Planning under uncertainty

AI-guided design of efficient clinical trials with BATON

Naim Rashid, PhD

Associate Professor, Department of Biostatistics, Gillings School of Global Public Health
Member, Lineberger Comprehensive Cancer Center · UNC Chapel Hill

NextGen TriBix Symposium · November 17, 2026

UNC, Lineberger, and TBCRC logos.

Section 1

1The stakes
What do we need to learn before committing?
2The design problem
3BATON
4BATON in practice
5Looking to the future

Drug development usually follows three phases before approval

Phase I
First trials in people. Is it safe, and at what dose?
→
Phase II
Is there a sign that it works?
→
Phase III
Often the pivotal trial. Is there convincing evidence of benefit?
→
Approval
Fewer candidates continue. Schematic.
Phases can overlap or be combined.
Where a Phase II–III gap exists, it averages 17 months.
Tufts CSDD · non-expedited approvals, 2008 to 2018 (Getz 2021)

The U.S. is concerned about its lead in early clinical research

U.S. Department of Health and Human Services

HHS reports that China’s global share of Phase I trials first exceeded the U.S. share in 2021.

Operation TrialBlazer roadmap, 2026

Australia · adult solid-tumor Phase I trials

Annual adult solid-tumor Phase I trial counts at Australian sites rose from 18 in 2012 to 75 in 2022. Only endpoints are plotted.

Adult solid-tumor Phase I trials with an Australian site. Hitchen et al., 2024

Developing one new therapy is slow and expensive

10.5 years
from Phase I to approval, based on successful transitions
$1.1 billion
median cost per approved drug, failed candidates included (2018 dollars)

Each year of development is a year patients wait.

BIO, Informa, QLS Advisors 2021 · Wouters, McKee, Luyten, JAMA 2020, corrected 2022

Many problems become apparent only in late-stage trials

A 10 by 10 grid of dots with 54 filled in navy and 46 in light gray.
54%
failed in late-stage development
640 therapeutics that entered pivotal trials in 1998–2008
Most often the drug did not show enough benefit (57% of failures). Hwang et al., JAMA Internal Medicine 2016

What can we do in the early trials to make a late failure less likely?

Late failures have many causes. Earlier-phase trials can also affect later-phase trials in ways that are hard to predict.

Early decisions become downstream commitments

Sotorasib · KRAS G12C lung cancer
Phase I960 mgHighest tested
Accelerated approval960 mg
Postapproval comparison
960 mg vs 240 mg
FDA-required
Phase II results 960 mg 240 mg
Progression-free survival 5.4 months 5.6 months
Severe side effects (grade 3+) 35.6% 19.2%
Tumor response 32.7% 24.8%
Descriptive comparison Trial · FDA
The highest dose patients tolerate, or the lowest dose with full benefit?

How best to design early trials for later success is not yet clear

More patients at 960 mg cannot show how 240 mg compares.

Different measures Schematic in two rows. A small early trial reads out a short-term signal, tumor shrinkage and side effects. A large later trial reads out a long-term outcome, survival. A dashed arrow from the short-term signal to the long-term outcome asks whether one predicts the other.

Separate designs
Each phase has its own targets.

Too many possible plans
Patients per phase, and when to move on.

More patients later may not answer what the early trial never asked.

Better early trials are possible. Designing them is the hard part.

So far
10.5 years
from Phase I to approval
54%
of late-stage candidates failed
960 mg
chosen early, compared with 240 mg only after approval
Trials that adapt as the data arrive can learn more from fewer patients.

One bottleneck is finding an efficient trial design that still meets the evidence requirements. That is a search problem.

Section 2

1The stakes
2The design problem
Is there enough evidence to advance the treatment?
3BATON
4BATON in practice
5Looking to the future

A clinical trial is an experiment built for a decision

?BenchmarkTreatment
The question
Does the treatment beat a benchmark or a control?
Single-armall get the treatmentRandomizedtreatmentcontrol
How patients are used
Everyone gets the treatment, or patients are assigned to treatment or control.
Does not workGOWorksSTOP
Two errors
Advance a treatment that fails, or drop one that works.

Two choices dominate: how many patients, and what evidence is enough to take the treatment to the next phase.

An adaptive trial can act on an interim result

Two schematic timelines. A traditional trial enrolls all patients then analyzes once. An adaptive trial enrolls in blocks separated by interim looks that can stop the trial early for futility, ending in a final go-or-stop analysis.

