A major feature of these cancers in HIV-infected persons is their association with immunosuppression. For KS and EBV-related NHL subtypes, risk increases as the CD4 count declines (5). This association with CD4 count has not been apparent for cervical cancer (5), although HIV-infected women with low CD4 counts have an elevated risk of persistent HPV infection and progression to pre-cancerous cervical lesions, compared to women with higher CD4 counts (6). Risk of all three AIDS-defining malignancies is also elevated among solid organ transplant recipients, another immunosuppressed population (7).

In developed countries, the availability of highly active antiretroviral therapy (HAART) beginning in 1996 has led to improvements in immunity and declining AIDS-related morbidity and mortality (8-10). HAART substantially reduces risk of KS and EBV-related NHL (11-13). For unknown reasons, KS and NHL incidence rates were falling even during the 1980s and early 1990s among people with AIDS in the U.S., but introduction of HAART in 1996 led to a further drop (Figure 1) (1). As a result of these declines, non-AIDS-defining malignancies now represent a much larger fraction of the overall cancer burden in people with AIDS compared with the pre-HAART era (Figure 1) (1). Among less immunocompromised individuals HIV-infected persons, non-AIDS-defining cancers now comprise the majority of all incident cancers (e.g., in the U.S.: 58% of all cancers, vs. 31% in the pre-HAART era) (14).

Five of these non-AIDS-defining cancers deserve special attention and will be highlighted in this review (Table 1). These malignancies (lung cancer, Hodgkin lymphoma, anal cancer, liver cancer, and non-melanoma skin cancers) are among the most common among HIV-infected persons, and each arises at increased frequency compared with the general population (7). As will be described, a concerning aspect of the epidemiology is that for two of the malignancies (Hodgkin lymphoma and anal cancer), incidence among HIV-infected persons has been increasing in the HAART era.

As reviewed below, the epidemiologic data suggest that HIV may facilitate, at least indirectly, the development of these non-AIDS-defining cancers. For Hodgkin lymphoma, HIV-infected persons present more frequently with extranodal disease and at an advanced stage than HIV-uninfected individuals (16). In contrast, data are lacking for carcinomas on whether the phenotype (e.g., histologic subtype, molecular tumor characteristics, or potential to metastasize) differs in any essential way from the phenotype of tumors in uninfected individuals. For lung cancer and liver cancer, HIV-infected individuals tend to present at a more advanced stage than uninfected individuals (17-20), but this difference could be secondary to poor access to medical care and delayed diagnosis. Likewise, lung cancers in HIV-infected individuals manifest greater genetic instability than those from uninfected persons (21), but this characteristic could merely reflect the advanced stage of the tumors. Prognosis among HIV-infected cancer patients is typically worse than in uninfected cancer patients (17,20,22), which plausibly could reflect advanced stage of cancer diagnosis, differences in cancer treatment, and AIDS-related mortality. One intriguing exception where HIV infection may alter the phenotype is non-melanoma skin cancer. Immunosuppressed solid organ transplant recipients have a greatly elevated risk for developing multiple, highly aggressive squamous cell skin cancers (23), and similar cases have been described in HIV-infected persons (24,25).

A reasonable working hypothesis is therefore that non-AIDS-defining malignancies in HIV-infected persons are not pathologically distinct entities. Rather, these tumors represent the same cancers as seen in uninfected individuals, but they occur at an increased rate among HIV-infected persons. This increased incidence could, in turn, reflect a high prevalence of known cancer risk factors, an independent effect of HIV on progression to cancer, or a biological interaction of HIV with the known risk factors. The following sections review the epidemiologic data and consider some potential biological mechanisms whereby HIV might facilitate development of these malignancies.

