A Spatial and Temporal Analysis of Four Cancers in African Gold Miners from Southern Africa¶
J. S. Harington; N. D. McGlashan; E. Bradshaw; E. W. Geddes; L. R. Purves (1975). A Spatial and Temporal Analysis of Four Cancers in African Gold Miners from Southern Africa. British Journal of Cancer, 31.
| Type | journal article |
| Scope | Wider apartheid literature |
| Method | descriptive quantitative |
| Period studied | 1964–1971 |
| Regions | Mozambique, Transkei, Ciskei, Lesotho, Malawi, Botswana, Natal, Transvaal, Orange Free State, Swaziland, Witwatersrand gold mines |
| Themes | mining migrant labour, demography urbanisation, methods data sources |
| Citation key | harington1975 |
Argument¶
This medical-geography study analyses 1,344 cancers diagnosed among African gold miners over 1964-71 (2.9 million man-years) and shows strong spatial gradients in four cancers (liver, oesophagus, respiratory, bladder) by home area. The authors use Poisson tests of observed against expected case numbers by territory and by local district. They argue that the patterns mirror those in the miners' home populations, so mining work is not the cause and the cancers are already present in occult form on recruitment. Primary liver cancer among Mozambican miners fell sharply over the period, for unknown reasons. The paper is not about economics, but it documents the scale, geography and structure of the migrant mine labour force.
Evidence and method¶
Chamber of Mines health returns and mine hospital records, Liver Cancer Unit records, Liver Cancer Registry (1964-68), compound records, and recruiting organisation data. Man-years of employment by home area (Table I). Comparison with published cancer registries and hospital surveys of resident populations.
Place in the debates¶
- Whether mine employment or the home environment causes cancer in migrant miners (see Living standards, health and social policy). Holds that cancers reflect home-area exposure (diet, childhood, social factors) and that mining has little or nothing to do with them.
Findings¶
-
Scale of the study population and cancer burden, with liver cancer dominant. (p. 1)
1344 cases of cancer of all sites were diagnosed in a population amounting to 2,926,461 man-years worked, giving a crude rate of 46 per 100,000 man-years.
-
Recruitment is organised across southern Africa through rural depots and central depots, with Mozambique the largest source (24.3% of man-years). (p. 2)
Recruiting is carried out by an organization which has depots in all parts of southern Africa, mainly in rural areas of the Transvaal, Natal, Orange Free State, Transkei
-
Migrants are mainly subsistence farmers supplementing their livelihood, on contracts of varied length. (p. 2)
The African miners are chiefly a migratory labour population who come to the mines in order to supplement their livelihood, based mainly on subsistence farming
-
Repeat contracting produces very high turnover. (p. 2)
Regular re-employment on the mines is the pattern of life for the great majority and this results in a labour turnover close to 100% per annum
-
Recruitment age limits are 18-40, but ages are poorly recorded, so only crude rates can be used. (p. 4)
The policy of the Chamber is to recruit able-bodied miners within the general age limit of 18 to 40 years.
-
Liver cancer is concentrated among Mozambican miners, at a much higher crude rate than elsewhere. (p. 5)
The crude rate of primary liver cancer in Mozambique was 68-6/100,000 miners employed, contrasted with 10-1/100,000 for all other areas combined.
-
Oesophageal cancer is concentrated in the Transkei and Ciskei. (p. 7)
79 (49%O) came from the Transkei and a further 30 (19%) from the Ciskei.
-
Within Mozambique, liver cancer rates vary almost ninefold between districts. (p. 7)
with crude rates varying almost 9-fold from the lowest to the highest.
-
The authors conclude mining operations do not cause these cancers, which are brought from home areas. (p. 13)
it seems clear that mining operations have little or nothing to do with the induction of the cancers which appear to exist in an occult form by the time the African reaches the mine.
