Data Hormesis: When More Data Makes Things Worse

More data should mean less risk. In nonclinical development, that's not always true. In this episode, we explore the concept of data hormesis — the inflection point where accumulating more data stops reducing uncertainty and starts creating it. If your team is running another study because the last one didn't give you the answer you needed, this episode is for you.

Key takeaways:

• Data hormesis is the point where more information stops helping and starts obscuring the path forward — just like a drug that's beneficial at low doses and toxic at high ones
• Early in development, data reduce uncertainty. Beyond a certain point, signals compete for attention rather than converging toward resolution
• When decisions aren't defined early, data accumulation quietly becomes a substitute for judgment rather than a tool to support it
• The patterns are recognizable: equivocal findings trigger rework instead of interpretation, borderline results lead to more studies without clarity on what would actually change
• Programs that move efficiently to IND aren't the ones with the most data — they're the ones that decided early which risks are acceptable and which questions are worth answering now
• Data don't create strategy. They make strategy visible.

Links:

• Data Is Not Strategy on Amazon: https://a.co/d/02xUsV6K
• Work with Dessi: www.toxistrategy.com

The Nonclinical is hosted by Dessi McEntee, MS, DABT — board-certified toxicologist and Fractional Head of Toxicology. Subscribe to the newsletter on LinkedIn, take the course at nonclinical.academy, or work with Dessi at toxistrategy.com.

Transcript

SPEAKER_00 0:00

Welcome to the deep dive. If you were listening to this, you are probably a biotech leader, maybe a founder, managing early stage drug development programs, and you know, working tirelessly to get your molecule to that critical IND stage.

SPEAKER_01 0:15

Yeah, that I and D submission is just it's everything. It's the whole focus.

SPEAKER_00 0:19

It really is. And today we're going to explore a really counterintuitive trap. Like this is something that derails perfectly viable non-clinical programs all the time.

SPEAKER_01 0:29

Right. Because we're going to dismantle this really widely accepted assumption. The idea that accumulating more data always equals, you know, reducing more risk.

SPEAKER_00 0:37

Exactly. And our source material for this deep dive is just fantastic. It's an article and newsletter excerpt called Data Hormesis When Data Becomes Toxic. And it's by Desi McInty.

SPEAKER_01 0:47

Oh, her work is incredible.

SPEAKER_00 0:49

It is. And for those who don't know, she has over 15 years of experience. She's a board certified toxicologist, a biotech board member, the fractional head of toxic imaging. And uh she's also the author of a book called Data Is Not Strategy. So she really knows her stuff.

SPEAKER_01 1:03

She definitely does. She builds bulletproof toxicology programs for a living. So when she points out a flaw in the industry, I mean you listen.

SPEAKER_00 1:10

You absolutely have to. And the core setup here, the thing she teases out, is that there's this invisible inflection point. Basically, a threshold where gathering more evidence actually starts to actively work against an early stage program.

SPEAKER_01 1:25

Which is so strange to think about, right? Because we're taught that data is always good. But to really understand how data can like sabotage a multi-million dollar program, we first have to understand a fundamental concept from toxicology.

SPEAKER_00 1:38

Right, this concept of hormesis.

SPEAKER_01 1:40

Exactly. Hormesis. It's a dose response relationship. Basically, an exposure to a substance is beneficial at low doses, but once it pushes past a certain threshold, the curve just pivots. Yeah, the curve pivots entirely. And that exact same substance becomes harmful. MATE brilliantly maps this dynamic to data collection in drug development.

SPEAKER_00 2:00

Wait, I love this. It's it's like watering a houseplant, right?

SPEAKER_01 2:02

Well, that's a great way to put it.

SPEAKER_00 2:03

Yeah, because a little bit of water is totally essential for growth.

SPEAKER_01 2:06

Right.

SPEAKER_00 2:06

But if you get nervous and you just keep pouring water in to be safe, well, the roots rot. You kill the plant.

SPEAKER_01 2:12

Right.

SPEAKER_00 2:12

The water itself didn't change. It's still just water. The threshold changed.

SPEAKER_01 2:16

Aaron Ross Powell Exactly. And early in development, that beneficial low dose of data, it does exactly what it's supposed to do. It reduces uncertainty, you know, it defines your boundaries, it characterizes risk, it gives the team something grounded to talk about.

SPEAKER_00 2:30

Aaron Powell Like exposure margins and target effects.

