Incidence des cancers de l'enfant en Bretagne entre 1991 et 2005. [Childhood cancer incidence in French Brittany, 1991-2005.]

International audienceBACKGROUND: The aim of this study was to describe childhood cancer incidence in French Brittany from 1991 to 2005, as well as its temporal and geographical variations. METHODS: Childhood cancer incidence was analyzed from the data from the Brittany child tumor registry. Crude rates, world age standardized rates and cumulative rates were estimated for all cancers and for each diagnosis group. Standardized rates were also estimated for each of the four Brittany districts. Poisson regressions were performed to estimate trends in annual rates and to compare incidence rates between 2000-2005 and 1991-1999 periods, and between districts. RESULTS: Between 1991 and 2005, 1176 incident cancer cases were recorded in children younger than 15. Age standardized incidence was 169.5 cases per million of children per year. The most frequent cancers were leukemia (30%), central nervous system tumors (24%), lymphomas (12%) and neuroblastomas (8%). For the period 1991-2005 in Brittany, the risk of cancer diagnosis among children aged less than 15years was 1/459. No significant increase trend was observed over the 1991-2005 period. There was however a significant or close to significant increase in incidence for lymphomas (RR: 1.38 [95%CI: 1.00-1.93]) and central nervous system tumors (RR: 1.24 [95%CI: 0.99-1.56]) between the 1991-1999 and 2000-2005 periods. A significant decrease trend was observed for renal tumors over the 1991-2005 period (Estimated Annual Percent Change=-7.6%, P=0.02). There was no significant difference of incidence between the four districts of the region. CONCLUSION: For the period studied, childhood cancer incidence in French Brittany was lightly higher than for the whole country. Although increases in lymphomas and central nervous system tumors coincided with increasing use of a standardised data collection system for medical information, the data collected provided no evidence in favor of more complete data collection following implementation of the system