Use of Joint Models in Health Technology Assessment (HTA) – The Future?


Moderator: Bram Ramaekers, PhD, Department of Clinical Epidemiology and Medical Technology Assessment (KEMTA), Care and Public Health Research Institute (CAPHRI), Faculty of Health, Medicine and Life Sciences, Maastricht University Medical Centre+, Maastricht, LI, Netherlands
Panelists: Louise Linsell, DPhil MSc BSc CStat, Visible Analytics, Oxford, OXF, UK; Keith R Abrams, PhD MSc CStat, Statistics / Centre for Health Economics, University of Warwick and University of York, Coventry / York, Warwickshire / North Yorkshire, UK; Grace Lee, PhD, Critical Path Institute, Tucson, AZ, USA

ISSUE: Joint modelling [JM] has received considerable attention in the clinical and methodological research literature. The ability to model one or more time-to-event outcomes simultaneously with one or more longitudinal biomarkers offers potential benefits. Not only can the strength of association between biomarker and clinical event be assessed, but when one of the outcomes is death, the biomarker profiles are automatically adjusted for informative dropout. In addition, individual-level predictions of time to an event, conditional upon a biomarker profile up to a specific time point, can be obtained. A number of packages in R and Stata have been developed to implement joint modelling, and for more complicated applications a Bayesian approach using Markov Chain Monte Carlo (MCMC) methods can be adopted. However, there has been less development of methods to assess covariate selection and model performance, and to date applications of JMs in a HTA context have been limited. This Issue Panel will Illustrate their use with motivating examples, and discuss their benefits and limitations.

OVERVIEW: Manuela Joore will introduce and provide an overview of Joint Modelling [JM] (10 minutes – including audience participation via Slido), and the three panellists will then introduce three potential application areas (in CVD, cancer & neurology) of JMs in HTA (10 minutes each) – Keith Abrams, adjusting longitudinal Quality of Life profiles for both informative dropout due to death and intermittent missing data in a CVD clinical trial; Louise Linsell, extrapolation of overall survival utilising longitudinal biomarker profiles in three phase I/II trials of adults and children with TRK fusion cancer, and Varun Aggarwal, modelling the natural history of disease of patients with Duchenne Muscular Dystrophy. Manuela Joore will then moderate a discussion involving the audience and panellists on benefits and limitations of using JMs in HTA , before eliciting the audience’s final views (via Slido).




Methodological & Statistical Research