bwd WAIS III) [51] Executive Functions: Verbal Fluency (Controlled Oral Word Association Test) [59] Inhibition (Stroop) [60] |
Reference | Training program/ task Cognitive Domain Trained | Description of Training | Training Frequency Training Duration | Total hours of training Supervised/ Home-based | Control Group |
---|---|---|---|---|---|
Suo et al. [19] 2016 | COGPACK Multidomain: memory, attention, response speed, executive functions, language | COGPACK: Exercises focused on memory, attention, response speed, executive functions, and language | 26 weeks 52 sessions 90 min/session | 78 Supervised | Active: watched videos on computer, followed by questions |
Rosen et al. [18] 2011
| Posit Science Multidomain: processing speed, accuracy in auditory processing | Auditory verbal repetition paradigm: 7 exercises aimed at improving processing speed and accuracy in auditory processing | 5 weeks 24 sessions 100 min/session | 36 Home | Active: 3 different computer-based activities (listening to audiobooks, reading online news, playing visuospatial computer game) |
Lampit et al. [16] 2015 | COGPACK Multidomain: memory, attention, response speed, executive functions, language | Exercises focused on memory, attention, response speed, executive functions, and language | 12 weeks 3×/week 60 min/session | 36 Supervised | Active: viewed 7 National Geographic videos per session on a computer with multiple choice questions |
Belleville et al. [21] 2014 | Customized program Executive Function: Attention |
Alphanumeric equation task: judge accuracy of visually presented letter and number equations. Visual detection task: detect the red rectangles (press a button) in a series of white and red rectangles Groups: 1. Single repeated: Complete both tasks individually (focused attention) 2. Divided fixed: Complete 2 tasks simultaneously with divided attention (50%) 3. Divided variable: Complete two tasks simultaneously with different attention allocation levels (80%, 50%, 20%) | 2 weeks 3×/week 1 h/session | 6 Supervised | No Control |
Lin et al. [17] 2014 | RehaCom Executive Function and memory | Computer-assisted exercise focused on memory and executive function | 10 weeks 6×/week 60 min/session | 60 Supervised | Passive |
Strenziok et al. [24] 2014 | Multidomain: Brain Fitness (BF): auditory perception; Space Fortress (SF): visuomotor and working memory Rise of Nation (RoN): attention, motor processing, working memory, reasoning, visuospatial short-term memory, task-switching | 1. Brain Fitness (BF): Adaptive auditory perception computer game 2. Space Fortress (SF): Complex skill acquisition computer game 3. Rise of Nations (RoN): Off-the shelf real-time strategy computer game | 6 weeks 36 sessions 60 min/session | 36 Supervised + Home (50–50%) | No Control |
Lövden et al. [23] 2010 | Customized program Multidomain: working memory, episodic memory, perceptual speed | Working Memory (3-Back, Memory updating, Alpha span) Episodic memory (Object-position memory, Number-noun pairs, Word lists) Perceptual speed (Choice reaction tasks, Comparison Tasks) | >4 months Average of 100 ± 3.7 sessions 60 min/session | Average of 100 Supervised | Passive: Pre-post MRI only |
Antonenko et al. [20] 2016 | Object-location Learning Paradigm Memory |
Object-location Learning Paradigm: Learn the correct spatial locations of buildings on a street map. Five blocks of 120 stimulus-location pairing with a response interval of 3 s. Each block was followed by a cued recall and a recognition task | 3 consecutive days 5 learning blocks/day | Unknown Unspecified | No Control |
Heinzel et al. [22] 2014 |
n
-Back training Executive Function: Working Memory | Adaptive n-back training, 3 runs (12 blocks/run) each session. Difficulty level increased according to individual performance (higher working memory load, shortened interstimulus interval (ISI). ISI ranged from 1500 to 500 ms in steps of 500 ms. | 4 weeks 3×/week 45 min/session | 9 Supervised | No control |
Reference | Structural changes | Functional changes | Changes in connectivity | Cognition Outcome | Cognition related to imaging outcome |
---|---|---|---|---|---|
