The Efficacy Versus Evidence Quality of Multicomponent Exercise for Cognitive Health in Older Adults: An Umbrella Review

Abstract

The world is undergoing an unprecedented demographic shift. The proportion of adults aged 60 and over is increasing at a faster rate than any other age group (Bloom et al., 2015;Tu et al., 2022). This global aging phenomenon presents profound implications for public health, social care, and economic stability. Among the most pressing challenges associated with an aging population is the rising prevalence of age-related cognitive decline (Bishop et al., 2010;Yang et al., 2023). This decline ranges from subtle memory lapses to severe neurodegenerative conditions, such as Alzheimer's disease and other dementias (Gonzales et al., 2022). Cognitive impairment significantly compromises an individual's quality of life (Hussenoeder et al., 2020). Furthermore, it increases the burden on families and caregivers while placing immense strain on global healthcare systems (Gauthier et al., 2006;Whitehouse and Moody, 2006;Tahami Monfared et al., 2022). The trajectory of cognitive aging exists on a continuum. In research and clinical settings, cognitive health is often evaluated through "global cognitive function," which refers to the overall functional integrity of the brain's cognitive processes. It represents a generalized construct that summarizes performance across multiple specific domains, including episodic memory, executive function, This umbrella review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement (Page et al., 2021). The review protocol was established a priori to define the research questions, search strategy, and inclusion criteria. The protocol for this study was pre-registered on the International Prospective Register of Systematic Reviews (PROSPERO), under the registration number: CRD420251161230. A systematic and comprehensive literature search was conducted in six electronic databases: PubMed, Web of Science, Embase, Scopus, SPORTDiscus, and the Cochrane Library. The inclusion of SPORTDiscus was intended to capture exercise-specific journals and literature that may be omitted by general medical databases. The search was performed from the inception of each database to September 2025. The search strategy combined keywords and subject headings related to three core concepts: (1) the population (e.g., "older adults," "elderly," "aging"); (2) the intervention (e.g., "multicomponent exercise," "combined training," "mixed exercise"); and (3) the study design (e.g., "systematic review," "meta-analysis"). Detailed search strategy for each database is available in the supplementary materials. For instance, the specific search string used for PubMed was: ("older adults" OR "elderly" OR "aged") AND ("multicomponent exercise" OR "combined training" OR "concurrent training") AND ("systematic review" OR "meta-analysis"). Additionally, the reference lists of included reviews and relevant publications were manually screened to identify any potentially eligible studies. Studies were included in this umbrella review if they met the following criteria:  Population: Older adults, defined as having a mean or median age of 60 years or older.Reviews focusing on specific cognitive statuses (e.g., cognitively healthy, Mild Cognitive Impairment [MCI], frailty) were all eligible.  Intervention: Multicomponent exercise (MCE), defined as a structured exercise program incorporating elements from at least two of the following modalities: aerobic, resistance/strength, balance, and flexibility training.  Comparator: Any non-MCE control group. This included inactive controls (e.g., no intervention, usual care) or active controls (e.g., single-intervention exercises like walking only). The inclusion of active controls (SIE) was intended to allow for the evaluation of the specific "added value" or synergistic effects of the multicomponent approach compared to single-intervention exercises.  Outcomes: The review must have reported on at least one standardized measure of cognitive function. This included Global Cognitive Function (typically assessed by screening tools such as the MMSE or MoCA), Executive Function, Memory (including subtypes such as working, immediate, delayed, and verbal memory), Attention, or Processing Speed.  