pathology reporting of neuroendocrine tumors

Pathology Reporting of Neuroendocrine Tumors:Application of the Delphic Consensus Process to the

Development of a Minimum Pathology Data Set

David S. Klimstra, MD,* Irvin R. Modlin, MD, PhD,w N. Volkan Adsay, MD,zRunjan Chetty, MD,y Vikram Deshpande, MD,J Mithat Gonen, PhD,z Robert T. Jensen, MD,#

Mark Kidd, PhD,w Matthew H. Kulke, MD, ** Ricardo V. Lloyd, MD, PhD, w wCesar Moran, MD,zz Steven F. Moss, MD,yy Kjell Oberg, MD,JJ Dermot O’Toole, MD,zz

Guido Rindi, MD,## Marie E. Robert, MD,*** Saul Suster, MD,w w w Laura H. Tang, MD, PhD,*Chin-Yuan Tzen, MD, PhD,zzz Mary Kay Washington, MD,yyy Betram Wiedenmann, MD,JJJ

and James Yao, MDzzz

Abstract: Epithelial neuroendocrine tumors (NETs) have been

the subject of much debate regarding their optimal classification.

Although multiple systems of nomenclature, grading, and

staging have been proposed, none has achieved universal accep-

tance. To help define the underlying common features of these

classification systems and to identify the minimal pathology data

that should be reported to ensure consistent clinical manage-

ment and reproducibility of data from therapeutic trials, a

multidisciplinary team of physicians interested in NETs was

assembled. At a group meeting, the participants discussed

a series of ‘‘yes’’ or ‘‘no’’ questions related to the pathology of

NETs and the minimal data to be included in the reports. After

discussion, anonymous votes were taken, using the Delphic

principle that 80% agreement on a vote of either yes or no would

define a consensus. Questions that failed to achieve a consensus

were rephrased once or twice and discussed, and additional votes

were taken. Of 108 questions, 91 were answerable either yes or no

by more than 80% of the participants. There was agreement

about the importance of proliferation rate for tumor grading, the

landmarks to use for staging, the prognostic factors assessable

by routine histology that should be reported, the potential for

tumors to progress biologically with metastasis, and the current

status of advanced immunohistochemical and molecular testing

for treatment-related biomarkers. The lack of utility of a variety

of immunohistochemical stains and pathologic findings was

also agreed upon. A consensus could not be reached for the

remaining 17 questions, which included both minor points

related to extent of disease assessment and some major areas

such as terminology, routine immunohistochemical staining for

general neuroendocrine markers, use of Ki67 staining to assess

proliferation, and the relationship of tumor grade to degree of

differentiation. On the basis of the results of the Delphic voting,

a minimum pathology data set was developed. Although there

remains disagreement among experts about the specific classifi-

cation system that should be used, there is agreement about the

fundamental pathology data that should be reported. Examina-

tion of the areas of disagreement reveals significant opportunities

for collaborative study to resolve unanswered questions.

Key Words: neuroendocrine tumor, consensus, carcinoid tumor,

Delphic, classification

(Am J Surg Pathol 2010;34:300–313)

Epithelial neuroendocrine tumors (NETs) are a diversegroup of pathologically related neoplasms that can

arise in most epithelial organs of the body but areCopyright r 2010 by Lippincott Williams & Wilkins

From the From the Departments of *Pathology; zEpidemiology andBiostatistics, Memorial Sloan-Kettering Cancer Center, New York,NY; Departments of wSurgery; ***Pathology, Yale UniversitySchool of Medicine, New Haven, CT; zDepartment of Pathology,Emory University, Atlanta, GA; yDepartment of Pathology,Toronto General Hospital, Toronto, Ontario, Canada; JDepartmentof Pathology, Massachusetts General Hospital; **Department ofMedical Oncology, Dana-Farber Cancer Institute, Boston, MA;#National Institute of Diabetes, Digestive and Kidney Diseases(RTJ), Bethesda, MD; wwDepartment of Pathology (RVL), MayoClinic, Rochester, MN; Departments of zzPathology; zzzGastro-intestinal Medical Oncology, M.D. Anderson Cancer Center,Houston, TX; yyDepartment of Medicine, Rhode Island Hospital/Brown University, Providence, RI; JJDepartment of MedicalSciences, University of Upsalla, Upsalla, Sweden; zzDepartment ofMedicine, St. James’s Hospital and Trinity College, Dublin, Ireland;##Department of Pathology and Laboratory Medicine, University ofParma, Parma, Italy; wwwDepartment of Pathology, Medical Collegeof Wisconsin, Milwaukee, WI; zzzDepartment of Pathology, CathayGeneral Hospital, Taipei, Taiwan; yyyDepartment of Pathology,Vanderbilt Medical School, Nashville, TN; and JJJDepartment ofInternal Medicine, Charite Hospital, University of Berlin, Berlin,Germany.

Supported by Novartis Corporation.David S. Klimstra, MD, Irvin R. Modlin, MD, PhD, Shared first

authorship.Correspondence: David S. Klimstra, MD, Surgical Pathology Service,

Department of Pathology, Memorial Sloan-Kettering Cancer Center,1275 York Avenue, New York, NY 10065 (e-mail: [email protected]).

Supplemental digital content is available for this article. Direct URLcitations appear in the printed text and are provided in the HTMLand PDF versions of this article on the journal’s Web site(www.ajsp.com).

