Tel.: +27-21-9380314, Fax: +27-21-9380310
Cancer incidence in a rural population of South Africa, 1998–2002
Article first published online: 16 FEB 2010
Copyright © 2010 UICC
International Journal of Cancer
Volume 127, Issue 10, pages 2420–2429, 15 November 2010
How to Cite
Somdyala, N. I., Bradshaw, D., Gelderblom, W. C. and Parkin, D. M. (2010), Cancer incidence in a rural population of South Africa, 1998–2002. Int. J. Cancer, 127: 2420–2429. doi: 10.1002/ijc.25246
- Issue published online: 16 FEB 2010
- Article first published online: 16 FEB 2010
- Manuscript Accepted: 22 JAN 2010
- Manuscript Received: 9 JUN 2009
- cancer incidence;
- rural population;
- South Africa;
- population-based registry
Cancer incidence rates and patterns are reported for a rural population, living in the Eastern Cape Province of South Africa for the period 1998–2002. The population-based cancer registry has operated for 20 years, using both active and passive methods for case finding, through collaborations with 19 health facilities: 11 district hospitals, 7 referral hospitals and 1 regional laboratory. The age standardized incidence rates for all cancers were 73.1 per 100,000 in males and 64.1 per 100,000 in females. The leading top 5 cancers for males were oesophagus (32.7 per 100,000), lung (5.8 per 100,000), prostate (4.4 per 100,000), liver (4.4 per 100,000) and larynx (2.5 per 100,000) whereas for females they were cervix (21.7 per 100,000), oesophagus (20.2 per 100,000), breast (7.5 per 100,000), ovary (0.9 per 100,000) and liver (0.9 per 100,000). The incidence of Kaposi sarcoma was low, and higher for males (1.6 per 100,000) than females (0.3 per 100,000). Lung cancer in both males and females was relatively low compared to the high incidence of oesophagus cancer.
Information about disease burden is important for monitoring the health of the nation. In South Africa, a middle income country, such information, including cancer incidence and mortality data, is relatively sparse. A national pathology-based cancer register, established in 1986, has provided limited information on the cancer burden based on voluntary reporting by pathology laboratories of invasive cancers diagnosed by histology, cytology or haematology.1 However, the registry has been unable to provide a complete measure of the incidence of cancers, and has not reported since 1999.2 In a country such as South Africa, with a diversity of cultures and living conditions, population-based cancer registries monitoring the incidence in different settings become very important. The need to strengthen the national cancer register as well as developing population-based registries in a variety of settings has been identified in the context of developing a National Cancer Control Programme.
Over 20 years ago, the Medical Research Council (MRC) of South Africa established a population-based cancer register in 4 magisterial areas of the former Transkei area of the Eastern Cape Province. The original purpose was to monitor geographic and temporal variations in the incidence of oesophageal cancer, which was known to be particularly common in this area.3, 4 The register has developed to include all cancers and in 1998 was extended to include 10 magisterial areas covering the areas of Butterworth, Centane (Kentani), Idutywa, Nqamakwe, and Willowvale in the South West, and Bizana, Flagstaff, Lusikisiki, Port St Johns, and Umzimkulu in the North East. These districts now comprise 7 local municipalities (Fig. 1). The population at the most recent census in 2001 was 1.3 million.5 This register is independent of the pathology based national cancer register which has limited geographic information.
The majority of inhabitants of this area are indigenous (black) Africans who speak isiXhosa and support both Christian and traditional norms and values. The demography in these rural magisterial areas is typical of a developing country with about 43% of the population under 15 years of age.5 The average life expectancy for the province has been estimated to have dropped to 48 years as a result of the AIDS epidemic.6 The population is generally poor and the unemployment rate is 27%.5 Family members seek employment in the urban areas including the gold mines in Gauteng and the Free State Provinces, and coal mines in the KwaZulu-Natal Province. Subsistence farming is widely practiced.
This article reports cancer incidence for the rural population of South Africa, living in 8 magisterial areas of the former Transkei region of the Eastern Cape Province, for the period 1998–2002. The areas include Butterworth, Centane (Kentani), Idutywa, Nqamakwe, and Willowvale in the South West, and Bizana, Flagstaff, Lusikisiki in the North East. The other 2 areas are not reported due to data quality concerns. This study will form a baseline for the newly extended surveillance area.
