- Top of page
- Statistical analysis
The effect of the evolving HIV epidemic on cancer has been sparsely documented in Africa. We report results on the risk of cancer associated with HIV-1 infection using data from an ongoing study. A case–control analysis was used to estimate the relative risk (odds ratio, OR) of cancer types known to be AIDS defining: Kaposi's sarcoma (n = 333), non-Hodgkin lymphoma (NHL, n = 223) and cancers of the cervix (n = 1,586), and 11 cancer types possibly associated with HIV infection: Hodgkin lymphoma (n = 154), cancers of other anogenital organs (n = 157), squamous cell cancer of the skin (SCC, n = 70), oral cavity and pharynx (n = 319), liver (n = 83), stomach (n = 142), leukemia (n = 323), melanoma (n = 53), sarcomas other than Kaposi's (n = 93), myeloma (n = 189) and lung cancer (n = 363). The comparison group comprised 3,717 subjects with all other cancer types and 682 subjects with vascular disease. ORs were adjusted for age, sex (except cervical cancer), year of diagnosis, education and number of sexual partners. Significantly increased risks associated with HIV-1 infection were found for HIV/AIDS associated Kaposi's sarcoma (OR = 47.1, 95% CI = 31.9–69.8), NHL (OR = 5.9, 95% CI = 4.3–8.1) and cancer of the cervix (OR = 1.6, 95% CI = 1.3–2.0); Hodgkin's disease (OR = 1.6, 95% CI = 1.0–2.7), cancers of anogenital organs other than the cervix (OR = 2.2; 95% CI = 1.4–3.3) and SCC (OR = 2.6, 95% CI = 1.4–4.9) were also significantly increased. No significant associations were found between HIV and any of the other cancers examined. Risks for HIV-related cancers are consistent with previous studies in Africa, and are lower when compared to those observed in developed countries. © 2008 Wiley-Liss, Inc.
Infection with HIV-1 increases the risk of Kaposi's sarcoma, non-Hodgkin lymphoma and cervical cancer, all 3 conditions classified as “AIDS-defining.”1 In addition, HIV-1 infection has been associated with increased risks of Hodgkin disease, leiomyosarcoma (in children) and squamous cell carcinoma of the conjunctiva.2 There has been less conclusive evidence regarding the association between HIV-1 and cancers of the anogenital organs other than the cervix.2 HIV has also been thought to increase the risk of several other infection related cancers such as stomach and liver (hepatocellular) cancer, as well as certain cancers where an infectious etiology is suspected (for example, oral cancers and squamous cell skin cancers). Squamous cell skin cancers are a recognized complication of immunosuppressed organ transplant recipients,3 but the evidence of an association with HIV-1 is sparse. In a number of record-linkage studies of HIV-infected individuals to cancer registers4 a few other cancer types have been reported in association with HIV-1, but in most linkage studies confounding by lifestyle cannot be adjusted for. Cancer types found at higher risk in relation to HIV-1 include lung cancer,5 other sarcomas,6 myeloma7 and melanoma.8 A recent record linkage study in Uganda between an HIV/AIDS register and the Kampala cancer registry showed some increases in thyroid and kidney cancers in subjects with HIV, but the numbers observed were small.9 In some regions of Africa where HIV-1 is endemic such as Central and East Africa, Kaposi's sarcoma has become the leading cancer type, comprising about a quarter of all cancer cases, and increases have also been noted in the incidence of non-Hodgkin lymphoma.10
Case–control studies in sub-Saharan Africa have consistently shown lower risks of cancer development in relation to HIV infection than in Western countries.10 An ongoing case–control study to measure the association between infectious agents (including HIV-1), lifestyle factors and cancer among the black patient population attending the 3 major tertiary hospitals began in Johannesburg in 1995. The first results (up to January 1999) on the association between HIV and several cancer types, after adjustment for confounding factors such as number of sexual partners, were published elsewhere.11 The study has continued recruiting subjects, and the sample size available for analysis has almost doubled. This updated report on 10 years of data, from 1995 to 2004, provides estimates on the association between HIV-1 and a number of cancer types known or suspected to be associated with HIV-1, after adjustment for a number of lifestyle factors. Some of these cancer types were previously not available in sufficient numbers for more detailed analyses.
