Development,Validation And Application Of Risk Predictive Models For Stress Urinary Incontinence During And After Pregnancy

Background Early prevention of stress urinary incontinence(SUI)during and after pregnancy is crucial.To date,no feasible tool has been found to be useful in identifying high-risk women with SUI during and after pregnancy.Objectives 1.To develop risk predictive models for SUI during and after pregnancy;2.To investigate the implementation and preventive effect of antenatal and postnatal PFMT among the high-risk women of SUI.Methods 1,The study selected 521 pregnant women at their first trimester(9-12 weeks of gestation)or second trimester(20-23 weeks of gestation),who underwent pelvic floor three-dimensional ultrasonography from July 2016 to February 2017 in obstetrics and gynecology of the two hospitals in Shenzhen.The general information,pregnancy-related information and pelvic ultrasound data of pregnant women were collected.The Chi-square test and two-sided Student's t-test were applied to examine the distribution differences of variables between the training dataset and the validation dataset.All significant continuous variables(P<0.05 at univariable analysis)in the training dataset were analyzed by receiver-operating characteristics(ROC).Three multivariable logistic regressions with backward elimination were built:the first on significant clinical variables(model A);the second on significant clinical and two-dimensional(2D)ultrasound variables(P<0.05 at ROC analysis for continuous parameters,and bladder neck(BN)funneling if P<0.05 at univariable analysis)(model B);and the third on significant clinical and 2D and 3D ultrasound parameters(model C).Areas under the curve(AUC)of the three models were calculated and compared using Medcalc software for Windows.The final predictive model with best performance and fewest variables was selected to establish a nomogram using R for Windows with the rms package.The developed nomogram was evaluated by both discrimination measured by the C-:index and calibration evaluated by the Hosmer-Lemeshow test and calibration plot,which compares the prediction probability and the actual observation probability.A bootstrap technique with 1,000 resamplings was used for internal validation to adjust overfitting and overoptimistic model performance.External validation of the nomograms was performed on the independent validation dataset.2.One month after delivery,all women were followed up by phone.Obstetrics-related information,pelvic floor muscle training during pregnancy and the occurrence of postpartum SUI were investigated.Univariable and multivariable analyses were conducted to analyze whether antenatal PFMT had a preventive effect on SUI at 1 month postpartum for high-risk women based on the risk predictive nomogram for SUI during pregnancy.3.The way of establishment and evaluation of the nomogram for SUI postpartum is the same as SUI during pregnancy.4.Three month after delivery,high-risk women based on the risk predictive nomogram for SUI after pregnancy were followed up by phone.Univariable and multivariable analyses were conducted to analyze whether postnatal PFMT had a preventive effect on SUI at 3 month postpartum:Results1.Development and validation of risk predictive nomogram for SUI during pregnancyThe overall incidence of SUI during pregnancy was 35.1%.ROC curve analysis showed that BMI at inclusive,BMI gain since pregnancy,and six ultrasound parameters,namely the beta angle and rest and Valsalva,the yangle at Valsalva and the HA(rest,Valsalva and AHA)had predictive value for SUI during pregnancy.The clinical model A,including pregnancy BMI gain,constipation,and previous delivery mode,had a C-statistic of 0.709(95%CI,0.658-0.757).Model B,including pregnancy BMI gain,constipation,previous delivery mode,? angle at rest,and BN funneling,had a C-statistic of 0.789(95%CI,0.742-0.831).The C-statistic of model C,including pregnancy BMI gain,constipation,previous delivery mode,? angle at rest,BN funneling,and HA at VM,was 0.813(95%CI,0.768-0.853).Three nomograms were developed using R software based on the three prediction models.The three nomograms showed good discrimination and satisfactory calibration by Internal and external verification.2.Application of risk predictive nomogram for SUI during pregnancyThe incidence of SUI during pregnancy in high-risk women was 51.9%(125/241),and in low-risk pregnant women was 20.1%(54/269).The preventive effect of antenatal PFMT for SUI at 1 month postpartum has not been observed(P=0.390).3.Development and validation of risk predictive nomogram for SUI after pregnancyThe incidence of postpartum SUI was 14.2%.Only the HA at Valsalva had predictive value on postpartum SUI with AUC of 0.592(95%CI,0.546 to 0.638).The final risk prediction model includes pre-pregnancy BMI,SUI during pregnancy,abortion history and vaginal delivery.The nomogram for SUI postpartum showed good discrimination and satisfactory calibration by internal verification.4.Application of risk predictive nomogram for SUI after pregnancyThe incidence of SUI at 1 month postpartum in high-risk women was 28.8%(53/184),and in low-risk women was 4.4%(12/274).The preventive effect of postnatal PFMT for SUI at 3 months postpartum has been observed after deducting the effect of confounding factor(P=0.018).Conclusions1.According to different clinical conditions,this study developed a seris of nomograms for the prediction of SUI during and after pregnancy,which are expected to become important predictive tools,as well as the basis for individualized prevention program implementation.2.High-risk pregnant women of SUI did not benefit from the antenatal PFMT,while the postnatal PFMT had a preventive effect for high-risk women of SUI at 3 months postpartum.Further clinical prospective investigations are necessary to verify the conclusions of this study.