On the pre-IND meeting

What an Oncology Pre-IND Meeting Actually Decides — and How to Prepare for Useful FDA Feedback

Published 2026-08-06 · Substantively reviewed 2026-09-01 against the August 2026 final FDA meetings guidance · Timelines below are FDA's published performance goals and procedural commitments under PDUFA VII, not legal advice. Practice varies by Center, office, and division.

The package is the meeting

Founders picture the pre-IND as a conversation: a room, reviewers on one side, the team on the other, questions going back and forth until everyone understands each other.

The reality is quieter and more consequential. Your meeting package is due 30 calendar days before the meeting date. That document is what the review division reads, discusses internally, and forms positions against. Before a live meeting, FDA intends to send preliminary written responses to your questions — so by the time anyone speaks, the agency's initial position already exists in writing.

Those preliminary responses are not the verdict. FDA is explicit that they should not be construed as final unless the agency and the sponsor agree that further discussion is unnecessary. What they do is set the agenda: a proposal that is underdeveloped in the package cannot be rescued in the room, because substantial new data and new proposals are not what the meeting is for.

If you take one thing from this piece: the package largely determines the quality of the meeting. The live discussion is where unresolved positions can still be clarified — and sometimes changed.

What the process actually looks like

Pre-IND is a Type B meeting. The PDUFA VII commitment letter — the operative source for meeting management — establishes the principal performance goals and procedural commitments below. The day-21, day-60, and meeting-minutes goals are measured at 90%; the other entries are sponsor deadlines or procedural commitments rather than percentage-based goals. Practice varies by Center, office, division, and the questions you bring.

1. Before you request anything. You need enough product definition, and enough of an integrated CMC, pharmacology/toxicology, and initial clinical position, to support a comprehensive review. An immature product or an open-ended program is not ready for a pre-IND, and asking early buys you vague answers. If the blocking issue is a genuinely novel one that could otherwise delay first-in-human testing, an INTERACT meeting is the earlier route — but it is a one-way door in sequence terms: the final guidance states that once you have had a pre-IND meeting or filed an IND, INTERACT is no longer the appropriate meeting type.
2. The meeting request. Purpose, objectives, proposed format, agenda, the disciplines you need in the room, and your initial questions. Question discipline starts here, not when you draft the package.
3. By day 21. FDA's goal is to grant or deny the request and specify format and date.
4. 30 calendar days before the meeting or written response. Your package is due.
5. No later than two calendar days before a live meeting. FDA intends to provide preliminary responses to your questions. They are not final unless both sides agree that further discussion is unnecessary.
6. By day 60. The goal for the live meeting — or for the written response, if that is the format granted.
7. 30 calendar days after a live meeting. The goal for FDA's official minutes. Where the format is written response only, the written response is itself the record — there are no separate minutes. If a live meeting is cancelled by agreement, preliminary responses can become the final record.
8. Within 20 calendar days of minutes or written response. You may submit a Request for Clarification. FDA's goal is to respond within 20 calendar days if the request is genuinely clarifying. New issues and new proposals are out of scope and need a different route.
9. After you submit the IND. A separate 30-day safety review. The investigation may begin only if FDA permits it or does not impose a clinical hold — with IRB and other requirements applying independently.

Two things in that sequence deserve emphasis, because they are where programs get surprised.

You may not get a live discussion — and pre-IND is named in the carve-out. FDA determines the format, and written response only is a formal meeting format, not a refusal to meet. The August 2026 final guidance is specific about who this applies to. For Type A, Type B excluding pre-IND, and Type B (end-of-phase) meetings, FDA may grant a written response only if the requester asked for one. For Type B (pre-IND), Type C, Type D and INTERACT meetings, FDA may grant a written response regardless of the format the requester asked for.

That distinction is worth sitting with. A sponsor who requests a live pre-IND meeting can be answered in writing instead, at the agency's discretion, without having asked for it — and the same sponsor could not be handled that way at end-of-phase. A rationale requesting discussion may still be submitted, and the agency may or may not convert it. Plan the package so it works with no conversation at all, because for pre-IND specifically that is a decision FDA can make alone.

An ambiguous answer is not automatically another cycle. The Request for Clarification route exists precisely for answers you cannot operationalize. It is narrow — clarification, not a second bite at a new proposal — but it is far cheaper than a new meeting.

