Parkinson's disease is the one condition in this whole territory where the treatment genuinely works. Levodopa is probably the most effective symptomatic drug in neurology. People go from barely able to rise out of a chair to walking across a room.
And there is nothing — not one licensed drug, not one supplement, not one intervention of any kind — shown to slow the disease down.
Those two facts get run together constantly, in both directions. "There's no treatment" is wrong and discouraging. "There's a treatment" gets heard as "it's being held back", which is also wrong. This post keeps them apart, and it opens our series on the neurological conditions beyond dementia. We are not doctors, and nothing here is treatment advice.
Parkinson's disease is the progressive loss of dopamine-producing neurons in the substantia nigra, a small region of the midbrain, driven by aggregation of the same misfolded protein — alpha-synuclein — that causes Lewy body dementia. They are relatives, not strangers, and that matters later on this page.
The motor picture:
The non-motor picture usually starts earlier and is usually what damages life more: constipation, loss of smell, acting out dreams in sleep, depression and anxiety, blood pressure that drops on standing, pain, fatigue, and in many people cognitive change later on.
In the Lewy body post we covered REM sleep behaviour disorder: across 1,280 patients at 24 centres, isolated RBD converted to an overt neurodegenerative syndrome at 6.3% per year and 73.5% by twelve years, with cognitive markers at baseline being the only thing that separated those heading for dementia from those heading for parkinsonism.
There is a second prodromal syndrome, and a 2026 systematic review in JAMA Neurology quantified it. Pure autonomic failure is a disorder of the autonomic nervous system — blood pressure collapsing on standing, bladder and bowel dysfunction, impaired sweating — with no motor or cognitive features. Across 9 studies and 900 individuals followed for a mean of 6.4 years:1
Read that breakdown again, because it is not what anyone expects. The commonest destination was not Parkinson's. It was multiple system atrophy — a rarer and considerably more aggressive condition. A prodromal syndrome does not tell you which synucleinopathy is coming, and the one it most often announces here is the one most people have never heard of.
The practical point for a reader: autonomic symptoms in midlife or later, with no movement problem, are worth describing to a doctor in those words rather than being managed symptom by symptom in isolation. Seventy per cent of these people did not convert in six years, so this is not a prophecy either.
Not everything that looks like Parkinson's is Parkinson's, and one cause is reversible. Drugs that block dopamine D2 receptors can produce a parkinsonian syndrome — the mechanism is exactly the one described in what antipsychotics actually do, applied to a healthy nigrostriatal system rather than a degenerating one. Antipsychotics are the best-known class; some anti-nausea drugs act the same way.
We are not going to put a number on how often this happens, because we have not found a figure we trust. The useful thing is the question: if parkinsonian features appear in someone who recently started a new medication, "could this be the drug?" belongs in the conversation with the prescriber — and never as a reason to stop a prescription unilaterally.
Levodopa is converted to dopamine in the brain and replaces what the dying neurons no longer make. It is the reference standard against which everything else is measured, and the research effort now goes mostly into the problem it creates rather than the one it solves: after years of oral dosing, the benefit starts to wear off between doses ("OFF time") and involuntary movements appear at peak dose (dyskinesia).
Three recent syntheses of what is being done about that:
Add-on drugs. A Bayesian network meta-analysis of 11 randomised trials and 2,824 patients compared the drugs added to levodopa for motor fluctuations.2 All of them beat levodopa plus placebo except zonisamide at 25 mg. Safinamide 100 mg significantly outperformed rasagiline 1 mg on OFF-time (mean difference −0.52 hours, 95% CI −0.97 to −0.07) and on UPDRS Part III (−2.00, −3.30 to −0.6). Worth noting the scale of the wins between active drugs: roughly half an hour a day, and two points on a motor scale. Real, and small compared with the gap between levodopa and nothing.
Reformulated levodopa. IPX203 is an extended-release carbidopa-levodopa designed to hold plasma levels longer. Across 4 randomised trials and 712 patients, it beat immediate-release carbidopa-levodopa on UPDRS (mean difference −6.80, 95% CI −9.38 to −4.23) and cut OFF time by 2.42 hours (−3.12 to −1.71) on sensitivity analysis.3
Continuous infusion. A systematic review of 10 observational studies and 867 patients on continuous subcutaneous levodopa — ND0612 and foslevodopa/foscarbidopa — found daily OFF time reduced by roughly 2.0 to 3.5 hours and quality-of-life scores improved by 5.6 to 10.8 points on the Parkinson's Disease Questionnaire, over follow-up out to 36 months.4 The cost is on the skin: infusion-site reactions occurred in 42% to 92% of patients, mostly mild to moderate. Also, these were observational studies, not randomised trials — a design that reliably flatters an intervention people have chosen and stayed on.
Tavapadon is a selective D1/D5 dopamine receptor partial agonist, developed on the premise that targeting those receptors gives motor benefit without the impulse control disorders and sleepiness associated with the older D2/D3 agonists. A 2026 systematic review with GRADE assessment pooled 7 trials and 1,540 participants, five of which entered the quantitative synthesis.5
The efficacy numbers are striking. In early-stage monotherapy, tavapadon improved the combined MDS-UPDRS Parts II and III score by 10.55 points versus placebo (95% CI −13.11 to −7.99). In levodopa-treated patients with fluctuations, good ON-time rose by 1.09 hours a day (0.57 to 1.61).
