The FDA's Hammer Problem

The case for approving medical devices by what they do instead of the diseases they treat

Pretend for a moment you’re a founder who has created an innovative new device called hammer. Hammer can do many things, but you created it to serve one general purpose: hitting nails into things. There are all different kinds of nails. Hammer can hit any nail though, and so you’ve decided to market hammer as such: the tool that can hit any nail.

Now imagine a regulator steps in. He likes it. He agrees it hits nails. He just won’t let you say it hits any nail--not without proof, and not proof once, but separately, for every kind of nail there is. Roofing. Then drywall. Then framing, finish, brad, and every subcategory in between. You learn each nail requires its own multi-year trial, even though the tool is identical each time.

You spend a fortune proving what you knew on day one: the hammer hits nails. After you spend that fortune, the tool is only approved to hit one specific subtype of nails. If only it’d been approved to hit nails of any kind, it could’ve immediately helped every builder who needs to hit nails.

You spend a fortune proving what you knew on day one: the hammer hits nails. After you spend that fortune, the tool is only approved to hit one specific subtype of nails.

If only it’d been approved to hit nails of any kind, it could’ve immediately helped every builder who needs to hit nails. This is just a little taste of what it is like to submit a medical device to the FDA under a tool claim.

Starting off, I thought that there was only one path to go to market for a medical device with a 1:many value proposition: pick a disease, run your trials, get approved, and loop the same for each additional disease.

The pathway makes sense for biologics that are targeting hyper-specific pathways inside of cells, but for a broadly applicable tool, which could potentially be used for an array of disease treatment scenarios, it is incredibly inefficient.

On a call with another founder I was exposed for the first time to the idea of running with a tool claim. A tool claim--also referred to in FDA guidance as a “general intended use”--means the device is approved based on the function it performs, rather than every disease where a doctor might decide to use it.

Here’s what that means in a bit more color. Picture a device like the one we’re working on at AION: a device that can be guided to apply energy to a specific region of the body on demand, non-invasively with fields. The fields being applied are stimulating the environment of the tissue in such a way so as to push its biophysical state in some direction, such as increasing the positive charge or the negative charge.

This is actually what the tool we’re developing can do. Now, as to what it could be used for, that is a much more broad and open-ended answer.

I’ve written extensively on the potential therapeutic applications of directed energy to control bioelectricity in past posts, so I won’t rehash them all here, but suffice it to say, there are at least dozens, if not hundreds, of potential indications that could be targeted as a byproduct of having control over the biophysical levers in the body.

Take one simple example in the form of “tumor treating field” devices from Novocure, a publicly traded company that did over $650M in revenue last year.

Novocure’s core technology is a non-invasive, wearable, portable device that delivers low-intensity electric fields (typically 150–200 kHz) through skin-applied transducer arrays. The fields exert physical forces on rapidly dividing cancer cells.

This is fundamentally a biophysical tool--it targets the universal process of cell division rather than a hyper-specific molecular pathway inside one cancer type. Preclinical data shows it works across a wide range of solid tumors because the mechanism is physical, not biochemical.

Despite its broad applicability, Novocure never pursued a tool claim. Instead, they followed the classic indication-by-indication playbook, with separate Phase 3 pivotal trials, separate Premarket Approval (PMA) submissions, and separate branded devices. Optune for brain, Optune Lua for thoracic, Optune Pax for abdominal, etc.

After nearly two decades, they have FDA approvals in only four main indications: glioblastoma, mesothelioma, lung cancer, and most recently, pancreatic cancer this February. They are still running dedicated trials for brain metastases, ovarian, gastric, liver, and others.

Financially, this path has still built a strong business but it required enormous capital, time, and risk to run multiple massive trials instead of one foundational approval that could have unlocked faster physician-directed expansion.

Novocure is therefore a near-perfect real-world case study of a device that could have been regulated and labeled as a tool (“delivers tumor treating fields to disrupt mitosis in solid tumors”), yet they deliberately chose the slower, more capital-intensive 1:many indication route.

So why didn’t they go for the tool claim? That is the exact question that I tried to answer when I talked to the people who worked on the Novocure submissions and more broadly, other past medical device reviewers.

There are three primary reasons Novocure and other companies like it go disease-by-disease instead of one sweeping tool claim. The first is related to company strategy, the latter two are related to the FDA.

Copycat problem

A broad approval opens the floodgates for 510(k) copycats almost immediately. The 510(k) pathway lets a follow-on device secure clearance for a similar tool claim by showing it is substantially equivalent to an already-approved predicate--requiring far less clinical work and capital than the original device needed.

Regulators allow this shortcut on purpose: requiring every follow-on to carry the same evidentiary burden as the first would quickly draw the criticism that the system blocks obvious equivalents and discourages iterative tool innovation in the first place.

Disease-by-disease preserves a capital moat where every competitor has to repeat the enormous clinical spend.

