The Spectrum of Cancers in West Africa: Associations with Human Immunodeficiency Virus

Background Cancer is a growing co-morbidity among HIV-infected patients worldwide. With the scale-up of antiretroviral therapy (ART) in developing countries, cancer will contribute more and more to the HIV/AIDS disease burden. Our objective was to estimate the association between HIV infection and selected types of cancers among patients hospitalized for diagnosis or treatment of cancer in West Africa. Methods A case-referent study was conducted in referral hospitals in Côte d’Ivoire and Benin. Each participating clinical ward enrolled all adult patients seeking care for a confirmed diagnosis of cancer and clinicians systematically proposed an HIV test. HIV prevalence was compared between AIDS-defining cancers and a subset of selected non-AIDS defining cancers to a referent group of non-AIDS defining cancers not reported in the literature to be positively or inversely associated with HIV. An unconditional logistic model was used to estimate odds ratios (OR) and their 95% confidence intervals (CI) of the risk of being HIV-infected for selected cancers sites compared to a referent group of other cancers. Results The HIV overall prevalence was 12.3% (CI 10.3–14.4) among the 1,017 cancer cases included. A total of 442 patients constituted the referent group with an HIV prevalence of 4.7% (CI 2.8–6.7). In multivariate analysis, Kaposi sarcoma (OR 62.2 [CI 22.1–175.5]), non-Hodgkin lymphoma (4.0 [CI 2.0–8.0]), cervical cancer (OR 7.9 [CI 3.8–16.7]), anogenital cancer (OR 11.6 [CI 2.9–46.3]) and liver cancer (OR 2.7 [CI 1.1–7.7]) were all associated with HIV infection. Conclusions In a time of expanding access to ART, AIDS-defining cancers remain highly associated with HIV infection. This is to our knowledge, the first study reporting a significant association between HIV infection and liver cancer in sub-Saharan Africa.


Introduction
In industrialized countries, the advent of antiretroviral treatment (ART) has been marked by a substantial improvement in the duration of life of HIV-positive people together with the decline of AIDS defining illnesses [1,2]. This quantitative and qualitative improvement in life expectancy has led many to consider that non-AIDS defining diseases including cancers would become soon a major cause of morbidity and mortality [3]. Infection with HIV is known to be associated with neoplasms such as Kaposi's sarcoma (KS), non-Hodgkin's lymphoma (NHL) and Invasive Cervical Cancer (ICC) thus contributing to the definition of the AIDS stage [4]. Since the advent of ART, the incidence of these AIDSdefining cancers dropped in northern countries, with the clearest declines being experienced for KS and NHL [5]. Population-based record linkage studies in North America and Europe have identified in parallel several non AIDS-defining cancers as significantly associated with HIV infection [6]. With now more than six million patients on treatment, access to ART and subsequently survival of those treated have considerably increased over the last ten years in sub-Saharan Africa [7,8]. Case-control studies and one record linkage study exploring the association between HIV infection and cancers have been conducted in Eastern and Southern Africa in a time of limited access to ART [9][10][11][12]. They have consistently reported a lower risk of AIDSdefining cancers in people living with HIV compared to pre-ART reports from industrialized countries [13]. Conversely, some non-AIDS defining cancers have been reported as associated with HIV in Africa such as Hodgkin's disease, cancers of the anogenital organs and skin cancers [11]. In many West African countries, the association of cancer with HIV infection remains poorly understood. Furthermore, the distribution of cancers and exposure to carcinogens is different from other parts of sub-Saharan Africa. Indeed, some cancers considered as traditional because of linkage to endemic biological environmental agents are particularly represented in West Africa such as primary liver cancer related to hepatitis B virus infections. Others, such as KS related to Human Herpes Virus 8 infection seems to be less frequent compared to Central and Eastern Africa [14]. Our aim was to document the association between HIV infection and AIDSdefining cancers as well as some non-AIDS defining cancers in West Africa in the ART era.

