Objective To determine the diagnostic accuracy of point-of-care ultrasound in suspected pulmonary embolism.

Design Systematic review and meta-analysis.

Data sources MEDLINE, Embase, CINAHL and Cochrane library were searched on 2 July 2020 with no restrictions on the date of publication. Subject headings or subheadings combined with text words for the concepts of pulmonary embolism, ultrasound and diagnosis were used.

Eligibility criteria and data analysis Eligible studies reported sensitivity and specificity of deep venous, lung, cardiac or multiorgan ultrasound in patients with suspected pulmonary embolism, using an adequate reference-test. Prospective, cross-sectional and retrospective studies were considered for eligibility. No restrictions were made on language. Studies were excluded if a control group consisted of healthy volunteers or if transesophageal or endobronchial ultrasound was used. Risk of bias was assessed using quality assessment of diagnostic accuracy studies-2. Meta-analysis of sensitivity and specificity was performed by construction of hierarchical summary receiver operator curves. I² was used to assess the study heterogeneity.

Main outcome measures The primary outcome was overall sensitivity and specificity of reported ultrasound signs, stratified by organ approach (deep venous, lung, cardiac and multiorgan). Secondary outcomes were stratum-specific sensitivity and specificity within subgroups defined by pretest probability of pulmonary embolism.

Results 6378 references were identified, and 70 studies included. The study population comprised 9664 patients with a prevalence of pulmonary embolism of 39.9% (3852/9664). Risk of bias in at least one domain was found in 98.6% (69/70) of included studies. Most frequently, 72.8% (51/70) of studies reported >24 hours between ultrasound examination and reference test or did not disclose time interval at all. Level of heterogeneity ranged from 0% to 100%. Most notable ultrasound signs were bilateral compression of femoral and popliteal veins (22 studies; 4708 patients; sensitivity 43.7% (36.3% to 51.4%); specificity 96.7% (95.4% to 97.6%)), presence of at least one hypoechoic pleural-based lesion (19 studies; 2134 patients; sensitivity 81.4% (73.2% to 87.5%); specificity 87.4% (80.9% to 91.9%)), D-sign (13 studies; 1579 patients; sensitivity 29.7% (24.6% to 35.4%); specificity 96.2% (93.1% to 98.0%)), visible right ventricular thrombus (5 studies; 995 patients; sensitivity 4.7% (2.7% to 8.1%); specificity 100% (99.0% to 100%)) and McConnell’s sign (11 studies; 1480 patients; sensitivity 29.1% (20.0% to 40.1%); specificity 98.6% (96.7% to 99.4%)).

Conclusion Several ultrasound signs exhibit a high specificity for pulmonary embolism, suggesting that implementation of ultrasound in the initial assessment of patients with suspected pulmonary embolism may improve the selection of patients for radiation imaging.

PROSPERO registration number CRD42020184313.