One-third of all people with HIV live in southern Africa, and South Africa has one of the world's largest HIV epidemics, which has not yet reached a plateau.12 During the period of this study, estimated HIV prevalence among women attending Public Health Service antenatal clinics across all 9 provinces in South Africa increased from 10.4% in 199513 to 29.5% in 2004.14 In Gauteng province, where this study was conducted, HIV prevalence among these women was 33.1% overall, and 38.5% among those aged 25–29 years in 2004.14 A South African National household survey, conducted in 2005, showed overall HIV prevalence among adults aged 15–49 years to be 20.2% in females and 11.7% among males, with an overall prevalence of 19.9% among black Africans.15 From 1997 to 2004, death rates from “natural causes” in South Africa increased 5-fold for women aged 25–34, and more than doubled for males aged 30–44. A large proportion of these deaths have been attributed to the AIDS epidemic.16–20 Given the statistics above, it is clearly of interest to document the cancer burden associated with HIV and to monitor whether increases in HIV prevalence during the period of this study have brought to light new associations between HIV and cancer.
- Top of page
- Statistical analysis
An ongoing cancer case control study commenced in 1995 in the three main referral hospitals in Johannesburg and Soweto to measure the relative importance of key risk factors for cancer including infectious agents like HIV. Subject recruitment has been described elsewhere,11 but briefly, nurses trained in interviewing and counseling questioned adult (18 years and over) black African patients (black patients all of whose ancestors came from the African continent) with newly diagnosed cancers at Chris Hani Baragwanath, Hillbrow, and Johannesburg Hospitals in Johannesburg, using a structured two-page questionnaire (available on request). During the study period (March 1995–June 2004), these state hospitals provided cancer chemotherapy and radiotherapy to patients mainly (75%) from southern Gauteng province. Additionally, from 1998 to 2001, patients presenting with cardiovascular diseases at the same hospitals were also included to serve as controls. The interview was conducted in the preferred language of the patient (usually Zulu or Sotho) following written or witnessed verbal (if illiterate) consent to participate. The questionnaires and included inter alia questions on smoking, frequency of alcohol consumption, birthplace, residence, education, and reproductive, contraceptive and lifetime sexual history. Blood samples were collected at the time of interview, prior to commencing treatment. The study was approved by the University of Witwatersrand Human Research Ethics Committee (Medical).
Cancers were classified by primary site (topography) and by morphology using the WHO International Classification of Diseases for Oncology (ICD-O-2) guidelines.21 In the first years of the study most of the diagnoses were taken from the patient records; from September 2000, every effort has been made to obtain the laboratory reports of the diagnoses: 97% were microscopically verified and most of the remaining diagnoses were supported by biochemical or clinical tests.
Serum specimens were stored at −20 to −30°C prior to being batched for HIV testing. Operational issues limited the availability of specimens to 88.2% of the subjects interviewed, ranging from an annual low of 76.3% (1997) to 98.5% (2003). Fewer than 1% of patients refused testing. From 1995 to 2001 HIV testing was mainly done using the Abbott Axysm HIV1/2 gO Microparticle Enzyme Immunoassay. Thereafter the Vironostika (HIV Uni-Form II plus O) microELISA has been used. Specimens having intermediate reactivity were classified as negative. The laboratory performing the tests was unaware of any disease or demographic details of the subjects.
A case–control analysis was used to calculate relative risks of developing specific cancers, defined a priori in relation to HIV-1 infection. These were as follows: (i) cancer types known to be HIV/AIDS related, comprising Kaposi's sarcoma (n = 333), non-Hodgkin lymphoma (n = 223, 12 of which were adult Burkitt lymphoma) and cancer of the cervix (n = 1,586) and (ii) other cancer types thought to be associated with HIV-1 infection, comprising Hodgkin lymphoma (n = 154), cancers of the anogenital organs other than the cervix (n = 157), squamous cell cancers of the skin (n = 70), cancers of the oral cavity and pharynx (n = 319), liver (n = 83), stomach (n = 142), lung (n = 363), leukemia (n = 323), melanoma (n = 53), sarcomas other than Kaposi's (n = 93) and myeloma (n = 189). Subjects diagnosed with lymphoma, sarcoma, leukemia and myeloma were further divided into more detailed (4-digit ICD-02) morphology types. Subjects with anogenital types were further divided into anal, penile, vulval and vaginal cancers.
