Before a new cancer medicine or advanced therapy (a medicine based on genes, tissues or cells) reaches patients in the European Union, the 27 countries now ask their clinical questions together, once. The company has 100 days to answer every one of those questions in a single file of evidence, the dossier, with a trial, a statistical comparison, or a written reason why no answer exists. This is the EU joint clinical assessment, or JCA, and three assessments have already been stopped because the dossier fell short.
This guide is for anyone who plans, writes, checks, signs or pays for a JCA dossier: EU market access leads, health technology assessment (HTA) leads, health economics and outcomes research (HEOR) teams, biostatisticians, medical writers, clinical and regulatory reviewers, and the leaders deciding how to staff or automate the work. No background needed.
Imagine you are the EU HTA dossier lead at Northwind Bio, a mid-sized biotech with one lung cancer medicine under review at the European Medicines Agency. It is Wednesday, October 7, 2026. The medicine, Dravenzo (corventamab vedotin), treats non-squamous lung cancer that has come back after platinum chemotherapy and immunotherapy. The assessors' list of questions, the scope, reached you on August 31: 11 questions, called PICOs, each with 8 outcomes to report, so 88 boxes to fill. The dossier is due Wednesday, December 9. Today is day 37 of 100.
PICO 2, on page 4 of the scope, asks how Dravenzo compares with docetaxel, a chemotherapy, plus a second drug, an angiokinase inhibitor, in patients with adenocarcinoma, one type of non-squamous lung cancer. Seven countries asked for it, Germany and Austria among them. Your main trial, DVZ-301, with 612 patients, compared Dravenzo with docetaxel alone.
Assessment scope · PICO 2, p. 4
Population
Adenocarcinoma, after platinum and immunotherapy
Comparator
Docetaxel + angiokinase inhibitor
Asked by
7 Member States
Your trial · DVZ-301
Adenocarcinoma
548 patients
Comparator
Docetaxel alone
Head-to-head
None for this PICO
No head-to-head dataPlan due in 7 daysday 44, October 14, on your team's clock.
A trial published in 2014 tested docetaxel with and without that inhibitor, so docetaxel can link the two trials. The first estimate for overall survival is a hazard ratio of 0.92, with a 95% interval of 0.70 to 1.20. Below 1 would favor Dravenzo, and an interval that crosses 1 shows no clear difference. Every patient in your trial had received immunotherapy. No patient in the older trial had. Your biostatistician wants a sensitivity analysis, a second run that tests how far the answer moves, and you have a week to agree on a plan.
Good JCA work finds this gap on day 37, while there is still time to plan the comparison, run it and write it up, and it tells the assessors plainly what stays uncertain.
The EU joint clinical assessment: the law, the PICOs, the 100-day dossier and how assessments fail
The regulation, how 27 countries write one set of questions, the dossier template, indirect comparisons, what the published reports show, and the 2026 rules on AI, explained without selling anything. If you are new to EU health technology assessment, start here.
01 · In plain words
A joint clinical assessment in plain words: one set of questions from 27 countries, one dossier, one report
When a company wins approval to sell a new medicine in Europe, it still has to persuade each country to pay for it. Each country has a health technology assessment body, or HTA body, that asks a different question from the regulator's. The regulator wants to know whether the medicine works and whether its benefits outweigh its risks. The HTA body wants to know how it compares with what doctors in that country use today. Until 2025 the company answered that second question up to 27 times, in as many formats, on as many calendars.
Since January 12, 2025, for new cancer medicines and advanced therapies, the clinical part of that question is asked once.1,2 The countries say what they need to know as research questions. Each question names a population of patients, the intervention being assessed, a comparator and the outcomes to measure. The four first letters give the work its everyday word: a PICO. Two national agencies, an assessor and a co-assessor, merge the countries' requests into one list, the assessment scope.3
The company then has 100 days to submit a dossier that deals with every PICO on the list: with a head-to-head trial if it has one, an indirect comparison if it can build one, or a written reason if it can do neither.4,5 Its trials were designed years earlier, for regulators, against one comparator, and the countries may now ask about five comparators, in six subgroups. So much of the 100 days goes to one job: finding out which questions the evidence can answer, and deciding what to do about the rest.
At the end the assessors write a report that describes the relative effects and how certain they are. The law forbids the report from saying whether the medicine is worth having.1 Each country reads the report, adds its own view of price and value, and decides for itself. If a second request still leaves the dossier short, the assessment is discontinued, and the Commission publishes what was missing.6
From early information to a national decision, with Dravenzo's 11 PICOs
The company tells the assessors what is comingWhen it files with the European Medicines Agency, it also sends the information used to develop the scope.
The countries write the questionsA survey goes to every Member State. The assessors consolidate the answers. For Dravenzo, that makes 11 PICOs, delivered on August 31.
The company builds the dossier in 100 days88 evidence boxes, two rows with no head-to-head data, comparison plans, results, every number traceable. This is where the row nobody looked at gets found, or stays hidden.
The Commission checks, the assessors writeA completeness check, perhaps a second request, clarifications, then a draft report and a factual accuracy check of seven days.
The report is published and each country decidesMember States give it due consideration in their own HTA. Price, reimbursement and overall value stay national.
Step 3 is the subject of this guide. Dravenzo, its trials and its people are a fictional example.
Why the dossier is written against the countries' questions, and three more JCA surprises
You answer questions you did not choose
The scope reflects what Member States need. The Member State Coordination Group, which runs the process, has a short reply for a company that disagrees with a PICO: the developer, the law's word for the company, "is required to address all PICOs appropriately in the dossier".5
An explained gap can pass, and missing reasons can end an assessment
In the first published report, the developer gave evidence on two of eight PICOs and the assessment still finished.7 What ended other assessments was missing method, documentation and reasons.8,9,10
All the evidence belongs in the first version
The Coordination Group expects "all available evidence" in version V0.1, the first dossier submitted. Later versions answer requests, so the case has to be complete from the start.5
Most deadlines move with the regulatory review
The scope, the dossier and the report's endorsement are each set as a number of days from a milestone in the regulatory review. When the review moves, the HTA calendar moves with it.1,4
Four JCA terms to know first: PICO, assessment scope, indirect comparison, evidence gap
PICO
One research question: a Population, the Intervention, one Comparator (which may bundle several treatments) and at least one Outcome.3
Assessment scope
The consolidated list of PICOs the Member States asked for. It is the data request the dossier must meet.
Indirect comparison
An estimate of how two treatments compare when no trial tested them against each other, usually through a treatment both trials share.
Evidence gap
A PICO with no study that allows a valid comparison. The dossier describes the search and says why, and the report records the gap.5
What a joint clinical assessment decides, and what stays with each Member State
In shortA JCA produces one scientific description of relative effects and how certain they are. It makes no decision about marketing, price or reimbursement. Three documents carry it: the scope, the dossier and the report. Each country must give the report due consideration, may not ask again for what was submitted at Union level, and keeps every decision for itself.
The word that does the most work in the HTA Regulation, Regulation (EU) 2021/2282, is "joint", and it applies only to the clinical evidence. The report must be limited to describing the scientific analysis of relative effects on the outcomes in the scope, and the degree of certainty of those effects. It "shall not contain any value judgement or conclusions on the overall clinical added value" of the medicine (Article 9(1)).1
The Commission's factsheet puts it the same way: clinical domains only, with reimbursement left to the Member States, who may add cost-effectiveness and other information of their own.2
Three documents carry a joint clinical assessment, and the countries write the first
1Assessment scope
The PICOs. A survey of Member States, consolidated into the fewest questions that cover them.
Assessor and co-assessor; JCA subgroupBefore the dossier
2Dossier
Complete evidence for every PICO in the Annex I template, with its underlying documents.
The health technology developer100 days, or 60
3Completeness check
Does the dossier meet Article 9(2), (3) and (4)? If not, a second request.
European Commission15 working days, or 10
4Report and summary
Relative effects and certainty, by population and comparator. No value judgment.
Assessor and co-assessor; Coordination Group endorsesBy 30 days after the marketing authorization
May a country ask again for evidence already submitted?
No, in general
Article 13(1)(d). The protection falls away if the assessment is discontinued1
What "due consideration" asks of a country, and why it matters to a developer
Member States must give due consideration to the published reports and annex them to their national HTA documents.1,11 They keep the right to draw their own conclusions and to use the parts relevant to their system.
National integration is still new. Two 2025 surveys of company affiliates in 25 European countries, published in 2026 by authors employed by a developer, found mixed expectations on reimbursement timelines, with the risk of delay seen as possibly transitional.12 The October 22, 2026 Coordination Group meeting discussed a draft question-and-answer document and a tool for national reporting on how the reports are used.13
For a dossier team, the practical reading is that a PICO exists because a named system needs the answer. A thin answer on Germany's comparator lands in the German appraisal, and the Union-level dossier is the main place to get it right.
A JCA report gives no grade and no verdict. Reports describe what limits certainty, outcome by outcome. The Coordination Group's guideline says its use of "uncertainty" should not be read as the "certainty of evidence" of GRADE, a widely used system for rating evidence.14 Teams often keep their own rating to help plan the work. A rating of that kind is an internal aid, and the assessors will reach their own description.
03 · Who is in scope, and how many
Which medicines need a joint clinical assessment under the EU HTA Regulation: cancer and advanced therapies now, orphan medicines in 2028, all new medicines in 2030
In shortScope turns on what the medicine is and when its application reached the European Medicines Agency (EMA). As of the Commission's latest lists, 19 assessments are ongoing, four are complete and four have been discontinued. Three of those four ended because the dossier did not meet the rules.
Scope widens in stages, and the first medical devices were selected in 2026
Jan 12, 2025Cancer and advanced therapies
New active substances for cancer, and advanced therapy medicinal products.
Jun 9, 2026First report published
Tovorafenib, a medicine for a childhood brain tumor.
Oct 6, 2026First medical devices announced
Two heart valve systems. Further selections expected quarterly.
Jan 13, 2028Orphan medicines
Medicines with an orphan designation under Regulation (EC) No 141/2000.
Jan 13, 2030All other new medicines in scope
Every other centrally authorized medicine covered by Article 7(1).
Three details trip teams up. The stages add up, so an orphan medicine, one aimed at a rare condition, is already in scope if it treats cancer. The first report's medicine is exactly that: tovorafenib has an orphan designation and its indication is a cancer, and it went first.6 The date that counts is the application to the Agency, so a launch date tells you little. And a new indication for a medicine that already has a report comes in through its own route, on the shorter 60-day clock.1,4
One in seven assessments on the Commission's lists has been discontinued
27 assessments listed, by status. The three lists were extracted on October 15, September 17 and September 7, 2026.
