Spectrum of Cancer Risk Late After AIDS Onset in the United States

Background  Persons living with AIDS today remain at elevated cancer risk. Highly active antiretroviral therapy (HAART), widely available since 1996, prolongs life, but immune function is not fully restored. We conducted this study to assess long-term cancer risk among persons with AIDS relative to the general population and the impact of HAART on cancer incidence.

Methods  Records of 263 254 adults and adolescents with AIDS (1980-2004) from 15 US regions were matched to cancer registries to capture incident cancers during years 3 through 5 and 6 through 10 after AIDS onset. Standardized incidence ratios (SIRs) were used to assess risks relative to the general population. Rate ratios (RRs) were used to compare cancer incidence before and after 1996 to assess the impact of availability of HAART.

Results  Risk was elevated for the 2 major AIDS-defining cancers: Kaposi sarcoma (SIRs, 5321 and 1347 in years 3-5 and 6-10, respectively) and non-Hodgkin lymphoma (SIRs, 32 and 15). Incidence of both malignancies declined in the HAART era (1996-2006). Risk was elevated for all non–AIDS-defining cancers combined (SIRs, 1.7 and 1.6 in years 3-5 and 6-10, respectively) and for the following specific non–AIDS-defining cancers: Hodgkin lymphoma and cancers of the oral cavity and/or pharynx, tongue, anus, liver, larynx, lung and/or bronchus, and penis. Anal cancer incidence increased between 1990-1995 and 1996-2006 (RR, 2.9; 95% confidence interval [CI], 2.1-4.0), as did that of Hodgkin lymphoma (RR, 2.0; 95% CI, 1.3-2.9).

Conclusion  Among people who survived for several years or more after an AIDS diagnosis, we observed high risks of AIDS-defining cancers and increasing incidence of anal cancer and Hodgkin lymphoma.

Persons infected with human immunodeficiency virus (HIV) have an increased risk of cancer.^1,2 Advanced HIV infection is characterized by profound immunosuppression (ie, AIDS), a risk factor for a number of malignancies. Three cancers are AIDS defining: Kaposi sarcoma (KS), non-Hodgkin lymphoma (NHL), and cervical cancer.³ These cancers are caused by oncogenic viruses, specifically, KS-associated herpesvirus for KS, Epstein-Barr virus for the major AIDS-defining NHL subtypes (diffuse large B-cell NHL and central nervous system [CNS] NHL), and human papillomavirus (HPV) for cervical cancer.^4-6 Those infected with HIV also have an elevated risk for other virus-related cancers (eg, anal cancer related to HPV, liver cancer related to hepatitis B and C viruses).^5-7 In addition, HIV-infected people have a higher prevalence of lifestyle-associated risk factors for cancer, including smoking and alcohol abuse.^8,9

Use of highly active antiretroviral therapy (HAART) among persons with HIV/AIDS can lead to partial immune restoration and prolonged survival.^10,11 Despite declines in cancer incidence attributable to the widespread introduction of HAART in 1996, cancer risk among HIV-infected people remains elevated during the HAART era.^2,12-14 Of note, prior studies of cancer risk in HIV-infected people have largely been limited to the first 1 to 5 years after an AIDS diagnosis.^1,2,14-16 As persons with AIDS continue to live longer, the number of people surviving for many years after an AIDS diagnosis will increase. This growing population comprises individuals who have lived for a prolonged period with immune suppression. The effects of these chronic immune disturbances, HAART, and long-term infection with oncogenic viruses are unknown. Estimates of cancer risk are needed to inform both public health programs and clinicians treating this population.

To assess the long-term cancer risk among persons with AIDS, we analyzed data from the HIV/AIDS Cancer Match Study,¹ a large, representative cohort of persons with AIDS in the United States. Previous analyses included persons diagnosed as having AIDS during the 1978-1996 period (with cancer information up to 5 years after AIDS onset)¹⁵ and persons diagnosed as having AIDS between 1980 and 2004 (with cancer information up to 2 years after AIDS onset).² In the current study, we evaluated the period from 3 to 10 years after AIDS onset to assess cancer risk relative to the general population and to determine the impact of HAART on cancer incidence over time.

