KS was endemic in East and Central Africa before the AIDS epidemic accounting for 5-18% of cancers 9 10 . During the AIDS epidemic, KS has become the most common cancer (Table 1) 52 53 54 . In Uganda, the annual age-standardized incidence rate per 100,000 men rose 12 times from 3.2 in 1960-66 to 39.3 in 1995-97 and from 0.1 to 21.8 per 100,000 women during the same periods (Figure 1) 52 . The proportion of KS in childhood cancers increased from 2% in the 1960s to 33% in the 1990s. The risk for KS with HIV/AIDS was elevated 6 times relative to KS rates in the contemporaneous general population in the Uganda HIV/AIDS Cancer match Study (Figure 2, SIR using contemporaneous, AIDS era, comparison data) 46 . The relatively modest increase in KS incidence in Africa compared to that observed in the West may be due to high background rates of KS and of HIV in the general population. In support for this explanation, the risk of KS in men, women, and children was much higher when the general population in the pre-AIDS KS years of 1960-1971 were used (Figure 2, SIR using pre-AIDS KS population incidence rates) as comparator. Another notable impact of HIV on KS was the loss of the large gender disparity in male/female incidence ratio from 20:1 in endemic KS to 2:1 in AIDS-related KS. The KS gender disparity (observed in classical KS as well) is not explained by differences in KSHV seroprevalence in men and women, which differ only by 20-50% 55 . Loss of gender disparity in the setting of immunosuppression suggests that immunocompetent women may be protected by gender-specific immunobiological factors 56 , but the nature of these factors is unknown. A hypothesis that women might be protected by female hormones was advanced, but it not supported by reports that KS occurs in pregnant women 9 57 and laboratory studies have failed to find sex-hormone receptors on KS tissues 58 . KS risk is linked to environmental factors 59 60 , KSHV viremia 61 62 , expression of HIV proteins 63 , immune reconstitution syndrome 64 and severe immunosuppression, some of which may contribute to gender disparity. Recently, Ruocco et al., 65 has advanced the quinine or "oncodrug" hypothesis for KS in Africa, but this remains to be tested, and it would not explain the gender disparity. AIDS-related KS is a unique model of the relationship between viral infection, immunity, environmental, and genetic factors in viral cancers. Studies of KS patterns during the rollout of cART programs to document changes in the incidence and male/female ratio and the association, if any, with antimalarial use, may provide valuable insights into the biology of KS in Africa.

In contrast to KS, the risk for NHL in the general population in sub-Saharan Africa is relatively low 66 , except for childhood BL, which is endemic in some countries where it may accounts up to 50-75% of childhood cancers. Whether HIV has increased the risk of NHL risk in sub-Saharan Africa 67 68 69 , especially of BL in particular, is unclear. The cumulative lifetime risk of NHL for AIDS patients in sub-Saharan Africa is about 3%, which is lower than about 10% observed in more developed countries at the onset of the epidemic 32 . In his autopsy study of 247 adult (>14 years) patients dying from AIDS-related conditions during 1991-1992, Lucas et al., 70 found NHL in only 2.8% of HIV-positive decedents in Côte d'Ivoire. He estimated that the crude incidence of NHL was 84/100,000 per year among HIV-positive adults, about 10 times greater than the expected pre-AIDS incidence of NHL. None of 78 autopsied HIV-positive children (median age = 17 months) had NHL. Case series data of 26 adults (aged >16 years) with BL in Kenya during 1992-96 were compatible with an increase three times that expected from previous estimates 71 . The median age of HIV-positive patients with BL was 35 years compared with 16-25 years for those who were HIV-negative, and the HIV positive cases presented with lymph node involvement, indicating a departure, most likely as a result of the HIV epidemic, from the presentation typically seen in endemic BL. Recent reports from the Kampala Cancer Registry in Uganda have noted that NHL incidence rates increased three times from 2.3 per 100,000 persons in 1961 to 6.6 in 1997 (Figure 1) 52 . This increase was mostly due to pediatric BL (from 0.9 to 3.8 over the time period) and to DLBCL in young adults, which is compatible with an increase due to AIDS, but less so than observed among HIV-positive adults in industrialised countries.

