Before availability of effective antiretroviral therapy (ART) for persons living with HIV (PLWH), their risk of developing Kaposi sarcoma (KS) was elevated 2800-fold compared with the general population, occurring in up to 30% of PLWH. Wide-scale availability of ART beginning in 1996 had a dramatic impact on the clinical outlook for HIV-infected individuals, and accordingly, US incidence rates of KS in PLWH in the ART era declined 83.5% between 1990 and 2002, with continued declines into the 2000s.1,2
AIDS-related KS incidence varies by HIV geography and socioeconomic status,3 with more recent studies in the ART era finding declining incidence is not uniform across the United States. One analysis using the National Cancer Institute's (NCI's) Surveillance, Epidemiology, and End Results (SEER) database found overall KS incidence 2005 through 2009 was positively associated with geographic areas of higher poverty.4 A more recent SEER-based study (2000–2013) found AIDS-related KS disproportionately affects African American men, particularly in the south.5 These studies have begun to shed light on persistent and potentially widening disparities in KS incidence and survival. However, as SEER is designed only to capture a broadly representative sample of the US population and does not include data from most states, its usefulness in valid estimation of rates and trends both nationwide and for most individual states may be limited, especially if there are substantial underlying geographic differences.
Therefore, we used official US nationwide cancer data to perform a comprehensive examination of overall burden as well as secular trends in incident AIDS-related KS rates in the United States between 2000 and 2014. This included identifying demographic subgroups as well as states and major urban areas with the greatest burden and most rapidly changing KS incidence rates.
We obtained KS incidence data from the US Cancer Statistics (USCS) registry, the source for official US federal government–reported cancer statistics. It compiles data on all incident cancer cases reported in 2 primary sources, the Center for Disease Control and Prevention's (CDC's) National Program of Cancer Registries, which includes all population-based state cancer registries, and the SEER program, which includes 14 population-based cancer registries and 3 supplemental registries.6 The publicly available USCS registry incorporates data from states whose cancer registries data meet all quality control standards, which includes ≥90% capture of all cases; ≤5% cases identified solely through death certificates; ≤5% cases missing age, sex, or race; and ≥97% of records passing all validity and logic checks. In 2012, the USCS registry covered 97% of the US population. Only deidentified aggregate data are available in the publicly available files that may be accessed at no cost.
We accessed these USCS registry data through the CDC's Wide-ranging Online Data for Epidemiologic Research platform (WONDER).7 All incident KS cases are coded within this registry using the International Classification of Disease for Oncology (ICD-O) as 9140/03. The USCS registry unfortunately does not include HIV status or medication usage. We therefore restricted our searches to identify cases of presumed AIDS-related incident KS to men aged 20–54 years given prior research findings that >91% of men and 100% of women with incident KS aged >55 years had classic or elderly related KS,5,8 with almost all cases of incident AIDS-related KS occurring in men (100:1 male-to-female ratio).9 Our restriction to men only was further necessitated because USCS data on incident KS are suppressed for women in most geographic and sociodemographic subgroups due to small numbers.
We calculated annual and multiyear incidence rates of AIDS-related KS (hereafter referred to as KS) in men aged 20–54 years using CDC WONDER, with the number of USCS-identified cases as the numerator and corresponding US census–estimated population size as the denominator. We obtained both age-group–specific (20–29, 30–44, and 45–54 years old) and age-adjusted rates, both for the United States overall and by the race/ethnic group [Asian Pacific Islander (API), American Indian/Alaska Native (AI/AN), Black/African American, Hispanic/Latino, and White/Caucasian-non-Hispanic], and also for all 50 states, US census region (Northeast, West, South, and Midwest), and for the 30 metropolitan statistical areas (MSAs) with the highest numbers of KS cases. Age-adjusted rates were directly standardized using the US 2000 standard population, with 95% confidence intervals (CIs) calculated using methods of Tiwari et al.10
We evaluated trends in annual KS incidence rates using the NCI's Joinpoint program (version 3.5.2; http://surveillance.cancer.gov/joinpoint). It applies piecewise linear regression to determine whether average rate changes over time are best described by a single line or multiple linear segments (ie, none or ≥1 joinpoints). We required at least 3 observations per segment with a maximum of 3 joinpoints over the 15-year study period, with the best model (ie, model in which additional joinpoints offer no significant improvement in fit) determined using log-transformed data.11 We calculated both annual percentage change (APC) in incidence rates over each segment and average APC (AAPC) over the cumulative study period for each model. Monte-Carlo permutation–based methods were used to assess trend significance, with 2-sided P values <0.05 considered statistically significant.