The HIV/AIDS Cancer Match Study¹ links records of persons with HIV or AIDS diagnosed between 1980 and 2008 from 15 US states and metropolitan regions to corresponding cancer registry records using a probabilistic algorithm. Cancer registry coverage varies by region but is complete through the 2004-2006 period for several registries. Following linkage of registry databases, only de-identified data were retained for analyses. Institutional review boards at participating sites approved the study.

The current study focuses on cancer risk late after an AIDS diagnosis, specifically for the period 3 to 10 years after AIDS onset. Onset of AIDS was defined using the 1993 Centers for Disease Control and Prevention definition.³ The study was restricted to the most common racial/ethnic categories (non-Hispanic whites, non-Hispanic blacks, and Hispanics) to allow for stable estimates of expected cancer counts. Children younger than 15 years were excluded. The current study was therefore limited to adults and adolescents who had AIDS onset in the 1980-2004 period and whose follow-up during the 3- to 10-year period after AIDS onset overlapped with cancer registry coverage (N = 263 254).

Malignancies reported to cancer registries were coded according to the International Classification for Diseases for Oncology.¹⁷ Cancers were categorized by site and histologic characteristics using the Surveillance, Epidemiology, and End Results (SEER) program's “site recode with Kaposi sarcoma and mesothelioma.”¹⁸ This classification was slightly modified by grouping some rare cancer types and expanding others (eg, NHL) into subtypes of special interest. We considered people at risk for each cancer if they had not previously had that cancer.

We initially evaluated cancer risk separately for the periods 3 to 5 and 6 to 10 years after AIDS onset. For each period, cancer cases and person-times were tabulated from the start of the period or the beginning of cancer registry coverage (whichever occurred later) until the first of the following: the end of the period, the end of cancer registry coverage, or death (as ascertained from the HIV/AIDS registries at the time of the match). For each cancer, we present the standardized incidence ratio (SIR), defined as the ratio of observed to expected number of cancer cases, as a measure of risk relative to the general population. Expected counts were obtained by applying gender, age, race/ethnicity, calendar year, and registry-specific incidence rates to person-time among people with AIDS. Because virtually all KS and CNS NHL cases in the general population are AIDS related, we used expected counts for these cancers from SEER registries prior to the epidemic (1973-1979).¹⁹ We calculated 2-sided exact Poisson 95% confidence intervals (CIs) for the SIRs.

Similarly, for the combined period 3 to 10 years after AIDS onset, we evaluated changes in cancer incidence between 2 “attained” calendar periods: the 1990-1995 pre-HAART era and the 1996-2006 HAART era. These attained calendar periods correspond to the years in the named range regardless of how long an individual had had AIDS, providing that the onset of AIDS was within 3 to 10 years previously. In these analyses, we calculated incidence rates as the observed counts divided by person-time at risk. Changes in incidence between the calendar periods were then assessed using a rate ratio (RR) obtained from Poisson regression. The Poisson models were adjusted for attained age, race/ethnicity, and gender/mode of HIV exposure (men who had sex with men [MSM] alone or combined with injection drug use [IDU], all other men, and women). To account for underdispersion or overdispersion in the Poisson models, the standard errors of the coefficients were adjusted by the Pearson χ² statistic divided by the degrees of freedom. When an RR was 0 or undefined due to an empty cell, we calculated a 1-sided exact 95% CI. We also present SIRs for the combined 3- to 10-year period restricted to the HAART era as a measure of cancer risk relative to the general population.

For selected cancers with substantial changes in incidence and sufficient sample size, we calculated incidence as a function of individual attained calendar years (1990-2006). We used joinpoint regression to fit log-linear Poisson models for these annual rates, which translated into annual percentage changes across calendar time.²⁰ All joinpoint models manifested adequate fit according to the Hosmer-Lemeshow statistic. We present only annual percentage changes that differed significantly from 0 (P < .05). We used the Joinpoint Regression Program (National Cancer Institute, Rockville, Maryland, version 3.4.2, October 2009) for joinpoint analyses.

