Accurate and precise measurement of plasma D-dimer levels is important in the diagnosis and management of venous thromboembolism. Considerable variability in D-dimer results obtained using different methods has, however, been reported in multicentre studies. This study explored in two separate multicentre exercises the degree of precision amongst laboratory D-dimer measurements, and the degree by which inter-method agreement could be improved using a calibration curve model. The first exercise demonstrated generally good within-centre precision, with 82% of the centres reporting results for two identical but differently coded samples within 10% of each other. However, six centres reported results which would have excluded deep vein thrombosis (DVT) for one sample but failed to exclude DVT for the other, identical sample. In the second exercise, overall between-method precision of D-dimer results for two samples was shown to improve markedly when a calibration model was applied, using the consensus median values obtained by all participants for three "calibration plasmas" to recalculate D-dimer values. For centres reporting results in fibrinogen equivalent units (FEUs), between-centre coefficients of variation (CVs) fell from 25.9% to 11.6% and 22.4% to 7.7%, respectively, for the two samples. For centres reporting in ng/ml, CVs fell from 45.3-21.6% and 40.8-11.6%, respectively. In conclusion, improved harmonisation of D-dimer results by different methods may be achieved by a calibration model and common calibrant plasmas.