February 27, 2007
HIV-related immunodeficiency is well known to dramatically increase the incidence of specific cancers, which have therefore been regarded as AIDS-defining malignancies (ADM). These include Kaposi's sarcoma, non-Hodgkin's lymphoma and invasive cervical carcinoma. There have been conflicting reports, however, as to whether HIV infection and the degree of HIV-related immunodeficiency also increase the risk of other malignancies. Loss of immune surveillance and effective cell-mediated immunity might be expected to result in an excess of all kinds of cancers. The longer survival and overall aging of the HIV-infected population may also contribute to a greater incidence of many age-related neoplasms. Finally, antiretroviral therapy (ART) has allowed many people with HIV to survive long enough to suffer morbidity and mortality from conditions other than those classically associated with immune deficiency.
In fact, large antiretroviral therapy trials and cohort studies now show greater absolute morbidity and mortality from these conditions than from opportunistic diseases. Yet no dramatic increases in non-AIDS-defining malignancies (nADM) -- that is, all cancers excluding the three that confer an AIDS diagnosis -- have been reported to date among persons with HIV, and it remains unclear whether immunodeficiency increases the risk of nADM. Several studies presented at the 14th Conference on Retroviruses and Opportunistic Infections CROI shed important new light on this topic.
Correlation Between HIV-Induced Immunodeficiency and the Risk of Fatal ADM and nADM: Results from the D:A:D Study
The Data Collection on Adverse Events of Anti-HIV Drugs (D:A:D) Study is the largest and longest of several investigations to recently address the question of non-AIDS-defining malignancies incidence.1 D:A:D is a collaboration between 11 observational cohorts of HIV-infected persons in Europe, North America, and Australia. Data were available for 23,441 persons seen between 1999 and early 2005 for a total of 76,893 person-years of prospective follow-up. Follow-up was calculated for each participant from study enrollment to death, six months after the most recent clinic visit or February 1, 2005 -- whichever came first.
D:A:D was designed to study serious cardiovascular complications of HIV infection; regular assessment for AIDS-defining or other clinical conditions, including malignancies, were not performed. For this reason, this analysis was limited to fatal malignancies, given that data could only be gleaned from death certificates. Demographic and clinical risk factors for a given malignancy as the cause of death were identified using multivariable Poisson analysis. Factors included in the model were the most recent CD4+ cell count and viral load, age, gender, race/ethnicity, HIV exposure category, prior AIDS diagnosis, exposure to combination antiretroviral therapy (cART), coinfection with hepatitis B virus (HBV) or hepatitis C virus (HCV), cohort, and calendar year.
There were 305 fatal malignancies. Notably, nearly twice as many patients died of non-AIDS-defining malignancies as died of AIDS-defining malignancies (193 versus 112 cases). AIDS-defining malignancies included non-Hodgkin's lymphoma (n=82), Kaposi's sarcoma (n=28) and cervical cancer (n=2). Fatal non-AIDS-defining malignancies included lung cancer (n=62), hepatocellular carcinoma (n=16) and other gastrointestinal neoplasms (n=25), followed by hematologic cancers other than non-Hodgkin's lymphoma, anal cancer, other urogenital neoplasms, cancer of the upper airway and a variety of other tumors. In comparison with those who died of AIDS-defining malignancies, patients who died of non-AIDS-defining malignancies were older, less likely to have a prior AIDS diagnosis and more likely to be receiving cART at the time of death. Patients with non-AIDS-defining malignancies also had a higher CD4+ cell count at the time of death (211 versus 75 cells/mm3), a lower viral load at the time of death (2.3 versus 3.5 log10 copies/mL) and a higher nadir CD4+ cell count (87 versus 30 cells/mm3) than those who died of AIDS-defining malignancies, but the ranges were very wide in both groups. Demographics, including exposure category and prior cART exposure, did not differ significantly between those with AIDS-defining malignancies and non-AIDS-defining malignancies.
The most striking findings were that:
In other words, people with HIV are much more likely to die of non-AIDS-defining malignancies than of AIDS-defining malignancies, and their risk of dying of non-AIDS-defining malignancies goes up significantly as the CD4+ cell count goes down. This is best illustrated by a plot of the event rate per 1,000 person-years (on a logarithmic scale) versus the CD4+ cell count stratum.
