MCI Exercise Study: Tai Chi Beats Cardio, With Caveats

Tai chi may help mild cognitive impairment more than standard cardio

If the goal is protecting cognition in mild cognitive impairment, the usual advice to walk more may be too generic. A 2026 meta-analysis by Wang et al. found that mind-body exercise improved cognition across all three major scales it analyzed, while structured aerobic exercise reached statistical significance on only one. That does not prove tai chi or yoga are inherently superior to walking or cycling. The comparison was indirect, the trials were heterogeneous, and the better aerobic studies tell a less tidy story. Still, the signal is strong enough to take seriously.

The paper pooled 26 randomized controlled trials with 2,555 older adults with MCI. Mind-body exercise, driven mostly by tai chi trials, improved MMSE by 1.27 points versus inactive control (95% CI, 0.99 to 1.55), MoCA by 1.89 points (95% CI, 0.78 to 3.00), and ADAS-Cog by -2.09 points (95% CI, -2.94 to -1.25), with all three results statistically significant. Structured aerobic exercise, mainly walking and cycling, improved ADAS-Cog by -1.41 points and reached significance there, but not on MMSE (mean difference 0.37, P=.21) or MoCA (mean difference -0.49, P=.26) [Wang et al., 2026].

That headline result is counterintuitive because aerobic exercise is the default recommendation for brain health. It also matters because clinicians often talk about exercise as if modality barely matters. This analysis suggests the details may matter a lot.

Why mind-body exercise might do better

The simplest hypothesis is that tai chi and related practices ask the brain to do more at once. They combine balance, motor sequencing, attention, spatial control, and breath regulation. Walking on a flat path can be healthy and useful, but it is usually less cognitively demanding unless you add hills, intervals, navigation, or some reason not to drift into autopilot.

That mechanistic story has some human data behind it. An fMRI study in people with MCI by Jiao Liu et al. (2021) found that mind-body exercise altered functional connectivity in the locus coeruleus and ventral tegmental area, two hubs involved in attentional arousal and dopaminergic signaling. That does not prove a treatment mechanism, but it does show that this kind of training reaches brain networks that plausibly matter for cognition.

Structural data point in the same direction. A 2022 imaging trial by Zheng et al. reported gray matter volume changes after six months of Baduanjin, a traditional Chinese mind-body exercise, along with cognitive improvement in older adults with MCI. Another randomized trial, Xia et al. (2023), found cognitive gains with Baduanjin in community-dwelling older adults with MCI, which helps a little with generalizability beyond tai chi itself.

There is also a biomarker thread, though it is still a thread, not a rope. Lin et al. (2024) reported higher circulating exosomal LRP1 after tai chi in people with amnestic MCI, with associations to memory function and hippocampal plasticity. LRP1 relates to amyloid clearance and synaptic biology, so the result is interesting. It is still a peripheral biomarker finding, not proof that tai chi slows Alzheimer disease.

Behavioral evidence supports the idea that cognitively demanding movement may train a broader system than plain aerobic load. In a 2022 randomized trial, exergaming-based tai chi improved both cognitive function and dual-task gait performance in older adults with MCI. Dual-task gait is useful here because it forces motor and executive systems to work together. If an exercise improves that, it is doing more than raising heart rate.

What the meta-analysis can and cannot prove

The most important limitation is structural. Wang et al. did not randomize participants to tai chi versus walking within the same trials. Instead, they pooled mind-body trials against their own inactive controls and aerobic trials against their own inactive controls, then compared the patterns. That is an indirect comparison. It can generate a strong hypothesis, but it cannot settle a head-to-head clinical question.

The paper also appears heterogeneous. The MoCA confidence interval for mind-body exercise runs from 0.78 to 3.00, a wide span that suggests substantial between-trial variability. The authors flagged heterogeneity as a limitation. That means the pooled estimate is an average across quite different studies, not a fixed dose-response law handed down from the exercise gods.

The aerobic MoCA result needs special caution. The pooled mean difference was -0.49 versus control, meaning aerobic participants scored slightly worse than inactive controls on that scale. The available evidence does not explain why. Differential dropout, odd control groups, or regression to the mean are more plausible than the idea that walking harms cognition. But unexplained anomalies make the comparison harder to trust.

Another limitation is geography. Much of the mind-body literature, especially the tai chi and Baduanjin trials, comes from Chinese populations. That matters because cultural familiarity, instructor quality, participant expectations, and adherence may differ in the United States or Europe. It also matters because tai chi dominates the pooled mind-body category. The data do not show that yoga has the same effect at the same dose.

The evidence against overselling the result

If you stopped at the 2026 meta-analysis, you might conclude that aerobic exercise mostly fails in MCI. That would be too strong. The EXERT study by Baker et al. (2025) is the main reason. It was a U.S.-based randomized trial in amnestic MCI with structured aerobic exercise and Alzheimer biomarker endpoints, and it found meaningful cognitive and biomarker outcomes. EXERT does not erase the Wang signal, but it does show that well-designed aerobic exercise can work in this population. A pooled null from mixed-quality aerobic trials is not the same thing as proof that cardio is second-rate.

The broader evidence hierarchy also raises a bias concern. An umbrella review by Sun et al. (2025) examined exercise meta-analyses for cognitive dysfunction and emphasized the usual problems: small-study effects, inconsistency, and likely publication bias. That warning lands hard on a literature dominated by small, positive, geographically concentrated mind-body trials. The directional signal may be real while the effect size is inflated. Those two ideas can both be true.

Then there is the direct-comparison problem. One of the few relevant head-to-head trials is Chen et al. (2025), which compared tai chi with aerobic exercise in older adults with type 2 diabetes using wearable monitoring in a mobile health model. It was not a pure MCI trial, so it cannot answer the exact question. But if direct comparisons in cognitively at-risk groups show similar outcomes between tai chi and walking, the most likely explanation for the meta-analytic gap shifts away from magic movement patterns and toward trial quality, supervision, dose, or adherence.

A related randomized trial in a closer population, Chen et al. (2023), found cognitive benefit with tai chi in adults 60 or older who had both type 2 diabetes and MCI. That is useful support for tai chi as an active intervention. What it does not prove, based on the evidence available here, is that tai chi beat its active comparator on cognition.

What to do with this now

The practical implication is modest but actionable. If an older adult has diagnosed MCI and currently relies mostly on walking or cycling for exercise, the published evidence supports adding a mind-body modality, especially tai chi, rather than assuming all exercise will affect cognition equally. In the Wang meta-analysis, the best-supported dose was at least 20 weeks, at least 60 minutes per session, and at least three sessions per week. A 2024 meta-analysis by Wenxin Chen et al. also supports the idea that duration, frequency, and exercise type matter for cognitive outcomes in MCI and dementia.

This is not a reason to abandon aerobic exercise. Aerobic training still has strong cardiovascular, metabolic, and functional benefits, and EXERT suggests it may help cognition under the right conditions. The more defensible conclusion is narrower: for cognition in MCI, modality may matter more than guidelines currently acknowledge, and mind-body exercise deserves a place near the front of the line.

Current evidence supports a prescription change in tone, not a declaration of victory. Tai chi looks better than standard cardio in indirect pooled data for MCI, but the field still needs large head-to-head trials in diverse populations before anyone should pretend the case is closed.

For now, the sober version is this: if you want one exercise plan for brain health in MCI, make it broader than cardio. Add movement that forces attention, balance, and coordination. The brain rarely rewards monotony.