INTERPRETING SURVIVAL DIFFERENCES AND TRENDS

Since 1990 a concerted action between European population-based cancer registries (the EUROCARE project) has been carried out with the aims of establishing whether there are differences in cancer patient surviv al in Europe, and the reasons for such differences. Survival differenc es actually exist for cancer sites for which the stage of disease at d iagnosis is the major prognostic factor (such as breast, stomach and c olon cancer). However, for most cancer sites, survival increases over time and the survival rates of different countries tend to converge to wards higher values. Interpreting survival differences and trends is n ot an easy task. Longer survival may be achieved by postponing death t hrough better treatment or by anticipating diagnosis. However, an earl ier diagnosis may or may not make a treatment more effective in postpo ning death. The computation of stage-specific or stage-adjusted surviv al is not sufficient for interpretation of survival differences, becau se staging procedures change over time and may vary in different hospi tals and countries. In addition to an early diagnosis and more effecti ve treatment, a number of factors may bias survival estimates. They ma y be classified into factors that can be controlled in the analysis (a t least partially), such as mortality from other causes, demographic f actors, epoch of diagnosis, different statistical methodology, and fac tors depending on the validity of cancer registry data, such as defini tion of the illness, exhaustiveness and quality of registration, compl eteness of follow-up, definition of the date of diagnosis, and definit ion of disease stage including the diagnostic procedure used to establ ish stage. To help disentangle the effects of early diagnosis and bett er treatment, several statistical approaches are being developed: mult ivariate analysis on relative survival data, new modeling analysis to separately estimate the proportion of cured patients and the length of survival for those patients destined to die, and the standardized col lection of information on stage at diagnosis and staging procedures.