Study Design: Published systematic reviews with or without a quantitative meta-analysis. Exclusion criteria included original research articles (e.g., individual RCTs), narrative reviews without a systematic search methodology and conference abstracts. All records retrieved from the database search were imported into a reference management software, and duplicates were removed. Two reviewers independently screened the titles and abstracts of the remaining records against the eligibility criteria. The full texts of potentially relevant articles were then retrieved and assessed for final inclusion by the same two reviewers. Any disagreements at either stage of the screening process were resolved through discussion or, if necessary, by consulting a third reviewer. A standardized data extraction form was developed and used by two independent reviewers to extract relevant information from the finally included studies. Discrepancies in extracted data were resolved by consensus. The extracted data included: (1) general characteristics (author, year of publication); (2) study design details (number and type of primary studies); (3) population characteristics (sample size, age, sex, cognitive status); (4) intervention and comparator details; (5) cognitive outcome measures; (6) the results of any quantitative meta-analyses (pooled effect sizes, 95% confidence intervals, heterogeneity statistics); and (7) the authors' conclusions regarding the methodological quality of their included primary studies. The methodological quality of each included systematic review was independently assessed by two reviewers using the AMSTAR-2 (A MeaSurement Tool to Assess systematic Reviews 2) tool. The AMSTAR-2 is a 16-item instrument that evaluates the rigor of the systematic review process, with seven items designated as critical domains. According to the AMSTAR-2 guidance (Shea et al., 2017), the overall confidence in the results was rated as: "High" (zero or one non-critical weakness), "Moderate" (more than one non-critical weakness), "Low" (one critical weakness with or without non-critical weaknesses), or "Critically Low" (more than one critical weakness). Any disagreements in the quality ratings were resolved by consensus. A narrative synthesis was conducted to summarize the characteristics, key findings, and methodological quality of the included reviews. For the quantitative synthesis, we specifically extracted the pooled effect sizes (Standardized Mean Differences [SMD] or Mean Differences [MD]) and their 95% confidence intervals as reported by the original meta-analyses. Systematic reviews that did not provide quantitative pooled estimates were excluded from the generated forest plots and were summarized narratively. Quantitative results from the meta-analyses were extracted and tabulated, organized by specific cognitive outcome domains. Regarding publication bias, we did not perform a de novo funnel plot or Egger's test on the set of included systematic reviews, given the aggregated nature of the data. Instead, we extracted the publication bias findings (e.g., results from Egger 's tests or funnel plot inspections) reported within each included review. These reported assessments were used to evaluate the 'Publication Bias' domain in the subsequent GRADE assessment. Effect sizes were extracted as Standardized Mean Differences (SMD), predominantly Hedges' g (to correct for small sample sizes) or Cohen's d, as reported by the primary reviews. The overall certainty of the evidence for the main cognitive outcomes was assessed using the GRADE (Grading of Recommendations, Assessment, Development and Evaluation) framework. For each outcome, the evidence from the included meta-analyses of RCTs started at "High" certainty. The rating was then downgraded by one level for "serious" limitations or by two levels for "very serious" limitations across five domains: risk of bias (informed by our AMSTAR-2 assessment and the reviews' own quality assessments), inconsistency (unexplained heterogeneity, informed by the I² statistic), indirectness (concerns regarding the generalizability or applicability of the evidence to the research question), imprecision (small number of studies or wide confidence intervals), and publication bias. Consequently, the final certainty of evidence was classified into one of four levels: "High" (very confident in the effect estimate), "Moderate" (moderately confident), "Low" (limited confidence), or "Very Low" (very little confidence). The study selection process is illustrated in the PRISMA flow diagram (Fig. 1). The initial search across six electronic databases (PubMed, Web of Science, Embase, Scopus, SPORTDiscus, and the Cochrane Library) yielded a total of 2,548 records. After duplicates were removed, 1,987 unique articles remained for screening. During the title and abstract screening phase, 485 records were excluded, leaving 76 full-text articles to be assessed for eligibility. Upon full-text review, a further 49 studies were excluded. Ultimately, 27 systematic reviews (Tseng et al., 2011;Carvalho et al., 2014;Asteasu et al., 2017;Northey et al., 2017;Falck et al., 2019;Sanders et al., 2019;Biazus-Sehn et al., 2020;Bliss et al., 2020;Wang et al., 2020Wang et al., , 2024;;Xiong et al., 2020;Cai et al., 2021;Ahn and Kim, 2022;Gallardo-Gómez et al., 2022;Li et al., 2022;Mello et al., 2022;Silva et al., 2022Silva et al., , 2023;;Xu et al., 2023;Alowaydhah et al., 2024;Cerda-Vega et al., 2024;Luo et al., 2024;Ni et al., 2024;Sirikul et al., 2024;Jia et al., 2025;Liu et al., 2025;Vafa et al., 2025) met the full inclusion criteria