Suo et al. [19] 2016 | Combined cognitive training and progressive resistance training led to increased cortical thickness in posterior cingulate cortex. Cognitive training alone led to atrophy. | - | Cognitive training groups showed Group X Time interaction indicating decreased connectivity between the posterior cingulate and superior frontal lobe (F(67) = 31.7, p < 0.001) and between the posterior cingulate and the anterior cingulate cortex (F(67) = 13.9, p < 0.001)a. Cognitive training group (alone or combined with exercise) showed a Group X Time interaction indicating increased connectivity between hippocampus and the left superior frontal lobe compared with non-computerized cognitive training (p = 0.012)a
| Computerized cognitive training (alone and with resistance training): Memory domain: Group X Time interaction (F(90) = 5.7, p < 0.02) showing no decline in cognitive training group compared to non-cognitive training groupsa
ADAS-Cog: No effect of cognitive training | Change in posterior cingulate grey matter correlated with improvement in the ADAS-Cog (r = 0.25, p = 0.030)a. Increased connectivity between hippocampus and superior frontal lobe was correlated with improved memory domain performance (r = 0.33, p = 0.005)a
|
Rosen et al. [18] 2011 | - | Significant increase of activation in left anterior hippocampus in experimental group compared with controls. | - | A non-significant but greater gain in memory performance in experimental group compared with control group (F(1,10) = 4.76, p = 0.054). Change scores showed improved memory performance in intervention group compared with decrease in performance in the control group (t(10) = 2.61, p < .0027, Cohen’s d = 1.38) | Non-significant trend showing changes in hippocampal activation correlated positively with changes in memory score on RBANS in all participants (r = 0.49, p = 0.10, Cohen’s
d = 1.14) |
Lampit et al. [16] 2015 | Significant increase in grey matter density in right post-central gyrus in training group compared with a decrease in control. Vertex-based analysis showed significant difference in rate of thickness change over time between training and control in both the left fusiform gyrus (T > 3.39) and the supramarginal and post-central gyri (T > 2.24). | - | Group x Time interaction showed functional connectivity decrease between posterior cingulate and right superior frontal gyrus in training group while functional connectivity increased in the control group (p = .006) at FU1. Group x Time interaction showed functional connectivity increase between right hippocampus and left superior temporal gyrus in CCT, while decreased in control at first FU1 (p = .029). No significant Group x Time interactions found for Magnetic Resonance Spectroscopy (MRS) and whole brain Diffusion Tensor Imaging (DTI) | Repeated-measured ANOVA showed improved global cognition in training group compared to control (Group X Time, F = 7.833, p = 0.003). Effect size on Global Cognition (d = 0.94 baseline versus FU1 and d = 2.18 baseline versus FU2) | Significant positive correlation between change in grey matter density in right post-central gyrus at FU2 and change in global cognition at FU1(r = 0.647, p = .023) and FU2 (r = 0.584, p = 0.046) in both training and control. Inverse correlation between functional connectivity between posterior cingulate and right superior frontal gyrus at FU1 and change in global cognition at FU2 (r = −.771, p = .003). Significant positive correlation in functional connectivity between the right hippocampus and left superior temporal gyrus at FU1 and change in global cognition at FU2 (r = 0.591, p = .043). |
Belleville et al. [21] 2014 | - | Single Repeated:
Alphanumeric single
task: Decreased post- training activation in inferior and right middle frontal gyrus (t = 5.91), left middle frontal gyrus (t = 4.57) and left thalamus (t = 5.37).
Visual detection single
task: no change
Dual task: no change Divided Fixed
Alphanumeric single task: no change
Visual detection single task: Decreased post-training activation in right cerebellum (t = 4.73) and right middle occipital gyrus (t = 4.68) when performing the visual detection task.
