Analytics

Real-world evidence (RWE) has been used in some form for many decades, but its role in drug development has evolved significantly over time. Early use of real-world data (RWD) was primarily focused on post‑marketing safety and pharmacovigilance, particularly from the mid‑20th century onwards, as regulators recognized the need to monitor medicines after approval using observational data.

From the early 2000s onwards, the use of RWE in drug development decision‑making expanded with the increasing availability of digital health data, national healthcare registers, and advances in epidemiological and statistical methods. Over the past 15-20 years, RWE has moved beyond safety monitoring to support a broader range of activities, including disease understanding, study design, and post‑approval evidence generation.

Today, RWE is used across the product lifecycle. During early development, it helps characterize disease epidemiology, treatment pathways, and unmet needs, supporting more informed and feasible trial design. After approval, RWE supports post-authorization safety studies, long‑term effectiveness assessments, and evidence generation required by regulators and health technology assessment bodies after launch.

Overall, RWE makes drug development more iterative and data‑driven. Rather than relying solely on clinical trials conducted at fixed points in time, evidence can continue to be generated as treatments are used in routine care. This supports faster, more targeted development decisions, better post‑approval oversight, and a stronger evidence base for regulatory and reimbursement decisions, especially in areas such as rare diseases, oncology, and chronic conditions.

AI is expected to change how real-world data (RWD) and real-world evidence (RWE) are generated, analyzed, and used, but it is not replacing established scientific methods or human involvement. Instead, its main impact is on scale, speed, and consistency.

At the data level, artificial intelligence (AI) can help handle the growing volume and complexity of RWD. Techniques such as natural language processing and machine learning enable more efficient extraction of relevant information from unstructured sources like clinical notes and free‑text fields, as well as improved linkage and harmonization of data across multiple sources.

At the analysis stage, AI can support faster data exploration and pattern recognition, helping researchers identify signals, patient subgroups, and treatment patterns that might be difficult to detect using traditional approaches alone. This is particularly valuable in large‑scale datasets such as national health registers, where AI can help prioritize analyses and improve the efficiency of study execution.

That said, regulatory and health technology assessment bodies are cautious. Agencies such as the European Medicines Agency, Food and Drug Administration, and National Institute for Health and Care Excellence emphasize that AI‑supported analyses must be transparent, reproducible, and clearly validated. “Black‑box” models without clear rationale are unlikely to be accepted for decision‑making on their own. Human oversight, a clear context of use, and robust study design are still essential for regulatory‑grade RWE.

Overall, AI is a powerful tool for RWD and RWE, improving efficiency and scalability, but scientific judgement and methodological rigor remain central to how RWE is assessed and used in regulatory and reimbursement decisions.