The role of cytomorphology and proliferative activity in predicting biologic behavior of pancreatic neuroendocrine tumors

A study by endoscopic ultrasound-guided fine-needle aspiration cytology




The biologic behavior of pancreatic neuroendocrine tumors (NETs) is difficult to predict in the absence of metastases or invasion into adjacent organs. In this study, the authors retrospectively evaluated the cytopathology and proliferative activity in cytology specimens obtained by endoscopic ultrasound (EUS)-guided fine-needle aspiration (FNA).


Thirty-five patients who were diagnosed with pancreatic NET based on EUS-guided FNA were studied retrospectively (2002-2007). Cytopathology and proliferative activity (Ki-67) in cytology specimens were reviewed. Patients were divided into 2 groups: Group A (all patients with simultaneous, suspicious, metastatic masses [unresectable tumors]) and Group B (patients with final histopathologic diagnosis). Moreover, the patients in Group B were classified into 4 risk subgroups according to the 2004 World Health Organization (WHO) classification (Subgroups 1a, 1b, 2, and 3).


Thirteen of 35 patients who were diagnosed with unequivocally malignant tumors were placed in Group A (unresectable tumors), and 22 of 35 patients were placed in Group B. In Group A, >2% Ki-67–positive cells were present in 12 of 13 patients (92.3%). In Group B, >2% Ki-67–positive cells were present in 0 of 6 patients in WHO Subgroup 1a (0%), in 3 of 7 patients in WHO Subgroup 1b (42.86%), in 7 of 7 patients in WHO Subgroup 2 (100%), and in 2 of 2 patients in WHO Subgroup 3 (100%). In Group A, nuclear pleomorphism/multinucleation was observed in 8 or 13 patients (61.53%). In Group B, nuclear pleomorphism/multinucleation was observed in 4 of 7 patients in WHO Subgroup 2 (57.14%) and in 2 of 2 patients in WHO Subgroup 3 (100%). In Group A, nucleoli were present in 7 of 13 patients (53.85%); whereas, in Group B, nucleoli were present in 6 of 7 patients in WHO Subgroup 2 (85.7%) and in 2 of 2 patients in WHO Subgroup 3 (100%). None of the remaining cytologic features that were evaluated (necrosis, mitoses, spindle cells, and molding/crush artifact) were observed consistently in malignant NETs.


The current results indicated that Ki-67 evaluation in routine EUS-FNA cytology specimens can be used as a potential prognostic marker in pancreatic NETs. Nuclear pleomorphism/multinucleation and the presence of nucleoli also are reliable for predicting malignant pancreatic NETs. Cancer (Cancer Cytopathol) 2009. © 2009 American Cancer Society.

Neuroendocrine tumors (NETs) of the pancreas are relatively uncommon tumors that account for 1% to 2% of all pancreatic neoplasms.1 They usually are identified preoperatively because of their propensity to secrete various endocrine hormones and to exhibit hormone-related clinical signs and symptoms. These functional tumors are classified based on the hormones they produce and the associated endocrine syndrome. Nonfunctioning tumors are either an incidental finding or are associated with an expanding mass rather than a hormonal syndrome. However, serologic or immunohistochemical evidence for elevated hormones may be identified.2

The precise localization of NETs is of major importance, because surgical resection is the only curative treatment.3 Endoscopic ultrasound (EUS)-guided fine-needle aspiration (FNA) cytology is an effective tool for the diagnosis of pancreatic NETs.4, 5 This procedure provides cytologic material adequate for cytomorphologic and immunocytochemical studies. There have been several reports that discussed the cytomorphologic features and diagnostic criteria of pancreatic NETs.6-8 However, a few authors have attempted to describe criteria suggestive of malignancy in FNA of these tumors.6, 7 In recent years, many proposals have been made to predict their biologic behavior. The most recent World Health Organization (WHO) classification2 included several clinicopathologic criteria: the presence of metastasis, greatest tumor dimension, angioinvasion, mitoses, and proliferative index (based on the MIB1 reaction).

In the current study, we conducted a retrospective review of the cytomorphology, pathology (22 patients), and proliferative activity (Ki-67) on EUS-FNA cytology specimens of 35 pancreatic NETs, including 22 tumors that were unequivocally malignant, in an attempt to determine which, if any, cytologic features were indicators of malignancy.


