A 2023 umbrella review (Blomstrand — 20 SR/MAs, 332 primary studies, healthy adults >=55 without MCI or dementia) reports that mind-body exercise (defined as movement sequences combined with breathing control and attention regulation — Tai Chi, yoga, Pilates) has the greatest cognitive effect of any exercise modality: (Blomstrand et al., 2023) «Mind-body exercise had the greatest effect on cognition with a pooled effect size of d = 0.48 (SE = 0.06; p < 0.001)», against aerobic and resistance. The review recommends mind-body exercise (Blomstrand et al., 2023) «serve as a complement to other types of exercise».
The decision question this page answers is narrower than the review’s headline: does the mind-body label do work that generic physical activity does not — is a person already doing aerobic/resistance activity given a distinct reason to add Tai Chi or yoga for cognition? The finding below is that the umbrella’s own numbers do not establish this: the apparent modality advantage is most parsimoniously a comparator + heterogeneity artifact, not evidence that the mind-body ingredient carries a cognition-specific effect. This is Is the Food Category Doing Any Work applied to an exercise modality.
The reported effect sizes (all pooled Cohen’s d, random-effects)
Effect sizes were pooled from each constituent review’s reported d — (Blomstrand et al., 2023) «No re-analysis of raw data from reviews included in this study was performed.» So every contrast below is a cross-review comparison of separately-pooled estimates, each measured against its own control, not a head-to-head test.
| Contrast (pooled across separate SRs) | d | SE | p | Reading |
|---|---|---|---|---|
| Overall exercise -> cognition | 0.22 | 0.04 | <0.01 | small |
| Global cognition (domain) | 0.43 | 0.11 | <0.001 | moderate |
| Executive function (domain) | 0.26 | 0.07 | <0.001 | small |
| Memory (domain) | 0.20 | 0.05 | <0.001 | small |
| Attention (domain) | 0.20 | 0.11 | 0.01 | small |
| Processing speed (domain) | 0.21 | 0.05 | <0.001 | small |
| Mind-body (type) | 0.48 | 0.06 | <0.001 | moderate — greatest |
| Aerobic (type) | 0.17 | 0.04 | <0.001 | small |
| Resistance (type) | 0.24 | 0.24 | 0.32 | non-significant |
| Mixed (type) | 0.18 | 0.05 | <0.001 | small |
(Blomstrand et al., 2023) «Aerobic exercise had a small effect on cogni-tion (d = 0.17; SE = 0.04; p < 0.001), as did resistance exer-cise (d = 0.24; SE = 0,24; p < 0.32). The effect of mixed exercise on cognition was also small (d = 0.18; SE = 0.05; p < 0.001).» Note the resistance estimate is non-significant (p = 0.32, SE = effect), so the modality ranking rests on a mind-body-vs-aerobic gap (0.48 vs 0.17), the aerobic arm being the only well-estimated comparator.
The umbrella’s own credibility signals
- Constituent quality is moderate, not high. AMSTAR-2 overall confidence: «high in six studies, moderate in 12 studies and low in two studies» (one further review removed for critical flaws). Review overlap was slight (CCA 1.84%), so double-counting across reviews is not the problem — the constituent evidence quality is.
- Publication bias is present and uncorrected. «Variation across studies due to heterogeneity was very high (I2 = 85%)», and «Egg-er’s test including all data revealed a significant devia-tion from zero (β0 = 0.23; CI = 0.107-0.350; t = 3.783; p < 0.001) confirming that small study effects may have influenced the results.» The pooled d values above are not corrected for this.
- The comparators are mostly passive. Most constituent reviews used passive control groups (no training, health education, stretching); only two (Clifford, Jiang) included active/exercise controls. The review states the consequence itself: reviews that «included controls who undertook another form of exer-cise … would likely result in smaller effect sizes».
Why the mind-body advantage is not established (echo + failed friction claim)
The umbrella asserts a modality advantage (friction: mind-body > generic PA). On appraisal the claim does not survive its own limitations. Five reasons, in order of force:
- Not head-to-head — the crux. d = 0.48 (mind-body) vs d = 0.17 (aerobic) is a between-subgroup comparison of effect sizes pooled from different sets of reviews, each vs its own control, with «No re-analysis of raw data». A modality difference is only credibly established within trials that randomise the same people to mind-body vs another active modality. No such comparison is made.
- Comparator confound (active-vs-passive). Mind-body trials (yoga/Tai Chi/Pilates) run largely against passive controls, so their effect bundles expectancy, social contact, novelty and attention with any modality-specific ingredient. The review’s own caveat — active controls shrink the gap — points the artifact in exactly the direction of the apparent advantage.
