Why Real-World Evidence Is Failing HEOR Teams at the Moment It Matters Most

Real-world evidence was supposed to solve one of health technology assessment's most persistent problems. Randomised controlled trials, however rigorously designed, capture what a therapy does under controlled conditions in selected patient populations over defined time horizons. They do not capture what happens when that therapy is used by the full spectrum of patients who present in clinical practice, managed by clinicians with varying levels of experience, across health systems with different treatment protocols and support infrastructure. Payers have always known this. They have always wanted evidence that bridges the gap between the trial population and the real-world population they are being asked to fund treatment for. Real-world evidence, generated from registries, electronic health records, claims databases, and observational studies, was positioned as the answer. And to a significant extent, it has delivered on that promise. The volume of published real-world evidence in most major therapeutic areas has grown dramatically over the past decade. The problem is not the evidence. The problem is what happens to it when HEOR teams try to integrate it into systematic reviews, economic models, and HTA submissions alongside conventional trial data. That integration process is where real-world evidence most frequently fails to deliver its potential value. And it is where artificial intelligence is beginning to create a genuinely different kind of capability.