For example, among patients with a CD4+ cell count of 100 to 199 cells/mm3, the risk of fatal AIDS-defining malignancies was 2.8 per 1,000 person-years, whereas the risk of fatal non-AIDS-defining malignancies was more than twice as high at 6.8 per 1,000 person-years. Among patients with a CD4+ cell count above 500 cells/mm3, the corresponding risks were 0.3 per 1,000 person-years for fatal AIDS-defining malignancies versus 1.1 per 1,000 person-years for fatal non-AIDS-defining malignancies.
The relative risk (RR) of both types of malignancy increased with greater immunosuppression, as measured by the most recent CD4+ cell count. In contrast, there was no correlation with HIV viral load. The magnitude of the effect of CD4+ cell count was greater for AIDS-defining malignancies (e.g., RR = 175 for CD4+ cell count below 50 vs. > 500 cells/mm3), but the effect was still substantial for non-AIDS-defining malignancies (RR = 15 for CD4+ cell count < 50 vs. > 500 cells/mm3).
In the Poisson model, every doubling of the most recent CD4+ cell count was associated with a 47% decrease in the risk of fatal AIDS-defining malignancies and a 39% decrease in the risk of fatal non-AIDS-defining malignancies. Both estimates had very narrow confidence intervals. The nadir CD4+ cell count was also associated with the risk of both types of malignancy, but not as strongly as the most recent CD4+ cell count.
Other predictors of fatal non-AIDS-defining malignancies included older age, smoking and HBV infection, although the associations with smoking and HBV became nonsignificant when lung and liver cancers, respectively, were excluded from the analysis. Audience members pointed out that anal cancer is pathogenically similar to cervical cancer and should probably be considered an AIDS-defining malignancies. The authors had anticipated this concern and reported an analysis in which anal cancers were excluded. Their results essentially demonstrated an identical strong association between advancing immunodeficiency and fatal non-anal non-AIDS-defining malignancies. The only other predictor of fatal AIDS-defining malignancies was a prior AIDS-defining malignancies diagnosis. Later calendar year was associated with reduced rates of both types of malignancy and may be a reflection of better treatment and less time spent at lower CD4+ cell counts.
The investigators concluded that fatal non-AIDS-defining malignancies have become more common than AIDS-defining malignancies among persons with access to cART. Moreover, non-AIDS-defining malignancies will likely continue to increase as the HIV-infected population ages. Immunodeficiency strongly influences the risk of both types of malignancy, so prevention must focus on both sustaining high CD4+ cell counts and reducing other known cancer risk factors (e.g., smoking and HBV infection). The only serious limitation of the study is that because of its focus on only fatal malignancies, it is not truly a study of cancer incidence but rather of cancer mortality. This could have biased the association between incidence and CD4+ cell count if patients with more advanced HIV disease were less likely to received aggressive treatment for malignancies or were more likely to die of their cancer despite treatment -- both plausible suppositions. The magnitude of any such bias, however, is extremely unlikely to account for the large effect size of CD4+ cell count seen in the D:A:D study.
Change in the Incidence of ADM and nADM Over Time Among HIV-Infected Individuals in the United States
This study -- a collaboration between the U.S. Centers for Disease Control and Prevention, the National Cancer Institute and three state health departments -- assessed cancer incidence by linking the HIV and cancer registries of New Jersey, Colorado and Florida.2 Data were collected on 57,350 persons with HIV who were registered from 1991 to 2002 and who were initially cancer free. The five-year risk of cancer from the time of registration was compared with rates in the general population, and standardized incidence ratios (SIRs) were calculated. The RRs of select cancers were calculated by multivariate Poisson regression according to AIDS onset, calendar year and baseline CD4+ cell count at HIV registration after adjustment for gender and age. Baseline CD4+ cell counts were available for only 27% of patients (median = 491 cells/mm3; interquartile range = 356-685 cells/mm3).
There were 871 incident cancers during follow-up. Not surprisingly, SIRs were significantly elevated for ADM (P < .05), especially Kaposi's sarcoma and central nervous system non-Hodgkin's lymphoma. SIRs were also significantly elevated for Hodgkin's lymphoma and for cancers of the lung, oral cavity, anus, liver and penis, indicating a higher risk for HIV-infected individuals compared with the general population. Being diagnosed with AIDS during the study period was associated with an increased subsequent risk of Kaposi's sarcoma (RR = 7.2), non-Hodgkin's lymphoma (RR = 4.1), Hodgkin's lymphoma (RR = 3.0), cervical cancer (RR = 2.2) and lung cancer (RR = 2.1) compared with HIV-infected individuals who did not receive an AIDS diagnosis.
