| Objective1.Based on the inflammatory indexes including interleukin,PCT,CRP,NLR,and clinical baseline data,Nomogram was constructed based on the risk factors of septic events,which can be used to evaluate the severity of postoperative traumatic inflammatory reaction in patients with hepatobiliary and pancreatic diseases in real-time and dynamically,to facilitate early and timely intervention and take effective treatment measures.2.Another nomogram model was constructed only based on the relevant indicators of clinical baseline data,which was used to evaluate the daily changes of the severity of traumatic inflammatory response after operation of hepatobiliary and pancreatic diseases,and compared with the model based on inflammatory indicators and clinical data.3.In the radical pancreaticoduodenectomy group as the subgroup,the continuous changes of inflammatory indexes during the perioperative period were observed,and the biomarkers that could reflect the severity of sepsis were found by multivariate regression analysis.to achieve the purpose of early evaluation of postoperative traumatic inflammatory reaction,early detection of pancreatic fistula,and other serious complications,to facilitate early intervention.4.In the hepatectom y group(subgroup),we looked for the risk factors of sepsis after hepatectomy and further verified that cytokines(IL-6,IL-2)and PCT alone or combined to evaluate the severity of postoperative traumatic inflammatory reaction of the hepatobiliary pancreas.MethodThis is an observational case-control study.The clinical data of hepatobiliary and pancreatic diseases in the same treatment group from June 2015 to September 2019 in the Department of Hepatobiliary and Pancreatic surgery of the first Medical Center of PLA General Hospital were collected retrospectively.All patients were prospectively examined by serology at the same time every day on the 1st day before operation and the 1st,3rd,5th,and 7th day after the operation.The detection indexes included cytokines(IL-1,IL-2,IL-6,IL-8,IL-10,and TNF-α),procalcitonin(PCT),C reactive protein(CRP),and neutrophil/lymphocyte ratio(NLR).The data collected included demographic data(sex,age,BMI,basic disease history,ASA grade),perioperative data(preoperative and postoperative AST,ALT,TBIL,DBIL,body temperature,operation time,anesthesia time,blood loss,blood transfusion history,operation method,pathological diagnosis,bacterial microbial culture)and so on.According to the occurrence of sepsis,the sepsis group and the control group were divided into sepsis group and control group,and the diagnostic criteria of sepsis were all Sepsis3.0.All data were processed by SPSS22.0 and R3.6.1software.1.The first part: the data sets of 658 patients who meet the inclusion and exclusion criteria are divided into training sets in chronological order to establish predictive models and verification sets for external verification,that is,522 cases of data sets from June 2015 to January 2017 are training sets,and 136 cases of data sets are validation sets from January 2017 to September 2019.In the training set,univariate Logistic regression analysis was used to screen the related risk factors of sepsis,multivariate Logistic regression analysis was used to determine the independent risk factors of sepsis,and an R software package was used to draw a nomogram diagram based on the results of multifactor Logistic regression to construct the prediction model.The 1000 cycle sampling results of the Bootstrap self-sampling method are used as internal verification set to evaluate the accuracy of the model.By calculating the consistency index(C-index),the closer the value is to 1,the more accurate the prediction ability is,that is,the closer the prediction graph display curve is to the standard curve,the better the prediction ability is.To judge the difference between the predicted probability and the actual probability,the calibration curve(Calibrationcurve)is used to evaluate.The working characteristic curve(ROC)and the area under the curve(AUC)were compared to compare the value of the two prediction models.2.The second part: Taking pancreaticoduodenectomy as a subgroup,using cytokines,PCT,CRP,NLR,and other inflammatory indexes and clinical characteristics before and 1 day after the operation,univariate Logistic regression analysis was performed between the sepsis group and the control group.The difference was statistically significant as a related risk factor,including all relevant risk factors of inflammatory indicators and clinical characteristics.Model A,considering the simplicity and practicability of the model,combined with clinical practice.Models B and C were constructed by univariate regression analysis of different deletions,respectively,and the efficacy of the three models in dynamically evaluating the severity of daily trauma and inflammation after PD was compared.Through the model,it is possible to find complications such as early pancreatic fistula.The third part: a total of 172 cases of hepatocellular carcinoma were diagnosed according to the inclusion criteria.Regression analysis and other statistical methods were used to screen the risk factors of sepsis after hepatectomy.The sepsis group,the S IRS group,and the control group were compared to further verify the value of cytokines and other inflammatory indexes in evaluating the severity of postoperative traumatic inflammation after hepatectomy.Results1.In the first part,a total of 658 cases met the inclusion criteria,522 cases established a predictive model for the training set in the early stage,and 136 cases were validated externally for the verification set in the later stage.The number of 522 cases in the sepsis group and the control group was 55 and 467 respectively,while the 136 cases were concentrated,19 in the sepsis group and117 in the control group.Univariate Logistic regression showed that there were significant differences in IL-2,IL-6,IL-10,PCT,CRP,NLR,history of blood transfusion,history of diabetes,age,operation time,bleeding,and