Contributed Commentary By Jennifer Bradford

October 29, 2020 | Quality, accurate, and comprehensive clinical trial data are essential for a successful clinical trial. Without this information, companies cannot ensure the safety and efficacy standards that will pass regulatory review. 

Clinical teams often spend hours managing the data cleaning process, taking time away from the focus on data analysis.  Manual efforts for quality assurance are helpful but can be costly and lead to significant delays.

With data and data sources, i.e. wearables, rapidly expanding, better management of data is essential. Artificial Intelligence (AI) augmented by Machine Learning (ML) can save time, promote collaboration across trial sites, and ensure accuracy and quality.

New Data Management Approaches for Quality Outcomes

Quality clinical data provides the basis for analysis, submission, approval, labeling, and marketing of a compound.  A data cleaning process is key for consistency and accuracy in data collection and management.

Data management teams need to employ a manual approach to raise queries to the trial site for problem-solving or inconsistencies. On some studies there can be high levels of manually generated queries. Understanding the context of these may help improve the automated edit checks and identify potential data issues earlier. 

When ML is applied to historic manual queries across different studies, it can create an understanding of the common issues across and within studies, enabling a targeted approach to process optimization for data cleaning.  Directing efforts towards particularly problematic or critical aspects of the data enables a move towards automated detection of data issues, through a human-machine hybrid approach driving time and cost savings in the process.

The elimination or reduction of manual queries can increase efficiencies and costs better managed.  It’s estimated each manual query can cost nearly $200 from start to finish. If eliminated or greatly reduced, thousands of dollars can be saved, and more time is available for analysis to reach a successful trial conclusion.

Powerful ML technologies have the potential to monitor data as it is generated—identifying issues and inconsistencies as trials are ongoing. Remote monitoring has the potential to monitor many different measurements from a patient as they go about their everyday lives.  This generates large volumes of data that that would be near impossible for a clinician to monitor and analyze across a number of patients on a regular basis. ML technologies could be used to flag certain changes, potential issues or anomalies, directing the medical team to take any necessary action.

Adoption of the AI/ML Approach

AI and ML adoption creates a learning curve. For all consumers of AI technology, i.e. a sponsor, medical teams at a site, or patients, it is important that they understand the limitations of the technology and the context in which it can be used. Specifically, AI is only as good as the data from which it was built, and it may not always have all the right answers.

Decision makers who use the output of AI technologies have the challenge of acting on results. AI is a machine with no emotions or empathy. Therefore, decision makers must learn to use these technologies in the way they were intended but always apply their own reasoning to augment and evaluate AI output.

To maximize the value of data from the clinical trial process and provide essential quality insights, all data generated during the study should be considered—from the data to the associated metrics and audit trail.  Data from multiple studies can enhance analyses and/or determine baseline metrics that may inform current studies.  To support effective integration into current processes, the results of these approaches must be disseminated to the right people in the most effective way.  The development and utilization of intuitive visualizations can provide such a mechanism, bringing together these powerful analyses into a single source.

The application of AI to the manual data queries generated during the data cleaning process may be a human-machine approach for the foreseeable future. While the technology can be supportive, the output is determined by the quality of the data input. A machine can take large volumes of data and sift through it at an alarming rate without the inherent bias of a human. 

Rather than replace humans, AI/ML can be used to empower experts, enhance the information available to them to support evidence-based decision-making for diagnosis, treatment decisions, and enhancing operational aspects of care. The application of AI/ML technologies heightens at a level of reasoning and expertise, but the combination of human/machine may elicit better predictions than either group could generate alone. 

Jennifer Bradford, PhD is Head of Data Science for the CRO PHASTAR.  She previously worked for the Advanced Analytics Group at AstraZeneca, leading the development of the REACT clinical trial monitoring tool, which she later customized and delivered to other sponsors as part of Cancer Research UK (CRUK).  Within CRUK and in close collaboration with the Christie hospital she worked on EDC, app development and wearables data analytics in the context of clinical trials. She has a degree in Biomedical Sciences from Keele University and a bioinformatics Masters and PhD from Leeds University. She can be reached at jennifer.bradford@phastar.com.