Why Clinical Trials Still Fail & How Faro Health Is Fixing the Design Problem With Scott Chetham

• Scott Chetham has spent over 20 years in clinical research, enrolling patients, designing programs, and watching trials collapse under the weight of their own complexity.

• At Faro Health, he's tackling the root cause: protocols get over-designed based on habit and fear, not data. His platform brings structured design intelligence to a process that still runs on Microsoft Word and copy-paste.

• With drug development averaging 10 to 12 years, billions of dollars, and a 6 to 9% success rate, Scott sees this as the first moment in his career where real, meaningful change is actually happening, and fast.

Let's dive in. No floaties needed…

Revenge of the Nerds

Scott Chetham, Co-Founder & CEO of Faro Health

He didn't have one breaking-point moment with clinical trials. It was more like a death by a thousand cuts: assumptions about patient enrollment that never held up, timelines that stretched far beyond estimates, and resources that ballooned to compensate. After more than two decades across hospitals, research labs, startups, and venture capital, Scott co-founded Faro Health in 2019 to fix the problem at the source: trial design itself.

He also sits on the board of CDISC, the global nonprofit behind clinical research data standards, giving him a front-row seat to how the entire industry is (slowly) learning to talk the same language.

You've worked every layer of trials, from enrolling patients to designing programs at scale. When did you first feel that the system was broken?

It's not a single moment; it's more like a death by a thousand cuts. You design a trial and make a lot of assumptions. You assume the protocol is feasible, that patients will be able to accommodate the extra visits, the extra testing, the extra care required. That often turns out not to be true. You assume patients can be found at certain hospitals and that those hospitals can conduct the trial. That turns out not to be true either.

It keeps going. You might estimate 12 to 18 months to enroll a trial, but it gets longer and harder, and you have to throw more resources at it. We make assumptions based on experience, and our experience as individuals is quite limited by where we've worked, within a particular country. It's not the same everywhere.

The reason this matters so much right now is simple math. Getting a drug to market takes 10 to 12 years, costs a couple of billion dollars, and you have somewhere between a 6 and 9% chance of success, and we're getting worse every year. There aren't many areas in the world where you'd invest that kind of money with a 6 to 9% chance of success. Imagine Boeing saying they're going to build a plane, but there's only a 6 to 9% chance it might ever fly. It just doesn't practically work. We have no choice but to fix this.

You've said that complexity is the strongest predictor of trial failure. What fear drives teams to overdesign trials & why is that behavior so hard to reverse?

At the highest level, the more data I collect from patients, the more hospitals and countries I go to, and the harder I make the inclusion and exclusion criteria. Every time I add something, it gets harder, and the likelihood of success goes down. That's well known.

Why is it hard to change? Because until now, this field has been based on experience. What we tend to do is open what worked last time, or what might have worked, and copy and paste. We use the same thing as a starter. There's also this fear of being wrong. The question is always: if I don't collect this piece of information, will it ever be available later? That fear drives the over-collection of data because there isn't data showing the consequence of doing it.

What we've seen is that when you bring data to the table at the time people are making these decisions, they do change. But without data, you just default to the way you've always done it.

What goes into writing a clinical trial protocol?

A protocol can be about 200 pages, sometimes longer. It describes in narrative form what you're testing, in whom, when, and how you'll analyze the results. The design part is where you make the foundational decisions. Let's say you have a new weight loss drug. The trial would measure how much weight a patient lost from week one to week 52. Simple enough. But then you need to prove the drug is safe by checking liver function, kidney function, and tolerability. You start building this big table of assessments, and it gets larger as new questions come up. Maybe patients lose muscle mass instead of fat. Now you need a test for that.

Then comes the question of who you're testing it in. Male or female? Older than 18? You don't want cancer patients in this trial, so the exclusion list grows. That's all design. Authoring is the wrapper around it: the scientific rationale, the background on the molecule, how you'll protect women of childbearing potential, contraception methods. It all gets laid into this massive document we call a protocol. It's time-consuming because it's incredibly dense.

When teams design trials, what operational constraints do they most underestimate?

Traditionally, we've very much underestimated how much burden we put on patients and on the site staff who work in the hospitals. It's easy to write down that you want a blood test, but that means the patient has to come in. A staff member has to look up what needs to be done, draw the blood, prepare it for shipping, label it, and enter it into a separate computer system. Clinical trial data doesn't go through the normal hospital collection pathway. Most of the time, it has to go off to a central lab for analysis.

There's a lot of extra labor for both the patient who has to show up and the staff who have to execute all of it. And traditionally, we do a very poor job of estimating what it truly means to be part of these trials.

You helped sponsors prove to regulators that certain trial requirements were operationally impossible. What does it take to convince the FDA that something won't work?

I can give a real example. This was an early-stage test in pediatrics, children as young as two up to 16 or 17. Because it was early, they had to collect blood samples before the drug was delivered in the morning, then again eight hours later, three to four days in a row. These facilities require a cold centrifuge immediately after drawing blood, which limits where this can happen.

