The effects of individual and area-level socioeconomic status on mortality from cancer of the head and neck in Belgium, 2001-2011

Background
Several studies have found a significant effect of area characteristics on cancer mortality. However, few studies have focused on specific cancer types. Since cancer consists of a heterogeneous group of diseases, type-specific cancer research might yield more information on cancer etiology, risk factors, and at-risk populations. Previous studies found a negative association between individual socioeconomic status (SES) and head and neck cancer. Possible explanations for higher head and neck cancer mortality among individuals with a low SES include poor oral hygiene, Human Papillomavirus (HPV), poor diet, exposure to certain chemicals, and alcohol consumption. In addition, several studies point to an independent effect of area deprivation on incidence and survival of cancer of the head and neck. Especially individuals with a low SES living in deprived areas were found to have a bad survival. Several mechanisms between area deprivation might underlie this association, such as lack of social support, stress and lower social capital. Areas with high social disorder were found to have higher levels of alcohol consumption, which might result in higher mortality of alcohol-related head and neck cancer in these areas as well. So although a few studies looked into incidence and survival, little study has been done on the effects of area SES on mortality from head and neck cancer.

Study aim
This study aims (i) to determine the spatial pattern of head and neck cancer among Belgian inhabitants aged 40-64 and 65+ years, (ii) to assess the effect of individual and area-level status SES on mortality of head and neck cancer, and (iii) to estimate the association (and thus the interaction) between the individual and area-level SES variables
These aims boil down to the following research questions:
What is the spatial pattern of mortality from head and neck cancer among Belgian men and women aged 40-64 and 65+? (i)
What is the association between individual SES and mortality from cancer of the head and neck? (ii)
What is the association between area-level SES and mortality from cancer of the head and neck, while adjusting for individual SES? (ii)
To what extent does the association with area-level SES differ by individual SES characteristics? (iii)
To what extent is the spatial pattern in mortality from cancer of the head and neck explained by both individual and area-level SES? (i)

Head and neck cancer consists of several subtypes with different etiologies and risk factors. Several cancers of the head and neck (such as larynx cancer) are associated to alcohol consumption, while other subtypes originate from the human papillomavirus. Therefore, groups of head and neck cancers related to alcohol and HPV will be studied more in depth as well, and the spatial pattern of cancer of the head and neck will be compared to that of other causes of death with similar risk factors.

Hypotheses
Based on findings from previous studies, the following hypotheses are constructed:
(i) Mortality from cancer of the head and neck will vary across sub-districts, due to both individual and sub-district SES, as individual and area-level risk factors associated to cancer of the head and neck are likely to be unequally distributed across space.
(ii) Mortality from cancer of the head and neck will be higher among individuals with a low SES, and among more deprived sub-districts.
(iii) Individuals with a low SES living in deprived sub-districts experience the highest mortality from cancer of the head and neck.

Data and methods
Data from the 2001 census are linked to register data on cause-specific mortality and emigration covering the period 2001-2011. The data include the total “de jure” Belgian population. The study population consists of men and women aged 40-64 and 65 years and older at the time of the 2001 census. Head and neck cancer mortality was defined according to the ICD-10 classification, ICD-10 codes C00-C14 and C30-C32. In addition, sub-types of cancer of the head and neck are grouped by risk factor (alcohol consumption and HPV). Individual SES characteristics include education, employment status, housing status and home ownership. At the area level, housing quality, unemployment rate and area level income are included. The sub-district is used as the geographical unit of analysis. Sub-districts are based on administrative districts at NUTS3 level (“arrondissement”) subdivided into urban and rural sub-districts. Large urban agglomerations are further subdivided into the central city and its agglomeration.

In order to get an overview of the spatial pattern of mortality from cancer of the head and neck, the (indirectly) standardized mortality ratio (SMR) for cancer of the head and neck is calculated by sub-district using the age and sex-specific rates of Belgium in 2001-2011 as the standard (i). The effect of individual and area-level SES on cancer of the head and neck is modeled using multilevel Poisson modeling (ii). Cross-level interactions between individual and area-level SES are used to estimate whether the effects of area characteristics differ across certain individuals (iii). The sub-district level variation in head and neck cancer mortality, and the extent to which individual and area-level SES account for this variation, are estimated by the Poisson multilevel models as well (i).




Preliminary results
The SMRs show distinct spatial patterns for cancer of the head and neck, especially among men and women aged 40-64 years. Preliminary analysis shows a negative association between individual SES and mortality from cancer of the head and neck, with the exception of women aged 65+. Individuals who are highly educated, employed and live in high quality owner-occupied housing experience lower mortality rates from head and neck cancer. Based on visual inspection, the spatial pattern of the sub-district level variables housing quality and unemployment rate shows similarities to that of head and neck cancer mortality, especially among men aged 40-64 years. This indicates there might be an association between sub-district SES and cancer mortality of the head and neck. In addition, the spatial pattern of alcohol-related mortality observed in Belgium closely resembles that of the mortality from head and neck cancer, especially for men aged 40-64 years. Alcohol consumption might thus be a possible risk factor.