ORIGINAL ARTICLE

300 | www.ajsp.com Am J Surg Pathol � Volume 34, Number 3, March 2010

particularly well-described in the lung, tubular gastro-intestinal tract, and pancreas. As the entity of ‘‘carcinoidtumor’’ was first proposed by Oberndorfer over 100 yearsago,56 the clinical and pathologic features of NETs havebeen described by many investigators, with most studiesfocusing on subsets of tumors restricted to 1 organ ororgan system.30,40,46 For this reason, and also because thelarger family of NETs is highly diverse in terms of origin,mechanism of development, functional status, histologicpatterns, and biologic behavior, many different diag-nostic terms and classification systems have arisen. Forexample, ‘‘neuroendocrine tumor,’’ ‘‘carcinoid tumor,’’‘‘endocrine neoplasm,’’ and ‘‘neuroendocrine carci-noma’’ have all been applied to small intestinal pri-maries.6,9,47,58,80 In general, carcinoid tumors of thetubular gastrointestinal tract have been separated frompancreatic endocrine neoplasms.8–10,28 Site-specific grad-ing, staging, and classification systems have been devel-oped by the World Health Organization (WHO) and theEuropean Neuroendocrine Tumor Society (ENETS), andadditional proposals are under development by theAmerican Joint Committee on Cancer (AJCC), and theNorth American Neuroendocrine Tumor Society (NA-NETS).9,10,21,28,75,76 Furthermore, multiple different in-dividual classification schema have been published,22,30,45

and numerous studies have investigated potential prog-nostic factors based on morphology, immunophenotype,and molecular biology.7,11,17,31,35,44,66,74,99 Although allof these different systems have merit and can be usedto stratify the biologic behavior of neuroendocrinetumors,14,24,29,30,64 the specific criteria differ between thevarious anatomic sites, and translation between the classi-fication systems is not always possible.54 As promisingnew therapies arise for this family of tumors,41,61,96 it isof increasing clinical relevance to carry out accurateclassification and prognostication to properly tailortreatment, not only for surgically resected primary tumorsbut also for metastatic disease, from which only biopsyspecimens may be available. The lack of a single uniformsystem of nomenclature, grading, and staging for NETswill hamper efforts to evaluate new therapies and tocompare the results of published therapeutic trials.54

A careful examination of the major publishedclassification systems for NETs reveals considerableoverlap in the essential information used to derive thespecific subgroups. Tumor size, extent of invasion, andpresence of nodal or distant metastases are well-acceptedstaging parameters for all primary sites.9,10,21,35,38,75,76

The proliferative index has emerged as a fundamentalgrading characteristic that appears in most majorschema.1,5,6,33,62,63,93 Thus, there likely exists a compen-dium of basic information that should be collected for allneuroendocrine tumor specimens that would allow trans-parent and accurate data comparisons. Whereas the stan-dardization of the terminology and classification criteria forall NETs of every anatomic site would be a laudable butlikely elusive goal, the development of a ‘‘minimum dataset’’ that could easily be translated into any chosenclassification system should be more easily attainable.

The purpose of this workshop or meeting was tobring together recognized experts in the field of neuroen-docrine tumors, including pathologists and surgeons,oncologists, and gastroenterologists, to (1) review existingclassification systems, (2) evaluate the published literaturerelated to prognosis-related parameters or treatment-related parameters, (3) determine which pathologic para-meters are currently (or likely will become) important topredict prognosis and guide therapy, (4) decide what basicinformation related to diagnosis, grading, staging, andprognosis should be included in pathology reports, and(5) draft a checklist of the minimal pathology data setfor NETs.

METHODSPhysicians with dedicated expertise in the field of

NETs were identified. An effort was undertaken to recruitindividuals from different countries and to include bothpathologists (12 participants) and clinicians (8 partici-pants, including 2 surgeons, 2 gastroenterologists, and4 medical oncologists). Panel members were recruitedfrom the United States, Canada, Europe (Sweden, Germany,Italy, and Ireland), and Asia. Participants in majororganizations that have developed (or will develop soon)classification systems for NETs were selected, includingthe WHO, the ENETS, the NANETS, the AJCC, and theCancer Committee of the College of American Patho-logists. An effort was made to ensure balanced represen-tation of individual groups and institutions to provide abreadth of input from clinical, scientific, and geographicperspectives. A biostatistician was also included. A groupmeeting was held in Miami Beach, FL from February 6to 8, 2009. Each participant was asked to prepare abrief presentation related to specific aspects of the topic,to collectively review the existing published data and toensure that the ensuing discussions would be based onuniform background information.