The comparison group comprised subjects with all the other cancer types not mentioned in (i), or (ii), as follows: breast-female (n = 1,261), breast-male (n = 28), esophagus (n = 778), prostate (n = 358), colon/rectum (n = 290), endometrium (n = 162), ovary (n = 158), larynx (n = 127), endocrine (n = 60), other minor cancer types (n = 495) and subjects diagnosed with vascular disease (n = 682).
- Top of page
- Statistical analysis
Two-sided p-values for heterogeneity or trend were calculated using Cochran Mantel-Haenszel statistics (Table I). In Table II, odds ratios and 95% Wald confidence intervals were calculated using unconditional logistic regression. The risk of HIV infection for a given HIV-related cancer type of interest compared to the comparison group of cancers (see earlier) was estimated by odds ratios. Two odds ratios are presented, the first was adjusted for age group at diagnosis (18–34, 35–54, 55–74 years), sex (where appropriate) and year of diagnosis (annually from 1995 to 2004, as a continuous variable from 0 to 9), and the second odds ratio was adjusted in addition for the number of sexual partners (0–2, 3–4, 5 or more, missing) and completed level of education [none, up to 6 years (i.e., completing primary school), up to 10 years (i.e., completing 3 years high school education) and 4 or more years of high school]. Additional adjustments were made for smoking behavior (never, ex-, current), however this adjustment did not materially change the odds ratios for any of the cancer types examined and the data are not presented.
Table I. Association between Selected Socio-Demographic Variables and HIV-1 Seroprevalence in Controls
| ||HIV-1 seroprevalence|
|Age group (years)|
| Missing||3|| ||14|| |
|p-value for trend test||<0.0001||<0.0001|
|Education|| || || || |
| No schooling||321||5.6||472||5.7|
| Primary (6 years)||650||7.2||1,039||9.0|
| Junior high school (7–9 years)||377||6.4||950||13.2|
| Senior high school (10 or more years)||159||11.9||385||17.4|
| Missing||29|| ||17|| |
|p-value for trend test||0.99||0.85|
|Number of sexual partners|
| Missing||54|| ||52|| |
|p-value for trend test||0.1||<0.0001|
| Current smoker||349||7.7||205||13.7|
| Missing||22|| ||16|| |
|p-value for trend test||0.24||0.17|
|Place of birth|
| Missing||29|| ||2|| |
|p-value for test of homogeneity||0.8||0.95|
|Place of residence|
| Missing||38|| ||23|| |
|p-value for test of homogeneity||0.98||0.04|
|Total||1,536|| ||2,863|| |
Table II. Association between Certain Cancers and HIV-1 Infection in South Africa 1995–2004
|Cancer site or type (and ICD-O-2 code)||N||HIV-1 +ve (%)||OR11 and 95% CI||OR22 and 95% CI|
|Oral cavity and pharynx (C00–C10)||319||7.5||0.9 (0.6–1.4)||0.8 (0.5–1.3)|
|Stomach (C16)||142||4.2||0.4 (0.2–1.0)||0.4 (0.2–1.0)|
|Liver (C22)||83||9.6||0.8 (0.4–1.6)||0.8 (0.4–1.7)|
|Lung (C33–C34)||363||8.5||1.1 (0.7–1.6)||1.1 (0.7–1.6)|
|Squamous cell skin (C44, M8050–8082)||70||21.4||2.6 (1.4–4.7)||2.6 (1.4–4.9)|
|Melanoma (M8720–8780)||53||15.1||1.6 (0.7–3.5)||1.7 (0.7–3.8)|
|Kaposi's sarcoma (M9140)||333||89.2||50.4 (34.2–74.3)||47.1 (31.9–69.8)|
|Other sarcomas3||93||14.0||0.9 (0.5–1.6)||0.8 (0.4–1.6)|
|Cervix (C53)||1,586||14.9||1.7 (1.4–2.0)||1.6 (1.3–2.0)|
|Anogenital organs other than cervix (C21, 51, 52, 60)||157||22.3||2.5 (1.7–3.8)||2.2 (1.4–3.3)|
|Hodgkin lymphoma3||154||19.5||1.5 (1.0–2.4)||1.6 (1.0–2.7)|
|Non-Hodgkin lymphoma3||223||44.4||6.1 (4.4–8.4)||5.9 (4.3–8.1)|
|Myeloma3||189||5.3||0.7 (0.4–1.4)||0.7 (0.4–1.4)|
|Leukemia3||323||7.4||0.6 (0.4–1.0)||0.6 (0.4–1.0)|
|Comparison group (males and females)||4,399||9.7||1.0 (–)||1.0 (–)|
|Comparison group, females only||2,863||11.0||1.0 (–)||1.0 (–)|
- Top of page
- Statistical analysis
