Lung ultrasonography as a direct measure of evolving respiratory dysfunction and disease severity in patients with acute pancreatitis

Abstract

Background

The value of lung ultrasonography in the diagnosis of respiratory dysfunction and severity stratification in patients with acute pancreatitis (AP) was investigated.

Methods

Over a 3-month period, 41 patients (median age: 59.1 years; 21 males) presenting with a diagnosis of potential AP were prospectively recruited. Each participant underwent lung ultrasonography and the number of comet tails was linked with contemporaneous clinical data. Group comparisons, areas under the curve (AUC) and respective measures of diagnostic accuracy were investigated.

Results

A greater number of comet tails were evident in patients with respiratory dysfunction (P = 0.021), those with severe disease (P < 0.001) and when contemporaneous and maximum CRP exceeded 100 mg/L (P = 0.048 and P = 0.003 respectively). Receiver-operator characteristic plot area under the curve (AUC) was greater when examining upper lung quadrants, using respiratory dysfunction and AP severity as variables of interest (AUC = 0.783, 95% C.I.: 0.544–0.962, and AUC = 0.996, 95% C.I.: 0.982–1.000, respectively). Examining all lung quadrants except for the lower lateral resulted in greater AUCs for contemporaneous and maximum CRP (AUC = 0.708, 95% C.I.: 0.510–0.883, and AUC = 0.800, 95% C.I.: 0.640–0.929).

Discussion

Ultrasonography of non-dependent lung parenchyma can reliably detect evolving respiratory dysfunction in AP. This simple bedside technique shows promise as an adjunct to severity stratification.

Introduction

Acute pancreatitis (AP) complicated by organ dysfunction has a high mortality, which ranges from 14% to 30%.^{1, 2, 3, 4} Up to 53% of all deaths occur during the first week of admission, as a result of progressive organ failure.^{5, 6} Respiratory dysfunction is the most prevalent and one of the most deleterious systemic manifestations of severe acute pancreatitis (SAP). Lung injury may be subclinical, manifest as mild hypoxaemia or, in extreme cases, severe acute respiratory distress syndrome (ARDS).^{7, 8, 9} Therefore, it is important to identify patients with developing respiratory dysfunction in order to implement appropriate monitoring and supportive measures early in the course of the disease.¹⁰

Pulmonary conditions with diffuse involvement of the interstitium and impairment of the alveolo-capillary exchange capacity can be classified under the term Alveolar-Interstitial Syndrome (AIS) and are common in the critically ill.¹¹ This syndrome includes acute pulmonary oedema, ARDS, interstitial pneumonia, exacerbation of chronic interstitial lung disease, and other miscellaneous conditions.^{9, 12, 13, 14} The use of transthoracic lung ultrasonography (LUSS) has been proposed as a useful non-invasive method for diagnosing AIS in critically ill patients, based on the detection and quantification of comet-tail artefacts, generated by the ultrasound beam reverberation.^{11, 12, 14} Although the accuracy of LUSS has been evaluated in generic critical care populations,¹⁵ the specific diagnostic features of lung sonography in patients with AP have not been examined to date.

The aim of the present study was to investigate the value of LUSS in the diagnosis of evolving respiratory dysfunction and severity stratification in patients with AP.

Methods

Study approval

The present study was performed with Research Ethics Committee approval (REC reference number: 13/SS/0136) and institutional regulatory approval (ACCORD Project No: 2013/0098). Written consent was obtained from all participants or in cases of Adults with Incapacity, the named individual responsible for their welfare.

Recruitment

Over a three-month period from September 2013 to December 2013, adult patients presenting with potential AP were identified. The initial screening criterion was a presentation serum amylase >100 U/L, following which a definite diagnosis of AP, the definition of organ failure and severity stratification were performed in accordance with the revised Atlanta guidelines for AP, published by the Acute Pancreatitis Classification Working Group in 2013.¹⁶ Patient recruitment to this study was by convenience, with no conscious or deliberate selection bias. Data collection was planned in advance and was performed prospectively.

