How Pfizer Uses AI to Transform Clinical Trial Data Oversight

At the 2024 SCOPE Summit, Demetris Zambas, VP and Global Head of Data Monitoring and Management at Pfizer explored the shift in Clinical Trial Data Management (CDM) towards digital methods and the imperative of maintaining data quality. He highlighted the challenges and opportunities this digital transformation presents, such as the need for accurate data entry and the potential of Generative AI to improve data review processes. Emphasizing the critical role of audit trails and centralized statistical monitoring in upholding data integrity, Zambas provided insights into using these tools to detect anomalies effectively.

The Imperative of Data Quality in Clinical Studies

Zambas began by referring to ICH E8, emphasizing that the “quality of a clinical study is considered fitness for purpose.” A clinical trial aims to yield reliable data to answer research questions while safeguarding participants. He stressed that the information produced must support robust decision-making, pointing to ICH E6 (R2) and the MHRA’s definitions to distinguish between reliable and potentially unreliable data. Zambas reinforced that data quality enables correct decision-making, whereas data integrity ensures clinical trial data is managed correctly.

The Evolution from Paper CRFs to Digital Innovation

Zambas discussed transforming paper-based to electronic data capture methods in clinical trials. He provided historical context, recalling how the industry’s reliance on paper CRFs led to potential data integrity issues, such as sites maintaining spare binders to replace pages that were too messy, leading to questionable audit trails. With the advent of electronic methods like EDC and eCOA, these practices have been largely mitigated, as the data is now directly captured and stored in a more secure and traceable manner.

However, Zambas also pointed out that new complexities come with digital evolution. He cited the risk of data entry timing—patients not adhering to the schedule, potentially filling out pain scales right before visits instead of the intended times, which could doubt data accuracy. Moreover, the surge in data volume from various sources, exacerbated during the COVID-19 lockdowns with remote data collection, has created challenges in tracking the data’s chain of custody. These examples illustrate the multifaceted risks inherent in modern digital data management and the ongoing need to develop rigorous validation and review processes to ensure clinical trial data integrity.

Methodological Advances and the Role of Generative AI

At the forefront of CDM innovation, Zambas highlighted the transformative role of Generative AI. He spoke about the methodology’s leap forward, with AI technologies set to fundamentally overhaul the data review processes by AI promising to enable a significant shift from reactive to forward-thinking data management. Despite these advances, Zambas noted a persistent vulnerability at the data’s entry point, emphasizing the need for stringent checks at the initial stages of data capture to ensure the integrity of the subsequent analyses.

Zambas illustrated this shift with a timeline, mapping out the evolution of CDM methodologies over the years. He traced the journey from the manual data trend detection era, where analysts would manually sift through data for insights, to the current landscape, where AI-powered processes can identify patterns and anomalies with little human intervention. This evolution signifies a paradigm shift in how data is reviewed and validated and suggests a future where CDM is as much about intelligent algorithmic prediction as it is about data collection.

The Crucial Use of Audit Trail Data

Zambas championed using audit trail data as an essential tool for risk management within clinical trials. He described how audit trails serve not just as a best practice but as a regulatory mandate, ensuring that each step in the data handling process can be verified and traced back to its origins. Zambas provided concrete examples where audit trail analysis has uncovered discrepancies and failures in processes, leading to corrective actions such as additional staff training or procedural amendments. The growing importance of audit trails is evident in their integration into routine trial reviews, particularly as a means to pinpoint and investigate data anomalies.

Zambas continued to expound on the practical applications of audit trails, detailing instances where they’ve been instrumental in improving the reliability of trial outcomes. He noted that a thorough review of audit trails could reveal inconsistencies that might otherwise go undetected, allowing for a more informed assessment of the data’s credibility. This practice has become increasingly crucial as trials become more complex, with audit trails acting as both a safeguard against data manipulation and a mechanism for upholding the integrity of the research.

Centralized Statistical Monitoring and Audit Trails

Zambas expanded into the intricacies of employing centralized statistical monitoring for audit trails, which is critical in identifying unusual patterns within electronic diary (eDiary) entries. He explained how this type of monitoring system is designed to flag data that deviate from expected trends, thereby acting as an early warning system against possible integrity issues. Through centralized oversight, such systems can sift through vast datasets to spotlight aberrations that merit further investigation.

To illustrate the effectiveness of this strategy, Zambas recounted a revealing case study from his experience. The data under scrutiny showed a series of eDiary entries suspiciously timestamped at one-minute intervals, an unlikely occurrence under normal circumstances. This pattern triggered an alert, leading to a probe that uncovered a coordinator’s misconduct in data entry. The coordinator was found to be systematically falsifying data, an act exposed through the meticulous examination of the audit trails. This incident highlighted the crucial role of audit trails not just in detecting but also in preventing the compromise of data integrity, demonstrating the tangible benefits of centralized statistical monitoring in maintaining the sanctity of clinical trial data.

The Future of Data Surveillance in Clinical Trials

In concluding his presentation, Zambas provided a visionary outlook on AI and Machine Learning (ML)’s role in data surveillance for clinical trials. He stressed the significance of these technologies as game-changers in detecting data anomalies and ensuring clinical trial data integrity. Zambas was confident that AI and ML would become indispensable tools for identifying discrepancies that could compromise the quality of trial results, thereby elevating the standard of risk management in clinical research.

Zambas further elucidated this point by presenting a conceptual diagram that encapsulated integrating various data streams into a comprehensive AI system. This diagram demonstrated the potential for AI to synthesize diverse data inputs—from patient records to sensor outputs—into a unified analysis platform. Such a system could then actively monitor and analyze these streams, applying sophisticated algorithms to detect irregularities and bolster risk-based quality management. This integration of AI promises to transform the landscape of CDM by providing a more dynamic, responsive, and ultimately more secure approach to handling and reviewing clinical trial data.