Multicenter, International Assessment of the Eighth Edition of the American Joint Committee on Cancer Cancer Staging Manual for Conjunctival Melanoma

Puneet Jain, MD; Paul T. Finger, MD; Bertil Damato, MD; Sarah E. Coupland, MD, PhD; Heinrich Heimann, MD; Nihal Kenawy, MD; Niels J. Brouwer, MD; Marina Marinkovic, MD; Sjoerd G. Van Duinen, MD, PhD; Jean Pierre Caujolle, MD; Celia Maschi, MD; Stefan Seregard, MD; David Pelayes, MD; Martin Folgar, MD; Yacoub A. Yousef, MD; Hatem Krema, MD; Brenda Gallie, MD; Alberto Calle-Vasquez, MD; for the American Joint Committee on Cancer Ophthalmic Oncology Task Force

C onjunctival melanomas comprise 2% of all ocular tumors and 5% of eye melanomas [1][2][3] ; however, their incidence is increasing. 4 Tumor-related mortality is reported to be 30%. 5 Reliable prognostic tools are needed to enhance clinical and basic science research.
Published by the American Joint Committee on Cancer (AJCC) and used by the Union for International Cancer Control, the TNM staging system is widely implemented around the world. In 2016, the AJCC Ophthalmic Oncology Task Force created the eighth edition of the conjunctival melanoma staging system with data from more than 50 eye cancer specialists in 18 countries. Changes from the seventh edition include further defining the tumors' circumferential extent to define the clinical T categories. 6 Widely accepted by medical oncology, ophthalmic oncology, radiation oncology, and medical journals, the AJCC-TNM staging system serves to standardize data reporting, case-to-case prognosis, and selection of the most suitable treatment for conjunctival melanoma. Specialists who stage cancers have recognized the need for standardized collaborative data sharing. 7,8 We performed, to our knowledge, the first multicenter, international study to evaluate the accuracy of the eighth edition of the AJCC Cancer Staging Manual in estimating mortality rates of metastasis from conjunctival melanoma.

Methods
Patients diagnosed with conjunctival melanoma from January 1, 20011, , to December 31, 2013, were included in this multicenter, international study. Data were collected retrospectively and entered into a secure online database. Data analysis was performed from July 7, 2018, to September 11, 2018. This study adhered to the tenets of the Declaration of Helsinki 9 and the Health Insurance Portability and Accountability Act of 1996. All participating centers obtained internal institutional review board approval to perform retrospective medical record reviews and contribute data to the AJCC Ophthalmic Oncology Task Force Conjunctival Melanoma Registry. The Princess Margaret Cancer Centre determined that individual patient consent was not required because there were no patient identifiers.

Conjunctival Melanoma Registry Centers
In 2015, at the first Eye Cancer Working Day, held at the Curie Institute (Paris, France), 158 eye cancer specialists were invited to participate in this conjunctival melanoma registry. Conceptualized as an internet-based data-sharing registry, the data fields were designed by participating eye cancer specialists. The resulting registry included data from 10 ophthalmic oncology centers in 9 countries (2 in the United States and 1 each in Canada, Colombia, Argentina, France, Netherlands, United Kingdom, Sweden, and Jordan) on 4 continents (North America, South America, Europe, and Asia).

Data Security
The following features were incorporated to ensure data security and study participant privacy. There were no personal identifiers. The participants were assigned a unique identification number at their local institution that linked patients to locally stored corresponding identification numbers to ensure accuracy of follow-up and outcomes. Secure Socket Layer encryption was used to prevent information from being viewed by third parties. Registry access was limited to participants with user accounts. Therefore, the login page required a unique username and password combination to access the application. The database incorporated record locking to prevent different users from accessing the same record for the same study participant at the same time. This locking prevented unintentional overwriting or data corruption. In addition, audit trails were automatically created. Such logs were stored in a separate, secured directory and were not available online.

Tumor Staging
Clinical (cT) and pathologic (pT) staging were performed according to the staging system for conjunctival melanoma in the eighth edition of the AJCC Cancer Staging Manual 6 ( Table 1). According to the AJCC, T1 disease is confined to the bulbar conjunctiva, T2 disease affects the nonbulbar conjunctiva and/or caruncle, and local invasion to adjacent tissues elevates the conjunctival melanoma to T3 disease. T4 disease denotes central nervous system invasion. 6 Definitions Ophthalmic oncology examinations were performed according to the standards and practice of each participating center. These practices included slitlamp and gonioscopic photography and high-frequency anterior segment ultrasonography to rule out intra-ocular invasion. Metastases were subgrouped at presentation or during follow-up. Systemic screening for metastasis was performed according to the custom and practice of the local institutions. Time to death was defined as the interval between date of diagnosis and the date of metastasis. Because at the time of the study no curative treatment for metastatic conjunctival melanoma existed, metastasis and mortality were deemed to be equivalent for the study.

Statistical Analysis
Means (SDs) or medians (interquartile ranges [IQRs]) were used to express continuous variables, whereas categorical vari-

Key Points
Question Can the eighth edition of the American Joint Committee on Cancer (AJCC) Cancer Staging Manual for conjunctival melanoma be used to accurately estimate metastasis and mortality rates?
Findings In this case series of 288 patients, there was a significantly higher cumulative mortality rate among patients with cT2 and cT3 conjunctival melanoma compared with those presenting with cT1 conjunctival melanoma.

