Regulatory decision-making has traditionally relied on randomized controlled trials (RCTs) as the gold standard for demonstrating safety and efficacy. However, the limitations of this approach—high costs, lengthy timelines, and narrow patient populations—have driven increasing interest in real-world data (RWD) and the real-world evidence (RWE) derived from it.

Real-world data refers to health-related information collected from sources other than traditional clinical trials, including electronic health records, registries, claims databases, and patient-generated data. When analyzed appropriately, this data can generate real-world evidence capable of supporting regulatory decisions across the product lifecycle.

The field has reached a critical inflection point. In December 2025, the FDA issued updated final guidance on using RWE to support regulatory decision-making for medical devices, superseding its 2017 guidance. Meanwhile, the European Medicines Agency has continued advancing its RWE framework, with the DARWIN EU network now conducting collaborative studies across member states.

This article examines the latest research and regulatory developments shaping RWD integration, the frameworks governing its use, and the challenges that remain for broader adoption. For broader context on AI's role in healthcare research, see The Role of Artificial Intelligence in Medical Research.

What Is Real-World Data Integration?

Real-world data integration refers to the process of combining, harmonizing, and analyzing data from multiple real-world sources to generate evidence suitable for regulatory purposes.

Unlike data from controlled clinical trials, RWD is collected during routine clinical care. Sources include:

  • Electronic health records (EHRs) – capturing diagnoses, treatments, laboratory results, and clinical notes
  • Administrative claims and billing data – documenting healthcare utilization and costs
  • Patient registries – systematic collections of data on patients with specific conditions or treatments
  • Wearable devices and digital health platforms – generating continuous physiological and behavioral data
  • Patient-reported outcomes – capturing symptoms and quality of life from the patient perspective

The integration challenge lies in combining these heterogeneous sources into coherent datasets that can answer specific regulatory questions.

FDA Guidance: Relevance and Reliability Framework

The FDA's updated guidance establishes a structured framework for evaluating whether RWD is suitable for regulatory decision-making.

Study Protocol Requirements

The guidance emphasizes that study protocols and analysis plans should be created prior to analyzing real-world data, regardless of whether the data already exist or will be collected prospectively. These documents should be finalized before reviewing outcome data and before performing prespecified analyses.

Relevance Assessment

Relevance evaluates whether the data can credibly address the regulatory question. Key considerations include:

  • Data availability – Whether the data contains sufficient detail to capture the exposure, outcomes, and covariates needed for the study population
  • Linkages – Whether data from different sources can be integrated to address the study question
  • Timeliness – Whether the time between data collection and release reflects the current clinical environment
  • Generalizability – Whether the study sample represents the intended use population

Reliability Assessment

Reliability evaluates whether the data is accurate and complete enough to support credible conclusions. This includes:

  • Data accrual – Adequacy of information about data sources and collection methods
  • Data quality and integrity – Quality control processes, completeness assessments, and consistency checks
  • Patient protections – Privacy safeguards and adherence to ethical standards
  • Prior demonstration – Whether the data source has previously generated usable RWE

Medical Journal IMJ Health Call for Papers

Global Regulatory Perspectives

United States: Expanding RWE Acceptance

The FDA has significantly expanded its RWE initiatives. In 2026, the agency published 73 examples of marketing authorizations using RWE from FY2020-2025, demonstrating a broad spectrum of clinical applications.

These examples highlight growing use of RWD to validate AI and machine learning-enabled device software. For instance, the FDA relied on retrospective medical record data from intensive care unit stays to support clearance of an ML-based software function predicting hemodynamic instability.

European Union: DARWIN EU and Harmonization

The EMA has established DARWIN EU (Data Analysis and Real World Interrogation Network) to provide access to real-world healthcare data across Europe. The network has launched collaborative studies on topics including GLP-1 receptor agonists and vaccine safety signal evaluation, with findings expected in 2026.

The EMA has also developed a Data Quality Framework for RWD, with a draft chapter on RWD quality undergoing public consultation. This framework aims to standardize quality expectations across data sources and study types.

