Combining lung ultrasound and Wells score for diagnosing pulmonary embolism in critically ill COVID-19 patients

We recruited all patients reverse transcriptase polymerase chain reaction (rtPCR)-confirmed SARS-CoV-2 infection treated at the medical intensive care unit (ICU) at the University of Freiburg Medical Center between March 8th and May 31th, 2020 if they fulfilled all of the following inclusion criteria:

All data for this study was taken from the electronic patients files. As data were collected retrospectively, no interventions were applied for the purpose of this study and all patients were treated according to current treatment standards and guidelines.

Our ICU is located at a university hospital offering a 24/7 ECMO center specialized in acute respiratory distress syndrome (ARDS). ARDS treatment is performed according to current guidelines, including early mobilization or prone positioning and early spontaneous breathing in patients without desynchronization with the ventilator [12]. In case of severe pulmonary failure, a multidisciplinary team including at least the intensivist in charge, an ECMO specialist, a registered nurse and a perfusionist decide about extracorporeal membrane oxygenation (ECMO) in severe courses where invasive mechanical ventilation is not sufficient. During the SARS-CoV2 pandemic, daily LUS was encouraged by local standard operating procedures when deterioration in respiratory function was evident.Imaging (sonography, including LUS, echocardiography and sonography) was performed by experienced intensivists. A CTPA was performed when indicated by the intensivist and radiologist in charge.

Experienced intensivists with advanced knowledge in sonography carried out the LUS. Evaluation and assessment of the lung sonography findings was mostly done with knowledge of the d-dimer value, but before performing a CTPA and therefore prospectively blinded with regard to the CTPA result. LUS examination was carried out using a Philips CX50 echocardiography machine with a multifrequency probe C5-1 (5-1 MHz) or L12-3 (12-3 MHz). Alternatively, a Philips Sparq with a multifrequency probe C6-2 (6-2 MHz) or L12-4 (12-4 MHz) was used. Due to our local standard, we used an adjusted BLUE protocol for investigating the lungs. The BLUE protocol [13] is a standardized diagram for the rapid identification of 97% of the causes of dyspnea in adult patients (pulmonary edema, pneumonia, PE, COPD, asthma, pneumothorax). Ultrasound examinations were performed along the midclavicular line in the bilateral anterior chest wall and the scapular line and interscapular regions in the posterior chest wall—each right and left side of the chest—at the bedside. Since the mechanically ventilated intensive care patients can usually only be examined either lying on the front or back, examination of all 12 lung fields was only possible in individual cases, mostly only 8 fields could be examined reliably. Furthermore, we focused on the COVID-19-typical signs as described before [8], such as multiple B-lines (comet-tail artefacts) in a variety of patterns (focal, multifocal and confluent), a thickening of the pleural line with irregularity and consolidations in a variety of patterns (see Fig. 2). On a LUS survey sheet we documented the number of fields examined as well as the patterns described above (B-Lines, consolidations, pleura-irregularities) for each individual field. Finally, the summary of the examination assessed whether it was suitable for COVID-19 and whether PE was highly likely, probable, possible or unlikely.

The diagnosis of PE suggested by Mathis et al., based on the number and size of the subpleural consolidations, was used but slightly modified. Mathis et al. describe consolidations with a size of more than 5 mm as typical for PE [14]. Due to the pronounced pleural changes in COVID-19 with a significant thickening of the pleura, we only considered triangular consolidations ≥ 1 cm as PE-typical, < 1 cm as non-typical for PE. The following criteria for detection of PE diagnosis were used.

PE is considered highly likely when two or more characteristic triangular lesions (≥ 1 cm) were demonstrated; PE is considered probable: if one characteristic triangular lesion (≥ 1 cm) was detected; PE is considered possible: if two (or more) non-typical lesions (< 1 cm) were detected; PE is considered unlikely: neither typical nor atypical consolidations.

To assess hemodynamic relevance of the PE, additionally, right heart echocardiography and sonography of the vena cava were performed.

Our local standard operating procedure strongly recommended that all patients with ARDS and SARS-CoV2 infection should undergo CTPA, unless endangering patient heath. CTPA scans were performed using a commercial CT scanner (SOMATOM Definition Flash; Siemens Healthineers GmbH, Forchheim, Germany) with the following scanning parameters: tube voltage, 100 kV; tube current, 90 mAs; rotation time, 0.28 s. 128 × 0.6 mm collimation with automated dose modulation (CARE dose4D, Siemens Healthineers GmbH, Forchheim, Germany). Patients without extracorporeal membrane oxygenation (ECMO) received the standard pulmonary angiography protocol with bolus-tracking method of 70 ml contrast agent (Imeron 400, Bracco Imaging, Germany). To consider an altered blood flow in patients with ECMO device the amount of contrast agent was adjusted to 100 ml and the ROI was placed in the air. The scan was manually started when an adequate contrast was visually detected in the pulmonary trunk. If tolerated by the patient ECMO flow was reduced to 70 to 50% of the initial value after scout acquisition for the time of the contrast-enhanced scan. Each pulmonary lung segment was separately evaluated for parenchymal abnormalities (ground-glass opacities and/or crazy-paving pattern, and air space consolidation) and PE.

