MedCity Influencers, BioPharma

Machine learning: An essential tool in the shift to risk-based quality management

Risk-based quality management is a data monitoring approach that accelerates the drug development pathway without compromising on safety – and one that is enhanced by the use of technologies such as artificial intelligence.

Clinical trials have become increasingly expensive, time-consuming, and complex, leading sponsors to look for more efficient ways to conduct their business. Risk-based quality management (RBQM) is answering that call.

It is a data monitoring approach that accelerates the drug development pathway without compromising on safety – and one that is enhanced by the use of technologies such as artificial intelligence (AI).

AI, ML, and healthcare

Machine learning (ML) uses computer algorithms to achieve AI, i.e., a computer learning and mimicking human cognitive skills. ML models are designed to extract knowledge from large datasets and use this knowledge to make predictions on new or unseen data. It is particularly suited to tasks that cannot be solved by simple rule-based solutions meaning it naturally applies to healthcare, where decisions rely on complex processes.

As the healthcare sector continues on its path to digital transformation, it is generating a vast amount of data, stimulating a growing interest in health-related ML applications.

Deep learning, which uses large neural networks to handle complex structured and unstructured data, for example, has recently been used to scan electronic health records for medical device surveillance and to predict kidney injury. Computer vision, another application of deep learning, is streamlining microbiological testing and breast cancer screening. ML is also being deployed in the clinical trial arena, including in the implementation of RBQM.


Recommended by ICH E6 (R3), RBQM is a way of identifying, visualizing, managing, and documenting the risks that could impact the outcome of a clinical trial. Rather than the resource-intensive process of monitoring, recording, and verifying every trial parameter, it enables researchers to follow the most critical safety and efficacy data.

It works by replacing source data verification (SDV) with powerful methods to detect risks in clinical trials. Centralized statistical monitoring uses statistical algorithms to highlight atypical data patterns, or risk signals, in near real-time. It enables sponsors and CROs to identify, investigate, and correct problems like fraud, sloppiness in data entry, or problems with training or study equipment before they can impact trial integrity.

The result is more efficient studies that protect participant safety and ensure data integrity.

ML and RBQM: Perfect data-driven partners

Clinical trial data are so numerous and complex that making sense of them can be a daunting challenge for the human mind. Yet this is where deep learning models, which learn from complex and vast volumes of data, excel.

Deep learning can power RBQM tools, mining and analyzing information to highlight relevant insights. The modality has the flexibility to work with different data formats, workflows, and processes, meaning it can extract text from study documents, clinical notes, and investigation reports and combine it with clinical data, for example. It can also leverage previous studies’ learning in new programs. It can extract knowledge from external clinical databases or clinical corpora to inject medical knowledge into model predictions.

What’s more, ML models are not static. With the addition of approaches such as “human-in-the-loop,” which allows clinicians to highlight data areas to focus on and suggest actions, they are continually learning and refining their insights.

The possibilities are almost endless. But, to succeed, anyone setting out on an ML-driven RBQM journey must clearly define their goal and what they expect their solution to discover. This requires a strong understanding of the domain and collaboration with the industry.

Deep learning RBQM opportunities 

In practical terms, ML-driven RBQM can help study teams reduce their manual review efforts and provide them with meaningful and targeted data insights. It can automate key tasks, such as the grouping of risks and centralized monitoring study setup, and extract valuable insights from past trials, allowing sponsors and CROs to effectively plan, manage and document their studies.

It can also make study conduct more efficient by differentiating between different types of risk signals. RBQM systems raise a flag, or risk signal, when data patterns indicate a potential issue. These are then used to monitor and track the investigations that determine whether corrective action is necessary, resulting in reams of free text being entered by various users documenting their findings. But deep learning models can analyze this text data, and flag signals that represent a real issue. It allows teams to prioritize the review of signals and ensure the most effective follow-up and documentation of findings.

Industry-leading RBQM platforms processes a large volume and diversity of data from various organizations, meaning there are always new opportunities to develop scalable, robust ML solutions. The sector is for example, currently working to expand ML capabilities to data management and clinical review. The aim is to make time- and resource-hungry data management tasks, such as data cleaning, medical and AE coding, and mapping raw clinical data to SDTM, more efficient.

Over the years, the pharma industry has built a data goldmine on these activities. This information can now be used to train ML, particularly deep learning models, to present pharma teams with AI-driven suggestions that can be reviewed, then accepted or rejected.  The clinical review mainly relies on clinical expertise and, as such, will strongly benefit from deep learning that can mine data to acquire medical knowledge and learn from complex human decision processes.

No matter the project or application, such approaches have one ambition – to use ML to present study teams with the right insights to make better and faster decisions, optimize processes and free up time to focus on what matters most.

Joint coming of age

The key features of ML have the power to significantly improve the way we monitor and manage clinical trials, empowering RBQM teams, data managers, and clinical reviewers.

They are driving a revolution in research that promises to enable the more efficient development of life-changing medicines – but only if we ensure the careful design needed to provide meaningful insights.

Photo: mrspopman, Getty Images

After finishing a Ph.D. in atomic physics, Nicolas Huet pursued his career in the field of machine learning. He worked for several years in the data analytics team of a consulting company before joining CluePoints at the end of 2018 to integrate the research team. He's now leading the machine learning team in charge of developing new machine learning solutions.

Dr. Laura Trotta joined CluePointsC in 2015 and moved into her current role as Vice President of Research in January 2022, where she leads a team of research scientists responsible for developing new statistical and machine learning algorithms to assess the quality of clinical trial data. Laura holds a Master’s degree in Biomedical Engineering and a Ph.D. in Applied Mathematics from the University of Liège, Belgium.

As CSO, Steve Young oversees the research and development of advanced methods for the CluePoints platform and provides RBQM methodology guidance to customers. Steve worked in clinical data management over 15 years and has since focused on leading the development of analytics and risk-based quality management (RBQM) solutions for eClinical Solution providers. Steve holds a Master’s degree in Mathematics from Villanova University.

This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.