Pancreatic neuroendocrine tumors (PanNETs) are a heterogeneous group of well-differentiated neuroendocrine neoplasms that arise in the pancreas, accounting for approximately 2%–5% of all pancreatic tumors. The majority of PanNETs show typical organoid growth patterns, such as trabecular, solid, nested, or glandular architectures. However, these tumors are increasingly recognized as morphologically heterogeneous. Over the past two decades, a growing number of unconventional morphologic variants have been reported, including hepatoid, oncocytic, plasmacytoid/dyscohesive, lipid-rich (clear-cell), paraganglioid, and other rare patterns. The majority of these reports are isolated case reports or small series [1–27].

These variants may mimic other primary pancreatic neoplasms or metastatic tumors, particularly in small biopsy or metastatic specimens, leading to diagnostic challenges. Moreover, accumulating evidence suggests that certain morphologic patterns are associated with distinct clinical behaviors, indicating that morphologic heterogeneity may reflect underlying biological differences in PanNETs.

The most comprehensive systematic evaluation of morphologic variants in PanNETs to date was performed by Xue et al. [1]. In that study, oncocytic, hepatoid, plasmacytoid/dyscohesive, and lipid-rich variants were associated with adverse pathologic features, suggesting that morphologic “aggressiveness” may reflect biologic behavior. However, it remains uncertain whether morphologic variant status offers prognostic information that is independent of established parameters, including tumor grade, Ki-67 index, and nodal status.

In this context, systematic evaluation of well-characterized institutional cohorts may help further define the clinicopathologic relevance of PanNET morphologic variants. In the present study, we retrospectively analyzed 82 resected PanNETs to characterize morphologic variants, determine their prevalence, and examine their clinicopathologic associations and survival observations.

The present study retrospectively evaluated 82 resected, well-differentiated PanNETs from a single institution, identified morphologic variants using literature-based criteria, and examined their associations with clinicopathologic parameters and survival outcomes, with a particular focus on variants previously linked to aggressive behavior.

In our cohort, nearly one-quarter of tumors exhibited non-classical morphologic features, a frequency comparable to previous large-scale series [1]. The spectrum of variants largely overlapped with those described in prior studies, supporting the reproducibility of these morphologic patterns in routine practice [1, 6, 13, 14, 17, 22, 30, 31]. Notably, several variants reported in the literature, such as ductuloinsular, mammary tubulo-lobular carcinoma-like, and sclerosing patterns, were not observed in our cohort, likely reflecting their extreme rarity rather than their true absence [1, 4, 5].

Oncocytic and hepatoid morphologies have been described as distinct PanNET variants based on architectural, cytologic, immunophenotypic, and ultrastructural features, including mitochondria-rich eosinophilic cytoplasm and, in hepatoid cases, a higher nuclear-to-cytoplasmic ratio, cytoplasmic lipid-like deposits, and hepatocellular marker expression [1, 14–17, 32]. Previous studies have also associated these morphologies with more aggressive clinicopathologic features [1, 17]. A more recent review of the literature by Xue and Reid, who were involved in the original characterization of these categories, now considers hepatoid morphology a subtype within the broader oncocytic spectrum rather than a distinct variant, based on their shared eosinophilic cytoplasm and overlapping morphologic features [32]. In our cohort, tumors with abundant eosinophilic cytoplasm similarly showed overlapping oncocytic and hepatoid features, with variable, often focal hepatocellular marker expression and no steatosis. In keeping with emerging literature and their broadly similar clinicopathologic behavior, these findings support interpreting such tumors as part of an eosinophilic morphologic spectrum rather than as rigidly distinct entities.

In the dyscohesive/plasmacytoid-rhabdoid variant, ultrastructural studies have shown that the intracytoplasmic globules underlying rhabdoid morphology correspond to paranuclear aggregates of intermediate filaments and may coexist with conventional neuroendocrine areas, supporting the interpretation of this pattern as a morphologic phenotype within the PanNET spectrum rather than a distinct entity [26, 27]. Solid pseudopapillary neoplasm-like architecture, described in a subset of cases [1], was not observed in our cohort.

The lipid-rich (clear cell) variant is a rare but well-recognized morphologic pattern. Although historically linked to von Hippel-Lindau syndrome, similar features have also been reported in sporadic PanNETs [6, 9, 11, 12]. In our cohort, which excluded syndromic cases, lipid-rich morphology was identified in two tumors, supporting its occurrence in sporadic PanNETs. We also observed focal clear cell-like changes adjacent to fibrotic areas, likely related to prior fine-needle aspiration, in otherwise conventional PanNETs, highlighting a potential diagnostic pitfall and emphasizing the importance of distinguishing true lipid-rich changes from reactive or procedure-related cytoplasmic clearing. The peliotic/angiomatoid and pseudoglandular variants are uncommon morphologic patterns, with limited data, and their clinicopathologic significance remains uncertain [1].

The paraganglioid variant is rare and represents a well-recognized diagnostic challenge due to its resemblance to paraganglioma [1, 30]. In our cohort, this variant was identified in two cases, one of which demonstrated prominent pleomorphic (symplastic) tumor cells, a feature described both in paraganglioid PanNETs and as a distinct pleomorphic PanNET variant [1, 21]. Such pleomorphism may further complicate diagnosis by mimicking high-grade carcinoma despite preserved well-differentiated neuroendocrine features and low proliferative activity. Although S100-positive sustentacular cells have been reported in some cases, this finding is inconsistent and typically less prominent than in true paragangliomas [1, 30]. In our cases, a sustentacular staining pattern was not identified; instead, diffuse S100 positivity was observed within tumor cells, further supporting their epithelial neuroendocrine differentiation rather than a true paraganglionic lineage.

