TY - JOUR
AU - Zhang,, Tong
AB - Abstract Objectives The long-lasting efficacy of working memory (WM) training has been a controversial and still ardently debated issue. In this meta-analysis, the authors explored the long-term effects of WM training in healthy older adults on WM subdomains and abilities outside the WM domain assessed in randomized controlled studies. Method A systematic literature search of PubMed, Web of Science, PsycINFO, Cochrane Library, ProQuest, clinicaltrials.gov, and Google Scholar was conducted. Random-effects models were used to quantitatively synthesize the existing data. Results Twenty-two eligible studies were included in the meta-analysis. The mean participant age ranged from 63.77 to 80.1 years. The meta-synthesized long-term effects on updating were 0.45 (95% confidence interval = 0.253–0.648, <6 months: 0.395, 0.171–0.619, ≥6 months: 0.641, 0.223–1.058), on shifting, 0.447 (0.246–0.648, <6 months: 0.448, 0.146–0.75, ≥6 months: 0.446, 0.176–0.716); on inhibition, 0.387 (0.228–0.547, <6 months: 0.248, 0.013–0.484, ≥6 months: 0.504, 0.288–0.712); on maintenance, 0.486 (0.352–0.62, <6 months: 0.52, 0.279–0.761, ≥6 months: 0.471, 0.31–0.63). Discussion The results showed that WM training exerted robust long-term effects on enhancing the WM system and improving processing speed and reasoning in late adulthood. Future studies are needed to use different tasks of the same WM construct to evaluate the WM training benefits, to adopt more ecological tasks or tasks related to daily life, to improve the external validity of WM training, and to identify the optimal implementation strategy for WM training. Long-term effect, Transfer effect, Working memory, Older adults Intact cognitive functioning is a critical component of well-being, autonomy, and successful aging (Rowe & Kahn, 2015; Tesch-Römer & Wahl, 2017), but a growing body of evidence suggests that aging in the human brain (occurring mainly in the prefrontal cortex and the medial temporal lobe system) is associated with cognitive declines (Bettio, Rajendran, & Gil-Mohapel, 2017) in several cognitive domains, such as spatial learning (Konishi, McKenzie, Etchamendy, Roy, & Bohbot, 2017), working memory (WM; Keller et al., 2015), executive function (Kirova, Bays, & Lagalwar, 2015), and episodic memory (Zheng et al., 2018). Aging-related cognitive decline and impairment affect older adults’ life expectancy, quality of life, and well-being (MacDonald, Hultsch, & Dixon, 2011; Wilson et al., 2013), causing great public health and economic concern worldwide (Jia et al., 2018; Lin & Neumann, 2013). As the incidence of cognitive impairment increases (Gillett et al., 2018), interventions for postponing or preventing cognitive decline are raising great research interest. Converging lines of evidence have indicated that engagement in cognitively stimulating activities may reduce the risk of cognitive deterioration and dementia (Daffner, 2010; Marioni et al., 2012; Verghese et al., 2003; Wilson et al., 2002), and there has been a growing interest in investigating the effect of WM training in healthy older adults over decades because researchers have discovered that WM is involved in other cognitive domains such as fluid intelligence (Borella, Carbone, Pastore, De Beni, & Carretti, 2017; Borella et al., 2014; Cantarella et al., 2017; Jaeggi, Buschkuehl, Jonides, & Perrig, 2008; Zinke et al., 2014), processing speed (Borella et al., 2014; Borella, Carretti, Riboldi, & De Beni, 2010; Cantarella et al., 2017; Jaeggi et al., 2019; Sandberg & Stigsdotter Neely, 2015), episodic memory (Borella, Carretti, et al., 2019; Du, Ji, Chen, Tang, & Han, 2018; Jaeggi et al., 2019; Toril, Reales, Mayas, & Ballesteros, 2016; Weicker et al., 2018), and its vulnerability to declining with age (Borella, Carretti, & De Beni, 2008). Baddeley and colleagues suggested one of the most influential models by several behavioral experiments and hypothesized that WM might consist of separate storage buffers for verbal and nonverbal stimuli as well as a central executive component in information management (Baddeley, 2003; Baddeley & Hitch, 1974; Hitch, Allen, & Baddeley, 2020). What makes the WM function properly work rely on the central executive process that actively operates the representations (Bledowski, Rahm, & Rowe, 2009). Accordingly, a number of researchers have sought to describe and organize the central executive function (Miyake et al., 2000; Nee et al., 2013). Miyake and colleagues (2000) adopted factor analysis to investigate the latent component of the commonly used executive function tasks and found that the executive function system was a separable but moderately connected system including updating, shifting, and inhibition subcomponents. Later, Nee and colleagues (2013) followed the Miyake’s categorization model and conducted a functional magnetic resonance imaging (FMRI) meta-analysis to identify the brain regions contributing each subcomponent. They found that all three components had shared activation and component-specific regions, and thus concluded that WM should be considered a dissociable but related system. We further extended this idea to WM training among healthy older adults to identify which component of WM was trainable and to what extent. The typical protocol used in WM training includes activities that involve practicing complex WM tasks to train WM-related mechanisms and processes (including passive maintenance, updating, shifting, and resistance; Nee et al., 2013), without any strategy being taught. Due to mixed results reported by original articles, the effects of WM training among healthy older adults have been investigated in several meta-analytic studies over the years. Later, we summarize the most important meta-analyses exploring the effects of WM training in healthy older adults in chronological order. To date, a total of four meta-analyses have explored the long-term effects in healthy older adults (Nguyen, Murphy, & Andrews, 2019; Sala, Aksayli, Tatlidil, Gondo, & Gobet, 2019; Schwaighofer, Fischer, & Bühner, 2015; Teixeira-Santos et al., 2019) and resulted in mixed results. Of the four meta-analyses, three studies have found small but significant effect sizes either in WM or transfer tasks (Nguyen et al., 2019; Sala et al., 2019; Teixeira-Santos et al., 2019), whereas one study failed to find any significant effects during follow-up visit (Schwaighofer, Fischer, & Bühner, 2015). In addition, the limitations of previous meta-analyses concerning the long-term effects should not be ignored: (a) lack of moderator analyses on long-term effects; (b) no clear definition of long-term effects (e.g., follow-up interval); (c) no clear WM theoretic model (e.g., subcomponents in WM); and (d) methodological inappropriateness (e.g., without considering different type of randomized controlled trials [RCTs] and drop-out rates). Thus, none of these meta-analyses have reached a clear conclusion on this topic. Detailed, Lampit, Hallock, and Valenzuela (2014) conducted a meta-analysis including all types of computerized cognitive training, nine of which were computerized WM training in older adults without dementia or other cognitive impairment. They found that small-to-medium near-transfer effects were obtained immediately after training, whereas almost no far-transfer effects were detected. However, they did not report the long-term effects and not all WM interventions were computer based. In the same year, Karbach and Verhaeghen (2014) conducted a meta-analysis comprising 13 studies to examine process-based WM training and found significant transfer effects on both near- and far-transfer tasks. They also did not mention the long-term effects. However, Melby-Lervag, Redick, and Hulme (2016) carried out a validation meta-analysis and found that the far-transfer effect diminished when considering the placebo effect of treated controls. In Schwaighofer, Fischer, and Bühner (2015), a meta-analysis comprising 47 studies in which 10 concerned older adults, they also found significant long-term effects on WM outcomes, but no sustainable long-term transfer effects. However, the small number of studies concerning the long-term effects in older adults limited the reliability of the targeting results. Mewborn, Lindbergh, and Stephen Miller (2017) adopted a multilevel meta-analysis to examine the training and transfer effects for all types of cognitive training including 16 WM training. However, they reported only one composite score of WM training on near- and far-transfer effects (g = 0.479). Thus, it was difficult to determine which test or component contributed to the overall effect and to what extent. Nguyen and colleagues (2019) focused on all types of computerized cognitive training including WM training, in cognitively healthy, mildly impaired, and mixed samples of older adults. Their results highlighted that the effect sizes from immediate and long-term efficacy analyses were comparable, suggesting that computerized cognitive training (CCT) gains could be maintained over time. However, analyses for long-term effects were not conducted at the level of WM subdomains due to limited comparisons (only three WM training). Sala and colleagues (2019) adopted a rigorous sensitivity analysis to control for statistical dependence of effect sizes, publication bias, and influential cases in their meta-analysis and found a near-zero effect size in the far-transfer task (g = 0.121) and only a modest effect size (g = 0.274) in post-test. When considering the long-term effects, they found significant effect sizes for the criteria task (g = 1.106), near-transfer task (g = 0.378), and far-transfer task (g = 0.241). However, they did not run any sensitivity or moderator analysis. The interpretation for the results of long-term effects should be cautious due to the lack of sensitivity analyses and moderator analyses. Finally, Teixeira-Santos and colleagues (2019) conducted the most technically sound meta-analysis so far by combining 10 studies that had reported long-term effects and observed small significant