Incidence of Types of Cancer among HIV-Infected Persons Compared with the General Population in the United States, 1992–2003 FREE

—The Editors

The advent of highly active antiretroviral therapy (HAART) has significantly improved the survival rate of persons with HIV in the United States (1–4) and led to a shift in the natural spectrum of HIV disease. Incidence of AIDS-defining infections and cancer has decreased (4–8), whereas the incidence of many chronic conditions, such as lipodystrophy, the metabolic syndrome, osteopenia, and cardiovascular disease, has increased (9–13). The incidence of several types of non–AIDS-defining cancer has also increased (14–42, 42). However, data from U.S. studies are frequently limited to persons with AIDS or small, demographically limited cohorts. In addition, few reports of cancer incidence have included persons at all stages of HIV infection or trends in cancer incidence after the introduction of HAART (30–34, 38).

To describe cancer diagnoses in persons at all stages of HIV infection, we analyzed data from 2 large prospective cohort studies in the United States: the Adult and Adolescent Spectrum of HIV Disease (ASD) Project and the HIV Outpatient Study (HOPS). We compared incidence rates of cancer among these persons with incidence rates in the general population, derived from the Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute (44–45) for 1992 to 2003. In addition, we determined incidence trends for selected types of cancer among persons at all stages of HIV infection and in the general population for 3 periods defined on the basis of HAART availability: 1992 to 1995 (pre-HAART), 1996 to 1999 (early HAART), and 2000 to 2003 (recent HAART). Finally, we examined the risk factors that could lead to increased rates of selected cancer types among HIV-infected persons.

The ASD Project was a multicenter, prospective, observational surveillance project established by the Centers for Disease Control and Prevention (CDC) in 1990 in collaboration with local health departments, including affiliated health care centers in 11 geographical areas (Atlanta, Georgia; Los Angeles, California; Dallas, Houston, and San Antonio, Texas; Denver, Colorado; Detroit, Michigan; New Orleans, Louisiana; New York City, New York; Seattle, Washington; and Bayamón, Puerto Rico). Detailed methods for the ASD Project are described elsewhere (46). Participating medical facilities identified HIV-infected patients age 13 years or older. The investigators conducted an initial medical record review for the 12 months preceding enrollment to document sociodemographic characteristics; HIV risk factors; occurrence of AIDS-defining opportunistic illnesses according to the 1993 revised AIDS surveillance definition from the CDC (47); other illnesses, including cancer; medication use; CD4 lymphocyte counts and other relevant laboratory test results; and hospitalizations or other uses of the medical care system. Thereafter, investigators conducted follow-up medical record reviews every 6 months to collect matching interval data. The ASD Project collected detailed information about cancer diagnoses, including anatomical site; morphology; and International Classification of Diseases, Ninth Revision, codes. The investigators enrolled more than 60 000 HIV-infected persons and collected more than 180 000 person-years of observation.

The HOPS is an ongoing, multicenter, prospective, observational cohort study that has continuously recruited and followed HIV-infected patients age 18 years or older since 1992 at 9 HIV specialty clinics in 8 U.S. cities (Oakland and San Leandro, California; Denver, Colorado; Tampa, Florida; Chicago, Illinois; Stonybrook, New York; Philadelphia, Pennsylvania; and Washington, DC). Detailed methods for HOPS are described elsewhere (1, 48). Data from physician–patient interactions are electronically collected at the time of clinical encounter and then submitted for central processing and analysis. Data collected include sociodemographic characteristics, HIV risk factors, symptoms, all diagnoses (both definitive and presumptive), medications prescribed, CD4 lymphocyte counts, and all other laboratory test results. Cases of cancer are defined by using clinical information about new cancer diagnoses. The HOPS investigators have enrolled over 8500 HIV-infected persons and collected more than 32 000 person-years of observation with more than 280 000 visits.

The SEER program is an ongoing active and passive cancer surveillance system that collects incidence and survival data from 13 geographically defined, population-based central cancer registries in the United States (Oakland, San Francisco, San Jose, and Los Angeles, California; Seattle, Washington; Detroit, Michigan; Atlanta, Georgia; rural Georgia; Connecticut; Iowa; New Mexico; Utah; Alaska; and Hawaii) (45). Geographical areas are selected for inclusion in the SEER program on the basis of ability to operate and maintain a high-quality, population-based cancer reporting system and presence of epidemiologically significant population subgroups. The SEER program includes HIV-infected persons who probably represent fewer than 1% of the total population. We excluded SEER data from Alaska, Hawaii, and rural Georgia from this analysis to best match SEER registries to the ASD Project and HOPS areas. The SEER program defines cancer cases by using anatomical site and histology codes.