A total of 8,487 subjects were tested for HIV, almost double the sample size of the previous study (n = 4,88311). The seroprevalence of HIV in the comparison group was 11% in females, 7% in males and 10% overall. The cancer profile of the subjects interviewed resembled the background distribution of histologically reported cancers in South Africa in 1999 (the latest period for which cancer registration data are available). In this study, compared to cancer registrations reported to the South African National (pathology based) Cancer Registry (NCR),22 the top 5 cancers in females were, respectively, as follows: cervix (32% (this study) vs. 36% (NCR)), breast (25% vs. 19%), oesophagus (6% vs. 6%), endometrium (5% vs. 4%) and ovary (4% vs. 3%). In males, the top 5 cancers in the study compared to the NCR were, respectively, as follows: oesophagus (17% vs. 15%), prostate (13% vs. 14%), lung (10% vs. 10%), oral-naso-pharynx (8% vs. 9%) and KS (6% vs. 5%). Microscopic verification of cancers in these tertiary treatment hospitals is a requirement for most cancer types before commencement of treatment, hence the high proportion of cancer microscopically verified subjects in the study (>90%). The prevalence of HIV infection in the comparison group resembles the pattern observed in other population-based HIV serological surveys of South African black populations,13–15 showing an age peak at 18–34 years and a decline thereafter.
In the absence of cancer registries in these three hospitals it is impossible to calculate response rates. In a recent audit in 2005–2006, of 195 out of 2,745 (7%) subjects who were approached at Johannesburg Hospital but who did not participate, the main reasons were as follows: out of age range (24%), unable to talk/deaf (22%), too ill/senile (19%), treatment already started (15%), refused (9%), foreigner (2.5%) and religious reasons (0.5%). There was no significant difference in the prevalence of HIV-1 among the control group subjects with cancer and vascular disease, after adjustment for age group, year of diagnosis and sex (x2 heterogeneity (8 d.f.) = 11.2; p = 0.2). In the male comparison group, HIV-1 seroprevalence (adjusted for age and year of diagnosis) was the following: vascular disease 4%, cancer of the prostate 4%, oesophagus 5%, colorectum 5%, larynx 8%, endocrine 3% and 5% for other minor cancer types. In female controls the HIV-1 seroprevalence was 21% in those with vascular disease, and for cancer of the ovary 17%, oesophagus 15%, larynx 25%, endocrine system 4%, endometrium 13%, colorectum 11%, breast 14% and other minor cancer types 14%.
Table I shows the relationship between HIV-1 seroprevalence and selected demographic characteristics in the comparison group. HIV-1 seroprevalence is inversely associated with increased age (p < 0.0001 in both males and females). After adjustment for age, HIV-1 prevalence was associated with a higher number of sexual partners in females (p < 0.0001) but not in males (p = 0.1) and with place of current residence in females only (p = 0.04). Seroprevalence was higher among females living in urban areas. Place of birth was not associated with HIV-1 prevalence (p = 0.8 in males, 0.95 in females). Although the prevalence of HIV-1 appeared to be higher in subjects of higher education, the relationship was not statistically significant after adjustment for age (p = 1.0 in males, p = 0.9 in females).
Table II shows the association between HIV and selected major cancer types adjusted for a number of lifestyle factors (see table footnote). Significant increased risks were found in relation to the three HIV/AIDS-related cancer types, namely Kaposi's sarcoma (ORadj = 47.1, 95% CI = 31.9–69.8), non-Hodgkin lymphoma (ORadj = 5.9, 95% CI = 4.3–8.1) and cancer of the cervix (ORadj = 1.6, 95% CI = 1.3–2.0). No difference was found in HIV seroprevalence for any of the B-cell related morphological categories of non-Hodgkin lymphoma when compared to the rest of the NHL morphological categories, except for (adult) Burkitt lymphoma, where 11 out of 12 cases were HIV-1 positive (crude OR = 103, 95% CI 13–798) (Table III). An increased risk was also found for Hodgkin disease in association with HIV-1 infection (Table II, ORadj = 1.6, 95% CI = 1.0–2.7).