When should we look, and what evidence is enough to keep enrolling?

A good design has to meet four requirements at once

Must meet
1
Rarely advance a treatment that fails
Type I error at most 10%
2
Usually advance one that works
Power at least 80%
Then
improve
3
Enroll few patients
4
Stop early if it fails

Feasible means meeting 1 and 2. Calibration then improves 3 and 4.

Simulation tests one design in two scenarios

One design, held fixed in every simulated trial

Fixed: settings, rules, model and assumptions. Only the treatment effect changes.

Ineffective
(null)
target: at most 10%
2,100 of 50,000 say go
4.2%
type I error
Effective
(alternative)
target: at least 80%
41,000 of 50,000 say go
82%
power
Illustrative numbers

Both targets met, so this design is feasible.

One EVOLVE protocol can host 10 to 15 subtrials

EVOLVE is a platform trial. Each biomarker-defined group of patients gets its own subtrial.
EVOLVE master protocolMetastatic breast cancer
Three example subtrial designs
Single-arm
8 settings
Randomized
8 settings
Seamless
13 settings

Each subtrial needs its own decision rules, with eight or more interacting settings.

I could not verify that a hand-tuned design met the power target

Eight settings in the EVOLVE calibration example
The outcome is progression-free survival, analyzed with a Bayesian piecewise exponential model. Interim reviews can stop the trial if benefit looks unlikely.
Max sample size
Spacing of interim looks
Events needed before a decision
Futility threshold
Efficacy threshold
Margin over the benchmark
Strength of the prior
Resolution of the survival model
Changing one setting can change which values work for the others.

Target power: at least 80%

Several more subtrials were waiting. How could we search so many settings systematically?

A grid search can produce 100 million candidate designs

Two panels. The haystack: a field of 100 million possible designs, from 10 options for each of 8 design choices, each tested on thousands of simulated trials, with one small square marked. The zoom shows the efficacy threshold on the horizontal axis and the futility threshold on the vertical axis. Each cell is one pair of threshold values with the other six settings held fixed. Green cells meet both statistical targets. Outlined green cells have lower expected enrollment.

Think hyperparameter tuning for trial rules.
Which candidate is worth the next simulation?

Section 3

1The stakes
2The design problem
3BATON
How do we search intelligently?
4BATON in practice
5Looking to the future

BATON in one picture

BATON stands for Bayesian Adaptive Trial Optimization. It applies Bayesian optimization to trial rules.
You supply a trial simulator · an analysis model · a performance evaluator · targets and an objective
Start from a known design or specify your own trial
Space-filling start
Futility thresholdAn illustrative initial sample spreads tested designs across the design space.Efficacy threshold
Fit model
Model-guided search
Futility thresholdThe same starting designs are shown in grey. Five numbered evaluations make broad jumps before approaching the green region. The green region meets both statistical targets. A diamond marks a candidate for verification.Efficacy threshold
Two settings shown. Each dot is one design tested in many simulated trials.
Meets both targetsCandidate to verify
Simulate → Fit → Choose, and repeat. Then check the best design with fresh simulations.

A nonlinear model predicts how untested designs might behave

XDesign settings YSimulated outcomes Initial
sample
.3.9326 72%3%16.2 .2.8365 88%13%19.5 Guided
search
.3.8305 77%8%14.6 .3.8325 simulate → 82%8%15.1 Next design .3.8315 Pending
x1x2x3x4
PowerType I
error
Expected
enrollment
Simulate → Fit Y from X → Choose
Illustrative values
Illustrative power estimates for nine tested settings. Error bars show simulation uncertainty and a dashed line marks the 80% power target. The same points with a nonlinear prediction and uncertainty band. The 80% power target remains marked.

BATON balances potential improvement with feasibility

Which candidate is worth the next simulation?

Feasibility map with two design settings. The white diamond marks the model’s current favorite. An orange point near the uncertain boundary is another candidate to test.

Test the orange point. It could reduce enrollment and still meet both targets.

Simulate it, update the predictions, and choose again.

