https://u-paris.hal.science/hal-01453080Huber, CatherineCatherineHuberMAP5 - UMR 8145 - Mathématiques Appliquées Paris 5 - UPD5 - Université Paris Descartes - Paris 5 - INSMI - Institut National des Sciences Mathématiques et de leurs Interactions - CNRS - Centre National de la Recherche ScientifiqueAcceleration of time to onset of a lung cancer due to occupational exposure to asbestosHAL CCSD2013Threshhold regression model occupational exposure asbestos[MATH.MATH-ST] Mathematics [math]/Statistics [math.ST]Huber, Catherine2017-02-02 15:22:532023-03-24 14:53:032017-02-02 15:22:53enConference papers1AbstractQuality of life is a central concern in medicine. Thus estimation of the number of years of life free of some disease lost due to an occupationalexposure is frequently required. The amount of compensation due to the worker relies upon this estimation. The motivating example is a Frenchcase-control study on the occurrence of lung cancer for workers exposed to asbestos (J.C. Pairon et al, 2009)[4]. For a case-control study the most usual model is the logistic. Several models, though, can be used in order to solve our problem. One of the simplest is the Cox model involvingthe occupational exposure as a covariate at the same level as other risk factors that could also induce lung cancer, like for example family historyof cancer and tobacco consumption. But we can also treat differently the occupational exposure and the other covariates : we adapt to case-control study (A. Chambaz et al. 2013 [1]) the threshold regression (TR) model, also named first hitting time model (FHT), which was initially developed forcohort studies. This model allows us to treat the occupational exposure as an accelerator of the time to onset of the disease, while the other covariatesare divided into two classes, acting differently on the time to onset: the built-in ones (like genetic factors for example) and the lifelong ones (liketobacco consumption for example). The time T of occurrence of the disease is modeled thanks to a stochastic process that represents the amount ofhealth of the subject at time t : the disease occurs when this amount of health hits the boundary 0 for the first time. Considering that, actually, thecovariates are not all of the same kind, we use the first hitting time model.