Thirty-five patients with pancreatic NETs were diagnosed by EUS-FNA cytology from 2002 to 2007 at the General Hospital of Athens, Greece (during that period, we diagnosed 71 NETs). The patients who were selected retrospectively through a computer search of the database in the Department of Cytology had either clinical follow-up (metastases; Group A) or pathologic confirmation (Group B). The remaining 36 patients were not included in our study because of inadequate clinical information (absence of proof of malignancy) or inadequate pathologic confirmation (lack of surgery). No informed consent was obtained from patients, because the studies were carried out during routine clinical examinations. All cytology specimens and microbiopsy procedures were performed in the endoscopy suite. A linear-array echoendoscope (EG3630 UR; Pentax, Tokyo, Japan) was advanced under indirect visualization through the cricopharyngeous into the esophagus. The echoendoscope was connected to a monitoring device (6000 Victor; Hitachi, Tokyo, Japan), and the EUS-guided FNA procedure was performed using 22-gauge needles (Medi-Globe GmbH, Achenmuhle, Germany). The aspirated samples were assessed immediately by an on-site cytopathologist in all cases. The aspirated material was smeared onto slides, and the smear preparation was followed by either air dying or immediate fixation in 95% solution for subsequent Papanicolaou staining methods. Additional aspirated material was fixed in formalin, embedded in paraffin, and processed for routine histologic examination using standard techniques. On average, 3 or 4 passes finally were performed to obtain diagnostic material.

Immunocytochemical studies were performed on cell blocks. For this purpose, 5-μm sections were cut, deparaffinized, and mounted on precoated slides. The antibody MIB1 for Ki-67 (dilution 1:20; Dako, Glostrup, Denmark) was used. Visualization was performed using the avidin-biotin complex method, which yielded a brown staining signal. The immunostain was evaluated by counting all positive and negative tumor cell nuclei, and the staining index was calculated as the percentage of positive cells. The threshold of positive tumor cells was Ki-67 >2% for biologic behavior based on the 2004 WHO classification.2

Special cytologic features that were evaluated for the definition of malignant potential were nuclear pleomorphism/multinucleation, the presence of nucleoli, necrosis, mitoses, molding/crush artifact, and spindle cells. All pathologic slides were reviewed and re-evaluated according to WHO criteria.2 These included the presence of metastases, the greatest dimension of the tumor, gross invasion, angioinvasion, the number of mitoses per 10 high-power fields (HPF), and the Ki-67 index. According to these criteria, NETs were classified into 4 risk subgroups: WHO Subgroup 1a, well differentiated NET in the absence of all criteria; WHO Subgroup 1b, well differentiated NET with uncertain behavior (greatest tumor dimension, ≥2 cm; ≥2 mitoses per 10 HPF; angioinvasion; or Ki-67 index >2%); WHO Subgroup 2, well differentiated neuroendocrine carcinoma (metastasis); and WHO Subgroup 3, poorly differentiated neuroendocrine carcinoma (≥10 mitoses per 10 HPF).


Cytologic-Histologic Findings

Cytologic findings with malignant potential and their distribution in each group of patients are summarized in Table 1. Group A included all patients (n = 13) who had hepatic metastases at the time of diagnosis and did not undergo surgery (unresectable tumors). Group B included patients (n = 22) with available histopathologic diagnosis: WHO Subgroup 1a, well differentiated NETs confined to the pancreas, <2 cm in greatest dimension, nonangioinvasive, <2 mitoses per 10 HPF, and <2% Ki-67–positive cells (n = 6); WHO Subgroup 1b, well differentiated NETs of uncertain behavior confined to the pancreas and the presence of ≥1 of the following: ≥2cm in greatest dimension, from 2 to 10 mitoses per 10 HPF, ≥2% Ki-67–positive cells, angioinvasion, or perineural invasion (n = 7); WHO Subgroup 2, well differentiated neuroendocrine carcinoma invading adjacent organs and/or metastases (n = 7); and WHO Subgroup 3, poorly differentiated neuroendocrine carcinoma with ≥10 mitoses in 10 HPF (n = 2). Unequivocally malignant tumors were those from patients in Group A and from patients in WHO Subgroups 2 and 3 in Group B (total, 22 patients).

