FROM GUIDANCE TO PRACTICE: METHODOLOGICAL LESSONS FROM FIRST-YEAR EUROPEAN UNION JOINT CLINICAL ASSESSMENT (JCA) SUBMISSIONS
Author(s)
Asma Jouini, Biology engineer1, Mikolaj Parkitny, MSc2, Aleksandra Caban, PharmD2, Piotr Wojciechowski, MSc3, Elzbieta Olewinska, MSc3, Aleksandra Matjaszek, MSc2, Mondher Toumi, MSc, PhD, MD4.
1Clever Access, Paris, France, 2Clever Access, Kraków, Poland, 3Clever-Access, Kraków, Poland, 4Aix-Marseille University, Marseille, France.
1Clever Access, Paris, France, 2Clever Access, Kraków, Poland, 3Clever-Access, Kraków, Poland, 4Aix-Marseille University, Marseille, France.
OBJECTIVES: The JCA sets methodological requirements across many guidelines but offers few real-life examples. Drawing on first-year oncology submissions, we distil recurring lessons for health technology developers (HTDs).
METHODS: We synthesised technical learnings from supporting first-year JCA submissions, the applicable guidelines, the Q&A documents, and assessor requests.
RESULTS: Lessons clustered in seven areas. (1) Scope and feasibility: excluding a PICO on absent comparator evidence requires systematic literature searches for both RCTs and non-RCTs. (2) Comparators: an individualised-treatment bundle requires one combined effect estimate per outcome against the whole bundle, not one per option; for several comparators connected by "OR" with same-level evidence, all must be presented, with an estimate against each plus an aggregated one where appropriate. (3) Indirect comparison validity: both sides of a comparison must share an identical type of effect measure (patient-reported outcomes are the main failure point), and the proportional-hazards (PH) assumption must be assessed for every time-to-event outcome, if violated, an appropriate non-PH method must be pre-specified and applied. (4) Subgroups: many are expected per PICO (≈10), reporting both factor levels, every outcome, the interaction test and code, scoped by a 10-patient/10-event threshold, at PICO and network level. (5) Data cuts: report the last pre-specified cut for all outcomes (plus the last-available cut for overall survival if different); at all reported interim and final cuts, all outcomes are expected even if the data cut-off was originally planned only for a subset of endpoints. (6) Safety: extend time-to-event analysis, with subgroups, to aggregated safety outcomes whenever exposure differs between arms, as event counts are insufficient. (7) Certainty: dual study-level (RoB 1) and outcome-level (RoB 2) appraisal resolves the template tension.
CONCLUSIONS: The principal challenge is operationalising broadly stated requirements where real-life examples are scarce. Anticipating these recurring assessor expectations early may reduce information requests and support more efficient, predictable JCAs.
METHODS: We synthesised technical learnings from supporting first-year JCA submissions, the applicable guidelines, the Q&A documents, and assessor requests.
RESULTS: Lessons clustered in seven areas. (1) Scope and feasibility: excluding a PICO on absent comparator evidence requires systematic literature searches for both RCTs and non-RCTs. (2) Comparators: an individualised-treatment bundle requires one combined effect estimate per outcome against the whole bundle, not one per option; for several comparators connected by "OR" with same-level evidence, all must be presented, with an estimate against each plus an aggregated one where appropriate. (3) Indirect comparison validity: both sides of a comparison must share an identical type of effect measure (patient-reported outcomes are the main failure point), and the proportional-hazards (PH) assumption must be assessed for every time-to-event outcome, if violated, an appropriate non-PH method must be pre-specified and applied. (4) Subgroups: many are expected per PICO (≈10), reporting both factor levels, every outcome, the interaction test and code, scoped by a 10-patient/10-event threshold, at PICO and network level. (5) Data cuts: report the last pre-specified cut for all outcomes (plus the last-available cut for overall survival if different); at all reported interim and final cuts, all outcomes are expected even if the data cut-off was originally planned only for a subset of endpoints. (6) Safety: extend time-to-event analysis, with subgroups, to aggregated safety outcomes whenever exposure differs between arms, as event counts are insufficient. (7) Certainty: dual study-level (RoB 1) and outcome-level (RoB 2) appraisal resolves the template tension.
CONCLUSIONS: The principal challenge is operationalising broadly stated requirements where real-life examples are scarce. Anticipating these recurring assessor expectations early may reduce information requests and support more efficient, predictable JCAs.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
HTA89
Topic
Health Policy & Regulatory, Health Technology Assessment, Medical Technologies
Disease
Oncology