Background: Previous studies describe decreased prostate cancer risk in HIV-infected men. In the United States, prostate-specific antigen (PSA) screening is common and increases the detection of prostate cancer. We evaluated whether the prostate cancer deficit among men with AIDS reflects differential PSA screening.
Methods: Data from the U.S. HIV/AIDS Cancer Match Study were used to calculate standardized incidence ratios (SIR) for prostate cancer, comparing men with AIDS (N = 287,247) to the general population. Furthermore, we estimated PSA testing rates in the Johns Hopkins HIV Clinical Cohort.
Results: Prostate cancer rates increased over time in the general population and, beginning in the 1990s, were consistently higher than among men with AIDS. Men with AIDS had the same prostate cancer risk as the general population in the pre-PSA era (<1992, SIR = 1.00), but significantly reduced risk during the PSA era overall (1992-2007, SIR = 0.50) and across age, race, HIV risk group, antiretroviral therapy era, and CD4 counts. Local and regional stage prostate cancer risk was lower among men with AIDS (SIRs, 0.49 and 0.14, respectively), but distant stage cancer risk did not differ (SIR = 0.85). Among HIV-infected men ≥40 years old, PSA testing was uncommon (18.7% per year), but increased 2.4-fold from 2000 to 2008, after age adjustment.
Conclusion: Prostate cancer risk was decreased by 50% among men with AIDS compared with the general population. This deficit was limited to the PSA era and early stage cancers.
Impact: Our findings suggest that the prostate cancer deficit in HIV-infected men is largely due to differential PSA screening. Cancer Epidemiol Biomarkers Prev; 19(11); 2910–5. ©2010 AACR.
People with HIV infection are at an increased risk of certain cancers, including both acquired immune deficiency syndrome (AIDS)–defining conditions (i.e., Kaposi sarcoma, non–Hodgkin lymphoma, and invasive cervical cancer) and non–AIDS-defining cancers (e.g., anal cancer, lung cancer, and Hodgkin lymphoma; refs. 1-6). In contrast, several epidemiologic studies have shown a significant inverse association between HIV and prostate cancer risk (2, 3, 5, 7-12). Two meta-analyses reported an overall 30% reduction in prostate cancer risk in HIV-infected men compared with the general population (5, 6).
The reason for the prostate cancer deficit in HIV-infected men is unclear. It has been hypothesized that HIV infection may prevent the development of prostate cancer through reductions in androgen levels (3, 6). Alternatively, the deficit may be an artifact driven by low rates of screening with the prostate-specific antigen (PSA) test among HIV-infected men (3, 6, 12).
The U.S. Food and Drug Administration approved the use of the PSA test for the monitoring of men with prostate cancer in 1986 and for the diagnosis of prostate cancer in 1992 (13). Subsequent widespread PSA screening in the general U.S. population resulted in a dramatic increase in prostate cancer incidence due to the diagnosis of early stage prostate cancers that were not clinically detectable (14). According to data from the 2000 National Health Interview Survey, 57% of U.S. men ages 50 years or older had previously had a PSA test (15). Currently, there are no published estimates of the prevalence of PSA screening among HIV-infected men. However, if the prevalence of PSA screening in HIV-infected men is substantially lower than the U.S. general population, a corresponding deficit in prostate cancer would be anticipated.
In this study, we explored the epidemiology of prostate cancer among HIV-infected men in the United States. Using data from the HIV/AIDS Cancer Match (HACM) Study, we evaluated prostate cancer risk among men with AIDS, both overall and according to demographic subgroups and cancer stage, to assess whether the patterns observed were consistent with the hypothesis that any deficit in prostate cancers among men with HIV is driven by differential rates of PSA screening. Additionally, we assessed the effect of prostate cancer on survival among men with AIDS. Finally, because nationally representative data on PSA screening among HIV-infected men were not available, we evaluated the frequency of PSA testing in an urban cohort of HIV-infected men in Baltimore, MD.
Materials and Methods
Our analyses used data from the HACM Study, a linkage of population-based HIV/AIDS and cancer registries in 15 U.S. areas (8). A probabilistic algorithm is used to match records based on name, Social Security number, sex, birth date, death date, and race. Only anonymized data are retained by investigators. The study was approved by institutional review boards at each of the participating registries.
This study included men with AIDS who contributed follow-up from the latter of the date of the start of registry coverage or four months after the date of AIDS onset to the earliest of date of death, the last date of complete registry coverage or 60 months after the date of AIDS onset. The 0- to 3-month period after AIDS diagnosis was excluded from this analysis because increased medical surveillance at the time of AIDS diagnosis artificially inflates cancer rates during this time period (16). Prostate cancers were identified using cancer registry data, specified according to International Classification of Diseases for Oncology (ICD-O-3) topography code C619 (17). Stage was coded by cancer registries using Surveillance, Epidemiology and End Results Program summary stage algorithms.
