The FDA is paying attention to RWE. Is your evidence generation ready?
AI and real-world evidence are converging. The organizations that combine them rigorously will define the next era of healthcare decision-making.
When AI Met Real-World Evidence
FDA Begins Accepting RWE
FDA starts accepting real-world evidence in drug and biologic applications, marking a paradigm shift in regulatory science.
First Synthetic Control Arm Approval
Merck’s Bavencio approved for metastatic Merkel cell carcinoma based on a single-arm trial with a synthetic comparator arm — a regulatory first.
Decentralized Trials Accelerate
Implementation of hybrid and decentralized trial elements jumps 154%, driven by pandemic necessity and enabled by AI-powered remote monitoring.
RWE Goes Mainstream
23-28% of FDA label expansions utilize RWE. The FDA launches its Rare Disease Innovation Hub. Half of novel drug approvals carry orphan drug designation.
FDA Removes Data Barrier
FDA announces it will accept RWE without always requiring individually identifiable patient data — a landmark privacy-enabling policy shift.
AI + RWE Convergence Matures
The global RWE market reaches $2.15B. Context-aware AI models achieve 94% accuracy in clinical phenotyping. Integrated evidence generation becomes standard practice.
The Causal Inference Problem
The fundamental challenge of real-world evidence is causation. Unlike randomized trials, observational data comes with confounding baked in. Patients aren't randomly assigned to treatments, and the factors that influence treatment selection often influence outcomes too.
The core challenge in merging AI and RWE isn't technical — it's organizational. Teams that treat evidence generation as a corporate strategic mandate rather than a Medical Affairs tactical activity consistently produce more impactful, regulatory-ready evidence.
AI can help address these challenges through:
- →Propensity models — Automated covariate balancing across treatment groups
- →Instrumental variable identification — Finding natural experiments in observational data
- →Automated sensitivity analyses — Stress-testing results against unmeasured confounding
But these tools only work when deployed by teams that understand the underlying epidemiological principles. A well-specified causal diagram is worth more than the most sophisticated algorithm applied blindly.
What Regulators Actually Want
There's a gap between what companies submit and what FDA/EMA reviewers need to see.
What Companies Typically Submit
What FDA/EMA Actually Need
Convenience samples, post-hoc analyses, loosely defined endpoints
Pre-specified protocols, clearly defined populations, registered analysis plans
Acknowledges confounding exists, adjusts with basic regression
Multiple approaches: propensity scoring, IV, sensitivity analyses, quantitative bias analysis
Assumes data is 'good enough' from the vendor
Systematic data quality assessment, phenotype validation, missingness characterization
Methods described in a few paragraphs
Full protocol, statistical analysis plan, code availability, RECORD/STROBE reporting
Single data source, limited external validation
Multi-database studies, sensitivity to data source, explicit target population definition
References
- US Food and Drug Administration. FDA eliminates major barrier to using real-world evidence in drug and device application reviews. FDA Press Release. December 2025. Accessed February 2026. fda.gov
- Cozen O'Connor. FDA's guidance on the use of real-world evidence for medical devices. Published 2026. Accessed February 2026. cozen.com
- Seyhan AA. Drug development cost reduction through real-world data and AI. Pharm Times. 2024. Accessed February 2026. pharmacytimes.com
- Digital Bricks AI. 7 AI trends to watch in 2026. Published 2026. Accessed February 2026. digitalbricks.ai
- Solici Research. Synthetic control arms: use of RWE in clinical trials. Published 2024. Accessed February 2026. solici.com
- Li J, et al. Real-world evidence supporting FDA label expansions, 2022-2024. Drug Saf. 2024. Accessed February 2026. pubmed.ncbi.nlm.nih.gov
- Yahoo Finance. Real-world data (RWD) market report. Published 2026. Accessed February 2026. finance.yahoo.com
Let's Bridge the Gap
Looking to combine AI capabilities with rigorous evidence generation? Let's explore how.