Diagnostic Accuracy of Bedside Lung Ultrasound and Chest Radiography Compared to Thoracic Computed Tomography in Critically Ill Patients

Bedside lung ultrasound (LUS) is being adopted in ICUs, with a higher diagnostic accuracy for consolidation, interstitial syndrome, pleural effusion, and pneumothorax when compared with chest radiography (CXR). Compare the diagnostic performance of LUS and CXR for the detection of pathologic abnormalities in critically ill patients in ICU, using thoracic computed tomography (CT) as a gold standard. Prospective and descriptive single-center study. Prior to CT, we performed bedside LUS and CXR, evaluating hemithoraces and characterized them as positive or negative for each of four abnormalities (consolidation syndrome, interstitial syndrome, pleural effusion, and pneumothorax). We decided therapeutic intervention after CT. A total of 58 patients were evaluated. With CT, consolidation was present in 100 (89.3%) of the evaluated hemithoraces, followed by pleural effusion in 74 (66.1%), interstitial syndrome in 40 (35.7%), and pneumothorax in 4 (3.6%). LUS performed better than CXR, exhibiting significantly higher sensitivity for all conditions. CXR had a marginally higher specificity than LUS for consolidation, interstitial syndrome, and pneumothorax. In 42 (72%) patients, imaging studies led to a specific action; in 33 (79%) patients, a subsequent measure was performed based on the information provided by LUS. In critically ill patients, bedside LUS shows better diagnostic performance than CXR for the diagnosis of common pathologic conditions and could be an alternative to CT scans.


Introduction
Lung imaging in critically ill patients is performed either by bedside chest radiography (CXR) or thoracic computed tomography (CT) [1]. These approaches have limitations that could lead to misdiagnosis and adverse outcomes [1][2][3][4]. CXR requires patients to be moved to a difficult position, while to perform CT, which is the most sensitive method, the patient needs to be transported out of the intensive care unit (ICU). Bedside lung ultrasound (LUS) is being increasingly used in ICUs, with a higher diagnostic accuracy for consolidation, interstitial syndrome, pleural effusion, and pneumothorax than CXR [3,5,6]. Although portable CXR lacks the sensitivity and specificity of CT, it is widely used because of its low cost and speed [7,8]. LUS does not require the demanding positioning of CXR, and unlike CT, the equipment can be brought into the ICU. LUS compares to the accuracy of chest CT for respiratory findings and provides physicians with a radiation-free tool [9][10][11][12][13][14][15].
International guidelines support the use of point-ofcare LUS to diagnose pathological abnormalities in critically ill patients. The reported sensitivity and specificity of LUS, when compared to CT, was high among hospitalized patients with acute respiratory failure, and its sensitivity was better than that of CXR; however, paired comparisons in individual patients are scarce, and published studies have This article is part of the Topical Collection on Imaging * Javier Roberti javierroberti@gmail.com 1 methodological limitations, such as a high risk of selection bias [9]. Our objective was to compare the diagnostic performance of LUS and bedside CXR for the detection of pathologic abnormalities in critically ill patients in the ICU, with chest CT as the gold standard.

Method
This prospective and descriptive single-center study was conducted between November 2018 and November 2019 at a medical-surgical ICU of a teaching hospital in Buenos Aires. The ICU admits patients who require hemodynamic, respiratory, circulatory, renal assistance, or special observation support.

Subjects
We included patients admitted to the ICU, aged ≥ 18 years who needed mechanical ventilation (MV) for ≥ 12 h. Patients could be enrolled if a physician not involved in this study had prescribed a thoracic CT because of clinical deterioration resulting in any of the following events: (i) a suspicion of respiratory infection because of fever ≥ 100.4 °F (≥ 38 °C), hemodynamic deterioration, systolic blood pressure of < 90 mmHg with new onset or increase of vasopressor drugs, alterations in gas exchange evidenced by a decrease in PaO 2 /FiO 2 or the need for an increase in FiO 2 , changes in the characteristics and volume of respiratory secretions; (ii) suspected barotrauma (pneumothorax); (iii) suspected increased extravascular lung water; and (iv) suspected pleural effusion. Prior to CT, bedside LUS and CXR were performed. Each hemithorax was evaluated and classified as positive or negative for each of four conditions, including consolidation, interstitial syndrome, pleural effusion, and pneumothorax. Patients were not considered if they were under 18 years of age, on a ventilator for < 12 h, or with a confirmed diagnosis of pneumothorax, pulmonary condensation, acute lung edema, or pleural effusion on admission.
The treatment was determined after CT was performed.

