I started tracking IND deficiencies because I kept hearing the same problems described in different words by different people. A regulatory affairs lead in Durham who lost four months to a clinical hold over a missing non-rodent tox study. A CSO in San Francisco who had to repeat a genotoxicity assay because the original wasn't GLP. A consultant in Boston who told me, unprompted, that she could predict with about 80% accuracy which INDs would get clinical holds just by looking at the nonclinical section.
The deficiencies aren't surprising. They're the same five things, over and over. What's surprising is how often they still happen — even at companies that have done this before. The complexity of an IND submission creates a kind of fog. There are hundreds of requirements across five CTD modules. People focus on the hard scientific questions and miss the structural ones.
Five deficiencies, ranked by how often I've encountered them. The ranking is subjective — I'm going by conversations and FDA enforcement data, not a formal meta-analysis — but the top three are so common that the order barely matters.
1. Incomplete nonclinical safety package
ICH M3(R2) Table 1 lays out the minimum nonclinical studies for each clinical phase. For Phase 1: repeat-dose toxicity in two species, safety pharmacology core battery, genotoxicity standard battery per ICH S2(R1), and pharmacokinetic characterization in the tox species. One table. Maybe 20 cells. And yet.
The gap I hear about most — by far — is the missing non-rodent species. I wrote about this in the clinical holds guide and I keep coming back to it because something about the rat-only trap is almost magnetic. Company runs rat tox, gets clean results, starts assembling the IND. Month three, someone actually reads M3(R2) and discovers they need dog data too. CRO lead times for beagle studies run 3-6 months before the study even starts. Add 2-3 months for the study itself. That's 5-9 months of delay from a single overlooked requirement.
Then there's the duration mismatch — teams run 14-day tox studies and write a Phase 1 protocol with 28 days of dosing. Table 1 is explicit about the mapping. 7-day clinical dosing needs 2-week tox. 14-day clinical needs 1-month tox. This should be easy to check. It still catches people.
The third common gap is safety pharmacology. The core battery per ICH S7A covers cardiovascular (with hERG per ICH S7B), respiratory, and CNS. Some teams do the hERG and skip the rest, or bury respiratory and CNS endpoints in the general tox study without calling them out explicitly. FDA reviewers look for safety pharmacology data as a distinct section. If they can't find it fast, they flag it.
The fix is boring but effective: map every cell in M3(R2) Table 1 to a specific study in your program. Study ID, CRO name, completion date. Any empty cell is a risk you're choosing to carry.
2. Inadequate starting dose justification
FDA's 2005 dose guidance is still the primary reference in 2026, which is kind of remarkable for a 21-year-old document. The math itself isn't complicated: NOAEL from your most sensitive tox species, converted to a human equivalent dose (HED) using body surface area scaling factors, divided by a safety factor (usually 10x).
What goes wrong is almost embarrassingly basic. I've reviewed dose justification sections where the calculation used the rat conversion factor (6.2) when the non-rodent species was a dog (factor of 1.8). These are published values in Table 1 of the FDA guidance. Printed numbers. Using the wrong one shifts your starting dose 3-4x in either direction.
Or the NOAEL comes from a non-GLP dose range-finding study instead of the pivotal GLP tox study. You can reference non-GLP data as supporting context, sure. But the NOAEL that feeds into the HED calculation — the one the whole starting dose rests on — needs GLP provenance.
The third failure mode is organizational rather than scientific: the justification is buried. I've seen it in protocol appendices, in the pharmacokinetics section of Module 4, in a standalone document cross-referenced by page number. Reviewers expect it in the Nonclinical Overview (Module 2.4), step by step, on one page. If the reviewer has to hunt for it — technically not a deficiency, but it slows the review, and slowed reviews generate more questions. Every time.
Write it as a standalone section in Module 2.4: NOAEL by species, GLP status of the source study, HED calculation with the explicit conversion factor, safety factor with justification, proposed starting dose. One page. Two minutes to read. That's the target.