Section 4

1The stakes
2The design problem
3BATON
4BATON in practice
What did it find for a real trial?
5Looking to the future

Sanity check: a known answer tells us how close BATON gets

Scatter plot of expected enrollment when treatment is inactive, with the exact Simon design on the horizontal axis and BATON on the vertical axis. Both axes use the same scale. Four points sit on the equality line. The largest gap is 1.98 expected patients, or 13.5% above the optimum.
Simon’s two-stage design
Stop after a first group if too few respond.
Only four settings
Every possible design can be listed, so the best is known exactly.
Three ways to define best
Optimal: fewest patients on average. Minimax: smallest maximum. Admissible: a compromise.

BATON against Optimal: 3.6% more patients on average.

4 exact matches (dark points). All 12 met both targets.

EVOLVE designs have too many settings to try every one

Simon: 4 settings. Every design can be listed.
EVOLVE (the breast cancer platform trial): 8 to 13 settings. Listing every design is not possible.
One randomized EVOLVE subtrial, 8 settings. The score blends max sample size and expected enrollment. Lower is better.
BATON
model-guided search
51.3
score
76 core-hours
Grid, same effort
2 values per setting, 256 designs
101.3
score
matched to BATON
Grid, finer
3 values per setting, 6,561 designs
70.5
score
3,594 core-hours

A coarse grid does worse. A finer grid costs about 47 times more and still does worse. It is the hyperparameter tuning problem.

EVOLVE had to settle each subtrial’s max sample size in advance

The team needed a max sample size about three months before first enrollment.

This EVOLVE design has eight settings. We estimated its performance by simulation.

Interim schedule

The calibrated rules fixed when the interim reviews happen.

Capacity across subtrials

The steering committee weighed the max sample size against expected enrollment before choosing a design for the first subtrial.

How many slots should a subtrial reserve when an early stop could leave some unused?

Same requirements, different trials

EVOLVE’s triple-negative breast cancer (TNBC) subtrial
Optimal (fewest patients if ineffective) Admissible (compromise)
Max sample size (slots reserved) 32 30
Average enrollment if ineffective 11.2 13.3
Power 82.1% 86.1%
Type I error 1.4% 1.5%
Would you reserve two fewer slots for about two more patients on average if the treatment is ineffective?

EVOLVE’s TNBC subtrial: three objectives, three designs

Trade-off plane for the TNBC single-arm cohort, most patients the trial can enroll against expected patients if the treatment is ineffective: Optimal at maximum 32 and expected 11.2, Admissible highlighted in orange at 30 and 13.3, and Minimax at 30 and 21.4. The dashed line is a visual guide.

What can the team prioritize?

  • Optimal (fewest patients if ineffective): max sample size 32, 11.2 expected
  • Admissible (the compromise): max sample size 30, 13.3
  • Minimax (smallest maximum, tied): max sample size 30, 21.4

Simon defined these for his two-stage design. BATON carries them to the wider family of adaptive trials by swapping in an objective built to mimic each one.

Admissible and Minimax tied at the lowest max sample size searched. Minimax also required eight events before an interim decision.

Physicians can choose the design that fits their priorities, from verified options.

The same workflow calibrated a more complex trial

Randomize patients from the start
Stage 1Screen the experimental arm against a historical benchmark.
Decision to proceedApply the planned rule.
No-go or ambiguousResolve under Stage 1 rules.
Go to Stage 2Compare experimental and reference arms.
Schema of the seamless design. Patients are randomized one to one from the start. The experimental arm passes a single-arm interim that can stop for futility, then a conversion gate. A no-go or ambiguous result ends with a Stage 1 declaration. A go result continues both arms into Stage 2 under between-arm rules, each with further interim looks.
The seamless design as calibrated, with 13 possible settings

Swap in new rules and a simulator. Reuse search and verification.

What if the trial assumes perfect classification?

Work in progress
Trial calculations usually assume the biomarker test classifies every patient correctly.
Power lost as the test misclassifies more patients
Bar chart of the change in power in simulated screening trials when a biomarker test misclassifies 5, 10 or 15 percent of patients, for three trial outcomes. Power falls by about 4 to 6 points at 5 percent, 8 to 11 points at 10 percent, and 13 to 15 points at 15 percent.
Patients misclassified by the test (%)
Trial outcomesurvivalresponseblood marker
Same specimen
Different test call
Different subtrial or treatment decision
With 10% misclassified, power fell by about 8 to 11 points. The false go rate barely moved.
Simulate plausible error rates in BATON. Recheck power and type I error, then recalibrate.

