Prognostic Validity of the American Joint Committee on Cancer Eighth Edition Staging System for Well-Differentiated Pancreatic Neuroendocrine Tumors: A Retrospective Multinational Multicenter Study


 Background:There is no widely-accepted staging system for pancreatic neuroendocrine tumors (pNETs). The aim of this study was to validate the American Joint Committee on Cancer (AJCC) 8th edition staging system for well-differentiated (G1/G2) pNETs.Methods:A multicenter dataset (n=1086) was used to evaluate the application of the AJCC 7th and 8th, the European Neuroendocrine Tumor Society (ENETS), and the modified ENETS (mENETS) staging systems for well-differentiated pNETs.Results:The proportion of patients with stage III tumors was extremely low (1.1%) according to the AJCC 7th staging system. For the ENETS staging system, patients with stage IIIA disease had worse estimated mean survival than patients with stage IIIB disease (78.9 vs. 107.3 months). When comparing with patients in stage I, the AJCC 7th, ENETS, and mENETS staging systems showed good performance in discriminating between stages; however, there was no significant difference in some stages when the reference was defined as the earlier stage. When the reference was defined as stage I or the earlier stage, there was a significant inter-stage difference in the AJCC 8th staging system.Conclusions:The AJCC 8th staging system is more suitable for pNETs than other TNM staging systems and may be adopted in clinical practice.


Introduction
Pancreatic neuroendocrine tumors (pNETs), which account for less than 2% of all pancreatic tumors, are a group of heterogeneous neoplasms with varying pathologic, functional, and clinical features [1,2]. With an estimated annual worldwide incidence of 2.5 to 5 per 1,000,000 individuals, pNETs have been increasingly diagnosed over the past several decades, likely due to more accurate classi cation and improvements in diagnostic tools [3,4]. Because of their rarity and heterogeneous behavior, it is di cult to stratify pNETs patients into prognostic groups using commonly accepted staging systems [5]. Thus, a robust and accurate staging system for risk strati cation and survival prediction is urgently required.
In 2006, the European Neuroendocrine Tumor Society (ENETS) was the rst to publish a consensus statement on the TNM staging classi cation of pNETs, as applied in other solid tumors [6]. Subsequently, the American Joint Committee on Cancer (AJCC) also introduced a TNM staging system for pNETs in 2010 (the AJCC 7th edition), which initially was applied to exocrine pancreatic adenocarcinoma based on studies by Bilimoria et al [5,7]. Considering the drawbacks of these systems, Luo et al. developed a modi ed ENETS (mENETS) staging system, which maintained the ENETS T, N, and M de nitions and adopted the AJCC system staging de nitions [8]. Although several classi cation schemes have been proposed, there is no commonly accepted staging system ( Table 1). The ENETS staging system pointed out that patients with stage I disease had a similar prognosis to those with stage II disease [9], and the hazard ratio (HR) of death for patients with stage IIIB disease was even lower than that for patients with stage IIIA disease [10]. According to the AJCC 7th staging system, few patients would have stage III disease because pNETs seldom invade the celiac axis or the superior mesenteric artery (unresectable tumor) without distant metastasis, and patients with stage II or III disease were prognostically indistinguishable [11].
In 2017, the AJCC formally introduced a newly revised 8th staging system especially for pNETs, which was somewhat similar to the ENETS criteria, that included several signi cant changes [12]. First, the AJCC 8th staging system was only applied to welldifferentiated pNETs, whereas poorly differentiated neuroendocrine carcinoma G3 (NEC G3) were excluded and considered according to the staging criteria for pancreatic exocrine adenocarcinoma. Second, it rede ned the T, N, M, and staging system; for example, stage II and stage III was de ned by combining the 2 previous subgroups of the ENETS system (stage IIA and stage IIB, and stage IIIA and stage IIIB, respectively) ( Table 1).
While the new AJCC 8th staging system for well-differentiated pNETs has been validated by some studies [13,14], validation using a large multi-center cohort has not been performed. Therefore, we utilized a multi-center database to validate the AJCC 8th staging system in the present study. To accomplish this, survival curves were compared according to the AJCC 7th and 8th, the ENETS, and the mENETS staging systems.

Patients and Data Collection
The study involved a cohort that included 1086 patients from a multi-center database comprising ve Chinese centers and one The main patient variables, including sex, age, tumor location, functional status, operation, type of operation, grade, T category, N category, M category, and overall survival, were all systematically collected and retrieved. We only included patients who had a pathologically con rmed diagnosis of pNET and had complete data to allow restaging per the AJCC and ENETS classi cations, including T stage, nodal status, distant metastases, and follow-up data. Patients with NEC G3 tumors were excluded. Electronic datasheets were provided to all participating centers and all de-identi ed data were reviewed and cross-checked for inconsistencies. The study was approved by the Institutional Review Board of all participating centers.