and were included in the final synthesis for this umbrella review. The detailed characteristics of the 27 included reviews are summarized in Table 1. The publication dates ranged from 2011 to 2025, providing a comprehensive overview of the evidence. The scope of these reviews was extensive, collectively synthesizing data from hundreds of primary randomized controlled trials and encompassing tens of thousands of participants. The target populations were consistently older adults. A key source of heterogeneity was the baseline cognitive status, with reviews focusing on cognitively healthy older adults, individuals with MCI, frail or cognitively frail older adults, or mixed populations. The reviews focused on MCE interventions, typically compared against non-exercise controls such as usual care, health education, or social activities. Global cognitive function was the primary outcome reported, with frequent analyses of specific domains including executive function, memory, attention, and processing speed. The quantitative findings from the included systematic reviews consistently demonstrated a beneficial effect of multicomponent exercise on cognitive function in older adults (Table 2). The most frequently assessed outcome was global cognitive function. Across numerous reviews, multicomponent exercise was found to have a statistically significant, small-to-moderate positive effect, with Standardized Mean Differences (SMDs) generally ranging from 0.24 to 0.65. One network meta-analysis reported a particularly large effect size (SMD = 1.52). Positive effects were also consistently reported for other key cognitive domains. Several meta-analyses showed significant improvements in executive function (SMDs ranging from 0.21 to 0.76) and various aspects of memory, including working memory (SMD = 0.38) and overall memory function (SMD = 0.21). While the direction of the effect was consistently positive, many of the pooled analyses reported moderate to substantial statistical heterogeneity (I ² > 50%), suggesting considerable variability in the results of the underlying primary studies. The included systematic reviews utilized a variety of established tools to assess the methodological quality and risk of bias of their primary studies (Table 3). The most frequently employed instruments were the Physiotherapy Evidence Database (PEDro) scale and various versions of the Cochrane Risk of Bias (RoB) tool. Other tools such as the Jadad scale and the Agency for Healthcare Research and Quality (AHRQ) guidelines were also used by some reviews. Three reviews were narrative or descriptive and did not report a formal quality assessment of primary studies. The authors' conclusions on the quality of the primary evidence were mixed. Several reviews characterized the quality as generally "good" to "excellent". For instance, Northey et al. (Northey et al., 2017) reported a mean PEDro score of 6.7 (out of 10), indicating good quality. However, a substantial number of reviews described the evidence base as being of "fair" to "moderate" quality, or as having a "moderate" to "high" risk of bias. Commonly cited methodological weaknesses in the primary studies included inadequate reporting of randomization and allocation concealment, lack of blinding, and potential selection bias. The methodological quality of the included reviews was assessed using the AMSTAR-2 tool (Table S2). Overall, the quality was highly variable. Based on the AMSTAR-2 criteria, one review was rated as "High" quality, four as "Moderate" quality, 11 as "Low" quality, and 11 as "Critically Low" quality (Fig. 2). Several critical methodological weaknesses were prevalent, leading to the downgrading of most reviews. The most common critical flaws included a failure to register a protocol a priori (Q2), not providing a list of excluded studies with justification (Q7), failing to discuss the likely impact of risk of bias on the results (Q12), and not adequately reporting on conflicts of interest (Q15) (Fig. 3). These widespread limitations temper the confidence in the conclusions of many of the included reviews. The pooled effects of multicomponent exercise on various cognitive domains are summarized in Table 4. For global cognitive function (Fig. 4A), the meta-analysis of 10 studies demonstrated a moderate and statistically significant positive effect (SMD = 0.45, 95% CI [0.32, 0.57], p < 0.001), though with substantial heterogeneity (I² = 71.4%). Regarding executive function (Fig. 4B), the pooled analysis of 5 studies showed a large, significant benefit (SMD = 0.31, 95% CI [0.12, 0.50], p = 0.002), also with substantial heterogeneity (I² = 69.9%). Within this domain, cognitive inhibition (Fig. 4C) showed a small but significant effect (SMD = 0.14), while the effect on cognitive flexibility (Fig. 4D) was not statistically significant (95% CI [-1.74, 1.29]). The intervention also showed significant benefits across multiple memory domains. The