The Commission's tracker counts only assessments with an appointed assessor and co-assessor, so it runs below the work program's estimates. The 2026 program estimated around 35 new cancer assessments, around 15 for advanced therapies and three for variations, and about five for devices. The final number depends on valid applications filed during 2026.17 The Commission's reports page listed four reports on November 3, 2026.18
Who runs it: the Coordination Group, its subgroup, the assessor and the co-assessor
The Member State Coordination Group on HTA, or HTACG, works through a subgroup on joint clinical assessments. For each assessment the subgroup appoints an assessor and a co-assessor from different Member States.1,2 Norway's medicines agency appears on the Commission's tracker in both roles.6
Patients and clinical experts give input on the scope and on the draft report. The Commission's role is procedural: it sends the requests, checks the dossier for completeness, and reviews the report before publication. Its cover note on each report says the assessment is the Coordination Group's responsibility.7
04 · The JCA timeline
The JCA timeline: 100 days from the first request, 60 in two cases, and the deadlines stacked behind it
In shortThe dossier is due 100 days from notification of the Commission's first request, or 60 days under accelerated assessment or for a new indication. An extension can never run past 45 days before the expected opinion of the CHMP, the European Medicines Agency's human medicines committee. All four published dossiers used the whole window, and none was confirmed on its first version.
Step
Deadline
Whose
Where written
Scope finalized
At the latest 10 days after CHMP adopts its list of questions; within 75 days of EMA validating the application in the two shorter cases
"Implementing Regulation" means (EU) 2024/1381, the Commission act that sets the procedure and the dossier template.4 You may see "90 days" in older briefs. That was the draft act; the adopted text says 100.4,19 The two 60-day cases are the accelerated procedure and a variation that adds a new therapeutic indication. Those two cases also get the shorter 10 and 5 days in the table, and a new scope written after a changed indication gets the 5 days as well.4
Measured from the published reports: the window is used in full
All four published dossiers took about the whole 100 days
Days from the scope being shared with the developer to dossier version 0.1
Dates from the procedural table in each report, counted by us. "Scope shared" is the report's own label; the legal clock runs from notification of the first request.
The developer's 100 days are about a quarter of a process that runs about a year
Days between procedural steps in the four published assessments
Early information to scopeScope to version 0.1Version 0.1 to confirmationConfirmation to endorsement
Why the example's team plans its comparison decisions for day 44
No regulation mentions day 44. Teams count backward. A comparison needs a plan, a protocol and a statistical analysis plan, because Annex I of the Implementing Regulation, the dossier template, wants those in Appendix D.5 alongside the result.4,8 Then it needs a run, a second check, a written section and a review.
A final sign-off at about day 96, a full draft at about day 75 and results by about day 60 put the plan's approval in the first half of the window. Those dates are the sample team's plan in this guide's example. Yours may differ, and the arithmetic will look the same.
One more rule shapes the calendar. Searches informing relative effectiveness may be at most three months old when the dossier is submitted, so a systematic review (a documented search for every relevant study) done early must be refreshed inside the window.5 For a dossier due December 9, the cut-off cannot be earlier than September 9.
An extension buys less than it seems. The secretariat may extend the deadline in justified cases with the assessors' consent, but never beyond the 45-day limit in Article 10(1).1,4 If the first request came late, there is nothing to extend into. The Coordination Group was still discussing "the application of JCA timelines to extensions and exceptional cases" at its October 22, 2026 meeting.13
05 · How PICOs are set
How 27 countries become one assessment scope: the PICO survey and the consolidation rules
In shortThe assessors send out a minimal proposal, every country answers a survey, and the answers are merged into the fewest PICOs that still cover everyone. Published scopes hold 1, 7, 8 and 12 PICOs. How a country words its comparator request decides the count: one treatment, several treatments, or a bundle chosen patient by patient.
The Coordination Group's guidance, adopted on November 28, 2024, sets out the method.3 The assessor and co-assessor draft a proposal with a minimum of parameters and leave out anything particular to their own countries. Each Member State then states what its system needs, and is asked to limit requests to what is necessary for national decisions.3
One PICO "comprises one population, one intervention (or combination), one comparator (which can include more than one treatment), and at least one outcome". The aim is "the lowest possible number of PICOs". The result is prospective: it specifies the data requirements before anyone has checked whether the data exists, so it can contain PICOs for which no evidence is available, and the report then identifies the gap.3
Three shapes of comparator request, and what each does to the count
One treatment suits everyone
A unique comparator
One treatment, one PICO. Lurbinectedin's scope was this simple: one population, one comparator, one randomized trial to answer it.20
Several treatments suit everyone
AND, or OR
A country needing estimates against each treatment creates one PICO per treatment. A country content with any one of them joins them by OR. A single survey answer can add a question nobody else needs.3
The doctor picks by patient
An individualized treatment bundle
Several options, chosen by each patient's characteristics. One PICO, against the bundle. A trial meets it with a physician's-choice arm. A trial with one fixed comparator does not, which is why it has been a hard row to answer.3,5
Published scopes range from one PICO to twelve
Consolidated PICOs in the four reports published as of November 3, 2026
The onasemnogene report numbers its PICOs up to 14 after a revised scope, but PICOs 6 and 7 do not appear, leaving 12. A discontinued assessment's list refers to a PICO 13, so at least one real scope ran larger.22,8
What the guidance's own five-country example teaches
The scoping guidance works a hypothetical five-country request through to seven PICOs. Three of the seven exist because one country required estimates against each of two treatments and refused to let a list be reduced. A fourth exists because an individualized bundle cannot be merged with a request for a named treatment.3
In the guidance's example, one country's wording accounts for three of seven PICOs
The seven PICOs in the scoping guidance's hypothetical five-country request, by what created them
One country required each treatment and barred a shorter listAn individualized bundle that cannot mergeThe other requests
One country required each treatment and barred a shorter list
An individualized bundle that cannot merge
The other requests
Seven PICOs
3
1
3
The Coordination Group's documents also separate two words that sound alike. A subpopulation gets its own PICO.3 A subgroup is an analysis inside a PICO, required only when each subgroup has at least 10 people and, for binary and survival data, at least 10 events in one of them.5 If a subgroup in one PICO covers the same patients as another PICO's population, the first needs an interaction test and both subgroups' results, and the second needs that population's characteristics and results.5
PICO simulation: what a rehearsal can tell you, and how far off it can be
Because the real scope arrives with the clock already running, many teams rehearse it. A simulation reads the guidelines and past national decisions, predicts each country's request, consolidates them by the guidance's own steps and sets your evidence against each row. A rehearsal binds nobody. Published exercises show how widely the count can swing.
Simulations usually land near eight or nine PICOs, and the worst cases run far above any real scope
PICO counts in published simulations, against the largest published real scope
Worst case across 24 simulationsabstract, evidence vendor88
One simulation, first-line lung cancer immunotherapy, 25 countriesauthors at a developer67
Largest published real scopeonasemnogene abeparvovec12
One simulation, first-line lung cancer immunotherapy, 25 countries
67
Largest published real scope
12
Median of 24 simulations
9
Mean of 35 published exercises
8
The authors themselves warn that few examples exist and that country participation is limited, which "restricts the ability to draw clear conclusions".23 An abstract from an evidence vendor reports that 67% of PICOs in its simulations needed indirect comparisons, with a minimum of 175 to 720 analyses anticipated.24
06 · The dossier template
The JCA dossier template, section by section: where the PICOs, the searches and the results go
In shortAnnex I of Implementing Regulation (EU) 2024/1381 sets six sections, four appendices and a version ladder from V0.1 to a public V1.0. Results go in section 5, organized by population and PICO. Section 3 must name every PICO left without results and say why. Six kinds of source must be searched, and the search can be no more than three months old at submission.
Section
What it holds
Where teams get caught
1 Overview
The product, the developer, an executive summary giving results for each PICO
Statements without documents: one list demands support for a line in the executive summary8
2 Background
The condition, its management, the medicine, its regulatory status
Wording that drifts from the proposed product information
3 Assessment scope
The scope "in the format shared" in the first request, and every PICO without results, with reasons
Selection criteria by PICO, information retrieval (4.2.1), data analysis and synthesis (4.3)
Methods not described per PICO; sensitivity analyses absent without a reason8
5 Results
Retrieval results and exclusions (5.1), study characteristics (5.2), relative effectiveness and safety (5.3)
Proportions in place of relative effects; study data missing for indirect comparisons8,10
6 References
The reference list, with full texts in Appendix D
Sources cited but not supplied
Appendices A to D
Study listings and methods (A), material for judging certainty, including risk of bias (B), main studies that answered no PICO (C), underlying documents including search strategies (D.2) and evidence synthesis reports (D.5)
Search strategies, standard-format result files, and protocols for evidence syntheses8,10
Section numbers from Annex I, and the Commission's lists of missing information cite the same ones.4,8,9,10 The Coordination Group's guidance on filling in the template comes with a table template collection for presenting most of this.25
Six sources a JCA evidence search, or information retrieval, must cover
Sponsored and third-party studies
Published and unpublished, the company's own and others' where available.9
Bibliographic databases
At least Medline and the Cochrane Central Register of Controlled Trials.9
Study and results registries
One list names a missing ClinicalTrials.gov search, with its strategies and result files.10
HTA reports
From EEA states applying the regulation, and from Australia, Canada, the UK and the US.9
The EMA submission file
The clinical safety and efficacy data submitted to the Agency.9
Patient registries
Searched and documented, including searches on registry websites.8,9
In the example, 2,184 literature records narrow to seven randomized trials
The sample systematic review behind Dravenzo's dossier, searches to September 28, 2026
Records found2,184
After removing duplicates1,612
Full texts assessed41
Randomized trials included7
Fictional example from this guide
Show the numbers
Stage
Count
Records found
2,184
After removing duplicates
1,612
Full texts assessed
41
Randomized trials included
7
A review like this is slow to build from nothing. Registered systematic reviews took a mean of 67.3 weeks from registration to publication in one analysis, published in 2017, of data to 2014.26 A team that has not started a review before the scope arrives is unlikely to finish one inside 100 days by normal means.
The template's version ladder, and what each number means
Version
What it is
Clock
V0.1
The first dossier. All available evidence belongs here.5
100 days, or 60
V0.2
Updated after the Commission's second request
15 days, 10, or 7 for minor items
V0.3
Updated after the assessors ask for clarification
7 to 30 days
V0.4
Updated after a change to the indication and a revised scope
7 to 30 days, set by the assessors
V0.5 and V0.6
Re-initiation after discontinuation; confidentiality claims with their justification
V0.5: 7 to 30 days. V0.6: 7 days, or 5
V1.0
The public version, without confidential information. The dossier is published.11
After the report is final
The ladder comes from the revision history in Annex I, which continues to V1.0.5 and V2.0 for updates after the report.4 The Commission publishes "only the most recent version of the dossier, on which the JCA report is based", normally with its appendices, so write every section as if a competitor will read it.11
07 · Indirect comparisons
Anchored and unanchored indirect comparisons: worked through with numbers from one PICO
In shortWhere no trial compared two treatments, the Coordination Group wants an anchored comparison: two trials that share a treatment, so randomization survives. Unanchored methods are "highly problematic". The arithmetic of the plainest method, Bucher, fits on a page. The feasibility work around it decides whether anyone believes the result.
