Geographic Patterns of Mortality Rates for Individual Cancers in Kagoshima Prefecture

Abstract

Based on the death records covering a 15-year period 1967 to 1982, the standardized mortality ratios (SMRs) for selected sites of cancer were calculated for each of 96 local community units (14 cities, 73 towns and 9 villages) in Kagoshima prefecture. These local communities characterized by the SMRs for different sites of cancer were classified by the cluster analysis using the statistical package BMDP 2M on MELCOM III computer at the Kagoshima University. Results are as follows: 1) With a criterion of chi-square value 30 in the cluster analysis, the 96 local com munities were classified into four regional groups (Cluster I, II, III and IV), each of which could be regarded as the group of local communities with similar characteristics of SMRs for different sites of cancer. The proportion of the number of local communities on Amami islands to total ones was heighest in Cluster I, while that of cities was highest in Cluster IV. 2) The levels of SMR for esophageal cancer were generally high across the four clusters for both sexes, the highest being in Cluster I. The SMR of this cancer was also related to the amount of sale for alcoholic beverages, suggesting that the high consumption of a strong drink "Shouchu" could provide a clue to the clustering of this cancer in the Amami-islands. 3) The levels of SMR for stomach cancer were generally low throughout the four clusters for both sexes. One of the probable causes responsible for the low mortality of this cancer could be explained for the small percentage of?gSake?hin total alcoholic beverages. 4) For cancer of the lung, the patterns of SMR were associated with the amount of sale for cigarettes among the four clusters for females but not for males with no evidence of increasing risk of this cancer with increasing cigarettes consumption. 5) The levels of SMR for cancer of the female breast were generally low throughout the four clusters, the lowest being in Cluster I. Furthermore, the SMR of this cancer correlated with the number of physicians for population unit and inversely correlated with the percentage proportion of farming population among the four clusters, suggesting a low risk of developing this cancer in the lower socio-economic class. 6) For cancers of the liver, pancreas, uterus and other sites and leukemia, no evidences of suspectable causal association were found between the SMRs of these cancers and the indices for correlation with socio-economic status or personal habits.