Statistical Studies of Gastrointestinal Diseases

Abstract

This dissertation comprises three statistical studies in the field of gastrointestinal diseases. Undertaken during the author's tenure as a doctoral student in the Computational Biology Group of Dr. Niko Beerenwinkel at the Department of Biosystems Science and Engineering at ETH Zurich; the presented studies span multiple diseases and approaches. Epidemiological questions in the context of IBD risk and colorectal cancer development are investigated through the extension of mathematical models and application of robust statistical methodology. Outlining the public health relevance of the individual diseases, as well as introducing the context in which the work has been carried out, Chapter 1 provides necessary motivation and background for the following chapters. A brief introduction into the clinical context of inflammatory bowel disease and patient wellbeing is then followed by a brief description of analysis methods utilized in the subsequent clinical research. Finally, Chapter 1 introduces the epidemiology and models historically used to investigate colorectal cancer. In Chapter 2, we make a first investigation into the interplay between smoking-related genetic variants and the effect of smoking on inflammatory bowel disease (IBD) severity. While previous work has focused on the simple effects of environmental or genetic factors on IBD disease course, we expand this approach and discover evidence of a modulating effect between a genetic smoking-risk score and the association of smoking with the expected number of surgeries in Crohn's disease patients. In Chapter 3, we characterize the self-reported wellbeing of the Swiss IBD cohort and investigate potential relationships between individual wellbeing and future disease complications. Chapter 4 presents an extension to the two-type branching process model and applies it to epidemiological data of colorectal cancer (CRC) growths. Utilizing data sources pertaining to precancerous adenomas as well as malignancies, we produce a practically identifiable model which can recapitulate trends of prevalence and average growth size seen in CRC. After performing inference with this model, we provide new evidence for the rates which define adenoma-to-malignancy conversion. Chapter 5 provides us with a conclusion; synthesising the main takeaway of this thesis as well as offering a path forward for new research.