overall analysis for memory (3 studies) yielded a small, significant effect (SMD = 0.18, 95% CI [0.06, 0.29]) (Fig. 5A). The effect was particularly strong for verbal memory (4 studies; SMD = 0.33) (Fig. 5B). However, analyses for working memory (Fig. 5C), immediate memory (Fig. 5D), and delayed memory (Fig. 5E) did not reach statistical significance, as their confidence intervals crossed zero. For attention and processing speed (Fig. 5F), the analysis of 4 studies indicated a large pooled non-significant pooled effect (SMD = 0.19, 95% CI [-0.12, 0.50]) associated with considerable heterogeneity (I² = 83.5%). Subgroup analyses were conducted to investigate the influence of baseline cognitive status on the intervention's effectiveness (Table S3). For global cognitive function (Fig. 6A), significant moderate effects were found in both the MCI subgroup (6 studies; SMD = 0.54) and the healthy subgroup (4 studies; SMD = 0.41), with substantial heterogeneity present in both analyses. We notably did not generate a quantitative forest plot for the "Frailty" subgroup due to the limited number of reviews reporting compatible Standardized Mean Differences (SMD) for this specific population. However, individual reviews focusing on frail older adults (e.g., Sirikul et al., 2024;Wang et al., 2020) consistently reported positive outcomes, with effect sizes generally ranging from small to moderate (SMD ~0.18 to 0.33), aligning with the trends observed in the Healthy and MCI subgroups. For executive function (Fig. 6B), both subgroups showed significant effects, with a small effect observed in participants with MCI (4 studies; SMD = 0.16) and a moderate effect in healthy participants (2 studies; SMD = 0.49). A GRADE assessment was performed to evaluate the overall confidence in the pooled effect estimates (Table 4). Based on the GRADE assessment, the certainty of evidence varied across domains. Global cognitive function was the only outcome supported by 'Moderate' certainty evidence; it was downgraded by one level primarily due to the risk of bias in the underlying primary studies. In contrast, the certainty of evidence for Executive Function, Memory (including overall, verbal, immediate, delayed, and working memory), and Cognitive Inhibition was rated as 'Low'. These outcomes were typically downgraded due to a combination of risk of bias and inconsistency (heterogeneity) or imprecision. Finally, the evidence for Attention/Processing Speed and Cognitive Flexibility was graded as 'Very Low', reflecting serious limitations in inconsistency (I² > 80%) and imprecision (wide confidence intervals). The results of the sensitivity analysis demonstrated that the overall findings were robust. As shown in the supplementary materials (Fig S1 -S7), the sequential removal of each individual study did not substantially alter the magnitude or direction of the pooled effect sizes for global cognitive function, executive function, or memory. In all iterations, the recalculated pooled effect estimates and their 95% confidence intervals remained consistent with the primary analysis, indicating that no single systematic review disproportionately influenced the overall results. This strengthens the confidence in the conclusion that multicomponent exercise has a positive effect on cognitive function in older adults. This umbrella review synthesized evidence from 27 systematic reviews and meta-analyses, providing a comprehensive assessment of the effects of multicomponent exercise (MCE) on cognitive function in older adults. To visually summarize the core findings of this umbrella review and their hierarchy of evidence, a graphical abstract is presented (Fig. 7). The findings overwhelmingly demonstrate that MCE is an effective strategy for improving cognition across a spectrum of older populations. The meta-analytic results revealed statistically significant, moderate positive effects on global cognitive function and executive function. While MCE also conferred a benefit for overall memory (specifically verbal memory), the analyses for working memory, delayed memory, and attention/processing speed did not reach statistical significance, suggesting that the efficacy of MCE may vary across specific cognitive sub-domains. Subgroup analyses further indicated that these benefits are present in both cognitively healthy individuals and those with Mild Cognitive Impairment (MCI). However, this positive conclusion is moderated by a critical appraisal of the evidence base. Despite the consistency of the findings, our methodological quality assessment using AMSTAR-2 revealed that the majority of the included reviews were of "Low" to "Critically Low" quality. Consequently, the certainty of the evidence for most cognitive outcomes, as assessed by the GRADE framework, was judged to be "Low" to "Very Low," with only global cognitive function achieving a "Moderate" rating. This discrepancy between the magnitude of the effect and the