Endoscopic ultrasound features of pancreatic solid lesions: Descriptive and predictive analysis on a multicenter sample

Abstract

BACKGROUND

Endoscopic ultrasound (EUS) is crucial for diagnosing solid pancreatic lesions, especially pancreatic ductal adenocarcinoma (PDAC), a highly aggressive cancer which represents the majority with a prevalence of approximately 85%.

AIM

To identify EUS features that differentiate PDAC from other lesions such as neuroendocrine tumors (NETs) and helping in the differential diagnosis, by analyzing a large sample of solid pancreatic lesions.

METHODS

This observational, retrospective, multicenter study analyzed the endosonographic characteristics of 761 patients with a radiological diagnosis of solid pancreatic lesion, who underwent pancreatic EUS for typing and staging with needle biopsies between 2015 and 2023. General patient characteristics (age and sex) and solid lesion features were collected and described, such lesion size (B-mode), vessel involvement (compression or invasion), ductal dilation, lymphadenopathy, echogenicity, echopattern, margin regularity, multifocality, internal vascularization and elastography. Subsequently, a predictive analysis was performed through univariate and multivariate logistic regression to identify predictive features for PDAC or NET diagnoses.

RESULTS

Our study enrolled 761 patients, predominantly male with a mean age of 68.6. PDACs were generally larger (mean 33 mm × 27 mm), often had irregular margins, and displayed significant upstream ductal dilation. Hypoechogenicity was common across malignant lesions. In contrast, NETs were smaller (mean 20 mm × 17 mm) and typically had regular margins with multiple lesions. Vascular involvement, although predominant in PDAC, is a common feature of all malignant neoplasms. Multifocality, however, although a rare finding, is more typical of NETs and metastases, and practically absent in the remaining lesions. Predictive analyses showed that ductal dilation and irregular margins were the most significant predictors for PDAC [odds ratio (OR) = 5.75 and 3.83], with hypoechogenicity, heterogeneous echopattern and lymphadenopathies also highly significant (OR = 3.51, 2.56 and 1.99). These features were inversely associated with NETs, with regular margins and absence of ductal involvement or lymphadenopathies (OR = 0.24, 0.86 and 0.45 respectively), as already shown by the descriptive analysis. Finally, age, despite achieving statistical significance, lacks clinical value given an OR trending towards 1.

CONCLUSION

This study provides a comprehensive overview of EUS features for solid pancreatic lesions, identifying distinct features like upstream ductal dilation and irregular margins for PDAC vs regular margins for NETs as strong diagnostic predictors. These findings enhance the understanding of pancreatic pathologies, offering valuable insights for improved differential diagnosis and clinical management, especially in complex cases. Further prospective studies could build on these results.

Key Words: Endoscopic ultrasound; Pancreatic solid lesions; Pancreatic ductal adenocarcinoma; Pancreatic neuroendocrine tumors; Pancreatic metastasis; Focal pancreatitis; Endoscopic ultrasound-fine needle aspiration/fine needle biopsy; Endoscopic ultrasound features of pancreatic solid lesions

Core Tip: This multicenter observational study provides a broad overview of the endo-sonographic characteristics of solid pancreatic lesions, with detailed descriptions for each pathology and a predictive analysis of the most significant variables for key differential diagnoses. While existing literature touches on this topic, no prior study has analyzed endo-sonographic features with such a large sample size and comprehensive analysis. Our findings can aid clinical practice by optimizing the diagnostic process for solid pancreatic lesions, preventing both false negatives (delayed diagnosis) and false positives (unnecessary surgical intervention), and simultaneously supporting biopsy diagnosis when it is doubtful or inconclusive.

INTRODUCTION

In industrialized countries, pancreatic cancer is currently the tenth most prevalent neoplasm but the third leading cause of cancer-related deaths, surpassed only by lung and colorectal malignancies[1,2].

Pancreatic neoplasms are diverse, characterized by varied demographics and pathophysiology. The most frequent neoplasm by far is pancreatic ductal adenocarcinoma (PDAC), accounting for approximately 85% of cases[3]; it also represents the neoplasm with the worst prognosis with a five-year survival rate at around 10%[4]. Unlike almost all other cancers, the prognosis of this disease has not improved in recent decades: Clinical research is struggling to find a cure due to its chemo- and radioresistance characteristics[5,6], while early diagnosis remains difficult to achieve in the general population due to its indolent course and rapid growth and spread to surrounding organs[7,8].

