Transforming Clinical Trials with AI: Clario’s Innovative Approach to Endpoint Data Solutions

Co-Authors: Kim Nguyen and Shyam Banuprakash, Clario

Overview of Clario’s Role in Clinical Trials

Business Challenge: The Need for Efficient Software Configurations

Solution Overview: Implementing Clario’s Genie AI Service

Workflow Steps: From Study Initiation to Documentation

Benefits and Results: Streamlining Processes and Enhancing Data Quality

Lessons Learned: Insights from Implementing Generative AI

Conclusion: The Future of Data Processing in Clinical Trials

About the Authors

Transforming Clinical Trials: The Power of Generative AI at Clario

This post was co-written with Kim Nguyen and Shyam Banuprakash from Clario.

Clario is a leading provider of endpoint data solutions for systematically collecting, managing, and analyzing specific outcomes (endpoints) in the clinical trials industry. With over 50 years of experience, Clario has supported clinical trials over 30,000 times, contributing to more than 700 regulatory approvals in over 100 countries. This legacy is built on generating high-quality clinical evidence for life sciences companies striving to bring new therapies to market.

In our previous post, we discussed how Clario developed an AI solution powered by Amazon Bedrock to accelerate clinical trials. Since then, our focus has expanded to innovative solutions that streamline the generation of software configurations and artifacts while delivering high-quality clinical evidence.

Business Challenge

Designing and customizing software configurations to manage clinical trial stages efficiently is essential. These configurations span from basic study setup to advanced features such as data collection customization. Clario collects data from various sources to build specific configurations, but the traditional workflow, reliant on manual extraction from documents, posed significant challenges:

1. Manual Data Extraction: Team members manually review PDF documents for structured data, leading to potential inaccuracies and inefficiencies.
2. Transcript Challenges: Transferring data manually into configuration documents opened doors for inconsistencies.
3. Version Control Difficulties: Iterating or updating studies complicated the maintenance of consistency across documents and systems.
4. Fragmented Information Flow: Data existed in silos, making integration difficult.
5. Software Build Timelines: The lengthy configuration processes directly impacted the timelines for generating necessary software builds.

Recognizing these challenges, Clario implemented stringent quality control measures. However, the reliance on manual processes presented ongoing risks related to precision and consistency.

Solution Overview

To tackle these business challenges, Clario developed the Genie AI Service, leveraging generative AI powered by large language models (LLMs) like Anthropic’s Claude 3.7 Sonnet on Amazon Bedrock. This innovative approach revolutionizes how Clario manages clinical trial software configuration.

Key Features of the Genie AI Service

• Automated Data Structuring: Utilizing a custom data parser, the Genie AI Service automatically structures information from PDF forms, consolidating it into validated tables.
• Centralized Data Source: The system aggregates data from various sources—transmittal forms, study details, and standard exam protocols—into an interactive review dashboard.
• Post-Validation Automation: After stakeholder verification, the system generates a Software Configuration Specification (SCS) document, culminating in AI-powered XML generation for Clario’s proprietary medical imaging software.

Workflow Steps

1. Study Initiation & Data Collection: Users enter a study code, and the system pulls relevant details via API calls, establishing a solid foundation for configuration.
2. Data Extraction: Leveraging Anthropic’s Claude Sonnet, the solution parses and organizes structured data from transmittal forms, applying domain-specific rules.
3. Review & Validation: Stakeholders validate and adjust AI-generated configurations using the interactive review interface before saving them to Clario’s database.
4. Document & Code Generation: The system automates the creation of essential documentation, generating XML files for software builds, complete with detailed conversion logs.

Benefits and Results

The Genie AI Service enhances data quality and streamlines the validation process, minimizing manual data entry. This results in reduced errors and improved communication across teams, fostering a culture of transparency.

Additional benefits include:

• Faster Configuration Execution: Study configuration times are significantly reduced, while maintaining high quality.
• Focus on Value-Added Activities: Teams can concentrate on optimizing study designs rather than manual tasks.

Lessons Learned

Clario’s transformation journey has imparted valuable lessons for future initiatives surrounding generative AI.

Key Insights

• Prompt Engineering: Crafting prompts with domain knowledge is crucial. Detailed examples provide the AI with the context required for success.
• Human Oversight: While AI accelerates extraction, human review is necessary for accuracy within structured workflows.
• Phased Implementation: Gradual rollouts with pilot teams allowed Clario to test functionality and ease transitions.

Challenges Faced

• Two-System Synchronization: Ensuring bidirectional integration between SCS documents and the solution requires careful refinements.
• Data Formatting Variability: Differences in transmittal forms across therapeutic areas present ongoing adaptation challenges.

Conclusion

The transformation of Clario’s software configuration process reflects a fundamental shift in how data processing is approached in clinical trials. The Genie AI Service exemplifies a hybrid model that leverages the power of LLMs combined with human expertise, orchestrated through Amazon ECS for reliable execution.

This initiative showcases how generative AI is not just a futuristic tool but a practical solution that creates immediate value. Organizations looking to undergo similar transformations should focus on well-defined use cases, fostering human-AI collaboration, and emphasizing measurable outcomes.

As Clario continues to advance in this space, the lessons learned will shape future generative AI implementations, paving the way for a more efficient and effective clinical trials landscape.

About the Authors

Kim Nguyen serves as the Sr Director of Data Science at Clario, leading a team focused on AI and machine learning solutions in clinical trials.

Shyam Banuprakash is the Senior VP of Data Science and Delivery at Clario, spearheading innovative data solutions within medical imaging.

Praveen Haranahalli is a Senior Solutions Architect at Amazon Web Services, specializing in secure cloud solutions and AI/ML implementations.

By leveraging their expertise, Clario is positioned to drive the next wave of innovation in clinical trials.