quality of the evidence is a central theme of our findings. The consistent finding that MCE enhances Global Cognitive Function and Executive Function aligns with the "Synergistic Hypothesis" of exercise adaptation. Unlike single-mode interventions, MCE targets the cardiovascular, musculoskeletal, and neuromuscular systems simultaneously. Aerobic components are known to increase cerebral blood flow and Brain-Derived Neurotrophic Factor (BDNF) levels, promoting vascular health and neurogenesis (Suzuki et al., 2013;Rondão et al., 2022;Venegas-Sanabria et al., 2022). Concurrently, resistance training stimulates the release of peripheral myokines (e.g., irisin, IGF-1), which can cross the blood-brain barrier to support synaptic plasticity (Rondão et al., 2022). By combining these modalities, MCE likely induces a "multi-pathway" stimulation that offers superior neuroprotection compared to aerobic or resistance training alone. However, our meta-analytic synthesis also revealed a divergence in efficacy across cognitive domains. While Global Cognition and Executive Function showed significant moderate improvements, specific sub-domains such as Working Memory, Delayed Memory, and Attention/Processing Speed did not reach statistical significance. This finding can be interpreted through the "Specificity Principle" of training (Hawley, 2008). Executive functions, which involve planning and inhibition, are inherently engaged during MCE (e.g., learning complex movement sequences, coordinating limbs). In contrast, domains like Working Memory and Processing Speed may require more specific, high-load cognitive demands-such as dual-tasking with high interference-to show robust improvements (Wollesen et al., 2020;Park et al., 2021;Ludyga et al., 2023). The high heterogeneity observed in these non-significant domains suggests that standard MCE programs, which primarily focus on physical outcomes, may not consistently provide the specific cognitive load required to boost these particular faculties. Furthermore, our subgroup analyses suggest that MCE is a versatile intervention, demonstrating efficacy in both cognitively healthy individuals and those with MCI. The comparable effect sizes on global cognition in both groups underscore MCE's dual role: as a preventative strategy to build cognitive reserve in healthy aging and as a therapeutic intervention to slow cognitive decline in at-risk populations. The evidence for frail individuals was slightly more ambiguous but still leaned positive, indicating the need for careful adaptation of programs for this vulnerable group. The substantial statistical heterogeneity observed in our analyses (I² > 70% for several outcomes) warrants careful interpretation regarding the "causes of heterogeneity." Our subgroup analysis (Figure 6) indicates that baseline cognitive status is one contributing factor; for example, distinguishing between participants with MCI and healthy older adults partially explains the variance in treatment response. However, considerable heterogeneity remains even within subgroups. This is likely attributable to the high variability in MCE protocols across the primary studies-specifically differences in the FITT principles (Frequency, Intensity, Time, and Type). The included reviews synthesized programs ranging from moderate to high intensity, with session durations varying from 30 to 90 minutes, and utilizing diverse equipment (e.g., machines vs. elastic bands). Furthermore, the "measurement heterogeneity" poses a challenge, as different neuropsychological tests (e.g., Stroop test vs. Trail Making Test) were often pooled under broad domains like "Executive Function." Despite this variability in the magnitude of the effect, the direction of the findings remains consistently positive across diverse protocols, suggesting that the benefit of MCE is robust to protocol variations. The primary strength of this umbrella review is its comprehensive and systematic approach. By synthesizing a large and recent body of evidence from 27 reviews, we provide a high-level, panoramic view of the current state of the science. The rigorous dual assessment of evidence quality-evaluating the methodological quality of the reviews themselves with AMSTAR-2 and assessing the certainty of the outcomes with GRADE-provides a transparent and critical appraisal of the evidence base, which is a significant contribution to the literature. However, several limitations must be acknowledged. First, as an umbrella review, our findings are inherently constrained by the quality and reporting of the included systematic reviews and the primary studies they contain. The "Low" to "Critically Low" AMSTAR-2 ratings for most reviews indicate that methodological flaws in the evidence base are common, which was the primary reason for downgrading the certainty of evidence in our GRADE assessment. Second, a specific limitation inherent to umbrella reviews is the overlap of primary studies