Statistical analysis

Our primary endpoint was DMFS, with disease-free survival (DFS) and overall survival (OS) as the secondary endpoints. DMFS was calculated from the first day of treatment to the date of first distant relapse or death from any cause, whichever occurred first. DFS was calculated from the first day of treatment to the date of the first relapse at any site or death from any cause, whichever occurred first. OS was calculated from the first day of treatment to the date of death from any cause. The Kaplan-Meier method and the log-rank test were used to estimate DMFS, DFS, and OS. Cox regression analyses were used to calculate the hazard ratios (HRs).

We assessed the differentially expressed proteins between the metastatic and nonmetastatic group with Student's t-test. The χ² test or Fisher's exact test were used to compare categorical variables. We used univariate Cox analysis to analyze the correlations between the clinical outcomes and factors such as the PSDM, age, sex, T stage, N stage, and EBV DNA load. The factors that correlated significantly with the outcomes were further tested with multivariate Cox regression analysis. The predictive efficiency with regard to metastasis was compared between the PSDM and single proteins, as well as between the PSDM and clinical factors, using time-dependent receiver operating characteristic (ROC) curves and time-dependent area under the curve (AUC) analysis. We also performed a subgroup analysis with the Kaplan-Meier method and log-rank test to evaluate whether the PSDM could distinguish patients with better outcomes from those with worse outcomes in the groups with different stages and EBV DNA levels. All statistical tests were two-sided and differences were deemed significant at a P value < 0.05. R software (version 3.6.1) was used to perform the statistical analyses.