Clinical review: The role of ultrasound in estimating extra-vascular lung water

The estimation of extra-vascular lung water (EVLW) is an essential component in the assessment of critically ill patients. EVLW is independently associated with mortality and its manipulation has been shown to improve outcome. Accurate assessment of lung water is possible with CT and MR imaging but these are impractical for real-time measurement in sick patients and have been superseded by single thermal dilution techniques. While useful, single thermo-dilution requires repeated calibration and is prone to error, suggesting a need for other monitoring methods. Traditionally the lung was not thought amenable to ultrasound examination owing to the high acoustic impedance of air; however, the identification of artefacts in diseased lung has led to increased use of ultrasound as a point of care investigation for both diagnosis and to monitor response to interventions. Following the initial description of B-lines in association with increased lung water, accumulating evidence has shown that they are a useful and responsive measure of the presence and dynamic changes in EVLW. Animal models have confirmed a correlation with lung gravimetry and the utility of B-lines has been demonstrated in many clinical situations and correlated against other established measures of EVLW. With increasing availability and expertise the role of ultrasound in estimating EVLW should be embedded in clinical practice and incorporated into clinical algorithms to aid decision making. This review looks at the evidence for ultrasound as a valid, easy to use, non-invasive point of care investigation to assess EVLW.


Introduction
Extra-vascular lung water (EVLW) is all the fl uid within the lungs that is present outside the vascular compartment. Its clinical correlate is pulmonary oedema and it is a function of hydrostatic pressure and pulmonary capillary permeability. In critically ill patients, high levels have been shown to be an independent predictor of outcome [1] and manipulation of the EVLW has also been shown to have a positive eff ect on outcome [2]. For many years the chest radiograph was used to diagnose changes in EVLW, but this has been shown to have a poor diagnostic accuracy [3]. Other indirect radiological esti mates of EVLW, such as CT scanning and MRI, are impractical in critically ill patients, leading to the develop ment of thermal dye techniques to allow more estima tion of EVLW at the bedside [4]. Due to the com plexity of the dye technique it has largely been super seded by the single thermo-dilution technique, which now constitutes the clinical gold standard of lung water estimation [5]. Th is technique, however, is invasive and requires regular calibrations, limiting its utility [6].
Traditionally the lung was not felt to be amenable to ultrasound due to its high air content and acoustic impedance; however, the description of common artefacts by Lichtenstein and colleagues [7] and their pathological correlates on CT scanning has led to an expansion in the use of bedside thoracic ultrasound, culminating in evidence-based consensus guidelines [8]. Th ere is now an accumulating body of evidence that lung ultrasound allows a semi-quantitative estimate of lung water and, given its ready availability and ease of use, is a useful point of care tool for this estimation. Th is paper seeks to lay out the evidence for the use of lung ultrasound in the estimation of EVLW.

EVLW measurement in critically ill patients
EVLW or EVLW indexed to body weight is an impo rtant measure of the state of the lungs in critically ill patients as a high level is independently associated with a worse clinical outcome, including duration of mechanical ventilation, ICU stay and mortality, whether associated with acute lung injury or not [1,6,9]. It also appears that manipulation of the EVLW may improve outcome in

