The epidemiology of cancers of the nasopharynx and salivary gland suggests a striking geographic and ethnic dependency, in that the tumors occur most frequently in Asia and in the Inuit population in Alaska, Greenland, and northern Canada.^11-16 Non-keratinizing nasopharyngeal carcinoma represents 75% or more of all nasopharyngeal carcinoma, with the highest proportion reported in regions with high nasopharyngeal carcinoma incidence.^11-14 Lymphoepithelial carcinomas comprise less than 1% of salivary gland carcinomas, with a higher incidence in Asian and Inuit populations.^15-17 There is a near constant association of EBV with nasopharyngeal carcinoma in all ethnic groups^{2, 8}, and an almost 100% association of EBV with lymphoepithelial carcinoma of the salivary gland in endemic areas (and variable association in areas of low incidence).^17-19

One unanswered question is whether HIV-induced immunosuppression might increase the risk of these rare cancers among racial/ethnic minorities in the U.S. population. Some registry-based studies, but not all have suggested an elevated risk for these cancer types among U.S., European, or African HIV populations.^20-24 However, prior studies have been limited by the rarity of nasopharynx and salivary gland cancers, which might explain some of the variability in results, and none of the studies evaluated risk for specific histologic subtypes of these cancers. Two previous studies utilized data from the HIV/AIDS Cancer Match Study, a large registry linkage study in the U.S.^{20, 21} In a 1996 paper, Melbye et al. did not observe an elevated risk of nasopharyngeal carcinoma in people with AIDS in the U.S. While the analysis was based on 50,050 people with AIDS, only four nasopharyngeal carcinoma cases were included, and the risk of salivary gland cancer was not assessed.²¹ In contrast, in a 2001 paper using a larger dataset (N=302,834 people with AIDS), Frisch et al. found elevated risks for both nasopharynx cancer (standardized incidence ratio [SIR] 2.6) and salivary gland cancer (SIR 2.5).²⁰

We included all individuals with AIDS reported to one of the participating HIV/AIDS registries. Using cancer registry data, we evaluated cancer risk beginning 5 years before AIDS onset or the start of cancer registry coverage (whichever came later) until 5 years after AIDS onset, death, or end of cancer registry coverage (whichever came first). For each person, we compared the observed number of cancers of the salivary gland and nasopharynx to the number expected, based on general population rates specific to the cancer registry region, sex, age, race/ethnicity, and calendar year. SIRs, calculated as the ratio of observed to expected cases, were used to compare risk of salivary gland and nasopharynx cancers in people with AIDS to the general population. To evaluate whether the risk for nasopharynx or salivary gland cancers changed with availability of effective HIV therapy in 1996, we used Poisson regression to calculate a relative risk (RR), derived as the ratio of the SIR in people diagnosed with AIDS in 1996 divided by the SIR in people diagnosed with AIDS in 1990-1995.

Salivary gland and nasopharyngeal cancers were reported among 43 and 39 individuals, respectively. Compared to the general population, individuals with AIDS had approximately two-fold risks for salivary gland cancer (SIR 1.8, 95% CI 1.3-2.4) and nasopharyngeal cancer (SIR 2.0, 95% CI 1.4-2.7) (Table 2). Among salivary gland and nasopharyngeal cancer cases, the vast majority were male (90.7% and 89.7%, respectively). The mean age (± standard deviation) at cancer diagnosis was 47.2 (± 12.0) years for salivary gland cancers and 45.6 (± 8.5) years for nasopharyngeal cancers. Two (5%) nasopharyngeal cancers occurred in Asians, 15 (39%) among non-Hispanic whites, 13 (33%) in non-Hispanic blacks, and 9 (23%) in Hispanics. No salivary gland cancers, were observed in Asians, while 22 (51%), 17 (40%), and 4 (9%) occurred among non-Hispanic whites, non-Hispanic blacks, and Hispanics, respectively. Risks for nasopharyngeal and salivary cancers did not change for people diagnosed with AIDS in 1996-2008 compared with people diagnosed with AIDS in 1990-1995 (RR 1.29 95% CI 0.7-2.4, and RR 0.6 95% CI 0.3-1.1, respectively).

