Construction Of Prediction Model For Ovarian Hyperstimulation Syndrome

Objective:Ovarian hyperstimulation syndrome is one of the most serious complications in assisted reproductive technology.At present,its pathogenesis is not fully clear.In clinical practice,prevention is the main focus and treatment is supplemented.This study mainly analyzes the risk factors that affect the occurrence of OHSS in patients with in vitro fertilization and embryo transfer,and constructs its occurrence prediction models at different clinical nodes(before the cycle starts and after the egg retrieval is completed),and Aims to early identify patients with high risk of OHSS and take intervention measures.Methods:A retrospective analysis of patients undergoing IVF/ICSI treatment at the Reproductive Center of Henan Provincial People’s Hospital from January 2017 to June 2020,17408 cycles were screened out according to the inclusion and exclusion criteria,and they were divided into OHSS group(n=4143)and non-OHSS group(n=13265),Compare the general conditions 、 ovulation induction and laboratory indicators between the two groups of patients,and obtain independent risk factors that affect the occurrence of OHSS through multivariate logistic regression analysis,Analyze the effectiveness of predictors based on the receiver operating characteristic curve(ROC),Use Empower Stats software machine learning to analyze the relative importance of each independent risk factor,establish a Norman nomogram model to predict,evaluate the predictive performance of the model based on the area under the curve(AUC).Results:1.Prediction of the occurrence of OHSS before the cycle startsAge,AMH,basal antral follicle(AFC),BMI,polycystic ovary syndrome(PCOS)history,basic FSH,basic LH,basic E2,basic P,and season are independent risk factors for the occurrence of OHSS.The ROC curve results show that AMH and AFC can predict the occurrence of OHSS alone,but their predictive value is moderate.The area under the curve is 0.799 and 0.764,respectively.According to the Youden Index(Yorden Index = Sensitivity + Specificity-1)when the maximum The best threshold values obtained are 3.645ng/ml and 12.5 respectively.The sensitivity is 72.3% and77.7%,and the specificity is 67.0% and 59.0% respectively,The other indicators age,BMI,basic FSH,basic LH,basic E2 and basic P cannot predict OHSS.The machine learning analysis of Empower Stats based on R software shows that AMH is the main factor affecting the occurrence of OHSS,followed by AFC,basic FSH,basic LH,BMI,PCOS history,age,season,basic E2,basic P.Set age,BMI,FSH,AMH,AFC,PCOS history,and season as independent variables to predict the occurrence of OHSS,and establish a nomogram prediction model.The obtained prediction model is: Y=1.33-0.06×age-0.08×BMI-0.17× Basic FSH+0.15×AMH+0.08×AFC+0.30×(PCOS=1)+0.15×(season=1)(AUC=0.811),the sensitivity is 83% at the optimal threshold,and the specificity is 66%.The model The predictive ability is medium.2.Prediction of the occurrence of OHSS after egg retrievalThe number of punctured follicles,the number of eggs obtained,the number of MII,the antagonist regimen,the number of Gn days,the total amount of Gn,and the E2 on HCG day are independent risk factors for the occurrence of OHSS.The ROC curve results show that the number of punctured follicles,the number of eggs obtained,the number of MII,and the E2 on HCG day can individually predict the occurrence of OHSS.The number of punctured follicles has a higher predictive value for the occurrence of OHSS,the area under the curve is 0.906,and the best threshold obtained from the maximum Youden index is 13.5,the sensitivity is 83.3%,the specificity is 81.7%,and other indicators predict The value is medium,the area under the curve is 0.866,0.856,0.819,and the optimal thresholds are 11.5,9.5,and1875.5pg/m L,respectively.The sensitivity is 78.0%,77.9%,73.6%,and the specificity is 72.8,respectively.%,76.6%,75.1%,and the remaining indicators Gn days and total Gn cannot predict the occurrence of OHSS.According to the machine learning analysis of Empower Stats based on R software,the number of punctured follicles is the main factor affecting the occurrence of OHSS,followed by HCG day E2,number of eggs obtained,number of MII,total Gn,number of Gn days,antagonist regimen,follicular phase long-acting long protocol and luteal phase short-acting long protocol.Conclusion:1.Before the cycle starts,AMH and AFC can individually predict the occurrence of OHSS,and the best thresholds are 3.645ng/m L and 12.5,respectively,and the predictive value is moderate.According to the patient’s age,BMI,FSH,AMH,AFC,PCOS,medical history,and season,a nomogram prediction model can be established.The model is stable and the prediction value is higher than that of a single factor.2.After the egg retrieval is complete,the E2 on HCG day,number of punctured follicles,number of eggs harvested,and number of MII on HCG days after ovulation induction can all individually predict the occurrence of OHSS.The best thresholds are 1875.5pg/m L,13.5,11.5,9.5,respectively.The value is higher,and the area under the ROC curve of the number of punctured follicles is 0.906,which can be used as the main predictor.