A Series B company I talked to in February had 14 completed preclinical studies sitting in their regulatory folder. Fourteen. They'd spent north of two years running them. When they finally mapped everything against ICH M3(R2) to prepare their IND, they found out they were missing a repeat-dose toxicity study in a second species. A non-rodent. The kind of study that takes 4-6 months to run.
They also didn't have a complete genotoxicity battery. They'd done the Ames test and the in vitro micronucleus, but never ran the in vivo bone marrow micronucleus. Three tests in the standard battery, they had two. Nobody caught it until a consultant flagged it eight weeks before their planned IND filing date.
That's roughly $400K in direct costs and five months of delay. And here's the part that gets me — the requirements weren't ambiguous. ICH S2(R1) spells out the standard battery. ICH M3(R2) Table 1 tells you the exact duration for repeat-dose tox by species. It's all written down. The problem is that "written down" means scattered across six or seven ICH guidelines, each one 30-80 pages, cross-referencing each other in ways that assume you've already read the others.
So I put together a checklist. Not the kind where someone tells you to "ensure adequate nonclinical coverage" — I mean the actual steps, with the specific ICH sections that require each one.
Step 1: Map your studies to ICH M3(R2) requirements by phase
ICH M3(R2) is the master document. Everything flows from it. It defines which nonclinical studies you need before Phase I, which ones before Phase II, and so on.
The part most people get wrong is the repeat-dose toxicity duration. There's a table — literally Table 1 in the guideline — that maps planned clinical trial duration to minimum animal study duration:
| Your clinical trial duration | Minimum rodent study | Minimum non-rodent study |
|---|---|---|
| Up to 2 weeks | 2 weeks | 2 weeks |
| 2 weeks to 1 month | 1 month | 1 month |
| 1 to 3 months | 3 months | 3 months |
| 3 to 6 months | 6 months | 6 months |
| Over 6 months | 6 months | 9 months |
Two species. Rodent and non-rodent. If you're planning a 4-week Phase I and your repeat-dose tox study is only 2 weeks, that's a gap. If you only ran the study in rats and not in dogs (or monkeys, for biologics), that's a gap. Sounds obvious when you read it here. Gets missed constantly when you're coordinating with three CROs and a dozen study reports.
Before Phase I you also need primary pharmacodynamic studies, a full safety pharmacology core battery, pharmacokinetic characterization, local tolerance data if you're using a novel route, and at minimum an Ames test plus one chromosomal damage test. I'll get into each of these below.
Step 2: Verify GLP compliance for every pivotal study
21 CFR Part 58. For each nonclinical study in your IND, you need either a statement that it was conducted under Good Laboratory Practice or a written explanation of why it wasn't.
Here's what most teams don't fully internalize: FDA can discount your entire toxicology study — clean results and all — if it's not GLP-compliant and you haven't explained why. This isn't a technicality. It's one of the specific deficiency categories that triggers clinical holds.
Go through every study report. Check the GLP compliance statement. If a study wasn't GLP (and some legitimate studies aren't — early pharmacology studies, for example), write the justification now, not when the reviewer asks. Your nonclinical overview in Module 2.4 should address this clearly.
Step 3: Check species selection
This is where biologics companies trip up more than small molecule teams. ICH S6(R1) governs nonclinical evaluation of biotechnology-derived pharmaceuticals, and it has a specific requirement: you need to test in pharmacologically relevant species. For monoclonal antibodies that means species where the target antigen is expressed and the antibody cross-reacts. Sometimes that's only one species (cynomolgus monkey), and if so, you need to justify why one species is sufficient.
For small molecules, the standard is one rodent (usually rat) and one non-rodent (usually dog). But "usually" does a lot of work in that sentence. If your compound has species-specific metabolism — and plenty do — you might need to justify your species choice with comparative in vitro metabolism data.
Quick sanity check: for each study type in your package, can you explain why you used the species you used? If the answer is "because that's what the CRO suggested," you might want to dig deeper.
Step 4: Cross-reference the safety pharmacology core battery
ICH S7A requires three organ system assessments before first-in-human dosing. Three. Not optional, not "recommended." Required.
| System | What you measure | Typical species | Method |
|---|---|---|---|
| Cardiovascular | Blood pressure, heart rate, ECG including QT | Dog | Telemetry in conscious animals |
| Central nervous system | Motor activity, behavior, coordination, reflexes | Rat | Functional Observation Battery or Irwin screen |
| Respiratory | Respiratory rate, tidal volume, minute volume | Rat | Whole-body plethysmography |
The cardiovascular one catches the most attention because of hERG. The in vitro hERG channel assay (per ICH S7B) is essentially mandatory before Phase I — it evaluates cardiac pro-arrhythmic risk, and if you skip it or run it without GLP compliance, well. That's a clinical hold.
But I've seen teams so focused on hERG that they forget respiratory. Whole-body plethysmography in rats. It's not glamorous, it doesn't make for good investor slides, but if you don't have it, your safety pharmacology section has a hole in it.
One thing to note: you can integrate safety pharmacology endpoints into your general toxicology studies instead of running standalone studies. ICH S7A allows this. But — and this is a meaningful but — you need to make sure the endpoints are actually captured with the same rigor. "We measured heart rate during the tox study" is not the same as continuous telemetry monitoring with ECG analysis.
