ICON plc logo

ICON plc: Scaling BRA Across 200+ Compounds with an AI-Powered CRO Platform

Multi-Therapeutic All Phases CRO Operations Multi-Sponsor
200+

Compounds managed on single platform

73%

Reduction in BRA setup time

50+

Sponsor organizations served

$8.4M

Annual cost savings across CRO operations

Overview

ICON plc is one of the world's largest contract research organizations, providing outsourced drug development and commercialization services to pharmaceutical, biotechnology, and medical device companies. With operations spanning 53 countries, ICON manages clinical trials and pharmacovigilance programs for more than 50 pharmaceutical sponsors simultaneously, covering over 200 active compounds across oncology, CNS, immunology, rare disease, cardiovascular, and other therapeutic areas.

In 2024, ICON engaged ArcaScience to deploy an enterprise-scale AI-driven benefit-risk analysis platform capable of supporting multi-tenant, multi-sponsor CRO workflows. The goal was to standardize BRA methodology across all therapeutic classes while maintaining each sponsor's specific regulatory requirements, output formats, and data access controls. The deployment aimed to transform ICON's pharmacovigilance operations from bespoke, compound-by-compound safety analysis into a scalable, template-driven platform serving the full spectrum of drug development phases from Phase 1 through post-marketing surveillance.

The Challenge

As a major CRO managing 200+ active compounds for over 50 pharmaceutical sponsors, ICON faced a fundamental scalability challenge in its pharmacovigilance operations. Each new compound onboarded to the CRO required a bespoke safety analysis setup, with dedicated safety scientists configuring data ingestion pipelines, signal detection parameters, and regulatory output templates from scratch. This process took an average of 6 to 8 weeks per compound, creating a significant bottleneck as ICON's portfolio continued to grow.

The problem was compounded by massive duplication of effort across similar therapeutic classes. Safety scientists working on oncology compounds for one sponsor were building BRA frameworks nearly identical to those being constructed by a separate team for another sponsor's oncology program. Yet ICON's legacy systems offered no mechanism for sharing validated templates, signal detection configurations, or analytical frameworks between project teams -- even within the same therapeutic area.

Sponsor audits had begun to reveal inconsistencies in BRA methodology between different ICON project teams. One sponsor's regulatory affairs team identified that ICON was using different disproportionality analysis thresholds for the same MedDRA preferred terms across two compounds in the same therapeutic class. Another sponsor flagged that ICON's signal evaluation methodology for their post-marketing PSUR did not align with the methodology used for their Phase 3 interim benefit-risk assessment -- despite both being managed by ICON. These inconsistencies posed regulatory risk and were eroding sponsor confidence in ICON's pharmacovigilance capabilities.

Additionally, each sponsor required regulatory outputs in their own preferred format. Some mandated ICH E2E-compliant documents, others required CIOMS Working Group-formatted assessments, and European sponsors increasingly expected EMA Module 2.5-aligned outputs. ICON's safety writers were spending significant time manually reformatting identical analytical content into different regulatory templates, adding weeks to delivery timelines without adding scientific value.

The Solution

ArcaScience deployed an enterprise CRO configuration of its AI-driven benefit-risk analysis platform, purpose-built to support multi-tenant, multi-sponsor workspaces at the scale ICON required. The implementation was completed in 16 weeks, with the first 40 compounds onboarded in the initial phase and the remaining 160+ migrated over the following 6 months.

Multi-Tenant Architecture with Sponsor Isolation

The platform was configured with strict multi-tenant data segregation, ensuring each sponsor's proprietary safety data, clinical trial results, and regulatory documents remained fully isolated. Role-based access controls allowed ICON safety scientists to work across multiple sponsor programs while ensuring no cross-contamination of proprietary data. Sponsor-specific audit trails tracked every data access, analysis, and output generation event, meeting each sponsor's GxP compliance requirements and enabling seamless responses to sponsor audits.

Template-Based BRA Frameworks

ArcaScience and ICON's pharmacovigilance leadership jointly developed a library of validated BRA framework templates organized by therapeutic class. Oncology templates included pre-configured signal detection parameters for common SMQs (hepatotoxicity, QT prolongation, immune-mediated adverse reactions), standard benefit-risk value trees for checkpoint inhibitors and targeted therapies, and pre-populated comparator safety profiles from published literature. Similar template libraries were created for CNS (including neuropsychiatric safety endpoints), immunology (immunosuppression-related infection monitoring), rare disease (small-population statistical methods), and cardiovascular (cardiac safety monitoring). New compounds could be onboarded by selecting the appropriate therapeutic class template and customizing sponsor-specific parameters, reducing setup time from weeks to days.