3. GLP compliance gaps
21 CFR Part 58 — all pivotal nonclinical safety studies must be conducted under Good Laboratory Practice. I wrote a full GLP checklist on this, but the short version: GLP is binary. Either the study was conducted under GLP or it wasn't. There's no partial credit.
The classic is the non-GLP hERG assay. Company runs an early hERG during lead optimization, clean result, includes it in the IND as pivotal cardiovascular safety data. FDA rejects it. The science was fine. The paperwork wasn't. A GLP hERG costs $30-50K. The clinical hold costs months. I genuinely do not understand how this keeps happening, but it does, so I keep writing about it.
Undocumented protocol deviations are subtler. A study director changes a blood collection timepoint mid-study — scientifically reasonable, but if there's no formal protocol amendment signed by the study director, it's a GLP deviation. And here's the thing that makes this particularly insidious: FDA inspectors don't just flag the undocumented change. They infer that if one change went undocumented, there may be others. The trust fracture is worse than the deviation itself.
Test article characterization gaps are the third bucket. No certificate of analysis, stability data that doesn't cover the study duration, no formulation homogeneity verification. Under Part 58, the test article needs identity, strength, purity, and composition data before the study starts. Small biotechs outsourcing to CROs are the most vulnerable here — the assumption is that the CRO handles it, but actually the sponsor is responsible for providing characterized test article.
Before any GLP study starts: confirm with your CRO that the study is on the GLP master schedule, that the QAU has been notified, and that phase inspections are scheduled. After the study: check the final report for the compliance statement, QA statement with inspection dates, and the list of protocol deviations. Read that list carefully.
4. CMC deficiencies — the slow-motion kind
CMC problems (21 CFR 312.23(a)(7)) are different from the first three deficiencies in a way that makes them worse: they're slow to fix. You can schedule a hERG assay in weeks. You can't accelerate a stability study.
The drug substance and drug product requirements in Module 3 trip companies on three things. Insufficient stability data — your trial runs 6 months but you only have 3 months of real-time stability. Incomplete impurity characterization — a new impurity above the ICH Q3A identification threshold that nobody profiled. And batch comparability — your tox batch came from one process, your clinical batch from another, and now you need analytical data showing equivalence before FDA will accept the bridge.
I won't pretend to be a CMC expert. I'm a preclinical person looking at these from the nonclinical side. But the batch comparability issue surfaces in my world all the time, because it turns a CMC question into a tox question: if you can't demonstrate comparability, FDA may require bridging toxicology studies with the new batch. That's $300-800K and 4-6 months you didn't budget for. The preventive move is obvious in retrospect — manufacture tox and clinical batches by the same process, or at least plan the comparability assessment before you need the answer.
5. Protocol design — where the nonclinical and clinical worlds collide
This one I almost merged into the section on dose justification, because the overlap is real. Protocol deficiencies rarely cause clinical holds alone, but they compound fast when combined with nonclinical gaps. The governing framework: 21 CFR 312.23(a)(6) and 312.42(b)(1)(ii).
Stopping rules are the big one. "The investigator will use clinical judgment" is not a stopping rule. It's an abdication. FDA wants dose-limiting toxicity definitions, escalation pause criteria, and a documented decision process for continuing or stopping. Specific, measurable, auditable.
The second issue is more interesting to me because it sits exactly at the nonclinical-clinical interface: the informed consent form must reflect your toxicology findings. Per 21 CFR 50.25, reasonably foreseeable risks need to be described in language a participant can understand. Your rat study showed hepatotoxicity at the high dose? That goes in the consent form. FDA cross-references Module 4 findings with the consent document. They will catch the omission. It's one of the few things I've seen reviewed with genuine rigor every time.
