What has failed, and what we learned
Every trial in people that tried to slow Parkinson's and did not work.
Facts on this page checked 18–22 Oct 2026
This page lists every trial in people that tested a way to slow Parkinson's and found that it did not work. The trials are grouped by approach. Each group shows what we learned and what happens next.
A trial that does not work is not wasted. It tells researchers what not to try again. Often it also shows why, and that shapes the next trial.
At a glance
- Tried and did not work
- 43 trials
- Has anything slowed Parkinson's?
- Not yet shown
- Still to come
- 28 trials with results due
Does this mean there is no hope?
No. Most drugs tested against any brain disease do not work. Each one that fails rules something out.
These failures have changed how trials are run. Newer trials of protein drugs first check that people have the misfolded protein. Results are still due from trials in all six approaches: see What's coming next.
Replace or repair what's lost
Can we put back lost cells or help the brain regrow?
10 trials did not work
The 10 trials that did not workNewest first
-
CAP-003 · a gene therapy for GBA gene Parkinson's, given by vein
Did not work
Withdrawn before anyone was treated, after a child died in a trial of another drug with the same virus shell. It was not a test of the gene itself.
-
ProSavin · a gene therapy giving brain cells three dopamine-making genes
Did not work
Safe in 15 people in 2014, with no comparison group. Its successor stopped in 2022 when the sponsor collapsed, with 6 people treated.
-
TRANSEURO · grafts of dopamine-making cells from donated fetal tissue
Did not work
Only 11 people could be grafted because fetal tissue was scarce. After 3 years there was no overall gain compared with 16 people who had no graft.
-
Fosgonimeton · an injected drug meant to boost a growth-factor pathway
Did not work
Stopped early at 28 of a planned 75 people, who had dementia linked to Parkinson's. It missed its main target. Too small to close the idea.
-
VY-AADC · a gene therapy that helps the brain make dopamine
Did not work
A trial with sham surgery was put on hold in 2020 after MRI changes. The maker ended the Parkinson's work in 2021. No result on benefit was found.
-
GDNF infusions · a growth-factor protein pumped into the brain
Did not work
Two controlled trials, in 34 and 35 people, did not improve symptoms. One showed more dopamine activity on scans, but patients were not better.
-
CERE-120 · a gene therapy that makes the brain produce neurturin, a growth factor
Did not work
Two trials, in 58 and 51 people, did no better than sham surgery. Autopsies of two people suggested too little of the target area was reached.
-
Spheramine · retinal cells that make levodopa, placed in the brain
Did not work
In 71 people, the implants did no better than a fake operation. Movement scores changed by about the same amount in both groups.
-
Olanow fetal grafts · grafts of dopamine-making cells from donated fetal tissue
Did not work
In 34 people, grafts did not clearly beat a placebo operation at 24 months. 56% of grafted people had unwanted movements that continued off medication.
-
Freed fetal grafts · grafts of dopamine-making cells from donated fetal tissue
Did not work
In 40 people, grafts did not beat sham surgery on the main measure. People aged 60 or under did better on one score. 15% had unwanted movements.
What we learned
- Better scans did not mean better patients. GDNF, a growth factor, was pumped into the brain. Scans showed more dopamine activity, but people did not get better. This is the warning for a GDNF gene therapy now being tested. (GDNF infusions; REGENERATE-PD)
- A fake operation can make people feel better, so trials need one. In some trials the people given sham surgery improved too. Without that comparison, a gain can be wrongly credited to the treatment. (CERE-120; AAV2-GAD)
- Fetal grafts caused problems and could not be scaled up. Some people developed uncontrolled movements, called dyskinesias, that continued off medication. Fetal tissue was also scarce: only 11 people could be grafted in the last big trial. Scientists turned to stem cells. (Freed fetal grafts; Olanow fetal grafts; TRANSEURO)
- Gene therapy can be stopped by safety checks or by the company behind it. One trial was put on hold after MRI changes. One ended when its sponsor collapsed. One closed after a child died in a trial of a different drug with the same virus shell. (VY-AADC; ProSavin; CAP-003)
What happens next
The next result for this approach is due in Mar 2027, from ASPIRO. Read about Replace or repair what's lost
Back to topClear the bad protein
Can we stop the toxic clumps forming and spreading?
