Who actually generates the clinical evidence in your category?

Device marketing teams tend to treat clinical evidence as something the company produces: a pivotal study, a post-market registry, perhaps a white paper built on both. Evidence is a line item, owned by clinical affairs, referenced by marketing when it is ready.

The public trial registry describes a different world. Most of the studies being run on devices are not being run by the companies that make them, and the imbalance has been widening for a decade.

That registry is public, queryable, and almost entirely absent from the marketing function. Regulatory teams read it. Clinical affairs read it. Marketing teams, who arguably have the most to gain from knowing what the wider literature is about to say, generally do not open it at all.

Who is actually running device trials?

Not manufacturers, mostly. Of the interventional device studies with a recorded lead sponsor class, industry sponsors 26.1 percent. Academic institutions and other non-commercial organisations sponsor 69.5 percent. Government, network and individual sponsors account for the rest, none exceeding three percent.

That ratio should reframe how a marketing team thinks about the evidence base in its category. The majority of it is being created by investigators the company does not employ, cannot brief, and usually does not know about until the study is registered or published.

How lopsided is the split?

The tail is thin. Across all sponsor classes, academic and other institutions account for 44318 studies and industry for 16665. Every remaining class combined is smaller than a rounding error against those two: other government 1597, other federal 572, network 287, NIH 253, individual 69.

The NIH figure deserves a second look, because it is the one most people guess wrong. Federal research funding is a large presence in drug development and a very small one in device trials by lead-sponsor count. A device company building a regulatory or payer narrative around federal research interest is building on 253 studies out of more than sixty thousand.

The concentration in two classes also means the practical question for any manufacturer is binary. Either a study in your category is yours, or it belongs to an academic centre. There is no meaningful third population.

Is industry’s share rising or falling?

Falling, and not slowly. Industry sponsored 23.7 percent of device studies started in 2015 and 20.9 percent of those started in 2025, a decline of 2.8 percentage points. Over the same period the total number of new device studies grew by 1583, from 3145 to 4728.

Both halves of that sentence matter, and they point the same way. The category is generating more evidence overall while manufacturers are generating a shrinking proportion of it. Academic research volume is growing faster than commercial research volume, so the gap compounds.

The shape of the line is worth reading carefully rather than as a straight decline. Industry share rose through 2018, dipped in 2019, recovered to a peak in 2021, and has fallen in every year since. The recent decline is the most sustained movement in the series, and it is the part least likely to be an artefact of the period.

Does the split look the same in every category?

No, and the variation is wide enough that the overall average is close to useless for any individual company. Across seven device-heavy condition areas, the industry-sponsored share ranges from 13.0 percent to 44.2 percent. A manufacturer planning against the all-category figure of 26.1 percent will be wrong in most categories, in one direction or the other.

Heart failure sits at the commercial end: 528 industry-sponsored studies out of 1194, or 44.2 percent. Chronic wound care is close behind at 41.7 percent, though on a much smaller base of 60 studies, which is few enough that the rate should be treated as indicative rather than precise. Osteoarthritis is 35.9 percent, diabetes 31.8 percent.

At the other end, anterior cruciate ligament research is 13.0 percent industry-sponsored — 20 studies out of 154. Sleep apnea is 28.9 percent and COPD 27.6 percent, both close to the overall average.

The pattern behind the spread is roughly what a commercial reader would expect, and it is worth naming because it generalises. Categories where the device is expensive, implanted, and tied to a procedure attract commercial research money. Categories where the device is cheaper, used in rehabilitation, or prescribed as part of a broader care pathway attract academic attention instead.

That has an uncomfortable implication for the sports medicine and rehabilitation end of the market. A company selling into anterior cruciate ligament care operates in a category where roughly seven of every eight studies are run by someone else. Its own evidence will always be a minority voice in its own category, and a marketing strategy built on out-publishing the academic literature is not going to work.

The constructive reading is that in those categories the independent evidence is abundant, free to cite, and more credible to a clinical buyer than anything the manufacturer funds. The scarce resource is not evidence. It is the work of finding, organising and interpreting evidence that already exists.

It is worth being precise about why the base sizes differ so much, because the numbers invite a wrong inference. Anterior cruciate ligament care shows 154 device studies against 2493 in diabetes. That gap is mostly a statement about how many people have each condition and how broadly the registry’s condition search matches, not about how commercially interesting either category is. Comparing absolute study counts across conditions is close to meaningless. Comparing the sponsorship ratio within each one is not, because the ratio is scale-free.