pre-operation.The results of multivariate Logistic regression analysis of clinical features and inflammatory indexes showed that the OR values of blood transfusion history and 95%CI:1.224 history were 3.583 and 2.485,respectively,and the operation time,preoperative DBIL,BMI,PCT,and CRP were also independent risk factors for sepsis.A multivariate Logistic regression model was established in the training cohort,which only included the line chart prediction model established by clinical characteristics.Model A),constructed the line chart prediction model(model B)based on clinical characteristics and inflammatory indicators.The Cutoff values of model A training set and verification set were106 and 106,sensitivity were 0.64 and 0.47,specificity were 0.82 and 0.85,model B training set and verification set Cutoff were 109 and 109,sensitivity were 0.84 and 0.74,specificity were 0.73 and 0.78,model An and model B were0.756 and 0.839 in the training set,respectively.Using the bootstrap method,the C-index of model An is 0.777 and the corrected C-index is 0.761,while the C-index of model B is 0.844 and 0.831.It can be seen that the two Nomogram diagrams show better evaluation efficiency and stability,and model B based on seven indicators of clinical characteristics and inflammatory indicators is better.2.In the second part,a total of 138 patients who underwent pancreaticoduodenectomy met the inclusion criteria,including 31 patients with sepsis(sepsis group)and 107 patients without sepsis(control group).There were significant differences in BMI,operation time,blood transfusion history,and PCT,CRP,before and after IL-6,between the sepsis group and the control group.Univariate Logistic regression analysis showed that 10 factors(BMI,blood transfusion history,operation time,ASA value,diabetes history,IL-2,IL-6,IL-10,PCT,and CRP)were risk factors for sepsis.Model A was obtained by multivariate logistic regression analysis of 10 indexes.The results showed that the values of diabetes 4.06 and IL-6 were 3.31(95%CI:1.36 12.96 and 0.044)respectively.In model B,the operation time,IL-2,IL-10,PCT,and NLR were removed,and the remaining indexes were analyzed by multivariate analysis.the results showed that the value of IL-6 was 4.06(95%CI:1.68 14.44,P < 0.01),which exceeded the critical value of IL-6,and the risk of sepsis was 4.06 times higher than that of the critical value.Model C removed the history of blood transfusion,IL-2,IL-10,and PCT.Results: ASA value,history of diabetes,IL-6,BMI,operation time and NLR were independent risk factors.Comparing model A,model B,and model C,the C index of complete sepsis model A with clinical features and preoperative markers was 0.831(0.753-0.908),and the C index of model B and model C was 0.809(0.732-0.885)and 0.814(0.734-0.894),respectively.The best predictive efficiency is model A.3.In the third part,In the liver resection group,172 cases met the inclusion criteria,including 141 males and 31 females.120 cases in the control group(69.8%),42 cases in the SIRS group(24.4%),and 10 cases of sepsis(5.8%)Comparing the sepsis group with the SIRS group,the percentages of open surgery were 90% and 83%,respectively(P=0.019).The preoperative DBIL sepsis group was significantly different from the S IRS group and the control group(P=0.001),The blood loss was compared between the three groups(P=0.004),There was no significant difference in gender,BMI,AST,ALT,and operation time among the three groups(P>0.05),Univariate and Logistic regression analysis showed that preoperative IL-2 showed the best diagnostic power(AUC=0.80);Postoperative PCT showed the best diagnostic power(AUC=0.85),while the combined model of IL-2,IL-6,and PCT reached(AUC=0.91).In the differential diagnosis of SIRS and sepsis,postoperative PCT showed the best performance(AUC=0.85).Conclusion1.Using clinical baseline data to draw a nomogram model is the most common clinical index,and it is easy to obtain;in the application of the model,the risk of septic events can be evaluated by summing up the corresponding different values of relevant factors,and it is easy to operate.the test results of each patient are different,and the substitution value is also different,which has the characteristics of individualized evaluation of traumatic inflammatory response.The model can be used for trauma evaluation every day after the operation,which is real-time and dynamic and can timely observe the changes of traumatic inflammatory reaction after large hepatobiliary surgery and early intervention to prevent serious adverse complications.The model has certain clinical reference and application value.2.Drawing a nomogram model based on clinical baseline data and inflammatory indicators is superior to drawing a predictive model based on clinical baseline data only in terms of stability and predictive ability.However,due to the need to detect the corresponding cytokines,it is difficult for grass-roots hospitals to carry out,and large hospitals with conditions can be used as a better choice as a model,which tends to be of reference value in the evaluation of traumatic inflammatory response after hepatobiliary and pancreatic surgery.3.The establishment of a complete model based on inflammatory indicators and clinical baseline data has a certain application value in predicting sepsis after pancreaticoduodenal surgery,compared with the line chart prediction model can be used as a backup model.The establishment of two models by randomly deleting some indexes can be used as a reference model.All three models can observe the changes of traumatic inflammatory reaction after pancreaticoduodenal operation in real-time,and it is possible to find complications such as pancreatic fistula,biliary leakage,and anastomotic fistula in the early stage,and effective measures can be taken in time.4.The diagnosis effect of sepsis after PCT liver cancer surgery is the best,However,the postoperative IL-2,IL-6,and PCT combined model has more advantages in the diagnosis of postoperative sepsis;... |
PDF Full Text Request