I'm joined this week by Meri Beckwith, who founded Lindus Health after being a clinical trial patient—because he was astounded at how badly clinical drug trials were being run by the world's leading companies.

We talk about why clinical trials companies are twenty years behind in adopting new technologies, how a flooded supply closet can cause a billion-dollar clinical trial to fail, and his speculation about the incentives that—so far—have stopped the world from getting better.

Quotes

---

Timestamps

• 1:51 - Eroom's Law
• 3:12 - The gap between new science and new medicine
• 4:51 - Clinical trials: the patient experience
• 8:19 - The lifecycle of a biotech startup
• 10:22 - Introducing: Clinical Research Organizations
• 13:19 - 10% of trials fail from clerical errors
• 17:27 - Introducing: Lindus Health
• 19:03 - Recruiting more patients won't save badly-planned trials
• 23:36 - Meri's skepticism of international trials
• 27:04 - Biotech culture vs tech culture
• 32:54 - What'll it take to get biotechs to act like tech companies?
• 35:13 - Adaptive clinical trials
• 39:10 - Sidebar on the FDA
• 43:06 - Decentralized trials
• 46:13 - Who makes trials expensive?
• 48:43 - (Not) picking up the billion-dollar bills
• 53:22 - What does "better" look like?
• 56:08 - "...a lot of alpha if you're willing to disregard all of this b——..."
• 58:42 - What's changing with AI?
• 1:00:57 - What's going on with China?
• 1:03:10 - What's ahead for Lindus?

Transcript

Introduction (0:57)

Ross:
Welcome to Development & Research, where we talk to people doing things differently in clinical trials. I'm here with Meri Beckwith—a former clinical trials patient who, hm, enjoyed the experience so much he decided to work in the industry.

So welcome, Meri. Can you tell us about that?

Meri Beckwith:
(laughs) Thanks for having me. So, for my background, I was a venture capitalist. I spent the early part of my career investing in medtech, biotech, and software companies.

Ross:
Does that mean investing in drugs, or “everything but drugs”?

Meri Beckwith:
Drugs, devices, and digital health—so like a pretty broad spectrum—but I always had an interest in clinical trials, because it was very clear to me that it’s a huge market and it’s fundamental to how all of the companies we're investing in actually get to market and realize value.

Eroom's Law (1:51)

Meri Beckwith:
I started in 2013, 2014—around the time the concept of Eroom's Law was brand new. Some of our listeners may have heard of it, but I'll just describe it briefly.

So, over the last 60 or 70 years you can see a doubling in the cost to develop new medicines, new drugs, new health interventions [every 9 years]. [Ross: "Eroom" is "Moore" backwards. The term was coined by Jack Scannell and co-authors in a 2012 paper in Nature Reviews Drug Discovery.]

Ross:
Doubling the cost—that’s the bad direction, right?

Meri Beckwith:
Yeah, so it's getting exponentially more expensive, or we're pumping exponentially more money into the system and getting flat output. And that’s to the tune of, the cost to develop a drug on average doubles every nine years—which is pretty scary, right? I'm sure a lot of listeners will have heard the stats—it takes 10 years and costs $2 billion to develop a drug—

Ross:
—and then you need to make $3 billion back for the time, failure rate, and all that.

Meri Beckwith:
Exactly. These are scary numbers, and if it continues—exponentials being what they are—and if this continues we just won’t have progress in medicine potentially. So, I had this intuition that something is deeply broken about how we developed new drugs. And seventy or eighty percent of the cost of drug development is clinical trials, so you can say it’s all a problem of clinical trials.

The gap between new science and new medicine (3:12)

Ross:
The slowdown might be surprising because we see huge leaps forward in terms of what we read about coming from academic labs: we see entirely new modalities, new approaches. Where does the disconnect come from?

It’s like, I read about a lab at Stanford a cure for seven types of colorectal cancer in mice—and then it goes into a kind of black hole of silence forever—and at some point years later, it’s in the hospital, and it costs a quarter-million dollars, but it’s actually able to help people. What is happening in that gap?

Meri Beckwith:
It’s always in mice or rats first [when you read about it], not in humans—but bluntly, medicines are basically worthless until you show safety and efficacy in humans. Ultimately even if something works in mice or a model organism, we just don’t know how it will behave in a human. So that “black hole” is the clinical trial process.

Canonically, for a new drug, it’s Phase I, Phase II, and Phase III—typically today they're done in sequence, though they could be done in parallel—they aren't—and that process takes, on average, 10 years. The total cost per approved drug, factoring in the cost of failures, is about $10 billion. There’s a few different ways of calculating that, but it’s of that order of magnitude. [Ross: Indeed, I’ve seen a few different ways of calculating the cost per approved drug, and the $10 billion that Meri originally said here is at the top of the range I’ve heard. Most sources tend to estimate something in the $1 billion to $3 billion range.]

Clinical trials: the patient experience (4:51)

Ross:
Let’s start from the inside of this process and work our way out. You were a participant in a number of clinical trials—can you talk about your experience with those and then situate that within the arc of those drugs being developed?

Meri Beckwith:
So, I had this intuition that there was something very broken in the clinical trials process. You can see that in Eroom’s Law and I could see it in the companies we invested in, where there were just constant issues: delays, nothing seemed to go to plan…

And I would ask everyone—in my VC days—”why do you think this is?” And no one really gave me a satisfying answer. A lot of the answers were, “Doing things in the real world is hard,” or, “We've picked all the low-hanging fruit so we're going after undruggable targets.”

But I thought, come on; Janssen—the world’s first pharma company—was basically a guy in his basement, zapping random chemicals into mice. We have exponentially more understanding of biology today; it has to translate into results.

Meri Beckwith:
So, that was all in the back of my head, and then during Covid, I was bored—like a lot of people—and I wanted to do what I could to help out, and so I volunteered for a bunch of the vaccine clinical trials. I was really shocked by how bad the experience was.

First, it was very hard to sign up, even though this was critical—time was of the essence—and they’re basically taking all comers. I remember having to download Internet Explorer onto my laptop because the website the [Contract Research Organization] had built didn’t have an SSL certificate—it was an old Wordpress template—and so Chrome blocked you from accessing it. And so I was like, “Wow. They can’t even build a website. This is insane.”

