What exercise has been shown to do, and what it hasn't

Exercise is the closest thing Parkinson's has to a consensus. Neurologists recommend it, physiotherapists build programmes around it, and it is the first thing most people are told after a diagnosis. That near-unanimity makes it worth asking a blunt question: shown to do what, exactly?

The answer splits cleanly in two, and the halves are not equally strong.

The part that is well established

The largest synthesis is a Cochrane network meta-analysis published in 2023, pooling 156 randomised trials and 7,939 participants. It found beneficial effects on the severity of motor signs and on quality of life for most types of exercise compared with doing nothing.

Its more interesting finding is the one about which exercise. Across outcomes, the reviewers "observed little evidence of differences between exercise types" — and concluded that while exercise matters for motor signs and quality of life, "the exact exercise type might be secondary".

That is a more useful result than it first appears. Much of the advice in this area is a case for one modality over another: boxing over cycling, dance over treadmill, this programme over that one. On the pooled evidence, the argument is mostly beside the point. The thing that separates benefit from no benefit is doing it.

Intensity looks like it matters more than type

The trial usually cited here is SPARX, published in 2018, which randomised 128 people with newly diagnosed Parkinson's — none yet on medication — to high-intensity treadmill exercise at 80–85% of maximum heart rate, moderate intensity at 60–65%, or a wait-list control, four days a week for six months.

Motor scores worsened by 3.2 points in the usual-care group and by 0.3 points in the high-intensity group. The high-intensity group met the trial's threshold; the moderate-intensity group did not.

One thing about that trial is almost always dropped when it is quoted. It was a futility trial. Its job was not to prove exercise works — it was to decide whether a phase 3 trial was worth running. The high-intensity result cleared that bar. It was designed to answer a question about what to do next, and it is routinely cited as though it answered the question itself.

The strongest single symptomatic result comes from elsewhere: a 2019 Dutch trial of home-based aerobic exercise, remotely supervised, in 130 people. At six months the difference in off-state motor score was 4.2 points in favour of the exercise group (95% CI 1.6–6.9, p=0.0020). Off-state matters here — it is the measure taken when medication is at its lowest, so it is harder to explain away as a drug effect.

For balance specifically, a 2012 trial randomised 195 people to tai chi, resistance training or stretching, twice weekly for 24 weeks. Tai chi beat both comparators on postural control, beat stretching on every secondary outcome, and lowered the incidence of falls compared with stretching — though not compared with resistance training. The effects were still present three months after the programme ended.

The part that has not been shown

None of this establishes that exercise slows Parkinson's.

That is a different claim, and it is the one people most want to make. It is being tested: SPARX3, the phase 3 trial that the 2018 futility result was designed to justify, published its protocol in 2022. Its results are not out. The Dutch trial's authors said the same thing in their own conclusion — that future studies should establish long-term effectiveness "and possible disease-modifying effects".

So the honest position today is that exercise has good evidence for treating symptoms and improving quality of life, and no completed randomised evidence that it changes the course of the disease. Those are often said in the same breath. They are not in the same category.

What else to hold in mind

The trials are small. Across the 156 in the Cochrane review, the mean was 51 participants, ranging from 10 to 474. The reviewers were explicit that "larger, well-conducted studies are needed to increase confidence in the evidence".

Who was studied. Mostly mild to moderate disease, and mostly without major cognitive impairment. The review notes that further studies recruiting people with more advanced disease and with cognitive impairment "might help extend the generalizability of our findings".

Safety was not well reported. Falls and pain were the most frequent adverse events; the evidence about the risk of adverse events was rated very low confidence, though the interventions were described as relatively safe. Very low confidence is not the same as reassurance — it means the studies did not measure this well enough to say.

Where that leaves it

Exercise has better evidence behind it than almost anything else offered alongside medication in Parkinson's, and the case for doing it does not depend on the disease-modification question at all. Motor signs and quality of life are worth improving on their own terms.

What the evidence does not currently support is the stronger version of the claim — that a person can slow their own disease by training hard enough. That version is being tested properly, and until it reports, it is a hypothesis with a trial attached rather than a finding.

Which programme to do, at what intensity, and what is safe given your own balance and cardiac fitness, is a question for a neurologist or a physiotherapist who knows Parkinson's. The evidence above is about groups; the prescription is about you.