Characterization of solid focal pancreatic lesions ...

Scandinavian Journal of Gastroenterology. 2014; Early Online, 1–10

ORIGINAL ARTICLE

Scand J Gastroenterol Downloaded from informahealthcare.com by Helse Bergen - Haukeland universitetssykehus on 05/18/14 For personal use only.

Characterization of solid focal pancreatic lesions using endoscopic ultrasonography with real-time elastography

ROALD FLESLAND HAVRE1,2, SVEIN ØDEGAARD1,2, ODD HELGE GILJA1,2 & LARS BIRGER NESJE1,2 1

Department of Medicine, Haukeland University Hospital, Jonas Lies vei, Bergen, Norway, and 2Department of Clinical Medicine, University of Bergen, Bergen, Norway

Abstract Objectives. The aim of this study was to evaluate the diagnostic potential of strain assessment in solid focal pancreatic lesions using real-time elastography in combination with endoscopic ultrasonography (EUS). Material and methods. Forty-eight solid focal pancreatic lesions in 39 patients were included prospectively over a 3-year period and studied by EUS with real-time elastography (EUS-RTE). Lesions previously described as cystic by CT were not included. Distribution patterns of tissue strain were assessed using strain ratio (SR) measurements, continuous visual analog scale (VAS), and a visual categorical score (VCS), based on color coding of relative strain. Final diagnosis was based on histopathology, fine-needle aspiration cytology, and/or follow-up for ‡6 months. Results. The 48 lesions included 11 adenocarcinomas, 7 malignant neuroendocrine tumors (NETs), 11 benign or indeterminate NETs, 8 focal pancreatic lesions, 2 microcystic adenomas, and 9 other benign lesions. Malignant lesions had significantly higher median SR (7.05 vs. 1.56) and VAS scores (93.0 vs. 63.5) than benign lesions. A receiver operation characteristic curve analysis showed sensitivity of 67% and specificity of 71%, when using SR = 4.4 as a cut-off for malignancy. The highest SR values were found in two benign microcystic adenomas. Conclusions. EUS-RTE with SR measurements and VAS evaluation demonstrated a significant strain difference between benign and malignant lesions. However, the variation within the entities was substantial and some benign lesions presented with low strain. Benign lesions were generally characterized by a strain similar to reference tissue, whereas malignant lesions were harder. The recorded strain pattern in individual lesions must be interpreted with caution.

Key Words: endoscopic ultrasonography, focal pancreatic lesions, real-time elastography, strain quantification, strain ratio, tissue characterization

Introduction Early diagnosis is important for the prognosis of pancreatic cancer. Surgical resection is currently the only possible curative treatment, but this is an extensive procedure with some hazard and may also lead to postoperative endocrine and exocrine pancreatic insufficiency [1,2]. With improved medical imaging, an increasing number of small pancreatic lesions may be discovered [3]. Focal lesions can represent not only malignant tumors but also benign cystic or solid lesions, such as slow-growing neuroendocrine tumors

(NETs), focal changes caused by previous acute or chronic pancreatitis, intrapapillary mucinous neoplasias, serous cystic adenoma (SCAs) and mucinous adenomas. Preoperative diagnosis may be challenging and new methods for noninvasive tissue characterization of focal pancreatic lesions have been introduced with the purpose of identifying patients with early cancers and avoiding unnecessary surgery of benign lesions [4,5] In several studies, endoscopic ultrasonography (EUS) has proved to be more accurate than CT and MRI for characterization of small focal pancreatic

Correspondence: Roald Flesland Havre, Department of Medicine, Haukeland University Hospital, Jonas Lies vei 65, 5021 Bergen, Norway. Tel: +47 90842938. Fax: +47 55992950. E-mail: roald.fl[email protected]

(Received 27 January 2014; revised 10 March 2014; accepted 12 March 2014) ISSN 0036-5521 print/ISSN 1502-7708 online 2014 Informa Healthcare DOI: 10.3109/00365521.2014.905627