and long-lasting transfer gains in WM tasks (g = 0.23), and very small and only marginally significant effects in reasoning tasks (g = 0.13). Additionally, no prior meta-analytic studies have verified the potential moderators of WM long-term effects. However, some original RCTs have examined the moderating effect on potential factors on long-term effects (Borella et al., 2014; Borella, Carretti, Zanoni, Zavagnin, & De Beni, 2013; Brum, Borella, Carretti, & Sanches Yassuda, 2018; Jaeggi et al., 2019). Borella and colleagues (2013) found age-related differences in the long-term effects of the criteria tasks (old-old [Cohen’s d = 1.44] with young-old [Cohen’s d = 2.01]), but no significant difference in transfer tasks except the Cattell task and processing speed tasks, indicating that the moderating effect of age might be task dependent. Brum and colleagues (2020) found that training sessions might moderate the long-term effects by comparing shorter sessions (n = 3) versus longer sessions (n = 6). Specifically, the longer training protocol generated higher follow-up gains for matrix reasoning, the Stroop test, and the backward digital span, whereas the shorter training protocol showed a higher follow-up gain in letter-number sequencing, indicating that the moderation of training sessions might also be task dependent. Jaeggi and colleagues (2019) conducted a 2 × 3 RCT (training frequency (everyday, twice per day, and every other day vs type of training [WM training vs knowledge training]) to explore whether training frequency moderates the long-term effects, but they failed to find any moderating effects of training frequency on the outcomes. On the other hand, many studies reporting the association between the potential factors and immediate training effects have suggested that variables such as education (Borella, Carretti, et al., 2017), type of control (Melby-Lervag et al., 2016), training frequency (Jaeggi et al., 2019), and age (Burki, Ludwig, Chicherio, & de Ribaupierre, 2014; Zajac-Lamparska & Trempala, 2016) might moderate the immediate training effect. To avoid missing the potential moderators, we thus aimed to consider all available moderators in the association between WM training and long-term effects. Regarding the transfer effects, most of the previous meta-analyses concerning long-term effects of WM divided the tasks into criteria, near-transfer, and far-transfer tasks (Sala et al., 2019; Schwaighofer, Fischer, & Bühner, 2015), whereas Teixeira-Santos and colleagues (2019) further divided the included tests based on cognitive constructs. It is difficult to draw consistent conclusions by the first method due to the heterogeneity of the criteria, near-transfer, and far-transfer tasks in the original studies and the different definitions of transfer effects in each meta-analysis. We thus adopt a more replicable method herein with reference to Teixeira-Santos and colleagues (2019) and divide the non-WM tasks according to their cognitive constructs. Accordingly, reasoning, episodic memory, and processing speed are adopted as transfer tasks or secondary outcomes because they have been the most commonly reported outcomes in previous reports and have close relationship with WM (Conway, Kane, & Engle, 2003; Melrose et al., 2020; Monge & Madden, 2016). Considering the limitations of previous meta-analyses, it is urgent to conduct a comprehensive meta-analysis to address these issues. In addition, there has been no definite agreement as to whether WM training in healthy older adults has stable long-term maintenance effects in each sub-WM component. The present study aims to (a) evaluate the long-term effects of WM training on the tasks of maintenance, updating, shifting, and inhibition and compare the long-term effects between four WM subcomponents; (b) evaluate the long-term effects of WM training on the tasks of episodic memory, processing speed, and reasoning; (c) explore potential factors that moderate the long-term effects; and (d) assess the drop-out rates between training and control groups. Method The review was preregistered in the International Prospective Register of Systematic Reviews (PROSPERO, http://www.cdr.york.ac.uk/prospero, CRD42019122061). The work was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA; Moher, Liberati, Tetzlaff, & Altman, 2010) guidelines (see RPISMA checklist, Supplementary Appendix Table A). Search Strategy Comprehensive searches were conducted in PubMed, Web of Science, PsycINFO, and Cochrane library with no publication date limit. Search terms were a combination of training-related terms (working memory OR n-back training OR cognitive intervention OR dual-tasking OR switching OR shifting OR memory updating OR inhibition OR resistance OR maintenance OR process-based training) and population-related terms (aging OR aged OR older OR elder OR old). Supplementary searches were also conducted in Google scholar and clinicaltrials.gov. All searches were limited to peer-reviewed publications. The full searching terms are shown in Supplementary Appendix Table B. Selection Criteria Studies were considered as eligible if they met the following inclusion criteria: (a) cluster/block designed or individualized parallel RCT, (b) interventional studies that focus on WM, (c) reporting pretraining and follow-up assessment on at least one outcome measure on the WM domain, (d) including healthy participants older than 60 years, (e) reporting adequate information to calculate effect sizes for at least one outcome measure on the WM domain, and (f) peer-reviewed journal articles. Studies were identified as ineligible if they were (a) case reports, (b) review articles or theoretical articles, (c) mixed with physical exercise, nutrition intervention, neurostimulation, and/or neuro-feedback training, and (d) quasi- and cross-over RCT. Two authors (J.H. and R.S.) independently screened results based on titles and abstracts. The remaining articles were further reviewed by full-text assessment. Disagreements between reviewers about eligibility were resolved by discussion with T.J. The full-text articles (n = 58) were excluded because of (a) not having long-term assessments (n = 24), (b) not being RCT design (n = 21), (c) not having WM-related outcomes (n = 4), (d) not having comparable groups (n = 1); (e) having the same data set published in previous paper (n = 1); and (f) having mixed with young adults or mild cognitive impairment (MCI) (n = 2) and having mixed WM training with physical exercise (n = 5). The detailed selection process is shown in Figure 1. In addition, the detailed exclusion reasons for full texts are also shown in Supplementary Appendix Table C. Figure 1. Open in new tabDownload slide The flow chart of the included studies. Figure 1. Open in new tabDownload slide The flow chart of the included studies. Coding and Data Analysis The type of group was categorized into three levels: “passive control group”; “active control group”; and “training group.” Participants in the passive control group received no intervention and only participated in pre-assessment, post-assessment, and follow-up assessment. We thus included the “no-contact group” and “waiting-list group” in this category. Participants in the active control group received the same amount of non-WM-related training or extremely low-load WM training and maintained the same level of expected malleable ability as those in the training group. The training group included any types of repetitive practice on WM tasks. However, WM training mixed with exercise, neurostimulation, music, and strategy-based training was excluded due to the confounding impact of these types of training. The training dose was defined by the total training sessions and total training hours. We dichotomized two variables based on the median. Training effects included four separable but related WM subcomponents representing the passive maintenance and active manipulation of representations (Courtney, Sala, & Sala, 2007; Miyake et al., 2000; Niendam et al., 2012). We included all tasks that mainly rely on one of the WM subcomponents. Maintenance refers to the storage of representations in passive or extremely low-load demand. In this category, we included all simple span tasks (e.g., digital span, letter span, word span, and block span). The update process refers to monitoring and coding incoming information and then replacing old or no-longer relevant information with new or more task-relevant information. In this case, we included running memory, the keep-trace task, the paced auditory serial addition task, n-back without lure, and the Sternberg task. The inhibition component is defined as the process mitigating interferences from external or environmental stimuli and from intrusive memory. Tasks that can be considered as inhibition included the Stroop task, n-back with lure, intrusive error in Categorization Working Memory Span (CWMS), Flanker task, go/no-go task, and D2 task. In addition, we used the outcomes representing the inhibition process (e.g., reaction time or accuracy differences of conflict condition minus congruence condition). In case these were not available, we included the condition that involved conflicting stimuli. The shifting or switching process refers to the process shifting back and forth between representations and tasks. We thus included both representation-level shifting (e.g., local-global task, TMT-B, plus–minus task, and letter-number sequencing) and task-level shifting (e.g., task-switching paradigm). In addition, we used the outcomes representing switching or shifting cost (e.g., differences in reaction time or accuracy in the shifting condition vs the nonshifting condition). In addition, if the included tasks reported both accuracy and reaction time outcomes, accuracy was chosen as the priority outcome. Transfer effects were evaluated by processing speed, episodic memory, and reasoning tasks. The processing-speed theory of aging hypothesizes that aging is associated with decreased speed in many processing operations (like a CPU) and this reduction