A uniform classification for p-NENs has been lacking to stratify p-NENs into prognostic groups, although several varying systems have been devised, analyzed, and compared for p-NENs [18, 19]. In 2010, the WHO grading system distinguished G1 p-NETs from G2 p-NETs and “G3 p-NECs” based on mitotic rate and Ki-67 proliferative index [6, 7], which has been proven to be prognostic for the OS of p-NENs [8–12]. Although the WHO 2010 grading classification for p-NENs represented an important step toward adopting a uniform grading system with widespread acceptance, its weakness appeared gradually.
Firstly, WHO suggested the higher of the two parameters be used to assign the final grade (typically, the Ki-67 index often pointed to the higher WHO grade) when mitotic rate and Ki‐67 index were sometimes discordant . This would inevitably increase the number of cases of “G3 p-NECs”, which was demonstrated by Basturk et al. that mitotic G2/Ki‐67 “G3 p-NECs” biologically behaved more like mitotic G2/Ki‐67 G2 p-NETs . They found that p-NENs with a Ki-67 proliferative index > 20%, if well-differentiated, were more aggressive than G2 but significantly less aggressive than “G3 p-NECs” with poorly differentiated features (large or small cell type) . Furthermore, the WHO 2010 grading classification just used the terminology “high-grade” and “poorly-differentiated” interchangeably for neoplasms in the G3 category, while recent studies have further focused on the heterogeneity of “G3 p-NECs”, in which some might primarily present a high Ki-67 proliferative rate but be morphologically well-differentiated . Sorbye et al. demonstrated the WHO 2010 “G3 p-NECs” were morphologically and biologically heterogenous, in which they reported a lower response rate after platinum-based systemic chemotherapy (15% vs. 42%, respectively; P < 0.05), but a longer MST (14mon vs. 10mon, respectively; P < 0.05) among tumors with a Ki‐67 < 55%, compared with those having a higher Ki‐67 index . Similar conclusions have also been reached that G3 p-NENs might consist of two distinct subgroups: well-differentiated p-NETs with a high proliferative rate (grade-discordant G2 p-NETs or morphologically G3 p-NETs) and true poorly-differentiated p-NECs (small-cell or large-cell G3 p-NECs) [23–25].
The previous work eventually formed the basis for the WHO grading classification published in 2017 (Table 1), which officially defined p-NENs into two broad categories (well-differentiated and poorly-differentiated) and four groups (NET G1/G2/G3 and NEC G3) in the light of both morphological differentiation and grading upon proliferation rate . However, this new system has not yet been validated. According to the comprehensive analysis of p-NENs in the present study, we revealed three major findings. First, the WHO 2017 grading classification could well distribute p-NENs into four significant groups with different clinical features and long-term survivals. Second, the new WHO system was superior to WHO 2010 criteria for better stratifying ability and more accurate predicting ability for the OS of p-NENs. Finally, patients with different WHO 2017 grading p-NENs could be well staged by the new AJCC 8th TNM staging manual.
According to the definitions of WHO 2017 and 2010 grading classification for p-NENs, their main difference was that the WHO 2010 “G3 p-NECs” group was now divided into WHO 2017 G3 p-NET and G3 p-NECs (Table 1). We have just reported in one study that comparisons of patient demographics and tumor characteristics of G3 p-NETs and G3 p-NECs weren’t significant (P > 0.05), although the tumor diameter of G3 p-NETs seemed be smaller than that of G3 p-NECs (4.5 cm vs. 5.6 cm, respectively; P = 0.059) . Hereby, in Table 2, comprehensive comparisons were made for related factors between well-differentiated neoplasms (i.e. G1/G2/G3 p-NETs) and poorly-differentiated ones (i.e. G3 p-NECs). We found that the patient age of G1/G2/G3 p-NETs was notably younger than that of G3 p-NECs (3.5 cm vs. 5.6 cm, respectively; P = 0.027) and the tumor diameter of G1/G2/G3 p-NETs was statistically smaller than that of G3 p-NECs (50yrs vs. 57yrs, respectively; P = 0.034). Meanwhile, compared with G1/G2/G3 p-NETs, G3 p-NECs present significantly more vascular infiltration (32.4% vs. 17.2%, respectively; P = 0.035), lymph involvement (46.3% vs. 29.3%, respectively; P = 0.019) and distant metastasis (29.6% vs. 16.7%, respectively; P = 0.041). Referring to the results above , statistical differences of these clinicopathological features might be caused by the integration of G1/G2/G3 neoplasms, forming the category of well-differentiated p-NENs, as McCall et al. have demonstrated in their study .
