29 May 2026
Pull up the approval history for Alzheimer’s disease-modifying therapies since 2003. Run the numbers. The overall Phase II and III success rate for novel compounds is 2.0%. Not 20%. Not 12%. Two percent — a figure so catastrophic that if any other industry produced it, the methodology would have been abandoned a decade ago. And yet, for most of the last twenty years, the standard response from sponsors and regulators alike has been to run another sequential, single-arm Phase 3 trial, wait five years, and absorb the loss when it fails.
A Comment published in The Lancet is making the case that this has to stop. The authors argue that multi-arm, multi-stage (MAMS) platform trials represent the rational structural response to neurodegenerative drug development — a field where the biology is merciless, the patient populations are finite, and the cost of running compounds sequentially through failed trials is measured not just in dollars but in irreversible disease progression for the patients who waited.
The argument deserves serious operational scrutiny. Because MAMS designs are not a magic trick. They are a regulatory and logistical bet — one that requires infrastructure, institutional commitment, and a degree of regulatory predictability that does not yet fully exist.
The Failure Architecture We Built
To understand why platform trials matter, you have to confront what the traditional sequential model actually produces in neurodegeneration. Between 2004 and 2021, 40 out of 97 Alzheimer’s compounds — 41% — failed in Phase III. Among disease-modifying agents specifically, the failure rate climbs to 51%. These are not close calls. Many of these compounds consumed a decade of development time, hundreds of millions in capital, and thousands of patient-years of exposure — only to demonstrate futility at the finish line.
The structural problem underneath those numbers is one of sequentiality. In a traditional development program, you test compound A, wait for the result, then decide whether to test compound B. If the neurodegenerative field has taught us anything, it is that this model catastrophically underuses the shared infrastructure — the patient registries, the biomarker frameworks, the site networks — that each trial laboriously constructs and then abandons. Every new sponsor builds the same house from scratch and burns it down when the trial concludes.
MAMS platform trials attack this directly. A single master protocol governs multiple treatment arms simultaneously. Interim analyses — conducted at pre-specified timepoints against adaptive stopping rules — allow arms to be dropped for futility or added as new candidates emerge. The control arm is shared across all experimental arms, which reduces the number of patients assigned to placebo without contributing to the comparative signal. The platform persists. The infrastructure accumulates value rather than evaporating at database lock.
The MND-SMART trial in motor neuron disease is the most instructive live example in neurology right now. The trial ran as a genuine adaptive platform — and when the first results were published, neither of the initial two drugs showed benefit. Many observers read that as a failure. It was the opposite. The platform generated a clean, credible null result for two compounds simultaneously, using a shared control arm and adaptive decision rules — and the infrastructure remained intact to test the next candidates. In a sequential world, those two null results would have taken separate trials, separate site activations, and a combined timeline measured in the better part of a decade.
The Regulatory Gap Nobody Wants to Discuss
Here is the counterintuitive reality that the enthusiasm for platform trials tends to obscure: regulatory guidance has not kept pace with the operational ambition.
The FDA released its draft guidance on Master Protocols for Drug and Biological Product Development on December 21, 2023. The document is genuinely useful — it acknowledges the legitimacy of randomized platform trials, addresses key statistical concerns around type I error control, and provides recommendations on trial design and analysis. But it remains a draft. And draft guidance, in the FDA’s interpretive ecosystem, creates a very specific operational problem for sponsors: it signals direction without conferring the predictability that a platform trial’s long-term governance structure actually requires.
Consider what a MAMS platform trial for Alzheimer’s disease actually needs from a regulatory standpoint. It needs pre-specified rules for adding new arms — and those arms may enter the platform years after the initial IND is filed, potentially with different chemistry and different mechanism-of-action profiles. It needs a shared statistical analysis plan that can accommodate multiple simultaneous comparisons without inflating familywise error. It needs a governance structure — an independent steering committee, a data monitoring committee with disease-area expertise — whose authority over arm addition and deletion decisions must be legible to the agency. And it needs a clear understanding of what a successful arm completion means for a subsequent NDA or BLA: will the agency accept the platform’s shared control data, or will it demand a standalone confirmatory trial?
The 2023 draft guidance addresses some of this. It does not answer all of it. And in neurodegeneration specifically — where trials routinely run five to seven years, where biomarker-defined patient populations are small, and where the biology can shift the endpoint landscape mid-trial — regulatory ambiguity carries compounding operational cost.
Sponsors who want to run MAMS neurology platforms today are essentially negotiating bespoke agreements with the agency through the Type B meeting process, trial by trial. That is not a system. That is improvisation at scale.
What Has to Change Before the Promise Becomes Practice
The Lancet Comment is right that MAMS platform trials represent the rational path for neurodegenerative disease. But the path requires three things that the field has not yet delivered in combination: a finalized regulatory framework, a sustainable funding model, and operational infrastructure that can span the multi-decade horizon these platforms require.
On the regulatory side, the FDA needs to finalize the master protocol guidance — with neurodegeneration-specific annexes that address the unique challenges of long-duration trials, fluid endpoint landscapes as biomarker science evolves, and the mechanics of arm addition years into a running IND. The EMA has moved somewhat faster here, with its qualification of novel methodologies process offering a vehicle for prospective platform trial design review. The FDA has no direct equivalent, and that asymmetry is already shaping where some sponsors choose to anchor their platform governance.
The funding architecture matters just as much. Academic-led platforms like MND-SMART have demonstrated that the model is operationally viable. But academic funding cycles — typically three to five years — are fundamentally misaligned with platform trial timelines that may run ten years or more. The Alzheimer’s disease field has a drug prioritization pipeline actively identifying candidates for a planned MAMS platform, but the gap between “candidates identified” and “platform running with regulatory clarity” remains substantial. Bridging that gap requires public-private partnership structures with committed multi-year capital — not the grant-renewal treadmill that has historically governed academic trial infrastructure.
And then there is the operational reality that neither the Lancet Comment nor the FDA draft guidance fully confronts: site burden. MAMS platforms concentrate complexity at the investigator site in ways that sequential trials do not. Site staff manage multiple concurrent treatment arms, adaptive randomization algorithms, arm-specific safety monitoring protocols, and — in Alzheimer’s and Parkinson’s disease — patient populations whose cognitive status may change the consent and assessment landscape mid-trial. The site activation and training infrastructure for a platform trial is not a scaled-up version of a conventional trial’s infrastructure. It is categorically different, and the field’s workforce has not yet fully adapted to it.
The 2.0% success rate in Alzheimer’s drug development did not emerge from bad science alone. It emerged from a system that tests compounds one at a time, learns slowly, and rebuilds its infrastructure from zero each cycle. MAMS platform trials break that cycle — but only if the regulatory architecture catches up with the science, the funding models extend to match the timelines, and the sites are resourced to carry the operational load. The MND-SMART trial proved the concept works even when the drugs fail. The question now is whether the institutions surrounding these trials — the FDA, the funders, the academic consortia — are willing to build something permanent enough to make the concept matter at scale.
Source: Clinical Trial Vanguard