Clinical Trial Design — From Fundamentals to FDA Submission

Phase 1 tests safety and dosing in a small group (20-80 patients), often in healthy volunteers or patients with advanced disease. The goal is to find the maximum tolerated dose (MTD) and characterize pharmacokinetics.

Phase 2 evaluates preliminary efficacy and further assesses safety in 100-300 patients. Phase 2a is often dose-finding; Phase 2b is a more rigorous efficacy signal study. This is the "go/no-go" decision point for most programs.

Phase 3 is the pivotal registration trial — large (300-3000+ patients), randomized, and designed to provide definitive evidence of efficacy and safety for FDA approval. One-sided alpha of 0.025 is standard for oncology.

Phase 4 (post-marketing) studies safety and effectiveness in the broader population after approval.

Population — who are the patients?
Intervention — what treatment is being tested?
Comparator — what is the control?
Outcome — what is the primary endpoint?
Time — what is the study duration?

PICOT is the standard framework for structuring a clinical research question. A well-defined PICOT drives every downstream design decision — from eligibility criteria to sample size to the statistical analysis plan.

Randomization assigns patients to treatment groups by chance, eliminating selection bias. Common schemes: 1:1 (equal allocation, most efficient), 2:1 (more patients on experimental arm), stratified randomization (balanced within key subgroups). Without randomization, you cannot tell if differences are due to the drug or patient characteristics.

Open-label: everyone knows the assignment. Common in oncology with different regimens.
Single-blind: patients don't know but investigators do.
Double-blind: neither knows — gold standard for reducing bias. Requires matching placebo.

Blinding matters most for subjective endpoints. For objective endpoints like OS, open-label is more acceptable.

A control group provides the counterfactual. Types: Placebo control (ethical when no effective standard exists), Active control (current best treatment, required when withholding treatment would be unethical), External/historical control (weaker evidence, sometimes used for rare diseases). FDA generally requires a concurrent control for registration trials.

The primary endpoint determines whether the treatment works. Choosing depends on: disease area (PFS for solid tumors, EFS for hematology), phase (ORR for Phase 2, PFS/OS for Phase 3), regulatory precedent, and clinical meaningfulness. A poorly chosen primary endpoint can doom an otherwise well-designed trial.

Overall Survival (OS): time to death. Gold standard but requires large N and long follow-up. Often key secondary with hierarchical testing.

Progression-Free Survival (PFS): time to progression or death. Most common Phase 3 primary endpoint for solid tumors. Requires BICR for registration.

Objective Response Rate (ORR): tumor shrinkage proportion. Phase 2 or accelerated approval. Quick but doesn't capture duration of benefit.

For registration: PFS by BICR is standard for most solid tumors. ORR for accelerated approval with confirmatory trial requirement.

BICR (Blinded Independent Central Review): independent radiologists blinded to treatment read tumor scans. FDA requires it for PFS-based registration because investigator assessments can be biased. Expensive and complex but essentially mandatory for Phase 3 oncology PFS endpoints.

A surrogate substitutes for a clinical endpoint (like OS). Examples: ORR for OS, pCR for breast cancer, MRD for hematologic malignancies. FDA accepts surrogates through Accelerated Approval — but a confirmatory trial showing actual clinical benefit is required post-approval.

DOR measures how long a tumor response lasts. Always secondary, never primary. Critical for single-arm trials where ORR is primary — a high ORR with short DOR is much less impressive than moderate ORR with durable responses. FDA looks at DOR closely for accelerated approval.

Four key inputs: 1. Effect size (e.g. HR 0.75 = 25% risk reduction), 2. Alpha (one-sided 0.025 for oncology registration), 3. Power (80% or 90%), 4. Variability. For time-to-event endpoints, the number of events drives the calculation, not just patient count. Tools like R's gsDesign package handle the math.

Two-sided 0.05: tests if treatment is different from control (better OR worse). Most non-oncology trials.
One-sided 0.025: tests only if treatment is better. Standard for oncology registration. Mathematically equivalent in terms of critical value, but FDA oncology guidelines explicitly specify one-sided 0.025.

Allows looking at data before trial completion — stop for efficacy, stop for futility, or continue. Multiple looks inflate type I error, so alpha spending functions control this. O'Brien-Fleming (most common) spends little alpha early, preserving most for final analysis. Pocock spends equally across looks.

HR compares event rates between groups. HR 1.0: no difference. HR 0.75: 25% risk reduction. HR 0.50: 50% reduction (very strong). Estimated from Cox proportional hazards model. The 95% CI matters — if it includes 1.0, the result is not statistically significant.

Testing multiple endpoints increases false positive risk. FDA requires adjustment. Methods: Hierarchical testing (test in order; if first fails, stop — most common in oncology), Bonferroni (divide alpha equally — simple but conservative), Graphical methods (alpha recycled between hypotheses — flexible, increasingly popular).