The chronic wound figure illustrates the opposite failure mode. At 41.7 percent it looks like one of the most commercially active categories in the table, but it rests on 60 studies in total. Twelve studies moving from one sponsor class to the other would change that number by more than ten points. A rate calculated on a base that small should inform a hypothesis and nothing more, which is why it is reported here with its denominator attached rather than as a clean percentage.

How much does the gap compound?

Faster than the share figures suggest, because both terms are moving. Industry share fell 2.8 percentage points while total annual studies grew by 1583. A shrinking slice of a growing pie means the absolute gap in study counts widens every year even in the years when the share looks flat.

The practical consequence is a widening literature the manufacturer has no hand in. For a marketing team the relevant quantity is not the ratio at all; it is the absolute number of studies in the category that the company did not commission and has not read. That number rises every year in almost every category here, regardless of what the percentage does.

Why does this matter to a marketing team?

Because the evidence that shapes buyer belief is increasingly evidence the manufacturer did not commission and cannot control. A clinician forming a view about a device category reads what has been published about it, and roughly seven in ten of those studies came from an academic centre with no commercial relationship to any manufacturer in the category.

This has three consequences that show up directly in a marketing plan.

The first is that third-party evidence is abundant and largely unexploited. Manufacturers spend heavily to produce their own studies and then cite almost exclusively from that small library, while a much larger body of independent work sits in the registry unread. Independent evidence is also more persuasive to a clinical buyer than sponsor-funded evidence, which is the opposite of how most device companies allocate their evidence budget.

The second is that competitive surprise arrives through this channel. A study registered by an academic centre comparing two approaches in your category is public at registration, often years before publication. Companies that monitor the registry find out early. Companies that wait for publication find out when their customers do.

The third is a positioning risk. When most category evidence is independent, the category’s language and framing get set by investigators rather than by marketers. A manufacturer that wants its own framing to become the standard has to engage with that literature rather than publish alongside it.

What would make this analysis wrong?

Three things, in descending order of how much they would change the conclusion.

The first and most serious is sponsor misattribution. Lead sponsor class records who registered a study, not who funded it. A manufacturer that funds an investigator-initiated trial run through a university appears in this data as the university. Commercial involvement in device research is therefore larger than 26.1 percent by a margin nobody outside the individual agreements can size.

That single limitation is enough to make the level unreliable. It is not enough to make the trend unreliable, and the distinction matters. For the falling industry share to be an artefact, the practice of routing commercial funding through academic sponsors would have to have grown substantially over the same decade. That is plausible. It is also testable by anyone with visibility into funding disclosures, which the registry does not carry.

The second is registration behaviour. Some studies must be registered by law and others are registered voluntarily to satisfy journals. If voluntary registration among non-commercial researchers has become more common — and journal requirements have tightened over this period — then part of the apparent growth in academic studies is growth in registration rather than growth in research.

The third is the breadth of the device filter. It captures a surgical implant, a continuous glucose monitor, and a software tool used in a behavioural study under one label. Condition-level cuts are more meaningful than the overall figure precisely because they narrow that range, which is why the category table above carries more weight here than the headline percentage.

One smaller point of method, stated because leaving it out would be the kind of quiet rounding this publication tries to avoid: sponsor shares are calculated against the sum across sponsor classes rather than the unfiltered total. Those differ by 5 studies with no class assigned. Immaterial, but stated.

Condition counts come from the registry’s condition search, which matches synonyms and related terms. The seven categories above overlap and do not partition the whole. They were chosen to span the device markets this publication covers, not sampled at random, and no significance test would be appropriate on a selection made that way.

What follows for an evidence strategy?

Stop treating the company’s own studies as the whole of the evidence base, and start treating the registry as a standing input to marketing rather than a regulatory filing system.

That reframing has a concrete shape. An evidence strategy built on this data has three components, and none of them require new research spend.

Inventory what already exists. Every category has a body of independent work that the manufacturer has never catalogued. Listing it, tagging it by claim, and keeping it current is a few days of work that most competitors have not done. Independent evidence also carries more weight with a clinical buyer than sponsor-funded evidence, so the material is not merely cheaper — it is better suited to the job.

Watch registrations, not publications. A study is public when it registers, which is typically years before it reports. A competitor’s category being studied by an academic centre is knowable long in advance. Companies reading only the published literature learn the result at the same moment their customers do.

Know the investigators. In a category where most research is independent, the people running that research set the vocabulary the market eventually adopts. A manufacturer that engages with that work early has some influence over the framing. One that appears only after the framing has set is arguing against an established consensus.

The underlying point is narrow but consequential. In most device categories the manufacturer is not the main producer of knowledge about its own category and has not been for some time. Marketing plans that assume otherwise are planning around a version of the market that the public record does not support.