4 trials did not work
The 4 trials that did not workNewest first
-
ION464 · a drug that makes cells produce less of the protein
Did not work
Tested only in MSA, a related disease. The company dropped it in 2025, so it will not give an answer for Parkinson's.
-
ORCHESTRA · minzasolmin, a pill
Did not work
A pill meant to stop the protein folding the wrong way. It did not help in about 500 people.
-
Buntanetap · a pill
Did not work
No overall change in 523 people after 6 months. The company points to benefit in smaller groups of patients, which is not proof.
-
SPARK · cinpanemab, an antibody drug
Did not work
Made no difference at any dose in 357 people. The company stopped work on it in February 2021.
What we learned
- Small differences in a drug matter. Two antibodies aimed at the same protein gave different results. Cinpanemab did nothing; prasinezumab came close. (SPARK; PASADENA and PADOVA)
- Working in the lab is not the same as working in people. A pill that stopped the protein misfolding in the lab did not help patients. We do not yet know if the dose, the timing or the idea was wrong. (ORCHESTRA)
- A benefit in a small group is a clue, not proof. When a company reports benefit only in some patients, that needs its own trial. (Buntanetap)
- Trials now check for the target first. Newer trials use a spinal-fluid test to make sure people really have the misfolded protein before they join. (A spinal-fluid test can detect the misfolded protein)
What happens next
The next result for this approach is due in Dec 2026, from EXIST. Read about Clear the bad protein
Back to topHelp cells clean up
Can we fix the cell's recycling and power?
11 trials did not work
The 11 trials that did not workNewest first
-
BIA 28-6156 · BIA 28-6156, a pill meant to raise the enzyme GCase
Did not work
No difference from placebo in 273 gene carriers over 18 months. The company has stopped work on it for Parkinson's.
-
AMBITIOUS · ambroxol, a cough medicine, at about ten times the cough dose
Did not work
No benefit for thinking in 65 gene carriers over a year. This is a preprint, not yet checked by other scientists.
-
NOPARK · nicotinamide riboside, a form of vitamin B3
Did not work
No benefit in 410 people. After a year, scores were slightly worse than on placebo.
-
Venglustat · venglustat, a pill that cuts the raw material for the enzyme GCase
Did not work
No slowing in 221 gene carriers over a year. Scores were slightly worse than on placebo.
-
FAIRPARK-II · deferiprone, a pill that removes iron
Did not work
Symptoms got worse, not better, in 372 newly diagnosed people. Iron fell as planned, so the drug reached its target and still did harm.
-
SURE-PD3 · inosine, a pill that raises urate, a natural antioxidant in blood
Did not work
Urate rose as planned, but Parkinson's did not slow in 298 people. Kidney stones were more common on inosine.
-
Hydrogen water · water with hydrogen gas dissolved in it
Did not work
A pilot in 17 people looked good. A larger trial in 178 people found no difference from placebo.
-
Creatine · creatine, a supplement
Did not work
No benefit in 1,741 people. The trial stopped early because creatine was not helping.
-
Coenzyme Q10, QE3 · coenzyme Q10, a supplement
Did not work
No benefit in 600 people at two doses. Scores were no better than on placebo, and slightly worse. It was safe, but it did not work.
-
MitoQ · mitoquinone, a pill that carries an antioxidant into mitochondria
Did not work
No difference from placebo in 128 newly diagnosed people over 12 months. It did not slow Parkinson's.
-
ADAGIO · rasagiline, a pill
Did not work
In 1,176 people, one dose hinted at slowing but the other showed nothing. In 2011 a US expert panel voted 17 to 0 that slowing was not proven.