Transthoracic lung ultrasound scans

All scans were performed by two consultant radiologists (JS, JTM) and a specialist registrar in radiology (ZAD). Bilateral intercostal LUSS were performed with participants in the supine position, after applying acoustic gel on the skin to provide an airless interface. In order to optimize imaging, the intercostal spaces were widened by raising each patient's ipsilateral arm up to or above the head level during the procedure, and the transducer was held perpendicular to the skin surface. For the purposes of the study, either a Micromaxx^® portable ultrasound system fitted with a C60/5-2 MHz transducer (SonoSite, Inc., WA, USA), an Acuson S2000™ system with a 4C1 transducer (Siemens Medical Solutions USA, Inc., CA, USA) or an Acuson Antares™ Premium Edition ultrasound system with a CH4-1 transducer (Siemens Medical Solutions USA, Inc., CA, USA) was used.

Each hemithorax was divided into anterior and lateral, upper and lower areas (Fig. 1). For each hemithorax, the anterior area was delineated between the clavicle and the diaphragm and from the parasternal to the anterior axillary line. The lateral area was delineated between the axilla and the diaphragm and from the anterior to the posterior axillary line. The upper quadrants were demarcated from the 1st to the 3rd intercostal space and the lower quadrants from the 4th to the 6th intercostal space. Each of the 8 chest areas where visualised during normal respiration. The pattern analysed was the comet-tail artefact arising from the lung–wall interface (the hyperechogenic interface between the chest wall and the lung surface), which was defined as a hyperechogenic narrow-based reverberation artefact, spreading like a laser-ray up to the edge of the screen¹⁴ (Fig. 2). For each of the 8 quadrants, the number of comet-tail artefacts present was recorded and was linked with contemporaneous, prospectively collected clinical data.

Figure 1.

a: Lateral view of the right lung. Schematic representation of the areas scanned in relation to ribs (numbered) and intercostal spaces. b: Anterior view of the lungs. Schematic representation of the areas scanned (right hemithorax) in relation to ribs and intercostal spaces. (PA: posterior axillary line; AA: anterior axillary line; PS: parasternal line; RUL: right upper lateral area; RLL: right lower lateral area; RUA: right upper anterior area; RLA: right lower anterior area)

Figure 2.

a: Sonographic image of normal lung. b: Anterior view of the lungs from a patient with AIS: Sonographic pattern of comet-tail artefacts vertical fanning out from the lung–wall interface and spreading up to the edge of the screen (B-lines)

Based on findings from previous studies,^{14, 15, 17} comet-tail artefacts may be present in dependent regions of normally aerated lungs, and can be observed in healthy patients. Therefore, in order to optimize the protocol for sensitive detection of the comet-tail artefacts, three scan zones for each participant were compared, namely: (i). All lung quadrants, (ii). All lung quadrants except for the lower lateral, and (iii). Upper lung quadrants only, and the results were analysed.

Definitions

C-reactive protein

Plasma levels of C-reactive protein (CRP) have been shown to correlate well with the presence of pancreatic necrosis and severity of AP. However due to the late peak (36–72 h after admission) admission levels of CRP may not be useful in assessing severity.¹⁸ Therefore, both CRP contemporaneous with the LUSS (highest value within the same 24-h period) and maximum CRP value during the first week of admission were used as surrogate markers of AP severity, by using 100 mg/L as the cut-off for the binary classification of the cohort.^{19, 20}

Alveolar-interstitial syndrome

Pulmonary diseases with involvement of the alveolar space and the interstitium are grouped under the term alveolar-interstitial syndrome (AIS) and include ARDS, pneumonia, acute cardiogenic pulmonary oedema, exacerbation of chronic interstitial lung disease and miscellaneous other pulmonary conditions.^{11, 14} The radiological diagnosis of AIS was based on the presence of alveolar opacities (ill-defined shadowing, confluent opacities with air bronchograms) and/or interstitial opacities (septal lines, linear, reticular, or nodular opacities) on chest X-ray (CXR), as previously proposed by Lichtenstein et al.¹⁴

