Development And Internal Validation Of A Predictive Model For The Risk Of Complications After Holmium Laser Transurethral Resection Of The Prostate

Objective:Transurethral Holmium laser prostate enucleation is currently the most common procedure for the treatment of benign prostatic hyperplasia(BPH),but its popularity is limited by its long learning curve and the high number of postoperative complications.Therefore,this study investigates the risk factors for complications after holmium laser enucleation of the prostate,establishes a prediction model based on this,and visualizes and evaluates the established model with line graphs to provide scientific treatment guidance for medical practitioners and patients.Methods: 345 patients with BPH(Benign prostatic hyperplasia)treated with Ho LEP(holmium laser enucleation of the prostate)in Jiangsu Subei People’s Hospital from January 2019 to August 2022 were selected for the study.The age,BMI(body mass index),PV(prostate volume),IPSS(international prostate symptom score),and QoL(quality of life scale)were collected from all study participants.The patients were retrospectively analyzed according to their preoperative clinical data such as body mass index,PV(prostate volume),IPSS(international prostate symptom score),QoL(quality of life scale),PSA(prostate-specific antigen),and whether they had any pre-operative underlying disease as predictors and were followed up by telephone or outpatient visits for 3 months after surgery as outcome variables.The group was divided into a complication group and an uncomplicated group according to whether complications occurred after surgery.Histograms were first plotted for each predictor,and predictors with apparently skewed distributions were simply transformed by taking the logarithm(log or In)to normalize the skewed information and equalize the variance.Next,a spline function between each predictor and the outcome variable was plotted.For nonmonotonic predictors,they were transformed into grouped variables based on clinical significance,spline function,or optimal threshold of the ROC curve,and both of these methods were used to reduce the influence of each variable on the model to increase the stability of the model.The transformed predictors were all included and the final predictors affecting postoperative complications in Ho LEP were screened using lasso regression,on which the model was built.The GVIF(Generalized variance inflation factor)and Cook’s distance of the constructed model were calculated to determine the presence of multicollinearity and strong influence points.A line graph of the occurrence of complications after Ho LEP was plotted for visualization.Bootstrap was used for internal validation,plotting the calibration curve of the model and calculating the C-index.The accuracy and discrimination of the model were further validated by plotting the clinical decision curves of the model.Results : A total of 345 patients were eventually included,with the following complications: 24.06%(83/345)for postoperative incontinence,0.87%(3/345)for severe perioperative anemia requiring blood transfusion,5.51%(19/345)for postoperative urinary tract infection,3.48%(12/345)for postoperative haematuria The incidence of postoperative urethral stricture requiring urethral extension was 2.03%(7/345),postoperative acute urinary retention requiring reintroduction of a urinary catheter was6.09%(21/345),postoperative epididymitis was 1.16%(4/345),bladder neck contracture was 3.77%(13/345),and some patients had two or more complications at the same time.A total of 112 patients ended up with postoperative complications and 233 without any complications,giving a complication rate of 32.46%.The preoperative influencing factors of the study subjects were all included in the lasso regression,and the penalty coefficient(Lambda)was calculated.A total of six predictors including International Prostate Symptom Score(IPSS),Quality of Life Score(QoL),body mass index(BMI),whether or not they had diabetes before surgery,prostate volume(PV),and prostate-specific antigen(PSA)were finally screened,and a prediction model was constructed based on the above The prediction model was constructed based on these predictors.The model was evaluated: the area under the ROC curve(AUC)of the prediction model was 0.7901 with a 95% confidence interval of 0.702-0.827.Bootstrap internal validation showed that the C-index was 0.7902 and the calibration model curve overlapped well with the actual model curve.The results of the clinical decision curve analysis of the model showed that the probability of complications ranged from 5% to 85%,possessing a high accuracy and net benefit.Conclusions:The preoperative International Prostate Symptom Score(IPSS),Quality of Life Score(QoL),body mass index(BMI),prostate volume(PV),prostate-specific antigen(PSA),and the presence of preoperative diabetes were found to be predictors of the probability of postoperative complications after Ho LEP,and the model was diagnosed and evaluated with no multicollinearity and no strong influence points.The internal validation of Bootstrap proved that the accuracy,discrimination,and net benefit of this model are high,which can help to provide scientific guidance to clinicians and pat.