[Prognostic value of a predictive model comprising preoperative inflammatory response and nutritional indexes in patients with gastric cancer].

Objective: To investigate the prognostic value of preoperative inflammatory and nutritional condition detection in the postoperative survival, and establish a prognostic model for predicting the survival of patients with gastric cancer. Methods: The clinicopathological data of 1123 patients with gastric cancer who had undergone radical gastrectomy in Tianjin Medical University Cancer Institute & Hospital from January 2005 to December 2014 were retrospectively analyzed. Patients with history of other malignancy, with history of gastrectomy, who had received preoperative treatment, who died during the initial hospital stay or first postoperative month, and missing clinical and pathological information were excluded. Cox univariate and multivariate analyses were used to identify independent clinicopathological factors associated with the survival of these gastric cancer patients. Cox univariate analysis was used to identify preoperative inflammatory and nutritional indexes related to the survival of patients with gastric cancer after radical gastrectomy. Moreover, the Cox proportional regression model for multivariate survival analysis (forward stepwise regression method based on maximum likelihood estimation) was used. The independent clinicopathological factors that affect survival were incorporated into the following three new prognostic models: (1) an inflammatory model: significant preoperative inflammatory indexes identified through clinical and univariate analysis; (2) a nutritional model: significant preoperative nutritional indexes identified through clinical and univariate analysis; and (3) combined inflammatory/nutritional model: significant preoperative inflammatory and nutritional indexes identified through clinical and univariate analysis. A model that comprised only pT and pN stages in tumor TNM staging was used as a control model. The integrated area under the receiver operating characteristic curve (iAUC) and C-index were used to evaluate the discrimination of the model. Model fitting was evaluated by Akaike information criterion analysis. Calibration curves were used to assess agreement between the predicted probabilities and actual probabilities at 3-year or 5-year overall survival (OS). Results: The study cohort comprised 1 123 patients with gastric cancer. The mean age was 58.9±11.6 years, and 783 were males. According to univariate analysis, age, surgical procedure, extent of lymph node dissection, tumor location, maximum tumor size, number of examined lymph nodes, pT stage, pN stage, and nerve invasion were associated with 5-year OS after radical gastrectomy for gastric cancer (all P<0.050). Multivariate analysis further identified age (HR: 1.18, 95%CI: 1.03-1.36, P=0.019), maximum tumor size (HR: 1.19, 95%CI: 1.03-1.38, P=0.022), number of examined lymph nodes (HR: 0.79, 95%CI: 0.68-0.92, P=0.003), pT stage (HR: 1.40, 95%CI: 1.26-1.55, P<0.001) and pN stage (HR: 1.28, 95%CI: 1.21-1.35, P<0.001) as independent prognostic factors for OS of gastric cancer patients. Additionally, according to univariate survival analysis, the preoperative inflammatory markers of neutrophil count, percentage of neutrophils, neutrophil/lymphocyte