IRSN, Institut de radioprotection et de sûreté nucléaire

Search our site :


Contact us :

En Fr

All our expertise to protect you


Thesis proposals

Extensions of Bayesian profile regression mixture models to estimate the health effects of low-dose radiological co-exposures. Application to nuclear fuel cycle workers.


Themes: ​Mathematic, Computing

Thesis location: Ionizing Radiation Epidemiology Laboratory (LEPID)​ - Fontenay-aux-Roses (92) 

Start: October 2021

Skills required

Master​'s Degree or engineering Degree in Statistics if possible applied to life sciences
Advanced knowledge in computer programming (Python), probabilistic modeling and bayesian statistic
Interest in epidemiology applications/Public Health

Fluency in English

Age limit: 26 years old unless otherwise stated.

Thesis subject

Historically, the majority of environmental health research and risk assessment approaches have focused on assessing the influence of one environmental risk factor on health, considered independently of the influence of others. This is not the situation in "real life" where exposures to multiple environmental risk factors are simultaneous. While some research has documented examples of synergies or antagonisms following joint exposures to different environmental agents, the health effects of mixed exposures remain poorly characterized. This is particularly the case in radiation epidemiology (IR), where, to date, the risks of cancers potentially related to multiple environmental exposures of different nature to IR have been little studied and therefore little known. Thus, the development of radiation protection standards remains mainly based on a single-factor exposure framework. The final objective of this biostatistics thesis will therefore be to improve the characterization of cancer risks associated with radiological and chemical co-exposures based on cohorts of nuclear fuel cycle workers. An objective of individualized risk prediction in a multi-exposed worker will also be aimed at. In a previous thesis, a Bayesian profile regression mixture model (noted BPRM hereafter) was proposed and a specific Monte-Carlo Markov chain algorithm was implemented in the Python language to estimate the risk of death by lung cancer in the French cohort of uranium miners who are chronically exposed to several sources of IR at low doses during their professional activity. Although this thesis has shown that BPRM models are a promising, results-rich, and original approach for exposome research in radiation epidemiology, several methodological limitations have nevertheless been identified, requiring further research. Thus, the methodological objectives of the thesis - aimed at reaching the above finalized objective - will be : a) to improve the Bayesian inference tool proposed so far for the inference of BPRM models based on a sub-model of instantaneous excess risk disease (1st year of thesis); b) to propose new extensions to the class of BPRM models in order to take into account the time dimension (2nd year of thesis) and measurement errors (3rd year of thesis) of radiological and chemical co-exposure data in health risk estimates; c) to analyze the advantages and disadvantages of using BPRM models in the context of taking into account a small number (case study: French cohort of uranium miners) and a larger number (case study: subset of workers in the TRACY U cohort) of exposure variables (two applications treated over the 3 years). 

Send Print

Involved IRSN laboratory

Send your resume, cover letter


Send to a friend

The information you provide in this page are single use only and will not be saved.
* Required fields

Recipient's email:*  

Sign with your name:* 

Type your email address:*   

Add a message :

Do you want to receive a copy of this email?