At ISPOR Europe 2026 in Vienna, Dr Farzana Malik is presenting a Cogience poster on Cogience.Predict, our programme for forecasting the PICO scope of an EU Joint Clinical Assessment before the assessment report is published.

The problem

Since January 2025, every new oncology medicine and advanced therapy entering the EU has gone through a Joint Clinical Assessment (JCA). The consolidated PICO scope (the populations, comparators and outcomes the dossier must address) reaches the developer only 60–100 days before the dossier is due. That leaves too little time to build the evidence from scratch, so teams have to anticipate the scope months in advance.

Published prediction methods have so far been checked after the fact. None has been tested against an answer that was fixed before the prediction was made.

What we did

  • Sealed predictions. For each product we locked the scoring rules, built the prediction, then time-stamped and froze it before the published JCA report was opened. Only then did we score it.
  • Reasoning from sources, not a trained model. Too few JCA reports exist to train a model on. Instead, reasoning AI models research the pivotal trial, the label, EU standard of care, reimbursement status and prior national HTA decisions, citing each source. The scope is argued from those findings, with a logged rationale for every population split and comparator.
  • A learning loop. Each scored report is traced back to its causes and written into the next version of the method before the next prediction is built.

Early results

This is an ongoing programme with four cases scored so far, so the figures are interim:

  • The comparators we predicted were almost always right: 11 of 12 (92%).
  • The comparator that drove each decision was found every time: the one with usable comparative evidence behind it.
  • The main error was in how assessors split populations, not in which comparators they chose. The predictions found about two thirds of published comparators, rising to three quarters at class level.

What it means for dossier teams

On the evidence so far, comparator evidence searches, indirect comparison feasibility and the outcome set can be started months before the scope is published, then re-cut by population once it lands. Further predictions are sealed and awaiting their reports, and each one tests the latest version of the method.

If you are at ISPOR Europe, come and find the poster. To discuss JCA scope prediction for your asset, get in touch.