Lung ultrasound in a tertiary intensive care unit population: a diagnostic accuracy study

Many patients admitted to the intensive care unit (ICU) meet criteria for acute respiratory failure [1]. Common causes for respiratory failure in these patients include cardiogenic pulmonary edema (CPE), acute respiratory distress syndrome (ARDS), atelectasis and pneumonia [2]. To date, to detect these conditions chest X-ray and thoracic computed tomography (CT) are still the most common diagnostic modalities. However, recent years have seen an increase in the use of lung ultrasound [3, 4].

Above-mentioned respiratory conditions are associated with increased attenuation on thoracic CT, which can be divided into ground-glass opacity and consolidations. Consolidations are, for example, often caused by pneumonia or atelectasis, in contrast to CPE, which is more associated with ground-glass opacities [5]. Lung ultrasound can accurately detect these thoracic CT findings and provide a less costly, ionizing radiation-free diagnostic modality without the risk of transportation [6]. Previous studies have already shown that so-called B line artifacts on lung ultrasound are associated with an increased amount of pulmonary edema and thus correlate with linear and ground-glass opacities on thoracic CT [7, 8]. Moreover, consolidations are readily detected by lung ultrasound because they are usually the result of replacement of air by fluid or cells and are, therefore, easily traversed by ultrasound waves [3, 9].

Previous studies relating lung ultrasound to thoracic CT findings in ICU patients, however, were hampered by multiple methodologic limitations. Most studies excluded patients with multiple diagnoses, were inadequately blinded, or did not include anterior consolidations and/or pneumothorax [8, 10‐15]. Subsequently, a study that is appropriately blinded and investigates the full diagnostic extent of lung ultrasound in an ICU population was needed [7, 16].

As follows, the primary aim of this study was to evaluate the diagnostic accuracy of a 6-zone lung ultrasound protocol in ICU patients, with thoracic CT as reference standard, using adequate blinding and including patients with multiple respiratory conditions. Secondary aims were: 1. To evaluate the diagnostic accuracy of an extended 12-zone lung ultrasound protocol and 2. To correlate lung ultrasound patterns with respiratory conditions in a tertiary ICU population, while taking into account that one patient could have multiple lung ultrasound abnormalities, but also be affected by multiple respiratory conditions.

For primary analysis, 87 patients were included in two distinct time periods: from January 2016 until July 2016 (24 weeks) and September 2017 to January 2019 (70 weeks). In eight patients thoracic CT and lung ultrasound were not performed within the timeframe of eight hours and were subsequently excluded. There was no missing data, except in five patients one of the two PLAPS-points could not be accurately visualized due to thoracic drains or surgical dressings and these were excluded from consolidation and pleural effusion analyses. Baseline characteristics are described in Table 1. The results are depicted in Table 2. For secondary analysis (patients who received a 12-zone lung ultrasound examination) 18 patients were included. The results are depicted in Table 3. The results per lung pattern are described below.

Prevalence of consolidations was 0.50 (156/311). Estimated sensitivity and specificity were 0.76 (95%CI: 0.68 to 0.82) and 0.92 (0.87 to 0.96), respectively. Prevalence of interstitial syndrome was 0.51 (80/158). Estimated sensitivity and specificity were 0.60 (95%CI: 0.48 to 0.71) and 0.69 (95%CI: 0.58 to 0.79), respectively. Prevalence of pneumothorax was 0.11 (17/158). Estimated sensitivity and specificity were 0.59 (95%CI: 0.33 to 0.82) and 0.97 (95%CI: 0.93 to 0.99), respectively. Out of the 17 pneumothoraces (hemithorax) three were clinically significant, i.e., replacement of the thoracic drain due to persistent pneumothorax in one case and surgical intervention due to persistent air leak in two cases. All three cases were correctly identified by lung ultrasound. Prevalence of pleural effusion was 0.71. Estimated sensitivity and specificity were 0.85 (95%CI: 0.77 to 0.91) and 0.77 (95%CI: 0.62 to 0.88), respectively.

