Prognosis of neurodegenerative diseases
: methodological and empirical results for Multiple Sclerosis and Parkinson’s disease

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)

Abstract

Neurodegenerative diseases, like Multiple Sclerosis and Parkinson’s disease, lead to
disability that worsens with time. Being able to predict prognosis in these diseases is
important for both patients and clinicians when making medication choices and planning for
the future. My aim was to look at drug effectiveness over a ten-year period in Multiple
Sclerosis, in the absence of a long-term clinical trial, and to look at prognosis in Parkinson’s
disease by deriving subtypes.

I developed a longitudinal model for the untreated natural history of patients with Multiple
Sclerosis using multilevel models. This model was used to predict the untreated trajectories
of treated MS patients over a ten year period. A comparison between the observed treated
trajectories and the predicted untreated trajectories gave an estimate of long-term drug
effectiveness. I carried out intention-to-treat and per-protocol approaches along with imputed
analyses to adjust for missing data. The medications were found to be effective in the long
term.

I used a k-means cluster analysis on the baseline phenotype of a large cohort of recently
diagnosed Parkinson’s patients to attempt to derive subtypes. These subtypes were
subsequently found to be associated with medication response. This approach was extended
in another large inception cohort using a development and validation approach. Before
combining the two cohorts in an analysis I had to harmonise the data from the cohorts as
olfaction was measured using two different tests. I used Item Response Theory to convert the
two tests onto the same scale. The harmonised baseline phenotypic data from both cohorts
was used to estimate subtypes. This approach was relatively stable when comparing the
actual and predicted subtypes in the smaller cohort. These subtypes were associated with
differing rates of motor progression and to medication response.
Date of Award24 Mar 2020
Original languageEnglish
Awarding Institution
  • The University of Bristol
SupervisorYoav Ben-Shlomo (Supervisor) & Kate M Tilling (Supervisor)

Cite this

'