The Delphic consensus process,19,25 was used toachieve agreement on issues related to the analysis ofthe pathology of NETs. Before the meeting, a series ofquestions were developed and distributed to the groupfor review. Each question was formatted to be answer-able as ‘‘yes’’ or ‘‘no.’’ The questions covered aspects ofterminology, grading and staging, immunohistochemicalevaluation, assessment of prognostic factors, and otherissues, all related to the information that should becontained in standard pathology reports of NET speci-mens. At the meeting, the group discussed each of thequestions to debate controversial issues and achievesagreement where possible. Consistent with the Delphicprocess, some questions required modification to enablegeneral agreement to be reached. After the discussionperiod, a formal vote was taken, and all participants wererequired to vote on each question. Each question wasread aloud, and anonymous voting was conducted usingcomputerized electronic voting devices. The results of thevoting were immediately tabulated. The voting categoriesare displayed in Supplementary (Table 1, Supplemental

Am J Surg Pathol � Volume 34, Number 3, March 2010 Pathology Reporting of Neuroendocrine Tumors

r 2010 Lippincott Williams & Wilkins www.ajsp.com | 301

Digital Content 1, http://links.lww.com/PAS/A49). Votesof ‘‘Agree strongly’’ and ‘‘Agree with minor reservation’’were considered ‘‘yes’’ votes; votes of ‘‘Disagree withmajor reservation (disagree moderately)’’ and ‘‘Disagreestrongly’’ were regarded as ‘‘no’’ votes. Consensus wasdefined to be achieved if at least 80% of the partici-pants voted ‘‘yes’’ or ‘‘no.’’ In the event of no immediateconsensus, the topic was reopened for discussion, and asecond vote was then taken. If 80% agreement was notattainable, further debate to allow modification of theproposition was then undertaken and a third vote taken.If a consensus was still not attainable, the question wasdeemed to have no agreement. The final phrasing of thequestions and the result of the final vote on each questionwere recorded.

RESULTSThe complete list of Delphic questions along with the

consensus answers are presented in Supplementary (Table 2,Supplemental Digital Content 2, http://links.lww.com/PAS/A50).Of 108 questions, consensus was reached based on thedefinition of 80% agreement for 91 (84%) of thequestions. Most of the questions upon which there wasno agreement (Supplementary Table 3, SupplementaryDigital Content 3, http://links.lww.com/PAS/A51) showedmarked polarization of opinion, participants usuallyselecting a voting category to register either a ‘‘yes’’ or‘‘no’’ vote [rather than choosing ‘‘Agree with majorreservation’’ or ‘‘Disagree with minor reservation (dis-agree mildly),’’ votes that would have suggested no strongopinion on the question]. For only 1 question with noagreement would the change of a single vote have allowedconsensus to be reached. A discussion of the general areasof consensus and disagreement is presented below, dividedinto broad topic areas.

TerminologyThe group did not attempt to develop a single

standard terminologic classification system for NETs, butdid address limited specific issues. The terms ‘‘endocrine’’and ‘‘neuroendocrine’’ are variously used for this familyof tumors,4,36 and there is justification for either term.There was agreement that these 2 terms could be usedsynonymously for differentiated neoplasms with epithelialand (neuro)endocrine differentiation in the gastro-enteropancreatic system. It was also agreed that ‘‘carci-noid tumor’’ has become archaic and that it should beavoided as the primary diagnostic term. However, therewas recognition that the term has become entrenched inthe lexicon, and as a practical matter it may still bereasonable for the foreseeable future to reference thisterm (for example, parenthetically) in the pathologyreport to ensure fluid communication. There was con-siderable debate about the use of ‘‘tumor’’ versus‘‘neoplasm’’ Although some participants pointed outthat all NETs are regarded to be neoplastic, justifyingthe designation (neuro)endocrine neoplasm (44% favoredthis term), others argued that the (neuro)endocrine tumorterminology has become widely disseminated in several

major recent classification systems (that of the EN-ETS,75,76 for example), and that to abandon the termat this point would require excessive reeducation (44%preferred to retain the term ‘‘tumor’’). The other termino-logic issue that could not be resolved involves the useof the term ‘‘carcinoma,’’ especially when applied towell-differentiated NETs that have displayed clinicallyobvious malignant behavior. The WHO classificationfor small bowel and pancreatic NETs uses the term‘‘tumor’’ for organ-confined disease and ‘‘carcinoma’’ forotherwise identical neoplasms with metastases or grosslocal invasion.9,28,38 A number (69%) of the participantsconsidered that this distinction reflects differences intumor stage rather than an inherent difference in tumorbiology, as suggested by the use of ‘‘carcinoma.’’Conversely, others argued for the use of the ‘‘carcinoma’’for all well-differentiated NETs. This discussion andthe lack of consensus on this issue amplified the widediversity of opinions about the implications of diagnosticterminology.

Diagnostic ImmunohistochemistryDemonstration of neuroendocrine differentiation

by the use of specific immunohistochemical stains canbe helpful in the diagnosis of NETs.50 The vast majorityof well-differentiated NETs express 1 or more generalneuroendocrine markers (such as chromogranin A andsynaptophysin), and it was agreed that the use ofimmunohistochemistry for diagnostic purposes shouldbe recommended in a majority of cases, including in allcases of biopsies of metastatic disease. There was noagreement that diagnostic immunohistochemistry shouldbe mandated for all NETs, however. Some participants(53%) considered that routine immunohistochemicalstaining is not necessary to diagnose histologically typicalexamples of well-differentiated NETs (carcinoid tumors)of the ileum, appendix, and stomach (such as the multipletumors in patients with hypergastrinemia-associatedneuroendocrine tumors) or certain pancreatic endocrinetumors such as clinically functional insulinomas. Theuse of diagnostic immunohistochemistry for cases withunusual or unclear histologic features was encouraged.The group agreed that the only 2 stains to be routinelyrecommended are chromogranin A and synaptophysin.There was agreement that staining routinely for otherneuroendocrine markers including chromogranin B,CD56 (neural cell adhesion molecule, N-CAM), CD57(leu7), and neuron-specific enolase was not indicated, andthat the specificity of these markers (especially neuron-specific enolase) was questionable. There was alsoconsensus that stains for keratins were not to be routinelyrecommended, although most well-differentiated NETsdo express keratins (in particular, cytokeratins 8 and 18,detected by Cam5.250), and that p53 immunostainingalso was not essential. The group further agreed that thereare only limited indications to carry out stains for specificpeptide hormones or bioamines, even when the clinicalhistory suggests the presence of a ‘‘functional’’ NET. Asfunctional NETs are not defined by immunohistochemical