across the included systematic reviews, which can lead to double-counting and potentially inflated statistical precision. To mitigate interpretation errors arising from this issue, we took the precaution of conducting a leave-one-out sensitivity analysis (see Section 3.8 and Supplementary Figures S1-S7). This analysis demonstrated that the removal of any single systematic review did not substantially alter the magnitude or significance of the pooled effect sizes for global cognitive function or executive function. This stability suggests that our conclusions are robust and are not disproportionately driven by a specific subset of duplicated primary trials. Nevertheless, readers should prioritize the consistent direction of the findings across diverse reviews over the precise width of the confidence intervals. Finally, the high heterogeneity across studies, while an important finding, limits our ability to make specific recommendations regarding the optimal design and dosage of MCE protocols. Crucially, the impact of the predominantly "Low" evidence quality ratings (assessed via GRADE) must be interpreted with nuance. This rating does not necessarily imply that the intervention is ineffective; rather, it reflects limited confidence in the precise magnitude of the effect estimate. In exercise science, achieving "High" certainty is inherently challenging because double-blinding (blinding participants to the intervention) is impossible, which automatically triggers specific risk-of-bias penalties in tools like AMSTAR-2 and GRADE. Therefore, while the "Low" quality indicates that future rigorous research (e.g., with blinded outcome assessors) might change the estimated effect size, it should not preclude the clinical implementation of MCE. Given the consistent positive direction of the findings across 27 reviews and the low risk of adverse events associated with exercise, the current evidence remains sufficient to support MCE as a pragmatic public health strategy. Despite the limitations in evidence quality, the consistency and magnitude of the observed effects have clear implications. The findings provide robust support for clinicians and public health bodies to confidently recommend multicomponent exercise to older adults to maintain or improve cognitive function.  For Prevention and Treatment: MCE should be considered a core component of health promotion strategies for cognitively healthy older adults and a frontline, non-pharmacological intervention for individuals with MCI.  Program Design: While an optimal "dose" cannot be definitively prescribed, effective programs should integrate aerobic, resistance, and balance training. The inclusion of components that require motor learning and cognitive engagement appears to be particularly beneficial.  Accessibility: MCE programs are highly adaptable and can be tailored to various functional levels, making them a feasible and scalable public health strategy that can be implemented in community centers, clinics, and long-term care facilities. This review highlights several critical gaps and directions for future research. High-Quality Primary Trials: There is a pressing need for more large-scale, high-quality RCTs with robust methodology, including a priori protocol registration, adequate blinding of outcome assessors, and transparent reporting according to CONSORT guidelines. 1. Standardization: To reduce heterogeneity and enable more meaningful comparisons, the field would benefit from establishing a core outcome set for cognitive assessment in exercise trials In conclusion, this umbrella review confirms that multicomponent exercise is a powerful non-pharmacological strategy for enhancing cognitive function in a broad spectrum of older adults.The evidence strongly supports its positive effects on global cognition, executive function, and , and overall memory performance. However, evidence for improvements in specific sub-domains, such as working memory and processing speed, remains inconclusive and highly heterogeneous. Consequently, the certainty of this evidence is frequently limited by the methodological weaknesses and heterogeneity of the primary literature. These findings underscore the urgent need for promoting multicomponent exercise in clinical and community settings for general cognitive health, while simultaneously calling for more rigorous, standardized, and mechanistic research to optimize intervention protocols for specific cognitive domains. The table provides a detailed summary of the 27 systematic reviews included in this umbrella review. For each review, the table outlines the number and type of primary studies, total participant sample size, key population characteristics, the cognitive status of the participants, a description of the intervention details focusing on multicomponent exercise, the types of comparators used, the main cognitive outcomes assessed, and the primary research question or objective of the review.