Another solid pancreatic neoplasm is the neuroendocrine tumor (NET), which is rarer than PDAC. It is a disease with a wide clinical spectrum, including functioning and non-functioning forms, with differentiation ranging from well-differentiated to undifferentiated forms[9]. Other less common pancreatic neoplasms include metastases from other primary sites (mainly originating from renal clear cell carcinoma)[10], solid pseudopapillary tumor, pancreatoblastoma, and other rare histology[11]. Finally, there are non-neoplastic solid lesions that mimic a mass, such as focal pancreatitis, or other rare non-neoplastic diagnoses[12].

Endoscopic ultrasonography of the pancreas is the most important diagnostic examination for a solid pancreatic lesion, often performed following a radiological examination such as computed tomography (CT) or magnetic resonance imaging (MRI). This is due to its high accuracy and the possibility of performing a needle biopsy for cytological or histological examination[13,14]. Endoscopic ultrasound (EUS) diagnosis has benefited over the years from technical improvements, such as the use of contrast agents or elastography[15]. Despite these characteristics, EUS also has limitations that must be considered, particularly concerning interobserver agreement (which varies depending on the pathology)[16,17] and the need for a high level of expertise[18].

The objective of our study is to describe and analyze the EUS features of solid pancreatic lesions, evaluating any differences between pathologies and the predictivity of these characteristics in making a differential diagnosis. This remains an ongoing challenge, useful for avoiding false negatives (diagnostic delay) or unnecessary surgical interventions due to false positives[19].

MATERIALS AND METHODS

This multicenter study was conducted in three Italian tertiary referral centers (Fondazione IRCCS Ca’ Granda - Ospedale Maggiore Policlinico di Milano, ASST Spedali Civili di Brescia, and ASST Sette Laghi di Varese). All patients who underwent pancreatic EUS between 2015 and 2023 with the indication for EUS to characterize and stage a solid pancreatic lesion identified in a previous radiological examination (abdominal ultrasound, CT, or MRI) were enrolled.

The instruments used were Pentax EG-3870UTK and Pentax EG-38J10UT for echoendoscope (Pentax Medical, Tokyo, Japan) and Hitachi Arietta v70 and Hitachi Hi Vision Avius^® for sonograph (Fujifilm Healthcare, Tokyo, Japan). All patients underwent needle biopsy [fine needle aspiration (FNA)/fine needle biopsy (FNB)] performed using a 19-22-25 gauge Expect^TM and Acquire^TM (Boston Scientific, Marlborough, MA, United States) and EchoTip ProCore^® (Cook Medical, Bloomington, IN, United States). Diagnoses were established through the cytological/histological report or based on a detailed clinical-instrumental follow-up of at least twelve months in the case of inadequate tissue collection at baseline. No patients were excluded from the initial collection, but not all endosonographic characteristics were described in every patient, as shown in Figure 1.

Figure 1 Endoscopic ultrasound features of the total study sample and the subgroups.

The EUS procedure was standardized across the three participating centers according to the European Society of Gastrointestinal Endoscopy guidelines[20,21]. In particular, EUS-FNA and EUS-FNB involved the use of a linear echoendoscope to visualize the lesion, followed by the insertion of a fine needle through the duodenal or gastric wall. Once the echoendoscope was properly positioned, the needle was advanced through the working channel and into the lesion. To obtain an adequate tissue sample, a to-and-fro motion was performed for an average of two to four passes, often with the aid of negative pressure or a stylet. The collected specimen was then extruded from the needle and submitted for pathological analysis. To minimize intra-operator variability, which is inherent to the imaging nature of the method, all procedures were performed by expert operators (> 250 EUS). Conversely, to minimize inter-operator variability and to avoid errors, two operators worked together on data entry from report interpretation, specifically to resolve any doubtful cases.

A database was retrospectively collected using the REDCap platform[22,23] (Nashville, TN, United States) and anonymized using the ARX software[24] (open source). All statistical analyses were performed with R 4.2.1 [R Core Team (2022). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/].

