WELL DIFFERENTIATED GRADE 3 PANCREATIC NEUROENDOCRINE TUMORS IN COMPARISON TO RELATED NEOPLASMS: A MORPHOLOGICAL STUDY

Abstract

Background

Pancreatic neuroendocrine neoplasms with Ki-67 >20% were reclassified in 2017 by the World Health Organization into well differentiated (WD) grade 3 (G3) and poorly differentiated (PD) G3 neuroendocrine carcinoma (NEC). We describe the cytologic features of WD-G3 in comparison to G2 and NEC.

Design

We reviewed fine needle aspirates from 65 pancreatic neuroendocrine neoplasms and compared cytomorphological features across G2, WD-G3, and PD small cell type (S), large cell type (L), and type not otherwise specified (NOS).

Results

The 65 neoplasms consisted of G2: 19; WD-G3: 32; and NEC: 14 (PD-S: 6, PD-L: 5, PD-NOS: 3). The medians (range) of Ki-67 proliferation were G2: 11% (3.2–17); WD-G3: 40% (21–89); PD-S: 80% (63–95); PD-L: 39% (25–61); and PD-NOS: 70% (30–80). Both G2 and WD-G3 were associated with plasmacytoid morphology and smooth nuclear contours, but WD-G3 had significant increases in abundant cytoplasm (72% vs 17%; P=.007), nuclear tangles (75% vs 42%; P=.006), and apoptosis (86% vs 58%; P=.005). In comparison to NEC, WD-G3 had increased plasmacytoid morphology (75% vs 7%; P<.001) smooth nuclear contours (94% vs 64%; P=.02), round nuclei (59% vs 21%; P=.01), and less pleomorphism (13% vs 50%; P=.004), molding (9% vs 79%; P<.001), and necrosis (13% vs 43%; P=.003). WD-G3 had less pleomorphism (13% vs 50%; P=.04), less necrosis (13% vs 60%; P=.04), and more plasmacytoid morphology (75% vs 20%; P=.03) than PD-L.

Conclusions

We identified cytologic features at different prevalences in WD-G3 compared with G2 and NEC that assist in recognition of this newly classified entity.

Introduction

Recently, a new diagnostic category, well differentiated (WD) pancreatic neuroendocrine tumor (PanNET) grade 3 (G3), was created by the 2017 World Health Organization (WHO) for WD gastro-pancreatic neuroendocrine tumors with Ki-67 proliferation or mitotic rates exceeding 20%.¹ Previously, neoplasms with proliferation rates over 20% were classified as poorly differentiated neuroendocrine carcinoma (NEC) of small (S) or large (L) cell types. However, a two-tiered subclassification of this category is necessary as it has become clear that the G3 tumors consist of WD and PD neuroendocrine neoplasms and theses two classes have underlying differences not only in morphology but also in genotype, immunophenotype, behavior, and treatment response.^2–7 WD PanNETs typically present as local or locally advanced disease. Mortality for patients with WD PanNETs that are not cured by surgery is usually related to progressive disease within the liver.⁸ Rarely, WD-G3 PanNETs present as de novo histologically with a uniform G3 proliferation rate or as morphologically distinct areas with G3 level proliferation in association with an otherwise G1 or G2 neoplasm. More commonly, these tumors have an increased proliferation rate over the course of disease progression.^{6, 7} In contrast, NEC is universally aggressive and advanced at presentation. Patients with NEC are managed with platinum-based therapies as a first line of treatment, but often show an abbreviated response with poor survival.⁷

Emerging genotypes for WD-PanNET include mutations of death domain-associated protein (DAXX), alpha thalassemia/mental retardation syndrome × (ATRX), and menin (MEN1), whereas NEC shows more similarity with conventional pancreatic adenocarcinoma and may exhibit mutations of retinoblastoma (RB1), KRAS, and TP53 genes.^{3, 4, 9} Loss of immunohistochemical labeling for DAXX, ATRX, and RB1 and abnormal expression of p53 have been shown to correlate well with mutation status.¹⁰