This was for the Novavax Phase III—loads of funding, Parexel, one of the world’s biggest CROs was running it—and I calculated the value of each participant to the company running the trial might be to the order of $30,000 to $40,000. Imagine if you’re any company and your lifetime value of a customer is $30,000—you would spare no expense to make sure that the signup flow and that whole customer journey is perfect.

And then during the trial, there were all sorts of problems. We were using the market-leading app for recording data, and I remember they gave us a piece of paper saying, “The app will ask you about your flu symptoms; just pretend it says Covid.” And I was like “What? You couldn’t have changed the—”

Ross:
—couldn’t have shipped an updated app for the trial?

Meri Beckwith:
Right. And the app didn’t work on many models of phones, so they were giving participants iPhones. A lot of the older fellow participants I spoke to didn’t like using an iPhone because they’d never used one—and so they just gave up. It was just very basic stuff.

You’ve got cutting-edge science on one hand, and, frankly, just a sort of emergent incompetence in the real-world operations. I thought, “These problems should be so easy to solve,” and that was the impetus for founding Lindus Health, because it was just so obvious you could do a better job if you joined up the whole process.

The life of a biotech startup (8:19)

Ross:
You raise an interesting point here: these are brilliant scientific advances, and then when they reach this critical stage where they're interacting with patients, where the importance of any particular patient’s experience through the process is of enormous importance to the company's commercial value—and to humanity—then the ball is just being dropped all over the place.

But these are not the same people who are working in the lab and who are developing the apps and sending it with a piece of paper to the patients. Can you talk us through the different actors involved here? What’s difference between a sponsor, a trial site, a CRO, and a pharmaceutical distributor?

Meri Beckwith:
Sure. "Sponsor" is the industry term of art for the pharma or drug company who has developed the product that's being tested in the clinical trial.

Ross:
I think of the normal story as being, a grad student leaves their professor’s lab at Stanford and starts a company of six scientists.

Meri Beckwith:
Yeah, that's a pretty typical biotech story.

So you have a lab, and they have a particular hypothesis about a piece of biology that is implicated in disease. A PhD student might leave, set up a lab with a professor’s, hire six people, raise some VC funding, take it through preclinical and then clinical trials, Phase I, maybe Phase II... [Ross: "Phase I" typically means initial safety studies, usually in healthy-normal patients. "Phase II" is usually trying to demonstrate efficacy in patients with the disease or condition, but not in a trial large enough to get final approval.]

The typical story is that that will take 15 or 20 years, and by the time they reach Phase II—halfway through the actual clinical trial process—they get bought out by a big pharma company. Then that pharma company would take it through Phase III and then distribution and sale.

Ross:
And those are household pharma names?

Meri Beckwith:
Yeah, any of the big pharma companies you might think of.

Ross:
So that's the sponsor—depending on the stage, it's either the six-person biotech or the big pharma company with $6 billion.

Meri Beckwith:
Right, so the sponsor is whoever actually owns the rights to the actual drug being tested.

Introducing: Clinical Research Organizations (10:22)

Meri Beckwith:
But then, who actually conducts the trial? The majority of the time, some or all of the actual trial conduct will be outsourced to a third-party organization or a collection of organizations. Most clinical trials are run by these "Contract Research Organizations", or CROs for short. CRO services is an industry with about $100 billion of annual revenue, and it effectively is the clinical trials industry.

A sponsor—the biotech startup or pharma company—will usually just give their CRO a budget and maybe a protocol, and this CRO will just have to go and execute the entire trial. They will often then subcontract a lot of that to other third parties, like clinical trial sites, third party vendors for technology or loads of other kind of random services. But the CRO typically has the majority of the budget, and overall operational control—or they should.

Ross:
So it’s like, I'm remodeling my kitchen, I get a general contractor, they're going to do a bunch of stuff and make a bunch of decisions—maybe I'm going to get involved—and they're also going to hire all these subcontractors?

Meri Beckwith:
Yeah. That's a great analogy.

Ross:
—and the kitchen that I get depends on all of the choices that were made all the way down the chain of contractors, where maybe I've got some control.

Meri Beckwith:
Exactly. And, like you mentioned at the start—a lot of the people who come up with these treatments are deeply embedded in the science and the potential medical impact. But there's this disconnect between spending half a billion dollars coming up with the absolute optimal drug for a target—and then having all that effort undone by poor execution at the clinical trial stage.

Frankly, clinical trial operations are not sexy. We go to conferences all the time, academic conferences about the therapy areas that we’re active in—you’ll have someone presenting about this new therapy class and everyone is on the edge of their seats. Someone—maybe me—starts talking about clinical trial operations, and everyone's faces just go completely blank, it couldn’t be more boring.

Ross:
And that’s if you're in the room, Meri. I hear these presentations and they end with, “...get our therapy into the clinic.” The end. Clap hands.

Meri Beckwith:
Right, because a lot of people just ignore what happens in the clinic, or it's seen as an assembly line and there's nothing you can do about what happens in the clinic. I think that's where this mindset comes from.

Ross:
An assembly line or maybe a pachinko machine, right?

Meri Beckwith:
Yeah, exactly.

10% of trials fail from clerical errors (13:19)

Meri Beckwith:
And that’s where we have these problems. There's a great paper—that I plastered all over our Seed and Series A decks—that looked at the reasons for failure of drugs in Phase II and III. About 10% of the trials that don't meet their primary endpoint weren’t because there wasn't a strong signal of efficacy—but because of an abundance of clerical or operational errors in how the clinical trial was conducted. This is a huge amount of value being blown away.

Ross:
You mean, the trial is getting there in statistical terms, but then it becomes invalidated somehow?

Meri Beckwith:
Yeah; the data is getting there, but then either the FDA makes them discard data because it’s suspect—maybe there’s evidence of fraud, or no proper audit trail, or inconsistencies, any of which cause you to have to discard data.

Or there are simple operational issues that cause you to have to discard data from a site. One real example is, these documents were stored on paper and the warehouse flooded—so the [Good Clinical Practices] certificates signed by the clinical staff were lost, and so we now don’t trust any of this data. It’s really absurd stuff. It’s not sexy, but there’s just so much value that we’re throwing away because of really dumb, really easily solvable operational issues.

Ross:
Is part of the issue that these things get run with the statistically “right” number of patients, and then when we lose 10% of the patient data, that's just unrecoverable in terms of reaching a statistical significance threshold?