Statistical analysis
Survival time was calculated as the number of months from the date of initial diagnosis until the date of last contact or death. Overall survival was evaluated using Kaplan-Meier curves, and log-rank tests were used to assess staging classi cation. Cox proportional hazards regression was performed for univariate analysis, and the HR and 95% con dence intervals (95% CI) were calculated. Multivariate analysis of each staging classi cation, controlling for sex, age, tumor location, functional status, type of operation, and grade, was performed using Cox proportional hazards regression. Values are expressed as number and percentage. All tests were 2sided and statistical signi cance was de ned as p < 0.05. Statistical analyses were performed using SPSS v22.0 (IBM, Armonk, NY).

Patient characteristics
In total, the study included 1086 patients from the multicenter database. Table 2 show their baseline characteristics. The median age was 54 years (mean, 53.27 years) and the proportions of male patients was 50.4%. Most patients had nonfunctioning pNETs (90.1%) and more than half (56.3%) had tumors located at the body and/or tail of the pancreas. Distal pancreatectomy and pancreaticoduodenectomy were performed in 531 (48.9%) and 398 (36.6%) patients, respectively, and laparoscopic surgery was conducted in 214 (19.7%). The median survival was 136.8 months (3-year survival rate, 90.6%; 5-year survival rate, 83.9%; 10-year survival rate, 57.6%).

TNM staging classi cation and survival
As shown in Fig. 1 (A and B) and  1E and Table 2).

ENETS stage group and survival
In comparison with patients with stage I disease, a signi cant difference was observed across all stage groups. However, patients with stage IIIA disease had worse estimated mean survival than patients with IIIB (78.9 vs. 107.3 months). Furthermore, there were also no signi cant differences across stages IIIA, IIIB, and IV when the reference was de ned as the earlier stage (P =0.978, P =0.411, and P =0.077, respectively) ( Fig. 1F and Table 2).

mENETS stage group and survival
The mENETS staging classi cation showed better distinction between different stages than the AJCC 7th and ENETS systems. There was a signi cant difference in survival between all stage groups compared to patients with stage I disease, and the relative risk for death was correlated with advanced stage disease. Nevertheless, there were also no signi cant differences in stages IIB, III, and IV when the reference was de ned as the earlier stage (P =0.158, P =0.691, and P =0.139, respectively) ( Fig. 1G and Table 2.)

AJCC 8th stage group and survival
Because stages II and III were both previously divided into 2 previous subgroups in the ENETS system (stage IIA and stage IIB, stage IIIA and stage IIIB, respectively), the AJCC 8th staging system divided all patients into four stages. Compared with stage I disease, there was a signi cant difference between all stages, which was also observed when the reference was de ned as the earlier stage ( Fig. 1H and Table 2).

