血小板与淋巴细胞比值和中性粒细胞与淋巴细胞比值对早期胃癌诊断的预测价值

The predictive value of platelet to lymphocyte ratio and neutrophil to lymphocyte ratio in the diagnosis of early gastric cancer

目的探讨血小板与淋巴细胞比值(PLR)和中性粒细胞与淋巴细胞比值(NLR)与早期胃癌的相关性,评估PLR和NLR对早期胃癌诊断的预测价值。方法纳入2017年1月1日至2020年12月31日在浙江省台州医院住院治疗的178例早期胃癌、129例慢性胃炎和122例胃上皮内瘤变(GIN)患者,运用Rand随机函数将所有患者按7∶3分为训练组(301例,其中早期胃癌125例,慢性胃炎90例,GIN 86例)和验证组(128例,其中早期胃癌53例,慢性胃炎39例,GIN 36例)。收集所有患者的年龄、性别、血常规指标、癌胚抗原水平、幽门螺杆菌(H.pylori)感染情况等资料,比较训练组中早期胃癌、慢性胃炎与GIN患者的血常规指标和临床特征,训练组中早期胃癌患者与慢性胃炎+GIN患者(以下简称非早期胃癌患者)的血常规指标,分析早期胃癌的独立危险因素。绘制受试者操作特征曲线(ROC),分析独立危险因素诊断预测早期胃癌的最佳临界值、曲线下面积(AUC)、OR和95%置信区间(95%CI)等,建立诊断预测模型,运用Hosmer-Lemeshow检验模型的拟合度,比较模型应用于训练组与验证组的AUC以评估模型的区分度。统计学方法采用Kruskal-Wallis H检验、Mann-Whitney U检验或Wilcoxon秩和检验、卡方检验,以及单因素和多因素logistic回归分析。结果训练组中,慢性胃炎、GIN、早期胃癌患者中男性和女性占比分别为50.0%(45/90)、50.0%(45/90),61.6%(53/86)、38.4%(33/86),69.6%(87/125)、30.4%(38/125),差异有统计学意义(χ^(2)=8.49,P=0.014);早期胃癌患者的男性占比高于慢性胃炎患者,差异有统计学意义(χ^(2)=8.48,P=0.004)。训练组中慢性胃炎、GIN、早期胃癌患者的H.pylori感染率、年龄、PLR、NLR、淋巴细胞计数、中性粒细胞计数、癌胚抗原分别为18.9%(17/90)、18.6%(16/86)和43.2%(54/125),54.0岁(45.5岁,64.0岁)、63.0岁(58.0岁,66.3岁)和66.0岁(58.5岁,71.0岁),113.70(84.48,136.09)、120.00(97.94,138.37)和124.29(101.97,173.57),1.55(1.17,2.23)、1.71(1.44,2.02)和2.04(1.57,2.62),2.00×10^(9)/L(1.50×10^(9)/L,2.40×10^(9)/L)、1.75×10^(9)/L(1.50×10^(9)/L,2.40×10^(9)/L)和1.60×10^(9)/L(1.30×10^(9)/L,2.05×10^(9)/L),3.00×10^(9)/L(2.38×10^(9)/L,3.90×10^(9)/L)、3.00×10^(9)/L(2.48×10^(9)/L,3.40×10^(9)/L)和3.30×10^(9)/L(2.60×10^(9)/L,4.30×10^(9)/L),1.70 g/L(1.10 g/L,2.50 g/L)、2.05 g/L(1.48 g/L,2.90 g/L)和2.50 g/L(1.55 g/L,3.40 g/L),差异均有统计学意义(χ^(2)=21.26,H=41.00、11.79、21.13、10.82、8.54、14.42;均P<0.05)。早期胃癌患者H.pylori感染率高于慢性胃炎和GIN患者,早期胃癌和GIN患者的年龄均大于慢性胃炎患者,早期胃癌患者的NLR和PLR均高于慢性胃炎患者,早期胃癌患者的NLR高于GIN患者,早期胃癌患者的淋巴细胞计数低于慢性胃炎患者、中性粒细胞计数和癌胚抗原高于慢性胃炎患者,差异均有统计学意义(χ^(2)=13.98、13.90,Z=-6.13、-4.15、-4.07、-3.25、-3.40、-3.18、-2.62、-3.74;均P<0.017)。早期胃癌患者的PLR、NLR、中性粒细胞计数和癌胚抗原均高于非早期胃癌患者[124.29(101.97,173.57)比117.97(101.57,137.32)、2.04(1.57,2.62)比1.66(1.25,2.17)、3.30×10^(9)/L(2.60×10^(9)/L,4.30×10^(9)/L)比3.00×10^(9)/L(2.40×10^(9)/L,3.60×10^(9)/L)、2.50 g/L(1.55 g/L,3.40 g/L)比1.90 g/L(1.23 g/L,12.70 g/L)],淋巴细胞计数低于非早期胃癌患者[1.60×10^(9)/L(1.30×10^(9)/L,2.05×10^(9)/L)比1.80×10^(9)/L(1.50×10^(9)/L,2.20×10^(9)/L)],差异均有统计学意义(Z=-3.23、-4.45、-2.91、-3.30,-2.35,均P<0.05)。ROC分析显示,PLR、NLR、癌胚抗原、中性粒细胞计数和淋巴细胞计数的最佳临界值分别为138.18、1.76、2.70 g/L、3.40×10^(9)/L、1.80×10^(9)/L。单因素分析显示,性别、年龄、H.pylori感染情况、中性粒细胞计数、PLR、NLR、淋巴细胞计数和癌胚抗原水平均与早期胃癌有关(χ^(2)=5.98、27.73、21.26、8.26、10.26、22.80、4.81、25.91,均P<0.05)。多因素分析显示,年龄≥70岁(OR=9.267,95%CI 3.239~26.514)、H.pylori感染(OR=3.353,95%CI 1.862~6.037)、NLR>1.76(OR=2.084,95%CI 1.190~3.648)、PLR>138.18(OR=2.452,95%CI 1.325~4.539)、癌胚抗原>2.70 g/L(OR=2.637,95%CI 1.490~4.667)是早期胃癌的独立危险因素(均P<0.05)。Hosmer-Lemeshow检验显示模型预测值与实际观测值比较差异无统计学意义(P>0.05),提示模型的拟合度较好。在训练组人群中,诊断预测模型的AUC为0.787(95%CI 0.737~0.832,P<0.001),将该模型应用于验证组进行验证,结果显示模型的AUC为0.664(95%CI 0.576~0.745,P<0.001),提示模型具有良好的区分度。结论NLR和PLR是早期胃癌的独立危险因素,可能有助于识别早期胃癌。本研究建立的诊断模型具有较好的区分度和拟合度,可为临床诊断早期胃癌提供重要的参考信息,有助于患者进行早期治疗并改善预后。

