How to Choose a JCA Dossier Preparation Platform for EU Joint Clinical Assessment

EU JCA Dossier Submission Requirements and Deadlines: An Overview

Selecting the right JCA dossier preparation platform determines whether health technology developers can meet the 100-day (or 60-day accelerated) submission window under Regulation (EU) 2021/2282 while producing complete, transparent, and audit-ready clinical evidence. The European Union Joint Clinical Assessment (JCA) requires developers to submit a structured dossier that answers every Population-Intervention-Comparator-Outcome (PICO) element in the final assessment scope. The Commission requests the dossier after the scope is finalized. Developers then have 100 days in standard cases (60 days for accelerated procedures or certain variations). Literature searches must be no more than three months old at submission. Completeness checks occur within 15 working days, and any gaps can trigger a second request or, in documented cases, discontinuation of the assessment.

National pricing and reimbursement decisions remain with Member States, yet the shared clinical report now forms the starting point for those discussions. Early oncology and advanced therapy products already operate under this regime. High-risk medical devices entered in 2026, with further expansion planned for orphans in 2028 and all new medicinal products in 2030.

Teams therefore need platforms that support scope anticipation, systematic literature review (SLR), multi-PICO evidence mapping, indirect comparisons where required, and fully traceable dossier assembly. General-purpose AI tools rarely meet the methodological and disclosure standards set by the Health Technology Assessment Coordination Group (HTACG). Purpose-built solutions designed for life-sciences evidence workflows do.

Why JCA Dossier Preparation Breaks Conventional Review Tools

The JCA breaks conventional tools because the scope arrives as a matrix of PICOs, the clock is short, and the dossier will be published with its AI use disclosed.

What the EU HTA Regulation Requires

Regulation (EU) 2021/2282 replaced voluntary European cooperation with a mandatory joint clinical assessment that every Member State must use in national appraisal (European Union, 2021). JCAs became mandatory on 12 January 2025 for new oncology medicines and advanced therapy medicinal products. The HTA Coordination Group (HTACG) began selecting high-risk medical devices from June 2026, orphan medicines enter scope in January 2028, and all remaining centrally authorized medicines follow in 2030 (PharmExec, 2026). Each step adds dossiers that must land inside a fixed window.

Challenges in JCA Dossier Preparation

The first constraint is scope volatility. An ISPOR Europe 2025 simulation that followed the published consolidation guidance still produced 67 PICOs from 25 countries, 60 of them driven by biomarker or histology subpopulations (ISPOR, 2025). A platform that stores the scope as a document rather than as structured objects cannot re-cut the evidence base when the consolidated PICO list differs from the one you predicted.

The second constraint is time. Developers have 100 days to prepare the dossier after receiving the final scope, or 60 days for accelerated procedures and extensions of indication. The dossier must also reach the HTACG at least 45 days before the envisaged CHMP opinion, and an ISPOR Europe 2024 timeline analysis found that this rule can cut effective writing time to about 65 days (ISPOR, 2024). The dossier guidance adds that literature searches can be no more than three months old at submission (European Commission, 2024).

The third constraint is disclosure. In July 2026, the HTACG adopted general principles on the use of artificial intelligence in JCA dossier preparation. The principles make the developer fully responsible for AI-generated content, require human oversight throughout every AI-assisted step, and require the dossier to report the tool name, version, date, developer, and intended purpose (HTACG, 2026). Because JCA dossiers are published, the HTACG also expects compliance with the EU AI Act, copyright, and data protection law. A platform that cannot produce that disclosure automatically pushes the work back onto your team.

What Does a JCA Dossier Preparation Platform Actually Do?

A Five-Layer Platform for JCA Dossier Preparation

The platform turns the anticipated scope into the index for every downstream artifact, so that a scope change becomes a re-cut of an indexed library rather than a new review. The companion article on JCA scoping to evidence synthesis describes the six-stage workflow in detail. The architecture that makes it work has five layers.

