Ryan is a second year PhD candidate in Public Health, investigating joint modelling within heart failure.
Joint Modelling is a statistical technique for modelling longitudinal repeat measures together with survival data.
The Longitudinal data is typically modelled using an LME (Linear mixed effects model), whereas the survival data is modelled using standard regression survival models such as Cox Proportionate Hazards Model
These models are then linked through an association structure via their shared random effects, i.e. the individual patients or participants, resulting in a joint model
For my PhD I am exploring the uses of Joint Modelling within heart failure, with the aims to produce a prognostic model using repeat measures
Some advantages of Joint Modelling are:
- It can reduce Bias both with respect to treatment effect and (bias) due to censoring
- It is more efficient than traditional Models
- And it allows the use of repeat measure while inherently accounting for measurement error and allowing for different follow-up times
On the other hand, some disadvantages are:
- It requires repeat measurements to be integrated in the study design
- More complex and harder to fit than traditional models thus taking longer to fit
- And not as well known as traditional model
Currently in heart failure the primary focus is on the association of longitudinal repeat measurements and an endpoint such as a composite endpoint
Longitudinal measurements include: Quality of Life and Bio Markers such as: Natriuretic Peptides e.g. NT-ProBNP and Renal markers such are Creatinine
Other novel applications include physical activity as reported by an implanted device such as an ICD
There are a limited number of studies looking at prognostic models and individual patient survival predictions
My research aims to focus more on the prognostic value of bio markers in Heart Failure using data collected from randomised control trials in an effort to improve the existing prognostic models and create new prognostic models using joint modelling.
From the image we can see the results from a joint model presented as a graph looking at a patient specific trajectory with the longitudinal measurements on the left and the survival probability on the right.
The model assumes that the patient has survived until the last measurement and calculated the survival probability from that point.
The joint model can also provide coefficients from both the longitudinal and survival models with the latter being exponentiated to produce a hazard ratio e.g. of treatment effect or the association of the biomarker through an association parameter.
Ryan can be contacted at:
This presentation was presented as part of an IHW PGR half-day conference, All aboard
- Tsiatis AA, Degruttola V, Wulfsohn MS. Modeling the Relationship of Survival to Longitudinal Data Measured with Error. Applications to Survival and CD4 Counts in Patients with AIDS. J Am Stat Assoc. 1995;90(429):27–37.
- Tsiatis AA, Davidian M %J SS. Joint modeling of longitudinal and time-to-event data: an overview. 2004;809–34.
- Ibrahim JG, Chu H, Chen LM. Basic concepts and methods for joint models of longitudinal and survival data. J Clin Oncol. 2010/05/05. 2010;28(16):2796–801.
- Papageorgiou G, Mauff K, Tomer A, Rizopoulos D. An Overview of Joint Modeling of Time-to-Event and Longitudinal Outcomes. Annu Rev Stat Its Appl. 2019;6(1):223–40.
- Sudell M, Kolamunnage-Dona R, Tudur-Smith C. Joint models for longitudinal and time-to-event data: a review of reporting quality with a view to meta-analysis. BMC Med Res Methodol. 2016;16(1):168.
- Rizopoulos D. JM: An R package for the joint modelling of longitudinal and time-to-event data. J Stat Softw. 2010 Jul;35(9):1–33.
- Rizopoulos D. The R package jmbayes for fitting joint models for longitudinal and time-to-event data using MCMC. J Stat Softw. 2016 Jan;72(1):1–46.
- Rizopoulos D. Joint Models for Longitudinal and Time-to-Event Data: With Applications in R. 2012.