Masoud Jamei 1, Céline Brochot 1, Pauline Bogdanovich 1, Sibylle Neuhoff 1, Oliver Hatley 1, Iain Gardner 1, Karen R. Yeo 1
1 Certara Uk Limited (Sheffield, United Kingdom)
Introduction: The Simcyp Simulator V19 was qualified for predicting drug-drug interactions for 6 CYP enzymes (CYP1A2, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP3A4) for three contexts of use (CoUs) by the European Medicine Agency (EMA) in 2025 [1]. Since the qualification process was started several versions of the Simcyp Simulator have subsequently been released. Given V19 qualification, it is imperative that new versions of the Simulator can demonstrate consistency with the qualified version.
The EMA Qualification states that ‘Every time a new Simcyp version is used in regulatory submissions a de novo justification of assumptions and methods for uncertainty quantification may not be needed if it is demonstrated that the CoU, Qualification matrix and scope complies with the V19 qualification space.’ Although it is stated that a de novo justification may not be needed, the acceptance criteria for compliance with the V19 qualification space is not specified. Therefore, the question becomes – what metrics will best evaluate the predictive performance of the subsequent versions when compared against those of V19? The typical prediction performance criteria e.g. Average Fold Error (AFE) or Absolute Average Fold Error (AAFE) can be used however these assessments may not reflect the variability changes from one version to another.
Methods: In order to tackle this issue, we propose testing the equivalence of the Simcyp predictions for the AUC and Cmax ratios obtained with subsequent versions (V20 to V24) to the V19 predictions using Two One-Sided Tests (TOST), as it is conducted in typical bioequivalence (BE) assessment. In this assessment the V19 results are considered as the ‘reference set’ and the V20-V24 results are treated as ‘test sets’. If the BE tests are passed, indicating that each version is ’equivalent’ to the ‘reference’ version (V19), it would seem appropriate to consider that all ’equivalent’ versions are qualified for the same CoUs.
Results: The same qualification matrix of 220 studies for AUC Geometric Mean Ratio (GMR) and 159 studies for Cmax GMR were simulated in (V20 to V24) (test results) as shown below.
Version AUC GMR Mean Cmax GMR Mean
V19 3.534 2.123
V20 3.459 2.083
V21 3.478 2.092
V22 3.520 2.086
V23 3.578 2.102
V24 3.513 2.117
Then the simulation results were compared against those of V19 (reference results) testing for BE within (80% – 125%) range. The BE assessment results clearly demonstrate that all versions’ predictions are equivalent to V19 predictions.
Conclusions: Every year a new version of the Simulator is released where current features and databases are expanded and/or updated based on the latest scientific advancements. Due to these changes a new version of the Simulator may not produce identical results to the previous version. To assess the performance of subsequent versions of the EMA qualified V19, the simulations result of V20-V24 of the Simcyp Simulator using the same dataset were compared against those of V19 using TOST. The results showed that all versions are equivalent producing consistent results.
Disclaimer: a version of this abstract is submitted to the 1st International PBPK Conference, Basel, Swizerland, April, 2026.
References:
EMA website visited on 20th Feb 2026; Opinions and letters of support on the qualification of novel methodologies for medicine development | European Medicines Agency (EMA)
Reference: PAGE 34 (2026) Abstr 12187 [www.page-meeting.org/?abstract=12187]
Poster: Drug/Disease Modelling - Absorption & PBPK