Remember, you can't unscramble an egg — but you can check it's not cracked before you crack it. Same goes for nonclinical. Go crack some eggs.
👀 The Read
This Company Lost 15 Months to a Finding They Could've Explained in the IND
One clinical hold. One cataract finding. Two choices that cost over a year.
📋 THE RECEIPT
Drug: pretomanid (PA-824) — TB Alliance
Year: 2008
Held for: an unexplained tox finding
Damage: 15 months on full clinical hold
Pretomanid is approved today — part of the BPaL regimen for drug-resistant TB. But in 2008, the FDA stopped it cold. Not because of manufacturing or the clinical protocol, but because of the tox package.
And the finding that cost them the year? Turned out it could’ve been mitigated long before the hold. They didn't lose a year to a safety problem, they lost it to a problem they created and then couldn’t properly justify in their IND.
What FDA saw. Cataracts — irreversible in rats and observed in recovery monkeys. A lens finding in two species for an ocular drug is a major regulatory red flag, particularly when that finding was not teased out prior to submission.
What it took to undo. Repeat tox in both species. Cross-species metabolism. Ophthalmic data from earlier clinical work. An independent eye board.
What they found. Both cataracts findings were human-irrelevant. And the kicker? Every one of those studies to prove it could have been run before submission, or even been prevented, had the nonclinical program been designed sufficiently from the get-go.
They didn't fail to explain a finding. They engineered an alarming one — before the first animal was ever dosed.
The smoking gun:
❌ Using the rat for an ocular compound. The rat is a poor, unreliable ocular model — its eye doesn't translate to humans, and it throws lens findings that mean little clinically. Every toxicologist knows it. For any compound where the eye is in play: monkey and rabbit, or dog and rabbit if pharmacology allows. Never the rat if it can be avoided.
❌ No baseline eye exams. No pre-dose screen — so when cataracts appeared, no one could say whether the drug caused them or the animals came with them. Cataracts occur at background rates. A finding with nothing to compare it against isn't data. It's a question mark.
Two seemingly minor nonclinical program choices that turned out to be an almost catastrophic show-stopper. So, how did they get through? Some things that come to mind:
→ no ownership of the overall program
→ lack of communication between pharmacology and tox
→ implications of study design made on time and money, not necessity
These are the most common things I see in early-stage startups who encounter speedbumps trying to race to the clinic. I see founders fall into this trap all the time.
TB Alliance was able to dig themselves back out of the hole in this case, but I’m sure it was quite the scare for the leadership team at the time. If this is resonating, keep reading to see how to counter a situation like this ⏬
🔧 The Move
Here’s a quick program audit you can run on your own program to identify any potential FDA hold gaps (2 minutes):
List your tox studies and all the endpoints included on them. For each, ask two things —
Did I choose that species because it's right for the pharmacology and the program, or because it was available and cheap?
Did I build in a pretreatment baseline exam so findings can be correlated, compared and interpretated for treated and recovery groups?
If you can’t answer both of those questions cleanly, you need program review.
📍 Need someone to review it? Reply to this email, it’s what I do.
👀 Sneak Peek: Program Reviews
A full program review in 5 business days or less, here’s a preview of what you get:
📤 Need it? Reply to this email now.