Plasma D-dimer, a highly sensitive marker of venous thromboembolism, was measured using an enzyme-linked immunosorbent assay pre-operatively, on the third postoperative day, and at the time of phlebography in 118 patients undergoing elective total knee arthroplasty. Deep venous thrombosis (DVT) was detected using systematic bilateral phlebography between the eighth and 12th postoperative day in 47 (39.8%) patients. D-dimer plasma concentrations did not differ between patients who had DVT and those who had no DVT, either pre-operatively (n = 118, P = 0.63) or at the time of phlebography (n = 111, P = 0.70). On the third postoperative day, D-dimer concentration was significantly higher (P < 0.01) in the patients who had DVT (median 3270 μg/l, range 1156–9996, n = 47) than in those who did not (2287 μg/l, 685–7062, n = 64). However, receiver operating characteristics curve analysis did not provide any useful cutoff values that would allow individual diagnoses to be made. In conclusion, the results of the present study suggest that plasma measurement of D-dimer concentration is of no value for predicting, diagnosing or ruling out DVT in patients undergoing total knee arthroplasty.