Malignancies, including the three that are part of the definition of AIDS (Kaposi sarcoma [KS], non-Hodgkin lymphoma [NHL] and cervical cancer) as well as certain non-AIDS-defining cancers, cause significant morbidity and an estimated one-third of deaths among HIV-infected individuals (1). In this issue of Cancer, Seaberg et al. (2) examined cancer incidence among HIV-infected and HIV-uninfected men who have sex with men (MSM) in the U.S. Multicenter AIDS Cohort Study (MACS). Cancer risk in HIV-infected men was compared internally to HIV-uninfected MACS participants, and externally to general population cancer data from the Surveillance, Epidemiology and End Results (SEER) registry program.
The MACS analysis highlighted a number of interesting and important trends in AIDS-defining and non-AIDS-defining cancers among HIV-infected individuals. The authors report that the incidence rates of KS and NHL, which fell substantially with the widespread use of highly active antiretroviral therapy (HAART) beginning in 1996, continued to decline through 2007, presumably due to improvements in HAART. These results are consistent with most (3;4) but not all (5) previous reports. Interestingly, KS and NHL rates began to decline even prior to the introduction of HAART (5), perhaps attributable to single- and dual-agent antiretroviral therapies in use prior to 1996 (5).
Although comparing cancer rates by calendar period is useful for gauging the overall population effectiveness of changes in antiretroviral therapy, a strength of cohorts such as the MACS is that they also have data on actual medication use and HIV disease markers. Even within the HAART era, not all HIV-infected people are on HAART, medication regimens vary, and the effectiveness of HAART in suppressing HIV replication and restoring immune function may be incomplete. Analyses that consider only calendar periods cannot determine whether the decline in risk over time is driven by a larger proportion of individuals receiving therapy or improvements in HAART regimens. Additional studies that examine how the risk of KS and NHL varies with actual HAART use, specific medications, and HAART effectiveness as assessed by HIV disease markers, would help to clarify why the risk of KS and NHL remains elevated in the most recent HAART calendar period: Is the continuing risk due to people not taking HAART, poor adherence with medications, ineffectiveness of HAART due to emergence of resistant HIV, or incomplete immune reconstitution even in people on virologically suppressive regimens? Data from France indicate that CD4 cell counts and plasma HIV RNA levels continue to be important predictors of KS and NHL risk for people on HAART (6). We recognize that a challenging issue in evaluating the effect of HAART on cancer risk is confounding by indication, i.e., that the sickest people are prescribed HAART, which can lead to an artifactual positive association between HAART use and adverse health outcomes. Nonetheless, this methodological issue has been addressed in recent analyses of observational data from HIV cohorts (7).
In contrast to the declining trends for KS and NHL among HIV-infected MACS participants, Seaberg et al. report a concerning rise in anal cancer incidence over time, including a suggestive increase within the HAART era itself. This finding is in line with results from most previous studies, which have observed stable or increasing risk of anal cancer in the HAART era or with HAART use (8-10). HIV-infected individuals, including women and heterosexual men as well as MSM (11), are at increased risk of anal cancer and other human papillomavirus (HPV)-related cancers (3;8). Prolonged HIV-induced immunosuppression, as indicated by an extended time spent with a low CD4 cell count, is a risk factor for anal cancer (6). However, because the restoration of the immune system by HAART does not appear to reduce the risk of the immediate precursors of anal cancer (12), it has been postulated that immune suppression may only impact the persistence of HPV infection and not subsequent progress to invasive malignancy (10;12;13). Because the latency period for persistent HPV infection resulting in anal cancer may be quite long, even decades, if initiation of HAART is delayed, therapy may be given to patients too late to stop development of anal cancer, and may indirectly facilitate carcinogenesis by prolonging survival. One possibility is that earlier HAART initiation (i.e., initiation at a high CD4 cell count) would allow for HPV clearance before development of early pre-cancer lesions.
Lung cancer risk appeared higher in HIV-infected MACS participants than in HIV-uninfected participants, although this differences was not statistically significant. Registry linkage studies, and studies comparing to external rates have consistently found that HIV-infected individuals had 2- to 3-fold higher lung cancer rates relative to the general population (5;9;14). These associations are partially driven by the high proportion of HIV-infected individuals who smoke cigarettes compared with the general population. In fact, lung cancer among HIV-infected individuals occurs almost exclusively among smokers. Nonetheless, studies have found increased risk of lung cancer with HIV even after directly or indirectly adjusting for differences in smoking prevalence (15;16). The mechanism by which HIV infection increases lung cancer risk remains unclear, though several hypotheses have been proposed. HIV-infected individuals experience increased pulmonary inflammation, repeated and chronic pulmonary infections and deficiencies in antioxidants and other nutrients, all of which may act together with tobacco use to increase lung cancer risk (17). Along these lines, our recent analyses of U.S. registry linkage data indicate that people with AIDS who have had repeated episodes of bacterial pneumonia are at a somewhat increased risk of lung cancer (18).
