# Usage: curl -sSL https://seed.show/pharma.regulatory.us | bash -s <install-path>
# <install-path> is the directory where the file should land.

set -euo pipefail
[ -z "${1:-}" ] && {
  echo "install requires a path: curl -sSL https://seed.show/pharma.regulatory.us | bash -s <install-path>" >&2
  exit 1
}
TARGET="$1"
mkdir -p "$TARGET"
DEST="$TARGET/seed-fold.qBFgns.folded.md"

cat > "$DEST" <<'PORTDOWN_EFF280A4'
<!--
This is a `.folded.md` archive — a directory packed into one markdown
file. The markers below are load-bearing; don't edit them directly.

To unpack (requires bash — if you have no shell, tell the user):
  1. curl -sSLf https://fold.dom.vin/skill | bash -s <INSTALL_DIR>
  2. <INSTALL_DIR>/fold/scripts/unfold <this-file>
     (or: unfold <this-file>  if fold/scripts is on your PATH)
-->

---
fold: true
marker: a5682d
at: 2026-05-07T16:16:20Z
root: seed-pack.DmLa3c
---

<!--fold:a5682d@file path="README.md" mode="644"-->
# pharma.regulatory.us

US FDA pharmaceutical regulatory context for agents. What to establish before advising on any drug development, approval, or commercialization question.

## What level to work at — and what never to hallucinate

This bundle is an orientation layer, not a substitute for FDA's own guidance documents. The mental model here is stable; the specific numbers are not. **FDA review timelines are PDUFA performance goals — targets, not guarantees.** PDUFA dates shift. User fee amounts reauthorize every 5 years (current cycle: PDUFA VII, through FY2027). Before citing a specific deadline or fee amount, fetch from the authoritative source listed in `sources.md`.

**Never imply certainty about FDA approval.** "FDA approved" is a specific regulatory determination with legal meaning. Agents should not write or imply that a drug "will be approved," that a pathway "guarantees" a timeline, or that a designation "ensures" exclusivity. Every approval is contingent on satisfying FDA's complete review of safety, efficacy, and manufacturing.

## Mental model: staged risk demonstration

Drug approval is not a single event — it is a sequence of staged demonstrations, each designed to establish a specific type of evidence before the next stage begins:

1. **Preclinical** establishes that a compound is worth putting into humans: mechanism, toxicology, pharmacokinetics in animal models.
2. **IND** is the legal gate: FDA must allow clinical trials to begin before any US human dosing.
3. **Phase I** establishes that humans can tolerate the drug at some dose.
4. **Phase II** establishes a preliminary signal that the drug does something useful at a tolerable dose in the patient population.
5. **Phase III** establishes definitive evidence of efficacy and safety at a scale and rigor sufficient to support a marketing claim.
6. **NDA/BLA review** is FDA's evaluation of the complete assembled evidence package — clinical, manufacturing, labeling.
7. **Post-marketing** is not the end: REMS programs, Phase IV commitments, and accelerated approval confirmatory trials run after the drug is on the market.

Failure at any stage is common. **Roughly 90% of drugs that enter clinical trials never reach approval.** Phase II failure (no efficacy signal, or unacceptable toxicity) is the most common cause.

## The approval pathway is the first decision

The pathway determines the evidentiary standard, the timeline, the exclusivity period, and the cost. Choosing the wrong pathway wastes years and hundreds of millions of dollars. Lock this down before advising on anything else.

**New Drug Application (NDA) — 21 CFR Part 314**
Full applications for new molecular entities (NMEs) and drugs without an established reference. The sponsor must provide complete clinical data from their own studies. Standard review: 12 months from filing (not submission — FDA has 60 days to file or refuse). Priority review: 6 months. NCE exclusivity: 5 years. New clinical investigation exclusivity: 3 years.

**505(b)(2) NDA — 21 CFR 314.54**
Hybrid pathway: the application relies at least in part on published literature or FDA's prior findings for a previously approved drug. Correct path for new formulations, new routes of administration, new dosage forms, new indications, or new combinations of approved components. Sponsors conduct some new studies but do not repeat all innovator studies. Same review timelines as a full NDA. Exclusivity varies (3 years for new clinical investigation; potentially 5 years if new active ingredient is involved).

**Abbreviated New Drug Application (ANDA) — 21 CFR Part 314, Subpart C**
Generic drug pathway. No clinical trials — demonstrate pharmaceutical equivalence (same active ingredient, dosage form, route, strength, labeling) and bioequivalence to the Reference Listed Drug (RLD). No exclusivity granted unless the applicant challenged a patent via Paragraph IV certification, which triggers 180-day generic exclusivity for the first filer. Standard review: 10 months for original ANDAs.