To estimate the frequency of PSA testing among HIV-infected men, we used data from the Johns Hopkins HIV Clinical Cohort (JHHCC) during 2000 to 2008. The JHHCC was established in 1989 as a longitudinal study of HIV-infected patients who received care from the Johns Hopkins AIDS Service (18). JHHCC participants are primarily low-income African Americans living in Baltimore, MD. Baseline and follow-up demographic characteristics, medical history, and laboratory test results are collected. Information on PSA testing obtained by clinicians was abstracted from electronic medical records. Of note, we were unable to determine if PSA tests were utilized for screening or if they were instead obtained for clinical indication (e.g., urinary obstructive symptoms). Thus, the rate of PSA testing in the JHHCC overestimates PSA screening rates in asymptomatic men.
Using HACM Study data, we examined prostate cancer risk in two calendar periods: the era prior to Food and Drug Administration approval of PSA testing for prostate cancer diagnosis (pre-PSA era, 1980-1991) and the PSA era (1992-2007). To compare the risk of prostate cancer in men with AIDS to the general population, we estimated standardized incidence ratios (SIR) for each attained calendar period. SIRs were calculated as the number of observed cases divided by the number of expected cases, where expected cases were based on prostate cancer incidence rates from cancer registry data and standardized by age, race, calendar year, and registry to the AIDS population. Additionally, we present prostate cancer rates directly standardized to the 2000 U.S. general population by age (5-y categories) and race.
We limited additional subanalyses to person-time contributed during the PSA era (1992-2007). Specifically, prostate cancer incidence rates in the PSA era were calculated across demographic characteristics, antiretroviral therapy calendar period [pre–highly active antiretroviral therapy (HAART) era, 1992-1995; early HAART era, 1996-1999; and late HAART era, 2000-2007], and CD4 count at AIDS diagnosis. Rate ratios (RR) were estimated with Poisson regression. We calculated SIRs within these strata, using expected cases estimated from rates in the general population, standardized by age, race, calendar year, and registry. Furthermore, to examine whether a deficit in prostate cancers in the PSA era reflected screening, we calculated an additional SIR for the PSA era that used pre-PSA historical cancer rates from the Surveillance, Epidemiology and End Results program (1985-1987) for comparison, standardized by age and race. This time period has been previously used to examine prostate cancer rates in the absence of screening, as PSA screening rates in U.S. men began to increase as early as 1988, prior to Food and Drug Administration approval for that purpose (19). For all SIRs, exact two-sided confidence intervals (CI) were calculated.
Finally, we used two approaches to assess whether men with AIDS were diagnosed later with prostate cancer than men in the general population, as well as the effects of cancer diagnosis on survival. First, for men with AIDS during the PSA era, we calculated SIRs for prostate cancer separately according to stage at cancer diagnosis (local, regional, distant, and unknown stage), in which a delay in diagnosis would translate into an elevated risk of advanced stage cancer. Second, proportional hazards regression models were used to evaluate the association between prostate cancer and survival among men with AIDS in the PSA era. Men were followed from 4 months after the date of AIDS diagnosis until the earliest of date of death, the last date of complete registry coverage or 10 years of follow-up. Hazard ratios (HR) for death from all causes were estimated for any diagnosis of prostate cancer, and for localized/regional, distant, and unknown stage prostate cancers. In these models, we treated the diagnosis of prostate cancer and antiretroviral therapy calendar period as time-dependent covariates. Regression models were additionally adjusted for age at AIDS onset, race, and HIV transmission risk group.
JHHCC PSA testing rates were calculated by calendar year and 5-year age group. Poisson regression was used to estimate changes in PSA testing rates over time, adjusting for age. Furthermore, we directly standardized PSA testing rates by age to the 2000 U.S. population (men ≥40 years old) for comparison with general population rates.
Figure 1 shows the incidence rates of prostate cancer in men with AIDS and in the general population, standardized by age and race to the 2000 U.S. general population. From 1988 to 1992, rates overlapped in the AIDS and general populations; however, rates were unstable and based on few cases in men with AIDS. Prostate cancer incidence increased during this time in the U.S. general population, and after 1992, rates were consistently higher than in men with AIDS.
SIRs confirmed that men with AIDS had the same prostate cancer risk as the general population in the pre-PSA era (SIR = 1.00; 95% CI, 0.58-1.60, based on n = 17 prostate cancer cases among 90,803 men followed during 1980-1991), but they had a significantly reduced risk during the PSA era (SIR = 0.50; 95% CI, 0.44-0.57, based on n = 230 cases among 287,247 men followed during 1992-2007). However, when compared with pre-PSA historical rates (1985-1987), men with AIDS during the PSA era had a modestly increased risk of prostate cancer (SIR = 1.17; 95% CI, 1.02-1.33).
