Net survival: comparison between relative and cause-specific survival estimates.

This paper presents a comparison between cause-specific survival and relative survival calculated according to the Hakulinen method. The data are from the Tuscany Cancer Registry and we analysed the following cancer sites: stomach, colon, rectum, cervix and corpus uteri, female breast and skin melanoma. When cause-specific survival was calculated including among the deceased only those subjects whose death certificate reported strictly the same neoplasm as cause of death (three digits ICD-9), very good agreement between the two methods was found for stomach, lung, and skin melanoma and a good agreement was found for female breast. For cancer of the colon and rectum, in order to obtain similar estimates, deaths coded as caused by cancer in other sites of the large bowel and partially the not specified intestine had to be included. For cancer of the cervix and corpus uteri, in order to obtain similar estimates, but with cause-specific remaining higher than the relative estimates, deaths coded as not specified uteri cancer had to be included. In conclusion, cause-specific survival may be used also with overall mortality data, providing the death causes by cancers most susceptible to misclassification are to be included. However, using cause-specific survival creates some difficulties in comparative studies due to the different reliability of mortality data and to the different criteria used to identify the main cause of death.