THE constancy of the sex ratio is a remarkable feature of the epidemiology of gastric cancer. Doll (1956) noted that female death rates in 13 countries ranged from 50-67 per cent of the corresponding rates for males. Haenszel (1958) found that the sex ratios in the 119 economic subregions of the United States for showed no significant departures from the national average. In the 11 areal aggregates of England and Wales mortality of males and females was highly correlated, as was that of males with that of married women by husbands' occupations in the main occupational groups, in the period around the 1951 census (Griffith, 1963). In many different comparisons therefore males and females appear to be subject to common influences so far as gastric cancer is concerned.Data from the Inter American Investigation of Mortality (Puffer and Griffith, 1967) also showed this constancy of the sex ratio. The mean of the ratios of the age-adjusted death rate of males and females in 12 cities was 1 64 and only in one city did the ratio differ significantly from this figure. Closer study revealed, however, that the value of the ratio in the combined material of all 12 cities differed by age over the 60-year age span investigated. Variations in the sex ratio with age had also been noted by Gordon, Crittenden and Haenszel (1961) Kurihara, 1962Kurihara, , 1964Kurihara, , 1966 give deaths by age, sex and site of cancer for 25 population groups in 24 countries for the 6 years 1958-63. The deaths by 5-year age groups from 25 through 84 years of males and of females from cancer of the stomach in each population during the 6 years were aggregated.

IN a previous commumcation, cancer deaths in A-nglesey during the ten-year period [1943][1944][1945][1946][1947][1948][1949][1950][1951][1952] were examined with respect to the nature of the soil on which stood the house recorded on the death certificate as the ordinary residence of the deceased person (Davies and Wynne Griffith, 1954). It was show-n that, relative to the generality of cancer, cancer of the stomach and cancer of the breast appeared to be associated with certain groups of soil series. The effect of social class on this association was also considered.The investigation suffered from the drawback that the comparisons were made between cancer of one body site and cancer as a whole and left uncertain whether or not cancer generally is associated in any way with any particular soil group. Until that is known, no inferences can be drawn as to the reasons for the relationship between soil and location of cancer in the stomach or breast.It is not practicable to relate the mortality from cancer of any particular site to the population at risk on the different soil groups, since the patchwork of soils in the county bears no relationship to administrative areas for which data relating to the age, sex, and social class of the inhabitants might be available. Further, the results of any samphng procedure applied to the present population of the county might not indicate the population at risk of ten years ago. For these reasons a sampling method was adopted which assumed that deaths from all causes other than cancer in an administrative area would be distributed relative to soil group in the same way as the population of corresponding sex, age and occupation (i.e. that the specific non-cancer death-rates do not vary with soil group.)The distribution of the population as thus deduced from the sample is then compared with that of the deaths from cancer of the stomach and from cancer of the breast. METHOD OF SAMPLING.For all deaths of persons normally resident in the county, the place of residence, age, sex, occupation, cause and date of death is known. From this " universe " an attempt was made to select a death from some cause other than cancer which matched a death from cancer of the stomach or of the breast with respect to age, sex, month of death and social class, thus eliminating the possible disturbing effect of these variables. The appropriate grouping of the soil on which stood the normal residence of the deceased was then obtained in the way described in the previous paper (Davies and Wynne Griffith, 1954). The rules adopted in drawing the 4 Ccontrol " sample were:

No abstract