Respiratory dysfunction

The worst value of PaO₂/FiO₂ for the 24-h period when each LUSS was performed was used as a metric of respiratory dysfunction. In accordance with the modified Multiple Organ Dysfunction Syndrome (MODS) score recommended in the revised Atlanta guidelines for AP,¹⁶ a threshold value of 300 was used to divide the cohort into two groups, since those with PaO₂/FiO₂ < 300 have a modified MODS score equal or greater than 2. When an arterial blood gas (ABG) measurement was not available, the partial arterial pressure of oxygen (PaO₂) was extrapolated from the pulse oximetric oxygen saturation (SatO₂) by applying the method described by Severinghaus,²¹ previously used in different settings.²² For patients receiving supplemental oxygen, the fraction of inspired oxygen (FiO₂) was estimated as described Banks et al.,¹⁶ modified as shown in Table 1.

Table 1.

Fraction of inspired oxygen for non-ventilated patients (modification of the values provided in the revised Atlanta guidelines for acute pancreatitis)

Supplemental oxygen (L/min)	FiO2 (ratio)
Room air	0.21
1–2	0.25
3–4	0.30
5–8	0.40
9–10	0.50
11–12	0.60
15	0.90

Statistical analysis

Data for continuous variables are presented as mean ± standard deviation (SD) or as median and interquartile range (IQR). Categorical variables are presented as absolute and relative frequencies. Comparisons between independent groups were performed using the Mann–Whitney U-test. Spearman's rho was used to examine correlations between pairs of non-parametric variables. All statistical tests were based on a two-sided α-value of 0.05.

For early and late scans, independent samples analysis was performed with the cohort stratified by, (i) respiratory dysfunction (PaO₂/FiO₂ < 300 versus PaO₂/FiO₂ ≥ 300), (ii) disease severity (severe versus non-severe), based on the revised Atlanta criteria for AP (modified MODS score <2 vs. score ≥2), (iii) CRP value contemporaneous with LUSS (contemporaneous CRP < 100 mg/L vs. CRP ≥ 100 mg/L), and, (iv) maximum CRP value within the first week of admission (maximum CRP < 100 mg/L vs. CRP ≥ 100 mg/L).

Receiver Operator Characteristic (ROC) curves were plotted for each comparison. Areas under the curve (AUC) and measures of diagnostic accuracy with respective 95% confidence intervals (95% C.I.) were calculated, with the use of ‘pROC’ package²³ of RStudio version 0.98.1091 (RStudio, Inc.), after identifying ROC curve thresholds by the use of Youden's index. AUC for paired ROC curves were compared with the DeLong method.²⁴ Additional statistical analyses were performed with IBM SPSS Statistics Version 22.0 (IBM Corp., Armonk, NY, USA). Figures were designed using GraphPad Prism Version 6.0 (GraphPad Software, Inc., La Jolla, CA, USA). Formal blinding of the LUSS readers was not feasible due to the nature of the investigations. The principles of the STARD (STAndards for the Reporting of Diagnostic accuracy studies) statement have been adhered to.²⁵

Results

Demographics

Forty-one patients were enrolled in the study. Based on the revised Atlanta guidelines, 32 of 41 (78%) participants had a clear diagnosis of AP, whereas the remaining 9 patients had hyperamylasaemia that did not satisfy the revised Atlanta definition of AP. All patients were included in the analysis. Of the patients with definite AP, 12 participants (37.5%) had mild AP, 15 (46.9%) had moderately severe and 5 (15.6%) had severe disease. The demographic characteristics are summarized in Table 2.

Table 2.