Meri Beckwith:
Yeah, for sure. I mean, you could brute force it and just have more patients in the trial, but there are other constraints—money, but also recruitment speed. The more patients you have, the slower it will be to recruit for a trial—and time is money, so it factors into the equation.

Our hypothesis [at Lindus] is that the point of highest leverage is just doing a better job on the unsexy clinical operations of running the study.

Ross:
Let's get back to the higher level. We can have fixable clerical errors and paperwork that are causing us to lose data. We should simply not do that, or make that less likely.

But what are some other points where you see the standard across the industry falling short?

Meri Beckwith:
I think the biggest area is just clinical trial data in general—how it is captured, stored, and analyzed. People like to say that generally tech adoption in healthcare is 10 years behind any other industry, but I'd posit that clinical trials are another 10 years behind healthcare.

For example—only now is the industry starting to coalesce around certain accepted data standards. The FDA, in partnership with an organization called CDISC, has done a great job beginning to mandate a data format called SDTM, which is a big step forward for the industry. We’ve been involved with that, and have some press coming out about it.

In a typical trial, you'll have data captured digitally—if you're lucky—but it'll be spread across three or more different systems, all of which store data in a slightly different way. And so, to really have an understanding of what has happened in your trial, it's a separate months- or years-long project to get the data in a consistent format, to even begin to analyze it—let alone to look for insights that could help you conduct the trial better.

I’m not a clinician or a data scientist by background, but one of the most shocking things to me—after building Lindus and peeling back layers of the onion—was seeing how incredibly loose and diverse the standards around clinical data capture still are.

Introducing: Lindus Health (17:27)

Ross:
So we can talk about other sorts of facets, failings, the successes of the industry, but what does a better world look like—on data standards, technology capture—when Lindus puts together a trial?

Meri Beckwith:
Sure. So, I should probably say what we do at Lindus: we run radically faster end-to-end clinical trials. We are directly competing with—and ‘disrupting’ if you like—the Contract Research Organization market.

Our product is the entire clinical trial—

Ross:
—Phase I, II, and II?

Meri Beckwith:
Well, we’ve been deliberate about not trying to be everything to everyone—so at the moment we don’t do Phase I or certain therapy areas. A lot of our work is in Phase II, III, and IV. [Ross: "Phase III" trials are the final trials that are run on a drug before the sponsor applies for approval. “Phase IV” is an industry term for research conducted after a drug is already on the market.]

We're not a software company, but you can think of it like an “outcome as a service”. Under the hood, how we're able to do trials better and more efficiently is a combination of a different business model and a software platform we built in-house that helps our team, our sites, clinicians, and patients conduct a faster, more efficient, higher quality trial at every step.

Ross:
So you’ll ship an app for patients that actually says “Covid”, you’ll ship an app for signups, that actually has an SSL certificate—

Meri Beckwith:
(laughs) It’s a low bar, yeah. There’s a patient-facing component, but also a clinician-facing component, an internal platform for us, and a sponsor-facing part for oversight.

Recruiting patients (19:03)

Ross:
And are you on the recruitment side as well?

Meri Beckwith:
Yep. We do active patient recruitment for most of the trials we run—and recruitment is a really important, often overlooked, aspect of trials.

If you talk to any executive in pharma and ask, "What's your biggest pain point for clinical trials?" they'll probably say, "Patient recruitment." And then, if you ask them, "What can I give you to solve this pain point?" they'll probably say, "More clinical trial sites."

Look, your average clinical trial might have 500 patients; you typically see 300 clinical sites. [Ross: Wow.] These are 300 physical locations, each of which often has a cupboard of physical documents that need to be created and signed—that might get flooded, per my example earlier—

Ross:
—for two patients per site?

Meri Beckwith:
For their two patients, yeah. It's violently inefficient, but that is how most trials are set up.

Ross:
And that means that the site administrators are doing it for the first time—for all 500 patients?

Meri Beckwith:
Right. That’s key, and one of the most absurd parts of this setup is, not only is it violently inefficient cost-wise, it makes the data manifestly worse. Firstly, it’s just hard to keep track of all that data, but secondly—you have a team of nurses and investigators at each site who interpret the protocol slightly differently, adding massive variance to what should be a clean experiment.

Ross:
It’s wild to think that this can even be considered a 'controlled' trial. Like, I have these treatment patients and these placebo patients, and every one of them was at a different hospital. I feel like the treatment and placebo should have been at the same hospital and gotten the same—

Meri Beckwith:
—yeah, and that is a really absurd state of affairs.

So [at Lindus] we turned the problem on its head and said, "You don't need sites, you need patients. Let's find the patients, then work with the minimum number of sites required, optimizing for quality, to actually deliver the trial."

The kicker is, the way things are done doesn’t benefit anyone! Well, the only people who benefit are the CROs. Sites—who are trying to run an operational business—they hate having to juggle 10 or 20 different clinical trials on the go at once. They'd rather have two or three big trials that your team can focus on.

So what we give people is, rather than 250 sites for 500 patients, we might have 20 or 10 sites and ensure that we find as many patients as possible for those 10 sites. The reason you have 250 sites in the first place is that you typically were relying on the sites to refer patients, but the sites might not be the best—or they're juggling their nine other clinical trials—and you don't find the patients you need. So that’s why you have this ridiculous setup to begin with.

Ross:
Whereas you're finding the patients through outreach and directing them to a smaller number of sites? Or are you asking them, "Where do you get your healthcare?" and then picking the sites based on that?

Meri Beckwith:
A bit of both. So we have the ability to screen about 40 million electronic health records across Western Europe and the US, so we can see where the patients are already, and pick relevant sites based on that. Or we can directly outreach to patients and then ask them. It's surprisingly easy.

You'll often hear pharma execs tell you, "There's so much competition for patients in our trials. We have two other rival companies, both launching a trial."—and that is just an insane failure of imagination. Each of your trials needs to enroll 500 patients. There are more than 1,500 patients globally who have this condition—who fit your eligibility criteria. It's just that you're all going after the same sites!

So you have a site with three trials signed up, and one patient comes in the door—who do I give this patient to? Just—go outside those sites! Or find patients independently of those sites!

Ross:
How much of this is the mindset of, like, picking the most prestigious hospital first and then—

Meri Beckwith:
—yeah, that is a big part of it. And, well, maybe for the average condition, there are four or five top key opinion leaders and you want one of them on board, and there you do get into competitive dynamics there.

