Clinical Trials: Phases, Design, and Regulatory Oversight in the US

Before a drug reaches a pharmacy shelf, it passes through a structured gauntlet of human testing that can span 10 to 15 years and cost upward of $1 billion (Tufts Center for the Study of Drug Development). Clinical trials are the mechanism that makes that passage legible — to regulators, to patients, and to science itself. This page covers the phase structure of US clinical trials, the design principles that determine what a trial can actually prove, and the federal oversight architecture that governs the entire process from first-in-human to post-market surveillance.

• Definition and scope
• Core mechanics or structure
• Causal relationships or drivers
• Classification boundaries
• Tradeoffs and tensions
• Common misconceptions
• Checklist or steps
• Reference table or matrix

Definition and scope

A clinical trial is a prospective study that assigns human participants to one or more interventions — a drug, a device, a behavioral protocol, a surgical procedure — in order to evaluate the effects on health-related outcomes. The US Food and Drug Administration, whose definition governs domestic regulatory authority, distinguishes clinical trials from observational studies on exactly this point: randomization and assignment, not just observation (FDA, "What Are Clinical Trials and Studies?").

The scope is broad. The ClinicalTrials.gov registry, maintained by the National Library of Medicine, verified more than 470,000 registered studies as of 2023, spanning oncology, infectious disease, cardiovascular conditions, rare pediatric disorders, and mental health. Not all of these are drug trials — device trials, diagnostic trials, and behavioral intervention trials operate under the same basic framework but with regulatory nuances specific to their product category.

The practical boundary of "clinical trial" matters because it determines which federal rules apply. Trials that meet the FDA's definition of an Investigational New Drug (IND) application trigger a specific regulatory pathway. Trials of medical devices engage a separate premarket approval or 510(k) process. Both categories require institutional review board oversight under the Common Rule (45 CFR 46), which is the federal regulation protecting human research subjects.

Core mechanics or structure

The phase system is the skeleton of drug development. Each phase answers a different question, and the answer to one question is the prerequisite for asking the next.

Phase 0 is a recent addition — small (fewer than 15 participants, typically), designed purely to confirm that a drug behaves in humans the way animal models predicted. Sub-therapeutic doses are used. No therapeutic intent. Phase 0 exists mostly to catch pharmacokinetic surprises early and cheaply.

Phase 1 is where safety lives. Enrolling 20 to 100 healthy volunteers (or, in oncology, patients who have exhausted other options), Phase 1 trials establish the maximum tolerated dose, characterize how the body absorbs and eliminates the drug, and document the early adverse event profile. Roughly 70% of drugs that enter Phase 1 advance to Phase 2 (FDA Drug Approval Process).

Phase 2 introduces efficacy into the equation. Hundreds of patients with the target condition receive the drug. The trial is still primarily about safety and dosing — but now researchers are watching for a signal that the intervention actually works. About 33% of drugs that enter Phase 2 make it to Phase 3.

Phase 3 is the pivotal trial. This is the randomized, controlled, often double-blind study that regulators require before approving a new drug. Sample sizes run from hundreds to tens of thousands of participants, depending on the disease. Phase 3 trials are expensive, logistically complex, and the point where most of the industry's attrition occurs. Only about 25 to 30% of drugs entering Phase 3 eventually reach approval (FDA).

Phase 4 begins after approval — post-market surveillance that monitors long-term safety in a real-world population far larger and more diverse than any Phase 3 trial enrolled. Some Phase 4 commitments are required by the FDA as a condition of approval.

The randomized controlled trial (RCT) design — random assignment to treatment or control, blinded outcome assessment — is the evidentiary standard for Phase 2 and Phase 3. Adaptive trial designs, which allow pre-specified modifications to the protocol based on accumulating data, have grown more common and are explicitly acknowledged in FDA guidance documents (FDA Adaptive Designs for Clinical Trials of Drugs and Biologics Guidance, 2019).

Causal relationships or drivers

The phase system's logic is causal: harm must be bounded before efficacy can be evaluated, and efficacy must be demonstrated before a drug's risks become societally acceptable at scale. This sequence is not arbitrary bureaucratic layering — it reflects a genuine epistemic order.

The randomization requirement exists because human disease is not randomly distributed. Patients who choose to enroll in trials, or who are selected by physicians, differ systematically from those who are not. Without randomization, observed differences in outcomes between treated and untreated groups are confounded by those selection differences. Randomization is the mechanism by which a trial earns the right to make a causal claim rather than an associational one. This connects directly to broader principles covered in research design and methodology.

Blinding drives the same logic. Placebo effects are real — response rates in placebo arms of clinical trials for depression, pain conditions, and irritable bowel syndrome frequently exceed 30% (Colloca and Miller, "The Nocebo Effect and Its Relevance for Clinical Practice," Psychosomatic Medicine, 2011). If patients know they received the active drug, they report better outcomes. If investigators know, they may unconsciously rate outcomes differently. Double-blinding removes both biases simultaneously.

Sample size is driven by statistical power — the probability that a trial will detect a true effect if one exists. FDA guidance generally requires 80% to 90% power, which, combined with the expected effect size and the variability of the outcome measure, determines enrollment targets. A trial underpowered for its primary endpoint is not just inconclusive — it is also arguably unethical, because it exposes participants to risk without generating valid knowledge. This tension between statistical rigor and participant burden sits at the center of research ethics and integrity.