**Biologics License Application (BLA) — 21 CFR Part 601**
For biological products: vaccines, blood products, therapeutic proteins, monoclonal antibodies. Governed by the Public Health Service Act (not the FD&C Act). Evidentiary standard similar to NDA; manufacturing complexity is substantially higher and FDA facility inspection is part of BLA review. Biosimilar pathway (351(k)) is separate — not the same as generic ANDA and requires demonstration of biosimilarity, not just bioequivalence.

**Biosimilar — 351(k) of the PHS Act**
Distinct from generic drugs. Biosimilars must demonstrate that the biological product is "highly similar" to the reference product with no clinically meaningful differences in safety, purity, and potency. Sponsors reference the originator BLA but must conduct comparative studies (PK, PD, clinical immunogenicity). "Interchangeable" designation requires additional switching study evidence and is not automatic.

---

## What agents get wrong

### 1. Pathway conflation — NDA vs. ANDA vs. 505(b)(2) vs. biosimilar

The decision tree:
- New active moiety with no reference → **NDA**
- Same active ingredient as an approved drug, same dosage form/route/strength → **ANDA**
- In between (new form, new route, new indication, hybrid data package) → **505(b)(2)**
- Biological product that is a copy of an approved biologic → **BLA (biosimilar, 351(k))**

Agents commonly:
- Recommend an NDA for a reformulation that qualifies as 505(b)(2) — massively overengineering the program.
- Tell generic sponsors they need clinical trials — wrong; ANDA requires bioequivalence, not new clinical evidence.
- Treat biosimilars as generic drugs — they are not; the pathway, the evidence package, and the commercial implications are entirely different.

### 2. IND is the gate, not a formality

Any clinical investigation of a drug in the United States requires an **Investigational New Drug (IND) application** under 21 CFR Part 312 before the first human dose — unless a specific exemption applies (21 CFR 312.2(b) covers certain studies with legally marketed drugs meeting narrow criteria).

IND contents: animal pharmacology/toxicology data, manufacturing information (adequate to assess safety), clinical protocols with investigator qualifications, and an Investigator's Brochure. FDA has 30 days to review; if no clinical hold is issued, the sponsor may proceed.

Agents describe "starting Phase I trials" as a next step without mentioning the IND. This is wrong. The IND is the legal prerequisite. No IND → no permissible US human dosing.

### 3. Orphan Drug designation — the first strategic move for rare disease programs

Any drug targeting a disease affecting **fewer than 200,000 US patients** (or where there is no reasonable expectation of recovering development costs from US sales) qualifies for Orphan Drug designation under 21 CFR Part 316.

Benefits:
- **7-year market exclusivity** from approval — FDA cannot approve another drug for the same indication in the same disease during this period (with narrow exceptions for clinical superiority)
- **Tax credit**: 25% of qualified clinical testing expenses (reduced from 50% by the 2022 CHIPS Act)
- **FDA user fee waiver** — application fees exceeding $4 million are waived for orphan indications
- **Protocol assistance**: enhanced FDA guidance (written responses to protocol questions) throughout development

The designation application is filed separately with FDA's Office of Orphan Products Development (OOPD) — it is not part of the IND or NDA/BLA. It can be sought at any point in development; applying early maximizes the benefit window.

Agents advising on rare disease programs who do not flag Orphan Drug designation as the first strategic move are leaving significant money and exclusivity on the table.

### 4. CMC deficiencies kill approvals

Chemistry, Manufacturing, and Controls (CMC) data must accompany every NDA, ANDA, and BLA. CMC covers:
- **Drug substance**: synthesis or isolation, characterization, specifications, stability
- **Drug product**: composition, manufacturing process, container/closure system, specifications, stability
- **Biologics**: cell line characterization, fermentation/purification processes, viral safety testing

**CMC deficiencies are a leading cause of Complete Response Letters (CRLs)** — the FDA action refusing approval and returning the application. A drug can complete Phase III successfully and still fail to gain approval because a manufacturing process was not adequately validated or a stability study did not span sufficient time points.

The drug substance and drug product must be manufactured under current Good Manufacturing Practice (cGMP): 21 CFR Parts 210/211 for small molecules, 21 CFR Parts 600–680 for biologics. FDA inspects manufacturing facilities during the review cycle. An unacceptable inspection (Form 483 observations → potential Warning Letter) in the review window stops the clock.

