From CentraleSupélec to Cambridge: An Engineering Student at the Intersection of Mathematics and Medicine

You completed your first two years at CentraleSupélec before going to Cambridge. How did your interest in applied mathematics in healthcare develop?

Image

I completed the first two years of the engineering program at CentraleSupélec before taking a full gap year (one year at Cambridge studying two different subjects) dedicated to research. The school’s broad-based curriculum—particularly through electives and projects—allowed me to explore many fields and confirm my interest in computational and applied mathematics.
This interest was already present during my preparatory classes (CPGE): as part of the TIPEs (Supervised Personal Initiative Projects), I worked on the mathematical modeling of epidermal healing, which was my first experience with research applied to healthcare. At CentraleSupélec, I then worked with Airparif on forecasting air quality in the Île-de-France region, a topic also related to public health. Gradually, I became interested in applying mathematical tools to medical problems.

Why did you choose to spend your gap year doing research at the University of Cambridge, specifically on a project related to cystic fibrosis?

When I started looking for a gap year opportunity, I wanted to delve deeper into biostatistics and their application to medical research. So I applied to several research projects in this field, and the one offered at the University of Cambridge, within the MRC (Medical Research Council) Biostatistics Unit, particularly interested me. The project focused on cystic fibrosis and how changes in various health indicators could provide insights into the progression of the disease and the risk of mortality. It was exactly the intersection of mathematics, statistics, and medicine that I was looking for. So I joined the team to develop and apply new statistical models to data from the UK Cystic Fibrosis Registry.

Specifically, what were you trying to understand using your statistical models?

In people with cystic fibrosis, various indicators are monitored regularly, including FEV1 (Forced Expiratory Volume in One Second) and BMI. 
We usually focus on their average levels or how they change over time. Our idea was to also study their variability over time in a single patient. 
Two patients may thus have comparable average lung function, yet exhibit very different trends: one may be relatively stable, while the other experiences much greater fluctuations. 
We therefore sought to determine whether these fluctuations also provide information about disease progression and the risk of mortality.
The model we developed allows us to simultaneously examine the average levels of FEV1 and BMI, their individual variability, and their association with survival. In our first study, we observed that greater variability in both FEV1 and BMI was associated with a higher risk of mortality. The long-term goal is to achieve a more comprehensive representation of a patient’s condition than one based solely on average values.

You worked with physicians specializing in cystic fibrosis. How was their expertise important to your work?

Clinical expertise became particularly important when we sought to enrich the model with additional medical information. We had numerous variables that could influence the disease, such as certain infections, the F508 genotype, and other clinical factors. Statistically, we could incorporate these variables in several ways, but mathematics alone cannot determine which representation is medically the most relevant. 
So we consulted with specialists to identify the factors to include in our analyses and to better understand their role in lung function, its variability, and survival. They also helped us identify important risk factors that we needed to consider in our analyses. These discussions were essential to avoid building a model that was statistically consistent but disconnected from clinical reality. 

What did this collaboration with physicians teach you about how to integrate scientific and medical approaches?

I was surprised by how easy the dialogue was. The physicians and researchers we spoke with were very familiar with quantitative research, and some even had strong expertise in biostatistics themselves. So I was able to present our model, our results, and our questions quite precisely, while staying focused on their medical significance. I was also accompanied by my advisor, who has extensive experience with these interdisciplinary collaborations. 
This taught me that a good medical model isn’t built solely on data—it must be tested against the knowledge of specialists. 
We saw a very concrete example of this during our analyses. For one of the infections we studied, we found an association with survival that seemed counterintuitive to us (it appeared to reduce mortality), to the point where we wondered if there might be a problem with our model. The doctors then explained to us that interactions between different infectious agents in patients with cystic fibrosis can be complex. In particular, certain bacteria can compete with one another: Pseudomonas aeruginosa (infection) can produce substances capable of inhibiting the growth of Staphylococcus aureus (infection). This exchange showed us that a statistically surprising result can have a biological explanation. The fact that several of the observed associations corresponded to what doctors already knew about the disease reinforced our confidence in the relevance of our model.

Your work led to a publication in *Statistics in Medicine*. What might be the next steps for this research and its applications in medicine?

So far, we have mainly studied the associations between lung function, BMI, their variability, and the risk of mortality. One of the next steps is to move toward prediction: using a patient’s medical history to estimate their future risk and assess whether taking their individual variability into account improves this prediction. 
In the longer term, this information could help better identify at-risk patients. We also hope to study the effect of new cystic fibrosis treatments on this variability. The ultimate goal is for these methods to one day contribute to tools that can be used in clinical practice.

How has your education at CentraleSupélec helped you in this research project?

When I arrived in Cambridge, I wasn’t yet familiar with the very specific statistical models we were using. However, my education at CentraleSupélec had given me the mathematical, statistical, and computational foundations needed to understand them and make rapid progress. I also drew heavily on the work methods I acquired during projects at the school. In particular, the mathematical modeling workshops taught me how to structure my work around a precise roadmap. Before each meeting with my advisor, I would prepare a summary of the results I had obtained, the challenges I had encountered, and the questions that needed further exploration. The numerous projects and presentations at CentraleSupélec had also prepared me to present my findings, both to my advisor and to a broader research team.

Did this experience change your career plans and your view of the role of engineers in the medical field?

Above all, this experience reinforced a conviction that had developed over the course of my studies. I’ve often said that if I hadn’t chosen to become an engineer, I probably would have liked to study medicine. I much prefer engineering, but I discovered that it’s possible to contribute directly to medical research while remaining an engineer and mathematician—something I consider a true privilege. 
My year at Cambridge thus reinforced my desire to pursue biostatistics and, eventually, to earn a Ph.D. in this field. I hope to continue exploring these topics through research and in my future studies. 

Learn more: Palma M., El Makkaoui O. et al., “A Bayesian Location-Scale Joint Model for Time-to-Event and Multivariate Longitudinal Data With Association Based on Within-Individual Variability,” Statistics in Medicine, 2026.