Lung cancer is one of the most common and deadliest cancers, as it is often diagnosed at an advanced stage. In lung cancer screening, individuals at risk are offered health check-ups using state-of-the-art medical imaging. This enables the detection of lung cancer at an early stage, which improves the possibility of undergoing curative treatment. Clinical trials have shown that lung cancer screening can greatly reduce the number of deaths. But which individuals benefit the most from screening?
New insights through advanced modelling
Advanced statistical models and machine learning methods allow us to understand the different reasons why outcomes differ between various groups of individuals. Recently, these methods have been successfully applied to evaluate the different kinds of personalised medical treatments each individual receives.
With colleagues from Erasmus MC, this led us to ask the question: ‘What insights can we obtain by applying these techniques to lung cancer screening?’ We analysed two large randomised clinical trials involving people with an extensive history of smoking. This showed us that screening is particularly effective in former smokers and less heavy smokers compared to current smokers or heavy smokers.
The relevant next question we asked is: ‘How can we explain this finding?’ To answer this, we included the form of lung cancer that was detected in the analysis. While this information is not available at the time of screening, incorporating it into our models afterwards allowed us to reveal the driving mechanism.
Former smokers or less heavy smokers tend to develop forms of lung cancer that are less aggressive, resulting in better treatment options. Moreover, these less aggressive types often develop in locations where they are easier to detect. In short, lung cancer screening is most effective for former smokers or less heavy smokers, because they develop better treatable forms of cancer.
This insight would not have been possible with traditional analyses that are still common in medical research. In these traditional analyses, separate comparisons are made of, for example, former smokers to current smokers, or among groups with different smoking intensity. Only by including all the different factors in advanced models, including the detected form of lung cancer, could we draw this conclusion. That is how econometric models contributed to medical research.
It is never too late to quit: lung cancer screening is particularly effective in former smokers
Convergent expertise leading to impactful research
To succeed, our research needed the combination of medical knowledge of Erasmus MC and modelling expertise of the Department of Econometrics at Erasmus School of Economics. We also obtained an Open Mind grant from the Convergence Alliance of Erasmus University Rotterdam, Erasmus MC, and TU Delft, which aims to transcend boundaries between institutions and disciplines. To our delight, it was reported on by mainstream media. Finally, this project demonstrates the wide applicability of the skills taught in Erasmus School of Economics’ econometrics programmes, in this particular example, to biomedicine.
It still matters to quit
Some long-term smokers think that the damage in terms of cancer risk has already been done, so they may as well keep smoking. But our findings suggest that this is a myth. By quitting smoking, individuals can not only lower their overall risk of developing lung cancer but also lower their risk of developing one of the more aggressive types of lung cancer.
About Andreas Alfons
Andreas Alfons is an associate professor at the Department of Econometrics at Erasmus School of Economics and the Academic Director of the Master Programme Data Science for Econometrics. His expertise lies in the intersection of statistics and machine learning. He is particularly interested in applications in the behavioural and medical sciences.
About Max Welz
Max Welz obtained his PhD at the Department of Econometrics at Erasmus School of Economics. He is now a Postdoctoral Researcher at the University of Zurich. His current research is focused on methods for robustly analysing high-dimensional questionnaire data, for which he uses modern techniques from statistics and machine learning.
- More information
This item is part of Backbone Magazine 2026 jaartal. The magazine can be found in E-building or around campus for free. Additionally, a digital copy is available here.
Backbone is the corporate magazine of Erasmus School of Economics, published annually since 2014. The magazine highlights successful and interesting alumni, covers the latest economic trends and research, and reports on news, events, student and alumni accomplishments.