For this study, we analyzed data on cases of cancer diagnosed from 1992 to 2003. To accurately calculate incidence, we excluded persons in whom cancer was diagnosed within the year before enrollment (prevalent cases), as well as participant observation time after cancer diagnosis. However, we included cases of cancer occurring after cancer of a different type and corresponding participant observation times. Definitions of cancer categories for all cancer types observed in this analysis were established on the basis of the anatomical site of primary significance and International Classification of Diseases, Ninth Revision, codes (Appendix Table 1). We calculated incidence rates for types of cancer that occurred at least 5 times in the combined ASD Project and HOPS population during the study period (Appendix Table 2). Preliminary examination of the combined ASD Project and HOPS data indicated that the 3 AIDS-defining cancer types and 9 non–AIDS-defining cancer types occurred most frequently (>25 incident cases); we included these types in the trend and multivariable risk factor analyses. We had sufficient power (≥80%) to detect 1.5-fold differences in incidence over time for 6 of the cancer types examined: colorectal, Hodgkin lymphoma, lung, cervical, Kaposi sarcoma, and non-Hodgkin lymphoma. We excluded recurrences, metastases, and cases of in situ cancer from the analysis.

We calculated incidence rates for persons 15 to 84 years of age as the collective number of newly diagnosed cancer cases annually for the overall analysis period (1992 to 2003) and the pre-HAART (1992 to 1995), early HAART (1996 to 1999), and recent HAART (2000 to 2003) subperiods to account for the transition to and stabilization of HAART use. Incidence rates were expressed per 100 000 prospective person-years of observation during these same periods. Person-years varied slightly for each cancer type due to the exclusion of prevalent cases. We computed person-years at risk for cancer after HIV diagnosis from date of enrollment. Termination was the date of cancer diagnosis, last follow-up visit, or death. We examined incidence rates in the ASD Project and HOPS separately and found them to be similar (Appendix Table 2). Because cancer incidence estimated from the ASD Project and HOPS databases were not statistically different, we combined their data to measure cancer incidence among the HIV-infected population captured by both of these studies. In addition, because the ASD Project and HOPS samples are demographically different, the combined ASD Project and HOPS database is more representative of the HIV-infected population than either study alone. We determined cancer incidence rates for the general population by using data from SEER; the SEER rates were directly standardized according to the distribution of 14 age groups, 3 race groups, and sex of the ASD Project and HOPS population. We derived 95% CIs by assuming that the number of cancer cases in each standardization group followed a Poisson distribution (49). To compare the incidence of cancer in the HIV-infected population with that in the general population, we calculated standardized rate ratios for the entire observation period (1992 to 2003) because annual data were too sparse (50–51).

To evaluate trends over time in incidence of specific types of cancer in the HIV-infected population, we used multivariable Poisson regression in a model controlling for age (15 to 34 years, 35 to 59 years, ≥60 years), race (white, black, other), sex, HIV risk group (men who have sex with men vs. other), nadir CD4 count (<0.200 × 10⁹ cells/L, 0.200 to 0.499 × 10⁹ cells/L, ≥0.500 cells/L), and antiretroviral therapy (any vs. none). The linear trend denotes the estimated average change in incidence rate from pre-HAART to early HAART and from early to recent HAART. To evaluate trends over time in the incidence of a specific type of cancer in the SEER population, we used a weighted multivariable Poisson regression to assess linear trends in a model weighted according to the age, race, and sex distribution of the combined ASD Project and HOPS population. To determine whether cancer incidence in the HIV-infected population changed over time relative to the general population, we included a 2-way interaction term between trend and study sample in the multivariable model. To describe factors associated with each type of cancer among HIV-infected persons, we used multivariable Poisson regression, including the covariates as already defined for linear trend, age, race, sex, HIV risk group, nadir CD4 count, antiretroviral use, and hepatitis B or C infection (liver cancer only). We chose these covariates because they had a known association with cancer or were markers of immune function, and we adjusted for them simultaneously.