Discussion
Although the ENETS system is widely used in Europe, and the AJCC system in the United States [15], there is still no widely accepted staging system for pNETs as they are heterogeneous neoplasms with different clinical features, biological behaviors, and prognoses [5]. The current study used a large, multicenter database to validate the newly developed AJCC 8th staging system for welldifferentiated pNETs.
The TNM staging system is widely used for solid tumor staging guidelines, including pNETs [16], with the main guidelines currently available being the AJCC 7th and 8th systems, the ENETS system, and the mENETS system. Each of these exhibits both similarities and differences.
Both the AJCC 7th and ENETS systems have been widely used in clinical practice for a long time. The AJCC 7th staging system adopted the staging system of exocrine pancreatic carcinomas for pNETs, T4 was de ned as patients with involvement of the celiac axis or superior mesenteric artery (unresectable tumor), which meant that a very low proportion of patients were considered stage III (T4, any N, M0) [17]. Consistent with the literature, our multicenter cohort had a low proportion of stage III disease (1.1%). In addition, the presence of extra-pancreatic spread was di cult to assess pathologically due to the expansive growth pattern common to pNETs [13]. Furthermore, some studies found that the AJCC 7th staging system has a poor ability to differentiate between the prognoses of some subgroups. Rindi et al [18] identi ed 1072 pNETs patients from eight European cancer centers (1990-2007) and found no signi cant difference in mortality between stages IIA and IIB (death rate per 100 persons per year = 3.4 vs 3.7, respectively; HR of death = 25.2, 95% CI = 5.9 to 106.9, P = 0.84), and stages IIB and III (death rate per 100 persons per year = 3.4 and 3.4, respectively; HR of death = 25.1, 95% CI = 5.4 to 116.5, P = 1.0). Our study also exhibited similar results, where there was no signi cant difference across stages IIA, IIB, III, and IV when the reference was de ned as the earlier stage (P = 0.052, p = 0.128, P = 0.249, and P = 0.138, respectively).
The ENETS system's ability to predict the prognosis of speci c groups of patients was considered to be possibly superior to the AJCC 7th system. However, some studies have suggested that ENETS system cannot appropriately discriminate the prognoses of stage I and stage II patients [9], and predicted some abnormal survival outcomes between stages IIIA and IIIB [10]. Luo et al. [8] investigated the SEER database (N = 2529 patients) and a multicenter database (N = 1143 patients) and con rmed that patients with stage I disease had a similar prognosis to those with stage IIA disease (with stage I as the reference: SEER series, HR = 0.99, P = 0 .955; multicenter series, HR = 1.41, P = 0.337) using the ENETS staging system. In addition, the HR of death for patients with stage IIIA disease was even higher than that for patients with stage IIIB disease (with stage I as the reference: SEER series, HR of death = 2.87 vs. 2.77, respectively; multicentric series, HR of death = 4.56 vs. 4.25, respectively). In our study, there were no signi cant differences in stages IIIA, IIIB, and IV when the reference was de ned as the previous stage (P = 0.978, P = 0.411, and P = 0.077, respectively).
Maintaining the ENETS's T, N, and M de nitions and adopting the AJCC system's staging de nitions, Luo et al. [4] developed the mENETS, which solved the problem of the low proportion of patients of stage III in the AJCC 7th staging system and the poor differentiation between stages I stage II in the ENETS staging system. However, it became clear that the system showed poor ability to distinguish between the prognoses of stages IA and IB. Meanwhile, the prognostic difference between IIB stage III stage was slightly worse than other subgroups. In the present study, there was also no signi cant difference across stages IIB, III, and IV when the reference was de ned as the previous stage (P = 0.158, P = 0.691 and P = 0.139, respectively). Thus, the use of complex staging methods may not have much added value when considering diagnoses in the clinical setting.
In 2017, the AJCC 8th staging system, which followed the ENETS's de nition of T and simpli ed the original staging criteria, was introduced by the AJCC speci cally for well-differentiated pNETs. The AJCC 8th staging system was no longer suitable for poorly differentiated neuroendocrine carcinoma (G3), which have a signi cantly worse prognosis, and divided all patients into four stages. In fact, some studies have con rmed that there was no signi cant difference in survival between patients with stage IIIA or IIIB disease according to the ENETS system [13]. Thus, simplifying and merging some subgroups may be acceptable and may be more practical for determining the clinical prognosis of patients with pNETs. In this study, we con rmed that the AJCC 8th staging system succeeded in classifying patients into 4 signi cantly different staging groups using our multicenter database, and the relative risk for death was closely correlated with advanced stage disease.
Eschewing the TNM staging system, the World Health Organization published their histologic grade (G) classi cation based on Ki-67 expression and mitotic counts. The classi cation was principally to be used for prognostic strati cation and determining the adjuvant chemotherapy strategy according to the accurate measurement of proliferation [14]. In the present study, patients with G1 disease exhibited a better estimated 5-year survival rate than patients with G2 (87.1% vs. 81.6%), and histologic grade was found to be associated with overall survival in the univariate analysis (G2: HR = 1.534, 95% CI = 1.126 to 2.091, P = 0.007). The functional status of pNETs also has a signi cant effect on the prognosis. Studies have con rmed that functional pNETs have better prognoses than nonfunctional pNETs [19]. Consistent with the literature, the estimated 5-year survival rate of patients with functional and non-functional pNETs in our study were 95.0% and 82.5%, respectively. Functional status was also identi ed as signi cant positive prognostic factor in the univariate analysis (HR = 0.272, 95% CI = 0.128 to 0.580, P = 0.001).
Although the AJCC 8th staging system is superior to other TNM staging systems and may bring a uni ed consensus on TNM staging, it failed to include the histologic grade and functional status, which ultimately lead to its limitations in prognosis prediction. Therefore, a new stage classi cation combining the TNM staging system and G grade is needed to help guide therapeutic decisions. Furthermore, novel surveillance guidelines need to be developed and trialed.
The major limitation of this study was the retrospective nature of the data analysis. Furthermore, the lack of postoperative course data and postoperative chemotherapy information, among other variables, meant that their signi cance could not be assessed out.
Furthermore, since some patients underwent enucleation, the postoperative lymph node assessment may not be accurate.

Conclusions
The AJCC 8th staging system is more suitable for pNETs compared to the AJCC 7th, the ENETS, and the mENETS staging systems, and may be adopted in clinical practice. Using this novel system, a consensus on the TNM staging classi cation of pNETs may soon be reached.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.