The diagnosis of pancreatic neuroendocrine neoplasms requires accurate determination of grade and differentiation. Grade can be determined objectively using markers of proliferation including mitotic activity and/or Ki-67 proliferation, but currently, differentiation is initially based on subjective morphologic assessment. Most WD PanNETs are low grade and have a typical plasmacytoid cytomorphology that is easily recognized on fine needle aspiration specimens.¹¹ NECs also have distinctive features: small cell types have small cells with minimal cytoplasm, nuclear molding, and coarse chromatin, while large cell types have large cells with abundant cytoplasm and larger nuclei with prominent nucleoli. The cytological features for WD-G3 have not yet been examined in the context of these related entities. We recently demonstrated that morphologic assessment of G3 neuroendocrine neoplasms was challenging for experienced cytopathologists, prompting a comparative analysis of morphologic differences within the G3 category.¹²

In the present study, we compare the cytomorphology of WD-G3 with G2 and NEC using a cohort of well characterized tumors classified by review of prior and concurrent histology, ancillary studies (Ki-67 proliferation index, immunohistochemistry (loss of RB1, DAXX, ATRX, and p53 protein expression)), mutations in DAXX, ATRX, RB1, TP53, MEN1, or KRAS detected by targeted mutation analysis, and consensus review.

METHODS

Case selection

Patients with fine needle aspiration specimens from pancreatic neuroendocrine neoplasms (primary or metastatic) were selected retrospectively by one author (C.S.S.) from the pathology database at MSKCC with IRB approval during the time period 8/2002–12/2016. Clinical data was obtained from the electronic medical record including age of diagnosis, sex, stage of presentation, and therapeutic agents received prior to biopsy. Therapeutic agents were divided among the following classes: hormone only (octreotide); alkylating agent (dacarbazine, temozolomide); cytotoxic chemotherapy (irinotecan, capecitabine, 5-fluorouracil, etoposide); targeted agent (everolimus, sunitinib); and platinum based (carboplatin, oxaliplatin).⁸

Cytologic features assessment

Slides were de-identified immediately following retrieval from the slide archive (T.D.). Two authors reviewed all slides from each case independently (C.S.S. and V.W.S) and were blinded to clinical history and ancillary information. Cases were re-reviewed together for consensus in instances of disagreement. The cytology preparations consisted of the following: air dried Diff Quik, Thinprep, and alcohol-fixed smear or cell block with hematoxylin and eosin stain. The following cytologic features were assessed: large nuclear size, nuclear pleomorphism, round nucleus, smooth nucleus contour, nucleus angulation, fine chromatin, single prominent nucleolus, presence of plasmacytoid morphology, abundant cytoplasm), nuclear tangles, nucleus molding, and necrosis. Large nuclear size was defined as greater than 5 times a lymphocyte. Single prominent nucleoli were determined by visibility at low power magnification. Abundant cytoplasm was defined as greater than 3 times the size of the nucleus. Apoptotic debris was counted in 10 high power fields (Olympus BX43, 40× objective) and recorded as present, absent, or >5. Mitoses were counted in 10 high power fields (Olympus BX43, 40× objective) as any present, absent, or >5. Comparison of cytological features was performed across the tiers of grade and differentiation. Categorical variables were compared using Fisher’s exact test and continuous variables were compared using Wilcoxon's rank-sum using STATA 13 (Stata Corp., College Station, Texas). The threshold for statistical significance was P<0.05.

Immunohistochemistry

Additional immunohistochemical studies were performed on core biopsies with available material. The antibodies used included RB1 (13A10; 1:50 dilution; Dako, Carpintaria, CA); Anti-ATRX (1:500 dilution, Sigma-Aldrich Corporation, St Louis MO); Anti-DAXX (1:300 dilution; Sigma-Aldrich Corporation, St Louis, MO); P53 (D07; 1:250 dilution; Dako, Carpintaria, CA); and anti-Ki-67 (30–9; pre-diluted; Ventana, Tucson, AZ). The BenchMark XT automated equipment (Ventana Medical System Inc., Tucson, AZ) was used for immunohistochemistry for DAXX, ATRX, p53, and Ki-67 whereas the Leica BOND automated system (Leica Biosystems, Wetzlar, Germany) was used for RB1. Abnormal staining was assessed as complete loss of DAXX, ATRX, and RB1 nuclear protein expression in the presence of positive nuclear staining in non-neoplastic cells. Abnormal p53 labeling was defined as strong nuclear positivity in >25% of nuclei or complete absence of staining in the presence of a positive internal control. The Ki-67 proliferation rate was previously determined for all included cases by manual counting on core biopsy, cell block, and/or alcohol fixed cytology smears, and methods are described in the authors’ previous work.¹³ For concurrently obtained cytology/biopsy, the highest Ki-67 rate was used to assign grade for that biopsy instance using 2017 WHO criteria.