A consideration in evaluating cancer risk late after AIDS onset is that people with AIDS, after an extended period, may have migrated out of the cancer registry region. Persons with AIDS in our study were known to reside in the cancer registry coverage area at the time of AIDS diagnosis, but we had no data on subsequent out-migration. Prior studies have demonstrated that among persons with AIDS who had died, as many as 10% had moved from their area of AIDS diagnosis before death.^21,22 Similar migration rates have been noted among the general population.²³ If a person had moved out of the cancer registry coverage area, his cancer would not have been ascertained, leading to a reduction in the observed cancer count but not in the estimated follow-up time, thus causing an underestimation of cancer risk. To correct this bias, we adjusted for potential out-migration by incrementally discounting each subject's person-time to reach a total reduction of 10% by 10 years after AIDS diagnosis. All SIRs and incidence measures were based on these adjusted person-time estimates. Results were similar in a sensitivity analysis in which we discounted the person-time by a factor of 20% (data not shown).

Demographic characteristics of the 263 254 persons with AIDS included in the study are summarized in Table 1. Most were men (80.1%), and the median age at AIDS onset was 36 years. Similar proportions of subjects were non-Hispanic white (39.7%) and non-Hispanic black (40.1%), and 20.2% were Hispanic. Almost half were MSM (44.8%), and 24.8% had IDU as their mode of HIV exposure (Table 1). At AIDS onset, CD4 counts were higher for subjects who contributed follow-up to the 6- to 10-year period than for those who contributed follow-up to the 3- to 5-year period (median CD4 lymphocyte count, 144/μL vs 132/μL), consistent with better immune status at AIDS onset conveying longer survival. In addition, few persons diagnosed as having AIDS in the 1980-1989 period survived long enough to provide follow-up in the 6- to 10-year period. Also, some people diagnosed as having AIDS in the 1996-2004 period did not have enough time to be followed up in the 6- to 10-year period. As a result, most observation time in the 6- to 10-year period was provided by people diagnosed as having AIDS in the 1990-1995 period (Table 1).

Risks for the 3 AIDS-defining cancers were significantly elevated in years 3 to 5 and 6 to 10 after AIDS onset (Table 2). Risk of KS was extremely high in both time periods (SIRs 5321 and 1347, respectively), and NHL risk was also elevated in both periods (SIRs 32 and 15, respectively). Likewise, risks were substantially higher than in the general population for the AIDS-defining NHL subtypes, namely diffuse large B-cell NHL, Burkitt NHL, and especially CNS NHL (Table 2). Cervical cancer risk was increased after AIDS during years 3 to 5 (SIR, 5.6) and years 6 to 10 (SIR, 3.6).

Among non–AIDS-defining cancers (Table 2), risks were elevated in both the 3- to 5- and 6- to 10-year periods for cancers of the oral cavity and/or pharynx, tongue, anus, liver, larynx, lung and/or bronchus, and penis and for Hodgkin lymphoma. Compared with the general population, risks for anal cancer were especially high: SIR, 27 in the 3- to 5-year period and 40 in the 6- to 10-year period. Among Hodgkin lymphoma subtypes, risk was most elevated for the mixed cellularity subtype (SIRs, 15 in years 3-5; 17 in years 6-10). Other malignancies for which risk was increased in only 1 of the 2 periods after AIDS onset included cancers of the lip and vagina and/or vulva, multiple myeloma, lymphocytic leukemia, and myeloid and/or monocytic leukemia (Table 2). In both time periods, risk was lower than in the general population for cancers of the breast and prostate; risk was lower for kidney cancer during years 3 to 5 after AIDS onset. Overall, for all non–AIDS-defining cancers combined, risk was significantly elevated after AIDS onset in both periods (SIRs, 1.7 in years 3-5; 1.6 in years 6-10).

