Recent studies have revealed that contrast-enhanced harmonic endoscopic ultrasonography (CEH-EUS) is beneficial in the differential diagnosis of malignant neoplasms of the pancreas and gallbladder from benign masses, in terms of the evaluation of microvasculature and real-time perfusion. In this study, we aimed to prove the clinical value of CEH-EUS in the differential diagnosis of pancreatic and gallbladder masses by direct comparison with that of conventional EUS.
We reviewed the sonographic images and medical information of 471 patients who underwent conventional EUS and CEH-EUS for the diagnosis of pancreatic and gallbladder masses at a single medical center (Severance Hospital, Seoul, Korea) between March 2010 and March 2016.
The enhancement pattern of CEH-EUS of the pancreatic solid masses showed higher sensitivity and specificity in differentiating pancreatic adenocarcinoma and neuroendocrine tumors (82.0% and 87.9% for pancreatic adenocarcinoma and 81.1% and 90.9% for neuroendocrine tumors, respectively), and the area under the receiver operating characteristic curves was higher than that of conventional EUS. The enhancement texture of CEH-EUS of the gallbladder masses showed a higher sensitivity in differentiating malignant masses than that of conventional EUS; however, the difference between the areas under the receiver operating characteristic curves was not statistically significant.
CEH-EUS can complement conventional EUS in the diagnosis of pancreatic and gallbladder masses, in terms of the limitations of the latter.
According to the statistical research announced by the Ministry of Health and Welfare Korea in March 2015, the crude incidence rate of pancreatic cancer and gallbladder cancer has increased gradually since 1999: from 6.4/100,000 to 10.2/100,000 (pancreatic cancer) and from 5.5/100,000 to 10.7/100,000 (gallbladder cancer). Compared with those of other digestive tract cancers, their 5-year survival rates are relatively low: 8.8% and 28.3%, respectively. Therefore, early detection of these cancers and distinguishing them from benign masses are gaining importance.
Endoscopic ultrasonography (EUS) is widely used in diagnosing pancreatic and gallbladder diseases because of its higher spatial resolution than other imaging methods [
Recently, CEH-EUS has been used to characterize solid pancreatic cancer and gastrointestinal stromal tumors [
A recent meta-analysis on contrast-enhanced EUS that analyzed reports on both contrast-enhanced Doppler and CEH-EUS showed that this method can differentially diagnose pancreatic adenocarcinomas with a pooled sensitivity and specificity of 94% and 89%, respectively [
Therefore, we aimed to prove the clinical value of CEH-EUS in the differential diagnosis of pancreatic and gallbladder solid masses using a multi-parameter analysis in this study.
We retrospectively reviewed the sonographic images and medical information of 471 patients who underwent conventional EUS and CEH-EUS for the diagnosis of pancreatic and gallbladder masses at a single medical center (Severance Hospital, Seoul, Korea) between March 2010 and March 2016. All endoscopic sonographies were performed by five pancreatobiliary endoscopy specialists (SWP, SYS, SMB, MJC and JYP), and each image was reviewed by two of the endoscopic physicians and blindly reviewed by one medical doctor (GLL).
A radial echo-endoscope developed for CEH-EUS (GFUE260; Olympus Medical Systems Co., Tokyo, Japan) and a low acoustic power setting (mechanical index: 0.2) were used. Ultrasonography image analysis was performed using an Aloka ProSound Alpha-10 system (Aloka Co., Tokyo, Japan). After fundamental B-mode EUS was used to investigate the masses, the setting was changed to the extended pure harmonic detection mode, which combines the filtered fundamental and second harmonic component frequencies with a transmitting frequency of 5–7.5 MHz. Thereafter, 2.5 mL of the contrast agent was injected into the antecubital vein in a bolus fashion through a 20-gauge intravenous cannula, followed by flushing with 5 mL of normal saline. SonoVue (Bracco) was used for the contrast agents. One ampule (2.5 mL) of SonoVue contains 25 mg lyophilized sulfur hexafluoride powder. The vascular structures were assessed in real time by examining continuous 0–90-s images after contrast agent injection. Early enhancement patterns were shown from 30 to 45 s of the arterial phase, and late enhancement patterns were shown from 60 to 120 s of the venous phase.
