Is there a geography to female breast cancer mortality rates in England and Wales...?

Will Dear and Kelvyn Jones, University of Bristol

(Project no. 30043)

We are interested in female breast cancer, its relationship to class, and its increasing prevalence in developed countries and wish to look at this from a spatial perspective. We will explore whether there are area effects (the 'geography') contributing to breast cancer patterns, once socio-economic factors have been taken into account. This study will examine England and Wales only in accordance with the restrictions of the LS. We will investigate whether there is a geography to mortality rates of female breast cancer, the nature of the geography, and whether this geography has changed over time. We will be looking at mortality over the period 1971-2002 and comparing 1971-1991 with 1991 to 2002. The geography will in fact consist of two separate analyses for which two separate tables are requested: the formal geography of county districts and the functional geography of Craig-Webber areas.

Kelvyn Jones has carried out similar research (Ecob & Jones 1998: 2055-2066), exploring whether there was a geography to mortality rates across England and Wales, using LS data. We anticipate highly interesting results since there appears to be little research into whether there is a geography of breast cancer rates.

We believe that the LS can provide all the data necessary for the study, and is a unique source of continuous data linkable to the required individuals over a considerable period of time. We would wish to receive the data in cross-tabulated format, with separate tables, one for the formal geography (Counties or County Districts) and one for the functional geography of Craig-Webber areas. The tables would include age of the individual, a socio-economic variable, and crucially the afore-mentioned geographical variable. We do not expect to use geography below County District Level and would accept aggregation up to County level if necessary. All the data requested, with the exception of mortality, will refer to the baseline of 1971.

Extensive statistical analysis will explore whether there is 'a geography' to breast cancer mortality rates in England and Wales. For the proposed analysis we will use the multinomial regression multilevel model to relate the outcome of alive, dead from breast cancer, dead from other to class period and age at baseline. It is planned that the model is two level with 'cells' (representing people of similar characteristics) at level 1 nested in 'places' at level 2. I fully understand that breast cancer is a relatively rare event but expect that 30 years of records and precision-weighted estimation will result in reliable results. This project is seen very much as an initial pilot which if successful will lead to further research by Kelvyn Jones on this topic.