The simplest immunologic mechanism to consider is HIV-induced depletion of CD4 cells, leading to deficient cell-mediated immunity. This mechanism is highly relevant for KS and NHL (5). The epidemiologic hallmarks of this mechanism are strongly increased risk of cancer compared to the general population and clearcut associations with markers of immunosuppression, such as CD4 count, time relative to AIDS onset, and the protective effects of HAART. Another hallmark is an elevated risk for the cancer among transplant recipients, who receive chronic immunosuppressive therapy to prevent graft rejection. It has been hypothesized that an intact immune system can control or eliminate early cancer precursor cells, protecting against a wide range of cancers (i.e., the immune surveillance hypothesis (73)). Nonetheless, evidence in HIV-infected persons and transplant recipients is strongest that immunodeficiency predisposes to the subset of malignancies caused by oncogenic viruses, likely because of loss of immune control of viral replication and transformation. Regarding the non-AIDS-defining cancers highlighted in this review, this mechanism may be most relevant for some non-melanoma skin cancers, particularly the rare subtypes such as Merkel cell carcinoma and sebaceous carcinoma, where oncogenic viruses could have a direct transforming role. Although limited data on these skin cancers preclude a definitive assessment, the substantially increased risk compared to the general population, in the absence of another obvious explanation, suggests that HIV-related immunosuppression is relevant.

It is useful to speculate on additional immunologic mechanisms related to HIV infection that could potentially play a role in the development of other non-AIDS-defining malignancies (Table 2). For anal cancer, there are several salient observations: first, low CD4 counts are associated strongly with detection of anal HPV infection and pre-cancerous lesions, but less strongly with risk of anal cancer itself; second, anal cancer risk increases with progressive time relative to an initial AIDS diagnosis; and third, anal cancer incidence has increased during the HAART era and is elevated among individual HAART users. A model that could account for these observations is that HIV-related immunosuppression is relevant at the earliest stages of anal carcinogenesis (such as HPV persistence and development of low-grade squamous intraepithelial neoplasia) but not involved in the later progression to invasive cancer (39). Under such a model, HAART would not result in a benefit if administered after the early steps in this process. Indeed, due to the survival benefits associated with HAART use, persons receiving HAART would live long enough to develop anal cancer. In other words, HAART use increases time spent living with early-stage neoplastic lesions, allowing their progression to invasive cancer. Likewise, the increase in anal cancer risk associated with AIDS onset may not reflect the effects of progressive immunosuppression but rather prolonged time spent living with immunosuppression and duration of HPV-related disease.

Given the rarity of liver cancer, a challenge has been to gather sufficient data to assess the effects of HIV infection and related immunosuppression. One possibility is that deficiency of cell-mediated immunity plays a role in the development of liver cancer by reducing immune control of chronic HCV and HBV infections. However, the situation may be more complicated than for anal cancer. Specifically, for liver cancer, immunosuppression not only predisposes to AIDS-related mortality but also liver-related mortality (56,57,74). In the pre-HAART era, one or both of these competing causes of mortality may have obscured an association between immunosuppression and liver cancer risk.

The elevated risk of lung cancer in HIV-infected smokers probably has a different explanation, since risk is not strongly related to markers of HIV-related immunosuppression such as CD4 count and prior AIDS onset. Also, risk is elevated for all lung cancer subtypes, pointing away from involvement of a single oncogenic virus (and thus, pointing away from immunosuppression). One possibility is that HIV infection promotes the development of lung cancer by causing chronic pulmonary inflammation and repeated lung infections, which could induce DNA damage in a synergistic manner with tobacco (75). In the general population, chronic lung infections due to Chlamydia pneumoniae and tuberculosis are linked with an increased risk of developing lung cancer (75). At all stages of infection, HIV is associated with chronic or repeated lung infections, due to micro-organisms that frequently cause pneumonia in the general population (e.g. S. pneumoniae) as well as opportunistic micro-organisms (e.g., Mycobacterium avium intracellulare, Pneumocystis jirovecii). HIV infection also leads to expansion of the pool of pulmonary macrophages and elevated levels in the lung of pro-inflammatory cytokines, such as interleukin 1 beta, tumor necrosis factor, and interferon gamma (76). If HIV amplifies the effects of tobacco by inducing chronic lung inflammation or causing repeated pulmonary insults due to infections, processes that occur at all levels of HIV-related immunosuppression, a strong association of lung cancer risk with CD4 count or advanced HIV disease might not be expected.