The echogenic patterns of conventional EUS were classified into four categories: anechoic, hypoechoic, isoechoic, and hyperechoic. For pancreatic solid masses, these echogenic patterns were defined by comparing them with those of the normal pancreatic parenchyma. For gallbladder masses, these echogenic patterns were defined by comparing them with those of the gallbladder wall.
The enhancement patterns of CEH-EUS were classified into four categories: nonenhancement, hypoenhancement, isoenhancement, and hyperenhancement. These enhancement patterns were also defined by comparing them with those of the normal pancreatic parenchyma (in pancreatic solid masses) and those of the gallbladder wall (in gallbladder masses).
Other parameters that could be described using the sonographic images, such as duct dilatation, tumor size, tumor demarcation, tumor marginal irregularity, hypoechoic foci, and focal wall thickness, were obtained and used for the analysis.
All analyses were performed using the SPSS v. 20 (SPSS Inc., Chicago, IL, USA). First, we compared conventional EUS and CEH-EUS according to their sensitivities and specificities that were calculated depending on the EUS findings. Their discrimination abilities were evaluated using the classification table. Thereafter, we determined the statistically significant variables and obtained the receiver operating characteristic (ROC) curve and area under the curve (AUC) using logistic regression with the other parameters. They were calibrated with the Hosmer-Lemeshow test. Using a pairwise comparison of the AUCs, the clinical value of CEH-EUS was determined.
Among the 471 patients reviewed, pancreatic masses were diagnosed in 279 patients and gallbladder masses in 192 patients. Among the 279 patients with pancreatic masses, 72 patients with cystic neoplasm of the pancreas were excluded. Among the 192 patients with gallbladder masses, 47 patients with gallbladder sludge not enhanced on CEH-EUS were excluded in the analysis.
The sensitivity and specificity for ductal adenocarcinoma were 82.0% and 87.9%, respectively. Conversely, those for neuroendocrine tumor were 81.1% and 90.9%, respectively.
The diagnostic ability can be improved by considering the enhancement texture and echogenic texture together with the enhancement or echogenic pattern. When ductal adenocarcinoma was classified as having hypoenhancement and inhomogeneous patterns on CEH-EUS, the sensitivity and specificity were estimated as 85.7% and 89.4%, respectively. When ductal adenocarcinoma was classified as having hypoechoic and inhomogeneous patterns using conventional EUS, the sensitivity and specificity were estimated as 52.1% and 78.74%, respectively. When neuroendocrine tumors were classified as having hyperenhancement and homogeneous patterns using CEH-EUS, the sensitivity and specificity were 85.3% and 91.0%, respectively. When the tumors were classified as having hypoenhancement and homogeneous patterns using conventional EUS, the sensitivity and specificity were 52.1% and 84.9%, respectively.
To evaluate the clinical value of CEH-EUS, we analyzed the enhancement patterns and echogenic patterns with other parameters that we obtained from EUS. To diagnose ductal adenocarcinoma using the sonographic images, we analyzed these parameters using logistic regression to determine which variables were statistically significant (
Using a pairwise comparison of the ROC curves for ductal adenocarcinoma, the difference between the areas under the ROC curves (AUROCs) was 0.0602, which was statistically significant (
When we analyzed the parameters from conventional EUS and CEH-EUS together, the enhancement pattern of CEH-EUS was the single powerful parameter; thus, the echogenic pattern on conventional EUS lost its power and was no longer significant in the diagnosis.
