Plasma biomarkers for Parkinson's disease (PD) diagnosis that carry predictive value for cognitive impairment are valuable. We explored the relationship of Mini-Mental State Examination (MMSE) score with plasma biomarkers in PD patients and compared results to vascular dementia (VaD) and normal controls. The predictive accuracy of an individual biomarker on cognitive impairment was evaluated using area under the receiver operating characteristic curve (AUROC), and multivariate logistic regression was applied to evaluate predictive accuracy of biomarkers on cognitive impairment; 178 subjects (41 PD, 31 VaD and 106 normal controls) were included. In multiple linear regression analysis of PD patients, α-synuclein, anti-α-synuclein, α-synuclein/Aβ40 and anti-α-synuclein/Aβ40 were highly predictive of MMSE score in both full model and parsimonious model (R2 = 0.838 and 0.835, respectively) compared to non-significant results in VaD group (R2 = 0.149) and in normal controls (R2 = 0.056). Α-synuclein and anti-α-synuclein/Aβ40 were positively associated with MMSE score, and anti-α-synuclein, α-synuclein/Aβ40 were negatively associated with the MMSE score among PD patients (all Ps < 0.005). In the AUROC analysis, anti-α-synuclein (AUROC = 0.788) and anti-α-synuclein/Aβ40 (AUROC = 0.749) were significant individual predictors of cognitive impairment. In multivariate logistic regression, full model of combined biomarkers showed high accuracy in predicting cognitive impairment (AUROC = 0.890; 95%CI 0.796-0.984) for PD versus controls, as was parsimonious model (AUROC = 0.866; 95%CI 0.764-0.968). In conclusion, simple combination of biomarkers inclusive of α-synuclein/Aβ40 strongly correlates with MMSE score in PD patients versus controls and is highly predictive of cognitive impairment.