SPEAKER_01 2:33

Yeah, exactly. Those crucial metrics. It proves basic visibility. But if we accept that there is a toxic threshold for data, just like the water for the plant, the immediate question for you as a biotech leader is: well, how do you know when your team has crossed it?

SPEAKER_00 2:49

Right. What are the symptoms?

SPEAKER_01 2:50

Yeah.

SPEAKER_00 2:50

Because it's not like an alarm goes off in the lab.

SPEAKER_01 2:52

Aaron Powell, I wish an alarm went off. But no, it's actually a shift in team behavior. So past the threshold, new data stops clarifying decisions and starts complicating interpretations.

SPEAKER_00 3:04

It's like static on a radio.

SPEAKER_01 3:05

Yes. The signals become impossible to separate from the noise. And instead of converging toward a resolution, findings just start competing for attention.

SPEAKER_00 3:14

And MATI notes that the questions the team asks actually change, right?

SPEAKER_01 3:18

They absolutely do. In the beneficial phase, teams ask, what decision is this data meant to support? But once they cross into data hormesis, they start asking, what else can we run?

SPEAKER_00 3:30

Ah, like let's do another endpoint, let's do another study.

SPEAKER_01 3:33

Exactly. Just to make the uncertainty go away, they're basically using activity to mask anxiety. And Mackint lists some real-world symptoms too, like equivocal findings suddenly trigger rework instead of interpretation.

SPEAKER_00 3:46

Wait, what does that mean in practice?

SPEAKER_01 3:48

So say you get a borderline result. It's not clearly good, not clearly bad. Instead of interpreting it, the team just orders three more studies.

SPEAKER_00 3:55

Oh wow. Without even knowing what a new result would change.

SPEAKER_01 3:57

Exactly. And the really dangerous part is when a CRO, a contract research organization, steps into that vacuum.

SPEAKER_00 4:04

Aaron Powell Right, because the sponsor, the biotech company, is basically avoiding making a decision.

SPEAKER_01 4:08

Yeah. Sponsor ownership gets totally unclear. The team is paralyzed, so the CRO steps in and just runs their standard, massive testing protocols.

SPEAKER_00 4:18

Aaron Powell Just to fill the void. But wait, let me push back on this for a second.

SPEAKER_01 4:22

Sure, go ahead.

SPEAKER_00 4:23

Because if you're in a field built entirely on scientific rigor, you know, peer review, documentation, how can gathering more evidence ever be a bad thing?

SPEAKER_01 4:34

It sounds counterintuitive, I know.

SPEAKER_00 4:37

Right. I mean, isn't doing another study just being a responsible scientist? You're leaving no stone unturned.

SPEAKER_01 4:42

That pushback is totally fair.

SPEAKER_00 4:44

Right.

SPEAKER_01 4:44

And it's exactly what leads us to the root cause of the problem, because data hormesis, it isn't driven by bad science. It's driven by a fundamental misunderstanding of what non-clinical development actually is.

SPEAKER_00 4:55

Yeah, unpack that for me.

SPEAKER_01 4:57

So McInty's core argument is that non-clinical development is often mistakenly treated as a process of gathering evidence.

SPEAKER_00 5:03

Like an academic exercise.

SPEAKER_01 5:05

Exactly, like a PhD thesis where you just gather everything. But in reality, it's a process of making choices under uncertainty.

SPEAKER_00 5:11

Oh, making choices under uncertainty, that's a huge distinction.

SPEAKER_01 5:14

It is. And when decisions aren't defined early, data accumulation quietly morphs into a substitute for leadership. It becomes a substitute for actual judgment.

SPEAKER_00 5:25

So it's like this illusion of productivity.

SPEAKER_01 5:32

Yeah, they're just pushing paper and running tests to avoid making a hard call. And the end state of this is just tragic.

SPEAKER_00 5:40

Tell me about it. What happens?

SPEAKER_01 5:41

Conflicting signals pile up, interpretations fragment across the team, and suddenly the regulatory narrative becomes completely impossible to defend.

SPEAKER_00 5:49

Aaron Powell Because it's not anchored to anything. It's like, well, it sounds like using a GPS without ever plugging in a destination.

SPEAKER_01 5:55

Oh, that's exactly what it's like.

SPEAKER_00 5:56

You're driving around, you know, meticulously writing down every street name and detour. You feel really productive, but you're fundamentally lost because you never decided where you were trying to go.