G1/G2/G3 p-NETs were usually slow-growing tumors with equal sex preference occurring over a broad age range, highest incidence peak between third and sixth decade, while G3 p-NECs had an incidence peak in the sixth to seventh decade, whose clinical presentation was very similar to pancreatic exocrine adenocarcinomas (p-EACs) . Our analysis indicated that patient gender among each new grading group had a slight female predominance with a peak median incidence age ranging from 45yrs to 57yrs and that p-NENs more frequently involved the body or tail of pancreas (Table 2). In terms of the survival of p-NENs, the WHO 2017 and 2010 grading classification both showed significantly decreased survivals as grade increased (Fig. 1, Fig. 2; respectively). Most importantly, the estimated 5-year OS of G3 p-NETs was statistically better than that of G3 p-NECs (35.1% vs. 11.1%, respectively; P < 0.001) but notably worse than that of G2 p-NETs (35.1% vs. 58.4%, respectively; P = 0.023) and G1 p-NETs (35.1% vs. 75.8%, respectively; P < 0.001; Fig. 1). This situation was in agreement with the reported results we mentioned above [24–26]. We then revealed that although the WHO 2017 and 2010 criteria could be independent predictor for the OS of p-NENs (P = 0.016, P = 0.022, respectively; Table 3), the 95% CIs of WHO 2017 grading classification (0.983–9.454) was slightly smaller than that of WHO 2010 criteria (0.201–13.374), indicating a relatively better predicting accuracy.
Another concern of our analysis was the TNM staging classification for p-NENs. In 2010, AJCC began to apply its TNM staging system to p-NENs , which derived from the staging algorithm for p-EACs and was proven to be convenient but a little oversimplified for p-NENs [29, 30]. In 2017, AJCC updated its staging manual for p-NENs (i.e. 8th edition), in which AJCC highlighted that the novel system for p-NENs should only be applied to G1/G2 p-NETs, while “G3 p-NECs” be staged by the revised one for p-EACs . The two new independent AJCC staging systems for p-NENs have been separately demonstrated to be superior to the AJCC 7th edition system in two previous studies [31, 32]. Recently, considering the heterogeneity with “G3 p-NECs”, we for the first time attempted to evaluate which new system G3 p-NETs should be better staged by . We concluded that the AJCC 8th staging systems introduced for G1/G2 p-NETs and “G3 p-NECs” were both practical for G3 p-NETs, while the one originally applied to G1/G2 p-NETs appeared to be superior in performance . Therefore, in the present study, we firstly staged G1/G2/G3 p-NETs together by one new AJCC system for G1/G2 p-NETs and staged G3 p-NECs by the other one for “G3 p-NECs” (Table 2). According to our analysis, both G1/G2/G3 p-NETs and G3 p-NECs could be well classified into four prognostic groups by their corresponding AJCC system, respectively, with statistically different stage distributions on their OS (P < 0.05; Fig. 3 and Fig. 4).
Our study had some limitations. First of all, it was also a retrospective study in which data analysis and patient’s recruitment were over a long duration. Secondly, the accumulative OS was estimated by K-M methods due to some cases with a short follow-up time. Then, our analysis derived from one single medical institution which might reduce the statistical power between factors and survival outcomes. Finally, all patients had surgically-resected disease and applicability to patients presenting with advanced disease needs to be validated. Therefore, a prospectively designed study from multi centers and with a long follow-up time is still needed to confirm our results.