Exercise and Brain Health in MCI: What the Best Trials Show

Exercise can help in mild cognitive impairment. It just may not help the way people think.

If you want the clean influencer version, here it is: exercise protects the brain, so do more cardio. The problem is that the best randomized evidence does not cooperate. In older adults with mild cognitive impairment, structured exercise often keeps cognition stable, but higher-intensity aerobic training has not clearly beaten lower-intensity supervised movement on standard cognitive tests.

That is the main lesson from the 2025 EXERT randomized trial, the most informative completed study in this area. It enrolled 296 adults ages 65 to 89 with amnestic mild cognitive impairment across 13 U.S. sites and randomized them to 12 months of supervised moderate-to-high intensity aerobic exercise or a low-intensity program of stretching, balance, and range-of-motion training, followed by 6 months of unsupervised practice. The primary endpoint was ADAS-Cog-Exec, a cognitive measure designed for this population. The result: no differential effect between groups at 12 months. Both groups stayed cognitively stable rather than following the decline the trial had expected in this diagnosis.

That finding does not prove intensity is irrelevant. It does show that the comparison matters. EXERT did not pit aerobic exercise against a couch. It compared one active, supervised intervention with another. The low-intensity arm included trainer contact, scheduled attendance, motor coordination, and balance work. Those are not trivial ingredients. They may be part of the treatment.

The strongest long-term meta-analysis points the same way

EXERT is not alone. In a 2024 systematic review and meta-analysis of 16 randomized trials with 11,402 participants, exercise-alone programs lasting at least 12 months produced a non-significant cognitive effect: standardized mean difference 0.05, with a 95% confidence interval from -0.04 to 0.13 and p=0.25. The same review found no significant reduction in incident mild cognitive impairment, with a risk ratio of 0.98 (95% CI 0.73 to 1.31), and no statistically significant reduction in dementia onset, with a risk ratio of 0.61 (95% CI 0.25 to 1.52). That last estimate looks tempting until you notice the interval. It spans a large possible benefit and a possible increase in risk. In plain statistical terms, it is too imprecise to settle much.

The same review found that multidomain interventions did better. When exercise was combined with cognitive training, nutrition, or vascular risk management, cognitive outcomes improved modestly but significantly: SMD 0.09, 95% CI 0.04 to 0.15, p<0.01. That is still not dramatic, but it is more convincing than the exercise-alone literature.

So the simple story is this: structured activity appears useful for maintaining function, but exercise by itself does not reliably produce measurable cognitive gains over active comparators in long trials. Add cognitive challenge or other interventions, and the signal gets stronger.

Why EXERT matters more than another feel-good exercise study

EXERT is important because it asked a harder question than most trials. It did not just test whether movement beats inactivity. It tested whether aerobic intensity adds something above a credible active control. It also collected Alzheimer's-related biomarkers, including p-tau181, Aβ42/40, and neurofilament light, which matters because cognitive tests may be too blunt over 12 to 18 months to detect modest biological changes.

That biomarker question is where the field gets interesting and frustrating. The trial was designed to examine both cognition and Alzheimer's biology, but the evidence summarized here does not establish a verified between-group biomarker separation. That distinction matters. If the aerobic group and the stretching-balance group had similar cognition but different biomarker trajectories, that would support a disease-biology effect that standard testing missed. If they did not separate on biomarkers either, the case for intensity gets weaker. Based on the evidence available here, the cautious position is that EXERT raised the right mechanistic question but does not yet justify a strong public claim that high-intensity training changed Alzheimer's biomarkers more than the active comparator.

The broader literature is less decisive than the hype suggests

If you read enough exercise-and-brain headlines, you might think the science has already ranked every modality and intensity zone. It has not. A 2025 umbrella review found that most meta-analytic evidence in exercise interventions for cognitive dysfunction rated low to very low certainty under GRADE. Studies varied in population, duration, control condition, and outcome measure. Many used the Mini-Mental State Examination, which has ceiling effects in milder impairment and can miss changes in executive function or episodic memory.

That means two things at once. First, some positive trials may overstate what exercise can do because weak control groups make almost any structured program look good. Second, some null trials may underestimate domain-specific benefits because their cognitive measures are too crude. Both problems exist. Pretending otherwise is how you get a best-selling brain protocol and a very confused reader.

What kinds of exercise look most promising?