Demographic characteristics of study participants

	Overall Patient sample	Hyperamylasaemiaa	Mild AP	Moderate AP	Severe AP
n	41	9	12	15	5
Age (years)
Median	59.1	59.1	69.7	57	49.4
IQR	49.2–67.5	50.7–64.6	50–81.6	49.2–67.5	45.8–63.9
Gender – ratio, (%)
Males	21/41 (51.2)	5/9	6/12	8/15	2/5
Females	20/41 (48.8)	4/9	6/12	7/15	3/5
BMI
Median	28	26	27	29	32
IQR	23–32	23–28	23–30	24–35	30–33
AP Aetiology – ratio, (%)
Gallstones	19/32 (59.4)	N/A	7/12	11/15	1/5
Alcohol	6/32 (18.8)	N/A	2/12	1/15	3/5
Idiopathic	5/32 (15.6)	N/A	2/12	2/15	1/5
Other	2/32 (6.25)	N/A	1/12	1/15	0/5
N/A	9/41 (22.0)	9/9	N/A	N/A	N/A
Severity – ratio, (%)
Mild	12/41 (29.3)	N/A	16/16	N/A	N/A
Moderately Severe	15/41 (35.6)	N/A	N/A	11/11	N/A
Severe	5/41 (12.2)	N/A	N/A	N/A	5/5
N/A	9/41 (22.0)	9/9	N/A	N/A	N/A
Amylase (IU/L)b
Median	507	145	1045	642	669
IQR	230–1143	138–197	434–1435	324–2314	654–1143
CRP (mg/L)c
Median	66	13	29	154	133
IQR	24–146	6–26	12–71	46–241	119–146
APACHE II scored
Median	10	12	9	10	17
IQR	8.5–16.5	9–17	6–12	9–15	10–31
SIRSe
Ratio (%)	23/41 (56.1%)	6/9	6/12	6/15	5/5
Modified MODS score
Median	1	2	1	1	3
IQR	1–2	1–3	1–2	1–2	2–8
PaO2/FiO2ratiof
Median	328	360	360	319	223
IQR	282–361	268–390	333–390	291–337	95–223

BMI, body mass Index; LOS, length of hospital stay; CRP, C-reactive protein; MODS, multiple organ dysfunction syndrome; APACHE, Acute Physiology and Chronic Health Evaluation score; PaO₂, partial arterial pressure of oxygen; FiO₂, fraction of inspired oxygen; IQR, interquartile range; N/A, not applicable.

^aParticipants who did not meet the criteria for acute pancreatitis, according to the 2012 revised Atlanta guidelines.

^bWorst value of serum amylase on the first 24 h of admission.

^cWorst value of CRP on the first 24 h of admission upon recruitment.

^dBased on most extreme values of the first 24 h of admission.

^eBased on most extreme values of the first 24 h of admission.

^fAs defined in methods section.

Fifty-one LUSS were performed, of which 33 (64.7%) were performed early in the disease course, and 18 (35.3%) were performed at a late stage. With the exception of one patient who underwent an early scan on day 5 and a late scan on day 11 after recruitment, scans performed within 4 days from patient recruitment were defined as early scans and those performed later than 4 days were defined as late. Early scans were performed after a median of 3 days from the initial finding of elevated amylase (IQR: 1.5–4 days), and late scans after a median of 6 days (IQR: 5–9.3 days). With regard to onset of symptoms, early scans were performed after a median of 5 days (IQR: 3–7 days), and late scans after a median of 8.5 days (IQR: 6–12.3 days). Twenty-three participants (53.7%) underwent an early scan only, 8 (22%) underwent a late scan and 10 (24.4%) underwent both. One late scan performed on day 4 after recruitment (on a participant who underwent an early scan on day 1) was excluded from further analysis.

Early scans

In patients with respiratory dysfunction (n = 7), a greater number of comet tails were evident on LUSS when compared to patients without respiratory dysfunction (n = 26), both when all lung quadrants except for the lower lateral were examined, and when upper quadrants alone were considered (P = 0.030 and P = 0.021 respectively). When all lung quadrants were taken into consideration the difference did not reach statistical significance (P = 0.067). Conversely, when contemporaneous CRP was examined, borderline significant difference was shown with the methods using all lung quadrants and all quadrants except for the lower lateral (P = 0.048 for both methods), but only a tendency for the method using only upper quadrants (P = 0.074). When severity status was used as the defining parameter for the cohort, patients with severe AP (n = 5) had a greater number of comet tails than others (n = 28), by all three methods. Lastly, the number of comet tails was significantly different by all three methods when maximum CRP value of the first week of admission was examined. The number of comet tails identified by each method is summarized in Table 3 and group comparisons are depicted in Fig. 3.