2

R. F. Havre et al.

lesions [6–9]. However, a combination of EUS and CT may also be useful in imaging small pancreatic lesions [10]. Real-time elastography (RTE) is a strain imaging technique, which can be combined with EUS, utilizing the normally occurring tissue movements from arterial pulsations and respiratory movements. By RTE, the ultrasound (US) images are overlaid by a semitransparent color-coded map showing the local tissue strain resulting from the impact of internal movements. This image is named an elastogram and displays the relatively harder and softer tissue areas on the basis of low and high strains, respectively [11–13]. Pathological pancreatic lesions frequently become harder than the normal pancreas, and measurement of strain differences between lesions and the surrounding tissue has been suggested as a method to improve preoperative diagnosis (Figure 1). Estimation of strain ratio (SR) provides semiquantitative information about strain variations within an elastogram, and two recent publications reported promising results for preoperative classification of pancreatic tumors based on SR measurements [14,15] To date, the accuracy of SR measurements in EUS elastography of focal pancreatic lesions has varied [14,16]. We have previously evaluated the impact of relative size and position of reference tissue and lesions on SR measurements in vitro [17]. We also validated a visual analog scale (VAS) for elastogram evaluation in a tissue-mimicking phantom [18]. VAS is generally the fastest and most intuitive evaluation method for images without any measurements or post processing. The aim of this study was to use a commercially available EUS system with strain elastography software (EUS-RTE) and different evaluation methods to prospectively estimate the relative strain in solid pancreatic lesions as compared with surrounding tissue. A particular goal was to evaluate if strain imaging could characterize different entities of pancreatic lesions and separate malignant from benign lesions. Methods The study included patients with an apparently solid focal pancreatic lesion discovered by other imaging modalities such as CT, MRI or external US, without signs of locally advanced stage or distant metastasis revealing a malignant character. Patients who had lesions with known histopathology or cystic appearance by cross-sectional imaging were not included. Patients who proved to have retroperitoneal lesions outside the pancreas were excluded from further analysis. A total of 45 patients were referred for

examination, and 39 met the inclusion criteria. All patients signed an informed consent to participate. US images and strain data were produced in real time and stored during the EUS examination. SR measurement was performed on cine-loops immediately after recording. Histopathology of surgical specimens or demonstration of malignant cells in samples from fine-needle aspirations (EUS-FNA) was considered diagnostic. Patients with negative or inconclusive EUS-FNA or without tissue sampling were followed up for at least 6 months. Focal lesions without signs of disease progression after 6 months were considered benign. The study was approved by the Regional Committee for Ethics in Medical and Health Science in Western Norway and was conducted according to the Helsinki Declaration. The study was registered as a prospective clinical trial at ClinicalTrials.gov with reference number NTC 01360411.

Equipment A Hitachi High-Vision 900 US scanner with software for RTE (version V16-04 STEP 2, Hitachi Medical Corporation, Tokyo, Japan) was used with a curvilinear Pentax EG-3870 UTK echoendoscope (HOYA Corporation, Tokyo, Japan) using US frequencies 5.0–10.0 MHz. The applied strain-imaging algorithm records differences in echo positions between consecutive radio frequency frames using the extended combined autocorrelation method [12,13]. We used the following parameter settings for RTE: Frame reject: 6, Noise reject: 4, Persistence: 3–5, and Smoothing: 2–3. VAS and categorical scores were performed at E-dynamic range level 4 (default level) [18]. If no strain signal was recorded in a defined region of interest (ROI), including lesion and reference tissue, the frame-reject level was reduced to level 3 or 2. SR measurements were recorded at E-dynamic range levels 2, 4, and 6, respectively. EUS scanning was performed from the duodenum or the stomach depending on the location of the lesion within the pancreas. The ROI for elastography was selected to include the lesion and a similar or larger area within the reference tissue. For the largest tumors exceeding a diameter of approximately 3 cm, we sometimes had to select only a part of the lesion to be able to include the reference tissue in adjacent pancreas. EUS instrumentation and strain assessment was performed by one of two doctors with long EUS experience, who had also performed EUS-RTE in 60 clinical examinations, including 24 examinations of focal pancreatic lesions prior to the study. Patients were examined as part of the diagnostic workup, and preoperative tissue sampling by EUS-FNA or

EUS-elastography of focal pancreatic lesions

3


A

B

C

Figure 1. (A) EUS elastogram of an adenocarcinoma. Color-coded strain image (left) and plain B-mode image (right) are shown. A section of the hypoechoic tumor is imaged with low strain indicating hard tissue. The SR was 17.53 in this frame, calculated from reference area B and tumor area A. (B) EUS elastogram of a hypoechoic lesion in a patient with previous acute pancreatitis is shown. The poorly demarcated hypoechoic lesion shows similar or moderately reduced strain compared to the reference tissue. SR = 1.56. (C) EUS elastogram of a highly differentiated malignant NET. The hypoechoic and sharply demarcated lesion corresponds to the area with low strain, which is indicated by blue color in the elastogram. SR = 5.47.

