Abstract In 100 consecutive patients with clinically suspected deep vein thrombosis (DVT) of the legs, plasma D-dimer measurements based on an enzyme linked immunosorbent assay (ELISA) and on latex agglutination (Diagnostica Stago) were compared to the results of real time B mode ultrasound imaging combined with Doppler examination, which in a previous study has proved to be a very accurate method competitive with venography for the diagnosis of DVT. Forty five patients had DVT identified with the ultrasonic tests. We have obtained for ELISA and latex tests of D-dimer respectively: accuracy: 60%, 56%; sensitivity: 98%, 98%; specificity: 29%, 22%; predictive value of a positive test: 53%, 50% and predictive value of a negative test: 94%, 92%. These results confirmed those of previous studies using ELISA or latex assays, with venography as a reference test. We conclude that a negative D-dimer test, defined by a value lower than 0.5 μg/ml, excludes the diagnosis of DVT in 94% of cases by ELISA method and in 92% of cases by latex method. A reduction of venography or other objective testing of the venous circulation could be obtained if these methods were not performed in the case of a negative D-dimer test. However the safety of withholding anticoagulant therapy in out patients with negative tests needs confirmation in a prospective trial.