Abstract
The estimation of extra-vascular lung water (EVLW) is an essential component in the assessment of critically ill patients. EVLW is independently associated with mortality and its manipulation has been shown to improve outcome. Accurate assessment of lung water is possible with CT and MR imaging but these are impractical for real-time measurement in sick patients and have been superseded by single thermal dilution techniques. While useful, single thermo-dilution requires repeated calibration and is prone to error, suggesting a need for other monitoring methods. Traditionally the lung was not thought amenable to ultrasound examination owing to the high acoustic impedance of air; however, the identifi cation of artefacts in diseased lung has led to increased use of ultrasound as a point of care investigation for both diagnosis and to monitor response to interventions. Following the initial description of B-lines in association with increased lung water, accumulating evidence has shown that they are a useful and responsive measure of the presence and dynamic changes in EVLW. Animal models have confi rmed a correlation with lung gravimetry and the utility of B-lines has been demonstrated in many clinical situations and correlated against other established measures of EVLW. With increasing availability and expertise the role of ultrasound in estimating EVLW should be embedded in clinical practice and incorporated into clinical algorithms to aid decision making. This review looks at the evidence for ultrasound as a valid, easy to use, noninvasive point of care investigation to assess EVLW.
patients with lung injury [2]. Fluid management aimed at a conservative as opposed to liberal fl uid strategy was associated with improved oxygenation index, lung injury score, reduced plateau pressure and increased ventilatorfree days in a pivotal study by the ARDSnet group, further highlighting the importance of limiting EVLW in the critically ill patient population [10]. Th is was late in the progression of the critical illness and contrary to data on early goal-directed therapy, suggesting that it is not just the amount of fl uid but also the timing of fl uid that is essential in the management of critically ill patients [11].
Exploring the range of methods available to predict fl uid responsiveness in critically ill patients is beyond the remit of this article but the options for monitoring changes in stroke volume and cardiac output are numer ous [12][13][14][15][16]. In contrast, the devices available to monitor EVLW in response to changes in fl uid status are limited. With a reduction in use of pulmonary artery catheters and given the complexity of dye indicator techniques, the single thermal dilution technique has become the pre dominant technique for determination of EVLW, derived from the mean transit time of the thermal signal through the lungs.
While this may represent the best compromise between practicability (versus CT/MRI scanning) and portability, it requires central venous and proximal arterial catheters and frequent recalibration for accurate EVLW measurement. Th ere may also be specifi c methodological issues in critically ill patients where assumptions about chamber ratios and the possibility of shunting and thermal sinks may reduce the accuracy of the results [17,18].
Given these problems and the importance of EVLW in this patient population, a search for more continuous, less invasive measurement is justifi ed. While other modali ties such as impedance tomography may be developing a role, only ultrasound gives a direct window on the lung. Improvements in ultrasound technology in recent years have led to smaller machines and better image quality. As a result of this, ultrasound has moved out of the radiology department to become a point of care test facilitating both rapid diagnosis and dynamic response to inter ventions. Th e role of ultrasound is now well established in acute settings like focused assessment with sonography for trauma (FAST) scanning in accident and emergency [19,20], echocardiogram in acute pulmonary embolism [21], thoracic ultrasound for pleural interventions [22] and for vascular access [23]. Th e evolving evidence base has resulted in the publication of point of care lung ultrasound guidelines [8] Ultrasound lung comets, lung rockets and the B-line pattern Th e discovery of ultrasound as a rapid, non-invasive and reproducible bedside investigative tool to assess the EVLW in critically ill patients is worth pursuing as in other similar scenarios, such as vascular access and FAST scanning, it has been shown to be a rapidly acquired skill. Th e ease of using thoracic ultrasound to measure EVLW was further reinforced by the 99% feasibility rate achieved in one of the earliest studies conducted by Lichtenstein and colleagues in establishing B-lines as markers of elevated EVLW [7].
Th e use of ultrasound has expanded further since the signifi cance of the 'comet tail' or B-line artefact was characterised. Lung in its healthy state is poorly penetrated by ultrasound due to the high acoustic impedance of air. In the diseased lung the acoustic impedance may change, allowing better penetration and resolution from ultrasound. In addition, artefacts may develop that can imply disease in the underlying lung. Aerated lung and lung with alveolar/interstitial oedema produce two distinct types of ultrasound artefacts. A repetitive horizontal feature seen in aerated lung called the A-line ( Figure 1) is replaced by a vertical narrow based artefact in lungs with alveolar/interstitial oedema called the Bline or comet tail ( Figure 2) [7]. Th e B-line was fi rst demon strated by Ziskin and colleagues in a patient with abdominal shotgun wound [24]. Areas of diff ering impedance (for example, the sub-pleural end of a thickened septum), while not visible in ultrasound, results in the beam refl ecting indefi nitely, leading to a ring down B-line artefact composed of all the microrefl ections. Lichtenstein and colleagues have shown that B-lines have a high sensitivity and specifi city in patients with diff use alveolar/interstitial oedema [7]. Since then, various studies have correlated the ultrasound evidence of elevated EVLW, that is, B-lines with radiographic and invasive methods of lung water estimation and hydrostatic pressure measure ments [25,26].
Nomenclature is important and in this area there are a wide range of terms used -B-lines, lung rockets, ultrasound comet tails. In recognition of this and the potential for confusion, a consensus document has suggested that the term B-line pattern be used [8].

