The prevalence of Parkinson's disease (PD) increases with age. Up to 50% of PD show cognitive decline in terms of a mild cognitive impairment already in early stages that predict the development of dementia, which can occur in up to 80% of PD patients over the long term, called Parkinson's disease dementia (PDD). So far, diagnosis of PD/PDD is made according to clinical and neuropsychological examinations while laboratory data is only used for exclusion of other diseases. The aim of this study was the identification of possible biomarkers in cerebrospinal fluid (CSF) of PD, PDD and controls (CON) which predict the development of dementia in PD. For this, a proteomic approach optimized for CSF was performed using 18 clinically well characterized patients in a first step with subsequent validation using 84 patients. Here, we detected differentially sialylated isoforms of Serpin A1 as marker for differentiation of PD versus PDD in CSF. Performing 2D-immunoblots, all PDD patients could be identified correctly (sensitivity 100%). Ten out of 24 PD patients showed Serpin A1 isoforms in a similar pattern like PDD, indicating a specificity of 58% for the test-procedure. In control samples, no additional isoform was detected. On the basis of these results, we conclude that differentially sialylated products of Serpin A1 are an interesting biomarker to indicate the development of a dementia during the course of PD.