What we learned
- Hitting the right gene is not enough. Three drugs aimed at GBA did not help people who carry a GBA change. They were venglustat, BIA 28-6156 and high-dose ambroxol. The ambroxol result is not yet checked by other scientists. (Venglustat; BIA 28-6156; AMBITIOUS)
- Mopping up "rust" did not slow Parkinson's. Coenzyme Q10, creatine, MitoQ, inosine and hydrogen water did not work. (Coenzyme Q10, QE3; Creatine; MitoQ; SURE-PD3; Hydrogen water)
- Hitting the target can still do harm. Deferiprone removed iron as planned, yet symptoms got worse. Nicotinamide riboside raised a cell energy chemical, yet people did slightly worse than on placebo. (FAIRPARK-II; NOPARK)
- A small early result often does not hold up. Coenzyme Q10 and hydrogen water both looked helpful in small trials. Larger trials then found no benefit. (Coenzyme Q10, Phase 2; Coenzyme Q10, QE3; Hydrogen water)
- Easing symptoms can look like slowing the disease. Deprenyl and rasagiline delayed the need for other drugs, but this did not prove they slow Parkinson's. (DATATOP; TEMPO; ADAGIO)
What happens next
The next result for this approach is due in Dec 2026, from N-DOSE. Read about Help cells clean up
Back to topLook beyond the brain
Do the gut, the immune system or metabolism drive it?
9 trials did not work
The 9 trials that did not workNewest first
-
EFFACE-PD · a faecal transplant, given once by colonoscopy
Did not work
Did not change movement scores over 12 months in 59 people.
-
AZA-PD · azathioprine, a drug that dampens the immune system
Did not work
Did not meet its main goal in 66 people over 12 months. Immune markers moved, but symptoms did not clearly improve.
-
Exenatide-PD3 · exenatide, a diabetes drug
Did not work
Did not slow Parkinson's in 194 people over two years. The journal has published a concern about one trial site.
-
Stockholm exenatide trial · exenatide, a diabetes drug
Did not work
Brain scans changed, but movement scores did not, in 59 people. This matches the larger UK trial.
-
Finnish faecal transplant trial · a faecal transplant, given once by colonoscopy
Did not work
No benefit at six months in 48 people. Stomach side effects were more common with donor stool.
-
NLY01 · a long-acting diabetes-type drug aimed at brain immune cells
Did not work
Did not beat placebo in 255 people over 36 weeks, according to the company. No journal paper has been published.
-
PD STAT · simvastatin, a cholesterol drug
Did not work
Did not slow Parkinson's in 235 people. Movement scores worsened slightly more on the drug than on placebo.
-
AKST4290 (TEAL) · a pill that blocks an immune receptor called CCR3
Did not work
Did not beat placebo in 110 people over 12 weeks. Both groups improved a little, and placebo slightly more.
-
Pioglitazone (FS-ZONE) · pioglitazone, a diabetes pill
Did not work
Unlikely to slow early Parkinson's at either dose in 210 people. The researchers did not recommend a larger trial.
What we learned
- A good early result is not proof. Exenatide seemed to help in a small trial with no placebo, then did not slow Parkinson's in a large one. (Exenatide pilot; Exenatide-PD3; Stockholm exenatide trial)
- Changing the brain is not the same as helping. Exenatide changed brain scans but not movement scores. (Stockholm exenatide trial)
- Faecal transplants mostly did not work. One small trial showed a gain, but two later trials did not. (GUT-PARFECT; Finnish faecal transplant trial; EFFACE-PD)
- Calming the immune system was not enough on its own. Azathioprine and a CCR3 blocker did not work. (AZA-PD; AKST4290 (TEAL))
- Links and animal results are not proof. A cholesterol drug, tried because users had lower risk, did not slow Parkinson's. Nor did a diabetes pill that protected brain cells in animals. (PD STAT; Pioglitazone (FS-ZONE))
What happens next
The next result for this approach is due in Nov 2026, from TB006. Read about Look beyond the brain
Back to topTreat the inherited causes
Can we treat the genes behind some Parkinson's?
2 trials did not work
The 2 trials that did not workNewest first
-
LRRK2 antisense drug · BIIB094, an injection into the spinal fluid that makes cells produce less LRRK2
Did not work
It lowered LRRK2 protein in an 82-person safety trial. Biogen stopped it for business reasons, not because it failed. It was never tested for benefit.
-
LUMA · BIIB122, an LRRK2 drug
Did not work
Cut its blood target by over 90% in 648 people, but did not slow Parkinson's. A small spinal-fluid test hinted that less drug reached the brain.