With the same methods used for ductal adenocarcinoma, we repeated the analysis for neuroendocrine tumors (
As mentioned before, the sensitivity for neuroendocrine tumors with a hypoechoic pattern was too low. Therefore, the hypoechoic pattern was not statistically significant at this time. Duct dilatation, tumor size, tumor demarcation, tumor marginal irregularity, and hyperenhancement pattern were statistically significant in the univariate analysis.
Using a pairwise comparison of the ROC curves for neuroendocrine tumor, the difference between the AUROCs was 0.0744, which was statistically significant (
First, we focused on the differentiation of malignant masses from benign masses. There were 35 cases of malignant masses consisting of adenocarcinomas and carcinosarcomas and 110 cases of benign masses consisting of adenomas, adenomyomas, adenomyomatosis, cholesterol polyps, fundic gland polyps, pyloric gland polyps, hyperplastic polyps, cholesterolosis, and cholecystitis.
Unlike the pancreatic solid masses, the gallbladder malignant masses did not show any specific enhancement or echogenic patterns (
For further evaluation of the value of CEH-EUS, we analyzed the texture patterns with other parameters, including sex, age, hypoechoic foci, tumor size, multi-lobulated appearance, and number of lesions, using logistic regression (
To compare the two groups directly, we obtained the following predicted probability equations for each and their ROC curves (
Predicted probability equation with CEH-EUS: 1 / (1 + exp (-A)), where A = -6.760 + (2.718 × Hypoechoic foci) + (0.104 × Mass size) + (3.457 × Inhomogeneous texture in CEH-EUS).
Using a pairwise comparison of the ROC curves for malignant gallbladder masses, the difference between the AUROCs was 0.0270, which was not statistically significant (
Thereafter, we further expanded our study to determine the clinical value of CEH-EUS in the differentiation of the surgical indication of gallbladder masses. We classified all gallbladder masses into two groups: neoplasm group, which requires surgery, and non-neoplasm group, which does not require surgery. The neoplasm group consisted of adenomas, adenocarcinomas, and carcinosarcomas, while the non-neoplasm group consisted of the other masses.
As shown in the malignant mass evaluation, the neoplasm and non-neoplasm groups also did not show any specific echoic or enhancement patterns. Therefore, we categorized them according to their vascular and echogenic textures: homogeneous or inhomogeneous. The neoplasms showed mostly an inhomogeneous enhancement pattern (80%); however, the non-neoplasms were not characterized by CEH-EUS. The sensitivity and specificity for the neoplasms with an inhomogeneous enhancement pattern were only 40.4% and 85.5%, respectively. Conversely, the non-neoplasms on conventional EUS at this time mostly showed a homogenous echoic pattern; however, the neoplasms did not show a specific pattern. The sensitivity and specificity for the neoplasms with an inhomogeneous echoic pattern were 60.9% and 82.0%, respectively, which imply that conventional EUS is still inappropriate as a diagnostic tool.
When analyzed with other parameters, only age and hypoechoic foci were the only statistically significant variables in both groups (
To prove the clinical value of CEH-EUS in the differential diagnosis of pancreatic and gallbladder solid masses, we designed the study with following two aims: to compare the (1) sensitivity and specificity and (2) the ROC curves in the diagnosis. In this study, we proved the superiority of CEH-EUS in the differential diagnosis of pancreatic masses over conventional EUS with higher sensitivity, specificity, and AUROC for ductal adenocarcinoma and neuroendocrine tumor. For gallbladder masses, CEH-EUS showed its powerful value with a high AUROC in diagnosing malignant masses; however, its superiority over conventional EUS was not proven.