Classification boundaries

Not every study that looks like a clinical trial is regulated like one. The key distinctions:

• IND vs. non-IND trials: Trials of marketed drugs used within their approved labeling may not require an IND. Trials of unapproved indications, unapproved doses, or novel combinations do.
• Drug vs. device trials: Device trials operate under the Investigational Device Exemption (IDE) pathway, not the IND pathway. The risk classification of the device (Class I, II, or III) determines IDE requirements.
• Interventional vs. observational: Observational studies assign no intervention — they watch what happens. They are subject to IRB review but not to IND requirements.
• Humanitarian Use Device (HUD): A separate pathway for devices targeting fewer than 8,000 US patients annually, requiring effectiveness evidence but not the full efficacy proof of a pivotal trial (FDA, Humanitarian Device Exemption).
• Expanded Access (Compassionate Use): Not technically a trial — allows patients outside trial eligibility to access investigational drugs, typically when no alternatives exist.

The scientific method explained at a conceptual level maps onto these categories: the degree of experimental control determines the strength of causal inference the study design can support.

Tradeoffs and tensions

Three fault lines run through clinical trial design, and none of them resolve cleanly.

Internal validity vs. external validity. A tightly controlled RCT with strict inclusion criteria produces clean, interpretable results — but the population enrolled may bear little resemblance to patients who will eventually use the drug. Older adults, pregnant women, and people with multiple comorbidities have historically been underrepresented in pivotal trials. The FDA's diversity action plan guidance (2022) attempted to address this by requiring sponsors to submit enrollment diversity plans, but implementation remains a work in progress.

Speed vs. certainty. Accelerated approval pathways — Breakthrough Therapy designation, Fast Track designation, Accelerated Approval — allow drugs to reach patients faster by accepting surrogate endpoints (a tumor shrinkage rate instead of overall survival, for example). The tradeoff is that the confirmatory Phase 4 trial, which is supposed to validate the surrogate endpoint against real clinical outcomes, has historically been delayed or incomplete. The FDA Omnibus Reform Act of 2022 gave the FDA explicit authority to require these confirmatory trials and withdraw approval if they are not completed (FDA, Omnibus Reform Act provisions summary).

Industry funding vs. independence. Approximately 70% of clinical trial costs in the US are funded by pharmaceutical and device companies (NIH, Estimates of Funding for Various Research, Condition, and Disease Categories). Industry-sponsored trials are more likely to report positive results than independently funded trials — a finding documented across meta-analyses of oncology and psychiatric drug trials. This is not necessarily fraud; it may reflect selective trial initiation, publication bias, or endpoint selection. The existence and scale of industry funding sits at the heart of discussions around conflict of interest in research.

Common misconceptions

"Phase 3 approval means a drug is safe for everyone." Approval means the benefit-risk profile was acceptable in the studied population under the studied conditions. Rare adverse events — those occurring in fewer than 1 in 1,000 patients — are structurally invisible in a trial of even several thousand participants. Phase 4 surveillance is not a formality.

"Placebo arms deny patients treatment." In trials where an effective standard of care exists, placebo-only controls are generally considered unethical under the Declaration of Helsinki. The standard design compares the new drug to the best available existing treatment. Placebo arms are used only where no proven treatment exists for the condition.

"A randomized trial always produces definitive results." A single RCT, however well-conducted, is one data point. Replication across different populations, different research teams, and different settings is what generates scientific consensus. This connects to broader issues in the replication crisis in science.

"Adaptive trials are less rigorous." Pre-specified adaptations — those written into the protocol before enrollment begins — do not compromise validity. They require more sophisticated statistical planning and FDA pre-agreement, but they can be more efficient and sometimes more ethical than fixed designs.

Checklist or steps

The IND-to-Approval sequence — key regulatory milestones:

1. Preclinical studies completed — animal toxicology, pharmacokinetics, pharmacodynamics data assembled.
2. IND application submitted to FDA — includes preclinical data, proposed clinical protocol, investigator qualifications, manufacturing information.
3. FDA 30-day review window — FDA may place a clinical hold within 30 days; silence after 30 days constitutes implicit approval to proceed (21 CFR 312.40).
4. IRB approval obtained — required before any participant is enrolled, per 45 CFR 46.
5. Phase 1 initiated — dose-escalation, safety monitoring committee established.
6. End-of-Phase 2 meeting with FDA — optional but strongly recommended; aligns Phase 3 design with regulatory expectations.
7. Phase 3 initiated — pivotal trial protocol pre-agreed with FDA.
8. New Drug Application (NDA) or Biologics License Application (BLA) submitted — complete dataset including all trials, all adverse events, proposed labeling.
9. FDA review — standard review is 10 months; Priority Review is 6 months from the date of filing.
10. Post-market surveillance plan activated — Phase 4 commitments, Risk Evaluation and Mitigation Strategy (REMS) if required.

Reference table or matrix

US Clinical Trial Phase Summary

Regulatory Designations and Their Effect on Timeline

The full landscape of how clinical trials fit within broader scientific research infrastructure — funding mechanisms, institutional roles, data governance — reflects just how many systems have to coordinate before a treatment becomes available. The trial itself is the visible tip of an iceberg whose base includes federal research funding agencies, peer review processes, and the unglamorous work of research data management.