Chapter 8
Pulmonary embolism
Giovanni Volpicelli
Dept of Emergency Medicine, San Luigi Gonzaga University Hospital, Torino, Italy.
Correspondence: Giovanni Volpicelli, Dept of Emergency Medicine, San Luigi Gonzaga University Hospital, Torino, Italy. E-mail: giovi.volpicelli@gmail.com
The diagnosis of pulmonary embolism is often a complex process that starts from clinical suspicion and is guided by risk stratification in selected populations. LUS may have a role in this process, as it is a valid technique that is highly specific for diagnosing typical peripheral lung infarctions and highly sensitive in ruling out alternative pulmonary diagnoses to pulmonary embolism. Either alone, or in particular in combination with cardiac and venous US, sonographic examination of the lung represents a valid alternative to a multidetector CT scan when the latter is unavailable or contraindicated. Moreover, when integrated in the pre-test clinical risk score assessment of pulmonary embolism, a multiorgan US based on evaluation of pulmonary peripheral infarctions and deep vein thrombosis is useful to improve the performance of the Wells score. Integration of US in the process to assess the risk score for pulmonary embolism may significantly reduce the number of unnecessary CT scans. Finally, LUS is also useful in all undifferentiated emergency situations, such as cardiac arrest, shock and respiratory failure, to orientate the diagnosis and include the possibility of unexpected pulmonary embolism.
Cite as: Volpicelli G. Pulmonary embolism. In: Laursen CB, Rahman NM, Volpicelli G, eds. Thoracic Ultrasound (ERS Monograph). Sheffield, European Respiratory Society, 2018; pp. 102–114 [https://doi.org/10.1183/2312508X.10006717].
The diagnostic process for pulmonary embolism is a typical example of a complex problem-solving procedure, particularly in an emergency situation. The clinical symptoms and signs of pulmonary embolism are not specific, and the diagnosis cannot rely on history and physical examination alone. In practice, patients are stratified into groups of expected prevalence before deciding how to proceed in order to rule the disease in or out. This stratification defines the pre-test probability, which can be assessed either empirically (“gestalt”) or by applying clinical prediction rules calculated using independent clinical predictors. These predictors have been strongly validated by multivariate regression models in cohorts of patients with suspected pulmonary embolism [1].
In the context of patients with suspected pulmonary embolism, LUS should be evaluated as a diagnostic method. More than simply finalising the diagnosis of pulmonary diseases and injuries, very often it may contribute to expedite the process, to guide further diagnostic steps and to help restrict the differential possibilities. The main principle of the efficacy of US in examination of the lung is based on the fact that changes in the condition of pulmonary aeration and density, even at a regional extension, generate different US patterns that may help rule in or out various diseases and injuries [2].
General considerations: the Bayesian era
To optimise the diagnostic process and favour the safest and quickest decision making, the modern clinician is required to orient their mind between accredited rules on a daily basis. We currently live in the era of Bayesian medicine, which is the best-known approach for complex problem solving. According to Bayes’ theorem, the diagnostic process depends on consideration of the specific population being examined and the level of risk of having the disease in these patients [3]. Moreover, a complex evaluation of costs, benefits and invasiveness of the investigating tools must be considered to orientate the diagnostic work-up. Thus, the clinician cannot rely only on the calculated absolute accuracy of a positive or negative test to evaluate sufficiently the probability of disease; consideration about the prevalence of the disease in the tested population, as well as evaluation of the availability and opportunity to submit the patient to a specific test, are issues that may have a significant influence on the process.
Thus, the three main steps that characterise the diagnostic work-up for pulmonary embolism are: 1) a suspicion of pulmonary embolism, 2) assessing the pre-test probability and 3) deciding the most appropriate test to finalise the diagnosis. This is a diagnostic cascade that cannot be interrupted. For instance, the diagnostic process for pulmonary embolism cannot perform correctly in a population without clinical symptoms or signs that support the suspicion. In addition, without stratifying the probability risk, the accuracy of the applied diagnostic tests will remain undefined.
The European Society of Cardiology (ESC) has produced specific guidelines for the diagnostic process of pulmonary embolism [4], which start from the consideration of suspected pulmonary embolism. However, in the real practice of the emergency setting, suspicion of pulmonary embolism is not always the first step to initiate the diagnostic process. Sometimes, pulmonary embolism is a hidden or unlikely possibility, inside a more complex diagnostic process for undifferentiated respiratory failure, shock and hypotension, pleuritic pain or syncope, often in complex multipathological elderly patients. In these situations, pulmonary embolism may represent an even harder diagnostic challenge. The diagnostic process is also challenged by the difficult conditions that may be encountered, for example, in the case of unstable patients who cannot be moved in the radiology unit, or when the high risks linked to irradiation, allergies and renal failure contraindicate a contrast-enhanced CT scan.