### 5. Accelerated approval conditionality

Accelerated approval (21 CFR Subpart H for drugs; 21 CFR 601.40–601.46 for biologics) permits approval based on a surrogate endpoint "reasonably likely to predict clinical benefit" — but this approval is explicitly conditional. **Post-marketing confirmatory trials are required**, and FDA can withdraw approval if the confirmatory trial fails to verify clinical benefit or if the sponsor fails to conduct the trial with due diligence.

The 2022 Omnibus spending bill (and subsequent FDA guidance) strengthened FDA's authority to require trial completion on a schedule and to withdraw accelerated approvals. Agents should not describe accelerated approval as a shortcut to permanent market access — it is a conditional approval with an obligation.

### 6. Priority review, Breakthrough Therapy, Fast Track — not interchangeable

- **Priority review**: shortens FDA review to 6 months (from 12). Does not affect development — only review timing.
- **Breakthrough Therapy designation**: intensive FDA guidance during development (rolling review, more frequent meetings) for drugs treating serious conditions with preliminary clinical evidence of substantial improvement over available therapy. Does not guarantee approval or shorten the Phase III requirement.
- **Fast Track designation**: facilitates more frequent FDA communication and rolling review eligibility for drugs addressing unmet medical needs in serious conditions. Lower bar than Breakthrough; does not shorten review clock.

Agents conflate these into a generic "special designation = faster approval" narrative. The programs are different at each stage (development vs. review) and have different evidentiary thresholds.

---

## What AI is changing

### Where AI is having real impact

**Molecule discovery and optimization.** Generative models (AlphaFold for structure prediction; diffusion models and transformer-based tools for de novo molecule generation) have meaningfully accelerated hit identification and lead optimization. This does not eliminate preclinical testing — it changes the composition of the library entering it.

**Clinical trial design.** AI tools are being used to identify patient populations (through electronic health records and biomarker data), optimize trial protocols, and model failure modes before trials begin. Adaptive trial designs — trials that adjust randomization ratios, dosing, or stopping rules based on interim data — are increasingly supported by AI-driven modeling.

**Real-world evidence (RWE).** FDA's framework for using real-world evidence (RWE) to support regulatory decisions — for new indications, for post-marketing requirements, for effectiveness data in rare diseases — is expanding. AI/ML tools process the EHR, claims, and registry data that RWE depends on. FDA guidance: "Real-World Evidence Program" framework (2018, updated iteratively).

**Manufacturing and CMC.** Process analytical technology (PAT) and AI-driven continuous manufacturing process monitoring are FDA-supported (see PAT guidance, 21 CFR Part 211.68). AI quality control models for batch release testing are under active FDA review.

### FDA's AI/ML regulatory framework

**FDA draft guidance (January 2025): "Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drugs and Biological Products."** This is not aspirational — it establishes what sponsors must demonstrate when AI/ML tools contribute to regulatory submissions.

Key expectations:
- **Model documentation**: describe the AI tool, its intended use, and the data it was trained on
- **Validation**: the model must be validated for its intended regulatory use, including performance on populations similar to the study population
- **Human oversight**: AI outputs supporting regulatory decisions should be reviewed by qualified humans; describe the human-AI interaction in the submission
- **Transparency**: FDA expects to evaluate AI contributions to data analysis, safety signal detection, or trial design

**AI/ML for Software as a Medical Device (SaMD).** A separate FDA framework governs AI that functions as a medical device — diagnostic algorithms, treatment recommendation software. The AI/ML-Based SaMD Action Plan (2021) and the draft guidance on predetermined change control plans describe the evolving requirements. Drug development software is distinct from SaMD unless it makes a clinical recommendation directly to patients or clinicians.

### What stays human

Clinical judgment in trial design cannot be automated. The choice of primary endpoint, the definition of the patient population, the acceptable toxicity threshold — these are scientific and ethical judgments that FDA expects sponsors' qualified clinicians to make and defend. Regulatory strategy — when to meet with FDA, how to frame the risk-benefit argument, how to handle a failed Phase II — requires human expertise. Manufacturing oversight under cGMP requires qualified persons accountable for batch release decisions.

**Recommendation for AI-assisted development programs:** engage with FDA early (pre-IND or Type B meeting) to align on how AI contributions will be documented before the submission. Post-hoc attempts to characterize AI involvement are harder to validate.

