Multi-Jurisdictional Drug Development: Navigating FDA, EMA, and MHRA Divergence with Context Engineering

The ICH was established in 1990 with an ambitious vision: create a unified set of technical requirements for pharmaceutical development that would be accepted by all major regulatory authorities, eliminating the need for duplicative studies and enabling truly global drug development. For certain aspects of drug development—Good Clinical Practice (GCP), stability testing, impurity limits—the ICH achieved meaningful harmonization. A stability study conducted according to ICH Q1A is accepted by the FDA, EMA, and MHRA without modification.

However, the areas where harmonization has failed are precisely the areas that determine development timelines and costs. Clinical trial design, endpoint selection, statistical analysis plans, and post-marketing safety commitments remain subject to jurisdiction-specific interpretation and requirements. The result is that a biotech company developing a novel oncology therapy cannot simply design one global Phase III trial and submit the same dossier to the FDA, EMA, and MHRA. Each agency will have different expectations, and discovering these differences late in development can be catastrophic 1.

The divergence is not merely bureaucratic—it reflects genuine scientific and policy differences. The FDA’s accelerated approval pathway emphasizes surrogate endpoints and post-marketing confirmatory studies, enabling earlier patient access with ongoing risk management. The EMA’s conditional marketing authorization requires more robust upfront evidence and is more restrictive about which surrogate endpoints are acceptable. The MHRA, post-Brexit, is charting its own course with the Innovative Licensing and Access Pathway (ILAP), which offers rolling review and early scientific advice but imposes unique real-world evidence requirements that neither the FDA nor EMA demand 2.

The cost of regulatory divergence is not hypothetical—it is measurable in delayed patient access, abandoned development programs, and geographic inequities in drug availability. A 2025 analysis of oncology drug approvals found that only 40% of drugs approved by the FDA in 2023-2024 achieved simultaneous or near-simultaneous EMA approval. The median lag between FDA and EMA approval was 14 months, driven primarily by the need for supplementary data or analyses that were not required by the FDA but were deemed essential by the EMA’s Committee for Medicinal Products for Human Use (CHMP) 3.

The strategic challenge is compounded by the fact that regulatory requirements are not static. The FDA’s January 2025 draft guidance on AI/ML-enabled drug development tools introduced new expectations for algorithm transparency and validation that were not anticipated when many current development programs were initiated. The EMA’s sustainability and patient-centricity initiatives are reshaping expectations for patient-reported outcomes and environmental impact assessments in regulatory submissions. The MHRA’s implementation of the new Clinical Trials Regulation in April 2026 will introduce procedural changes that affect trial authorization timelines and sponsor obligations 4.

RuleMedi’s solution to multi-jurisdictional complexity is not to harmonize the regulators—that is beyond any company’s control—but to enable dynamic adaptation to jurisdiction-specific requirements through context engineering. This approach treats regulatory jurisdiction not as a binary choice (FDA or EMA) but as a contextual parameter that shapes every aspect of the AI system’s reasoning and output.

The architecture is built on Google Cloud’s Vertex AI platform, specifically leveraging Context Caching to manage the massive regulatory frameworks that govern drug development in each jurisdiction. The FDA’s guidance documents, EMA’s scientific guidelines, and MHRA’s regulatory procedures are each ingested into separate context layers—comprehensive, structured representations of the regulatory expectations, precedents, and interpretive nuances specific to that authority. When RuleMedi’s COMPASS agent is tasked with generating a clinical development plan, it does not apply a generic, jurisdiction-agnostic template. It dynamically loads the relevant jurisdictional context and reasons within that framework 5.

For example, when a user asks COMPASS to design a Phase III trial for a novel immunotherapy in metastatic melanoma, the system first asks: “Which regulatory jurisdictions are you targeting for initial approval?” If the answer is “FDA and EMA,” COMPASS loads both the FDA oncology guidance context and the EMA oncology guideline context simultaneously. It then identifies areas of alignment and divergence. Both agencies accept progression-free survival (PFS) as a primary endpoint for accelerated approval in melanoma—this is an area of alignment. However, the FDA accepts a single-arm trial with external control if the effect size is large and the unmet need is high, while the EMA requires a randomized controlled trial except in the most exceptional circumstances—this is a divergence.

COMPASS presents the user with a strategic decision framework. Option A: Design a randomized controlled trial that satisfies both agencies, with PFS as the primary endpoint. This maximizes the probability of simultaneous approval but increases trial size, cost, and timeline. Option B: Design a single-arm trial optimized for FDA accelerated approval, with a plan to conduct a post-approval confirmatory study that will support subsequent EMA submission. This accelerates U.S. market access but delays European approval by 18-24 months. Option C: Design a randomized trial for EMA, then submit the same data to the FDA with a request for priority review based on the robust evidence. This may satisfy both agencies but forgoes the FDA’s accelerated pathway.

The power of context engineering extends beyond initial trial design to ongoing adaptation as regulatory requirements evolve. RuleMedi’s CATALYST agent continuously monitors regulatory signals from the FDA, EMA, and MHRA—not just final published guidance, but draft documents, advisory committee discussions, and public statements by agency officials. When CATALYST detects an emerging regulatory expectation that could impact an ongoing development program, it automatically assesses the jurisdictional scope and client-specific impact.

In early 2025, the FDA published draft guidance on the use of real-world evidence (RWE) to support regulatory decision-making for oncology drugs. The guidance introduced new expectations for data quality, confounding control, and sensitivity analyses that went beyond previous FDA statements on RWE. CATALYST identified this guidance as high-impact for clients with oncology programs planning to use RWE for post-approval label expansions or confirmatory studies.

