The plasma level of D-dimer, a fibrin degradation product (FDP), is nearly always increased in the presence of acute pulmonary embolism (PE). Hence, a normal D-dimer level (below a cutoff value of 500 micrograms/L by enzyme-linked immunosorbent assay [ELISA]) may allow the exclusion of PE. To assess the negative predictive value of a D-dimer concentration below 500 micrograms/L in outpatients with suspected PE, and the safety of withholding anticoagulant treatment from such patients, we performed D-dimer assays, lower limb venous compression ultrasonography, and lung scans in 671 consecutive outpatients presenting in the Emergency Center of the Geneva University Hospital with suspected PE. Pulmonary angiography was reserved for patients with an inconclusive noninvasive workup. Patients with a normal D-dimer concentration were discharged without anticoagulant treatment and followed for 3 mo. The prevalence of PE was 29%, and D-dimer (using a cutoff of 500 micrograms/L) had a diagnostic sensitivity for PE of 99.5%. Overall diagnostic specificity of D-dimer was 41%, but it was lower among older patients. Of the 198 patients with a D-dimer concentration below the cutoff value, 196 were free of PE, one had a PE, and one had incomplete information because of loss to follow-up. Thus, the negative predictive value of D-dimer concentration fell between 197 of 198 and 196 of 198 cases of PE (99% [95% CI: 96.4 to 99.9]). Using a cutoff value of 4,000 micrograms/L, the overall specificity of D-dimer concentration for PE was 93.1%. In conclusion, a plasma D-dimer concentration below 500 micrograms/L allows the exclusion of PE in 29% of outpatients suspected of having PE. Withholding anticoagulation from such patients is associated with a conservative 1% risk of thromboembolic events during follow-up.