Neuroendocrine tumors (NETs) of the gastroenteropancreatic (GEP) system comprise a rare but challenging group of malignant neoplasms and occur at virtually any site of the GEP system. In 2006, a new TNM classification system was proposed for the staging and grading of upper GEP NETs.
The prognostic relevance of the TNM classification system was analyzed retrospectively in 202 patients from a referral center with histologically proven foregut NET. Patients were classified according to previous classification systems and the TNM classification. Survival data were acquired and statistical analyses were performed by using log-rank and Cox regression testing.
Primary tumors were gastric (n = 48), duodenal (n = 23), and pancreatic (n = 131). During the observation period, 21% of patients died. The overall 5- and 10-year survival rates were 75% and 64%, respectively. Previous classification systems discriminated between low-grade and high-grade malignant NETs but did not allow further prognostic differentiation. In contrast, the proposed TNM classification was able to differentiate significantly between different tumor stages (stages I-III vs stage IV; P < .01) and cellular proliferation rates according to Ki-67 labeling (grade 1 vs grade 2, P = .04; grade 1 vs grade 3 and grade 2 vs grade 3, P < .01). Cox regression analysis confirmed an increased risk of reduced survival for patients with stage III or IV NET and grade 2 or 3 NET.
Neuroendocrine tumors (NETs) of the gastroenteropancreatic system (GEP) are rare neoplasms with an estimated incidence of 1 to 2 per 100,000.1–3 During the last 30 years,1 the incidence of GEP NET has increased.2 Both because of their long clinical course and because of their heterogeneity, GEP NETs provide a substantial challenge to all disciplines involved in the management of these patients. Among GEP NETs, previous classification systems have paid more attention to “carcinoid” tumors of the gastrointestinal system,1, 2 which did not include pancreatic NETs. Recently, however, pancreatic NETs have been reported with increased frequency and have received special attention4–10; however, it is unclear whether this reflects a true rise in incidence or an increased and better recognition of the entity explained in part by the use of more recent classification systems.11–15 The importance of classification in the clinical management of GEP NETs is underlined by recent studies reporting on the incidence, characteristics, and prognosis of this NET subgroup.6, 8, 9 Gastric NETs have been characterized well as either mostly multifocal tumors that occur in the context of atrophic corpus gastritis (Type I), or as part of multiple endocrine neoplasia type 1 (MEN-1) syndrome (Type II gastric NET), or as solitary sporadic NET (Type III)16–18 and have been observed with increasing incidence during recent decades.18, 19 Duodenal NETs are relatively rare (1%-4% of all GEP NETs)1, 8, 10, 20, 21 and also occur in the context of MEN-1 syndrome as both gastrinomas22–24 and somatostatinomas.25
Williams and Sandler first attempted a systematic classification of GEP NETs.26 They subdivided NETs (then called “carcinoid” tumors) according to the section of the embryonal primitive gut (ie, foregut, midgut, and hindgut) from which they stemmed. Although this classification is of little if any prognostic significance, it still is in use for the practical reason of anatomic characterization of primary tumor localization in patients with NET. In 1980, the World Health Organization (WHO) suggested a classification system in which carcinoid tumors (including NETs derived from gastrin-producing G-cells) were separated from pancreatic tumors and a few other endocrine tumors, such as Merkel cell carcinoma, paragangliomas, and others.13 This classification also was unsatisfactory with respect to both adequate histologic classification and prognostically relevant clinicopathologic categorization. Therefore, Capella et al.11 attempted a new clinicopathologic classification system, which considered macroscopic features (eg, size and metastasis), histopathologic features (eg, cellular differentiation, neuroinvasion, angioinvasion, and lymphangioinvasion, proliferative capacity), and clinical features (eg, the presence of hormone hypersecretion syndromes). This classification generally separated benign neuroendocrine tumors from those with benign or uncertain behavior, low-grade malignant neuroendocrine carcinomas, and high-grade malignant neuroendocrine carcinomas. In 2000, the WHO published a new classification system for the histologic typing of endocrine tumors that was based on the work of Capella et al.12 This system identified well differentiated endocrine tumors (WDET), well differentiated endocrine carcinomas (WDEC), and poorly differentiated endocrine carcinomas (PDEC). Along with the aforementioned features, the latter 2 systems also considered the localization of the primary tumor as an additional classification criterion. However, a widely accepted, standardized classification system for GEP NETs that can serve as a basis for clinical management and appropriate design of clinical trials was not available until 2006, when Rindi et al. presented a proposal for a TNM classification system. This was analogous to the TNM classification systems used for other solid tumors and was the result of a consensus conference of international experts held by the European Neuroendocrine Tumor Society.27 The proposed system provides a staging classification (Table 1) and also suggests a grading system (Table 2), which characterizes the proliferative potential of the neuroendocrine tumor cells using either the mitotic count or the Ki-67 labeling index.
