On February 26, FDA Commissioner Marty Makary and principal deputy Vinay Prasad published a perspective in the New England Journal of Medicine that formally ended what they called the “two-trial dogma.” Going forward, the agency’s default position is that one adequate and well-controlled pivotal study, combined with confirmatory evidence, will be sufficient for marketing authorisation of novel drugs.
The policy was written for pharmaceuticals, not medical devices. Brain-computer interfaces follow a separate regulatory track — premarket approval, De Novo classification, or 510(k) — and device makers have historically been able to reach approval on a single well-designed pivotal study. But the announcement matters for the BCI industry for a subtler reason: it signals a broader shift in how the FDA thinks about evidentiary thresholds, and that shift may eventually influence the agency’s expectations for novel implantable devices.
The timing is notable. Synchron is in active discussions with the FDA about its pivotal trial design following positive 12-month results from the COMMAND early feasibility study of its Stentrode device. Neuralink has filed for premarket approval and is expanding its PRIME study across four countries. Paradromics received its investigational device exemption in November 2025 and is recruiting for speech-restoration trials. All three companies are, in practical terms, one pivotal study away from a potential commercial product.
Under the old pharmaceutical convention, the FDA could request two independent pivotal trials demonstrating efficacy — a requirement rooted in 1960s-era legislation designed to guard against statistical flukes. Each trial costs between $30 million and $150 million and takes years to enrol and complete. While that two-trial expectation was never the formal standard for devices, the FDA has at times requested additional clinical evidence for high-risk implantables, particularly those without predicates. If the agency’s appetite for redundant proof is genuinely declining across the board, BCI companies may benefit from that changed posture — though there is no guarantee.
The new drug policy does not eliminate rigour. Makary and Prasad specified that the single trial must be paired with “confirmatory evidence,” which the agency defines broadly: pharmacokinetic data, evidence from related indications, results from the same pharmacological class, or robust animal models. For devices, the nearest analogue would be bench testing, cadaver studies, and safety data from earlier feasibility trials — all of which BCI companies have been accumulating for years.
The more pressing question for brain implant makers is not how many trials the FDA wants, but what those trials should measure. As a February analysis in STAT News observed, BCI companies find it easier to demonstrate safety than to define and measure improved patient outcomes. A typing speed of 90 characters per minute, as BrainGate recently demonstrated, is impressive in a laboratory context. Whether it constitutes a clinically meaningful endpoint that the FDA will accept for a permanent surgical implant is a separate and largely unsettled matter.
Synchron CEO Tom Oxley has said publicly that his company is in ongoing conversations with the FDA about what endpoints make sense for a Stentrode pivotal trial. The Stentrode’s endovascular approach — threading a stent-mounted electrode array through the jugular vein into the superior sagittal sinus — avoids open craniotomy, which simplifies the safety argument. But the device records from fewer neurons than cortical implants, which may limit the functional outcome measures it can achieve.
Neuralink faces the inverse trade-off. Its N1 chip samples from 1,024 electrodes placed directly on motor cortex, providing access to richer neural data but requiring open brain surgery. The company’s first patient experienced thread retraction — some electrode filaments pulled away from the cortex after implantation — an event Neuralink disclosed in 2024. Any pivotal trial will need to show that such complications are rare and manageable, particularly if the device is intended for patients who are not terminally ill.
Paradromics, with its Connexus device, is targeting speech restoration in patients with ALS-related anarthria. Speech decoding is arguably the most demanding BCI application: it requires real-time interpretation of attempted phoneme production from motor cortex. The company’s clinical endpoints will likely need to demonstrate not just that the device works, but that it works reliably enough and fast enough to replace or meaningfully supplement existing assistive communication tools.
The one-trial default also raises questions about postmarket surveillance. Makary and Prasad noted that the FDA plans to expand its collection of postmarketing data as the single-trial standard rolls out. For brain implants, which remain in the body for years and interact with tissue that changes over time — as a UMC Utrecht study published this month on nighttime BCI signal drift demonstrates — postmarket monitoring is not a formality. Signal degradation, electrode migration, immune response, and long-term biocompatibility all manifest on timescales that no pivotal trial can fully capture.
The Makary-Prasad policy does not directly change anything for device approvals. But it reflects an institutional willingness to accept leaner evidentiary packages when the science supports it. For an industry where each pivotal trial represents years of work and tens of millions of dollars, even a modest easing of the FDA’s evidentiary expectations could meaningfully shorten the path from laboratory to patient. Whether that easing extends to implantable BCIs is a question the next round of pivotal trial negotiations will begin to answer.