|• Kaposi's sarcoma||173||1,251|
|• Non-Hodgkin's lymphoma||203||7.3|
|• Cervical cancer||28||2.9|
|• Hodgkin's lymphoma||36||5.6|
|• Lung cancers||109||2.6|
|• Oral cavity cancers||26||1.7|
|• Anal cancer||18||9.2|
|• Liver cancer||14||2.7|
|• Penile cancer||3||5.4|
|CNS = central nervous system.|
Beginning with the widespread adoption of highly active antiretroviral therapy (HAART) in 1996, the risk declined for Kaposi's sarcoma (RR = 0.4, compared with the period prior to 1996) and non-Hodgkin's lymphoma (RR = 0.7). These declines were greatest for HIV-infected individuals who had developed AIDS.
The risk for cervical cancer and lung cancer did not change from 1996. The risk for Hodgkin's lymphoma increased, but not significantly (RR = 2.2, 95% confidence interval [CI] = 0.9-5.7). The risk was inversely related to the baseline CD4+ cell count for Kaposi's sarcoma (P < .001) and non-Hodgkin's lymphoma (P = .02), but these associations were attenuated after 1996 (P = .007 and P = .22, respectively). No association with CD4+ cell count was seen for other cancers; the abstract reported that data for this analysis "were sparse."
This study has the advantages of being large in size, having a long follow-up and presumably having relatively accurate data. Of course, the true denominator of HIV-infected individuals is unknown, as only those tested for HIV are entered into the registries. The underreporting of cancer incidence is likely to be less significant. Although the data were limited to three states, there is no reason to think these findings would not be more broadly generalizable to all American HIV-infected individuals.
Unfortunately, the poster for this abstract was never hung at CROI, and the authors did not appear to present it. Thus, significant methodologic questions remain unanswered. For example, how was the low frequency of available baseline CD4+ cell counts handled in the analysis? No information is given on whether data were collected or controlled for with regard to other cancer risk factors, including race and ethnicity, smoking and hepatitis coinfection.
This study found 404 incident AIDS-defining malignancies s and, presumably, 467 incident non-AIDS-defining malignancies. State HIV registries, which include persons not in HIV care but who test positive for HIV infection, likely include a higher proportion of persons with high CD4+ cell counts than do clinical cohorts like those in the D:A:D Study. Indeed, the median baseline CD4+ cell count for those with available data was nearly 500 cells/mm3. Nevertheless, the American study still found a greater number of incident non-AIDS-defining malignancies than AIDS-defining malignancies in this HIV-infected population, echoing the D:A:D finding that non-AIDS-defining malignancies are now more frequent among HIV-infected individuals in relatively resource-rich settings.
This study also identified significantly elevated SIRs for HIV-infected individuals compared with the general population for the same set of non-AIDS-defining malignancies observed in the D:A:D cohort, thus corroborating the inference that HIV infection increases the risk of these neoplasms. Having an AIDS diagnosis -- a proxy for more advanced immunosuppression -- also predicted a higher risk for lung cancer, the most frequent non-AIDS-defining malignancies.
There was likely not enough CD4+ cell count data in this cohort to tease out the relationships between CD4+ cell count strata and the risk of different neoplasms. The decline in excess HIV-associated risk for Kaposi's sarcoma and non-Hodgkin's lymphoma, but not for lung or cervical cancer, following the introduction of HAART may reflect the much larger effect size of immunosuppression on non-AIDS-defining malignancies than on AIDS-defining malignancies, as observed in the D:A:D study. The investigators of this study argued that the increased incidence in Hodgkin's lymphoma that they observed since 1996 suggests that incomplete immune restoration increases the risk for this malignancy. However, given the low relative risk and the wide confidence interval, this should be taken as speculation.
French/Australian Study Finds No Association Between HAART and the Risk for Hodgkin's Lymphoma
This case-control study explored the competing hypotheses that the increased incidence of Hodgkin's lymphoma seen in HIV-infected individuals may be due to immunosuppression or, alternatively, to the effects of HAART.3 Hodgkin's lymphoma, although not an AIDS-defining malignancies, is approximately 11 times more frequent among persons with AIDS compared with the age-adjusted rate for the general population, and this malignancy is a disproportionate cause of mortality among HIV-infected individuals.