Tumor Classification

For tumor classification, a complete pathology review was performed for each patient using slides from the fine needle aspirate specimen, core biopsies, and resections, and compiled results of ancillary testing (previously performed and for this study) on any available tumor specimens (immunohistochemistry and targeted sequencing analysis by Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets, a custom next-generation sequencing array of 410 cancer-associated genes).^{10, 12, 14} Testing had been performed for 28 patients on material from 21 liver core biopsies, 1 lymph node excision, 5 pancreatectomies, and 1 cell pellet from a liver FNA. Cases lacking material for ancillary testing were reviewed for consensus opinion on morphologic differentiation by 2 out of 3 cytopathologists (C.S.S., D.C, and M.D.R.). All neoplasms were definitively classified as G2, WD-G3, or NEC by applying the criteria outlined in Table 1.

Table 1.

Criteria used to determine differentiation

	Well differentiated	Poorly differentiated
Excluded by	IHC loss of RB RB1 or KRAS mutation	Loss of staining for DAXX or ATRX Mutation for DAXX, ATRX, or MEN1 G1/G2 on prior pancreas resection or biopsy Ki67 <20% on all tumor samples
Supported by (at least one of the following)	G1/G2 on prior pancreas resection or biopsy	Adenocarcinoma component
	Ki67 proliferation rate <20% on all tumor samples Loss of DAXX or ATRX protein expression by immunohistochemistry DAXX, ATRX, or MEN1 mutation 3 cytologists agree well differentiated by morphology	Ki67 proliferation rate >20% on all tumor samples Loss of RB protein expression by immunohistochemistry Protein overexpression of p53 by immunohistochemistry RB1, TP53, or KRAS mutation 3 cytologists agree poorly differentiated by morphology

Results

Clinical characteristics

The study included 65 FNAs from pancreatic neuroendocrine neoplasms obtained from 62 patients: G2: 19 cases, WD-G3: 32 cases, and NEC: 14 cases (PD-S: 6 cases; PD-L: 5 cases; and PD-not otherwise specified (NOS): 3 cases) (Table 2). Comparing the clinical features between G2 and WD-G3, we found no significant differences in age of diagnosis, sex, stage of presentation, anatomic site of FNA, prior resection of primary tumor, and biopsy indication for FNA (Table 2). The majority of clinical parameters were similar between WD and NEC cases with two exceptions: the distribution of anatomic site for FNA differed (P=.04) and the FNA biopsy was more commonly taken at the time of initial diagnosis rather than the time of disease progression (P<.001). Overall, advanced disease at presentation was heavily represented in all groups, with 69% and 93% of respective WD and NEC having presented at stage IV (AJCC 7^th edition) (Table 2). The majority of all FNAs (80%) had a concurrently obtained core biopsy. At the time of FNA, the majority of WD and NEC had either no prior therapy or only hormone based (cytostatic) therapy (WD cases= 66% and NEC cases = 87%) (Table 2). No significant differences were seen in the classes of therapeutic agents used before biopsy (P>.05).

Table 2.

A comparison of clinical and pathologic characteristics between well and poorly differentiated neuroendocrine neoplasms