-
The decline of Mozambican liver cancer rates is unexplained and not attributable to screening or recruitment changes. (p. 13)
A steep drop in the crude incidence rate of primary liver cancer in mine workers from Mozambique was found from 1964 to 1971.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| Chamber of Mines of South Africa monthly health statistics / health returns from mine medical officers (case records 1964-71) | chamber of mines or recruiting records | 1964-71 | Supplies the cancer case counts by site and home area, which are checked and screened for diagnostic proof. |
| Employment (man-years) by home area of African gold miners, Table I | chamber of mines or recruiting records | 1964-71 | Provides the denominator (2,926,461 man-years) for crude rates and for Poisson expected counts. |
| Liver Cancer Unit records (Crown Mines Hospital), Liver Cancer Registry (1964-68) and mine case records | company records or reports | 1964-71 | Used to verify clinical diagnoses of primary liver cancer by biopsy or post-mortem, and to exclude unsupported cases. |
| Geddes (1972) and Geddes & Falkson (1973): differential diagnosis of suspected primary liver cancer | secondary literature compilation | from 1965 | Shows 30% of suspected cases proved to be other conditions, justifying case screening. |
| Robertson, Harington and Bradshaw (1971a) survey of cancer in African gold miners (Oettle survey from 1964 to 1968) | secondary literature compilation | 1964-68 | The earlier survey that the present study extends; used for recruitment description and comparison of cancer proportions. |
| Berman (1935) Malignant Disease in the Bantu | secondary literature compilation | 1925-33 | Baseline cancer proportions (liver 84%, bladder 3.5%, no oesophageal or lung) for temporal comparison. |
| Cancer Incidence in Five Continents Vols I (1966) and II (1970), UICC | international trade or finance data | 1960s | Median ages of cancer cases in resident populations are compared with those of miners (Table III). |
| Unpublished data of Schonland and Bradshaw (1968), Natal cancer | secondary literature compilation | 1964-66 | Comparative median ages of cancer cases in Natal and England (Table III) and support for raised oesophageal cancer in Natal. |
| Purchase and Gonçalves (1971) hospital survey of southern Mozambique | secondary literature compilation | c.1970 | Independent local pattern of liver cancer used to corroborate the miner-based Mozambique gradient. |
| Rose and McGlashan (1975) oesophageal cancer case collection network in the Transkei | secondary literature compilation | c.1970s | Age-standardised resident population data used to corroborate the Transkei oesophageal cancer gradient among miners. |
| Chamber of Mines / Crown Mines compound records and recruiting organisation (WNLA/NRC) medical data | chamber of mines or recruiting records | 1964-71 | Source of case records, compound records and medical data compiled for analysis. |
| Wilson (1972) Labour in the South African Gold Mines 1911-69 | secondary literature compilation | 1911-69 | Cited for the employment of labour on the mines, as background to recruitment. |
| Prates and Torres (1965) cancer survey in Lourenço Marques | secondary literature compilation | 1960s | Cited for Mozambique's world-highest recorded primary liver cancer incidence. |
New data¶
A case-level dataset of 1,344 cancers diagnosed in African gold miners over 1964-71 (2,926,461 man-years), classified by site and by home territory, with local breakdowns for Mozambique concelhos and Transkei magisterial districts. It was compiled from Chamber of Mines health returns, mine case records and Liver Cancer Unit records, with verified diagnoses.
Inference¶
Quantification: Statistical analysis.
Crude incidence rates per 100,000 man-years are compared across home territories. Poisson tests of observed against expected case numbers (Choynowski approach) identify significant spatial gradients at the territory and local scales. Linear regression on annual crude rates tests temporal trends. Findings are corroborated by comparison with resident-population studies, and the author infers a home-environment origin for the cancers.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Primary liver cancer is concentrated among miners from Mozambique, with a crude rate several times that of all other areas, and the excess is statistically significant. (p. 5) | Table IV case counts and man-years by home area from Chamber of Mines returns, with Poisson significance tests. | Systematic data |
| Oesophageal cancer is concentrated among miners from the Transkei and Ciskei, with significantly more cases than expected. (p. 7) | Table IV Poisson tests of observed against expected cases, supported by external studies of resident populations. | Systematic data |
| Local gradients within Mozambique (higher on the eastern coast, lower inland) and within the Transkei (higher in the south-west, lower in Pondoland) match patterns in the resident populations. (p. 11) | Tables V and VI and Figs 3-4 (Poisson tests by concelho and magisterial district), compared with Purchase and Gonçalves (1971) and Rose and McGlashan (1975). | Partial data |
| The crude incidence rate of primary liver cancer among Mozambican miners fell sharply from 1964 to 1971, and the decline is not attributable to screening, diagnosis or recruiting changes. (p. 13) | Table VII annual crude rates, a regression line and a 95% confidence limit, with the explanation of no extraneous factors asserted. | Partial data |
| Mining work has little or nothing to do with causing these cancers, which originate in the home environment before recruitment. (p. 13) | Inferred from the uneven geographical distribution of cases, the early age of recruits and the correspondence with home-area rates; no exposure data are analysed. | Illustrative or anecdotal |
| Bladder cancer rates were highest among Mozambican miners, and respiratory cancer rates were highest among Natal miners. (p. 7) | Table IV crude rates and Poisson tests, based on very small counts (65 bladder and 73 respiratory cases). | Partial data |
| Liver cancer miners are much younger (median 33.6 years) than other cancer groups and than resident populations, suggesting heavier carcinogenic exposure in Mozambique. (p. 4) | Table II and III median ages, based on crude age estimates with errors of 10 or more years. | Illustrative or anecdotal |
The author's caveats¶
The authors acknowledge that miners' ages are unregistered and only crudely estimated (errors of at least 10 years are common), so age-specific rates are impossible and only crude rates are used. Rates cannot be compared with older resident populations. Diagnosis was uncertain for primary liver cancer in the earlier years, and under-reporting of oesophageal cancer is possible. Some miners are repatriated before diagnosis. The Mozambique population at risk in Table V was extrapolated from a sample. They note that deviations about the regression line may not be random.
Gaps in the evidence¶
There is no age-structured denominator, no data on miners' length of service or occupational exposures (silica, dust, chemicals), and no individual histories of diet or smoking. These are needed to separate home-area from workplace causation, and the argument is asserted rather than tested. Repatriated or undiagnosed cases and post-contract mortality are missing, so incidence is likely understated. Data on time spent in mining, linked to health records, existed in Chamber of Mines and recruiting files and could in principle be used now. The explanation of the Mozambique liver cancer decline is unsupported by any environmental or dietary data.
Replicability: Data described but not public.