Commentary Contributed by Rich Davies, CluePoints 

May 24, 2024 | The clinical trials landscape is evolving with more data, increased investment in personalized medicines and a shift toward decentralized and hybrid clinical trials. At the same time, the industry is attempting to become more efficient and harness new technologies and data science in operations. 

Risk-based data management (RBDM) is the targeted and efficient application of clinical data management practices to clinical trials. It is designed to maximize the value of data management activities on any given trial and move away from a one-size-fits-all approach. RBDM is influenced by quality-by-design (QbD) based on a risk assessment, which informs where we should be concentrating our activities.  

An Evolving Clinical Trial Landscape 

The typical Phase III protocol now collects an average of 3.5 million data points—seven times more than 15 years ago. Several factors are driving this increase. They include direct access to patient data from tools like electronic patient-reported outcomes (ePROs) and ongoing data from wearable and other devices. At the same time, audit trails have become a hot topic, as has the need to utilize the large datasets they create. 

Increased investment in personalized medicines is resulting in more complex protocols. The percentage of drugs in research and development that are personalized medicines increased from 3% in 2000 to 61% in 2020. It is fantastic that we can more accurately target indications with these types of therapies. However, increasing protocol complexity has the potential to introduce stress to different stakeholders. For example, there might be a higher patient burden or more decision points for site teams as they are executing a protocol that could lead to data quality being compromised.  

How and where we collect data has also changed. Decentralized clinical trials (DCTs) give us the ability to target a wider patient population and increase diversity and inclusion. But they also bring in potential different dimensions of risk that we need to be aware of and manage. For example, if we are asking patients to collect more data themselves, and be more responsible for that data, we are potentially increasing patient burden which could have an impact on the data collected. If we are sending home healthcare professionals to manage supplementary data and they misunderstand the protocol or their responsibilities, that could create a bias. 

Different Focus Areas Contributing to the RBDM Concept 

There are several areas of industry focus which are currently circling data management. But ultimately, they all contribute to the concept of RBDM. 

As the industry seeks to improve efficiencies in clinical trials, sponsors are critically evaluating data management processes and focusing on marginal gains which combine to create a more streamlined development process. This efficiency scenario is often done with an eye on the rapidly shifting technology landscape, including exploring how artificial intelligence (AI) and machine learning (ML) solutions can provide time, quality, and efficiency benefits. Organizations are also looking to embed the concept of data science within their operations, using new technologies to focus on finding the quality issues that really matter. 

As the industry becomes more risk savvy and focused on the critical issues within a trial, organizations that have successfully deployed risk-based quality management (RBQM) are exploring replicating that risk-based process down into clinical data management. All these factors overlap. For example, if we can make efficiencies, it frees resources to start thinking about risk-based approaches.  

There is a good deal of redundancy in current data management practices. As we move to RBDM there is a real opportunity to break down some of the silos in businesses, so we get more efficiencies and fewer redundancies. We can look at resource-intensive processes like data review and manual query creation and ask, “Does it have to be the way it has always been or is there a better way of doing it today?” 

A risk-based approach can deliver further efficiencies and give us the opportunity to work more critically with new data sources. As more data comes down the pipeline, data science can then help us identify more systemic, important abnormalities in data, giving us the capability to address the issues that matter, faster. These efficiencies can be supported by the exciting developments we are seeing from a technology perspective.  

If we look at the issue of manual queries, we traditionally run our SAS and SQL listings, review the output then decide whether to raise a query. If we do, we go back into EDC and there is no feedback loop, so we do not typically tell those listings whether that query was answered positively or not. Every time we run those listings, the logic is the same and we get the same results out, meaning data managers are having to repeat the same activity. 

If, instead of testing data against listings and SAS programs, we feed it into a deep learning (DL) model trained on historical data, it can suggest queries. A feedback loop allows the DL model to learn which queries to raise, based on which ones have resulted in data changes. 

The Benefits of RBDM 

Section 5 of ICH E6 (R2) mapped out the seven steps of taking a risk-based approach – identification of critical data and processes, identification of risks, evaluation of risks, risk control, risk communication, risk review and risk reporting. This was applied cross-functionally meaning data management was only part of the process. 

What we are seeing now with organizations that are implementing RBDM is they are looking at what they have successfully put in place for RBQM and duplicating that down.  

A data management risk assessment allows us to confirm critical data from a data-management perspective. It helps to cascade down critical data enabling data management to understand where that data lives and how they are going to apply their practices to it. 

It is also an opportunity to assess each moving part in the data collection strategy and the complexities of data flow. It can even be used to decide which systems to implement additional audit review activities on. All of this allows us to start thinking about flexing downstream data management practices and to promote the concept of risk-based thinking across the data management team. 

The benefit for data management is we can then start to think about what we do once we have that deep insight and perception of risk. We can start to appreciate what is critical and non-critical data and flex the way we deploy data-management activity. Benefits of this approach include opportunity to reduce and target operational practices, such as deployment of EDC edit checks, reviews of listing-type outputs and adopting sampling strategies for dealing with some of the data.  

An Opportunity for an RBDM Revolution 

There is a huge opportunity to pivot the way we work in data management. By moving away from a one-size-fits-all approach to data management, RBDM allows us pull back from tedious, time-consuming, and lower value tasks. By doing so we can release bandwidth to deal with new data and more complex operational scenarios, leverage new ways to support traditional pain points and increase identification of the issues that matter. 

Richard Davies joined CluePoints in September 2018 and is based in the UK. As VP, Solution Expert his role is to support organizations adopting CluePoints’ solutions from a technical, functional and business process perspective, particularly as they execute their vendor selection programs. Additionally, he has product management responsibilities and provides a bridge between customers, prospects and ongoing product development. He has worked for technology vendors for over 20 years but prior to this, he worked in data management with Fisons Pharmaceuticals and Astra. Richard graduated from DeMontfort University, Leicester, with a Bachelor of Science degree in Computer Science. He can be reached at rich.davies@CluePoints.com.