This post-hoc analysis evaluated patients with gastric cancer who were enrolled in a phase III open-label multi-center randomized controlled trial (the CLASSIC trial) that compared capecitabine plus oxaliplatin after D2 gastrectomy versus D2 gastrectomy alone. A consecutive subset of the CLASSIC trial participants (n=144, enrolled between May 2006 and April 2009 at the Yonsei Cancer Center) were selected for the present study. All patients had undergone radical gastrectomy with D2 lymph node dissection for their primary gastric tumors and were randomized to receive either adjuvant chemotherapy or surgery alone. Among the 144 patients, the present study excluded five patients because immunohistochemical (IHC) analysis of their tumors was not technically feasible. Thus, the present study evaluated 139 patients with a median follow-up duration of 98.5 months at April 2017. Data regarding age, sex, tumor stage, and pathological findings were retrospectively obtained from databases of the CLASSIC trial and Severance Hospital. Tumors were staged according to the sixth version of the American Joint Committee on Cancer (AJCC) criteria.

Stored formalin-fixed paraffin-embedded tissues from the patients’ primary gastric tumors were collected, and a microtome was used to create 4-μm sections for IHC. The presence of carcinoma in the sections was confirmed using a pathological review of hematoxylin and eosin staining results. The sections were de-paraffinized and rehydrated using xylene and ethanol, and antigen retrieval was performed using a target retrieval solution (citrate pH 6; Agilent, Santa Clara, CA). The IHC staining was performed using primary antibodies against TS (TS106, Thermo Scientific, Waltham, MA), ERCC1 (8F1, Abcam, Cambridge, UK), and PD-L1 (SP142, Spring Bioscience, Pleasanton, CA) with chromogenic detection that was performed using a peroxidase/DAB kit (EnVision Detection Systems, Agilent). We used the weighted histoscore method (H-score) to determine the TS and ERCC1 expression scores. For TS and ERCC1 expression, sections of HeLa cell blocks and the normal tonsil were used as positive controls for TS and ERCC1 antibody, respectively. Mouse IgG2a (Dako Cytomation, Glustrup, Denmark), excluding the primary antibody, was used as the negative control. For TS and ERCC1 expression, the staining intensity of tumor cells was classified as 0 (negative), 1 (light brown), 2 (brown), or 3 (dark brown). The protein expression was interpreted using the histoscore, calculated by the multiplying staining intensity and proportion of stained tumor cells as follows: (0×% of negative cells)+(1×% of light brown cells)+(2×% of brown cells)+(3×% of dark brown cells) [21]. For example, a specimen with 20% of cells classified as an intensity of 3, 20% as 2, 30% as 1, and 30% as 0 (unstained cells) would have a histoscore of (3×20)+(2×20)+(1×30)=130. Histoscores range from 0 to 300. Tumors with H-score ≥ 100 were defined as having high TS or ERCC1 expression, whereas tumors with H-score < 100 were defined as having low TS or ERCC1 expression. The proportion of membranous PD-L1 staining on tumor cells or intercalated inflammatory cells was evaluated, and tumors with ≥ 1% of cells expressing PD-L1 were defined as PD-L1-positive tumors [22]. Representative IHC images for TS, ERCC1, and PD-L1 are shown in Fig. 1. All pathological reviews were performed by pathology experts who were blinded to the patients’ clinical and survival data.

Fisher exact test or the chi-square test was used to compare clinical variables according to treatment group and biomarker status. Disease-free survival (DFS) was defined as the time from surgery to recurrence, second primary cancer, or death, and overall survival (OS) was defined as time from the date of the operation to the date of death from any cause. Data for patients who had not had an event were censored as of the date of final observation. DFS and OS after surgical resection were compared in each biomarker group using the Kaplan-Meier method and log-rank test. Cox regression analysis was performed to determine the hazard ratios (HR) and 95% confidence intervals (CIs) for DFS and OS according to the clinical variables and biomarkers. The biomarkers’ values for predicting DFS and OS were estimated using multivariate Cox regression analysis, which was adjusted for the T category and N category. Differences were considered statistically significant at two-tailed p-values < 0.05, and all data were analyzed using IBM SPSS software ver. 24.0 (IBM Corp., Armonk, NY). GraphPad Prism software (GraphPad Software Inc., San Diego, CA) was used to create the Kaplan-Meier curves.