Step 9 is the one that gets misread

Pre-IND advice does not authorize anything. It is non-binding advice based on the information you presented, and it can change as evidence changes. The formal gate is the IND review, and it is a separate proceeding on a separate clock.

This distinction has practical consequences. "FDA was supportive at pre-IND" is not "we are cleared to dose patients." The first is input; the second requires an IND that survives its own review. Programs that blur the two build financing narratives and hiring plans on a foundation that has not actually been laid.

The meeting starts before the meeting request

A pre-IND package should be the output of an integrated program decision, not the first occasion on which the functions compare their assumptions. Before requesting the interaction, the team should be able to state its confidence — and its remaining uncertainty — across mechanism, translation, safety, CMC, clinical design and execution.

That includes the evidence that did not work: rejected models, negative findings, unvalidated biomarkers and assumptions for which the program still lacks a discriminating experiment. It also includes practical readiness: test article, manufacturing path, starting-dose rationale, protocol risk controls and the decisions that different FDA answers would change.

If those pieces have not been reconciled across Research, Nonclinical, Clinical, Regulatory, CMC and Operations, the package will contain multiple programs pretending to be one.

How to spend the question ceiling

Under the August 2026 final guidance (Formal Meetings Between the FDA and Sponsors or Applicants of PDUFA Products), the recommendation is no more than ten total questions, and sub-questions count toward that ten. The guidance is explicit about the arithmetic: a question with three parts should be numbered 1, 2 and 3, not 1a, 1b and 1c. If you have been budgeting ten headline questions and treating their sub-parts as free, the final guidance closes that reading. CBER's Office of Therapeutic Products applies a ten-question maximum to its 60-minute pre-IND meetings and written responses. Treat ten as a practical ceiling rather than a statutory limit — and for genuinely complex questions, fewer is usually better, because a small number of well-built questions gets better answers than ten thin ones.

That ceiling is the scarcest resource in early development, and it forces a decision, not a wish list.

Three tests make the allocation easier:

• Does the answer retire a decision? If every possible answer leads to the same next step, the question is not worth a slot.
• Is there a proposal in it? "Does the Agency concur that X is acceptable for Y purpose, given Z?" gets you a usable answer. "What are the Agency's expectations regarding X?" gets you a restatement of published guidance — which you could have read.
• Have you already answered it? Questions that published guidance resolves are a preventable way to waste the ceiling.

Then build a decision-to-question map: for each question, the answer you expect, the adverse answer, and what you will do differently under each. If you cannot name the divergent action, you have not yet found the question.

What the division is trying to assess

At pre-IND, a review division is asking whether it can evaluate the risk of the proposed investigation in humans. That resolves into a common core, plus modules that apply depending on what you are developing.

Pre-IND is not a Regulatory Affairs workstream with occasional functional input. It is a program-level synthesis. Regulatory owns the interaction mechanics; the scientific and operational positions remain owned by the functions that must defend and execute them.

The common core. Product definition and CMC sufficient to assess identity, purity, potency, stability, and control. Nonclinical pharmacology and toxicology adequate to support the proposed starting dose, in relevant species and conditions. The initial clinical protocol: population, dose escalation, safety monitoring, stopping rules. And enough development context for the division to see where this is going.

CMC deserves a specific note. Insufficient information about identity, purity, stability, potency, or control can prevent FDA from assessing subject risk and can contribute to a clinical hold. How much risk it carries varies substantially by modality and manufacturing maturity — but it is read whether or not you asked about it.

Conditional modules.

Dose optimization

The August 2024 final guidance, Optimizing the Dosage of Human Prescription Drugs and Biological Products for the Treatment of Oncologic Diseases, asks sponsors to justify dosage using relevant nonclinical and clinical data — PK, pharmacodynamics, safety, tolerability, activity, and dose- and exposure-response — rather than settling on a dose by escalating to the maximum tolerated dose and moving on. It recommends comparing more than one dosage, and identifies a randomized parallel dose-response trial as a recommended design.

Two points of precision, because both are commonly misstated. The guidance does not specifically address the first-in-human starting dosage, and it does not specifically address radiopharmaceuticals, cellular and gene therapy products, oncolytic viruses, microbiota products, or cancer vaccines — though some of its recommendations may still apply. And comparative dosage work does not have to sit entirely in Phase 1: FDA's own position is that adequate planning can align dosage optimization with expedited development. What creates cost is not the expectation itself; it is meeting the expectation late, after a single dose has already been carried into a program that assumed it.

