The basic data for the present atlas are about 10 years old, indicating the slow progress in creating this atlas. Cancer registries and official statistical offices typically need some years to get their annual statistics ready for publication. The requirement of having all cancer cases and deaths exactly located to small-areas further slowed the data preparation phase. In a multinational study like the present one hardly anyhing can be finalised before having the very last data file at hand. Because much new methodological development (mapping techniques, methods of presentation etc.) was included in the preparation of this atlas, some of the work processes took a considerably long time. After each further analytical step new problems with the basic data were revealed, and we had to go back to the national collaborators to get the data corrected. This type of iteration work easily takes several calendar years, if there is a serious aim to obtain high data quality like we had.

The present atlas illustrates geographical cancer patterns from a period just before the big changes in the societies of former socialistic countries. In some years there should be changes in the rates attributable to the development of medical systems in these countries, and later also changes attributable to altered living conditions. When that time comes it is important to have the base situation systematically documented, also in a format of cancer maps. When preparing similar maps in the future, the experience, computer programs and personal contacts built during the preparation of the present atlas may be utilised to speed up the process. One of the tasks before going deep into more analytical approaches would be to really measure the level of data quality and completeness; our relatively good impression of the data quality is mainly based on indirect parameters of data quality (Winkelmann et al. 1998). For certain cancer sites (e.g., nervous system tumours, leukaemia, thyroid cancer) a division into subtypes would give a better basis for aetiological studies; this division may be realistic for cancer incidence rates worked out by specified (research-oriented) cancer registries, but not for official mortality statistics.

During the work of the present map production, an extensive small-area based cancer incidence and mortality data base was created. These data are available for analytical studies, if so agreed with the national providers of the data. It may not add much to the present knowledge to study confirmed associations — such as smoking and lung cancer, or farming and lip cancer — in this ecological setting. More potential topics for further research might be found, e.g., in the dietary factors: the present knowledge does not allow proper explanation of the variation seen for instance in stomach cancer, colorectal cancer or kidney cancer. The extremely low incidence of liver cancer in Norway probably reflects different coding practices, but might the similarly low incidence of gallbladder cancer be true?

The large differences in bladder or testicular cancer rates obviously cannot be explained by the known factors of these diseases. Much of the variation in prostate cancer incidence can be explained by differences in diagnostic activity, but understanding the reason behind the relatively large variation in prostate cancer mortality requires more epidemiological knowledge. What is the role of genetic susceptibility? Might cancer maps even give clues for identifying regional founder effects? Clustering of childhood leukaemia has been linked to postponed infection exposure in certain areas with strong population mixing. Might similar factors, or viruses or bacteria stand behind some of the observed high-rate spots in cancer maps? May some of the spots be attributable to certain age strata only? This is an area with many questions open for further research, and it will depend on the interest and creativity of researchers whether that kind of studies will actually materialise.