Klimstra et al Am J Surg Pathol � Volume 34, Number 3, March 2010

302 | www.ajsp.com r 2010 Lippincott Williams & Wilkins

labeling, but rather by clinical symptoms and serology,28

a statement about the functional nature of the tumor isnot needed in the pathology report. Furthermore, thereare no histologic features that define a functional NETand there is no difference in biologic behavior betweenfunctional and nonfunctional NETs, independent ofother prognostic factors such as size, stage, and grade.30

Irrespective of the clinical syndrome or the bioactiveagent produced, both types are NETs and should beregarded as a single entity. Limited peptide immunohisto-chemistry could be done (eg, for insulin or gastrin) if thereis a clinical indication to show the production of a specificpeptide in a functional tumor, such as in a patient withmultiple endocrine neoplasia type 1 (MEN1) with morethan 1 pancreatic NET.

For patients presenting with metastatic diseasefrom an unknown primary, an attempt should be madeto determine the primary site because of the differingtherapeutic options available for pancreatic versusintestinal NETs.41,61,70,96 In the case of well-differentiatedNETs, immunohistochemical staining can be helpful todetermine the primary site; stains for CDX2 and TTF1should be carried out, which would point toward anorigin in the intestines or pancreas (for CDX2) and lung(TTF1), respectively.32,78,84 The sensitivity and specificityof these stains require more study, however, and negativestaining does not exclude origin in these sites. It should benoted that staining for TTF1 is not helpful for high-gradeneuroendocrine carcinomas, such as small cell carcinoma,as extrapulmonary examples may express this mar-ker.13,51,81,92 Peptide immunohistochemistry was felt tohave limited value in the determination of the primary sitefor NETs presenting with metastatic disease and was notrecommended. The group also recommended against theuse of silver-based histochemical stains such as argentaffinor argyrophil reactions, as these stains have largely beenreplaced by immunohistochemical markers such aschromogranin and synaptophysin.

The possibility of carrying out immunohistochem-ical staining for somatostatin receptors (SSTRs) wasdiscussed,65,91 given the therapeutic implications of SSTRexpression.55,59,72,89 However, it was agreed that thesestains are not currently routinely indicated, because wellcharacterized commercially available SSTR subtypeantibodies do not exist, and the information about SSTRexpression to some extent can be inferred based onsomatostatin receptor scintigraphy (SSTR 2 and 5), whichis now widely carried out in patients with NETs.60,89

Further evaluation of the clinical relevance of SSTRsubtype evaluation of individual NETs may lead toincreased application of this assessment to determine thespecific utility of SST analogues with different receptorprofiles.91

StagingThe prognostic importance of tumor stage is widely

reflected in all of the classification systems for NETs. Thediscussions focused on the specific features of the primarytumors that should be included in pathology reports that

will allow proper tumor staging to be carried out usingany of the well-established systems. For all NETs, thesize of the tumor (in 3 dimensions) should be reported. Inthe tubular gastrointestinal tract, the depth of maximalinvasion through the wall should be stated, using thesame landmarks as for the staging of exocrine carcinomasof the corresponding locations.21,75,76 Reporting of themaximal thickness of the tumor was felt to be unnecessaryhowever. For appendiceal NETs, invasion into themesoappendix should be reported, and the extent (limitedvs. extensive) should be documented. There was noagreement concerning the measurement of the amountof mesoappendiceal invasion however. For pancreaticNETs, the presence of extrapancreatic invasion should bereported.

Lymph node metastases should be reported, includ-ing the number of involved nodes and the total number ofnodes examined. The number of nodes needed to ensure‘‘adequate’’ staging has not been defined, but attemptsshould be made to identify as many lymph nodes in theresected specimen as possible. There was no agreementabout the necessity of distinguishing micrometastasesfrom macrometastases (47% favored making a distinc-tion), in part because no universally acceptable definitionof micrometastasis has been proposed for NETs. Inaddition, it was not agreed whether the size of the largestlymph node metastasis should be documented (56%believed the size should be reported). The group agreedthat TNM staging should be carried out using 1 of theproposed staging systems,21,75,76 but it was also recog-nized that several systems exist, none of which is currentlyuniversally accepted. For this reason, it is critical toindicate in the report which specific TNM staging systemhas been used.

The status of the resection margins should bereported, both for primary tumor resections and forresected metastases. For resected tumors with closemargins (ie, within less than 0.5 cm), it was recommendedthat the distance to the margin be indicated. It was notagreed to be necessary to document the distance if thetumor is more than 0.5 cm from the margin, but 74% stillthought it should be reported.