In the comparison of the diagnostic ability of the enhancement pattern and echogenic pattern of the pancreatic solid masses, the enhancement pattern showed higher sensitivity (82.0%) and specificity (87.9%) for ductal adenocarcinoma than the echogenic pattern (relatively low sensitivity [49.0%] and high specificity [93.3%]). Previous studies revealed that only with the enhancement pattern can the sensitivity and specificity be improved up to 95% [
After considering the texture with echogenic and enhancement patterns, CEH-EUS was still better in diagnosing ductal adenocarcinoma and neuroendocrine tumor than conventional EUS; however, the sensitivity and specificity were not as high as those in previous studies that reported nearly 95% for both. Therefore, we analyzed the parameters that physicians actually consider in the diagnosis together. In the logistic regression, the AUROC of CEH-EUS was 0.949 for ductal adenocarcinoma and 0.945 for neuroendocrine tumor. Considering that a model with AUC >0.75 is usually considered a powerful predictable model, we can conclude that the predictive model with the parameters of CEH-EUS is very powerful and rather precise. Although the model was not validated externally and the total patient numbers were not enough to create a scoring model, it can be concluded that the model is powerful in diagnosing ductal adenocarcinoma and neuroendocrine tumors precisely using the parameters of CEH-EUS. The calculated AUROC of conventional EUS for ductal adenocarcinoma and neuroendocrine tumor was 0.890 and 0.871, respectively, indicating that it is still a useful predictive model but is less powerful than CEH-EUS, with a statistical significance.
To prove the clinical value of CEH-EUS, we next present two example cases. The first case involves a 34-year-old male patient experiencing abdominal pain for 1 month. He had no family history of pancreatic cancer, and the levels of the tumor markers carcinoembryonic antigen and CA19-9 were within the reference ranges. On computed tomography (CT) scan, a 15-mm pancreatic mass was found. Considering his age, tumor marker level, and tumor size, the tumor was less likely pancreatic cancer, but rather possibly pancreatitis or another benign disease. He underwent CEH-EUS, which showed a hypoenhancement pattern, dilated pancreatic duct, and poorly demarcated and rough mass margin. Using the probability equation we created, his pancreatic lesion had a probability rate of 93.2% for ductal adenocarcinoma and 0.7% for neuroendocrine tumor. Therefore, he underwent surgery, and the pathology revealed pancreatic ductal adenocarcinoma. The second case is that of a 78-year-old female patient who was also experiencing abdominal pain for 2 weeks. She had no family history of pancreatic cancer, and the tumor marker levels were also in the reference ranges. On CT scan, a 15-mm pancreatic mass was found. Unlike the previous patient, this patient was old enough to raise suspicion for pancreatic cancer. On CEH-EUS, she had a hyperenhancement pattern, normal pancreatic duct, and well-demarcated and smooth mass margin. Using the probability equation we created, her pancreatic lesion had a probability rate of 7.8% for ductal adenocarcinoma and 89.8% for neuroendocrine tumor. Because of her old age, we planned surgical resection, and the pathology revealed a neuroendocrine tumor.
These two patients had similar pancreatic mass sizes. However, some features on CEH-EUS were different, and despite their unmatched age, the probability equation proved its precise ability of predicting ductal adenocarcinoma and neuroendocrine tumor. This would provide physicians more information on the diagnosis of pancreatic masses and help them in the treatment decision-making.
When we classified the malignant gallbladder masses with an enhancement pattern as what we did for the pancreatic solid masses, they did not have any specific enhancement pattern. This is similar to the result of a previous study [
Expanding the study into the differential diagnosis of neoplasms, CEH-EUS still was not superior to conventional EUS. This would be because of the different characteristics between adenomas and adenocarcinomas; thus, grouping these two pathologies in the same group did not profit EUS.
There were some limitations in this study. This study performed a retrospective data analysis, and we did not measure the degree of enhancement quantitatively. However, the results were sufficient to prove the superiority of CEH-EUS in the differential diagnosis of pancreatic solid masses over conventional EUS. With the high sensitivity of the enhancement texture of malignant gallbladder masses, it could be useful in screening such masses. In addition, we can diagnose malignant masses with a high probability (AUROC of 0.939), if we consider the enhancement texture, size, and hypoechoic foci of the mass on CEH-EUS. Although the superiority was not statistically significant, CEH-EUS still shows its powerful clinical value.
