The incidence of Grade (G) 3 Gastroenteropancreatic (GEP) neuroendocrine neoplasms (NENs) has increased over the last decades [1, 2]. Furthermore, in the largest epidemiological Italian series from AIRTUM (Italian association of tumor registry), poorly-differentiated (PD) GEP NENs showed an unexpected higher incidence than well-differentiated (WD) [3] NENs. This result seems in contrast with clinical practice; the methods applied, notably different ICD-O-3 codes, may have conditioned the results.

The G3 GEP NENs category has been extensively investigated over the last years [4,5,6,7,8,9,10,11]. Several reports remarked that some GEP NENs defined as G3 on the basis of their Ki-67 level in accordance with the 2010 WHO classification [12] were actually WD rather than PD and they correlated with a significantly better survival compared with PD. This new entity, named “Neuroendocrine tumors (NET) G3”, has been forecasted to be included in the new pancreatic NENs classification [13]. Poorly differentiated pancreatic GEP NENs will be defined as neuroendocrine carcinomas (NECs) G3. Beyond tumor differentiation (WD vs. PD), even Ki-67 level (21–55% vs. >55%) has been reported as a factor able to separate different populations in terms of survival [14]. These data are important not only from a prognostic point of view but also for their implications in clinical practice [8, 15, 16].

On the basis of the aforementioned considerations, it is clear that each pathologist reporting a G3 GEP NEN should describe both tumor differentiation and Ki-67. However, while the assessment of Ki-67 can be standardized provided that the WHO indications are followed [12], tumor differentiation evaluation is less reproducible [17], especially in the absence of a specific pathology training.

When the distinction between WD and PD is not possible even after an expert review, the use of specific immune-histochemical markers, such as TP53, Rb1 and somatostatin receptor type 2 (SSTR2A) and 5 (SSTR5), can be helpful, as suggested by some authors [3, 17,18,19]. Loss of TP53 and Rb1 oncosuppressor function is associated with PD phenotype, whereas expression of SSTR2A and 5 is indicative of WD. However, an inconsistency between molecular biomarkers and tumor differentiation in a small subset of patients has been reported [18]. These controversial issues imply that both morphological and/or molecular approaches may have limitations, therefore we strongly suggest that each controversial case should be reviewed and discussed within a multi-specialist team for an appropriate therapeutic decision.

Although from a prognostic point of view it looks quite clear that the G3 GEP NENs category should be differentiated in separate sub-categories on the basis of tumor differentiation and Ki-67, it is less clear how much these factors alone can affect therapeutic choices. To date, several studies focused on therapeutic approaches accordingly to WD or PD G3 GEP NENs. However a few of them are uniquely based on the diagnostic report, lacking a central pathological and molecular revision, which may result in limited reproducibility, [6, 8] while others focused on a thourough pathology characterization without clinical and therapeutic information [17, 18]. Current evidence about G3 GEP NENs heterogeneity is still far from producing validated specific therapies for specific sub-categories, however it represents a good basis to suggest on one hand a different clinical approach for WD vs. PD G3 GEP NENs and on the other hand to design prospective therapeutic clinical trials on solid hypotheses. Unfortunately the design of several prospective ongoing or incoming clinical trials on G3 GEP NENs could not lead to definitive conclusions.

The ongoing NORDIC study with everolimus and temozolomide (NCT02248012) is a first-line phase II and uses one biologic together with one chemotherapeutic agent. Therefore, although investigating a well defined sub-category (20–55% Ki-67) it will not clarify whether this therapy is superior to platinum/etoposide and biological therapy is better than chemotherapy. A randomized design with pre-planned analyzes of sub-categories would have allowed to draw some clinical conclusions. Likewise the US trial with modified FOLFIRINOX is a phase II and it is enrolling all GEP NENs with >20% Ki-67 and/or 20 mitoses/10 high power fields without specifying if a central pathology review is planned and if tumor differentiation and/or Ki-67 level sub-categories will be evaluated (NCT03042780). More relevant to the clinical practice is the ongoing US randomized phase II trial comparing cisplatin/etoposide with capecitabine/temozolomide in patients with 20–100% Ki-67 advanced GEP NENs. Small-cell diseases are excluded; a centralized pathology review and a correlation between tumor differentiation (WD vs. PD) and clinical outcome has been planned (NCT02595424).

The new immune check point inhibitors, which showed efficacy in several types of neoplasms, were not extensively investigated in NENs so far. However the G3 GEP NENs can be considered for inclusion in some ongoing clinical trials. Some of them are specific for NENs not requiring a molecular prescreening (NCT02955069), whereas some others are based on microsatellite instability (MSI) or mismatch repair (MMR) deficiency detection. By means of PCR or IHC of MMR proteins an MSI was reported in around 10% of GEP NENs, mainly colorectal [19, 20].

Until we will have definite evidence about the efficacy of a therapy instead of another it is recommended that each clinician translates the current evidence into a cautious and thoughtful clinical approach to patients with a >20% Ki-67 GEP NEN, notably availing him/herself of other tools, such as clinical behavior of the disease, molecular imaging and clinical status of the patient and using clinical judgment.