Table 1. Proposed TNM Staging for Foregut Neuroendocrine Tumors of the Stomach, Duodenum, and Pancreas by Rindi et al.*
T0 indicates no evidence of primary tumor; Tis, tumor in situ/dysplasia (size <5 mm); T1, gastric or duodenal tumor invading the lamina propria or submucosa and size <10 mm or pancreatic tumor limited to the pancreas and size <20 mm; T2, gastric or duodenal tumor invading the muscularis propria or subserosa, or size >10 mm, or pancreatic tumor limited to pancreas and size between 20 mm and 40 mm; T3, gastric or duodenal tumor penetrating the serosa, or duodenal tumor infiltrating the pancreas, or pancreatic tumor limited to pancreas and size >40 mm, or pancreatic tumor invading the duodenum or the common bile duct; T4, gastric, duodenal, or pancreatic tumor invading adjacent structures.
N0 indicates absence of regional lymph node metastasis; N1, invasion of regional lymph nodes.
M0 indicates absence of distant metastasis; M1, presence of distant metastasis.
Table 2. Proposed Grading System for Foregut Neuroendocrine Tumors of the Stomach, Duodenum, and Pancreas by Rindi et al.*
In the current study, we present what to our knowledge is the first validation of the new TNM classification system27 compared with the 2 previous clinicopathologic classification systems proposed by Capella et al.11 and the WHO in 2000.12 Detailed survival analyses were performed, and the prognostic relevance of the new classification system was determined to investigate its value for risk stratification of GEP NET.
MATERIALS AND METHODS
The medical records of 202 patients with histopathologically proven NET of the upper GEP tract who were treated at our Department of Hepatology and Gastroenterology between 1980 and 2003 were analyzed retrospectively. Histopathologic confirmation of the diagnosis of an upper GEP or foregut NET (gastric, duodenal, or pancreatic) was required for inclusion in the study. The patient files were reviewed systematically for the date of initial diagnosis, localization of the primary tumor, histopathologic diagnosis, tissue site from which samples were taken for histopathologic diagnosis, clinical staging at initial diagnosis (ie, results of imaging studies and surgical procedures), and the presence and quality of a clinically manifest hormone hypersecretion, ie, a functional syndrome. All patient files were reviewed, and disease- and tumor-specific information was recorded (HJ). The data were re-evaluated for correctness and consistency (U-FP). Survival data were obtained by chart review.
Special attention was paid to the histopathologic data acquired from the diagnosing institution. The histopathologic diagnosis was made at our center in 114 of 202 patients (56%) and at external institutions in the remaining patients. When it was determined externally, the histopathologic diagnosis was reviewed, revised, and/or completed by a staff pathologist from our own institution (MK or CR) when necessary using hematoxylin and eosin staining of formalin-fixed, paraffin-embedded specimens. Neuroendocrine differentiation was confirmed immunohistochemically using antibodies directed against chromogranin A and synaptophysin. In particular, histologic differentiation grade, immunohistochemistry, and the Ki-67 labeling index were amended when sufficient tumor tissue was available. On the basis of the available information, the clinicopathologic classification proposed by Capella et al.,11 the WHO classification12 for histologic typing of endocrine tumors, and the recently published TNM classification system suggested by Rindi et al.27 were applied whenever possible.
