Excelya at PSI 2026 Conference: Clinical Statistics Insights

AI is moving deeper into clinical statistics and programming

AI was everywhere at PSI 2026. The promise of AI agents working alongside statisticians and programmers is broad, especially as workflows evolve toward AI double programming, AI-assisted bibliographic review and more automated evidence generation. However, restrictions affecting foreign access to some AI tools, together with concerns around sensitive data leakage through web search, also raised an important question for clinical research: should the industry develop more sovereign AI tools designed specifically for biostatistics, statistical programming and regulated clinical development?

Joint Clinical Assessment is reshaping evidence planning

The first Joint Clinical Assessment procedures show that a new operating model is needed to manage extremely tight timelines, including the 100 day window to submit a dossier from the initial request. Statisticians are becoming increasingly important in anticipating the multiple potential PICOs that may be requested by JCA subgroup members. As a result, statistical teams now play a more strategic role during the scoping phase, where early planning can determine whether evidence packages are ready, robust and responsive.

Real-world evidence expectations are becoming clearer

Real-world evidence generated strong discussion at PSI 2026, particularly around recent guidance and the EMA reflection paper on the use of real-world data in non-interventional studies to generate real-world evidence for regulatory purposes. Adopted by CHMP PROM in March 2025, the paper aims to support industry and academia while increasing consistency on the regulatory side. It clarifies that real-world data should answer questions that cannot be addressed by clinical trials alone, and that evidence standards may differ depending on the data source. This gives sponsors more direction on how real-world evidence studies should be designed, assessed and interpreted.

EMA guidance is refining non-inferiority and equivalence analysis

The PSI Conference remains a valuable forum for discussion with regulators and industry leaders on recently published and upcoming guidance. One important example was the EMA draft guideline on non-inferiority and equivalence. The draft aims to streamline how these analyses are described, justified and interpreted in randomized clinical trials. It presents non-inferiority and equivalence as analysis types rather than trial types, highlights the importance of conservative analyses, confirms that the primary analysis must match the primary estimand, addresses intercurrent events and avoids making per-protocol or intention-to-treat analyses mandatory in all cases.