メニュー 戻る Poster-Presentations-Details

PP11-81: Utilizing a Bayesian Hierarchical Model to Design Quality Into a Clinical Trial and Allow Compliance with ICH E6 R2 Quality T





Poster Presenter

      Christine Wells

      • Senior Statistical Scientist
      • Roche Products Ltd
        United Kingdom

Objectives

The objective of this work is to introduce the concept and implementation of the use of Bayesian Hierarchical Modeling to Quality Tolerance Limits as mandated by ICHE6 R2

Method

ICHE6 R2 has mandated the use of Quality Tolerance Limits. Roche have utilized a Bayesian Hierarchical Model methodology, inspired by the Bayesian Meta analysis example in Berry et al (2011). Fixed parameters specify the prior for all unknown parameters, a conservative prior can be used or it can be

Results

Roche have been utilizing this Methodology on all Roche Studies since August 2019. Currently we are utilizing 3 global parameters, under/over reporting of AEs, over reporting of SAEs and Protocol Deviations (split into Baseline Deviations and Ongoing Deviations). Study specific QTLs will follow in the next wave. The parameters are assessed at a study level and a threshold is used to identify possible systematic issues. Distributional plots and parameter estimates for historical studies (Reference studies) are created and this information is used to set the QTLs and associated thresholds for the parameters. The modelling is then run across the ongoing study (Target Study) where the study level median rate is examined and if necessary action is taken down to the site level where issues can be targeted for review. Any mitigating actions are then documented and collected and will be reported in the CSR if they identify Systematic error. When the modelling is run on a 12 weekly basis we also begin to build up control charts of the study median for the associated parameter, which then signifies that mitigating action is taking effect. We are seeing that this methodology gives us great insight into the performance of sites and the data under analysis.

Conclusion

This Robust Methodology allows teams to model the AE rate of historical studies to inform the expected AE rate of the study under investigation. Teams can have full knowledge of the median rate of the parameter of interest throughout the study and hence full insight of the site adherence to reporting processes. This therefore further safeguards patient safety. The following people have input to this work: Chris Wells (Roche), Catherine Tomlinson and Heather Turner (PRISM Training and Consultancy), Tammy McIver and John Kirkpatrick (Roche), Sukalpo Saha (TATA Consulting Services)