Construction Of Individual Clinical Predictive Model For Lymph Node Metastasis Of Rectal Cancer Based On Clinical Data

Objective: To explore the related factors of lymph node metastasis of rectal cancer,and to construct a predictive model for preoperative diagnosis of lymph node metastasis of rectal cancer based on clinical data.Methods: The data of patients with rectal malignant tumors diagnosed by colorectal anal surgery in the first affiliated Hospital of Guangxi Medical University from September 1,2018 to August 31,2019 are analyzed retrospectively.All patients are pathologically proved to be rectal cancer and undergo rectal tumor resection.The preoperative serological test results,the report of CT enhanced imaging of abdomen and pelvis and the postoperative pathological results including lymph node analysis are intact,and the patient have not received any non-operative treatments for rectal cancer,including neoadjuvant chemotherapy,radiotherapy alone,combined radiotherapy and chemotherapy,targeted drug therapy or traditional Chinese medicine therapy.Finally,the clinical data of a total of 201 patients with rectal cancer are collected,and coded by random number method in SPSS25.0,and randomly assigned to verification group and training group according to the proportion of 3:7.There are59 patients in verification group and 142 patients in training group.First of all,the clinical data of 142 patients in the training group are analyzed,according to the data type and distribution,the appropriate univariate statistical method is selected,and then multi-factorial analysis is conducted according to the results of univariate analysis,so as the regression equation is established based on the relevant factors screened out by the multi-factorial analysis.According to the factors related to colorectal cancer lymph node metastasis screened out by multivariate analysis,the "glmnet" software package is used in R4.0.2 software to draw the nomogram,so as to finally build the individualized clinical prediction model of colorectal cancer lymph node metastasis.The consistency index(CINDEX)of the prediction model is calculated by using the "p ROC" software package,the subject working curve(ROC curve)is drawn,the area under the surface is calculated,and the calibration curve is drawn for internal verification,and the above prediction model is verified again in the verification group,so as to further demonstrate the accuracy and effectiveness of the model.The DCA decision curve and clinical impact curve which draw by "rmda" package are used to evaluate clinical benefit of this new predictive model.In addition,we also evaluate the clinical practicability of the model by comparing with enhanced CT in the diagnosis of metastatic lymph nodes.Results: Among the 201 patients,85 patients have lymph node metastasis,with an incidence of 42.3%.In the 142 patients in the training group,60 patients had lymph node metastasis,with an incidence of 42.3%.In the verification group,25 patients had lymph node metastasis,with an incidence of 41.7%.Univariate analysis shows that LD(lactate dehydrogenase),PT(prothrombin time),INR(international standardized ratio),CEA(carcinoembryonic antigen),CA242(carbohydrate antigen 242),CA724(carbohydrate antigen 724)and tumor differentiation are related to lymph node metastasis of rectal cancer.Logistic regression analysis shows that the degree of tumor differentiation,CEA(positive CEA),PT(prothrombin time)and LD(lactate dehydrogenase)are included in the regression equation for lymph node metastasis of rectal cancer.In R4.0.2 software,the line map is established with the selected independent related factors,and verified by the internal verification method of the training group,the results shows that the clinical prediction ability of the line map is good.In the internal verification method,the area under the ROC surface of the prediction model is0.732,the sensitivity is 0.671,the specificity is 0.733,and the calibration curve shows that the prediction model fits well;in the verification group,the area under the ROC surface is 0.728,the sensitivity is 0.824,and the specificity is 0.640,the result of the verification group is similar to that of the training group,and the calibration curve of the verification group also confirms that the model fits well.The constructed prediction model is used to draw the decision-making curve and clinical influence curve,and the results show that the prediction model can provide some benefits for the clinic.In both groups,the accuracy of enhanced CT in the diagnosis of rectal cancer lymph node metastasis was lower,and the diagnostic efficiency was not as good as that of the new predictive model.Conclusion: In this study,a new prediction model of lymph node metastasis of rectal cancer was established based on clinical data.The diagnostic performance of rectal cancer with or without lymph node metastasis is better than enhanced CT.