Association between Mortality Due to Nasopharyngeal Carcinoma and Race in the United States from 2007 to 2016

Background: Asians and Pacific Islanders (API) exhibit increased incidence of nasopharyngeal carcinoma (NPC). However, they are often excluded when the disease is studied. Risk-factors and incidence are well-researched while cancer-specific mortality trends remain unclear. We aimed to determine whether insurance status modifies the association between race and cancer-specific mortality in NPC patients. Methods: This retrospective cohort study used secondary data analysis from the Surveillance, Epidemiology, and End Results Program database. Patients ≥18 years with histologically confirmed primary NPC from 2007 - 2016 were included. The main outcome assessed was 5-year survival and the main exposure variable was race (API, white, black). Insurance status was classified into uninsured, any Medicaid, and insured (with any insurance). Potential confounders included age, sex, marital status, stage at diagnosis, and surgical treatment. Adjusted Cox regression analysis was used to calculate hazard ratios (HR) and corresponding 95% confidence intervals (CI). Results: 1610 patients were included (72.98% male, 27.02% female). 49.8% were API, 40.5% were Whites, and 9.8% Blacks. Maximum follow-up was 5-years. The adjusted hazards of 5-year cancer-specific death for API and Blacks compared with Whites were 0.77 (95% CI 0.62 – 0.96) and 0.92 (95% CI 0.65 – 1.31), respectively. Cases decreased with age in API and Blacks. 8.2% of cases had localized disease, 45.3% had local spread, and 44.6% had distant metastasis. Insurance status did not modify the association between race and mortality. Conclusion: Race is an important prognostic factor to account for in NPC patients. Investigating risk-factors and subtypes stratified by race may explain our findings.


Introduction
Nasopharyngeal carcinoma (NPC) is a unique subset of head and neck cancers. 86,000 cases of NPC present annually with distinct ethnic and geographic predilections (Chua et al., 2016;Ferlay et al., 2015). It is most prevalent in Southern China with 25 cases per 100,000 people annually. High-risk groups exhibit peak incidence at 50-59 years of age. Low-risk groups exhibit a steady increase of incidence with age. In the US, incidence increases with age, despite a minor peak in adolescents and young adults. Men are 2-3 times more likely to be affected than women. Increased risk due to immigration from endemic areas decreases with each passing generation (Mousavi et al., 2010;Chen et al., 2019, Chang et al., 2021. Immigrants from China in the US have the highest incidence of NPC and are 108 times more likely to be affected than