But hey, you still don’t need to have all the patients come through that key opinion leader’s site! Sign them up, get a few patients from them, and then have your workhorse sites where you can find patients, bring them in, and get them enrolled in the trial. I'm just really shocked by how uncreative the industry can be when it comes to these things.

Skepticism of international trials (23:36)

Ross:
You mentioned “more than 1,500 patients for—whatever it is—globally”. That requires looking outside the country you thought you were going to start in, right?

Meri Beckwith:
So there was a wave, starting in the 2000s, of trials looking outside the US—to Eastern Europe, APAC, India, and Africa. That trend has actually been receding recently—because, frankly, there were a bunch of cases of fraud in some of those trials.

It was driven by access to more patients, more treatment-naive patients, and new sites where there wasn't as much competition for clinical trial patients. But frankly, the US is the #1 market in healthcare—and for the FDA, US data will always be paramount and non-US data is often a compromise.

That’s a very blanket statement, and obviously every condition is different and the way that each condition is treated in different healthcare systems is a little bit different. But I have to say—oh my God, trying to have 200 sites in 30 countries is a nightmare. Every country has slightly different regulations around clinical trials, different conventions for ethics committees—and so, you save yourself a ton of time and a ton of money by keeping the list of countries small.

Our whole thing at Lindus is that, if you improve that patient signup flow and broaden access outside of fishing in the same pool of sites, maybe you maybe don't need to go to other countries. As long as you're enrolling diverse patients from countries that have diverse populations, it all works out.

Ross:
Who are the sponsors who take Lindus up on this, who are sort of interested in this— Sorry, I want to ask that question the other way: who are the sponsors who aren't interested in what Lindus has on offer?

Meri Beckwith:
I mean, no one gets fired for choosing insert-big-CRO-here—we've been told that many times, verbatim, by potential customers. They’ll say, "Look, I know what you guys are offering is manifestly better, but I have a board member who will get upset if we don't choose this other big CRO"—and that board member is often a shareholder of said big CRO…

Ross:
(laughs) Or the investor’s last successful company used that big CRO and now they think—

Meri Beckwith:
—yeah. We're new and doing things differently, people are understandably conservative about health, and clinical trials in particular. The trials that are in conditions or drugs with the medical danger—or the worst safety profile—are often the most hesitant to try new approaches, which makes total sense. And so, historically, a lot of our work was in non-drug trials—like digital health interventions, or simple medical device interventions, and diagnostics. That was a lot of—deliberately a lot of our early customers were in those areas, and we are super grateful to all of them.

We still do a lot of work in those areas as well as, now, drug therapeutic trials. But we still tend to shy away from oncology, immunotherapies, cell therapies and gene therapies, or conditions where patients are critically ill. As a startup, you always have to focus on the segments of the market that are going to be the best fit.

Biotech culture vs tech culture (27:04)

Meri Beckwith:
I think your question kind of gets at something really interesting about the biotech industry. So I imagine a lot of your listeners are a split between Tech People and Life Sciences People—or maybe Tech and they’re curious about Life Sciences—I'd always had a foot in both camps because of my career background, and I always just marvel at how radically different the vibe is.

Maybe people have heard of the biggest healthcare biotech conference, called JPM—J.P. Morgan—which tells you something anyway, right? Just the branding.

Ross:
Imagine if the biggest internet-tech conference were called “Bank of America”.

Meri Beckwith:
Right. And famously it happens in San Francisco in mid-January every year, and everyone talks about, “All of a sudden there are all these people in suits walking around San Francisco. What the hell is going on?” I think that encapsulates just how different the vibe is—everything is very formal.

Frankly, entrepreneurs or execs in biotech are a lot older. I've lost count of the number of times I've had people tell me, in biotech, “You're so young.” And I'm like, “I'm 32! I’m old!” I'm seeing some founder friends tonight, and I'm the oldest one by far.

And think it gets at something very interesting about how different building a company in biotech is from one in digital tech. Biotech takes 10 years—as we said—to get a trial from the start of Phase I to approval, but it’s often longer than that because the biotech will have been working on it for years before Phase I—

Ross:
—with a founder who's coming out of a postdoc position at a graduate lab, which is not what we think of as the modal founder of an internet-tech company. [Ross: I’m being oblique here, but what I mean is that a postdoc will be five to ten years older than a college dropout or graduate when they start their company.]

Meri Beckwith:
Exactly, yeah. And then, in biotech, founders are much less likely to be in control even by Series B.

The classic story is: the postdoc founder might get 30% of equity on spinout, the university takes 40%, and the rest goes to immediately raising a Seed or a Series A round. Biotech rounds you tend to dilute 50 to 75% per round—which is a far cry from 5 to 20% in tech! [Ross: meaning, a biotech’s new investors will own half to three-quarters of the company after putting a round of money in, by contrast to a small fraction of an internet-tech company.]

Ross:
Is that because of the size of the round or because of the valuations?

Meri Beckwith:
Both. So biotech rounds will tend to be bigger than a tech Series A, and the valuation will be lower so the dilution is much higher.

Ross:
Can you use numbers here? I think they'll be surprising to our listeners.

Meri Beckwith:
So a Series A will be raising $30 million to $50 million—and then the [pre-money] valuation will be, like, the same. So yeah, pretty scary. I mean, if you have tech founder listeners, they'll be terrified at the prospect of that—understandably!

And so you'll have the investors with the majority of the cap table, and maybe the academic founder has 5 or 10% pretty quickly, at least by the time the drug even gets to Phase I. And by that time, there's usually an “experienced CEO” brought in to take the product through Phase I, II, and III. And often that CEO is from a CRO or [big] pharma background. So it's not founder mode.

What I'm getting at is that the story is so different from the model of a successful tech company, where founders remain in control for longer and there's less dilution. I think one of the things I’ve found curious about biotech is—how many $10 billion-plus companies were built in biotech in the last decade? Probably three or four, right? And how many in tech? A hundred. You just have many fewer generational outcomes.

Ross:
It's a little bit unfair because of where you picked your point. I mean—how many half-billion-dollar exits have there been is a much larger number than three or four.

Meri Beckwith:
Yeah, yeah, but the distribution looks different, I guess.

Ross:
And those are not founder, CEO, legacy vision types. Those might be a success for all the investors and the founder was, I guess, involved.