in speed leads to ongoing impairment of higher-order cognitive function (Salthouse, 1996). We thus included the time-constrained or demanding tasks that heavily rely on mental speed (e.g., TMT-A, pattern separation, number cancelation, and digital-symbol substitution). Episodic memory is also an aging-sensitive cognitive domain and is defined as a uniquely different memory system that enables participants to remember past experience within their temporal and spatial context (Tromp, Dufour, Lithfous, Pebayle, & Després, 2015; Tulving, 2002). WM has been shown to explain individual differences in reasoning ability (Dehn, 2017). Reasoning was grouped according to the tests adopted to evaluate reasoning ability (e.g., Cattell test, all versions of the Raven reasoning test, and the syllogisms task). Two authors independently extracted data from each study. The summary statistics of each outcome were the means, SD, and number of participants at randomization. The long-term effect was defined as the maintenance of training effects over time measured by the following equation and was dichotomized by the median months of follow-up assessment after training (median = 6 months). Cohen′s d=M(mean change in the TG)−M(mean change in the CG)Pooled SD Pooled SD=(N1−1)∗SD12+(N2−1)∗SD22N1+N2−2 where TG, training group; CG, control group; N1, sample size in TG; N2, sample size in CG; SD1: standard deviation of the mean change in the TG; SD2: standard deviation of the mean change in the CG; mean change, the difference between the follow-up and pretraining assessments. Random-effects models were used to estimate the pooled estimation of effect sizes. Considering different types of RCTs, cluster RCTs were adjusted by calculating “effective sample size (ESS)” (Rao & Scott, 1992; White & Thomas, 2005). Intracluster correlation (ICC) describes the similarity of individuals within the same cluster (Eldridge, Ukoumunne, & Carlin, 2009) and was estimated by external similar studies (Campbell, Grimshaw, & Steen, 2000). ESS=ASS1+(M−1)∗ICC where ASS, actual sample size; M, cluster or block size. The variation in effect sizes across studies was assessed by the homogeneity statistic, Q. The I2 statistic was also calculated to estimate the proportion of true heterogeneity in the observed variance. Egger’s regression intercept test was used to estimate publication bias. Study Quality Assessment Two independent reviewers used a modified Physiotherapy Evidence Database (PEDro) scale (maximum score = 9) to assess the included study quality (Lampit, Hallock, & Valenzuela, 2014; Maher, Sherrington, Herbert, Moseley, & Elkins, 2003; Zhu, Yin, Lang, He, & Li, 2016). Each item scored 1 point if the original studies fully met the criteria and scored zero if the original studies partially met, unclearly met, or did not meet the criteria. Studies that scored 7 or higher were considered as having a low risk of bias, whereas others that scored 6 or lower were considered as having a high risk of bias. In addition, we compared high-quality studies with low-quality studies and checked the robustness of each outcome. Moderator Analysis For moderator analysis, a subgroup meta-analysis was performed for significant outcomes. The moderators included age of participants (60–70 years vs 70 years or older), total sessions (<10 sessions vs ≥10 sessions), total hours (<10 hr vs ≥10 hr), type of control (active vs passive), frequency (<3 sessions per week vs ≥3 sessions per week), and study quality (low vs high). Follow-up assessment after training was also dichotomized into shorter follow-up (<6 months) and longer follow-up (>6 months). Other moderators were excluded due to heterogeneity of measurement (e.g., years of education) or limited comparisons (e.g., baseline depression). We performed all analyses in Comprehensive Meta-analysis version 2 (CMA v2). Results Study Characteristics of the Included Studies Altogether, 22 eligible studies comprising 26 comparisons were included in the meta-analysis, comprising 643 participants in the WM groups and 633 in the control groups (Appendix Table D). The mean participant age ranged from 63.77 to 80.1 years (median age = 69.57). The sample size ranged from 30 to 183 (median sample size = 40). Eighteen of 22 studies made a two-group comparison. One study (Borella et al., 2014) used a four-group design to compare the effectiveness of WM between old-old individuals and young-old individuals. We separately coded this study into two comparisons (old-old training group vs old-old control group and young-old training group vs young-old control group). Borella, Carretti, and colleagues (2019) also adopted a four-group comparison (Mozart + WM, Albinoni + WM, white noise + WM, and active control) exploring the added effects of music before WM training. To minimize the effects of music enhancement, we only compared the white noise + WM group and the active control group. One study (Weicker et al., 2018) allocated healthy older adults into three groups (WM intervention group, active low-level intervention group, and passive control group), and we separately coded this study into two comparisons (WM training group vs active control group and WM training group vs passive control group). In one study (Brum, Borella, Carretti, & Sanches Yassuda, 2018), two experiments (WM training with shorter sessions vs active control and WM training with longer sessions vs active control) were separately coded as two comparisons. Cantarella and colleagues (2017) conducted two RCTs to explore the emotional valence (neutral vs positive) and training effects. Of 26 comparisons, 4 comparisons used passive control groups and 22 comparisons used various active control groups. The control activities included health promotion education, completion of questionnaires, very low-level practice, pseudo-training, and eccentric muscle relaxation. Regarding training tasks, seven studies adopted CWMS, three studies employed n-back, two studies used matrix tasks, one study used running memory tasks, one study used set shifting training, and most studies exploited multiple tasks, including executive function training that focused on mental set shifting, WM, and inhibition; card games that were based on storage, manipulation, and selective attention; WM-focused video games; WM tasks that required verbal WM, visuospatial WM, and executive control processes; adaptive verbal and nonverbal WM tasks; square sequence and animal-picture shifting; and item-memory tasks and keep-trace tasks. The total number of training sessions ranged from 3 to 25 (median = 10 sessions), and the total training hours ranged from 1.5 to 17.25 (median = 10 hr). The training frequency ranged from 1.5 to 5 sessions per week. The pre- to follow-up interval ranged from 3 to 18 months. The study quality ranged from 5 to 8 (median = 7). Eight studies (36.4%) were considered to have a low risk of bias. Over 90% of the studies (20/22) did not carry out concealed allocation except for the studies by Chiu and colleagues (2018) and Borella and colleagues (2013). Over 70% of the studies (17/22) did not adopt accessor blindness (Carretti, Borella, Zavagnin, & de Beni, 2013; Borella, Cantarella, Carretti, De Lucia, & De Beni, 2019; Borella, Carbone, et al., 2017; Borella et al., 2013, 2014; Buschkuehl et al., 2008; Cantarella et al., 2017; Dahlin, Nyberg, Bäckman, & Neely, 2008; Du et al., 2018; Gronholm-Nyman et al., 2017; Jaeggi et al., 2019; Matysiak, Kroemeke, & Brzezicka, 2019; McAvinue et al., 2013; Sandberg & Stigsdotter Neely, 2015; Toril et al., 2016; Zinke et al., 2014), and approximately 60% (13/22) of the studies lost over 15% of the participants at randomization (Brum et al., 2020; Buschkuehl et al., 2008; Chiu et al., 2018; Dahlin et al., 2008; Du et al., 2018; Jaeggi et al., 2019; McAvinue et al., 2013; Sandberg & Stigsdotter Neely, 2015; Zinke et al., 2014). In addition, one study did not use intention-to-treat (ITT) or modified ITT analyses (Buschkuehl et al., 2008) and another did not have similar baseline on key indicators (Sandberg & Stigsdotter Neely, 2015). Detailed information for study quality is shown in Supplementary Appendix Table E. Long-Term Effects on WM Subdomains Updating Eight comparisons comprising 482 participants (Carretti et al., 2013; Dahlin et al., 2008; Du et al., 2018; Jaeggi et al., 2019; Matysiak et al., 2019; Sandberg & Stigsdotter Neely, 2015; Weicker et al., 2018) reported outcomes with respect to updating. The meta-synthesized long-term effects on updating were 0.45 (95% confidence interval [CI] = 0.253 to 0.648, p < .001; <6 months: 0.395, 95% CI = 0.171 to 0.619, ≥6 months: 0.641, 95% CI = 0.223 to 1.058; Figure 2A). No significant heterogeneity was detected (Q(7) = 3.498, I2 = 0, p = .835). In addition, no significant publication bias was detected (intercept = 0.243, 95% CI = −1.584 to 2.07, p = .345; Figure 3A). Figure 2. Open in new tabDownload slide The forest plot for working memory subcomponents. (A) The pooled effect size for updating. (B) The pooled effect size for shifting. (C) The pooled effect size for inhibition. (D) The pooled effect size for maintenance. Figure 2. Open in new tabDownload slide The forest plot for working memory subcomponents. (A) The pooled effect size for updating. (B) The pooled effect size for shifting. (C) The pooled effect size for inhibition. (D) The pooled effect size for maintenance. Figure 3. Open in new tabDownload slide The funnel plot for working memory subcomponents. (A) The publication bias for updating. (B) The publication bias for shifting. (C) The publication bias for inhibition. (D) The publication bias for maintenance. Figure 3. Open in new tabDownload slide The funnel plot for working memory subcomponents. (A) The publication bias for updating. (B) The publication bias for shifting. (C) The publication bias for inhibition. (D) The publication bias for maintenance. Shifting Nine comparisons comprising 437 participants (Brum et al., 2020; Chiu et al., 2018; Gronholm-Nyman et al., 2017; Matysiak et al., 2019; McAvinue et al., 2013; Sandberg & Stigsdotter