We also created choropleth maps displaying the average age-adjusted incident KS rates in subgroups for all states over 4 consecutive 3-year time periods starting in 2003; the first-year state-level data were available for all 50 states, using ArcGIS 10.5.12 We used “equal interval” classification methods to divide KS incidence rates observed in the baseline time period (2003–2005) into 4 initial categories where data values (max − min) are divided equally across categories. This method, in contrast to other methods (ie, quartiles and Jenks), makes changing patterns over time easier to perceive.13 We used these cutoffs to similarly color-code rates in all subsequent periods (ie, 2006–2008, 2009–2011, and 2012–2014). Additional categories were included to reflect rates in states where there no cases (ie, rate = 0/100,000), rates in states that were known >0 but not calculable given data suppression (ie, cases >0 but <15), and an additional upper category covering rates in postbaseline time periods that were higher than upper baseline rates.
Incident AIDS diagnoses in all 50 states for 2003 to 2014 were obtained using publicly available, deidentified aggregate data files for the CDC's National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention (NCHHSTP) registry; it includes federally mandated data collected on reportable infections including incident AIDS/HIV stage 3 diagnoses.11 The data on incident AIDS infections maintained by NCHHSTP were obtained using the freely available AtlasPlus interface (https://www.cdc.gov/nchhstp/atlas/index.htm). We obtained annual incidence data in men aged 25–54 years (vs. 20–54 years for KS, as AIDS data for 20–24 years are included with teens). We applied the same methods used to calculate adjusted AIDS incidence rates for all states and similarly used ArcGIS 10.5 to generate heat maps with cutpoints determined using baseline-period data.
Because our research was solely restricted to data obtained at no cost from publicly available, deidentified aggregate data registries maintained by the federal government (ie, USCS and NCHHSTP) that are internet accessible and do not require submission of a protocol or institutional review board (IRB), local IRB review is automatically waived.
Overall KS Incidence in the United States
We identified 12,549 incident KS diagnoses in men aged 20–54 years in the United States between 2000 and 2014 (Table 1); men represented 95% of these cases (data not shown). The age-adjusted incidence rates of KS declined from 1.44/100,000 in 2000 to 0.95/100,000 by 2014. Joinpoint analysis identified 4 specific time periods best described linear change in rates over time, including significantly declining APCs of −6.3 (95% CI: −10.9 to −1.5) and −2.1 (95% CI: −3.8 to −0.5) between 2007 and 2010 and 2010 and 2014, respectively (see Table 1 and Figure 1, Supplemental Digital Content, https://links.lww.com/QAI/B345). The overall AAPC or decline was −3.1%/year over the 15-year study period.
By Age Group
There were 7557 incident KS cases within the 30–44-year-old age group (60.2% of total); this age group had the highest incidence rate (1.61/100,000) over the study period (see Table 2, Supplemental Digital Content, https://links.lww.com/QAI/B345). They were the only age group that experienced statistically significant declining rates over time (AAPC = −5.4%, 95% CI: −6.1 to −4.6) (see Table 1 and Figure 2, Supplemental Digital Content, https://links.lww.com/QAI/B345). By contrast, there was a significant 2.7% average annual increase in rates in 20–29 years, with essentially no change (AAPC =−0.03%, N.S.) in rates in those 40–54 years. Given this divergence, the initially disparate annual incidence rates in the 30–44- and the 45–54-year-old age groups had converged by the end of the study period (eg, 1.08/100,000, 95% CI: 0.97 to 1.20 in 30–44 years old vs. 1.07, 95% CI: 0.94 to 1.22 in 45–54 years old in 2014).
