Efficacy of machine learning models versus Cox regression model for predicting prognosis of esophagogastric junction adenocarcinoma / 南方医科大学学报

OBJECTIVE@#To compare the performance of machine learning models and traditional Cox regression model in predicting postoperative outcomes of patients with esophagogastric junction adenocarcinoma (AEG).@*METHODS@#This study was conducted among 203 AEG patients with complete clinical and follow-up data, who were treated in our hospital between September, 2015 and October, 2020. The clinicopathological data of the patients were processed for analysis using R language package and divided into training and validation datasets at the ratio of 3:1. The Cox proportional hazards regression model and 4 machine learning models were constructed for analyzing the datasets. ROC curves, calibration curves and clinical decision curves (DCA) were plotted. Internal validation of the machine learning models was performed to assess their predictive efficacy. The predictive performance of each model was evaluated by calculating the area under the curve (AUC), and the model fitting was assessed using the calibration curve.@*RESULTS@#For predicting 3-year survival based on the validation dataset, the AUC was 0.870 for Cox proportional hazard regression model, 0.901 for eXtreme Gradient Boosting (XGBoost), 0.791 for random forest, 0.832 for support vector machine, and 0.725 for multilayer perceptron; For predicting 5-year survival, the AUCs of these models were 0.915, 0.916, 0.758, 0.905, and 0.737, respectively. For internal validation, the AUCs of the 4 machine learning models decreased in the order of XGBoost (0.818), random forest (0.758), support vector machine (0.0.804), and multilayer perceptron (0.745).@*CONCLUSION@#The machine learning models show better predictive efficacy for survival outcomes of patients with AEG than Cox proportional hazard regression model, especially when proportional odds assumption or linear regression models are not applicable. XGBoost models have better performance than the other machine learning models, and the multi-layer perception model may have poor fitting results for a limited data volume.

Influencing Factors of HIV/ AIDS Death Based on Bayesian Cox Proportional Hazard Regression Model / 中山大学学报(医学科学版)

ObjectiveTo understand the composition of related characteristics of HIV/AIDS cases in Lanzhou and analyze the influencing factors of AIDS-related deaths. MethodsThe information of HIV/AIDS cases reported in Lanzhou from 2011 to 2019 was collected， the method of survival was used analysis and Bayesian Cox Proportional Hazard Regression Model was constructed to analyze the related factors of death. ResultsA total of 2 312 HIV/AIDS patients were selected in this study， including 45 AIDS-related deaths. The results of multivariate regression showed that the older the patients were， the higher the risk of death was； the risk of death of AIDS patients at the time of diagnosis was 13.91 times higher than that of HIV-infected patients； Patients who received CD4 testing had a lower risk of death than those who did not； The risk of death was 0.22 times higher among those who received antiretroviral therapy than those who did not receive antiretroviral therapy. ConclusionsAge at diagnosis， course of disease， antiviral therapy were the influencing factors of AIDS-related death in HIV/AIDS patients in Lanzhou. Therefore， it is necessary to strengthen health education for AIDS-related groups， advocate early detection， early diagnosis， and early treatment， expand the coverage of AIDS testing and treatment， prolong the survival time of AIDS patients.

Comparison of two prediction models for mortality ofacute paraquat poisoning / 预防医学

Objective@#To compare the effects of Cox proportional hazard regression model (Cox model) and extreme gradient boosting model ( XGBoost model ) on the prediction of the mortality of acute paraquat poisoning (APP).@*Methods@#The APP cases admitted to Qingdao Eighth People's Hospital and Shandong Provincial Hospital from January 1st of 2018 to December 1st of 2020 was recruited and divided into a training group and a verification group by a random number table. The Cox model and XGBoost model were established to select the predictors for APP mortality. Receiver operating characteristic ( ROC ) curve was drawn to analyze the predictive power of the two models, and the calibration was evaluated using Hosmer-Lemeshow test.@*Results@#Totally 150 APP cases were recruited. There were 75 cases each in the training group and in the verification group, with 52 and 55 cases died respectively, accounting for 69.33% and 73.33%. The Cox model showed that paraquat intake, the time from taking poison to seeing a doctor, the time for the first perfusion, the time for the first vomiting, aspartate aminotransferase, alanine aminotransferase, serum creatinine, blood urea nitrogen, white blood cell, lactic acid, creatine kinase isoenzymes, glucose, serum calcium and serum potassium were the predictors of APP mortality ( all P<0.05 ). The XGboost model showed that the predictive power of the factors in a descending order were the time from taking poison to seeing a doctor, the time for the first vomiting, the time for the first perfusion, lactic acid, white blood cell, paraquat intake, serum creatinine, serum potassium, serum calcium, creatine kinase isoenzymes, glucose, aspartate aminotransferase, blood urea nitrogen and alanine aminotransferase. The area under curve ( AUC ) of the XGBoost model for predicting was 0.972, which was greater than 0.921 of the Cox model ( P<0.05 ). The predicted results of the Cox model and XGBoost model were consistent with the actual situation ( P>0.05 ). @*Conclusion@#The Cox model and XGBoost model are consistent in predicting the mortality of APP, but the latter is better.

Expression and clinical significance of long non-coding RNA HOTTIP in tissues of patients with endometrial carcinoma / 中国肿瘤生物治疗杂志

@#[Abstract] Objective: To investigate the expression and clinical significance of long non-coding RNA (lncRNA) HOTTIP in tissues of patients with endometrial carcinoma. Methods: A total of 109 cases of patients with endometrial carcinoma who underwent surgery in Xingxiang Central Hospital from April 2012 to April 2014 were selected. The endometrial carcinoma tissue and its corresponding adjacent tissue (more than 5 cm from the cancer margin) were obtained. The expressions of HOTTIP in endometrial carcinoma and adjacent tissues were detected by qRT-PCR. All patients were followed up from the first postoperative day. The follow-up deadline was April 30, 2019. The end-point event was death and the patient's survival time was recorded. Results: The relative expression level of HOTTIP in endometrial carcinoma tissues was (2.55±0.21), which was higher than that in the adjacent tissue (1.03±0.16) (t=60.631, P<0.01). The differences of the relative expression levels of HOTTIP in endometrial carcinoma tissues between different FIGO stage, histological grade, depth of myometrial invasion, lymphatic vascular infiltration status and lymph node metastasis were statistically significant (all P<0.05). Kaplan-Meier survival analysis showed that the 5-year survival rate and the survival time in the low expression group were higher than those in the high expression group [78.57% vs 37.04%, (70.67±4.94) months vs (42.14±3.65) months], the difference was statistically significant (χ2=12.839, P<0.01). Cox proportional hazards regression model analysis showed that the FIGO stage [HR=2.248 (95%CI: 1.034-4.887)], myometrial invasion depth [HR=3.055 (95%CI: 1.668-5.592)], lymph node metastasis [HR=3.811 (95%CI: 1.786-8.131)] and the expression of HOTTIP [HR=2.649 (95%CI: 1.026-6.842)] were all independent influence factors for the prognosis of patients with endometrial carcinoma. Conclusion: lncRNA HOTTIP is highly expressed in endometrial carcinoma tissues and associated with malignant progression of patients. It is an independent influencing factor for patients’ prognosis.