Among men with AIDS followed during the 1992 to 2007 PSA era, prostate cancer incidence was 28.3 per 100,000 person-years (Table 1). Prostate cancer risk increased steeply with age, and was highest in black men (incidence RR, 2.13; 95% CI, 1.59-2.84 compared with white men), and among men in “other/unknown” risk groups (RR, 3.11; 95% CI, 1.34-4.13 compared with men who have sex with men). Furthermore, the risk of prostate cancer increased with increasing CD4 cell counts at AIDS (RR, 1.54; 95% CI, 1.05-2.26, comparing CD4 counts ≥200 cells/mm3 to <50 cells/mm3), and was significantly greater during the early HAART (RR, 1.72; 95% CI, 1.16-2.56) and late HAART calendar periods (RR, 3.12; 95% CI, 2.12-4.59) compared with the pre-HAART calendar period.
When compared with the general population, a significant deficit in prostate cancer risk was observed in 40- to 79-year-olds (SIRs, 0.31-0.59) but not in younger men. Deficits were observed across HIV transmission risk groups, with the greatest deficit observed among injection drug users (SIR = 0.29). Furthermore, prostate cancer risk was significantly decreased across racial groups, antiretroviral calendar periods, and regardless of CD4 cell count at AIDS diagnosis, although the deficit seemed to weaken with increasing CD4 count.
When considered by prostate cancer stage, the reduction in risk was limited to local stage (SIR = 0.49; 95% CI, 0.42-0.57) and regional stage (SIR = 0.14; 95% CI, 0.06-0.27) disease. Risk of distant stage prostate cancer was similar between men with AIDS and men in the general population (SIR = 0.85; 95% CI, 0.50-1.34).
Among men with AIDS followed during the PSA era, prostate cancer was associated with a 34% increase in subsequent all-cause mortality (HR = 1.34; 95% CI, 1.05-1.71). More specifically (Table 2), local/regional stage prostate cancer was not associated with death (HR = 0.93), but distant stage prostate cancer was associated with elevated mortality (HR = 4.73). Prostate cancer with an unknown stage at diagnosis was also associated with an increased risk of death (HR = 2.16).
During 2000 to 2008, 1,216 PSA tests were done among men in the JHHCC who were age 40 years and older, corresponding to a rate of 18.7% per year (6,488 person-years of follow-up). Over the full time period, the rate of testing was highest in 60- to 69-year-olds (37.9% per year), followed by 50- to 59-year-olds (28.0% per year) and 40- to 49-year-olds (9.5% per year). There was a significant increasein PSA testing rates during 2000 to 2008 (P trend < 0.0001), with rates more than doubled in 2008 compared with 2000 (RR, 2.4; 95% CI, 1.9-3.2), after age adjustment. The age-standardized rate of PSA screening for 2000 to 2008 was 22.7% per year overall, increasing from 17.1% per year in 2000 to 29.7% per year in 2008.
In our study, men with AIDS had a 50% lower risk of prostate cancer than men in the general population, which is consistent with previous studies that have also observed deficits in prostate cancer in HIV-infected men compared with the general population (2, 3, 5, 7-11). In the general U.S. population, PSA screening among men is very common (20) and has contributed to a steep increase in prostate cancer incidence (14). In contrast, we observed lower rates of PSA testing in a cohort of low-income HIV-infected men in Baltimore, suggesting that PSA testing may be less common in men with HIV. Although previous studies have proposed biological explanations for the reduced risk of prostate cancer in men with AIDS (3, 6), our results suggest that this deficit may be largely driven by differential PSA screening rates in the population with AIDS and in the general population.
PSA screening patterns strongly determine diagnosed rates of prostate cancer. In the United States, prostate cancer rates began to increase following the identification of PSA as a marker of prostate cancer in 1987 (21) and peaked in 1992 (22). Although rates have been stable since 1995, they remain substantially elevated in comparison to rates in 1986 (22). It has been estimated that approximately 40% of prostate cancers diagnosed among men in the United States are detected through screening (19). One study estimated that PSA screening has resulted in the diagnosis of prostate cancer in more than one million additional men in the last 20 years, the majority of whom were age 50 years or older (22). Furthermore, an estimated 67% of all screen-detected prostate cancers are overdiagnoses, meaning that these cancers would never have developed into symptomatic cancers (23).