The most traditional Phase 1 dose escalation: enroll 3 patients at a dose; if 0/3 have DLT, escalate; if 1/3, enroll 3 more; if ≤1/6, escalate; if ≥2, dose is too toxic. Pros: simple, well-understood. Cons: inefficient, poor at finding true MTD. FDA now encourages model-based designs (CRM, BOIN).

CRM (Continual Reassessment Method): uses a statistical model updated after each cohort to assign the next dose closest to target toxicity rate. More efficient than 3+3.

BOIN (Bayesian Optimal Interval): as simple to implement as 3+3 but performs nearly as well as CRM. Uses pre-calculated decision boundaries. FDA has published a favorable review. Both treat more patients at optimal doses.

A DLT is a severe adverse event during the observation window (usually first cycle, 21-28 days) considered drug-related. Typical: Grade 4 hematologic toxicity lasting >7 days, Grade 3+ non-hematologic toxicity, any Grade 5 (death). The DLT definition drives dose escalation and MTD determination.

FDA's 2023 guidance was a major shift: MTD is not necessarily the optimal dose — explore doses below MTD. Randomized dose-comparison may be required in Phase 2. Dose optimization should be completed before Phase 3. FDA may require post-marketing dose optimization studies.

An IND (Investigational New Drug) application is required before any US clinical trial. Contains: CMC (manufacturing), nonclinical data (animal studies), clinical protocol, and investigator information. FDA has 30 days to review — no "clinical hold" means you proceed. The IND stays active throughout development.

Allows FDA approval based on a surrogate endpoint reasonably likely to predict clinical benefit. Requires a post-marketing confirmatory trial. Common surrogates: ORR (oncology), pCR (breast cancer), MRD (hematologic). Recent FDORA law (2023) gives FDA more authority to enforce timely confirmatory trials.

For drugs showing substantial improvement over existing therapies. Benefits: intensive FDA guidance, rolling review, potentially shorter timelines. Higher bar than Fast Track. Does not guarantee approval or change evidentiary standard — Phase 3 data still needs to be convincing.

Pre-IND: discuss nonclinical data and Phase 1 design. End-of-Phase 2 (EOP2): the most critical meeting — align on Phase 3 design, primary endpoint, and statistical plan. Pre-NDA/BLA: discuss the submission package. FDA provides written responses to briefing documents.

The SAP specifies exactly how data will be analyzed. Finalized before database lock. Covers: analysis populations (ITT, mITT, per-protocol, safety), primary analysis method, secondary analyses, multiplicity adjustment, subgroup analyses, and sensitivity analyses. FDA reviews it closely — changes after unblinding are not acceptable.

Defines precisely what treatment effect you're measuring. Five attributes: Population, Treatment, Endpoint, Intercurrent event handling, Summary measure. Strategies for intercurrent events: treatment policy (ITT-like), hypothetical, composite, while on treatment, principal stratum. Required in all new FDA submissions since ICH E9 R1 (2021).

ITT (Intent-to-Treat): all randomized patients in assigned group — primary population for registration. mITT: subset excluding those who never received treatment — FDA increasingly skeptical for primary analysis. Per-Protocol: completers without major deviations — sensitivity analysis. Safety: all who received at least one dose.

Allows pre-planned modifications based on interim data: sample size re-estimation, dose modification, population enrichment, endpoint modification. Key requirement: all adaptations must be pre-specified and type I error controlled (usually demonstrated by simulation). FDA's 2019 guidance provides the framework.

Tests one drug across multiple tumor types sharing a common molecular target. Example: BRAF inhibitor across melanoma, NSCLC, thyroid, CRC — all with BRAF V600E. Each basket analyzed independently or with Bayesian borrowing. FDA has approved drugs this way (larotrectinib for NTRK fusions regardless of tumor type).

A perpetual multi-arm trial where treatments are added and removed over time against a shared control arm. Benefits: shared infrastructure, efficiency. Examples: I-SPY 2 (breast cancer), RECOVERY (COVID-19). Challenges: complex governance, statistical methods for adding/dropping arms, regulatory pathway.

Stratification factors balance randomization for important prognostic variables: disease stage, prior therapy, geographic region, biomarker status. Rule: 2-4 factors maximum. Too many with too few patients per stratum causes imbalanced randomization. FDA often flags over-stratification.

An independent Data Safety Monitoring Board reviews unblinded data during the trial. Reviews interim analyses, recommends stopping for efficacy/futility/safety, monitors adverse events. Required for trials with interim analyses and most Phase 3 registration trials. Charter must be written before trial starts.

Crossover (control patients switching to experimental treatment after progression) dilutes the OS benefit. A trial can show clear PFS benefit but non-significant OS due to crossover. Statistical adjustments: RPSFT and IPCW methods. FDA accepts these as supportive but not primary analysis.

A Pre-IND meeting is a formal interaction with FDA before submitting your Investigational New Drug (IND) application. For small biotechs, this meeting is critical because it allows you to present your development plan, nonclinical data, and CMC strategy to the review division and receive written feedback. FDA's responses become de facto guidance for your program. The meeting helps avoid costly protocol amendments, reduces the risk of a clinical hold, and can accelerate your timeline by months. Most Pre-IND meetings are conducted via written response only (Type B), meaning FDA responds to your questions in writing within 60 days without a face-to-face meeting.

