Before any diagnostic testing, patients were assigned a probability score, using the Wells Criteria, to categorize the probability of PE as low or moderate to high.
Patients with low probability Wells scores: D-dimer testing was done on all patients with a low clinical probability of PE. Patients with a negative D-dimer were randomly assigned either to no additional diagnostic testing and no anticoagulation treatment or to additional diagnostic testing with an initial VQ scan. If the VQ scan was negative, then PE was excluded. If the VQ scan showed one or more segmental perfusion defects that were normally ventilated, then the scan was considered diagnostic for PE. If there were perfusion defects that did not meet the criteria for a high probability scan, then the scan was considered non-diagnostic. Patients with non-diagnostic scans underwent ultrasonography of the proximal veins of the legs. If deep vein thrombosis was present, PE was diagnosed. If ultrasonography was normal, the test was repeated after seven and 14 days. In all patients with a positive D-dimer, a VQ scan was performed.
Patients with moderate to high probability Wells scores: A VQ scan was performed on all patients with a moderate to high probability for PE. Patients with high probability scans were treated; patients with normal scans were not treated. Patients with non-diagnostic scans and normal venous ultrasonography were randomly assigned to receive either no additional testing or serial ultrasonography.
Outcomes: All patients were followed for six months for the development of venous thromboembolism after initial diagnostic testing.
The study enrolled 1,126 patients. Overall, 160 patients (14.2%) had PE diagnosed at initial presentation or by venous ultrasonography. Of 952 patients who did not receive an initial diagnosis of PE, 11 (1.2%) had PE diagnosed at follow-up.
Patients with low probability Wells scores: Low clinical probability was present in 670 patients (60%). In patients with low clinical probability of PE, 373 (56%) had negative D-dimer tests and 297 (44%) had positive D-dimer tests. Of the 373 patients with low probability and negative D-dimer results, 187 were randomized to no additional testing and 186 received a VQ scan. The frequency of venous thromboembolism at six-month follow-up was similar in these two groups (-0.5% [CI, -3.0% to 1.6%]). Three patients with negative D-dimer tests were diagnosed with PE by VQ scan. Results were fairly complete (five patients without a six-month follow-up in the no additional testing group and one without a follow-up in the VQ scan group).
Twenty-four patients with low clinical probability and positive D-dimer results (n=297) were diagnosed with PE. Three patients did not complete the six-month follow-up. Of the remaining 294 patients, five patients had venous thromboembolism at six months.
Patients with moderate to high probability Wells scores: There were 456 patients (40%) had moderate or high clinical probability for PE. Each of these patients had a VQ scan. Non-diagnostic VQ scans and normal venous ultrasonography were performed on 226 patients. Of these 226 patients, 86 had a negative D-dimer and 140 had a positive D-dimer. Of the 86 patients with negative VQ scans, normal venous ultrasonography, and a negative D-dimer, 83 were randomly assigned to no additional testing or serial venous ultrasonography (42 and 41 respectively). At six months follow-up, one patient assigned to no additional testing had venous thromboembolism, and no patients in the additional testing group had venous thromboembolism.