Projection of cancer incidence: experiences and some results in Finland.

The prediction of the effects of known or postulated intervention is an important administrative (and sometimes scientific) problem, which can be successfully handled by the presently available methodological means. The long-term prediction of cancer risk is a challenging problem without a general solution. Direct extrapolation of past trends of cancer incidence or mortality may lead to grossly erroneous results. The long latent period from the beginning of exposure to the diagnosis of cancer can be utilized in making predictions. The risk factors used as auxiliary variables have two prerequisites: (1) they should be preponderant enough and (2) they should not be subject to unpredictable intervention during the period of prediction. Some primary sites can be thought to meet these general conditions, but it is unlikely that overall cancer risk or risk of cancer at all specific primary sites could be predicted in this way.