Real-World Evidence, AI, and the Board: What Life Sciences Directors Should Ask

Evidence drawn from health records, claims, and registries now informs regulatory decisions on drug safety and effectiveness. As AI takes on more of that analysis, the board's question is how the company knows the evidence is credible.

In Brief

The Problem. Regulators in the United States and Europe now rely on real-world evidence for safety actions and, in defined cases, for new indications, and more of that evidence is generated or analyzed with artificial intelligence (AI).

Why It Persists. Directors tend to treat evidence generation as a scientific matter for management, so the board rarely sees where label claims, safety commitments, or deal values rest on real-world evidence, or how its credibility was established.

What Boards Should Do. Map where the company relies on real-world evidence, require management to show how each AI model behind it was validated for its context of use, read data licenses as risk documents, and test counterparties' evidence in every licensing deal or acquisition.

Real-world data (RWD) and real-world evidence (RWE) have gained steady regulatory acceptance as a complement to traditional clinical trial data. The United States Food and Drug Administration (FDA) defines RWD as data on patient health status or the delivery of care, routinely collected from sources such as electronic health records (EHRs), claims, and registries. Analysis of those data produces RWE: clinical evidence about the use and potential benefits or risks of a medical product (U.S. Food and Drug Administration, n.d.).

When I first wrote on this subject in 2023, I described RWE as sometimes surpassing randomized clinical trials (RCTs). The evidence since then supports a more precise claim, and a more useful one for a board. RWE has earned a place beside the randomized trial, not above it. It can settle questions that trials cannot reach, for new molecular entities (NMEs) and for established products, when the data are fit for purpose and the study is designed with discipline. As AI produces more of that evidence, the quality of the evidence becomes a board matter, not only a scientific one.

The Evolution of Regulation

The regulatory history of RWD and RWE reflects growing recognition of their value alongside trial data. In the United States, the FDA Amendments Act of 2007 directed the agency to establish active postmarket risk identification and analysis. The FDA launched its Sentinel Initiative in May 2008, and the full Sentinel System launched in February 2016 (U.S. Food and Drug Administration, 2022). The 21st Century Cures Act of 2016 directed the agency to evaluate wider use of RWE, and in 2018 the FDA published a framework for using RWE to support new indications and post-approval study requirements (U.S. Food and Drug Administration, n.d.).

Since then the rules have moved from principle to practice:

• In July 2024, the FDA finalized guidance on assessing EHR and medical claims data to support regulatory decisions for drugs and biologics (U.S. Food and Drug Administration, 2024).
• In September 2025, the International Council for Harmonisation (ICH) adopted its M14 guideline on non-interventional studies that use RWD to assess the safety of medicines (Eglovitch, 2025).
• In December 2025, the FDA finalized updated guidance on using RWE to support regulatory decisions for medical devices, replacing its 2017 guidance (U.S. Food and Drug Administration, 2025b).

The European Medicines Agency (EMA) launched its initiative for patient registries in September 2015 (European Medicines Agency, n.d.). It also coordinates the European Network of Centres for Pharmacoepidemiology and Pharmacovigilance (ENCePP) (ENCePP, n.d.). Its own data network, the Data Analysis and Real World Interrogation Network (DARWIN EU), is now part of routine regulatory work. Between February 2025 and February 2026, the EMA conducted 88 regulator-led RWD studies, a 49 percent increase on the prior period. DARWIN EU had grown to 40 data partners in 18 European countries, covering more than 250 million patients (European Medicines Agency, 2026).

Other regulators have moved in the same direction. Health Canada announced on April 16, 2019 that it would strengthen how RWE is considered at each stage of a drug's life cycle (Health Canada, 2019). The United Kingdom's Medicines and Healthcare products Regulatory Agency (MHRA) issued guidance on RWD in clinical studies on December 16, 2021, and consulted in 2025 on a draft guideline for external control arms (Medicines and Healthcare products Regulatory Agency, 2021). A 2022 review of guidance worldwide found that frameworks are multiplying but that acceptance of RWE, especially before approval, still differs across countries (Burns et al., 2022).