From a diagnostic perspective, these morphologic variants represent important pitfalls, particularly in limited biopsy or metastatic settings. Eosinophilic tumors may mimic hepatocellular or oncocytic neoplasms; dyscohesive/plasmacytoid variants may resemble poorly cohesive/signet-ring cell carcinoma and hematolymphoid malignancies; lipid-rich tumors may simulate clear cell neoplasms, including adrenocortical tumors and clear cell renal cell carcinomas; pseudoglandular tumors may mimic adenocarcinoma; and paraganglioid tumors may be confused with paraganglioma. A lack of awareness of these morphologic variants may lead to the exclusion of neuroendocrine tumors from the differential diagnosis, resulting in misdiagnosis and inappropriate clinical management. Careful cytologic assessment, recognition of residual conventional PanNET areas, and confirmation with neuroendocrine markers are therefore essential for an accurate diagnosis in both primary and metastatic settings.

From a clinicopathologic standpoint, morphologic variants previously categorized as more aggressive in the literature were associated with higher Ki-67 indices and higher WHO grades in our cohort, consistent with increased proliferative activity. Tumors with these morphologic patterns showed a trend toward larger tumor size, although the difference did not reach statistical significance. These findings are broadly consistent with previous reports describing associations between selected morphologic variants and adverse histopathologic features [1, 14]. A prior study also reported higher rates of nodal and metastatic adverse events among aggressive variants [1]. We did not observe significant associations between morphologic variant status and nodal involvement or distant metastatic disease at diagnosis. This difference may reflect differences in cohort composition, endpoint definitions, and limited statistical power. Notably, aggressive morphologic variants were also more frequent in male patients, consistent with prior reports; although the biological basis of this association remains unclear [1, 14].

Despite their association with adverse histopathologic features, morphologic variants previously categorized as aggressive were not significantly associated with OS or RFS in our cohort, although a trend toward a less favorable OS was observed. Established prognostic parameters, including the Ki-67 index and nodal status, showed associations with recurrence outcomes, with nodal status remaining independently associated in multivariable analysis. Notably, while the Ki-67 index did not reach conventional statistical significance in the multivariable models, its point estimate suggested a possible association with poorer RFS. This finding is consistent with the established prognostic role of proliferative activity in PanNETs, although confirmation in larger cohorts with greater event numbers is warranted. However, Xue et al. reported higher rates of adverse outcomes among morphologic variants categorized as aggressive in their study [1]. Differences between studies may partly reflect methodological variations, as that study assessed composite adverse outcome measures that incorporated nodal, metastatic, and follow-up adverse events, whereas our study evaluated conventional time-to-event endpoints. Nevertheless, the limited number of variant cases and survival events in our cohort restricts definitive conclusions regarding the prognostic significance of specific morphologic variants, which should be further evaluated in larger, multi-institution cohorts.

The lipid-rich variant warrants particular mention, as it has been categorized among aggressive morphologies in prior reports based on descriptions of adverse outcomes despite low Ki-67 indices [1]. In our cohort, lipid-rich tumors similarly demonstrated a low proliferation profile (grade 1-2; median Ki-67 index, 3%). Both tumors were small (1.2 and 2.5 cm); neither showed distant metastasis, and no recurrence occurred during prolonged follow-up (approximately 4.8 and 7.6 years), despite one case being node-positive. Although no adverse outcomes were observed in our limited number of cases, the small sample size precludes definitive conclusions regarding the prognostic behavior of lipid-rich PanNETs. Variability in tumor size, stage distribution, and statistical power may partly explain the prognostic discrepancies reported across studies.

Similarly, morphologic variants categorized as indeterminate, particularly the peliotic/angiomatoid pattern (three cases), did not demonstrate an overtly aggressive clinical course in our cohort, although case numbers were limited, and only a single recurrence without disease-related death was observed. These observations further underscore the heterogeneity of morphologic variants and suggest that non-classical morphologic patterns may not uniformly correspond to similar clinicopathologic behavior. Taken together, our findings support the classification of morphologic variants as a useful component of histopathologic characterization, while suggesting that their prognostic significance should be interpreted cautiously and in the context of established clinicopathologic parameters.

The strengths of the present study include a well-characterized cohort of resected PanNETs with a detailed histopathologic review, long-term follow-up, and a systematic assessment of morphologic variants using classification based on previously published criteria. Unlike prior studies, we evaluated OS and RFS as conventional time-to-event outcomes, allowing assessment of prognostic associations beyond composite metastatic outcome measures. However, several limitations should be acknowledged. This was a retrospective single-institution study with a limited number of morphologic variant cases, reflecting the rarity of these patterns in resected PanNET cohorts and restricting statistical power for subgroup analyses. In addition, the relatively low number of recurrence and survival events may have reduced the sensitivity in detecting subtle or borderline prognostic effects of morphologic variant status. Finally, more granular outcome measures, such as disease-specific survival or separate analyses of locoregional versus distant recurrence, were not evaluated.

In conclusion, morphologic variants of PanNETs represent an important component of the histologic spectrum of PanNETs and are associated with distinct clinicopathologic features. Recognition of these variants remains important in diagnostic practice, particularly in limited biopsy or metastatic settings where they may mimic other neoplasms. While certain morphologic patterns were associated with adverse histopathologic features in our cohort, their prognostic significance remains uncertain and should be interpreted cautiously in the context of established clinicopathologic parameters such as nodal status and proliferation index. Larger, multi-institutional studies with greater numbers of variant cases and outcome events are needed to better clarify the prognostic relevance of specific morphologic variants.