Improved statistical method for early cancer drug trials better identifies safe and active doses while protecting patients

Share
Improved statistical method for early cancer drug trials better identifies safe and active doses while protecting patients

Improved statistical method for early cancer drug trials better identifies safe and active doses while protecting patients

Early-phase (Phase I) cancer trials are designed primarily to find the maximum tolerated dose of a new drug. But the FDA's Project Optimus initiative has pushed for reform, recognizing that the highest tolerated dose isn't always the most effective or sustainable for patients.

This paper extends an existing Bayesian trial design called BOLD with a 'backfill' component that continuously evaluates drug activity at lower doses while dose escalation proceeds. Instead of just finding the top safe dose, BF-BOLD also assesses which doses below that threshold actually show biological activity — without requiring complex statistical modeling.

Simulations show the backfill design improves both safety (fewer patients receiving overdoses) and activity assessment, leading to better-characterized recommended Phase II doses.

Key Findings

  • BF-BOLD extends the Bayesian Ordered Lattice Design with per-patient activity evaluation during dose escalation
  • Method identifies optimal biological dose (OBD) by assessing activity rates below the MTD
  • Simulations show improved safety and activity assessment compared to conventional designs
  • Backfill design reduces overdose rates in early trials
  • Approach is straightforward and does not require complex statistical modeling

Implications

Better Phase I trial design has downstream effects on all of drug development — getting the right dose into Phase II improves trial success rates, reduces unnecessary patient toxicity, and aligns with FDA reform goals. This has broad applicability across oncology drug development, particularly for novel targeted agents and immunotherapies where activity vs. toxicity trade-offs are complex.

Caveats

Preprint — not peer reviewed. Evidence is simulation-based, not from a real trial. Real-world implementation depends on trial sponsor and regulatory acceptance. Complexity of backfill logistics (patient enrollment, dose administration) not fully addressed.

Source: medRxiv — 2026-04-06

Read more