The variables collected for each patient, in addition to the diagnosis, included demographic characteristics (age and sex) and EUS features (Figure 2). Specifically, these included the maximum and minimum lesion dimensions (B-mode), vessel invasion or compression (defined by the reduction of the vessel caliber or by the presence of an abnormal and irregular wall, with loss of the interface between the vessel and the parenchyma[25]) upstream ductal dilation (above 3 mm for the main pancreatic duct and above 9 mm for the main bile duct[26,27]), presence of lymphadenopathy, echogenicity, echopattern, margin regularity (defined by the presence of an indistinct contour between the lesion and the surrounding parenchyma, generally due to the infiltrative nature of a malignant neoplasm[28]), and presence of multiple lesions (yes or no). In addition, two other variables were collected, though not for all patients due to their absence in the EUS report: Internal vascularization of the lesion (assessed by color Doppler as hypovascularized or non-hypovascularized) and stiffness (qualitatively evaluated by the operator via elastography as hard or non-hard). For the missing data (excluding vascularization and elastography) in the various features, the EUS images were recovered and re-evaluated by the operators.

Figure 2 In the first row, there is a hypoechoic neoplasm with irregular margins, a common bile duct invasion, and the measurement of the lesion in B-mode. In the second row, you can see an arterial invasion, a portal vein invasion, and a homogeneous neuroendocrine tumor with regular margins.

Statistical analysis

Initially, a descriptive analysis was conducted for the entire cohort and for subgroups based on pathology: PDAC, NET (all differentiation grade), metastases (main primitive tumors from the kidney, lung, ovary and colon), focal pancreatitis (both autoimmune and other etiology), and null diagnoses (when a diagnosis could not be determined on histology or radiological follow-up of at least twelve months). Subsequently, univariate and multivariate predictive analyses were performed using logistic regression to identify any predictive variables for the diagnosis of PDAC or NET, as these were the pathologies with a good prevalence in our sample. Univariate and multivariate logistic regression analyses were performed to identify factors associated with the diagnosis of PDAC or NET. For each predictive variable, the odds ratio (OR), its 95% confidence interval, and the P value were calculated. Variables found to be significant in the univariate analysis (P < 0.05) were included in the multivariate model to determine independent predictors.

Ethical statements

This study was approved by the Ethical Committee Lombardia 3, dated April 17, 2024 (ID 4760 - opinion number 4760_17.04.2024_P). Informed consent was not obtained in accordance with local ethical and privacy authorities, as it was deemed a disproportionate effort for the study and due to the demise of many patients enrolled before the study began.

RESULTS

For this study 761 patients were enrolled, with a slightly higher prevalence of males and a mean age of 68.6 years. In the descriptive analysis, shown in Table 1, the peculiarities of each pathology can be observed: (1) Mass size was significantly larger in PDAC compared to other solid lesions and smaller in null diagnoses; (2) Vessel invasion and compression were also more frequent in PDAC, although these features were present in other malignant neoplasms in the sample, such as NETs and metastases; (3) Upstream ductal dilation was the feature with the largest difference between PDAC and other subgroups, being present in the majority of the subgroup; (4) Hypoechogenicity was a common feature across virtually all solid lesions, particularly in malignant neoplasms, while a heterogeneous echopattern is observed in almost all PDACs and focal pancreatitis, conversely, NETs exhibit this characteristic in only one out of two cases; (5) Irregular margins represented a peculiarity of PDAC and focal pancreatitis, with a distinctly lower prevalence in other lesions; (6) Lymphadenopathies were present in varying percentages across all solid lesions, despite the morphological description bias explained below; (7) The presence of multiple lesions was a characteristic more typical of metastases, followed by NETs and null diagnoses, while it was practically absent in PDAC and focal pancreatitis; (8) Among the features collected only partially, hypovascularization was typical of PDACs and focal pancreatitis, and represents approximately one in three lesions in NETs and metastases. Conversely, hard elastography was found in the majority of all lesions with the exception of metastases; and (9) Rare diagnoses, such as retroperitoneal fibrosis, pancreatoblastoma, solid pseudopapillary tumor, schwannoma, accessory spleen, and others were not included in the analysis of Table 1, while differential diagnoses of NET and metastases are shown in Tables 2 and 3.