Table III. Distribution of HIV Infection According to Morphological Subtypes in Selected Cancers
|Description||ICD-O-2 Morphology code||HIV|
| Not otherwise stated (NOS)||9650||93||19|
| Hodgkin disease, mixed cellularity NOS||9652||14||6|
| Lymphocyte depletion NOS||9653||0||1|
| Lymphocytic depletion, reticular||9655||6||0|
| Lymphocytic predominance NOS||9657|
| Nodular lymphocyte predominance||9659|
| Nodular sclerosis, NOS||9663||11||4|
| Nodular sclerosis, lymphocytic predominance||9665||1||0|
|Non-Hodgkin Lymphoma B-cell|
| Malignant lymphoma NOS||9590||14||13|
| non-Hodgkin, NOS||9591||64||50|
|Non-Hodgkin Lymphoma Other|
| Diffuse NOS||9595||2||7|
| Large cell diffuse, NOS||9680||12||17|
| Large cell cleaved, diffuse||9681||0||1|
| Centroblastic diffuse||9683|| || |
| Immunoblastic, NOS||9484||13||0|
| Lymphoblastic||9685|| || |
| Burkitt lymphoma||9687||1||11|
| Small cell lymphocytic, NOS||9670|| || |
| Diffuse small cleaved cell||9672|| || |
| Lymphocytic, diffuse intermediate differentiation||9673|| || |
| Centrocytic lymphoma||9674||18||0|
| Follicular, NOS||9690|| || |
| Lymphocytic, well differentiated nodular||9693|| || |
| Mycosis fungoides||9700|| || |
| Waldenstrom's macroglobulinemia||9761|| || |
| Acute myeloid||9861||75||8|
| Chronic myeloid||9863||95||10|
| Acute promyelocytic||9866||6||1|
| Myeloid NOS||9860||17||0|
| Acute myelomonocytic||9867|
| Chronic myelomonocytic||9868|
| Acute NOS||9801||8||1|
| Acute lymphoid||9821||24||1|
| Chronic lymphoid||9823||53||1|
| Hairy cell leukaemia||9940||2||1|
| Leukaemia NOS||9800|| || |
| Chronic NOS||9803|| || |
| Lymphoid NOS||9820||20||0|
| Prolymphocytic||9825|| || |
| Acute megakaryoblastic||9910|| || |
| Acute myelofibrosis||9932|| || |
| Adult T-cell leukaemia||9827||0||1|
| Sarcoma NOS||8800||25||5|
| Aggressive fibromatosis, malignant||8821||3||1|
| Liposarcoma NOS||8850||9||1|
| Embryonal rhabdomyosarcoma||8910||1||1|
| Synovial sarcoma NOS||9040||5||2|
| Malignant tumour, spindle cell type||8004|| || |
| Neuroendocrine carcinoma||8246|| || |
| Extra-adrenal paraganglioma, malignant||8693|| || |
| Fibromyxosarcoma||8811|| || |
| Abdominal fibromatosis, malignant||8822|| || |
| Fibrous histiocytoma malignant||8830|| || |
| Liposarcoma well differentiated||8851|| || |
| Pleomorphic liposarcoma||8854|| || |
| Leiomyosarcoma NOS||8890|| || |
| Epithelioid leiomyosarcoma||8891||38||0|
| Rhabdomyosarcoma NOS||8900|| || |
| Synovial sarcoma biphasic||9043|| || |
| Clear cell sarcoma||9044|| || |
| Epithelioid hemangioendothelioma, malignant||9133|| || |
| Myxoid chondrosarcoma||9231|| || |
| Neurofibrosarcoma||9540|| || |
| Neurilemmoma, malignant||9560|| || |
| Leukaemia, eosinophilic||9880|| || |
| Spindle cell sarcoma||8801|| || |
| Giant cell sarcoma||8802|| || |
| Dermatofibrosarcoma NOS||8832||0||3|
| Plasmacytoma NOS||9731||9||2|
| Multiple myeloma||9732||169||8|
| Plasma cell leukaemia||9830||1||0|