A public service announcement on AI-written simulations

The help: AI can write a trial simulator quickly, including a custom one for BATON.

The danger: The code runs and the results look plausible, yet it may assume every biomarker call is correct.

The check: Which patients does the code actually put in the subtrial?

If you cannot explain or defend a choice you made with AI’s help, you may be delegating too much.

You can’t evaluate what you don’t know, and you should never outsource your judgment to an external tool.

Section 5

1The stakes
2The design problem
3BATON
4BATON in practice
5Looking to the future
Can the same search design the whole program?

What does Phase III inherit?

54%
of 640 therapeutics failed in late-stage development
Phase I 20–100 Informs dose choice
Phase II Up to several hundred Tests for activity
Phase III 300–3,000 Tests benefit at that dose
Typical enrollment. One dot is 25 participants.
What should earlier trials learn to help Phase III confirm real benefit?

FDA trial sizes · Hwang et al. (2016) · FDA case studies (2017)

An early dose rule can shape Phase III confirmation

Illustrative example

Two doses shrink tumors about equally. The higher dose has more side effects.

A · Tolerability firstChoose the dose by a toxicity limit.The higher dose advances. More patients may cut back or stop.
B · Activity and tolerabilityCompare the doses on both measures.The lower dose advances. It may take more patients early.
To test in simulationFrom Phase I onward, how often does each rule yield valid Phase III confirmation, and with how many patients?

How PARTNER counted early patients in the final comparison

PARTNER · breast cancer trial
Treatment arm Stages 1–2Earlier patients Stage 3New patients Final comparisonOutcomes from both groups
ControlChemotherapy alone
Selected scheduleOlaparib 48 hours after chemotherapy
Other scheduleOlaparib before chemotherapy
Schedule dropped · outside the primary comparison
Compare how often pathology found no invasive cancer at surgery (pathologic complete response).
Abraham et al., Nature 2024 · Dots are schematic. Orange marks earlier patients, blue marks new ones

Extend BATON from one trial to the whole program

BATON would vary patients per phase · which dose advances and when to stop · how much early evidence counts
Each candidate program is simulated from Phase I to confirmation. The best is rechecked with fresh simulations.
Phase IChoose the dose
Phase IITest for activity
Phase IIIConfirm benefit
Outcomes from earlier patients count in the final analysis →
earlier patientsnew Phase III patients · dots are schematic
Can we confirm benefit with fewer new patients or less time, while controlling false positives?

Where BATON goes next

In EVOLVE, BATON found designs that met prespecified targets in simulation.
BATON-DPhase I dose finding and seamless Phase I/II researchWhich doses are worth advancing?
BATON-CRobustness work in progressWhich trial rules still work if our assumptions are off?
BATON-PhaselessProposed whole-program calibrationCan we preserve Phase III confirmation with fewer patients overall?
Design early trials around the evidence Phase III will need.

Acknowledgments

EVOLVE / ADAPT

  • MPIs: Lisa Carey (UNC), Ian Krop (Yale), Chuck Perou (UNC), Eric Winer (Yale), Antonio Wolff (Hopkins)
  • EVOLVE steering committee and TBCRC site investigators
  • ARPA-H ADAPT program leadership

Statistics / BATON

  • Amber Young (Rashid Lab PhD student), who led BATON's design and validation, and co-leads BATON-C
  • Dinelka Nanayakkara (Rashid Lab PhD student)
  • Didong Li (UNC)
  • ARPA-H Statistics Working Group: Susan Hilsenbeck (Baylor), Nabihah Tayob (DFCI), Ruizhe Chen (Hopkins)
  • Rashid Lab

BATON funding · ARPA-H ADAPT (ARPA-H-MAI-24-01-03, award 140D042590009) · DoD HT9425241103100121
NIH/NCI U01 CA274298, P50 CA257911, T32 CA106209, P30 CA016086, and P50 CA058223
UNC Lineberger Comprehensive Cancer Center and its University Cancer Research Fund

Thank you

Questions

naim@unc.edu · rashidlab.github.io
BATON R package: github.com/rashidlab/BATON
Manuscript in revision

