Freddy Sitas and Valerie Beral devised and are the principal investigators of our study. Moosa Patel, Paul Ruff, Ranjan Sur, Martin Hale, David Saffer, Pradeep Rowji and Myles Connor assisted with the refinement of the study, patient recruitment, follow-up, counseling and with the classification of difficult conditions. Ute Jentsch managed the laboratory testing of the blood specimens. Rosana Pacella-Norman and Henri Carrara managed the interviewers, processing of specimens, data entry, verification and aspects of the analysis. Valerie Beral, Diana Bull, Rob Newton and Freddy Sitas produced the final tabulations. All authors contributed to the analysis of the data and to the writing of the article.
Epidemiology and Cancer Prevention
The spectrum of HIV-1 related cancers in South Africa †
Article first published online: 17 OCT 2000
Copyright © 2000 Wiley-Liss, Inc.
International Journal of Cancer
Volume 88, Issue 3, pages 489–492, 1 November 2000
How to Cite
Sitas, F., Pacella-Norman, R., Carrara, H., Patel, M., Ruff, P., Sur, R., Jentsch, U., Hale, M., Rowji, P., Saffer, D., Connor, M., Bull, D., Newton, R. and Beral, V. (2000), The spectrum of HIV-1 related cancers in South Africa . Int. J. Cancer, 88: 489–492. doi: 10.1002/1097-0215(20001101)88:3<489::AID-IJC25>3.0.CO;2-Q
- Issue published online: 17 OCT 2000
- Article first published online: 17 OCT 2000
- Manuscript Accepted: 23 MAY 2000
- Manuscript Revised: 18 APR 2000
- Manuscript Received: 24 JAN 2000
- South African Medical Research Council
- Cancer Association of South Africa, the University of the Witwatersrand
- South African Institute for Medical Research
- Department of Health
- Imperial Cancer Research Fund (U.K.)
Despite the high prevalence of infection by the Human Immunodeficiency Virus (HIV) in South Africa, information on its association with cancer is sparse. Our study was carried out to examine the relationship between HIV and a number of cancer types or sites that are common in South Africa. A total of 4,883 subjects, presenting with a cancer or cardiovascular disease at the 3 tertiary referral hospitals in Johannesburg, were interviewed and had blood tested for HIV. Odds ratios associated with HIV infection were calculated by using unconditional logistic regression models for 16 major cancer types where data was available for 50 or more patients. In the comparison group, the prevalence of HIV infection was 8.3% in males and 9.1% in females. Significant excess risks associated with HIV infection were found for Kaposi's sarcoma (OR=21.9, 95% CI=12.5–38.6), non-Hodgkin lymphoma (OR=5.0, 95%CI=2.7–9.5), vulval cancer (OR=4.8, 95%CI=1.9–12.2) and cervical cancer (OR=1.6, 95%CI=1.1–2.3) but not for any of the other major cancer types examined, including Hodgkin disease, multiple myeloma and lung cancer. In Johannesburg, South Africa, HIV infection was associated with significantly increased risks of Kaposi's sarcoma, non-Hodgkin lymphoma and cancers of the cervix and the vulva. The relative risks for Kaposi's sarcoma and non-Hodgkin lymphoma associated with HIV infection were substantially lower than those found in the West. Int. J. Cancer 88:489–492, 2000. © 2000 Wiley-Liss, Inc.