We could assess for trends only in Whites, Blacks, and Hispanics given mandatory data suppression due to sample size (n < 15) for multiple years for API and all years for AI/AN (data not shown). Although the largest number of incident KS cases occurred in White men (n = 5,098, 40.6% of total), Black and Hispanic men had higher age-adjusted incidence rates overall and in all individual study years (eg, 3.07/, 1.62/, and 0.73/100,000 in Blacks, Hispanics, and Whites, respectively, between 2000 and 2014) (see Table 2 and Figure 3, Supplemental Digital Content, https://links.lww.com/QAI/B345). There were variable significant declines in incidence rates observed in all 3 groups, with the largest decline in Hispanics (AAPC = −5.1, 95% CI: −6.4 to −3.8) (see Table 1, Supplemental Digital Content, https://links.lww.com/QAI/B345).
By Race/Ethnicity and Age Group
Overtime, the proportion of Whites vs. Blacks across age groups showed an inverse trend, with the largest proportion of Blacks in the youngest age group [n = 805 (52.7%) vs. n = 327 (21.5%) in 20–29 years old in Blacks vs. Whites, respectively], whereas the proportionately largest group of Whites was in the oldest age group [n = 878 (26.8%) vs. n = 1761 (53.8%) in 45–54-year-old Blacks vs. Whites, respectively] (see Table 2, Supplemental Digital Content, https://links.lww.com/QAI/B345). Our trend evaluation in combined race/ethnicity and age groups showed declining rates, with the largest significant declines in Whites and Blacks aged 30–44 years between 2000 and 2014 (AAPCs = −6.6, 95% CI: −7.9 to −5.2 and −5.3, 95% CI: −6.2 to −4.3, respectively) (see Table 1, Supplemental Digital Content, https://links.lww.com/QAI/B345 and Figs. 1A–C). Only 20–29-year-old Blacks experienced a significant increase in rates over time (AAPC = 4.0, 95% CI: 1.8 to 6.3). Given these divergent trends, incidence rates in 20–29-year-old Blacks surpassed rates in Hispanic and Black men aged 45–54 years and Whites of all age groups by the end of the study period (Figs. 1A–C).
Our trend analyses in the 5 states with the largest number of cases demonstrated significant declines in California, Florida, and New York (AAPCs = −4.98, −6.78, and −5.22, respectively); mixed trends in Texas, with both significant decline and significant increase in rates (APCs = −10.64 and 37.61 for 2007–2012 and 2012–2014, respectively); and a nonsignificant +1.82% annual increase in Georgia over the study period (see Table 2, Supplemental Digital Content, https://links.lww.com/QAI/B345 and Fig. 2).
Choropleth maps presenting state-level incidence rates across time are shown in Figure 3. Visual map comparison indicates most states with higher KS incidence rates between 2003 and 2005 (which were predominantly concentrated in the northeast, southern, and western regions) experienced declines over time. For example, 3 states (ie, California, New York, and Florida) with rates in the highest category (between 2.22 and 2.80/100,000 during 2003–2005) no longer had rates in the highest category by 2009–2011, whereas rates in Georgia were persistently high.
Choropleth maps presenting age-adjusted rates of incident AIDS/HIV stage 3 are also presented in Figure 3. As expected and similar with KS, states with higher AIDS incidence rates were concentrated in more populous states in the northeast, south, and western United States. There were substantial wide-scale declines in incident AIDS rates over time, with less variability in distribution of state-level rates by the end of the period. For example, there were no states with AIDS incidence rates in the upper 2 categories by 2012–2014, including Georgia whose rates had fallen from the highest category (25.8–33.6/100,000) between 2003 and 2005 to the second highest category (10.1–17.8/100,000) by 2012–2014.
