An analysis of 130 neuroendocrine tumors G3 regarding prevalence, origin, metastasis, and diagnostic features

We analyzed clinicopathological and immunohistochemical features in 1513 cases of a total of 1745 NENs, collected in our consultation center between April 2009 and April 2021, including NETs, NECs, and mixed neuroendocrine-non-neuroendocrine neoplasms (MiNENs). Two hundred thirty-two of 1745 NENs, encompassing cases of neuroendocrine microadenoma, insulinomatosis, glucagon cell hyperplasia and neoplasia, medullary thyroid carcinoma, pituitary adenoma, parathyroid neoplasm, Merkel cell carcinoma, and neuroendocrine precursor lesions (e.g. gastric neuroendocrine cell hyperplasia, pulmonary diffuse idiopathic neuroendocrine cell hyperplasia, duodenal G-cell hyperplasia, and neoplasia) as well as ectodermal NENs (paraganglioma, pheochromocytoma, olfactory neuroblastoma, and neuroblastic neoplasm), were excluded, because they were of no relevance for this study. The same consultation series was used to analyze mesenchymal neoplasms with neuroendocrine features as mimickers of NENs [13].

The patients’ age and sex and the origin of the samples are shown in Supplemental Table 1. Biopsies accounted for 44% (N = 645) of all specimens and were mostly obtained from the liver (N = 239).

The site of primary was determined in 1361 patients based on the organ tissue surrounding the tumor, the immunohistochemical findings indicating the origin of the lesion in case of metastases, and the information provided in the referral letters.

Hematoxylin and eosin (HE) staining and periodic acid-Schiff (PAS) staining were done on 2-µm-thick sections from formalin-fixed paraffin-embedded tissues. Immunohistochemical stainings were performed using a fully automated slide preparation system (Benchmark XT, Ventanta/Roche, Arizona, USA). Details regarding the immunohistochemical stainings are given in Supplemental Table 2. Immunohistochemical expression of CDX2, ISLET-1, TTF-1, und serotonin was recorded as positive when more than 10% of the total tumor cells were stained. SST2 was evaluated using a previously reported four-tiered scoring system [14]. Briefly, positive tumors were scored 2 (membranous incomplete staining in less than 50%) or 3 (circumferential membranous staining in more than 50%) [14]. Abnormal p53 expression was defined as moderate to strong nuclear immunoreactivity in more than 20% of tumor cells [15]. Rb1 nuclear expression in less than 10% of tumor cells was defined as loss/abnormal expression [5]. Nuclear Ki67 labeling was counted in more than 500 tumor cells in the area with highest density (hot spot), and its percentage was given as Ki67 index (%). Immunohistochemistry of p53, Rb1, and SST2 was performed in 55% (72/130), 35% (46/130), and 66% (86/130) of all NET G3 cases, respectively. All cases were reviewed at least by two endocrine and pancreas pathology experts including AK, BK, MS, WW, and GK. When there was a disagreement on the diagnosis, consensus was reached after joint discussion at a multiheaded microscope. The final diagnosis of all cases (including the cases before 2017) was based on the current WHO classifications of tumors of the digestive system and tumors of the endocrine organs [1, 2]. All well-differentiated NENs, no matter what organ they came from, were classified and graded as NETs [16]. MiNENs were graded in low, intermediate, and high according to a recent proposal by La Rosa et al. [17] (Supplemental Table 1).

NET was diagnosed, when the tumor showed an organoid architecture with either solid, solid-trabecular, or solid-glandular patterns (Fig. 1A), occasional intratumoral necroses, rather uniform nuclei with coarse granular chromatin and small nucleoli (Fig. 1B). The diagnosis of small cell-type NEC was based on a diffuse cell sheet pattern with geographic necroses. The cells had fusiform, occasionally also enlarged nuclei with finely granular chromatin, scant cytoplasm, and nuclear moulding. In the large cell-type NEC, the architecture was usually characterized by irregular nests (Fig. 1C) that often had necroses and occasionally showed peripheral palisading. The nuclei had prominent nucleoli with vesicular chromatin (Fig. 1D) [2].

From the referral letters, the requests of the consultation and the suggested or suspected diagnoses were extracted and categorized as follows: assessment of the neoplasms’ (1) classification into NET or NEC, (2) entity (i.e., NEN versus non-NEN); (3) primary site; (4) SST2 expression; and (5) other criteria such as infiltration into adjacent tissue, presence of lymphovascular invasion, margin status, and/or treatment options (Supplemental Table 1).

One hundred thirty NETs G3 were identified among 1513 NENs, including 1103 NETs (71%), 255 NECs (17%), and 155 MiNENs (10%). NETs G3 accounted for 12% of all NETs and 9% of all NENs (Supplemental Table 1).

NET G3 samples were mainly of metastatic origin (90/130, 69%), with the liver (67/90, 74%) as leading site of metastasis, followed by lymph nodes (10/90, 11%) and various other organs (13/90, 14%). Most samples of primary NETs G3 (N = 40) were obtained from the pancreas (40%), followed by the lung (23%), the stomach (19%), and the rectum (8%) (Table 1). Single NETs G3 came from the duodenum/papilla of Vater (5%), ileum (3%), and one from the prostate (Table 1). In contrast to NETs G3, NETs G1/G2 (739/973, 76%) and MiNENs (120/155, 77%) were more often obtained from primary sites, while NECs were equally frequent in primary (125/255, 49%) and metastatic sites (130/255, 51%).