Infection with HIV-1 has been associated with increases in the risk of developing Kaposi's sarcoma, non-Hodgkin lymphoma, squamous conjunctival cancers and leiomyosarcoma in children (IARC Working Group, 1996), all of which being cancers that are thought to be associated with infective agents. In a previous study conducted in Johannesburg between 1992 and 1994 (Sitas et al., 1997), a significant excess risk associated with HIV infection was found only for Kaposi's sarcoma (OR=61.8) and non-Hodgkin lymphoma (OR=4.8) but not for any other cancer types or sites. In studies from central Africa, infection with HIV was found to be associated with an increased risk of Kaposi's sarcoma (Ziegler et al., 1997), conjunctival cancer (Newton et al., 1995) and non-Hodgkin lymphoma (Newton et al., 2000). There are however a number of other cancer types that are known to be associated with an infectious agent, but the evidence to date of an association with HIV has been contradictory. In South Africa, the HIV epidemic was first noted in 1982 and is primarily due to HIV-1. In 1996 in the province of Gauteng, where Johannesburg is situated, 18.8% of the women 15–44 years of age that were surveyed at the antenatal clinics tested positive for HIV. Nationally, about 16% of surveyed women attending antenatal clinics were infected with HIV and it was estimated that about 2.4 million adults and 157,000 children were infected (Republic of South Africa Epidemiological Comments, 1997). As the HIV-1 epidemic matures, there is some uncertainty about the ability of HIV to increase the risks of certain cancers. Our study was carried out to monitor the effect of some of the main known risk factors for cancer, for example smoking, domestic pollution, alcohol consumption sexual and reproductive history, occupation and certain infections including HIV, in relation to a number of cancer types.
An ongoing cancer epidemiological study began in 1995 in the 3 main referral hospitals of Johannesburg and Soweto to examine the relative importance of key risk factors for cancer. The methods have been described elsewhere (Sitas et al., 1999) but briefly, until January 1999, trained nurses systematically interviewed 6,478 incident black (mainly cancer) patients aged 18 years or older who presented at the medical, surgical, radiation therapy and oncology wards of Chris Hani-Baragwanath, Hillbrow and New Johannesburg hospitals. Included in this patient group were 303 females admitted for vascular disease because of an interest in certain of these factors (smoking alcohol, reproductive history etc.) on vascular disease among women. Questions were asked in the home language of the patient (usually Zulu or Sotho) about their demographic and behavioral characteristics, including place of birth, current residence, years of education, and reproductive and lifetime sexual history. The questionnaires were anonymous and patients gave written or verbal consent (if illiterate) to participate in the study. This study was approved by the University of the Witwatersrand ethics committee.
For a variety of reasons, (patient refusal, inadequate volume or on occasion storage freezer and transport breakdowns), sufficient sera were available for HIV testing from 75% of patients (4,890). Testing for specific HIV-1 or -2 antibodies was done using the Abbott Axysm® HIV 1/2 group O Microparticle Immunoassay. Specimens with a sample optical density/cut-off value of greater than or equal to 1 were considered positive for HIV antibodies, assumed to be HIV-1. For about 300 samples where insufficient serum was available, the Capillus HIV 1/2 Latex agglutination Kit (Cambridge Diagnostics, United Kingdom) was used. Agglutinated sera were considered as positive for HIV-1 antibodies. Equivocal results were classified as HIV negative.
Because of the uncertainty regarding the cancer types that might be associated with HIV-1 infection, we chose to calculate the odds ratios associated with HIV infection for each of the major cancer types where there was a reasonable number of patients, i.e., 50 or more (Table I). The comparison group included 556 female subjects with a mixture of other cancers or vascular disease and 288 male subjects with other cancer. The conditions of the patients included in the control group (and their HIV seroprevalence) were 136 with cancers of the oral cavity, oro, naso pharynx and sinuses (5.9%); 54 digestive cancers other than colon and rectum (3.7%); 47 bone, connective tissue and sarcomas other than Kaposi's (6.4%); 65 melanoma and non-melanoma skin cancers (9.2%); 13 male breast (15.4%); 36 miscellaneous genital (13.9%); 40 urinary cancers (5%); 22 brain & nervous system cancers (13.6%); 31 endocrine (3.2%); cancers of secondary sites (12.5%) and 65 cancers of ill-defined origin (3.1%). Among the patients with vascular disease 139 had a variety of hypertensive disorders (13.7%), 4 ischaemic heart disease (0%); 19 pulmonary heart disease (5.3%); 50 other heart diseases and cardiomyopathies (16%) and 91 cerebrovascular diseases (8.9%). There was no significant heterogeneity in HIV seroprevalence rates between the various cancers or other conditions included in the comparison group. Heterogeneity was assessed using unconditional unmatched logistic regression to estimate HIV prevalence rates, after adjustment for age, sex and date of diagnosis (Sitas et al., 1999).