Dual task (50/50): Small increase in post-training activation in right and left middle frontal gyrus (areas 11, 47; t = 4.41 and t = 4.52 respectively). Divided Variable
Alphanumeric single task: no change
Visual detection single task: no change
Dual task:
Significant increased activation in right middle frontal gyrus (area 10; for 20% attention allocation t = 5.35 and 50% attention allocation t = 4.78). No reduced post-training activation in 80% attention allocation. | - |
Alphanumeric single task: All groups showed improved reaction time (RT; F(1,34) = 9.75, p < .001, η2 = .22) and accuracy (AC; F(1,34) = 14.8, p = .001, η2 = .30)
Visual detection single task: No change
Dual task (cost score)
b: Single repeated: No improvements in dual tasking Divided Fixed: Reduced dual-task cost (F(1,34) = 6.97, p < .001, η2 = .45) Divided Variable: Reduced dual-task cost and were able to modify attentional priority (F(2,33) = 5.17, p < .001, η2 = .34) | Single Repeated:
Alphanumeric single task: Significant positive correlation between right inferior and middle frontal gyrus activation and reaction time (r = .56, p < .05). Divided Variable: Significant negative correlation (post training) between activation of right superior and middle frontal gyrus (Brodmann area 10) and attentional cost (r = −.55, p < .05) |
Lin et al. [17] 2014 | - | - | Training group: Significant increased functional connectivity in (all p’s < 0.005): 1.Left hippocampus-right inferior frontal gyrus 2.Left hippocampus-right middle frontal gyrus 3.Right hippocampus-left middle frontal gyrus 4.Right hippocampus-left inferior frontal gyrus 5.Right hippocampus-left superior frontal gyrus 6.Right hippocampus-left parietal lobe Control group: Significantly decreased functional connectivity (all p’s < 0.005): 1. Left hippocampus-right middle occipital gyrus 2. Right hippocampus-right posterior lobe or cerebellum 3. Right hippocampus-left superior temporal gyrus | Training group: 1. Significant improved scores on 5/7 subtests from Wechsler Memory Scale, namely: Mental control (p = 0.003), Logical memory (p < 0.001), Digits forward and backward (p = 0.014), Visual reproduction (p = 0.008), and Associated learning (p < 0.001). 2. Improved Memory quotient (p = 0.005) 3. Improved performance on Trail Making Test-A (p < 0.001) Control group: no significant changes between baseline and 10-week scores | Training group: significant positive correlations between (all p’s < 0.001): 1. Memory quotient and functional connectivity of left hippocampus-right frontal lobe (r = 0.64) 2. Memory quotient and functional connectivity of right hippocampus-left frontal lobe (r = 0.85) 3. Memory quotient and functional connectivity of right hippocampus-left parietal lobe (r = 0.79) 4. Trail Making Test-A score and functional connectivity of left hippocampus-right frontal lobe (r = 0.94) 5. Trail Making Test-A and functional connectivity of right hippocampus-left frontal lobe (r = 0.68) Control group: no significant correlations between cognition and functional connectivityc
|
Strenziok et al. [24] 2014 | - | - | Ventral Network: Axial diffusivity (AD) in the right occipito-temporal white matter significantly increased after BF compared with a decrease after SF and RON (p < 0.05) Dorsal Network: Functional connectivity between right superior parietal cortex (SPC) and left posterior inferior temporal lobe (ITL) decreased in SF and increased in RON (p = 0.02). Functional connectivity between right SPC and left anterior ITL decreased in BF and showed an increase in RON (p = 0.03) | Univariate ANOVA showed main effects of training group: Reasoning on Everyday Problems Test: Main effect of training group (F(2,39) = 5.34, p < 0.01, partial η2 = 0.215). BF and SF showed improved performance after training and RON showed no effect. Spatial Working Memory: Main effect of training group (F(2,39) = 5.03, p < 0.001, partial η2 = 0.205). SF improved performance after training, RON decreased performance, and BF showed no effect. Matrix Reasoning: Main effect of training group (F(2,39) = 3.40, p < 0.044, partial η2 = 0.148). Largest gains seen in BF and a smaller gain in RON. The SF group showed a decrease in reasoning after training | Cognition and White Matter Integrity Positive correlation between change in thalamic AD and change in working memory performance in all participants (r = 0.44, p < 0.005). Negative correlation between changes in occipito-temporal AD and everyday problem solving (r = −0.32, p < .05) and spatial working memory accuracy (r = −0.35, p < .05). Negative correlation between changes in occipito-temporal-parietal AD and spatial working memory accuracy (r = −0.40, p < 0.05). Cognition & Functional Connectivity Positive correlation between changes in SPC-posterior ITL connectivity and changes in everyday problem solving time (r = −0.57, p < .001). |