Design and Study Population
A case-referent study was conducted in the three public referral hospitals of Abidjan, Côte d'Ivoire and in the referral hospital of Cotonou, Benin from October 2009 to October 2011. During the study period, clinical wards likely to manage patients with cancer were asked to include all patients attending with a diagnosis of cancer. In each participating ward, medical referents specifically designated for this study, contacted all adult patients ($18 years old) seeking care with a suspected or confirmed diagnosis of cancer.

Study Conduct
After obtaining patient's written informed consent, preincluded subjects were administered a structured two-page questionnaire assessing socio-demographic characteristics (i.e. age, sex, lifetime number of sexual partners, place of living). The form was also designed to record from medical files clinical, biological, radiological and surgical data leading to the diagnosis of cancer. The medical files were regularly reviewed by medical referents and clinical monitors for diagnostic confirmation according to the latest available information. For diagnostic purposes, a cytological and/or histological examination, if medically indicated, was systematically proposed and financially supported by the research project. Confirmed cancer cases were classified according to their primary anatomical site of cancer (topography) and histological morphology when available using the WHO third edition of the International Classification of Diseases for Oncology (ICD-O-3) [15]. Medical referents were also responsible for proposing a systematic rapid HIV test (DetermineH, Abbott Diagnostics) at the time of interview. They collected capillary blood by a finger prick test, after patient's informed consent was obtained. In case of indeterminate result, the test was repeated once. In case of positive result or if still indeterminate after two attempts, a venous blood sample was collected for confirmation with a Genie2H (Bio-Rad, Marnes-La-Coquette, France) test allowing the identification of patients infected with HIV-1 alone, HIV-2 alone as well as dually infected patients. HIV testing was not repeated if the medical records indicated that the patient had previously been diagnosed HIV-positive. The study was approved by the national ethic committees of Benin and Côte d'Ivoire.

Statistical Analysis
The frequency of HIV infection in selected types of cancer thought or known to be positively associated with HIV infection (i.e. KS, NHL, ICC, cancers of the anogenital organs, oral cavity, pharynx and larynx, Hodgkin lymphoma, squamous cell skin carcinoma, lung cancer, primary liver cancer and leukaemia) was compared to the frequency of HIV infection in a referent group of cancers. This referent group was chosen to be a mixture of cancer cases in order to compare to previously published case-control studies conducted in Southern Africa [10,11]. Our referent group was thus constituted with cancers not reported in the current literature as positively or inversely associated with HIV infection (i.e. colorectal, breast, prostate, pancreas, oesophagus, endocrine system, ovary, endometrium, stomach, sarcomas other than KS, myelomas, kidney, bladder, biliary ducts & gallbladder, melanoma and mesothelioma). An unconditional logistic regression model was used to estimate the association by the computation of the odds ratios (ORs) with their 95% confidence intervals (95% CIs). In multivariate analysis, ORs were adjusted on age taken as continuous variable, gender (except for ICC) and lifetime number of sexual partners (,5 partners versus $5 partners). All statistical analyses were performed using SAS software, version 9.2 (SAS Institute Inc, Cary, NC, USA).
Overall, 710 (69.8%) patients diagnosed with cancer had never been tested for HIV infection prior to the study and 49 of them (6.9%) were diagnosed HIV-positive in the course of the study. Among the 128 HIV-positive patients with cancer, 38.3% had never been tested for HIV infection prior to the study. In the 79 patients already known to be HIV-positive, 25 (31.6%) patients were already on ART, 26 (33.0%) were not currently on ART and for the remaining 28 (35.4%) patients; no information was available concerning ART exposure. Figure 2 summarises the distribution of ART use in all HIV-positive patients combined (i.e. those newly diagnosed, those already known to be HIV-infected) according to the different AIDS and non-AIDS defining cancers. Of