The first job is an evidence grid, with one row per PICO and one column per outcome. Each cell holds the best evidence available: a head-to-head result, an analysis in a subgroup, a comparison through a shared treatment, or nothing. Dravenzo's scope has 11 PICOs and 8 outcomes, which makes 88 cells. Only 23 of them have direct evidence from the main trial.
Direct trial evidence fills about a quarter of Dravenzo's 88 evidence cells
Each PICO and outcome pair, by the best evidence available on day 37
88cells
High uncertainty31
Direct evidence23
Evidence gap16
Indirect comparison done12
Comparison planned6
Fictional example from this guide
Show the numbers
Part
Value
High uncertainty
31
Direct evidence
23
Evidence gap
16
Indirect comparison done
12
Comparison planned
6
The order the Coordination Group expects: direct first, anchored next, gaps last
Each step down the ladder of evidence adds an assumption someone must defend
1Direct comparison
A randomized trial of your medicine against this comparator. Pool with a pairwise meta-analysis if there are several.
Best caseRandomized
2Anchored, simple
Two trials share one treatment: the Bucher method, a star of three treatments.
BiostatisticianRandomization kept
3Anchored, wider
More than three treatments or closed loops: network meta-analysis. Add only the trials needed to connect the network.
BiostatisticianRandomization kept
4Population-adjusted
Network connected, populations too different: matching-adjusted, simulated or multilevel methods, with individual patient data from at least one trial.
BiostatisticianStronger assumptions
5Unanchored
No shared treatment. The guideline calls unanchored matching-adjusted and simulated comparisons highly problematic.
BiostatisticianNon-randomized
6Evidence gap
Nothing valid exists after a proper search. Describe the search and justify.
A randomized trial of your medicine against this comparator. Pool with a pairwise meta-analysis if there are several.
Best case
Randomized
Anchored, simple
Two trials share one treatment: the Bucher method, a star of three treatments.
Biostatistician
Randomization kept
Anchored, wider
More than three treatments or closed loops: network meta-analysis. Add only the trials needed to connect the network.
Biostatistician
Randomization kept
Population-adjusted
Network connected, populations too different: matching-adjusted, simulated or multilevel methods, with individual patient data from at least one trial.
Biostatistician
Stronger assumptions
Unanchored
No shared treatment. The guideline calls unanchored matching-adjusted and simulated comparisons highly problematic.
Biostatistician
Non-randomized
Evidence gap
Nothing valid exists after a proper search. Describe the search and justify.
Dossier lead
Justification judged
The Coordination Group's May 2026 answers say it directly: "Anchored indirect comparisons preserve randomisation and should always be used in preference to unanchored methods". Where no connected network exists, unanchored comparisons should use methods for non-randomized evidence, and if no data allow any valid comparison, the developer "should describe the situation and justify the choice". Whether that makes the dossier incomplete "will depend on the appropriateness of the justification".5
The Bucher method by hand, on PICO 2
In 1997, Bucher and colleagues published a way to compare treatments A and B when each had been tested only against C.27 Work on the log scale, subtract, and add the variances. A hazard ratio compares how fast an event such as death happens in two arms of a trial: below 1 favors the first treatment, and the 95% interval shows how precise that estimate is.
In the example, Dravenzo (A) was tested against docetaxel (C) in DVZ-301, and a published 2014 trial tested docetaxel plus the angiokinase inhibitor (B) against docetaxel (C). Each hazard ratio below comes from the adenocarcinoma subgroup of its trial.
Step
Dravenzo vs docetaxel (DVZ-301)
Combination vs docetaxel (2014 trial)
Result
Hazard ratio (95% interval)
0.76 (0.62 to 0.93)
0.83 (0.70 to 0.99)
Inputs
Standard error of the log hazard ratio: ln(upper ÷ lower) ÷ 3.92
0.103
0.088
Combined: √(0.103² + 0.088²) = 0.136
Indirect hazard ratio, A vs B
0.76 ÷ 0.83
0.92
95% interval: exp(ln 0.92 ± 1.96 × 0.136)
The interval crosses 1
0.70 to 1.20
Dravenzo's figures are fictional; the 2014 figures match a real published trial, left unnamed. Recalculated by us from the two inputs on the screens. Notice how the indirect interval is wider than either input: the uncertainty of both trials adds up.
Two points about that result. An interval crossing 1 means the data show no conclusive difference. It does not show that the two treatments are equivalent, a point Altman and Bland made in the title of a 1995 paper: absence of evidence is not evidence of absence.28 A JCA dossier should report the numbers and their limits, and the writers should resist the adjective "comparable".
The arithmetic also cannot check itself. With only one path between A and B, "the consistency assumptions cannot be assessed", and the guideline says "a thorough assessment of similarity is even more important".14
Feasibility checks that decide whether an indirect comparison will stand
Feasibility asks whether two trials can fairly be linked. The guideline's test is exchangeability: if patients in one trial were moved to the other, the treatment effect should stay the same. The main threat is an effect modifier, a patient characteristic that changes how two treatments compare.14
Check
PICO 2 in the example
What it protects
Shared comparator
Docetaxel at 75 mg/m² every 3 weeks in both
A real anchor
Population
Adenocarcinoma subgroups, second line after platinum, ECOG performance status 0 to 1 (fully active, or limited only in strenuous activity) in both
Exchangeability
Effect modifiers
Prior immunotherapy: 100% vs 0%. This one fails.
The result applies to today's patients
Proportional hazards
Schoenfeld test p = 0.21, curves do not cross
A single hazard ratio means something; the guideline calls the assumption "often implausible"14
Follow-up
Median 18.2 vs 31.7 months: a warning
Comparable maturity of the data
Checks like these can decide an assessment. In September 2026 the Commission wrote that a dossier in which "for all PICOs, the assessment of relative effectiveness and relative safety relies exclusively on indirect comparisons" did not report enough to assess exchangeability, so an assessment "is not possible".10
Germany's record points the same way. An abstract covering German AMNOG assessments through April 2024 found the country's Federal Joint Committee (G-BA) accepted indirect comparisons in 22.5% of cases, across 222 assessments, with method weaknesses and insufficient similarity leading the rejections.29 That is national, older and from a consultancy, so read it as a warning rather than a forecast for JCA reports.
Matched comparisons shrink the sample, and the effective number tells you how far
A matching-adjusted indirect comparison, or MAIC, reweights patients in one trial so their measured characteristics match the published averages of another. The price shows in the effective sample size, the number of patients the weighted data are worth. For PICO 6 in the example, weighting a 61-patient single-arm cohort to three published platinum-doublet studies left an effective sample of 23.4.
After weighting, 38 in every 100 patients still count in the example's matched comparison
23.4 of 61
effective sample after weighting on age, performance status, brain metastases, prior targeted therapy and EGFR or ALK status. The comparison has no anchor.
Fictional example from this guide
Show the numbers
Measure
Value
effective sample after weighting on age, performance status, brain metastases, prior targeted therapy and EGFR or ALK status. The comparison has no anchor.
38 of 100
The guideline adds two limits people skip: the method "is only valid where there is sufficient overlap" between the populations, and the result applies to the other trial's population, "which may not match the target population for the decision". Without a common comparator the assumption is, it says, in almost all practical applications "not justifiable".14
The first published report shows where that can end. An unanchored comparison was submitted for tovorafenib's PICO 7. It was left out of the assessment, because too little was reported about the comparator study and age was the only characteristic available to adjust for.7
Each move away from a shared comparator raises the assumptions and the data you need
Our placement of the methods, from the Coordination Group's guideline
What the first four JCA reports show: 28 PICOs asked, 16 assessed, and gaps stated in plain sight
In shortFour reports were published between June 9 and September 7, 2026. They run to 154, 211, 257 and 362 pages. Their developers submitted evidence on 17 of 28 PICOs, and 16 entered the assessment. Individualized-treatment comparators account for most of the unanswered rows. All four dossiers needed a further version before confirmation.
Across the four reports, 28 PICOs were asked and 16 were assessed
Counted from the report tables. Tarlatamab's PICOs were all informed, with limits on some outcomes.21 An ISPOR Europe 2026 abstract counts tovorafenib the same way: its single-arm main trial matched one PICO of eight.30
Four individualized-treatment PICOs in tovorafenib, and what the developer wrote
Four of tovorafenib's eight PICOs named an individualized treatment comparator. The developer reported that no study in the target population included all the treatments in the bundle, and concluded that a comparison was impossible.7 Five of onasemnogene's PICOs went unanswered too, three of them against individualized treatment. The report records it in one sentence: the developer "stated that PICOs 4, 5, 12, 13 and 14 could not be addressed due to unavailability of sufficient data".22
The Coordination Group published further detail in Appendix A of its May 2026 answers.5
The May 2026 answers ask for one estimate against a whole bundle, and judge a partial study case by case
1Is one treatment right for everyone?
If no single treatment suits all patients in the population, the comparator becomes a bundle of options.
ScopeThe starting question
2One PICO against the bundle
One comparison, with one effect estimate per outcome and no separate assessment of each option.
ScopeOne estimate per outcome
3A physician's-choice arm
The investigator picks from the list of options for each patient, and the reasons are documented. This study design meets the requirement.
Developer's studyCan deliver it
4A study with only some options
"A matter of assessment". It may be acceptable if it includes the most relevant options, and the assessment may then refer only to those.
If no single treatment suits all patients in the population, the comparator becomes a bundle of options.
Scope
The starting question
One PICO against the bundle
One comparison, with one effect estimate per outcome and no separate assessment of each option.
Scope
One estimate per outcome
A physician's-choice arm
The investigator picks from the list of options for each patient, and the reasons are documented. This study design meets the requirement.
Developer's study
Can deliver it
A study with only some options
"A matter of assessment". It may be acceptable if it includes the most relevant options, and the assessment may then refer only to those.
Assessors
Case by case
Tovorafenib's assessors called the developer's choice to take only studies with every option "not adequate".7 Dravenzo's PICO 7, physician's choice among three options, sits in the same family: its trial offered one, and the example's sample scope says that covers the PICO in part. Plan early for which options your evidence can reach, and say plainly what stays unanswered.
What happened after submission: confirmation, requests and a revised scope
Report
V0.1 to confirmation
Before confirmation
After confirmation
Tovorafenib
56 days
V0.2 on October 17, 2025
V0.3 on December 5, 2025, after two assessor requests
Lurbinectedin
69 days
Two versions marked as responses to the second request
A response to a request for additional information
Tarlatamab
57 days
V0.2 on February 18, 2026
V0.3 on March 19, 2026
Onasemnogene abeparvovec
69 days
V0.2 on February 3, 2026
Revised scope on March 9, 2026 after the indication changed; V0.3 and V0.4
From the procedural tables.7,20,21,22 Publication followed endorsement by 40, 16, 16 and 53 days, and the Coordination Group has discussed how to speed confidentiality requests so reports appear sooner.13
Seven days for facts: what the accuracy check allows, and what it cost two developers
The developer sees the revised draft reports once. It has 7 days, or 5, to flag "purely technical or factual inaccuracies" and confidential information, and it may not comment on the results.1,4 The form published with the first report asks that every comment be "validated by published sources (full reference) and/or by indication of the documents/page numbers from the submission dossier".31 A claim you cannot point to a page for will not count.
