Can Better Data Management Save Clinical Trials?

By Dan Schell, Chief Editor, Clinical Leader

Let me paraphrase something for you to get us started:

“There's a whole new world waiting for us through guidelines like this. We will be taking incremental steps, but in doing so, we will actually make a very big leap down the road.”

Essentially, that's what Jessica Jolly said during our Clinical Leader Live on October 31, 2024, when we were discussing the ICH M11 Guideline. Was she being hyperbolic? I don't think so. Guidelines and initiatives like the ICH M11 could — and hopefully will — revolutionize how we develop and execute clinical trials. And according to Jolly and her fellow panelist Hassan Kahlid, it is standardization and automation that will get us there — but I'm guessing you already knew that.

WE SUCK AT BEING EFFICIENT

If there's one thing we all know in the clinical trial industry, it's that we have a problem. Actually, we have several problems, but this one has to do with efficiency. Our current process is slow. Sponsors know it, CROs know it, sites know it, and unfortunately patients know it and feel the brunt of this delay more than anyone.

This universal acknowledgement of the problem has led to a plethora of proposed solutions, including the current ICH M11 guideline, which focuses on a clinical electronic harmonized protocol. “The point of the Guideline is that by creating a common language and framework, it enables the ecosystem to bealloperable from a content perspective and reusable, but it maintains flexibility so that it can accommodate all of the diverse needs of our stakeholders that are running various interventional trials,” says Jolly, who is a senior data and science executive with 25 years of global experience working across all facets of TechBio and Biopharma. The CDISC Unified Study Definition Model (USDM) and the digital data flow (DDF) initiative are similar solutions that align with the goals of ICH M11.

When we talk about transforming clinical study design, standardization is the precursor to automation, or namely, an autogenerated digital protocol. The latter would deliver increased efficiencies such as reducing manual data entry and requiring fewer quality checks. Kahlid, who is senior engineer, machine learning and data science at AstraZeneca, cautions, though, that even with standardization, you can’t just use any system or data flow. “All of this data we are talking about — it's very sensitive. You have to be GxP compliant, and you need a secure system. And you need to ensure all those systems are properly documented and validated.”

WHAT IF WE PULL THIS OFF?

OK, so let's say that we're able to do all of this. We standardize, we automate, and everything is compliant. Is there a way to quantify those efficiencies that we'll be gaining? Kahlid says it’s all about speed, especially after the data lock when you start analyzing and preparing the report for the regulators. “Sometimes this takes months or even years in some cases, which is time wasted for the company — and for patients. Automating some of the basic and monotonous tasks can help move this process along from months to days, and ultimately, to hours.”

Both Jolly and Khalid note that the ICH M11 isn't a silver bullet; in fact, now, it's still just a guidance. But as we move, in any way, toward a digital protocol that may involve automation or an AI component, there is going to be a huge culture component that needs to be addressed. Namely, some people are going to be afraid that parts of their job — or their job itself — will disappear. We shouldn’t undermine the importance of employee psychological safety with these kinds of evolutions,” Jolly says. “We need to take the time to make the people who are leveraging these technologies feel a sense of security. We need to explain that with standardized data and automation, we can actually reduce — not remove — human involvement from the loop and have the tech handle the repetitive tasks. Therefore, we're making humans smarter because they're now focused on oversight and reviewing things that the chat GPT's of the world may not have been trained on.”

AVOIDING DEEP DO DO

Halfway through the Clinical Leader Live, I realized I was confused about something. I remembered that Jolly had referred to data as “currency” and had said that it was necessary we find ways to get more clinical data to data scientists. I wanted to know what data that they didn't have that they still wanted. “It's not that we don't have the data, we just don't always know where it is, nor do we always know how to make it usable and searchable,” she explained.

From there, she and Khalid both talked about the importance of governance by design when dealing with all these disparate data sources. “It's got to be governed by design because otherwise we're just going to get into ‘deep do do’ in terms of intellectual property and attribution,” Jolly says. “We don't always know where our data is, nor do we get maximum value out of it to generate those insights that could help us with a lot of the initiatives we're trying to do today.”

Kahlid again expressed caution, saying that just because we have a dataset, doesn’t mean we can distribute it freely to data scientists. We need to make sure that there are not any laws or regulations dictating what data we can distribute and through which systems. “There are many other issues that need to be considered before we unlock any insights,” he says.

There were plenty of other topics that we covered during this Clinical Leader Live, including some of the opportunities that this whole landscape evolution offers to the clinical trial workflow. The panelists also addressed a few of the questions sent in from the audience, and I encourage you to listen/watch to the entire event as it contains some great nuggets of actionable information.