Materials and Methods
Cancer care facilities
The public healthcare delivery system is based on a network of primary healthcare clinics, healthcare centers, district hospitals and referral hospitals provided by the government. Basic laboratory services are available in most public healthcare centers. Patients suspected to have cancer in the primary and secondary care facilities are mostly referred to Mthatha General Hospital Complex that serves as a central referral hospital for the region, where the regional state laboratory, the Nelson Mandela Pathology Laboratory, is also located. For specialized services including oncology and radiotherapy, patients are referred to hospitals such as Inkosi Albert Luthuli (oncology services since 2003), King George V (with specialist thoracic-surgery), King Edward VIII, and Addington hospitals in Durban, KwaZulu-Natal Province, and Frere (with radiation and oncology services) and Cecilia Makiwane hospitals in East London, Eastern Cape Province. The government provides transport to send cancer patients to these hospitals. However, patients must travel between 200 and 700 kilometers to get these specialized services and care.
South Africa has a private health sector incorporating medical practitioners, hospitals and laboratory services that caters for ∼20% of the South African population who can afford such care or/and have health insurance. The registry does not have access to data from these facilities. However, there are no private hospitals in the area and the proportion of this study population who use private care would be extremely small.
The registry collaborates with all hospitals in the surveillance area as well as the major public sector centers to the north and south of it, to which patients from the area would be expected to be referred. In total, 19 hospitals (11 district, 7 referral and a regional laboratory under the National Health Laboratory Services (NHLS) collaborate with the registry. Both active and passive methods are used for case finding based on the hospital's recorded residential address of cancer patients. The active case finding system involves annual visits to the collaborating hospitals with field trips twice a year. Active case finding extends to all of the referral hospitals outside the registration area described above. Passive case finding is undertaken by part-time oncology nurses who complete specially designed cancer notification forms and send them to the registry on a monthly basis. The nurses were trained in cancer data abstraction by the registry manager. Death certificates are not used as source of information as many deaths occurring in the registry area are not medically certified.
Data were manually abstracted from the records and included demographic variables, tumor characteristics including the site, type and behavior. Cancer sites were manually coded in the registry for topography and morphology according to the third edition of the International Classification of Diseases for Oncology (ICD-O)7 and captured using the latest version of CanReg,8 a customized software computer program designed by the Unit of Descriptive Epidemiology of the International Agency for Research on Cancer (IARC) for cancer registration.
The 2001 census was used to estimate population at risk in 1998–2002. More than 97% of the population are Xhosa speaking black Africans. The age and sex distribution of the population, shown in Figure 2, is typical of a South African rural population. It reflects that the area is a labor reservoir, in which there are more children and older persons, particularly women, than there are working age adults, especially males. Labor migration is historically significant in South Africa. Such a migratory pattern may result in a lower cancer incidence being experienced in the area as it is possible that people from this area that develop a cancer while working in an urban area, do not return to their rural home. On the other hand, some patients diagnosed elsewhere might return home to die and would inflate the incidence rate if they had not been counted in area at the time of the census. However, this is likely to have a limited effect as only those patients who receive treatment at the collaborating health facilities are included in the register. The number of children under 5 years of age is markedly smaller than the next age group. This is likely to be a result of declining fertility on the one hand and under-enumeration of young children on the other.
Data for the period 1998–2002 were cleaned, and a search made for duplicate records based on name, age, sex and diagnosis. Validity checks identified impossible codes and unlikely combinations. Only malignant cases were included in the analysis; cases from outside the registration area were excluded. The number of incident cases is presented by age, sex and site as well as method of diagnosis. Incidence rates were estimated, as well as the registration area as a whole, as age specific, crude and age standardized rates (using World Standard Population).
A total of 2,501 new cancer cases were reported for the period 1998–2002. There were 1,022 (40.9%) males and 1,479 (59.1%) females. The annual number of cases was fairly consistent during this period with an annual average of 500 per annum (204 males and 295 females). The numbers in 1999 and 2000, however, were lower than average (452 and 430, respectively). The average age at diagnosis was 56.2 years with the majority of cases in the age range 50–74 years. A total of 1,172 cases had a laboratory report, of which 981 also had a hospital report. A total of 2,301 cases had a hospital report, with hospitals in the registration area contributing 1,473 cases (64.0%) and the referral hospitals, including the state pathology laboratory, contributing 1,032 cases (44.9%) (204 cases were notified from both sources). Frere Hospital, which is the only radiation and oncology center in the region, accounted for 593 cases. With 1,329 cases notified from hospital records only, 191 from laboratory reports only, and 981 from both sources, the maximum likelihood estimate of the number of cases missed (not reported from either) is 259, indicating a completeness of case ascertainment of 90.6% (95% CI 89.5–91.7%).