Table 1. Cytologic Findings in Patient Groups A and B
 Findings: No. (%)
Tumor GradeNuclear Pleomorphism or MultinucleationNucleoli
  1. WHO indicates World Health Organization.

Group A, n=138 (61.53)7 (53.85)
Group B, n=22    
 WHO 1a, n=60 (0)1 (16.66)
 WHO 1b, n=70 (0)3 (42.85)
 WHO 2, n=74 (57.14)6 (85.71)
 WHO 3, n=22 (100)2 (100)
 Findings: No. (%)
Tumor GradeSpindle CellsMitosesNecrosisMolding
Group A, n=130 (0)3 (23)0 (0)0 (0)
Group B, n=22    
 WHO 1a, n=60 (0)0 (0)0 (0)0 (0)
 WHO 1b, n=70 (0)0 (0)0 (0)0 (0)
 WHO 2, n=71 (14.28)0 (0)0 (0)1 (14.28)
 WHO 3, n=21 (50)0 (0)1 (50)0 (0)

The FNA smears were highly cellular in all cases. The aspirates exhibited a loosely cohesive pattern, a single-cell population, and rosette-like structures. Emphasis on the following cytomorphologic features with malignant potential was given: 1) nuclear pleomorphism/multinucleation was evident in 14 of 22 malignant tumors (63.63%) (Fig. 1A), 2) nucleoli were present in 15 of 22 tumors (68.18%) (Fig. 1B), 3) mitoses were observed in 3 of 22 tumors (13.6%), 4) molding/crush artifact and necrosis were observed in 1 of 22 tumors (4.5%) (Fig. 1C), and 5) spindle cells were observed in 2 of 22 tumors (9%) (Fig. 1D). Tumor cells were mostly round to polygonal, and the amount of cytoplasm varied from abundant eosinophilic/basophilic and finely granular to scanty with only a rim of basophilic cytoplasm. A thin capillary network was noted in most tumors.

Figure 1.

(A) Nuclear pleomorphism and multinucleation are prominent in this sample (hematoxylin and eosin [H&E] stain; original magnification, ×200). (B) Occasional nucleoli are observed in tumor cells (H&E stain; original magnification, ×400). (C) Molding is well marked in this neuroendocrine tumor (H&E stain; original magnification, ×400). (D) This is an endocrine tumor with spindle cells (H&E stain; original magnification, ×200).

Proliferative Activity

Proliferative activity in each patient group is summarized in Table 2. In Group A, >2% Ki-67–positive cells were present in 92.3% of tumors; in Group B, >2% Ki-67–positive cells were present in 100% of tumors from patients in WHO Subgroups 2 and 3. In total, 21 of 22 malignant tumors (95.45%) had a high proliferative index even in EUS-FNA cytology specimens (Fig. 2). In contrast, Ki-67 was totally absent from tumors that were classified as WHO Subgroup 1a and was expressed in a small proportion of tumors (42.86%) classified as WHO Subgroup 1b (uncertain behavior).

Figure 2.

(A,B) Ki-67 is observed in >2% of nuclei in tumor cells.

Table 2. Proliferative Activity
Tumor GradeNo. of Ki-67-positive Cells >2% (%)
  1. WHO indicates World Health Organization.

Group A, n=1312 (92.3)
Group B, n=22 
 WHO 1a, n=60 (0)
 WHO 1b, n=73 (42.86)
 WHO 2, n=77 (100)
 WHO 3, n=22 (100)


The current study highlights for the first time in the literature, to our knowledge, the role of proliferative activity in routine practice of EUS-FNA cytology in the definition of biologic behavior of pancreatic NETs. Greater than 2% Ki-67–positive cells were very common in cytology specimens of unresectable tumors and malignant NETs (WHO Subgroups 2 and 3). Also, Ki-67 was totally absent from well differentiated NETs (WHO Subgroup 1a). The percentage distribution of Ki-67 in tumors with uncertain behavior was ambiguous. These findings strongly support the routine use of Ki-67 staining in EUS-FNA cytology specimens of pancreatic NETs for better documentation and more informative reporting in clinical practice. However, the limited number of our cases (n = 35), because of the lack of clinical or surgical follow-up in a total of 71 pancreatic NETs in the same period in our department was a potential weakness. Nevertheless, this is the first attempt to correlate proliferative activity in cytology specimens with the final histopathologic diagnosis and unequivocally malignant NETs (metastatic tumors).

