A comprehensive Bayesian network meta-analysis of 38 randomized controlled trials involving 4,047 community-dwelling older adults has identified specific exercise modalities and weekly volumes that are most effective for preserving global cognitive function. Published in the European Review of Aging and Physical Activity, the study distinguishes between healthy older adults and those with pre-sarcopenia, a condition characterized by reduced muscle strength or performance. The research demonstrates that while all structured exercise formats offer cognitive benefits compared to non-exercise controls, the optimal type of training varies significantly based on the participant’s muscle health status. This finding provides a critical framework for clinicians seeking to prescribe precision exercise interventions for cognitive health in aging populations.
The study was designed to address two major knowledge gaps in the existing literature: the lack of comparative evidence across different exercise modalities and the unclear dose-response relationship for cognitive outcomes. Previous reviews often pooled heterogeneous groups of older adults, overlooking the distinct physiological and neurocognitive responses between healthy individuals and those with suspected sarcopenia. By stratifying the analysis by phenotype and age, the researchers aimed to determine how modality and dose interact to influence cognitive outcomes. The primary outcome was global cognition, assessed using validated instruments such as the Mini-Mental State Examination or the Montreal Cognitive Assessment.
The network meta-analysis incorporated data from trials conducted between 2005 and 2025 across 16 countries, with significant representation from Asia and Europe. The interventions included aerobic exercise, resistance training, multimodal programs combining multiple modalities, and other mind-body or skill-based activities like Tai Chi and yoga. To harmonize the heterogeneous reporting of exercise intensity across trials, the researchers standardized all interventions into a unified metric of energy expenditure, measured in MET·min/week. This approach allowed for comparable dose-response analyses across different exercise types, enabling the identification of minimum effective doses and optimal ranges.
In the pooled analysis of all participants, multimodal exercise, aerobic exercise, and resistance training each yielded significant and clinically meaningful improvements in global cognition compared to non-exercise controls. Multimodal programs showed a standardized mean difference of 0.68, aerobic exercise 0.67, and resistance training 0.58. In contrast, other interventions such as Tai Chi, yoga, or dance showed smaller and statistically non-significant effects. The evidence network was well-connected, with aerobic exercise versus control and aerobic versus multimodal exercise being the most frequently tested contrasts, which influenced the density and stability of the effect estimates.
When stratified by health status, the pattern of effects diverged markedly. In healthy older adults, aerobic exercise and resistance training produced the largest gains, with standardized mean differences of 0.88 and 0.80, respectively. Both surpassed the conventional threshold for a large standardized mean difference, indicating substantial cognitive benefits. Multimodal programs delivered a moderate but imprecise effect in this group. Conversely, in pre-sarcopenic adults, multimodal interventions achieved the highest benefit with a standardized mean difference of 0.60. Aerobic and resistance training also showed moderate and significant gains in this subgroup, but to a lesser extent than the multimodal approach.
The dose-response modeling revealed that cognitive improvement rose steadily with total weekly exercise dose up to approximately 1,400 MET·min/week. Each additional 100 MET·min/week was associated with a 0.31-point increase in cognitive scores, supporting the conclusion that higher volumes, within feasible limits, are beneficial. In healthy adults, the dose-response curve showed a near-linear pattern, with gains evident at the lower boundary of the World Health Organization recommended range and continuing beyond 1,200 MET·min/week. This suggests that progressive volume increases in aerobic or resistance training are both safe and advantageous for those with preserved physiological reserves.
In contrast, the pre-sarcopenic subgroup displayed a non-linear curve, with cognitive gains peaking around 700 to 800 MET·min/week. This range is close to the clinically derived minimal clinically important difference reference. Beyond this point, the curve dipped slightly across 1,000 to 1,200 MET·min/week before a possible rebound at very high volumes, although these estimates were less precise due to limited data. This suggests an optimal range of efficiency at moderate doses for this phenotype, where increases in volume offer diminishing returns and may reduce program adherence.
Age also appeared to modify the comparative effects of exercise interventions. In participants aged 70 years and older, the network was more control-centered, with larger nodes for multimodal and aerobic interventions. These modalities yielded the most stable and sizeable gains over control in this age group. In participants younger than 70 years, the evidence base was dominated by resistance training, which produced the most reliable improvements. The asymmetry in evidence geometry suggests that younger-old adults may derive greater cognitive gains from high-intensity, single-modality programs, while the oldest-old benefit more from integrative, multimodal approaches.
The study also identified a duration-dependent divergence in the comparative effectiveness of exercise modalities. For interventions lasting 12 weeks or less, aerobic exercise demonstrated the highest probability of being the most effective, followed by resistance training. In contrast, for interventions exceeding 12 weeks, multimodal interventions attained the highest ranking probability. This time-dependent effect suggests that aerobic and resistance modalities may be more effective in the short term, whereas prolonged interventions appear to favor multimodal approaches, likely due to the gradual accumulation of cross-domain adaptations.