	G2	WD-G3	G2 vs WD-G3 P value	WD (G2 and G3)	NEC	WD vs NEC P value
Cases No.1	19	32		51	14
Female No. (%)	10 (53)	19 (59)	.7	29 (56)	9 (5)	.8
Age of diagnosis median years ( range)	56 (28–73)	63 (37–82)	.09	61 (28–82)	63 (26–74)	.8
AJCC 7th edition stage at presentation
I	1 (5)	1 (3)	.9	2 (4)	0 (0)	.8
II	1 (5)	3 (9)		4 (8)	0 (0)
III	1 (5)	1 (3)		2 (4)	1 (7)
IV	12 (63)	23 (81)		35 (69)	13 (93)
Anatomic site for FNA specimen No. (%)
Pancreas No.	2 (11)	2 (6)	.3	4 (8)	2 (14)	.04
Liver No.	17 (89)	30 (94)		47 (92)	11 (78)
Other No.	0	0		0	1 (7)
Primary tumor was previously resected No. (%)	5 (26)	10 (31)	.8	15 (33)	2 (13)	.3
FNA and core biopsy were obtained concurrently No. (%)	16 (84)	26 (84)	.9	42 (84)	10 (71)	.5
Biopsy indication for FNA Initial diagnosis	9 (47)	7 (21)	.06	16 (31)	12 (86)	<.001
No. (%)
Evaluation of disease progression No. (%)	10 (53)	25 (78)		35 (69)	2 (14)
Therapy prior to FNA No. (%)
None	11 (58)	17 (53)	.7	28 (55)	12 (86)	.06
Only hormone	4 (21)	5 (16)	.4	9 (18)	0	.2
Alkylating agent	4 (21)	8 (26)	.7	12 (24)	0	.06
Cytotoxic chemotherapy	4 (21)	9 (29)	.7	13 (26)	2 (13)	.5
Targeted agent	2 (11)	5 (16)	.7	8 (16)	0	.2
Platinum based	2 (11)	5 (16)	.7	7 (14)	2 (13)	.6

¹3 patients had two interval biopsies.

Abbreviations: FNA=fine needle aspiration, WD= well differentiated, NEC=neuroendocrine carcinoma, vs= versus, No. = number, G2=grade 2, G3=grade 3

Tumor classification

Approximately half of cases (52%) had multiple criteria supporting classification (Table 3). A prior history of a G1 or G2 neoplasm was the most common supporting criterion for G2/WD-G3, a requirement that was met by 63% of cases. In G2/WD-G3, mutations of DAXX or ATRX were detected in 12 of the 22 (55%) tested cases. MEN1 mutations were present in 14 G2/WD-G3, 11 of which also had DAXX or ATRX mutations. Immunohistochemistry for DAXX and ATRX was performed on 30 cases and loss of either DAXX or ATRX protein expression totaled 6 cases (19%) (Supplemental table 1). There were 3 cases where both mutation testing and immunohistochemistry for DAXX and ATRX was performed, and loss of protein expression corresponded with mutation in 2 cases; the discrepant case is further described in a subsequent paragraph.

Table 3.

Criteria used for classification for each diagnostic category

	G2	WD G3	PD small cell	PD Large cell	PD NOS
Total	19	32	6	5	3
>1 factor supporting WD No. (%)	15 (80)	13 (46)	0	0	0
IHC loss of DAXX or ATRX No. (%)	0	6 (19)	0	0	0
DAXX, ATRX, or MEN1 mutation No. (%)	6 (31)	9 (28)	0	0	0
All Ki67 proliferation No. (%) rates <20%	19 (100)	0	0	0	0
G1/G2 on prior biopsy or resection No. (%)	8 (42)	24 (75)	0	0	0
>1 factor supporting PD No. (%)	0	0	4 (67)	0	2 (67)
Loss of RB IHC staining No. (%)	0	0	3 (50)	1 (20)	2 (67)
RB1, TP53, or KRAS mutation No. (%)	12(5)	22(32)	1 (17)	0	2 (67)
Adenocarcinoma component No. (%)	0	0	0	0	0
Consensus on morphology differentiation1	0	5 (16)	1 (17)	4 (80)	0

Abbreviations: WD - well differentiated, PD - poorly differentiated, No. - Number, NOS - not otherwise specified.

¹Consensus review was used in cases without ancillary supporting information.

²TP53 mutations were seen in 3 cases with additional criteria supporting WD.

Classification as NEC was supported by mutation analysis in 21% of cases and immunohistochemistry (loss of RB1 expression) in 43% of cases (Supplemental tables 1 and 2). A minority (5, 13%) of WD-G3 cases was classified by fulfilling only the criterion of consensus on morphology (16%), but nearly half of NEC cases required classification by consensus review (43%) (Table 3). Immunohistochemistry for p53 was performed on 21 cases and none had abnormal expression.