4

R. F. Havre et al.

Table I. Diagnostic groups and methods used for establishing final diagnosis in a study of focal pancreatic lesions with of EUS-RTE.


Adenocarcinoma Malignant NET Benign or indeterminate NET Microcystic adenoma Focal lesion in pancreatitis Other benign lesions Total

Histopathological diagnosis (+2 positive cytology)

Diagnosis based on follow up

Total

11 7 1 2 2 0 23

0 0 10 0 6 9 25

11 7 11 2 8 9 48

fine-needle biopsy was performed only if indicated for a diagnostic purpose. Strain assessment Three methods for strain assessment were applied: SR, VAS, and a visual categorical score (VCS) as described by Janssen et al. [19]. The SR is a built-in RTE functionality comparing mean strain in two separate, user-defined areas. Areas A and B were selected with similar size and distance from the US probe, since this has been shown to improve repeatability of the method [17]. Reference tissue adjacent to visible vessels or ducts was avoided. SR was calculated as mean strain in area B divided by mean strain in area A, producing a higher SR value for harder lesions with low strain. The VAS score was performed after completion of the examination by reviewing the elastography cineloops used for VCS and SR evaluation. A mark was set on a 100 mm line to indicate the suggested lesion hardness as compared to the surrounding tissue. Lesions softer than the reference tissue were scored between 0 and 50 mm, harder lesions from 50 to 100 mm. The VCS uses numbers 1, 2, and 3 to classify the distribution pattern of colors within the lesion: 1 = homogeneous distribution, 2 = nonhomogeneous by two colors, and 3 = honeycomb pattern. Order of letters A, B, and C is used for identifying a decreasing relative content of colors within the lesion: A = blue, B = green/yellow, and C = red.

more than one pancreatic lesion was examined. The GEE-analysis of SR was done on log-transformed data in order to obtain a distribution closer to normal for all entities. The analysis of VAS data was done on untransformed data. Data from SR or VAS are shown in box plots where the central line indicates the median, the box indicates interquartile range (IQR), and whiskers indicate extreme values within 1.5 the width of the IQR. For analysis of SR and VAS scores in malignant and benign lesions, we also applied the Mann–Whitney test with a significance level of p < 0.05. Correlation was done using Spearman’s rho. For evaluation of VCS, we used a 2 2 table. For data analysis and plots, we used SPSS version 19 (SPSS, Chicago, IL, USA). Results Forty-eight lesions in 39 patients, including 19 females, 20 males with a median age of 55 years (27–81 years) were prospectively included from March 2008 to June 2011. The final diagnosis was 30.0 25.0 20.0 Strain ratio

Final diagnosis

15.0 10.0

Statistical analysis Data are presented as median values with range in brackets, if not otherwise indicated. Calculation of SR for the different entities was based on median SR from individual lesions. The general estimating equation (GEE) method was applied for statistical analysis between entities. This method compensates for clustering of data. Data from two lesions in one patient were considered clustered, whereas data from two different patients were unclustered. In nine patients,

o13 p < 0.001

5.0 .0 Benign, n = 28

Malignant, n = 18

Figure 2. Box-plot showing SR of benign and malignant lesions. Line represents the median value, and the box represents IQR from 25–75 percentile. Data from malignant lesions are shown in gray. Whiskers are values up to 1.5 IQR, and values exceeding this limit are marked as outliers.


