The geography of survival from colo-rectal cancer.

This study investigates variations in survival following surgery for colo-rectal cancer in the Wessex region (part of southern England), using 5147 cases diagnosed between 1 September 1991 and 31 August 1995. Survival curve estimation by life tables and Cox's proportional hazards model were used to examine geographical variation in cancer survival, with a specific focus on distance between place of residence and treatment centre, and district of treatment. We also consider whether area deprivation has an impact on survival. In seeking to answer these questions we control for possible confounders, including: age, gender, site of tumour, stage of disease at operation, hospital size and surgery type (whether elective or non-elective). District of treatment, distance and deprivation all show a relationship to outcome using survival curves, but when adjusting for other covariates using the Cox model, and considering deaths from all causes, only district of treatment was a very significant covariate (p<0.0001). Distance, deprivation, and gender were only weakly significant (p<0.10). Considering only deaths related to operation (within 30 days) district of treatment remained significant, but while distance had some effect on outcome, deprivation and gender ceased to be significant covariates. There is some evidence that those who live furthest from centres of treatment have the worst outcomes but the ‘geography of survival’ manifests itself more through where patients are treated than through area (deprivation) effects or relative location. The results have important policy implications, as they show variations among treatment centres having controlled for potentially confounding factors.