Meri Beckwith:
Exactly, and I think a lot of that just stems from the clinical trial process. Because it's so waterfall, so uncertain, so slow, it just kills ambition.

Whether your company achieves a $500 million, $1 billion exit or not hinges on a Phase II readout, and everything builds up to that—

Ross:
—where there were 300 patients, and 300 nurses who were collecting their data, typing them into records, that is what your billion dollars is pivoting on.

Meri Beckwith:
Exactly. It's all really tenuous.

So no wonder people are very conservative—but it's this learned-helplessness mentality. If you flipped it on its head and said, “Why don't we have a portfolio approach? Why don't we do whatever we can to have five or 10 shots on goal for the same amount of time and money?” Frankly, drug development is a luck-based endeavor.

Ross:
Because of how little we understand about biology?

Meri Beckwith:
Exactly. Even with the best scientists in the world and the best models, the play of chance is huge. So I don't understand why the industry doesn't recognize that, given that the optimum strategy is a portfolio approach.

Ross:
There's something of a vicious cycle here. You were talking about how there are enormous headwinds structurally across the industry, how that militates against ambition, how founders losing control of their companies squashes the opportunity to have a firebrand iconoclast founder deciding to do things differently when everything is being run by professional investors.

That seems like a recipe for not changing the way that things are done—and because the environment is toxic to ambition, there's no ambition to make the environment anything other than it is.

Meri Beckwith:
Exactly. That’s kind of what I’m getting at. That, to me, explains why tech and biotech look and smell so different.

What'll it take to get biotechs to act like tech companies? (32:54)

Meri Beckwith:
I frankly think biotech, and life sciences, and humanity as a whole, would be much better served if it looked more like tech—if you had more ambitious big bets, if founders stayed in control longer, and companies took more risk.

Ross:
Let's get into that. We can say that, as a whole, the world would be so much better if this was different. But at the end of the day, someone needs to make some decision differently tomorrow than they were going to yesterday.

Where do you see the windows, the cracks in the wall, for that kind of change? Who needs to do something differently and is in a position to?

Meri Beckwith:
I'm so glad you asked! I genuinely believe clinical trials are the linchpin. As I described at the start, there's where so much of the value and cost of this process is concentrated. The way great products are built in tech is through rapid and constant iteration. You can't do that in biotech with today's clinical trial infrastructure.

It's completely insane that the drugs we all take are the results of basically three iterations: one for dose, safety, and maybe one or two for efficacy. It's completely nuts.

Ross:
—and the dosing wasn’t checked on efficacy, usually.

Meri Beckwith:
Yeah, the dosing is checked first, mostly optimized for safety with some pharmacokinetic model about how the dose would interact with the biology of interest. Then maybe one or two dose iterations in the phase two or three.

We think we can cut this Gordian knot by giving biotechs the ability to think and act more like tech companies—faster, more rapid iteration, doing it safely, ultimately allowing them to produce better products, unlocking more ambition and more capital. That is the mission of Lindus Health.

Ross:
And the professional investors who now own half the company after their Series A, I can't imagine that the majority of the industry that we just described is falling over itself to beat down your door.

Meri Beckwith:
We don't need the majority. Even in the last year, we've seen some really exciting developments on that front, where you have a couple of larger pharma customers who are now working with us, who are kind of doing exactly what we wanted them to do—which is saying, “Okay, we need to run this, effectively this Phase III, but could you, instead of running just a Phase III, what about we run a platform Phase III—”

Ross:
Can you say more about that?

Adaptive clinical trials (35:13)

Meri Beckwith:
In your typical Phase III, it's very waterfall. You have a pre-specified, I don't know, 1,000 patients, this exact design, and that's it, and you just run the experiment—

Ross:
—why do I have 1,000 patients?

Meri Beckwith:
You've run some biostats calculations that show that given what—frankly—little data you have about the effect size and the variance, statistically, you think about 1,000 patients is enough to show an effect if those assumptions are correct. (laughs) That's a big if.

Ross:
Great. So I write down “1,000”, and then I send that off to my CRO, and the CRO says, “Yep, 1,000.” And then they conduct the trial. And then—six months after our final patient is done—we manage to merge our different data platforms and look at the data, and we look at it and we see actually 3,000 patients were needed.

Meri Beckwith:
Yes, exactly—because this assumption was off, or maybe all the variance that all those sites introduced are messing with the data, or a million other things. So that's obviously not good.

Ross:
What's the alternative approach here? Can I just run the trial, check every day whether I've got a statistically significant result, and then close the thing the day that I do? Does that work?

Meri Beckwith:
No, that's also tricky because you essentially introduce bias if you're stopping the trial at an optimum point. Imagine over the course of time, through random chance, you happen to get a strong efficacy signal, and then it weakens over time just because of natural variation.

So you can't do what you described, but you basically need to—I'm going to get roasted in the chat here because I'm getting over my skis—but as long as you pre-specify what the statistical plan is and you stick to it, then in theory, you're still running a fair experiment.

Ross:
You give yourself an extra handicap such that when you stop with your conditional stopping rule... it all works.

Meri Beckwith:
Exactly, I think what we're doing with this pharma company is we're saying, “Let's run an adaptive trial where we have an indeterminate number of arms.” Each arm is actually a slightly different formulation of their compound.

Ross:
On dose or the mixture—?

Meri Beckwith:
On basically the delivery method of the compound, yeah, and the formulation. It's a combination of several compounds, with a different blend of active ingredients, varying the ratio.

But let's run one arm with blend A, and then another arm with blend B and C in parallel. Then we can adjust the protocol if we need to. It's basically iterating to the optimal combination of blend of active ingredients, and we might also set up new arms for slightly different definitions of the patient population. So iterating to the optimal blend of active product and target population.

This is really exciting because from a financial perspective, this will be just many more patients and frankly a bigger spend, but the outcome should be much better, as in a much more effective product and overall a much more successful drug.

Ross:
Quality of product, probability of success—

Meri Beckwith:
Both are much, much higher. And this is exactly the kind of mindset shift we want to help catalyze in the industry.

Sidebar on the FDA (39:10)

Ross:
Can you talk about what you expect for regulatory acceptance in this case, or generally across these approaches?

Meri Beckwith:
Yes. So, sidebar on the FDA: the FDA has some people who are really, really good at designing regulatory frameworks for how clinical trials should work. They have guidance and frameworks for the kind of "adaptive" trial that I've just described. They have published briefing documents that are very supportive of these approaches because they recognize the benefits. So that's a big positive.