In the data we collected for classification, there were two instances of conflicting results in the 16 cases where both IHC and molecular testing had been performed. In one instance, a case of WD-G3 had both DAXX and ATRX mutations and DAXX nuclear staining was absent while ATRX was retained. A second instance was a TP53 mutated WD-G3 case without abnormal p53 protein expression. Neither discrepancy had an effect on tumor classification because other criteria supported what we believed was the correct classification. All other results for mutation testing and immunohistochemical staining for protein expression showed correlation including the two cases with an RB1 mutation which had loss of RB1 protein expression.

Three cases met criteria for both PD and WD because they had co-existing mutations of TP53 and/or KRAS and either DAXX, MEN1, and/or ATRX. One case had a prior G2 resection and a Ki-67 proliferation rate <20% and therefore was classified WD. The other two patients had G3 tumors with consensus WD morphology and had prior therapy with alkylating or platinum-based agents.

Morphologic comparison

Morphologic features and Ki-67 proliferation rates were compared across the tiers of differentiation and grade and results are summarized in Tables 4–6. Features classically described in WD-PanNET, including plasmacytoid morphology and smooth nuclear contours associated significantly with G2/WD-G3 cases compared to NEC (Figure 1; Table 4). G2 and WD-G3 showed low rates of molding (5% and 9%, respectively) and pleomorphism (5% and 13%, respectively) (Table 4). As expected, necrosis was not seen in G2 but was present occasionally in WD-G3 (13%). Most of the features that occurred at a significantly higher proportion in WD-G3 compared to G2 reflect proliferative activity, including apoptosis (any (P=.005) and >5/hpf (P=.001)), and nuclear tangles (P=.006) (Table 4, Figure 2). In addition, more WD-G3 cases had abundant cytoplasm compared to G2 (P=.007) (Figure 2; Table 4). As expected, evidence of proliferation was associated with NEC including conspicuous mitoses (>5/10 high power fields (hpf) (P=.03) Other features significantly associated with PD compared to G2/G3 included pleomorphic nuclei (P=.004), angulated nuclei (P<.001), molding (P<.001), lack of plasmacytoid morphology (P<.001), and less round nuclei (P=.01)(Figure 1). Comparing WD-G3 with PD-L, we found WD-G3 had less pleomorphic nuclei (P=.04), more plasmacytoid cells (P=.03), and fewer cases with necrosis (P=.04) (Figure 2; Table 5). As expected, frequent molding and lack of abundant cytoplasm were seen in PD-S (Figure 1; Table 6). There were too few cases of NEC to compare morphological features within the NEC category.

Table 4.

Comparison of G2, WD-G3, and PD neuroendocrine neoplasm cytologic features

	Well differentiated G2	Well differentiated G3	Poorly differentiated	P value
Total (n=65)	19	32	14
Ki67% median (range)	11% (3.2–17)	40% (21–89)	62% (25–95)	<.001
Large nuclear size No. (%)	0 (0)	3 (9)	2 (14)	.2
Nuclear pleomorphism No. (%)	1 (5)	4 (13)	7 (50)	.004
Round nucleus No. (%)	14 (74)	19 (59)	3 (21)	.01
Smooth nucleus contour No. (%)	18 (95)	30 (94)	9 (64)	.02
Angulated nuclei No. (%)	3 (21)	12 (38)	12 (87)	<.001
Fine chromatin No. (%)	3 (15)	3 (4)	1 (7)	.8
Single prominent nucleolus No. (%)	1 (5)	4 (13)	2 (14)	.6
Plasmacytoid morphology No. (%)	16 (84)	24 (75)	1 (7)	<.001
Abundant cytoplasm No. (%)	11 (17)	23 (72)	3 (21)	.007
Nuclear tangles No. (%)	8 (42)	24 (75)	13 (93)	.006
Molding No. (%)	1 (5)	3 (9)	11 (79)	<.001
Apoptosis No. (%)
Present	11 (58)	28 (86)	14 (100)	.005
>5/ 10 hpf	3 (18)	14 (44)	11 (79)	.001
Mitoses No. (%)
Present	10 (27)	18 (56)	9 (64)	.8
>5/10 hpf	1 (5)	6 (3)	6 (43)	.03
Necrosis	0 (0)	4 (13)	6 (43)	.003

Table 6.