Both the ASD Project and HOPS have been reviewed annually since their inception by the CDC and each participating site's institutional review board. The ASD Project was granted a waiver of informed consent by the relevant institutional review boards; informed consent was obtained and renewed annually for HOPS. The institutional review board of the National Cancer Institute reviewed SEER and granted the program a waiver of informed consent.

The CDC supported data collection through cooperative agreements with local health departments (ASD Project) and a contract with the Cerner Corporation (HOPS). The funding source had no role in the design, analysis, or interpretation of the data or in the approval of the manuscript.

From 1992 to 2003, 54 780 HIV-infected persons (47 832 persons from the ASD Project and 6948 from HOPS) contributed 157 819 person-years of observation. The median follow-up time was 2.0 years in the ASD Project and 2.6 years in HOPS. Of the 3550 incidents of cancer identified in the combined ASD Project and HOPS data for this analysis, 2842 (80%) were categorized as AIDS-defining and 708 (20%) as non–AIDS-defining. Of the 27 types of cancer observed, 5 non–AIDS-defining types of cancer occurred seldom or never: penile (4 cases), uterine (2 cases), ovarian (2 cases), gallbladder (0 cases), and biliary (0 cases).

Table 1 shows the age, sex, and race distributions of the combined ASD Project and HOPS and SEER populations. The populations differed in age, race, and sex distributions; HIV-infected persons were more likely to be male (76%) and nonwhite (61%). The age distribution of the combined ASD Project and HOPS population was clustered among young adults (range, 25 to 50 years), whereas the age distribution of the SEER population was more evenly distributed. Distribution of observation time was evenly divided among the 3 periods for both populations (Table 1).

The incidence of several non–AIDS-defining types of cancer was significantly higher in the HIV-infected population than in the general population: anal (standardized rate ratio [SRR], 42.9 [95% CI, 34.1 to 53.3]), vaginal (21.0 [CI, 11.2 to 35.9]), Hodgkin lymphoma (14.7 [CI, 11.6 to 18.2]), liver (7.7 [CI, 5.7 to 10.1]), lung (3.3 [CI, 2.8 to 3.9]), melanoma (2.6 [CI, 1.9 to 3.6]), oropharyngeal (2.6 [CI, 1.9 to 3.4]), leukemia (2.5 [CI, 1.6 to 3.8]), colorectal (2.3 [CI, 1.8 to 2.9]), and renal (1.8 [CI, 1.1 to 2.7]). Incidence was significantly lower for prostate cancer (standardized rate ratio, 0.6 [CI, 0.4 to 0.8]) (Appendix Table 2 and Appendix Figures 1 and 2). We found no significant difference between the 2 populations in the rates of other types of cancer we examined.

The analysis of trends over time in cancer rates for the HIV-infected population indicated that incidence rates decreased significantly for Kaposi sarcoma and non-Hodgkin lymphoma and increased significantly across the 3 periods for anal, prostate, and colorectal cancer; melanoma; and Hodgkin lymphoma (Table 2; Appendix Tables 3 and 4 and Appendix Figure 3). We observed no change in the rates of cervical cancer and several non–AIDS-defining types of cancer (liver, lung, oropharyngeal, and breast cancer).

Among the general population, standardized incidence rates for Kaposi sarcoma, non-Hodgkin lymphoma, cervical cancer, lung cancer, and oropharyngeal cancer decreased significantly across the 3 periods (Table 2). Incidence rates for anal, colorectal, liver, and prostate cancer increased significantly across the 3 periods, and incidence rates for Hodgkin lymphoma, melanoma, and breast cancer remained unchanged.

The relative incidence of Kaposi sarcoma and non-Hodgkin lymphoma among our HIV-infected study population decreased over time compared with the general population (Table 3; Appendix Table 5 and Appendix Figure 4). Conversely, the higher overall relative incidence of anal cancer in the HIV-infected population increased over time, with the adjusted rate ratio almost doubling during the analytic period.

Acquisition of HIV through male–male sex was associated with increased risk for Kaposi sarcoma (relative risk, 2.88; P < 0.001) and non-Hodgkin lymphoma (1.53; P < 0.001). Antiretroviral therapy was independently associated with decreased risk for Kaposi sarcoma (relative risk, 0.61; P < 0.001), non-Hodgkin lymphoma (0.68; P < 0.001), cervical cancer (0.48; P = 0.019), breast cancer (0.35; P = 0.013), colorectal cancer (0.50; P = 0.027), and lung cancer (0.52; P < 0.003). A low nadir CD4 count was associated with increased risk for Kaposi sarcoma (relative risk, 8.34; P < 0.001), non-Hodgkin lymphoma (6.03; P < 0.001), cervical cancer (3.70; P = 0.010), anal cancer (5.82; P = 0.017), colorectal cancer (6.27; P = 0.013), and lung cancer (2.42; P = 0.017). Co-infection with hepatitis B or C was associated with increased risk for liver cancer (relative risk, 3.63; P < 0.001).