Objective To explore the association of platelet to lymphocyte ratio(PLR)and neutrophil to lymphocyte ratio(NLR)with early gastric cancer(EGC),and to assess the predictive value of PLR and NLR in EGC diagnosis.Methods From January 1,2017 to December 31,2020,178 patients with EGC,129 patients with chronic gastritis(CG),122 patients with gastric intraepithelial neoplasia(GIN)admitted and treated at Taizhou Hospital of Zhejiang Province were enrolled.According to Rand random function and with the ratio of 7 to 3,the patients were divided into training group(n=301,125 cases of EGC,90 cases of CG,86 cases of GIN)and validation group(n=128,53 cases of EGC,39 cases of CG,36 cases of GIN).The age,gender,routine blood test,carcinoembryonic antigen(CEA)level,Helicobacter pylori(H.pylori)infection status and other data of the patients were collected.The routine blood test and clinical characteristics of EGC,CG and GIN patients of the training group,and the routine blood test of EGC patients and CG+GIN patients(hereinafter referred to as non-EGC group)of training group were compared to analyzed the independent risk factors of EGC.Receiver operator characteristic curve(ROC)was drawn.The optimal cut-off value,area under the curve(AUC),OR,95%confidence interval(95%CI)of independent risk factors were analyzed for EGC diagnosis and prediction.A diagnostic prediction model was established,and the model was apply to the validation group for validation.Hosmer-Lemeshow test was used to test the fitting degree of the model.Compared the AUC of the model applied to training group with validation group to evaluate the discrimination of model.Kruskal-Wallis H test,Mann-Whitney U test or Wilcoxon rank sum test,chi square test,and univariate and multivariate logistic regression analysis were used for statistical analysis.Results In the training group,the proportions of males and females in CG,GIN and EGC patients were 50.0%(45/90)and 50.0%(45/90),61.6%(53/86)and 38.4%(33/86),69.6%(87/125)and 30.4%(38/125),respectively,and the difference was statistically significant(χ^(2)=8.49,P=0.014).The proportion of males in EGC patients was higher than that in CG patients,and the difference was statistically significant(χ^(2)=8.48,P=0.004).The H.pylori infection rate,age,PLR,NLR,lymphocyte count,neutrophil count,and CEA level of CG,GIN and EGC patients in the training group were 18.9%(17/90),18.6%(16/86)and 43.2%(54/125);54.0 years old(45.5 years old,64.0 years old),63.0 years old(58.0 years old,66.3 years old)and 66.0 years old(58.5 years old,71.0 years old);113.70(84.48,136.09),120.00(97.94,138.37)and 124.29(101.97,173.57),1.55(1.17,2.23),1.71(1.44,2.02)and 2.04(1.57,2.62),2.00×10^(9)/L(1.50×10^(9)/L,2.40×10^(9)/L),1.75×10^(9)/L(1.50×10^(9)/L,2.40×10^(9)/L)and 1.60×10^(9)/L(1.30×10^(9)/L,2.05×10^(9)/L),3.00×10^(9)/L(2.38×10^(9)/L,3.90×10^(9)/L),3.00×10^(9)/L(2.48×10^(9)/L,3.40×10^(9)/L)and 3.30×10^(9)/L(2.60×10^(9)/L,4.30×10^(9)/L),1.70 g/L(1.10 g/L,2.50 g/L),2.05 g/L(1.48 