Step 5: Validate the genotoxicity battery
ICH S2(R1) gives you two options for the standard battery. Most companies go with Option 1:
- Bacterial reverse mutation assay (Ames test) — 5 bacterial strains minimum, tested with and without metabolic activation (S9 fraction)
- In vitro cytogenetic test — chromosome aberration, micronucleus, or mouse lymphoma assay
- In vivo test — typically rodent bone marrow micronucleus
All three. Not two. The company I mentioned at the beginning had two out of three and it cost them five months.
The Ames test has specific requirements that are easy to get partially wrong. You need five strains: S. typhimurium TA98, TA100, TA1535, TA1537 (or TA97), and either E. coli WP2 uvrA or S. typhimurium TA102. Miss one strain and the test is incomplete. Top dose for soluble, non-cytotoxic compounds: 5 mg/plate. These details matter.
Timing: you need at minimum the Ames test and one chromosomal damage test before Phase I. The full battery (all three tests) should be done before Phase II. If you're enrolling women of childbearing potential, don't wait — complete the battery early.
Check your genotoxicity data against these requirements. Specifically.
Step 6: Check exposure margins and NOAEL
This is the part that connects your animal data to your proposed human dose. For every repeat-dose tox study, you should have a No Observed Adverse Effect Level (NOAEL). Your starting clinical dose needs to be derived from the NOAEL using appropriate scaling — and FDA expects body surface area (BSA) scaling for interspecies dose conversion, not simple mg/kg.
"Selected starting dose too high or inadequately justified" is one of the specific deficiency categories the FDA uses when placing INDs on clinical hold. This isn't a gray area. If your starting dose math doesn't trace back to your NOAEL through proper allometric scaling, that's a problem the reviewer will find.
Also check whether your toxicokinetic data demonstrates adequate exposure. The animals in your tox studies need to have actually been exposed to the drug at relevant levels. Plasma concentration data, AUC values, Cmax — these should be in your study reports and should be cross-referenced in your pharmacokinetics summary.
Step 7: Verify reproductive toxicity timing
This one is phase-dependent, so the timing question is real. ICH M3(R2) doesn't require reproductive tox studies before Phase I for most drugs (unless the target population is women of childbearing potential). But you need embryo-fetal development studies in two species — typically rat and rabbit — before Phase II if WOCBP will be enrolled.
I've seen teams get caught by this because they planned a Phase I in male-only subjects, then wanted to expand to include women in Phase II, and didn't have the reproductive tox data ready. That expansion becomes a 6-month wait.
Plan backwards from your Phase II protocol. If there's any chance you'll enroll women of childbearing potential, start reproductive tox studies early.
Step 8: Review your CTD Module 2.4 and 2.6 structure
All of the above studies need to end up organized in the Common Technical Document format. Module 2.4 is your Nonclinical Overview — a written narrative that integrates all your nonclinical findings. Module 2.6 is the written and tabulated summaries for pharmacology, pharmacokinetics, and toxicology.
The data can be perfect and still get flagged if the narrative doesn't tie it together coherently. Does your nonclinical overview explain why your preclinical package supports the proposed clinical trial design? Does it address risks identified in your safety pharmacology and tox studies? Does it reference each study with its GLP status?
This is where a lot of the "back-and-forth" happens that regulatory affairs people describe. One RA lead I talked to put it this way: "If the regulatory requirements had been identified early, we would be able to set up the protocol to pass from the very beginning." Which, yeah. That's the whole point.
What gets missed most often
I keep a running list based on conversations with RA professionals and published FDA clinical hold analyses. The pattern is surprisingly consistent:
Repeat-dose tox duration mismatch. Study ran for 2 weeks, clinical trial plans for 4 weeks. Table 1 in M3(R2). Seems like the easiest thing to check, yet it shows up again and again.
Incomplete genotoxicity battery. Two tests instead of three. Or the Ames test missing a strain. Or tested without metabolic activation.
GLP gaps without justification. Study wasn't GLP and nobody wrote down why. The fix is a paragraph of text. The cost of not writing it is a clinical hold.
Safety pharmacology missing an organ system. Cardiovascular done, CNS done, respiratory... not done. Three out of three, not two.
Starting dose not traceable to NOAEL. The dose calculation exists but doesn't use BSA scaling, or doesn't reference the most sensitive species.
According to published FDA analyses, fewer than 10% of INDs get placed on clinical hold. But first-in-human trials from sponsors with limited regulatory experience are disproportionately represented. If you're a Series A biotech filing your first IND, you're in the higher-risk group by definition.
A faster way to do this
I built Regfo because I watched this process play out manually too many times. The checklist above works — but it takes a regulatory professional 200+ hours to cross-reference study reports against 24 ICH guidelines and 373 specific requirements by hand. For a 10-person biotech burning $500K/month, that's not a knowledge problem. It's a time problem.
Regfo reads your study reports, extracts 46 structured data points from each one (species, strain, GLP status, dose levels, NOAEL, target organs, exposure margins, and more), and checks them against the same ICH/FDA requirements I listed above. It gives you a compliance score and tells you specifically what's missing, with citations to the exact guideline section.
It takes about 30 seconds per study report instead of the 4-6 hours it takes manually.
If you're in the middle of a preclinical gap analysis right now, you can try the protocol checker at regfo.com — paste your protocol text and see what it finds. No login required for the first check.