Time trends of childhood cancer mortality rates: a report from the Childhood Cancer Registry of Piedmont, Italy, 1971-1998.

Time changes in mortality rates can be used to suggest future trends in the prognosis of childhood cancer. Estimates provided from recent studies led to divergent previsions of future trends. In this study we used data from the population-based Childhood Cancer Registry of Piedmont (CCRP) in order to measure the decrease in mortality. The study included 1,060 deaths in the period 1971-1998. Analyses were carried out using Poisson regression models and considered separately the total tumours, acute lymphoblastic leukaemia (ALL) and tumours of the central nervous system (CNS). We observed a progressive decline in the mortality rates: for all tumours the estimated annual percentage change was -3.6% (P < 0.05), for ALL it was -5.2%, (P < 0.05) and for CNS tumours it was -3.8% (P < 0.05). The statistical test for effect modification due to age at death was not significant. For all cancer types, the Poisson model fits the data well, corresponding to an exponential decrease of mortality rates and leading to the conclusion that current therapies, despite their proven effectiveness, are not sufficient to cure all cases of childhood cancer.