thebench.bio #1 Ecosystem for the Bio-Tech Industry
All Clinical Trial Design questions

Question 8 of 9 Intermediate Troubleshooting

Enrollment is lower than expected and event rates change. What design implications would you evaluate?

Short answer

Falling recruitment and changing event rates affect power, feasibility and interpretability. Reassess with prespecified, governance-controlled methods rather than opportunistically changing objectives.

Why

In event-driven studies, power relates more closely to accumulated informative events than to enrollment count alone. Falling baseline risk, changes in follow-up and competing risks can delay the endpoint. Revising sample size after looking at unblinded treatment differences may inflate bias or type I error if not accounted for.

How to work it through

  1. Audit screened-to-enrolled conversion, geographic or site bottlenecks, endpoint ascertainment and event counts against blinded assumptions.
  2. Model alternative timelines and feasible increases in sites or follow-up.
  3. Consider prespecified blinded sample-size reassessment or an adaptation reviewed by independent oversight, ensuring alpha control and operational firewalls.
  4. Document protocol and statistical-analysis amendments before relevant unblinding.

An example

A cardiovascular study expected a 12% control event rate but blinded pooled incidence suggests 7%. Extending follow-up or adjusting event targets may be scientifically reasonable under preplanned rules. Switching the primary endpoint after inspecting arm-specific results is not.

Where the answer stops holding

Longer follow-up can change the clinical estimand and expose competing events. Recruitment incentives may alter case mix and site quality. More subjects do not automatically fix differential missingness.

A strong answer covers

  • Falling recruitment and changing event rates affect power, feasibility and interpretability
  • Reassess with prespecified, governance-controlled methods rather than opportunistically changing objectives
  • Audit screened-to-enrolled conversion, geographic or site bottlenecks, endpoint ascertainment and event counts against blinded assumptions

Common mistakes

  • Confusing event rate with treatment-effect size
  • Using operational pressure as sole justification for unblinded redesign