To determine whether apparent involvement of DYRK1A in Alzheimer's disease (AD) pathology makes it a candidate plasma biomarker for diagnosis, we developed a method to quantify plasma DYRK1A by immunoblot in transgenic mouse models having different gene dosages of Dyrk1a, and, consequently, different relative protein expression. Then, we measured plasma DYRK1A levels in 26 patients with biologically confirmed AD and 25 controls (negative amyloid imaging available on 13). DYRK1A was detected in transgenic mouse brain and plasma samples, and relative levels of DYRK1A correlated with the gene copy number. In plasma from AD patients, DYRK1A levels were significantly lower compared with controls (P<0.0001). Results were similar when we compared AD patients with the subgroup of controls confirmed by negative amyloid imaging. In a subgroup of patients with early AD (CDR=0.5), lower DYRK1A expression was confirmed. In contrast, no difference was found in levels of DYRK1B, the closest relative of DYRK1A, between AD patients and controls. Further, AD patients exhibited a positive correlation between plasma DYRK1A levels and cerebrospinal fluid tau and phosphorylated-tau proteins, but no correlation with amyloid-β42 levels and Pittsburgh compound B cortical binding. DYRK1A levels detected in lymphoblastoid cell lines from AD patients were also lower when compared with cells from age-matched controls. These findings suggest that reduced DYRK1A expression might be a novel plasma risk factor for AD.