Cancer has been the primary cause of death in Japan for many years and accurate cancer incidence dataare necessary in order to make plans for cancer control. Although population-based cancer registries are thebest answer, regrettably there are still many regions with low accuracy registries. In an alternative estimation,cancer incidences have been analyzed by age-period-cohort (APC) models, allowing future prediction of cancerincidences in 2004. Considering the unexpectedly rapid aging of the Japanese population after this figure wasreported, it would be worthwhile to examine more recent data. In this study, we therefore projected majorcancer incidences based on the earlier results leaving estimated values for the age and cohort effects. Relating tothe period effect, the most adequate scenario was selected from 12 projection methods. Furthermore, incidenceswhen registration rates varied between 70 and 100% were calculated. As a result, different trend from reportedincidences were observed for liver cancer in males, and trends of registration rates differed by sites. Until stableaccurate registration data become available, it is difficult to judge whether predicted increase is real or only looksso because the registration rate is not 100%. However, it is clearly necessary to continuously observe variationin cancer incidences.