The most consistent signals do not come from pure aerobics. In a 2024 systematic review and meta-analysis of randomized trials in mild cognitive impairment and dementia, dual-task training and mind-body exercise, including Tai Chi, showed more consistent benefits for executive function than aerobic exercise alone. The review also suggested that program features such as frequency, session length, and combined physical-cognitive demand matter.

That does not mean Tai Chi has won the tournament bracket. A 2023 network meta-analysis of nonpharmacological interventions in older adults with mild cognitive impairment found substantial heterogeneity across intervention types, control conditions, outcome measures, and study durations. The current evidence can support a short list of plausible winners. It cannot support a precise ranking with much confidence.

The practical implication is still useful: if your goal is brain health, a program that forces attention, coordination, and motor planning may do more than steady-state cardio alone. That could mean dual-task walking, dance-based training, martial-arts forms, or a well-taught Tai Chi class. It means the brain probably likes being asked to do something while the body moves, not merely being transported by the legs like extra luggage.

Exergames are a delivery tool, not a proven upgrade

Technology-mediated exercise gets a lot of marketing oxygen, so it deserves a hard look. A 2022 randomized trial in older adults with mild cognitive impairment found that exergaming-based Tai Chi improved cognitive function and dual-task gait performance compared with a standard control condition. A 2025 systematic review and meta-analysis of digitalized traditional Chinese exercises also reported improvements in physical function, cognition, and quality of life in older adults.

Promising, yes. Case closed, no. The current evidence does not establish that exergaming beats in-person Tai Chi or other matched mind-body exercise when the only thing that changes is the screen. Its main advantage is scalability and adherence. If an app gets someone to practice consistently, that matters. But the format has not earned the right to claim independent cognitive superiority.

The biology is plausible, but plausibility is not proof

The mechanistic story for exercise is stronger than the trial story. A 2023 cohort study found that skeletal muscle mitochondrial function predicted cognitive impairment severity and was associated with Alzheimer's biomarker profiles. A 2024 study in cognitively unimpaired older adults linked exercise with changes in blood-based Alzheimer's biomarkers. Narrative reviews such as a 2026 review describe plausible pathways involving BDNF, neuroinflammation, the HPA axis, and myokine signaling. An in vitro study from 2022 showed secretion of irisin and cathepsin B from contracting engineered muscle.

That body of work supports biological plausibility. It does not prove that these mechanisms translate into meaningful cognitive protection in people with established mild cognitive impairment. Observational associations are not interventions. Biomarkers are not patient outcomes. Engineered muscle is not a 78-year-old with memory complaints and osteoarthritis. The mechanistic case is real, but it is still a bridge, not the destination.

Where resistance training fits

Resistance exercise may be especially relevant in people with sarcopenia or metabolic disease, where the muscle-brain axis has more room to matter. A 2025 narrative review argued that resistance training in mild cognitive impairment plus sarcopenia could act through anti-inflammatory pathways, myokines, and structural brain changes. A 2025 review on type 2 diabetes described shared pathways linking sarcopenia and cognitive decline, including insulin resistance, chronic inflammation, and mitochondrial dysfunction. There is also subgroup evidence such as a 2025 secondary analysis showing Tai Chi improved balance in older adults with type 2 diabetes and mild cognitive impairment.

What is missing is a completed, high-quality randomized trial showing that resistance training as the primary intervention improves cognition in people with both mild cognitive impairment and sarcopenia. The hypothesis is credible. The trial proof is still thin.

What the evidence supports right now

For people with mild cognitive impairment, the best current evidence supports structured physical activity as a way to preserve function, not as a proven way to restore lost cognition or prevent dementia. The strongest long-term randomized data say exercise alone has little measurable advantage over active comparators on standard cognitive tests. Multidomain programs do better. Mind-body and dual-task approaches have the most consistent signals for executive function, but the evidence still does not justify chest-thumping certainty about one perfect modality.

If you are choosing a program for brain health, the practical answer is boring in the best way. Pick something supervised or structured enough that you will actually do it. Include balance, coordination, or cognitive challenge. Keep resistance training in the mix, especially if muscle loss or metabolic disease is part of the picture. And treat any claim that one exercise intensity zone will save your neurons with the suspicion it deserves until a trial shows both biological and clinical benefit against a real active control.

Exercise looks useful for maintaining cognition in mild cognitive impairment. The case that aerobics alone can outperform other structured movement is much weaker than the sales pitch.