1. Source ingestion. The platform indexes national HTA decisions, clinical guidelines, drug registries, and trial databases across Member States, and version-stamps every search so that a refresh is a repeat execution.
2. PICO data model. Each anticipated and consolidated PICO is a structured object with population, intervention, comparator, outcome, and study design fields. Every study, data point, and decision carries a reference to the PICO cells it informs.
3. AI services. Models anticipate comparators and outcomes, triage titles and abstracts, extract arm-level data, and flag PICO cells with no direct evidence. Each service exposes its model reference, threshold, and confidence.
4. Human review and sign-off. A named reviewer confirms, edits, or rejects every AI output. The platform stores the decision, the reason, and the reviewer identity.
5. Dossier output and disclosure. Evidence tables, PRISMA flow diagrams, gap justifications, and the AI-use statement generate from the same indexed library.

Validation against the first three published JCA reports shows what this architecture delivers. Given only the product name, licensed indication, and mechanism of action, a platform built on these layers identified 10 of 11 JCA-defined comparators across tovorafenib, lurbinectedin, and tarlatamab, and produced a reviewer-ready anticipated scope in under one hour against a manual baseline of three to five weeks (MadeAi, 2026). The full results appear in AI-powered EU JCA readiness: what three published assessments reveal. The sample is small, but it shows what you should expect a vendor to demonstrate on your own asset.

Seven Criteria for Choosing a JCA Dossier Preparation Platform

Score every candidate on seven criteria, weight them by your portfolio, and reject any platform that fails a criterion marked as a gate.

1. PICO-Native Data Model (Gate)

Ask the vendor to add a comparator to a completed scope and show which studies, extractions, and evidence tables update. If the answer involves re-running the review, the platform stores documents, not structure. This criterion is a gate because every other capability depends on it.

2. Multi-Source Scope Anticipation

The platform should draw on HTA reports, guidelines, and registries across all 27 Member States and link every proposed comparator and outcome to its source document with a rationale. Ask for comparator recall against published JCA scopes and how the vendor measured it.

3. Search Currency and Living Review Support (Gate)

The three-month search rule makes a scheduled, version-stamped refresh non-negotiable. Ask how long a full refresh takes on a review of your size and whether the platform can run a delta refresh without reopening extraction. Systematic literature review services that already run living reviews can supply the baseline timings you should expect.

4. Extraction Accuracy With Routed Verification

A 2026 systematic review of large language model data extraction found categorical and string variables extracted more reliably (74 to 96 percent) than numerical data (47 to 88 percent), with omissions the dominant error type (Journal of Biomedical Informatics, 2026). A credible platform routes numeric extraction to human verification by default and shows the confidence score beside every value. Ask the vendor for accuracy figures from your therapeutic area, not a benchmark from another domain.

5. HTACG-Ready AI Disclosure

The platform must generate the AI-use statement the HTACG principles require, with tool name, version, date, developer, intended purpose, and a description of the human oversight applied at each step. It must also keep a decision-level audit trail that a reviewer can reproduce from a dated snapshot. A platform that offers a document-level change log only does not meet this bar. 

6. Portfolio Reuse

The same evidence map should serve the JCA dossier, later national submissions where Member States request extra comparators, and the next indication in the same disease. Ask what the marginal effort is for a second indication that shares comparators with the first.

7. Security, Hosting, and Legal Compliance

Check data residency, whether your data trains vendor models, and how the platform handles copyright on full-text articles. The HTACG’s May 2026 data transparency principles state that evidence synthesis approaches are not commercially confidential and will be published (RAPS, 2026), so the platform must separate confidential development plans from the methods it will disclose.

Table 1: Weighted evaluation matrix for platform selection

What Benefits Should You Expect from the Right Platform?

The main benefit is faster adaptation to scope changes. As a result, teams can update evidence in days rather than reopen the entire review. Furthermore, the platform offers several operational benefits.

• Earlier start. Teams begin protocol and search against the anticipated scope months before the official PICO survey, which is where market access solutions and evidence teams first share one data model.
• Protected expert time. Specialists move from assembling comparator lists and screening titles toward interpretation, gap justification, and assessor questions.
• Faster responses to assessors. A decision-level audit trail answers “why was this study excluded” in minutes rather than days.
• Lower marginal cost per indication. The second asset in a disease inherits the search infrastructure and extracted comparator data.