Statistical analyses were performed using SPSS software (version 13.0; SPSS, Chicago, Ill) and SAS statistical software (version 8.02; SAS Institute, Cary, NC). When data distribution was approximately normal, the results are reported as the mean ± standard error. Otherwise, median values and full ranges are reported. Overall survival analyses were performed using the Kaplan-Meier method. The log-rank test was used for univariate analyses of potentially prognostic factors. The relative risk of overall death for each of the new staging or grading groups compared with the lowest risk group was analyzed using Cox proportional-hazards modeling. Cox regression analyses were conducted for both TNM staging and tumor grading before (ie, not adjusted) and after adjustment for age and sex (ie, adjusted). A P value <.05 was considered statistically significant; all tests were 2-sided.
In total, 202 patients with foregut NETs who were treated at our institution between January 1980 and May 2007 were analyzed retrospectively. The mean follow-up was 47 ± 3 months (median, 28 months; range, 1–19 months). The mean patient age at initial diagnosis was 55 ± 1 year (median, 58 years; range, 7–79 years). The ratio of men to women for the entire study cohort was 1:0.9 (106 men and 96 women). No patients with hereditary foregut NET (eg, MEN-1 syndrome) were included. The distribution according to primary tumor localization, TNM stage, and tumor grade, together with the ratio of men to women, for each subgroup is provided in Table 3. A functional syndrome caused by hormone hypersecretion at initial diagnosis was present in 46 patients (23%) (Table 4).
Table 3. Localization of Primary Tumors and Classification of Patients With Foregut Neuroendocrine Tumors According to the TNM Staging and Grading Criteria Proposed by Rindi et al., Including Sex Distribution*
Table 4. Characteristics of Clinically Symptomatic Hormone Hypersecretion Syndromes (Functionality) in the Studied Patients With Foregut Neuroendocrine Tumors
No. of patients
Primary tumor localization
Sufficient information for adequate categorization was available for classification according to Capella et al.11 in 171 of 202 patients (85%), the 2000 WHO classification12 in 123 of 202 patients (61%), and the TNM classification according to Rindi et al.27 in 193 of 202 patients (96%). Grading according to immunohistochemical Ki-67 labeling was available in 158 of 202 patients (78%). TNM stage was determined according to pathologic (pTNM) and clinical (cTNM) classifications. However, only cTNM staging was validated because, complete pTNM staging was available only for an insufficient number of patients (100 of 202; 50%).
In total, 43 of 202 patients (21%) died during the median observation period of 28 months (range, 1–219 months). Of all deaths, 70% (n = 30) were caused by NET. The mean overall survival for the whole cohort was 145 ± 9.8 months. The 2-year, 5-year, and 10-year survival rates for overall survival were 87%, 75%, and 64%, respectively. When only NET-related deaths were considered as events for survival analysis, the mean survival was 167 ± 9 months with 2-year, 5-year, and 10-year overall survival rates of 89%, 81%, and 75%, respectively.
Survival Analysis of NET Classified According to Capella et al.
The survival analyses of 171 patients who had NETs classified according to Capella et al.11 indicated a significantly poorer survival for patients who had high-grade malignant neuroendocrine carcinoma (NEC) compared with those who had either benign NET (chi-square test, 11.569; P = .001), NET with benign or uncertain behavior (chi-square test, 11.151; P = .001), or low-grade malignant NEC (chi-square test, 29.812; P < .001) (Table 5) (Fig. 1a). Although the trend is obvious in the graph shown in Figure 1, the survival difference did not reach significance between benign NET and low-grade malignant NEC (chi-square test, 2.444; P = .118) or between NET with benign or uncertain behavior and low-grade malignant NEC (chi-square test, 1.957; P = .162); there was no survival difference at all between benign NET and benign or low-grade malignant NET.
Table 5. Mean Survival and 2-Year, 5-Year, and 10-Year Survival Rates for Patients With Foregut Neuroendocrine Tumors According to Different Classification Systems
No. of deaths/total no. (%)
Mean ± SD survival [95% CI], mo
Survival rate, %
SD indicates standard deviation; 95% CI, 95% confidence interval; ND, not determined; WHO, World Health Organization; WDET, well differentiated endocrine tumor; WDEC, well differentiated endocrine carcinoma; PDEC, poorly differentiated endocrine carcinoma.
