Establishment Of Predictive Factors And Preoperatively Predictive Nomogram Model For Perioperative Complications In Endoscopic Transsphenoidal Pituitary Adenoma Surgery

Background: Pituitary adenomas(PAs)are the most common type of neoplasms in sellar region and comprise about 15% of primary intracranial tumors.Endoscopic transsphenoidal pituitary adenoma surgery(eTSS)is the first-line treatment for patients with clear indication of surgery.Although eTSS has advantages including high tumor resection rate and low incidences of complications,there are some potential complications in eTSS(e.g.intraoperative cerebrospinal fluid leakage,postoperative delayed hyponatremia,and postoperative early polyuria).Occurrence of these complications will decrease the medical quality,sharpen the pain,and even cause death.Preoperative prediction of these complication has significant application value in adjusting operation strategy,postoperative management,and follow-up plan.However,there are limited researches about the predictors for these complications and lack of clinical predictive models for them.Objective: The current research aims to explore the predictors of intraoperative cerebrospinal fluid leakage,postoperative delayed hyponatremia,and postoperative early polyuria in eTSS for PAs,and further develop and validate predictive nomograms for these three complications respectively.Methods: Consecutive patients,who underwent eTSS for PAs between January 2018 and October 2020 at the Department of Neurosurgery in Jinling Hospital,were included.Data was collected,including medical record,laboratory test,and radiology image.For postoperative delayed hyponatremia,external validation dataset was obtained from the First Affiliated Hospital of Nanjing Medical University.For the other two complications,random division method was applied to produce training and validation datasets.In the training dataset,univariable analysis,univariable logistic regression analysis,and LASSO regression analysis were utilized to filter potential predictors.Selected variables were included in multivariable logistic regression analysis to identify independent predictive factors.Nomogram models were constructed based on these predictors,and evaluated on discrimination,calibration,and clinical benefit.Results: Tumor height(OR = 1.11,P < 0.001)and albumin(OR = 0.87,P = 0.048)are independent predictors for intraoperative cerebrospinal fluid leakage.The AUCs of nomogram predicting intraoperative cerebrospinal fluid leakage are 0.733,0.643,and 0.644 in training dataset,validation dataset 1 and 2,respectively.Hyponatremia on post-operative day1–2(OR = 2.64,P = 0.033),monocyte percentage(OR = 1.22,P = 0.047),and preoperative prothrombin time(OR = 1.78,P = 0.008)are independent predictors for postoperative delayed hyponatremia.The AUCs of the nomogram can reach 0.688 in training dataset and0.617 in external validation dataset.Preoperative IGF-1(OR = 1.001,P = 0.049)and fibrinogen(OR = 1.90,P = 0.035)are independent predictors for postoperative early polyuria.The AUCs of this nomogram are 0.667 in training dataset and 0.704 in validation dataset.Good calibrate plots and decision curves for these three models are computed.Further subgroup analyses show robust predictive performance of these nomograms in some subgroups including clinical subtypes.Conclusion: Novel independent predictors for intraoperative cerebrospinal fluid leakage,postoperative delayed hyponatremia,and postoperative early polyuria in eTSS were identified in this research.Three well-performed nomograms were constructed and validated,from which patients can benefit with optimization of treatment decisions.