The risk of NHL with HIV has been quantified in a few studies (Table 1). In Rwanda, Newton et al. 72 found an odds ratio (OR) for HIV of 12.6 (95%CI, 2.2-54.4), based on 19 clinically and histologically diagnosed NHL cases. In South Africa, Sitas et al. 73 found an OR for HIV of 5.0 (95%CI, 2.7-9.5), based on 105 histologically confirmed adult NHL cases compared with 844 hospitalized control subjects who had cancers unrelated to HIV (in adult men) or vascular disease (in adult women), enrolled at a tertiary hospital. These results have been confirmed in more recent analyses including larger numbers from South Africa 74 . The SIR for NHL was elevated 6.7 (95%CI, 1.8-17) times than in the general population in the Uganda HIV/AIDS-cancer match study 46 . However, the finding of marginal or null association with HIV has been reported at least one study in Uganda: OR of 2.2 (95%CI, 0.9-5.1), based on 38 histologically confirmed cases 51 .

The reasons for the comparatively lower risk of NHL in sub-Saharan Africa relative to rates observed in the West may be artifactual due to under-diagnosis or may be due to competing mortality or environmental/genetic factors. Survival with HIV may be too short due to competing mortality from common infections, such as malaria and tuberculosis 75 , to permit the significant development of NHL. This reason was supported by findings of fewer subjects with severely depressed CD4⁺ counts in sub-Saharan Africa than in the West 76 . However, report of a median survival of 10 years from HIV sero-conversion in a longitudinal data from a rural cohort in southwest Uganda 77 is comparable to survival observed in western countries before effective antiretroviral drugs were introduced and the individuals have had the opportunity to develop NHL. Under-diagnosis, especially the limited availability of pathological diagnosis, may contribute 78 . Might racial factors contribute? Generally, NHL incidence is lower in African-American men and women than in white men and women 28 , perhaps because of differences in environment, health care access and/or genetic factors. Possibly, the lower NHL risk with HIV in sub-Saharan Africa may reflect part of a general pattern of lower NHL risk in Blacks.

Is it possible that decreased risk of NHL might be due to prevalent exposure linked to treatment of common infections in the region? For example, antimalarials, including chloroquine, are widely used in Africa through formal prescription as well as self-medication for fever is prevalent 79 . In this regard, we are intrigued by recent animal data showing that administration of antimalarials to mice transgenic for c-myc inhibits B cell lymphomagenesis 80 . We are also intrigued by the widely accepted view that immune activation may contribute to HIV/AIDS-related lymphomagenesis 81 . If so, we would expect a high incidence of NHL in sub-Saharan Africa because immune activating infections, such as malaria, tuberculosis, are common in Africa. We observe the opposite. Is it possible that treatments for these conditions, by reducing immune activation, may coincidentally lower the risk of NHL as well? This hypothesis is compatible with recently published findings of reduced HIV-induced immune activation in a Ugandan cohort receiving antitubercular therapy 82 . Clearly, firm conclusion on these issues must await new data on the impact of HIV on NHL that will emerge from new studies that are being conducted with improved diagnostic and data capture methods.

Cervical cancer is the most common cancer in women in most countries in sub-Saharan Africa 47 73 , but the impact of HIV on invasive CC is unclear. Screening programs for CC in sub-Saharan Africa are not well developed, so we would not expect to be able to readily demonstrate HIV-related increase in CC incidence. No dramatic increase has been noted in registry-based studies and clinical reports do not suggest a dramatic increase in number of cases, or drastic change in mean age incidence or disease stage at diagnosis 83 . The associations between CC and HIV have been small or null. For example, in Uganda, Newton et al., 47 observed an HIV prevalence of 32% in 65 women with CC attending four large referral hospitals in Kampala versus 21% observed in 112 controls with non-HIV related cancers or noncancerous conditions recruited at the same hospitals (OR, 1.6; 95% CI, 0.7-3.6). Similarly marginal results have been reported in a study conducted in South Africa, where the prevalence of HIV among 1323 cases of CC was 12.6% versus 9.0% observed in a comparison group of women hospitalized with a mixture of non-HIV-related cancers or vascular disease (OR, 1.6; 95% CI, 1.1-2.3) 47 73 . The risk for CC was 2.4 times (95% CI, 1.1-4.4) in women with HIV/AIDS in Kampala compared with women in the general population in the same area in the Uganda HIV/AIDS cancer match 46 . These results are compatible with a modest elevation in CC risk with HIV, and the conclusion that the impact of immunosuppression on CC risk is likely small.

