*2.3. Interpretation of Immunohistochemical Staining*

Immunohistochemical staining was interpreted by consensus between two authors (O.-J.L. and D.-H.K.) in a double-blinded fashion to minimize inter-rater variability. Samples with a Cohen's kappa coefficient of less than 0.20 were removed from further analysis. Although immunoreactivity for β-catenin was found in the membrane, cytoplasm, and nucleus, only cytoplasmic staining was assessed for scoring. The expression of NME1 protein in tumor cells was evaluated based on cytoplasmic staining. Cytoplasmic staining of both proteins was semi-quantitatively evaluated using a score calculated by multiplying the intensity score (0, none; 1, weak; 2, moderate; 3, strong) with the proportion score of positive cells (0, absent; 1, 0–10%; 2, 10–50%; 3, 50–80%; 4, >80%). For NME1, its expression was defined as reduced if a composite score was less than two in a tumor. β-catenin expression was considered to be overexpressed in a tumor with a composite score greater than or equal to two. Staining was performed in triplicate and average values of scores were used to determine the expression levels. Cutoff values for the abnormal expression of NME1 and β-catenin were determined considering an internal control consisting of 23 normal lung cores. Representative positive stainings for β-catenin and NME1 expression are shown in adenocarcinoma and squamous cell carcinoma (Figure 1A). Details of the immunohistochemical staining procedure and interpretation for Ki-67 (MKI67) proteins were reported previously [27].

#### *2.4. Study Design*

Patients were randomly selected without stratification or matching by age. The median duration of follow-up was 56 months. The clinical endpoint of the study was recurrence-free survival (RFS), which was defined as the time from the date of the diagnosis to the first recurrence. Variables such as age, sex, histology, pathologic stage, NME1 expression, and adjuvant chemotherapy were initially considered for the analysis of RFS. FFPE tissue samples were obtained from 425 patients because at least 365 patients were needed for analysis of the effect of NME1 expression on RFS under 2-sided α = 0.05 and β = 0.1 (i.e., 90% power).

#### *2.5. Statistical Analyses*

To find factors associated with NME1 reduction in NSCLC, chi-square test (or Fisher's exact test) and Student's *t*-test (or one-way ANOVA) were used for univariate analyses of continuous and categorical variables, respectively. A linear relationship between two continuous variables was analyzed using Pearson's correlation coefficient. The prognostic significance of NME1 on RFS was evaluated by Kaplan–Meier survival curves. The difference between two survival curves was assessed using the log-rank test. Variables with *P* ≤ 0.25 in the univariate analysis were included in the multivariate model. Hazard ratios of predictor variables for survival were estimated using the Cox proportional hazards model after controlling for potential confounding factors. The effect of β-catenin expression and adjuvant chemotherapy on NME1 function was analyzed using a stratified Cox proportional hazards model. No replacement was made for missing values. All statistical analyses were two-sided with a type I error of 5%.
