Biomarkers of neurological injury can potentially predict postoperative cognitive dysfunction. We aimed to identify whether classical neuronal damage-specific biomarkers, including brain fatty acid-binding protein, neuron-specific enolase and S100 calcium-binding protein β, as well as plasma-free haemoglobin concentration as a measure of haemolysis, could be used to predict the risk of long-term cognitive decline after coronary artery bypass grafting with or without cardiopulmonary bypass. We assessed cognitive function using the CogState brief computerised cognitive test battery at 3 months and at 15 months after surgery. Blood samples were obtained pre-operatively, after sternal closure, and at 6 h and 24 h postoperatively. We found signs of cognitive decline at 3 months in 15 of 57 patients (26%), and in 13 of 48 patients (27%) at 15 months. Brain fatty acid-binding protein was already significantly higher before surgery in patients with postoperative cognitive dysfunction at 15 months, with pre-operative plasma levels of 22.8 (8.3-33.0 [0-44.6]) pg.ml-1 compared with 9.7 (3.9-17.3 [0-49.0]) pg.ml-1 in patients without cognitive dysfunction. This biomarker remained significantly higher in patients with cognitive decline throughout the entire postoperative period. At 3 months after surgery, high levels of plasma-free haemoglobin at sternal closure were associated with a negative influence on cognitive performance, as were high baseline scores on neuropsychological tests, whereas a higher level of education proved to beneficially influence cognitive outcome. We found that postoperative cognitive dysfunction at 3 months was associated with cognitive decline at 15 months after surgery, and served as a valuable prognostic factor for declines in individual cognitive performance one year later. Classical neuronal injury-related biomarkers were of no clear prognostic value.