What do clinical trial phases actually mean?
Each phase answers a different question, and the questions get progressively harder — which is why most drugs that enter testing never reach patients.
Preclinical. Laboratory and animal studies before any human receives the compound. Establishes basic safety and plausible activity.
Phase I — is it safe, and what dose? Typically 20–100 participants, frequently healthy volunteers, though in oncology they are usually patients. The aim is not to see if the drug works. It establishes tolerability, side effects, and how the body absorbs, distributes and eliminates it, usually by escalating the dose cautiously until unacceptable effects appear.
Phase II — does it appear to work, and at what dose? Several hundred participants who have the condition. Looks for a treatment effect and refines dosing. This is where most drugs fail, because promising biology frequently does not translate into benefit.
Phase III — is it better than what we already have? Hundreds to thousands of participants, across multiple sites, usually randomised, controlled and blinded, compared against existing standard treatment or placebo. This is the evidence regulators require. It is large, long and expensive precisely because detecting a modest real effect reliably requires it.
Phase IV — what happens in the real world? Post-marketing surveillance after approval. This catches rare side effects, effects in populations under-represented in trials, long-term outcomes, and interactions. Drugs have been withdrawn at this stage, which is why it is a genuine phase and not an afterthought.
Why the design features matter:
Randomisation prevents selection differences between groups.
Blinding prevents expectation influencing outcomes, from both the participant and the assessor — hence double-blind.
A control group is what allows any effect to be attributed to the drug rather than to time, attention or regression to the mean.
Common misreadings: a Phase I result says nothing about effectiveness; a positive Phase II result frequently does not survive Phase III; and surrogate endpoints — a biomarker moving — are not the same as patients doing better.