inconclusive cases were followed up without sign of malignant disease. The median follow-up time was 12 months (6–25). Patients without surgery or proven malignancy were followed up as clinically indicated by CT (13), MRI (12), Octreotide scintigraphy (5), external contrast enhanced ultrasonography (1), or clinical follow up without specific imaging for 9 and 12 months, respectively (2). Methods used for establishing final diagnosis are listed in Table I. Figure 1 shows examples of EUS elastograms of an adenocarcinoma (Figure 1A), hypoechoic lesion following acute pancreatitis (Figure 1B), and a malignant NET (Figure 1C). SR results for the malignant and benign entities are shown in Figure 2. Median SR in malignant lesions was 7.05 (3.02–27.57) and for benign lesions 1.56 (0.07–35.55), (p < 0.001). VAS results for benign and malignant lesions are shown in Figure 3. The median VAS score for malignant lesions was 93.0 (89–98) and for benign lesions 62.4 (17–97), (p < 0.001). Further analyses of SR and VAS scores in subclasses of benign and malignant lesions were performed and are shown in Figures 4 and 5, respectively. Benign lesions without cytological or histological verification are classified as “other benign lesions”. No significant strain difference was demonstrated between the entities “Benign or indeterminate NETs” or “Focal lesions in pancreatitis”, when compared to “Other benign lesions”. There was, however, a significant difference between “Malignant NETs” and “Adenocarcinomas” as compared to

100 p < 0.001 Elasticity VAS score

60

40

20

0 Malignant

Figure 3. Box-plot showing VAS evaluation of EUS strain images in benign and malignant pancreatic lesions. Data from malignant lesions are shown in gray. Box-plots are explained in legends of Figure 2.

based on histology of a surgical specimen in 21 cases (43.7%), EUS-FNA showing malignant cells in 2 cases (4.2%) and follow up in 25 cases (52.1%). EUS-FNA was performed in 15 cases (31.3%), and samples were positive for malignancy in 5 cases, negative in 8 cases and inconclusive in 2 cases. Two of the eight patients with negative EUS-FNA samples underwent surgery, one had adenocarcinoma and the other had focal pancreatitis. The remaining six patients with negative EUS-FNA and the two 40.00

Strain ratio

30.00

O5

20.00

10.00

O12

O29

= on ,n

s,

si

iti be ni g

n

le

ea t nc r pa O

th er

in io n

Fo ca ll

es

9

8 n

= ,n ET N

in ed er m

or i

nd et

M gn Be ni

=

11

7 = ,n ET

na nt N al ig

ci no ca r

Ad en o

ic r

oc

ys

tic

ad e

no m

m a,

a,

n

n

=

=

2

11

.00

M


80

Benign

5

Figure 4. Box-plot showing SR in EUS strain images by subclasses of pancreatic lesions. Data representing malignant lesions are shown in gray. Malignant NETs had significantly higher SR than benign or indeterminate NETs (p = 0.01 MW). Box-plots are explained in legends of Figure 2.

6

R. F. Havre et al.

VAS score

100 80 60 40 20

9 =

8

n n,

s, n

le

si o

at iti

ig

n

nc re

en rb th e

Fo

ca l

le

O

si o

de in or n ig Be n

=

11 = n pa in n

te r

M

m

al

in

ig

na

at e

nt

N

N

om ci n ca r no

Ad e

ET ,

a,

ET ,

n

n

=

=

11

2 = n a, om en ad cy st ic ic ro M

Figure 5. Box plot showing VAS evaluation of EUS strain images in subclasses of pancreatic lesions. Data from malignant lesions are shown in gray. Box-plots are explained in legends of Figure 2.

“Other benign lesions”. There was also a significant difference between the benign entity “Serous microcystic adenoma”, comprising only two lesions, and “Other benign lesions” using SR and VAS. In an attempt to find the optimal cut-off values between malignant and benign lesions in regard to sensitivity and specificity, a receiver operation characteristic (ROC) curve was produced for VAS and median SR. These curves display corresponding sensitivity and specificity at different cut-off levels for pancreatic malignancy (Figure 6). For SR, a cut-off value of 4.4 gave a sensitivity of 67% and specificity of 71.4%. Selecting a lower SR cut-off value of 3.05 gave a sensitivity of 94.4% and a specificity of 60.7%. Area under SR ROC-curve, ROC-AUCSR was 0.814 (0.693–0.936). For VAS, the sensitivity was 100%

