We combined data from 18 studies (13 studies in overall meta-analysis, and 5 additional studies included in stratified analyses) to estimate the relative incidence of specific non-AIDS cancers among HIV-infected individuals, compared to the general population. HIV-infected individuals had twice the risk of a non-AIDS cancer than the general population. SIRs for all non-AIDS cancers were greater among men than women and among those with AIDS than those without AIDS; however, no substantial difference was observed by HAART era. The random-effects SIRs showed a substantial increase in the incidence rate of many individual cancer sites among those with HIV when compared to the general population. Additionally, the SIRs for certain cancers seemed to be modified by gender, AIDS and HAART era. These differences do not by themselves imply that the incidence rate of specific cancers is different by gender, or HAART era, as the reference group of general population non-AIDS cancer incidence may differ in these strata.
Though the body of research supporting an increased risk of non-AIDS cancers among HIV-infected individuals continues to expand, explanations for this increase have not yet been well-elucidated. HIV-associated immune suppression may increase susceptibility to cancers that are caused by oncogenic viruses. Indeed, we observed the greatest SIRs to be in cancer sites that have an infectious etiology. Anal, vaginal, penile, nasopharyngeal, laryngeal and oral cancers are all associated with infection with human papillomavirus (HPV), and were all found to occur at a greater rate among those infected with HIV than in the general population. Additionally, liver cancer is associated with hepatitis B (HBV) and hepatitis C (HCV) virus, and nasopharyngeal cancer and Hodgkin lymphoma are both associated with Epstein-Barr virus (EBV); the incidence rates for these three cancer sites were all substantially higher in those with HIV than in the general population. HIV-infected persons may be disproportionately infected with oncogenic viruses. For example, in a study of homosexual men, HIV-infected men were found to have 1.5 times the risk of anal HPV infection as HIV-uninfected men50
, and in a study of women at high risk for HIV, HIV-infected women were found to have 1.8 times the risk of anal HPV infection as HIV-uninfected women51
. Conversely, several cancers without strong evidence for an infectious etiology, such as colorectal and pancreatic cancers, had summary SIRs that were near null. Decreased immune function paired with increased incidence of these infections may be responsible for the increased rates of virally-associated cancers among those infected with HIV. The association between decreased immune function and increased risk of anal cancer, Hodgkin lymphoma and liver cancer is further supported by our observation that the SIRs of these cancers were increased among those with AIDS compared to those without AIDS. Furthermore, rates of anal cancer, Hodgkin lymphoma, and liver cancer, as well as other cancers associated with oncogenic viruses have been found to be elevated among organ transplant recipients, providing support for the role of suppressed immunity in their etiology15
HIV-infected individuals may also have a higher prevalence of other cancer risk factors (e.g., cigarette smoking) than the general population. One study reported the prevalence of cigarette smoking to be 59% among persons with HIV/AIDS in New York State, which is three times the prevalence in the general population52
. Another study reported the prevalence of smoking among all participants of the Swiss HIV Cohort Study to be 72%, and as high as 96% among injection drug users in the cohort38
. We observed elevated summary SIRs for several smoking-related cancers, including lung, kidney, stomach, laryngeal and oral cancers. While some studies suggest that the increased risk of lung cancer among HIV-infected individuals is likely explained by differences in smoking6, 12
, other studies have shown an increase in lung cancer, even after accounting for differences in smoking40, 53
. The SIRs for most smoking-related cancers (lung, kidney, laryngeal, and stomach cancers) were observed to be greater among women than men. Perhaps the greater SIR observed among women is due to a greater relative increase in smoking among HIV-infected women compared to the general population compared to HIV-infected men, though we did not have data to examine the prevalence of cigarette smoking among men and women in each study. Interestingly, the SIR for lung cancer was observed to be greater among those with AIDS than those without AIDS, suggesting that immune suppression may play a role in the development of lung cancer among those who are HIV-infected. It has been hypothesized that HIV-associated suppression of the immune system may lead to reduced tumor surveillance, allowing tumors of the lung to continue to develop when they otherwise would be destroyed by the immune system54, 55
We observed the incidence of breast and prostate cancer to be lower among HIV-infected women and men compared to the general population. This may suggest a potential protective effect of HIV on the development of these cancers, perhaps through changes in hormone levels42
. However, the decreased incidence of these cancers among HIV-infected individuals may also be due to differential screening by HIV status27, 42
. Those with HIV, particularly injection drug users and the poor, may be less likely to be regularly screened for cancers. For example, the Women’s Interagency HIV Study found women with HIV and at high risk for HIV were significantly less likely to have a mammogram than women in the general U.S. population56
. Therefore, more sub-clinical disease may be detected among those without HIV, which would result in an increased incidence.
Our analyses were limited to the information provided by each of the included studies. We were unable to look at other factors that may modify the association between HIV and non-cancer AIDS, such as age, likely route of transmission, smoking status, or race/ethnicity. We observed notable between-study heterogeneity; this heterogeneity may be due to differences in study design, populations studied, or in unmeasured characteristics that may vary across studies. Indeed, it is likely that one or more of these factors explains a portion of the observed between-study heterogeneity. Additionally, we observed departures in symmetry for the funnel plots of a few cancer sites, which may be due to publication bias.
The use of general population comparisons is a limitation of the literature, and thus of our review of that literature. In occupational epidemiology, such external comparisons with general populations are generally considered inferior to internal comparisons within employed populations. The biases have been grouped under such rubrics as the "healthy worker" and "healthy worker survivor" biases. The analog of internal comparisons in occupational studies would be to compare HIV-infected individuals with and without AIDS directly, or to compare HIV-infected men and women directly, rather than to compare the comparisons with the general population.
Another limitation is the examination of effect-measure modification for ratio effect measures. It may have been preferable to examine modification of rate differences57
. Fortunately, the indications are of a greater elevation of rate ratios in groups with higher baseline rates (HIV-infected men and AIDS patients). In this case, positive rate ratio modification implies positive rate difference modification.
Finally, for some cancer sites, our meta-analyses were based on very few cases from few studies. Summary SIRs for nasopharyngeal, eye, bone, small intestine and gall bladder cancers were all based on 10 cases or fewer, and summary SIRs for nasopharyngeal, eye, small intestine, lip, oral and pharyngeal, oropharyngeal, gall bladder and rectal cancers were based on results from only two or three studies. Thus, the estimates for these cancer sites are particularly imprecise.
This work expands upon a previous meta-analysis that presented summary SIRs for non-AIDS cancers among HIV-infected individuals15
. In addition to the seven studies included in the prior meta-analysis, we included six studies that have presented SIRs for non-AIDS cancers. Additionally, our study differs from the prior meta-analysis in that we examined whether SIRs differed by gender, AIDS status and HAART era, while the previous study did not. These stratified analyses highlight differences in SIRs of non-AIDS cancers, relative to the general population, that may be associated with gender, immune suppression and HAART use.
This study found an increased general-population SIR for many types of non-AIDS cancers. It remains unclear whether HIV-infected individuals are truly at a greater risk for non-AIDS-defining cancers, or if confounding by unadjusted cancer risk factors may be responsible for the apparent elevated incidence. Future pooling projects (rather than meta-analyses) that compare HIV-infected to HIV-uninfected individuals will be better able to elucidate the effect of HIV infection on the development of non-AIDS cancers.