|Cancer type or site||ICD-10||HIV + ve/total cases||HIV %||HV %||OR1||OR2|
|Oesophagus||C15.-||20/400||5.0||97.3||1.1 (0.6–1.9)||1.1 (0.6–2.0)|
|Stomach||C16.-||2/64||3.1||95.1||0.5 (0.1–2.1)||0.5 (0.1–2.1)|
|Colorectum||C18-C20.-||11/131||8.4||97.8||1.1 (0.6–2.2)||1.0 (0.5–2.0)|
|Larynx||C32.-||3/56||5.4||96.8||0.8 (0.2–2.8)||0.8 (0.2–2.7)|
|Lung||C33-C34.-||9/134||6.7||95.3||1.1 (0.5–2.4)||1.0 (0.5–2.3)|
|Kaposi's sarcoma||C46.-||89/119||74.8||99.4||20.4 (12.1–34.6)||21.9 (12.5–38.6)|
|Female breast||C50.-||43/687||6.3||99.1||0.9 (0.6–1.4)||0.8 (0.5–1.3)|
|Vulva||C51.-||10/53||18.9||95.6||4.5 (1.8–11.4)||4.8 (1.9–12.2)|
|Cervix||C53.1||167/1323||12.6||99.3||1.4 (1.2–2.3)||1.6 (1.1–2.3)|
|Endometrium||C54.1||9/99||9.1||96.7||1.5 (0.7–3.4)||1.8 (0.8–4.2)|
|Ovary||C56.-||9/90||10.0||97.2||1.4 (0.7–3.2)||1.4 (0.6–3.0)|
|Prostate||C61.-||7/312||2.2||97.3||0.4 (0.2–1.1)||0.4 (0.2–1.2)|
|Hodgkin||C81.-||14/109||12.8||100.0||1.3 (0.7–2.6)||1.4 (0.7–2.8)|
|Non-Hodgkin lymphoma||C82.-, C83.-, C85.-||23/105||21.9||100.0||3.8 (2.1–6.9)||5.0 (2.7–9.5)|
|Myeloma||C90.-||5/114||4.4||100.0||0.8 (0.3–2.2)||0.9 (0.4–2.5)|
|Leukemia||C91-C96.-||10/243||4.1||100.0||0.5 (0.3–1.1)||0.5 (0.3–1.2)|
|Controls male||24/288||8.3||90.9||Reference category||Reference category|
|Controls female||50/556||9.0||86.0||Reference category||Reference category|
Unconditional unmatched logistic regression (SAS Institute Inc, 1997) was used to calculate the odds ratios and 95% confidence intervals of developing cancers of different types in relation to HIV infection. Odds ratios were first adjusted for age (18–34, 35–54, 55+ years and age unknown), sex (where appropriate) and date of diagnosis (before and after September 1996) and then number of lifetime sexual partners (0–4, 5+, unknown) and place of usual residence (urban/rural/unknown). For females with cancers of the breast, cervix, endometrium, vulva and ovary, parity (0–2, 3+ children and unknown) was included in the model. A separate term was added for each adjustment factor in the model.