Where Does the Platform Fit in Practice?

Four scenarios recur across pharma evidence teams.

• Oncology asset with a multi-biomarker scope. The team records predicted subpopulation PICOs before the survey opens. When the consolidated scope adds one comparator, the team filters the library, runs a targeted search, and regenerates evidence tables within two weeks.
• Orphan portfolio preparing for 2028. One living library indexed by PICO serves several assets in the same rare disease, so each new JCA concentrates effort on the intervention arm.
• Post-JCA national follow-through. A Member State requests an additional comparator. The team extends the existing map rather than commissioning a new review.
• Label change after the first assessment. The implementing act provides for re-initiation when the indication changes, so a versioned evidence map lets the team submit a delta rather than a rebuild.

Best Practices for Selecting and Implementing a Platform

Define the data model first, validate on your own completed reviews, and treat disclosure as a deliverable from day one.

1. Write the PICO object, extraction schema, and provenance fields before you issue a request for proposal. A platform that cannot represent that structure will recreate the silos you are trying to remove.
2. Run the live demo on your own asset. Supply only the product name, licensed indication, and mechanism of action, and compare the anticipated scope against a published JCA in the same therapeutic area.
3. Validate AI components against a completed internal review. Recall and precision vary by therapeutic area and study design.
4. Insist on decision-level audit logs. Document-level change tracking cannot reproduce a dossier from a dated snapshot.
5. Check the disclosure export against the HTACG principles line by line. The joint position statement from Cochrane, Campbell, JBI, and the Collaboration for Environmental Evidence and the NICE position statement on AI in evidence generation set the same expectation (Cochrane et al., 2025; NICE, 2025).
6. Name a refresh owner and a refresh cadence before go-live, with triggers for incremental versus full updates.
7. Pilot on one asset, then scale. Measure time to anticipated scope, time to delta refresh, and reviewer edit rate on extraction.

Risks and Limitations

Key Considerations for AI-Assisted JCA

A platform reduces assembly time; it does not remove the need for clinical and methodological judgment. Predicted PICOs can still miss unexpected national questions, and assessor-defined composite endpoints cannot be derived from published sources alone. Over-reliance on automated extraction for numeric outcomes invites exactly the omission errors the published evidence documents. Living reviews also require sustained organizational ownership; without it, they drift out of the three-month window.

Guidance from HTA bodies is still consolidating. The safe posture is to apply the more stringent of existing national standards and the HTACG principles, and to disclose AI use fully even where a national body has not yet asked.

Future Trends in JCA Dossier Preparation

Three developments will shape the next three years. Portfolio load rises sharply as orphan medicines enter scope in 2028 and all remaining medicines follow in 2030. Reporting standards for AI-assisted synthesis are consolidating around frameworks such as RAISE, so disclosure becomes easier to standardize and harder to skip. And the same PICO-indexed evidence base increasingly feeds national appraisals and real-world evidence programs, so the return on maintaining it keeps rising. As an AI platform for Life Sciences matures from review acceleration toward evidence operations, platform choice becomes an infrastructure decision rather than a project decision.

Conclusion

Choose a JCA dossier preparation platform on structure, currency, and disclosure, in that order. The right platform carries one PICO data model from anticipation to submission, refreshes within the three-month window on a schedule, and exports an AI-use statement that satisfies the HTACG principles without manual rework. Everything else, including speed, is a consequence of those three properties.

A practical next step: take one completed review and check whether it can be re-cut against a new comparator without reopening extraction. If it cannot, the gap is in the data model, and that is the first thing to test in any vendor demo. Teams exploring AI in Life Science evidence workflows can run that test on their own or with pharma evidence generation solutions partners that already operate living reviews for JCA, including MadeAi.

Author’s Note: This article was supported by AI-based research and writing, with Claude 5.1 assisting in the creation of text and images.