Survival Analysis of NET Classified According to the WHO Classification of 2000
The survival analyses of 123 patients who had NETs classified according to the WHO classification12 indicated a significantly poorer survival for patients who had PDEC compared with those who had WDET (chi-square test, 21.793; P < .001) and WDEC (chi-square test, 22.886; P < .001) (Table 5) (Fig. 1b). When WDET and WDEC were compared, there was a trend toward statistical significance, although it was not completely reached (chi-square test, 3.374; P = .066).
Survival Analysis of NET Classified According to the New TNM Staging System
The survival analysis of 193 patients who had NETs classified according to the new TNM staging system27 indicated a significantly poorer survival for patients who had stage IV tumors compared with those who had stage I, II, and III tumors (stage I: chi-square test, 12.152; P < .001; stage II: chi-square test, 14.037; P < .001; stage III: chi-square test, 8.497; P = .004) and for patients who had stage I tumors compared with those who had stage III tumors (chi-square test, 3.913; P = .048) (Table 5) (Fig. 2a). There was no statistically significant difference between stage I and stage II tumors (chi-square test, 1.459; P = .227) or between stage II and stage III tumors (chi-square test, 1.871; P = .171).
One hundred sixty-seven of the 171 patients who had NETs classified according to Capella et al.11 were analyzed in a subgroup analysis for direct comparability between classification systems. Again, significantly poorer survival was observed for patients who had stage IV tumors compared with patients who had tumors in all other stages (stage I: chi-square test, 9.121; P = .003; stage II: chi-square test, 13.287; P < .001; stage III: chi-square test, 9.984; P = .002). The smaller number of patients available for analysis in this subgroup, however, lead to nonsignificant differences between stage I and stage II (chi-square test, 0.733; P = .392), between stage I and stage III (chi-square test, 1.966; P = .161), and between stage II and stage III (chi-square test, 1.706; P = .191).
One hundred nineteen of the 123 patients who had NETs classified according to the 2000 WHO classification12 were available for analysis according to the new TNM staging system, which still indicated significantly poorer survival for patients who had stage IV tumors compared with patients who had stage I, II, and III tumors (stage I: chi-square test, 8.169; P = .004; stage II: chi-square test, 7.824; P = .005; stage III: chi-square test, 7.715; P = .005). However, also because of the smaller number of patients, there no longer was a statistical difference between stage I and stage II (chi-square test, 0.889; P = .346), between stage I and stage III (chi-square test, 1.023; P = .312), or between stage II and stage III (chi-square test, 0.115; P = .735).
Survival Analysis of NETs Classified According to the New Grading System
In total, 158 patients were available for survival analysis according to the new grading system (Table 5) (Fig. 2b).27 This analysis indicated that survival was significantly poorer for patients who had grade 3 tumors compared with patients who had grade 1 and grade 2 tumors (grade 1: chi-square test, 35.641; P < .001; grade 2: chi-square test, 31.165; P < .001) and for patients who had grade 2 tumors compared with patients who had grade 1 tumors (chi-square test, 4.217; P = .040).
Cox Regression Analysis for TNM Staging and Tumor Grading
Cox regression analysis was performed and did not include stage I tumors for TNM staging, because the absence of any events excluded this category from calculation within the model. The relative risk of death increased to approximately 4-fold and 30-fold for grade 2 NET and grade 3 NET, respectively, compared with grade 1 NET, whereas the relative risk of death for stage III NET and stage IV NET increased to 3-fold and 9-fold, respectively, compared with stage II NET (Table 6).