In the largest analyses of cancer incidence trends among HIV-infected persons in the United States, we observed significantly higher rates of several types of cancer from 1992 to 2003 among HIV-infected persons than in the general population. Non–AIDS-defining types of cancer with higher incidence rates were anal, colorectal, liver, lung, oropharyngeal, renal, and vaginal cancer; Hodgkin lymphoma; leukemia; and melanoma. These findings are consistent with previous reports of cancer incidence among persons with HIV or AIDS (Table 4). We extend these previous findings by examining cancer incidence trends over time among persons with all stages of HIV infection, showing that incidence rates increased significantly for melanoma; Hodgkin lymphoma; and colorectal, anal, and prostate cancer—despite the advent of HAART. Immune dysfunction (52); concomitant infection with oncogenic viruses (53–55); and lifestyle factors, such as smoking, may account for the higher cancer incidence among HIV-infected persons.

Immunosuppression may accelerate the progression of melanoma and other types of cancer in individuals who are already predisposed to them, as described in studies of transplant recipients (56–62). In our study, a low nadir CD4 cell count was associated with significantly increased risk for colorectal cancer, whereas use of antiretroviral therapy significantly decreased risk, suggesting that pathogenesis of colorectal cancer may be immune-mediated; this finding is consistent with previous reports (63). However, incidence rates for melanoma and colorectal cancer increased significantly across the 3 periods, suggesting a role of other contributing factors (such as behavioral or lifestyle factors).

Although survival in Hodgkin lymphoma has improved, rates of Hodgkin lymphoma have increased in the HAART era (64). The strong association of Hodgkin lymphoma with Epstein–Barr virus infection in HIV-infected individuals and the influence of immunosuppression may explain this finding (35); Righetti and colleagues (65) suggested that immune reconstitution while receiving HAART increases B-cell stimulation and the number of Epstein–Barr virus–infected cells, which may in turn increase risk for Epstein–Barr virus–associated cancer.

Anal cancer is the only type that increased in both incidence among HIV-infected persons and relative incidence compared with the general population over time. Although this finding is concerning, it is not surprising. The predominance of men who have sex with men among the HIV-infected population and the resultant higher prevalence of anal human papillomavirus (HPV) infection are associated with anal intraepithelial neoplasia (66–67). The interaction between HIV and HPV allows for persistence of HPV infection in HIV-infected persons, who are more commonly infected with the oncogenic HPV subtypes 16 and 18, leading to development of dysplasia (14, 68–69). Because HAART does not alter the incidence or progression of anal intraepithelial neoplasia (20, 70), persons who are successfully treated with HAART but are co-infected with HIV and HPV are expected to remain at greater risk for anal cancer over time and incidence rates are expected to increase as HIV-infected persons live longer.

We found a lower rate of prostate cancer among HIV-infected persons than in the general population, in contrast to smaller published reports (19, 40). However, our results were consistent with those from an AIDS–cancer registry match (32) in which the investigators attributed the lower observed risk in the general population to less prostate cancer screening. In our analyses, rates of prostate cancer among HIV-infected persons increased over time but remained consistently lower than in the general population. We know of no reason why HIV-related immunosuppression would decrease prostate cancer risk (71). Differential screening among men with and without HIV may explain these results, although many men with HIV are under closer medical supervision. In addition, persons with HIV are known to have lower androgen levels, which may in turn decrease their risk for prostate cancer. Because androgens have long been known to contribute to the risk for prostate cancer (19), patients receiving replacement therapy with exogenous testosterone or anabolic steroids may need to be monitored carefully.

