Background There have been recent suggestions that the majority of cancers are caused by random cell mutations, as opposed to environmental and inheritable factors. We hypothesise that if this is the case, then there would be no geographical patterning to cancer mortality rates at the neighbourhood level. Our paper presents a test of this hypothesis for England and Wales.
Methods Mortality data for England and Wales (2006–2009) are used for Middle Super Output Areas (mean population size 7200). We focus on the 16 most prevalent cancers. Age- and sex-standardised mortality ratios were calculated. The global Moran’s I is used to estimate the spatial autocorrelation of each cause of death, detailing the degree of geographical clustering each displays. We also use the local Moran’s I to explore the patterns of any clustering found.
Results Lung cancer was found to have a strong geographical clustering at the neighbourhood level (I = 0.431, 95% CI = 0.417–0.445). However, few other cancers were found to have any clustering, with I values mostly below 0.1. Only those causes related to poor diet and increased alcohol consumption (associated to social factors) displayed some degree of clustering (stomach cancer I = 0.185, 95% CI = 0.171–0.199; liver cancer I = 0.117, 95% CI = 0.103–0.131). Results were consistent after completing sensitivity analyses.
Discussion Our results suggest that the majority of the most prevalent cancers have somewhat random geographical patterns, supporting our hypothesis. Nevertheless, social and environmental factors were still important for some cancers including lung, stomach and liver cancers. The results have important implications for policy, as if cancer development is mostly random then there is greater need for focusing on early detection and increasing awareness as opposed to tackling environmental risk factors.
- Stochastic Processes