“Being just sort of in an adjunctive therapy to accent drugs really limits the ability for any copy cat. Whereas if you were even able to get some sort of de novo submission as a sort of a general tumor treating field claim, well then anyone can come copy you in a very short order with a 510k.“

— FDA Attorney

In fact, this is exactly what happened to Raymond Damadian when he invented the MRI!

FONAR, the company Damadian founded, spent years and millions securing the very first FDA Premarket Approval (PMA) for a commercial whole-body MRI scanner in 1984. That approval was a classic “tool claim”: a broad, general clearance for magnetic resonance diagnostic imaging of the head and body, based on how tissue water structure and ion dynamics (the very bioelectric signals Damadian had targeted) altered relaxation times. It was the regulatory equivalent of proving the device could read cellular states non-invasively. Once that predicate existed, the FDA eventually reclassified MRI systems from Class III to Class II. The floodgates opened.

GE and Siemens, who had already been copying the hardware, simply rode FONAR’s regulatory precedent. They secured their own initial PMAs, but every subsequent model--right down to today’s high-field SIGNA and MAGNETOM lines--cleared in months via the 510(k) pathway as “substantially equivalent” to the predicates FONAR had created with no new clinical trials required. Damadian won the patent lawsuits and the $128 million settlement, but commercially he was scooped.

I wanted to give this reason first because it is actually the strongest argument against going for a tool claim. Damadian invented the MRI. He was right, he was first, and the broad clearance he opted to pursue ended up paving the way for his competitors to get to market faster and cheaper, effectively crippling his advantage in the process.

Tool claim or not though, that competitor scoop was arguably inevitable. GE & Siemens had the distribution channels, manufacturing infrastructure and balance sheets to beat FONAR any way the race was run. And so avoiding a tool claim for the sole purpose of preserving a regulatory moat has its counter arguments.

It also proposes a question: what are you optimizing for as the founder? If the answer is a cornered resource moat, then disease-by-disease is the rational choice. It forces every competitor to repeat your spend and buys you years in first mover advantage. That said, it buys that advantage by keeping your company’s solution out of the hands of patients who could potentially benefit from it for the longest possible time!

For us at AION and I suspect for Damadian as well, to optimize for this is to admit to optimizing against the reason you built the thing. I haven’t reconciled this yet strategically in our case, and writing this post is a starting attempt at doing so.

With that, let’s dive into the two other reasons Novocure didn’t pursue a tool claim: FDA baggage.

Drift in approval culture

From what I've come to understand across close to a dozen conversations now, review culture inside the FDA's Center for Devices and Radiological Health (CDRH)--the center responsible for regulating medical devices--has drifted to preventing many companies from pursuing tool claims, despite no statute or regulation actually forbidding them.

“FDA has a lot of unwritten rules, and so the review culture...over, I would say, the last decade or so has come to frown upon generalized intended uses.”

— former FDA medical device reviewer

The shift is the product of an accumulation of unwritten norms, reviewer training emphases, and institutional risk aversion within CDRH that now frowns on generalized intended uses.

In practice, this means that when a sponsor walks in with a platform technology, the conversation often pivots quickly from “what does this device do?” to “what specific diseases or conditions are you claiming it treats?” CDRH reviewers may not outright reject a broad tool claim, but they will probe for specificity in a way that feels like an inevitable funnel toward narrower labeling.

A bankrupting “yes”

The last reason against going for a tool claim as a broadly applicable medical device flows from the second: CDRH’s current review culture makes most tool claims economically infeasible.

What this means is that when the reviewers ask, “What specifically do you foresee this being used for?” and you name the range of uses--they reply, “Great! Run an adequately powered efficacy study for each one.” Technically a yes. Functionally a no with a billion-dollar deductible.

“If you want to go run like a billion dollar trial and make sure you’re properly powered for every possible thing under the sun--Yeah, you could go get generalized intended use--Sure. And of course, no one does that.”

— FDA consultant to early stage biotechs

Giving Mike the mic

And so here’s the break point between how I understand the FDA works in practice and how it is instructed to work on paper.

This is also the discrepancy that the same founder who introduced me to the idea of a tool claim for the first time is actively working to fix.

Mike Petegorsky is the co-founder of a company called Proxima Health. Mike’s path to Proxima started when his wife was pregnant and he began researching non-obvious threats to his unborn daughter’s health. His research pointed toward microplastics, PFAS, and other environmental toxins, leading him to a harder question: if we can’t fully prevent exposure to these toxic chemicals, can we remove them from our bodies? That inquiry led him to bringing blood purification technology to the forefront of chronic disease prevention and treatment.

Mike previously practiced as an intellectual property litigator at a top law firm, and also led legal and regulatory at a psychedelic medicine company, and so when he started to bump into the limitations laid out above, he took it upon himself to work towards a solution.