and the specificity was 82% with a cut-off at VAS = 87.5. The ROC-AUCVAS was 0.869 (0.758–0.980). If the two microcystic SCAs were removed from the data analysis, the distinction between malignant and benign lesions (GEE, n = 46, ln(Median SR) and VAS) was still significant (p < 0.001). In a ROC curve analysis, the ROC-AUCSR improved to 0.860 (0.755– 0.965), providing a sensitivity of 72% and a specificity of 77% for a SR cut-off value of 4.32. For the VAS score, the distinction between malignant and benign lesions was even better with a ROC-AUC VAS of 0.891 (0.786–0.996), providing a sensitivity of 1.00 and a specificity of 0.85 for a cut-off of VAS = 87.5. There was a correlation between median SR and lesion size. Median SR and short diameter had Spearman’s rho: r = 0.453 (p = 0.001) and long

1.0 Median SR VAS score Reference line

0.8

Sensitivity


7

0

0.6

0.4

0.2 AUC SR: 0.81 (0.69–0.94), p < 0.001 AUC VAS: 0.87 (0.76–0.98), p < 0.001 0.0 0.0

0.2

0.4

0.6

0.8

1.0

1-Specificity Figure 6. ROC curve showing SR (short dotted line) and VAS (long dotted line) for differentiation between benign and malignant lesions. The curve shows changes in sensitivity and specificity for malignancy with different cut-off values of the two continuous scales. The area under the ROC curve (ROC-AUC) represents the overall accuracy of the test based on the current material.



7

Figure 7. B-mode image (right) and elastogram (left) of a serous cystic adenoma in the pancreatic body/tail are shown. Lesion size was 35 30 mm. The microcystic structure of the lesion gives a solid appearance in EUS B-mode and a homogeneous low-strain signal in the elastogram. The tumor margins are better visualized in the elastogram than in the B-mode image.

Figure 8. B-mode image (right) and elastogram (left) of an oligocystic lesion in the pancreatic neck described as a serous cystic adenoma by histology are shown. The lesion is a complex cyst with a central solid region. By CT imaging, the lesion was described as solid, and thus, included in the study. The strain image shows reproducible low-strain signal in this benign lesion.


8

R. F. Havre et al.

diameter r = 0.490 (p < 0.001). The correlation between VAS score and size was weaker. Correlation between VAS and short diameter was r = 0.286 (p = 0.049) and for VAS and long diameter was r = 0.291 (p = 0.044). We collected categorical data for 46 of 48 lesions. Lesions classified as 1A (homogeneously blue, 10 cases), 2AB (more blue than green, 18 cases), and 3AB (honeycomb pattern blue>green, 2 cases) were regarded positive for malignancy. Lesions classified as 1B (homogeneously green, 5 cases), 2BA (more green than blue, 7 cases), 2A/B/C (various representation of all colors, 3 cases), and 2BC (more green than red, 1 case) were regarded as negative for malignancy. On this basis, a 2 2 table of the VSC analysis showed sensitivity of 100%, specificity of 57%, accuracy of 74%, positive predictive value of 60%, and negative predictive value of 100%. Discussion In the present study, we found a relatively lower lesion strain with a correspondingly higher SR for malignant lesions compared to the entity of various benign lesions. Previous publications on categorical scoring systems by Giovannini et al. [20] and Janssen et al. [19] showed good sensitivities of 92.3% and 93.8%, respectively, for identification of malignant lesions. However, the specificity was only 80.0% and 65.4%, respectively, for the two categorical scoring systems. Later studies have shown diverging results on the diagnostic accuracy of elastography in pancreatic lesions using categorical visual scores for classification of these lesions [4,19,21]. In a multicenter study, including 285 patients in 13 centers, a cut-off level of 175 in a histogram of 256 hues was used in order to differentiate malignant from benign pancreatic lesions using EUS-RTE. Sensitivity, specificity, and accuracy were 96.7%, 63.8%, and 90.7%, respectively [22]. In a recent meta-analysis, the sensitivity and specificity for malignancy in EUS-RTE of focal pancreatic lesions were reported to be 95% and 69%, respectively, with a ROC-AUC close to 0.87 [23]. In another EUS-RTE meta-analysis comparing the diagnostic performance of categorical color-pattern analysis and hue histograms, the pooled visual color pattern analysis reported sensitivity of 99% and a pooled specificity of 74% compared to histogram analysis with sensitivity of 85–93% and specificity if 64–76% [24]. Our study showed a high sensitivity of 94.4% with a limited specificity of 60.7%, when using a SR malignancy cut-off at 3.05. This corresponds well with sensitivity of 95% and specificity of 69% in a recent meta-analysis [23]. In both cases, false positives are