With regard to cancers where there has been some evidence in the published literature of an increased risk in relation to HIV-1 infection, significant increased risks were found in relation to cancers of the anogenital organs (ORadj = 2.2; 95% CI = 1.4–3.3). The HIV seroprevalence in subjects with anogenital cancer was similar across the different organs: anal (5/28, 18%), penile (4/21, 19%), vaginal (2/10, 20%) and vulval cancer (24/98, 25%). An association was observed between HIV-1 infection and squamous cell carcinomas of the skin (ORadj = 2.6, 95% CI = 1.4–4.9). Eleven out of 70 squamous cell skin carcinomas were reported in the pathology reports to be in close proximity to the anogenital region. Two of these subjects were infected with HIV-1. No associations were noted between HIV-1 and cancer of the oral cavity, liver, lung, melanoma, other (non-Kaposi) sarcomas or myeloma. An inverse association was observed between HIV-1 and stomach cancer (ORadj = 0.4, 95% CI = 0.2–1.0) and HIV-1 and leukemia (OR = 0.6, 95% CI = 0.4–1.0), but in both the association was of borderline significance (p = 0.05). Of the 22 subjects diagnosed with leiomyosarcoma (Table III), none were HIV-1 positive. No cases of conjunctival cancers were available for analysis in this series. None of the 12 testicular cancers (in the comparison group) were HIV-1 positive.
Figure 1 and Table IV show the prevalence of HIV-1 and risk (OR) by age group. The risks of developing AIDS-related cancers (KS, cervical and non-Hodgkin lymphoma) in relation to HIV infection appeared similar across age groups (p-interaction = 0.2).
Table IV. Age Specific Risks (OR1 and 95% CI) between HIV-1 and HIV/AIDS Related Cancers
|Age group||Kaposi's sarcoma||Non-Hodgkin lymphoma||Cervical cancer|
|18–34||58.6 (27.6–124.5)||6.6 (3.7–12.0)||1.5 (0.9–2.4)|
|35–54||55.9(32.4–96.4)||7.1 (4.6–11.2)||1.7 (1.3–2.2)|
|55+ years||24.1 (9.8–59.2)||4.5(2.1–10.0)||1.9 (1.2–3.0)|
|χ2 interaction of risk by age group(3 d.f.), p||4.7, 0.2||1.5, 0.6||1.5, 0.7|
- Top of page
- Statistical analysis
This study reconfirms the strong association observed between HIV-1 and both Kaposi's sarcoma and non-Hodgkin lymphoma, as found previously in this case–control study11 and in other similar studies from Africa.9, 23, 24
Indeed, only two female and nine male cases of histologically verified Kaposi's sarcoma were reported in South Africans aged 15- to 34-years by the National (pathology based) Cancer Registry in 1988,25 increasing to 167 female and 177 male cases in the latest report in 1999.22 Kaposi's sarcoma comprised about 40% of all cancers in a rural area in Kwa-Zulu / Natal in 2001–2002,26 compared with 0.4% of all cancers in males in “Northern” KwaZulu-Natal in the mid-1960s.27 Although the risk of developing KS in older age groups appears lower, there is no effect of interaction by age in this study, p = 0.19). The 5-fold increased risks for developing non-Hodgkin lymphoma in HIV-1 individuals is consistent with other studies from Africa, but again an order of magnitude lower to what is observed in developed countries. For the first time in this series an association between HIV-1 and Hodgkin lymphoma was observed. Aside from the strong association between adult Burkitt lymphoma and HIV-1 (100-fold increased risk) no other morphological type of lymphoma shows any striking association, including those with differing broad morphological types of Hodgkin's disease. We previously argued that infection with Epstein Barr virus (known to cause certain types of lymphoma) or other related viruses may cause greater morbidity if acquired later in life.11 In a separate analysis of a subgroup of subjects from this study, seroprevalence rates of herpesviruses 1 to 6 was over 90%,28 suggesting that in this setting other related viruses may be acquired very early in life.
