We determined the distribution of the demographic characteristics (i.e., sex, age at diagnosis [patients diagnosed before 54 years and after 55 years of age], tumor sites [i.e., choroidal melanoma and ciliary body/iris melanoma], year at diagnosis [1999–2005 and 2006–2012], and diagnostic verification methods of uveal melanoma patients identified during the 14-year study period from 1999 to 2012). Differences in the distributions of categorical variables were assessed using χ
2 tests, and mean difference (
±SD) of age was tested by independent
t-test. The survival duration for each uveal melanoma patient was determined as the time from the date of initial uveal melanoma diagnosis to the date of death or end of follow-up (December 31, 2013). The KCCR is also linked to database of the Ministry of the Interior to identify the follow-up loss (e.g., immigration). In our study, there was no loss to follow-up. We estimated the probability of survival from all-causes death and cancer death using the Kaplan-Meier method, according to age at diagnosis, tumor sites, year at diagnosis, and diagnostic verification methods, stratified by sex. Subgroup analysis was performed with the purpose of exploratory assessment. Sex disparities in the survival probability from all-cause death and cancer death of patients with uveal melanoma were addressed and tested using log-rank tests. The life table method was used to estimate 5-year survival and 95% confidence intervals (CIs). In addition, we performed the Cox-proportional hazards model for all-cause deaths and cancer deaths to determine the difference of the survival probability from all-cause death and cancer death between men with uveal melanoma and women with uveal melanoma. There is a limitation to identify uveal melanoma death from causes of death information, because the cause of death data was recorded and classified using only ICD-10 code, which contains only tumor site information and does not contain morphologic information. Therefore, melanoma and eye cancer were treated as the main cause of death for patients diagnosed with uveal melanoma. Of patients with uveal melanoma, the total number of deaths was 90. Of these, 78 patients died from cancer, and 12 died from noncancer deaths. Of these 78 cancer deaths, 8 cases who also had secondary cancers and died of diagnosed secondary cancer other than eye cancer or melanoma, were excluded from cancer death, because we wanted to capture primary cancers with uveal melanoma-related cancer death. Then, we considered the death for remaining 70 cases from other tumor as metastasis that primary tumors were originated from uveal melanoma, because there was no case with multiple primary tumors which died from cancer except for the two cases with unspecified or ill-defined tumor (ICD-10 code C80, C97).
26 In short, we treated these 70 patients who died from cancer death as uveal melanoma-related cancer death. The Cox-proportional hazards model was performed to measure the hazard ratios (HRs) for all-cause death and cancer death as events of interest after adjusting for covariates (age, year of diagnosis, tumor site, and diagnostic verification method). We used the backward elimination procedure to fit the best parsimonious model. We checked for the proportional assumptions using martingale residuals and Scheonfeld residuals, and the proportional hazard assumptions were satisfied.
P < 0.05 was considered statistically significant, and Bonferroni correction was applied to the subgroup analysis. We used Stata 12.0 (StataCorp LP, College Station, TX, USA) and SAS 9.3 (SAS Institute, Cary, NC, USA) software for statistical analyses.