The cause of Hodgkin's lymphoma is thought to be related to Epstein Barr virus (EBV) in more than 90% of HIV-infected individuals, whereas EBV accounts for only approximately 30% of Hodgkin's cases among persons without HIV. In contrast with Kaposi's sarcoma and non-Hodgkin's lymphoma, the incidence of Hodgkin's lymphoma has continued to increase through the introduction of HAART and beyond.
The study investigators used data from the Australian HIV Observational Database (AHOD) and the French Aquitaine cohorts to identify 41 incident cases of histologically confirmed Hodgkin's lymphoma occurring between January 1, 1996, and June 30, 2004, and at least three months after the patient's inclusion in the cohort.
Mixed cellularity was the dominant pathology in 39%, nodular sclerosis in 15% and lymphocyte depletion in 10%; 37% could not be typed because of extranodal disease. Potential controls were identified for each case patient from among individuals without Hodgkin's lymphoma according to study center, the CD4+ cell count at diagnosis (within 100 cells/mm3), the year of Hodgkin's lymphoma diagnosis and follow-up time in the cohort (within 1 year). Two controls were randomly selected to match each case. Differences between cases and controls with regard to putative risk factors were analyzed using standard t-test and Chi-square statistics, and matched odds ratios (ORs) were calculated.
The median age of individuals was 39 years. The median CD4+ cell count was 241 cells/mm3 for cases and 256 cells/mm3 for controls, and the median follow-up for these groups was 11.7 and 11.2 years, respectively. Two-thirds of cases were on HAART at the time of Hodgkin's lymphoma diagnosis, as were 75% of controls at the time of matching. There were more men among the cases (98%) than the controls (85%); otherwise, cases and controls were well matched for demographic, clinical and laboratory variables.
There was a slight increase over time in the rate of Hodgkin's lymphoma. Based on incident cases in the Aquitaine cohort, the mean Hodgkin's lymphoma incidence rates were calculated to be 0.7 per 1,000 person-years for the period from 1996 to 2000 (95% CI = 0.2-1.2) and 0.9 per 1,000 person-years for the period from 2001 to 2004 (95% CI = 0.4-1.5). No significant risk factors for Hodgkin's lymphoma were identified following the comparison of cases with controls. In particular, the following variables were tested for their association with Hodgkin's lymphoma:
The small numbers of patients may have limited the ability to detect significant associations. For example, male gender was associated with Hodgkin's lymphoma risk with an OR of 4.70, but the confidence interval (0.65-34.4) spanned 1 and thus was not significant. Nevertheless, for all other factors tested, the ORs were close to unity, suggesting not even a trend toward an association with Hodgkin's lymphoma.
The authors concluded that Hodgkin's lymphoma does not appear to be associated with the duration of immunosuppression, the CD4+ cell count nadir, viral load or the use of antiretroviral drugs. They speculated that other unmeasured factors, including EBV infection, immune activation and altered qualitative immune function, may explain the increasing incidence over time of HIV-associated Hodgkin's lymphoma.
This study presents an intriguing contrast to the previous two, suggesting that the emerging mode of increased non-AIDS-defining malignancies risk associated with HIV infection, particularly in individuals with low CD4+ cell count counts, may not be simple and may not apply equally to various non-AIDS-defining cancers.
The large American study found a doubling of Hodgkin's lymphoma incidence from the late 1990s to the early 2000s and a three-fold higher risk of Hodgkin's lymphoma among HIV-infected individuals compared with the general population. However, the increase in Hodgkin's lymphoma incidence seen in the French/Australian study, with a smaller number of person-years, did not reach statistical significance. Moreover, the authors found no support for an association between HAART and the risk of Hodgkin's lymphoma, but they also did not identify a protective effect of HAART.
The small number of cases and controls may have masked associations with risk factors. In particular, the study design -- matching controls to cases by CD4+ cell count -- precluded the possibility of finding an association between Hodgkin's lymphoma risk and CD4+ cell count. Hence, the possible influence of the degree of immunosuppression remains unaddressed.
Finally, because Hodgkin's lymphoma is a disease of lymphocytes and one that is associated in this setting with EBV infection, it may represent a unique type of malignancy that is unrepresentative of other non-AIDS-defining malignancies. As the case of anal cancer -- another disease associated with chronic viral infection -- also suggests, the dichotomous categories of AIDS-defining malignancies and non-AIDS-defining malignancies may need to give way as clinicians develop a more sophisticated understanding of how HIV infection increases the risk of different neoplasms.