A causal relationship between HIV-1 and cervical and anogenital cancers has been difficult to establish because the effect of confounding by sexual lifestyle factors could not be ruled out.2 However, since the review by IARC (and a previous study in Johannesburg29 showing a nonsignificant risk of cervical cancer in HIV positive individuals), relative risks of developing cervical cancer in developing countries have ranged consistently between 1.2 and 2, compared to 5–10 in western countries.4, 30 In studies from Africa, stronger associations have been observed between HIV-1 and preinvasive cervical lesions than with invasive cervical cancer,31 the lower risks observed in association with cervical cancer in Africa assumed to be due to AIDS-related competing comorbidities.32, 33 In this study, the association between HIV-1 and cervical cancer appeared the same across all age groups; there is no significant interaction with age, after adjustment for year of diagnosis (p-interaction = 0.6, Table IV). Increases in the incidence of cervical cancer have not yet been noted, but if the association between HIV-1 and cancer of the cervix is causal, then in the absence of a comprehensive cervical screening program, one should expect some increases in the future incidence of this condition. Likewise, if the association between HIV-1 and cancer of the other anogenital organs is causal, then one should also expect small increases in incidence in these conditions, albeit that these are rare (in black males and females, cumulative risks (0–74 years) in 1999 were 0.2 and 0.2%, respectively.22
This study showed an association between HIV-1 and squamous cell skin cancer, in keeping with other studies of HIV-infected individuals.8 To our knowledge, very few cancer registries worldwide collect information on squamous cell skin cancers. The South African National pathology-based Cancer Registry collects information on incident squamous cell skin cancers; however, there is no special code to discern specific anatomical sites such as the perianal region. Cumulative risks (0–74 years) for squamous cell skin cancer in black males and females were 0.13 and 0.21%, respectively, in 1993–1995 increasing to 0.4 and 0.25%, respectively in 1999.26 Further investigation of the topographical site of squamous cell skin cancers in pathology series “pre-” and “post”-HIV/AIDS may be warranted.
We have attempted to subdivide major cancer sites or types into smaller groups, being mindful that these subdivisions would diminish our statistical power to discern an association. No significant differences in HIV-1 risk were found between subjects with different oral cancer types, different anogenital types (data not shown) or between the different morphological types of leukemia.
In contrast to some data from developed countries, no association was found between HIV-1 and cancers of the oral cavity, liver, lung, other sarcomas, melanoma or myeloma. The borderline negative association between HIV-1 and stomach cancer and HIV-1 and leukemia may be due to chance and requires confirmation in independent studies.
The use of hospital-based case–control studies in etiological studies is known to have certain inherent limitations. This study has been designed to assess the relative importance of a number of lifestyle factors including HIV infection on the leading cancers, and so differential interviewer or subject recall bias is minimized. Subjects were selected by systematically identifying eligible patients from clinical records: over 90% of cancers were microscopically verified. The cancer profile of the study subjects resembles the profile that would be expected from national pathology-based cancer registration. The comparison group for these analyses comprised subjects with cancer not thought (from the current assessment of the literature) to be associated with HIV or an infectious agent. Indeed, the prevalence of HIV-1 in this comparison group resembles the prevalence found in other population-based studies. An additional noncancer control group of subjects with vascular disease was used in this analysis, and no difference in the adjusted HIV-1 seroprevalence was observed between the cancer and vascular disease controls.
In summary, in keeping with other studies from Africa, in HIV-1 positive subjects, the risk of developing Kaposi's sarcoma was about 50-fold higher, six-fold higher for developing non-Hodgkin lymphoma and about 1.5- to 2.5-fold higher for developing cervical, anogenital, squamous cell skin cancers and Hodgkin lymphoma. As observed previously in Africa,10 these risks remain about an order of magnitude smaller than what has been observed in developed countries. None of the other cancer types examined (lung, melanoma, oral, liver, stomach, leukemia and sarcomas) showed a significant increased risk in relation to HIV-1.
In the time period of this study, highly active antiretroviral therapy (HAART) was not routinely available in South African public hospitals. In regions of the world where these regimens have become available, the incidence of Kaposi sarcoma, primary central nervous system lymphoma and non-Hodgkin lymphoma has declined.34 As some HAART has now become available to public hospital patients in South Africa, it will be interesting to examine future cancer patterns among HIV positive patients participating in this ongoing case–control study.