Use of Joint Modelling in Health Technology Assessment (HTA)

Abstract:

Joint Models (JM) are used to simultaneously analyse data on longitudinal biomarkers and time-to-event outcomes. By doing so they enable either prediction of future clinical events, conditional upon biomarker profiles, and/or estimation of biomarker profiles after allowing for dropout due to death when the time-to-event is death. To date they have been developed both within a frequentist and a Bayesian framework, though the latter more easily enables a number of extensions, such as imposing a hierarchical structure and allowing for intermittent missing data. In this talk I will describe the development and application of JMs to a number of settings in Health Technology Assessment (HTA). Specifically the use of a JM approach to; (i) modelling HRQoL with both dropout due to death and intermittent missing data in a trial to evaluate Transcatheter Aortic Valve Implantation (TAVI); (ii) allow for differential associations between tumour burden/size and overall survival (OS) in a tumour agnostic study of Larotrectinib in order to enhance extrapolation of OS; (iii) allow for informative observations in Electronic Health Record (EHR) data and applied in pre-trial screening of routine estimated glomerular filtration rate (eGFR) measurements for patients with Chronic Kidney Disease (CKD); (iv) developing a natural history model in Duchenne Muscular Dystrophy (DMD). I will also discuss further extensions, and in particular how JMs can be estimated in large RWE and EHR databases without a computationally prohibitive burden.

If you are not a member of University of York staff and are interested in attending a seminar, please contact alfredo.palacios@york.ac.uk or shainur.premji@york.ac.uk so that we can ensure we have sufficient space.