g/L,2.90 g/L)and 2.50 g/L(1.55 g/L,3.40 g/L),respectively,and the differences were statistically significant(χ^(2)=21.26,H=41.00,11.79,21.13,10.82,8.54 and 14.42;all P<0.05).The H.pylori infection rate of EGC patients was higher than that of CG and GIN patients,the ages of EGC and GIN patients were older than that of CG patients,the NLR and PLR levels of EGC patients were higher than those of CG patients,the NLR level of EGC patients was higher than that of GIN patients,the level of lymphocyte count of EGC patients was lower than that of CG patients,and the levels of neutrophil count and CEA were higher than those of CG patients,and the differences were statistically significant(χ^(2)=13.98 and 13.90,Z=-6.13,-4.15,-4.07,-3.25,-3.40,-3.18,-2.62 and-3.74;all P<0.017).The levels of PLR,NLR,neutrophil count and CEA of EGC patients were all higher than those of non-EGC patients(124.29(101.97,173.57)vs.117.97(101.57,137.32);2.04(1.57,2.62)vs.1.66(1.25,2.17);3.30×10^(9)/L(2.60×10^(9)/L,4.30×10^(9)/L)vs.3.00×10^(9)/L(2.40×10^(9)/L,3.60×10^(9)/L);2.50 g/L(1.55 g/L,3.40 g/L)vs.1.90 g/L(1.23 g/L,2.70 g/L)),and the lymphocyte count level was lower than that of non-EGC patients(1.60×10^(9)/L(1.30×10^(9)/L,2.05×10^(9)/L)vs.1.80×10^(9)/L(1.50×10^(9)/L,2.20×10^(9)/L)),and the differences were statistically significant(Z=-3.23,-4.45,-2.91,-3.30 and-2.35;all P<0.05).The results of ROC analysis showed that the optimal cut-off value of PLR,NLR,CEA,neutrophil count and lymphocyte count was 138.18,1.76,2.70 g/L,3.40×10^(9)/L,1.80×10^(9)/L,respectively.The results of univariate analysis indicated that the gender,age,H.pylori infection,neutrophil count,PLR,NLR,lymphocyte count and CEA were all related to EGC(χ^(2)=5.98,27.73,21.26,8.26,10.26,22.80,4.81 and 25.91;all P<0.05).The results of multivariate analysis demonstrated that age≥70 years old(OR=9.267,95%CI 3.239 to 26.514),H.pylori infection(OR=3.353,95%CI 1.862 to 6.037),NLR>1.76(OR=2.084,95%CI 1.190 to 3.648),PLR>138.18(OR=2.452,95%CI 1.325 to 4.539),CEA>2.70 g/L(OR=2.637,95%CI 1.490 to 4.667)were independent risk factors for EGC(all P<0.05).The Hosmer-Lemeshow test showed that there was no statistically significant difference between the predicted value of the model and the actual observed value(P>0.05),which indicated that the fitting degree of the model was good.In the training group,the AUC of the diagnostic prediction model was 0.787(95%CI 0.737 to 0.832,P<0.001).The model was applied to the validation group for validation,and the result showed that the AUC of the model was 0.664(95%CI 0.576 to 0.745,P<0.001),which indicated that the discrimination of the model was good.Conclusions PLR and NLR are independent risk factors of EGC,and may help to identify EGC.In this study the established diagnostic model has good discrimination and fitting degree,which can provide important reference information for early clinical diagnosis of EGC,which may facilitate early treatment and improve prognosis of patients.