What we learned
- Hitting the target is not the same as slowing Parkinson's. The drug in LUMA switched off its blood target but did not slow Parkinson's. A small spinal-fluid test hinted that less drug reached the brain. (LUMA; DNL151 safety studies; DNL201)
- "Safe" is not the same as "works". Early trials showed the drugs were well tolerated and hit their targets. They did not test whether Parkinson's slowed. The trial for carriers, BEACON, measures safety and markers, so it cannot prove benefit. (DNL201; DNL151 safety studies; BEACON)
- Stopped is not the same as failed. The spinal-fluid drug in LRRK2 antisense drug lowered LRRK2 protein safely. Its company dropped it for business reasons, not because it failed. A carrier trial inside LUMA ended early with no answer, because very few people joined. (LRRK2 antisense drug; LUMA)
- Gene findings need replication in thousands of people. Several genes named as causes did not hold up when larger studies looked again. (UCHL1 I93M (PARK5); GIGYF2 and EIF4G1: disputed genes; HTRA2/Omi (PARK13) did not hold up; DNAJC13 and TMEM230: disputed genes)
What happens next
The next result for this approach is due in Mar 2027, from BEACON. Read about Treat the inherited causes
Back to topKeep vulnerable cells alive
Why do these cells die first, and can we shield them?
5 trials did not work
The 5 trials that did not workNewest first
-
PROSEEK · vodobatinib, a leukaemia-type pill
Did not work
The drug reached the brain but did no better than placebo in 513 people. The trial was stopped in April 2024.
-
Nilotinib · a leukaemia drug
Did not work
Almost none of the drug reached the brain. It did not help in these two trials, which had 151 people in all.
-
STEADY-PD III · isradipine, a blood-pressure pill
Did not work
Did not slow Parkinson's in 336 people over three years. The dose may have been too low to reach its target.
-
PRECEPT · CEP-1347, a pill meant to stop cells dying
Did not work
Stopped early in 806 people because it was clearly not working. It had protected cells in animals.
-
TCH346 · a pill meant to stop cells dying
Did not work
Made no difference in 301 people with early Parkinson's. It had protected cells in animals.
What we learned
- Protecting cells in animals does not mean protecting them in people. Two drugs saved cells in animals. Neither helped in people. (PRECEPT; TCH346)
- The dose we can safely give may be too low. The blood-pressure drug isradipine did not slow Parkinson's. The calcium idea may still stand, because the dose people could take may have been too low. (STEADY-PD III)
- A drug must reach the brain. Nilotinib barely got into the brain, so it was never a fair test. (Nilotinib)
- Reaching the brain is not enough. Vodobatinib did get in, and still did not help. This weakens the c-Abl idea at the doses tested. (PROSEEK)
- Safe is not the same as works. Risvodetinib was safe, but it missed its main goal in a short study. (Risvodetinib)
What happens next
The next result for this approach is due in Nov 2026, from Low-dose lithium. Read about Keep vulnerable cells alive
Back to topExisting Parkinson's medicines
2 trials did not work
The 2 trials that did not workNewest first
-
LEAP · levodopa, the main Parkinson's medicine
Did not work
Starting levodopa straight away did no better than starting it 40 weeks later, in 445 people. It eases symptoms but does not slow Parkinson's. There was no sign that it speeds the disease up at this dose.
-
PROUD · pramipexole, a standard Parkinson's medicine
Did not work
Starting pramipexole straight away, rather than 6 to 9 months later, made no lasting difference in 535 people. It eases symptoms but does not slow Parkinson's.
What we learned
- Easing symptoms is not the same as slowing the disease. These medicines help with movement. That can look like slowing. Trials that start the drug early in some people and later in others can tell the two apart. Neither showed slowing. (PROUD; LEAP)
- Starting levodopa early did not seem to harm the disease. There was no sign that it speeds Parkinson's up, at the dose tested. (LEAP)
What "did not work" means here
It means the trial did not show that the treatment slowed Parkinson's. It does not always mean the idea is wrong. The dose may have been too low, or the drug may not have reached the brain.
A few trials ended before they could give an answer. Some were stopped by the company for business reasons. Where we know this, the trial's own line says so.
This is not a list of treatments to avoid. Some of these drugs are used safely for other conditions.