For Hodgkin lymphoma, an additional possibility is that some cases arise as part of an immune reconstitution syndrome (32). This hypothesis is consistent with the nonlinear relationship between CD4 count and Hodgkin lymphoma risk, and with rising Hodgkin lymphoma incidence during the HAART era. Specifically, the increase in CD4 count associated with HAART use among extremely immunodeficient individuals may shift them to a level of immunosuppression (i.e., a CD4 count 200-250 cells/mm³, Figure 2) that puts them at greatest risk for Hodgkin lymphoma. In that setting, Hodgkin lymphoma may develop because the malignant Hodgkin Reed-Sternberg cell may already be present, and partial restoration of immunity allows recruitment of surrounding immune cells and manifestation of the tumor (32). In other cases, Hodgkin lymphoma may develop in a more direct manner with progressive or prolonged immunosuppression. The situation may be analogous to what is observed for KS, another malignancy in which a substantial component of the lesional cells are not actually malignant. While KS most often arises with progressive loss of cell-mediated immunity, cases with onset or worsening related to HAART-induced immune reconstitution are described (77-79).

In the HAART era, several mechanisms in Table 2 are likely to become increasingly important in the development of cancer. The role of these various mechanisms in the etiology of non-AIDS-defining cancers should be evaluated in large, well-designed epidemiologic studies. Ideally, the studies should include detailed data on validated cancer cases, and extensive longitudinal measurements of CD4 count, HIV viral load, and use of antiretroviral medications. Given the rarity of most cancer outcomes, use of pooled data from consortia may be required, and given the complexity of the analyses, replication will likely be necessary.

Counseling by clinicians will play a role in prevention programs. The cornerstone of prevention for lung cancer is tobacco cessation, and given the role of tobacco in other illnesses, programs to encourage HIV-infected individuals to quit smoking should be a priority. Although counseling and available pharmacotherapy are helpful (81), development of more effective interventions in smoking cessation, including those targeted specifically at HIV-infected persons, will be important. Encouraging HIV-infected people to limit the number of their sexual partners may reduce their risk of acquiring new HBV and HPV infections, but condom use is of unproven benefit (e.g., for anal HPV (82)). For individuals with liver disease, abstinence from alcohol may help prevent progression to liver cancer. Finally, encouraging HIV-infected persons to minimize unnecessary sun exposure is prudent to reduce skin cancer risk.

Vaccination strategies may also help prevent cancers. HBV vaccination prevents HBV infection and is associated with reduced occurrence of liver cancer (83). HBV vaccine is less immunogenic among HIV-infected individuals than healthy persons but should still be offered (84). A vaccine against HPV subtypes 16 and 18 is available to prevent cervical infection (85). This vaccine could conceivably prevent anal infections due to these oncogenic subtypes (although this is unproven), and thus might prevent anal cancer (86). To be effective, HBV and HPV vaccines would need to be administered before acquisition of these infections, which often occurs before acquisition of HIV. Thus, it remains unclear whether vaccination of persons who are already HIV-infected could prevent a sizeable number of liver or anal cancers. Among individuals with chronic liver disease, treatment of HCV and HBV infection may prevent progression to liver cancer (87).

Screening is targeted at identifying early-stage pre-malignant lesions, with the goal of eliminating those lesions to prevent progression to cancer. Only two of the non-AIDS-defining malignancies in Table 3 have known pre-malignant lesions that can be detected clinically and are thus amenable to screening intervention. For anal cancer, anal Pap smear screening can detect potentially pre-cancerous anal lesions (i.e., high-grade squamous intraepithelial lesions) that can be ablated with localized therapies (88). Issues regarding Pap smear screening include the high prevalence of anal intraepithelial neoplasia lesions potentially requiring intervention, uncertainty regarding the natural history of these lesions (i.e., rates of progression from high-grade squamous intraepithelial lesions to cancer), need for repeated screening to detect incident lesions, need for expertise in evaluating and treating pre-malignant lesions detected by Pap smears, and lack of proven effectiveness in reducing cancer burden (89). Nonetheless, under a broad range of assumptions, anal Pap smear screening in HIV-infected men-who-have-sex-with-men appears to be cost-effective (90). Testing of anal specimens for carcinogenic HPV genotypes may also be useful in risk stratification, as is the case for cervical cancer (91). Skin cancer screening can identify pre-malignant lesions, such as dysplastic nevi (a precursor to melanoma) and actinic keratoses (a precursor to squamous cell carcinoma). Given the simplicity and low cost of skin cancer screening, it seems feasible to incorporate it into periodic health examinations in HIV-infected persons. As in the transplant setting (92), HIV-infected patients with prior skin cancers should likely receive more intensive screening by dermatologists. Because the vast majority of skin cancer cases arise in non-Hispanic whites, it is reasonable to focus predominantly on this group.