Introduction Generalize Linear Latent and Mixed Models are scarcely used in cancer epidemiology, having been basically used multilevel and generalised linear and mixed models. Using flexible models allow including random effects, common factors coupled to a multilevel structure for unobserved heterogeneity. Cancer is the main cause of death worldwide. In Córdoba, (Argentina) bladder cancer is the fourth most incident cancer among men and ninth in the overall population. Previous studies have suggested that fruits, lean meats, some cereals and cereal products, and vegetable oils would prevent against these tumours, while some fatty meats and use of sweeteners, may increase the risk. In the present work new methodological strategies are used in order to explore the dietary influence on the disease outcome.

Objective To define some possible promoting dimensions related to meat intake and combine with a disease model including some bio-socio-cultural characteristics.

Methods A case-control study, conducted in Córdoba, including 221/472 cases/controls is used. Subjects were interviewed using a validated FFQ containing biological, educational and lifestyle characteristics, and 127 food items. A two steps model was proposed: common factor modelling as confirmatory factor analysis to explore the dimensionality of constructs from the diet information; and a disease model, which arises from the composition of exposure and measurement models.

Results and Conclusion Two constructs were identified, a promoting and a protective one. The direct and indirect covariates effects were also estimated as risk. This work improves the understanding about the diet-cancer relationship.