Where the FDA process can be difficult is on the individual trial level—and I’m sure this will be familiar to many of your listeners. Say, for example, you go to the FDA for scientific advice on a draft protocol of a Phase II trial. They will often bring in a third-party scientific expert to dialogue with you about the protocol—and that external expert will then force many changes to the protocol that make it very hard and expensive to deliver. That is where the problems start, and where the FDA process essentially causes this dysfunction we’ve been talking about in the industry.

You'll often have the same kinds of people who roll their eyes at conferences when someone asks about operational concerns—and they're dictating design changes to a trial without really thinking through: "How do you actually recruit patients?", "How do you work with sites?". Maybe you've specified a patient population that doesn't exist because this just isn’t how these patients are treated in the real world…

Ross:
What are the kinds of changes might be demanded in this process?

Meri Beckwith:
I was speaking to a biotech exec this week—the FDA is forcing them to have additional endpoints that are very hard to collect. [Ross: An "endpoint" is a medical test or measurement that your trial is based on.] Their trial should be a pretty simple, straightforward Phase II, but they're having to do a [pharmacokinetic] analysis—which requires taking a bunch of blood draws from participants—even though that analysis was already done in Phase I.

The typical Phase-II approach would be recording adverse events and measuring overall efficacy—but now they're drawing blood frequently. That's really expensive—and what patient wants to come in once a week two weeks to have a bunch of blood drawn?

And their reviewers are also making the company do a bunch of extra endpoints that are hard to collect, like an MRI. Very few of their potential sites have MRI machines, and so all of a sudden, the complexity to deliver this trial has increased by orders of magnitude.

Ross:
And the constraints for doing anything better have been boxed in, because now we’re stuck with exactly these centers, exactly these patients—

Meri Beckwith:
—and even with these trials you can still do things much better, but you’re exactly correct. That’s an example of science over practicalities—and how this mindset leads to really bloated, unwieldy clinical trial designs with tons of endpoints.

Ross:
You said that you see that coming in the individual trial review case—whereas the frameworks that are put forward by FDA staff tend to be welcoming to these approaches?

Meri Beckwith:
Exactly, yeah. And so you have this weird duality where—well, it doesn't make sense to think of any big organization as a monolith. Whether the FDA is friend or foe in this, I guess they're both.

Ross:
—out of opposite sides of their brain. Like with any large organization, we shouldn't be thinking about it as one solitary actor.

Meri Beckwith:
Yeah.

Decentralized trials (43:06)

Meri Beckwith:
For another example, you can look at the FDA's guidance on things like risk-based monitoring [Ross: discussed elsewhere 1, 2], central monitoring, and decentralized trials. These are all techniques that leverage technology and allow you to run trials at significantly lower cost—

Ross:
—by changing how you collect data?

Meri Beckwith:
Yes, a "decentralized" trial would have some or all of the trial happen remotely, so a patient never visits a physical site. Fully-remote trials are only possible for a subset of trials, but we think a lot of trials could benefit [from some remote elements].

Like—in my vaccine Phase III, we had to come into the hospital once a week or once a month just to fill out some paper forms. And this was during COVID.

Ross:
Just for the papers?

Meri Beckwith:
No physical exam—just filling out paper forms with a pen. And I was like, “Why?”

I actually thought, at one point, maybe this is a stealth human challenge trial [Ross: meaning, a trial where vaccinated patients are exposed to an infectious source under very controlled conditions.]—I'm in this windowless room, there's no ventilation, and there's 10 of us filling out these paper forms...

I asked these investigators, “Hey, you don't want me here—why aren’t we just getting this to you online?” and they said, “Oh—I don't know. I mean, we were just told to give you these paper forms.”

Ross:
—and these are the site leads?

Meri Beckwith:
These are the doctors at the site, so they aren't the CRO or the sponsor. They just do what they're told. "Only following orders"... But a decentralized trial would have us doing that remotely, on an app or a web form.

I think the key thing here is that it's not all-or-nothing, and we think often the most efficient study design is you have your dosing, your medical exams in person, but you have follow-up visits remotely. Patients massively prefer it, because who wants to take a day out to go to a hospital for filling out forms? It's obviously a lot cheaper because you're not paying the hospital costs—

Ross:
Yeah, I think that's the thing that boggles my mind when I start to think about it.

It's not only were you there and you had to be a patient who was sufficiently committed to the cause to go there weekly to fill out the forms—you've then got the investigator and the research nurse and the hospital staff and the room in the hospital where you all had to be. Delivery of healthcare in the US and the UK is not cheap, and those things, I can't imagine, were cheap to the bottom line of the trial.

Meri Beckwith:
Yeah, exactly. So decentralizing is one technique you can use to remove these bottlenecks. The FDA is very supportive of it.

Pharma are very skeptical of it—for basically no good reason, just inertia.

Who makes trials expensive? (46:16)

Meri Beckwith:
I used to think it was just CROs not proposing these solutions or not enacting them because their business model means they make more money in the on-site scenario, because it's more forms to fill out, there's more forms to type up into a database. But—

Ross:
Sorry, their costs are higher, this is better for them because…

Meri Beckwith:
Yeah, a CRO's business model is teams of people filling out— it's all hourly billing, so the more manual work there is, the more paperwork there is, that's all gravy for them. The scenario of that trial with 200 sites, paper records that get flooded, that's all ideal for them.

Ross:
—and now everyone has to read a paper record. You’re filling it out in pen—

Meri Beckwith:
—and then they’re typing it back in, and then you're paying a separate team at the CRO to come and monitor it, so to look at the paper record if it hasn't been flooded, and (laughs) looks at the database, and literally cross-checks them and goes, “Okay, this is accurate.” This is—

Ross:
—and then bills you by the hour for that.

Meri Beckwith:
Yeah, exactly. And they do that every three months. “Oh, there's a mistake here. This participant's weight was recorded wrong. Well, they left the trial two months ago, so there’s nothing we can do.” Shrug, better throw it all out.

Ross:
At best, get another patient.

But if we'd said 1,000 patients at the start, and we discover a set of 50 that had this issue, maybe we're cooked.

Meri Beckwith:
Yeah, pretty much.