So you'd have two-year-olds coming into a site very early in the morning, doing testing, then waiting around for eight hours for another blood draw, three to four days in a row. I have children. There's no way that would happen. It's a massive ask for everyone involved.

What the team was able to show the FDA was that this was just extremely hard on children and their families. They came up with a novel mechanism where, instead of every child doing three or four days in a row, each child would only stay once. They randomized patients into different parts of the testing schedule, and scientifically, they worked out how to get equivalent information. The FDA agreed.

It worked because there was data, and they could clearly present that this just wasn't possible. You read it on paper, and it's abstract until you see it laid out day after day. Then it becomes obvious.

You've argued that we're wasting capacity on trials that should be shut down. What incentive keeps those zombie trials alive?

Unfortunately, yes, zombie trials are real. Here's what happens. A new indication emerges. Let's say in oncology, someone wants to develop a better treatment for a particular cancer. Six or seven companies might launch programs at the same time and start clinical trials simultaneously. Meanwhile, one company's drug is effective, gets approved by the FDA, and becomes the standard of care.

Now you've got an approved drug that's standard of care, but you also have older drugs still in trials, compared against nothing. The problem is that's the wrong test. Now, you’d need to show your drug is either superior or inferior to the approved one. But companies keep those older trials enrolling, often for financial reasons. Shutting a program down can hurt your stock price. They should be shut down, but there are more instances than many of us would like. I do think we're getting a little better.

Everyone talks about AI in medical writing. What do most people misunderstand about generating a 200-page clinical protocol with real quality?

The important thing is we don't have trouble authoring documents; we have trouble designing them and then putting words around the design. They're different things.

You can ask any LLM to design a protocol. It will look partly convincing, but it's garbage. When you train these models across roughly 60,000 protocols in the public domain, you get regression to the mean. Ask for a solid tumor protocol, and you'll get the exact average of all solid tumor protocols out there. But that's not what you want for a particular molecule with novel safety parameters and a novel mechanism. We're doing inherently net-new biology.

You have to couple design with authoring in a way that wraps language around established design concepts. LLMs write convincing things, but writing isn't the problem; effective design is. We have to educate people on that constantly. There's a separate problem in the submission space, where companies already have all the prior design documents, collected data, and everything from the trial. There, AI just rearranges existing content into the format regulators want. That's very time-consuming, often several months, and AI can do it in days. But they inherited everything. We solve the other problem: how do you design better, and then how do you write language around what you designed?

If everything works the way you hope, what will be fundamentally different about how drugs move from idea to patient five years from now?

It won't be just Faro that solves it; it'll be a combination of solutions. But I think we can get that 10- to 12-year timeline down by at least half. That would be a monumental achievement.

It'll happen through speeding up protocol design and authoring, providing data to make protocols more effective and efficient, enrolling more patients, finding the right patients at the right time, and pre-identifying them. You lose months just putting together contracts for hospitals, and automation can compress that time. It's death by a thousand cuts, and it's about taking all those thousand cuts away.

I've been doing this for over 20 years. This is the first time I'm seeing real, meaningful change happening quickly. Everyone's aligned on the fact that we have to solve this. I'm hoping for 50%.

What founder lesson did you learn the hard way while building Faro?

There are many hard lessons. You're on a continuous learning curve, and it doesn't stop. In some ways, you have to be pretty arrogant to start a company, because you have to think you can do it better than anyone else. But you also have to quickly realize that you need to hire people who can do every job better than you. Getting comfortable with hiring people who can make you redundant or make you look good is a key part of the journey.

The other thing is you can't talk to customers too much. Even though I have done this job for many years, I need a breadth of experience from other people. You don't want to build a solution that only works for your situation; it has to work for everybody. But you also have to get good at identifying what feedback is actionable and what's noise. Without focus, you get pulled in a million directions.

One of the big mistakes I made was not reinforcing focus often enough. You have to keep repeating the message almost weekly to the entire company, because it drifts. Internal messaging is like playing telephone; one manager repeats it to the next, and it shifts. I've found it's just easier for me to communicate the same message constantly, so it doesn't waver.

What has most shaped how you think about the future of medicine lately?

I tend to watch adjacent industries very closely. I've learned a lot by following the fast evolution of AI in legal tech. It's not 100% applicable to life sciences, but it's regulated, high-risk, plays with language, and the meaning of words matters immensely. Legal tech is probably about two years ahead of us in life sciences, and it's been fascinating to watch what's worked and what hasn't, how it's being deployed, what the systems architecture looks like.

I often follow adjacent industries more closely than my own. Right now, legal tech has my fascination.

Additional Reads

• San Diego startup Faro Health snags $20M for software to improve clinical drug trials

• Breaking Barriers in Clinical Trial Design: Insights from Faro Health CEO Scott Chetham

• How Faro Health is Reinventing Clinical Trials with AI

Quick Poll

Rate This Edition