The dose escalation timing question — how long between cohorts, what safety data gets reviewed before moving up — is less dramatic but equally common as a finding. Build the safety review intervals into the protocol. Have a Phase 1 investigator or clinical pharmacologist review the full protocol before it goes into the IND. This is not a document to draft in isolation.
What connects all five
Every one of these deficiencies is knowable in advance. That's what makes them so frustrating. ICH M3(R2) tells you which studies you need. Part 58 tells you how to conduct them. The FDA dose guidance tells you how to calculate the starting dose. 21 CFR 312.23 tells you what goes in the IND. None of it is hidden.
The problem — and I've been thinking about this more since I started working on compliance tooling — is that "knowable" and "known" are different things. The requirements are scattered across maybe 30 documents, each 50-100 pages, cross-referencing each other in ways that assume you've read the other 29. The people who have actually internalized all of it charge $400/hour, and honestly, most of them specialize in two or three of the five deficiency areas above. Nobody holds the full picture in their head.
There's a related problem I didn't list above because it doesn't fit neatly into one category: the information asymmetry between the team running the studies and the team assembling the IND. At small biotechs those are sometimes different people, sometimes the same person at different points in time, and the handoff — or the memory gap — is where requirements fall through. The nonclinical scientist who ran the rat tox study six months ago knows the NOAEL. The regulatory affairs person writing Module 2.4 today may not know whether that NOAEL came from a GLP or non-GLP study. That's deficiency #2 and #3 colliding.
Anyway. Read M3(R2) Table 1. It's one table. Takes 10 minutes. If your IND-enabling program doesn't map cleanly to every cell in that table, you have a gap.
FAQ
What percentage of IND submissions receive clinical holds from FDA?
Available data suggests 5-10% of original IND submissions receive clinical holds before first patient dosing. FDA reviews several thousand original INDs each year — across all sponsors, disease areas, and modalities. The rate skews higher for first-time sponsors and small biotechs without dedicated regulatory affairs teams. In-house experience matters: companies filing their first IND face a hold rate meaningfully above average, while sponsors with 3+ prior INDs see lower rates. Oncology programs under ICH S9 face reduced nonclinical requirements (no carcinogenicity studies required pre-IND, less stringent tox duration requirements), which pulls down holds in that subset. The more actionable statistic: over 80% of holds across all programs trace to five categories — missing nonclinical data, inadequate dose justification, GLP gaps, CMC issues, and protocol design problems — every one of which is identifiable before filing if you check against ICH M3(R2) Table 1.
How long does it take to fix a clinical hold due to missing tox data?
Count on 6-12 months minimum if the missing study is a non-rodent repeat-dose toxicology study. Breaking that down: 2-4 months to secure a CRO contract and slot (beagle and NHP facilities book out fast), 1-3 months for study initiation and test article shipment, 2-4 months for the study itself (a 4-week GLP repeat-dose study in dogs runs roughly 4 weeks on-study plus 2-4 weeks for report finalization), then 2-4 weeks to write the complete response and 30 days for FDA review. For rodent studies the timeline compresses to 4-8 months because slots are more available. For a missing GLP hERG assay — cardiovascular safety pharmacology — the timeline is shorter: 6-12 weeks for a GLP-compliant in vitro assay, then the same 30-day FDA review window. The total cost including CRO fees runs $200K-$600K depending on study type, before counting burn rate during the hold.
Can I submit an IND without a non-rodent toxicology study?
Technically yes — you can submit an IND package without non-rodent data. FDA won't refuse to accept the submission. What happens next: FDA will issue a clinical hold under 21 CFR 312.42 within 30 days because ICH M3(R2) Table 1 explicitly requires repeat-dose toxicology in two species (one rodent, one non-rodent) before Phase 1 in most programs. The only legitimate exceptions are programs where a second species is scientifically inappropriate — for example, a biologic that has no pharmacological activity in any non-rodent species. If that's your situation, you need to document the justification thoroughly and get FDA's written agreement at a pre-IND meeting before filing. Without that pre-IND agreement, submitting without the non-rodent study is essentially filing an incomplete application and waiting 30 days to receive confirmation that it's incomplete.