Association between Mortality Due to Nasopharyngeal Carcinoma and Race in the United States from 2007 to 2016
non-immigrant Whites (Dickson and Flores, 1985;Mousavi et al., 2010;Lee et al., 2019). Chinese Americans overall maintain a 10-20 times higher incidence of NPC compared to White or Black Americans (Chang et al., 2021). Incidence in the US has decreased steadily while prognosis and outcomes have improved (Lee and Ko, 2005;Chen et al., 2019).
Surveillance, Epidemiology, and End Results Program (SEER) combines data from diverse population groups including Chinese immigrants and Pacific Islanders into a single Asians and Pacific Islanders group (API). Despite increased incidence, the current literature on cancerspecific mortality in API is inconsistent. Some studies suggest Asians have lower cancer-specific mortality compared with Whites and Blacks (Sun et al., 2007;Ou et al., 2007;Trinh et al., 2015;Chang et al., 2021). Other studies report no significant differences (Dickson and 916 Flores, 1985;Bhattacharyya, 2004). There are reports of increased cancer-specific mortality among API up to 10 times versus other races (Huang et al., 2013;Miller et al., 2008;Levine et al., 1987). Breaking down the API population into ethnicities and countries of origin suggested that the increased incidence and mortality burdens in NPC are not equally distributed among various API groups. Chinese, Filipinos, and Vietnamese may show particularly high cancer-specific mortality among API (Levine et al., 1987;Miller et al., 2008;Trinh et al., 2015). Blacks below 30 years of age have been shown to have an increased incidence of NPC. However, Black race is not associated with cancer-specific mortality when controlled for covariates (Lee and Ko, 2005;Cannon et al., 2006;Richey et al., 2006;Ferrari et al., 2012).
Most of the previous scientific studies of race and NPC do not account for presentation, viral status, or histologic subtypes. Substantial research has been done on East Asian populations due to the region's high prevalence while American populations are less examined (Ferlay et al., 2015;Wang et al., 2013). Epidemiological studies on NPC in the United States often exclude API (Ferlay et al., 2015;Sun et al., 2007;Ou et al., 2007;Bhattacharyya, 2004;Cannon et al., 2006;Wang et al., 2013). Even in studies focusing on API, small sample sizes prevent researchers from examining the mortality of the individual ethnicities and respective countries of origin within the API population with enough power (Miller et al., 2008). Most similar studies have relatively small sample sizes (Ferlay et al., 2015;Sun et al., 2007;Ou et al., 2007;Bhattacharyya, 2004;Cannon et al., 2006;Chan et al., 2017). Population-based cancer registries offer pooled data and provide another opportunity to study this rare disease (Bhattacharyya, 2004;Richey et al., 2006;Ou et al., 2007;Sun et al., 2007;Ferlay et al., 2015;Raghupathy et al., 2014;Chan et al., 2017;Chang et al., 2021). However, such registries over-represent select urban areas and developed countries and include just 21% of the global population (Raghupathy et al., 2014). Important prognostic elements such as histologic subtype, treatment modality, and viral status in NPC are not well-researched on a large scale (Geara et al., 1997;Lo et al., 1999;Chan and Lo, 2002;Lin et al., 2004;Le et al., 2005;Stevens et al., 2005;Leung et al., 2006;Sun et al., 2007;Ou et al., 2007;Chan et al., 2013;Raghupathy et al., 2014;Chan et al., 2017). Furthermore, insurance status is an important social determinant of health that may modify the association between race and mortality in patients with NPC. However, its role in the association between race and cancer-specific mortality in NPC patients has not been investigated. Current mortality rates and relevant factors must be determined for providers to deliver fully comprehensive care to API patients.
The objective of our study was to determine the prognostic relation of race and cancer-specific mortality in patients with NPC, and to check whether insurance status modifies this association.

Study design and population
This was a retrospective cohort study and secondary data analysis of the SEER database. SEER is a program of the National Cancer Institute (NCI) focused on cancer surveillance. Mortality data reported by SEER are provided by the National Center for Health Statistics, and population data is taken from the Census Bureau (SEER 2019). SEER collects survival and incidence data on cancer in 34.6 percent of the American population. Insurance status was reported to SEER from 2007 -2016. The inclusion criteria of our study were patients with a histologically-confirmed diagnosis of a primary NPC (C110-C113, C118-C119 sites and 8000 -9975 /3 histology) recorded between 2007 and 2016 who were 18 years or older and whose race was recorded in SEER. Exclusion criteria were patients who died of an unknown cause, patients missing follow-up data, and patients who were American Indian/Alaskan Natives (AI/AN) or of unknown race.

Variables
Variables on SEER database include gender, age, ethnicity, histological type, marital status, tumor size, surgery history, disease stage, cause of death, and survival time. Race was classified as White, Black, and API. The primary outcome was 5-year mortality calculated from the date of diagnosis to the date of cancer-specific death. Maximum follow-up time was five years. Ethnicity was grouped as Hispanic or Non-Hispanic. Insurance status was classified as uninsured, on Medicaid, and insured (including any Non-Medicaid insurance). Age, sex, marital status, stage of diagnosis, and surgical treatment at the primary site were included as covariates. Age was categorized as less than 50, 50-59, 60-69, 70-79, and 80 years or older. Sex was grouped as male and female. Marital status was classified as partnered and unpartnered. Surgical intervention was classified into any intervention and none. Disease stage at diagnosis was classified as localized disease (disease confined to the nasopharynx), regional disease (regional lymph node invasion), and metastatic disease.