Key FDA Guidance: Formal Meetings Between the FDA and Sponsors or Applicants of PDUFA Products — covers meeting types (A, B, C), request procedures, briefing document format, and timelines.
Download PDF

Your Pre-IND briefing package should include: (1) a concise summary of your product (mechanism of action, indication, unmet need), (2) nonclinical pharmacology and toxicology data or plans, (3) CMC overview (drug substance, drug product, manufacturing), (4) a proposed Phase 1 clinical protocol synopsis, and (5) specific questions for FDA. Each question should be focused — FDA responds question-by-question. Avoid overly broad questions like "Do you agree with our development plan?" Instead, ask targeted questions such as "Is a 28-day repeat-dose GLP toxicology study in two species sufficient to support a first-in-human single ascending dose study?"

Key FDA Guidance: IND Meetings for Human Drugs and Biologics: Chemistry, Manufacturing, and Controls Information — details what CMC information FDA expects at each stage.
Download PDF

An IND submission for Phase 1 includes: Form FDA 1571 (cover sheet), Table of Contents, Introductory Statement and General Investigational Plan, Investigator's Brochure, Clinical Protocol (Phase 1 dose-escalation design, inclusion/exclusion criteria, safety monitoring), Chemistry, Manufacturing, and Controls (drug substance, drug product, stability, GMP compliance), Pharmacology and Toxicology (GLP tox studies, PK/PD data, safety pharmacology), and Previous Human Experience (if any). FDA has 30 calendar days to review; if no clinical hold is issued, you may proceed.

Key FDA Guidance: Content and Format of Investigational New Drug Applications (INDs) for Phase 1 Studies of Drugs — the definitive reference for what FDA expects in your initial IND.
Download PDF

The nonclinical package for FIH dosing typically includes: in vitro pharmacology (target binding, selectivity), in vivo efficacy models, safety pharmacology (cardiovascular, CNS, respiratory — ICH S7A/B), single-dose toxicity, repeat-dose GLP toxicology (duration depends on clinical duration — typically 28 days for short-term FIH studies), genotoxicity (Ames test, in vitro chromosomal aberration, in vivo micronucleus), and PK/ADME studies. For oncology, the requirements may be less stringent under ICH S9. The starting dose is typically derived from the NOAEL in the most sensitive species, converted to human equivalent dose (HED), with a safety factor applied.

Key FDA Guidance: ICH M3(R2) — Nonclinical Safety Studies for the Conduct of Human Clinical Trials and Marketing Authorization for Pharmaceuticals — the international standard for what nonclinical studies are needed and when.
Download PDF

An Exploratory IND (also called Phase 0) allows very early human studies with limited drug exposure — typically microdose PK studies or pharmacodynamic studies at sub-therapeutic doses. The advantage is significantly reduced nonclinical requirements: you may only need an extended single-dose toxicity study in one species instead of full repeat-dose GLP studies. This pathway is ideal for biotechs that want to test target engagement, biodistribution, or PK in humans before committing to a full IND program. However, the clinical study must have no therapeutic intent and very limited dosing.

Key FDA Guidance: Exploratory IND Studies — explains the reduced regulatory requirements, study designs, and nonclinical support needed for exploratory studies.
Download PDF

The traditional approach uses the NOAEL (No Observed Adverse Effect Level) from GLP toxicology studies: convert the animal NOAEL to a Human Equivalent Dose (HED) using body surface area scaling (FDA Guidance for Industry: Estimating the Maximum Safe Starting Dose), then apply a safety factor (typically 1/10 for most drugs, 1/6 for oncology under ICH S9). For biologics, MABEL (Minimum Anticipated Biological Effect Level) may be more appropriate, especially for immunomodulatory agents. FDA's Project Optimus initiative (2024 final guidance) also encourages dose optimization beyond just MTD — sponsors should consider efficacy-response relationships and select doses that maximize benefit-risk, not just the maximum tolerated dose.

Key FDA Guidance: Optimizing the Dosage of Human Prescription Drugs and Biological Products for the Treatment of Oncology Indications (Project Optimus, 2024 Final) — FDA's framework for dose optimization in oncology.
Download PDF

Yes. A sponsor-investigator is an individual who both initiates and conducts a clinical investigation — common at academic medical centers. The IND requirements are the same, but FDA provides specific guidance on responsibilities: the sponsor-investigator must ensure GMP compliance for the investigational product, submit annual reports, report safety information, and maintain regulatory oversight. For small biotechs partnering with academic investigators, understanding the distinction between a commercial IND and an investigator-sponsored IND is critical for resource planning.

Key FDA Guidance: Investigational New Drug Applications Prepared and Submitted by Sponsor-Investigators — covers the unique regulatory responsibilities when the investigator is also the sponsor.
Download PDF