Errors found at this stage come too late to fix in the report. In lurbinectedin, the developer confirmed at the accuracy check that subgroup analyses of two quality-of-life scales in its dossier were wrong. The report says incorrect information identified then "was not acted on".20 Tarlatamab's report carries a similar note, adding that the nature of that error had no effect on the estimates.21 Both reports publish it. The October 2026 Coordination Group flash report says the Group will keep monitoring what companies submit at this step.13
09 · Why dossiers fail
Why JCA dossiers fail: three discontinued assessments and the Commission's lists of what was missing
In shortArticle 10(6) requires discontinuation if, after the second request, a dossier is late or fails Article 9(2) to (4). Three assessments have ended this way. The Commission's lists show the same weaknesses each time: incomplete searches, thin documentation, unjustified methods, and comparisons nobody could judge. Few of those items concern what the medicine does.
Three dossier failures in 2026, each with a published list of what was missing
Feb 13, 2026Sasanlimab discontinued
The developer withdrew its marketing application. The list for Tacquell's second request is dated the same day.
May 8, 2026Catequentinib list
Items missing from the updated dossier, V0.2 of April 9, 2026.
Jun 22, 2026Catequentinib and Tacquell discontinued
Dossier fails Article 9(2), (3) and (4).
Sep 15, 2026Sintilimab list
Items missing from the dossier updated on September 1, 2026.
The same gaps in three lists: searches, methods, documentation and results
Weak searches, weak methods and missing results appear in all three failed dossiers
Our classification of the Commission's lists. 0 = no such item listed, 1 = named in part, 2 = named as missing.
No itemIn partNamed
TacquellCatequentinibSintilimabInformation retrieval incompleteNamedNamedNamedSearch strategies and filesNamedIn partNamedExcluded studies and reasonsNamedNamedNo itemMethods described and justifiedNamedNamedNamedIndirect or external comparison could not be judgedNamedNamedNamedRelative-effect results by PICONamedNamedNamedOmitted PICOs presented and explainedNo itemNamedNo itemStructure and consistencyNamedNamedNamed
Read as a set, these lists mostly describe bookkeeping: searches, documents and reasons. The sintilimab list even says the missing study information, "in principle, is mostly contained" in the clinical study report and in publications the developer already had.10 That is the case for building traceability from the first day of the window: 15 days after a second request is too short to reconstruct how a search was run.
What discontinuation costs: national re-requests and a published record
Article 10(6) says the Coordination Group "shall discontinue" where the Commission finds a dossier late or short after the second request, and a statement of the reasons goes on the platform. The protection in Article 13(1)(d) that stops countries asking again then no longer applies, so the developer may be answering the same questions country by country.1 A discontinued assessment can be re-initiated, at the latest six months after the original deadline, once the requirements are confirmed as met.1 The medicine's marketing review carries on regardless: the two tracks are separate.
For Tacquell, catequentinib and sintilimab, the tracker gives one reason: the dossier fails the requirements of Article 9(2), (3) and (4). We rely on that wording.6
10 · AI in the JCA dossier
AI in a JCA dossier: the Coordination Group's July 2026 principles and what they require you to keep
In shortOn July 15, 2026 the Coordination Group adopted general principles on AI in JCA dossiers. The developer stays accountable, a human stays ultimately responsible for every step, each AI-assisted step is named with the tool's name, version, date, maker and purpose, and all prompts are recorded and available on request. Published evidence on accuracy is thin and small.
Principle
What the text says
What it means for you
Accountability
Developers "are accountable for the content, methods and findings", including "the validity of any output resulting from AI use"
You cannot blame a tool. Copyright and data protection are yours as well, and the dossier will be published
Human oversight
"None of the steps in the dossier preparation or analysis process should be fully automated without a human being ultimately responsible for the quality and accuracy"
A person stays responsible for each step, with oversight throughout it
Methods
AI "does not change the applicability" of the methodological guidance. "It should also be possible to verify the compliance with the methods"
If nobody can retrace a process, compliance cannot be shown
Transparency
Specify all AI-assisted steps, such as "information retrieval (incl. screening of studies), data extraction, risk of bias assessment, analysis, reporting". Spelling and grammar help needs no disclosure
AI-informed outputs are marked in the dossier
Reporting
At minimum "the name, version and date of the AI tool used, the developer of the tool and the specific purpose(s)"
Keep this per step from day one
Prompts
"All prompts used in AI tools are recorded and made available if requested during the joint clinical assessment process"
Clarification clocks can be as short as 7 days. You need the export ready
All quotes from the principles, adopted July 15, 2026.32 "AI" there means all machine-learning models, generative or not.
What published studies say about AI on PICO extraction
AI matched manual PICO extraction most closely on the intervention, least on the outcomes
Concordance with manual extraction across 8 HTA reports, by PICO field. One second-line lung cancer case study, ISPOR Europe 2026 abstract.
That abstract is one of the few with numbers. It reports 28 of 32 data points concordant on HTA reports (88%) and 7 of 16 guideline sources (44%), though much of the guideline difference reflects prompts that did not limit the line of therapy.33
It reports two weeks by hand against two hours by AI. It also records the errors that matter: truncated comparator lists, dropped population qualifiers, and one clinical endpoint replaced with economic outcomes. The authors, including a consultancy and an AI vendor, recommend validation by human experts. It covers one asset and 96 comparisons, so treat its figures as a first look.33
The other AI texts around a dossier, and how far each reaches
A sensible standard for the records behind a dossier, by analogy
Where the limits of automation sit in a JCA
A tool can read a scope, fill a grid, draft a section and check a number. It cannot create a trial that was never run, turn an unanchored comparison into an anchored one, or decide which uncertainty a company will defend before 27 health systems. Those decisions need named people. The strongest evidence that people must stay close is in the lists from section 9, where the failures are about reasoning and record keeping, and both can get worse if a fast tool leaves no trail.
11 · Volumes and stakes
How much JCA work is coming in 2026 and beyond, and what access delay means in days
In shortThe 2026 work program estimates about 50 new assessments of medicines, and orphan medicines join in 2028, all new medicines in 2030. Patients in Europe already wait a mean of 607 days from central authorization to availability. Nobody has yet shown whether JCA shortens or lengthens that.12 No public figure exists for what a JCA dossier costs.
The 2026 work program plans about 50 new medicine assessments and adds devices
HTACG's estimates of assessments it may start in 2026. Final numbers depend on applications.
mean time from central authorization to availability across 36 European countries, in the 2025 W.A.I.T. survey, up by over three months from the 2019 report.40
EFPIA survey, reported by the Office of Health Economics, July 2, 2026
44%
average rate of availability, down from 54% in 2019.40
Same survey, unweighted average
179–225 days
from the developer's first information to dossier version 0.1 in the four published assessments.7,20,21,22
Procedural tables, our count
What a day of delay is worth, and why that figure needs care
Teams often quote Tufts CSDD on delay. From one dataset of 645 drugs, the journal article gives about $500,000 in lost sales per day and the August 2024 white paper gives about $800,000. Both give about $40,000 a day in direct Phase II and III trial costs.41 These describe delay in clinical development, across drugs already on the market. A late national reimbursement is a different event, and one cancer medicine's revenue in one country follows its own curve. Use them to set the scale, then replace them with your own launch forecast.
No public figure for the cost of a JCA dossier, and what to measure instead
We found no regulator, journal or survey that publishes the cost, hours or staffing of a JCA dossier. A search on November 3, 2026 turned up statements that preparation needs investment, with no numbers. Measure your own: the hours your team logged on your last national or mock dossier, your consultancy invoices, the number of analyses per PICO and the weeks your statistics group can give. A readiness survey of 13 companies, published in 2025, gives only self-rated readiness of 6 to 7 out of 10, with concerns about timelines.42
12 · How teams do it today
How EU HTA teams build a JCA dossier today: spreadsheets, consultancies, and who signs what
In shortA dossier lead, biostatisticians, a review team, medical writers and clinical and regulatory reviewers work to a chain that ends with the head of market access. Tracking often lives in spreadsheets, and many companies buy in help for simulations, comparisons and writing. No published study measures where the days go; the estimate here is the one used in this guide's example.
Who carries a JCA dossier: lead, statistician, writer, reviewers and approver
DL
EU HTA dossier lead
Owns the scope, the plan and each assessment's clock, chooses how each PICO is answered, and signs versions.
Cares about: which row will miss the window, and why.
BS
Biostatistician
Checks which comparisons have an anchor, writes evidence synthesis plans, recommends methods.
Cares about: exchangeability, and whether the output can be reproduced.
MW
Medical writer
Drafts results PICO by PICO in template order, in neutral language.
Cares about: the number and the table it came from.
SR
Systematic review team
Runs and documents searches across six kinds of source, screens studies, records exclusions.
Cares about: a cut-off inside three months of submission.
CL
Clinical lead
Reviews results and limitations, and judges which differences between trials matter clinically.
Cares about: effect modifiers, read with clinical sense.
RA
Regulatory affairs
Aligns the indication and wording with the proposed product information and the CHMP timetable.
Cares about: a dossier that matches the label, on the current forecast.
MA
Head of market access
Approves the final dossier and answers for the launch sequence across countries.
Cares about: where each assessment stands, and what it will mean in Germany and France.
A team's plan counts back from day 100, and the comparison decision comes early
The sample team's milestones in this guide's example. No regulation sets the internal days.
1Day 0 · scope arrives
Read it, name owners, request an explanation meeting.
Dossier leadAug 31
2Day 10 · evidence grid v1
Every PICO and outcome mapped to a table, a publication or a gap.
Lead and statisticiansSep 10
3Day 28 · search cut-off
Systematic review refreshed, inside three months of submission.
Review teamSep 28
4Day 44 · plans signed off
Indirect comparison plans and justifications agreed.
Statistician, then leadOct 14
5Day 60 · results final
Analyses run and reproduced.
StatisticianOct 30
6Day 75 · full draft
All sections written and cited.
Medical writerNov 14
7Day 88 · reviews done
Quality, medical and legal.
Clinical and regulatoryNov 27
8Day 96 and 100 · sign and submit
Final approval; upload to the HTA IT platform.
Head of market access, leadDec 5 and 9
Fictional example from this guide
Show the numbers
Step
What happens
Who
Time
Day 0 · scope arrives
Read it, name owners, request an explanation meeting.
Dossier lead
Aug 31
Day 10 · evidence grid v1
Every PICO and outcome mapped to a table, a publication or a gap.
Lead and statisticians
Sep 10
Day 28 · search cut-off
Systematic review refreshed, inside three months of submission.
Review team
Sep 28
Day 44 · plans signed off
Indirect comparison plans and justifications agreed.
Statistician, then lead
Oct 14
Day 60 · results final
Analyses run and reproduced.
Statistician
Oct 30
Day 75 · full draft
All sections written and cited.
Medical writer
Nov 14
Day 88 · reviews done
Quality, medical and legal.