References

  1. Biotechnology Innovation Organization, Informa Pharma Intelligence, QLS Advisors. Clinical Development Success Rates and Contributing Factors 2011–2020. 2021.
  2. Wouters OJ, McKee M, Luyten J. JAMA 2020;323(9):844–853, corrected 2022.
  3. U.S. Department of Health and Human Services. Operation TrialBlazer roadmap. 2026.
  4. Hitchen et al. Adult solid-tumor Phase I trials with an Australian site. 2024.
  5. Getz K. Characterizing white space in the quest to drive development speed. Applied Clinical Trials, 2021.
  6. Hwang et al. JAMA Internal Medicine 2016.
  7. Hong et al. KRAS G12C inhibition with sotorasib in advanced solid tumors. NEJM 2020;383:1207–1217.
  8. Sotorasib 960 mg and 240 mg randomized dose comparison. 2024.
  9. FDA Oncologic Drugs Advisory Committee briefing document, sotorasib, Table 29.
  10. FDA. Adaptive Designs for Clinical Trials of Drugs and Biologics. Guidance for industry, 2019.
  11. Simon R. Optimal two-stage designs for phase II clinical trials. Controlled Clinical Trials 1989.
  12. Jung SH, Lee T, Kim K, George SL. Admissible two-stage designs for phase II cancer clinical trials. Statistics in Medicine 2004.
  13. Rashid N. When Biomarkers, Assays, and Protocols Co-evolve. FDA CDRH OSEL seminar, 2026.
  14. FDA. Step 3: Clinical Research (trial sizes by phase).
  15. FDA. 22 Case Studies Where Phase 2 and Phase 3 Trials Had Divergent Results. 2017.
  16. Abraham et al. The PARTNER trial. Nature 2024.
  17. BATON software and manuscript (in revision).

Backup

Reference slides

Backup: the surrogate and the acquisition

Two-panel schematic. Top: a Gaussian process prediction for type I error with an uncertainty band and a feasibility threshold. Bottom: expected improvement times modeled feasibility, peaking near the constraint boundary.

  • Gaussian process models for the weighted loss and each constraint.
  • Proportion constraints use design-specific binomial Monte Carlo variances.
  • Expected constrained improvement multiplies expected objective gain by modeled feasibility.
  • The reported calibrations used DiceKriging GP fits.

Backup: ten starts on one Simon benchmark

Best feasible expected enrollment against cumulative evaluations for ten seeded runs on one benchmark. Eight runs finished at the known optimum.

All ten runs finished within 1% of the optimum under narrower search bounds. Eight matched the known design.

Backup: what the constraints and parameters actually are

Calibrated (searched) parameters, single-arm:

  • maximum enrollment; interim spacing; minimum-events gate before decisions count
  • posterior futility threshold; posterior efficacy threshold
  • decision margin over the historical benchmark
  • analysis-prior concentration; hazard-model resolution (piecewise intervals)

Seamless adds conversion and second-stage settings (13 possible, up to 11 active in the reported calibrations).

Constraints (headline scenarios):

  • type I error at most the ceiling (0.10 in the scenarios shown)
  • power at least the floor (0.80)
  • seamless designs add gate-behavior constraints (null conversion rate bounded, alternative conversion rate floored)

Objectives, by philosophy: expected null enrollment (H0-Optimal), maximum enrollment (Minimax), weighted blends (Admissible family), expected alternative enrollment (H1-Optimal).

Scenario A truths: null median 3.8 months, alternative 9 (median ratio 0.42), Weibull shape 1.3, accrual 2 per month; maximum enrollment searched in [30, 120]. Two Simon scenarios use a 0.05 ceiling with 0.90 power.

Backup: known failure modes and limits

  • Budget exhaustion: 3 of 10 reproducibility runs used the full 300-evaluation budget rather than stopping early; quote evaluations used, not “convergence.”
  • Approximate optimization: worst single-run benchmark gap was 13.5%. For certified designs, the constraint statements have 95% familywise confidence within each certification family at the simulated scenarios.
  • Correlated constraints: the acquisition multiplies marginal feasibility probabilities; latent surfaces for power and type I error can be correlated. Affects the search only; reported designs are verified directly.
  • No common random numbers in the loop: candidate comparisons carry full independent Monte Carlo noise; verification uses a shared seed set.
  • One assumed truth per calibration: every current result conditions on an assumed clinical scenario; robustness across scenarios is BATON-C, in progress.