In total there were 147 respiratory conditions in 79 patients. The majority (49.4%, 39/79) had two respiratory conditions. The most prevalent lung pathology was atelectasis, followed by pneumonia. The least prevalent lung pathology was lung infarction. Figure 2 shows a heatmap combining respiratory conditions and ultrasound patterns. Dendrograms show hierarchical clustering of respiratory conditions with respect to ultrasound patterns and vice versa.

The main findings of this prospective observational study are: 1. In a tertiary study population lung ultrasound is an accurate diagnostic modality to detect consolidation and pleural effusion, whereas the diagnostic accuracy to detect pneumothorax and interstitial syndrome is slightly lower. However, all clinically significant pneumothoraces were correctly identified by lung ultrasound. 2. The diagnostic accuracy did not seem to differ between the 6-zone BLUE-protocol and an extended 12-zone lung ultrasound protocol. 3. The majority of patients had at least two respiratory conditions, correlated to (partially) similar lung ultrasound patterns (Fig. 2). These findings are novel as, to our knowledge, this is the first study that investigated the full extent of lung ultrasound in ICU practice. Previous studies solely focused on one diagnosis (for example, ARDS, ILD or pneumonia) did not include anterior consolidations, or were inadequately blinded [8, 10, 12‐15].

Sensitivity and specificity of lung ultrasound to detect consolidation are in line with previous research. However, in contrast to two studies with similar design, anterior consolidations were included as well [8, 13]. Anterior consolidations were defined as a C-profile or B-profile without lungsliding. Of note, a B-profile without lungsliding has been shown to be correlated to ARDS and pneumonia [7, 23]. Sensitivity and specificity for interstitial syndrome were lower than reported in the previous literature. We assume the reason to be twofold. First, sensitivity and specificity were affected by regarding a B-profile without lungsliding to be an anterior consolidation, whereas this might not always be the case. Second, ultrasound examinations were performed with a linear transducer, whereas recent literature suggests that B-lines are often better visualized using a microconvex or abdominal transducer [23, 24]. Sensitivity to detect pneumothorax was low, but specificity high. The number of areas of the lung covered by lung ultrasound examination offers an explanation; most false negatives were small apical or retrosternal pneumothoraces and these are easily missed by strictly following the BLUE-protocol. However, as our results show, an A-profile without lungsliding in ICU patients is highly indicative of pneumothorax. Sensitivity and specificity to detect pleural effusion were only slightly lower than previously reported [8, 13]. Again we believe this reason to be twofold. First, a small amount of pleural effusion in proximity of the spine is easily missed if patients are only examined in supine position, this reduces the sensitivity of lung ultrasound. Second, only a minimal amount of pleural effusion at one of the lung ultrasound points already renders the ultrasound examination positive, whereas in that case pleural effusion might not be visible on thoracic CT and, consequently, lowers the specificity.

Our study showed that an extended 12-zone lung ultrasound protocol is of little benefit in diagnosing lung pathology. Most diagnostic accuracy parameters were comparable to the 6-zone BLUE protocol (Table 2 and 3). This might be explained by the fact that almost all acute respiratory disorders involve a large portion of the subpleural space and are therefore easily accessible to ultrasound. In other words, if one or two of the anterior BLUE-points are involved in an acute respiratory disorder, the extra points scanned below the anterior axillary line are most likely involved as well. Moreover, if the PLAPS-point is involved, the extra point scanned more superiorly below the posterior axillary line is likely involved as well. Therefore, the 6-zone lung ultrasound protocol suffices and any additional points do not increase the diagnostic yield. Other studies have demonstrated the same results [25‐27].