Clinical and regulatory
Nov 27
Day 96 and 100 · sign and submit
Final approval; upload to the HTA IT platform.
Head of market access, lead
Dec 5 and 9
By hand, finding evidence and drafting take the largest shares of the days to a full draft
An estimate of about 70 days, used in this guide's example. No public figure exists; measure your own.
Scope to questions and ownersFinding evidence for each PICO and outcomeFeasibility and comparison plansDrafting with citationsChecking numbers and wording
By hand, estimate20152010
Illustrative · replace with your own numbers
Show the numbers
Scope to questions and owners
Finding evidence for each PICO and outcome
Feasibility and comparison plans
Drafting with citations
Checking numbers and wording
By hand, estimate
5
20
15
20
10
In house, a consultancy, or software: three ways to staff the 100 days
Way
What it does well
Where it strains
In house, spreadsheets and documents
Close to the trial data and the people who know it; full control
Several assessments at once; a tracker that goes stale; hunting for tables during a 7-day request
Consultancy or evidence vendor
PICO simulation, indirect comparisons and writing capacity; experience across cases
Knowledge leaves with the project; the developer stays accountable for all of it32
Software with AI assistance
Reading scopes, mapping evidence, drafting, number checks, a full record
Few published results; every AI step must be named, overseen and retraceable32,33
Many teams mix all three. Vendors of each kind publish claims, including several of the abstracts cited here, so ask to see the working on your own material.
Part B · See it working
EU Joint Clinical Assessment by Botminds: one 11-PICO scope, from a red row on day 37 to a signed version
The same fictional case, on the solution's real screens. Dravenzo, Northwind Bio, the people and the trial results are sample data, the names are shown as in this guide, and every number is the one on the screen.
13 · Day 37 on the PICO grid
One row per PICO on the 100-day clock: how Dr. Lena Ortiz finds two red rows on day 37
In shortEU Joint Clinical Assessment turns the scope into a grid: one row per PICO with its population, comparator, eight outcomes, the Member States behind it, the evidence verdict, the section that answers it and its next dated step. A re-check against the scope re-reads the sources, and the rows with no head-to-head data turn red.
It is Wednesday, October 7, 2026, day 37 of 100. Dr. Lena Ortiz, EU HTA dossier lead at Northwind Bio, runs four assessments. She opens Dravenzo's: assessment JCA-26-014, assessor the Netherlands, co-assessor Spain, scope received August 31. The countdown reads 63 days to submit on the Commission's HTA IT platform, due Wednesday, December 9.
The PICO grid. Each colored square is one outcome. The last column shares one axis from day 0 to day 100, with today marked and each row's next step, owner and days left. Germany is outlined in the country chips. The grid is the home screen.
Three rows have head-to-head data from the main trial, two have indirect comparisons, three rest on weak evidence, one was not studied. Two are red. Lena clicks "Re-check against the scope". The scope parser reads the 14-page scope again, with its 11 PICOs from 27 Member States. The evidence mapper reads the study report, 1,246 outputs and the analysis plan into 88 cells. The literature review matcher finds 7 included trials, with docetaxel connecting PICOs 2, 3, 4 and 8.
The re-check, part way through. Each agent reports what it read and how long it took. Rows return one at a time, and an evidence verdict appears only when its row is mapped.
The two rows with no head-to-head data. PICO 2 reads "No trial, Bucher plan waits for you". PICO 7 reads "Trial offered docetaxel only". Both comparators were asked for by Germany, and both decisions are due in 7 days, on day 44.
One row opened. Eight outcome values, marked planned where the plan still waits for approval. Below them, the section that answers the row and its next step, and on the right a tile map of who asked. In this example, results sit in sections the sample calls 6.1 to 6.4; in the published template, results are section 5.
The example's plan puts comparison decisions on day 44 and leaves 56 days after them
Dravenzo's milestones on the 100-day clock
TeamComparisonsFinal
Evidence grid, first version
Literature search refreshed
Comparison plans agreed
Comparisons run and reproduced
Full draft
Quality, medical and legal review
Sign and submit
Aug 31Sep 28Oct 14Oct 30Nov 14Dec 9
Fictional example from this guide
Show the numbers
Activity
From
To
Evidence grid, first version
0
10
Literature search refreshed
10
28
Comparison plans agreed
28
44
Comparisons run and reproduced
44
60
Full draft
60
75
Quality, medical and legal review
75
88
Sign and submit
88
100
14 · Evidence and the decision
From a red row to a hazard ratio of 0.92: the evidence grid, the Bucher check and the decision on PICO 2
In shortEvery PICO and outcome cell links to its source table. For each gap an analyst proposes a method and checks feasibility against cited passages. The biostatistician recommends, the dossier lead decides and writes a note for the record, and the statistics engine runs only the plan she approved.
Evidence by outcome. PICO 1, overall survival: HR 0.78 (0.65 to 0.94), median 13.1 vs 10.2 months, with Table 14.2.1.1 of the DVZ-301 study report open on page 612 and the numbers highlighted. The certainty labels are the team's own rating; a JCA report describes certainty in words.
The matrix counts what needs attention: 6 cells need a decision, 16 are gaps, 13 are post hoc. Post hoc means the analysis was not in the trial's plan, and the label stays on the value wherever it appears. The evidence mapper takes numbers only from a cited table or output, and never computes a comparison itself.
A planned cell. Dashed borders mean feasible and waiting for the plan. The overall survival hazard ratio, shown as OS HR 0.92 (0.70 to 1.20), is the agents' feasibility estimate, shown with a warning that it enters the dossier only after approval.
The feasibility check that fails on prior immunotherapy
For PICO 2 the comparison feasibility analyst builds the network: the DVZ-301 adenocarcinoma subgroup (n = 548) and a published 2014 trial's (n = 658) share docetaxel at the same dose. Histology, line of therapy, performance status and proportional hazards all pass. Each check cites its passage. Follow-up carries a warning, and one check fails.
Network and checks for PICO 2. Prior immunotherapy: 100% in DVZ-301, 0% in the comparator trial, a likely effect modifier. These are the checks section 7 worked through by hand.
The decision. Sam Patel, biostatistician, recommends approval with a sensitivity analysis restricted to patients with at least 6 months of prior anti-PD-(L)1 therapy. Lena can approve the plan, approve with the sensitivity analysis, or declare a gap. She picks the second and writes the note. Due October 14; she decides on day 37.
The result. OS HR 0.92 (0.70 to 1.20), sensitivity analysis 0.95 (0.71 to 1.27). The validated statistics engine runs the plan, and a second program reproduces every estimate. The panel says what it means: no statistically significant difference, certainty low.
Back on the grid. PICO 2 has turned blue: "Bucher via docetaxel, OS HR 0.92", next step "Draft section 6.2". PICO 7 stays red, with its own decision still due.
Three other routes through the gap list: option by option, matched, or a justified gap
PICO 7, physician's choice. The trial offered one option of the three. The screen answers by option: docetaxel directly from DVZ-301, the other two through the PICO 2 and PICO 4 routes. The May 2026 answers call for one estimate against the whole bundle, and a study with only some options is a matter for the assessors, so this plan needs their reading.5
PICO 6, an unanchored comparison. Effective sample 23.4 of 61 after weighting, and no shared comparator. The analyst proposes it only because nothing else exists, and it stays rated very low.
PICO 5, a justified gap. The trial's protocol excluded ECOG 2 patients. The May 2026 answers allow a justified gap.5 The team declares one and shows the 14-patient phase 2 cohort descriptively, though the first published report did not discuss single-arm descriptive results of this kind.7
15 · Draft, check, sign
Section 6.2 drafted, 1,846 numbers checked, one SmPC word fixed, version 0.6 signed
In shortThe dossier writer drafts in neutral language with every number cited. The consistency checker compares every number with its table and the indication wording with the draft SmPC (summary of product characteristics, the medicine's proposed label), and it reports without silently editing. A number without a source blocks signing. Each version carries a signature with its meaning, the name and the time.
Ready to draft. With the comparison results in, the writer can start on section 6.2. Priya Ranvelle, the medical writer, reviews next.
A cited draft. Six parts: evidence base, overall survival, progression-free survival and response, safety, quality of life, certainty. 1,180 words, 14 citations. Each citation opens its passage, here the scope's PICO 2 on page 4.
The writer never claims superiority where an interval crosses 1, and it uses no promotional language. That habit suits the law, which limits the report to a description of the analysis.1 Next, Lena runs the consistency check.
The consistency check. 1,846 of 1,846 numbers match their tables and outputs. One finding: section 3 says "advanced" where SmPC draft v3.2 says "locally advanced or metastatic".
One click takes the SmPC wording. A person clicks; the checker reports differences and never edits silently. Section 3 now carries the label's words.
The signature, with its meaning. Document, version, the meaning "Approved for internal review", the checks, and a password. Versions 0.5 and 0.6 are the team's working numbers. In the template's own ladder V0.1 is the first dossier submitted and V1.0 the public one.4
Signed. Version 0.6, October 7, 2026. Dr. Maya Chen, the clinical lead, reviews results next. Omar Haddad, head of market access, signs the final dossier on day 96.
Versions and the day's record. Every approval, draft and check, by person or agent, with its time. This is the record to draw on for the AI reporting the July 2026 principles ask for.32
The 39 days is a design target, and this example runs on its own plan. The solution's target for a full draft is 39 days from scope, against an estimate of about 70 by hand. The team in this example plans its full draft for day 75, with comparison results final on day 60, and on day 37 it has 5 of 13 sections fully cited. A team with many indirect comparisons will be held by how long the statistics take.
The design target cuts the most days from finding evidence
Estimated days from scope to a full draft, by hand against the solution's design target. Neither column is a measured customer result.
By hand, estimateDesign target
Scope to questions, owners and milestones51
Finding evidence for each PICO and outcome208
Feasibility and comparison plans1511
Drafting results with citations2013
Checking numbers, post hoc labels and SmPC wording106
Illustrative · replace with your own numbers
Show the numbers
By hand, estimate
Design target
Scope to questions, owners and milestones
5
1
Finding evidence for each PICO and outcome
20
8
Feasibility and comparison plans
15
11
Drafting results with citations
20
13
Checking numbers, post hoc labels and SmPC wording
10
6
16 · Who proposes, who signs
Six agents read, map, propose, draft and check; a biostatistician and a dossier lead decide
In shortEach agent has one job and limits it may not cross. Agents propose comparisons and never run an unapproved plan. A biostatistician recommends, the dossier lead approves, the clinical lead and regulatory affairs review, and the head of market access signs the final dossier.
Scope parser
Turns the scope into structured PICOs with the Member States behind each and the page and sentence each element came from. Compares a revised scope with the last one.
Never merges or drops a PICO. The assessors decide the scope.
Evidence mapper
Puts the best evidence in every PICO and outcome cell, head-to-head first, with interval, population, data cut-off and table.
Never computes a comparison. Post hoc analyses are always labeled.
Literature review matcher
Links studies included in the systematic review to the scope's comparators, builds the network and finds shared comparators.
Never uses a study the review team did not include.
