Clinical evidence strategies that support both the EU/UK and the US

"Do we need to do a clinical trial?"

It's the most common question in early MedTech, and often the most expensive one to get wrong. Let me get this out of the way early, not every device needs a prospective clinical trial. 

That doesn’t mean you don’t need some clinical data for your product submission, there is a difference there. More on that later.

There are many flavors of clinical studies to choose from. You will match the right one to your particular device’s needs to demonstrate its clinical performance and safety. You will need to choose carefully as to what clinical evidence approach you’ll take. In the end, it depends on the market you’re going after and what you’re claiming your product does. 

The US and the EU/UK approach requirements for clinical evidence quite differently, with the US not requiring any substantial clinical evidence for some devices prior to regulatory clearance. 

In this blog, I’ll walk you through how to approach clinical testing so that you can make sure your protocol is accepted with a smile from regulators on both sides of the Atlantic.

Your product determines your study type

When someone asks me if they need a clinical study, the first question to ask is: is it true your device’s clinical performance and safety can ONLY be proven by a prospective clinical study? 

For example, it is difficult to demonstrate that a digital therapeutic device works only using retrospective data. This is because you need to provide your device as an intervention and see how your potential patients react. 

However, if you have a diagnostic software examining image data, that is usually a great use case for retrospective data as that is what your device encounters on the market, can be easily gathered retrospectively, and doesn’t typically require an interventionist approach.

So if the other options are viable, you should consider them all to figure out which one is right for you. 

Choosing the right study will come down to finding the most appropriate way to substantiate your intended use statement and your claimed clinical benefit. Let’s walk through some options.

Retrospective study. Work from data that already exists like a sequestered multi-site dataset for software, chart review or hospital data, and measure performance against a reference standard or a comparator group. The catch here is that you can’t always adjust for all the edge cases you may want as they may not be in your data set. Also, these are inherently non-interventionist and depend on your ability to gather the appropriate data.

Reader study. Stated simply, clinicians read cases with and without your device, which allows you to measure what your device does with a clinician rather than what it does alone. If your claim contains the word "aids" or is in decision support, this study is likely one you’ve considered. Regulators often warn that whichever type of reader study design you pick the design should mirror how the device actually gets used to ensure no issues on bias.

Prospective observational study (silent study). The device runs live in the real environment, but its output is hidden from the care team and changes nothing about what happens to the patient. You get to see your actual population, workflow and prevalence without exposing anyone to an unvalidated output and see how your device would perform. One honest caveat: you get performance, not impact, because nobody acted on the output.

Randomised controlled trial. The filet mignon of studies. Randomize, control, measure what happens to the patient, which makes it a design that supports most causal claims about outcome. Usually this is the most resource intensive (read: time and money) of all the studies to run but the most robust for your submissions and claims. Every regulator likes these usually.

There are more subtypes of study to choose from but I won’t touch on them all as there are just too many options. No regulator will outright tell you which study type to choose but FDA has an extensive list of guidance documents for different clinical trial topics and IMDRF has some as well. 

An important side note: when thinking about your study make sure that you’re choosing the right endpoints. You want to test specifically what will back up your intended use and benefits you’re claiming. 

Also, make sure that the subjects you’re using to test make sense for your device. Are race, gender, age, disease severity, clinical condition, etc. makes sense for the clinical environment and eventual patients your device will encounter in the real world?

The EU likes standalone clinical data.

The way that the EU thinks about clinical data is that they ask: Can the clinical evidence you provide stand alone in demonstrating your device is safe and performs as intended? Under MDR, conformity rests on clinical data providing sufficient clinical evidence, and you specify and justify the level of evidence yourself. Essentially, you have to prove it to the notified body. 

The EU does have an “equivalence” pathway you can use for clinical data. It is a distant cousin to the US FDA’s predicate and 510(k) pathway, but only applies to the clinical evaluation. Using equivalence to a device on the market, you claim your product is the same and you can leverage their on market clinical data to lessen how much pre-market clinical evidence you have to put in your submission. 

There is a catch though. I have found it is very difficult to claim equivalence now. Most notified bodies prefer you provide your own standalone clinical data in support of your submission, and make the equivalence requirements very difficult to pass. 

If you’re creating implantables or class III devices, this won’t matter as the EU requires prospective clinical investigations by default for high risk products. 

All this data ends up in your clinical evaluation report, a comprehensive document outlining the literature and studies related to your device and state of the art to demonstrate your clinical evidence supports the clinical performance and safety. 

So if you’re asking what kind of clinical data you’ll need for the EU, the answer is you’ll need some sort of testing that covers your performance and safety. Whether that is retrospective data testing or a full RCT depends on your device.

The US clinical evidence depends on your regulatory pathway.

The US doesn’t specifically ask for standalone clinical data for all of its submissions like the EU does. Instead, you have submission pathways like the 510(k), where clinical data can even be omitted as long as performance and testing data matches what was done for the predicate device. 

A 510(k) argues “my device is exactly like this other device on the market, so I’ll test my product the same way they did and if it is not inferior, then that is sufficient.” This argument is called  substantial equivalence and must be compared to a specific legally marketed device in the US (called a predicate). 

FDA's own position is that in many cases substantial equivalence can be shown through robust non-clinical safety and performance data alone. However, many 510(k)s do require clinical data. 

So how can you tell?