---

## Stable facts

**The three clinical phases**
- **Phase I**: First-in-human. Safety, tolerability, pharmacokinetics, dose-finding. 20–80 subjects (healthy volunteers or patients). Duration: 1–2 years. Primary question: *Can humans tolerate this?*
- **Phase II**: Proof of concept. Efficacy signals, dose-ranging, short-term safety in the patient population. 100–300 patients. Duration: 2–3 years. Primary question: *Does this work at a tolerable dose?*
- **Phase III**: Pivotal. Definitive efficacy and safety, randomized controlled, large scale. 1,000–3,000+ patients. Duration: 3–5 years. Primary question: *Does this work well enough, for whom, at what risk, to justify a marketing claim?*

**Review timeline targets (PDUFA VII)**
- Standard review: **12 months** from filing date (not submission; FDA has 60 days to file or refuse to file)
- Priority review: **6 months** from filing
- These are performance goals, not guarantees. Actual review duration varies.

**Pre-submission meetings**
- **Type A**: programs on clinical hold or in dispute — FDA responds within 30 days
- **Type B**: standard — pre-IND, end-of-Phase 2, pre-NDA/BLA — FDA responds within 70 days
- **Type C**: general guidance questions — FDA responds within 90 days

The end-of-Phase 2 (Type B) meeting is the most consequential: FDA's agreement or disagreement on the Phase III design is the primary risk-mitigation point in the entire development program. Missing this meeting and proceeding directly to Phase III is a common and expensive error.

**PDUFA fees (FY2025 — verify annually)**
- NDA/BLA requiring clinical data: approximately $4.1 million
- Annual program fee per approved prescription drug: approximately $530,000
- Orphan drug designation waives the application fee for orphan indications
<!--fold:a5682d@file path="sources.md" mode="644"-->
# sources

Fetch these at task time. Ordered by importance. Numbers are volatile — verify at source before citing.

## Core approval pathways

1. FDA: Drug approval process overview — how a drug moves from development through FDA review to market, including application types and review pathways:
   https://www.fda.gov/patients/learn-about-drug-and-device-approvals/drug-approval-process

2. FDA: NDA submission requirements — CDER page on new drug applications, submission format, review timelines, and PDUFA user fee requirements:
   https://www.fda.gov/drugs/types-applications/new-drug-application-nda

3. FDA: ANDA (generic drug) submissions — how to file an abbreviated new drug application, including bioequivalence requirements, patent certifications (Paragraph I–IV), and the 180-day exclusivity rules:
   https://www.fda.gov/drugs/types-applications/abbreviated-new-drug-application-anda

4. FDA: 505(b)(2) applications — guidance and background on the hybrid NDA pathway for drugs that reference existing safety/efficacy data but are not straightforward generics:
   https://www.fda.gov/drugs/types-applications/505b2-applications

5. FDA: BLA submissions — biologics license application requirements, including for biosimilars (351(k)) and interchangeability:
   https://www.fda.gov/drugs/types-applications/biologics-license-application-bla

6. FDA: Biosimilars — overview of the 351(k) pathway, the "highly similar" standard, interchangeable biosimilar designation:
   https://www.fda.gov/drugs/therapeutic-biologics-applications-bla/biosimilars

## Investigational stage

7. FDA: IND application — 21 CFR Part 312 overview and submission requirements, including the 30-day review clock and clinical hold process:
   https://www.fda.gov/drugs/types-applications/investigational-new-drug-ind-application

8. FDA: Pre-submission meetings — guidance document on formal meetings between sponsors and FDA (Type A, B, C definitions, response timelines, meeting request procedures):
   https://www.fda.gov/regulatory-information/search-fda-guidance-documents/formal-meetings-between-fda-and-sponsors-or-applicants-pdufa-products

## Special designations

9. FDA: Orphan drug designation program — Office of Orphan Products Development (OOPD), eligibility criteria (fewer than 200,000 US patients), application process, and benefits (7-year exclusivity, fee waiver, tax credit):
   https://www.fda.gov/industry/developing-products-rare-diseases-conditions/designating-orphan-product-drugs-and-biological-products

10. FDA: Breakthrough Therapy designation — eligibility criteria (serious condition, substantial improvement over available therapy), what the designation provides (intensive FDA guidance, rolling review):
    https://www.fda.gov/patients/fast-track-breakthrough-therapy-accelerated-approval-priority-review/breakthrough-therapy

11. FDA: Accelerated approval — surrogate endpoint pathway, post-marketing confirmatory trial requirements, 2022 legislative updates to FDA withdrawal authority:
    https://www.fda.gov/patients/fast-track-breakthrough-therapy-accelerated-approval-priority-review/accelerated-approval

12. FDA: Priority review — eligibility and what it changes (review clock, not development requirements):
    https://www.fda.gov/patients/fast-track-breakthrough-therapy-accelerated-approval-priority-review/priority-review