Chest Radiography
Anterior CXR was performed with portable X-ray equipment (Siemens Polymobile, Erlangen, Germany). CXRs were assessed by two expert radiologists, well-known specialists, blinded to the LUS and CT findings. Consolidation, interstitial syndrome, pneumothorax, and pleural effusion were defined using the terminology of the Nomenclature Committee of the Fleischner Society (Table 1).

Lung Ultrasound
For LUS, a microconvex 5-9 MHz transducer was used. Ultrasonography findings were evaluated by two expert operators who were blinded to the CT and CXR findings. Patients were examined in the supine position. The lateral decubitus position was used for posterior lung surface examination. The anterior surface of each lung was defined by the clavicle, parasternal, anterior axillary lines, and by the diaphragm and divided into two areas, i.e., upper and lower. The lateral surface was defined by the anterior and posterior axillary lines and divided into upper and lower areas. Finally, the posterior lung surface was defined by the posterior axillary and paravertebral lines and divided into upper and lower areas. The apex from the supraclavicular space was scanned from the supraclavicular space. Those informing LUS were teachers in the course ultrasound in critical care at the Argentine Association of Critical Care and the University of Buenos Aires. When the ultrasound was performed, those who reported the X-ray were blinded to the ultrasound and vice versa. Only after CT scan was performed, was diagnosis confirmed and medical measures were taken.

Lung Ultrasound Definitions
Consolidation: Defined as the presence of a shredded line or tissue-like image at the PLAS point or a C-line at the pleural line. Lung areas were identified using the same anatomical landmarks as with CRX. Pneumothorax: Diagnosed when the A-line sign (only A-lines visible) and associated with the absence of lung sliding and the presence of the lung point sign in B mode and the stratosphere sign in M mode.
Pleural effusion: Determined as an anechogenic or hypoechoigenic pattern at the pleural space. It may contain isoechogenic particles or septations in inflammatory pleural diseases.
Interstitial syndrome: Defined as the presence of 3 or more B-lines simultaneously visible between two ribs in a long-axis scan.

Multiple-Detector Computed Tomography (MDCT)
MDCT was performed with a Siemens Somatom Sensation (16 slices, Erlangen, Germany). Scans were obtained in the supine position from the apex to the base of the lungs. The assessment included thin MDCT high-resolution computed tomography (MDCT-HRCT) and spiral MDCT scans. MDCT-HRCT scans were used to show diffuse lung parenchymal involvement as ground-glass opacities, septal or non-septal lines, and fibrotic changes (structural distortion). Scans were evaluated for mediastinal and pleural conditions and lung lesions (atelectasis, alveolar consolidation, and parenchymal bands) as described by the Nomenclature Committee of the Fleischner Society. Two individual radiologists, blinded to the LUS and CXR findings, assessed the MDCT-HRCT and MDCT images together, and conclusions were reached by agreement.

Ethical Aspects
The study was evaluated and approved by the Independent Ethics Committee of the German Hospital. Informed consent was obtained from the patients or their representatives. The investigators of this study complied with Law 25.326/ Habeas Data.

Statistical Analysis
The frequency and distribution of categorical variables were expressed in absolute quantity and percentages, while continuous variables were expressed as the mean ± standard deviation (SD) and interquartile range (IQR), according to the noted characteristics of the distribution. Continuous variables were compared using Student's t-test or Wilcoxon rank-sum test, depending on their distribution. For categorical variables, the chi-square test or Fisher's exact test was used. The ROC area, sensitivity, and specificity were calculated using standard formulas. A p-value < 0.05 was considered statistically significant. All analyses were performed in Stata v14 (College Station, TX, USA) and MedCalc® Statistical Software version 20.011 (MedCalc Software Ltd., Ostend, Belgium; https:// www. medca lc. org; 2021).

Results
A total of 58 patients (126 hemithoraces) were evaluated by 3 imaging techniques. Table 2 shows patient characteristics. Of the 58 patients, 34 (59%) were men, and the median age was 70 years. The mean value of Apache II was 16.4 ± 5.7, and the median value of Charlson was 4.7 (IQR, 4-7). The most frequent causes of admission to the ICU were liver transplant, with 4 (7.1%) cases; upper gastrointestinal bleeding, with 3 (5.4%); and abdominal surgery, with 3 (5.4%).
The sensitivity, specificity, and correctly classified percentages of the four pathologies by LUS and CXR are shown in Table 3. LUS performed better than CXR, exhibiting significantly higher sensitivity for all conditions. CXR had a marginally higher specificity than LUS for consolidation, interstitial syndrome, and pneumothorax.

Decision-Making
In 42/58 (72%) patients, the information provided by imaging studies led to a specific measure, while no change in patient management was made for the rest. The most frequent measures were blood culture tests and empiric therapy. Regarding those cases for whom a specific measure was taken, in 29/42 (69%) cases, the medical decision was made based on a diagnosis determined only by LUS but later confirmed by CT. In 4/42 (10%) patients, both LUS and CRX provided the right diagnosis, while the latter was the only accurate method in 2/42 (5%). In the remaining 7/42 (17%) patients, only the information provided by the CT was considered relevant to order a therapeutic and/or diagnostic intervention. Overall, in 33/42 (79%) patients, a subsequent measure was performed based on the information provided by LUS. LUS was consistent with CT in 11 of 16 patients (69%) who did not require any specific measure.

Discussion
Our findings show that LUS yielded high diagnostic accuracy for common lung conditions in critically ill patients. Indeed, LUS showed a higher percentage of correct diagnoses of consolidation, interstitial syndrome, pleural effusion, and pneumothorax than CRX as a diagnostic tool. In this study, LUS had a sensitivity and specificity of 89.7% and 94.1%, respectively, for the detection of consolidation, comparable to previous studies [5,6,10]. The same occurred with the LUS accuracy for the detection of interstitial syndrome and pleural effusion when compared with CRX; LUS was significantly better, as reported in the literature [5,6,11]. Indeed, CRX showed low sensitivity for these abnormalities, with 64.2%, 43.9%, and 58% for consolidation, interstitial syndrome, and pleural effusion, respectively [5,6,11]. Furthermore, it has also been suggested that LUS may be better than CRX in distinguishing between different types of pleural effusions on the basis of internal echogenicity, homogeneity, and pleural thickness [12]. LUS sensitivity for pneumothorax was 100%. This discrepancy with a comparable study in which LUS had a sensitivity of 75% [6] may be due to a low number of patients in our study, yielding a wide confidence interval and being insufficient to reach statistically reliable results in terms of sensitivity. Additionally, it has been shown that LUS outperforms CRX when detecting residual pneumothoraxes after drainage and that residents can learn to operate on LUS after only 2 h of training [13,14].
Replacing CXR with LUS in the ICU reduces radiation exposure, which is a substantial improvement in patient safety [15]. Indeed, the carcinogenic effects of X-rays are well known. A single chest CT scan has an actual radiation dose equivalent to 400 CXRs [16]; indeed, medical radiation from CRX and nuclear medicine scanning is the most important source of radiation exposure in Western countries [16]. In our study, 610 CRXs were performed, and some patients spent several days in the ICU with substantial radiation exposure. Routine daily CXRs for critically ill patients have been and are still the standard practice in many institutions, such as ours. Other studies have shown that the use of LUS reduced the use of CRX [17] or CRX and CT [17,18]. Brogi et al. found that the routine use of LUS contributed to a significant reduction in the number of CXRs and in costs without affecting the outcome [17]. Additionally, the pointof-care LUS protocol led to a significantly lower utilization of chest radiography, ultrasound performed by nonintensivist specialists, and CT scans, leading to lower radiation exposure, less intrahospital transportation of unstable patients, and cost savings [18].
In this study, LUS was found to be a useful method to inform intensivists on the decision to prescribe a diagnostic and/or therapeutic procedure. LUS proved to inform the decision correctly in the vast majority of patients. In 76% of patients, performing CT did not provide any additional information that would change the decision made based on the LUS findings. Agreement between LUS and CT was high; coinciding with the literature, LUS alone informed the decision-making process [19]. This saved time and provided a better use of healthcare resources by avoiding unnecessary CRXs and CTs. Similar to other studies, we also found LUS to be superior to CRX [20].
This study has some limitations that need to be acknowledged. First, we studied a small number of patients, which led to some constraints in the analysis. The location of the pathological finding and the analysis to identify factors associated with subsequent measures were both affected by the sample size. The time interval between LUS and CT was minimized but could not be fully controlled; and inter-rater reliability across different individuals was not assessed. Finally, this is a single-center study, which could limit the extrapolation of its findings to a number of institutions, especially those related to decision-making.
In conclusion, in critically ill patients, bedside LUS, which can be performed by critical care professionals, shows better diagnostic performance than chest radiography for the diagnosis of common pathologic conditions and could be an alternative to CT scans.