Table 3 compares cancer incidence in the 3 to 10 years after AIDS onset for the pre-HAART era (1990-1995) and HAART era (1996-2006). Incidence of KS declined by 80% (RR, 0.2; 95% CI, 0.2%-0.2%) between the pre-HAART and HAART eras. A joinpoint model fitted to the incidence data revealed that this change corresponded to a steep decline during the 1994-1997 period (44.3% per year, Figure 1A). Similarly, NHL incidence was 70% lower in the HAART era (RR, 0.3; 95% CI, 0.2-0.3) than in the pre-HAART era, and the joinpoint model indicated that this corresponded to a 36.9% annual reduction during the 1995-1998 period (Figure 1B). Declines were similar in magnitude for diffuse large B-cell NHL and CNS NHL (Table 3 and Figure 1C and D). In contrast, the change in Burkitt NHL incidence was weaker (RR, 0.6; 95% CI, 0.4-1.0), and cervical cancer incidence did not change significantly (RR, 0.8; 95% CI, 0.5-1.2). During the HAART era, persons with AIDS continued to have significantly higher risk for all AIDS-defining cancers than people in the general population (Table 3).

Among non–AIDS-defining malignancies, anal cancer exhibited a 3-fold increase in incidence between the pre-HAART and HAART eras (RR, 2.9; 95% CI, 2.1-4.0) (Table 3). During the HAART era, anal cancer risk remained significantly elevated relative to the general population (SIR, 32). The joinpoint model for anal cancer demonstrated a 10.7% annual increase in incidence over the entire 1990-2006 period despite a suggested decline in the most recent years (Figure 2A). In a separate analysis among men, anal cancer incidence increased 11.3% per year (95% CI, 5.9%-16.8%) during the HAART era (1996-2006). Incidence of local anal cancer increased significantly (14.1%; 95% CI, 7.3%-21.3%) among men during the same years. Regional and distant anal cancer increased annually among men during the same period (5.7%; 95% CI, −1.0% to 12.9%), although this increase was not significant. There were no anal cancer cases among women prior to 1996. During the 1996-2006 period, anal cancer increased 4.3% per year (95% CI, −10.5% to 21.5%) among women.

Hodgkin lymphoma incidence increased significantly between the pre-HAART and HAART eras (RR, 2.0; 95% CI, 1.3-2.9) (Table 3). The Hodgkin lymphoma joinpoint model showed a 14.1% annual increase during the 1990-2000 period, with no significant change subsequently (Figure 2B). The increase in Hodgkin lymphoma incidence was most apparent for the mixed cellularity subtype. During the HAART era, the risk of Hodgkin lymphoma remained significantly elevated relative to the general population (SIR, 11).

Incidence of lung cancer, the most common non–AIDS-defining cancer, was marginally lower in the HAART era than in the pre-HAART era (RR, 0.8; 95% CI, 0.6-0.9) (Table 3). However, this decline over time was not manifest in the joinpoint model (Figure 2C). There were increases in the HAART era for other less common non–AIDS-defining cancers including cancers of the tongue (RR, 2.9) and prostate (RR, 1.6).

Overall, the incidence of all non–AIDS-defining cancers increased 20% between the pre-HAART and HAART eras (RR, 1.2; 95% CI, 1.0-1.3) (P = .01). This increase corresponded to a 4.3% annual increase during the 1990-2006 period, although there was a suggestion that incidence declined in the most recent few years (Figure 2D). During the HAART era, people with AIDS had a 60% higher risk for non–AIDS-defining cancers than the general population (SIR, 1.6; 95% CI, 1.6-1.7).

To our knowledge, this study represents one of the longest and most complete cancer risk follow-up studies of persons with AIDS. Our population-based data demonstrate that people with AIDS in the United States continue to be at elevated risk for a spectrum of AIDS-defining and non–AIDS-defining cancers years after AIDS diagnosis. While the incidence of KS and NHL has declined substantially in the HAART era, the incidence of non–AIDS-defining cancers overall has risen. Two malignancies in particular, anal cancer and Hodgkin lymphoma, have increased in incidence in recent calendar years.