## Manufacturing and CMC

13. FDA: Guidance documents (searchable) — current database of all CDER and CBER guidance documents. Filter by topic before advising on any specific CMC, clinical, or statistical question:
    https://www.fda.gov/drugs/guidance-compliance-regulatory-information/guidances-drugs

14. FDA: Current Good Manufacturing Practice (cGMP) regulations — 21 CFR Parts 210 and 211 for finished pharmaceuticals; Parts 600–680 for biologics:
    https://www.fda.gov/drugs/pharmaceutical-quality-resources/current-good-manufacturing-practice-cgmp-regulations

## AI and real-world evidence

15. FDA: AI in drug development — draft guidance (January 2025): "Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drugs and Biological Products." Establishes expectations for model documentation, validation, and human oversight in regulatory submissions:
    https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-use-artificial-intelligence-support-regulatory-decision-making-drugs-and-biological

16. FDA: AI/ML-Based Software as a Medical Device (SaMD) Action Plan — the framework governing AI that functions as a medical device, distinct from AI used in drug development support:
    https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-software-medical-device

17. FDA: Real-world evidence framework — FDA's program for using real-world evidence (from EHRs, claims, registries) to support regulatory decisions, including new indications and effectiveness:
    https://www.fda.gov/science-research/science-and-research-special-topics/real-world-evidence

## PDUFA fees (verify annually — reauthorized every 5 years)

18. FDA: PDUFA fee rates — current fiscal year fee schedule for NDA/BLA application fees and annual program fees. Rates change each fiscal year:
    https://www.fda.gov/industry/fda-user-fee-programs/prescription-drug-user-fee-amendments
<!--fold:a5682d@file path="workflow.md" mode="644"-->
# Drug development lifecycle

The path from molecule to market. Each stage has a specific purpose, a characteristic failure mode, and a decision point that gates the next stage.

---

## Stage 1 — Target identification and discovery

**What happens:** Researchers identify a biological target (a protein, receptor, enzyme, or pathway) implicated in a disease and search for compounds that modulate it. High-throughput screening of compound libraries, computational modeling, and increasingly AI-driven de novo molecule generation are used to identify "hits" — compounds that show activity against the target in vitro.

**Key output:** A hit series — multiple structurally related compounds showing target engagement.

**Failure mode:** Target is not actually causally linked to the disease (target validation failure); or the compound class is not drug-like (poor bioavailability, toxicity, or inability to reach the target tissue). Most targets fail here before significant spend.

**Decision point:** Is the target credible? Is there a tractable chemical series?

**Typical duration:** 2–4 years. No regulatory interactions at this stage.

---

## Stage 2 — Lead optimization and preclinical development

**What happens:** Medicinal chemists optimize the lead compound series for potency, selectivity, bioavailability, half-life, and metabolic stability. In vitro and in vivo (animal model) testing establishes:
- Pharmacokinetics (PK): absorption, distribution, metabolism, excretion (ADME)
- Pharmacodynamics (PD): what the drug does to the body at the target
- Toxicology: acute and subacute toxicity, genotoxicity (Ames test), safety pharmacology (cardiovascular, CNS, respiratory panels)
- Preliminary efficacy in relevant animal disease models

**Key output:** A single candidate molecule (development candidate / IND candidate) with a defined synthetic route, characterized safety profile in two animal species, and preliminary formulation.

**Failure mode:** Unacceptable toxicity in animal studies; poor PK (too rapidly metabolized, poor oral bioavailability, CNS penetration issues); off-target activity in safety pharmacology panels. A "clean" clinical candidate at this stage is rare — lead optimization cycles can take years.

**Decision point:** Is this compound safe enough and drug-like enough to put into humans?

**Regulatory note:** FDA does not review preclinical data in real-time. This data will be assembled into the IND package. For some high-risk programs, sponsors can request a pre-IND meeting (Type B) to discuss the preclinical package before filing.

**Typical duration:** 2–4 years.

---

## Stage 3 — IND filing

**What happens:** The sponsor files an Investigational New Drug application (21 CFR Part 312) with FDA. The IND package contains:
- Preclinical data: animal pharmacology, toxicology, ADME
- Manufacturing and quality information: synthesis, characterization, formulation, stability, purity specifications
- Clinical protocol: Phase I study design, patient eligibility criteria, dose escalation scheme, stopping rules
- Investigator information: qualifications of the principal investigator and clinical site
- Investigator's Brochure: comprehensive summary of preclinical data for investigators

**FDA review clock:** 30 days. If FDA does not issue a clinical hold within 30 days, the sponsor may proceed with the first study. A clinical hold requires the sponsor to address the hold before dosing.

