The Trial Architecture Neurodegenerative Disease Has Always Needed — And Kept Refusing to Build

Picture a trial coordinator at a Phase 2 site running a Parkinson’s disease study sometime around 2018. She has enrolled 60 patients over 14 months — a heroic effort given the recruitment arithmetic of neurodegenerative trials — and the drug has just failed its primary endpoint. The compound is shelved. The patients’ data is locked in a sponsor’s archive. The site goes dark. And somewhere in the pipeline, three more single-arm, single-hypothesis trials are queuing up to repeat the same sequence.

This is not a story about one bad molecule. It is the structural reality of how the field has approached neurological disease for decades — one trial, one target, one failure at a time. The consequences are written in the numbers: between 2002 and 2012, the Alzheimer’s disease drug development pipeline posted a 99.6% failure rate, with a single approval emerging from roughly 244 compounds that entered trials. Parkinson’s and motor neuron disease (MND) tell comparable stories. The biology is intractable, yes — but the trial architecture has made the intractability worse.

A Comment published in The Lancet this week makes a direct argument: multi-arm multi-stage (MAMS) platform trials can break the cycle. The authors are not the first to say it. But the timing — arriving as neurological conditions account for 43% of all global nervous system health loss and as adaptive trial frameworks finally reach regulatory maturity — gives the argument a different weight. The question worth asking is not whether MAMS designs are theoretically superior. They demonstrably are. The question is why the field built its infrastructure around the approach least suited to the biology it is trying to crack.

One Disease, Many Failures, One Architecture Problem

To understand what MAMS offers neurology, start with what conventional trial design costs it. A standard Phase 2 randomized controlled trial in Alzheimer’s disease tests one mechanism against placebo, typically over 18 to 24 months, in a patient population that requires expensive and operationally demanding biomarker stratification at entry. When the trial fails — and the base rate says it will — the scientific community learns approximately one thing: that this compound, at this dose, in this population, over this window, did not move this endpoint. Everything else is noise.

MAMS trials are built on a different epistemological premise. Multiple arms test multiple candidates simultaneously against a shared control. Interim analyses at pre-specified stages allow non-performing arms to be dropped and promising ones to be expanded, without inflating family-wise error rates if the design is properly pre-specified. The shared control group is not just operationally efficient — it is scientifically generative, because the arms accumulate comparative evidence that a sequence of standalone trials cannot produce.

The PRE-MASTODON-AD initiative, funded through NIHR, is attempting to establish exactly this kind of platform for Alzheimer’s disease and mild cognitive impairment — building the infrastructure before committing to a full trial slate, so that candidate treatments can enter a running platform rather than each requiring a bespoke operational build. That distinction matters enormously at the site level. A site that is already credentialed, IRB-approved, and screening-capable for a MAMS platform can add a new arm in weeks. A site starting from scratch on a standalone trial needs 12 to 18 months before the first patient walks in.

The Alzheimer’s Clinical Trials Consortium (ACTC), funded through the National Institute on Aging, represents the broader infrastructure ambition — a standing academic network capable of conducting trials across the full Alzheimer’s continuum, from prevention to late-stage disease. Infrastructure of this kind is the precondition for MAMS to work at scale. Without pre-established sites, shared protocols, and harmonized biomarker pipelines, a multi-arm design is an aspiration on paper.

But here is the counterintuitive reality that the field keeps stepping around: the bottleneck in neurodegenerative trial design has never been scientific imagination. It has been organizational inertia. Sponsors build standalone trials because their internal processes, budget cycles, CMC timelines, and regulatory submissions are structured around standalone trials. A MAMS platform requires upfront coordination costs that no single sponsor wants to absorb, shared data governance that legal teams flag as competitive risk, and interim decision frameworks that require regulators and sponsors to agree on rules before the data exists. These are solvable problems. They have been solved — in oncology. The question that The Lancet comment implicitly raises is why neurology keeps treating them as unsolvable.