Japan: Longitudinal Trends

A systematic review of Japanese regulatory approvals from 2019-2024 examined the evolving use of RWD/E in new drug and regenerative medical product applications. This six-year analysis characterized how RWE utilization has developed alongside regulatory initiatives, including the publication of basic guidance documents for registry data use.

Comparative Analysis: FDA vs. EMA

A 2025 comparative review of RWE standards across regulatory and health technology assessment bodies identified important differences:

  • FDA applies more flexibility in evaluating RWE as the basis for drug approvals compared to the EMA
  • EMA demonstrates more advanced registry integration maturity and governance expectations, likely due to established European registry and data-sharing networks
  • Both agencies are increasingly accepting external control arms, though the FDA is less restrictive

Emerging Methodologies in RWD Integration

Target Trial Emulation

Target trial emulation is a methodological framework that applies the principles of randomized trial design to observational data analysis. By explicitly defining the target trial's eligibility criteria, treatment strategies, and outcomes, researchers can design observational analyses that approximate the rigor of RCTs.

A 2026 study applied this framework across four decentralized health systems—INSIGHT, OneFlorida+, UPHS, and YNHHS—to identify repurposable medications for Alzheimer's disease and related dementias. The multi-site approach enhanced the generalizability and credibility of findings.

AI-Powered Data Integration

Artificial intelligence is increasingly being applied to RWD integration challenges.

A 2026 study in Nature Communications described EmulatRx, a multi-agent AI framework designed to support clinical trial design using real-world data. The system integrates clinical trial knowledge graphs, literature retrieval, statistical analysis tools, and subgroup analysis capabilities to generate comprehensive trial design reports.

The framework demonstrated capacity to optimize eligibility criteria using Shapley-based attribution methods and identify heterogeneous treatment effects that might be masked in aggregate population analyses.

Federated Learning and Interoperability

The SHIELD project, part of Horizon Europe's effort to reduce non-communicable diseases, has developed a federated system integrating retrospective and prospective clinical data using LLM-based multi-agent systems. The system harmonizes data into OMOP Common Data Model and FHIR standards, demonstrating feasibility of scalable and interoperable data integration.

This federated approach addresses privacy concerns by enabling analysis across distributed data sources without centralizing sensitive patient information.

Adaptive ETL Orchestration

Real-time integration of biomedical data presents challenges related to latency, data freshness, and resource utilization. A 2026 IEEE study proposed an adaptive Extract-Transform-Load (ETL) framework using reinforcement learning to dynamically adjust synchronization intervals based on system conditions. The approach demonstrated reduced synchronization latency compared to fixed-interval scheduling.

Applications in Medical Device Regulation

The FDA's updated guidance provides specific recommendations for medical device submissions using RWE.

When IDE Requirements Apply

The guidance clarifies that if data are being gathered to determine safety and effectiveness, and the process for gathering the data would influence treatment decisions, an Investigational Device Exemption (IDE) may be required. Three hypothetical examples illustrate when an IDE may or may not be needed for clinical studies using RWD.

Real-World Data in EUAs

Clinical data routinely collected from devices authorized under Emergency Use Authorizations may be considered RWD and used to support regulatory decision-making if determined to be relevant and reliable.

Postmarket Applications

RWD can support post-approval studies imposed as conditions of device approval, potentially precluding the need for or addressing postmarket surveillance requirements. It can also generate evidence for expanding labeling to include additional indications or updated safety and effectiveness information.

Applications in Drug Development and Pharmacovigilance

Oncology

RWE has become increasingly important in oncology regulatory submissions. A review of European regulatory and health technology assessment decisions for oncology medicines examined the use of RWE in EMA reviews and HTA assessments by NICE, G-BA, and HAS.

In the UK, a systematic review of NICE oncology technology appraisals assessed how RWD has been used for both clinical evidence and cost-effectiveness analyses, documenting statistical methods and committee feedback. For related insights on oncology research, see Oncology Research 2026: Immunotherapy, Targeted Therapy, and Precision Medicine Advances.

Rare Diseases

Rare diseases present particular challenges for traditional RCTs due to small patient populations and ethical constraints. RWE generation in rare diseases has evolved to include digital twin technology—individualized computational replicas integrating multi-omics, clinical, and longitudinal data—that can serve as synthetic controls and support outcome prediction.