Table 3.

Descriptive summary of the number of comet tails identified by each method (early scans)

	Respiratory dysfunction
PaO2/FiO2 < 300			PaO2/FiO2 ≥ 300			Sub-group comparisona P-value	Correlationb with PaO2/FiO2 (P-value)
n	Median	IQR	n	Median	IQR
All lung quadrants	7	9	7–10	26	2	1–8	0.067	−0.456 (0.008)
All lung quadrants except for the lower lateral	7	7	5–9	26	2	0–6	0.030	−0.470 (0.006)
Upper lung quadrants	7	5	3–9	26	1	0–3	0.021	−0.409 (0.018)

	Disease severity
Severe			Non-severe			Sub-group comparisona P-value	Correlationb with modified MODS score (P-value)
n	Median	IQR	n	Median	IQR
All lung quadrants	5	11	10–20	28	2	1–8	0.001	0.388 (0.026)
All lung quadrants except for the lower lateral	5	10	9–13	28	2	0–6	0.001	0.426 (0.014)
Upper lung quadrants	5	8	7–9	28	1	0–3	<0.001	0.396 (0.023)

	Contemporaneous CRP
≥100			<100			Sub-group comparisona P-value	Correlationb with CRP on the day of LUSS (P-value)
n	Median	IQR	n	Median	IQR
All lung quadrants	13	9	6–10	20	2	1–8	0.048	0.406 (0.019)
All lung quadrants except for the lower lateral	13	6	2–9	20	2	0–5	0.048	0.354 (0.043)
Upper lung quadrants	13	4	1–6	20	1	0–3	0.074	0.306 (0.083)

	Maximum CRP
≥100			<100			Sub-group comparisona P-value	Correlationb with maximum CRP (P-value)
n	Median	IQR	n	Median	IQR
All lung quadrants	20	8	4–10	13	2	0–2	0.005	0.395 (0.023)
All lung quadrants except for the lower lateral	20	6	2–9	13	0	0–2	0.003	0.391 (0.024)
Upper lung quadrants	20	3	1–6	13	0	0–2	0.040	0.323 (0.066)

^aIndependent samples Mann–Whitney U test.

^bSpearman correlation coefficient.

Figure 3.

Scatterplot of the total number of comet tails per patient group as determined by: a. The respiratory dysfunction status (PaO₂/FiO₂ < 300 vs. PaO₂/FiO₂ ≥ 300). Red lines represent median and IQR for each patient group. A: PaO₂/FiO₂ ≥ 300; B: PaO₂/FiO₂ < 300; ns: P ≥ 0.05; *: P < 0.05 – b. The severity status (Severe vs. Non-severe). Red lines represent median and interquartile range for each patient group. ****: P < 0.001 – c. Contemporaneous CRP – d. Maximum value of CRP. **: P < 0.01

With respect to respiratory dysfunction and disease severity, the AUC was greater for the method examining upper lung quadrants alone (AUC = 0.783, 95% C.I.: 0.544–0.962, P = 0.023 and AUC = 0.996, 95% C.I.: 0.986–1.000, P < 0.001, respectively). Pairwise AUC comparisons between methods examining all quadrants, all quadrants except the lower lateral and upper quadrants alone, did not reveal superiority for either variable of interest (Fig. 4).

Figure 4.

Receiver operator characteristic curves of the number of comet tails by: a. Respiratory dysfunction status – b. Disease severity status – c. Contemporaneous CRP – d. Maximum value of CRP

The greatest AUC for contemporaneous and maximum CRP resulted form examining all lung quadrants except for the lower lateral (AUC = 0.708, 95% C.I.: 0.510–0.883, P = 0.047, and AUC = 0.800, 95% C.I.: 0.640–0.929, P = 0.004 respectively), as depicted on Fig. 4. For contemporaneous CRP, pairwise comparison did not reveal superiority of either of the three methods whereas for maximum CRP, examining all quadrants except for the lower lateral was superior to upper quadrants alone (P = 0.011).

