Potential of joint modelling of longitudinal observations and time-to-event data to improve prognosis in chronic heart failure studies

Field, Ryan James (2024) Potential of joint modelling of longitudinal observations and time-to-event data to improve prognosis in chronic heart failure studies. PhD thesis, University of Glasgow.

Abstract

Background:
Heart failure is a clinical syndrome with inter-relationships between numerous biochemical, physical, and imaging characteristics, and adverse outcomes. Patients with heart failure typically have poor prognosis, and previous prognostic models are limited to using only baseline measurements of these characteristics. In recent years there has been a rise in interest and usage of joint modelling. Joint modelling seeks to combine two or more models, typically containing longitudinal observations and time-to-event (survival) data. The purpose of which is to reduce bias and increase efficiency, allowing for these repeat measurements of longitudinal observations whilst accounting for correlation and measurement error. The inter-relationships within heart failure and properties of joint modelling make heart failure an excellent candidate for joint modelling. It is therefore the aim of this thesis to explore the use of joint modelling in heart failure and to see whether it can be used to improve prognosis.

Methods:
This research comprised of a systematic review paired with an exemplar to introduce joint modelling and how joint models are currently being applied to heart failure; whilst also illustrating how joint modelling can be applied to clinical trial data. Following this, seven joint models were fit under a Bayesian framework, using data from a randomised control trial, and validated with data from different randomised control trials. These joint models were then compared to models fitted using current standards of prognostic model methodology to evaluate and assess how joint modelling can potentially improve model performance. Finally, a web application was developed to illustrate how these joint models can translate into real world applications.

Results:
On average the joint models performed better, in a statistical sense, than the traditional models (considered the current standard for prognosis) and performed adequately when validated with data from another randomised control trial. The web application effectively shows how these joint models can be used in practice and highlights the potential of the dynamic nature of joint models when used in a prognostic setting.

Conclusion:
This thesis illustrates how joint modelling can improve on the current standard of prognostic models, adding repeated measurements and allowing for dynamic predictions over time, whilst outperforming the traditional models. However, with limitations around the use of latent parameters such as random effects, and the novel nature of these models with their limited use, it may be prudent to wait until these types of models mature, are evaluated further, and the statistical packages used to fit these models mature before implementing them in clinical practice.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: R Medicine > R Medicine (General); R Medicine > RA Public aspects of medicine > RA0421 Public health. Hygiene. Preventive Medicine
• Colleges/Schools: College of Medical Veterinary and Life Sciences > School of Health & Wellbeing > General Practice and Primary Care
• Funder's Name: British Heart Foundation (BHF)
• Supervisor's Name: Lewsey, Professor James and Jhund, Professor Pardeep
• Date of Award: 2024
• Depositing User: Theses Team
• Unique ID: glathesis:2024-84042
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 19 Jan 2024 16:05
• Last Modified: 19 Jan 2024 16:15
• Thesis DOI: 10.5525/gla.thesis.84042
• URI: https://theses.gla.ac.uk/id/eprint/84042

Related URLs

• Article DOI
• Enlighten Publications Record
• Article DOI
• Article DOI
• Article DOI