numerous, reducing the specificity considerably. A higher portion of benign lesions may partly cause the low specificity seen in our study because patients with irresectable malignant tumors were excluded. Variations in sensitivity and specificity between different studies may also to some degree reflect differences in equipment and examination techniques. Usually, the main purpose when imaging pancreatic lesions is early identification of cancer or lesions with malignant potential in order to improve prognosis. On the other hand, preoperative imaging should ideally also reduce unnecessary surgery for benign lesions. The choice of cut-off points will influence on sensitivity and specificity in disclosing malignancy. By using a higher SR cut-off alternative of 4.4, corresponding more with studies by Dawwas et al. [16] and Iglesias-Garcia et al. [14], a more balanced relationship between sensitivity (67%) and specificity (71%) could be achieved. However, the correct cut-off level for malignancy in EUS-RTE may still be a matter for discussion. Two serous cystic adenomas had SR and VAS scores higher than the malignant lesions, indicating harder lesions. These lesions represented one microcystic serous adenoma (Figure 7) and one oligocystic serous adenoma with solid tissue components (Figure 8). Both lesions were presented repeatedly with homogeneous low strain, indicating hard tissue. The presence of small cystic structures was only visible in one of the two adenomas (Figure 8). Even if the lesion in Figure 8 revealed itself as oligocystic on EUS, it delivered a reproducible hard elastogram, and our study design was “intention to diagnose”, which is relevant and different from other studies that may have excluded lesions on the basis of histological diagnosis or preferred to only compare focal lesions in pancreatitis to adenocarcinomas [15]. Our visualization of hard tissue in SCAs corresponds to observations of microcystic adenomas made by Janssen et al., who categorized three microcystic adenomas in the same category as malignant lesions (3AB) [19]. By excluding the SCAs from the analysis, our results come closer to the accuracy of EUS elastography of solid focal pancreatic lesions described in multicenter trials [20,22]. Strain imaging depends on several factors contributing to increased tissue hardness: increased fibrosis and desmoplastic reaction. The latter is typical of malignant tumors of the breast, bowel, and pancreas [25,26]. Increased interstitial fluid pressure (IFP) may also increase tissue incompressibility. In previous studies, the IFP in several malignant tumors was higher than in normal tissues [27–29]. Increased IFP may thus be an important contributor to the low strain found in some benign and malignant pancreatic lesions.


EUS-elastography of focal pancreatic lesions Dietrich et al. found increased tissue hardness in mass forming autoimmune pancreatitis, both in the lesions and in the surrounding pancreatic tissues [30]. Mateen et al. observed low tissue strain and high shear-wave speed, indicating hard tissue, during the acute phase of pancreatic inflammation, using acoustic radiation force impulse imaging. However, follow up indicated reduction in tissue stiffness within 4–6 weeks corresponding to the resolving edema [31]. Accordingly, increased tissue hardness as imaged by EUS-RTE may be caused by other physiological mechanisms than increased fibrosis. In theory, SR measurements have the potential of being more objective for strain evaluation than the VAS score. Selection bias of frames and reference areas may influence the results. In a meta-analysis focusing on EUS-RTE for differentiation between pancreatic adenocarcinomas and pancreatic inflammatory masses, visual color pattern evaluation was not inferior to color hue analysis, but heterogeneity in diagnostic standards was identified as a source of differences in sensitivity between the included studies [24]. Limitations This study was performed as a part of the clinical workup of patients with focal pancreatic lesions without signs of invasive malignant growth. Like in most of the original publications on EUS-RTE on focal pancreatic lesions, the examiners were not blinded to patient presentation and previous imaging results. Another limitation of the study is the lack of cytological or histological result for nearly 50% of the lesions. We decided only to use tissue sampling on clinical indication and in agreement with our pancreatic surgeons, and not in lesions already planned for surgery due to concern about possible tumor seeding. In five cases (10.4%), the recorded strain in the ROI was initially inadequate for a stable elastogram. In three of these cases, adequate strain signal was obtained when lowering the frame-reject from level 5 to 3 or 2. Conclusion Strain assessment of solid pancreatic lesions using EUS-RTE showed significant differences between the entities of malignant and benign lesions. The semiquantitative SR method was not clearly superior to the visual evaluation methods (VAS and VCS) in identifying malignant lesions. All methods showed substantial variations in strain within entities, particularly within the group of benign lesions. A low SR between a lesion and the surrounding tissue strongly indicated a benign entity. However,