The prevalence of cervical intraepithelial neoplasia (CIN) or human papilloma virus (HPV) infection is elevated 2-6 times in HIV-positive women than in HIV-negative women in east 84 85 , west 86 87 , and southern Africa 88 , although not all studies agree 89 90 91 . In their study of Nairobi prostitutes in 1992, Kreiss et al., 89 found cervical HPV DNA in 37% of HIV infected women versus 24% in HIV non-infected women (OR, 1.7, 95% CI 0.8-3.6). The OR for HPV DNA in women attending an antenatal clinic in Mwanza in Tanzania was 1.02 (95%CI, 0.6-1.6) 90 . The link between squamous intraepithelial lesions (SIL) and immunity has not been fullu characterized 92 . The risk of low grade SIL was 6.1 (95%, CI = 1.2-41.4) times elevated in women with CD4+ cell count < 200/mm³ compared to those with higher counts 92 , although these findings were based on 20 subjects with SIL. In a larger study including 710 HIV positive women in Rwanda, Anastos et al., 93 , found cervical HPV DNA (HPV 16 contributed 14%) in 67% of the women, of whom 8.8% also had CIN grade 3. Detection of HPV other than 16 was inversely associated with CD4 counts. Interestingly, a positive association between malaria infection in the past 6 months and risk of CIN3, which has not been reported before, was observed. The risk of high grade SIL was 2.4 times higher in HIV positive women with <200 CD4/ml than HIV positive women with >500 CD4/ml in a study of 1,010 HIV positive women in South Africa 94 . In this study, detection of HPV 16 and 66 were inversely related to CD4 count. Taken together, the modest increase in CC with HIV is compatible with a small impact due to HIV, but the relationship with immunity is observed only with pre-invasive CC lesions.

Despite the small risk increases reported, CC is the most important cancer in HIV infected populations in Africa because women account for >50% of HIV epidemic. The public health impact of CC could be addressed by harnessing the historical interest and increased funding of HIV/AIDS treatment and prevention programs to support new initiatives for CC screening and treatment 95 . This approach has been attempted in Zambia with great success 95 96 . The Zambian model has brought 58,000 women, including HIV negative women, for CC screening up from 0 less than 5 years ago by leveraging available, broad-based capacity-building efforts of vertical HIV/AIDS care and treatment programs. These modest programs are saving about one CC death per 46 HIV-positive women screened, demonstrating potential public health benefit. The program will yield timely data on HPV infection and risk for cervical dysplasia among HIV-infected women in Africa 93 , and might make it possible to examine novel hypotheses, such as the interaction between malaria and CC 93 .

Squamous cell carcinoma of the conjunctiva (SCCC) is a rare tumor of the ocular surface, which is linked to ultraviolet radiation exposure and, based on elevated risk with HIV 97 , appeara to be etiologically linked to immunosuppression and/or an underlying, albeit not yet well-characterized infection. The link with HIV infection was first reported by Ateenyi-Agaba in 1995 50 when he observed that 75% of patients with SCCC at Mulago Hospital in Uganda were HIV-seropositive compared with only 19% of cases with nonmalignant eye conditions. This finding was confirmed in other case-control studies conducted in the tropics and subtropics 97 98 , but it has not been reported in South Africa 74 , suggesting that ultraviolet exposure may be a necessary component of the HIV impact. The risk of SCCC in Uganda 15 times from before to during AIDS. The incidence rate of ocular tumors, which are mostly due to SCCC, in Kyadondo County rose from 0.2 per 100,000 person-years in 1960-66 to 3.0 in 1995-97. The proportion of SCCC in eye tumors during the same period increased from 23.5 to 71% in men and 0 to 85% in women 52 . The dramatic increase in SCCC focused attention on the role of immunity and infection in SCCC, prompting a search for infectious etiology, including of mucosal or genital high-risk and cutaneous HPV 99 100 101 102 103 . de Koning et al. have summarized the comprehensive literature on the association between HPV and SCCC 103 The studies conducted have reported a higher prevalence of cutaneous HPV, but not genital HPV types, in SCCC (OR ranges 8 - infinity) 103 . The frequency or load of HPV DNA does not appear to vary with histological grade of tumor 103 . The studies conducted thus far remain are relatively small, and the positive results may be biased, while differences in laboratory methodology for HPV detection is a limitation likely shared by all the studies. Nonetheless, the consistent finding of elevated risk of SCCC with HIV supports the hypothesis that a known or novel HPV or other infectious agents may be involved in SCCC etiology, while the absence of impact in South Africa, suggests effects of immunosuppression are expressed on a background of exposure to ultraviolet radiation.

Hodgkin lymphoma, although not designated as AIDS defining, is consistently elevated with HIV in most studies 15 104 . The risk was elevated in the Uganda HIV/AIDS cancer match study (OR, 5.7; 95%CI, 1.2-17) and in a case-control study conducted in South Africa (OR, 1.4; 95%CI, 1.0-2.7). Interestingly, some 105 but not all studies 106 conducted in the West have reported that risk for Hodgkin lymphoma with use of cART. Whether Hodgkin lymphoma risk will further increase as cART becomes widely available in sub-Saharan Africa will be clarified by follow-up studies of cohorts on cART in Africa.