Neely, 2015; Weicker et al., 2018) reported outcomes for shifting. The composite long-term effects on shifting were 0.447 (95% CI = 0.246 to 0.648, p < .001; <6 months: 0.448, 95% CI = 0.146 to 0.75, ≥6 months: 0.446, 95% CI = 0.176 to 0.716, Figure 2B). No significant heterogeneity was detected (Q(8) = 4.815, I2 = 0, p = .777). Moreover, no significant publication bias was found (intercept = −0.519, 95% CI = −3.864 to 2.826, p = .725; Figure 3B). Inhibition Fourteen comparisons comprising 768 participants (Borella, Cantarella, et al., 2019; Borella et al., 2010, 2013, 2014; Brehmer, Westerberg, & Bäckman, 2012; Brum et al., 2018; Chiu et al., 2018; Jaeggi et al., 2019; Matysiak et al., 2019; Sandberg & Stigsdotter Neely, 2016; Weicker et al., 2018) reported measurements on inhibition. The comprised long-term effects on inhibition were 0.387 (95% CI = 0.228 to 0.547, p < .001; <6 months: 0.248, 95% CI = 0.013 to 0.484, ≥6 months: 0.504, 95% CI = 0.288 to 0.712, Figure 2C). No significant heterogeneity was found (Q(13) = 15.703, I2 = 17.215, p = .266). However, significant publication bias was detected (intercept = 1.478, 95% CI = −0.829 to 3.785, p = .188; Figure 3C). Maintenance Twenty-two comparisons comprising 968 participants (Borella, Cantarella, et al., 2019; Borella et al., 2010, 2013, 2014; Borella, Carretti, Meneghetti, et al., 2017; Borella, Carretti, Sciore, Capotosto, Taconnat, Cornoldi, & De Beni, 2017; Brehmer et al., 2012; Brum et al., 2018; Buschkuehl et al., 2008; Cantarella et al., 2017; Carretti et al., 2013; Chiu et al., 2018; Dahlin et al., 2008; Du et al., 2018; Matysiak et al., 2019; McAvinue et al., 2013; Sandberg & Stigsdotter Neely, 2016; Toril et al., 2016; Weicker et al., 2018; Zinke et al., 2014) reported measurements on maintenance. The composite long-term effects on maintenance were 0.486 (95% CI = 0.352 to 0.62, p < .001; <6 months: 0.52, 95% CI = 0.279 to 0.761, ≥6 months: 0.471, 95% CI = 0.31 to 0.63, Figure 2D). No significant heterogeneity was found (Q(21) = 13.375, I2 = 0, p = .895). In addition, no significant publication bias was detected (intercept = 0.506, 95% CI = −1.568 to 2.581, p = .616; Figure 3D). Comparison between WM subcomponents No between-group difference was detected (Q(3) = 0.857, p = .836). When stratified by follow-up period after training, no between-group differences were detected either for the shorter follow-up or for the longer follow-up (all ps > .05). Long-Term Effects on Other Cognitive Domains Episodic memory Twelve comparisons (Borella, Cantarella, et al., 2019; Brehmer et al., 2012; Buschkuehl et al., 2008; Dahlin et al., 2008; Du et al., 2018; Jaeggi et al., 2019; McAvinue et al., 2013; Sandberg & Stigsdotter Neely, 2016; Toril et al., 2016; Weicker et al., 2018) reported measurements on episodic memory. The composite long-term effects of episodic memory were 0.16 (95% CI = −0.015 to 0.335, p = .074; <6 months: 0.182, 95% CI = −0.029 to 0.392, ≥6 months: 0.111, 95% CI = −0.205 to 0.427, Figure 4A). No significant heterogeneity was detected (Q(11) = 3.254, I2 = 0, p = .985). Furthermore, no significant publication bias was found (intercept = 0.605, 95% CI = −0.544 to 1.754, p = .268; Figure 5A). Figure 4. Open in new tabDownload slide The forest plot for other cognitive domains. (A) The pooled effect size for episodic memory. (B) The pooled effect size for processing speed. (C) The pooled effect size for reasoning. Figure 4. Open in new tabDownload slide The forest plot for other cognitive domains. (A) The pooled effect size for episodic memory. (B) The pooled effect size for processing speed. (C) The pooled effect size for reasoning. Figure 5. Open in new tabDownload slide The funnel plot for other cognitive domains. (A) The publication bias for episodic memory. (B) The publication bias for processing speed. (C) The publication bias for reasoning. Figure 5. Open in new tabDownload slide The funnel plot for other cognitive domains. (A) The publication bias for episodic memory. (B) The publication bias for processing speed. (C) The publication bias for reasoning. Processing speed Eleven comparisons (Borella et al., 2010, 2013, 2014; Borella, Carretti, et al., 2017; Brum et al., 2018; Cantarella et al., 2017; Dahlin et al., 2008; Du et al., 2018; Jaeggi et al., 2019; McAvinue et al., 2013; Sandberg & Stigsdotter Neely, 2016; Weicker et al., 2018) reported measurements of processing speed. The composite long-term effects on processing speed were 0.242 (95% CI = 0.065 to 0.419, p = .007; <6 months: 0.158, 95% CI = −0.137 to 0.454, ≥6 months: 0.288, 95% CI = 0.067 to 0.509, Figure 4B). No significant heterogeneity was detected (Q(10) = 5.211, I2 = 0, p = .877). Furthermore, no significant publication bias was found (intercept = 0.746, 95% CI = −0.923 to 2.416, p = .338; Figure 5B). Reasoning Nineteen comparisons (Borella et al., 2010, 2013, 2014; Borella, Carretti, Meneghetti, et al., 2017; Borella, Carretti, et al., 2017; Brehmer et al., 2012; Brum et al., 2018; Cantarella et al., 2017; Carretti et al., 2013; Dahlin et al., 2008; Du et al., 2018; Matysiak et al., 2019; Sandberg & Stigsdotter Neely, 2016; Weicker et al., 2018; Zinke et al., 2014) reported outcomes for reasoning. The pooled long-term effects on reasoning were 0.148 (95% CI = 0.006 to 0.291, p = .041, <6 months: 0.17, 95% CI = −0.107 to 0.448, ≥6 months: 0.14, 95% CI = −0.026 to 0.306). No significant heterogeneity was found (Q(18) = 15.247, I2 = 0, p = .645). Moreover, no significant publication bias was detected (intercept = −1.149, 95% CI = −3.728 to 1.429, p = .36). Subgroup Analyses of Potential Moderators Subgroup analyses were conducted among those significant outcomes in four WM components and two transfer cognitive domains (processing speed and reasoning). Lower training frequency and fewer total sessions were associated with higher maintenance on the inhibition task (p = .007 and p = .028, respectively). No other moderation effects were detected. The detailed results for subgroup analyses are given in Table 1 and Supplementary Appendix Table F. Table 1. Subgroup Analyses of Potential Moderators . . Maintenance . Updating . Shifting . Inhibition . Processing speed . Reasoning . Age 60–70 years old 0.5*** (0.328 to 0.672) 0.433** (0.167 to 0.699) 0.394** (0.132 to 0.655) 0.275* (0.052 to 0.499) 0.398** (0.139 to 0.656) 0.18* (0.016 to 0.344) 70 and older 0.465*** (0.252 to 0.678) 0.471** (0.171 to 0.765) 0.477** (0.124 to 0.829) 0.514*** (0.279 to 0.749) 0.107 (−0.139 to 0.346) 0.052 (−0.236 to 0.34) Training frequency 0–3 sessions per week 0.51*** (0.289 to 0.731) 0.939** (0.25 to 1.629) NA 0.735***(0.438 to 1.033) 0.318* (0.074 to 0.563) 0.11 (−0.119 to 0.338) ≥3 sessions per week 0.473*** (0.304 to 0.641) 0.407*** (0.201 to 0.603) 0.447*** (0.246 to 0.648) 0.262**(0.089 to 1.435) 0.158 (−0.099 to 0.414) 0.173 (−0.009 to 0.355) Training sessions 0–10 sessions 0.446*** (0.337 to 0.709) 0.939** (0.25 to 1.629) 0.276 (−0.131 to 0.685) 0.582***(0.341 to 0.822) 0.318* (0.074 to 0.563) 0.138 (−0.037 to 0.314) ≥10 sessions 0.523*** (0.253 to 0.639) 0.407*** (0.201 to 0.603) 0.501*** (0.27 to 0.733) 0.256**(0.065 to 0.447) 0.158 (−0.099 to 0.414) 0.167 (−0.077 to 0.411) Total hours 0–10 hr 0.498*** (0.34 to 0.656) 0.448*** (0.228 to 0.668) 0.394** (0.132 to 0.655) 0.4*** (0.24 to 0.561) 0.25* (0.056 to 0.443) 0.153* (0.003 to 0.304) ≥10 hr 0.456*** (0.202 to 0.71) 0.46* (0.013 to 0.906) 0.524** (0.209 to 0.84) 0.261 (−0.151 to 0.672) 0.2 (−0.238 to 0.639) 0.105 (−0.333 to 0.543) Type of control Passive 0.536*** (0.326 to 0.745) 0.467*** (0.245 to 0.689) 0.256 (−0.066 to 0.579) 0.383* (0.088 to 0.677) 0.164 (−0.183 to 0.511) 0.091 (−0.127 to 0.308) Active 0.452*** (0.278 to 0.625) 0.388 (−0.044 to 0.819) 0.569*** (0.311 to 0.827) 0.376*** (0.17 to 0.583) 0.269* (0.063 to 0.475) 0.192* (0.003 to 0.38) Study quality High 0.489*** (0.323 to 0.655) 0.531* (0.107 to 0.956) 0.455 (0.199 to 0.712) 0.446 (0.236 to 0.656) 0.318* (0.074 to 0.563) 0.126 (−0.047 to 0.298) Low 0.482*** (0.256 to 0.709) 0.428*** (0.107 to 0.956) 0.433 (0.109 to 0.757) 0.287 (−0.017 to 0.59) 0.158 (−0.099 to 0.414) 0.197 (−0.056 to 0.449) . . Maintenance . Updating . Shifting . Inhibition . Processing speed . Reasoning . Age 60–70 years old 0.5*** (0.328 to 0.672) 0.433** (0.167 to 0.699) 0.394** (0.132 to 0.655) 0.275* (0.052 to 0.499) 0.398** (0.139 to 0.656) 0.18* (0.016 to 0.344) 70 and older 0.465*** (0.252 to 0.678) 0.471** (0.171 to 0.765) 0.477** (0.124 to 0.829) 0.514*** (0.279 to 0.749) 0.107 (−0.139 to 0.346) 0.052 (−0.236 to 0.34) Training frequency 0–3 sessions per week 0.51*** (0.289 to 0.731) 0.939** (0.25 to 1.629) NA 0.735***(0.438 to 1.033) 0.318* (0.074 to 0.563) 0.11 (−0.119 to 0.338) ≥3 sessions per week 0.473*** (0.304 to 0.641) 0.407*** (0.201 to 0.603) 0.447*** (0.246 to 0.648) 0.262**(0.089 to 1.435) 0.158 (−0.099 to 0.414) 0.173 (−0.009 to 0.355) Training sessions 0–10 sessions 0.446*** (0.337 to 0.709) 0.939** (0.25 to 1.629) 0.276 (−0.131 to 0.685) 0.582***(0.341 to 0.822) 0.318* (0.074 to 0.563) 0.138 (−0.037 to 0.314) ≥10 sessions 0.523*** (0.253 to 0.639) 0.407*** (0.201 to 0.603) 0.501*** (0.27 to 0.733) 0.256**(0.065 to 0.447) 0.158 (−0.099 to 0.414) 0.167 (−0.077 to 0.411) Total hours 0–10 hr 0.498*** (0.34 to 0.656) 0.448*** (0.228 to 0.668) 0.394** (0.132 to 0.655) 0.4*** (0.24 to 0.561) 0.25* (0.056 to 0.443) 0.153* (0.003 to 0.304) ≥10 hr 0.456*** (0.202 to 0.71) 0.46* (0.013 to 0.906) 0.524** (0.209 to 0.84) 0.261 (−0.151 to 0.672) 0.2 (−0.238 to 0.639) 0.105 (−0.333 to 0.543) Type of control Passive 0.536*** (0.326 to 0.745) 0.467*** (0.245 to 0.689) 0.256 (−0.066 to 0.579) 0.383* (0.088 to 0.677) 0.164 (−0.183 to 0.511) 0.091 (−0.127 to 0.308) Active 0.452*** (0.278 to 0.625) 0.388 (−0.044 to 0.819) 0.569*** (0.311 to 0.827) 0.376*** (0.17 to 0.583) 0.269* (0.063 to 0.475) 0.192* (0.003 to 0.38) Study quality High 0.489*** (0.323 to 0.655) 0.531* (0.107 to 0.956) 0.455 (0.199 to 0.712) 0.446 (0.236 to 0.656) 0.318* (0.074 to 0.563) 0.126 (−0.047 to 0.298) Low 0.482*** (0.256 to 0.709) 0.428*** (0.107 to 