For each method, the derived threshold for the number of comet tails, the corresponding sensitivity (SE) and specificity (SP), positive (PPV) and negative predictive values (NPV), along with respective 95% C.I., as well as the correlation coefficients of the number of comet tails with PaO₂/FiO₂, modified MODS score, contemporaneous and maximum CRP are presented in Table 4.

Table 4.

Areas under the curve, thresholds for the number of comet tails and parameters of diagnostic accuracy for each method, with regard to respiratory dysfunction, disease severity status and CRP. In order to examine the diagnostic association between the scan findings and each dichotomous variable, a value equal to or greater than the respective threshold of comet tails has been applied

		AUC (95% C.I.)	Threshold	Sensitivity (ratio; 95% C.I.)	Specificity (ratio; 95% C.I.)	PPV (ratio)	NPV (ratio)	Likelihood ratio
Respiratory dysfunction	All lung quadrants	0.728 (0.481–0.926)	7	0.857 (6/7; 0.571–1.000)	0.654 (17/26; 0.462–0.809)	0.400 (6/15)	0.944 (17/18)	2.5
	All lung quadrants except for the lower lateral	0.767 (0.544–0.948)	5	0.857 (6/7; 0.571–1.000)	0.692 (18/26; 0.500–0.846)	0.429 (6/14)	0.947 (18/19)	2.8
	Upper lung quadrants	0.783 (0.544–0.962)	3	0.857 (6/7; 0.571–1.000)	0.731 (19/26; 0.539–0.885)	0.462 (6/13)	0.950 (19/20)	3.2
Severity	All lung quadrants	0.968 (0.893–1.000)	9	1.000 (5/5; 1.000–1.000)	0.821 (23/28; 0.679–0.964)	0.714 (5/7)	1.000 (26/26)	5.6
	All lung quadrants except for the lower lateral	0.961 (0.868–1.000)	9	0.800 (4/5; 0.400–1.000)	0.964 (27/28; 0.893–1.000)	0.800 (4/5)	0.964 (27/28)	22.2
	Upper lung quadrants	0.996 (0.986–1.000)	6	1.000 (5/5; 1.000–1.000)	0.964 (27/28; 0.893–1.000)	0.833 (5/6)	1.000 (27/27)	27.8
CRP contemporaneous with early USS	All lung quadrants	0.706 (0.494–0.892)	8	0.692 (9/13; 0.462–0.923)	0.750 (15/20; 0.550–0.950)	0.643 (9/14)	0.789 (15/19)	2.8
	All lung quadrants except for the lower lateral	0.708 (0.510–0.883)	5	0.692 (9/13; 0.462–0.923)	0.750 (15/20; 0.550–0.900)	0.643 (9/14)	0.789 (15/19)	2.8
	Upper lung quadrants	0.689 (0.513–0.889)	3	0.615 (8/13; 0.308–0.846)	0.750 (15/20; 0.550–0.900)	0.615 (8/13)	0.750 (15/20)	2.5
Maximum CRP of the first 7 days of admission	All lung quadrants	0.787 (0.610–0.937)	4	0.750 (15/20; 0.550–0.900)	0.846 (11/13; 0.615–1.000)	0.882 (15/17)	0.689 (11/16)	4.9
	All lung quadrants except for the lower lateral	0.800 (0.640–0.929)	3	0.650 (13/20; 0.450–0.850)	0.846 (11/13; 0.615–1.000)	0.867 (13/15)	0.611 (11/18)	4.2
	Upper lung quadrants	0.714 (0.529–0.864)	3	0.550 (11/20; 0.350–0.750)	0.846 (11/13; 0.615–1.000)	0.846 (11/13)	0.550 (11/20)	3.6
CXR diagnosed AIS	All lung quadrants	0.859 (0.705–0.980)	8	0.733 (11/15; 0.467–0.933)	0.923 (12/13; 0.769–1.000)	0.917 (11/12)	0.750 (12/16)	9.5
	All lung quadrants except for the lower lateral	0.818 (0.639–0.949)	3	0.733 (11/15; 0.467–0.933)	0.846 (11/13; 0.615–1.000)	0.846 (11/13)	0.733 (11/15)	4.8
	Upper lung quadrants	0.810 (0.631–0.946)	3	0.667 (10/15; 0.400–0.867)	0.923 (12/13; 0.769–1.000)	0.909 (10/11)	0.706 (12/17)	8.7