Statistical Analysis
STATA was used to perform statistical analyses. Percentages were used to analyze nominal variables. Chi-squared tests were used to compare distributions of possible confounders by race. Kaplan-Meier survival curves were constructed to visually measure the survival rates of NPC patient according to race over the course of 5 years. The log-rank test was used to assess for differences in survival between each race. Unadjusted and adjusted Cox regression models were used to test for the associations between race and survival. Adjusted hazard ratios (aHR) and their corresponding 95% confidence intervals (CI) were calculated to interpret the differences in risk of death of API and Blacks as compared to Whites. Interaction was tested by adding an interaction term race*insurance status to the statistical model. The proportional hazard assumptions were tested graphically. Alpha of 0.05 was used for all statistical tests. All tests measuring p-value and Cis are two-sided. P-values less than or equal to alpha are assumed statistically significant. All Cis which do not include 1 are assumed to have active follow-up reported in SEER. Exclusions were made for unknown or AI/AN race (n=32), unknown insurance status (n=68), and causes of death unrelated to NPC or unknown (n=28). Our final cohort included 1,610 patients.  underwent surgery at the primary site (16%) compared to Blacks (9%) and API (5%) (p-value < 0.001). Surgical intervention data was not available for 1 API. Figure 1 represents survival estimates based on a log-rank test. The survival curves differed according to race. API had a statistically significant increased 5-year survival rate compared to Whites (p-value = 0.008).