About 89% of the patients with cancers had their diagnoses verified by histology, haematology or cytology (Table I). A total of 4,883, 3,404 female and 1,479 male, patients were tested for antibodies against HIV, and the distribution of these by site is shown in Table I. Among patients in the comparison group (see Methods), the crude prevalence rate of HIV was 9.0% (50/556) in females and 8.3% (24/288) in males. In the comparison group, in both sexes there is characteristic peak prevalence of HIV infection in the 25–34 year-old-age group (Table II) and the prevalence of HIV infection at ages 18–44 years was 12.8% in males and 18.5% in females. In the comparison group, significantly increased risks were found between HIV infection and 5 or more lifetime partners (OR=2.6, 95%, CI=1.5–4.5). Those born in a rural area (OR=0.7), of no education (OR=0.8) and females of higher parity (3 or more children, OR=0.7) had lower risks of HIV infection but none of these reached statistical significance, after adjustment for all these factors, age and date of diagnosis. These data are similar to those found by Newton et al. in a similar study in Uganda (Newton et al., 2000).
|N subjects||Prevalence % (and 95% CI)||N subjects||Prevalence % (and 95% CI)|
|18–24||25||12.0 (3.9–31.3)||18||16.7 (5.5–40.9)|
|25–34||38||26.3 (14.8–42.4)||22||13.6 (4.5–34.8)|
|35–44||86||16.3 (9.9–25.7)||38||10.6 (4.0–24.9)|
|45–54||128||11.7 (7.2–18.5)||84||7.2 (3.3–15.0)|
|55–64||142||4.8 (2.3–9.7)||78||7.7 (3.5–16.1)|
|65+||132||0.8 (0.1–5.1)||48||5.1 (1.1–15.1)|
|Total||556||9.0 (6.9–11.7)||288||8.3 (5.7–12.1)|
Given the high proportion of patients who had their cancers histologically verified, there was virtually no difference in the odds ratios if only these subjects were included, but because of the slightly reduced numbers, the confidence intervals also became slightly wider. Significantly increased risks in association with HIV infection were noted for Kaposi's sarcoma (89 out of 112 patients with Kaposi's were HIV seropositive; OR=21.9, 95%, CI=12.5-38.6), non Hodgkin lymphoma (23/105; OR=5.0, 95%, CI=2.7–9.5), cervical cancer (167/1323; OR=1.6, 95%, CI=1.1–2.3) and cancer of the vulva (10/53; OR =4.8, 95%, CI=1.9–12.2). No excess risks were found for any of the other 12 major cancer types. There were no patients in this series with squamous cell conjunctival cancer. Among those with non-Hodgkin's lymphoma, 4 individuals were diagnosed as having Burkitt's lymphoma, 3 of whom were HIV seropositive (OR 19.4, 95%, CI=1.8–211.4, based on all 4 cases). Among the cases with cervical cancer, when compared with those aged 18–34 years, the risk of HIV infection was the same in those aged 35–54 (OR=0.9, 95%=0.5–1.7) but appeared to be higher in those aged 55 years and older (OR=1.6, 95%, CI= 1.1–2.1), p trend=0.02.
The seroprevalence of HIV among the females (aged 18–44 years) in the comparison group resembled the prevalence rates from females attending antenatal clinics in 1996 (18.5% vs. 18.8%), the midpoint of the study. There are no contemporary HIV seroprevalence data available for males. In an earlier study on HIV and cancer, the seroprevalence of HIV in males in the comparison group (using a similar study design) resembled closely the seroprevalence of HIV in a group of male “blue collar” workers (Sitas et al.1997).
There are certain similarities between the results found in our study and earlier studies carried out in Rwanda in 1992 (Newton et al., 1995), Johannesburg between 1992-1994 (Sitas et al., 1997) and in children in Uganda (Newton et al., 2000). Although all studies found elevated risks for Kaposi's sarcoma, the relative risk of Kaposi's sarcoma in HIV infected individuals in Africa (35 in Rwanda, 62 in South Africa and 54 in Uganda) appears to be much lower than the relative risk found among HIV infected individuals in the west (usually 300 or more (Goedert et al., 1998; Beral et al., 1990)). Multiple pathology and rapid mortality associated with HIV infection may hide some of the cancers that might occur in HIV infected individuals in Africa. However, central Africa was the region of highest incidence of Kaposi's sarcoma before the advent of HIV, and even in South Africa a cumulative incidence rate (0–64 years) of 1–3 per thousand was estimated (Cook-Mozzaffari et al., 1998). Assuming a pre-AIDS baseline cumulative rate of developing Kaposi's sarcoma in South Africa of 1 per thousand (Cook-Mozzaffari et al., 1998; Sitas et al., 1998), and the odds ratios found here, then the cumulative risk of developing Kaposi's sarcoma in HIV positive individuals in the South African black population is calculated to be about 20 per thousand. By contrast, in most areas in the west, in the 1970s Kaposi's sarcoma was extremely rare and the cumulative incidence rate in the UK for example was about 0.005 per thousand (Beral et al., 1990; Grulich et al., 1992). So, if relative risks in the order of 300 or more applied, the cumulative risk in HIV infected individuals in the West may well be of a similar order of magnitude as found in Africa.