**Failure mode:** Clinical hold — FDA identifies a safety signal (inadequate animal data, missing toxicology study, safety pharmacology concern) or protocol deficiency that must be resolved before human dosing. Also: inadequate CMC information for Phase I.

**IND goes into effect but is not "approved."** The IND is an application that becomes effective when FDA does not object — not an approval. This distinction matters for how agents characterize regulatory status.

**Ongoing obligation:** The IND remains active throughout development. Annual reports (IND annual report) are required. Protocol amendments must be submitted before implementing major changes. Safety reports (IND safety reports, SUSARs) must be filed within strict timelines when unexpected serious adverse events occur.

**Typical duration:** Preparation 6–12 months; FDA review 30 days.

---

## Stage 4 — Phase I: First-in-human

**What happens:** First administration of the investigational drug to human subjects. The primary goal is safety and tolerability, not efficacy. Typical design: single ascending dose (SAD) → multiple ascending dose (MAD), starting well below the estimated human equivalent of the no-observed-adverse-effect level (NOAEL) from animal studies.

Endpoints: dose-limiting toxicities (DLTs), maximum tolerated dose (MTD) or recommended Phase II dose (RP2D), PK parameters (Cmax, AUC, half-life), and initial PD biomarkers if available.

Subjects: usually 20–80. Healthy volunteers for most small molecule programs; patients for oncology (where exposing healthy volunteers to cytotoxic agents is unethical) or where PK is disease-modified.

**Failure mode:** Unexpected serious adverse event; PK shows the drug cannot reach therapeutic exposure; PD biomarkers show no target engagement. Oncology Phase I programs specifically look for signals of anti-tumor activity — absence of PD biomarkers is a major red flag.

**Decision point:** Is the drug safe at a dose that engages the target? Is there a plausible dose for Phase II?

**Regulatory note:** FDA can place the IND on clinical hold at any time if safety signals emerge. Phase I can be conducted at multiple sites under a single IND (multi-site Phase I).

**Typical duration:** 1–2 years.

---

## Stage 5 — Phase II: Proof of concept

**What happens:** The drug is tested in the patient population for the first time at scale. Two sub-phases:
- **Phase IIa**: Proof-of-concept — does the drug show a signal of efficacy at the selected dose in the target patient population? Usually 50–150 patients, shorter duration.
- **Phase IIb**: Dose-ranging — establishes the optimal dose(s) to bring into Phase III. 100–300 patients.

Key design elements: randomized (often vs. placebo or standard of care), controlled, with predefined primary endpoints tied to biological activity or clinical outcome. Biomarkers are critical here — failure to see movement in a validated PD biomarker is a signal to stop before Phase III.

**Failure mode:** No efficacy signal in the target population (most common cause of drug failure); unexpected safety findings at therapeutic doses; PK variability makes dosing unreliable; wrong patient population selected (biomarker-unselected trials that needed a biomarker-selected population).

**Decision point — end of Phase II meeting (Type B):** This is the most important regulatory interaction in the entire development program. The sponsor meets with FDA to discuss: the Phase III trial design, the primary endpoint, the patient population, the statistical analysis plan, and the CMC requirements for Phase III/NDA. FDA's written agreement on the Phase III design provides the maximum protection against a "failed trial" determination at NDA review. Skipping this meeting is a major strategic error.

**Typical duration:** 2–3 years.

---

## Stage 6 — Phase III: Pivotal trials

**What happens:** Definitive evidence generation. Phase III trials are large, randomized, controlled, and designed to the exact specifications agreed upon with FDA at the end-of-Phase 2 meeting. The primary endpoint established here forms the basis of the efficacy claim in the label.

Typical scale: 1,000–3,000+ patients per trial; often two adequate and well-controlled trials are required (the "two-trial rule"), though FDA may accept one pivotal trial with strong statistical evidence and supporting data. Multicenter, often multinational. Duration 3–5 years.

**Statistical design:** Pre-specified primary and key secondary endpoints; pre-specified analysis plan; alpha spending rules for interim analyses. Any deviation from the pre-specified plan that is not pre-registered creates risk at NDA review.

**Safety database requirements:** FDA typically expects at least 1,500 patients exposed to the drug (for a standard chronic-use indication) with at least 300–600 exposed for at least 6 months, and 100 for at least 12 months. Rare disease programs may be granted smaller databases based on disease prevalence.

**Failure mode:** Primary endpoint not met (the drug doesn't work, or the effect size is too small to achieve statistical significance at the designed sample size); safety finding that changes the benefit-risk calculation (new serious adverse event at scale, adverse event rate higher than predicted from Phase II).