Annual age-adjusted KS incidence rates in the 4 census-defined geographic regions of the United States are presented in Table 2, Supplemental Digital Content, https://links.lww.com/QAI/B345. Rates in 3 regions (northeast, central, and west, respectively) all demonstrated significant (P < 0.05) declines in age-adjusted KS incidence over at least 50% of the study period (see Figure 4 and Table 1, Supplemental Digital Content, https://links.lww.com/QAI/B345). Given the large decline in rates in the West and Northeast, which started with the highest KS incidence, their rates converged or were lower than rates in the South by the end of the study period (see Figure 4, Supplemental Digital Content, https://links.lww.com/QAI/B345). Figure 4A–C shows changing incidence by race in each of the 4 regions. In particular, Figure 4A shows a nonsignificant decline (AAPC = −0.86) of KS incidence among African Americans compared with significant declines in every other race/region combination.
Metropolitan Statistical Areas
The 30 MSAs with highest age-adjusted KS incidence rates are presented in Table 2, Supplemental Digital Content, https://links.lww.com/QAI/B345. Rates ranged between a high of 4.25/100,000 (95% CI: 3.96 to 4.55) in Atlanta–Sandy Springs–Roswell, Georgia and a low of 1.11/100,000 (95% CI: 0.95 to 1.29) in Detroit–Warren–Dearborn, MI. Our trend analysis in the 5 MSAs with the most KS cases showed significantly decreasing incidence rates in 4 (APCs = −3.8, −8.6, −5.5, and −6.2 in the Los Angeles–Long Beach–Anaheim, Miami–Fort Lauderdale–Palm Beach, New York–Newark–Jersey City, and San Francisco–Oakland–Hayward between 2000 and 2014, respectively), whereas rates in the Atlanta–Sandy Springs–Roswell MSA demonstrated a nonsignificant 1.5% average annual increase (see Table 1 and Figure 5, Supplemental Digital Content, https://links.lww.com/QAI/B345). Given these differences, rates in the Atlanta-area MSA surpassed those in the San Francisco- and Miami-area MSAs that had historically higher rates every year post-2007. The New Orleans–Metairie MSA, which had a less total KS cases than found in the 5 leading MSAs (n = 111), had the fifth highest age-adjusted KS incidence rates over the cumulative study period (2.60/100,000, 95% CI: 2.13 to 3.13) (see Table 2, Supplemental Digital Content, https://links.lww.com/QAI/B345). However, as its annual incidence rates were suppressed in all individual study years given n <15 incident cases/year (data not shown), we could not assess trends.
Our comprehensive evaluation of trends in incident KS in PLWH using nationwide USCS cancer registry data showed growing disparities. We found, for example, significant increases in KS incidence rates in men aged 20–29 years, particularly in Blacks, whereas significant declining rates were observed in the United States overall and among Whites and Blacks aged 30–44 years between 2000 and 2014. Substantial geographic variations were also found; most high AIDS-related KS incidence states were concentrated similarly to high AIDS incidence states in the northeast, southern, and western United States, with all states reporting lower rates by the end of the study periods except Georgia (where rates rose). In addition, KS incidence rates decreased among Blacks overall in all geographic areas, except the South, where regional incidence rates were generally unchanged. There were also large decreases in incidence found in several states and MSAs that historically had among the highest KS incidence rates including New York and California as well as the San Francisco- and Miami-area MSAs. Consequently, by the end of the study period, Georgia and its large Atlanta-area MSA overtook previously higher risk states and MSAs to have the highest rates nationwide.