The Regulatory Gap That Still Needs Closing

Adaptive trial design now has its most authoritative regulatory home yet. The ICH E20 Guideline on Adaptive Designs for Clinical Trials was endorsed as a Step 2 draft by the ICH Assembly in June 2025, released for public consultation by the EMA on June 30, 2025, with comments due by November 30, 2025, and the FDA announced availability of the same draft guidance on September 30, 2025. For sponsors navigating global submissions, ICH E20 creates the closest thing to a shared adaptive design rulebook that the industry has ever had.

That regulatory progress is genuinely significant. But a guideline on adaptive designs is not the same as a guideline on platform trials — and MAMS architecture introduces regulatory questions that E20 only partially addresses. How should a shared control arm be handled when the platform has been running for three years and the standard of care has shifted? What is the regulatory status of a compound that clears one stage but the platform sponsor adds a new comparator arm before the next stage initiates? When two arms in the same MAMS platform are owned by different sponsors, who holds the IND and what are the liability partitions?

The FDA’s own testimony to Congress in July 2021 acknowledged that clinical trial experience for neurodegenerative diseases has been “challenging with many failed studies, reflecting the difficulty in drug development when the disease is not well understood” — and attributed many failures to “inadequate disease characterization.” That diagnosis points directly at what MAMS platforms could fix: by running multiple mechanistic hypotheses simultaneously against harmonized biomarker endpoints, they generate disease characterization data that standalone trials structurally cannot. But the agency’s acknowledgment of the problem has not yet translated into bespoke platform trial guidance for neurodegeneration. The space between recognizing failure and creating the regulatory scaffolding that reduces it remains wide.

What Oncology Built That Neurology Hasn’t

The field that provides the most instructive contrast is not a competitor — it is a template. Oncology’s platform trial infrastructure, developed over roughly 15 years, now runs trials like I-SPY 2 in breast cancer and the STAMPEDE platform in prostate cancer, the latter of which has enrolled tens of thousands of patients across multiple arms and delivered definitive answers on combination regimens that would have taken 20 years of sequential standalone trials to produce. STAMPEDE’s design — a permanent shared control with arms entering and exiting as evidence accumulates — is exactly the model the Lancet comment is calling neurology toward.

What made those platforms possible was not a single regulatory breakthrough. It was the prior accumulation of disease biology infrastructure: agreed biomarker endpoints, active patient registries, credentialed sites with standing IRB protocols, and — critically — academic-commercial partnerships willing to share control arm data. Neurology is now reaching that prerequisite threshold in Alzheimer’s disease, with amyloid and tau PET imaging established as surrogate endpoints and the ACTC providing a standing site network. Parkinson’s and MND are at earlier stages, but the directional pressure is the same.

The coordinator from that 2018 Parkinson’s trial is probably still running neurology studies. The sites are still there, the patients are still enrolling, and the biology is still fighting back. What has changed — just barely, just enough — is that the architecture around her is starting to look different. Whether the field moves fast enough to match the urgency written into those global burden statistics is a question that the next five years of platform trial investment will answer.

References

1. The Lancet — “Multi-arm multi-stage platform trials for neurological disease: accelerating progress”
2. Alzheimer’s Research & Therapy — “Failure rates in Alzheimer’s disease drug development pipeline, 2002–2012”
3. IHME — “Neurological Conditions: Global Burden of Disease Data”
4. NIHR — “PRE-MASTODON-AD: Preliminary development of a Multi-Arm Multi-Stage Trial platform for Dementia and mild cognitive impairment due to Alzheimer’s Disease”
5. NIH Reporter — “Alzheimer’s Clinical Trials Consortium (ACTC), NIA funding”
6. FDA Congressional Testimony — “The Path Forward: Advancing Treatments and Cures for Neurodegenerative Diseases,” July 29, 2021
7. EMA / ICH — “ICH E20 Guideline on Adaptive Designs for Clinical Trials, Step 2b Draft, June 2025”