The presence of multiple respiratory conditions per patient in a tertiary ICU population is a common occurrence, but makes a diagnostic accuracy study difficult to perform. We addressed this issue by showing the correlation between lung ultrasound findings and respiratory conditions in a heatmap (Fig. 2). One should be careful to interpret the results in an absolute sense as the purpose is to provide the clinician with a tool to incorporate lung ultrasound in the diagnostic process. For example, if a C-profile is present, the differential diagnosis should contain pneumonia, but the possibility that a patient has an additional other diagnosis should not be disregarded. In our study, a patient who underwent an esophagectomy and gastric pull-up which was postoperatively complicated by an anastomotic dehiscence developed ARDS, a pneumothorax and compression atelectasis; all respiratory conditions could lead to acute respiratory failure. If one follows a deterministic protocol in any ICU setting, such as the one proposed by Mojoli et al.[19], one could easily miss other diagnoses if present. With the findings of the current study, we would like to emphasize the importance of using lung ultrasound as a diagnostic aid rather than as deterministic model with an exhaustive list of diagnoses. In the heatmap, respiratory conditions and lung ultrasound findings are clustered by a dendrogram. For example, with regard to ultrasound findings, ARDS and cardiogenic pulmonary edema are similar respiratory conditions. Both are characterized by bilateral B-lines, posterior consolidations and pleural effusion. In fact, the distinction between these conditions is notoriously difficult and is the subject of ongoing research (28, 29).

Due to its usefulness, the role of bedside ultrasound has been increasing in the intensive care setting. Not only lung ultrasound, but also transthoracic echocardiography have become standard care in many ICUs. Reasons for this surge in popularity are obvious: immediate bedside availability in deteriorating patients, absence of ionizing radiation and avoiding transportation to the radiology department. Despite these advantages, to deliver best patient care and avoid conflicts, well-deliberated agreements between ICU physicians and other medical specialties are necessary, especially in cardiology and radiology. For example, in case of ultrasound of the heart the intensivist should be able to qualitatively assess the systolic function of the heart or exclude important conditions such as cardiac tamponade, whereas quantitative assessment of systolic function and other more advanced measurements remain mostly reserved for echocardiography performed by the cardiologist. Emphasizing the importance of careful deliberation, in case of more advanced trained ICU physicians – for example with an European Diploma in Advanced Critical Care Echocardiography (EDEC) – departments can decide to change these agreements.

This study has several limitations. First, it was a single-center observational study conducted in a tertiary ICU population; a substantial part of the study population was immunocompromised or underwent complicated surgical procedures. Second, lung ultrasound examinations were performed by multiple ultrasound operators with different levels of experience. More experienced operators are more likely to detect subtle lung ultrasound abnormalities and, thus, are better able to correctly classify each examination. However, all operators followed a basic ICU point-of-care ultrasound course and were able to perform examinations. With this in mind, our study results are indeed better generalizable than results of a study with just a single experienced lung ultrasound operator. Moreover, ultrasound examinations are inherently dynamic and interpretation is operator dependent. Because of the operator dependency, it requires all medical staff to be trained in performing ultrasound examinations. In contrast, CT is often performed by a technician and images can then be remotely reviewed by a single radiologist. However, in an acute situation it is important to rapidly obtain a diagnosis and, in our opinion, ICU physicians should be able to perform basic ultrasound examinations. Major strengths of our study are that this is, to our knowledge, the first study that included anterior consolidations and pneumothorax, is adequately blinded and does not exclude patients with multiple diagnoses. Due to the fact that this study allowed for the possibility of multiple respiratory conditions, these results are very representative of and applicable to routine ICU practice. Most importantly, this is also one of the largest studies regarding the diagnostic accuracy of lung ultrasound performed on a tertiary ICU population.

In conclusion, lung ultrasound is an adequate diagnostic modality in a tertiary ICU population to detect consolidation, clinically significant pneumothorax and pleural effusion and is less suitable to detect interstitial syndrome. A 12-zone lung ultrasound protocol shows no benefits over the 6-zone lung ultrasound protocol. Moreover, one should be careful not to interpret lung ultrasound findings in deterministic fashion as most patients have multiple respiratory conditions associated with (partially) similar lung ultrasound patterns.