RSS sections and groups meeting reports

Merseyside local group meeting: Joint modelling

Written by Laura Bonnett on . Posted in Sections and local group meeting reports

On 20 September 2016, the RSS Merseyside local group hosted a full afternoon session on joint modelling. There were 22 people in attendance, approximately half of whom were RSS members. As usual, most attendees were from the University of Liverpool. However, there were also visitors from the University of Leicester, Liverpool School of Tropical Medicine, and the Health and Safety Executive.  

Virginie Rondeau, director of research at INSERM in Bordeaux began the afternoon with an extended talk considering the use of tumour dynamics and new lesions to predict survival with multivariate joint frailty models. Virginie compared the predictive ability of cancer progressions based on categorical criteria and quantitative measures of tumour size and/or appearance of new lesions on overall survival. Outlining work with her PhD student, Agnieszka Krol, Virginie proposed a joint model for a simultaneous analysis of three types of data: longitudinal marker, recurrence events and a terminal event. Results from a simulation study showed that the proposed trivariate model was appropriate for its practical application. Virginie concluded her presentation by suggesting statistical tools for evaluating predictive accuracy of joint models to compare the developed model to models based on categorical criteria and their components.

After a well-earned coffee break the RSS Merseyside local group welcomed three speakers from the University of Liverpool’s Joint Modelling Group. The first presenter was PhD student Maria Sudell. Maria outlined how joint analysis of longitudinal and time-to-event data is often preferable over separate longitudinal or time-to-event analyses. She also explained that the current literature on joint modelling is limited for situations where joint longitudinal and time-to-event data are recorded from multiple sites or studies, such as during meta-analyses. Maria went on to present a range of possible one-stage models for the meta-analysis of individual participant joint longitudinal and time-to-event data including use of fixed and random effects at study level, and stratified baseline hazards. Maria concluded her whistle-stop tour with the results of a simulation study and the application of the various methods to a dataset constructed to establish how the efficacy of pharmacological treatment for high blood pressure depends on patient characteristics.

Graeme Hickey, post-doctoral research associate, was next to take to the floor. He began by reminding the audience that there are limited options for fitting joint models of longitudinal data and competing risks in standard statistical software programs. He also intimated that a lack of awareness and understanding of the different models and their corresponding software applications prohibits their adoption when undertaking the statistical analysis of a clinical dataset. Graeme proceeded to review five published models, each of which has either in-built software or purpose-built code available for model estimation. To illustrate how the models could be used in practice, Graeme fitted each model in turn to a real dataset from a randomised control trial of new anti-epileptic drug treatment. He concluded that further work is required before integration into routine biostatistical analyses can be realised.

The afternoon concluded with a presentation by senior lecturer, Ruwanthi Kolamunnage-Dona, who is the lead of the Joint Modelling Group. Ruwanthi’s talk focussed on modelling variable dropout in randomised controlled trials with longitudinal outcomes. Ruwanthi explained that these dropouts usually arise due to patients withdrawing or being lost to follow-up before completing the measurement schedule. The reasons for dropout are sometimes clearly known and recorded during the trial, but in many instances these reasons are unknown. Usually, such reasons are non-ignorable. However, Ruwanthi explained that the standard methods for analysing longitudinal outcome data assume that missingness is non-informative. Therefore, she explored the impact of informative dropout due to competing reasons on the evaluation of treatment effect in the largest randomised placebo-controlled study to data comparing the addition of nebulised magnesium sulphate to standard treatment in acute severe asthma in children. Ruwanthi explained how she had jointly modelled the longitudinal outcome and informative dropout process to incorporate the information regarding the reasons for dropout by treatment group. Ruwanthi concluded her talk by emphasising the importance of identifying reasons for dropout and undertaking an appropriate statistical analysis accounting for such competing dropout reasons.

The RSS Merseyside local group committee received positive feedback on the session. Although audience members were disappointed that joint modelling cannot currently handle a large number of outcomes it did spark interest in collaborations between research groups. The hunt for relevant further datasets has also begun!

The next meeting of the RSS Merseyside local group will take place on Thursday 17 November at 2pm at the University of Liverpool with David Firth from the University of Warwick presenting a talk on ‘Football Leagues: A better mid-season ranking than the points tally?’ Further details of this, and all future talks, are available on the Merseyside local group website:


Merseyside local group

Join the RSS

Join the RSS

Become part of an organisation which works to advance statistics and support statisticians

Copyright 2019 Royal Statistical Society. All Rights Reserved.
12 Errol Street, London, EC1Y 8LX. UK registered charity in England and Wales. No.306096

Twitter Facebook YouTube RSS feed RSS feed RSS newsletter

We use cookies to understand how you use our site and to improve your experience. By continuing to use our site, you accept our use of cookies and Terms of Use.