The relative risk of non-Hodgkin lymphoma associated with HIV infection was found to be similar to that reported in previous studies from Africa in South Africa (4.8) (Sitas et al., 1997), Uganda (3.8) (Newton et al., 2,000) and Rwanda (12.6) (Newton et al., 1995). In these studies, the relative risk of non-Hodgkin lymphoma associated with HIV infection is an order of magnitude lower than those reported from Western countries (IARC Working Group, 1996; Goedert et al., 1998). Reasons for the lower relative risks for NHL in Africa could include HIV related co-morbidity, but it may be possible that infection with Epstein Barr Virus (thought to be a cause of certain types non-Hodgkin lymphoma IARC Working Group 1997) or other viruses may be more likely to lead to disease in the West if acquired in young adulthood, especially around the same time as the acquisition of HIV. The relative risk for Burkitt's lymphoma in relation to HIV infection was 19.4 (95%, CI 1.8–211), higher than for other non-Hodgkin's lymphomas. The number of cases of Burkitt's lymphoma was small (4 cases); nevertheless, the relative risk estimate is substantially lower than has been found in western countries (Beral et al., 1991).
An excess risk of cancer of the cervix associated with HIV was not observed in a previous study in this setting (Sitas et al., 1997) nor in Rwanda (Newton et al., 1995). Now similar excess risks have been found both in Johannesburg, OR=1.6 (95%, CI=1.1–2.3) and in Uganda, OR=1.9 (95%, CI=1.1-3.3) (Newton et al., 2000). There have been suggestions from Western countries that cervical cancer may be associated with HIV infection, (see, e.g., Goedert et al., 1998; Serraino et al., 1999), but most of the previous studies have tended to have small numbers of patients with cervical cancer and some studies have combined invasive cervical cancers with cervical carcinomas in situ. Vulval cancer appears to be associated with HIV in our study (OR= 4.8). However in our previous study included only 12 cases of vulval cancer of which 2 were HIV positive (OR=1.2, 95%, CI=0.1–10.9). The data from Rwanda and Uganda have not been published. It is unclear at this stage whether the excess risks found in relation to vulval and cervical cancers are real or due to incomplete adjustment for sexual activity. No human papillomavirus status was available for this group of patients and it may be important to understand the relationship of these 2 viruses in the causal pathway of cervical and vulval cancer.
About 100 new cases of histologically diagnosed Kaposi's sarcoma were recorded per year at the (pathology based) South African National Cancer Registry prior to 1993 (Sitas et al., 1998). This number has increased to 315 in 1996 (unpublished data), and the number of cases of Kaposi's sarcoma diagnosed without pathology is likely to be higher. No marked increases have been noted for any of the other HIV related cancers in South Africa. The latest HIV seroprevalence data from women attending antenatal clinics (Republic of South Africa, 1997) suggest that in South Africa the HIV epidemic will soon reach the proportions observed in other central and east African countries. In countries with prolonged exposure to HIV like Uganda and Zimbabwe (Parkin et al., 1999; Chokunonga et al., 1999), Kaposi's sarcoma is now the leading cancer, accounting for about a third of all cancers. In South Africa, between 1993 and 1995 Kaposi's sarcoma comprised only 0.3% of female and 0.7% of male cancers (Sitas et al.1998), so cancer trends in South Africa need to be carefully monitored over time.