A major strength of the study by Seaberg et al. is the internal comparison of cancer risk within the MACS. Many prior studies reporting on excess cancer risk in HIV-infected individuals have relied on comparisons between cohort data and general population cancer rates (9) or matches between HIV/AIDS and cancer registries (4;5;14). HIV-uninfected individuals enrolled in the same cohort study and drawn from the same source population are a more appropriate comparison group than the general population, because they are more similar to HIV-infected individuals in their demographic, socioeconomic, and behavioral characteristics. Further, studies that include individual-level data on all cohort participants, like the MACS, allow for careful control of any residual differences in these characteristics. For example, Seaberg et al. were able to restrict the analysis to smokers and adjust for level of cigarette smoking in their lung cancer analysis, whereas data on smoking are not available in population-based registries.
Assessing the associations between HIV infection and cancer risk within a cohort study is not without weaknesses. Registry-based studies can include more than 100 times the number of HIV-infected people that are included in cohort studies such as the MACS, enabling the detection of much smaller differences in cancer risk between HIV-infected and HIV-uninfected populations. In the MACS analysis, many of the comparisons between HIV-infected and HIV-uninfected individuals were likely underpowered; thus, some associations with HIV infection may have been missed. Also, the MACS study includes only one HIV risk group (MSM) and lacks data on cancer risk in women. Future consortia that combine data from U.S. and international cohorts, include both HIV-infected and comparable HIV-uninfected individuals, and include people from diverse demographic groups would be ideal for studying HIV infection and cancer risk, bridging the gap between registry-based studies (with robust statistical power and population representativeness) and HIV cohort studies (with detailed individual-level data).
An additional point to note in the current MACS analysis is that HIV-uninfected men were observed to have 47% fewer cancers than expected based on SEER general population rates (standardized incidence ratio [SIR]=0.53; 95% confidence interval [CI]: 0.43-0.66). While Seaberg et al. postulate that MACS participants may have “generally low risk for most cancers,” this apparently low risk could plausibly reflect under-ascertainment of cancers. Cancers in the MACS were ascertained through “study interviews, medical record abstraction and review, report by self or next of kin, and review of vital status records including data from the National Death Index” (2). Cancers not reported by participants or occurring after loss to follow-up would not have been detected by study personnel, potentially leading to artificially low cancer rates.
Two examples where under-ascertainment seems possible are lung and anal cancers. As cigarette smoking prevalence was far greater in HIV-uninfected MACS participants (40% current smokers) than in the U.S. general population (24% current smokers in 2000 (19)), an elevated lung cancer risk would be expected. However, the risk of lung cancer appeared reduced among HIV-uninfected men in the MACS (SIR=0.30; 95% CI 0.11-0.66). Further, male homosexuality and receptive anal intercourse are strong, established risk factors for anal cancer (20), but HIV-uninfected MSM in the current MACS analysis manifested only a modest and non-significant elevation compared to men in the general population (SIR=2.50; 95% CI 0.30-9.03). Thus, we believe that the results for lung and anal cancers likely reflect some degree of under-ascertainment of these cancers, and, by implication, perhaps other cancers as well. If present, the under-ascertainment of cancers is likely non-differential by HIV status, and thus would likely attenuate internal comparisons between HIV-infected and HIV-uninfected men. This issue emphasizes the importance of complete and accurate cancer ascertainment in HIV cohort studies. One approach would be to supplement the ongoing data collection activities by linking the data from cohort participants with population-based cancer registries.
Among HIV-infected individuals, HAART has improved immune function, reduced the risk of AIDS, including AIDS-defining malignancies, and dramatically prolonged survival. As a result, HIV-infected individuals are living to older ages where the risk of non-AIDS-defining cancers steeply increases. Thus, in the coming years, even if the number of AIDS-defining malignancies continues to decline, the number of non-AIDS-defining cancers diagnosed in HIV-infected individuals will increase. The importance of prevention, early detection and treatment efforts targeted toward HIV-infected individuals has also grown. As lung cancer is the most frequent non-AIDS-defining malignancy, smoking cessation efforts are essential to reduce morbidity and mortality. Further, HIV-infected women should adhere to specific Pap smear guidelines tailored toward woman with HIV infection to prevent cervical cancer. In addition, future research is needed to determine the efficacy of anal cancer screening with the Pap test in HIV-infected MSM, as well as other risk groups. Given the ever improving life expectancy afforded by effective HAART, age-specific screening recommendations for colon, prostate, and breast cancers should be applied to HIV-infected individuals as they are for the general population.
The success of HAART at prolonging survival now brings new challenges as HIV-infected individuals age. It is important that we continue to build on the findings of Seaberg et al. and others by working toward a better understanding of the etiology of cancer in the setting of HIV infection. We must also devote continuing effort and attention to cancer prevention with the goal of reducing cancer incidence, morbidity and mortality among HIV-infected individuals.