Published: 03-09-2025 10:59 | Updated: 03-09-2025 10:59

Thesis on survival extrapolation and its application on health technology assessment

A new thesis from Karolinska Institutet focuses on advanced statistical methods to better predict long-term patient outcomes, supporting healthcare decision-making. The application focused on chronic myeloid leukemia with the goal to inform clinical and policy insights.

Portrait of the PhD student Enoch Yi-Tung Chen
Enoch Yi-Tung Chen Photo: Gunilla Sonnebring

PhD student Enoch Yi-Tung Chen, Department of Medical Epidemiology and Biostatistics, has worked on how to make better predictions about how many years patients will live in order to avoid unrealistic survival prediction. 

What are the most important results in your thesis?

“When researchers try to predict how long patients with a particular disease are likely to live, one useful approach is to think of the risk of dying as coming from two separate sources: first, the baseline risk of dying that comes from being human (the general population mortality rate), and secondly, the extra risk of dying from the disease itself. This approach is known as ”relative survival extrapolation”.

Using data from the Swedish Cancer Register, we found that modelling the baseline and extra risks of dying separately provides more accurate survival predictions. We also worked with the extension of relative survival extrapolation into a multistate framework.

We applied this multistate approach to examine the quality-adjusted life years (QALYs) of patients with chronic myeloid leukaemia (CML), a blood cancer that can be managed well with tyrosine kinase inhibitors. We saw that while CML patients lose very few life years compared to the general population, on average they still lose more QALYs, indicating that their quality of life could be improved. Finally, we examined the economic impact of CML on Sweden’s healthcare system in light of these updated models. The number of people living with CML is expected to nearly double between 2015 and 2030, but treatment costs are declining, and this is expected to help reduce the overall economic burden.”

Why did you become interested in this topic?

"It was during my master’s studies in epidemiology at KI that I realised the importance of considering both health risks and resource use when informing policy. This insight led me to focus on developing statistical methods for health economic modelling."

What do you think should be done in future research?

"I believe future research should focus on improving survival extrapolation in health technology assessment, including advancing methodological innovations such as Bayesian and machine learning approaches, refining multistate modelling, and better utilising population-based registers. Greater emphasis on measuring health-related quality of life and promoting transparency in modelling will enhance both accuracy and credibility. Sustained integration of research across academia, industry, and regulators is vital to ensure that methodological progress translates into reliable, patient-centred decision making.”