Meaning
Higher T-staged tumors were associated with both an earlier and greater incidence of metastasis; therefore, the results from this multicenter, international registry study support the use of the eighth edition AJCC staging system for conjunctival melanoma.
ables were expressed as proportions. Group comparisons were made using the t test or Wilcoxon rank sum test for nonparametric variables. Analysis of variance or the Kruskal-Wallis test was used to compare variables across more than 2 groups. The χ 2 test or Fisher exact test was used to analyze group differences across categorical variables. Separate survival analyses were performed using metastasis and mortality as the censoring variable, and Kaplan-Meier curves were plotted to depict cumulative survival rates at various time points.
Because of the number of patients analyzed in this study, the different tumor subgroups were combined into T1, T2, and T3 for plotting meaningful Kaplan-Meier curves. Comparison between the survival rates of different subgroups was analyzed using the log-rank test. The survival probability for each outcome was assessed using Cox proportional hazards regression models and displayed using hazard ratios (HRs). Covariates used for adjusting HRs were those with a P < .10 in univariate models and those that were associated with mortality in previous studies 3,5,7 (tumor location, tumor invasion, ulceration, and lymphatic invasion).
Data were exported into Microsoft Excel (Microsoft Corp) for analysis. Data were analyzed using Stata, version 12.1 (Stata-Corp) statistical analysis software package, and a 2-tailed P < .05 was considered to be statistically significant.

Demographics
A total of 288 eyes from 288 patients (mean [SD] age, 59.7 [16.8] years; range, 15-95 years; 147 [51.0%] male) with conjunctival melanoma were studied. All patients had the relevant data set; therefore, no patient was excluded from the study. There was slight preponderance of right eye involvement (n = 154 [53.4%]). The median follow-up time (from the time of definitive treatment) was 4.4 years (IQR, 2.3-6.9 years; range, 1 month to 14.3 years).  Table 2). There were no cT4 tumors.

Cumulative Mortality Rates According to Pathologic Stage
The cumulative mortality rates with respect to pathologic categories and at various time points were 0% at 1 year, 3.1% (95%    Abbreviation: NA, not applicable. a Unspecified cT1 disease in 28 (9.7%) and unspecified cT2 disease in 5 (1.7%).
On multivariable Cox proportional hazards regression analysis, patients with higher cT stages (cT2, cT3) had 3 times greater risk of mortality (HR, 3.25; 95% CI, 1.48-7.11; P = .003) compared with those with cT1. Patients with conjunctival melanomas reported to have ulceration at presentation also had a significantly higher risk of mortality (hazard ratio, 7.58; 95% CI, 1.02-56.32; P = .04) ( Table 3). This study also notes that involvement of the plica, caruncle, and lymphatic invasion likely did not influence mortality rates. Although caruncular and plical involvement were not a significant risk factor for metastasis, there were 15 T3-staged tumors (4.5%). 10

Discussion
Multicenter, international tumor registries are providing new medical evidence by allowing analysis of large numbers of rare tumors. 2,5,7,8,[11][12][13][14] This registry provided evidence from a large case series of eyes with conjunctival melanoma. The findings suggest that the eighth edition of AJCC Cancer Staging Manual for conjunctival melanoma is valid. Specifically, the rates of 1-, 5-, and 10-year mortality increased with the advancing initial cT stage of the primary tumor. The study also revealed independent factors associated with increased mortality, which included tumor thickness, tumor invasion, and ulceration (Table 3). Caruncular or plical involvement did not affect mortality.
The finding that the cumulative rate of mortality with respect to clinical categories showed a steep increase from T1 (3%) to T2 (28%) at the 5-year follow-up suggests the validity of the staging system. This increase in mortality supports the importance of tumor location, with nonbulbar tumors (T2) tending to be more aggressive or harder to control than bulbar tumors (T1). The cumulative 10-year mortality was 3-fold greater for T2 disease compared with T1 disease (43.6% vs 15.2%). Patients with T3 disease were the only ones with mortality at 1 year (cumulative mortality rate of 21%). The 5-year cumulative mortality rates were comparable for patients with T2 and T3 tumors. However, on subgroup analysis, the number of patients in the T3 group was insufficient for 5-year analysis. Although there are no multicenter data on cumulative mortality rates to date, previous single-center studies 2,3,5,7,13,14 have suggested that larger tumors are associated with a higher incidence of metastasis and mortality. All patients in whom primary tumor could not be assessed (ie, cTx and pTx categories) were censored for the calculation of mortality rates.
In this study, the cumulative mortality rates were comparable for both clinical and pathologic T categories, which suggests that clinical staging is at least as useful as pathologic staging for mortality. Univariate and multivariable HR analyses revealed trends that were comparable to previous studies (Table 3). 2,7,10,[13][14][15][16][17] However, contrary to the AJCC definitions and prior studies, 10,12,15 our data analysis did not support conjunctival melanoma involvement of the plica and caruncle being associated with higher mortality rates. A small case series suggested that, like other mucous membrane melanomas, immunotherapy may be effective in the treatment of advanced local and metastatic conjunctival melanoma. 18,19 This study supports the continued use of the conjunctival melanoma staging system published in the eighth edition of the AJCC Cancer Staging Manual because use of the tool helped to accurately estimate melanoma-related mortality. The study suggests a potential for modification of weighting of caruncular and plical involvement as well as T4 definitions. Universal staging allowed for large multicenter, international cooperation investigating a rare ophthalmic cancer. This study provided data analysis to improve our knowledge of conjunctival melanoma.

Limitations
The main limitation of this study is its retrospective design. Another limitation is the small number of patients in some cT categories; 75.7% of the tumors were cT1 staged, and there were no cT4 tumors. Because there were no criteria for case selection, this distribution represents the clinical experience of the multiple international centers involved in this study. Ethnic/ racial backgrounds were not collected as data; however, because the data were derived from 4 continents, the registry included a broad spectrum of patients. In contrast, Zhou et al 13 reported on mostly cT2 and cT3 tumors from the Chinese population. 5,7