0.956) 0.433 (0.109 to 0.757) 0.287 (−0.017 to 0.59) 0.158 (−0.099 to 0.414) 0.197 (−0.056 to 0.449) Note: Values in bold mean significant between-group difference. *p < .05, **p < .01, ***p < .001. Open in new tab Table 1. Subgroup Analyses of Potential Moderators . . Maintenance . Updating . Shifting . Inhibition . Processing speed . Reasoning . Age 60–70 years old 0.5*** (0.328 to 0.672) 0.433** (0.167 to 0.699) 0.394** (0.132 to 0.655) 0.275* (0.052 to 0.499) 0.398** (0.139 to 0.656) 0.18* (0.016 to 0.344) 70 and older 0.465*** (0.252 to 0.678) 0.471** (0.171 to 0.765) 0.477** (0.124 to 0.829) 0.514*** (0.279 to 0.749) 0.107 (−0.139 to 0.346) 0.052 (−0.236 to 0.34) Training frequency 0–3 sessions per week 0.51*** (0.289 to 0.731) 0.939** (0.25 to 1.629) NA 0.735***(0.438 to 1.033) 0.318* (0.074 to 0.563) 0.11 (−0.119 to 0.338) ≥3 sessions per week 0.473*** (0.304 to 0.641) 0.407*** (0.201 to 0.603) 0.447*** (0.246 to 0.648) 0.262**(0.089 to 1.435) 0.158 (−0.099 to 0.414) 0.173 (−0.009 to 0.355) Training sessions 0–10 sessions 0.446*** (0.337 to 0.709) 0.939** (0.25 to 1.629) 0.276 (−0.131 to 0.685) 0.582***(0.341 to 0.822) 0.318* (0.074 to 0.563) 0.138 (−0.037 to 0.314) ≥10 sessions 0.523*** (0.253 to 0.639) 0.407*** (0.201 to 0.603) 0.501*** (0.27 to 0.733) 0.256**(0.065 to 0.447) 0.158 (−0.099 to 0.414) 0.167 (−0.077 to 0.411) Total hours 0–10 hr 0.498*** (0.34 to 0.656) 0.448*** (0.228 to 0.668) 0.394** (0.132 to 0.655) 0.4*** (0.24 to 0.561) 0.25* (0.056 to 0.443) 0.153* (0.003 to 0.304) ≥10 hr 0.456*** (0.202 to 0.71) 0.46* (0.013 to 0.906) 0.524** (0.209 to 0.84) 0.261 (−0.151 to 0.672) 0.2 (−0.238 to 0.639) 0.105 (−0.333 to 0.543) Type of control Passive 0.536*** (0.326 to 0.745) 0.467*** (0.245 to 0.689) 0.256 (−0.066 to 0.579) 0.383* (0.088 to 0.677) 0.164 (−0.183 to 0.511) 0.091 (−0.127 to 0.308) Active 0.452*** (0.278 to 0.625) 0.388 (−0.044 to 0.819) 0.569*** (0.311 to 0.827) 0.376*** (0.17 to 0.583) 0.269* (0.063 to 0.475) 0.192* (0.003 to 0.38) Study quality High 0.489*** (0.323 to 0.655) 0.531* (0.107 to 0.956) 0.455 (0.199 to 0.712) 0.446 (0.236 to 0.656) 0.318* (0.074 to 0.563) 0.126 (−0.047 to 0.298) Low 0.482*** (0.256 to 0.709) 0.428*** (0.107 to 0.956) 0.433 (0.109 to 0.757) 0.287 (−0.017 to 0.59) 0.158 (−0.099 to 0.414) 0.197 (−0.056 to 0.449) . . Maintenance . Updating . Shifting . Inhibition . Processing speed . Reasoning . Age 60–70 years old 0.5*** (0.328 to 0.672) 0.433** (0.167 to 0.699) 0.394** (0.132 to 0.655) 0.275* (0.052 to 0.499) 0.398** (0.139 to 0.656) 0.18* (0.016 to 0.344) 70 and older 0.465*** (0.252 to 0.678) 0.471** (0.171 to 0.765) 0.477** (0.124 to 0.829) 0.514*** (0.279 to 0.749) 0.107 (−0.139 to 0.346) 0.052 (−0.236 to 0.34) Training frequency 0–3 sessions per week 0.51*** (0.289 to 0.731) 0.939** (0.25 to 1.629) NA 0.735***(0.438 to 1.033) 0.318* (0.074 to 0.563) 0.11 (−0.119 to 0.338) ≥3 sessions per week 0.473*** (0.304 to 0.641) 0.407*** (0.201 to 0.603) 0.447*** (0.246 to 0.648) 0.262**(0.089 to 1.435) 0.158 (−0.099 to 0.414) 0.173 (−0.009 to 0.355) Training sessions 0–10 sessions 0.446*** (0.337 to 0.709) 0.939** (0.25 to 1.629) 0.276 (−0.131 to 0.685) 0.582***(0.341 to 0.822) 0.318* (0.074 to 0.563) 0.138 (−0.037 to 0.314) ≥10 sessions 0.523*** (0.253 to 0.639) 0.407*** (0.201 to 0.603) 0.501*** (0.27 to 0.733) 0.256**(0.065 to 0.447) 0.158 (−0.099 to 0.414) 0.167 (−0.077 to 0.411) Total hours 0–10 hr 0.498*** (0.34 to 0.656) 0.448*** (0.228 to 0.668) 0.394** (0.132 to 0.655) 0.4*** (0.24 to 0.561) 0.25* (0.056 to 0.443) 0.153* (0.003 to 0.304) ≥10 hr 0.456*** (0.202 to 0.71) 0.46* (0.013 to 0.906) 0.524** (0.209 to 0.84) 0.261 (−0.151 to 0.672) 0.2 (−0.238 to 0.639) 0.105 (−0.333 to 0.543) Type of control Passive 0.536*** (0.326 to 0.745) 0.467*** (0.245 to 0.689) 0.256 (−0.066 to 0.579) 0.383* (0.088 to 0.677) 0.164 (−0.183 to 0.511) 0.091 (−0.127 to 0.308) Active 0.452*** (0.278 to 0.625) 0.388 (−0.044 to 0.819) 0.569*** (0.311 to 0.827) 0.376*** (0.17 to 0.583) 0.269* (0.063 to 0.475) 0.192* (0.003 to 0.38) Study quality High 0.489*** (0.323 to 0.655) 0.531* (0.107 to 0.956) 0.455 (0.199 to 0.712) 0.446 (0.236 to 0.656) 0.318* (0.074 to 0.563) 0.126 (−0.047 to 0.298) Low 0.482*** (0.256 to 0.709) 0.428*** (0.107 to 0.956) 0.433 (0.109 to 0.757) 0.287 (−0.017 to 0.59) 0.158 (−0.099 to 0.414) 0.197 (−0.056 to 0.449) Note: Values in bold mean significant between-group difference. *p < .05, **p < .01, ***p < .001. Open in new tab Drop-out Rates The pooled drop-out rates were 24.8% (95% CI = 20.9%–29.2%) in the training group and 26.2% (95% CI = 21.9%–31%) in the control group. No between-group difference was detected (Q(1) = 0.207, p = .649). Discussion Based on results from 22 randomized controlled trials of moderate quality, our meta-analysis of WM training in healthy older adults identified that most WM interventions were effective in terms of long-term performance improvement in the aspects of four WM components as well as processing speed and reasoning. Importantly, the pattern of results for long-term effects was highly consistent across studies. Our results showed that WM training produced long-term enhanced functioning of the WM system, which demonstrated the plasticity of WM in late adulthood. These results were partially consistent with previous meta-analyses reporting long-term effects on WM (Nguyen et al., 2019; Sala et al., 2019; Schwaighofer, Fischer, & Bühner, 2015; Teixeira-Santos et al., 2019). In addition, we found that the maintenance effects for all WM subcomponents were stable in either shorter or longer follow-up assessments at similar levels. Notably, the finding regarding the stability of WM training should be interpreted with caution because of the reduced sample size at follow-up assessment. The systematic bias caused by the drop-out may be amplified due to the RCTs with small sample sizes and lack of multiple tests that assess the same cognitive construct. For possible interpretation, according to Engle, Tuholski, Laughlin, and Conway (1999), WM capacity is closely related to controlled attention, which reflects the basic ability to keep the representation active, especially in the face of lure, distraction, or competing information. WM training enhanced the basic mental process that is highly involved in all WM subcomponents and might lead to long-lasting training effects. In this meta-analysis, we also addressed the limitations of previous meta-analyses. Melby-Lervag and colleagues (2016) identified two main issues in previous meta-analyses that reported positive effect size of WM training. The first one noted by the authors was the importance of considering different types of control groups (active vs control). To address this issue, we computed the effect sizes stratified by the types of control and conducted subgroup analyses among all outcomes. We found no significant differences in all outcomes, which partially corroborated previous meta-analyses reporting immediate training effect sizes (Karbach & Verhaeghen, 2014). Notably, our pattern did not comply with Melby-Lervag and colleagues (2016), and the possible reasons might be the differences in the research population (healthy older adults vs participants in all age groups), method of outcome categorization and target question (long-term vs immediate training effects). The second issue mentioned by Melby-Lervag and colleagues (2016) was related to the calculation of Cohen’s d without considering the unbalancing of the baseline score. This issue was severe even in RCTs with small sample sizes, which may hint at the failure of the randomization process. In this meta-analysis, we calculated our effect sizes considering the baseline score and thus may draw a more reliable conclusion. However, these methodology issues should be cautiously addressed in future WM training programs. More interestingly, no prior meta-analyses have considered the confounding effects of RCT type (cluster vs individual). Our analyses showed that 13.7% (3/22) of the total comparisons adopted cluster RCT, and if statistical analysis at the level of individual was conducted for cluster RCT, this may lead to a high level of precision in the analysis (Higgins, Eldridge, & Li, 2019). We thus followed the Cochrane guidebook to adjust the sample size by calculating the effective sample size and to obtain a more precise effect size for cluster RCTs. Whether transfer effects after training exist is also the core question for clinicians and researchers. Interestingly, we found that WM could exert significant long-term improvement in processing speed and reasoning, but not in episodic memory. Both processing speed and reasoning are basic and related abilities that determine participants’ capacity. Because processing speed can be regarded as a basic cognitive process that interprets individual differences in WM occurring with aging, enhancing WM performance increases the efficiency of mental operations fostering the ability to move among the basic information processes (Park et al., 2002). For reasoning ability, reasoning tasks (e.g., Cattell task and RAVEN) rely on the maintenance of activation to goal-relevant information and manipulation of the final goal and subgoals in the concurrent information and/or distraction (Engle et al., 1999). Therefore, the reasoning tasks that are needed to maintain task-related information (e.g., rules) in the process of concurrent information (e.g., searching other rule) and distractors (e.g., inhibiting irrelevant features) may require the function of the central executive function of WM (Conway, Cowan, Bunting, Therriault, & Minkoff, 2002). In contrast, the cognitive tasks representing episodic memory may rely more on task-specific processing and strategies, leading to transient transfer effects. WM training protocols usually adopt a process-based procedure and teach no specific strategies. We thus may fail to find any transfer effect on episodic memory. One study hypothesized that subcomponents of WM could facilitate the utility of mnemonics for episodic memory in older adults. The researchers thus designed