For one client—a biotech company with an FDA-approved immunotherapy planning a post-approval study using electronic health record (EHR) data to support a new indication—CATALYST performed an automated impact assessment. It compared the client’s planned study design against the new FDA expectations and identified three gaps: the planned data source did not include patient-reported outcomes, the confounding adjustment strategy did not account for unmeasured confounders using sensitivity analyses, and the statistical analysis plan did not include the FDA’s recommended approaches for handling missing data in RWE studies.

CATALYST generated a jurisdictional context report showing that the EMA had not published equivalent guidance and continued to apply its existing, more conservative framework for RWE acceptance. This created a strategic fork: the client could adapt the study design to meet the new FDA expectations, potentially making the study more robust but also more expensive and complex. Or the client could proceed with the original design, which still met the EMA’s expectations but now carried higher risk of FDA rejection or requests for supplementary analyses.

A UK-based biotech company developing a gene therapy for a rare metabolic disorder faced the classic multi-jurisdictional challenge. The disease affected approximately 500 patients annually across the U.S., EU, and UK combined—too small a population to conduct separate trials for each jurisdiction, but large enough that regulatory divergence could not be ignored. The company needed a single pivotal trial that would support simultaneous submissions to the FDA, EMA, and MHRA.

The company engaged RuleMedi’s CONDUCTOR agent to orchestrate the multi-jurisdictional development strategy. CONDUCTOR loaded the jurisdictional contexts for all three agencies and performed a comprehensive alignment analysis. It identified several areas of divergence that required strategic decisions:

Endpoint Selection: The FDA’s draft guidance on gene therapies for rare diseases indicated openness to biochemical endpoints (reduction in toxic metabolite levels) as primary endpoints if correlated with clinical benefit. The EMA’s CHMP guideline on advanced therapy medicinal products (ATMPs) expressed preference for clinical endpoints (survival, organ function) even in rare diseases. The MHRA’s position, articulated through its ILAP framework, aligned more closely with the FDA’s pragmatic approach but emphasized the need for long-term follow-up data.

Comparator Selection: The FDA indicated that a single-arm trial with natural history comparison could be acceptable given the rarity of the disease and the lack of approved therapies. The EMA’s preference was for a randomized controlled trial, but acknowledged that external control using natural history data was acceptable if the natural history was well-characterized and the treatment effect was large and durable. The MHRA aligned with the EMA’s position but requested specific statistical methods for external control analysis.

Manufacturing and Quality: All three agencies required compliance with GMP, but the EMA’s specific requirements for viral vector characterization were more prescriptive than the FDA’s, and the MHRA had recently published additional expectations for process validation that went beyond both the FDA and EMA.

CONDUCTOR generated a unified trial design that satisfied all three agencies’ core requirements while minimizing redundancy. The primary endpoint was a composite of biochemical and clinical measures—the biochemical component satisfied the FDA’s pragmatic approach, while the clinical component addressed the EMA’s preference for patient-relevant outcomes. The trial design was single-arm with a robust natural history comparator, using the statistical methods specifically recommended by the MHRA and accepted by the EMA. The manufacturing process was designed to the most stringent requirements (EMA’s viral vector characterization and MHRA’s process validation), ensuring that the same manufacturing data package would satisfy all three agencies.

The company submitted the trial protocol to all three agencies for scientific advice. The FDA’s feedback was positive, with no major concerns. The EMA requested minor clarifications on the natural history comparator methodology, which were addressed in a protocol amendment. The MHRA provided the most detailed feedback, requesting additional long-term follow-up data and a more comprehensive pharmacovigilance plan, both of which were incorporated.

The trial was conducted as a single global study, with patients enrolled in the U.S., UK, and EU. The results showed a large, clinically meaningful treatment effect with an acceptable safety profile. The company submitted to all three agencies within a 60-day window. The FDA granted accelerated approval 8 months after submission. The EMA granted conditional marketing authorization 10 months after submission.

The MHRA granted approval through the ILAP pathway 9 months after submission. The total time from trial completion to approval in all three jurisdictions was 10 months—a timeline that would have been impossible without the unified, context-aware development strategy enabled by RuleMedi’s multi-agent architecture.

The technical foundation for RuleMedi’s multi-jurisdictional capabilities is Google Cloud’s global infrastructure and AI services. Vertex AI’s Context Caching enables the simultaneous loading of multiple jurisdictional contexts—FDA, EMA, MHRA, and others—without performance degradation. Each context is a comprehensive representation of that agency’s regulatory framework, including guidance documents, precedent decisions, advisory committee transcripts, and scientific literature on regulatory science specific to that jurisdiction.

When COMPASS or CONDUCTOR performs a multi-jurisdictional analysis, it does not sequentially query each jurisdiction’s context and then manually compare the results. It loads all relevant contexts into a unified reasoning space and performs parallel analysis, identifying alignments and divergences in real-time. This is computationally intensive—a single multi-jurisdictional query may involve processing millions of tokens across multiple regulatory frameworks—but Vertex AI’s distributed infrastructure and caching mechanisms make it feasible at scale 6.

The era of truly harmonized global drug development remains aspirational. The FDA, EMA, and MHRA are not converging—they are diverging, each pursuing jurisdiction-specific priorities in patient access, innovation policy, and risk management. For pharmaceutical and biotech companies, this divergence is not a problem to be solved—it is a reality to be navigated.

RuleMedi’s context engineering architecture transforms multi-jurisdictional complexity from an existential threat into a manageable strategic challenge. By dynamically adapting to jurisdiction-specific requirements, proactively monitoring regulatory evolution, and enabling simultaneous multi-agency development strategies, RuleMedi empowers companies to pursue global market access without multiplying development costs or timelines.