GradingAnother fundamental predictor of outcome in

NETs is the grade of the tumor. The group recognized3 grading categories of NETs: low grade, intermediategrade, and high grade.38,75,76,87 It was further agreed thatin general, well-differentiated NETs are regarded to beeither low grade or intermediate grade, whereas poorlydifferentiated NETs are high grade by definition. Thegroup recognized the dramatic difference in clinicalbehavior between well-differentiated NETs (which canprogress very slowly over years to decades, even in thepresence of metastatic disease) and poorly differentiatedNETs (which are high-grade carcinomas characterized byrapid dissemination, resistance to treatment, and aquickly fatal course). Thus, the fundamental distinctionof well differentiated from poorly differentiated NETs



was recognized to be a critical pathologic determination.Grade progression in well-differentiated NETs wasdiscussed, and it was agreed that this phenomenon canoccur.43,52,69,100 Thus, it is possible that different regionswithin a single tumor (or different sites of metastasis) candisplay different grades. However, there has been littlehistologic documentation of grade progression and it isnot clear how often NETs may undergo this change.There have been rare reports of high-grade transforma-tion of well-differentiated NETs,86 but this occurrenceseems to be particularly rare; there are insufficient datato determine the biologic behavior of such tumors. Thus,there was no agreement about whether well-differentiatedNETs could ever be biologically high grade (68% feltthat well-differentiated NETs can never be high grade).It was noted that in some circumstances, more than1 primary NET may arise within a single organ. In thecase of familial multicentric NETs, it was agreed thatthe different tumors can also have different grades,but there was no agreement on this point regardingsporadic multicentric NETs. There was a consensus thatno NETs should be reported as ‘‘benign’’ (other thantightly defined subsets such as pancreatic or pulmo-nary NETs less than 0.5 cm or gastric NETs less than0.5mm).35,37,38,83

The group recognized that different grading systemsexist for well-differentiated NETs of different anatomicsites, and although there was consensus that a uniformgrading system applicable to all primary sites wouldbe desirable, the data to justify such a system do notcurrently exist. The difficulty of determining whichgrading scheme to apply to metastases of unknownprimary sites was acknowledged. Despite all of thevagaries of grading NETs, there was unanimous agree-ment that a grade should be stated in the report, withthe particular grading system used to be specified.Furthermore, the raw information used to derive thegrade should also be reported to enable comparisonbetween different grading systems. It was recognized thatthe proliferative index is inherently linked to the grade ofNETs but it was also appreciated that the proliferativefraction may not be homogeneous throughout a givenNET and that there may be differences in proliferationbetween the primary tumor and metastatic sites. Thesimplest method to determine the proliferative index isto count mitotic figures, and this piece of data wasunanimously accepted to be necessary for completepathology reports. The group agreed that the calculationof the number of mitoses per unit area of the tumor wasthe most reliable, and that the number of mitoses in 10high-power fields (2mm2) was a standard format used inmany classification and grading systems and should beendorsed. It was pointed out that some grading schemesuse the number of mitoses in 50 high-power fields,22,30

and the group agreed that a total of 50 fields should becounted, but ultimately the mitotic rate should beexpressed based on the number in 10 high-power fields.It is not adequate to count a single randomly chosen areaof the tumor; rather, the most active regions (‘‘hot spots’’)

should be identified by scanning the slides at intermediatemagnification, and the formal mitotic count shouldinclude the most active areas to ensure assessment ofthe worst case scenario for therapeutic strategy.

Significant discussion revolved around the use ofimmunohistochemical staining for Ki67 as an alternativemethod to quantify the proliferative fraction.33 TheKi67 index (expressed as the percentage of neoplasticcells showing nuclear labeling) is used in several gradingand classification systems and is incorporated into therecent ENETS proposals.28,75,76 In Europe, Ki67 is widelyused and reported as a critical pathologic predictor ofprognosis. In North America, however, the use of Ki67staining has largely been restricted to selected institutionsfor specific clinical or pathologic indications, and routinereporting of this marker is not currently the standard ofpractice. The group was divided regarding the advisabilityof incorporating Ki67 labeling indices into reports (53%said no), and there was no agreement that the stainshould be carried out in all cases, especially for resectionspecimens, in which an accurate mitotic count can becarried out readily. However, it was agreed that Ki67staining can be highly useful for biopsy specimens. Thevastly different labeling rates between well-differentiatedNETs (with a labeling index typically in the 1% to 20%range) and high-grade neuroendocrine carcinomas (suchas small cell carcinoma and large cell neuroendocrinecarcinoma, which typically show 50% to 95% labeling)make Ki67 staining very effective for distinguishing thesegroups.49,67 In addition, mitotic counting of small speci-mens is difficult, as many of these samples will not containsufficient neoplastic tissue to enable 50 high-power fieldsto be counted. There was consensus; therefore, that Ki67staining should be reported for biopsies of metastases orin other circumstances in which a larger resected portionof the tumor is unlikely to become available.