The direction across these documents is consistent. Regulators accept RWE when the data are relevant and reliable for the question asked, and when the study is planned and documented in advance (U.S. Food and Drug Administration, 2024). They treat it as evidence fit for a defined purpose.

Two Decisions Real-World Evidence Has Already Made

On effectiveness, the clearest example remains Prograf (tacrolimus). In July 2021, the FDA approved its use to prevent organ rejection in adult and pediatric lung transplant recipients. The approval rested on a non-interventional study using the U.S. Scientific Registry of Transplant Recipients, which covered all U.S. lung transplants, with randomized trials in other transplant settings as confirmatory evidence. The FDA stated that a well-designed non-interventional study using fit-for-purpose RWD, compared with a suitable control, "can be considered adequate and well-controlled under FDA regulations" (U.S. Food and Drug Administration, 2021).

On safety, a recent European example shows RWE changing practice. A DARWIN EU study of more than 40,000 patients informed updated recommendations from the EMA's Pharmacovigilance Risk Assessment Committee (PRAC) on blood monitoring for clozapine, including less frequent blood count checks (European Medicines Agency, 2026). Neither decision displaced the randomized trial. Each answered a question a new trial would have answered slowly, at great cost, or not at all.

Drug Safety Use Cases

RWD is a valuable resource for understanding drug safety in real-world populations, particularly for NMEs, whose safety profile at approval rests on trials of limited size and duration. Sentinel and DARWIN EU exist for this purpose (U.S. Food and Drug Administration, 2022). The main uses are these:

• Post-marketing surveillance. Analysis of EHRs, claims, registries, and other sources can identify adverse events, drug-drug interactions, and other concerns that clinical trials did not detect.
• Rare adverse events. Trials may lack the sample size or duration to detect rare or long-term adverse events. Larger and more diverse real-world populations can reveal them.
• Serious adverse events from polypharmacy. Trial inclusion criteria often exclude patients with comorbidities and concomitant medications. RWD from those patients can surface serious adverse events caused by drug-drug interactions, and can inform modeling that anticipates them.
• Safety in specific subpopulations. Patients with comorbidities, pregnant women, older adults, and children are often underrepresented in or excluded from trials. RWD can show how a drug affects them in practice.
• Comparison with alternative treatments. RWD can compare the safety of an NME with existing treatments or standards of care, which helps clinicians choose.
• Pharmacovigilance and signal detection. Monitoring real-world data sources lets regulators and companies detect potential safety signals early and act on them.
• Risk management and mitigation. A clearer real-world safety profile supports targeted risk management plans and mitigation strategies.
• Label updates and regulatory decisions. RWD can support new safety information, precautions, and contraindications on labels, and can inform post-marketing commitments and risk evaluation and mitigation strategies (REMS).

Drug Effectiveness Use Cases

RWD can also help evaluate and improve the effectiveness of NMEs in the patients who actually receive them. Where the data are fit for purpose, RWE can support a new indication or satisfy a post-approval requirement, the two uses named in the FDA's 2018 framework (U.S. Food and Drug Administration, n.d.). The main uses are these:

• Comparative effectiveness research. RWD can compare an NME with existing treatments or standards of care in routine practice.
• Understanding real-world patient populations. RWD shows how a product performs across comorbidities, demographics, and treatment histories, and can identify the subgroups that benefit most.
• Treatment adherence and persistence. Poor adherence and persistence reduce effectiveness. Real-world patterns can inform strategies to improve outcomes.
• Real-world endpoints. Patient-reported outcomes, quality-of-life measures, and healthcare resource use can show the overall value of a product in practice.
• Treatment optimization and personalized medicine. RWD can identify factors that influence response, such as genetic variation, drug-drug interactions, or lifestyle, and guide personalized treatment.
• Post-marketing studies and evidence generation. RWD can support post-marketing studies whose evidence informs regulatory decisions, label updates, and value-based payment.
• Clinical practice guidelines. Guidelines built on both trial and real-world evidence reflect the patients clinicians actually treat.