9

strain measurements in individuals still need to be interpreted with caution.

Acknowledgments We thank Dr Dag Hoem at Department of Surgery and Eva Fosse RN at National Centre for Ultrasound in Gastroenterology, Medical Department, Haukeland University Hospital for their contribution. We also thank Professor Geir Egil Eide at Centre for Clinical Research, Haukeland University Hospital, for counseling on statistical methods. The equipment used in this study was partially financed by donations from “Kjøpmann Olaf Rundshaugs Legat” and Olav Raagholt and Gerd Meidel Raagholts Research Foundation, Norway. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. The study was supported by MedViz (http://medviz.uib.no/), an interdisciplinary research cluster from Haukeland University Hospital, University of Bergen and Christian Michelsen Research AS.

References [1] Ishikawa O, Ohigashi H, Eguchi H, Yokoyama S, Yamada T, Takachi K, et al. Long-term follow-up of glucose tolerance function after pancreaticoduodenectomy: comparison between pancreaticogastrostomy and pancreaticojejunostomy. Surgery 2004;136:617–23. [2] Tran TC, van Lanschot JJ, Bruno MJ, van Eijck CH. Functional changes after pancreatoduodenectomy: diagnosis and treatment. Pancreatology 2009;9:729–37. [3] Klibansky DA, Reid-Lombardo KM, Gordon SR, Gardner TB. The clinical relevance of the increasing incidence of intraductal papillary mucinous neoplasm. Clin Gastroenterol Hepatol 2012;10:555–8. [4] Giovannini M, Hookey LC, Bories E, Pesenti C, Monges G, Delpero JR. Endoscopic ultrasound elastography: the first step towards virtual biopsy? Preliminary results in 49 patients. Endoscopy 2006;38:344–8. [5] Nesje LB, Varhaug JE, Husebye ES, Odegaard S. Endoscopic ultrasonography for preoperative diagnosis and localization of insulinomas. Scand J Gastroenterol 2002;37:732–7. [6] Cannon ME, Carpenter SL, Elta GH, Nostrant TT, Kochman ML, Ginsberg GG, et al. EUS compared with CT, magnetic resonance imaging, and angiography and the influence of biliary stenting on staging accuracy of ampullary neoplasms. Gastrointest Endosc 1999;50:27–33. [7] Kinney T. Evidence-based imaging of pancreatic malignancies. Surg Clin North Am 2010;90:235–49. [8] Michl P, Pauls S, Gress TM. Evidence-based diagnosis and staging of pancreatic cancer. Best Pract Res Clin Gastroenterol 2006;20:227–51. [9] Morgan KA, Adams DB. Solid tumors of the body and tail of the pancreas. Surg Clin North Am 2010;90:287–307.


