Reference
Clinical Trial Glossary
Plain-language definitions of the research terms used across this site. Where a term matters for reading retatrutide evidence specifically, we say how.
Last reviewed: 24 October 2026
- Adverse event
- Any unwanted medical occurrence recorded during a trial, whether or not it was caused by the treatment. This is why trial tables list things like upper respiratory infection that occur at similar rates on placebo. A “side effect” implies causation; an adverse event does not. Causality is inferred by comparing against the control group.
- Blinding (masking)
- Keeping participants, investigators, or both from knowing which treatment someone received. Double-blind means both are masked, as in the retatrutide Phase 3 trials. Open-label means everyone knows — as in the retatrutide versus semaglutide diabetes study. Blinding matters most for subjective measures such as pain.
- Comparator
- What a treatment is measured against. A placebo comparator answers “does it do anything?” An active comparator — another real treatment — answers the more useful question, “is it better than what we already have?”
- Confidence interval
- A range expressing the uncertainty around an estimate. The practical rule for a ratio such as a hazard ratio: if the interval crosses 1.0, the result is compatible with no difference. Every cardiovascular confidence interval reported in TRIUMPH-3 crosses 1.0, which is why those data establish neither benefit nor harm.
- Dose escalation
- Starting at a low dose and increasing stepwise. In retatrutide Phase 3 trials, participants began at 2 mg weekly and increased every four weeks to their assigned target. It exists to manage tolerability and to let investigators observe each step. It is a research design feature conducted under supervision, not dosing guidance for individuals.
- Endpoint
- What a trial measures. The primary endpoint is the one the trial is designed and statistically powered to answer — it is the result the trial stands or falls on. Secondary endpoints are additional pre-specified measures. Co-primary endpoints means two must both be satisfied, as in TRIUMPH-4's pain and weight measures.
- Estimand
-
The precise definition of what a reported number represents, especially how it handles people who
stopped treatment. This is the single most useful term for reading this class of evidence.
- Efficacy estimand — the effect had participants remained on treatment.
- Treatment-regimen estimand — the average effect across everyone randomised, regardless of adherence.
- Expanded access
- A pathway through which an investigational medicine may sometimes be made available outside a clinical trial, typically for serious conditions where no alternative exists. It is operated by the sponsor under regulatory authorisation and normally initiated by a treating physician. It is not a consumer route and not a form of approval.
- Hazard ratio
- A comparison of how quickly events occur in two groups. Below 1.0 suggests fewer events in the treatment group; above 1.0 suggests more; 1.0 means no difference. A hazard ratio is only as meaningful as its confidence interval and the number of events behind it.
- Investigational medicine
- A substance under study that has not received marketing approval from a regulatory authority. It may lawfully be given to people only inside an authorised clinical trial under an approved protocol. There is no approved labelling, no approved dosing information, and no lawful general sale. Retatrutide is an investigational medicine.
- MACE
- Major Adverse Cardiovascular Events — a composite endpoint bundling several serious cardiovascular outcomes. MACE-3 typically covers cardiovascular death, heart attack and stroke; MACE-5 adds heart failure events and coronary revascularisation. Composites increase the number of events a trial can observe, but mean a result can be driven by its least severe component.
- Marketing approval
- A regulator's decision that a product may be sold for a stated medical use in that jurisdiction. Distinct from a clinical trial authorisation, which permits research only. Retatrutide has clinical trial authorisations in several countries and marketing approval in none.
- Multiplicity
- The statistical problem that testing many things makes a false positive more likely by chance. Trials control for it by pre-specifying a limited hierarchy of endpoints. When a result is marked “not controlled for multiplicity” — as several TRIUMPH-4 secondary endpoints are — it carries less statistical weight.
- Phases (1, 2, 3)
- Phase 1 — tens of participants; safety, tolerability and how the body handles the molecule. Phase 2 — hundreds; dose range and early signal of effect. Phase 3 — thousands; large randomised trials against a comparator, measuring pre-declared endpoints. Completing Phase 3 is not approval.
- Placebo
- An inactive treatment given to a control group. Placebo groups routinely report real symptoms and real improvements — in TRIUMPH-4 the placebo group reported a 2.4-point knee pain improvement. This is precisely why a control arm exists: it separates treatment effect from everything else happening in a trial.
- Post-hoc analysis
- An analysis performed after results are known, rather than planned in advance. Not worthless — it generates hypotheses — but much weaker than a pre-specified analysis, because with enough looks at a dataset something interesting always appears. We label post-hoc findings wherever we report them.
- Pre-specified
- Declared in the protocol before results are seen. Pre-specification is what stops researchers choosing the question after seeing the answer, and is a large part of why registered trials carry more weight than observational claims.
- Protocol
- The complete written plan for a trial, fixed before it begins: who is eligible, what they receive, what is measured and when, and how results will be analysed. How protocols work →
- Randomisation
- Assigning participants to groups by chance, so the groups are comparable at the outset — including on factors nobody thought to measure. The ratio states the proportions: TRIUMPH-1 used 1:1:1:1 across three doses and placebo; TRIUMPH-3 used 1:1:2.
- Topline results
- Headline findings released by a sponsor before full publication. Useful and usually accurate on the headline numbers, but not independently reviewed, and typically without the full data, subgroup detail or limitations that a peer-reviewed paper carries. We label these Sponsor announcement throughout.
A note on how we use these terms
We apply these labels consistently rather than decoratively. If a figure on this site comes from a post-hoc analysis, it says so. If a result was not controlled for multiplicity, it says so. If a confidence interval crosses 1.0, we say what that means rather than quoting the point estimate alone. Our editorial policy →