AUC, area under the curve; 95% C.I., 95% confidence interval, ×2000 bootstrap; PPV, positive predictive value; NPV, negative predictive value; N/A, not applicable.

Late scans

None of the 19 patients who underwent LUSS late in the course of the disease had concurrent respiratory dysfunction, therefore it was not feasible to examine the diagnostic accuracy of late scans in this context. No significant correlation between the number of late LUSS comet tails and either corresponding disease severity (contemporaneous modified MODS score), respiratory dysfunction (PaO₂/FiO₂), disease severity on admission (as defined by the revised Atlanta criteria), or contemporaneous and maximum CRP was discovered, regardless of the method used (data not presented).

X-rays

Twenty-eight of the 33 participants (84.8%) who underwent an early LUSS, also had at least one prior CXR as part of their routine clinical management, and 15 (53.6%) of these showed AIS. Although LUSS were performed a median of 2 days (IQR: 1–3 days) after the CXR, a satisfactory diagnostic association was demonstrated between CXR diagnosed AIS and LUSS findings (Table 4).

Discussion

Respiratory dysfunction is the most frequent and clinically relevant manifestation of extra-pancreatic organ injury in AP.^{6, 7, 26} In patients with SAP, it constitutes a major determinant of mortality and often precedes renal, cardiovascular and hepatic dysfunction.^{7, 27, 28} Several hypotheses have been explored and several associated inflammatory mediators and signalling pathways have been identified regarding the underlying pathophysiology,^{8, 29, 30} however the exact mechanisms remain incompletely understood. The spectrum of sequelae of AP-associated respiratory dysfunction ranges from subclinical hypoxaemia to severe ARDS,^{31, 32} and radiologically evident complications such as atelectasis, pleural effusion, and pulmonary infiltrates and/or oedema may frequently be present, amongst others.^{33, 34} Nevertheless, hypoxaemia may precede radiological findings and has been described as an early indicator of pathological pulmonary changes.¹⁰ Consequently, a timely and accurate detection of evolving respiratory dysfunction may aid in identifying high risk patients and may expedite the implementation of appropriate management.

Although the lung parenchyma has historically been considered not amenable to assessment by ultrasound, lung sonography has undergone rapid development and its diagnostic potential has improved in recent years.¹³ Several studies exploring the value of LUSS in the diagnosis of AIS in acutely ill patients have emerged,^{11, 12, 14, 15, 35} with reports of high sensitivity and specificity.^{11, 14, 36} The sonographic diagnosis of AIS relies on the detection of multiple and diffuse comet tail B lines,¹² described as vertical artefacts fanning out from the lung–wall interface and spreading up to the edge of the screen, resulting from thickened interlobular septa and extravascular lung water.¹⁴ In addition, LUSS has been proposed as advantageous to other imaging techniques; bedside X-rays pose diagnostic challenges^{14, 15} whereas CT can be time-consuming and hazardous for critically ill patients requiring transfer,^{11, 37} and is associated with significant radiation exposure.