Based on the evaluation of the information on the tumor origin provided in the referral letters, and the expression patterns of site-specific markers conclusions on the origin of metastasis were possible in 337/489 (69%) cases, including 67 of 90 (74%) metastatic NETs G3. In the remaining 152 cases (including 23 NETs G3), the primary site of the tumor remained unknown. Table 2 summarizes the origin and prevalence of NET G3 liver metastases (N = 67) compared to that of NET G1/G2 (N = 135), NEC (N = 54), and MiNEN (N = 23). Most hepatic metastases of NET G3 could be assigned to the pancreas (49%, 33/67) and the lung (13%, 9/67). Only few metastases could be ascribed to the stomach, rectum, and other sites such as the breast and presacral region. The metastatic rate of pancreatic NETs G3 (49%) was higher than that of NETs G1/G2 (52/135, 39%), NECs (5/54, 9%), and MiNENs (7/23, 30%). Liver metastases from ileal NETs (5/67, 7%) were more commonly of NET G1/G2 grade (27/135, 20%) than of NET G3 grade. Among the hepatic metastases from lung NENs (N = 38), NEC metastases (17/38, 31%) exceeded NET G3 (9/38, 23%) and all the other NEN metastases. No NET G3 metastases could be ascribed to duodenal, appendiceal, and colonic primaries. This was not surprising, since NET G3 primaries in these organs were rare or absent (see below). Liver NEN metastases, whose primary remained unknown, mostly concerned NECs (27/54, 50%) followed by NETs G1/G2 (42/135, 31%), NETs G3 (14/67, 21%), and MiNENs (17/23, 4%) (Table 2). In metastases, in which the origin was uncertain, immunohistochemistry was performed. A pancreatic origin was strongly suggested by the expression of ISLET-1 in 88% (23/26) of NET G3 liver metastases, an ileal origin by the expression of CDX2 and serotonin in 80% (4/5) and 100% (5/5), respectively, and a pulmonary origin by TTF1 in 76% (6/7).

When the numbers of NET G3 primaries and NET G3 metastasis were considered together, it is obvious that most NETs G3 occurred in or derived from the pancreas (54/130, 42%), followed by the lung (26/130, 20%), stomach (10/130, 7%), ileum (7/130, 5%), rectum (4/130, 3%), and other sites including duodenum/papilla of Vater (2/130, 2%), presacral lesion, breast, and prostate (1/130, 1%, each). Eighteen percent (23/130) NETs G3 remained unknown regarding their origin (Table 1).

Ki67 index values of NET G3 ranged from 21 to 70% with a median of 30%. This median Ki67 index separated NETs G3 from NECs and MiNENs (p < 0.0001), while the range, as expected, showed a broad overlap between the three groups (Table 3). There was no difference between the median Ki67 index of NET G3 primaries versus metastases. The analysis of p53 and Rb1 revealed that these markers were not abnormally expressed in NET G1/G2 and only rarely in NET G3 (p53 8%, Rb1 7%) in contrast to NECs and MiNENs, in which abnormal p53 and Rb1 expression was high, reaching 80% and 60%, respectively (p < 0.0001 for both Table 3 and Fig. 2). In subsets of NETs G3 from non-pancreatic and non-pulmonary organs, abnormal p53 and Rb1 expression was observed in 10% and 20%, respectively. When we looked at NET G3 primaries and metastases separately, abnormal expression of p53 and Rb1 was only observed in specimens from metastases, mainly from the liver. SST2 expression rate was high in NETs G3 from the pancreas (70%) and from other digestive organs (59%, including the ileum, colon, and rectum) and low in NETs G3 from the lung (12%, Table 3). In the single NENs from the breast (Fig. 3a–c), prostate, and presacral regions (Fig. 3d–f), that corresponded to NETs G3, a normal expression of p53 and Rb1 and a strong SST2 expression (Fig. 3e) were observed. The breast NET G3, in addition, expressed ER and PgR, the presacral NET G3 serotonin (Fig. 3f), while the prostatic NET G3 failed to express NKX3.1.

In total, 2120 requests were evaluated. The most frequently addressed requests concerned the differential diagnosis of NET versus NEC (34%), followed by the distinction of NEN from non-NEN (30%), the presumable origin of the primary (28%), the expression of SST2 (7%), and other minor issues such as the hormone expression (2%) (Supplemental Table 1). In NET G3, the issues of tumor origin (63%), differential diagnosis NEN versus non-NEN (39%), and NET versus NEC (37%) played the main role.

In 1121 of 1513 NENs, a diagnosis was suspected or suggested by the submitting pathologist. The concordance rates between referral diagnosis and final diagnosis per tumor entity are listed in Supplemental Table 3. Lowest concordance was found in NETs G3 (20%) compared to all other entities, i.e., NET G1/G2 (66%), NECs (39%), and MiNENs (23%) (for details, see Supplemental Table 3). In NENs identified as NET G3, NECs (49%, 40/81) or NETs not otherwise specified (43%, 35/81) accounted for most inconsistent diagnoses (for details, see Supplemental Table 3). NET G3, on the other hand, was suggested as diagnosis in 10 NETs G2, 5 NECs, 7 MiNENs, and various other carcinomas including medullary thyroid carcinoma, adenocarcinoma, squamous cell carcinoma, acinar cell carcinoma, and SMARCB1-deficient neoplasm (data not shown). Of the 82 NETs G3 with an inconsistent diagnosis, 59 cases were from the years 2009 to 2017 (6.7 cases per year) and 23 cases from 2018 to 2021 (5.8 cases per year).