**CMC escalation:** Phase III manufacturing must be at commercial-representative scale and the process must be validated or in process validation. The CMC data submitted at NDA/BLA will reference the Phase III manufacturing process. Changes to manufacturing after the NDA submission trigger comparability studies.

**Typical duration:** 3–5 years per trial; pivotal program often includes 2 trials running in parallel.

---

## Stage 7 — NDA/BLA submission and FDA review

**What happens:** The sponsor assembles the complete evidence package and submits to FDA in electronic Common Technical Document (eCTD) format. The NDA/BLA contains:
- **Module 1**: Administrative (labeling, patent and exclusivity certifications)
- **Module 2**: Summaries (quality overall summary, nonclinical overview, clinical overview, clinical summary)
- **Module 3**: Quality/CMC data
- **Module 4**: Nonclinical study reports
- **Module 5**: Clinical study reports

**FDA filing review (60-day clock):** FDA has 60 days after receipt to determine whether the submission is sufficiently complete to permit substantive review ("file" it) or to refuse to file (RTF). An RTF sends the application back — a major setback.

**FDA review clock:** 12 months (standard) or 6 months (priority review) from the filing date (day 60 after receipt). The review clock can be paused by a Major Amendment to the submission or by an Information Request.

**Advisory Committee (AdCom):** For many novel agents, FDA convenes a panel of outside experts to review the data and vote on approvability. AdCom votes are advisory, not binding — FDA has approved drugs over negative AdCom votes and rejected drugs with positive votes — but a negative AdCom is a serious setback.

**Facility inspections:** FDA conducts pre-approval inspections (PAIs) of manufacturing facilities during the review cycle. An unacceptable inspection (Form 483 observations, Warning Letter) stops the review.

**Possible outcomes:**
- **Approval:** Complete approval; the drug may be marketed. Labeling negotiations (what claims are permitted, what warnings are required, REMS if applicable) are part of the approval action.
- **Complete Response Letter (CRL):** FDA refuses to approve in the current cycle and specifies what must be addressed. The sponsor can resubmit (Class 1 resubmission: minor — FDA reviews in 2 months; Class 2: major — FDA reviews in 6 months).
- **Withdrawal:** Sponsor withdraws the application, typically when it becomes clear the approval will not be achieved on current terms.

**Typical duration:** Submission preparation 1–2 years; FDA review 6–12 months.

---

## Stage 8 — Post-marketing

**What happens:** Approval is not the end of the regulatory relationship. Post-marketing obligations include:

**Phase IV commitments:** Studies FDA required as a condition of approval. Common for pediatric populations (under the Pediatric Research Equity Act / Best Pharmaceuticals for Children Act), for long-term safety follow-up, or for conditions FDA identified as needing further characterization.

**Accelerated approval confirmatory trials:** If the drug was approved under accelerated approval based on a surrogate endpoint, the confirmatory trial demonstrating actual clinical benefit is required. FDA can withdraw approval if the trial is not conducted on schedule or if it fails. The 2022 Omnibus and subsequent FDA guidance strengthened FDA's authority to enforce timelines.

**REMS (Risk Evaluation and Mitigation Strategy):** Required when FDA determines a risk management program is needed to ensure benefits outweigh risks. Can range from a Medication Guide (a patient handout) to a REMS with Elements to Assure Safe Use (ETASU) — restricted distribution, mandatory enrollment of prescribers or pharmacies, mandatory monitoring. REMS modification or removal requires FDA approval.

**Pharmacovigilance:** Sponsors must maintain an adverse event reporting system. Serious unexpected adverse reactions must be reported within 15 days. Annual safety reports (Periodic Adverse Drug Experience Reports, PADERs; or for BLAs, Periodic Safety Update Reports, PSURs) are required. FDA can require labeling changes based on post-marketing safety data.

**Label updates:** FDA can require labeling updates (new warnings, new contraindications, new drug interactions) based on post-marketing signal detection or new studies. Sponsors can also request supplemental NDAs (sNDAs) to add new indications, which require their own evidence packages.

**Generic/biosimilar entry:** NCE exclusivity (5 years for new chemical entities) and orphan exclusivity (7 years) block generic ANDA filings and biosimilar 351(k) applications during the protected period. After exclusivity expiration, generic and biosimilar entrants are free to file.

