Summary.-The 5-year study of cancer in black gold miners, 1964-68, previously reported (Robertson et al., 1971) has now been extended for a separate 8-year period, 1972-79. This allows analyses of all cancers together and of 6 less common sites of cancer severally: lymphosarcomas, colon and rectum, leukaemia, stomach, pancreas and buccal cavity and also of those too rare to classify. The malignancies are considered by territory of origin of the gold miners. Lesotho miners have significantly fewer (P < 0.05) tumours of the lymphatic and haemopoietic tissues and Natal miners have the highest incidence rates for 5 of the 6 sites (excluding leukaemia). A simple grouping method is applied to determine which of the 11 sites of cancer in the miners have similar distributions in their 10 territories of origin. The aetiological implications of clusters over space of certain sites of cancer are discussed. Finally, temporal change over the years 1964-79 shows a significant decrease overall (P <0.01) in cases of lymphosarcomas and colo-rectal cancers and an increase (P<0.05) in stomach cancer. The rare tumour, Kaposi's sarcoma, has also decreased significantly between the 2 periods studied.ANALYSES published by Robertson et al., (1971), Harington et al. (1975 and Bradshaw et al. (1982) have considered the spatial and temporal patterns of the 4 most common sites of cancer recorded among black gold miners recruited from homes in, respectively, 11 and 10 territories in 2 consecutive 8-year periods, 1964-1971 and 1972-1979 (see Fig. 1 for total gold-mine labour force from each home territory).
METHODSThe methods of analysis and the limitations of the data-base, especially with regard to both age of recruitment and age of death, have been reported previously (Harington et al., 1975). Briefly, it should be stressed that recruiting was of fully medicallyexamined men of apparent age between 18 and 40, but in the absence of birth-certificate records, actual age was neither known nor recorded by the mines of employment. Although an estimated age of each death was recorded, it was not possible to calculate age-specific mortality rates. We have previously suggested that the crude death rates were probably equivalent to an agespecific rate for the age group 25-35 years (Harington et al., 1975). Diagnoses were made in the well equipped hospitals of the major mining groups, a high proportion being confirmed histologically in the laboratories of the South African Institute for Medical Research. The 4 most common sites (in rank order, primary liver, oesophagus, respiratory system and bladder) comprised 75*1 % of all cancers in the earlier period and 79-1% in the later period.The purpose of this paper is to examine the remaining data which refer to less common sites of cancer. These are sites of which, in an 8-year period, inadequate numbers of cases occurred for analysis but for which useful information emerges in a longer period of record. Unfortunately data are now extant t Correspondence: Dr J. S. Haringtoin.