Our study has limitations. First, the case ascertainment standard for SEER is 98% (44), whereas case ascertainment in the ASD Project and HOPS is undoubtedly lower (75% to 85%, as determined by an ASD Project–local cancer registry match in select project areas). In addition, the completeness of cancer ascertainment has not been formally evaluated in the ASD Project or HOPS. Because the HIV incidence rate is in the numerator, the potential bias would be to underestimate the standardized rate ratio—suggesting that standardized rate ratios showing increased incidence are minimum estimates of the rate disparity. Standardized rate ratios may be further underestimated by double-counting of cases in areas where SEER registries overlap with the ASD Project or HOPS sites, because SEER does not collect data on HIV status. Second, although the ASD Project and HOPS have large cohorts, they are not representative of all persons with HIV infection in the United States, and SEER is not representative of the general U.S. population; our findings may therefore have limited generalizability. The SEER program covers 14% of the U.S. population, has a higher proportion of foreign-born persons, and tends to be more urban than the general population (44). In a comparison of SEER data with data from the National Program of Cancer Registries, the incidence rates of head or neck cancer, anal cancer, and Hodgkin disease were similar; however, the rate of lung cancer was slightly underestimated and the rates of melanoma and liver cancer were slightly overestimated (72). The ASD Project and HOPS also represent different geographical areas from SEER. Third, we had inadequate information on smoking behavior; because the prevalence of smoking varies among specific groups in the United States (for example, by sex, race, or HIV risk), some characteristics associated with cancer in our analyses may have been surrogates for smoking status. Finally, we could not account for changes in screening practices over time.

In conclusion, our findings indicate that HIV-infected persons are at higher risk than the general population for many non–AIDS-defining types of cancer. In addition to encouraging tobacco cessation, HIV care providers should be aware of these elevated risks and screen for preventable diseases, such as cervical and colorectal cancer (73–75). Screening programs for early detection and treatment of precancerous anal lesions should be evaluated and will probably become more important as the HIV-infected population ages and lives longer. Furthermore, primary prevention strategies to reduce HPV infection and HPV-associated diseases, such as vaccination and circumcision, warrant further evaluation.

AIDS Research Consortium of Atlanta, Atlanta, Georgia: Melanie Thompson, MD, and Erica Sinclair, MS.

Denver Department of Health and Hospitals, Denver, Colorado: David Cohen, MD; Arthur Davidson, MD; and Cornelius Rietmeijer, MD.

Los Angeles County Department of Health Services, Los Angeles, California: Wendy Garland and Amy Wohl, PhD.

Louisiana Department of Health, New Orleans, Louisiana: Anne Morse, BS.

Michigan Department of Health, Detroit, Michigan: Linda Wotring, PhD, and Eve Mokotoff, MPH.

New York City Department of Health, New York City, New York: Chris Murrill, PhD, and Marie Antoinette Bernard, MD.

Universidad Central de Caribe, Bayam|$$|Aaon, Puerto Rico: Alejandro Amill; Robert Hunter, MD; and Maria de los Angeles Gomez, PhD.

Puerto Rico Department of Health, San Juan, Puerto Rico: Sandra Miranda.

Seattle-King County Department of Health, Seattle, Washington: Susan Buskin, PhD, and Elizabeth Barash.

Texas Department of Health, Austin, Texas: Sylvia Odem, MPH.

Parkland Hospital, Dallas, Texas: Philip Keiser, MD.

Department of Health and Human Services, Houston, Texas: Adebowale Awosika-Olumo, MD.

Cerner, Vienna, Virginia: Kathleen C. Wood and Rose K. Baker.

Northwestern University Medical School, Chicago, Illinois: Frank J. Palella, Joan S. Chmiel, and Janet Cheley.

Rose Medical Center, Denver, Colorado: Kenneth A. Lichtenstein, Kenneth S. Greenberg, Benjamin Young, Barbara Widick, Cheryl Stewart, and Peggy Zellner.

Infectious Disease Research Institute, Tampa, Florida: Bienvenido G. Yangco and Kalliope Halkias.

Dupont Circle Physicians Group, Washington, DC: Douglas J. Ward and Charles A. Fiorentino.

State University of New York, Stonybrook, New York: Jack Fuhrer, Linda Ording-Bauer, Rita Kelly, and Jane Esteves.

Temple University Hospital, Philadelphia, Pennsylvania: Ellen M. Tedaldi, Ramona Christian, and Faye Ruley.

Adult Immunology Clinic, Oakland, California: Joseph B. Marzouk, Roger T. Phelps, and Mark Rachel.

Fairmont Hospital, San Leandro, California: Robert E. McCabe and Mark Rachel.

University of Illinois at Chicago, Chicago, Illinois: Richard M. Novak, Jonathan P. Uy, and Andrea Wendrow.