Table 1 Descriptive analysis of the entire sample and main subgroups.

Descriptive analysis	Total (n = 761)	PDAC (n = 480)	NET (n = 123)	Metastases (n = 40)	Focal pancreatitis (n = 27)	No diagnosis (n = 59)
Sex (female)	361 (47.4)	236 (49.2)	54 (43.9)	18 (45.0)	7 (25.9)	29 (49.2)
Age (year)	68.6 ± 11.7	70.7 ± 11.7	62.3 ± 14.8	70.4 ± 9.5	63.6 (9.1)	67 (13.6)
Major axis (mm)	29.1 ± 15.0	33.2 ± 15.0	20.3 ± 15.0	24.8 ± 13.4	27 (14.2)	18.9 (14.9)
Minor axis (mm)	24 ± 13.3	27.5 ± 13.3	17.2 ± 13.0	20.7 ± 10.8	20.9 (12.5)	14.6 (11.9)
Arterial invasion	117 (15.4)	102 (21.3)	8 (6.5)	5 (12.5)	0	0
Venous invasion	186 (24.4)	157 (32.7)	17 (13.8)	5 (12.5)	2 (7.4)	0
Vessels compression	324 (42.6)	255 (53.1)	29 (23.6)	14 (35)	5 (18.5)	12 (20.3)
Lymphadenopathy	290 (38.1)	236 (49.2)	17 (13.8)	11 (27.5)	5 (18.5)	12 (20.3)
Ductal dilation	450 (59.1)	379 (79.0)	21 (17.1)	11 (27.5)	10 (37.0)	15 (25.4)
Hypoechoic	697 (91.6)	461 (96.0)	110 (89.4)	34 (85.0)	21 (77.8)	47 (79.7)
Echopattern heterogeneous	648 (85.2)	463 (96.5)	63 (51.2)	28 (70.0)	26 (96.3)	41 (69.5)
Irregular margins	623 (81.9)	460 (95.8)	51 (41.5)	23 (57.5)	23 (85.2)	45 (76.3)
Multiple lesions	44 (5.8)	9 (1.9)	17 (13.8)	9 (22.5)	1 (3.7)	7 (11.9)
Hypovascularization	184 (62.6), n = 294	124 (80.0), n = 155	25 (31.6), n = 79	8 (38.1), n = 21	7 (87.5), n = 8	14 (70.0), n = 20
Hard elastography	128 (81.5), n = 157	83 (96.5), n = 86	24 (68.6), n = 35	7 (50.0), n = 14	7 (77.8), n = 9	4 (50.0), n = 8

Quantitative variables are shown with mean ± SD, qualitative variables are shown with as n (%). Hypovascularization and Hard elastography also show the total number of cases in the disease subgroup in parentheses, as these are not described in all patients. PDAC: Pancreatic ductal adenocarcinoma; NET: Neuroendocrine tumor.

Table 2 Descriptions of the differential diagnoses of the neuroendocrine tumor group, n (%).

NET	Total (n = 123)
Non functional	112 (91.1)
Insulinoma	8 (6.5)
Glucagonoma	2 (1.6)
Gastrinoma	1 (0.8)

NET: Neuroendocrine tumor.

Table 3 Descriptions of the differential diagnoses of the metastases group, n (%).

Metastases	Total (n = 40)
Kidney	23 (57.5)
Colon	7 (17.5)
Lung	4 (10.0)
Breast	2 (5.0)
Ovary	2 (5.0)
Melanoma	1 (2.5)
Non-Hodgkin's lymphoma	1 (2.5)

Predictive analysis

Following the descriptive analysis, Tables 4 and 5 present the univariate and multivariate predictive analyses, respectively, using PDAC and NET diagnoses as the outcome variables.

Table 4 Predictive analysis through logistic regression to identify predictive features of pancreatic ductal adenocarcinoma.