Previously and widely used conventional EUS is limited in the evaluation of the microvasculature of diseases in the pancreas and gallbladder. In this study, we suggest that CEH-EUS has a high clinical value in the differential diagnosis of pancreatic masses and malignant gallbladder masses; thus, it can supplement the limitations of conventional EUS. Further studies that would consider the quantitation of enhancement and other diagnostic features are needed to overcome these limitations.
Supplementary materials are available at
Comparison of the receiver operating characteristic (ROC) curve with predicted probabilities from contrast-enhanced harmonic endoscopic ultrasonography (CEH-EUS) and conventional EUS. (A) For ductal adenocarcinoma. The area under the ROC curve for CEH-EUS is 0.949 (SE 0.0140) and that for conventional EUS is 0.889 (SE 0.0220). (B) For neuroendocrine tumors. The area under the ROC curve for CEH-EUS is 0.945 (SE 0.0145) and that for conventional EUS is 0.870 (SE 0.0274).
Comparison of the receiver operating characteristic (ROC) curve with predicted probabilities from contrast-enhanced harmonic endoscopic ultrasonography (CEH-EUS) and conventional EUS for gallbladder malignant masses. The area under the ROC curve for CEH-EUS is 0.939 (SE 0.0218) and that for conventional EUS is 0.912 (SE 0.0331).
Patient Characteristics
Pancreatic solid mass ( |
Gallbladder mass ( |
|||
---|---|---|---|---|
Age (mean±SD) | 58.7±13.9 | 55.4±15.2 | ||
Sex (M/F) | 110/97 (53.1%/46.9%) | 58/87 (40.0%/60.0%) | ||
Diagnostic modality | ||||
EUS | 207 (100) | 145 (100) | ||
CT | 205 (99.0) | 84 (57.9) | ||
MRI | 163 (78.7) | 35 (24.1) | ||
PET-CT | 144 (69.6) | 42 (29.0) | ||
Tumor size (mm) (mean±SD) | 25.3±15.0 | 15.6±10.9 | ||
Tumor location | ||||
Head/uncinate | 82 (39.6) | Neck | 22 (15.2) | |
Neck | 25 (12.1) | Fundus | 51 (35.2) | |
Body | 48 (23.2) | Body | 72 (49.6) | |
Tail | 52 (25.1) | |||
Pathology | ||||
DA | 95 (45.9) | Neoplasm (adenocarcinoma and adenoma) | 48 (33.1) | |
NET | 57 (27.5) | |||
MFP | 23 (11.1) | |||
SPN | 14 (6.8) | Non neoplasm | 97 (66.9) | |
Others | 20 (9.7) |
SD, standard deviation; EUS, endoscopic ultrasonography; CT, computed tomography; MRI, magnetic resonance imaging; PET-CT, positron emission tomography-computed tomography; DA, ductal adenocarcinoma; NET, neuroendocrine tumor; MFP, mass forming pancreatitis; SPN, solid pseudopapillary neoplasm.
Univariate and Multivariate Analysis of Variables for Ductal Adenocarcinoma
Variables | Univariate | Multivariate |
||||
---|---|---|---|---|---|---|
Conventional EUS |
CEH-EUS |
|||||
Odds ratio (95% CI) | Odds ratio (95% CI) | Odds ratio (95% CI) | ||||
Age | 1.049 (1.025–1.073) | <0.001 | 1.053 (1.024–1.084) | <0.001 | 1.058 (1.018–1.100) | 0.004 |
Sex | 0.785 (0.466–1.325) | 0.365 | ||||
Duct dilatation | 6.657 (2.606–17.008) | <0.001 | 5.748 (1.829–18.060 | 0.003 | 6.344 (1.470–27.379) | 0.013 |
Tumor size | 1.043 (1.021–1.065) | <0.001 | 1.029 (1.003–1.056) | 0.030 | 1.016 (0.984–1.049) | 0.333 |
Tumor demarcation | 3.442 (1.729–6.852) | <0.001 | 0.608 (0.210–1.760) | 0.359 | 0.446 (0.112–1.771) | 0.251 |
Tumor marginal irregularity | 13.372 (6.782–26.366) | <0.001 | 14.805 (5.859–37.408) | <0.001 | 19.167 (5.678–64.705) | <0.001 |
Hypoechoic | 13.429 (1.732–104.144) | 0.013 | 15.955 (1.167–218.136) | 0.038 | ||
Hypoenhance | 35.249 (16.155–76.912) | <0.001 | 35.071 (12.261–100.315) | <0.001 |
EUS, endoscopic ultrasonography; CEH-EUS, contrast-enhanced harmonic endoscopic ultrasonography; CI, confidence interval.
