Epidemiological Study of Uveal Melanoma from US Surveillance, Epidemiology, and End Results Program (2010–2015)

Purpose Uveal melanoma is the most common intraocular malignancy, and the American Joint Committee on Cancer (AJCC) changed its staging methodology from 2010, incorporating notable changes into the T-staging. There were few literatures evaluating the epidemiological trend and risk factors of survival in multicenter longitudinal studies regarding the new staging system. Methods We performed population-based cohort analyses using the Surveillance, Epidemiology, and End Results (SEER) database to identify patients with primary uveal melanoma from 2010 to 2015. Patients and potential prognosis indicators were extracted from SEER 18. Incidence rates, incidence rates ratios (IRR), annual percent changes (APC) in rate, hazard ratios (HR), 5-year accumulative overall survival (OS), and disease-specific survival (DSS) were calculated. Results A total of 2631 patients for incidence analysis and 1142 patients for survival analysis were retrieved. The overall incidence of uveal melanoma was 4.637 per million (95% confidence interval (CI), 4.458–4.821), which was significantly elevated by average APC of 4.215% (p = 0.03). Females had significantly lower incidence (4.076 per million, IRR, 0.768, 95% CI, 0.710–0.832) with noticeable differences among age, race, origin, and laterality in sex-stratified analyses as well. Survival analyses revealed 5-year accumulative OS and DSS for patients with uveal melanoma of 61.8% and 66.5%, respectively. Age, AJCC stage, and radiation therapy were found to be consistent predictors in both univariate and multivariate analysis models. Conclusion Incidence of uveal melanoma increased by significant APC and varied between genders. Determinants of survival included age at diagnosis, AJCC stage, and radiation therapy.

Uveal melanoma was commonly classified into three sizes: small, medium, and large, historically [26][27][28]. However, since 2010, the American Joint Committee on Cancer (AJCC) has changed the staging system for uveal melanoma from AJCC sixth to the seventh edition, which incorporated notable changes into the T-staging (i.e., changes to the size criteria for T1-T4, ciliary body involvement, and amount of episcleral extension) [29,30]. ere were few literatures comparing the outcomes of new classification methodology with past results. We consider that it is a high priority to update the epidemiological trends in uveal melanoma and evaluate prognostic predictors of patient survival regarding the new staging system. e Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute (NCI) is an important source of the longitudinal epidemiological study from multicenter population-based registries, which is updated annually and is available free of charge to the public. In this study, we used data from SEER 18 to analyze recent epidemiological trends of uveal melanoma, disease characteristics, and various potential predictors implicating patient survival.

Methods
A population-based longitudinal analysis for patients with a diagnosis of primary uveal melanoma was performed using the NCI SEER 18 database (http://www.seer.cancer.gov) via the SEER * Stat software (version 8.3.5) in client server mode.
is cancer registry captures 18 distinct population groups in 198 counties in the United States, which covers a 28% of the US population, including 23% of African Americans and 40% of Hispanics. It collects patients' data such as demographic information, cancer characteristics, initial treatment, and follow-up. Internal review board permission was not required because the database compiles publicly available information without personal identifiers.
We identified cases as primary uveal melanoma according to the International Classification of Disease-Oncology, third edition (ICD-O-3), morphology code 8720-8790 and site code C69. 3-C69.4. e inclusion criteria of survival analysis were cases need to be microscopically confirmed during 2010-2015, active follow-up, and survival of not less than 2 months after diagnosis. We excluded the cases that were only confirmed via autopsy after death. Variables such as year of diagnosis, age at diagnosis, sex, race (White, Black, American Indian/Alaska Native (AIAN), and Asian or Pacific Islander (API)), origin (non-Hispanic and Hispanic), primary laterality, marital status, summary stage, AJCC stage, months of survival, surgery, metastasis at diagnosis, radiation treatment, chemotherapy, and causespecific classification of death were extracted.
Age-adjusted incidence rates (cases per million personyears, using 2000 US Standard Population as reference population), incidence rate ratios (IRR), and annual percent changes (APC) were calculated via the SEER * Stat software. Age at diagnosis, race, origin, and primary laterality were taken into account among sex-stratified analyses. e optimal cutoff values for age range were determined using the X-tile software (http://www.tissuearray.org/ rimmlab) in survival analyses. Survival status, survival time, and age of patients were loaded as parameters of "Censor," "Survival Time," and "Marker" in the X-tile software, in order to discover the difference between age ranges. e overall survival (OS) and disease-specific survival (DSS) were plotted by the Kaplan-Meier method, with differences tested by log-rank algorithm. Multivariate Cox proportional hazards regression model was adopted to assess the predictive performance of covariates. Statistical analyses and graphics were performed using IBM SPSS statistics, version 24.0 (SPSS, Inc, Chicago, IL); p value less than 0.05 was considered to be statistically significant unless otherwise specified.