Predictive analysis (PDAC)	Univariate			Multivariate
	Odds ratio	95%CI	P value	Odds ratio	95%CI	P value
Age	1.04	1.03-1.06	< 0.01	1.03	1.01-1.05	< 0.01
Sex	1.21	0.90-1.62	0.21
Major axis (mm) of the mass	1.07	1.05-1.09	< 0.01	1.01	0.97-1.05	0.69
Minor axis (mm) of the mass	1.08	1.06-1.10	< 0.01	1.01	0.97-1.06	0.59
Arterial invasion	4.79	2.72-8.41	< 0.01	1.88	0.88-4.02	0.10
Venous invasion	4.22	2.75-6.49	< 0.01	1.14	0.59-2.22	0.70
Vessels compression	3.48	2.51-4.82	< 0.01	1.51	0.88-2.58	0.14
Lymphadenopathy	4.07	2.88-5.75	< 0.01	1.94	1.25-3.02	< 0.01
Ductal retrodilation	11.10	7.84-15.70	< 0.01	5.75	3.81-8.69	< 0.01
Hypoechoic	4.63	2.65-8.09	< 0.01	3.51	1.69-7.29	< 0.01
Heterogeneous	14.10	8.21-24.30	< 0.01	2.56	1.22-5.38	0.01
Irregular margins	16.70	10.00-27.60	< 0.01	3.83	1.97-7.44	< 0.01
Multiple lesions	0.13	0.06-0.28	< 0.01	0.48	0.18-1.28	0.14
Hypovascularization	5.27	3.14-8.83	< 0.01
Hard elastography	16.00	4.58-55.70	< 0.01

PDAC: Pancreatic ductal adenocarcinoma; 95%CI: 95% confidence interval.

Table 5 Predictive analysis through logistic regression to identify predictive features of neuroendocrine tumor.

Predictive analysis (NET)	Univariate			Multivariate
	Odds ratio	95%CI	P value	Odds ratio	95%CI	P value
Age	0.95	0.94-0.97	< 0.01	0.97	0.95-0.99	< 0.01
Sex	0.84	0.57-1.24	0.39
Major axis (mm) of the mass	0.94	0.92-0.95	< 0.01	0.96	0.91-1.02	0.19
Minor axis (mm) of the mass	0.93	0.91-0.95	< 0.01	1.04	0.97-1.11	0.25
Arterial invasion	0.34	0.16-0.71	< 0.01	0.69	0.26-1.83	0.46
Venous invasion	0.45	0.26-0.77	< 0.01	2.26	0.91-5.61	0.08
Vessels compression	0.36	0.23-0.56	< 0.01	0.74	0.35-1.61	0.46
Lymphadenopathy	0.21	0.13-0.37	< 0.01	0.45	0.24-0.84	0.01
Ductal retrodilation	0.10	0.06-0.17	< 0.01	0.28	0.15-0.50	< 0.01
Hypoechoic	0.74	0.39-1.40	0.35
Heterogeneous	0.10	0.06-0.15	< 0.01	0.44	0.24-0.81	< 0.01
Irregular margins	0.08	0.05-0.13	< 0.01	0.24	0.14-0.42	< 0.01
Multiple lesions	3.63	1.91-6.89	< 0.01	0.88	0.39-1.99	0.76
Hypovascularization	0.16	0.09-0.29	< 0.01
Hard elastography	0.38	0.16-0.90	0.03

NET: Neuroendocrine tumor; 95%CI: 95% confidence interval.

Upstream ductal dilation and irregularity of margins represented the most important predictive factors for PDAC, showing notable ORs in both univariate and multivariate analyses, and a highly significant P value. Additionally, hypoechogenicity, heterogeneous echopattern and the presence of lymphadenopathies maintained strong statistical significance despite lower OR values. The case was different for vessel invasion or compression: Despite robust ORs and P values in the univariate analysis, statistical significance was not confirmed in the multivariate analysis. Mass size and age did not show relevant clinical significance (OR < 1.1), sex did not assume any predictive value while multiple lesions suggested predictivity for a negative diagnosis, although with a non-significant P value in the multivariate analysis. Finally, hard elastography and hypovascularization showed remarkable ORs with strong statistical significance in the univariate analysis, albeit with wider confidence intervals due to the smaller sample size; the missing data for these two features precluded their inclusion in the multivariate analysis.