Lövden et al. [23] 2010 | - | - | Mean Diffusivity (MD) Group X Time interaction found for segment 1 (genu) of corpus callosum, showing a decrease in MD (t(11) = 2.39, p = .036). No changes in control group Fractional Anisotropy (FA) Group X Time interaction found for segment 1 of corpus callosum, showing an increase in FA (t(11) = 3.12, p = .010) | Unknown: analysis combined younger and older subsets | Unknown: analysis combined younger and older subsets |
Antonenko et al. [20] 2016 | Hippocampal volume: no difference pre to post training (p = 0.505) | Mean Diffusivity (MD): A significant decrease in fornix MD was found at post-training compared with pre-training (p = 0.036). No difference in hippocampal MD from pre- to post-training (p = 0.669). Fractional Anisotropy (FA): A non-significant increase in fornix FA was found between pre- and post-training (p = 0.114) |
% Correct during training: Task performance significantly improved in a curvilinear convex manner over the 3 training days learning | - Higher increase in fornix FA from pre to post assessment was significantly related to better average recall performance on the object-location task during training, at 1-day post and follow-up (r = 0.431, p = 0.031) - Change in fornix FA did not correlate with episodic memory performance on the control task (Rey Auditory Verbal Learning Test; p = 0.214) - Change in fornix MD did not correlate with recall performance p = 0.728 - Change in hippocampal MD or volume did not correlate with recall performance (p = 0.688 and p = 0.758, respectively) | |
Heinzel et al. [22] 2014 | No significant change in grey matter volume of working memory network post training (t(14)= 0.83, p = .421) | No significant 2(time) ×3(working memory load) interaction (F = .24, p = .714, partial η2 = .024). Significant main effect of time (F = 12.68, p = .003, partial η2 = .475) driven by BOLD decrease in 1-back (t = .99, p = .029). | A 2(time)×3(load) repeated measures ANOVA showed no changes in connectivity in working memory network (F(2,28) = 1.08, p = .355, partial η2 = .071) |
n-Back: paired t-tests showed improved performance on 1-Back (t(18) = 3.37, p = .003), 2-ack (t(18) = 7.47, p < .001), and 3-Back (t(18) = 4.86, p < .001)d. Repeated-measures MANOVA (factor time) showed improvements in neuropsychological measures after training. Post hoc paired t-tests showed improvements in Digit Span Fwd (t(18) = 2.97, p = 0.008), D2 test (t(18) = 6.48, p < 0.001), Digit Symbol (t(18) = 2.76, p = 0.013), Stroop Interference (t(18) = 3.28, p = 0.004), and Figural Relations (t(18) = 4.73, p < 0.001). No improvements after training were found in Digit Span Bwd, Verbal Fluency, and Raven’s SPM.d
| Non-significant trend between BOLD activation at baseline and relative improvement in Digit Span Fwd (r = .43, p = .067) |
Structural imaging (n = 4)
Task-based fMRI (n = 3)
Connectivity
Resting-state fMRI (n = 5)
Structural connectivity (n = 4)
Correlation between imaging outcomes and cognitive function outcomes (n = 8)
Quality assessment of the included studies
Quality item | Suo et al. [19] 2016 | Rosen et al. [18] 2011 | Lampit et al. [16] 2015 | Belleville et al. [21] 2014 | Lin et al. [17] 2014 | Strenziok et al. [24] 2014 | Lövden et al. [23] 2010 | Antonenko et al. [20] 2016 | Heinzel et al. [22] 2014 |
---|---|---|---|---|---|---|---|---|---|
PEDro Scale Items | |||||||||
1 | + | + | + | + | + | − | + | + | − |
2 | + | + | + | + | + | + | − | − | − |
3 | + | + | + | + | + | − | − | − | − |
4 | − | + | + | + | + | + | + | − | − |
5 | + | + | + | − | − | − | − | − | − |
6 | + | − | − | − | − | − | − | − | − |
7 | + | + | + | − | + | − | − | − | − |
8 | + | + | − | + | − | + | + | + | + |
9 | + | + | − | + | − | + | + | + | + |
10 | + | + | + | − | − | + | + | − | − |
11 | + | + | + | + | + | + | + | + | + |
Additional Items | |||||||||
12 | + | + | + | + | + | + | + | + | + |
13 | + | − | − | − | − | − | − | − | − |
14 | + | − | − | − | − | − | − | − | − |