Among HIV-infected individuals, HAART use has been associated with major decreases in KS and NHL risk, and these decreases have translated into declines in incidence measurable at the population level.^2,14,24,25 We demonstrated that while persons with AIDS remain at considerable risk for KS, incidence of this malignancy first started to decline in 1995, possibly due to use of early combination antiretroviral therapy (eg, dual nucleoside therapy) and continued to decline during the first years when HAART was available. Likewise, steep declines in the incidence of diffuse large B-cell NHL and CNS NHL may be partly explained by introduction of HAART in 1996. Human immunodeficiency virus–induced immunosuppression, indicated by low CD4 count, is directly related to risk of KS and these NHL subtypes, and declining incidence in people with long-standing AIDS is plausibly linked to the effectiveness of HAART in improving immune function.²⁶ In contrast, we noted a less dramatic decline in the incidence of Burkitt NHL, the incidence of which is not closely associated with CD4 count.²⁶ Despite overall declines in NHL incidence, NHL was the most common malignancy found during the HAART era.

Notable excess risks were observed for cancers of the anus, penis, cervix, vagina and vulva, and certain sites in the head and neck (oral cavity, pharynx, and tongue), all of which can be attributed to persistent infection with oncogenic HPV subtypes.^5,27 Anal cancer risk was strongly elevated up to 10 years after AIDS onset and, consistent with other reports,^12,28,29 we found that anal cancer incidence increased in the HAART era. This rise was likely not due to only increased anal cancer screening, which has been recommended for HIV-infected MSM³⁰: the increasing trend was observed for both men and women and for regionally advanced disease (which would be detected without screening) as well as for localized disease. Instead, the continued rise of anal cancer incidence in recent years, even in the presence of widespread HAART use, suggests that the key steps susceptible to immune control might have occurred years earlier and that prolonged survival among people with AIDS has now allowed for the manifestation of invasive cancer. While some investigators have advocated anal cancer screening among HIV-infected people,^30,31 it remains controversial.³²

A similar explanation related to the long latency of the carcinogenic effects of HPV may explain why cervical cancer incidence in women has not declined in the HAART era. Women with HIV should be screened annually for cervical cancer.³³ A prophylactic HPV vaccine is now available to prevent cervical infection, but its impact will not be immediately realized, and its efficacy in preventing infection in HIV-positive women or in preventing anal infection is unknown.³⁴

Hodgkin lymphoma risk was also substantially elevated among people with AIDS, and we noted a doubling in Hodgkin lymphoma incidence in this population in the HAART era. Our joinpoint model demonstrated that this increase was largely due to a significant rise throughout the 1990-2000 period. Mixed cellularity Hodgkin lymphoma was the most common subtype and, in the context of AIDS, is often associated with Epstein-Barr virus infection.^1,35 Interestingly, Biggar et al³⁵ reported that AIDS-related immunosuppression appears to have a nonlinear effect on Hodgkin lymphoma risk. Specifically, Hodgkin lymphoma risk increases as the CD4 count declines, reaching a peak when the CD4 count is approximately 225 to 249 lymphocytes/μL, and then falls again at very low CD4 counts.³⁵ A subsequent study did not confirm this nonlinear relationship.³⁶ Nonetheless, among people with advanced AIDS, it is possible that use of increasingly effective HIV therapies during the HAART era has resulted in a shift in immune function associated with higher Hodgkin lymphoma risk.