Earlier national and regional estimates of presumptive AIDS-related KS in the United States used SEER data,5,14 with the most recent evaluation covering years 2001–2013.5 It similarly reported a significant decline in rates in men <55 years, with declining trends also reported in earlier SEER-based analyses covering earlier time periods. Notably, the annual and overall age-adjusted rates of KS estimated using the 2001–2013 SEER data are higher than we observed using USCS data 2000–2014 (eg, ∼1.8 vs. ∼1.3/100,000 using SEER vs. USCS data in 2002, respectively). This is likely attributable to SEER data being limited to 14 states and 3 additional select cancer registries and thus not including data from multiple lower incidence rate states disproportionately concentrated in the central (noncoastal) United States (eg, Iowa, Missouri, North and South Dakota). We also confirmed earlier SEER-reported findings, indicating KS rates may be increasing in 20–29-year-old Black men, particularly in the South.5 Compared to the most recent reported SEER-based analysis, we identified lower KS incidence rates in the Southern United States likely attributable to SEER's inclusion of only 2 southern states, Georgia and Louisiana, which have higher incidence rates that most other southern states including North and South Carolina, VA, and Tennessee. We are unable to definitively determine specific causes for our observed geographic variability given the aggregate nature of USCS data. However, it is likely multiple factors contribute to differences including the total population size and the percent of the population living in urban areas where higher KS incidence rates have been reported14; variability in new HIV infection rates as well as in access to ART; and differing prevalence of human herpes virus-8 (HHV-8) infection across geographic and risk subgroups in the United States as HHV-8 is necessary though insufficient for KS development.15 Of note, in mapping changes in state AIDS incidence over time and comparing those with KS incidence, we found that although both were generally decreasing, the patterns were not entirely parallel.15
Our study has several strengths including the previously noted comprehensive nationwide data capture using USCS data. It provides the first truly national estimates updated through 2014 and with a 15-year trend analysis in a contemporary ART era, and a novel geographic comparison of state-level incident KS and AIDS rates over time. Another is our reporting of age-adjusted KS incidence rates for all 50 states, the 4 major geographic regions, and the 30 leading MSAs, including trend analyses that are crucial for informing local and national health care planning and surveillance.
Our study has several limitations including previously noted lack of data to definitively confirm presumptive HIV/AIDS infection (eg, viral load and CD4 count) in our incident KS cases. However, our use of previously validated age- and gender-based restriction criteria suggests almost all our identified incident KS cases in men aged 20–54 years should be AIDS-related,8,16 with this demographic constituting a large majority of AIDS-related KS across all age groups. Nonetheless, our estimates of the burden of presumptive AIDS-related KS are undoubtedly modest underestimates, particularly given our exclusion of elderly male PLWH who are also at increased risk of KS compared with the general male elderly population.17,18 However, the total number of incident KS cases in elderly PLWH is far lower than in the younger age groups. This is likely attributable to much lower rates of new HIV infection in the elderly and because most AIDS-related KS cases in elderly PLWH occur within 5 years of a new AIDS diagnosis and continue declining thereafter; prior cohort research showed ∼83% of cases in elderly PLWH occur in those 50–59 years18 (with our study including those 50–54 years), whereas <2% occur in those ≥70.17 Although KS also occurs in the youngest PLWH (<20 years), it is exceptionally rare in the United States, with only 44 cases nationwide from 2000 to 2014. We also cannot exclude the possibility that a few transplant-related KS cases are in our study cohort. However, as transplantation is such a rare KS cause in the United States, it would not meaningfully impact our estimates. Another limitation is that owing to our necessary exclusion of women given their frequent privacy-related data suppression, we cannot determine whether our observed presumed AIDS-related KS trends in men are similar in women.19 Furthermore, we were unable to evaluate trends in annual incidence rates separately in men in the API and AI/AN subgroups, as their data were typically suppressed due to low sample size. Finally, although our study is population-based and spans a considerable contemporary ART-era time period, our suggestive and significant findings must be interpreted keeping in mind our cross-sectional study design and reliance on aggregate USCS data that refer only to residence at the time of incident KS diagnosis. For example, although our finding that KS incidence rates in the Southern United States have not experienced the same declines noted in all other geographic regions corroborates recent data suggesting that higher UV exposure further increases risk of a person with AIDS developing KS,20 we cannot directly examine this using USCS data, although we did note that KS incidence decreased among Whites in the Southern United States who have similar exposures to African Americans in the Southern United States, whose rates have not decreased.
In summary, this 50-state, population-based evaluation of nationwide trends confirms an overall decreasing rate of incident presumptive AIDS-related KS nationally, while also highlighting potentially widening disparities between Black men in the Southern United States and men aged 20–29 years (and particularly among Black men in this age group). Future research that helps identify specific potential causes for these disparities in addition to ongoing monitoring of the epidemiology of incident AIDS-related KS at the local, state, and national levels is critically needed to inform development of targeted interventions to effectively address persistent or increasing disparities observed among multiple high risk subgroups of PLWH.