Discussion
Our data revealed that API had higher 5-year cancer-specific survival than Whites. No difference in cancer-specific mortality was found between Blacks and Whites. Patients without insurance and patients with Medicaid had increased cancer-specific mortality compared to patients with insurance. There was no evidence that insurance status modified the association between race and survival. Age, tumor stage, and marital status predicted cancer-specific mortality, while surgical history and sex were not associated with survival.
There is no consensus within the scientific literature regarding the association between race and cancer-specific mortality in NPC patients. However, current evidence has revealed associations between Asian race and the incidence of NPC (Dickson and Flores, 1985;Chang et al., 2021). In agreement with our findings, some studies have found that Asian NPC patients had greater survival rates compared to other racial groups in the US (Ou et al., 2007;Ferlay et al., 2015;Chan et al., 2017). Other studies did not find a significant difference in cancer-specific mortality among races in the US (Bhattacharyya, 2004;Richey et al., 2006;Sun et al., 2007). One study found that overall survival was higher in Asians while cancer-specific survival was not (Sun et al., 2007). This suggests that increased overall survival in Asians may confound the reports in which Asians exhibit lower cancer-specific mortality. Another study reported that Black, White, and Asian NPC patients under 30 shared similar cancer-specific survival possibly due to an overall lack of insurance and comorbidities within the age group (Richey et al., 2006). A third study initially found better cancer-specific mortality in Asians compared to Whites, but after matching Asian and White patients according to epidemiological factors, including disease subtype and staging, the study found no significant difference in cancer-specific death (Bhattacharyya, 2004). This may indicate that Asians are more likely to present early and with more favorable subtypes of NPC. Other studies support this conclusion and show that Asians are more likely to present with Type 2b non-keratinizing undifferentiated carcinoma (or World Health Organization [WHO] subtype III) NPC, which is more sensitive to treatment (Arnold et al., 2013;Ferlay et al., 2015). WHO subtypes II and III are associated with positive Epstein-Barr virus (EBV) status. Literature shows EBV+ NPC exhibits better outcomes compared to NPC with other viral statuses (Pathmanathan et al., 1995;Raghupathy et al., 2014;Le et al., 2005;Chen et al., 2019;Lee et al., 2019;Chang et al., 2021). EBV+ status, Chinese ethnicity, and male gender are the most significant predictors of NPC incidence globally (Yu and Yuan, 2002;Raghupathy et al., 2014;Chen et al., 2019;Lee et al., 2019;Chang et al., 2021). These findings are paralleled in the US API population (Sun et al., 2007). This combination of findings provides a likely explanation for the lower cancer-specific mortality of API observed in our study. With EBV and NPC endemic in China, it is appropriate that API race in the US is associated with increased incidence of viral NPC and histomolecular subtypes purporting better prognoses. Possible explanations for the racial disparities in NPC incidence may include variable exposures to life-style related risk factors such as diet, smoking, pollution, ethanol use, and viral status. Nonracial factors associated with worse cancer-specific mortality in NPC may include age, substance abuse, and low socioeconomic status (Richey et al., 2006;Chen et al., 2019;Lee et al., 2019).
Several mechanisms have been proposed to explain the higher survival in API patients. The trends in NPC incidence and survival in API are related to the prevalence of EBV in Asian-Americans. EBV infection leads to genetic alterations that promote tumorigenesis (El-Naggar et al., 2017). Diets high in volatile nitrosamines and salt-cured fish are associated with either latent EBV activation or other carcinogenic effects (Guo et al., 2009;Chua et al., 2016;Lee et al., 2019). These exposures are more frequent in developing countries, many in East Asia. Familial, molecular, and chromosomal models for pathogenesis reportedly also contribute to carcinogenesis (Chan and Lo, 2002;Chen et al., 2019). It is proposed that dietary elements lead to a loss of heterozygosity that causes premalignant and low-grade lesions to develop. Further mutations make these lesions susceptible to EBV infection that provides cells with proliferative and survival advantages and mutagenicity with latent viral genes (Rohlfing et al., 2017).
Our data revealed differences in cancer-specific survival according to health insurance status. This agrees with current scientific literature. Insured patients exhibit earlier presentation, better survival, fewer surgical complications, shorter lengths of stay, and fewer comorbidities compared to Medicaid or uninsured patients (Rohlfing et al., 2017;Amini et al., 2018;Obeng-Gyasi et al., 2018). Adjusting for possible confounders, our data showed that having Medicaid or no insurance was associated with higher cancer-specific mortality compared with having insurance.
We recognize some limitations of this study. Hispanic ethnicity was not included in our statistical model due to small sample size. Ethnicity is a social determinant of health in the US, and its inclusion would have provided relevant data. API was assessed as a single group as coded in SEER which excludes important variations of risk between API subgroups. SEER does not include NPC tumorigenesis and progression so we did not assess relevant carcinogenic exposures. Possibly confounding comorbidities were not reported to SEER. We excluded all radiation and chemotherapy history from study due to unreliable coding and reporting within SEER. Positive and negative margins for tumor resections as well as types of surgeries are valuable prognostic factors that unfortunately were unavailable. We were only able to dichotomize surgical history into yes or no. We also excluded histological classifications of NPC from study due to limited sample sizes, outdated classifications, and limited reliability of histology within SEER. This meant excluding histological subtypes differing among racial lines which significantly affect prognosis (Yu et al., 2009;Arnold et al., 2013). SEER does not have information on viral status. EBV is a well-established prognostic factor in NPC while Human Papilloma Virus is likely implicated as well (Le et al., 2005;Guo et al., 2009;El-Naggar et al., 2017).
Our study shows that race plays a significant role in predicting survival rates among patients with NPC. Particularly, API showed decreased cancer-specific mortality risk compared to Whites. Understanding racial and ethnic trends in mortality due to cancer are useful in guiding screening and early detection. Early detection of NPC and awareness of risk-factor exposure are effective preventative measures. Further study is of racial influences on the progression of NPC may be fruitful. There are many differences in genetics and risk-exposures between the various groups within API and we suggest conducting separate sub-group analyses in future studies. Ethnicity is associated with differences in care and must be assessed in the future as well. Analyses of mortality differences between histological subtypes and virally associated disease are also warranted. Furthermore, time to treatment and time to diagnosis are critical prognostic factors in cancer care and epidemiology that could not be addressed in this study. These factors differ along racial lines and addressing them is suggested for future studies. Rather than assessing insurance status alone, it would be valuable to assess mortality rates according to multiple socioeconomic metrics.