- Publication bias, confirmed and uncorrected here, erases the effect elsewhere. The review cites a broader umbrella (23 MAs, ages 6-80) in which «these effects became negligi-ble after correcting for publication bias». With Egger significant (p < 0.001) and no correction applied, the pooled magnitudes are an upper bound, not a point estimate.
- Pooling magnitudes across I2 = 85% is configure-not-aggregate territory. The pooled d averages over 40+ heterogeneous memory tests and heterogeneous protocols; a washed-out average across a heterogeneous set strips the mechanism and answers nobody’s question (the wiki’s own synthesis-mode rule). The number is fragile; the shape (all effects small, modality unresolved) is the robust object.
- Internal inconsistency in the headline. The largest effect the review found was on global cognition (d = 0.43) — and in the type-stratified analysis global cognition was investigated only for aerobic (d = 0.51) and resistance (d = 0.68), not mind-body. The «greatest effect on cognition» claim for mind-body therefore rests on the executive-function / processing-speed / memory subsets, excluding the domain where exercise’s largest effects actually sat.
Reading: this is an echo of held fabric — physical activity is associated with lower dementia risk / better cognition, weakly and with confounding (Dementia Prevention and Modifiable Risk Factors: activity RR 0.80; Multidomain Lifestyle Intervention and Cognitive Decline: structured exercise did not move cognition in RCTs) — plus a modality-specific friction claim that fails its own appraisal. The mind-body label’s apparent extra work runs through the comparator and the measurement heterogeneity, not through a demonstrated modality-specific mechanism. Same structure as the food-category diagnostic: a category boundary that looks load-bearing until you ask whether the better-designed contrast exists (it does not) -> Is the Food Category Doing Any Work.
Independence — this is NOT type-E corroboration
This umbrella pools the same observational and small-RCT exercise-cognition literature the fabric
already holds. An umbrella over that literature is not an independent line of evidence: it shares the
constituent studies, their confounding structure, and their passive-control design. It therefore does
not raise confidence in exercise helps cognition by convergence, and it earns no [E-independent]
mark and no sources: slot on the dementia or multidomain pages. Its value is type F (it bounds and
distinguishes the modality sub-claim) plus type G (it surfaces the head-to-head gap), not type E.
Decision relevance (Layer-1: peripheral, small, low-certainty)
- For a reasonably-healthy older adult already doing aerobic/resistance activity, this evidence gives no distinct reason to add mind-body exercise specifically for cognition at the margin. The modality advantage is not established, and even the pooled cross-type effects are small (d ~0.17-0.24) with the clinical-translation question open in the review’s own words.
- Mind-body exercise may still be chosen for other, separately-evidenced outcomes — balance and fall prevention (Exercise for Preventing Falls in Older Adults), joint function, or pain — where it has its own evidence base. That is a different decision from a cognition lever.
- The big rocks are unchanged. For cognition/dementia the dominant modifiable levers remain the cardiometabolic and physical-inactivity rocks on Dementia Prevention and Modifiable Risk Factors; modality choice within being active is a refinement, not a rock.
Gap — what would settle it (G)
A head-to-head RCT of mind-body exercise vs an active-exercise control (matched contact time, dose, and expectancy) on cognitive outcomes in healthy older adults. Until such a comparison exists, the modality difference cannot be separated from the comparator. — a directional-only candidate lever, held as a named gap, not a finding.
Self-critique
- Counter-passage check: run. The review’s Discussion and Conclusions were read end to end. It does assert the modality advantage and recommends mind-body exercise — but it also states the passive-control caveat, the I2 = 85%, the significant Egger test, «No re-analysis of raw data», and cites a broader umbrella where effects vanish after bias correction. The page leads with the review’s headline claim and then appraises it, rather than quoting the caveats against a claim the review does not make.
- Over-claim check. The page does not say mind-body exercise is useless for cognition, nor that the effect is zero — only that a modality-specific advantage over generic PA is not established and is most parsimoniously a comparator/heterogeneity artifact. The uncertainty is stated as uncertainty (confidence: low, single source).
- Independence guard. Explicitly refused type-E: no
[E-independent], nosources:padding on the dementia/multidomain pages. This is the load-bearing discipline for an umbrella of already-held literature. - Symmetric standards. The publication-bias and passive-control critiques are the same appraisal applied to any unblindable-exercise effect estimate; they are not extra scrutiny reserved for a claim the maintainer dislikes. The resistance-arm non-significance (p = 0.32) is reported as cutting against a clean modality ranking, not suppressed.
- Residual: one gold-tier source, moderate-quality constituents. The conclusion rests on the review’s own internal limitations plus the food-category diagnostic; a head-to-head RCT could overturn it in either direction.