Development and application of competing risk and multistate models in cancer epidemiology

picture:

in doctoral studies, Nikolaos Skorlis in Department of Medical Epidemiology and Biostatistics She worked with competing risk models and multi-state models in flexible frontier survival models. Among others, he studied various aspects of modeling, such as time scale selection, multi-state structure, and how to introduce constraints in the estimation of various parameters. In addition, try to make it easier to communicate the forms to a wider audience by developing an interactive web tool (MSMplus).

Can you tell us something about the most important findings of your dissertation?

When using competing risk and multi-state models, there are a number of choices that need to be made regarding the structure of the multi-state model and interstate velocity modeling. Choosing time scales, making assumptions about similarities between transitions to different states, considering congruences between each multi-state structure and a set of research questions, and targeting specific scales of clinical significance, taking into account the limitations of each structure and the type of data an individual has is very challenging. Therefore, it is always important to choose appropriate models and perform sensitivity analyzes when applying the model.

Does your research make it easier for other researchers to apply different models and present their findings?

Yes, because a large part of my research focuses on evaluating and exploring different modeling options when applying these statistical models. I hope that other researchers can use this to develop and select multi-state models in combination with appropriate modeling options that can be evaluated either by simulation techniques or sensitivity analysis. Because it is so complex, I was really excited to develop MSMplus so that researchers could use it when presenting their findings.

How do you think research should continue in the future in this area?

I think it is important that future research focus on simultaneous modeling of several time scales in a model with competing risks and conditions. There is increasing interest in using multi-case models in phase II and III cancer clinical trials and an optimized design is needed. Last but not least, it is possible to use complex models with many parameters for complex research questions and to deal with large amounts of data that are now available, especially in Sweden where we have such good records.

thesis

“Development and application of competing risk and multistate models in cancer epidemiology.“

Nikolaos Skorlis. Karolinska Institutet (2023), ISBN: 978-91-8016-934-9