Backup: bidirectional stopping variants

The single-arm and between-arm headline scenarios use futility-only interim stopping (interims can stop for futility; efficacy is decided at the end). The seamless design is calibrated with bidirectional stopping throughout, because its conversion gate is efficacy-driven.

  • The choice between stopping structures is itself a design decision upstream of the philosophy choice.

Backup: ten seeds, one design

The whole pipeline, rerun end to end at ten independent random seeds:

Runs returning a feasible design 10 / 10
Runs within 1% of Simon’s optimum 10 / 10
Runs returning Simon’s exact optimal design 8 / 10
Expected null enrollment, mean (spread) 17.77 (0.06)
Evaluations used, median 250 of a 300 budget

The other two seeds return a near-neighbor design (max sample size 29 instead of 25) whose expected enrollment differs by 0.15 patients. What reproduces is the design’s behavior, and usually the design itself.

Rerunning the search shows how much the returned design and its behavior change with the seed.

Backup: the randomized-subtrial frontier

Trade-off plane for the ER-positive doublet-versus-triplet cohort, total patients in both arms: H0-Optimal at maximum 103 and expected 59.2, Admissible at 91 and 63.9, Minimax at 86 and 70.8.

Cutting the max sample size from 103 to 91 costs about 5 extra patients on average when the treatment does not help. The next 5 cost about 7 more.

Backup: the ARPA-H ask and the N-of-1 addition

(ii) Clinical trial characteristics:
  1. Clinical trial for metastatic cancer patients.
  2. Clinical trial that includes multiple lines of therapy under a single protocol.
  3. Flexible clinical trial protocol that enables changes in design (e.g., randomized, adaptive, etc.).
  4. Strong capability to collect tumor and blood-based measurements.
  5. Possess a treatment line with a Progression Free Survival (PFS) rate of <9 months average.
  6. Mechanisms for biomarker evaluation.
  7. Modular clinical trial design enabling rapid dissemination and protocol uptake by different sites.
(iii) General characteristics: key clinical needs; evolutionary design examples and rapid-enrollment plans; treatment lines; biomarker testing and integration plans.
New ask (program-manager email): integrate an "N-of-1" trial component.

excerpts from the ARPA-H solicitation and program-manager correspondence

The ask: design an adaptive multi-line trial to discover, evaluate, and implement novel biomarkers, with an additional N-of-1 subcomponent
Animated this-is-fine cartoon from the original slide.

Backup: ADAPT technical areas and the EVOLVE team

Cropped official ADAPT technical-areas graphic: three circles for TA1 therapy recommendation, TA2 evolutionary clinical trial, and TA3 cancer treatment and analysis platform, with their sub-bullets; source slide number, footer, and highlight box removed.

Official Meet the Team slide: MPI photographs and roles beside tables for the steering committee, fifteen participating TBCRC sites, and the working groups.

Backup: Across 12 benchmarks: feasible in all 12, mean objective gap 3.6%

One run per scenario; ratio = BATON’s expected null enrollment / Simon’s exact optimum.

Claim Value
Feasible design returned 12 / 12
Exact Simon design returned 4 / 12
Mean ratio, BATON / Simon (SD) 1.036 (0.041)
Worst-case ratio 1.135
Best-case ratio 1.000

Feasibility is verified directly. Objective optimization remains approximate.

Backup: The seed sketches the map. Then BATON proposes each batch.

Heatmap of a fictional design-space slice, darker meaning better, with fifteen scattered space-filling evaluation dots and no lattice. The same surface with thirty model-guided evaluations clustering on the dark high-quality basin; repeated cells overlap. No connecting lines.

Initial points learn broadly; later points concentrate where the answer may improve.

Backup: Calibrating a seamless design

Three trial architectures with rising parameter counts, ending in a seamless design that begins single-arm and converts to a randomized comparison at an interim gate: thirteen parameters.

Stage 1Randomize from the start. Screen the experimental arm against the historical benchmark, then apply the conversion rule. Stage 2For trials that continue, compare the randomized arms. The reported calibrations stopped at the first Stage 2 look.

The seamless design had 13 possible settings. At most 11 were active in the reported calibrations.

Backup: verifying the final design with fresh simulations

Three narrowing stages in the EVOLVE single-arm example: search many designs with 3,000 simulated trials per design, recheck promising designs with 10,000 per design, and verify the selected design with five independent runs of 10,000 per scenario.

Fresh simulations check both targets before BATON reports a design.