*2.5. Statistical Analysis*

Statistical analysis was conducted separately for each cohort.

Categorical variables are presented as counts and proportions and continuous variables are displayed as mean (standard deviation). Median (interquartile range) is used to describe variables with a highly skewed distribution.

Chi-square test was used to test the association between categorical variables; the Mann–Whitney U test was used to compare continuous variables with skewed distribution. A logistic regression analysis was carried out to identify predictors of PD-L1 expression. The variables considered in the logistic regression model were *RAD51Bme* (continuous), sex, age, smoking status and histological subtype.

The area under the receiver operating characteristics curve (AUC, 95% CI) was analysed to assess the performance of the *RAD51B* promotor methylation level as a predictive biomarker for PD-1 blockade response. Specificity, sensitivity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were determined for PD-L1, according to positive vs. negative immune scores and for *RAD51B* methylation by applying an empirical cut-off obtained by ROC curve analysis (sensitivity + (1-specificity)). This cut-off value combines the maximum sensitivity and specificity, ensuring the perfect categorization of the samples as positive and negative for the methylation test. For the analysis of combined *RAD51Bme*+/PD-L1+, the test was considered positive when at least one of the variables was plotted, as positive in individual analysis. Diagnostic biomarker performance was calculated, taking into consideration that all the patients included were subjected to anti-PD-1 treatment.

Progression-free survival (PFS) and overall survival (OS) were estimated by means of the Kaplan–Meier method for the Treated Cohort (#2). PFS was defined as the length of time from the beginning of anti-PD-1 blockade until disease progression or death from the disease and OS as the length of time from the beginning of anti-PD-1 blockade until death from any cause. The differences between groups were tested using the log-rank test. Hazard ratios (HRs) from multivariable Cox regression were used to quantify the association between clinicopathological features and survival. *RAD51B* promoter methylation level was considered positive if the quantitative value was above the 75th percentile. A *p*-value smaller than 0.05 (two-sided) indicated statistical significance.

All analyses were performed using IBM SPSS Statistics version 26.0 (SPSS, Chicago, IL, USA) and GraphPad Prism 7.01 (GraphPad Software, La Jolla, CA, USA).
