乳酸及复合模型对非体外循环冠状动脉旁路移植术后机械辅助通气时间延长的预测价值

Evaluation of predictive value of lactate and complex model for prolongation of mechanical ventilation after off-pump coronary bypass grafting

目的基于剂量反应和决策曲线,探讨乳酸(Lac)及复合模型对非体外循环冠状动脉旁路移植(OPCABG)术后机械辅助通气时间延长(PMV)的预测价值。方法回顾性分析北部战区总医院心外科2019年1-12月收治的683例OPCABG患者的临床资料。根据术后机械辅助通气时间是否>24 h,将其分为PMV组(n=107)与非PMV组(n=576)。收集两组患者的基线资料及其他临床指标,筛选有统计学差异的因素;采用多因素logistic回归分析OPCABG术后PMV的影响因素,并构建预测PMV发生风险的复合模型。应用受试者工作特征(ROC)曲线、限制性立方样条模型、决策曲线分析(DCA)等方法评估乳酸及复合模型对OPCABG术后PMV的预测价值。结果两组性别、年龄、体重指数(BMI)、心功能NYHA分级、既往心肌梗死史、既往经皮冠状动脉介入治疗(PCI)史、吸烟史、高血压、糖尿病、高胆固醇血症、3支冠脉狭窄>50%、左主干狭窄>50%、桥血管数、术后即刻乳酸、术前肌酐、术后血红蛋白、术前总胆红素(TBil)、术前直接胆红素(DBil)比较差异均无统计学意义(P>0.05)。两组术前使用主动脉内球囊反搏(IABP)比例、左心室射血分数(LVEF)、肺动脉收缩压(PAP)、术前红细胞宽度(RDW)、术后6 h乳酸、术前血红蛋白(Hb)、术前超敏C反应蛋白(hs-CRP)、术后hs-CRP、术前肌钙蛋白T(TnT)、术前N末端前体脑利钠肽(NT-proBNP)比较差异均有统计学意义(P<0.05)。多因素logistic回归分析结果显示,术前RDW(OR=1.242,95%CI 1.001~1.539)、术后6 h乳酸(OR=1.370,95%CI 1.171~1.604)、PAP(OR=1.043,95%CI 1.002~1.087)及术前NT-proBNP(OR=2.065,95%CI 1.333~3.200)升高是OPCABG术后发生PMV的独立危险因素(P<0.05),而术前使用IABP为其保护因素(OR=0.146,95%CI 0.071~0.301,P<0.05)。乳酸和复合模型预测PMV发生的ROC曲线下面积(AUC)分别为0.582(95%CI 0.518~0.646)、0.727(95%CI 0.674~0.781)。术后6 h乳酸与PMV发生的关联强度呈非线性剂量反应关系(P<0.01)。与术后6 h乳酸相比,决策曲线分析显示,复合模型在阈值概率为0.05~0.75时具有更高的净获益率。结论与术后6 h乳酸相比,复合模型预测OPCABG术后PMV的发生具有更高的临床价值。

Objective To investigate the predictive value of lactate(Lac)and complex model for the occurrence of prolonging mechanical ventilation(PMV)after off-pump coronary bypass grafting(OPCABG)based on dose-response analysis and decision curve analysis.Methods A retrospective analysis was conducted on 683 patients who underwent OPCABG from January to December 2019 in the Department of Cardiovascular Surgery of the General Hospital of the Northern Theater Command.These patients were divided into PMV group(n=107)and non-PMV group(n=576)based on whether duration of prolonged mechanical ventilation was longer than 24 h.The arterial blood lactate at 0 h and 6 h after admission to the ICU of cardiovascular surgery,baseline data and other clinical indicators of the patients were recorded.Preoperative,intraoperative and postoperative risk factors affecting PMV time were analyzed by univariate analysis,and a prediction model was established by indicators selected by logistic regression.Restricted cubic spline model,decision curve analysis(DCA)and receiver operating characteristic(ROC)curve were used to evaluate the predictive value of arterial blood lactate and complex model.Results There were no significant differences between PMV group and non-PMV group in gender,age,BMI,NYHA cardiac function classification,history of myocardial infarction,history of PCI,smoking history,hypertension,diabetes,hypercholesterolemia,>50%stenosis with of 3 coronary artery branches and left aortic stenosis>50%,number of blood vessel bridge,arterial blood lactate at 0 h after admission to ICU,preoperative serum creatinine,hemoglobin after operation,preoperative total bilirubin,preoperative direct bilirubin(P>0.05).There were statistically significant differences in the use of IABP,left ventricular ejection fraction(LVEF),pulmonary arterial pressure,preoperative red blood cell distribution width(RDW),arterial blood lactate at 6 h after admission to ICU,preoperative hemoglobin,preoperative hypersensitive CRP(hs-CRP),postoperative hs-CRP,preoperative troponin T(TNT)and preoperative amino-terminal pro-brain natriuretic peptide(NT-proBNP)(P<0.05).Multivariate logistic regression analysis showed that preoperative RDW,arterial blood lactate 6 h after admission to ICU,pulmonary arterial pressure,preoperative NT-proBNP were risk predictors of PMV,and use of IABP was protective predictor of PMV(P<0.05),OR of five factors were 1.242(95%CI 1.001-1.539),1.370(95%CI 1.171-1.604),1.043(95%CI 1.002-1.087),2.065(95%CI 1.333-3.200),0.146(95%CI 0.071-0.301),respectively(P<0.05).The area under the receiver operating characteristic curve of arterial blood lactate 6 h after admission to ICU and complex model were 0.582(95%CI 0.518-0.646),0.727(95%CI 0.674-0.781),respectively.Intensity of association between Lac 6 h and the development of PMV exhibited a non-linear dose response relationship(P<0.01).Decision curve analysis showed that compared with Lac 6 h,the complex model had a higher net benefit when the threshold probability was between 0.05 and 0.75.Conclusions Compared with Lac 6 h,complex model has a higher predictive value for the occurrence of PMV after OPCABG.