Tables 1 and 2 show the number of reported cancers by primary site, age group, and sex, together with the percentage frequency, crude and age standardized incidence rates. Figure 3 shows the ranking of the 10 cancers with the highest age-standardized incidence rates, by sex. In males, the most frequently reported cancers were oesophageal (43.4%, ASR 32.7 per 100,000), lung (7.2%, ASR 5.8 per 100,000), prostate (6.8%, ASR 4.4 per 100,000) and liver (6.1%, ASR 4.4 per 100,000). In females, the most common cancers were cervix (33.2%, ASR 21.7 per 100,0000), oesophageal (32.4%, ASR 20.2 per 100,000) and breast (11.0%, ASR 7.5 per 100,000). There were relatively few Kaposi sarcoma cases, in both males (2.1%, ASR 1.6 per 100,000) and females 0.5%, ASR 0.3 per 100,000). Cases 1,130 (45.2%) had been diagnosed on the basis of clinical information only including X-ray and biochemical/biological tests without histological confirmation whereas 54.8% were cases with histological or cytological confirmed diagnosis. Out of the total of 923 oesophageal cancer cases, 115 (12.5%) had histological confirmation of diagnosis. Of these cases (87.6%) were squamous cell carcinomas and only one case (0.8%) adenocarcinoma; 13 cases (11.5%) had a nonspecific histological diagnosis (e.g., carcinoma NOS).
Figure 4 shows the age specific incidence rates for the most common cancers. The incidence of cancer of the oesophagus increases steadily with age to reach a maximum at age 65–69, although the rates in females are lower at all ages. Lung cancer incidence rates in males increase markedly from age 35–39 reaching a peak at 55–59, whereas in females, the rates are 10-fold lower. Prostate cancer is confined to ages over 50, and rates increase steadily with age. Cervix cancer incidence rates increase to a peak at ages 60–64; in contrast, breast cancer incidence rates are lower and become relatively constant after ages 40–44.
There were 73 childhood cancers in total which accounted for 2.8% of all the cancers reported during the 1998–2002 period. The most common childhood cancers observed were nephroblastoma (15 cases), brain tumours (15 cases), leukemia (14 cases), and retinoblastoma (11 cases) and neuroblastoma (10 cases).
The overall age standardized incidence per 100,000 population for cancer (excluding non-melanoma skin cancer) was 72.4 for males and 63.7 for females respectively. These rates appear low, but there are few data from rural populations in low income countries with which they can be compared. They are similar to the rates reported from Gambia in 1997–989 and rates from the rural population of Barshi in India in 1988–92.9 The common cancers observed during this period were oesophagus, cervix, breast, lung, prostate and liver, and the ranking was largely dominated by oesophageal and cervical cancers. Table 2 compares the rates in the Eastern Cape (this study) with those recorded by the National Cancer Registry (histological diagnosed cases)1 and by the cancer registry of Swaziland in 1996–99.10 It can be seen that the pattern in this area is distinct.
Oesophageal cancer incidence rates for this region are higher than those observed elsewhere in Africa10 or for the Black population in the USA.11 Oesophageal cancer incidence rates for this region have been consistently high for a period of more than 50 years,3, 4, 12–15 despite a low frequency of histological confirmation; diagnosis is generally clinical with confirmation by radiology (barium swallow), since few cases receive any definitive treatment. Oesophageal cancer is related to tobacco smoking and alcohol drinking, and the difference in rates between males and females is consistent with the different prevalence of use of tobacco and alcohol, although neither habit is common in the local population. Results from a household survey conducted during 2002 by the Eastern Cape Department of Health (Bisho, South Africa)16 and Equity Project indicated that in the former Transkei region 16% of men drink alcohol regularly while 13% partake in communal drinking. The prevalence is much lower among women with only 4% drinking regularly and another 4% partaking in communal drinking. This survey also found that 31% of men smoke tobacco and only 5% women, although even in men, daily consumption is modest (3.2 cigarettes per day). Dietary deficiencies and fungal toxins are some of the other risk factors that have been thought to be associated with the development of oesophageal cancer in this region.17, 18 Many early studies were concerned with the geographical differences in incidence within the former Transkei, and a possible correlation with exposure to fumonisins from moldy maize.18 However, an association between exposure to fumonisins and individual risk of oesophageal cancer has not been demonstrated.19
Cervical cancer was the most common cancer among women reported in this region with an ASR of 21.7 per 100,000, about 3 times higher than that of breast cancer. The incidence is, however, lower than that recorded in South Africa as a whole (Table 2).1 Cervix cancer is classified as an AIDS-defining cancer, although the risk of invasive cervical cancer in Africa seems little influenced by HIV, and trends do not reflect changing prevalence of infection.20 Rather, incidence is related to prevalence of infection with Human Papillomavirus (HPV), modified by the protective effect of screening. South Africa has adopted a national policy of offering, a free screen to asymptomatic women aged 30 years, followed by 2 further screens 10 years apart. Reviews of the programme suggest major challenges in implementation, however, even in the more resourced areas of the country.21, 22 While in the long-term vaccination against the HPV will provide effective prevention, implementation will require careful community preparation, and, in any case, secondary prevention through screening urgently needs to be strengthened for the current generation of women. Although studies in South Africa have shown the feasibility of screening using visual inspection with acetic acid (VIA)23 current evidence suggest that the preferred screening option will involve inexpensive tests for HPV.24 The registry has an important role to play in monitoring trends in cervical cancer incidence in the context of the HIV/AIDS epidemic on the one hand and the possible improvements in cervical cancer prevention on the other.
