Development and Validation of a Prostate Cancer Risk Prediction Model for the Elevated PSA Population

Junhui Wu¹Xiaodong Jin¹Jiali Li¹Lingqian Zhao²Chenkai Zhao¹Nengfeng Yu¹Yubing Li¹Jiasheng Yan¹Junlong Wang¹Fei Yang¹Wenhao Zhang^1*

• ¹First Affiliated Hospital, Zhejiang Chinese Medical University, Hangzhou, Zhejiang Province, China
• ²Zhejiang Provincial People's Hospital, Hangzhou, Zhejiang Province, China

Introduction: To develop and validate a dynamic clinical prediction model integrating prostate-specific antigen (PSA) and peripheral blood biomarkers for distinguishing benign from malignant prostate diseases in patients with elevated PSA levels. Methods: A retrospective study was conducted of clinicopathological data and preoperative blood specimen information of patients who underwent ultrasound-guided prostate biopsy in The First Affiliated Hospital of Zhejiang Chinese Medical University due to elevated PSA between January 2018 and November 2024.Univariate analysis, Least Absolute Shrinkage and Selection Operator regression, and multifactorial logistic regression analysis were utilized to identify independent risk factors associated with benign or malignant prostate disease in patients with elevated PSA (PSA > 4.0ng/ml). The construction of a clinical prediction model was then undertaken, with the subsequent calibration and integration into a network calculator. Results: A total of 529 patients were included based on predefined inclusion and exclusion criteria, comprising 268 (50.7%) with benign pathology and 261 (49.3%) with malignancy. After analysis, independent risk factors associated with benign or malignant prostatic diseases in patients with elevated PSA levels were identified, including PSA, white blood cell, neutrophil-to-lymphocyte ratio, lymphocyte-to-monocyte ratio, eosinophil count, basophil count, and serum albumin. Utilizing these independent risk factors, a clinical prediction model for the risk of PSA-elevated prostate benign-malignant disease was constructed, yielding an area under the curve of 0.906, a predictive model specificity of 77.6%, and a sensitivity of 95%. The calibration curve and clinical decision curve indicated that the model exhibited superior calibration ability. A dynamic prediction model was formulated based on the clinical prediction model integrated into a network calculator. Conclusion: This study establishes a non-invasive prediction model integrating PSA and peripheral blood biomarkers, providing a clinically practical tool for risk stratification in patients with elevated PSA levels.