以多元回歸分析進行腎腫瘤患者接受單側腎切除後的術後腎絲球過濾率預估

許瓊文¹、陳階曉^1,2

¹中國醫藥大學北港附設醫院泌尿科

²中國醫藥大學生物醫學工程系

Estimating Postoperative Glomerular Filtration Rate after Unilateral Nephrectomy in Renal Tumor Patients Using Multiple Regression Analysis

Chiung-Wen Hsu¹, Chieh-Hsiao Chen^1,2

¹Department of Urology, China Medical University Beigang Hospital

²Department of Biomedical Engineering, China Medical University

Purpose:

This study aims to develop an accurate predictive model for postoperative renal function following nephrectomy. Given the vital role of the kidney in excretion and metabolism, predicting the function of the remaining kidney is crucial for ensuring patient quality of life and reducing complications. While previous research has focused on general trends, there remains a lack of individualized predictive models for clinical use.

Materials and Methods:

The study analyzed data from 192 patients who underwent nephrectomy at China Medical University Hospital between 2011 and 2018. Predictive variables included age, sex, BMI, comorbidities, preoperative estimated glomerular filtration rate (eGFR), and nuclear medicine effective renal plasma flow (ERPF). The research employed both multiple regression analysis and artificial neural network (ANN) techniques to compare their predictive precision and clinical applicability.

Results:

Multiple regression analysis identified preoperative eGFR, contralateral and ipsilateral ERPF, bladder disease, ASA score, and cerebrovascular accident (CVA) as significant predictors. A simplified bivariate model was developed to enhance clinical utility, yielding the formula:

Post-operative eGFR = 0.85 × preserved side preoperative eGFR + 0.21 × preoperative eGFR.

Compared to the ANN model, the multiple regression model demonstrated superior performance with this sample size, offering higher stability, interpretability, and less risk of overfitting.

Conclusion:

The developed model, particularly emphasizing the preserved side's preoperative eGFR, provides a reliable and scientifically grounded tool for predicting renal function after unilateral nephrectomy. Although limited by its single-center data and sample size, the model shows significant potential to help clinicians formulate treatment plans and improve patient outcomes. Future multi-center studies are recommended to further validate these findings.