Part 1: The Problem & ADAPT

Metastatic Cancer: Incremental Progress

Luyendijk M et al. JNCI 2023

Despite Precision Oncology Advances

  • Metastatic cancer remains a leading cause of death
  • Survival gains have been modest
  • New medicines approved, but 5-year survival gains incremental

Despite more drugs, we haven’t fundamentally changed how trials learn.

Question: Why aren’t we seeing bigger gains?

Resistance: A Dynamic Process

Karagiannis D, Rampias T. Cancers 2022

Resistance Mechanisms Emerge Over Time

  • ESR1 mutations, PI3K pathway activation
  • Tumor microenvironment remodeling
  • Clonal evolution escapes initial therapy

Traditional trials fail to capture this

  • Fixed regimens miss interception
  • Baseline-only biomarkers can’t track evolution

Clinical Consequence: Delayed Intervention

Current Standard of Care

  • Treatment decisions fixed at baseline
  • Response assessed at predefined intervals
  • Course correction only after clinical progression

Result: Patients may remain on ineffective therapy while resistance mechanisms consolidate

What Patients Need

  • Earlier detection of treatment failure
  • Biomarker-guided therapy adaptation
  • Trials designed to learn during treatment

ADAPT Goal: Enable real-time treatment optimization based on evolving tumor biology

The Opportunity: ARPA-H ADAPT

March 2024: ARPA-H launches ADAPT program to address these gaps

Gaps in Current Trials

Gap Current → Needed
Design Fixed → Adaptive
Biomarkers Baseline → Continuous
Infrastructure Retrospective → Real-time
Translation Siloed → Embedded

ADAPT Goal: Dynamic Biomarkers

ADAPT: Three Technical Areas

MBC TA2 Team: TBCRC EVOLVE-BDT

TBCRC EVOLVE-BDT: Our Trial

EVOLVE-BDT: A Biomarker-Stratified Platform

  • Population: 2L metastatic breast cancer (ER+/HER2- and TNBC)
  • Biomarker stratification: Patients matched to subtrials by molecular profile
  • Resistance monitoring: Serial tumor and blood collection
  • TA1 integration: New arms added to test proposed resistance targets

Bayesian adaptive design enables early stopping + continuous learning (~700 patients)

Why Bayesian Adaptive Design?

Traditional vs Adaptive

Aspect Fixed Adaptive
Enrollment All pts Stop early
Analysis End only Continuous
Arms Static Add/drop
Savings None 30-50%

Enables: Early futility/efficacy stopping, adaptive randomization, biomarker-driven arm selection

What This Means for Your Patients

Traditional Fixed Trial

Mrs. Johnson, 58, ER+/HER2-, ESR1 mutation

  • Enrolls in standard Phase II (2L MBC)
  • Trial enrolls all 65 patients before analysis
  • Takes 32 months to reach enrollment target
  • Ineffective arm discovered only at study end

EVOLVE-BDT Adaptive Trial

Mrs. Johnson, 58, ER+/HER2-, ESR1 mutation

  • Enrolls in biomarker-matched sub-trial
  • Interim analyses after 15 PFS events accumulate
  • 92% chance of early futility stop if ineffective
  • Expected enrollment: ~46 patients vs 65

Bottom line: 19 fewer patients on ineffective therapy. Answers months sooner.

The Design Challenge

Once we established the scientific framework, we encountered a fundamental obstacle:

We could design the science. We could not design the trial.

Bayesian adaptive trials require 8+ interacting parameters—efficacy thresholds, futility boundaries, interim timing, sample size constraints—each affecting regulatory compliance and patient outcomes.

Manual calibration proved intractable: weeks of iteration, inconsistent results across statisticians, no guarantee of optimality.

We needed a systematic approach to navigate this complexity.

Part 2: BATON - AI-Powered Calibration

The Calibration Challenge: Four Competing Goals

Designing adaptive trials requires achieving ALL FOUR simultaneously

Out of 1,000 possible designs, only ~70 satisfy all four requirements → AI finds the needle in the haystack

BATON: Intelligent Search Through Design Space

70% fewer evaluations — minutes instead of days

Traditional Calibration vs BATON

Manual Approach

Aspect Limitation
Process Trial-and-error
Duration Days to weeks
Scope One at a time
Reproducibility Person-dependent
Documentation Often incomplete
Regulatory risk Variable

BATON Approach

Aspect Capability
Process Systematic
Duration Minutes to hours
Scope 1000s evaluated
Reproducibility Deterministic
Documentation Full audit trail
Regulatory risk Documented

Designs that would have taken months to discover—or never been found at all

BATON = Bayesian Adaptive Trial OptimizatioN — uses Bayesian optimization to efficiently search complex design spaces

BATON: Fewer Patients, FDA-Compliant Designs

Bottom line: BATON finds designs that spare ~25 patients/arm from ineffective treatment

Each point = one candidate design evaluated (~5,000 simulated trials per design)

An Unexpected Finding

BATON revealed a hidden trade-off with real patient consequences:

Minimizing enrollment caps can prevent patients from learning sooner that their treatment works.

Caps set too low sacrifice early success declarations—patients wait longer for answers.

Why this matters: BATON surfaced this before launch, recommending a slightly higher cap that preserves early efficacy stopping while still protecting against futility.

Summary

The Big Picture: Real Impact

Design calibration: minutes, not days | Documentation: complete FDA audit trail | Reproducibility: 100%

Key Takeaways

Thanks to Our Team and Funders

Resources

  • BATON R package
  • evolveTrial R package
  • Manuscript in prep

Contact

naim@unc.edu

EVOLVE-BDT: A True Testament to Team (Data) Science