Statistical methods for disease burden and cost-effectiveness studies, and diagnostic testing
18 september 2020
Online defence - - - - -
15 - 17 uur
Prof. N. Hens (Universiteit Hasselt), Prof J. Bilcke
Joint PhD defence Zoë Pieters - Faculty of Medicine and Health Sciences and Universiteit Hasselt
Statistical methods are widely used in the medical field to properly assess underlying relations between quantities, or to evaluate the efficacy of pharmaceutical interventions. In this dissertation, we focused on how statistics can be applied throughout the process of conducting a health economic evaluation and how it affects the data analysis of data derived from an antibody-based assay.
When performing a health economic evaluation, the first step is to identify all relevant information, through a systematic search, and combining them using a suitable statistical method. We conducted a meta-analysis using a random intercept logistic regression model, to properly estimate the case fatality risk of enteric fever in Gavi-eligible countries. Another important phase in health economic evaluations is sensitivity analysis in which the effect of varying the values of input parameters, e.g. the case fatality risk mentioned above, on the model outcomes is recorded. A special case of sensitivity analysis is threshold analysis, which is used to determine the threshold value of an input parameter at which a health care strategy becomes cost-effective. Current methods resulted in either incorrect estimates or were time consuming. Therefore, we proposed a statistical method called generalized additive models that overcomes the aforementioned issues. Lastly, we combined existing statistical methods, as well as the one previously described, to perform a health economic evaluation for vaccination against herpes zoster in 50- to 85-year-old Belgian cohorts.
The second part of the dissertation focusses on setting up a general framework for the simulation of data derived from an enzyme-linked immunosorbent assay, also known as ELISA. We generated a standard curve in silico, which describes the relationship between the concentration and the absorbance of interleukin 6, to determine its concentration in an unknown sample. We implemented the current data analysis adopted in a research setting to describe potential variations in the results.