Systematic analysis of mortality trends of cervix and corpus uteri cancers is difficult in Italy, as in many other countries, because of the poor specification of uterine cancer subsites in official death statistics. The aim of this article is to propose a method for the analysis of uterine cancers mortality based on high quality incidence and prevalence data from population-based cancer registries. The method assumes that the excess mortality of cancer patients, compared to death rates expected in the general population, is attributable to the specific cancer. The method is applied to estimate mortality trends for cancers of cervix, corpus and uterus as whole, during the period 1987-1999, in an area covered by 8 Italian cancer registries. Official mortality rates for the 2 subsites were about 60% lower than excess mortality rates, due to the very high proportion of deaths attributed to not specified subsite. Age adjusted cervical cancer excess mortality rates decreased from 3.7 to 2.7 x 100,000 women. Excess mortality for corpus uteri cancer remained approximately stable between 3 and 3.3 x 100,000 women in the period 1990-1999. The results support the efficacy of organized screening in reducing cervical cancer mortality. The same method can be used to assess mortality rates for every cancer entity identifiable in cancer registries data, not otherwise available from official death records.