The Data Was Already There: RWE's Real Bottleneck Is People
By John Oncea, Chief Editor, Clinical Tech Leader
Years ago, researchers at Abbott’s pharmaceutical division in Basel noticed something that didn’t fit the story their trial was supposed to be telling. One of the company’s products, developed for an entirely different purpose, had a side effect on liver enzymes that appeared to protect against liver injury. Buried inside the real-world data, a pattern emerged: certain cancer patients taking the drug for unrelated reasons were tolerating their chemotherapy longer and doing better than expected.
The connection wasn’t obvious, and the drug being developed had nothing to do with oncology. But its protective effect on the liver meant patients could stay on chemotherapy regimens that would otherwise have been cut short by side effects, and staying on treatment longer improved their cancer outcomes.
“This was only done because a student at the time took an interest and just happened to find a needle in a haystack,” said James Richardson, an industry pharmacist who spent a decade in Bayer’s medical affairs organization before moving into academia and, this month, into a global medical communications leadership role at UCB, a multinational biopharmaceutical company headquartered in Brussels. “They started plowing into the data, and then they found out why.”
Richardson didn’t share this story to celebrate a lucky break. Rather, he was making a point about where real-world evidence is actually stuck. Specifically, it isn’t a data problem; it’s a people problem.
When AI Finds A Signal, Someone Still Has To Ask “Why”
Clinical research organizations don’t lack real-world data anymore. Between EMRs, claims databases, registries, and connected devices, the raw material for discoveries like Abbott’s already exists inside most companies’ systems. What’s missing, Richardson argues, is the capacity to notice it.
“I think the data in these links exist, but we as people don’t have the bandwidth to actually look at it,” he said. “I think AI is powerful enough to really start churning through these things.”
That’s not how the role of AI in research is usually pitched. Richardson isn’t describing AI as a replacement for statisticians or a shortcut past regulatory scrutiny; he’s describing it as a way to surface “fuzzy sets of data,” correlations nobody thought to look for because nobody had the time to look.
But finding a pattern and understanding it are two different jobs. “The limitation is when these things come up, is there enough people to look at why there could be a linkage there? Does it make sense? Is it just serendipity?” Richardson said. “That is the limitation now: the number of people that can make sense of the data that AI might find in this kind of scenario. But it exists. So, the data’s there to look at.”
Asked why organizations are short on that kind of analytical capacity – hiring, education, or simply too much data to absorb no matter how well-staffed a team is – Richardson replied, “The short answer is I don’t know.” But he did offer a hypothesis: the industry doesn’t yet have enough visible success stories to justify the investment.
The Missing Ingredient Isn’t Technology, It’s Incentive
That’s the second half of Richardson’s argument, and it’s the less comfortable one for a technology-focused audience: better tools alone won’t fix this. Organizations invest in analytical capacity when they can see a return, and right now, Richardson said, there simply aren’t enough public examples of real-world data investigation paying off.
“There’s not enough real positive stories that exist to whet the appetite of companies to throw money at it,” he said, pointing back to Abbott as one of the few visible cases. “I think that whet their appetite enough to hire some people to then really start looking in this area. I think other companies can probably follow suit, but I think there needs to be more stories about the process of why new indications might be found for products like this. If there’s incentive, then companies will put money into it, but they need to know that there’s a good enough incentive to do that.”
It’s a subtle but important distinction for clinical research leaders weighing AI investment. The pitch shouldn’t be “AI will find things humans can’t.” It’s already capable of that, according to Richardson. The pitch that actually moves budget is proof that finding those things pays off in indications, in outcomes, and in patients who stay on treatment longer because someone noticed a pattern buried in the noise.
Where The Champions Are And Why They’re Isolated
Richardson also pushed back on the idea that this shift needs a single company or technology vendor to lead it. Instead, he described a landscape of individual champions inside pharma and medtech organizations, each pushing for better use of real-world data and AI-assisted analysis, largely without coordinated support.
“Companies are looking for internal champions to push things in the right area,” he said. “I think these internal champions could really do with some help with external partners as well, because often they’re working in isolation.” He described informal networks among peers at different companies who treat the challenge as shared rather than competitive. “We don’t see it as a competitive business between companies,” he said. “We see it as more than a zero-sum game where we can all help each other and move forward.”
That collaborative instinct, paired with a genuine data-analysis capacity gap, is where Richardson sees the real opportunity for the industry over the next few years – not in acquiring more real-world data, but in building the human infrastructure to interrogate it, and in publicizing the wins clearly enough that the next Abbott-style discovery doesn’t depend on one curious student stumbling onto it.
“I think it’s still very early for most companies,” Richardson said of AI adoption in real-world evidence work. “I think they’re moving in the right direction, and we need help. I think we really need help.”