What does FDA mean by "adequate and well-controlled" for nonclinical studies?
For nonclinical GLP toxicology studies, FDA's standard comes from 21 CFR Part 58 (Good Laboratory Practice) rather than the "adequate and well-controlled" language in 21 CFR 314.126, which applies to clinical trials. In the nonclinical context, FDA wants studies that: used characterized test article with identity, strength, purity, and stability data before study start; followed a protocol with pre-defined endpoints, dose groups, and collection timepoints; documented all deviations with formal protocol amendments; included appropriate controls (concurrent vehicle control, positive controls where relevant); and were conducted with QAU oversight including phase inspections. The final study report must include a GLP compliance statement and a QA statement listing all inspections and dates. Studies missing any of these elements risk rejection — not as data, but as pivotal studies supporting the safety basis for human dosing.
How do I fix a GLP compliance gap in my IND?
First, get the QA audit report from your CRO and read the deviation list carefully. Minor deviations — a late calibration record, a slightly late report — can often be addressed with a deviation report that acknowledges the issue and explains why scientific integrity wasn't compromised. Document these in the IND with a brief explanation. Significant deviations — undocumented protocol changes, test article characterization gaps, missing QAU inspections — are harder to explain away. FDA reviewers don't just flag the documented deviation; they assume undocumented deviations may exist. If the deviation is material to the study's conclusions, you may need to repeat the study under full GLP. The preventive move: confirm GLP master schedule registration before each study starts, get phase inspection reports during the study (not after), and check the compliance statement before the final report is signed. Catching a GLP gap at the CRO is weeks of work. Catching it at the IND review is months.
What is NOAEL and why does it matter for IND submission?
NOAEL stands for No-Observed-Adverse-Effect Level — the highest dose in your GLP toxicology study at which no adverse effects were observed. It's the anchor for your starting dose calculation. FDA's 2005 dose guidance (still current in 2026) requires you to convert the NOAEL from your most sensitive tox species to a human equivalent dose (HED) using body surface area scaling factors from Table 1 of that guidance (rat: 6.2, dog: 1.8, monkey: 3.1). You then divide the HED by a safety factor — typically 10x for standard Phase 1 non-oncology programs — to get the maximum recommended starting dose (MRSD). If the NOAEL comes from a non-GLP study, FDA won't accept it as the basis for human dosing. If you use the wrong species conversion factor, your starting dose shifts 3-4x. If your NOAEL is "greater than" the highest dose tested and you didn't adequately expose the animals, FDA will question whether you've established a true NOAEL. Show the complete calculation in Module 2.4, not buried in an appendix.
Can I use published literature instead of my own studies in an IND?
Yes, published literature can support your IND under 21 CFR 312.23(a)(8), but it can't replace GLP pivotal studies for your specific compound. You can use literature to support the scientific rationale, characterize the drug class, or provide context for findings in your own studies. What literature can't do: stand in as the GLP repeat-dose toxicology data, safety pharmacology core battery, or genotoxicity standard battery for your compound. FDA needs data on your specific molecule at your specific doses in animals exposed to your specific test article. Published data on a structurally similar compound may strengthen your IB and Nonclinical Overview, but it won't satisfy the M3(R2) Table 1 requirements. One legitimate use of literature: if published studies demonstrate that a second tox species is scientifically inappropriate for your compound (no pharmacological activity, no target expression), that literature can support a waiver request — but you still need FDA's written agreement at a pre-IND meeting before relying on it.
Related reading:
- FDA IND Submission Checklist 2026 — complete requirements by CTD module
- Clinical Holds: Why FDA Stops Trials — prevention strategies
- ICH S2 Genotoxicity Testing Guide — S2(R1) battery requirements
- How to Check Preclinical Studies Against ICH Guidelines — 8-step gap analysis