Another important consideration was that theoptimal method to determine the Ki67 index has yet tobe defined. Only 53% of participants believed that theKi67 index was sufficiently reproducible between pathol-ogists to discriminate clinically relevant subsets of NETs.Options to quantify Ki67 labeling include a general‘‘eyeballed’’ estimate of the percentage of positive cells,systematically counting manually a defined number oftumors cells (2000 as per the ENETS proposals75,76) andcalculating the positive percentage, or using a computer-ized digital image analysis system to measure the positivepercentage.18,48 Manual counting was felt by some to betoo time-consuming for general pathology practice, and itwas agreed that image analysis, although attractive and inuse at some institutions, was not yet sufficiently availablefor routine practice. There was no agreement that these 2techniques could be advocated at this time. Seventy-fivepercent felt that manual counting should not be done, and63% believed that image analysis is not yet ready forgeneral use. Thus, from a practical standpoint, providingan ‘‘eyeballed’’ estimate of the labeling percentage wasagreed to be the only method that could be stronglyadvocated at present. However, there was recognition of



many shortcomings to this approach. The interobserver(and likely intraobserver) variability was felt to be high,especially when a difference between 1% and 3% is usedto separate 2 grades in some systems.75,76 Furthermore, itwas recognized that there is intratumoral heterogeneityin Ki67 labeling (Fig. 1), meaning that a small sampleof a NET (eg, a biopsy of a metastasis) may not representthe most highly proliferative region of the tumor.15 Thegroup recommended to count the most densely stainingregions (‘‘hot spots’’) and to count a variety of areaswithin the tumor; it was specifically noted that countingof random areas or single regions is inadequate. Theresult should be reported as a single percentage reflectingthe average of the regions counted, rather than a range ofvalues. Although many issues were recognized regardingthe use of Ki67 staining, it was agreed that there currentlyexists no better marker of proliferation for routine use.

If multiple sites of disease are sampled, it was agreedthat separate mitotic counts should be provided for eachanatomic site (for example the primary lesion, lymphnode metastasis, liver metastasis). There was no agree-ment whether separate Ki67 indices should be reportedfor different sites of disease (61% said yes).

Some grading systems for NETs use the presence oftumor necrosis (ie, nonischemic or infarct-like necrosis) asa criterion to increase the tumor grade.30,87 Thus, it wasagreed that the presence of nonischemic tumor necrosisshould be documented in the pathology report.

The group acknowledged that the size of the pri-mary tumor may be associated with the mitotic rate and,therefore, the grade, but it was agreed that the tumor sizewas a parameter that helps define the stage of the tumorrather than the grade. Thus, grade and stage should beseparately determined and reported; the size of the NETshould not be used in the determination of the grade.

Metastasis-specific IssuesSeveral issues specifically relate to the reporting

of metastatic disease. For resected metastases, it wasrecommended to report the anatomic site, the number oflesions resected, and the size (in 1 dimension) of thelargest metastasis. Some comment about the percentageof involvement of the resected tissue (such as the liver)should be provided. Owing to the potential for hetero-geneity among different sites of metastasis, the groupadvised sampling more than 1 metastasis for histology,if multiple metastases are resected. It was recognizedthat currently there is no system for substaging distantmetastatic disease, meaning that patients with limitedspread outside the regional lymph nodes are groupedtogether with those having extensive, widely distributedmetastases. The desirability of a substaging system formetastatic disease was acknowledged, as it would betterstratify the extent of disease for comparison of therapeutictrials.

FIGURE 1. Immunohistochemical staining for Ki67. Two images captured from a single pancreatic neuroendocrine tumor (NET)reveal heterogeneity of labeling, from 1.4% in 1 region (A) to 21.5% in another (B).



Grading of metastases creates some special chal-lenges, as has been partially discussed above. Althoughin most cases well-differentiated NETs can be distin-guished from poorly differentiated (high grade) neuro-endocrine carcinomas, the distinction of low gradeand intermediated-grade NETs can be very challengingbased on biopsy material. Although the group agreedthat an attempt should be made to grade NETs onbiopsy specimens, it was only felt to be realistic whenadequate neoplastic tissue is represented in the biopsy,and distinction of low and intermediate-grade NETs isnot possible based on fine needle aspiration cytologyspecimens.

Other Prognostic Factors andTherapeutic Biomarkers

Many different potential prognostic factors forNETs have been reported. Histologic findings thatwere of recognized significance (and therefore, shouldbe reported) include vascular and perineural invasion.Immunohistochemical stains for endothelial markers(such as CD34 or D2-40) that can aid in identifyingvascular invasion were recommended only for cases witha histologic suspicion of vascular invasion. It was not feltnecessary to differentiate between lymphatic and bloodvessel invasion, as this distinction often cannot be madereliably on the basis of routine histologic sections. Thevarious architectural growth patterns of NETs were notconsidered to be prognostically relevant, nor was nuclearatypia. Unusual histologic features in NETs (includingclear cell or oncocytic morphology, and gland-formation)should be mentioned in the report, however, as suchfeatures may obscure the neuroendocrine nature of thetumor. It was not thought necessary to quantify angio-genesis or apoptosis. Although the immunoexpression ofcytokeratin 19 (CK19) reportedly has adverse prognosticimplications in pancreatic NETs,17 the data were not feltto be sufficiently compelling to recommend the routineuse of this marker.

Advances in targeted therapy have risen the possi-bility that molecular testing or immunohistochemistrymay become useful to determine the treatment of NETs.However, at present no special immunohistochemicalstains or molecular tests are routinely recommended.