It’s for this reason that when I started to bump into those same limitations and draft this blog, I tapped Mike to lay out the below:

Law vs practice

The legal foundation here is simple, and shows where written law and current practice have come apart.

The FDCA does not say that every medical device must be authorized one disease at a time. The statute says a medical product is a “device” if it is:

• “Intended for use in the diagnosis of disease or other conditions, or in the cure, mitigation, treatment, or prevention of disease, in man or other animals,” or

• “intended to affect the structure or any function of the body of man or other animals”

That second prong means that a device is approvable if it affects a structure or function of the body, even without being tied to a specific disease.

Congress wrote the law this way because some medical technologies are tools that perform a broadly-applicable function, and should not be reduced to a single disease.

FDA’s own written framework recognizes this distinction. The agency’s 1998 General/Specific Intended Use guidance distinguishes between “general” intended uses (i.e., tool claims) and “specific” intended uses (i.e., disease claims), both of which are permitted. Under the law, a device can be described by the medical function it performs, rather than by every disease in which that function might be useful.

Yet in practice, sponsors are often pushed in the opposite direction. Even though no statute or regulation requires a disease-specific intended use, CDRH review divisions pressure companies to "pick a disease." The result is a gap between what the law permits and what companies can practically pursue: tool claims remain legally available on paper, but in many areas are so difficult to pursue that sponsors abandon them.

The damage is hard to overstate:

• Clinical trials are significantly more expensive and take years longer

• Companies must repeat trials for each additional disease, even where the device is used the same way

• Authorized uses are artificially narrowed, pushing clinically-accepted uses off-label and outside insurance coverage

• Physicians don’t adopt new technologies with limited labeling and insurance coverage

• Patients lose access to life-saving technologies

This damage isn’t just hypothetical: it’s occurred with focused ultrasound devices, transcranial magnetic stimulation (TMS) and deep brain stimulation (DBS) devices, and countless others that have been forced to pursue narrow indications one by one, spending extra years and millions of dollars and keeping additional uses off-label.

Off-label use by physicians is legal, but manufacturers cannot promote or support it. Insurance coverage is poor without a labeled indication, and doctors are often reluctant to adopt at scale when they lack clear labeling and company backing. A tool claim solves this by making the core function legally promotable and reimbursable.

The system also creates the wrong incentive for builders. Instead of developing broad platform technologies that could help many, they are pushed into the narrowest, immediately approvable uses that only help a few.

But there’s good news: this regulatory drift can easily be fixed.

How FDA can solve this

Because the FDCA already allows tool claims, no new statutes or regulations are required. Fixing it requires only that FDA say what the law already allows, and direct reviewers accordingly.

FDA can act the same way it did in December 2025 on real-world evidence, when it removed a long-standing, self-imposed barrier through a policy statement and updated guidance, expressly to get tools to patients faster. The same mechanism works here.

FDA could issue new guidance explicitly supporting tool claims, describing the types of devices for which they’re available, and explaining how to structure trials to support the claims.

Or even more simply, FDA could issue a policy statement returning to first principles under the FDCA and the 1998 General/Specific Intended Use Guidance, and direct reviewers accordingly.

Whatever the mechanism, FDA should make clear that companies have a choice: they can pursue a disease claim, with the potential moat and also the costs that come with it; or they can pursue a tool claim, with faster/cheaper approval and broader post-approval use.

The key is that the evidentiary burden should follow the claim. Disease claims require disease evidence. Tool claims require evidence that the tool safely and reliably performs the claimed function.

The benefits of returning to tool claims would be substantial:

• Massively reduce development time and cost

• Expand physician choice and patient access

• Reduce off-label use driven by artificial label narrowing

• Encourage sponsors to develop broadly-applicable “platform” technologies rather than the narrowest approvable use

None of this means that the FDA should rubber-stamp unsupported claims. A tool claim still must be bounded:

• Tool claims should be available for devices that perform the same medical function, through the same mechanism, with the same material risks, across multiple clinical contexts.

• If the proposed use creates materially different risks in different clinical settings, CDRH can require evidence addressing those risks.

• If the sponsor wants to claim disease modification, symptom improvement, survival benefit, or diagnostic accuracy for a specific disease, FDA can demand the clinical evidence necessary to support that claim.

• Companies cannot imply disease-treatment outcomes they haven’t proven.

The point is that companies should get to choose the type of claim, and the approval burden should follow the claim.

The best version of FDA is not an agency that makes every new tool pretend to be a single-indication drug. The best version of FDA is an agency that asks the right question: what does this product claim to do, what risks does that create, and what evidence is necessary to prove it?

---

Thanks, Mike.

I wanted to write this up because fundamentally, the device we’re building at AION is a hammer. It hits a lot of nails. I can spend the next two decades and a few hundred million dollars proving it on one nail at a time, or I can aim to say what’s already true: it hits nails.

-Benjamin Anderson

Follow me on X.