Drug trials are described as phase one, two and three, which sounds like a ladder of increasing certainty. Each phase is really a different question, and the ordering follows from how much risk is acceptable at each point.
Laboratory work sets the starting dose
Before any person is involved, a candidate compound is studied in cells and in animals to establish how it behaves in a body and at what exposure harm appears.
Those results are what justify a first human dose, which is set deliberately low relative to anything expected to be active.
This stage cannot establish benefit in humans, and it is not designed to.
Phase one asks what the body does to the drug
The first human phase involves a small group, often healthy volunteers, and its questions are about safety, tolerability and handling.
Investigators track absorption, distribution, metabolism and elimination, and escalate dose cautiously while watching for adverse effects.
Effectiveness is not the object here. The group is too small and, where volunteers are healthy, there is no condition present to improve.
Phase two looks for a signal and a dose
The second phase enrols people who have the condition in question, in the low hundreds, and asks whether the drug appears to do anything useful.
It also narrows the dose, since the range that is tolerable may be wider than the range that is effective.
Many compounds stop here, and that is the phase's function: to fail candidates before the expensive stage begins.
Phase three is the comparison that counts
The third phase is large and randomised, comparing the drug against a placebo or against existing standard treatment, usually across multiple sites.
Size matters for two reasons. It allows a modest true effect to be distinguished from chance, and it exposes uncommon adverse effects that a few hundred participants would never reveal.
Randomisation and blinding are what stop expectation, on either side, from shaping the recorded result.
Monitoring continues after approval
Effects that occur rarely, or only after long exposure, or only in groups underrepresented in trials, can remain invisible through phase three. A trial of a few thousand people cannot reliably detect something that happens once in fifty thousand.
Post-marketing surveillance collects reports from ordinary clinical use, where the population is larger, less selected and often taking other medicines alongside.
This is why the safety information attached to an approved medicine continues to change over its lifetime, and why an approval is better read as a decision made on the evidence available than as a final verdict.