The retrospective study of cancer incidence around the Gilly-sur-Isère municipal waste incineration plant—initiated amid a crisis following its closure in late 2001—aimed to determine whether there was an excess of cancer cases in the vicinity of the facility. It involved compiling as comprehensive a list as possible of cancers occurring between 1994 and 2002 in the study area (exposed to atmospheric emissions from the incinerator) in order to compare the observed incidence with that established by French cancer registries. This report describes the main challenges encountered and the solutions implemented during the cancer case survey. Medical data were collected from multiple sources: anatomical cytopathology and hematology laboratories, medical information departments (DIM), and specialized departments in hospitals and clinics, primary health insurance funds (CPAM), private practitioners, and specialized cancer registries. The 59 computer files collected were processed: searching for missing addresses (using administrative data from the admissions offices of healthcare facilities and the National Health Insurance Fund), selecting patients from the study area, standardizing cancer codes, merging the files into a single database, analyzing the available information on each cancer, and deduplicating records. The majority of cancers were then validated by consulting medical records to eliminate false positives (metastases from a known primary cancer, recurrences). Of the 2,360 cancers that underwent validation, 34% were excluded; ultimately, the database included 2,055 cancer cases. Three types of sources alone accounted for 94% of cases: laboratories, hospital medical records (DIM), and the CPAM. With an average of 3.2 sources per case and a 91% rate of cases with histological confirmation, the database could be considered to have the same comprehensiveness as a registry. The use of administrative data and the review of medical records proved necessary given the uncertainties regarding patients’ addresses at the time of diagnosis, coding errors for cancers in some of the collected files, and difficulties in identifying false positives. In conclusion, a retrospective cancer incidence study is feasible at the local level provided that significant human resources are allocated (2.5 full-time equivalents over three years to identify 2,055 cases). Recommendations and practical advice for conducting such a study are provided at the end of the report. (R.A.)