Computation of cancer incidence rates for defined small geographical areas--matching of numerator and denominator.

The authors present the problem of validity of numerator and pertinence of denominator for small geographical areas on sample of average annual crude incidence rates of stomach cancer in Slovenia during the period 1968--1974. At census in 1971 this country counted 1 727 137 of population. According to the residence of the patients the rates have been calculated for 60 municipalities, and besides for 29 so called "epidemiological regions", defined by their geographical and ethnic characteristics. Although in both instances great differences in incidence rates have been found, only a few areal units exhibited a significantly lower or higher rate from the national average. Specially areas with small population base "disappeared" in the average, although the value of the rate for them was rather low or high. A formula is presented by which at given rate for total country it is possible to determine the minimal size of population which still could provide estimation of statistically significant lower rate. Thus, for areas with small population a period of observation through many years is required in order to obtain the necessary minimal population base. The question is pointed out, whether this approach is reasonable as to the epidemiological interest, considering that some factors related to cancer may change during the long period of time. On the other hand, when merging small areas into larger ones, attention must be paid that these are not too heterogeneous as to their cancer incidence rate, geographical, ethnic, socio-economic and life-style features.