Drug Repositioning

A comprehensive review examined the role of RWE in validating AI-generated drug-repositioning candidates. Validation using EHR and insurance databases enabled retrospective assessment of drug efficacy across large populations, with successful applications identified in neurodegenerative, metabolic, infectious, autoimmune, and psychiatric diseases.

Challenges and Limitations

Data Quality and Standardization

The effectiveness of RWD integration remains dependent on data quality. The EMA's public consultation on its RWD reflection paper received 695 comments from 39 stakeholders, with many requesting additional guidance on data quality expectations and methodology.

The EMA noted that AI, ML, and NLP technologies are not mature enough for comprehensive recommendations, though study protocols should detail methodologies used to evaluate performance, risk of bias, and impact on results.

Generalizability

A model that performs well in one healthcare system may perform poorly elsewhere. Differences in patient demographics, clinical practices, data structures, and disease prevalence can affect performance.

Prospective Validation

One of the biggest gaps remains the difference between promising retrospective performance and demonstrated prospective clinical utility. An algorithm can achieve excellent accuracy on historical data without improving real-world outcomes.

Regulatory Divergence

Despite shared principles, regulatory entities maintain varying standards for data robustness and validity. Harmonizing policies across jurisdictions poses challenges due to differences in regulatory mandates and decision-making frameworks.

The Role of RWE in Precision Medicine

RWE is increasingly recognized as complementary to RCTs in supporting precision medicine. A 2025 review described how RWE can address questions that clinical trials typically do not answer about treatment benefits and risks, ultimately impacting public health by guiding decision-making across the healthcare ecosystem.

The integration of RWD with genomic data, biomarker information, and patient-reported outcomes creates opportunities for more personalized regulatory assessments and post-market surveillance. For related insights on biostatistics and research methods, see Biostatistics and Health Research Methods: Data-Driven Decision Making in Medicine.

Future Directions

Real-Time Regulatory Reporting

The FDA's 2026 initiative on real-time clinical trials represents an important shift toward more responsive regulatory models. Proof-of-concept trials designed to report endpoints and data signals to regulators in real time could fundamentally change how evidence is generated and evaluated.

AI-Assisted RWE Generation

As AI capabilities advance, the integration of RWD with AI-powered analytics will likely accelerate. The EmulatRx framework and similar systems demonstrate the potential for AI to support trial design, patient stratification, and outcome prediction using real-world data.

Global Harmonization

The International Council for Harmonisation (ICH) has made RWE integration a strategic priority. The EMA's Network Data Steering Group has developed operational principles and templates for collaboration, with initial collaborative studies expected to publish findings in 2026.

Digital Twins and Synthetic Controls

Digital twin technology—individualized computational replicas of patients—represents an emerging frontier in RWE generation. These models can serve as synthetic controls, personalize treatment arms, and support outcome prediction, reducing trial size and duration.

Real-world data integration is transforming regulatory decision-making by enabling evidence generation from routine clinical practice rather than relying exclusively on controlled trial environments.

The FDA's updated guidance provides a structured framework for evaluating RWD relevance and reliability, while expanding examples demonstrate the breadth of applications across medical devices and drug development. The EMA's DARWIN EU network and Data Quality Framework represent significant investments in European RWD infrastructure.

Emerging methodologies—including target trial emulation, AI-powered data integration, and federated learning—are addressing historical limitations of observational research. These approaches enable more rigorous analyses while respecting privacy constraints.

However, challenges remain. Data quality varies across sources, regulatory standards differ across jurisdictions, and prospective validation of RWE-derived conclusions is still limited. The field is evolving toward greater standardization and harmonization, but this transition will require continued collaboration between regulators, industry, and academic researchers.

For researchers and regulatory professionals, the integration of RWD represents both an opportunity and a responsibility: the opportunity to generate evidence that better reflects real-world patient experiences, and the responsibility to ensure that this evidence meets the rigorous standards required for regulatory decision-making. For those considering doctoral research in this area, Top 10 Pharmaceutical Research Topics for PhD offers guidance on selecting impactful research directions.

Continue Exploring Our Blog