There are four good places to check for this:

• Your classification regulation. Find the FDA’s product code and read the regulation in 21 CFR. If your class carries special controls they're listed there, they sometimes name clinical performance testing outright, and they're binding. Easiest place to check first.
• Device-specific FDA guidance. Under the FDA product code page, you’ll see relevant guidances and standards. The FDA calls guidance "non-binding recommendations" but you can treat them as gospel, because that's what your reviewer does. If the guidance “recommends” clinical testing, you’ll probably need clinical testing.
• The 510(k) summary database. Read the summaries of recent clearances under your product code, especially your predicate device’s summary. If your predicate device has done clinical testing, there is a good chance you will need to too. To be sure, check the last ten cleared products under that product code and see if they also did a clinical trial. That is an obvious 
• FDA presubmission. Your best source? The FDA themselves. Once you’ve decided on your predicate and product code. Propose what testing you want to do and ask if the FDA is ok with it. Don’t ask the FDA if you should do a clinical study, give them a justified testing strategy that is lean and have them push back if it isn’t sufficient. Check our blog on pre-submissions for advice here.

If you aren’t 510(k), then your options may look different. Let’s take De Novo for example, which is a different approach entirely. 

De Novos are required when there is no predicate or product code for your device, so nobody's comparing you to anything. In that way, the FDA is deciding whether there's a reasonable assurance of safety and effectiveness from the data in your submission. This reads very similar to Europe’s standalone approach. So, clinical data is not a legal prerequisite here but expect most every De Novo to require substantial clinical evidence for the submission.

Premarket Approvals (PMAs) are for Class III products and, similar to the EU, you are required to do a prospective clinical study for these types of devices. No real arguments you can make here. 

Using clinical data from outside your target market

The question we’ve all been waiting for: Can you use external data for the market you’re submitting to. The answer is an outstanding yes. I have seen the US accept clinical data from the EU, and vice versa. Even data from other countries like Australia, China, and Africa. 

However, what portion of it can make up your core clinical evidence of your submission depends on the type of study and type of device. The study should be well designed and done under Good Clinical Practice (GCP). More on that can be found in the ICH E6 R3 document.

FDA has published a set of FAQs to help elucidate this a bit more but it stops short of giving you any explicit framework to work off of. I find that these criteria apply well to either market accepting foreign clinical data though. The important ones are:

1. Is it done according to GCP? If regulators can’t be assured you ran the study according to proper protocol, using ethics committees / institutional review boards where needed, with the right protections in place, then they won’t accept the study on merit.
2. Is the study applicable to the US population and US medical practice? This seems straight forward but works both ways for the US and EU. If you set up a study at only one site in Denmark but clinical practice involving devices like yours is significantly different between Denmark and the US, how can you guarantee that when you release the product in the US you will see the same results when US physicians use it? A scalpel will be used in the same way in most every hospital. However, a decision support tool for deciding when a pregnant woman should go for a C-section many differ quite significantly from country to country.
3. Is the study valid for your intended use? If I have a device that treats a mental health condition and set my study to only include subjects aged 25-60 but want to claim that adolescents can use it, I am in serious trouble. This is because we know performance may very likely vary in adolescents given their social experiences and cognitive behaviors are fundamentally different from adults. 

Regulators want to be assured that when you release your device, it will work as intended and be safe for the users/patients in the market they regulate. This means your study design and location should consider the representative population that will be subjected to the device when it’s released.

Building the right plan

How do we approach these markets then? Start by knowing which type of product you’re bringing to each market. 

Think first about your device’s intended use and claims. Will they be the same in each market? Or, for example, is the only available FDA predicate have a slightly different intended use or claims than you intended for the EU market that you’ll have to use. These can affect how you run your study and even whether you need a whole separate device technical file to submit to each market.

Next, think about your clinical use environment. Is it broadly the same in the market you wish to collect data from and where you plan to market it? Are there some key differences? 

You can create justifications here for regulators to consider. Maybe there are differences between the way clinicians would use a device between markets but you would imagine they’re negligible. Or better yet, someone published a peer-reviewed publication saying they’re negligible. Use this as an argument.

Finally, choose your study type, build a clinical validation/investigation protocol, and share it with regulators from each market. Do a pre-submission with the FDA or do a structured dialogue with your notified body and see if they agree with your proposed plan. Again, don’t just ask them if you think you need a randomized, controlled trial. They’ll just say yes.

If they say yes, you’re well on your way. It’s not always the case you can manage with one study for both markets. Sometimes the predicate comparison needs endpoints the EU benefit argument can't accommodate, and forcing them together gives you two mediocre datasets instead of one good one. But you can almost always avoid running two full ones. 

The real mistake is committing blindly to multiple studies, no studies, or the wrong study and hoping for the best. By talking to regulators early you’ve saved yourself the potential risk of reworking a clinical study, which in the end can save you millions, years to market, and even the success of your company in some cases.

Deciding what evidence you actually need?

A guidance document or a chatbot will give you an answer on study design, and it can be confidently wrong. This is because it doesn't know which arguments reviewers have actually accepted and which they've thrown out. That knowledge isn't public. You learn it by chatting directly with FDA or notified bodies to know how they think.

Dovetail is built to tackle this head on. We're not a consulting firm and we're not an eQMS. We're a team of regulatory experts, a validated platform, and an AI layer trained on devices we've actually cleared. One system, not three. Expert-led, AI-enabled, audit-ready. We've been in the room with the FDA and with notified bodies, and we build the documentation you hold in-house, so we know it will pass.

Want to learn more? Reach out to us today.

Dr Spencer Todd is the CEO and co-founder of Dovetail (formerly FormlyAI), a regulatory partner for medical device companies. Before founding Dovetail, he spent years at the FDA and as a medical device consultant.