Real-World Data Won't Replace Trials. Here's Its Real Job

By John Oncea, Chief Editor, Clinical Tech Leader

Real-world data is often seen as a shortcut: a way to shrink trial size, skip a comparator arm, or answer a regulatory question without running another randomized study. Natasa Rajicic, ScD, a biostatistician with 25 years of experience across Pfizer, Cytel, and independent consulting, doesn’t buy the shortcut framing, but she’s practical about where real-world data genuinely earns its place in a development program.

“It depends,” was her first answer when I asked how well organizations use real-world data today. That wasn’t a hedge; rather, it’s an accurate answer from someone who has watched the same category of evidence work well in some contexts and fail, sometimes expensively, in others.

Where It Works: After The Drug Is Already On The Market

Rajicic’s clearest example of real-world data delivering value has nothing to do with proving initial efficacy. It’s in pharmacovigilance, after a drug is already approved and being used broadly. “Real world data will tell you of new safety marks or signals that were not just maybe some rare stuff that was just never...” she said, trailing into the point: signals too rare to show up reliably in a controlled trial population can surface once a drug reaches a much larger, more varied group of real patients.

The same post-marketing data can reveal how clinicians are actually using a product in practice, sometimes uncovering an unanticipated indication or a subgroup of patients who benefit disproportionately, information that wouldn’t necessarily emerge from the original trial’s inclusion criteria. “There’s all kinds of post-marketing and medical affairs and medical access impact on real-world data from real-world data,” she said. It’s not a single use case. It’s a category of ongoing value that keeps paying off well after initial approval.

Where It’s Overestimated: As A Comparator In Confirmatory Trials

The more contested use case is earlier in development, using real-world data as a comparator arm in place of randomized control, particularly for rare diseases where enrolling a traditional control group is difficult. Rajicic has seen this attempted more often in recent years. Her assessment of the track record is candid: “There are still so many examples of it when it didn’t work, or the FDA wouldn’t accept it, especially in Phase 3 applications, even for super rare diseases. So, I still see it with a lot of challenges.”

That’s a meaningful caution for any sponsor building a regulatory strategy around real-world comparators as a way to reduce trial size or timeline, especially in rare disease programs where the temptation to substitute real-world data for a control arm is strongest, precisely because recruiting a traditional comparator population is hardest.

The Right Framing: Augment, Don’t Compete

Rajicic was direct about where real-world data belongs relative to randomized trials: “It doesn’t compete with them.” Her preferred framing is augmentation. Real-world data can supplement an NDA application, and in rare disease programs specifically, it can allow the randomized portion of a development program to be smaller than it would otherwise need to be, because the real-world evidence fills part of the evidentiary gap rather than trying to replace the trial entirely. “You may end up with a package that is smaller than clinical development normally would be,” she said, “because you have real-world data that supplements it.”

That’s a materially different pitch than the one sponsors sometimes hear from vendors: not real-world data instead of a trial, but real-world data alongside a trial, reducing what the randomized portion has to carry on its own. It’s a narrower promise, but it’s the one that has actually held up with regulators, in Rajicic’s experience.

Why The Question You Ask Matters As Much As The Data

Rajicic’s most important qualifier came early in our conversation, almost as an aside: real-world data is valuable “if asked the right questions.” That phrasing puts the burden on the sponsor, not the dataset. The same real-world data source can produce a defensible pharmacovigilance signal and an indefensible efficacy claim, depending entirely on what question it’s being asked to answer and how rigorously that question was framed before the analysis started.

This is where her broader philosophy about pre-trial planning connects directly to real-world data strategy. A team that treats real-world evidence as an afterthought, something to pull together after a regulatory setback to shore up a weaker package, is far more likely to overreach with it than a team that defines up front exactly what gap the real-world data is meant to fill and why a randomized comparator alone can’t fill it. The organizations Rajicic has seen get real-world data to work well are the ones that scope its role early and narrowly, rather than treating it as a flexible, all-purpose substitute for evidence they haven’t yet generated.

A Practical Filter For Evaluating RWD Strategy

For those assessing where real-world data fits into a program, Rajicic’s experience suggests a workable filter. Post-approval pharmacovigilance, indication expansion, and subgroup identification: these are places where real-world data has a strong, repeatable track record. Using real-world data as a stand-in for a randomized comparator in a confirmatory trial, particularly for regulatory submission: this is where sponsors, including experienced ones, keep running into resistance from the FDA, rare disease status notwithstanding.

The dividing line isn’t about the sophistication of the analytics platform processing real-world data. It’s about what question the data is being asked to answer, and whether that question requires the specific protections randomization provides. “A valuable tool if it’s used” alongside randomized trials, as Rajicic put it, is a very different proposition than a tool meant to substitute for them. Sponsors who keep that distinction clear tend to get more durable value out of their real-world data investments, and fewer surprises when it comes time to submit.

None of this requires abandoning real-world data as a strategic asset. It requires being honest about which job it’s actually suited for. Post-market safety signals, indication expansion, and NDA supplementation for rare disease programs have a track record. Standing in for a randomized comparator in a confirmatory trial, on its own, generally does not. For a technology leader deciding where to invest in real-world data infrastructure, that distinction should shape not just the analytics being built, but the regulatory conversations happening well before any of that data gets collected.