And so the obvious thing to do is not do that, and just look at the data in real time, and then you could intervene when you notice these issues. And that brings me on to—I mentioned "risk-based remote monitoring". This is another technique that the FDA has guidance for, and it’s just manifestly better. (laughs)

There are meta-analyses showing that Phase III trials run using risk-based remote monitoring produced the same or higher quality data than the in-person monitoring I just described and were vastly cheaper.

The FDA guidance document says, “We are very disappointed that the industry does not adopt these approaches,” which is crazy that you have the regulator admonishing the industry for not adopting technology and not doing things that are better economically. And I think that gets to the heart of something very strange going on with pharma incentives.

(Not) picking up the billion-dollar bills on the sidewalk (48:43)

Ross:
Coming from the investor/financial world, it seems borderline unbelievable that there would be billion-dollar bills lying on the sidewalk here—because if you can run a $2 billion trial and have it cost $1 billion instead, that was valuable to the tune of $1 billion.

Who is choosing not to take that? It seems crazy to believe that in this capitalist marketplace, everyone is making this billion-dollar mistake.

Meri Beckwith:
Yeah, true. I agree. (laughs)

You know, I think the answer is just that there is so much inertia. A pharma company is basically a collection of monopolies that then has a monopoly over distribution, right?

As a pharma company, you might have franchises of 20 drug areas. Maybe some of those you have competition, but realistically, you can just do whatever you want. And so, frankly—

Ross:
—think like airlines chopping up the country into their own routes—

Meri Beckwith:
—exactly, yeah. Or train companies with different catchment areas. So there's very little pressure on them to improve or be more efficient.

When scenarios do arise where there is competition, you see them act differently. COVID being a prime example. Oh, we had decentralized trials—but obviously by necessity. We had just much, much faster drug development during COVID because every company—

Ross:
—because there were 12 players in the game on vaccines.

Meri Beckwith:
Exactly, yeah. The GLP-1 rush has been similar where you have seen trials run more efficiently, faster—

Ross:
—in obesity, everyone is trying to bring a drug to market. We have three pharma majors and how many startups?

Meri Beckwith:
Yeah, and there’s different combinations of formulation, and specific therapy area, but it’s been similar.

Ross:
But you have been seeing innovation in the trial design on those?

Meri Beckwith:
A little bit? And maybe it’s hard to disaggregate—maybe it’s just the therapy area or the intervention lends itself well to it, But you do tend to see those trials tend to be run more efficiently, faster... But it's hard for me to say.

My current hypothesis for why you have these billion-dollar bills lying on the floor is that there are insane amounts of bureaucratic inertia within pharma that massively discourage anyone from trying anything new.

Ross:
Everyone's trying individually not to get fired—or everyone's trying to avoid making their company unfundable to the next round of investors.

Meri Beckwith:
Exactly, yeah. They're kind of the same thing.

The craziest thing is that I've spoken to board members of major pharma companies, and they kind of know this is happening. They've said to me, “Yeah, we know that what we're doing is far from optimal, but turning the oil tanker is really, really hard.”

You've got layers and layers of people whose sole job is to not fuck up. There's just structure throughout the whole organization, and there's no incentive to do something like run a whole trial more efficiently. I'm being deliberately cynical a little bit, but—

Ross:
—I want to push back and say it's possible that you're not being cynical enough. You're talking about the boards of major pharmas, and I think one of these $10 billion companies that we've seen be created in the last few decades is Moderna. The chair of Moderna goes on a panel and says, “It is to the advantage of the incumbents that trials cost so much and take so long”...

“We have as an industry accepted the notion that this is what it is. Frankly, our good friends in the pharmaceutical industry benefit from that... Going slow and having it [cost] a lot of money favors incumbents.”

— Noubar Afeyan, Moderna chairman (reported in STAT, May 7, 2024)

Ross:
And when these Phase III cost half a billion dollars and take however many years, those are run by the fifteen or so drug companies in the world that can afford a half-billion-dollar outlay on something that's only 60% likely to work, and have the investment timeline of six years to make it work—

If those cost $25 million and got done in three years instead, those companies wouldn't have a monopoly on absorbing these startups before they take it to market. And Moderna, I think, stands out as one of the few examples where a company could just blitz through the process fast enough without getting eaten—because of exactly the moment that they were in.

Doing things better (53:22)

Meri Beckwith:
There are other examples of it outside of Covid, as well. There's this example of The Medicines Company—they had a drug for a big common condition. They got quotes to run their Phase III from CROs, and they were quoted about a billion dollars. (laughs) It's just a crazy amount of money.

And so they basically ran the trial themselves in partnership with elements of the UK Health Service and ran it for, I can't remember, well under $100 million, and then got acquired by Novartis for $13.6 billion shortly after the readout. So that is the model; it is absolutely possible.

Ross:
A thing that you might be concerned about as a company who's looking at this menu of options and one of them costs a billion dollars and one of them costs $100 million is, am I going to be the uninvestable weirdo if I'm on the $100 million? If you get through the trial, is that enough?

Meri Beckwith:
I think that because drug development is so luck based, people become—frankly—very superstitious.

So basically put it this way: we've spoken to tons of people in pharma who are like, “Imagine you have two drugs. They have exactly the same phase two data. One was spun out of a lab with some prestigious [Key Opinion Leader] involved, and the other was less prestigious. People will always pay more for the drug with people with fancier names attached.”

Ross:
And if they're only on the market for one, it's possible the other one doesn't sell basically at all.

Meri Beckwith:
Yeah, exactly. And it just completely dies.

But in theory, if the data is exactly the same, the data's the data, right? If it was conducted in the same way, those assets should be worth the same.

Ross:
And if there were a different amount of costs sunk into that data, if one dataset cost 10 times as much to generate, that also shouldn't come through, because the data should be the data.

Meri Beckwith:
Exactly. You just have this weird superstition.

So a company that was doing things a little differently, maybe set up their trials in a different way or used different methodologies, people would pay less for that drug conceivably at exit because of this weird superstition about drug development. (laughs)

Ross:
If it was a remote trial with an adaptive endpoint, and the acquirer is evaluating it against a statistically equivalently strong trial that fixed 1,000 patients who all went to the number one hospital in the city—they're going to go for that one every time over the second one. Even on stronger data.

Meri Beckwith:
Exactly, yeah, I think so.