Other InformationIt was agreed that abnormalities in the neuroendo-

crine cells in the resected organs away from the NETshould be described in the report. These include, forexample, gastric neuroendocrine cell hyperplasia asso-ciated with Type 1 or Type 2 gastric well-differentiatedNETs, or pancreatic endocrine microadenomas asso-ciated with well-differentiated pancreatic endocrine tu-mors in patients with MEN1.2,3,37,68,83 Many NETs mayexhibit stromal fibrosis, but it was agreed that there wasno indication to attempt to quantify stromal fibrosis.

DISCUSSIONMany different groups have explored the topic of

NET classification. Various systems of nomenclature,grading, and staging have been published, but none hasgained universal acceptance. Discussion of the systemsproposed by the WHO, the ENETS, the AJCC, and othergroups9,10,21,22,30,35,38,45,75,76 revealed that essentially all ofthese classifications are useful to stratify survival afterresection of NETs and to predict the rate of progressionof metastatic disease. A graphic comparison of majorgrading, staging, and classification systems is presented inFigure 2. This study was carried out in an effort to definethe essential pathologic information needed for clinicaldecision-making that in fact constitutes the data under-lying most of these NET classifications. There was nointent to introduce another classification system, as it wasperceived that adding to the plethora of available systemsalready in use would only cause greater confusion.

The application of the Delphic process19,25,77 in thisstudy seems to be novel for ‘‘consensus’’ groups preparingjoint statements related to NETs. Many group effortsinvolve open and extensive discussions, allowing allparticipants to present and defend their opinions. How-ever, the decision-making process is often not anonymousand introduces issues such as peer pressure, weight ofpersonal opinion, and vigor of personality into thedeliberations. The use of anonymous voting with anelectronic system in this study ensured that all partici-pants had an equal opportunity to express an independentopinion. It is possible that this methodology preventedagreement from being reached on some issues, as theholders of minority opinions could less readily beswayed to change their votes. Nonetheless, agreementwas achieved for 84% of the questions.

Major areas of agreement were achieved related toterminology, use of diagnostic immunohistochemistry,grading parameters, staging information, relevance ofother histologic prognostic features, and utility of ther-apeutic biomarkers. The group agreed upon the inter-changeable use of ‘‘endocrine’’ and ‘‘neuroendocrine’’terms to refer to these tumors and the desirability ofeventually eliminating the term ‘‘carcinoid.’’12,82 Thespecific immunohistochemical stains useful to defineneuroendocrine differentiation were agreed upon, as werethe limited circumstances under which staining forbioactive peptide or amine secretory products (hormones)may be helpful. As some novel therapeutic agents haveshown greater efficacy against NETs from specificprimary sites (eg, temozolamide for pancreatic NETs42),the need to attempt to determine the origin of meta-stases from occult primary NETs was recognized, andimmunohistochemical staining for CDX2 and TTF1 wasrecommended as a diagnostic aid.32,78,84 Recent data alsosuggest that staining for Isl1 may also be useful to definepancreatic origin.79 The group also recognized manypotential immunohistochemical biomarkers of therapeu-tic responsiveness on the horizon, including somatostatinreceptor subtypes, MGMT, VEGF-r, and mTOR path-way members26,34,42,53,73,94,97; however, none of these has



been adequately validated to justify their routine use, anddata showing the correlation between immunohistochem-ical staining and therapeutic response are still emerging.However, it seems likely that in the future, with increasingadvances toward specific molecular characterization ofNETs in terms of transcript identification (KiSS, MAGE,NAPI) and functional biologic components (somatostatinreceptor subtypes), that such information may becomerelevant.20

Formal TNM staging of NETs has been relativelylimited compared with the staging of carcinomas. Forexample, in the most recent AJCC staging classification,gastroenteropancreatic NETs are specifically excludedfrom the TNM system.27 A relatively crude staging systemexists in the Surveillance, Epidemiology, and End ResultsProgram database of the NCI, which categorizes the

extent of disease as ‘‘localized,’’ ‘‘regional,’’ or ‘‘dis-tant.’’95 However, there is widespread recognition thatNETs should be staged using the TNM system, and thereis currently an ENETS proposal for staging gastro-enteropancreatic NETs.75,76 The upcoming revision of theAJCC staging system will also include NETs.21 Althoughthere may be some subtle differences between the ENETSand AJCC systems, the fundamental staging landmarksparallel those used for carcinomas of the correspondingorgans, and the group recommended to include thesize of the tumor, the extent of invasion based on theselandmarks, the status of regional lymph nodes, andthe presence of distant metastatic disease (if known) inthe pathology reports, in addition to indicating a specificTNM stage. Reporting of margin status was alsorecommended.

G r a d e

S t a g e

WD NETLow Grade

WD NETIntermediate Grade

WD NE Carcinoma MD NE Carcinoma

PD NE Carcinoma

PD NE Carcinoma

Others

WHO Lung Typical Carcinoid Atypical CarcinoidHG NE Carcinoma

Small Cell/Large Cell

G1 G2 G3ENETS

T1 T

2 T3 T

4 N1 M

1

Well Differentiated(Neuro)endocrine Tumor

Localized

W H OPoorly Differentiated

(Neuro)endocrineCarcinoma

W H O

Well Differentiated(Neuro)endocrine Carcinoma

SE

ER

AJC

C

Regional D

istant

FIGURE 2. Comparison of various grading, staging, and classification systems for neuroendocrine tumors (NETs). The gradingsystems are displayed along the x-axis (top) and the staging systems along the y-axis (left). The WHO systems for gastrointestinaland pancreatic NETs include a combination of grading and staging information and are displayed with the box (lower right). Thevarious nomenclature of each system is used. The overlaps between the different systems are approximate. Abbreviations: AJCCindicates American Joint Committee on Cancer; ENETS, European Neuroendocrine Tumors Society; HG, high grade; MD,moderately differentiated; NE, neuroendocrine; NET, neuroendocrine tumor; PD, poorly differentiated; SEER, Surveillance,Epidemiology, and End Results Program of the National Cancer Institute; WD, well differentiated; WHO, World HealthOrganization.



