Phase 3 Trial

Phase 3 is the confirmatory stage. The design is fixed in advance, the population is broad enough to resemble the intended market, and the trial is powered for a single prespecified primary endpoint with a controlled type I error rate. Regulators have historically expected two adequate and well-controlled trials, or one exceptionally persuasive trial with supporting evidence, and these are the studies that populate the efficacy and safety sections of the approved label.

Size follows the endpoint. STEP 1 randomised 1,961 adults with obesity for 68 weeks and reported roughly 15 percent mean weight reduction on semaglutide against roughly 2 percent on placebo, because the outcome is continuous and the effect is large. SELECT randomised 17,604 people with cardiovascular disease and overweight or obesity but not diabetes and ran for years, reporting a roughly 20 percent relative reduction in major adverse cardiovascular events in 2023, because it had to accumulate rare events rather than measure a difference in every participant.

Phase 3 gives the effect estimate you should actually quote, with a confidence interval wide enough to be honest and narrow enough to be useful, and it assembles the first safety database counted in thousands of patient-years. Its limits are structural rather than accidental. Even a trial this size cannot detect an adverse event occurring once in ten thousand, and its exclusion criteria have already removed much of the population that will later be prescribed the drug.

Phase 3 is routinely used as a synonym for approved, which it is not. A compound can be in Phase 3 for years, and a completed Phase 3 can miss its primary endpoint outright. Subgroup results from a Phase 3, however plausible the biology, remain exploratory unless they were prespecified within the testing hierarchy.