Breast cancer accounted for 11.0% of cancers with ASR of 7.5 per 100,000, somewhat lower than the rate recorded nationally (Table 2). The incidence rates peak as early as age 40–44 and remain at a similar level across older ages whereas in South Africa peak incidence occurs at ages 55 and above.1 This may reflect increasing rates in younger generations of women in the Eastern Cape region.
The estimated completeness of case ascertainment, based on capture–recapture by the 2 sources used (hospital notifications and pathology records), is 90.6%. This estimate depends on the assumption of independence of the 2 sources (that detection by each source is independent of the other).25 It may be something of an overestimate, in fact, if there is positive dependence (more likelihood of being notified from hospital if there is a laboratory report, for example), but we have no means of evaluating any possible bias in the estimate. Despite this reasonably high level of ascertainment from the data sources covered, it is possible that under-diagnosis of cases may be an issue in this rural population, with difficulties in accessing medical services. For example, the lung cancer rates in this region are low when compared with the national figures (Table 2)1 and much lower than rates in other southern African centre of Harare in Zimbabwe (17.4 per 100,000).26 However, low rates have also been reported by registries in West African countries such as Mali (2.7 per 100,000) and Uganda (3.9 per 100,000),27 and may simply be the consequence of low prevalence and intensity of tobacco smoking, rather than a failure of diagnosis or registration. In contrast, the incidence of liver cancer in this region is very low (ASR in males 4.4 per 100,000 and 0.9 per 100,000 in females) compared to national and Swaziland rates (Table 2) and Harare (14.4 and 12.7 per 100,000 in men and women respectively, in 1998–2002).26 Prevalence of hepatitis B infection in South Africa was considered to be endemic prior to the introduction of a vaccination in 1995. Robson and Kirsh reported the prevalence among black South Africans to be around 8%.28 Although there is no population based data, the prevalence was probably also relatively high in the study population. The low liver cancer incidence rates observed in this study may therefore represent under-diagnosis of a cancer for which hospitalization rates may be rather low. Similarly, the high proportion of some cancers diagnosed by histology, such as brain and nervous system, suggests that limited access to noninvasive diagnostic technology may also result in under-diagnosis of cases. Without including information from death certificates, the more lethal cancers may well be under represented.
Kaposi sarcoma accounted for only 2.1% of male cancers and 0.5% of female cancers during 1998–2002. This was unexpected given the rapidly growing prevalence of HIV in South Africa,29 as well as the clear association of Kaposi sarcoma with HIV.30 ASRs per 100,000 of 1.6 in males and 0.3 in females for Kaposi sarcoma were very low when compared with other African countries with a high HIV prevalence. Kaposi sarcoma is only a common manifestation of HIV/AIDS when Human Herpes Virus 8 (HHV-8) infection is also prevalent. The low incidence of Kaposi sarcoma observed in South Africa in the pre-AIDS era,31, 32 despite the moderate prevalence of HHV-8 observed in opportunistic sero-survey data,33–35 would be consistent with the contention that other cofactors play a role in the development of Kaposi sarcoma.36 While increases in the incidence of Kaposi sarcoma have been reported by the national cancer register since the start of the HIV/AIDS epidemic,1 the national rates remain comparatively low (Table 2).
Childhood cases (<15 years) accounted for 2.9% (73 cases) of the total cancers during the 1998–2002 period. There has been a considerable increase on cases reported when compared with the previous report15 and probably indicates an improvement in registration as a consequence of the extension of networks included in the Western Cape Paediatric Oncology Registry. However, the rates of childhood cancer may be understated. Cancers with genetic predisposition (retinoblastoma and nephroblastoma) when combined, constituted 35.6% of the childhood cancers reported.
