Effects of deprivation on the geographical variability of larynx cancer incidence men,Girona (Spain) 1994–2004 Carme Saurina a,b, * , Marc Saez a,b , Rafael Marcos-Gragera a,c , Maria Anto `nia Barcelo ´ a,b , Gemma Renart a,b , Carmen Martos d,e a Research Group in Statistics, Applied Economics and Health (GRECS), University of Girona, Spain 1 b CIBER of Epidemiology and Public Health (CIBERESP), Spain c Epidemiology Unit and Girona Cancer Registry (UERCG), Oncology Director Plan, Health Department, Girona,Spain d Direccio ´ n General de Salud Pu ´ blica,Gobierno de Arago ´ n, Zaragoza, Spain e Instituto Aragone ´s de Ciencias de la Salud, Zaragoza, Spain 1. Introduction Our objective in this paper is to assess the association between the incidence of larynx cancer and socioeconomic conditions in the province of Girona from a spatial viewpoint. The question we set out to answer is which part of spatial variability can be accounted for by the geographical variation in the socioeconomic conditions of the province. The standardised incidence ratio (SIR), given that its variance is inversely proportional to expected values,depends heavily on population size [1,2].Moreover,variability in observed cases is usually greater than that expected, leading to what is known as extra-Poisson variability. In fact, it is important to distinguish two sources of extra variation when spatial data are available. First of all, so-called ‘spatialdependence’ is usually the most important source of extra variation and is a result of the correlation of the spatial unit with neighbouring spatial units. Thus,the SIRs from adjoining or nearby areas are more similar than those from areas that are spatially distant.Part of this dependence is not really structural,but is due mainly to the existence of non-controlled variables such as socioeconomic and environmental factors. In relation to the second source, there is also an independent and spatially uncorrelated extra variation called ‘uncorrelated’ or ‘non- spatial heterogeneity’, a consequence of non-observed variables without spatial structure, such as a contaminant focus, that could affect risk[3–5]. Associations have been observed between socioeconomic factors and health response variables,both with mortality [6] and, to a lesser degree, with morbidity. However, on most occasions a causal nature has not been attributed to these associations,because the socioeconomic factors are also in fact related to exposure to environmentalhazards [7], to which a certain aetiological role could indeed be attributed. Examples of these arguments can be found in the analysis of the geographical distribution of cancer incidence. Many studies have introduced socioeconomic conditions into their models insofar as they are believed to be associated with exposures considered to be Cancer Epidemiology 34 (2010) 109–115 A R T I C L E I N F O Article history: Accepted 14 January 2010 Keywords: Larynx cancer incidence Deprivation Extra variation Besag York and Mollie ´ model A B S T R A C T Objective:To assess the association between the incidence of larynx cancer and socioeconomic conditions in the province of Girona from a spatial viewpoint. Materials and methods: Incidence cases of larynx cancer (CL) in 1994–2004 were provided by the Girona Cancer Registry. A census tract (CT) was assigned to all patients. Socioeconomic data were extracted from the 2001 Census. A deprivation index for each CT was obtained by principal component analysis, using four socioeconomic indicators. The standardised incidence ratio (SIR) was calculated using the CL incidence rates in the men of the province of Girona assuming a Poisson distribution. Relative risk was obtained applying the Besag, York and Mollie ´ model. The deprivation index was introduced into the model and was categorised in quartiles. Results: Four hundred and seventy-six incident cases in men were registered. CTs in the lowest deprivation index had a lower risk of larynx cancer, with a risk increase in the higher quartiles. In the highest quartile it wa 1.91 times greater than in the lowest. This association was significant when the whole province was considered. Discussion: The deprivation index explains only part of the geographical variability of CL incidence.Other risk factors without spatial structure may contribute to this explaination. ß 2010 Elsevier Ltd. All rights reserved. * Corresponding author at: Research Group in Statistics, Applied Economics and Health (GRECS), University of Girona,Campus de Montilivi,17071 Girona,Spain. Tel.: +34 972418736; fax: +34 972418032. E-mail address: carme.saurina@udg.edu (C. Saurina). 1 http://www3.udg.edu/fcee/economia/english/grecs.htm. Contents lists available at ScienceDirect Cancer Epidemiology The International Journal of Cancer Epidemiology, Detection,and Prevention j o u r n a l h o m e p a g e : w w w . c a n c e r e p i d e m i o l o g y . n e t 1877-7821/$ – see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.canep.2010.01.006