Morphologic features and Ki67 proliferation rate for poorly differentiated neuroendocrine carcinoma subtypes

	Small cell	Large cell	Type not otherwise specified	P-value1
Total (n=15)	6	5	3
Ki67% median (range)	80%(63–95)	45% (25–61)	70% (30–80)	.003
Large nuclear size No. (%)	0	2 (40)	0
Nuclear pleomorphism No. (%)	2 (33)	4 (80)	2 (67)
Round nucleus shape No. (%)	0	2 (40)	1 (33)
Smooth nucleus contour No. (%)	2 (40)	4 (67)	2 (67)
Angulated nuclei No. (%)	6 (100)	3 (60)	3 (100)
Fine chromatin No. (%)	0 (0)	0 (0)	1 (33)
Single prominent nucleolus No. (%)	0	2 (40)	0
Plasmacytoid cells No. (%)	0	1 (20)	0
Abundant cytoplasm No. (%)	0	3 (60)	0
Nuclear tangles No. (%)	6 (100)	4 (80)	3 (100)
Molding No. (%)	6 (100)	2 (40)	3 (100)
Apoptosis No. (%)
Present	6 (100)	5 (100)	3 (100)
>5/ 10 hpf	6 (100)	3 (60)	2 (100)
Mitoses No. (%)
Present	3 (50)	3 (60)	3 (100)
>5/10 hpf	3 (50)	2 (40)	1 (33)
Necrosis	3 (50)	3 (60)	0 (0)

¹Numbers were too limited to test.

Figure 1.

A. A typical well differentiated (WD) G2 neoplasm with round nuclei and smooth contours B. G2 neoplasm with abundant cytoplasm and nuclear tangles. C-E. Three poorly differentiated neuroendocrine carcinomas, large cell type on Thinprep (C), cell block (D), and Diff Quik preparations (E). F-G. Two poorly differentiated neuroendocrine carcinomas, small cell type on Thinprep (F) and Diff Quik (G) preparations. H-I. Poorly differentiated neuroendocrine carcinomas, not otherwise specified. KRAS, TP53, and RB1 mutations were detected in the neoplasm in Figure 1I.

Figure 2.

The spectrum of well differentiated G3 pancreatic neuroendocrine tumors in twelve representative cases. A-C. Plasmacytoid forms were commonly seen. E-G. Many cases also had non-plasmacytoid cellsnuclear tangles, apoptotic debris, and angulated nuclei. H-I. Mitotic activity was conspicuous. J-M. Abundant cytoplasm and pleomorphism were often seen.

Table 5.

A comparison of morphologic features between well differentiated G3 and poorly differentiated carcinoma

	Well differentiated G3	Poorly differentiated large cell	P value
Total	32	5
Large nuclear size No. (%)	3 (9)	2 (33)	.1
Pleomorphic nuclei No. (%)	4 (13)	3 (50)	.04
Round nucleus No. (%)	19 (59)	2 (33)	.6
Smooth nucleus contour No. (%)	30 (93)	4 (67)	.4
Angulated nuclei No. (%)	12 (38)	3 (50)	.4
Fine chromatin No. (%)	3 (9)	0 (0)	.9
Single prominent nucleolus No. (%)	4 (13)	2 (33)	.2
Plasmacytoid cells No. (%)	24 (75)	1 (20)	.03
Abundant cytoplasm No. (%)	23	3	.59
Nuclear tangles No. (%)	24 (75)	4 (67)	.9
Molding No. (%)	3 (9)	2 (33)	.12
Apoptosis
Present No. (%)	28 (88)	5 (100)	.9
>5/ 10 high power fields No. (%)	14 (44)	3 (60)	.6
Mitoses
Present No. (%)	18 (56)	3 (60)	.9
>5/10 high power fields No. (%)	6 (19)	2 (40)	.3
Necrosis No. (%)	4 (13)	3 (60)	.04

Plasmacytoid morphology was absent in 11 (22%) of G2/WD-G3 and 9 of these cases were G3. The most common features seen in WD-NET without plasmacytoid morphology were angulated nuclei, nuclear tangles, and apoptosis (Figure 2).