SPEAKER_01 6:06

That is the perfect analogy. And because you never defined the destination, the strategic spine, as she calls it, regulators look at this massive, messy pile of data and they push back. They issue clinical holds.

SPEAKER_00 6:19

On data that probably didn't even need to be collected in the first place.

SPEAKER_01 6:23

Exactly. And then the creeping belief sets in internally that the drug just isn't viable. The board gets nervous, investors get nervous.

SPEAKER_00 6:32

But the tragedy is the molecule might be fine, right?

SPEAKER_01 6:35

That's the most heartbreaking part. Mackinty points out that in many cases, the molecule's perfectly viable. The failure wasn't the science, the failure was the absence of non-clinical leadership when decisions mattered most.

SPEAKER_00 6:47

Wow. Okay, so now that we know why these viable molecules fail from data overload, we need to talk about the path forward.

SPEAKER_01 6:54

Right. How do we avoid the trap?

SPEAKER_00 6:56

Yeah. How do the most efficient IND-bound programs successfully navigate all this uncertainty?

SPEAKER_01 7:02

Well, she provides a really clear contrast. The programs that move efficiently toward an IND are explicitly not the ones with the most data.

SPEAKER_00 7:10

Aaron Powell Really? They don't have the biggest binders.

SPEAKER_01 7:12

Nope, not at all. What successful nonclinical leaders do instead is they decide explicitly and very early on three specific things.

SPEAKER_00 7:21

Okay, what are they?

SPEAKER_01 7:22

First, they decide which risks are acceptable. Trevor Burrus, Jr.

SPEAKER_00 7:25

Because you can't eliminate all risk.

SPEAKER_01 7:27

Exactly. It's biology. Second, they decide which questions are worth answering right now.

SPEAKER_00 7:32

Oh, timing is key.

SPEAKER_01 7:33

Yeah. And third, and this is the big one, they decide which outcomes would meaningfully change the path forward.

SPEAKER_00 7:40

Right. If a test result won't actually change your strategy, why are you running the test?

SPEAKER_01 7:44

Aaron Ross Powell Exactly. It's a waste of time and money, and it invites that toxic data hormesis, which really spotlights her foundational takeaway from her book, Data is Not Strategy. Which is she says that data is meant to serve its intended role. Her exact point is they don't create strategy, they make strategy visible.

SPEAKER_00 8:03

They make strategy visible. I love that. It's like the data is a flashlight, not the steering wheel.

SPEAKER_01 8:07

That's exactly it. You still have to drive the car. You have to make the decisions. And avoiding data hormesis isn't about doing less science or lowering rigor.

SPEAKER_00 8:16

Right. We're not talking about cutting corners.

SPEAKER_01 8:18

No, not at all. It's about having the restraint to not let the data become overgrown. It's about leveraging the data in its most beneficial state.

SPEAKER_00 8:26

So, practically speaking, if you're a biotech founder listening to this right now, and you know, your team is begging you to run just one more assay before submitting to the FDA, what is the exact question you should ask them? Like, how do you test if they are falling into data hormesis?

SPEAKER_01 8:42

You have to look them in the eye and ask, what specific decision will the result of this assay change? Wow.

SPEAKER_00 8:49

Simple but brutal.

SPEAKER_01 8:51

It is. If they say, well, it'll just give us a better picture or it'll make us feel more confident, you are in the toxic zone.

SPEAKER_00 8:58

You're overwatering the plant.

SPEAKER_01 9:00

Exactly. But if they can say, if it's X, we go to human trials, if it's Y, we till the program, then it's necessary data.

SPEAKER_00 9:07

That is such a clarifying test. So to wrap up this deep dive, you really have to remember that your primary job as a leader isn't to just blindly eliminate uncertainty through endless data collection.

SPEAKER_01 9:18

Right, because that's impossible.

SPEAKER_00 9:20

It is. Your job is to manage that uncertainty through clear anchored decision making, because clarity matters way more than another data set when you're approaching that IND phase.

SPEAKER_01 9:30

It really does. Clarity over volume every time. Exactly.

SPEAKER_00 9:33

So I want to leave you with a final lingering thought to mull over. Something that really builds on this idea of data acting as a substitute for judgment.

SPEAKER_01 9:42

Oh, this is a good one.

SPEAKER_00 9:43

Think about this. If you were to look at your current early stage development budget right now, how much of your capital is funding actual rigorous scientific inquiry? And how much is secretly funding security blankets for a team that is just too afraid to make a definitive choice?