The bigger trial said no. Then the paper was flagged.

Last week's issue covered a French trial in which a diabetes drug appeared to hold Parkinson's symptoms steady for a year. It ended by noting that a larger trial of a related drug was due, and that its result would matter a great deal.

That result is in. It is negative. And in the months since it was published, the paper reporting it has been placed under formal review.

The hope that led here

Exenatide had one of the better stories in Parkinson's research.

A phase 2 trial published in 2017 followed about 60 people for 48 weeks and found that those on exenatide were, on average, around 4 points better on the motor scale than those on placebo — and, importantly, the gap persisted 12 weeks after everyone stopped the drug. For a field that had watched candidate after candidate fail, this was a real result, and it drove a decade of follow-on work.

It was still a small trial. That is what phase 2 means: a signal worth chasing, not a conclusion. Chasing it properly is exactly what this new trial did.

What they did

The trial enrolled 194 people with Parkinson's across six research hospitals in the UK. Participants were randomly assigned to a weekly injection of 2 mg exenatide or a matching placebo, double-blind, and followed for 96 weeks — close to two years, roughly twice the length of the earlier trial.

The primary endpoint was the motor examination score on the MDS-UPDRS, measured in the practically-defined off state: participants withheld their usual Parkinson's medication overnight before assessment. This is a deliberately demanding choice. It strips out the effect of levodopa and tries to reveal the underlying state of the disease.

What they found

Both groups got worse, at almost exactly the same rate.

The numbers slightly favoured placebo, but nowhere near enough to mean anything — a p-value of 0.47 is what you expect when nothing is happening. The confidence interval is the more useful number here: it comfortably includes zero, and its narrowness means the trial was large enough to rule out the sort of benefit the 2017 study had suggested.

The absence of an effect held across the secondary measures too — non-motor symptoms, quality of life, medication requirements, and imaging. There was no subgroup that clearly benefited. The drug was safe and well tolerated. It simply did not work.

Then the paper was flagged

On 18 May 2026, more than a year after publication, The Lancet issued an expression of concern about the paper.

The reason was not the statistics. Regulators had inspected King's College Hospital NHS Foundation Trust, one of the six sites, and found department-wide problems with how clinical trials there were being conducted, overseen and governed. Some findings were graded critical and others major — the two most serious categories in a good clinical practice inspection.

The Lancet asked the corresponding author's institution to investigate, and attached the notice while that work goes on.

It is worth being precise about what this does and does not mean.

It does not mean the result is wrong. An expression of concern is a flag, not a verdict. It says a question has been raised that has not yet been answered.

It does not mean anyone falsified anything. Inspection findings of this kind more often concern record-keeping, consent documentation, staff training, and oversight than deliberate misconduct.

It does mean the conclusion is provisional. One of six sites is under question. Until the investigation reports, nobody — including the trial's own authors — can say with certainty how much the affected data influenced the result.

Why it matters

A decade-long hypothesis was tested properly. That is how the process is meant to work, and it is worth saying plainly: a negative trial from a well-run study is a real contribution. It stops other groups spending years on the same idea, and it stops patients enrolling in studies of something that does not help.

It shows what phase 2 results are worth. The 2017 trial was not fraudulent or incompetent. It was small, and small trials produce unstable estimates. Chance, a strong placebo response, and the enthusiasm that surrounds an early positive result can combine to produce a promising number that evaporates under scrutiny. This happens constantly, in every area of medicine.

And it shows the correction machinery working in public. An inspection found problems. A journal attached a warning. An investigation is under way, and its outcome will be published. This is slow and unglamorous, and it is the part of science that most rarely gets reported — which is exactly why it is worth reporting.

Reading the two trials together

Several explanations are live, and the honest position is that nobody yet knows which is right.

It may be that assessing people on their usual medication, as the French trial did, is more forgiving — capturing a modest symptomatic effect rather than true disease modification. It may be that 12 months is short enough for a small early difference to appear and 96 weeks long enough for it to wash out. It may be that the two drugs genuinely differ in how much reaches the brain. Or the French result may simply have been a false positive, of the same kind the 2017 exenatide trial now appears to have been.

Until the investigation into the exenatide paper concludes, there is one more possibility that cannot be dismissed: that the negative result is itself unreliable. It is unlikely — one site out of six, in a trial whose result was resoundingly flat — but "unlikely" is not "ruled out".

What this doesn't tell us

Whether the exenatide result will stand. That depends on an investigation that has not reported. If a substantial share of data came from the affected site, the analysis may need redoing.

Whether GLP-1 drugs as a class are finished in Parkinson's. Two drugs, two trials, two opposite answers. That is not enough to close a question.

Whether either drug might work in people at an earlier stage. Both trials enrolled people who already had a diagnosis, by which point a great deal of damage is done.

What to watch next

The investigation's outcome. The Lancet will publish it, and it will either lift the notice, correct the paper, or retract it. That is the single most consequential thing pending in this story.

Whether the lixisenatide finding replicates in a larger trial. That is now the decisive question for this line of research.

And whether trial design shifts earlier. The argument that treatments must start before diagnosis — in the prodromal stage, identified by tests like the seed amplification assay we covered two weeks ago — gets stronger every time a trial in diagnosed patients comes back empty.