We would like to introduce the new R package {jmpost}.
{jmpost} supports fitting joint models (JM) involving two sub-models and a link function.
The longitudinal sub-model describes the non-linear time dynamics of a Gaussian-distributed biomarker (Y); the time-to-event sub-model describes the hazard of a terminal event (right-censored); and the link function describes the association term relating the two sub-models.
For the time being, the package considers the case of tumor size as a biomarker and death as a terminal event. Tumor size and death events are frequently collected in oncology drug development e.g. in Phase 3 trials. A possible motivation to fit such a JM would be to estimate and/or predict the treatment effect on survival, leveraging the correlation between longitudinal tumor size data and overall survival. Several research advances have been published on this topic (see main page).
With {jmpost} we aim to enable data scientists to apply this methodology more easily.
The development team is led by Craig Gower-Page. The current list of contributors to the package includes Daniel Sabanes Bove, Georgios Kazantzidis, Xiaoyan Fang, Isaac Gravestock, and Francois Mercier.
The package development is in progress, but now is a great time for us to make the community aware of it. Please have a look at {jmpost}, and try it out (e.g. fitting a joint model and looking at estimated longitudinal and survival fit). Please let us know if you hit any bugs, have any ideas, or positive feedback (e.g. everything works). All feedback is valuable and very welcome - best post it on GitHub.
It is also worth noticing that we use the {jinjar} package to help with writing the Stan templates and combining them in the backend of our package.
Note: The package was introduced by Francois Mercier and Daniel Sabanes Bove from Roche, on 8 June 2023, during the Stan for Pharmacometrics one-day conference. Overall about 80 participants joined from both the Pharmacometrics and Statistics domains. Due to its focused scope and speakers (including Andrew Gelman and Aki Vehtari), it was a great conference with interesting discussions. Thanks to the organizing team (especially Julie Bertrand from IAME) for putting the conference together! We are looking forward to the next event already!