ratio, platelet/neutrophil ratio and preoperative nutritional indicators of serum albumin and body mass index were potential prognostic factors for gastric cancer (all P<0.05). On the basis of the above results, three models for prediction of prognosis were constructed. Variables included in the three models are as follows. (1) Inflammatory model: age, maximum tumor size, number of examined lymph nodes, pT stage, pN stage, percentage of neutrophils, and neutrophil-lymphocyte ratio; (2) nutritional model: age, maximum tumor size, number of examined lymph nodes, pT stage, pN stage, and serum albumin; and (3) combined inflammatory/nutritional model: age, maximum tumor size, number of examined lymph nodes, pT stage, pN stage, percentage of neutrophils, neutrophil-lymphocyte ratio, and serum albumin. We found that the predictive accuracy of the combined inflammatory/nutritional model, which incorporates both inflammatory indicators and nutrition indicators (iAUC: 0.676, 95% CI: 0.650-0.719, C-index: 0.698),was superior to that of the inflammation model (iAUC: 0.662, 95% CI: 0.673-0.706;C-index: 0.675), nutritional model (iAUC: 0.666, 95% CI: 0.642-0.698, C-index: 0.672), and TNM staging control model (iAUC: 0.676, 95% CI: 0.650-0.719, C-index: 0.658). Furthermore, the combined inflammatory/nutritional model had better fitting performance (AIC: 10 762) than the inflammatory model (AIC: 10 834), nutritional model (AIC: 10 810), and TNM staging control model (AIC: 10 974). Conclusions: Preoperative percentage of neutrophils, NLR, and BMI have predictive value for the prognosis of gastric cancer patients. The inflammatory / nutritional model can be used to predict the survival and prognosis of gastric cancer patients on an individualized basis.目的： 探讨术前炎性反应和营养状况的检测对胃癌患者术后生存的预测价值，并建立预测胃癌患者生存的预后模型。 方法： 本研究为回顾性观察性研究，收集2005年1月至2014年12月就诊于天津医科大学肿瘤医院接受胃癌根治术的1 123例胃腺癌患者的临床病理资料。排除有恶性肿瘤病史、胃切除手术史、术前新辅助治疗、围手术期死亡及病例资料不完整者。通过Cox单因素和多因素分析来确定影响本组胃癌患者生存的独立临床病理因素；Cox单因素分析来确定与本组胃癌患者生存有关的术前炎性反应标志物和营养标志物。采用Cox风险比例回归模型进行多因素生存分析（采用基于最大似然估计的向前逐步回归法），将影响生存的独立临床病理因素分别纳入到3种新的预后模型中：（1）炎性反应模型：纳入临床病理因素和单因素分析有意义的术前炎性反应标志物；（2）营养模型：纳入临床病理因素和单因素分析有意义的术前营养指标；（3）炎性反应-营养模型：纳入临床病理因素结合单因素分析有意义的术前炎性反应标志物和营养标志物；同时，将仅仅纳入肿瘤TNM分期中的pT和pN分期的模型作为对照模型。通过评估受试者工作曲线下的综合面积（iAUC）和C-index来评估模型的区分度；通过赤池信息量标准（AIC）分析以评估模型拟合性能。校准曲线用于评估3年或5年的总生存（OS）预测概率和实际概率之间的一致性。 结果： 本研究纳入的1 123例患者中，年龄（58.9±11.6）岁，男783例。单因素分析结果显示：年龄、手术方式、淋巴结清扫范围、肿瘤位置、肿瘤最大径、送检淋巴结数目、pT分期、pN分期以及侵犯神经与胃癌根治术后患者的5年OS有关（均P<0.05）。多因素分析进一步确定，年龄（HR：1.18，95%CI：1.03~1.36，P=0.019）、肿瘤最大径（HR：1.19，95%CI：1.03~1.38，P=0.022）、送检淋巴结数目（HR：0.79，95%CI：0.68~0.92，P=0.003）、pT分期（HR：1.40，95%CI：1.26~1.55，P<0.001）和pN分期（HR：1.28，95%CI：1.21~1.35，P<0.001）为影响患者OS的独立预后因素。术前炎性反应标志物中性粒细胞计数、中性粒细胞百分比、中性粒细胞/淋巴细胞比值（NLR）、血小板/中性粒细胞比值以及术前营养指标血清白蛋白值和体质指数与预后有关（均P<0.05）。3种模型纳入变量：（1）炎性反应模型：将年龄、肿瘤最大径、送检淋巴结数目、pT分期、pN分期、中性粒细胞百分比和NLR纳入；（2）营养模型：将年龄、肿瘤最大径、送检淋巴结数目、pT分期、pN分期和白蛋白值纳入；（3）炎性反应-营养模型：将年龄、肿瘤最大径、送检淋巴结数目、pT分期、pN分期、中性粒细胞百分比、NLR和白蛋白值纳入。构建的3种预后预测模型结果显示：炎性反应-营养模型预测准确性（iAUC：0.676，95%CI：0.650~0.719，C-index：0.698）优于炎性反应模型（iAUC：0.662，95%CI：0.673~0.706；C-index：0.675）、营养模型（iAUC：0.666，95%CI：0.642~0.698，C-index：0.672）以及TNM分期对照模型（iAUC：0.676，95%CI：0.650~0.719，C-index：0.658）；同时，炎性反应-营养模型具有较好的模型拟合度（AIC：10 762），优于炎性反应模型（AIC：10 834）、营养模型（AIC：10 810）和TNM分期对照模型（AIC：10 974）。 结论： 术前中性粒细胞百分比、NLR以及体质指数对胃癌患者预后具有预测价值。炎性反应-营养模型可用于个体化地预测胃癌患者的生存和预后。.