Stefano Garzon

Research · Oct 23, 2026

Clinical trial phases: the question each phase answers

Concept

Drug development proceeds through stages that answer different questions. The classical scheme uses phases I to IV. The International Conference on Harmonisation (ICH) links each phase to a study type: human pharmacology (phase I), therapeutic exploratory (phase II), therapeutic confirmatory (phase III), and therapeutic use (phase IV).

The boundaries are not rigid. A study can blend phase I into phase II, and pharmacology studies can occur in later phases. Surgical, behavioral, or lifestyle interventions rarely fit the sequence. For this reason Friedman and colleagues prefer the terms early phase studies and late phase trials.

Phase I: tolerability and dose

The phase I question is how much drug can be given safely and how the body handles it. Participants are usually healthy volunteers or patients who failed standard therapy. These studies estimate tolerability, pharmacokinetics, pharmacodynamics, bioavailability, and distribution.

A frequent goal is the maximally tolerated dose. In the classical 3+3 scheme, three participants receive a dose. Without unacceptable toxicity, the dose increases. With toxicity in one, three more are treated at the same dose. Further toxicity stops the escalation. The design assumes that the maximally tolerated dose causes unacceptable toxicity in about one third of participants. Bayesian methods, such as the continual reassessment method, update the dose estimate after each cohort.

Phase II: biological activity and the go/no-go decision

The phase II question is whether the drug has activity and whether it is worth taking to a confirmatory trial. The control may be concurrent, historical, or a before-and-after comparison. Several dose arms are common.

Participants are selected with narrow criteria. Outcomes are biomarkers, less severe clinical events, or minor safety signals. Phase II lacks power for major clinical outcomes. Its role is to refine the probability of success in phase III.

In oncology, Gehan’s two-stage design shows the logic. To rule out a response rate of 20% or less, 14 patients are enrolled. With no response, the drug is dropped. With any response, 10 to 20 more patients refine the estimate. Friedman notes that phase II studies are usually smaller than they should be.

Phase III: confirmatory efficacy

The phase III question is whether the intervention improves clinical outcomes compared with a control, by how much, and with which adverse effects. This is the randomized controlled trial that supports practice and approval.

Even large trials have limits. Follow-up is short relative to chronic use. Size ranges from hundreds to tens of thousands of participants, while post-approval use reaches millions. Rare or late adverse events can go undetected.

Late phase trials also correct established practice. CAST tested suppression of ventricular ectopy after myocardial infarction. With encainide or flecainide there were 43 arrhythmic deaths versus 16 with placebo. CRASH randomized 10,008 adults with head injury. Two-week mortality was 21.1% with corticosteroids and 17.9% with placebo (RR 1.18; 95% CI 1.09–1.27). In AIM-HIGH (3,414 patients) and HPS2-THRIVE (25,673 patients), niacin did not reduce vascular events in statin-treated patients.

Phase IV: wide use and late safety

The phase IV question is what happens when the intervention is used in broad populations over long periods. It includes post-approval studies and surveillance. Examples cited by Friedman: COX-2 inhibitors showed cardiovascular harm in large colon cancer prevention trials; thiazolidinediones were linked to heart failure after marketing.

Devices and procedures need additional attention. Results depend on operator skill, and trials usually enroll experienced operators.

Common errors when reading a paper

What to check in a trial

  1. The stated phase and whether the design matches that phase’s question.
  2. The type of control: randomized concurrent, historical, or before-and-after.
  3. The primary outcome: clinical or biomarker.
  4. Follow-up duration relative to the expected duration of use in practice.
  5. Trial size and the ability to detect infrequent adverse events.
  6. Whether post-approval or phase IV studies exist for the same intervention.

Related posts: power and event rates, interim analyses.

References