Animal models using ultrasound to estimate EVLW
Th e ultimate gold standard for EVLW is gravimetric assess ment of the lungs post mortem [27]. Indeed, indicator dye and thermal dilution techniques have been validated in animal models against gravimetric methods showing a bias that could be off set for clinical use [17]. For ultrasound to be accepted as a reasonable alternative to other measures of EVLW it should undergo the same validation in animal models.
Th is validation was undertaken by Jambrik and colleagues in an oleic acid-induced model of acute lung injury (ALI) in pigs [28]. In this porcine model, oleic acid was administered intravenously in 12 animals to induce ALI while 5 animals where used as controls. Th oracic ultra sound was performed to detect B-lines using a 3.5 Hz probe at the third and fi fth intercostal spaces in the anterior axillary line in each hemithorax. Th ere was a signifi cant increase in B-lines with time after oleic acid administration. Th ere was also a signifi cant co-relation between the number of B-lines at the time of animal sacrifi ce and wet/dry ratio by gravimetric method.
While the study by Jambrik and colleagues validated Blines to measure EVLW, the more clinically relevant question was if it could detect elevated EVLW before functional impairment becomes apparent. Gargani and colleagues in an oleic acid-induced ALI model in pigs studied the evolution of B-lines in relation to lung compliance and PaO 2 /FiO 2 (partial pressure of oxygen/ fractional inspired oxygen concentration) ratios [29]. B-lines, compliance and PaO 2 /FiO 2 ratios were measured every 15 minutes for a total of 90 minutes after the administration of oleic acid in 10 pigs while 5 pigs acted as control. B-lines were measured in six sites in each hemithorax in the third and fi fth intercostal spaces using a 7.5 Hz vascular probe. Mild to moderate ALI was demon strated in histology with a signifi cant increase in the wet/dry ratio in the lungs of the oleic acid administered pigs when compared to the controls. A signifi cant increase in B-lines was demonstrated in this study at 15 minutes post oleic acid administration with associated reduction in compliance. Th e increase in B-lines consistently preceded the development of functional impairment signifi ed by a fall in PaO 2 /FiO 2 .
Th ese data suggest that a worsening B-line pattern correlates well with increases in lung water but also that B-lines develop before overt impairment in oxygenation, suggesting that this may be a useful sign to detect increases in EVLW very early and therefore a potential guide to fl uid management.

Validation of B-lines outside critical care
Th ere is a wealth of data on the validity of B-line pattern correlating with high EVLW outside of critical care. Emer gency presentations with shortness of breath, patients with known cardiac disease, fl uid overload in the context of chronic haemodialysis and in situations of extreme physiology have all been studied with lung ultrasound. While critically ill patients may have some diff erences in the underlying pathophysiology, the principles of increased lung water are similar and are a function of both hydrostatic pressures and pulmonary capillary permeability. It is therefore pertinent to examine the data from these settings.

Extreme physiology
In a study of climbers in Nepal who had a clinical diagnosis of high-altitude pulmonary oedema (HAPE) there was a clear diff erence in the number of B-lines seen when compared to control subjects without HAPE. Indeed, there was a near linear correlation between the number of B-lines and the oxygen saturations as deter mined by pulse oximetry [30] and a temporal correlation with improvement in symptoms. Similarly, in another group of 18 climbers in Nepal, the B-line count increased from sea level with increasing altitude and again showed a strong negative correlation with oxygen saturation, making the authors conclude that clinically silent HAPE occurs [31]. Increases in the B-line count have also been demonstrated in elite apnoea divers who are known to suff er with respiratory diffi culty after dives due to centralisation  of blood fl ow from the peripheries to the pulmonary circulation. Th e B-line count increased signifi cantly following dives and normalised within 24 hours in a subset [32]. In 31 'Iron man' competitors at sea level there was an increase to what was felt to be a signifi cant level in the B-line count in 23 of them immediately following the race but then decreasing 12 hours after. Th ere were strong correlations between the B-line count and measures of cardiac function, including N-terminal probrain natriuretic peptide (NT pro-BNP) [33].

