The new prognostic index of advanced gastric cancer using the data from JCOG1013.

342 Background: We previously reported that performance status (PS) ≥1, no prior gastrectomy, number of metastatic sites ≥2, and high serum ALP level are poor prognosis factors and proposed a JCOG prognostic index by analyzing the data of advanced gastric cancer (AGC) patients enrolled in JCOG9912 (Takahari D et al. Oncologist 2014). Recently, Neutrocyte/ Lymphocyte Ratio (NLR) and diffuse type of cancer have also been reported to be prognostic factors of AGC. Methods: Prognostic factors were assessed in patients with AGC who enrolled in JCOG1013, the phase III study comparing docetaxel, cisplatin, and S-1 (DCS) versus cisplatin and S-1 (CS) in first-line treatment (Yamada Y et al. Lancet Gastroenterol Hepatol. 2019) was retrospectively evaluated using a Cox proportional regression model. The overall adequacy of risk prediction in the modified JCOG prognostic index and the original index were assessed by C-statistics. Results: Among 741 patients, 730 with all data for this analysis were selected. The median overall survival (OS) was 14.9 months (95% confidence interval [CI], 14.1–15.8). The median OS of the good (n=233), moderate (n=444), and poor (n=53) risk groups by the original JCOG index was 19.0, 14.2, and 9.1 months, respectively. The HRs compared with the good group were 1.59 [95% CI, 1.33–1.89; p < 0.0001] in the moderate group, and 2.47 [95% CI, 1.81–3.38; p < 0.0001] in the poor group with C-statistics of 0.572. In the multivariable analysis of 7 factors which showed p<0.1 in univariable analysis, PS ≥1 (hazard ratio [HR], 1.490; 95% CI, 1.264–1.755; p < 0.0001), diffuse type (HR, 1.337; 95% CI, 1.122–1.594; p = 0.012), number of metastatic sites ≥2 (HR, 1.256; 95% CI, 1.058–1.492; p = 0.0093), and NLR≥3.1 (HR, 1.539; 95% CI, 1.309–1.809; p < 0.0001) were significantly associated with worse prognosis, whereas no prior gastrectomy, peritoneal metastasis, and high serum ALP level were not. As the modified JCOG prognostic index, patients were classified into three groups according to the number of these four newly identified prognostic risk factors: good (no), moderate (1 or 2), and poor (3 or 4). Median OS of the good (n=221), moderate (n=441), and poor (n=68) risk groups was 20.6, 13.7, and 9.4 months, respectively. The HRs compared with the good group were 1.70 [95% CI, 1.42–2.03; p < 0.0001] in the moderate group, and 3.11 [95% CI, 2.33–4.14; p < 0.0001] in the poor group with C-statistics of 0.591. Three groups by the modified JCOG prognostic index also showed clear separation of OS in either subgroup of DCS or CS. Conclusions: Modified JCOG prognostic index, including NLR ≥3.1 and diffuse type instead of no prior gastrectomy and high serum ALP level, showed clear stratification of OS in JCOG1013.