---

## Decision point summary

| Stage | Gate question | If fail |
|---|---|---|
| Discovery → Preclinical | Is the target credible? Is there a tractable compound? | Stop program |
| Preclinical → IND | Is the compound safe enough for human dosing? | Optimize further or abandon |
| IND → Phase I | Did FDA allow clinical trials? | Address clinical hold |
| Phase I → Phase II | Is the drug tolerable at a dose that engages the target? | Stop or reformulate |
| Phase II → Phase III | Is there a proof-of-concept efficacy signal? | Stop or redesign |
| End-of-Phase 2 meeting | Did FDA agree on the Phase III design? | Redesign before committing Phase III spend |
| Phase III → NDA/BLA | Did the pivotal trial(s) meet the primary endpoint? | Stop, seek different indication, or renegotiate endpoint |
| NDA/BLA → Approval | Did FDA approve? | Address CRL and resubmit |
| Post-marketing | Are confirmatory trials complete (if accelerated approval)? | Risk of withdrawal |

---

## Typical timeline and cost

End-to-end (discovery through approval): **10–15 years**. Highly variable by indication (oncology tends to be faster due to priority review and surrogate endpoints; CNS and metabolic disease tend to be slower). Post-marketing obligations extend the regulatory relationship indefinitely.

Cost: estimates for full development of a single NME range from $1 billion to $2.6 billion (DiMasi et al., 2016, inflation-adjusted) when accounting for cost of capital and failed programs. The cost is dominated by Phase III and the high failure rate earlier in development — most cost is spent on drugs that do not reach the market.

The cost asymmetry is why pathway selection matters enormously: a 505(b)(2) program that avoids repeating Phase I/II for already-characterized safety/efficacy data can compress the timeline and cost by 30–50% relative to a full NDA for the same clinical profile.
<!--fold:a5682d@end-->
PORTDOWN_EFF280A4

# ── post ──
MARKER=$(awk '/^---$/ { f++; if (f==2) exit; next } f==1 && /^marker:[[:space:]]/ { sub(/^marker:[[:space:]]+/, ""); print; exit }' "$DEST")
[ -z "$MARKER" ] && { echo "seed: archive has no marker — corrupt" >&2; exit 1; }
awk -v m="$MARKER" -v outdir="$TARGET" '
  BEGIN {
    # Match <!--fold:<m>@file path="X"--> with an optional mode attr after
    # the path (fold emits  mode="644"  on executables).
    file_re = "^<!--fold:" m "@file path=\"([^\"]+)\"( mode=\"[0-9]+\")?-->$"
    end_re  = "^<!--fold:" m "@end-->$"
  }
  $0 ~ end_re { if (current) close(current); exit }
  $0 ~ file_re {
    if (current) close(current)
    line = $0
    sub(/^<!--fold:[^@]+@file path="/, "", line); sub(/".*$/, "", line)
    current = outdir "/" line
    dir = current; sub(/\/[^\/]*$/, "", dir)
    if (dir != current) system("mkdir -p \"" dir "\"")
    printf "" > current
    next
  }
  current { print >> current }
' "$DEST"
SEED_EXTRACTED=$(find "$TARGET" -type f -not -path "$DEST" 2>/dev/null | wc -l)
if [ "$SEED_EXTRACTED" = "0" ]; then
  echo "seed: archive contained no files — refusing to delete the source" >&2
  echo "  archive preserved at: $DEST" >&2
  exit 1
fi
rm -f "$DEST"

echo "" >&2
echo "✓ seed unpacked → $TARGET ($SEED_EXTRACTED files)" >&2
find "$TARGET" -type f | sort | while IFS= read -r _sf; do
  echo "  ${_sf#${TARGET}/}" >&2
done
echo "" >&2
if [ -f "$TARGET/SKILL.md" ]; then
  echo "This seed contains a skill (SKILL.md). Install it in your agent's skills directory." >&2
  echo "" >&2
fi
echo "Install the seed skill if not already installed:" >&2
echo "  https://seed.show/skill" >&2
echo "" >&2
echo "Publisher prompt:" >&2
sed 's/^/  /' >&2 <<'__SEED_PROMPT_END_AC1F2B__'
You are an expert US FDA pharmaceutical regulatory advisor. This seed contains three files: README.md (the orientation layer — mental model, what agents get wrong, what AI is changing, stable facts), workflow.md (the full drug development lifecycle from discovery through post-marketing with decision points and failure modes at each stage), and sources.md (authoritative FDA URLs to fetch before advising on any specific question). Start by reading README.md to calibrate your level, then fetch the relevant source before answering. Never imply certainty about FDA approval — 'FDA approved' is a specific legal determination, not a guarantee. Review timelines are PDUFA performance targets, not guarantees; fees change annually.
__SEED_PROMPT_END_AC1F2B__
exit 0