Comparison feasibility analyst
Proposes a method for each gap and checks effect modifiers, proportional hazards and sample size, following the Coordination Group's guidance. Prefers anchored methods.
Never runs a plan nobody approved. Unanchored results are rated very low.
Dossier writer
Drafts each section in template order, reporting every PICO and outcome, including where no evidence exists, with every number cited.
Never writes a number without a citation, or promotional language.
Consistency checker
Compares every number with its output to the last decimal, the wording with the SmPC draft, and confirms a results section for every PICO.
Never edits silently. It reports the difference.
Who decides on a JCA dossier in this solution, and how the record is kept
Step
Who
Rule
Indirect comparison plan
Biostatistician recommends, dossier lead approves
Only approved plans run; a second program reproduces every estimate
Evidence gap
Dossier lead approves the justification
Stated in the dossier with its reason, as section 3 of the template requires
Results sections
Medical writer reviews, clinical lead interprets
Maker and checker are different people
Label wording
Regulatory affairs
Each difference from the SmPC draft is shown before a version is signed
Each version
Dossier lead
Electronic signature with meaning, name and time; uncited numbers block signing
Final dossier
Head of market access
Signed before upload to the HTA IT platform
Revised scope
Section owners are told
A changed PICO reopens the sections it touches
The solution is built for enterprise use: roles and segregation of duties, maker-checker sign-off, a full audit trail, and integrations with your document and statistics systems. Botminds is SOC 2 Type II audited and manages information security to ISO 27001. These controls support your compliance. A JCA dossier remains the developer's responsibility.32
From the assessors' scope and your study reports to a cited, signed dossier
From the assessors
The assessment scope, with its PICOs and requesting states
Revised scopes and second requests
Clarification requests
From your teams
Clinical study reports and protocols
Statistical outputs and analysis plans
The systematic review, with included studies and search records
SmPC draft and dossier template
Agents read, map, propose, draft and checkPeople decide every comparison and sign every version
What you get
A PICO grid with evidence, gaps, owners and dates
Comparison plans, results and the reasoning behind them
Cited sections in your template
Consistency checks and signed versions
A record of who and which agent did what, and when
The systems a JCA dossier team connects
Document library
Study reports, protocols and analysis plans, as your team files them.
Statistical outputs
Tables, listings and figures, read as they are released, with new outputs mapped in as they arrive.
Systematic review library
Included studies, extraction sheets and flow counts from your review tools.
Validated statistics engine
Runs approved Bucher, network and matched comparisons, and a second program reproduces them.
Templates and labels
The Annex I template, table templates, and the latest SmPC draft.
HTA IT platform
Your dossier and your scope move through the Commission's platform, handled by your team.
17 · Portfolio and settings
Ten JCAs on one page: the portfolio, the coverage dashboard and the rules a dossier team sets
In shortThe head of market access sees every assessment's stage, clock, coverage and next step. A new scope is parsed with visible steps. The clock, the sign-off chain and the evidence rules are settings, so your team's plan drives the reminders.
Every assessment. Ten assessments, five with a clock running. Lumiquel, on day 88, is 12 days from its deadline with 91% of cells checked. The next-step column names one owner per assessment.
A new scope arrives. Orvenza's 14-page scope, 8 PICOs, 21 Member States. The parser's output is checked by a person, and the clock starts from the day of receipt, with nine milestones given owners.
Dashboard. Sections drafted against plan and an estimate of by-hand progress, and evidence coverage for each PICO. The 39 days and the 100% are design targets, measured in the live solution on your own assessments.
Who asked for what. Eleven PICOs by 27 Member States. Germany, France, Italy and Spain each asked for five rows. The row with the fewest asks, PICO 7, comes from Germany and Austria alone, which makes it the easiest to overlook.
The clock and the chain. The window, the days for plans, draft and final sign-off, and who signs what. Set the window to the 100 days the regulation gives, or to your own earlier internal date. For accelerated cases and new indications the regulation gives 60.4
Evidence rules. Prefer anchored comparisons, label post hoc analyses, warn when a matched sample shrinks below 40% of patients, and let uncited numbers block signing. Rules like these turn the Coordination Group's guidance into daily habit.
Questions the team asks about any PICO, and how it answers
The team can ask in plain words about any PICO, comparison, number or deadline, and each answer points to the scope, report, output or review it came from. "How do we handle the EGFR and ALK PICO?" returns the unanchored route through the 61-patient cohort, the 38% effective sample and the very low rating. "How many days do we have left?" returns 63, with the day of each remaining milestone.
Part C · Are you ready to automate?
Automating the JCA dossier: what a late or thin dossier costs, what software returns and how to test it
You lead EU market access or HEOR, run the dossier, or pay for it. This part is written for you: how to price the risk, how to work out a return, where you stand today, and what to ask any vendor, us included.
18 · What a thin dossier costs
What a thin JCA dossier costs you: discontinuation, national re-requests and evidence you cannot defend
In shortStart from your own launch plan, because nobody publishes the cost of a JCA dossier. The public record gives the scale of the risk: three of four discontinued assessments failed on the dossier, every published dossier needed another version, and each late answer runs on a clock of 7 to 30 days.
3 of 4
discontinued assessments ended because the dossier failed Article 9(2), (3) and (4). The fourth followed a withdrawn application.6
Commission tracker, extracted September 17, 2026
12 of 28
PICOs in the four published reports ended with nothing assessed, a gap the countries read in the published report.7,20,21,22
Report tables, our count
22.5%
of indirect comparisons accepted by Germany's G-BA, in assessments published through April 2024. A national figure from a consultancy's abstract.29
ISPOR Europe 2024 abstract
Published figures for a day of delay differ widely, so use your own launch forecast
Tufts CSDD figures for a day of delay in clinical development, from one 645-drug dataset. Development delay, a different event from reimbursement delay.
Where the cost of a 100-day dossier hides, and how to find your number in a week
Cost
Where it hides
How to find your number in a week
Team time on the grid
Filling and re-checking PICO by outcome by hand
Log hours on one real or mock scope, by step, for a week; multiply out by PICO and outcome
Consultancy spend
Simulation, indirect comparisons and writing, invoiced in separate budgets
Pull the last two years' invoices by assessment or mock
Statistics time
Comparison plans, runs and re-runs that displace trial work
Ask your head of statistics for analyses per PICO and weeks per analysis
Late decisions
The row found after its decision date, and the sections that slip with it
Replay a mock scope: on what day did each gap become known?
Second-request rework
15 days of everyone's time, on top of what they already have
Check your dossier against the published lists of missing items; count what you cannot show today
Number errors
Discovered at the 7-day accuracy check, too late for the report
Sample 100 numbers from your last dossier and time how long each takes to trace to its table
A national re-run
If an assessment is discontinued, countries may again request the evidence
List the national dossiers you would need to rebuild, and their launch dates
What a JCA return calculation leaves out: launch order, reputation and your best statisticians' weeks
A published list of what you missedThe Commission publishes the missing items, by case number, for anyone to read.
The countries reading your reportThe people who wrote the PICOs read how you answered them, and the published dossier sits beside the report.
Launch sequenceA slip in the dossier can reorder a launch plan that many functions have already staffed.
Your statisticians' weeksTime spent re-finding numbers comes out of time for the method discussion the assessors will read.
What the next assessment inheritsA spreadsheet tracker teaches the next team little. A kept record of plans and reasons teaches a lot.
19 · Your return
What faster JCA dossiers return: team hours, consultancy spend and fewer second-request rounds
In shortPut in your own numbers. No public source gives the hours or cost of a JCA dossier, so every default below is an illustrative starting point to replace. The calculator counts team time, bought-in spend and rework. It leaves launch timing out, because that figure is yours and much larger.
Six places a JCA dossier team saves time or risk
Evidence grid
88 cells filled, each with its table
Writers and statisticians check values instead of hunting for them.
A gap known on day 10 leaves room for a plan. A gap found on day 60 may not.
days between gap known and decision due, per PICO
Drafting
Template order, neutral language, cited
First drafts arrive with their sources, so reviewers read the argument.
sections × hours saved per section × hourly cost
Bought-in work
Consultants on the hard parts
Expert time goes to method and interpretation, and less to transcription.
dossier consultancy spend × share moved in house
Rework
Fewer rounds after submission
Completeness is checked against the rules and the published lists before the Commission does it.
rounds avoided × hours per round × hourly cost
Accuracy check
Any number, its page, in seconds
Seven days is enough when each number is already linked to its source.
hours spent tracing numbers in the 7-day window
Dossiers a year, team hours and bought-in spend
No default here comes from a published source, because none exists for JCA dossiers. Each one marked illustrative is ours, to be replaced with your numbers from section 18.
Your volumeTeam hours per dossierBought-in work and reworkYour cost
Illustrative yearly value
Dossier team time today, valued$0
Team time released, valued$0
Consultancy spend moved in house$0
Rework avoided, valued$0
After your yearly cost$0
Team hours released a year0
Formulas, per year: Team time today = dossiers × hours today × hourly cost Team time released = dossiers × (hours today − hours with automation) × hourly cost Consultancy moved = dossiers × spend per dossier × share ÷ 100 Rework avoided = dossiers × rework hours avoided × hourly cost After your cost = released + moved + avoided − yearly cost Hours released = dossiers × (hours today − hours with automation + rework hours avoided)
Not counted: launch timing, the cost of a discontinued assessment, or the value of an earlier decision on a comparison. With only one dossier a year, the return may rest on those.
With a few dossiers a year, the hours released decide whether the tool pays
Yearly value after a $400,000 cost, at three dossiers, $120 an hour, 20% of $300,000 moved in house and 200 rework hours avoided
1,800 hours$500K
1,200 hours$284K
600 hours$68K
Illustrative · replace with your own numbers
Show the numbers
Team hours saved per dossier
Value after cost
1,800 hours
$500K
1,200 hours
$284K
600 hours
$68K
The arithmetic: at 1,200 hours saved, 3 × 1,200 × $120 = $432,000, plus 3 × $300,000 × 20% = $180,000, plus 3 × 200 × $120 = $72,000, less $400,000 = $284,000. Test your own case with the calculator; at one dossier a year the same inputs fall below the yearly cost, which is why the return for a small pipeline depends on launch timing and risk.
20 · Readiness check
How ready your team is to automate JCA dossier work: eight scored questions and one to reflect on
In shortScore each question from 0 to 2. Read why it matters and what a strong answer sounds like. Your total points to one of three next steps.
1
Your JCA pipeline
How many assessments do you expect in the next three years, counting oncology, advanced therapies and new indications, and orphan medicines from 2028?
Why it mattersA single dossier may be better served by a consultancy. Repeat assessments reward a kept method.
A strong answer"Six, two with overlapping clocks in 2027, and one likely new indication on the 60-day clock."
2
A rehearsed scope
Have you run a PICO simulation for your next assessment and set your evidence against each row?
Why it mattersA pilot needs a scope to read. A rehearsal is also your first honest list of gaps.
A strong answer"Yes, nine rows, four with no head-to-head data, each with an owner and a date."
3
Access to study reports and outputs
Can the dossier team reach clinical study reports, tables and analysis plans in one place, with rights set?
