Abstract
OBJECTIVE: To identify independent risk factors for Parkinson disease mild cognitive impairment (PD-MCI) and develop a prediction model integrating clinical indicators, blood biomarker, and neuroimaging data, aiding in detection and intervention. METHODS: A retrospective study was conducted with 150 PD patients. The PD-MCI group (n = 64) and PD with normal cognition (PD-NC, n = 86) were identified using the Montreal Cognitive Assessment scale. Data on demographics, motor symptoms, cognitive function, quality of life, blood markers, and diffusion tensor imaging along perivascular spaces (DTI-ALPS) were collected. Univariate analysis identified significant variables, and multivariate logistic regression identified independent risk factors. A nomogram prediction model was developed using R software. Model performance was evaluated using Receiver Operating Characteristic (ROC) curves, bootstrap resampling calibration curves, and decision curve analysis (DCA). RESULTS: Significant differences between the groups were found in levodopa equivalent daily dose (LEDD), PD Quality of Life Questionnaire, creatinine, cystatin C, and ALPS index. Multivariate regression identified higher LEDD (OR = 1.01, 95%CI 1.00-1.03, p = 0.005) and creatinine levels (OR = 1.34, 95%CI 1.10-1.66, p = 0.005) as independent risk factors. The nomogram model demonstrated strong discriminatory ability (AUC = 0.864, 95%CI 0.807-0.922) and good calibration. DCA showed a significant net benefit within clinical threshold ranges. CONCLUSION: This study developed a PD-MCI prediction model incorporating DTI-ALPS and clinical blood biomarkers. It confirmed that LEDD and creatinine levels are independent risk factors, with high clinical value for early screening and individualized treatment.