PSA screening might be more common among men in the U.S. general population than among HIV-infected men. In 2001, 57% of U.S. men ages 40 years and older had a PSA test in the previous year (20). This estimate is much higher, for example, than the age-standardized PSA testing rate of 22.7% per year observed in the JHHCC. Furthermore, our estimates provide an upper bound of PSA screening in the JHHCC because our data captured some PSA tests that might have been obtained for clinical indications rather than for screening in asymptomatic men. We speculate that the low rates of screening reflect the clinicians' perception that the historically poor prognosis of HIV-infected men would limit the utility of the test, coupled with challenges in addressing other medical and social issues in this population. Men in the JHHCC were receiving medical care, and PSA screening rates could be even lower among other HIV-infected populations that lacked access to care, e.g., due to lack of medical insurance. Although annual PSA testing rates in the JHHCC cohort remained lower than the general population, there was a 2-fold increase from 2000 to 2008. Given this pattern, it is likely that the rate of PSA testing was less frequent in the years preceding 2000.
Our results imply that prostate cancer rates have been lower among men with AIDS largely because of the dramatic increase in PSA screening in the general population. Specifically, the deficit in prostate cancer that we observed among men with AIDS was limited to the PSA era. Indeed, PSA era incidence rates of prostate cancer in men with AIDS were actually slightly higher than in the general population when we used comparison rates from before the mass uptake of PSA screening. The explanation for this small excess is unclear, but it could reflect limitations in our approach of using historical rates or residual confounding by other factors for which we could not control in the comparison.
Further arguing that screening patterns largely explain the overall reduced risk of prostate cancer, we observed a significant deficit in localized and regional stage prostate cancers among men with AIDS, but no difference for distant stage prostate cancer compared with the general population. If the deficit in prostate cancer risk had a biological explanation, such as lower androgen levels or a direct effect of HIV itself, one would have predicted uniformly reduced risk across prostate cancer stages. Screen-detected cancers are more likely to be early stage and asymptomatic compared with clinically detected cancers (24). Therefore, the disproportional deficit of localized/regional prostate cancer among men with AIDS adds support to the hypothesis that the difference is explained by low rates of screening. Likewise, the deficit in prostate cancer risk was greatest in groups that would have been least likely to have been screened, including injection drug users and men with very low CD4 cell counts. Additionally, we observed that the deficit in prostate cancers was seen across antiretroviral therapy calendar periods. Further arguing against a role of immune suppression, the risk of prostate cancer in immune-suppressed transplant recipients is no different from the risk in the general population (5).
Of final note, relative survival following prostate cancer diagnosis in men with AIDS was no worse than in prostate cancer patients in the general population. In men with AIDS, localized/regional prostate cancer was not associated with an increased risk of death (Table 2), similar to what is observed in the general U.S. population (25). Even with lower rates of PSA screening, we did not find an elevated risk of advanced stage prostate cancer in men with AIDS, and the 5-fold increase in mortality following distant stage prostate cancer in men with AIDS was similar to that reported for men in the general population (standardized mortality ratio = 5.22; ref. 25). Thus, although the public health benefits of PSA screening in the general population are still under debate (26, 27), our results on mortality following a prostate cancer diagnosis do not suggest that men with AIDS are being harmed by the relatively low screening rates.
The main strength of our study was the use of nationally representative data from the HACM Study, which includes men with AIDS in the United States over the duration of the AIDS epidemic, before and after the introduction of PSA screening. The main limitation of our study was our inability to assess PSA screening prevalence patterns for a larger and more representative HIV population or directly assess prostate cancer incidence in relation to screening. Furthermore, although novel, the PSA testing results from the JHHCC should be interpreted with caution. The JHHCC includes HIV-infected men that are primarily African American and low-income, and is not representative of all HIV-infected men in the United States; thus, the PSA testing rates presented here may not be generalizable to all men infected with HIV.
In conclusion, there is a deficit in prostate cancers among American HIV-infected men, which is limited to local and regional stage cancers that are most likely to be detected by PSA screening. A sharp increase in prostate cancer incidence in the general population has largely been driven by widespread use of PSA screening, a practice that seems to be less frequently utilized for HIV-infected men. Combined, these factors suggest that the deficit in prostate cancer observed among HIV-infected men is largely an artifact due to differential PSA screening practices, and is likely not due to a protective effect of HIV against the development of prostate cancer. Furthermore, it seems that the lack of PSA screening in HIV-infected men has not resulted in an increased risk of advanced stage prostate cancer or death.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
The authors thank the HIV/AIDS and cancer registry staff in Colorado, Connecticut, Florida, Georgia, Illinois, Massachusetts, Michigan, New Jersey, Texas, New York, NY; Los Angeles, San Diego, and San Francisco, CA; Seattle, WA, and Washington, DC. We also thank Tim McNeel for database management.
Grant Support: Intramural Research Program of the National Cancer Institute. The Johns Hopkins HIV Clinical Cohort is supported by the following NIH grants: R01 DA11601, K24 DA00432, R01 AA16893.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Received July 12, 2010.
- Revision received August 4, 2010.
- Accepted August 9, 2010.