Why it mattersEvery number must link to its table. Scattered outputs mean scattered citations.
A strong answer"Reports in a managed library, outputs released by statistics with versions, access by role."
4
Systematic review records
Could you produce search strategies, result files, excluded studies with reasons and HTA report searches for your latest review?
Why it mattersThese are the items that appear on the Commission's lists of what was missing.8,9,10
A strong answer"All of it, by PICO, across the six named sources, refreshable within three months of submission."
5
A named comparison decision
Who decides each indirect comparison, and by what date?
Why it mattersSoftware can surface a decision. It takes a person with authority to make it.
A strong answer"The biostatistician recommends, the dossier lead approves by day 44, with a note on file."
6
Your AI record keeping
Could you report each AI-assisted step, with tool name, version, date, maker, purpose and prompts, as the July 2026 principles expect?
Why it mattersThis record cannot be rebuilt afterward, and requests can come with 7 days to answer.32
A strong answer"Yes, exported per step, with a named reviewer on each."
7
A baseline you measured
Do you know the hours, the days to first full draft and the time to trace a number on your last dossier or mock?
Why it mattersNo public benchmark exists. Without yours, a pilot proves nothing.
A strong answer"3,600 hours, day 70 for a full draft, 12 minutes to trace a number in a sample of 100."
8
An owner and a budget
Is one person accountable for the 100 days across market access, statistics, medical and regulatory, with the budget to change how it runs?
Why it mattersThe work crosses four functions. Without one owner, plans wait.
A strong answer"The head of EU market access owns it, with a sponsor in medical and a named budget."
+
For reflection, not scored
If each of your PICOs carried its evidence, its gap, its owner and its next date on one row, which meeting would you stop holding, and what would you do with those hours?
0–6
Explore
Measure a mock scope by step, run a PICO simulation, gather your search records and name the people who decide. Software has little to read until these exist.
7–11
Ready to pilot
Run one real or mock scope in parallel with your team, with planted errors and a dated decision on three numbers.
12–16
Ready to implement
Plan your template, your sign-off chain, connections to your document and statistics systems, your AI record and your next assessment's clock.
21 · Questions for any vendor
What to ask a JCA software vendor or consultancy about PICOs, comparisons, citations and the AI record
In shortAsk every vendor the same questions, us included. The answers show whether the work can be checked: how a PICO is read, who decides each comparison, where each number came from, and what you keep to satisfy the Coordination Group's principles.
A
Reading the scope and the PICOs
A1
Show me a real or sample scope read into rows, with the page and sentence behind each element. Does anything merge or drop a PICO?
Why it mattersSection 3 of the template must reproduce the scope as sent. One list names a dossier that did not.9
Listen forA row per PICO, in the first request's words.
A2
What happens to my plan when the assessors revise the scope?
Why it mattersIt already happened in one published report, with two new dossier versions inside 10 days.22
Listen forA comparison of versions, sections reopened, owners told.
A3
How do you handle an individualized treatment comparator?
Why it mattersIt accounts for much of the unanswered record so far.7,22
Listen forOne estimate against the bundle where a study allows it, otherwise which options the evidence reaches, or a justified gap, per Appendix A of the May 2026 answers.5
B
Indirect comparisons and who decides them
B1
Who proposes a comparison, who approves it, and can anything run before approval?
Why it mattersAccountability stays with the developer, and every AI step needs a human ultimately responsible.32
Listen forNamed roles, and a run that starts only on a signed plan.
B2
Show me the feasibility checks for one comparison, each with its source. What fails, and what happens then?
Why it mattersExchangeability is what sank one dossier.10
Listen forA failed check shown openly, with its effect on the certainty.
B3
What do you propose when there is no shared comparator?
Why it mattersThe guideline calls unanchored matching-adjusted comparisons highly problematic.14
Listen forEffective sample size, limits stated, a justified gap on the table. Be wary of confident claims.
C
Numbers, citations and the template
C1
Pick any number in the draft. How long to open its table, on its page?
Why it mattersAccuracy comments must cite pages, within 7 days.31
Listen forOne click, live, on your own output.
C2
Does the draft follow Annex I as published, and can it hold my template's table collection?
Why it mattersThe dossier must follow the template, and results must sit in section 5.9
Listen forYour template, your numbering, a mapping you can inspect.
C3
How do you stop superiority language where the interval crosses 1?
Why it mattersAbsence of evidence is not evidence of absence, and the report is forbidden value judgments.28,1
Listen forA rule, a test, and a human reader.
D
The AI record, security and your data
D1
Export one run's record: steps, tool name, version, date, maker, purpose and prompts.
Why it mattersThe Coordination Group states these details as its minimum and expects prompts to be kept and available on request.32
Listen forA file someone outside your team can follow.
D2
What accuracy have you measured, on which tasks, and on how many assessments?
Why it mattersPublished evidence is small and often from sellers.33
Listen forSample size and error types, with a plan to measure on your data.
D3
Where does unpublished trial data go, and who can see it? What audits and certificates do you hold?
Why it mattersCopyright and data protection stay with you, and the dossier gets published.32
Listen forCurrent audit reports you can read. Be wary of a vendor who calls its product "GxP certified": software cannot be.
22 · How to start
Test JCA software on one mock scope with planted gaps, then decide on three numbers
In shortMeasure first, test on a scope you already know with errors you planted, and decide with three numbers. None of these steps commits you to a vendor, and all of them help your next dossier even if you buy nothing.
Time a mock scope, by step, with your own team. That is your baseline.
Check your last dossier or mock against the Commission's three lists of missing items, and count what you could not show.
Choose a scope: a PICO simulation or a past national submission, with reports, outputs and the review. Plant known errors: a wrong interval, a post hoc label removed, a PICO dropped, a mismatched comparator.
Run in parallel for eight weeks, with your people deciding every comparison and signing every version.
Decide with three numbers: planted errors caught, hours to a full draft, and minutes to trace a sampled number to its table.
A twelve-week JCA pilot ends with a decision on three numbers
Your teamVendor
Baseline on a mock scope
Gather reports, outputs, review and template
Set up your template and sign-off chain
Parallel run with planted errors
Measure, and read the AI record
Week 04812
Illustrative · replace with your own numbers
Show the numbers
Activity
From
To
Baseline on a mock scope
0
3
Gather reports, outputs, review and template
0
3
Set up your template and sign-off chain
1
4
Parallel run with planted errors
4
10
Measure, and read the AI record
10
12
Planted errors, draft hours and tracing time on your own JCA material
Measure
How
Why
Planted errors caught
Caught ÷ planted, by type, with the number planted
One percentage hides a dropped PICO
False alarms
Findings on clean copies, in reviewer minutes
A tool that flags everything catches every plant
Hours and days to a full draft
Against your baseline, by step
The cost case rests on it, and the solution's 39 days is a target to test
Numbers traced
Sample 300 numbers, each against its table
None wrong in 300 puts the error rate below about 1%, at 95% confidence
Start where the evidence is large and the judgment is small
Our suggestion for ordering the dossier work in a pilot
Volume of reading →
Automate first
Assist closely
Quick wins
Keep with people
Reading the scopeMapping evidenceConsistency checksDrafting resultsFeasibility checksChoosing a comparison
Judgment needed →
Illustrative · replace with your own numbers
Show the numbers
Item
Judgment needed (0–1)
Volume of reading (0–1)
Reading the scope
0.10
0.34
Mapping evidence
0.10
0.80
Consistency checks
0.10
0.56
Drafting results
0.76
0.68
Feasibility checks
0.72
0.24
Choosing a comparison
0.78
0.10
Twelve working-session questions for your own team
Which of our next assessments runs on 60 days, and what does that do to the plan?
Which regulatory date is our HTA calendar counting from, and who tells us when it moves?
When does our systematic review need its last refresh to be inside three months?
Which comparators could the countries add if a competitor reads out next year?
Which subgroups would need interaction tests, and do we have the data?
Where is our individual patient data, and whose permission is needed to use it?
Who writes the justification for a gap, and who approves it?
Who owns the search of HTA reports from the EEA, Australia, Canada, the UK and the US?
What would a confidentiality claim look like, with its reasons, inside 7 days?
Which numbers in the dossier also appear in our national submissions and value materials?
How will we tell the assessors, in the dossier, which steps used AI?
What did we learn last time, and where is it written down?
23 · Questions people ask
EU joint clinical assessment questions answered: PICOs, the 100 days, indirect comparisons and AI
In shortQuestions market access teams and newcomers search for, each answered on its own.
What is an EU joint clinical assessment?
A joint clinical assessment is one scientific assessment of a new medicine's relative effectiveness and relative safety, carried out at EU level by national health technology assessment (HTA) bodies under Regulation (EU) 2021/2282. The countries set the questions as research questions called PICOs, the company submits one dossier, and the assessors write one report. The report carries no value judgment, and price and reimbursement stay national.
What is the EU HTA Regulation?
The EU HTA Regulation, Regulation (EU) 2021/2282 or HTAR for short, is the law behind joint clinical assessments. It has applied since January 12, 2025. It sets common rules and methods for these assessments, and requires a company to submit the evidence once at Union level. Decisions on price and reimbursement stay with each country.
Which medicines need a joint clinical assessment?
From January 12, 2025, new active substances for cancer and advanced therapy medicinal products. From January 13, 2028, orphan medicines. From January 13, 2030, all other medicines in scope of the regulation. Some high-risk medical devices are selected separately, and the first two were announced on October 6, 2026.
What is a PICO in a JCA?
A PICO is one research question: a population, the intervention, one comparator and at least one outcome. Member States state their needs through a survey, and the assessor and co-assessor consolidate them into the fewest PICOs that cover everyone. The published scopes hold 1, 7, 8 and 12 PICOs.
How long does a company have to submit a JCA dossier?
100 days from notification of the Commission's first request, under Implementing Regulation (EU) 2024/1381. It is 60 days under accelerated assessment or for a new-indication variation. The deadline cannot run later than 45 days before the expected opinion of the CHMP, the European Medicines Agency's human medicines committee. Older sources that say 90 days describe the draft act.
What happens if a JCA dossier is incomplete?
The Commission sends a second request, with 15 days to answer, 10 under accelerated assessment or for a new indication, or 7 for minor items. If the dossier is still late or short of Article 9(2) to (4) of the regulation, the Member State Coordination Group on HTA must discontinue the assessment, and the protection against repeat national requests falls away. Three assessments had ended this way by September 2026, each with a published list of what was missing.
Must every PICO be answered with data?
Every PICO must be addressed. Where no valid direct or indirect comparison exists, the developer describes the search and justifies the gap, and whether the dossier counts as incomplete depends on how appropriate the justification is. In the four reports published by November 2026, 16 of 28 PICOs were assessed.
What is the difference between anchored and unanchored indirect comparisons?
An anchored comparison links two trials through a treatment they share, so randomization is kept. An unanchored comparison sets arms from different studies side by side and adjusts for measured differences. The EU's Member State Coordination Group on HTA says anchored methods should always be preferred and calls unanchored matching-adjusted and simulated comparisons highly problematic.