We are indebted to Sisters S. Ngcukana and G. Mokwatle who conducted the most of the interviews; to Sister M. Kubekha, manager of the SAIMR phlebotomy service at Chirs Hani Baragwanath Hospital; Ms. V. Davis, Ms. L. Bester, Ms. M. Ford, Ms. P. Rikhotso, Ms. P. Mabaso, Ms. N. Mchunu, Ms. S. Korber, Ms. P. Caenazzo and Ms. S. Zaba for specimen processing and HIV testing; Ms. J. Knuppel for typing the article; Ms. F. Mgnomezulu and Ms. H. Mambatha for data coding, entry and verification; Mr. J. Wessie and J. Madhoo for computer support and to all the clinicians and matrons in charge of the wards at Chris Hani-Baragwanath, Hillbrow and Johannesburg hospitals and to the patients for their participation in the study. The Cancer Epidemiology Research Group is supported by the South African Medical Research Council, the Cancer Association of South Africa the University of the Witwatersrand and the South African Institute for Medical Research. The National Cancer Registry is also funded by the Department of Health. Additional support for our study was received from the Imperial Cancer Research Fund (U.K.).
- Kaposi's sarcoma among persons with AIDS: a sexually transmitted infection? Lancet, 335, 123-128 (1990). , , and ,
- AIDS-associated non-Hodgkin lymphoma. Lancet, 337, 805-809 (1991). , , and ,
- AIDS and cancer in Africa: the evolving epidemic in Zimbabwe. AIDS, 13, 2583-2588 (1999). , , , , , and ,
- The geographical distribution of Kaposi's sarcoma and of lymphomas in Africa before the AIDS epidemic. Br J Cancer, 78, 1521-1528 (1998). , , and ,
- Spectrum of AIDS-associated malignant disorders. Lancet, 351, 1833-1839 (1998). , , , , , , and ,
- Kaposi's sarcoma in England and Wales before the AIDS epidemic. Br. J. Cancer, 66, 1135-1137 (1992). , and ,
- International Agency For Research On Cancer. Working Group. IARC Monographs on the evaluation of carcinogenic risks to humans, vol 67. Human immunodeficiency viruses and human T-cell lymphotropic viruses, Lyon: IARC (1996).
- International Agency For Research On Cancer. Working Group. IARC monographs on the evaluation of carcinogenic risks to humans, vol 70. Epstein Barr Virus and Kaposi's sarcoma herpesvirus/human herpesvirus 8. Lyon: IARC 1997).
- Cancer and HIV in Rwanda. Lancet, 345, 1378-1379 (1995). , , , , , and ,
- Human immunodeficiency virus infection and cancer in Ugandan adults and children. Int. J. Cancer, this issue. , , , , , and ,
- AIDS related cancers in Africa: maturation of the epidemic in Uganda. AIDS, 13, 2563-2570 (1999). , , and ,
- Republic Of South Africa, Department Of Health. Epidemiological comments. 23, 4-16 1997).
- SAS Institute, Inc. SAS/STAT software: changes and enhancements through release 6.12. Cary, NC (1997).
- Risk of invasive cervical cancer among women with or at risk for, HIV infection. Int. J. Cancer, 82, 334-337 (1999). , , , , , , , , , , and ,
- Association between human immunodeficiency virus type 1 infection and cancer in the black population of Johannesburg and Soweto, South Africa. Br. J. Cancer, 75, 1704-1707 (1997). , , , , and ,
- Incidence of histologically diagnosed cancer in South Africa 1993-1995. Johannesburg: National Cancer Registry, South African Institute for Medical Research 1998). , and ,
- Antibodies against human herpesvirus 8 in black South African patients with cancer. N Engl J Med, 340, 1863-1871 (1999). , , , , and ,
- Risk factors for Kaposi's sarcoma in HIV positive subjects in Uganda. AIDS, 11, 1619-1626 (1997). , , , , , and ,