Emergency department and acute cardiology patients
Th ere are numerous studies examining the role of B-line estimation in this setting (Table 1). Th e fi rst demonstration of the use of lung ultrasound to distinguish between chronic obstructive pulmonary disease (COPD) and pulmonary oedema in acutely dyspnoeic patients was demonstrated by Lichtenstein and Meziere in 1998 [34]. Sixty-six consecutive patients with acute dyspnoea, 40 of whom had pulmonary oedema and 26 of whom had exacerbations of COPD, were studied and compared to 80 controls. A B-line pattern was demonstrated in all the patients with pulmonary oedema, only two of the patients with COPD and one control patient, giving a sensitivity of 100% and specifi city of 92% [34]. Jambrik and colleagues [26] studied 121 patients admitted to a combined cardiology/pulmonology unit, performed lung ultrasound on admission and then compared the B-line count to a blinded assessment of pulmonary oedema from the admission chest X-ray. Th ey found a strong correlation between the B-line count and a validated radiological assessment of pulmonary oedema [26,35]. In a compara tive study of 72 patients, 53 with known left ventricular (LV) systolic dysfunction and 19 with normal LV func tion, patients underwent lung ultrasound and echo cardio graphy before and after exercise. Th ere were strong positive correlations between the B-line count and several measures of cardiac dysfunction, most notably the Echo-derived measure of pulmonary capillary wedge pressure [36,37]. Another analysis was performed on the data from a cohort of 340 patients admitted to a cardiology clinic in Italy [38]. Patients admitted with clinical features consistent with acute pulmonary oedema had a much higher B-line count (50 versus 20) and there was also a correlation between the count and New York Heart Association classifi cation, with higher counts being associated with a higher class [38]. Th e same group examined data on 290 patients admitted with acute dyspnoea or chest pain who were followed up to 16 months and found that B-line count was the strongest predictor of subsequent death or cardiovascular events, even above the ejection fraction [39]. Comparison to other well validated markers of acute LV impairment, such as NT pro-BNP, has also been performed. In 129 patients admitted with acute dyspnoea, a B-line count of greater than 4 was found to have a sensitivity of 81% and specifi city of 85% and a cutoff of 9 was associated with a sensitivity of 73% and specifi city of 100% [40]. Th e area under the receiver operating characteristic curve (AUROC) was 0.893 for lung ultrasound compared to 0.978 for NT pro-BNP [40].
Lichtenstein and Meziere [41] validated a protocol for assessment of acute dyspnoea patients that included a cohort of 64 patients with pulmonary oedema. Th ey used a more simplistic approach of examining the pattern of the anterior lung and defi ning it as either A-line preponderant or B-line preponderant. When validated against American Heart Association recommendations for the diagnosis of pulmonary oedema together with echo measurements, they found that anterior bilateral B-lines had a sensitivity of 97% and specifi city of 95% for diagnosing pulmonary oedema [41]. In a similar cohort of 300 consecutive patients admitted to an emergency medical unit, patients underwent lung ultrasound in eight zones (four on either side) and the presence of B-lines in more than two zones was found to have a high sensitivity (85.7%) and specifi city (97.7%) for diagnosing alveolar-interstitial syndrome [42]. In a group of 81 patients thought clinically likely to have decompensated heart failure, a modifi ed lung ultrasound protocol was performed using 11 intercostal spaces [43]. Th ere were positive correlations between the ultrasound B-line count and the radiological score, clinical score and NT pro-BNP level. Th e B-lines resolved following treatment on follow-up scans [43]. More recently, the combination of B-lines and NT pro-BNP diff erentiated acute decom pensated heart failure from both asthma and COPD in a group of 218 emergency department patients [44]. Using the combi nation gave sensitivity of 100% and specifi city of 100% [44].
Th ese data support the role of ultrasound in helping to detect increases in EVLW manifested as acute pulmonary oedema.

Haemodialysis patients
Patients with end stage renal failure on dialysis represent a stable cohort who will have signifi cant fl uid shifts around the time of their renal replacement therapy and constitute a useful cohort to study real time changes in fl uid status and lung water.
Noble and colleagues found that in 34 of 40 patients with end-stage renal failure examined pre-dialysis, at the mid-point and post-dialysis, there was a signifi cant reduction in B-lines from both pre-dialysis to mid-point and post-dialysis [45]. Th is has been complemented more recently [46] in 41 patients scanned before and after haemodialysis; there was a strong correlation between the B-line count and the accumulated weight before treatment and that lost after dialysis. All ultrasound examinations in this study occurred within a 10 minute period, demonstrating practical, real-time changes.
Mallamaci and colleagues performed chest ultrasound in 75 patients before and after haemodialysis [47]. Th ey grouped patients into three ultrasonographic categories of increasingly severe pulmonary congestion (described as mild, <14 B-lines; moderate, 14 to 30 B-lines; severe, >30 B-lines) and found that B-line count signifi cantly reduced after dialysis. Th ey also found signifi cant correlations of pre-dialysis B-lines with early left ventricular fi lling velocity, early fi lling to early diastolic mitral annular velocity ratio, pulmonary pressure, and left ventri cular end diastolic volume.
Th ese studies give the most convincing data on the cotemporaneous relationship between EVLW, ultrasonographic B-line measurement and hydrostatic pressures, supporting the role of ultrasound as a useful dynamic means of estimating the eff ects of changes in fl uid status on the lung.

