Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.749773
Title: Definition of trajectories of joint space narrowing in knee osteoarthritis in the presence of measurement error
Author: Parsons, Camille Michelle
ISNI:       0000 0004 7234 1795
Awarding Body: University of Southampton
Current Institution: University of Southampton
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Abstract:
Osteoarthritis (OA) is the most prevalent joint disease in older adults, with the knee being the most commonly affected joint. The disease progression is slow, often over decades. During progression the breakdown of cartilage occurs within the affected joint, causing the joint space to narrow such that ultimately the joint will fail. This process leads to great pain and disability on an individual level, and an ever increasing economic burden to health services. The gold standard measurement currently used within epidemiological studies to monitor the progression of the disease is joint space narrowing. However, changes in joint space measurements over time are small and sensitive to measurement error. This makes identification of real deterioration difficult, and maybe masking the identification of factors associated with progression. Therefore the research aims of this thesis are to develop statistical methods that identify change in continuous knee joint space measurements over time within individuals diagnosed with knee osteoarthritis. Two datasets were used in this thesis; the Strontium ranelate Efficacy in Knee Osteoarthritis trial (SEKOIA), a 3-year international, multicentre, double-blind, placebo-controlled phase 3 trial, and the Osteoarthritis Initiative (OAI), an open-access longitudinal dataset following study participants in the United States of America. Three different statistical methods were applied to both datasets; the reliable change (RC) index, frequentist linear mixed effect (LME) modelling and Bayesian hierarchical modelling, to determine how each method accounted for measurement error within the knee joint space width measurements. When compared with crude differences in knee joint space width, implementation of the novel approach of the RC index dramatically reduced the proportion of study participants in both datasets that were identified as having statistically reliable change after accounting for measurement error. The RC index allows for calculation of a magnitude of change threshold that gives a figure above which it can be said reliable change had occurred; the threshold was found to be approximately 1mm in a year in both SEKOIA and the OAI. Using the frequentist techniques of LME modelling, on average, joint space width decreased by 0.14mm per year in SEKOIA and 0.08 in the OAI, and results from Bayesian hierarchical modelling indicated a similar pattern. Each statistical method used within this project handles measurement error in different ways. The RC index only considers two time points at a time, but accounts for measurement error by using the variability of the measurement within the calculation. Both the frequentist linear mixed effect and Bayesian modelling account for measurement error in a similar way, by using all available time points from the total population within the modelling process to ‘smooth’ individual trajectories of change. These methods could prove valuable in monitoring of disease progression and in identification of risk factors for the progression of knee OA which have not yet been discovered. The RC index threshold of change could be directly applied within a patient population to aid clinicians in monitoring progression of knee OA. In a research environment, the simpler LME modelling technique could be implemented to ‘smooth’ data for measurement error and obtain individual estimates of change. The more complex Bayesian methodology can be used to investigate phenotypes for disease progression. Additionally, application of these statistical methods in other clinical areas may aid in monitoring of other conditions.
Supervisor: Inskip, Hazel ; Cooper, Cyrus Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.749773  DOI: Not available
Share: