Zoek in de site...

Seminar: "Cohort heterogeneity and competing risks in survival analysis" (Lecture)

Date
Friday 5 April 2019Add to my calendar
Time
11:00 to
Location
UMC Study center, Galenus room, route 94
Speaker
prof. Ton Coolen (King's college London, UK)
Description

Ton Coolen apr2019Modern medicine presents us with data of unprecedented complexity and dimensionality. Yet most mathematical and statistical methods used routinely to analyse such data date from the 1970s. Uncritical use of simple tools in complex situations may not only cause important signals to be missed, but can also lead to erroneous conclusions - thereby causing patient harm. There is a clear need for innovation. My aim is to illustrate the potential of some more recent and advanced mathematical and statistical methods, that are designed to meet the needs of post-genome medicine, using applications to cancer data. These include epidemiological data and data from clinical trials. I focus on two specific aspects of modern medical data that conventional survival analysis tools (such as Cox regression and Kaplan-Meier estimators) cannot handle: latent cohort or disease heterogeneity, and informative censoring by competing risks.

About Ton Coolen

Prof. Coolen was trained in theoretical physics in Utrecht, where he obtained his PhD in 1990 on the modelling of neural networks. After brief postdoctoral periods in Utrecht and Nijmegen, he moved to Oxford to spend four years as a postdoctoral researcher, working with David Sherrington on disordered many-variable systems. In 1995 he joined King’s College London, where he created the Disordered Systems group, and became Professor of Applied Mathematics in 2000. At King’s he initiated various teaching and research initiatives to boost the interface between the mathematical, physical and biomedical sciences. In 2012 he created the Institute for Mathematical and Molecular Biomedicine at King’s, which he has been leading since then.

His main research field during the last decade has been the development of novel Bayesian models and mathematical tools for the description of survival/failure processes and outcome prediction in data-driven personalised medicine, dealing specifically with the complexities of modern cancer data, such as latent cohort/disease heterogeneity, informative censoring by competing risks (or comorbidities), and inference from high-dimensional covariate streams. This work is done in close collaboration with various clinical and commercial partners, in the UK (he is honorary professor at UCL’s Cancer Institute) and abroad (mainly in the EU, Japan, and the USA).