Validation of B-lines in critical care
In a seminal study, Lichtenstein and colleagues [7] enrolled 282 consecutive patients (without pneumo thorax) admitted to intensive care and assessed 250 of them for alveolar-interstitial syndrome using conventional chest X-ray against ultrasound examination of the anterolateral chest wall. Alveolar-interstitial syndrome describes a heterogenous group of pathologies associated with increases in lung water (acute respiratory distress syndrome, cardiogenic pulmonary oedema). In 92 patients diagnosed as positive for the syndrome on chest radiograph, they found a pattern consistent with the syndrome in 86 patients using ultrasound (sensitivity of 93.4%). Absence of positive fi ndings on ultrasound (or confi ned to the last lateral intercostal space) was noted in 120 of 129 patients with a normal chest X-ray (specifi city of 93.0%). Whilst 29 patients were excluded owing to inconclu sive radiography, only 3 patients were excluded because of a non-diagnostic ultrasound exam. Th e authors concluded that this was the fi rst example in the literature of a relationship between the B-line pattern and alveolar-interstitial syndrome.
Whilst the role of the pulmonary artery catheter (PAC) in the management of critically ill patients has been debated in recent years [48], the measurement of pulmonary artery occlusion pressure (PAOP) remains a reasonable estimate of left heart fi lling pressures. EVLW is a function of this increase in pressure. A comparison between ultrasound fi ndings and PAOP from PAC has been made in 102 mechanically ventilated patients [49]. Ultrasound provided good correlation, distinguishing between dry interlobular septa (with a predominance of anterior A-lines on ultrasound and a low PAOP value) and alveolar-interstitial syndrome (with a predominance of anterior B-lines on ultrasound and a higher PAOP value). Further to this, they found B-predominance to be present in a wide range of PAOP values. Th us, bedside lung ultrasound could be used to help guide fl uid therapy. Ultrasound could provide a useful non-invasive surrogate of PAOP without the potential harm from a PAC. Similar strong correlation has been shown with the PiCCO device (Pulsion Medical Systems, Munich, Germany) [50]. Agricola and colleagues [50] followed 20 patients undergoing cardiac surgery and studied the assessment of interstitial pulmonary oedema, comparing chest ultrasound against chest radiography, PAC readings and the PiCCO single thermodilution method. Th ey reported a signifi cant positive linear correlation between presence of B-lines, EVLW using the PiCCO single thermodilution method and wedge pressure as measured using the pulmonary artery catheter. Th ere was also high sensitivity (90%) and specifi city (89%) of the negative test result for detection of EVLW <500 ml. Th e study did exclude patients with signifi cant lung pathology, reducing the possibility of false-positive results.

Limitations, pitfalls and future directions
Lung ultrasound is a new technique and has its limitations. It is a user-dependent technique. While previous studies have shown good inter-observer variability, it is important that this is replicated in a broader clinical setting and that training processes and competencies are validated to ensure high quality assurance. It is also important to note that B-lines can be detected in other diseases such as pulmonary fi brosis, although diff erent spacing and artefact patterns may help to diff erentiate these [51]. B-line artefacts may also be aff ected by other dynamic changes, such as the application of positive endexpiratory pressure (PEEP) and the respiratory cycle. Above all, it is vital that future research addresses the inconsistency of approaches to quantifying B-lines and attempts to examine the levels of agreement between these measured values against other gold standard techniques such as thermodilution.

Conclusion
Lung ultrasound is a readily available bedside tool and the correlation of B-lines with EVLW has been demonstrated in animals, extreme environments and a range of clinical settings showing good sensitivity and specifi city. Th e presence or absence of B-lines is a useful tool for ruling in or ruling out pulmonary oedema and alveolar interstitial syndrome in the acute setting and there is evidence of real time matching of B-line quantity with changes in EVLW and total body water. Correlation between PAOP and B-lines in critically ill patients would suggest it is a useful surrogate of left sided fi lling pressures and has a role in guiding fl uid fi lling and resuscitation.
Th e use of ultrasound in detecting EVLW should be embedded in clinical practice by the development of a protocol to enable its incorporation into a clinical practice algorithm. Th e protocol should follow the recent guidelines [8] to allow development and implementation of a standardised technique of lung ultrasound.