Table 6. Relative Risk of Death for Patients With Foregut Neuroendocrine Tumors
NETs of the GEP system comprise a heterogeneous group of solid tumors with a large spectrum of clinical pictures and differing clinical courses.1–3, 7, 10, 28 This complex situation has lead to the formulation of several guidelines for the diagnosis and treatment of GEP NET by various national and international expert panels.29–40 The basis for treatment of patients with NET is a proper, ultimately histopathologic diagnosis. Modern diagnosis and evidence-based treatment of GEP NET requires a commonly accepted and easily applicable classification system, which to our knowledge has not been achieved to date by any of the older classifications.9, 11–13, 26, 27, 33 The initial attempt by Williams and Sandler26 has survived in the sense that primary tumor localization within the gastrointestinal tract still plays a role in defining the involved anatomic structure and provides some hints regarding potential problems, such as obstructive jaundice in NETs of the pancreatic head. However, classification according to these criteria alone has not lead to significant prognostic risk stratification. The WHO classification of 198013 was unable to overcome these difficulties, and only Capella et al. first described a classification system that considered various clinicopathologic criteria.11, 13 This classification system built the basis of the most recent WHO classification,12 which was published in 2000 and has proven its relevance in several studies.8–10, 41–43 However, neither of these systems allows for the direct recognition of details of the diagnosis, and they are not accepted widely for prognostic risk stratification. Finally, the common practice of stratifying patients with gastric, duodenal, and pancreatic NETs according to the relevant TNM adenocarcinoma counterpart is a conceptually faulty and eventually ineffective approach to this disease. The newly proposed TNM classification system was developed by an international expert panel not only to improve the histologic classification of NETs but also to define exactly and comparably the tumor-specific variables that are important for the prognosis and development of treatment plans.9, 10, 17, 27, 41, 44, 45
Such a standardized and easily applicable classification system is available for most solid tumors and largely has contributed to improvement of treatment strategies by providing a common basis for risk stratification and clinical trials.46 However, even in tumor entities with a longstanding TNM classification system, such as breast cancer47 and colorectal cancer,48 similar, more elaborate approaches for risk stratification are under discussion. However, to the best of our knowledge, the new classification has not been validated yet for its prognostic significance.27
In the current study, we were able to demonstrate that the classification of GEP NETs according to the newly proposed TNM classification system improved the assessment of survival in patients with GEP NETs, in that it allows more subtle and clinically relevant risk stratification. The new TNM staging system is able to prognostically significantly differentiate between stage IV tumors (ie, tumors with distant lymph node or organ metastasis) and all other tumor stages. This is in accordance with many data from the literature that indicate a significant influence of distant metastasis on poorer outcome in patients with GEP NETs.1, 4–10, 17, 18, 21, 22, 41, 49–51 However, definition of distant metastasis (equivalent to M category) or even metastasis only (equivalent to M and/or N category) is incongruent in most studies; therefore, the results are only roughly comparable. This situation becomes even more confusing when the characteristics of patient cohorts studied in clinical trials are considered.52–57 Furthermore, our analysis revealed that the new TNM classification also statistically significantly separated stage III foregut NETs from stage I tumors by log-rank testing and indicated an increased relative risk of death for each increasing tumor stage in Cox regression analysis; statistically significant full separation between stage I and II from stage III could not be reached because of the small sample size, although trends were obvious (Tables 5 and 6) (Fig. 2a). This means that patients who have small and limited tumors have a significantly better prognosis than patients who have tumors with locally invasive growth (stage IIIA) or with lymphangioinvasion and spread to regional lymph nodes (stage IIIB), which also is reflected by the survival rates. Compared with reports from the literature, this has not been demonstrated unequivocally before,1, 4–6, 8–10, 15, 16, 19, 20, 37, 45 but studies are only poorly comparable.10 In particular, patients with locally limited gastrinomas and insulinomas have a good prognosis, and pure locoregional lymph node metastasis is not an indicator of poor prognosis.19, 20, 45 However, these entities most likely were underrepresented in our cohort, because a subgroup analysis of gastrinomas from our cohort (Table 4) indicated a better prognosis for these patients compared with patients in the cohort who had other pancreatic NETs (data not shown). Therefore, the TNM staging system needs further validation in the specific tumor subgroups with sufficient numbers to clarify the prognostic relevance of categorization with this system.