Hepatocellular carcinoma (HCC) was relatively common in men in Africa before the AIDS epidemic, in part, because of the high prevalence of hepatitis B virus (HBV) infection and exposure to aflatoxin 66 . There is no evidence that HCC risk has increased during the AIDS epidemic, although data for deep seated tumors in Africa should be considered largely incomplete. The rates of liver cancer in Uganda, one of the first countries to be touched by the AIDS epidemic 107 , were stable among men, but they increased by 50% among women from 1960-80 to 1991-05 108 . Given the high prevalence of chronic HBV infection, aflatoxin exposure, and to a lesser extent hepatitis C infection in sub-Saharan Africa 109 110 , a surge in HCC cases might have been expected. This reasoning is supported by recent data suggesting that HIV might be associated with significant liver fibrosis 111 . However, as liver cancer has a long induction period, the associations with HIV might not be expected early in the HIV/AIDS epidemic, and lower risk might be related to under-diagnosis, non-pathological confirmation, and short survival. Continued surveillance of HCC might provide improved understanding of the impact of HIV in HCC risk in sub-Saharan Africa.

Evaluation of cancer statistics must focus on the quality of data available to support scientific and public health initiatives. Although based on the best data currently available, the impact of HIV on the incidence and burden of cancer is probably underestimated. Only about one third of people with HIV in sub-Saharan Africa know about their infection 112 and only about one third of those who need HIV-specific treatments are receiving them 46 113 . In addition, cancer registration is relatively underdeveloped 114 . Although record-linkage methods, which have been used effectively and efficiently in the West, are feasible in Africa 46 , lack of high-quality computerized medical records 115 precludes use of some other effective methods.

The study of cancer in HIV persons in Africa is valuable for several reasons. First, the large size of the HIV epidemic underscores its public health significance, including in its impact on cancer. Two, Africa encompasses extraordinary genetic diversity of pathogens and hosts, which hold promise for unique opportunities to learn about the biology of infection, immunology and cancer. Two HIV types are relevant to the epidemic (HIV-1 and HIV2), but HIV-1 is responsible for 95% of HIV infections globally. HIV-1 is divisible into ten subtypes (A-H, J, K) and some circulating recombinant forms are recognized 116 , which have different transmission potentials and pathogenesis and could, plausibly, be associated with differential risk of cancer, and vary by geography. HIV-2 infection has a restricted distribution confined to West Africa, where co-distribution with HIV-1 offers opportunity to investigate HIV-type specific effects. For example, one small study including 40 women found that HIV-2 was associated with increased HPV clearance 117 and women infected with HIV-2 were less likely to develop high-grade SIL (HR, 0.3; 95%CI, 0.1-0.9) than those infected with HIV-1 118 . This finding should be interpreted with caution because the small size of the study, and it was not statistically significant when CD4 counts were taken into account. HIV-1, subtype C accounts for about half (48%) of all global HIV-1 infections 116 , and is the major subtype in southern Africa. Subtypes A and D are predominant in equatorial sub-Saharan Africa, where they account for 12% and 2%, respectively, of HIV-1 infections. Subtypes B and G account predominate in the West and the Far East Africa and they account for 11% and 5%, respectively, of HIV-1 infections. Inter-HIV-1 subtype recombinants are playing an increasingly important role. They account for 20% of infections occurring in all regions. There are rare subtypes such as F, H, J, and K, which together account for <1% of infections 116 . Most current knowledge about the link between HIV and cancer risk is based on populations with subtype B, highlighting our still very limited knowledge of interactions between HIV, host immunogenetics, and cancer.