Discussion
In a time of expanding access to ART, AIDS defining as well as some non-AIDS defining cancers were significantly associated with   HIV infection in this multi-country survey in West Africa. KS and NHL were associated with HIV infection, with a similar magnitude of association than in previous case-control studies from sub-Saharan Africa performed prior to the roll out of ART [10,11,16]. In Côte d'Ivoire, access to ART has started in 2002 and is free of charge since 2007 [17]. Three years later, KS and NHL were still highly associated with HIV infection according to our findings. Reports from population-based record linkage studies in industrialized countries have constantly documented a prompt decline in the occurrence of KS and NHL since the advent of ART [5,18] [7]. This first hypothesis is supported by the low frequency of ART use among HIV-positive patients reported in the present study. Second, the association between KS, NHL and HIV infection has been constantly reported to be lower in Africa than in industrialized countries and might have shaded the potential impact of ART [13]. NHL presented with a wide variety of histological subtypes in our series, as already documented elsewhere in sub-Saharan Africa [19]. Burkitt lymphoma was common within the NHL series with a well-documented morphological type, especially in HIV-positive patients, consistent with prior reports from South Africa [11,20]. Unlike other NHL subtypes such as diffuse large B-cell lymphoma or primary brain lymphoma, Burkitt lymphoma does not seem to be influenced by cellular immunity [21,22]. The high frequency of Burkitt lymphoma might contribute to the lower-than-expected association between NHL and HIV report in sub-Saharan African studies, together with a competitive mortality due to communicable diseases such as tuberculosis. In addition, the high frequency of Burkitt lymphoma observed in HIV-positive patients might also challenge the potential positive impact of ART on NHL in the coming years.
According to the international agency research on cancer (IARC) latest estimates in sub-Saharan Africa, ICC is the leading cause of cancer in women regardless of HIV infection [23]. Several genital human papillomavirus (HPV) types have been identified as the necessary cofactor for ICC and its histological precursor and classified as carcinogenic according to the IARC classification [24]. HIV infection and its associated immunosuppression have been repeatedly linked with carcinogenic HPV infection in sub-Saharan Africa [25][26][27]. However, previous casecontrol studies conducted in other parts of Africa found no or weak association between HIV and ICC [13]. In our present study, ICC was strongly associated with HIV infection. A previous case-control study assessing the relationship between HIV infection and ICC was conducted in Côte d'Ivoire from April 1997 to October 1999. The prevalence of HIV infection was 16.7% among 132 patients with ICC compared to 8.3% among 120 patients in the control group, giving an adjusted OR of 3.4 [95% CI 1.4-8.3] [28]. The HIV prevalence reported in their control group was consistent with the HIV prevalence in adult women in Côte d'Ivoire ten years earlier. Conversely, we report today a higher frequency of HIV infection in women diagnosed with ICC. As the majority of these women with ICC was unaware of their HIV status and thus untreated, the high frequency of HIV infection we report is not likely to be related to a prolonged life expectancy while on ART. According to ICC natural history, progression to invasive stages takes years to decades to occur. A latency period is thus expected between the HIV epidemic that peaked during the late 909s in most countries in West Africa and its impact on the occurrence of ICC. While HIV prevalence figures in the general population of adult women in Côte d'Ivoire and Benin have fallen during the last ten years, women with ICC continue to present with a high and sustained HIV prevalence explaining the particularly high association between ICC and HIV infection reported in our study. However, comparing cancer rates by calendar period from case-referent studies conducted in different settings with various methodologies does not allow any definite conclusion on trends in cancer incidence according to HIV status. Assessing the true impact of HIV infection and exposure to ART on cancer trends such as ICC in sub-Saharan Africa will ultimately need HIV and cancer registries matched studies. Cancer of the anogenital organs which share an important risk factor with ICC, namely HPV infection, were highly associated with HIV infection, a finding consistent with a previous report from South Africa [11]. Primary liver cancer has been reported as associated with HIV infection in industrialized countries. Indeed, a meta-analysis of the incidence of non-AIDS cancers according to HIV status from 18 studies reported an increased standardized incidence ratio of 5.6 [95% CI 4.0-7.7] for primary liver cancer [6]. To our knowledge, this is the first study reporting an association between primary liver cancer and HIV in sub-Saharan Africa. Immunosuppression induced by HIV infection might play a direct role in the occurrence of this cancer as suggested by a meta-analysis reporting a higher incidence of liver cancer in both HIV-infected patients as well as transplant patients compared to the general population [29]. In industrialized countries, the prevalence of hepatitis B virus (HBV) and hepatitis C virus (HCV), infectious agents leading to liver cancer, are higher among HIV-infected patients compared to the general population as they share the same routes of transmission (i.e. through blood contamination or sexual relation). In sub-Saharan Africa, especially the West Africa part, HBV infection is highly prevalent and usually acquired during childhood whereas this occurs later in life in industrialized countries. These different transmission dynamics between the two types of settings might directly impact on the risk of HBV infection according to HIV status. Indeed, several studies reported similar seroprevalence figures of HBV infection in HIVpositive adults compared to HIV-negative ones [30,31]. However, the exact impact of HIV infection on HBV infection and the risk of subsequent liver cancer remain to be explored in sub-Saharan Africa. Dedicated case-control studies taking into account the main known risk factors for primary liver cancer (HBV, HCV, as well as aflatoxin exposure) are needed to explore more accurately this association.