"...a lot of alpha if you're willing to disregard all of this b——..." (56:08)

Meri Beckwith:
There aren't many real examples of that scenario, but the flip side is there's a lot of alpha if you're willing to disregard all of this bullshit and just look at the data—

Ross:
—if you can take it through whatever phase it is that you need to take it through, right?

It's these Keynesian beauty contests where you care about how someone else is going to evaluate the thing—there might be this collective action problem about, well, it's not clear which individual actors can intervene to pick up that alpha. You need a coalition who is capable of bringing a drug to the point where the market can tell you that, yes, this is okay.

I tend to think that by the time that the FDA has approved your drug, people stop caring. So you need to bring it there. You at least don't have to climb the hill of commercialization—those drugs can be sold off. But it's possible you need to take these things through approval.

Meri Beckwith:
I think that's one of the ways you break this Gordian Knot, is you make it possible for one company to take something all the way through to approval. Whereas today, that's very, very difficult.

Ross:
I mean, when that's going to cost an outside envelope of $2 billion and it has—from the outset a 12% chance of succeeding—there's no individual company that can raise for that. But what does it have to cost?

Meri Beckwith:
(laughs)

Ross:
Like, for this [hypothetical] company that is going to raise and blitz through—I mean, we’re not talking about raising $2 billion here. How far can we get these costs down?

Meri Beckwith:
I think you can easily get an order of magnitude. As the Medicines Company example showed, you can easily get an order of magnitude lower cost.

Take that as a hypothesis. Let’s say you have one biotech doing things the traditional way—portfolio of whatever, one single lead asset. It’s going to cost, maybe if you wrap in some failures, $2 billion to get that lead asset through. For that $2 billion, you could literally have 10 times the number of assets you’re taking through, more shots on goal.

In a paradigm where this is largely a luck-based endeavor, that is just the optimal strategy. It’s about shots on goal, not over-investing in something that is ultimately luck-determined.

And so I am hopeful that the industry collectively realizes this over the next 30 years, and we get a sea change in what biotech looks like.

What's changing with AI? (58:42)

Meri Beckwith:
I think paradigms like AI for drug discovery, where the intelligence is increasingly commoditized, and you just have something that spits out hypotheses about targets and drugs that need to be tested—that feels like a pretty good catalyst for this, as well as the work we’re doing to make it possible to run these trials much cheaper and much faster.

Ross:
That seems like a complement to this bottleneck. If successful drug candidates are commoditized, we still have this process that takes a billion dollars for the drugs that work, and that isn't going away unless—unless this changes.

Meri Beckwith:
Exactly. Hence why we’re doing what we’re doing.

Ross:
It feels like, yeah, the AI could just give us all the drugs, and they still wouldn't be available to healthcare providers until we make it through that 10 year process—unless we do something better.

Meri Beckwith:
Exactly.

Demis Hassabis, when he accepted his Nobel Prize, said, “AI will cure all disease in ten years,” and it's like, nope. (laughs) Sadly, not unless we do something about the clinical trial process.

Ross:
So, it sounds like you're hopeful about the way, the direction that the world is moving. Is that mainly through the efforts of Lindus, or do you see a broader sea change across the world? Is there a broader appetite for this thing in this moment?

Meri Beckwith:
I think a bit of both. I'm stoked to get to work on this directly with Lindus—and then there is just a growing sense that the current model isn't working.

I don't think I've met a single person in the four years of doing this, thousands of conversations, who was like, “No, I'm pretty happy with my CRO, happy with the way things work.”

Ross:
(laughs) Well, they don't talk to you. The ones who are happy with their CRO don't come to Lindus.

Meri Beckwith:
Fair, yeah, maybe. (laughs)

But as I said, I think AI can be a huge catalyst. The commoditization of the intelligence part shines the light more on the operational bottlenecks, which I think are key to unlocking, frankly, just a much better century of progress in life sciences than the one we've had, as great as it's been.

What's going on with China? (1:00:57)

Ross:
One thing that is fascinating to me about looking forward into the future—that we've seen already over the horizon—is the rise of China in the space of drug discovery and early-stage drug development. What does that look like from where you sit?

Meri Beckwith:
Yeah. From what I can see, China is succeeding at exactly this, with essentially looser regulations, which does translate to lower costs, faster timelines—

Ross:
—looser regulations on who is allowed to well what? Or—

Meri Beckwith:
All of the above. I'm not sure about the regulations around drug sales, but certainly around participating in clinical trials, how you set up and run a trial...

Ross:
There isn't this external review process [like with the FDA]?

Meri Beckwith:
I actually don't know the way ethics committees work, and they can be a blocker in the West, but I don't know how they work in China.

Ross:
I've seen some of the registrations about trial start dates and end dates, and—it's not unheard of to hear a Chinese company which initiated a Phase I, and then a year later initiated a Phase II, which is unheard of in the conventional paradigm in the West.

Meri Beckwith:
It's definitely faster in China. Labor costs play a role—

Ross:
—but there’s also a regulatory side, and there’s a cultural difference in the industry?

Meri Beckwith:
I suspect so, and I think that because a lot of their drugs are kind of repurposed or slightly tweaked versions of established compounds, you naturally shift the focus to operational excellence over deep scientific understanding.

That mindset shift plays a role—back to the start of our conversation—describing the reception you get at a conference when you start talking about site strategy and patient enrollment...

What's ahead for Lindus? (1:03:10)

Ross:
One final question: if this is the time that I hear about Lindus Health—like I'm reading about a cure for colorectal cancer in rats, and then it's going to go into a black hole—then at some point in the future, I will or won't see a glorious future of new medicines. What's ahead for Lindus?

Meri Beckwith:
More of the same, really. We want to continue showing that we can drive step changes in speed and efficiency across an increasingly broad portfolio of trials.

Today, we're still fairly focused on certain therapy areas, particularly metabolic, cardio-metabolic diagnostics, dermatology, respiratory, infectious disease, but showing we can broaden that—as well as more of these really interesting collaborations where we are giving our customers a new paradigm of how to run research—more adaptive trials, looking at more variables in real time, helping them iterate to an optimal drug product.

Ultimately, our goal is to become the default choice for anyone looking to run a clinical study, because why wouldn't you? For all the reasons we've been poking at in this conversation. It will take time, but it's worth doing, and we'll get there.

Ross:
Meri Beckwith, thank you very much.

Meri Beckwith:
Thanks a lot for having me.

---