The present study’s protocol was approved by the institutional review board (IRB) of Severance Hospital (4-2017-1111). This study is the retrospective post-hoc analysis of CLASSIC trial, and the informed consent for CLASSIC trial enrollment was obtained from all patients. The requirement for additional informed consent for this study was waived by the IRB because this study was performed more than 5 years after the CLASSIC trial enrollment and acquisition of the tumor tissue.

The post-hoc analysis evaluated data from 139 patients with stage II-III gastric cancer who underwent radical gastrectomy with or without adjuvant chemotherapy (Table 1). Eightyfive patients (64%) underwent subtotal gastrectomy and 50 patients (36%) underwent total gastrectomy. The patients were randomized to the adjuvant chemotherapy subgroup (n=66) or the surgery alone subgroup (n=73). In the adjuvant chemotherapy subgroup, 52 patients (80.3%) completed eight cycles of capecitabine plus oxaliplatin chemotherapy and the remaining patients stopped adjuvant chemotherapy because of consent withdrawal (n=7), chemotherapy toxicity (n=3), or disease recurrence during chemotherapy (n=4). In accordance with previous reports regarding this population, the two treatment groups did not have any significant differences in their clinical variables and histological findings (Table 1).

All tumors were analyzed for TS and ERCC1 expression, and the patients were dichotomized according to the IHC score (H-score ≥ 100 vs. H-score < 100). There were no differences in the clinicopathological characteristics of the patients with low TS expression and high TS expression (Table 2). Compared to cases with low ERCC1 expression, cases with high ERCC1 expression were more likely to involve female patients (p < 0.001), signet ring cell carcinoma (p=0.012), the diffuse type based on the Lauren classification (p=0.001), no lymphovascular invasion (p=0.017), and total gastrectomy (p=0.025). Positive tumor expression of PD-L1 was associated with non-signet ring cell carcinoma (p=0.01), the intestinal type based on the Lauren classification (p=0.003), and N2 category disease (p=0.003). There were no significant associations between the TS, ERCC1, and PD-L1 expression statuses in the patients’ tumors (S1 Table).

The values of TS, ERCC1, and PD-L1 expression for predicting DFS and OS after surgical resection were evaluated among all patients. Compared to low TS expression, high TS expression was associated with significantly worse DFS (p=0.039) but not OS (p=0.141) (Fig. 2). In contrast, ERCC1 expression was not associated with DFS or OS. Positive PD-L1 expression was associated with significantly better DFS (p=0.003) and OS (p=0.015) (Fig. 2). In the univariate Cox analysis of DFS and OS (Table 3), higher T category and N category tended to predict poorer DFS (T category, p=0.064; N category, p=0.070) and poorer OS (T category, p=0.051; N category, p=0.203). In the multivariate Cox analysis, which were adjusted for T category and N category (Table 4), high TS expression tended to be associated with poorer DFS (HR, 1.80; 95% CI, 0.99 to 3.28; p=0.053) but not poorer OS. Positive PD-L1 expression independently predicted favorable DFS (HR, 0.33; 95% CI, 0.17 to 0.65; p=0.001) and favorable OS (HR, 0.38; 95% CI, 0.18 to 0.78; p=0.009). The expression of ERCC1 was not associated with DFS or OS in the Cox regression analysis.

In the adjuvant chemotherapy subgroup, high TS expression was associated with significantly worse DFS (p=0.029) and OS (p=0.009), but it was not associated with survival in the surgery alone subgroup (Figs. 3 and 4). Similarly, high ERCC1 expression in the adjuvant chemotherapy subgroup was associated with significantly worse DFS (p=0.029) (Fig. 3). In the surgery alone subgroup, positive PD-L1 expression was associated with favorable DFS (p=0.010) and favorable OS (p=0.015) (Fig. 4), with the DFS and OS values being non-significantly higher than in the adjuvant chemotherapy subgroup (DFS, p=0.141; OS, p=0.354) (Fig. 3). In the multivariate Cox analysis, which was adjusted for T category and N category (Table 5), high TS expression in the adjuvant chemotherapy subgroup independently predicted worse DFS (HR, 2.51; 95% CI, 1.14 to 5.52; p=0.022) and worse OS (HR, 3.34; 95% CI, 1.42 to 7.87; p=0.006). However, high ERCC1 expression showed a trend of being associated with poor DFS, but it was not statistically significant on multivariate Cox analysis (HR, 2.08; 95% CI, 0.94 to 4.58; p=0.070). In the surgery alone subgroup, high PD-L1 expression significantly predicted favorable DFS (HR, 0.25; 95% CI, 0.09 to 0.73; p=0.011) and favorable OS (HR, 0.22; 95% CI, 0.07 to 0.57; p=0.015).