Epidemiology and public health   The principal aim of this thesis will describe the care pathway of diffuse large B cell lymphoma (DBCL) and follicular lymphoma (FL) in the general population and identify the factors predicting the place of care and survival in a population based cohort of non Hodgkin lymphoma (NHL) patients from three counties in France between 2002 and 2008. The specific aims will 1/ identify the socioeconomic and medical factors predicting the place of care (teaching vs. non teaching hospitals) and 5 years survival, 2/ identify the socioeconomic and medical factors predicting longest delays of care and to study the influence of longest delays on survival 3/ analyse all these factors on a frailty population defined by a high age and/or other criteria of vulnerability (such as comorbidity). To answer these questions, the followings methods are proposed:    1/ case registration of all identified NHL (DLBCL and FL) by three haematological malignancies registries in France (Côte d’Or, Basse Normandie and Gironde). 1977 patients are included. Individual medical data (e.g. patient characteristics, management and treatment) from NHL incident cases diagnosed between 2002 and 2008 was stored in a standardized database. 2/ we collected and geocoded the patient’s place of residence at diagnosis. Each patient was affected to an IRIS (smallest area allowing for aggregate statistics) using a geographical information system. 3/ we applied successively Townsend deprivation index and then the new French deprivation index available in 2012 in univariate and multivariate analyses. 4/ we used adjusted multilevel logistic regression models to take into account the existence of aggregate data. The overall survival probability at 1, 3 and 5 years were estimated by the Kaplan Meier method and then Cox regression model for multivariate analysis. Net survival (Pohar Perme) and Relative survival (Esteve’s method) was also being performed. Finally, to palliate to a non-negligible proportion of missing values on important prognostic factors, we used Multiple imputation by Chained Equation (MICE) method.  We will present results of analysis of the influence of socio-geographical, medical determinants and treatment provider on relative survival of DBCL patients. Cette présentation a été donnée dans le cadre du BRIO SIRIC scientific day 3 organisé annuellement par le SIRIC BRIO et qui a pour but de réunir tous les acteurs du SIRIC BRIO et plus largement de la cancérologie à Bordeaux.
Mot(s) clés libre(s) : cancer, Oncologie, Thérapie, Recherche