Our analysis also demonstrated that the new TNM staging system offers an easily applicable tool for the description of tumor load in GEP NETs. The system will enable the assessment of patient prognosis and, thus, most likely will influence patient management and future research. Particularly cTNM staging is easily possible in almost all patients with NET once initial staging has been performed, including the acquisition of a representative histology. This likely will allow relevant prognostic stratification of NET patients even by less experienced institutions with information that usually is available for all patients. This is supported by the finding, that in almost all of the patients included in our study, cTNM staging was possible, whereas the information required for other classification systems was available only in fewer patients, thereby weakening the value of those systems. However, complete and proper pTNM staging most likely will not be possible in all patients (approximately 50% in our cohort); because, in patients with more advanced and incompletely resectable tumor stages (particularly stage III tumors), not all clinical tumor manifestations may be available for pathologic examination and classification. A similar situation arises in patients with locally resected tumors (eg, by endoscopy), in whom the locoregional lymph nodes may not be available for histopathologic staging and, thus, proper pTNM staging may not be possible. cTNM staging according to the new criteria, however, proved to be a promising tool in our experience. Furthermore, TNM criteria are well established in oncologic practice of solid tumors; therefore, their use may be easier than using the more complex and specialized criteria of the earlier classification systems.39, 42–44
One objective of the NET classification systems by Capella et al.11 and the WHO of 200012 has been a more specific characterization of tumor cell biology of NETs.11–13, 27, 58 One of the most important details, which is easily detectable in routine pathology, is the proliferative capacity of the tumor reflected either by the Ki-67 labeling index or by the mitotic count.9, 11, 12, 27, 41–43 The introduction of an NET-specific grading system relying on one of these relatively easily reproducible histopathologic techniques27 is an attempt to further specify a tumor-specific modality with important influence on survival that is independent from staging categories, as demonstrated in the current study. Here, we demonstrated that the Ki-67 labeling index is correlated inversely with patient prognosis. This makes the Ki-67 labeling index an important independent prognostic biomarker for patients with foregut NETs (Tables 5 and 6), as suggested previously by other studies.9, 10, 41–44
Our study also has a few limitations. Retrospective analyses always carry the risk of selection biases. However, they are extremely helpful in the assessment of rare diseases, particularly in rare diseases with relatively long survival, because of the limited data available. The selection bias most relevant to our study cohort is the referral bias to our tertiary gastroenterologic-endocrinologic-oncologic center. This includes more advanced and sometimes unsuccessfully treated patients with a poorer prognosis.7, 10 However, we included all patients who were seen at our center with foregut NETs of the stomach, duodenum, and pancreas for which a histopathologic diagnosis was available. Furthermore, a Ki-67 index was not available for all patients. It is unlikely that this represents a systematic selection bias; rather, it may represent a random limitation of histopathologic diagnoses made in external institutions. In most of these patients, differentiation was characterized by tumor cell histology. Tumor tissue for additional Ki-67 labeling was unavailable because of limitations of the biopsy specimen or because of a lack of availability from the institution, making supplementation of the Ki-67 index by histopathologic staining impossible. This also highlights the finding that different pathologists made the histopathologic diagnoses of GEP NET, including the aforementioned details. In all tumors that had an unclear histopathologic diagnosis, a review and re-evaluation was performed by expert pathologists from our institution, as described above (see Materials and Methods). However, this situation most likely reflects the true situation in the routine management of GEP NETs for which assessment by expert pathologic expert frequently is unavailable. Despite these limitations, the new TNM classification proved its significance and applicability in routine patient management.
In summary, the current study demonstrated that the newly proposed TNM classification system for staging and grading of foregut NETs27 is valid for the prognostically significant risk stratification of patient subgroups with foregut NETs in all categories and produces at least equal or clearer results compared with previous classifications systems. The proposed TNM system has the advantage of familiarity to pathologists and clinical disciplines involved in the management of these patients, because it uses the general terminology of the TNM classification as already published by the International Union Against Cancer and in use for many other solid tumors. This should make it a classification system that deserves wider acceptance and likely will lead to a more stringent classification and stratification of GEP NETs. This is urgently needed for the development of future evidence-based treatment strategies of these complex neoplasms.