Centralized Signal Detection

The platform deployed a centralized signal detection engine serving all 200+ compounds simultaneously. EBGM disproportionality analysis, temporal scan statistics, and literature-based signal corroboration ran continuously across FAERS, EudraVigilance, VigiBase, and each sponsor's proprietary safety database. Critically, the centralized architecture allowed cross-compound signal correlation -- when a safety signal was detected for one compound in a therapeutic class, the platform automatically flagged related compounds in the same class for enhanced monitoring. This cross-compound intelligence capability was unique to the CRO deployment model, giving ICON a safety signal detection advantage that no individual sponsor could achieve independently.

Automated Multi-Format Regulatory Outputs

The Regulatory Outputs module was configured with output templates matching each sponsor's preferred regulatory format. The same underlying analytical content could be automatically generated in ICH E2E format, CIOMS Working Group format, EMA Module 2.5 format, or FDA-preferred structured benefit-risk assessment format. Safety writers no longer needed to manually reformat documents -- they could focus on clinical interpretation and narrative quality while the platform handled formatting, cross-referencing, table generation, and regulatory compliance checks. Full audit trails were maintained for every output, compliant with 21 CFR Part 11 and EU Annex 11.

Platform Modules Used

Data Intelligence Engine Decision Intelligence Regulatory Outputs

Implementation Timeline

16 weeks

Therapeutic Areas

Oncology

CNS / Neurology

Immunology

Rare Disease

Cardiovascular

Regulatory Deliverables

ICH E2E, CIOMS Working Group, EMA Module 2.5, FDA structured BRA, PSURs, PBRERs, Phase 1-4 interim assessments

Results & Impact

200+

Compounds on a Single Platform

ICON successfully migrated its entire active pharmacovigilance portfolio onto the ArcaScience platform within 10 months. Over 200 compounds spanning oncology, CNS, immunology, rare disease, and cardiovascular therapeutic areas are now managed through a unified interface. The cross-compound signal correlation capability has identified 14 class-level safety signals that would not have been detected by analyzing compounds individually -- a unique competitive advantage for ICON's CRO model.

73%

Reduction in BRA Setup Time

New compound onboarding dropped from an average of 6 weeks (42 days) to just 11 days using the template-based BRA framework approach. Therapeutic class templates eliminated the need for bespoke configuration of signal detection parameters, benefit-risk value trees, and comparator safety profiles. The time savings compounded across ICON's portfolio -- with 35+ new compounds onboarded annually, the platform saves approximately 1,085 person-days per year in setup effort alone.

50+

Sponsor Organizations Served

More than 50 pharmaceutical, biotechnology, and medical device sponsors now receive AI-driven BRA deliverables through ICON's ArcaScience-powered platform. Multi-tenant data isolation and sponsor-specific access controls have passed 23 sponsor audits without findings. Three sponsors have cited ICON's AI-driven BRA capability as a deciding factor in selecting ICON for new pharmacovigilance contracts, contributing to an estimated $12M in new business wins in the first year.

$8.4M

Annual Cost Savings

Total annual cost savings across ICON's CRO operations reached $8.4M, driven by reduced safety scientist headcount requirements (avoided hiring 28 planned FTEs), elimination of duplicated BRA framework development across therapeutic classes, and automated multi-format regulatory output generation. The per-compound pharmacovigilance cost decreased by 41%, allowing ICON to offer more competitive pricing to sponsors while improving margins on existing contracts.

"ArcaScience gave us a competitive advantage we didn't know we needed. Our sponsors now expect AI-driven BRA as standard -- and we're the only CRO that can deliver it at scale. The cross-compound signal correlation alone has changed the conversation with our sponsor safety teams. We're not just executing pharmacovigilance anymore -- we're providing intelligence that no single sponsor could generate on their own."

Chief Medical Officer

ICON plc

Technical Details

Data Sources

  • FAERS (FDA): Complete spontaneous adverse event report database for all 200+ active compounds, with historical data spanning 5-20+ years depending on product lifecycle stage
  • EudraVigilance (EMA): European post-marketing ICSRs across all sponsor compounds with MedDRA-coded adverse reactions, stratified by reporting region and seriousness criteria
  • VigiBase (WHO-UMC): Global pharmacovigilance data from 140+ national drug safety authorities, providing international signal detection coverage
  • Sponsor Proprietary Databases: 50+ sponsor-specific safety databases (including Argus, ARISg, ArisG LifeSphere, and custom systems) integrated via validated data connectors with sponsor-specific access controls
  • Clinical Trial Databases: Patient-level data from 300+ active clinical trials across Phase 1-4, including SDTM-formatted adverse event datasets
  • Published Literature: Continuous monitoring of PubMed, Embase, Cochrane, and 200+ specialty journals across all therapeutic areas, with automated relevance filtering and signal corroboration
  • Real-World Evidence: CPRD, Optum Claims, Truven MarketScan, and IQVIA databases for background incidence rates and comparator safety profiling across therapeutic classes