Limitations of the study include the heterogeneity of exercise interventions, with wide variation in session duration, frequency, and program length. The grouping of Tai Chi, yoga, and dance under a single ‘other’ category may have obscured potentially meaningful differences between these modalities. Additionally, estimates for the pre-sarcopenic subgroup were associated with wider credible intervals due to the smaller number of trials enrolling these participants. The use of standardized mean differences may also obscure domain-specific cognitive patterns, such as differences between executive function and memory.
Despite these limitations, the translational message is clear: exercise functions as a scalable, safe, and clinically effective strategy to support late-life cognition. Matching modality and dose to physiological reserve and phenotype is key to maximizing benefits while minimizing burden. For healthy older adults, aerobic or resistance training at World Health Organization recommended volumes delivers maximal neurocognitive benefits. For pre-sarcopenic individuals, integrated multimodal programs of approximately 700 to 800 MET·min/week appear to balance efficacy with sustainability. These findings provide actionable targets for clinicians and policymakers to advance precision exercise prescriptions for cognitive health in aging populations.
Future trials should include more head-to-head comparisons across exercise modalities, improve reporting of exercise dose and adherence, and incorporate longer follow-up periods to examine the durability of cognitive effects. By aligning comparative effectiveness, dose-response evidence, and population heterogeneity, this study offers a practical blueprint for precision exercise approaches in cognitive aging. The results underscore the importance of tailoring exercise interventions to specific cognitive goals, timeframes, and individual physiological profiles to optimize outcomes in diverse aging populations.
A closer examination of the trial architecture illustrates the distribution of participants across different exercise modalities. Across the 38 trials, 29 followed two-arm designs and nine employed multi-arm protocols, contributing a total of 38 non-exercise control arms alongside distinct intervention arms. Aerobic exercise was evaluated in 20 arms encompassing 907 participants, resistance training appeared in 16 arms with 785 participants, multimodal regimens spanned nine arms with 457 participants, and other formats such as mind-body activities comprised seven arms with 351 participants. Intervention parameters varied widely across individual programs, with session durations spanning from 15 to 110 minutes, weekly frequencies ranging from one to seven sessions, and overall program lengths running from 4 to 52 weeks. Global cognitive outcomes were tracked primarily through the Mini-Mental State Examination across 26 trials and the Montreal Cognitive Assessment across 12 trials, with investigators prioritizing the latter in primary analyses because of its heightened sensitivity to subtle cognitive decline.
Methodological evaluation using the Cochrane Risk of Bias tool revealed distinct patterns in trial rigor and reporting practices. Random sequence generation was rated as low risk in 97.4 percent of trials, and incomplete outcome data, selective outcome reporting, and other sources of bias were judged to be at low risk across all included studies. However, blinding of participants and personnel was rated as unclear in 100 percent of trials, reflecting the inherent practical impossibility of concealing physical exercise assignments from participants or trainers. Furthermore, allocation concealment was judged low risk in only 28.9 percent of the trials, remaining unclear in 71.1 percent, while blinding of outcome assessors remained unclear in 36.8 percent. Because the primary cognitive screening instruments rely entirely on assessor administration, these gaps in allocation concealment and outcome blinding could introduce potential selection or detection bias into pooled estimates.
Statistical modeling also confirmed precise quantitative differences when trials were stratified by participant age and intervention duration. In adults under 70 years, resistance training produced a standardized mean difference of 0.92 with a 95 percent credible interval of 0.55 to 1.28, and aerobic training achieved an effect size of 0.85 from 0.49 to 1.21. For individuals aged 70 and older, multimodal exercise yielded an effect of 0.68 from 0.40 to 0.96, aerobic training reached 0.65 from 0.37 to 0.92, and resistance training produced 0.51 from 0.23 to 0.79. Cumulative ranking curves showed aerobic exercise leading short-term interventions of 12 weeks or less with a SUCRA value of 0.919, whereas multimodal programs dominated regimens exceeding 12 weeks with a SUCRA value of 0.886.
Subject of Research: Optimal exercise type and dose for cognitive function in healthy and pre-sarcopenic older adults
Article Title: Optimal type and dose of exercise to improve cognitive function in healthy and pre-sarcopenic older adults: a bayesian network meta-analysis of randomized controlled trials
Article References: Yang, Y., Pan, N., Liu, Y., Xu, W., & Ossowski, Z. (2026). Optimal type and dose of exercise to improve cognitive function in healthy and pre-sarcopenic older adults: a bayesian network meta-analysis of randomized controlled trials. European Review of Aging and Physical Activity, 23(1), Article 9. https://doi.org/10.1186/s11556-026-00404-2
Image Credits: AI Generated
DOI: 10.1186/s11556-026-00404-2
Keywords: cognitive function, exercise prescription, older adults, pre-sarcopenia, network meta-analysis, dose-response, aerobic exercise, resistance training, multimodal exercise, aging, Optimal, type
News Source: Beatrice Stafford. (November 13, 2026). Exercise Dose and Type for Older Adult Cognition. Scienmag.