Regarding the diagnosis of NETs, the relationships with EUS features are almost mirror images of those for PDACs. Upstream ductal dilation, irregular margins, echopattern heterogeneous and the presence of lymphadenopathies showed ORs strongly predictive of a negative diagnosis (i.e., less likely to be NETs). While arterial invasion and vessel compression showed predictive values for a negative diagnosis in univariate analysis, these were not confirmed in multivariate analysis. Venous invasion, likely due to its low prevalence in NETs, exhibited paradoxical values, indicating a negative diagnosis initially and then a positive one. Similar to PDACs, age and mass size had modest ORs with no clinical significance. The presence of multiple lesions showed a significant OR (predictive of positive diagnosis) in univariate analysis, but this was not confirmed in multivariate analysis. Finally, lesion hypoechogenicity and patient sex showed no predictive significance.

DISCUSSION

The sample collected in this study allowed for a comprehensive overview of the EUS features of solid pancreatic lesions. This represents a significant strength, as our study boasts a larger sample size compared to much of the existing literature, enabling a more complete and in-depth analysis of these EUS characteristics.

As mentioned, the two most frequent diagnoses, namely PDAC and NETs, exhibit mirror-image characteristics: PDACs are typically larger, present with irregular margins, show ductal invasion (and more rarely, vessel invasion), have a hypoechoic and heterogeneous appearance and are seldom multiple. In contrast, NETs are generally smaller, more often have regular margins, show a low rate of vessel invasion or compression, and more frequently present as multiple lesions and in younger patients.

Among the peculiarities of rarer solid lesions, we found irregular margins and a notable rate of upstream ductal dilation for focal pancreatitis, while smaller size and a complete absence of vessel invasion were typical of null diagnoses (the typical small lesions that are difficult to biopsy and often disappear on subsequent radiological follow-ups).

Lymphadenopathies warrant a separate discussion. With being a retrospective study and collecting data from EUS reports, there was a lack of strong consistency in the definition of this variable. In a minority of cases, lymphadenopathies were not distinguished between suspicious/pathological and reactive and neither the size nor the shape of the reported lymph nodes was indicated in the report. This certainly represents a limitation of this study, linked to its retrospective design and the absence of standardization in describing EUS findings in reports. This limitation must be considered when interpreting our results, which show a prevalence of lymphadenopathies of 49% in PDACs but also 18% in focal pancreatitis, which is a benign condition.

Another limitation of this study, also related to the retrospective design and the lack of standardization in report writing, is the missing data concerning the internal vascularization and stiffness of the solid lesion influencing both the descriptive and predictive analyses, preventing a complete evaluation of these two features. Finally, a limitation related to the procedure's technique is the lack of contrast-harmonic EUS usage. This has been shown to be useful in the differential diagnosis of solid pancreatic lesions, but it is not clearly indicated by the European guidelines and is not routinely used in the three centers of this study.

In the literature, there are general studies regarding the sonographic appearance of solid lesions, primarily describing them as hypoechoic[29], with upstream ductal dilation for PDAC[30], and sharp margins for NETs[31,32]. However, few studies are analogous to ours, with more in-depth analyses of various features, despite some methodological differences and variations in sample size[33-35].

The strength of our study compared to those in the literature lies in its size and its multicenter design. The predictive analysis of such a large and varied sample has allowed identification of EUS features most predictive of PDAC and NETs, with a quantitative measure such as the OR, which was more accurate and had lower confidence intervals than the above-mentioned works. These findings provide a greater understanding of these pathologies, leading to improved management linked to early diagnosis and a more precise evaluation of the mass based on its echoendoscopic characteristics. This precision is crucial as it can help avoid diagnostic errors such as false negatives (preventing delayed diagnoses) or false positives (sparing patients with focal pancreatitis from unnecessary surgery mistakenly identifying it as PDAC) and avoid excessive repetitions of EUS for FNA-FNB sampling in cases of inconclusive findings, especially in doubtful cases. These data could be corroborated by confirmatory studies, with the development of more advanced predictive models, prospective designs, or the enrollment of a larger number of centers.

CONCLUSION

Solid pancreatic lesions have peculiar characteristics on EUS, some of which are strongly predictive of a particular diagnosis, such as upstream ductal dilation in PDAC or the regularity of margins in NETs. The importance of describing these features in reports can lead to improved management of patients affected by these pathologies.