In 2000 and 2004, the WHO classification of NETs introduced a risk stratification, based on an earlier proposal by Capella et al9 that distinguished tumors with benign, uncertain, low-grade malignant, or high-grade malignant behavior. Because this system was based expert opinion rather than on sound data, it is not universally accepted. In particular, the value of the WHO 1b subgroup is disputed. In a recent study by Schmitt et al,10 authors have demonstrated the predictive value of some criteria and especially Ki-67 index (>2%) which are associated with a shortened disease-free survival and time tumor-induced death. Furthermore, they provided evidence for the usefulness of all 4 subgroups of the WHO 2004 classification. Moreover, the value of 1b subgroup was highlighted in this study by Schmitt et al10 Previously published findings by Heymann et al11 reported the prognostic power of the original Capella classification, but the proliferative index was not included as a criterion. In the study by Schmitt et al,10 tumors were classified as WHO Subgroup 1b (separated from WHO Subgroup 1a) on the basis of an elevated proliferative index alone. The follow-up data showed that all of those tumors recurred. The inclusion of the proliferation index, therefore, seems useful.

We recently published the cytomorphologic findings in pancreatic NETs from a large series.6 The most helpful cytologic feature that facilitated the cytologic diagnosis of NETs was a monotonous, poorly cohesive population of small cells with plasmacytoid morphology. The tumor cells were relatively small and medium in size with pale, soft cytoplasm and indistinct cell borders. The nuclei were round to oval with finely stippled and uniformly distributed chromatin (salt-and-pepper pattern). The nucleoli usually were small and inconspicuous. Occasionally, in that study,6 we observed mitotic activity, anisonucleosis, and prominent nucleoli, but we did not correlate them with clinical follow-up or WHO classification in histopathologic reports.

According to the findings mentioned above, 3 other published studies have attempted to describe criteria suggestive of malignancy in FNA of these tumors.7, 12, 13 Gu et al7 observed the presence of nucleoli in almost 70% of their cases and reported pleomorphism, binucleation, and multinucleation in percentages that ranged from 33.3% to 40%. Those authors also observed molding and necrosis in 6% and 1% of tumors, respectively, and spindle cells and mitoses were observed in 1% and 2%, respectively.7 However, none of the patients with those cytologic findings had clinical or cytologic evidence of metastasis. Moreover, the authors did not finally correlate their findings with any histopathologic report or classification based on WHO criteria. Similarly, the 2 other studies12, 13 had the same limitations.

We retrospectively re-examined the specific cytomorphologic features of NETs with emphasis on their potential malignant significance in the current study. For this reason, we selectively studied only those cases that were unequivocally malignant (metastatic) and those with histopathologic WHO classification. Special emphasis was given to the following cytomorphologic features: nuclear pleomorphism/multinucleation (63.63%), nucleoli (68.18%), spindle cells (9%), necrosis (4.5%), molding/crush artifact (4.5%), and mitoses (13.6%). We observed that nuclear pleomorphism, multinucleation, and nucleoli seemed to be reliable prognostic markers in pancreatic NETs. The other features that we examined had restricted significance in their ability to predict biologic behavior. Thus, we strongly recommend that cytologists pay attention to such cytomorphologic findings and include specific comments regarding these prognostic markers in their cytologic reporting.

In conclusion, the current study demonstrates the role of proliferative activity (Ki-67) in EUS-FNA cytology specimens as a reliable predictive factor for pancreatic NETs. In addition, specific cytologic features like nuclear pleomorphism, multinucleation, and nucleoli also can contribute to the prediction of these tumors. Moreover, a few malignant samples revealed necrosis and mitoses. Because of the limited number of cases, larger series will be needed to confirm our findings.

Conflict of Interest Disclosures

The authors made no disclosures.