Muscles and the Brain

Why this is worth being careful about

The claim that lifting weights protects your brain is everywhere, and it sits on two very different bodies of evidence that get quoted as though they were one.

The first is trials: give older people an exercise programme, measure their cognition, see what changes. That evidence is real and it is human.

The second is mechanism: muscle releases signalling molecules, some of those reach the brain, and in animal models they do impressive things to memory pathways. That evidence is real and it is almost entirely not human.

Most articles blend the two, so the mouse biology lends borrowed authority to the human result. This piece keeps them apart, because the honest picture is more useful than the exciting one.

The trial evidence, at proper scale

A systematic review and network meta-analysis compared exercise-based interventions for cognitive and psychological outcomes in older adults with mild cognitive impairment. It followed a registered protocol and searched PubMed, Embase, the Cochrane Library and Web of Science from inception to March 2026.

A registered protocol means the analysis plan was published before the results were known. That matters more than most readers realise, because it removes the freedom to go looking for a favourable answer afterwards.

Seventy-four randomised controlled trials involving 5,578 participants and 12 intervention types were included. The intervention nodes covered aerobic exercise, resistance training, mind-body exercise, multicomponent exercise, digital or virtual-reality exercise, cognitive training, and combined exercise with cognitive training. Risk of bias, rankings and certainty were assessed with RoB 2, SUCRA and GRADE.

Twelve different things compared against each other in one model. That is what a network meta-analysis buys you: comparisons between interventions that were never tested head to head.

Against wait-list control, mind-body exercise was associated with higher MoCA scores, a mean difference of 3.94 with a confidence interval from 2.28 to 5.59, and higher MMSE scores, mean difference 4.52, interval 2.34 to 6.70, at moderate certainty. Resistance training was associated with lower ADAS-Cog scores, where lower means better.

Moderate certainty, on a GRADE scale that runs to high. That is the ceiling this field reaches, and it is the honest level to hold this at.

Note also which modality topped that particular outcome: mind-body exercise, the tai chi and yoga end of the spectrum, not the weight room. Resistance training performed on a different cognitive scale. Anybody telling you one specific type of training is definitively best for your brain is reading past what this analysis actually supports.

This was conducted in older adults who already had mild cognitive impairment, so it does not describe a healthy 30-year-old. Network rankings are less stable than they appear, MoCA and MMSE are screening instruments rather than measures of daily life, and cognition is a matter for a clinician.

The molecule everybody quotes

If you have read about muscle and the brain, you have read about irisin. Here is exactly what that evidence is.

A systematic review summarised preclinical in vivo studies of exercise-induced FNDC5 expression and irisin-related signalling in experimental models of Alzheimer's disease, searching PubMed and EMBASE following PRISMA.

Preclinical. In vivo. Experimental models. Three phrases in one sentence, all meaning the same thing: animals.

Ten preclinical studies met the inclusion criteria. The evidence indicated that exercise was associated with increased FNDC5 expression, activation of irisin-related pathways, and downstream neuroprotective mechanisms including PPARGC1A, AMPK, SIRT1 and BDNF signalling. Those adaptations were associated with improvements in synaptic plasticity and reductions in neuroinflammation and oxidative stress.

That is a genuinely interesting mechanism and it is ten studies in animals.

I am including it because you will meet this claim, presented as though it describes you. It does not yet. Mouse neurobiology has a long history of not surviving translation to people, and a pathway that works in a mouse model of Alzheimer's is a hypothesis about humans, not a finding about them.

This is a qualitative synthesis of ten animal studies. It establishes a plausible mechanism, not an effect in any person, and nothing in it should change what you do.

Muscle as an organ that signals

The broader idea underneath is better established, and worth understanding on its own terms.

A review describes skeletal muscle, once regarded solely as contractile tissue, as a dynamic endocrine organ secreting exercise-induced myokines: bioactive peptides with autocrine, paracrine and endocrine functions that regulate glucose and lipid metabolism, mitochondrial function, inflammation and communication between organs.

An endocrine organ. Your muscle is not just machinery that moves you, it is tissue that talks to the rest of you chemically when you use it.

The review covers established and emerging myokines including interleukin-6, irisin, myostatin, GDF11, IL-15, BDNF, meteorin-like, SPARC, FGF21, BAIBA, LIF, apelin and musclin, across skeletal muscle, liver, adipose tissue and bone. BDNF, brain-derived neurotrophic factor, is the one most directly tied to the brain.

And the authors state the limitation themselves, which is why this review is worth citing rather than a more excitable one: the current literature indicates substantial heterogeneity in causal evidence, receptor or target certainty, and translational relevance.

Substantial heterogeneity in causal evidence. From the people who study it for a living.

So the framework is sound: contracting muscle releases signals that act elsewhere in the body. Which specific signal does which specific thing in a human brain is a great deal less settled than the popular version suggests.

What I would actually take from this

Train, and do not do it because of irisin. Do it because 74 randomised trials in nearly 5,600 people found exercise interventions improved cognitive scores in a population at risk, which is a far better reason than any molecule.

Do not pick your training type on brain grounds. Mind-body exercise topped one outcome, resistance training another, and the certainty is moderate at best. There is no cognitive case for abandoning what you already enjoy and will keep doing.

Treat the mouse work as interesting, not as instruction. Ten animal studies describe a mechanism worth researching. They do not describe your brain.

And understand what these trials cannot tell you. Improving a screening score in people with mild cognitive impairment is not the same as preventing dementia in a healthy adult, and no study here tested that.

These are my conclusions from the material rather than clinical recommendations. Nothing in this article is medical advice, and anything concerning memory, cognition or dementia risk belongs with a qualified clinician.