Then the rest of the table:
That last line is the one to carry away. A ten-point improvement on a motor scale is a big number. "No outcome reached high certainty" is the authors telling you how much weight that number can bear. Both belong in the same paragraph, and in most coverage only the first one appears.
A 2025 review of neuroprotection in Parkinson's opens by stating that no effective disease-modifying treatments have been discovered, and lists why: clinical and biological heterogeneity, poor preclinical models, no established progression biomarkers, and — a problem worth appreciating — the existence of effective symptomatic therapies, which hampers detection of actual disease modification.6 Levodopa works so well that it masks whether anything underneath has changed.
A 2026 review put the failure pattern in one sentence that we think is the most useful thing on this page. Candidate disease-modifying therapies in Parkinson's — alpha-synuclein antibodies, mitochondrial antioxidants, lysosomal interventions and anti-inflammatory agents — "frequently achieve target engagement without slowing clinical progression."7
Target engagement means the compound reached the target and did to it what the mechanism said it would. The drug worked, on its own terms. The disease carried on anyway.
That is the entire logic of the supplement industry, tested at trial scale, and found insufficient. "Supports mitochondrial function", "reduces oxidative stress", "lowers neuroinflammation" are target-engagement claims. In Parkinson's, where serious money and serious trials have gone into exactly those mechanisms, engaging the target has repeatedly not been enough.
In 2015, the NINDS Exploratory Trials in Parkinson Disease programme published Long-term Study 1 in JAMA. It randomised 1,741 people with early, treated Parkinson's across 45 sites in the US and Canada to placebo or creatine monohydrate at 10 g a day, for a minimum of five years and a maximum follow-up of eight.8
It was terminated early for futility. At the planned interim analysis of 955 participants with a median four years of follow-up, mean summed ranks on the composite outcome were 2,360 for placebo and 2,414 for creatine — higher scores being worse, so numerically slightly worse on creatine. The global statistical test gave t = −0.75, two-sided P = 0.45. No detectable differences in adverse events. The authors' conclusion: the findings do not support the use of creatine monohydrate in patients with Parkinson disease.
We sold creatine monohydrate until September 2026, when we discontinued it. We have written at length about what the cognitive research on creatine actually shows, which is a genuinely mixed but not empty literature in healthy people and in sleep deprivation. None of that transfers here. A 1,741-person trial running up to eight years at 10 g a day, stopped for futility, is about as clean a negative result as neurology produces, and anyone who has ever seen creatine described as neuroprotective should know it exists.
A 2026 Bayesian dose-response network meta-analysis pooled 25 randomised controlled trials of exercise for depressive symptoms in middle-aged and older adults with Parkinson's, standardising dose as MET-minutes per week and assessing certainty with GRADE.9 The relationship was significant and non-linear:
Two honest caveats. The outcome here is depressive symptoms, not motor progression — this is not evidence that exercise slows the disease. And an inverted dose-response in a meta-regression can reflect how few trials sit at the high-dose end as much as a real ceiling. But the shape is worth knowing, because "more exercise is always better" is the default assumption and this analysis does not support it.
A separate 2026 meta-analysis of 13 randomised trials and 430 patients on virtual-reality-based therapy found improved global cognitive function (SMD 0.40, 95% CI 0.11–0.70, I² = 0%) and improved depressive symptoms (SMD −0.77, −1.42 to −0.12) — but the prediction interval for depression crossed zero (−1.82 to 0.27), and there were no significant effects on executive function, memory, attention or quality of life.10
That prediction interval is worth a sentence of its own. A confidence interval describes uncertainty about the average effect across the trials done. A prediction interval describes the range the effect might take in the next setting. When one excludes zero and the other does not, the honest summary is that the average is positive and the next clinic might see nothing.
We sell nothing that treats, slows or prevents Parkinson's disease. No ingredient in our range has been tested in a Parkinson's population. The condition is treated with prescription dopaminergic therapy, by a neurologist, and the specialist Parkinson's nurse and physiotherapist usually matter as much as the drug.
The two findings above are the ones we would keep. Engaging a mechanism is not the same as changing a disease — that is the lesson of the antioxidant and anti-inflammatory failures, and it applies directly to how supplements are sold, ours included. And the biggest negative trial in this field was run on a compound we used to sell, which is why it is on this page rather than left out.
On what is worth doing: the general risk-factor evidence in modifiable dementia risk is about dementia rather than Parkinson's and should not be stretched to cover it. The comparison we keep returning to — exercise versus supplements — holds here too, and our standing argument about mechanism stories is in why we don't sell brain blends.
This article is for informational purposes only and has not been evaluated by the FDA. It is not intended to diagnose, treat, cure, or prevent any disease. We are not doctors. Parkinson's disease requires specialist diagnosis and management. Nothing here is a reason to start, stop or change any prescribed medication — levodopa and dopamine agonists in particular should never be stopped abruptly, which can be dangerous. If you notice slowness, stiffness, tremor, or a change in handwriting, walking or facial expression, that warrants medical assessment.