Each use depends on the same thing. The data must be complete enough, the methods sound enough, and the analysis transparent enough to support the decision being made.

Where AI Changes the Question

AI now assists at many steps: extracting outcomes from clinical notes, linking records across sources, matching comparable patient groups, and screening large safety databases. That brings speed, and it adds a link in the chain of evidence that regulators, and boards, must be able to examine.

In January 2025, the FDA issued draft guidance on using AI to support regulatory decisions on the safety, effectiveness, or quality of drugs and biologics. It proposes a risk-based framework for establishing the credibility of an AI model for a specific context of use (U.S. Food and Drug Administration, 2025a). As of October 21, 2026, that guidance remains in draft. On January 14, 2026, the FDA and the EMA jointly issued ten guiding principles of good AI practice in drug development, including a clearly defined context of use and management across the full life cycle (Al-Faruque, 2026). Together they signal what a sponsor should be ready to show: what the model does, how much rests on it, and how it performed.

The Strongest Objection: Observational Evidence Can Mislead

The strongest objection to RWE is that it is observational. Without randomization, unmeasured differences between patient groups can produce an effect that is not real. That objection is serious, and the best test of it is instructive. Researchers at Brigham and Women's Hospital, working with FDA staff, used three U.S. claims databases to emulate 32 randomized trials. Across all 32, the correlation with trial results was 0.82. Among the 16 trials whose design could be closely emulated, it rose to 0.93; among the 16 that could not, it fell to 0.53 (Wang et al., 2023).

The lesson is not that RWE is reliable or unreliable. It is that its reliability depends on whether the data and design fit the question. That is exactly the judgment a board cannot make itself, and exactly the judgment it must confirm management has made, with documentation, before the company relies on the result.

Why This Belongs in the Boardroom

Directors do not need to evaluate study designs. They do need to know where the company's regulatory strategy, labeling, and commercial claims rest on RWE, and whether management can show how that evidence was produced. I would point a board to three exposures.

Reliance. A label expansion or a post-marketing commitment may depend on RWE. If so, the quality of the data and methods is a regulatory and reputational risk of the first order.

Data rights. Much RWD is licensed from third parties. The license terms decide whether the company may train models on the data, what happens when a source changes, and who bears the risk of unreliable data.

Transactions. In a licensing deal or an acquisition, the value of an asset may rest on RWE or on AI-generated analysis produced by the counterparty. Diligence should test that evidence with the same rigor the company applies to clinical trial data.

RWE will keep gaining ground as data collection and analytical methods improve. Companies that can show how their evidence was produced will gain the most from that shift.

For the Board

• Which of our current or planned regulatory submissions, label changes, or safety commitments rely on real-world evidence?
• For each, how closely does the study design fit the question, and who outside the study team reviewed it?
• Where is AI used to generate or analyze that evidence, and can management describe how each model's credibility was established for its context of use?
• Do our data license agreements address model training, change notification, data quality, and audit rights?
• In our last licensing or acquisition review, how did we test the counterparty's real-world or AI-generated evidence?
• Who is the accountable executive for evidence quality, and when did the board last hear from that person?

Subscribe to Oversight Rounds, from Avancer for monthly FDA and European Union updates, and find the life sciences entries in the Avancer AI Governance Navigator.