Endpoints and model-derived measures

FDA evaluates endpoints in relation to the trial objective, the development phase, the disease setting, and the intended regulatory use. A pre-IND may address your initial Phase 1 objectives and endpoints and the shape of the broader plan; it does not finally settle an endpoint strategy for later phases.

The distinction worth carrying into the package is between four different things a model output can be: a tool for enrollment or stratification; an exploratory biomarker; a safety or activity measure; and a measure intended to support a regulatory decision. They carry different evidentiary weight and different questions. Treating a scientifically meaningful model output as automatically an acceptable endpoint is where translation breaks down.

Population and feasibility

Eligibility criteria, line-of-therapy windows, biomarker prevalence, and referral patterns determine whether the protocol you designed can recruit. FDA does comment on population and eligibility where safety and interpretability are implicated. Operational accrual, though, is mostly yours to own — and in AI-originated programs a planning risk is confusing a computationally defined target population with one that can actually be identified, recruited, and treated at the intended sites. If that distinction is missed, the result can be slower-than-planned accrual, with downstream timeline and capital consequences.

Combinations

Where you propose a combination, the pre-IND questions are the scientific rationale, the safety basis, the dosage strategy, and how the development plan will establish each component's contribution where that applies. Working that out later, after the regimen is fixed, is materially harder.

Assay-selected populations

Where patient selection depends on an assay, a companion-diagnostic co-development obligation can arise — when the assay is essential to the safe and effective use of the therapeutic product. Not every exploratory or enrichment assay becomes a companion diagnostic. Knowing which one you have determines whether coordinated therapeutic–diagnostic co-development is required.

AI used in the development path

FDA's January 2025 draft guidance, Considerations for the Use of Artificial Intelligence To Support Regulatory Decision-Making for Drug and Biological Products, sets out a risk-based credibility framework that applies when a model produces data or information supporting a regulatory decision about safety, effectiveness, or quality. It expressly excludes AI used solely for drug discovery, or for operational efficiency where the use does not affect patient safety, drug quality, or the reliability of study results.

That scope line matters commercially. A model used only to select a target does not, by that fact, pull a credibility package into your IND. A model used for patient selection, dose selection, endpoint assessment, or manufacturing may — depending on its context of use and how much the regulatory decision rests on it. Engineering documentation should therefore be translated into FDA's concepts: question of interest, context of use, model influence, decision consequence, model risk, credibility evidence, and life-cycle maintenance where applicable.

Preparing between preliminary responses and the meeting

If you get a live meeting, the preliminary responses change what the meeting is for — and the window to use them is short. The work is:

• Triage. Which answers are sufficient as written? Mark them and stop discussing them.
• Set the real agenda — and send it. Which answers are unclear, or reveal a disagreement worth the remaining time? That, and only that, is your agenda. This is no longer just good practice: under the August 2026 final guidance the requester tells FDA whether the meeting is still needed and sends a revised meeting agenda marking which questions are considered resolved and which are still to be discussed. For pre-IND that notification is due as soon as possible after the preliminary responses arrive — which, at two calendar days out, means the triage has to be ready before they land.
• Assign. Who answers on CMC, nonclinical, clinical, statistics — by name, before the call.
• Build the counter-case and the fallback. For each contested position: the strongest argument against you, and the alternative you would accept if the division does not move. Deciding that live is how teams commit to positions they cannot support.
• Plan the close. Before the meeting ends, summarize agreements, disagreements, clarifications, and action items out loud. That summary shapes the minutes — which become the record.

A fallback is not real until the affected functions can execute it. For every answer that could change the program, identify the document that changes, the accountable owner, the decision deadline and the downstream consequence for CMC, protocol, sites, supply and capital. Where the scenarios are foreseeable, draft the alternate language and review path before the meeting.

The objective is not to predict FDA's answer. It is to prevent a predictable answer from starting an avoidable internal decision cycle.

Presentations are generally unnecessary and consume time you need for discussion. Do not introduce substantial new evidence; that is what the package was for.

After: the record is the asset

The minutes — or the written response, where that was the format — are the durable artifact. That document travels: into the IND, into diligence, into the next interaction with the division, sometimes to an acquirer years later.

Reconcile your own notes against the minutes as soon as they arrive. Where something is genuinely unclear, use the Request for Clarification window. Then push the outcome back into the development plan, the decision log, the risk register, and the IND package — while it is fresh, and while the reasoning is still recoverable.

