Placebo-controlled trial

In a placebo-controlled trial, participants are assigned to the drug or to a placebo that looks, feels and is taken the same way but contains no active ingredient. Comparing the two groups separates what the drug did from everything else that happens to people over the same period: lifestyle advice, attention from a study team, seasonal weight changes, side effects that would have happened anyway, and the natural drift of any measurement.

The three parts that make it work

• Randomisation. Chance decides who gets what. That balances the groups on everything, including factors nobody thought to measure.
• Blinding. In a double-blind trial, neither participants nor investigators know who is on what. That stops expectations from shaping the reporting of symptoms or the judgement of outcomes.
• Same everything else. Both groups get the same visits, tests and advice, so the only systematic difference is the drug.

What the placebo arm shows: STEP 1

STEP 1 randomised 1,961 adults with obesity or overweight, without diabetes, 2:1 to semaglutide 2.4 mg or placebo for 68 weeks, both with lifestyle intervention.

Two things stand out. Nearly a third of the placebo group lost at least 5 percent of their weight, from lifestyle support and attention alone. And 17 percent of people taking a dummy injection reported nausea. Without a placebo arm, all of that would have been credited to, or blamed on, the drug. The difference between the arms, 12.4 percentage points of weight (95% confidence interval 13.4 to 11.5), is the drug's effect.

The limits of blinding

Blinding is harder when a drug has obvious effects. Someone who has lost 15 percent of their weight, or feels sick in week two, may well guess their group. Trials soften this by relying on objective endpoints, such as weight measured at a clinic visit or heart attacks confirmed by a committee that does not know the allocation. Self-reported symptoms are the outcomes most open to expectation.

When there is no placebo

Some trials compare two active drugs head to head, like SURMOUNT-5 (tirzepatide against semaglutide), which answers a different question. Others have no control at all: everyone gets the drug and the result is compared with where they started. Those uncontrolled results look larger, because they include everything the placebo arm would have absorbed. Observational studies of people who chose a treatment are weaker still for cause and effect, because the people who chose it may differ in ways that matter.

For many research peptides, the evidence is animal studies, case reports and uncontrolled before-and-after numbers. A clear effect in those is a reason to run a placebo-controlled trial, not a substitute for one. The GLP-1 clinical trials page lists the placebo-controlled trials behind the approved drugs.

Your own log has no placebo arm, which is worth remembering. The dose log keeps doses, weight and symptoms on one timeline, so you can at least see when a change started and what else changed at the same time.

Related terms: clinical trial phases, estimand, confidence interval, surrogate endpoint. Back to the glossary.