The presence of substantially increased cancer risk in people with AIDS compared to the general population can suggest an important role for immunosuppression and an associated viral etiology. Our finding of markedly elevated risk for lymphoepithelial carcinoma of the salivary gland in the setting of HIV-induced immunosuppression thus points to an underlying viral etiology, and EBV is an attractive candidate.

The oropharyngeal compartment is a central component of EBV persistence within individuals, and oropharyngeal shedding of EBV is the primary mechanism of transmission in the human population. Although it is unclear whether EBV persists in epithelial cells in the oropharynx or transiently shuttles between lymphocytes and oropharyngeal epithelial cells, EBV is associated with a variety of epithelial neoplasms including nasopharyngeal carcinoma and lymphoepithelial carcinomas of stomach, salivary gland, lung, and thymus.² The EBV association in lymphoepithelial carcinomas of the stomach and thymus is independent of race and geographical origin,² but the association of EBV with lymphoepithelial carcinoma of the salivary gland and lung appears to be more frequent in Asian and Inuit populations. ^{2, 11-16}

Following the first description of EBV association with salivary lymphoepithelial carcinoma,²⁶ numerous studies have documented the presence of EBV within lymphoepithelial carcinoma tumor cells (Table 3). These studies suggest a nearly universal presence of the virus in salivary gland lymphoepithelial carcinomas from endemic areas, such as the Arctic region, southeastern China, and Japan.^27-31 Data from other populations are sparse,^32-36 with available results suggesting perhaps more variable detection of EBV.¹⁹ Published studies may overestimate the prevalence of EBV infection in these tumors, because some investigators may only have reported EBV positive cases. In addition, the majority of the studies were small and examined the expression of only a restricted set of two viral genes that encode non-translated small RNAs (EBER-1 and EBER-2) using in situ hybridization.^{10, 26-31, 36-43} Fewer studies evaluated the expression of additional viral latency associated genes, although both latency patterns I (seen also in Burkitt lymphoma) and II (seen in Hodgkin lymphoma and nasopharyngeal carcinoma) have been detected in salivary gland lymphoepithelial carcinomas.^{9, 32, 33, 41, 44-49}

While this study examined salivary gland cancers in HIV-infected people, the occurrence of EBV-related salivary gland cancers has also been described among other patients with disturbed immunity, such as chronic autoimmune disorders.³³ Furthermore, EBV has also been implicated in HIV-related sialadenitis, a chronic inflammatory condition of the salivary gland.⁵⁰ Of note, chronic or autoimmune sialadenitis has been reported to precede development of salivary gland carcinoma.^51-53

The increased risk of squamous cell carcinomas of both salivary gland and nasopharynx might be attributed to tobacco or alcohol use, which are common in HIV-infected people and linked to other cancers of the head and neck.⁵⁴ Also, HPV is etiologically implicated in a subset of head and neck squamous cell carcinomas, specifically those arising in the oropharynx and tonsil. ^55-57 HPV has also been detected in nasopharyngeal carcinomas ⁵⁸ and was found more commonly in squamous cell nasopharyngeal carcinomas arising in U.S. whites than among Asians.⁵⁹ The observation of only moderately increased risk for non-keratinizing carcinoma of the nasopharynx, which is etiologically associated with EBV in other settings, is perhaps because U.S. and Asian populations differ in genetic predisposition, age at initial EBV infection, or other factors.

This study is the first to provide a detailed examination focusing on the risk of salivary gland and nasopharyngeal cancers, including data on specific subtypes of these cancers, in HIV-infected people. Epidemiologic assessment of rare cancers is greatly facilitated by database linkages such as provided in the HIV/AIDS Cancer Match Study. Our large sample size allowed calculation of stable SIR estimates, although the number of cases for most histologic subtypes was still small. We did not have a sufficient number of cases to examine risk of these cancers according to demographic characteristics or CD4 count. The small number of cases may also partly explain why we did not demonstrate a decline in cancer risk among people with AIDS diagnosed in 1996 or after, as has been demonstrated for other virus-related cancers.⁶⁰ Finally, while we could not adjust for differences in smoking or alcohol use between our subjects and the general population, it is unlikely that the approximately 40-fold risk for lymphoepithelial carcinoma of the salivary gland could be entirely explained by such differences.