A segmental or subsegmental PE was defined as central filling defect within a vessel surrounded by contrast material when orthogonal or parallel to the long axis of the vessel as well as eccentric filling defect rendering an acute angle with the vessel wall as well as complete occlusion of a dilated vessel [15].

The Wells score is a well-established screening tool for PE and is used in everyday care to assess the clinical pre-Test probability at our institution [16]. The following questions are scored: Clinical signs and symptoms of deep vein thrombosis (DVT)? (+ 3), PE is #1 diagnosis or equally likely? (+ 3), heart rate > 100 bpm? (+ 1.5), Immobilization at least 3 days OR surgery in the previous 4 weeks? (+ 1.5), previous objectively diagnosed PE or DVT? (+ 1.5), hemoptysis? (+ 1) and malignancy with treatment within 6 months or palliative? (+ 1). The largest study [16] using a three-tier Model and demonstrated risk stratification with: (a) Low score of 1–2 points having a 1.3% prevalence, (b) Moderate score of 2–6 points having a 16.2% prevalence or (c) High score of > 6 points having a 37.5% prevalence. In order to reduce inter-observer variance for our research, the Wells score was retrospectively re-assessed for the day of CTPA by a single intensivist considering all available data.

Statistical analyses were performed using SPSS, version 23.0, (IBM, New York City, USA). Graphs were drawn with Prism, version 8.4.3 (GraphPad, San Diego, USA).

Continuous variables were compared using student’s T-Test, Fisher`s exact test was used for contingency table analyses. Significance level was set at p < 0.05. The study confirms with the 1975 declaration of Helsinki and it was approved by the ethics committee of the Albert Ludwig University of Freiburg (file number 234-20).

A total of 113 patients were admitted to the ICU within the observed time period, of these 25 patients were diagnosed with severe respiratory failure due to COVID-19, all of which underwent LUS. Four Patients had to be excluded from the study because they did not receive a CTPA, in one patient LUS could not be reliably evaluated due to huge parenchymal pulmonary bleeding and hemothorax (patients characteristics of patients excluded are given in supplemental material Table 1). Therefore, 20 patients could be enrolled in this study (see Fig. 1). Mean age (± S.D.) of the patients was 61.6 ± 9.9 years, SAPS II score was 48.4 ± 12.4 on day 1 after admission to ICU. Eighteen of twenty patients (90%) were intubated. Eleven from 20 (55%) patients were on ECMO during their ICU stay. For more detailed patient characteristics see Table 1.

The final diagnosis of PE was confirmed by CTPA. In total 300 lung segments were evaluated. In 12 out of 20 patients (60%), segmental and subsegmental PE were detected. In patients with PE 62 segments in total with an average of 2.90 ± 3.38 lung segments were affected (range 1 to 9). The analysis of distribution of PE on lung lobe level, showed the right lower lobe to be affected in 10 out of 12 patients with PE, followed by the right upper lobe which was affected in 8 out of 12 patients. The complete PE distribution of the entire lung is shown in Fig. 1 of the supplement.

LUS examinations showed abnormal lung ultrasound findings, with pleural abnormalities including thickening of the pleural line with pleural line irregularity in nineteen of twenty cases (95%). B-lines, in a variety of patterns including focal, multifocal, and confluent could also be documented in 19 from 20 (95%) patients. Multifocal B-lines were found in 14 patients and confluent in 11 patients, however B-line pattern was heterogeneous in individual patients. Subpleural consolidations were found in 18/20 (90%) patients. Typical consolidations with a size/depth of > 1 cm were detectable in 13/20 (65%) patients. Ten (50%) patients showed atypical consolidations (< 1 cm). Typical and atypical consolidations could occur co-dominantly in the same individual. The COVID-19-typical lung sonographic findings are summarized in Table 2. Sonographic image-examples of these different pleural morphologies are shown in Fig. 2.

Average Wells score in the entire cohort was 2.3 ± 0.8. In the group with evidence of PE, the wells score was significantly higher than in the group without PE (2.7 ± 0.8 in patients with respectively 1.7 ± 0.5 in patients without PE, p = 0.042).

Using LUS and the criteria described above and suggested by [14], PE was considered in 12/20 (60%) patients, probable in 1/20 (5%), possible in 2/20 (10%), and unlikely in 5/20 (25%) patients. When forming a two-tier scale of probable PE (considered/probable PE in LUS versus possible/unlikely PE in LUS), pulmonary embolism could be predicted with an AUC of 0.729 and a sensitivity of 77% and a specificity of 71%, see Fig. 2. The Wells score estimated the risk for PE as very high (Wells score > 6) in 0/20 patients, as moderate (score 2–6) in 10/20 patients and as low (score < 2) in 10/20 patients. With the Wells score, PE could be predicted with an AUC of 0.813 and a sensitivity of 90% and a specificity of 70%, see Fig. 2. When combining the two modalities, comparing patients with considered/probable PE using LUS plus a Wells score ≥ 2 to patients with possible/unlikely PE using LUS plus a Wells score < 2, PE could be predicted with an AUC of 0.944 and a sensitivity of 100% and a specificity of 80%, see Fig. 3.