Importantly, we observed a 20% overall increase in the incidence of non–AIDS-defining cancers in this population. Because cancer risk was calculated as a rate (ie, number of events per unit of follow-up time), this rise in incidence cannot be explained simply by a decline in AIDS-related mortality and a consequent increase in time at risk for cancer. The median attained age increased from 33 to 44 years over the 24-year calendar period of the study, reflecting aging of the US AIDS population with improved survival. Nonetheless, our multivariable analyses comparing cancer risk in different calendar periods adjusted for attained age. Thus, the RRs that we report correspond to changes in cancer risk that cannot be explained by aging within the cohort. Instead, we believe that the recent overall increase in incidence of non–AIDS-defining cancers is largely driven by the rises in anal cancer and Hodgkin lymphoma incidence. Indeed, when those malignancies were excluded, the change in incidence of all non–AIDS-defining cancers between the pre-HAART and HAART eras was no longer apparent (RR, 1.0; 95% CI, 0.9-1.1).

Other specific non–AIDS-defining cancers deserve brief comment. Lung cancer was the most frequent non–AIDS-defining malignancy, accounting for 41% of all non–AIDS-defining cancers. While lung cancer incidence declined slightly with widespread HAART use, persons with AIDS remain at higher risk for lung cancer than the general population. Among HIV-infected individuals, the high risk of lung cancer is partly explained by tobacco use, but HIV may amplify the carcinogenic effects of smoking.³⁷ Liver cancer risk was elevated among people with longstanding AIDS, reflecting the carcinogenic effects of hepatitis C and B viruses and alcohol use.⁷ Finally, people with longstanding AIDS had a lower risk of prostate and breast cancers than the general population, as has been noted in other studies,^16,38 and we found that prostate cancer incidence increased between the 1990-1995 and 1996-2006 periods. These patterns could reflect hormonal effects of HIV infection or other factors that may have changed over time, or could partly reflect low rates of screening compared with the general population.³⁸

The strengths of our study include its large size and population-based nature, representing all major US regions and demographic groups affected by the AIDS epidemic, as well as the long period of evaluation of cancer risk (up to 10 years after AIDS onset). While we could not verify that people with AIDS remained in the cancer registry surveillance catchment areas, we adjusted for plausible rates of out-migration. A limitation of our study is that we lacked individual-level data on important cancer risk factors such as smoking and HAART use. Nonetheless, our results accurately reflect the aggregate impact of these factors and changes in their prevalence over time at the population level.

In summary, individuals with AIDS remain at substantially increased risk for cancer for up to 10 years after AIDS onset. Declines in KS and NHL incidence in recent years can be attributed to introduction and wide availability of HAART, but the continuing occurrence of these cancers in people with AIDS points to the need for improvements in access to HAART, more effective regimens targeting drug-resistant HIV strains, and perhaps interventions to boost immune restoration even with HAART use.^39-41 Of concern, we also observed an increasing incidence of anal cancer and Hodgkin lymphoma. As persons with AIDS continue to live longer after an AIDS diagnosis and as they age, it is possible that cancer risk will increase further.

Correspondence: Eric A. Engels, MD, MPH, Infections and Immunoepidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, 6120 Executive Blvd, EPS 7076, Rockville, MD 20892 (engelse@mail.nih.gov).

Accepted for Publication: January 28, 2010.

Author Contributions: Dr Simard had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Simard, Pfeiffer, and Engels. Acquisition of data: Simard and Engels. Analysis and interpretation of data: Simard, Pfeiffer, and Engels. Drafting of the manuscript: Simard. Critical revision of the manuscript for important intellectual content: Simard, Pfeiffer, and Engels. Statistical analysis: Simard and Pfeiffer. Obtained funding: Engels. Administrative, technical, and material support: Simard. Study supervision: Engels.

Financial Disclosure: None reported.

Funding/Support: This study was supported by the Intramural Research Program of the National Cancer Institute, National Institutes of Health, Bethesda, Maryland.

Additional Contributions: We thank the staff at the HIV/AIDS and cancer registries at the following locations: Colorado; Connecticut; Florida; Illinois; Georgia; Massachusetts; Michigan; New Jersey; New York, New York; Los Angeles, San Diego, and San Francisco, California; Seattle, Washington; Texas; and Washington, DC. Tim McNeel, BA, assisted with database management.

This article was corrected online for typographical errors on 8/9/2010.