Research methods
Phase 2 vs phase 3 evidence
Why phase 2 trial results in metabolic peptides should not be conflated with phase 3 evidence. Sample size, population, statistical power and replication considerations.
A phase 2 result is a hypothesis-generating signal. A phase 3 result is the confirmatory test of that hypothesis. The strength of the inference is not the same.
Sample size and statistical power
Phase 2 trials are typically a few hundred participants and powered to detect dose-response, not population-level treatment effect with high precision. Phase 3 trials are typically in the thousands and powered against the regulatory standard for the indication.
Population: hand-picked vs realistic
Phase 2 populations are often more controlled, with stricter inclusion criteria; phase 3 populations are broader and closer to the eventual licensed indication. Effect sizes can compress between phase 2 and phase 3 simply because the population is more diverse.
Replication is the unsung hero
A single phase 2 trial result, however impressive, has not been replicated in the phase 3 programme. Replication across independent phase 3 studies is what justifies confident generalisation.
Implications for incretin coverage
Retatrutide's NEJM phase 2 weight reduction is a striking result. Until the TRIUMPH phase 3 programme reads out and is reviewed by regulators, it remains a phase 2 result. Until SURPASS-CVOT and SUMMIT-style outcome trials read out for any new molecule, weight efficacy does not equal outcome benefit.
Key takeaways
- Phase 2 magnitude is a hypothesis. Phase 3 magnitude is the test of that hypothesis.
- A new molecule should be discussed at its current evidence stage, not its anticipated stage.
- Headlines should specify whether a result is phase 2 or phase 3.
- Formal statistical training.
- Individual treatment, prescribing or dosing advice.
- Permission to skip the original publication of any trial cited.
Sources are tagged by class. Regulatory documents, peer-reviewed publications and trial registry records are preferred. Manufacturer disclosures and conference material are accepted where peer-reviewed publication is not yet available; preprints are flagged.