Establishing and running a population-based cancer registry in a rural setting with limited resources is challenging. Not only does it require a reliable system to capture and process all the cancer cases that occur in the area, but is highly dependent on the clinical capacity and health services infrastructure in the area, as well as individual health seeking behaviour of the community. Furthermore, it is taxed by work-related population migration, as well as migration directly related to access to health services. Special efforts have been made to develop collaborative networks with the health facilities inside the registration area and in referral centres to maximize the completeness of registration and ensure the quality of data. Nevertheless, incidence rates for some cancers may be underestimated. In particular lymphomas and haematological cancers might be under-counted as a result of limited diagnostic and treatment services.
The cancer registry in the former Transkei region of the Eastern Cape Province remains the only cancer registry established in a rural setting in Africa. It has been possible to maintain such a register because South Africa has a national public health infrastructure with hospital services reaching all districts. Despite some uncertainty in the rates, the registry contributes comparative information that can assist in tracking the diversity of cancer patterns. It also highlights the occurrence of cancer in rural areas and the need to implement cost-effective cancer control programmes as well as appropriate health services in under-resourced areas. This study will form a baseline for the newly extended surveillance area.
We are grateful to the Eastern Cape Non-communicable Diseases Directorate, medical superintendents, doctors and nursing personnel for their contribution and co-operation during cancer data collection. We also appreciate the diligent work of data collectors; Mrs Z Mavukwana, Mrs S Grootboom, Mrs C Fadana, Mrs N Lwana. Mr E Chokunonga of the Zimbabwe National Cancer Registry, Mr Paul Opoku of Komfo Anokye Teaching Hospital Cancer Registry; Ghana and Mrs E De Kock and Mrs Karin Barnard of the MRC Burden of Disease Research Unit are thanked for their technical support. The International Agency for Research on Cancer (IARC) and International Union Against Cancer (UICC) for training support. The registry is financially supported by the South African Medical Research Council (MRC) and Cancer Association of South Africa (CANSA).
- 1Cancer in South Africa: Incidence of histologically diagnosed cancer in South Africa, 1998–1999. Johannesburg: National Cancer Registry Report, National Health Laboratory Service, 2004., , , .
- 2Lifestyle induced cancer in South Africa. In: SteynK, FourieJ, TempleN, eds. Chronic diseases of lifestyle in South Africa: 1995–2005. Cape Town: Medical Research Council 2006. 142–185., , .
- 5Statistics South Africa. Census 2001: Census in Brief. Report No 03–02-03(2001). Pretoria: Statistics South Africa, 2003.
- 6The demographic impact of HIV/AIDS in South Africa. National and provincial indicators 2006. Cape Town: Centre for Actuarial Research, South African Medical Research Council, Actuarial Society of South Africa, 2006., , , .
- 7FritzA, PercyC, JackA, ShanmugaratnamK, SobinL, ParkinDM, WhelanS, eds. International classification of diseases for oncology. Geneva: World Health Organization, 2000.
- 9Cancer incidence in five continents, Vol. I to VIII. IARC CancerBase No. 7, Lyon, 2005., , , .
- 11ParkinDM, WhelanSL, FerlayJ, TeppoL, ThomasDB, eds. Cancer incidence in five continents volume VIII, IARC Scientific Publication No. 155: International Agency for Research on Cancer: Lyon, 2002.
- 14Cancer patterns in Transkei. S Afr J Sci 1981; 77: 555–561., .
- 16Eastern cape primary health care evaluation surveys: results from the 2002 household survey. Report prepared for Eastern Cape Department of Health and Equity Project, 2004., , , , ,
- 19IARC. Some naturally occurring substances: food items and constituents, heterocyclic aromatic amines and mycotoxins. IARC monographs on the evaluation of carcinogenic risks to humans, Volume 56. Lyon: International Agency for Research on Cancer, 1993.
- 26GLOBOCAN 2000. Cancer incidence, mortality and prevalence worldwide. Version 1.0. IARC CancerBase No. 5. Lyon: International Agency for Research on Cancer, 2000., , , .
- 27CuradoM, EdwardsB, ShinH, StormH, FerlayJ, HeanueM, BoyleP (Eds.). Cancer incidence in five continents, Vol. IX. IARC Scientific Publications No. 160. Lyon: IARC, 2007.
- 29Department of Health. National HIV and syphilis prevalence survey South Africa, 2005. Pretoria: Department of Health, 2006.
- 35Risk factors for high anti-HHV-8 antibody titers (≥1: 51,200) in black, HIV-1 negative South African cancer patients: a case control study. BMC Infect Dis 2003: 12; 3: 21., , , , , , , , , .