That record also survives a change of ownership. In an acquired or partnered program, reconstruct the complete agency history — not only the latest development plan — and identify where earlier advice depended on assumptions that have since changed. Advice from another jurisdiction is useful evidence, but it is not automatically portable to FDA. In a global program, sequence authority interactions deliberately: decide which questions should be resolved once, which require region-specific evidence, and how advice from one authority will — or will not — inform another submission.

One discipline worth adopting for investor conversations: describe what the minutes actually say. FDA agreed with X; did not object to Y; qualified Z; left W unresolved. That is more precise than "we have alignment with the division," and it is the version that survives a diligence process where someone reads the minutes themselves.

What this costs when it goes wrong

For a live Type B meeting, the PDUFA sequence from request to minutes is roughly 90 calendar days; where the format is a written response, the response targeted by day 60 is itself the record. But that is the regulatory clock, and it is not the clock a founder actually lives on. The sponsor's calendar adds the readiness work before the request and the reaction after: redesign, re-analysis, sometimes a second interaction. The gap between those two clocks is where programs lose time they did not plan to lose — and unlike the regulatory clock, that gap is substantially under your control.

The costs compound in a specific order. Calendar first. Then capital, because an early-stage oncology company's burn continues whether or not the quarter produced usable regulatory answers. Then the narrative: a Series B conversation turns on whether the regulatory path is credible, and what you can honestly say about it is fixed by what the minutes say.

The most expensive version is the quietest. Advice is given, clearly and correctly. Its implications are not fully understood. The program proceeds, and the constraint surfaces much later, when the record shows the division raised it at the time. Nothing about that failure looks like a failure while it is happening — which is exactly what makes it worth engineering against.

The standard to hold yourself to

Before the package is locked, take each question and write three things: the answer you expect, the answer a skeptical reviewer would give, and what you will do differently under each. Where the first two diverge, you have found either a question that needs rewriting or a position that needs evidence. Do the same for the sections you are not asking about — the division reads the whole package.

That is what calibration means at this stage: knowing which of your positions will survive contact with the division, which are conditionally true and on what condition, and which are assertions you have not yet earned — while there is still time to fix it in a document rather than in a program.

Common questions about pre-IND meetings

These are the mechanics. The harder question — which of your positions will survive contact with the division — is judgment, and it is the substance of oncology drug development consulting at the front end of a program.

---

The AI-Oncology Calibration Sprint is built for this kind of decision: two weeks, fixed scope, senior clinical-regulatory judgment on the question in front of your program — with the strongest counter-case and the conditions under which the recommendation would change. No one can guarantee an outcome with the agency. What you can do is identify which positions are most vulnerable to regulatory scrutiny while changing them still costs less.

Sources

• FDA — PDUFA Reauthorization Performance Goals and Procedures, Fiscal Years 2023 Through 2027: operative performance goals for meeting management, written responses, preliminary responses, minutes, and Requests for Clarification.
• FDA — Formal Meetings Between the FDA and Sponsors or Applicants of PDUFA Products (August 2026, final guidance): meeting types, formats, and question recommendations, including the ten-question ceiling and the sub-question numbering convention, the meeting-format table placing Type B (pre-IND) among the types FDA may answer with a written response regardless of the format requested, the INTERACT sequencing limit, and the revised-agenda step following preliminary responses.
• FDA — Optimizing the Dosage of Human Prescription Drugs and Biological Products for the Treatment of Oncologic Diseases (final, August 2024).
• FDA — Considerations for the Use of Artificial Intelligence To Support Regulatory Decision-Making for Drug and Biological Products (January 2025, draft guidance): risk-based credibility framework and its scope exclusions.
• FDA — OTP Pre-IND Meetings: the ten-question maximum cited above, as applied by CBER's Office of Therapeutic Products. OTP-specific, not a general CDER oncology rule.
• FDA — Companion Diagnostics.
• FDA — IND Applications for Clinical Investigations: Overview and Clinical Hold: the separate 30-day IND review and clinical-hold gate.
• FDA OCE — Oncology Regulatory Expertise and Early Guidance (OREEG).

Jesús Gómez-Navarro, M.D., is a medical oncologist and drug development executive, and founder of OncAdios LLC. He advises AI-oncology and biotech companies as a fractional CMO, board director, or co-founder.

← Back to Writing  ·  What a first-in-human trial is actually for →  ·  When model validation is not enough →  ·  The Calibration Sprint →