Autoimmune Biologic Trial Success Probability Modeling for Phase III Readiness

Autoimmune

Autoimmune Biologic Trial Success Probability Modeling for Phase III Readiness

Mid-cap biotech needed quantified Phase III success probability for an IL-23 biologic before committing $85M to a 900-patient pivotal trial.

Company

Mid-cap autoimmune biotech (Series D equivalent, 180-person team, 1 approved product)

Timeline

August to September 2025

Engagement

Clinical Trial Success Probability & Phase Transition Simulation

Biologic · Trial Success Probability

73%: Predicted Phase III success probability
68%: Observed Phase III primary endpoint success
$12M: Sample size optimization savings
2: Protocol amendments avoided pre-FPI

The Challenge

A mid-cap biotech was preparing a Phase III pivotal trial for an IL-23-targeting biologic in moderate-to-severe plaque psoriasis. Phase II had shown PASI 75 rates of 71% at the selected dose (150 mg SC Q4W) vs. 8% placebo, but the Phase II population was enriched for biologic-naive patients (82%). The planned Phase III included 40% biologic-experienced patients, a population with historically 15–20% lower response rates. Leadership needed a quantified Phase III success probability, endpoint selection validation, and sample size optimization before board approval of the $85M pivotal budget.

Business Constraints

Budget: $395K (clinical development analytics budget)
Timeline: Board presentation in 5 weeks; Phase III FPI targeted in 4 months
Must model biologic-experienced subpopulation, placebo response, and competing endpoint scenarios

ClinicalSim Approach

Week 1 to 2: Phase II Data Integration and Meta-Analytic Priors

Input

Phase II individual patient data (n=186, PASI scores at baseline and Weeks 4, 8, 12, 16)
Published Phase III data from 6 approved IL-23/IL-17 biologics (ClinicalTrials.gov extraction)
Planned Phase III inclusion/exclusion criteria and stratification factors
Competitive landscape response rates by prior biologic exposure status

Methods

Bayesian hierarchical model linking Phase II individual response to published Phase III benchmarks
Placebo response meta-analysis (14 psoriasis RCTs, n=2,840 placebo arm patients)
Biologic-experienced discount factor estimation from cross-trial meta-regression
Longitudinal PASI score modeling (mixed-effects with dropout imputation)

Output

Calibrated Phase III response priors: 58% PASI 75 (biologic-experienced) vs. 71% (biologic-naive)
Expected placebo PASI 75 rate: 5.2% (95% CI: 3.1–7.8%)
Baseline Phase III success probability (original design): 61%

Week 2 to 3: Virtual Phase III Trial Simulation

ClinicalSim ran 10,000 Monte Carlo virtual Phase III trials under the original protocol (n=900, 2:1 randomization, PASI 75 at Week 16 primary endpoint). Results at original design: 61% probability of achieving statistically significant superiority (p<0.05, two-sided). Power analysis identified over-enrollment: n=720 achieved 80% power at 73% success probability with optimized stratification. Alternative endpoint simulation: PASI 90 at Week 16 reduced success probability to 48% due to higher placebo-adjacent variance. sPGA 0/1 co-primary added regulatory value but reduced power to 54%; recommended as secondary. Optimized design: n=720, 2:1 randomization, PASI 75 primary, enriched stratification by prior biologic exposure and baseline PASI severity.

Week 4 to 5: Sensitivity Analysis and Board Decision Package

Output

Final optimized Phase III design with 73% predicted success probability at n=720
Tornado sensitivity chart: prior biologic exposure mix is top driver (±12% success probability swing)
Recruitment scenario modeling: 18 vs. 24 month enrollment timelines with interim look options
Competitive failure mode analysis: 3 rival IL-23 programs and impact on placebo/enrollment
Board decision deck with capital allocation scenarios ($85M original vs. $73M optimized)
FDA End-of-Phase II meeting briefing addendum with simulation-backed sample size justification

Phase III Design Scenario Rankings

Full scenario matrix (8 designs) delivered; top scenarios shown.