References

Al-Faruque, F. (2026, January 14). EMA, FDA issue joint AI guiding principles for drug developers. Regulatory Affairs Professionals Society. https://www.raps.org/resource/ema-fda-issue-joint-ai-guiding-principles-for-drug.html

Burns, L., Le Roux, N., Kalesnik-Orszulak, R., Christian, J., Hukkelhoven, M., Rockhold, F., & O'Donnell, J. (2022). Real-world evidence for regulatory decision-making: Guidance from around the world. Clinical Therapeutics, 44(3), 420–437. https://doi.org/10.1016/j.clinthera.2022.01.012

Eglovitch, J. S. (2025, September 15). ICH adopts M14 guideline on observational studies. Regulatory Affairs Professionals Society. https://www.raps.org/resource/ich-adopts-the-m14-guideline-on-observational-stud.html

ENCePP. (n.d.). European Network of Centres for Pharmacoepidemiology and Pharmacovigilance. European Medicines Agency. Retrieved October 21, 2026, from https://encepp.europa.eu/index_en

European Medicines Agency. (n.d.). Patient registries. Retrieved October 21, 2026, from https://www.ema.europa.eu/en/human-regulatory-overview/post-authorisation/patient-registries

European Medicines Agency. (2026). Real-world evidence framework to support EU regulatory decision-making: 4th report on the experience gained with regulator-led studies from February 2025 to February 2026 (EMA/66577/2026). https://www.ema.europa.eu/en/documents/report/real-world-evidence-framework-support-eu-regulatory-decision-making-4th-report-experience-gained-regulator-led-studies-february-2025-february-2026_en.pdf

Health Canada. (2019, April 16). Optimizing the use of real world evidence to inform regulatory decision-making. Government of Canada. https://www.canada.ca/en/health-canada/services/drugs-health-products/drug-products/announcements/optimizing-real-world-evidence-regulatory-decisions.html

Medicines and Healthcare products Regulatory Agency. (2021, December 16). MHRA guidance on the use of real-world data in clinical studies to support regulatory decisions (updated July 17, 2025). GOV.UK. https://www.gov.uk/government/publications/mhra-guidance-on-the-use-of-real-world-data-in-clinical-studies-to-support-regulatory-decisions

U.S. Food and Drug Administration. (n.d.). Real-world evidence. Retrieved October 21, 2026, from https://www.fda.gov/science-research/science-and-research-special-topics/real-world-evidence

U.S. Food and Drug Administration. (2021). FDA approves new use of transplant drug based on real-world evidence. https://www.fda.gov/drugs/news-events-human-drugs/fda-approves-new-use-transplant-drug-based-real-world-evidence

U.S. Food and Drug Administration. (2022, February 4). FDA's Sentinel Initiative: Background. https://www.fda.gov/safety/fdas-sentinel-initiative/fdas-sentinel-initiative-background

U.S. Food and Drug Administration. (2024). Real-world data: Assessing electronic health records and medical claims data to support regulatory decision-making for drug and biological products [Guidance for industry]. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/real-world-data-assessing-electronic-health-records-and-medical-claims-data-support-regulatory

U.S. Food and Drug Administration. (2025a). Considerations for the use of artificial intelligence to support regulatory decision-making for drug and biological products [Draft guidance for industry]. Retrieved October 21, 2026, from https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-use-artificial-intelligence-support-regulatory-decision-making-drug-and-biological

U.S. Food and Drug Administration. (2025b). Use of real-world evidence to support regulatory decision-making for medical devices; Guidance for industry and Food and Drug Administration staff; Availability. Federal Register, 90(241). https://www.govinfo.gov/content/pkg/FR-2025-12-18/html/2025-23252.htm

Wang, S. V., Schneeweiss, S., Franklin, J. M., Desai, R. J., Feldman, W., Garry, E. M., Glynn, R. J., Lin, K. J., Paik, J., Patorno, E., Suissa, S., D'Andrea, E., Jawaid, D., Lee, H., Pawar, A., Sreedhara, S. K., Tesfaye, H., Bessette, L. G., Zabotka, L., . . . Quinto, K. (2023). Emulation of randomized clinical trials with nonrandomized database analyses: Results of 32 clinical trials. JAMA, 329(16), 1376–1385. https://doi.org/10.1001/jama.2023.4221