Better Baseline
See how fewer events than expected can weaken a trial before enrollment begins.
Plan a binary-outcome trial, freeze enrollment, and explore which assumptions put its statistical power at risk.
Execution mode: LIVE — fresh calculations require the backend service
Synthetic Presets & Scenario Selection
Select a versioned synthetic preset or reset the current scenario to initial defaults. Editing assumptions will update field provenance to manual_user.
Planned 20% control risk collapses to 10% under fixed enrollment (N=903/arm, 0% dropout). Illustrates dramatic loss of statistical power.
Inputs Hash: sha256:d4dfd5525e9b355437d458dd211996674bd3b263509125fdceaf1b3bf8643380
Historical case study: AERIFY: fewer events than expected
Itepekimab (REGN3500 / SAR440340), developed by Regeneron and Sanofi, was studied for Chronic obstructive pulmonary disease (COPD).
- AERIFY-1 (NCT04701983): Met the primary endpoint at 52 weeks.
- AERIFY-2 (NCT04751487): Did not meet the primary endpoint at 52 weeks.
Reported endpoint: Annualized rate of acute moderate or severe COPD exacerbations.
The sponsors reported fewer exacerbations than anticipated and reduced statistical power in both trials. Pandemic-era enrollment was described as a possible contributor, not an established cause. Regeneron and Sanofi results announcement — May 30, 2025.
This historical context does not establish why AERIFY-2 missed its endpoint. Recurrent exacerbation rates cannot be entered as binary risks without an explicit validated model. Original binary planning assumptions and reconstructed trial power are not available here.
What earlier stress-testing could have revealed
Before enrollment, a team could ask: if fewer events occur than our plan expects, will our fixed sample still give us the chance of detecting the assumed benefit? Better Baseline makes this sensitivity visible for binary-outcome designs, so teams can discuss more conservative assumptions or enrollment margins with a statistician.
Applying that workflow to AERIFY would require a validated recurrent-event model and its original planning assumptions. This prototype illustrates the planning question; it does not establish that a different design would have changed AERIFY-2's result.
Explore the mechanism with default sample data
The default sample data are synthetic binary assumptions, not AERIFY measurements or a reconstruction. A planned control risk of 20% falls to 10% with RR 0.75 and zero dropout; enrollment stays at 903 per arm and approximate conditional power falls from 80% to 47%. Power is the chance of detecting the assumed difference under the model, not the probability a drug works.
Calculate & Lock Design, then compare the planned and stressed risks with enrollment fixed. This is an educational example, not regulatory advice.
Planned design and enrollment locking
Planning assumptions
Binary adverse endpoint, two-sided test, equal 1:1 arm enrollment, and symmetric dropout.
Percent fields use the displayed percent units: enter 20 for 20%. Relative risk uses a decimal ratio.
Planned design summary
Stale result. These values belong to the previous calculation; recalculate to match current assumptions.
Illustrative mock response. These fixed fixture values are not a fresh statistical calculation.
| Count | Control arm | Treatment arm |
|---|---|---|
| Required evaluable participants | 903 | 903 |
| Planned enrolled participants | 903 | 903 |
Enrollment inflation represents expected retention, not guaranteed evaluable counts or observed retained participants.
- Nominal analytic power at evaluable N
- 80.10%
- Hypothetical treatment endpoint risk
- 15.00%
- Analytical method
- Cohen’s-h two-sided normal approximation (cohen_h_normal_v1)
- Model version
- binary-v1
- Inputs hash
- sha256:d4dfd5525e9b355437d458dd211996674bd3b263509125fdceaf1b3bf8643380
- Source
- Baseline collapse — synthetic; synthetic
- Scenario ID
- baseline-collapse-v1
- Follow-up
- 12 months (illustrative)
- Source generator
- local_fixture
- Provenance method version
- binary-v1
Method warnings
- Hand-authored illustrative preset mock.
Power is conditional on the specified assumptions and method. It does not predict whether treatment works.
Stress Controls & Fixed Enrollment Invariant
Vary underlying control risk or attrition under fixed sample size. Notice that adjusting stress parameters never alters the locked enrolled counts.
Numeric value is a fraction (0.10 = 10%). Risk range: 1–99%.
Dropout numeric value is a fraction (0.10 = 10%). Slider range: 0–50%.
Analytic chart requests
Status: IDLEPlanned design and selected analytic scenario
Calculate a design and evaluate a stress scenario.
Analytic conditional power by baseline risk
Calculate a design and request its analytic curve.
Sensitivity heatmap
Calculate a design and request its analytic grid.
Selected-scenario Monte Carlo simulation
Simulation unavailable: calculate and select a valid frozen design first.
No simulation result yet.
Scenario export
JSON v1 records complete assumptions, locked enrollment, analytic grid and curve, selected simulation, separate MCSEs, warnings, hashes, software versions, RNG and field provenance.
Run the displayed seed and replicate configuration before exporting.
Reconstruction requires the recorded software environment. Schema validation does not independently verify statistical correctness.
Detached artifact validation
Reopening validates the file without changing workspace inputs, restoring enrollment or rerunning calculations.
Field provenance and audit trail
Current result hash: sha256:d4dfd5525e9b355437d458dd211996674bd3b263509125fdceaf1b3bf8643380
| Field | Source | Description |
|---|---|---|
| /planned_design/model_version | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /planned_design/control_event_rate | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /planned_design/effect/kind | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /planned_design/effect/value | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /planned_design/alpha | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /planned_design/target_power | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /planned_design/allocation_ratio | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /planned_design/dropout/control | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /planned_design/dropout/treatment | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /fixed_enrollment/control | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /fixed_enrollment/treatment | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /stress_scenario/control_event_rate | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /stress_scenario/dropout/control | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /stress_scenario/dropout/treatment | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /stress_scenario/effect_override | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /simulation/replicates | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |
| /simulation/seed | synthetic_preset | Hand-authored synthetic assumption from playbook; no clinical source |