Utility of ultrasound imaging of the lower extremities in the diagnostic approach in patients with suspected pulmonary embolism


Dr Philip Wells Suite 453 737 Parkdale Avenue, Ottawa, Ontario, Canada K1Y 1J8 (fax: 613-761-4840; e-mail: pwells@ottawahospital.on.ca).


In a recent study, we demonstrated that use of a clinical model to determine pretest probability was valuable in the management of patients with suspected pulmonary embolism (PE) [1]. The goal of that study was to utilize lower extremity ultrasound imaging, repeated over 2 weeks if the initial ultrasound was negative, as the predominant diagnostic strategy and, thereby, limiting the need for angiography. The safety of that management strategy was validated; however, the approach was cumbersome as it required the performance of a large number of ultrasounds, the vast majority of which were normal. It has been reported that, in the absence of risk factors for venous thromboembolism (VTE) and symptoms of deep vein thrombosis (DVT), very few patients with suspected PE have abnormal ultrasound results [2]. As this could change the diagnostic, we undertook an evaluation of this observation in our study cohort.

All patients were evaluated by a physician to determine the clinical pretest probability for PE using a clinical model. Ventilation-perfusion lung scans and bilateral compression ultrasound from the common femoral vein to the trifurcation of the calf veins were then performed in all patients [3]. The diagnostic algorithm is outlined in Fig. 1.

Figure 1.

 Diagnostic strategy used in patients with suspected pulmonary embolism (PE).

Anticoagulant therapy was withheld in patients unless, as outlined in the algorithm, PE or DVT was diagnosed. Patients were defined as PE+ if one or more of the following occurred: abnormal pulmonary angiography, abnormal ultrasound or venography, high probability ventilation-perfusion lung scan plus moderate or high pretest probability; or VTE within the 3 month follow-up period. All other patients were classified as PE–.

Patients were followed-up for 90 days and those with suspected VTE, at any time in the 90-day period, underwent objective testing performed in a standardized way [1]. We determined the accuracy of ultrasound for detection of DVT for each of the following four categories of patients with suspected PE: those patients with (i) both clinical symptoms of DVT and risk factors for VTE, (ii) symptoms of DVT but no risk factors, (iii) risk factors but no symptoms of DVT and (iv) neither symptoms nor risk factors. In our analysis, risk factors included previous DVT/PE (objectively diagnosed by ultrasonography, venography, angiography or high probability ventilation-perfusion lung scan), major surgery within 4 weeks (12 weeks in the original study), immobilization for more than 3 days in the preceding 4 weeks, cancer, the postpartum period (the 6 weeks postdelivery), lower limb paralysis, lower limb fracture or cast within 12 weeks, strong family history (>1 first degree relative with VTE, or a family member with a thrombophilia).

Of the 1401 eligible patients 147 refused consent, two had inadequate ventilation-perfusion lung scans and 13 were lost in follow-up due to moving from the study region. Thus, 1239 patients were evaluated of whom 155 had previous DVT or PE. Of the 1239 valuable patients, 334 (27%) had normal perfusion scans, 736 (59%) had nondiagnostic probability ventilation-perfusion lung scans, and 169 (14%) had high probability ventilation-perfusion lung scans. Including events in the 3 month follow-up period, four patients (1.2%) were PE+ in the normal perfusion scan group, 62 patients were PE+ in the nondiagnostic probability ventilation-perfusion lung scan group (8%), and 151 (89%) were PE+ in the high probability ventilation-perfusion lung scan group. Overall 17.5% (217) of patients were PE+. The initial ultrasound was abnormal in 97 of the PE+ patients for a sensitivity of 45%. Serial testing detected a further 13 cases of DVT. Thus, the sensitivity of serial ultrasound testing for PE was 52% (110/217) and the specificity was 99% (997/1007–15 patients with normal lung scans had normal impedance plethysmography tests and ultrasound was not performed), the negative predictive value of ultrasound was 89% and the positive predictive value, 91%. The prevalence of abnormal ultrasound results according to the presence of signs and symptoms of DVT and risk factors for VTE is outlined in Table 1. Combining all patients with either symptoms of DVT or risk factors 13.9% had DVT by ultrasound. Only 7% (30/421) of patients with neither risk factors nor symptoms of DVT were PE+; ultrasound was abnormal in 23% (7/30 – four on serial testing) which represents only 1.7% of this patient group (7/421 – four on serial testing). Compared with the abnormal ultrasound rate of 12.8% (103/803, 95%CI=10.6–15.4%) in patients with either risk factors or clinical symptoms of DVT this is a statistically significant difference (P=0.001). In the patients with nondiagnostic ventilation-perfusion lung scans the ultrasound results were very similar to those obtained in the patients described above (data not shown).

Table 1.   Influence of presence or absence of risk factors and clinical signs and symptoms of DVT on the ultrasound result according to PE+ or PE– diagnosis (all patients regardless of lung scan results) Thumbnail image of

One objective of our original study was to develop a management strategy in patients with nondiagnostic ventilation-perfusion lung scans that would result in a low VTE rate during 3-month follow-up. We aimed for a rate similar to that seen in patients in whom PE was considered to be ruled out on the basis of a normal ventilation-perfusion lung scan as physicians are willing to rule out PE in patients with normal ventilation-perfusion lung scans [4,5]. We accomplished this low follow-up VTE event but many patients in whom ultrasound was performed had normal results. We acknowledge that by not obtaining gold standard tests in all our patients some cases of PE were missed. However, some VTE will have no clinical consequences and in this regard clinical outcome may be considered more relevant than definitive diagnosis. This has recently been demonstrated in a study assessing VTE in patients postorthopaedic surgery [6].

In this posthoc analysis, we sought to determine the prevalence of abnormal ultrasound results according to whether patients had risk factors for venous thrombosis or symptoms of DVT in order to compare our findings with those of a previous analysis [2]. Our study, which was much larger than that of Rosen also suggests that performing venous ultrasound imaging may not be warranted in patients who do not have either clinical symptoms of DVT or risk factors for DVT/PE regardless of the ventilation-perfusion lung scan result. Routine ultrasounds may not be justified in this group as only 1.7% will be abnormal and indicate the presence of PE. It could be argued that because this rate is similar to the rate of PE in patients with normal lung scans, further tests are not indicated in patients with nondiagnostic probability ventilation-perfusion lung scans if they do not have risk factors for DVT and they do not have clinical symptoms of DVT. Other strategies are needed in these cases. Whether clinical probability, sensitive D-dimer tests or other imaging tests should be used needs to be determined.


Funding for this study was provided by the National Health Research and Development Program of Canada (project no. 6606-5283-403).

Dr Philip Wells is a Canada Research Chair, Dr Kearon is the recipient of a Research Scholarship from the Heart and Stroke Foundation of Canada. Drs Weitz and Ginsberg are recipients of Career Investigator Awards from the Heart and Stroke Foundation of Ontario, and Dr Weitz holds a Canada Research Chair and the Heart and Stroke Foundation of Ontario JF Mustard Chair in Cardiovascular Research; Dr Hirsh is a Distinguished Professor of the Heart and Stroke Foundation of Canada.