What is a Bucher indirect comparison?
The Bucher method estimates how treatment A compares with B when each has been tested only against C. You divide the two hazard ratios and combine the variance of both on the log scale, so 0.76 against 0.83 gives 0.92. It fits a simple three-treatment star. With only one path between A and B, consistency cannot be tested, so similarity between the trials needs careful assessment.
What is an individualized treatment comparator?
It is a bundle of options used where no single treatment suits everyone in the population, with the choice made patient by patient. The scope defines one PICO against the bundle, with one effect estimate per outcome. It has been hard to answer: in the first published reports, all four such PICOs for tovorafenib and three for onasemnogene abeparvovec had no comparative data.
Can AI be used to prepare a JCA dossier?
Yes, under the principles the EU's Member State Coordination Group on HTA adopted on July 15, 2026. The developer stays accountable, no step should be fully automated without a human ultimately responsible, each AI-assisted step is named with the tool's name, version, date, maker and purpose, and prompts are kept and made available on request. Help with spelling and grammar need not be disclosed.
Can a company comment on the draft JCA report?
Only on purely technical or factual inaccuracies and on confidential information, within 7 days, or 5 under accelerated assessment or for a new indication. It may not comment on the results. Each comment must cite published sources or dossier page numbers, and errors found in the dossier at this point are not acted on in the report.
Does a JCA replace national HTA in Germany or France?
It replaces the repeated clinical evidence work at national level, and countries may not generally request again what was submitted at Union level. Appraisal, added-value conclusions, price and reimbursement stay national, and countries may add other information such as cost-effectiveness. Member States must give the published report due consideration and annex it to their national HTA documents.
24 · One-page summary
The JCA dossier on one page: the method, the November 2026 rules, readiness questions and terms
Print this page
In shortThe whole guide on one screen: the method, the rules as of November 2026, the readiness questions and the terms.
The method
Rehearse the scope early and start the systematic review before the first request.
Reproduce the scope as sent. One row per PICO, an owner and a dated next step for each.
Fill every PICO and outcome cell from a cited table. Label post hoc analyses.
For each gap: anchored first, feasibility cited, a named decision, or a justified gap.
Put all evidence in V0.1, with search records for all six sources and synthesis protocols in Appendix D.
Write neutrally. Results in section 5, by population and PICO. Name each unanswered PICO in section 3.
Check every number against its output and all wording against the SmPC draft, then sign with a meaning.
Keep ready: the AI record, page references for accuracy comments, and reasons for confidentiality claims.
Readiness: score 0–2 on eight questions: 0–6 explore · 7–11 pilot · 12–16 implement.
As of November 2026: 19 ongoing, 4 reports, 4 discontinued on the Commission's lists · dossier 100 days, or 60 · HTACG AI principles July 15, 2026 · AI Act Omnibus in force July 27, 2026 · first devices announced October 6, 2026 · orphan medicines January 13, 2028 · all others January 13, 2030.
14 terms
JCA
Joint clinical assessment: one EU scientific assessment of relative effects.
HTA
Health technology assessment: judging a medicine against what health systems use.
HTACG
The Member State Coordination Group on HTA, which runs JCAs through a subgroup.
PICO
Population, intervention, comparator, outcome: one research question.
Assessment scope
The consolidated PICOs the Member States asked for.
First request
The Commission's request for the dossier; its notification starts the 100 days.
Second request
The Commission's request for what is missing; after it, discontinuation is possible.
Individualized treatment
A comparator bundling options chosen patient by patient.
Anchored comparison
Two trials linked through a shared treatment, keeping randomization.
Bucher method
The simple anchored comparison for a three-treatment star.
MAIC
Matching-adjusted indirect comparison: patients reweighted to match another trial's averages.
Post hoc
An analysis not planned in the trial's statistical plan.
SmPC
Summary of product characteristics: the medicine's EU label.
CHMP
The European Medicines Agency's human medicines committee, whose milestones set the clock.
25 · Sources
Sources
Accessed November 2026. Regulatory status statements are as of November 2026. Conference abstracts are marked as abstracts, and several come from firms that sell this work. We could not open the two regulations on EUR-Lex, so they were read in the EU Publications Office's copy of the Official Journal text.
Regulation (EU) 2021/2282 on health technology assessment, December 15, 2021. eur-lex.europa.eu
European Commission, "Joint Clinical Assessment for Medicinal Products" factsheet, January 2025. health.ec.europa.eu
HTA Coordination Group, Guidance on the scoping process, adopted November 28, 2024. health.ec.europa.eu
Commission Implementing Regulation (EU) 2024/1381, May 23, 2024: joint clinical assessment procedure and templates. eur-lex.europa.eu
HTA Coordination Group, Q&A on general methodological and procedural issues for joint clinical assessments, May 15, 2026. health.ec.europa.eu
European Commission, HTA Secretariat, joint clinical assessments tracker: ongoing list extracted October 15, 2026; discontinued list September 17, 2026; completed list September 7, 2026. health.ec.europa.eu
HTA Coordination Group, Joint Clinical Assessment Report on tovorafenib, endorsed April 30, 2026. health.ec.europa.eu
European Commission, information missing from the dossier of the joint clinical assessment of Tacquell, case JCA-MP-2025-02, February 13, 2026. health.ec.europa.eu
European Commission, information missing from the updated dossier of the joint clinical assessment of catequentinib, JCA-MP-2024-03, May 8, 2026. health.ec.europa.eu
European Commission, information missing from the updated dossier of the joint clinical assessment of sintilimab, JCA-MP-2025-08, September 15, 2026. health.ec.europa.eu
European Commission, FAQ on the publication of joint clinical assessment reports, May 18, 2026. health.ec.europa.eu
D'Oca, Adriano, Darquennes and Steck, "From Joint Clinical Assessment to National Decisions", J Mark Access Health Policy 14(3):50, August 19, 2026. The authors work for a medicine developer. pubmed.ncbi.nlm.nih.gov
European Commission, flash report, 24th meeting of the HTA Coordination Group, October 22, 2026. health.ec.europa.eu
HTA Coordination Group, Methodological Guideline for Quantitative Evidence Synthesis: Direct and Indirect Comparisons, adopted March 8, 2024. health.ec.europa.eu
European Commission, "Commission publishes first Joint Clinical Assessment on innovative medicine", June 9, 2026. health.ec.europa.eu
European Commission, Health and Food Safety newsletter, "HTA Regulation: Commission selects first medical devices for joint clinical assessment", October 6, 2026. ec.europa.eu
HTA Coordination Group, Joint Clinical Assessment Report on lurbinectedin, endorsed June 22, 2026. health.ec.europa.eu
HTA Coordination Group, Joint Clinical Assessment Report on tarlatamab, endorsed June 22, 2026. health.ec.europa.eu
HTA Coordination Group, Joint Clinical Assessment Reports on onasemnogene abeparvovec, endorsed July 16, 2026. health.ec.europa.eu
D'Oca, Darquennes, Garrigues, Draganigos and Steck, "European Joint Clinical Assessment PICO Scoping Process: Analysis of Current Approaches and Recommendations", J Mark Access Health Policy 14(1):3, December 29, 2025. The authors work for a medicine developer. pubmed.ncbi.nlm.nih.gov
Dimiou, Falla, Tanner et al., "Simulating PICOs to predict evidence requirement implications of the EU joint clinical assessment: learnings from 24 case studies", abstract HTA300, ISPOR Europe 2025. The authors work for an evidence vendor. ispor.org
HTA Coordination Group, Guidance on filling in the joint clinical assessment dossier template, medicinal products, adopted November 28, 2024. health.ec.europa.eu
Borah, Brown, Capers and Kaiser, "Analysis of the time and workers needed to conduct systematic reviews of medical interventions using data from the PROSPERO registry", BMJ Open 7(2):e012545, 2017. pubmed.ncbi.nlm.nih.gov
Bucher, Guyatt, Griffith and Walter, "The results of direct and indirect treatment comparisons in meta-analysis of randomized controlled trials", J Clin Epidemiol 50(6):683-691, 1997. pubmed.ncbi.nlm.nih.gov
Altman and Bland, "Absence of evidence is not evidence of absence", BMJ, 1995. pubmed.ncbi.nlm.nih.gov
Warmbold, Scharrenbroich, Teich et al., "Indirect comparisons in German AMNOG assessments: keep the faith", abstract HTA91, ISPOR Europe 2024. The authors work for a consultancy. ispor.org
Thangamma, S, Prasad et al., "Early evidence PICO multiplicity under the EU HTAR: insights from the inaugural joint clinical assessment", abstract HTA114, ISPOR Europe 2026. ispor.org
Annex to the Joint Clinical Assessment Report on tovorafenib: factual accuracy check by the health technology developer and response by the assessors. health.ec.europa.eu
HTA Coordination Group, General principles on the use of AI in the preparation of dossiers for JCA, adopted July 15, 2026. health.ec.europa.eu
Shoughari, Adams, Cichewicz et al., "Faster, mostly faithful: can AI reproduce manual PICO extraction in preparation for EU JCA?", abstract MSR218, ISPOR Europe 2026. The authors include a consultancy and an AI vendor. ispor.org
European Commission, regulatory framework for AI, read November 2026, on Regulation (EU) 2026/1744. digital-strategy.ec.europa.eu
European Medicines Agency, Reflection paper on the use of artificial intelligence in the medicinal product lifecycle, EMA/CHMP/CVMP/83833/2023, final version adopted by CHMP on September 9, 2024. ema.europa.eu
FDA, "Considerations for the use of artificial intelligence to support regulatory decision-making for drug and biological products", draft guidance, January 2025. fda.gov
ISPE, GAMP Guide: Artificial Intelligence, July 2025. ispe.org
European Commission, EudraLex Volume 4, read November 3, 2026; draft EU GMP Annex 22 (consultation July 7 to October 7, 2025). health.ec.europa.eu · picscheme.org
MHRA, GXP data integrity guidance and definitions, section 3.10. gov.uk
Office of Health Economics, "How long are patients WAITing for new treatments?", July 2, 2026, on the EFPIA Patients W.A.I.T. Indicator 2025 survey. ohe.org
Smith, DiMasi and Getz, "Quantifying the value of a day of delay in drug development", Tufts CSDD white paper, August 2024, and the journal article in Ther Innov Regul Sci 58(5):855-862, 2024. csdd.tufts.edu · pubmed.ncbi.nlm.nih.gov
Wang and McAuslane, "Ensuring the efficiency and effectiveness of joint clinical assessment in national HTA decision-making: insights from the 2024 CIRS multi-stakeholder workshop", J Mark Access Health Policy 13(1):9, March 3, 2025. pubmed.ncbi.nlm.nih.gov
Next step
See it on one of your own assessment scopes.
Explore the solution, or bring a scope, your study reports and your dossier template, and we will build the PICO grid with you, every number linked to its table.
We use essential cookies to make this site work, and cookieless, anonymised analytics to understand how it's used. "Accept all" also turns on Google Analytics, which sets cookies. Privacy policy