Stratifying severity in AP is challenging. Early stratification into moderately severe (transient organ dysfunction) and severe AP (persistent organ dysfunction) disease groups for clinical and academic purposes is bound by limitations inherent in the definitions, such as the requirement for a period of at least 48 h before a definite conclusion about severity has been reached. Similar to several previous classification systems,^{31, 38, 39, 40} the revised Atlanta guidelines use the long established PaO₂/FiO₂ ratio as a measure of respiratory dysfunction, which although undoubtedly invaluable, constitutes a relatively non-specific prognostic marker, a fact already noted by Ranson et al. four decades ago.⁴¹ The introduction of transthoracic lung ultrasound as a method supplementary to PaO₂/FiO₂ would aim to increase prognostic efficiency in AP. The hope is that it may potentially provide a valid and accurate measure of respiratory dysfunction and severity early in the disease course, by differentiating between established pathological pulmonary changes and transient physiological impairment. If validated, this non-invasive, easily repeatable bedside method would allow for a more timely patient stratification in comparison to and/or complementary to existing scores.

The findings of the present study demonstrate a useful diagnostic association of early LUSS with disease severity and respiratory dysfunction, for all three methods applied, particularly when all lung quadrants excluding the lower lateral and when upper quadrants only were examined. More comet tail artefacts are present in patients with respiratory dysfunction, severe disease and/or elevated CRP, and thresholds for the number of comet-tails have been proposed. The routine use of LUSS in the early stage of AP appears to be attractive as an initial proposition for further research, as a non-invasive, easily repeatable bedside adjunct that provides an accurate evaluation of disease severity and respiratory dysfunction.

The present study has several potential limitations. The cohort size is relatively small, however as a pilot evaluation of this diagnostic approach it is considered adequate. The number of late scans performed is insufficient to comment on its utility in the late phase of the disease and the small number of patients who underwent both early and late scans (n = 10) has not allowed for a valid comparison between these two groups. However, given that organ dysfunction and mortality are early events in AP, this fact does not obviate the utility of this approach, and a possibly slower resolution of the pathological pulmonary changes in comparison to the anticipated clinical recovery of patients with non-severe AP may raise questions on the potential application of late scans in monitoring disease progress.

Moreover, the inter-observer variability of scan findings has not been examined, and despite the fact that satisfactory results are already available from previous studies,³⁵ such data would have provided useful information on a method generally characterized by operator dependency. With regard to methodology, the convention used to extrapolate PaO₂/FiO₂ from SatO₂ may have introduced unknown error in the analysis, as the Severinghaus equation has not been validated for calculation of the modified MODS score in AP patient cohorts. Nevertheless, SatO₂ has previously been used for similar approximations, with reasonable accuracy.^{42, 43} Additionally, the timing of CXRs for study participants was not altered from ordinary clinical care. LUSS were in general performed after CXRs, therefore the accuracy of the association between the findings of the two imaging methods may be confounded by this discrepancy.

The fact that early scans were performed a median of 3 days after the first finding of an elevated serum amylase, may have concealed an even greater prognostic association with early respiratory dysfunction. The exact prognostic features of early LUSS deserve further investigation and future studies are likely to benefit from standardizing the timing of scans within the first 48 h. Nevertheless, the fact that scans performed very early may carry a high false negative ratio has to be taken into consideration, as at that stage developing pulmonary involvement may not yet be identifiable with this method.

In conclusion, ultrasonography of non-dependent lung parenchyma can reliably detect evolving respiratory dysfunction in AP, and shows promise as an adjunct to severity stratification. This simple, easily repeatable bedside test warrants further validation in a larger cohort.

Meetings

Results from this study have been presented in the following meetings:

• 1.Annual Scientific Meeting of the Pancreatic Society of Great Britain and Ireland (26th–28th of November 2014) – Poster and short oral presentation
• 2.Alpine Liver and Pancreatic Surgery Meeting 2015 (ALPS 2015 – 4th–8th of February 2015) – Oral presentation
• 3.27th European Congress of Radiology (4th–8th of March 2015) – e-poster
• 4.Digestive Disorders Federation 2015 meeting (DDF 2015–22nd–25th of June 2015) – Oral presentation

Sources of funding

No external funding was received for this study.

Conflicts of interest

None declared.