The median (range) of Ki-67 proliferation rates for each tier were as follows: G2: 11% (3.2–17): WD-G3: 40% (21–89); PD-S: 80% (63–95); PD-L: 39% (25–61); and PD-NOS: 70% (30–80). Notably, there were significant differences among the NEC subtypes and WD-G3 had a similar median and range to PD-L.

Discussion

Separating WD-NET from NEC has major prognostic implications and directly impacts therapeutic decision making.⁸ Molecular analysis and immunohistochemistry for loss of mutant protein expression (e.g. DAXX, ATRX, MEN1, RB1) can be very useful in supporting differentiation when a characteristic abnormality is identified, but these tests are not universally available, may not be feasible, and may not identify a genotype with known significance.¹⁰ In order to enhance the cytologic diagnosis of these neoplasms, we compared G2 and WD-G3 neuroendocrine cytomorphology. We gathered as much ancillary information as was feasible using available tumor material and records of prior testing to provide the best classification of the included neoplasms.

Classic descriptions of WD-NET cytology point out plasmacytoid forms with round nuclei resembling benign islet cells.¹⁵ The majority of G2 and WD-G3 were in keeping with this description, but the increased proliferation of WD-G3 is accompanied by morphologic changes, which might be expected with increased proliferation as they are also seen in NEC (apoptosis, mitoses, and nuclear tangles). In addition to the presence of plasmacytoid forms, abundant cytoplasm was much more common in WD-G3 cases than G2 (Figure 2). Among the features seen at a higher rate in WD-G3 over G2, abundant cytoplasm was the only feature associated with WD-G3 and not NEC (Table 4). Although, it should be noted this feature was uniformly present in the 5 cases of PD-L. Comparing WD-G3 and PD-L, we found similar Ki-67 proliferation rates but WD-G3 was associated with a higher rate of plasmacytoid morphology and less pleomorphic nuclei and necrosis; however, we emphasize that necrosis was present in both but usually focal in WD-G3. Non-plasmacytoid WD tended to have other atypical features including nuclear angulation, nuclear tangles, and apoptosis which could present a problem with morphologic classification, particularly if they are G3.

Most NEC cases were biopsied at initial disease presentation (86%) from a liver metastasis (78%). The differences between G2/WD-G3 and NEC included nuclear atypia (e.g. nuclear pleomorphism and molding) and features expected to be seen in tumors with high proliferation (e.g. necrosis, apoptosis, mitoses, and nuclear tangles). In contrast, plasmacytoid morphology and round nuclei were rare to infrequent in NEC.

Many of the WD-G3 FNAs were biopsies of disease progression, and due to concern that morphologic features might be influenced by therapy, we recorded therapy classes used prior to biopsy and we found similar classes of agents were used in G2 and WD-G3.

Our aim was to describe morphology for neoplasms that were well characterized for grade and differentiation in this study. A Ki-67 proliferation rate was available for all tumors and it was used to determine tumor grade because it is predictive in cytology and biopsy specimens.¹³ There is no gold standard for determining differentiation, so we collected as much information as feasible to support tumor biology. A prior history of G1/G2 neoplasm was obtained in the majority of WD-G3. When there was no prior history of G1/G2 neoplasm and proliferation was >20%, targeted genotyping plus immunohistochemistry provided support for WD-G3 in several cases. Loss of DAXX or ATRX protein expression was included as support for WD because it has been shown to be a reliable surrogate test for mutation status, however we identified one instance where ATRX protein expression was retained despite the presence of a mutation.

We acknowledge that most neuroendocrine neoplasms encountered in routine practice will likely be straightforward to diagnose. WD-G3 cytology often maintains at least focal plasmacytoid morphology but can also have variable mitotic activity, conspicuous apoptosis, abundant cytoplasm, prominent nucleoli, molding, necrosis, and large nuclei which may raise concern for NEC, but in the context of a prior history of PanNET, a diagnosis of NEC should not be made.

In summary, using a well-characterized cohort of pancreatic neuroendocrine neoplasms, we have identified several morphologic features associated with WD-G3 neoplasms that may assist in the morphologic distinction of differentiation for this newly classified entity. De novo G3 neoplasms that are non-plasmacytoid, have pleomorphic nuclei, or have abundant cytoplasm will likely present the greatest challenge for classification and ancillary studies such as mutational analysis and/or immunohistochemistry for mutated proteins are recommended.