AI Models Applied

  • EBGM Disproportionality Analysis: Empirical Bayesian Geometric Mean analysis running continuously across all 200+ compounds, with therapeutic class-specific stratification and automated threshold calibration based on compound maturity and reporting volume
  • Cross-Compound Signal Correlation: Proprietary algorithm comparing signal profiles across compounds within the same therapeutic class, identifying class-level safety signals that transcend individual compound analysis
  • Temporal Scan Statistics: Sequential probability ratio tests (SPRT) and MaxSPRT for detection of emerging time-dependent safety signals across both clinical trial and post-marketing data streams
  • Neural Narrative Clustering: Deep learning models for unstructured ICSR narrative processing across 12 languages, with automated causality assessment extraction and MedDRA auto-coding at 96.8% accuracy
  • Template Recommendation Engine: Machine learning model that recommends optimal BRA framework templates for new compounds based on mechanism of action, therapeutic class, indication, and phase, accelerating setup time
  • Multi-Format Output Generation: NLG (Natural Language Generation) models trained on ICH E2E, CIOMS, EMA Module 2.5, and FDA structured BRA formats, enabling automated regulatory document generation with sponsor-specific formatting preferences
  • MCDA Benefit-Risk Synthesis: Multi-criteria decision analysis with swing weighting and probabilistic sensitivity analysis, calibrated per therapeutic area with class-specific value trees

Multi-Tenant Architecture

The enterprise CRO deployment required a purpose-built multi-tenant architecture to meet ICON's operational and regulatory requirements:

  • Sponsor Data Isolation: Cryptographically enforced tenant boundaries ensuring zero cross-sponsor data leakage, with dedicated encryption keys per sponsor workspace and hardware security module (HSM) key management
  • Role-Based Access Control (RBAC): Granular permissions allowing ICON safety scientists to work across multiple sponsor programs while maintaining strict data isolation, with 14 predefined roles and custom role support
  • Sponsor Audit Portal: Self-service audit interface allowing sponsor QA teams to independently verify data access logs, analysis provenance, and regulatory output generation history for their compounds
  • Scalable Infrastructure: Auto-scaling compute architecture supporting concurrent signal detection analysis across 200+ compounds without performance degradation, with 99.95% uptime SLA
  • Compliance Framework: Platform validated to FDA 21 CFR Part 11, EU GMP Annex 11, SOC 2 Type II, ISO 27001, and GDPR requirements, with sponsor-specific compliance attestations available on demand

Regulatory Context

This engagement spans the full regulatory landscape for CRO pharmacovigilance operations:

  • ICH E2E: Pharmacovigilance Planning -- standardized signal detection methodology and risk management integration across all therapeutic areas and development phases
  • ICH E2C(R2): Periodic Benefit-Risk Evaluation Report (PBRER) format and content requirements for post-marketing compounds
  • CIOMS Working Group IV: Benefit-risk balance methodology for marketed medicinal products, applied as an alternative output format for sponsors requiring CIOMS-formatted deliverables
  • EMA Module 2.5: Clinical overview format for European regulatory submissions, with automated generation of benefit-risk sections aligned to EMA expectations
  • EMA GVP Module VII: Periodic Safety Update Report requirements for European post-marketing pharmacovigilance
  • FDA Guidance: "Benefit-Risk Assessment for New Drug and Biological Products" (2021) -- structured framework applied as default for US-focused sponsors
  • Outcome: 23 sponsor audits passed with zero critical findings. Methodology consistency scores improved from 67% to 98% across ICON project teams. Three regulatory authorities (FDA, EMA, PMDA) have positively commented on the standardized BRA methodology in submission reviews.

Related Case Studies

Accelerating Oncology BRA with AI-Driven Signal Detection

AstraZeneca

68% reduction in BRA cycle time

Transforming Rare Disease Benefit-Risk Assessment

Sanofi

52% faster regulatory submissions

View All Case Studies

Explore More Success Stories

ArcaScience Evidence Hub

20+ enterprise case studies

Scale Your CRO Pharmacovigilance with AI-Driven BRA

See how ArcaScience can help your CRO standardize benefit-risk methodology, reduce compound onboarding time, and deliver AI-powered pharmacovigilance at enterprise scale. Talk to a scientist about your multi-sponsor operational challenges.

Talk to a Scientist Download PDF