Retrospective census of cancers between 1994 and 2002 around the municipal solid waste incinerator of Gilly-sur-Isère

The retrospective cancer incidence study carried out around the municipal solid waste incinerator of Gilly-sur-Isère (Savoie, France) was ordered in a context of crisis during its closing in the late 2001. Its purpose was to determine whether or not there was an excessive number of cancers around the incinerator. In the absence of cancer registry in Savoie, this study consisted in counting as exhaustively as possible the cancers that occurred between 1994 and 2002 in the study area, which was exposed to the atmospheric fallouts from the incinerator. Thus, it was planned to compare the observed cancer incidence to the French cancer registries'. This work describes the main difficulties encountered as well as the solutions found during the census of cancer cases; the results of the incidence study are not included. The collection of medical data was carried out thanks to multiple sources of information: pathology and hematology laboratories, hospitals' and clinics' departments of medical information, health insurance funds, liberal practitioners or specialised cancer registries. The collected medical data files were dealt with: looking for the missing addresses, selecting patients from the study area, homogenizing cancers coding, merging files into a single database, analysing available information on each cancer and de-duplicating the database. Most cancers were validated by consulting medical folders so as to exclude the false cases like metastasises of a known primary cancer or recurrences. Two thousand eight hundred and forty-five cancers were initially collected, and 28% of them were excluded because they did not correspond to the case definition (no proof of cancer, diagnosis date before the study period...); the final database was made of 2055 cancer cases. Quality indicators showed that the database could be considered as exhaustive and valid as a registry's. Three types of sources allowed to identify 94% of cases: laboratories, hospitals' departments of medical information and health insurance funds. Using administrative data and consulting medical folders turned out to be necessary considering uncertainties about: the patients' residence at the time of the diagnosis, errors in coding cancers in some databases that were collected and difficulties to identify false cases. This census required very important means.