Company
Mid-size biopharmaceutical (Top 50 global)
Therapeutic Area
Oncology — Non-Small Cell Lung Cancer (NSCLC)
Product
PD-1/LAG-3 bispecific antibody (checkpoint inhibitor)
Development Stage
Pre-BLA submission (Phase 3 complete)
ArcaScience Modules Used
Data Intelligence, Decision Intelligence, Automated Outputs
Timeline
12-week engagement (vs. 30-week projected manual process)

The Challenge

Complex Safety Profile in a Crowded Therapeutic Space

The client was preparing a BLA submission for a first-in-class PD-1/LAG-3 bispecific antibody for second-line NSCLC. While the pivotal Phase 3 trial (N=1,247) demonstrated statistically significant improvement in overall survival versus standard-of-care pembrolizumab, the novel mechanism introduced a safety profile with unique characteristics that required careful characterization.

The specific challenges facing the regulatory team were multifaceted:

The Solution

ArcaScience Platform Deployment

ArcaScience deployed its full platform suite—Data Intelligence, Decision Intelligence, and Automated Outputs—to create a comprehensive, quantitative benefit-risk assessment that met regulatory expectations while fitting within the compressed timeline.

Implementation Process

Weeks 1-2
Data Integration and Decision Framework Setup
Connected to the company's clinical data warehouse and safety database (Argus). Ingested individual patient-level data from all 14 studies. Established the MCDA framework with 18 criteria (7 benefit, 11 risk) identified through advisory board input and regulatory feedback.
Weeks 3-4
Safety Signal Analysis and Characterization
AI-powered signal detection analyzed the integrated safety database, identifying 3 new signals not previously characterized. NLP models extracted detailed case-level information on hepatotoxicity events, including time-to-onset patterns, dose-response relationships, and outcome trajectories. Comparative analysis against published pembrolizumab and nivolumab safety data provided class-level context.
Weeks 5-7
Quantitative Benefit-Risk Modeling
Built MCDA model with performance data from the pivotal trial and integrated safety database. Conducted discrete choice experiment with 125 oncologists and 85 NSCLC patients to derive preference weights. Executed 50,000 Monte Carlo simulations to characterize uncertainty. Generated interactive sensitivity analyses including tornado diagrams, spider plots, and threshold analyses.
Weeks 8-10
Document Generation and Regulatory Preparation
Automated generation of the benefit-risk summary for Module 2.5 of the CTD. Created interactive presentation materials for the advisory committee. Prepared detailed appendices with full methodology documentation. Generated response templates for anticipated FDA questions on the hepatotoxicity signal.
Weeks 11-12
Review, Refinement, and Finalization
Cross-functional review by regulatory affairs, clinical development, safety, and medical affairs. Incorporated feedback and refined visualizations. Final QC and regulatory formatting. Delivered complete package 4 weeks ahead of the BLA submission deadline.

Results

60%
Reduction in BRA
preparation time
12
weeks
Total engagement duration
(vs. 30-week manual estimate)
93%
Probability of favorable
benefit-risk balance

Quantitative Outcomes

Metric Manual Estimate ArcaScience Actual
Time to complete BRA 30 weeks 12 weeks
FTE effort 12 FTEs 5 FTEs
Data sources integrated 3-4 (manually) 14 studies + 2 registries + literature
Sensitivity scenarios 10-15 (manual) 50,000 (Monte Carlo)
New safety signals identified 0 (known signals only) 3 (including 1 requiring label language)
Cost of BRA preparation $1.8M estimated $720K actual

Regulatory Outcome

The BLA was accepted for filing without a Refuse to File letter. During the review cycle, the FDA issued 14 information requests; notably, none were related to the benefit-risk assessment—a significant achievement given the pre-BLA feedback on hepatotoxicity concerns. The advisory committee voted 11-2 in favor of approval, with multiple committee members commenting favorably on the transparency and rigor of the benefit-risk analysis.

The product received FDA approval with standard prescribing information that included the hepatotoxicity risk in Warnings and Precautions rather than as a Boxed Warning—an outcome the team attributed in part to the quantitative demonstration that the overall benefit-risk profile remained favorable even under the most conservative risk assumptions.

Unexpected Value

The AI-powered safety signal analysis identified a previously unrecognized pattern of late-onset (>12 months) autoimmune thyroiditis that occurred at a rate of 4.7% in the treatment arm. Because this signal was identified during BLA preparation rather than post-marketing, the company was able to proactively include monitoring recommendations in the prescribing information and risk management plan. This proactive approach was noted favorably by both FDA reviewers and the advisory committee, and prevented what would likely have been a post-marketing safety signal requiring a label supplement.

"ArcaScience transformed what we expected to be a 30-week manual exercise into a 12-week process that produced a far more rigorous and comprehensive benefit-risk assessment than we could have achieved on our own. The quantitative approach gave us confidence in our regulatory strategy and provided the advisory committee with exactly the kind of transparent, data-driven analysis they were looking for."
— Vice President, Regulatory Affairs
"The signal detection capabilities identified a safety signal we would have missed entirely with our traditional approach. Catching the late-onset thyroiditis signal before approval allowed us to include appropriate monitoring guidance in the label from day one. That's not just good science—it's the right thing for patients."
— Chief Medical Officer

Key Takeaways

  1. Quantitative BRA can be achieved on regulatory timelines when supported by AI-driven data integration and analysis. The 60% time reduction was achieved without sacrificing analytical rigor.
  2. AI-powered signal detection adds value beyond efficiency. The identification of three new safety signals, including one that influenced labeling decisions, demonstrates that AI analysis can uncover clinically meaningful findings that manual review might miss.
  3. Regulatory agencies respond positively to structured quantitative approaches. The absence of benefit-risk-related information requests and the advisory committee's favorable comments validate the investment in rigorous methodology.
  4. Proactive risk characterization benefits all stakeholders. By identifying and addressing the thyroiditis signal before approval, the company protected patients, demonstrated good pharmacovigilance practice, and avoided post-marketing regulatory actions.
  5. The cost savings are substantial and real. The $1.08M savings on a single BRA preparation represents a strong ROI, and the platform continues to provide value through living benefit-risk updates for the product's post-marketing lifecycle.

Ready to Transform Your Benefit-Risk Strategy?

Learn how ArcaScience can support your next regulatory submission with AI-driven benefit-risk analysis.

Schedule a Consultation  |  info@arcascience.ai