LUS in pulmonary embolism
In general, the main diagnostic possibilities for demonstrating pulmonary embolism are: 1) detecting the presence of a thrombus in the main pulmonary artery tree, 2) detecting the presence of a source thrombus in the deep venous system or in the right heart, 3) observing the upstream effect of the artery occlusion (i.e. the haemodynamic consequence visible in the right cardiac function) and 4) visualising the downstream effect of the artery occlusion in the form of peripheral pulmonary infarcts. The last condition is potentially detected by LUS. Indeed, peripheral infarctions are due to a cascade of events following the thrombotic occlusion of a secondary pulmonary artery, which causes, in rapid sequence, collapse of surfactant, inflow of interstitial fluid and erythrocytes, and congestion of the alveolar spaces. Histological investigation of the infarcted lung reveals the alveolar spaces to be congested with erythrocytes, but the structure of the organ is fully preserved. From the US point of view, this phenomenon is simply a condition of loss of aeration and an increase in density of a small region of the peripheral lung. As such, it is potentially readily detected by a change in the US pattern of the affected area when the latter is explored by the sonographic probe.
The usefulness of LUS for pulmonary embolism has been investigated in the literature over the last 15–20 years, and it is clear that US represents a new and efficient tool in the armamentarium of physicians facing clinical conditions where pulmonary embolism is included in the differential diagnosis. The specific properties of the modern PoCUS make this tool particularly suitable at the bedside for guiding complex diagnostic processes [5]. US is a powerful, noninvasive diagnostic tool that may be repeated at the bedside to monitor changes in time, and may reduce costs and time, and is even more efficient when used by clinicians.
In the existing literature, the validation of LUS for pulmonary embolism followed three sequential steps. The use of LUS alone was the first step, demonstrating that US had the potential to visualise lung infarcts and that it was a valid alternative to CT scans in the diagnostic work-up of pulmonary embolism. Next, multiorgan US was used to show that a combination of LUS with cardiac and venous US could improve the performance of each organ examination considered alone, particularly considering the potential of LUS to detect alternative pulmonary diagnoses. This multiorgan and multitask examination is even more valid as an alternative to CT scans, not only to confirm but especially to rule out the disease. Finally, the US-enhanced prediction rule was used to demonstrate that LUS, in combination with venous US, may be useful to improve the performance of the pre-test prediction rule in the diagnostic process for pulmonary embolism. Thus, LUS contributes to the definition of the pre-test clinical probability of disease and guides further diagnostic steps in patients with suspected pulmonary embolism. These three steps are discussed further.
LUS alone: visualisation of lung infarcts
Several studies explored the accuracy of LUS for pulmonary embolism by investigating its potential in visualising the peripheral infarctions. The most significant was a multicentre study performed on 352 patients with suspected pulmonary embolism and admitted to internal medicine and pulmonology wards [6]. LUS was performed at the bedside and targeted to the visualisation of typical triangular or rounded pleural-based lesions. The main hypothesis was that at least two of these lesions were pathognomonic of the disease. The US technique consisted of scanning the whole chest in the anterior, lateral and posterior areas by longitudinal and oblique intercostal scans. The typical US lesions indicating peripheral lung infarctions appeared as small, well-defined consolidations with sharp margins, usually of triangular or rounded shape, with interruption of the pleural line and the absence of air bronchogram and vascularisation. These infarcts are considered significant when the diameter at the pleural surface is >5 mm (figure 1). In the study, the authors investigated four US patterns, assigning to each of them different diagnostic criteria in terms of probability of the disease: 1) confirmation of pulmonary embolism when at least two of these typical lung infarctions were detected, 2) probable pulmonary embolism when one typical lesion was detected combined with a corresponding low-grade pleural effusion, 3) possible pulmonary embolism when small (<5 mm) nonspecific subpleural lesions were detected or pleural effusion alone and 4) pulmonary embolism ruled out when US showed a normal lung pattern (i.e. a regular pleural line with sliding and A-lines, or other patterns sometimes indicative of alternative conditions, in the absence of typical consolidations) [6]. In most cases, the gold standard used was single-row CT. To compensate for the low sensitivity of this radiological tool, a combination of at least two of the following adjunctive criteria confirmed the diagnosis in the case of a negative CT scan: 1) high clinical suspicion, 2) a positive D-dimer result in the case of outpatients, 3) positive leg-vein thrombosis, 4) characteristic changes in echocardiography, 5) a high-probability V/Q scan, 6) angiography of the pulmonary artery or 7) necropsy. The study demonstrated that the combination of the two US criteria, confirmed pulmonary embolism and probable pulmonary embolism, yielded the best sensitivity (74%) and specificity (95%), whereas the criteria of confirmation alone showed an even better specificity (99%) at the expense of a much lower sensitivity (43%) [6]. Notably, in this study, only 352 patients were definitively enrolled out of a projected number of 710 with suspected pulmonary embolism during the study period. The authors reported different reasons to explain this high rate of exclusion. Two of the main reasons, accounting for 77% of the exclusion rate, were the impossibility of performing CT (the patient could not be transported or had a contrast medium allergy or severe renal failure, or the unavailability of CT at the time of admission; total 44%) and unavailability of an expert in US (33%) [6]. Thus, often a CT scan cannot be performed, and implementing US education will be fundamental to improve the standard of care.