Globally coordinated efforts aimed at interrupting the spread of, and mortality from, HIV/AIDS in sub-Saharan Africa present opportunities to study cancer in individuals with HIV/AIDS on the continent 19 . Millennium Development Goal (MDG) 6, established by the UN General Assembly Special Session in 2000 to halt and reverse the spread of HIV-1, malaria, and other diseases, led to the creation of the Global Fund for AIDS, Tuberculosis, and Malaria and unprecedented opportunities for funding. The US President's Emergency Plan for AIDS Relief (PEPFAR) launched shortly thereafter provided additional funding streams for HIV prevention and clinical care 119 120 . These initiatives, which largely ignore cancer 121 , have created a funding stream that has seen access to HAART increase from <1% in 1999 to 35% in 2009 19 and resulted in dramatic declines in the rate of new infections, stabilization of the epidemic, reduction of AIDS-related mortality, and increase in life expectancy 122 . This funding stream has strengthened infrastructure for HIV/AIDS disease surveillance, diagnosis, and data capture and established large or networked cohorts 115 123 . The resulting infrastructure and cohorts could be converted to provide new data on cancer in sub-Saharan Africa. For example, more than 13,000 PEPFAR clinics provide HIV prevention and treatment to millions of individuals in sub-Saharan Africa. Similarly, the International Epidemiology Databases to Evaluate AIDS (IeDEA), a global consortium funded by the National Institutes of Health, has linked 183 clinics in 17 countries in sub-Saharan Africa serving 286,793 individuals 124 . These large linked clinics offer three broad-based opportunities: to define: a) the spectrum of common and rare cancers in HIV-infected individuals and to obtain precise estimates of risk and heterogeneity of risk; b) temporal trends on common and rare cancers, including before and after introduction of cART; and c) test specific hypothesis, such as investigating the risk of KS with immune reconstitution syndrome 64 125 , the role of antimalarials in NHL or other cancers, and the role of genetic and/or viral co-infections in cancer 20 , or assess impact of interventions. We think the opportunities may be best considered along the lines of a resource-focused and disease-focused approach, as discussed below.

The resource-focused approach would aim to leverage large networked or linked clinics 124 126 127 to routinely collect data that can be used for multiple studies and purposes. In the West, efforts such as the Swiss Cohort Studies 128 and the U.S. HIV/AIDS Cancer Match Study 29 provide useful models to consider. Because cancer is rare even in people with HIV, combining data from different cohorts allows investigators to address questions pertaining to HIV/AIDS that cannot be answered in small single institution cohorts. We are encouraged by establishments of such cohorts in Africa, prominently including the IeDEA consortium (http://www.iedea-hiv.org), which was launched with funding from the National Institutes of Health, to collect, harmonize, and standardize data from the continent to allow comparative analysis of common and dissimilar impacts 129 . The African Organization for Research and Training in Cancer (AORTIC) (http://www.aortic.org/) has a mission is to promote cancer awareness and improve cancer diagnosis and treatment in Africa. With its network across the continent and bi-ennial scientific meetings, AORTIC provides a resource that could be leveraged to initiate, implement, and report on Africa-wide studies of cancer. The recent focus large biomedical centers, such as the National Cancer Institute at NIH, by establishing Centers for Global Health 130 is timely and likely to speed up the conversion of infectious disease-specific programs in developing countries into programs for study and control of cancer.

In contrast to the resource-focused approach, in the disease-focused approach, specific hypotheses are raised and specific studies designed to answer those questions. For example, the question of the types and risk of NHL is currently being addressed by a consortium of investigators focused on lymphomas (http://www.ssalc.org/acsr_drupal/). These investigators will bring clarity to the question of the spectrum of lymphomas diagnosed with HIV in Africa and their diagnostic support is likely to spur epidemiological and clinical studies of NHL. We noted outstanding questions about the etiology of SCCC 97 99 . Disease-specific hypothesis-driven studies to investigate the role of immunosuppression, ultraviolet exposure, pathogens in the pathogenesis of the disease will bring clarity to our understanding of SCCC. Other interesting questions that could benefit from a disease-focused approach include understanding the association between HIV and lung and breast cancer. The risk for lung cancer is increased with HIV in the West, but given the much lower prevalence and intensity of cigarette smoking in sub-Saharan Africa, studies conducted there might shed light on the biology of the disease among non- or low-intensity smokers. The risk of breast cancer is decreased with HIV in the West 43 . Hessol et al. and co workers 44 have linked low breast cancer risk with HIV to infection with CXCR4-using variants of HIV. They speculated that CXCR4-using variants might reduce breast cancer risk by binding to apoptosis receptors on hyperplastic and neoplastic breast duct cells. This hypothesis brings new perspectives to the question about the biology for breast cancer with potential for treatment or intervention using CXCR4-agonists. Data about breast cancer in sub-Saharan Africa are sparse with some reporting increased risk 131 132 and others no increase 46 . Thus, clarifying the epidemiology of breast cancer in sub-Saharan Africa where CXCR4-using variants are prevalent may help evaluate the feasibility of testing and confirming or refuting this hypothesis. Finally, anal cancer is increased with HIV in the West 41 . Homosexual practices are being increasingly acknowledged in Africa, opening opportunity to investigate the association.