10

R. F. Havre et al.

[10] Miura F, Takada T, Amano H, Yoshida M, Furui S, Takeshita K. Diagnosis of pancreatic cancer. HPB (Oxford) 2006;8:337–42. [11] Frey H. [Realtime elastography. A new ultrasound procedure for the reconstruction of tissue elasticity]. Radiologe 2003; 43:850–5. [12] Shiina T, Doyley M, Bamber JC. Strain imaging using combined RF and envelope autocorrelation processing. Paper presented at: Ultrasonics Symposium, 1996; IEEE Proceedings 1996. [13] Shiina T, Yamakawa M. Fast reconstruction of tissue elastic modulus image by ultrasound. Paper presented at: Engineering in Medicine and Biology 2005. Shanghai, China. [14] Iglesias-Garcia J, Larino-Noia J, Abdulkader I, Forteza J, Dominguez-Munoz JE. Quantitative endoscopic ultrasound elastography: an accurate method for the differentiation of solid pancreatic masses. Gastroenterology 139:1172–80. [15] Itokawa F, Itoi T, Sofuni A, Kurihara T, Tsuchiya T, Ishii K, et al. EUS elastography combined with the strain ratio of tissue elasticity for diagnosis of solid pancreatic masses. J Gastroenterol 2011;46:843–53. [16] Dawwas MF, Taha H, Leeds JS, Nayar MK, Oppong KW. Diagnostic accuracy of quantitative EUS elastography for discriminating malignant from benign solid pancreatic masses: a prospective, single-center study. Gastrointest Endosc 2012;76:953–61. [17] Havre RF, Waage JR, Gilja OH, Odegaard S, Nesje LB. Real-time elastography: strain ratio measurements are influenced by the position of the reference area. Ultraschall Med 2012;33:559–68. [18] Havre RF, Elde E, Gilja OH, Odegaard S, Eide GE, Matre K, et al. Freehand real-time elastography: impact of scanning parameters on image quality and in vitro intra- and interobserver validations. Ultrasound Med Biol 2008;34:1638–50. [19] Janssen J, Schlorer E, Greiner L. EUS elastography of the pancreas: feasibility and pattern description of the normal pancreas, chronic pancreatitis, and focal pancreatic lesions. Gastrointest Endosc 2007;65:971–8. [20] Giovannini M, Thomas B, Erwan B, Christian P, Fabrice C, Benjamin E, et al. Endoscopic ultrasound elastography for evaluation of lymph nodes and pancreatic masses: a multicenter study. World J Gastroenterol 2009;15:1587–93.

[21] Hirche TO, Ignee A, Barreiros AP, Schreiber-Dietrich D, Jungblut S, Ott M, et al. Indications and limitations of endoscopic ultrasound elastography for evaluation of focal pancreatic lesions. Endoscopy 2008;40:910–17. [22] Saftoiu A, Vilmann P, Gorunescu F, Janssen J, Hocke M, Larsen M, et al. Accuracy of endoscopic ultrasound elastography used for differential diagnosis of focal pancreatic masses: a multicenter study. Endoscopy 2011;43: 596–603. [23] Pei Q, Zou X, Zhang X, Chen M, Guo Y, Luo H. Diagnostic value of EUS elastography in differentiation of benign and malignant solid pancreatic masses: a meta-analysis. Pancreatology 2012;12:402–8. [24] Li X, Xu W, Shi J, Lin Y, Zeng X. Endoscopic ultrasound elastography for differentiating between pancreatic adenocarcinoma and inflammatory masses: a meta-analysis. World J Gastroenterol 2013;19:6284–91. [25] de Araujo VC, Furuse C, Cury PR, Altemani A, Alves VA, de Araújo NS. Desmoplasia in different degrees of invasion of carcinoma ex-pleomorphic adenoma. Head Neck Pathol 2007;1:112–17. [26] Ohtani H. Pathophysiologic significance of host reactions in human cancer tissue: desmoplasia and tumor immunity. Tohoku J Exp Med 1999;187:193–202. [27] Heldin CH, Rubin K, Pietras K, Ostman A. High interstitial fluid pressure - an obstacle in cancer therapy. Nat Rev Cancer 2004;4:806–13. [28] Lunt SJ, Kalliomaki TM, Brown A, Yang VX, Milosevic M, Hill RP. Interstitial fluid pressure, vascularity and metastasis in ectopic, orthotopic and spontaneous tumours. BMC Cancer 2008;8:2. [29] Wiig H, Noddeland H. Interstitial fluid pressure in human skin measured by micropuncture and wick-in-needle. Scand J Clin Lab Invest 1983;43:255–60. [30] Dietrich CF, Hirche TO, Ott M, Ignee A. Real-time tissue elastography in the diagnosis of autoimmune pancreatitis. Endoscopy 2009;41:718–20. [31] Mateen MA, Muheet KA, Mohan RJ, Rao PN, Majaz HMK, Roa GV, et al. Evaluation of ultrasound based acoustic radiation force impulse (ARFI) and eSie touch sonoelastography for diagnosis of inflammatory pancreatic diseases. JOP 2012;13:36–44.