A recent meta-analysis by SQUIZZATO et al. [7] analysed the results of studies on LUS for pulmonary embolism. Ten studies were evaluated for a total of 887 patients investigated. Overall, LUS showed a sensitivity of 87% and a specificity of 82% for the diagnosis of pulmonary embolism. The analysed studies covered a period of 20 years, which is quite long considering the progress in medical imaging technology and the rapid changes in the consideration of LUS signs of pulmonary embolism. For instance, the performance of the old single- or two-row detector CT is much inferior to the modern multidetector CT (MDCT) devices. Notably, the accuracy of LUS changed significantly when more advanced CT devices were used as the gold standard in pulmonary embolism studies. Pooled data from the meta-analysis demonstrated higher sensitivity (98%) and specificity (94%) when LUS was compared with MDCT [7]. Another major limitation of this meta-analysis was that in some studies the authors did not explain in detail the US criteria for the diagnosis of pulmonary embolism. Of course, the standardisation of these criteria is fundamental. Modern literature is more focused on US standardisation than the older studies. The study of MATHIS et al. [6] responds better than others to this need. The main criteria introduced in the study to support the diagnosis (i.e. typical peripheral consolidation of >5 mm) is currently widely accepted among experts.
Lungs are wide organs. Thus, exploring the lung surface by US to look for consolidations should be extended to the whole chest, including the anterior, lateral and posterior walls. A whole-chest examination may be complex to perform in some patients and in some specific settings, particularly in an emergency setting and in the critically ill. There are some situations where LUS may reveal infarctions more reliably if guided in a restricted area by the patient’s symptoms. This is typically the case for pleuritic pain. The usefulness of US to reveal pulmonary radio-occult conditions causing pleuritic pain has been clearly demonstrated [8, 9]. In these cases, US is very efficient simply because pain reveals the involvement of the parietal pleura. Thus, when the pain is due to a pulmonary disease, it is peripheral and potentially visible by LUS. Pleuritic pain is also a common and frequent symptom in pulmonary embolism. Unpublished data obtained by analysing three databases from multicentre studies on LUS for pulmonary embolism have shown that LUS performed better in terms of sensitivity and specificity in the subgroup of patients complaining of pleuritic pain (table 1) [10–12].
All patients | No pleuritic chest pain | Pleuritic chest pain | |
---|---|---|---|
Patients n | 872 | 655 | 217 |
Sensitivity % (95% CI) | 57.0 (51.0–62.9) | 49.5 (42.7–56.4) | 81.5 (70.0–90.1) |
Specificity % (95% CI) | 94.9 (92.9–96.6) | 94.8 (92.3–97.7) | 95.4 (90.7–98.1) |
PPV % (95% CI) | 84.1 (78.1–89.0) | 82.2 (74.5–88.4) | 88.3 (77.4–95.2) |
NPV % (95% CI) | 82.4 (79.4–85.2) | 79.5 (75.8–82.8) | 92.4 (87.0–96.0) |
PLR (95% CI) | 11.3 (7.83–16.20) | 9.50 (6.24–14.46) | 17.71 (8.51–36.84) |
NLR (95% CI) | 0.45 (0.40–0.52) | 0.53 (0.47–0.61) | 0.19 (0.11–0.31) |
The meta-analysis was performed on the databases of three multicentre studies [10–12]. Data represent work in progress and are unpublished. PPV: positive predictive value; NPV: negative predictive value; PLR: positive likelihood ratio; NLR: negative likelihood ratio. |