Results
From the SEER 18 database between 2010 and 2015, a total of 2631 cases were pooled for incidence analyses. Further, after filtered by inclusion criteria mentioned above, 1142 cases were extracted for survival analyses. In incidence analysis, patient age ranging from 60 to 80 years (49.0%), White race (96.3%), and non-Hispanics (94.3%) counted for the majority proportion of uveal melanoma patients. Overall incidence of uveal melanoma was 4 (Table S1, Table 1). Given the significant gender variation, we conducted sex-stratified subgroup analyses including age range, race, origin, and primary laterality. Similar incidence patterns remained ( Table 1).
A total of 1142 cases were selected for survival analyses, among which 44.8% were females and 55.2% were males. e mean age at diagnosis was 61.5 years. Surgical treatment was carried out for 42.7% of patients, and 60.5% were treated with different kinds of radiation therapy. e majority of cases were classified in AJCC stage II (38.3%) and the rest distributed in stage I (15.1%), stage II (19.4%), and stage IV (2.4) with 24.9% unknown stage. Other tumor characteristics about race, origin, primary laterality, marital status, summary stage, metastasis, and chemotherapy were presented in Table S2.
Survival curves from Kaplan-Meier (univariate analyses) ( Figure 1) revealed that the 5-year accumulative OS and DSS for uveal melanoma was 61.8% and 66.5%, respectively (Table 3). Both OS and DSS showed significant higher survival rates in young patients (age range was determined by the X-tile software, Figure   Abbreviation: SEER, Surveillance, Epidemiology, and End Results; IRR, incidence rates ratios; AIAN, American Indian/Alaska Native; API, Asian or Pacific Islander. a Total amount may not be 2631 due to exclusion of cases with unknown information. b Incidence rates are based on the number of persons diagnosed as having uveal melanoma per 1,000,000 person-years, age adjusted using the 2000 US population standard. Bold letter indicates that measurements are statistically significant compared with references (p < 0.05).