a combined cognitive training (updating, switching, and mnemonic in sequence) versus mono-mnemonic training study, reporting a smaller memory composite score in the combined cognitive training group relative to the mono-mnemonic training group (Li et al., 2016). Regarding the results of moderator analyses, we found that lower training frequency and fewer total sessions were associated with higher long-term effects on the inhibition task, which is partially consistent with previous meta-analysis regarding the effects of computerized cognitive training in healthy older adults (Lampit et al., 2014). Tight training schedule may cause cognitive fatigue, low level of training adherence, and high level of stress (Fiatarone Singh et al., 2014). In addition, this phenomenon might not be unique to healthy older adults, as the training programs in children (Wang, Zhou, & Shah, 2014) and young people (Penner et al., 2012) also identified similar association between lower doses and greater training efficacy. However, our results of moderator analyses should be interpreted with caution because of the lack of statistic power and multiple adjustment. Future studies are needed to investigate the optimal dose to maximize the training effects in healthy older adults and to identify other potential moderators (e.g., booster session). With respect to the study quality, though we did not find any between-group differences (high vs low) in all outcomes, we still must point out the methodology issues. Almost 90% of the total studies did not adopt concealed allocation, leading to potential selection bias and inflated effect sizes (Schulz, Chalmers, Hayes, & Altman, 1995). Approximately 70% studies did not blind accessors, which may have caused the assessors’ biases to impinge on their measurements of outcomes. Another issue, namely, high drop-out rates, should also be addressed. Although we did not find between-group differences in drop-out rates, it is necessary that measurements of WM outcomes are made on all participants who are randomized to groups. Participants who are not followed up may differ systematically from those who are, and this potentially causes bias. Buschkuehl and colleagues (2008) did not adopt intention-to-treat analysis. Analysis of data based on how participants were trained (instead of according to how participants should have been trained) may produce biases (Hollis & Campbell, 1999). Sandberg and Stigsdotter Neely (2015) failed to obtain a similar baseline, indicating inadequate randomization and potential bias arising by chance with random allocation. With respect to publication bias assessed, we found no significant publication bias across outcomes, indicating the robustness of our outcomes. Our analyses provide surprisingly consistent evidence for the long-term effectiveness of WM interventions for four WM subcomponents and other cognitive domains. Future studies are needed to use different tasks of the same WM construct to evaluate the WM training benefits and adopt more ecological tasks or tasks related to everyday functioning and thereby to improve the external validity of WM training. In addition, implementation studies are needed to effectively promote training protocols in a large-scale aged population at risk of dementia. Funding This work was supported by the National 13th Five-Year Grand Program on Key Infectious Disease Control (2017ZX10202102-005-003 to B.S., 2018ZX10301-407-005, 2018ZX10302103-001-003 and 2018ZX10301407-005-001 to T.J., and 2017ZX10202101-004-001 to T.Z.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Acknowledgments The review was preregistered in the International Prospective Register of Systematic Reviews (PROSPERO, http://www.cdr.york.ac.uk/prospero, CRD42019122061). The data that support the findings of this study are available on request from the corresponding author (T.Z.). Author Contribution J.H. conceptualized the study. J.H. and T.J. searched the literature, selected studies, and extracted the data. J.H., R.S., and J.F. contributed to the analysis and interpretation of the data and provided important scientific input. J.H. and R.S. analyzed the findings and wrote the first draft of the manuscript with input from B.S. T.J. and T.Z. supervised the study. All authors collaboratively discussed key decisions throughout the course of the review, provided critical feedback on preliminary manuscript and interpretation of results, and approved the final version. Conflict of Interest None reported. References Baddeley , A . ( 2003 ). Working memory: Looking back and looking forward . Nature Reviews Neuroscience , 4 ( 10 ), 829 – 839 . doi:10.1038/nrn1201 Google Scholar Crossref Search ADS PubMed WorldCat Baddeley , A. D. , & Hitch , G. ( 1974 ). Working memory. In G. H. Bower (Ed.), The psychology of learning and motivation (Vol. 8 , pp. 47 – 89 ). New York: Academic Press. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Bettio , L. E. B. , Rajendran , L., & Gil-Mohapel , J. ( 2017 ). The effects of aging in the hippocampus and cognitive decline . Neuroscience and Biobehavioral Reviews , 79 , 66 – 86 . doi:10.1016/j.neubiorev.2017.04.030 Google Scholar Crossref Search ADS PubMed WorldCat Bledowski , C. , Rahm , B., & Rowe , J. B. ( 2009 ). What “works” in working memory? Separate systems for selection and updating of critical information . The Journal of Neuroscience , 29 ( 43 ), 13735 – 13741 . doi:10.1523/JNEUROSCI.2547-09.2009 Google Scholar Crossref Search ADS PubMed WorldCat Borella , E. , Cantarella , A., Carretti , B., De Lucia , A., & De Beni , R. ( 2019 ). Improving everyday functioning in the old-old with working memory training . The American Journal of Geriatric Psychiatry , 27 ( 9 ), 975 – 983 . doi:10.1016/j.jagp.2019.01.210 Google Scholar Crossref Search ADS PubMed WorldCat Borella , E. , Carbone , E., Pastore , M., De Beni , R., & Carretti , B. ( 2017 ). Working memory training for healthy older adults: The role of individual characteristics in explaining short- and long-term gains . Frontiers in Human Neuroscience , 11 , 99 . doi:10.3389/fnhum.2017.00099 Google Scholar Crossref Search ADS PubMed WorldCat Borella , E. , Carretti , B., Cantarella , A., Riboldi , F., Zavagnin , M., & De Beni , R. ( 2014 ). Benefits of training visuospatial working memory in young-old and old-old . Developmental Psychology , 50 ( 3 ), 714 – 727 . doi:10.1037/a0034293 Google Scholar Crossref Search ADS PubMed WorldCat Borella , E. , Carretti , B., & De Beni , R. ( 2008 ). Working memory and inhibition across the adult life-span . Acta Psychologica , 128 ( 1 ), 33 – 44 . doi:10.1016/j.actpsy.2007.09.008 Google Scholar Crossref Search ADS PubMed WorldCat Borella , E. , Carretti , B., Meneghetti , C., Carbone , E., Vincenzi , M., Madonna , J. C., . . . Mammarella , N. ( 2017 ). Is working memory training in older adults sensitive to music? Psychological Research , 83 ( 6 ), 1107 – 1123 . doi:10.1007/s00426-017-0961-8 Google Scholar Crossref Search ADS PubMed WorldCat Borella , E. , Carretti , B., Riboldi , F., & De Beni , R. ( 2010 ). Working memory training in older adults: Evidence of transfer and maintenance effects . Psychology and Aging , 25 ( 4 ), 767 – 778 . doi:10.1037/a0020683 Google Scholar Crossref Search ADS PubMed WorldCat Borella , E. , Carretti , B., Sciore , R., Capotosto , E., Taconnat , L., Cornoldi , C., & De Beni , R. ( 2017 ). Training working memory in older adults: Is there an advantage of using strategies? Psychology and Aging , 32 ( 2 ), 178 – 191 . doi:10.1037/pag0000155 Google Scholar Crossref Search ADS PubMed WorldCat Borella , E. , Carretti , B., Zanoni , G., Zavagnin , M., & De Beni , R. ( 2013 ). Working memory training in old age: An examination of transfer and maintenance effects . Archives of Clinical Neuropsychology , 28 ( 4 ), 331 – 347 . doi:10.1093/arclin/act020 Google Scholar Crossref Search ADS PubMed WorldCat Brehmer , Y. , Westerberg , H., & Bäckman , L. ( 2012 ). Working-memory training in younger and older adults: Training gains, transfer, and maintenance . Frontiers in Human Neuroscience , 6 , 63 . doi:10.3389/fnhum.2012.00063 Google Scholar Crossref Search ADS PubMed WorldCat Brum , P. S. , Borella , E., Carretti , B., & Sanches Yassuda , M. ( 2018 ). Verbal working memory training in older adults: An investigation of dose response . Aging & Mental Health , 24 ( 1 ), 81 – 91 . doi:10.1080/13607863.2018.1531372 Google Scholar Crossref Search ADS PubMed WorldCat Burki , C. N. , Ludwig , C., Chicherio , C., & de Ribaupierre , A. ( 2014 ). Individual differences in cognitive plasticity: An investigation of training curves in younger and older adults . Psychological Research , 78 ( 6 ), 821 – 835 . doi:10.1007/s00426-014-0559-3 Google Scholar Crossref Search ADS PubMed WorldCat Buschkuehl , M. , Jaeggi , S. M., Hutchison , S., Perrig-Chiello , P., Däpp , C., Müller , M., . . . Perrig , W. J. ( 2008 ). Impact of working memory training on memory performance in old-old adults . Psychology and Aging , 23 ( 4 ), 743 – 753 . doi:10.1037/a0014342 Google Scholar Crossref Search ADS PubMed WorldCat Campbell , M. , Grimshaw , J., & Steen , N. ( 2000 ). Sample size calculations for cluster randomised trials. Changing Professional Practice in Europe Group (EU BIOMED II Concerted Action) . Journal of Health Services Research & Policy , 5 ( 1 ), 12 – 16 . doi:10.1177/135581960000500105 Google Scholar Crossref Search ADS PubMed WorldCat Cantarella , A. , Borella , E., Carretti , B., Kliegel , M., Mammarella , N., Fairfield , B., & De Beni , R. ( 2017 ). The influence of training task stimuli on transfer effects of working memory training in aging . Psychologie Française. doi:10.1016/j.psfr.2017.04.005 OpenURL Placeholder Text WorldCat Carretti , B. , Borella , E., Zavagnin , M., & de Beni , R. ( 2013 ). Gains in language comprehension relating to working memory training in healthy older adults . International Journal of Geriatric Psychiatry , 28 ( 5 ), 539 – 546 . doi:10.1002/gps.3859 Google Scholar Crossref Search ADS PubMed WorldCat Chiu , H. L. , Chan , P. T., Kao , C. C., Chu , H., Chang , P. C., Hsiao , S. S., . . . Chou , K. R. ( 2018 ). Effectiveness of executive function training on mental set shifting, working memory and inhibition in healthy older adults: A double-blind randomized controlled trials . Journal of Advanced Nursing , 74 ( 5 ), 1099 – 1113 . doi:10.1111/jan.13519 Google Scholar Crossref Search ADS PubMed WorldCat Conway , A. R. A. , Cowan , N., Bunting , M. F., Therriault , D. J., & Minkoff , S. R. B. ( 2002 ). A latent variable analysis of working memory capacity, short-term memory capacity, processing speed, and general fluid intelligence . Intelligence , 30 ( 2 ), 163 – 183 . doi:10.1016/s0160-2896(01)00096-4 Google Scholar Crossref Search ADS WorldCat Conway , A. R. , Kane , M. J., & Engle , R. W. ( 2003 ). Working memory capacity and its relation to general intelligence . Trends in Cognitive Sciences , 7 ( 12 ), 547 – 552 . doi:10.1016/j.tics.2003.10.005 Google Scholar Crossref Search ADS PubMed WorldCat Courtney , S. M. , Sala , J. K. R., & Sala , J. B. ( 2007 ). A hierarchical biased-competition model of domain-dependent working memory mainatenance and executive control. In N. Osaka, R. H. Logie, & M. D’Esposito (Eds.), The cognitive neuroscience of working memory . Stanford: Oxford University Press. pp. 369–384 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Daffner , K. R . ( 2010 ). Promoting successful cognitive aging: A comprehensive review . Journal of Alzheimer’s Disease , 19 ( 4 ), 1101 – 1122 . doi:10.3233/JAD-2010-1306 Google Scholar Crossref Search ADS PubMed WorldCat Dahlin , E. , Nyberg , L., Bäckman , L., & Neely , A. S. ( 2008 ). Plasticity of executive functioning in young and older adults: Immediate training gains, transfer, and long-term maintenance . Psychology and Aging , 23 ( 4 ), 720 – 730 . doi:10.1037/a0014296 Google Scholar Crossref Search ADS PubMed WorldCat Dehn , M. J . ( 2017 ). How working memory enables fluid reasoning . Applied Neuropsychology Child , 6 ( 3 ), 245 – 247 . doi:10.1080/21622965.2017.1317490 Google Scholar Crossref Search ADS PubMed WorldCat Du , X. , Ji , Y., Chen , T., Tang , Y., & Han , B. ( 2018 ). Can working memory capacity be expanded by boosting working memory updating efficiency in older adults? Psychology and Aging , 33 ( 8 ), 1134 – 1151 . doi:10.1037/pag0000311 Google Scholar Crossref Search ADS PubMed WorldCat Eldridge , S. M. , Ukoumunne , O. C., & Carlin , J. B. ( 2009 ). The intra-cluster correlation coefficient in cluster randomized trials: A review of definitions . International Statistical Review , 77 ( 3 ), 378 – 394 . doi:10.1111/j.1751-5823.2009.00092.x Google Scholar Crossref Search ADS WorldCat Engle , R. W. , Tuholski , S. W., Laughlin , J. E., & Conway , A. R. A. ( 1999 ). Working memory, short-term memory, and general fluid intelligence: A latent-variable approach . Journal of Experimental Psychology. General , 128 ( 3 ), 309 – 331 . doi:10.1037//0096-3445.128.3.309 Google Scholar Crossref Search ADS PubMed WorldCat Fiatarone Singh , M. A. , Gates , N., Saigal , N., Wilson , G. C., Meiklejohn , J., Brodaty , H., . . . Valenzuela , M. ( 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 . Journal of the American Medical Directors Association , 15 ( 12 ), 873 – 880 . doi:10.1016/j.jamda.2014.09.010 Google Scholar Crossref Search ADS PubMed WorldCat Gillett , S. R. , McClure , L. A., Callas , P. W., Thacker , E. L., Unverzagt , F. W., Wadley , V. G., . . . Cushman , M. ( 2018 ). Hemostasis biomarkers and incident cognitive impairment: The REGARDS study . Journal of Thrombosis and Haemostasis , 16 ( 7 ), 1259 – 1267 . doi:10.1111/jth.14138 Google Scholar Crossref Search ADS PubMed WorldCat Gronholm-Nyman , P. , Soveri , A., Rinne , J. O., Ek , E., Nyholm , A., Stigsdotter Neely , A., & Laine , M. ( 2017 ). Limited effects of set shifting training in healthy older adults . Frontiers in Aging Neuroscience , 9 , 69 . doi:10.3389/fnagi.2017.00069 Google Scholar Crossref Search ADS PubMed WorldCat Higgins , J. P. T. , Eldridge , S., & Li , T. ( 2019 ). Chapter 23: Including variants on randomized trials. In J. P. T. Higgins, J. Thomas, J. Chandler, M. Cumpston, T. Li, M. J. Page, & V. A. Welch (Eds.), Cochrane handbook for systematic reviews of interventions version 6.0 (updated July 2019). Cochrane . Retrieved from www.training.cochrane.org/handbook Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Hitch , G. J. , Allen , R. J., & Baddeley , A. D. ( 2020 ). Attention and binding in visual working memory: Two forms of attention and two kinds of buffer storage . Attention, Perception & Psychophysics , 82 ( 1 ), 280 – 293 . doi:10.3758/s13414-019-01837-x Google Scholar Crossref Search ADS PubMed WorldCat Hollis , S. , & Campbell , F. ( 1999 ). What is meant by intention to treat analysis? Survey of published randomised controlled trials . BMJ , 319 ( 7211 ), 670 – 674 . doi:10.1136/bmj.319.7211.670 Google Scholar Crossref Search ADS PubMed WorldCat Jaeggi , S. M. , Buschkuehl , M., Jonides , J., & Perrig , W. J. ( 2008 ). Improving fluid intelligence with training on working memory . Proceedings of the National Academy of Sciences of the United States of America , 105 ( 19 ), 6829 – 6833 . doi:10.1073/pnas.0801268105 Google Scholar Crossref Search ADS PubMed WorldCat Jaeggi , S. M. , Buschkuehl , M., Parlett-Pelleriti , C. M., Moon , S. M., Evans , M., Kritzmacher , A., . . . Jonides , J. ( 2019 ). Investigating the effects of spacing on working memory training outcome – A randomized controlled multi-site trial in older adults . The Journals of Gerontology, Series B: Psychological Sciences & Social Sciences. 75(6), 1181–1192. doi:10.1093/geronb/gbz090 OpenURL Placeholder Text WorldCat Jia , J. , Wei , C., Chen , S., Li , F., Tang , Y., Qin , W., . . . Gauthier , S. ( 2018 ). The cost of Alzheimer’s disease in China and re-estimation of costs worldwide . Alzheimer’s & Dementia , 14 ( 4 ), 483 – 491 . doi:10.1016/j.jalz.2017.12.006 Google Scholar Crossref Search ADS WorldCat Karbach , J. , & Verhaeghen , P. ( 2014 ). Making working memory work: A meta-analysis of executive-control and working memory training in older adults . Psychological Science , 25 ( 11 ), 2027 – 2037 . doi:10.1177/0956797614548725 Google Scholar Crossref Search ADS PubMed WorldCat Keller , J. B. , Hedden , T., Thompson , T. W., Anteraper , S. A., Gabrieli , J. D., & Whitfield-Gabrieli , S. ( 2015 ). Resting-state anticorrelations between medial and lateral prefrontal cortex: Association with working memory, aging, and individual differences . Cortex , 64 , 271 – 280 . doi:10.1016/j.cortex.2014.12.001 Google Scholar Crossref Search ADS PubMed WorldCat Kirova , A. M. , Bays , R. B., & Lagalwar , S. ( 2015 ). Working memory and executive function decline across normal aging, mild cognitive impairment, and Alzheimer’s disease . Biomed Research International , 2015 , 748212 . doi:10.1155/2015/748212 Google Scholar Crossref Search ADS PubMed WorldCat Konishi , K. , Mckenzie , S., Etchamendy , N., Roy , S., & Bohbot , V. D. ( 2017 ). Hippocampus-dependent spatial learning is associated with higher global cognition among healthy older adults . Neuropsychologia , 106 , 310 – 321 . doi:10.1016/j.neuropsychologia.2017.09.025 Google Scholar Crossref Search ADS PubMed WorldCat Lampit , A. , Hallock , H., & Valenzuela , M. ( 2014 ). Computerized cognitive training in cognitively healthy older adults: A systematic review and meta-analysis of effect modifiers . PLoS Medicine , 11 ( 11 ), e1001756 . doi:10.1371/journal.pmed.1001756 Google Scholar Crossref Search ADS PubMed WorldCat Li , B. , Zhu , X., Hou , J., Chen , T., Wang , P., & Li , J. ( 2016 ). Combined cognitive training vs. memory strategy training in healthy older adults . Frontiers in Psychology , 7 , 834. doi:10.3389/fpsyg.2016.00834 OpenURL Placeholder Text WorldCat Lin , P. J. , & Neumann , P. J. ( 2013 ). The economics of mild cognitive impairment . Alzheimer’s & Dementia , 9 ( 1 ), 58 – 62 . doi:10.1016/j.jalz.2012.05.2117 Google Scholar Crossref Search ADS WorldCat MacDonald , S. W. , Hultsch , D. F., & Dixon , R. A. ( 2011 ). Aging and the shape of cognitive change before death: Terminal decline or terminal drop? The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 66 ( 3 ), 292 – 301 . doi:10.1093/geronb/gbr001 Google Scholar Crossref Search ADS WorldCat Maher , C. G. , Sherrington , C., Herbert , R. D., Moseley , A. M., & Elkins , M. ( 2003 ). Reliability of the PEDro scale for rating quality of randomized controlled trials . Physical Therapy , 83 ( 8 ), 713 – 721 . doi:10.1093/ptj/83.8.713 Google Scholar Crossref Search ADS PubMed WorldCat Marioni , R. E. , van den Hout , A., Valenzuela , M. J., Brayne , C., & Matthews , F. E.; MRC Cognitive Function and Ageing Study. ( 2012 ). Active cognitive lifestyle associates with cognitive recovery and a reduced risk of cognitive decline . Journal of Alzheimer’s Disease , 28 ( 1 ), 223 – 230 . doi:10.3233/JAD-2011-110377 Google Scholar Crossref Search ADS PubMed WorldCat Matysiak , O. , Kroemeke , A., & Brzezicka , A. ( 2019 ). Working memory capacity as a predictor of cognitive training efficacy in the elderly population . Frontiers in Aging Neuroscience , 11 , 126 . doi:10.3389/fnagi.2019.00126 Google Scholar Crossref Search ADS PubMed WorldCat McAvinue , L. P. , Golemme , M., Castorina , M., Tatti , E., Pigni , F. M., Salomone , S., . . . Robertson , I. H. ( 2013 ). An evaluation of a working memory training scheme in older adults . Frontiers in Aging Neuroscience , 5 , 20 . doi:10.3389/fnagi.2013.00020 Google Scholar Crossref Search ADS PubMed WorldCat Melby-Lervag , M. , Redick , T. S., & Hulme , C. ( 2016 ). Working memory training does not improve performance on measures of intelligence or other measures of “far transfer”: Evidence from a meta-analytic review . Perspectives on Psychological Science , 11 ( 4 ), 512 – 534 . doi:10.1177/1745691616635612 Google Scholar Crossref Search ADS PubMed WorldCat Melrose , R. J. , Zahniser , E., Wilkins , S. S., Veliz , J., Hasratian , A. S., Sultzer , D. L., & Jimenez , A. M. ( 2020 ). Prefrontal working memory activity predicts episodic memory performance: A neuroimaging study . Behavioural Brain Research , 379 , 112307 . doi:10.1016/j.bbr.2019.112307 Google Scholar Crossref Search ADS PubMed WorldCat Mewborn , C. M. , Lindbergh , C. A., & Stephen Miller , L. ( 2017 ). Cognitive interventions for cognitively healthy, mildly impaired, and mixed samples of older adults: A systematic review and meta-analysis of randomized-controlled trials . Neuropsychology Review , 27 ( 4 ), 403 – 439 . doi:10.1007/s11065-017-9350-8 Google Scholar Crossref Search ADS PubMed WorldCat Miyake , A. , Friedman , N. P., Emerson , M. J., Witzki , A. H., Howerter , A., & Wager , T. D. ( 2000 ). The unity and diversity of executive functions and their contributions to complex “Frontal Lobe” tasks: A latent variable analysis . Cognitive Psychology , 41 ( 1 ), 49 – 100 . doi:10.1006/cogp.1999.0734 Google Scholar Crossref Search ADS PubMed WorldCat Moher , D. , Liberati , A., Tetzlaff , J., & Altman , D. G.; PRISMA Group. ( 2010 ). Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement . International Journal of Surgery (London, England) , 8 ( 5 ), 336 – 341 . doi:10.1016/j.ijsu.2010.02.007 Google Scholar Crossref Search ADS PubMed WorldCat Monge , Z. A. , & Madden , D. J. ( 2016 ). Linking cognitive and visual perceptual decline in healthy aging: The information degradation hypothesis . Neuroscience and Biobehavioral Reviews , 69 , 166 – 173 . doi:10.1016/j.neubiorev.2016.07.031 Google Scholar Crossref Search ADS PubMed WorldCat Nee , D. E. , Brown , J. W., Askren , M. K., Berman , M. G., Demiralp , E., Krawitz , A., & Jonides , J. ( 2013 ). A meta-analysis of executive components of working memory . Cerebral Cortex , 23 ( 2 ), 264 – 282 . doi:10.1093/cercor/bhs007 Google Scholar Crossref Search ADS PubMed WorldCat Nguyen , L. , Murphy , K., & Andrews , G. ( 2019 ). Immediate and long-term efficacy of executive functions cognitive training in older adults: A systematic review and meta-analysis . Psychological Bulletin , 145 ( 7 ), 698 – 733 . doi:10.1037/bul0000196 Google Scholar Crossref Search ADS PubMed WorldCat Niendam , T. A. , Laird , A. R., Ray , K. L., Dean , Y. M., Glahn , D. C., & Carter , C. S. ( 2012 ). Meta-analytic evidence for a superordinate cognitive control network subserving diverse executive functions . Cognitive, Affective & Behavioral Neuroscience , 12 ( 2 ), 241 – 268 . doi:10.3758/s13415-011-0083-5 Google Scholar Crossref Search ADS PubMed WorldCat Park , D. C. , Lautenschlager , G., Hedden , T., Davidson , N. S., Smith , A. D., & Smith , P. K. ( 2002 ). Models of visuospatial and verbal memory across the adult life span . Psychology and Aging , 17 ( 2 ), 299 – 320 . doi:10.1037/0882-7974.17.2.299 Google Scholar Crossref Search ADS PubMed WorldCat Penner , I. K. , Vogt , A., Stöcklin , M., Gschwind , L., Opwis , K., & Calabrese , P. ( 2012 ). Computerised working memory training in healthy adults: A comparison of two different training schedules . Neuropsychological Rehabilitation , 22 ( 5 ), 716 – 733 . doi:10.1080/09602011.2012.686883 Google Scholar Crossref Search ADS PubMed WorldCat Rao , J. N. , & Scott , A. J. ( 1992 ). A simple method for the analysis of clustered binary data . Biometrics , 48 ( 2 ), 577 – 585 . doi:10.2307/2532311 Google Scholar Crossref Search ADS PubMed WorldCat Rowe , J. W. , & Kahn , R. L. ( 2015 ). Successful aging 2.0: Conceptual expansions for the 21st century . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 70 ( 4 ), 593 – 596 . doi:10.1093/geronb/gbv025 Google Scholar Crossref Search ADS WorldCat Sala , G. , Aksayli , N. D., Tatlidil , K. S., Gondo , Y., & Gobet , F. ( 2019 ). Working memory training does not enhance older adults’ cognitive skills: A comprehensive meta-analysis . Intelligence , 77 , 101386 . doi:10.1016/j.intell.2019.101386 Google Scholar Crossref Search ADS WorldCat Salthouse , T. A . ( 1996 ). The processing-speed theory of adult age differences in cognition . Psychological Review , 103 ( 3 ), 403 – 428 . doi:10.1037/0033-295x.103.3.403 Google Scholar Crossref Search ADS PubMed WorldCat Sandberg , P. , & Stigsdotter Neely , A. ( 2015 ). Long-term effects of executive process training in young and old adults . Neuropsychological Rehabilitation , 26 ( 5–6 ), 761 – 782 . doi:10.1080/09602011.2015.1108205 Google Scholar PubMed OpenURL Placeholder Text WorldCat Schulz , K. F. , Chalmers , I., Hayes , R. J., & Altman , D. G. ( 1995 ). Empirical evidence of bias. Dimensions of methodological quality associated with estimates of treatment effects in controlled trials . JAMA , 273 ( 5 ), 408 – 412 . doi:10.1001/jama.273.5.408 Google Scholar Crossref Search ADS PubMed WorldCat Schwaighofer , M. , Fischer , F., & Bühner , M. ( 2015 ). Does working memory training transfer? A meta-analysis including training conditions as moderators . Educational Psychologist , 50 ( 2 ), 138 – 166 . doi:10.1080/00461520.2015.1036274 Google Scholar Crossref Search ADS WorldCat Teixeira-Santos , A. C. , Moreira , C. S., Magalhães , R., Magalhães , C., Pereira , D. R., Leite , J., . . . Sampaio , A. ( 2019 ). Reviewing working memory training gains in healthy older adults: A meta-analytic review of transfer for cognitive outcomes . Neuroscience and Biobehavioral Reviews , 103 , 163 – 177 . doi:10.1016/j.neubiorev.2019.05.009 Google Scholar Crossref Search ADS PubMed WorldCat Tesch-Römer , C. , & Wahl , H. W. ( 2017 ). Toward a more comprehensive concept of successful aging: Disability and care needs . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 72 ( 2 ), 310 – 318 . doi:10.1093/geronb/gbw162 OpenURL Placeholder Text WorldCat Toril , P. , Reales , J. M., Mayas , J., & Ballesteros , S. ( 2016 ). Video game training enhances visuospatial working memory and episodic memory in older adults . Frontiers in Human Neuroscience , 10 , 206 . doi:10.3389/fnhum.2016.00206 Google Scholar Crossref Search ADS PubMed WorldCat Tromp , D. , Dufour , A., Lithfous , S., Pebayle , T., & Després , O. ( 2015 ). Episodic memory in normal aging and Alzheimer disease: Insights from imaging and behavioral studies . Ageing Research Reviews , 24 ( Pt B ), 232 – 262 . doi:10.1016/j.arr.2015.08.006 Google Scholar Crossref Search ADS PubMed WorldCat Tulving , E . ( 2002 ). Episodic memory: From mind to brain . Annual Review of Psychology , 53 , 1 – 25 . doi:10.1146/annurev.psych.53.100901.135114 Google Scholar Crossref Search ADS PubMed WorldCat Verghese , J. , Lipton , R. B., Katz , M. J., Hall , C. B., Derby , C. A., Kuslansky , G., . . . Buschke , H. ( 2003 ). Leisure activities and the risk of dementia in the elderly . The New England Journal of Medicine , 348 ( 25 ), 2508 – 2516 . doi:10.1056/NEJMoa022252 Google Scholar Crossref Search ADS PubMed WorldCat Wang , Z. , Zhou , R., & Shah , P. ( 2014 ). Spaced cognitive training promotes training transfer . Frontiers in Human Neuroscience , 8 , 217 . doi:10.3389/fnhum.2014.00217 Google Scholar PubMed OpenURL Placeholder Text WorldCat Weicker , J. , Hudl , N., Frisch , S., Lepsien , J., Mueller , K., Villringer , A., & Thöne-Otto , A. ( 2018 ). WOME: Theory-based working memory training – A placebo-controlled, double-blind evaluation in older adults . Frontiers in Aging Neuroscience , 10 , 247 . doi:10.3389/fnagi.2018.00247 Google Scholar Crossref Search ADS PubMed WorldCat White , I. R. , & Thomas , J. ( 2005 ). Standardized mean differences in individually-randomized and cluster-randomized trials, with applications to meta-analysis . Clinical Trials (London, England) , 2 ( 2 ), 141 – 151 . doi:10.1191/1740774505cn081oa Google Scholar Crossref Search ADS PubMed WorldCat Wilson , R. S. , Boyle , P. A., Segawa , E., Yu , L., Begeny , C. T., Anagnos , S. E., & Bennett , D. A. ( 2013 ). The influence of cognitive decline on well-being in old age . Psychology and Aging , 28 ( 2 ), 304 – 313 . doi:10.1037/a0031196 Google Scholar Crossref Search ADS PubMed WorldCat Wilson , R. S. , Mendes De Leon , C. F., Barnes , L. L., Schneider , J. A., Bienias , J. L., Evans , D. A., & Bennett , D. A. ( 2002 ). Participation in cognitively stimulating activities and risk of incident Alzheimer disease . JAMA , 287 ( 6 ), 742 – 748 . doi:10.1001/jama.287.6.742 Google Scholar Crossref Search ADS PubMed WorldCat Zajac-Lamparska , L. , & Trempala , J. ( 2016 ). Effects of working memory and attentional control training and their transfer onto fluid intelligence in early and late adulthood . Health Psychology Report , 4 ( 1 ), 41 – 53 . doi:10.5114/hpr.2016.56846 Google Scholar Crossref Search ADS WorldCat Zheng , L. , Gao , Z., Xiao , X., Ye , Z., Chen , C., & Xue , G. ( 2018 ). Reduced fidelity of neural representation underlies episodic memory decline in normal aging . Cerebral Cortex , 28 ( 7 ), 2283 – 2296 . doi:10.1093/cercor/bhx130 Google Scholar Crossref Search ADS PubMed WorldCat Zhu , X. , Yin , S., Lang , M., He , R., & Li , J. ( 2016 ). The more the better? A meta-analysis on effects of combined cognitive and physical intervention on cognition in healthy older adults . Ageing Research Reviews , 31 , 67 – 79 . doi:10.1016/j.arr.2016.07.003 Google Scholar Crossref Search ADS PubMed WorldCat Zinke , K. , Zeintl , M., Rose , N. S., Putzmann , J., Pydde , A., & Kliegel , M. ( 2014 ). Working memory training and transfer in older adults: Effects of age, baseline performance, and training gains . Developmental Psychology , 50 ( 1 ), 304 – 315 . doi:10.1037/a0032982 Google Scholar Crossref Search ADS PubMed WorldCat Author notes These authors contributed equally to this study. © The Author(s) 2020. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - The Long-Term Efficacy of Working Memory Training in Healthy Older Adults: A Systematic Review and Meta-Analysis of 22 Randomized Controlled Trials
JF - The Journals of Gerontology Series B: Psychological Sciences and Social Sciences
DO - 10.1093/geronb/gbaa077
DA - 2020-09-14
UR - https://www.deepdyve.com/lp/oxford-university-press/the-long-term-efficacy-of-working-memory-training-in-healthy-older-mhRvlXnzth
SP - e174
EP - e188
VL - 75
IS - 8
DP - DeepDyve
ER -