Computerized cognitive training in cognitively healthy older adults: a systematic review and meta-analysis of effect modifiers

Abstract

Background: New effective interventions to attenuate age-related cognitive decline are a global priority. Computerized cognitive training (CCT) is believed to be safe and can be inexpensive, but neither its efficacy in enhancing cognitive performance in healthy older adults nor the impact of design factors on such efficacy has been systematically analyzed. Our aim therefore was to quantitatively assess whether CCT programs can enhance cognition in healthy older adults, discriminate responsive from nonresponsive cognitive domains, and identify the most salient design factors.

Methods and findings: We systematically searched Medline, Embase, and PsycINFO for relevant studies from the databases' inception to 9 July 2014. Eligible studies were randomized controlled trials investigating the effects of ≥ 4 h of CCT on performance in neuropsychological tests in older adults without dementia or other cognitive impairment. Fifty-two studies encompassing 4,885 participants were eligible. Intervention designs varied considerably, but after removal of one outlier, heterogeneity across studies was small (I(2) = 29.92%). There was no systematic evidence of publication bias. The overall effect size (Hedges' g, random effects model) for CCT versus control was small and statistically significant, g = 0.22 (95% CI 0.15 to 0.29). Small to moderate effect sizes were found for nonverbal memory, g = 0.24 (95% CI 0.09 to 0.38); verbal memory, g = 0.08 (95% CI 0.01 to 0.15); working memory (WM), g = 0.22 (95% CI 0.09 to 0.35); processing speed, g = 0.31 (95% CI 0.11 to 0.50); and visuospatial skills, g = 0.30 (95% CI 0.07 to 0.54). No significant effects were found for executive functions and attention. Moderator analyses revealed that home-based administration was ineffective compared to group-based training, and that more than three training sessions per week was ineffective versus three or fewer. There was no evidence for the effectiveness of WM training, and only weak evidence for sessions less than 30 min. These results are limited to healthy older adults, and do not address the durability of training effects.

Conclusions: CCT is modestly effective at improving cognitive performance in healthy older adults, but efficacy varies across cognitive domains and is largely determined by design choices. Unsupervised at-home training and training more than three times per week are specifically ineffective. Further research is required to enhance efficacy of the intervention. Please see later in the article for the Editors' Summary.

Figure 1. Summary of trial identification and selection.

Note that a single study could be excluded on more than one criterion, but appears only once in the chart.

Figure 2. Overall efficacy of CCT on all cognitive outcomes.

Effect estimates are based on a random-effects model, and studies are rank-ordered by year of publication.

Figure 3. Funnel plot for overall effects after removal of one outlier .

Figure 4. Efficacy of CCT on measures of verbal memory.

Effect estimates are based on fixed-effects (top) and random-effects (bottom) models, and studies are rank-ordered by year of publication.

Figure 5. Efficacy of CCT on measures of nonverbal memory.

Effect estimates are based on a random-effects model, and studies are rank-ordered by year of publication.

Figure 6. Efficacy of CCT on measures of working memory.

Effect estimates are based on a random-effects model, and studies are rank-ordered by year of publication.

Figure 7. Efficacy of CCT on measures of processing speed.

Effect estimates are based on fixed-effects (top) and random-effects (bottom) models, and studies are rank-ordered by year of publication.

Figure 8. Efficacy of CCT on measures of executive functions.

Effect estimates are based on a random-effects model, and studies are rank-ordered by year of publication.

Figure 9. Efficacy of CCT on measures of attention.

Effect estimates are based on a random-effects model, and studies are rank-ordered by year of publication.

Figure 10. Efficacy of CCT on measures of visuospatial skills.

Effect estimates are based on a random-effects model, and studies are rank-ordered by year of publication.

Figure 11. Subgroup analyses of moderators of overall efficacy of CCT in older adults.

^a Q-test for between-group heterogeneity, mixed-effects model. ^bOne study that combined data from both home- and group-based training was excluded from this analysis. ^cTotal number of training hours. ^dSession length could not be determined for one study.

Figure 12. Overview of efficacy and moderators of efficacy for CCT in older adults.

Numbers refer to SMDs from an individual meta-analysis (see Figures S2, S3, S4, S5, S6, S7, S8 for details). Colored cells indicate significant outcomes, with effect sizes color coded: yellow, g<0.3; pink, g = 0.3–0.6; red, g≥0.6. White depicts non-significant results, and grey shows where no studies were available for analysis. *p<0.05, **p<0.01 for within-subgroup results (between-subgroup results are reported in Figures 11 and S2, S3, S4, S5, S6, S7, S8). ^aBased on a fixed-effects model because of evidence of potential publication bias in these outcomes. ^bSMD based on a single trial.