Cement, cancers and clusters: an investigation of a claim of a local excess cancer risk related to a cement works.

Abstract

We investigated claims by a campaigning group of a cancer cluster associated with a local cement works. To investigate cancer rates in the town we defined the study area as the Census wards matching the geographical area code supplied to the campaigning group. Standard methods were applied to registered cases of cancer for the area for the years 1974-1989 to derive observed and expected numbers. The significance of the relative risk was assessed using the Poisson distribution. By selecting a different denominator population we attempted to reproduce the results of the campaign group. Cancer rates around the cement works were investigated for four cancer types plausibly associated with emissions, using cancer registrations for the years 1985-1994. Cases were mapped to 1981 Census ward boundaries, and the same statistical methods were used, but expected counts were also adjusted for deprivation. Rates were calculated for an inner 2 km zone and outer zone 2-5 km from the works. Relative risk was calculated and the ratio of risks was examined for evidence of increased risk closer to the works. Relative risks were not significantly elevated either in the town or around the cement works itself. We were able to reproduce the likely errors that resulted in the elevated relative risks for five cancer groups claimed by the campaigning group. We found no evidence of increased incidence of cancer around the cement works. Incorrect handling of cancer registration data can result in spurious cancer clusters and unnecessary public alarm.