Late-stage oncology trials can fail when treatment effects are diluted across heterogeneous patient populations. In this case study, Marc Buyse shows how IDDI helped improve the probability of success in a phase 3 non-small cell lung cancer trial by identifying patient subgroups most likely to benefit from treatment.
Looking to accelerate decisions in oncology prevention trials? Explore how early endpoints can support earlier, data-driven development decisions.
Working in heterogeneous or rare disease populations? See how early endpoint strategy can support regulatory approval.
Key Challenges Following the Phase 2 Results
- Uncertainty around which patient populations benefited from the vaccine
- Risk of diluting treatment effect across heterogeneous populations
- Difficulty translating phase 2 observations into phase 3 strategy
- High failure rates in late-stage oncology trials
The IDDI Approach
Drawing on deep expertise in adaptive design and biostatistics, IDDI:
- Embedded a Bayesian phase 2 confirmation within the phase 3 design
- Evaluated differential treatment effects across patient subgroups
- Enriched the phase 3 population based on observed response patterns
- Improved trial efficiency while maintaining the planned sample size
Sponsor Benefit
- Increase probability of success in phase 3 oncology trials
- Avoid exposing non-responding patients to ineffective treatment
- Improve trial efficiency without increasing patient numbers
- Support more informed, data-driven development decisions
Planning a phase 2 or phase 3 clinical trial?
Talk to our statistical experts about adaptive designs that improve success rates while maintaining rigor.
Read the full video transcript
A notable project in which IDDI provided key statistical advice was for the design of a phase 3 trial of a therapeutic vaccine for patients with advanced non-small cell lung cancer. A phase 2 trial had been carried out by the company and suggested that a subset of patients did not benefit from the vaccine, but these observations, although compelling from a statistical point of view, had to be replicated before being acted upon. What we suggested was a very innovative trial design, which embedded a phase 2 confirmation of the differential efficacy of the vaccine using a Bayesian approach, followed by a phase 3 trial aimed at confirming the effect of the vaccine either in all patients, or in the subset that benefitted the most. As it happened, the subset of non-responding patients was confirmed, which resulted in the phase 3 trial enriched in responding patients, and that vastly increased the chance of success of the trial even though the same total number of patients were included in the trial.
