We present the main results of the first population-based cancers survival study gathering all French registry data. Survival data on 205,562 cancer cases diagnosed between 01/01/1989 and 31/12/1997 were analysed. Relative survival was estimated using an excess rate model. The evolution of the excess mortality rate over the follow-up period was graphed. The analysis emphasised the effect of age at diagnosis and its variation with time after diagnosis. For breast and prostate cancers, the age-standardised five-year relative survivals were 84% and 77%, respectively. The corresponding results in men and women were 56% versus 58% for colorectal cancer and 12% versus 16% for lung cancer. For some cancer sites, the excess mortality rate decreased to low values by five years after diagnosis. For most cancer sites, age at diagnosis was a negative prognostic factor but this effect was often limited to the first year after diagnosis. (R.A.)