Discussion
Several published studies reported the mean age-adjusted incidence of uveal melanoma at different time periods in the United States Overall age-adjusted incidence remained stable over the past 4 decades with significant higher incidence in male subjects, claimed by the abovementioned studies, and some other studies demonstrated the disease had no sex preference [15]. e median age of diagnosis is around 60 years; however, the peak range seemed to be between 70 and 79 years [4,6,33]. In our current study, we identified an overall incidence of uveal melanoma of 4.637 per million (95% CI, 4.458-4.821), which is similar to earlier reports, and confirmed the sexual variation of incidence. Elderly people, White race, and non-Hispanic population tended to be more vulnerable to have uveal melanoma. However, we did notice a considerable overall APC of 4.125% (p � 0.03) during 2010-2015. Meanwhile, the increasing trends existed in mid-aged (60-80 years) people, female elderly people, White males, and non-Hispanic population. Unlike retinoblastoma, there are few studies that focus on laterality of uveal melanoma. Also, sex-stratified analysis of uveal melanoma laterality is scarce as well. It is quite understandable that males had a significantly higher incidence of uveal melanoma in both eyes than females, considering the men's higher total age-adjusted incidence. However, our relatively short follow-up duration, higher p value set-up (0.05), and marginally statistical significance cannot be neglected.
ough a higher IRR of a male left eye (1.141 (1.023-1.272)) and total right eye APC (4.389%, p � 0.046) might be true reflection of reality, they might also be false positive as well. Hence, the data should be interpreted with caution, and further studies are needed to test the conclusion.
In terms of potential prognostic predictors, it has been implicated that older age at diagnosis and male gender correlate with reduced survival [34], which is consistent with the findings in UK [35], Sweden [36], and Denmark [37]. ough no sex differences were found in the COMS, maximum basal tumor diameter together with age was the strongest predictor of mortality for uveal melanoma. [38] Yet researchers are not able to determine to what extent these associations are results of bias generating from confounding factors [39]. In our study, we adopted the X-tile software to optimize the age range cutoff instead of subjective classification. According to the results of X-tile, we divided patients into 3 age groups of OS and DSS analyses (due to different causes of mortality): group a, 15- [40] We identified that 5-year accumulative OS and DSS for uveal melanoma was 61.8% and 66.5%, respectively. Compared with previous studies, the survival period seemed to be lower, which may be due to short recruiting time window. In our univariate analyses, older age, White race, higher AJCC stage, severe summary stage, distant metastasis, and no radiation treatment showed significantly lower survival rates in OS and DSS analyses. However, only factors of age, AJCC stage, and radiation persisted to demonstrate prediction potential in multivariate analyses. Despite our restriction of observation time period, this study might act as a supplement to other long-time follow-up literatures using the old staging system. We observed reduced survival period in patients who underwent primary-site surgery.
ough it lost statistical significance in multivariate analyses, it still brought up the controversial topic of treatment decision-making. Our analyses indicated that patients who underwent radiation therapy showed improved survival period, which supported the COMS trial comparing radiation with enucleation during 15 years of follow-up [25].
ere has been a significant shift from local resection and enucleation toward radiotherapy in the United States [41]. Radiation therapy displayed wonderful local control and globe preservation, but long-term vision loss is inevitable. Some alternative treatments like transpupillary thermal therapy also showed efficacy in residual uveal melanoma [42,43]. Moreover, a series of novel approaches are currently developing. For example, human tissue factor VII is commonly overexpressed in uveal melanoma and contributes to tumor growth, thrombosis, angiogenesis, and metastasis [44]. ICON-1, which is a synthetic structural variant of factor VII, binds to tumor cells and initiates a signal cascade targeting immune cells to pathological tissue (NCT02771340). However, we still wait for long-term outcomes from different research studies.
While SEER registries give us the easy access to largescale population-based data from multicenters, which is very helpful to do longitudinal analysis on tumor epidemiological study, there are some inherent limitations as well. e database lacks information such as detail surgical depiction, comorbidities, hospital volume, and tumor recurrence.
ere are also concerns regarding misclassification among different registries. In terms of statistical methodology, under a certain sample scale, a relatively higher p value cutoff (0.05) might cause false significance, which could give chances of deception and misinterpretation of data. ough we have compared similar studies using the SEER database with different case numbers and follow-up duration, most of

Conclusion
In summary, we identified the incidence of 4.637 per million population of uveal melanoma during 2010-2015. ere is a significantly increasing APC of 4.215% and continued gender preference toward incidence. Age at diagnosis, AJCC stage, and radiation therapy may be potential predictors of prognosis. ese findings may raise public attention to monitor epidemiological trends, prognostic factors, and treatment selection of uveal melanoma. Our study might supplement previous long-time follow-up literatures using the old staging system. Further studies are needed to confirm our results.

Data Availability
e raw data about uveal melanoma cases supporting this research are from http://www.seer.cancer.gov. e proceeded data are all available from the corresponding author upon request (Yejuan@zju.edu.cn).