Limitations
Our study population might not reflect the exact distribution of cancers occurring in these two countries of West Africa. Since there were no functional population-based cancer registries in the catchment area of the four referral hospitals and no hospital-based cancer registry during the study period we were not able to ensure that all cancer cases were screened during the study period. Additionally, we disregarded patients with cancers attending exclusively the private health sector. Thus, the present report does not pretend to be fully representative of the distribution of cancers in Côte d'Ivoire and Benin. However, the pattern of cancers reported was quite similar to the distribution of cancers reported in the past by the former cancer registry of Côte d'Ivoire and from most recent estimated data reported by the Globocan program for Côte d'Ivoire and Benin [14,23]. Based on multiple sources of information (referral hospitals, government and private hospitals, health centers and private clinics), Echimane et al had indeed registered 1,871 cancers in the urban area of Abidjan during the 1995-97 period. Prostate cancer (15.0%), liver cancer (14.6%) and NHL (10.0%) were the three most frequent malignancies registered in men and breast cancer (25.2%), ICC (23.6%) and NHL (7.2%) the three most frequent malignancies in women, concordant with the cancer distribution reported in our study 15 years later. One could argue that the choice of a referent group constituted only of cancer patients might not be appropriate for HIV prevalence comparisons. A previous South African casecontrol study measuring the association between HIV infection and selected type of cancers has both recruited their control group among cancer patients and patients consulting for cardiovascular disease [11]. They found no significant difference in the prevalence of HIV infection among these two groups of control subjects, after adjustment for age, year of diagnosis and gender. However, we acknowledge that potential biases might have been introduced by using cancer cases as a referent group. Indeed, the HIV prevalence of our hospital-based referent group was higher than the estimated HIV prevalence of 3.4% [95% CI 3.1-3.9] reported in the adult  year old) general population of Côte d'Ivoire in 2009, thus potentially underestimating ORs estimates [32]. However, differences in the age distribution and other characteristics related to hospital subjects limit comparisons between our hospital-based group to the general adult population. The limited number of cancer cases in specific anatomic localizations prevented a precise assessment of the association between these cancers and HIV infection. The extension of such a case-referent study to other referral hospital from other countries with generalized HIV epidemic in West Africa will provide a unique opportunity to explore rarer cancers types suspected to be HIV-related. Finally, the cross-sectional nature of the study prevents from drawing any causal relation between HIV infection and cancer. In a context where the collection of prospective cancer cases is particularly challenging, a cross-sectional approach that could be repeated over time using the same methodology might however provide useful information on cancer trends, paving the path for the development and strengthening of cancer registries.

Conclusion
In a time of expanding access to ART, AIDS-defining cancers remain highly associated with HIV infection in West Africa. Some non-AIDS defining cancers such as cancer of anogenital organs or primary liver cancer are also associated with HIV infection. It is a public health commitment for oncologists and other practitioners in charge of cancer patients to offer systematic screening of HIV infection. Indeed, this practice was not routine prior to the conduct of this study with nearly half of the HIV-positive patients newly diagnosed as a direct consequence of the study. The low proportion of refusals we observed indicates that is now acceptable and workable to perform provider-initiated HIV testing in patients suspected or diagnosed with cancer.