1. Engels EA, Biggar RJ, Hall HI, et al. Cancer risk in people infected with human immunodeficiency virus in the United States. Int J Cancer. 2008;123:187–194.
2. Robbins HA, Shiels MS, Pfeiffer RM, et al. Epidemiologic contributions to recent cancer trends among HIV-infected people in the United States. AIDS. 2014;28:881–890.
3. Beral V, Peterman TA, Berkelman RL, et al. Kaposi's sarcoma among persons with AIDS: a sexually transmitted infection? Lancet. 1990;335:123–128.
4. Boscoe FP, Johnson CJ, Sherman RL, et al. The relationship between area poverty rate and site-specific cancer incidence in the United States. Cancer. 2014;120:2191–2198.
5. Royse KE, El Chaer F, Amirian ES, et al. Disparities
in Kaposi sarcoma incidence and survival in the United States: 2000–2013. PLoS one. 2017;12:e0182750.
6. U.S. Cancer Statistics Working Group. United States Cancer Statistics: 1999–2014 Incidence and Mortality Web-Based Report 2017. Available at: www.cdc.gov/uscs
. Accessed March 15, 2018.
7. US Department of Health and Human Services. United States Cancer Statistics: 1999–2012. 2017. Available at: https://wonder.cdc.gov/
. Accessed February 3, 2017.
8. Dal Maso L, Polesel J, Ascoli V, et al. Classic Kaposi's sarcoma in Italy, 1985–1998. Br J Cancer. 2005;92:188–193.
9. Haverkos HW. The epidemiologic scope of Kaposi's sarcoma. Oncology (Williston Park). 1996;10(6 suppl):9–12; discussion 37–18.
10. Tiwari RC, Clegg LX, Zou Z. Efficient interval estimation for age-adjusted cancer rates. Stat Methods Med Res. 2006;15:547–569.
11. Kim HJ, Fay MP, Feuer EJ, et al. Permutation tests for joinpoint regression with applications to cancer rates. Stat Med. 2000;19:335–351.
12. ArcGIS 10.5
. Esri Trademarks. Redlands, CA; 2018.
13. Brewer CA. Basic mapping principles for visualizing cancer data using Geographic Information Systems (GIS). Am J Prev Med. 2006;30(2 suppl):S25–S36.
14. Armstrong AW, Lam KH, Chase EP. Epidemiology of classic and AIDS-related Kaposi's sarcoma in the USA: incidence, survival, and geographical distribution from 1975 to 2005. Epidemiol Infect. 2013;141:200–206.
15. Uldrick TS, Whitby D. Update on KSHV epidemiology, Kaposi sarcoma pathogenesis, and treatment of Kaposi sarcoma. Cancer Lett. 2011;305:150–162.
16. Shiels MS, Pfeiffer RM, Hall HI, et al. Proportions of Kaposi sarcoma, selected non-Hodgkin lymphomas, and cervical cancer in the United States occurring in persons with AIDS, 1980–2007. JAMA. 2011;305:1450–1459.
17. Mahale P, Engels EA, Coghill AE, et al. Cancer risk in older persons living with human immunodeficiency virus infection in the United States. Clin Infect Dis. 2018;67:50–57.
18. Yanik EL, Katki HA, Engels EA. Cancer risk among the HIV-infected elderly in the United States. AIDS. 2016;30:1663–1668.
19. Hiatt KM, Nelson AM, Lichy JH, et al. Classic Kaposi Sarcoma in the United States over the last two decades: a clinicopathologic and molecular study of 438 non-HIV-related Kaposi Sarcoma patients with comparison to HIV-related Kaposi Sarcoma. Mod Pathol. 2008;21:572–582.
20. Cahoon EK, Engels EA, Freedman DM, et al. Ultraviolet radiation and Kaposi sarcoma incidence in a nationwide US cohort of HIV-infected men. J Natl Cancer Inst. 2017;109.