Summary A case-control study of 913 black cancer patients (aged 15-50 years) was undertaken to measure the association between human immunodeficiency (HIV) infection and cancers believed to have an infective aetiology. Controls were patients with cancers believed not to be infective in origin. The prevalence of HIV in the controls of 7.3% (24 of 325) was similar to the background HIV seropositivity in this population. Odds ratios (ORs) and 95% confidence intervals (Cl) adjusted for age, year of diagnosis, marital status and sex were calculated. There was a strong association between HIV infection and Kaposi's sarcoma (KS), with 27 of 33 cases being HIV seropositive, OR = 61.8 and an elevated association with non-Hodgkin's lymphoma (NHL), with 27 of 40 cases being HIV seropositive [OR = 4.8 (95% Cl 1.5-14.8)]. The elevated odds ratio for KS associated with HIV infection accords with the observed increases in the incidence of KS in several sub-Saharan African countries where the prevalence of HIV is high. The odds ratio for NHL associated with HIV infection was lower than that reported in developed countries, and the reason for this is not clear. No other cancers, including cervical and liver cancers, showed significantly elevated odds ratios associated with HIV infection.Keywords: human immunodeficiency virus type 1; South Africa; case-control studyThe association between infectious agents and a number of cancers has been documented elsewhere (Weiss, 1984; ZurHausen, 1991). In South Africa about a third of the female and a sixth of the male cancers are thought to be infective in origin (Sitas et al, 1996). HIV-induced immune suppression in populations with a high prevalence of HIV has led to an increased incidence of a number of these infection-associated cancers, notably Kaposi's sarcoma (KS) and non-Hodgkin lymphoma (NHL) (and perhaps other cancers) in, for example, San Francisco single men (Rabkin et al, 1994) and sub-Saharan Africa (Bourdeaux et al, 1988;Wabinga et al, 1993;Bassett et al, 1995). In South Africa, the first notifications of HIV were in 1981, but by 1994 the HIV seroprevalence in female antenatal clinic attenders in Johannesburg and Soweto had increased to 7.9% in blacks and 0.6% in whites (Department of Health, 1995). An attempt was therefore made to quantify the association between HIV infection and certain infection-related cancers.
MATERIALS AND METHODSThe study, which was approved by the University of the Witwatwersrand Ethics Committee, was conducted between (Beral, 1991; Rabkin and Blattner, 1991). In addition, certain cancers were added to the case group 1704 HIV and cancer in South Africa 1705 because of varying degrees of suspicion of an infectious cause namely stomach (International Agency for Research on Cancer, 1994), oral (International Agency for Research on Cancer, 1995), oesophageal (Dillner et al, 1995), squamous cell skin, penile and vaginal cancers (Zur Hausen, 1991). 'Controls' comprised patients with cancers other than those suspected to have an infectious cause (see Tab...