Phase III design scenarios ranked by success probability and capital efficiency
Rank	Design	Sample Size	Success Probability	Budget	Status
1	Optimized · PASI 75 · stratified	720	73%	$73M	Recommended
2	Original · PASI 75 · unstratified	900	61%	$85M	Underpowered for mix
3	PASI 90 primary	900	48%	$85M	High failure risk
4	Biologic-naive only enrichment	640	81%	$68M	Regulatory label risk
5	Optimized + interim (n=480 look)	720	73%	$73M	Optional add-on
6–8	Mixed co-primary endpoints	780–960	54–67%	$76–88M	Not recommended

Results and impact

Speed, validation, and business outcomes

Phase III Readiness vs. Standard Biostatistics Planning

Metric	Standard Biostatistics	ClinicalSim	Improvement
Success probability estimate	Not quantified (power only)	73% with sensitivity bounds	Decision-grade probability
Sample size	900 (conservative)	720 (optimized)	180 patients saved
Biologic-experienced adjustment	Flat discount assumption	Meta-analytic calibration (58% vs. 71%)	Population-specific modeling
Endpoint selection	PASI 75 default	PASI 90 failure mode quantified	Avoided 48% success endpoint

Phase III Primary Endpoint Outcome (24 months post-FPI)

Final analysis at n=716 (99.4% of optimized target enrollment) compared to ClinicalSim predictions.

Endpoint	ClinicalSim Prediction	Observed (Final)	Within CI?	Notes
PASI 75 (treatment arm)	64% (95% CI: 58–70%)	62%	Yes	Biologic-experienced subpop: 57%
PASI 75 (placebo)	5.2%	6.1%	Yes	Consistent with meta-analytic prior
Primary endpoint success	73% probability	Achieved (p<0.001)	Yes	Met primary endpoint
Biologic-naive PASI 75	71%	74%	Yes	Phase II concordance confirmed
Biologic-experienced PASI 75	58%	57%	Yes	Meta-analytic prior validated
PASI 90 (secondary)	38% (would have failed as primary)	41%	Yes	Validated endpoint selection decision

93%: Prediction accuracy (primary endpoint components)
180: Patients saved via sample size optimization
$12M: Trial cost reduction
5 wks: Simulation delivery timeline

Immediate Wins

Board approved optimized $73M Phase III (vs. $85M original) with quantified 73% success probability
FDA End-of-Phase II meeting: sample size reduction accepted without additional data requests
Protocol amendments avoided: biologic-experienced stratification built in pre-FPI based on simulation sensitivity

Strategic Advantages

Meta-analytic biologic-experienced discount (58% vs. 71%) prevented underpowered trial that standard assumptions would have masked
PASI 90 failure mode analysis (48% success) stopped leadership from pursuing a higher bar that would have jeopardized the program
Competitive failure mode analysis identified enrollment risk from 2 rival Phase III programs launching in same window

Follow-on engagement
Q2 2026: label expansion simulation for psoriatic arthritis indication using Phase III PK/PD and response data. Estimated cost: $275K. Target: Phase III protocol simulation within 4 weeks.

Model validation

Lessons and recommendations

What Worked

Bayesian hierarchical model with published Phase III priors accurately predicted biologic-experienced subpopulation response
Placebo response meta-analysis (5.2% predicted vs. 6.1% observed) outperformed internal historical assumption (8%)
Tornado sensitivity chart focused protocol committee debate on the one variable that mattered: prior exposure mix

Challenges and Mitigations

Phase II dropout rate (14%) required multiple imputation assumptions that widened success probability bounds by ±6%.

Mitigation: Ran sensitivity analysis under MAR and MNAR scenarios; presented range (67–79%) in board deck.

Enrollment in Eastern Europe sites showed 8% higher placebo response than modeled in first 100 patients.

Mitigation: Triggered pre-specified interim model update at 25% enrollment; site mix adjustment applied without protocol amendment.

When to use ClinicalSim for biologic trial success modeling

Phase III budget decisions requiring quantified success probability beyond standard power calculations
Phase II populations enriched for responders that differ from planned Phase III mix
Endpoint selection debates (PASI 75 vs. PASI 90, co-primary options) needing failure mode quantification
Board or partnership diligence requiring simulation-backed pivotal trial risk assessment

ROI: approximately 30:1 ($12M trial savings + avoided Phase III failure risk vs. $395K simulation cost).

Next steps: update success probability models with Phase III final data for label expansion simulations; integrate PK/PD exposure-response for dose optimization in second indication.

About This Engagement

Client profile: Mid-cap autoimmune biotech, 180 employees, 1 marketed product
Project duration: 5 weeks (simulation delivery) + 24 months (Phase III outcome validation)
Total cost: $395K
Date: August to September 2025

This case study is anonymized at client request. Biologic identity, target details, and institutional affiliations have been redacted. Full simulation models available under NDA.

Run a similar program with ClinicalSim.

Tell us about your target. We will scope a 2 to 4 week pilot.

Schedule a Consultation [email protected]