Univariate and Multivariate Analysis of Variables for Neuroendocrine Tumor
Variables | Univariate | Multivariate |
||||
---|---|---|---|---|---|---|
Conventional EUS |
CEH-EUS |
|||||
Odds ratio (95% CI) | Odds ratio (95% CI) | Odds ratio (95% CI) | ||||
Age | 0.986 (0.964–1.007) | 0.195 | ||||
Sex | 1.635 (0.920–2.906) | 0.094 | ||||
Duct dilatation | 0.069 (0.009–0.515) | 0.009 | 0.093 (0.011–0.758) | 0.026 | 0.125 (0.012–1.292) | 0.081 |
Tumor size | 0.922 (0.889–0.955) | <0.001 | 0.946 (0.914–0.978) | 0.001 | 0.945 (0.908–0.983) | 0.005 |
Tumor demarcation | 0.182 (0.062–0.533) | 0.002 | 0.777 (0.193–3.135) | 0.723 | 0.933 (0.180–4.826) | 0.934 |
Tumor marginal irregularity | 0.047 (0.014–0.158) | <0.001 | 0.075 (0.020–0.281) | <0.001 | 0.183 (0.040–0.840) | 0.029 |
Hypoechoic | 0.543 (0.184–1.600) | 0.268 | 1.186 (0.349–4.035) | 0.784 | ||
Hyperenhance | 43.000 (17.826–103.722) | <0.001 | 26.771 (9.633–74.397) | <0.001 |
EUS, endoscopic ultrasonography; CEH-EUS, contrast-enhanced harmonic endoscopic ultrasonography; CI, confidence interval.
Univariate and Multivariate Analysis of Variables for Gallbladder Malignant Masses
Variables | Univariate | Multivariate |
||||
---|---|---|---|---|---|---|
Conventional EUS |
CEH-EUS |
|||||
Odds ratio (95% CI) | Odds ratio (95% CI) | Odds ratio (95% CI) | ||||
Sex | 1.263 (0.577–2.762) | 0.559 | ||||
Age | 1.093 (1.052–1.136) | <0.001 | 1.049 (0.998–1.103) | 0.060 | 1.056 (0.998–1.117) | 0.058 |
Hypoechoic foci | 27.818 (10.097–76.643) | <0.001 | 8.273 (2.223–30.785) | 0.002 | 9.816 (2.777–34.700) | <0.001 |
Tumor size | 1.114 (1.066–1.165) | <0.001 | 1.062 (1.002–1.124) | 0.041 | 1.086 (1.018–1.158) | 0.013 |
Multi-lobulated | 3.185 (1.419–7.149) | 0.005 | 2.563 (0.710–9.255) | 0.151 | 1.332 (0.344–5.163) | 0.678 |
Multiple lesions | <0.001 | 0.998 | ||||
ConvEUS inhomogeneous | 16.164 (6.371–41.012) | <0.001 | 4.166 (1.163–14.918) | 0.028 | ||
CEH-EUS inhomogeneous | 42.327 (5.593–320.290) | <0.001 | 25.681 (2.579–255.689) | 0.006 |
EUS, endoscopic ultrasonography; CEH-EUS, contrast-enhanced harmonic endoscopic ultrasonography; CI, confidence interval.