There was universal recognition of the importanceof grading NETs. The distinction of relatively slowlyprogressive well-differentiated NETs from highly aggres-sive poorly differentiated NETs (small cell carcinoma andlarge cell neuroendocrine carcinoma) was highlightedas a critical pathologic determination.6,14,23,62 The rele-vance of tumor cell proliferation for grading NETs wasalso agreed. Mitotic rate was regarded as an essentialcomponent of pathologic data, and specific recommen-dations for the technique of mitotic counting wereestablished. The existence of 3 grades of NETs was alsoagreed upon, and it was accepted that inclusion of thegrade of the tumor, along with a reference to the specificgrading system being used, should be included in allreports on NETs. Although there was complete agree-ment that grade can progress as NETs grow anddisseminate,43,52,69,100 this was not considered a uniformevent in the evolution of the disease. An important issuethat was widely accepted was that intratumoral hetero-geneity of proliferative rate requires careful considerationgiven its potential impact on overall assessment of theNET aggressiveness. Thus, separate grading and mitoticrate counting of metastases should be carried out, andeach major site of disease should be assessed indepen-dently. The inclusion of information about nonischemicnecrosis is used in some grading systems and therefore,should be mentioned in the reports.30,87 The use ofimmunohistochemical staining for Ki67 as a measure ofproliferative rate was endorsed by the group for specificsituations (see below for further discussion about Ki67staining). The stain is especially useful for the distinctionof high-grade neuroendocrine carcinoma from well-differentiated NETs and should always be used whenthis differential diagnosis arises (particularly on biopsyspecimens).49,67 In addition, in cases with minimal tumorfor which adequate mitotic counting is not possible, Ki67staining should be used to determine the proliferativerate, as it labels a greater proportion of cells than thosewith active mitotic figures.

Other prognostic factors that were agreed to havesufficient relevance for inclusion in reports includedthe presence of vascular and perineural invasion. Nucleargrading was not felt to be relevant, based on the ob-servation that nuclear pleomorphism can be seen insome well-differentiated NETs that lack aggressivebehavior.98 Quantification of the extent of fibrosis,angiogenesis, and apoptosis was not considered to benecessary. Unusual histologic features, although unlikelyto have prognostic significance, were believed to beworthy of documentation in the reports.

The 17 questions for which no agreement could bereached (Supplementary Table 3, Supplemental DigitalContent 3, http://links.lww.com/PAS/A51) include somerelatively minor issues related to the precision with whichthe extent of disease should be reported but also moresubstantive issues regarding terminology, the need forroutine immunohistochemical staining, the concept ofprogression of well-differentiated NETs to high-gradeneuroendocrine carcinoma, and the use of proliferation

markers. Some of these more substantive disagreements(particularly those related to terminology) reflect theore-tic concerns bolstered by ‘‘personal opinions’’ and long-held traditions of practice. Whether to use ‘‘tumor’’ or‘‘neoplasm’’ and whether the term ‘‘carcinoma’’ should beapplied to well-differentiated NETs are important con-siderations that could not be resolved, but the basis forthe disagreement is unrelated to a lack of understandingof the disease or the absence of sufficient data. Weconsider these to reflect the evolution of semanticterminology and not a real disagreement in terms of thehistopathology or biology of NETs. It was clear to allthat much of the terminology (tumor or carcinoid) reflectsarchaic medical lexicon and will with the passage of timefade into medical history rather than be banned fromusage.12,16,82 The routine use of immunohistochemicalstaining for general neuroendocrine markers was alsocontroversial, the divergent opinions reflecting variedlevels of comfort with diagnoses rendered solely on thebasis of routine histology.

Some areas of disagreement were based on a lack ofsufficient data to support specific concepts or procedures.The concept of neoplastic progression in NETs wasaccepted, but the group could not agree whether well-differentiated NETs can progress to fully high-gradeneuroendocrine carcinomas.86 Some preliminary datadocumenting this occurrence were presented, but therehave been few published examples, and it seems that, atbest, this transformation is probably rare and does notrepresent the pathway of development of a vast majorityof high-grade neuroendocrine carcinomas.

Another controversial area was the use of Ki67staining to determine the proliferative rate.33 Althoughthis stain is widely used, particularly to distinguish well-differentiated NETs from poorly-differentiated NETssuch as small cell carcinoma,49,67 there was a division ofopinion about when to use it and whether it should becarried out for all specimens of NETs, especially thosewith sufficient material to allow mitotic rate counting.The ENETS grading system includes Ki67 labeling indexas 1 of the parameters, and oncologists in Europe relyextensively on this information for treatment.33,71,75,76,90

In the US, however, the use of Ki67 has been morelimited, reflecting concerns with rigor of the assessmentof the parameter. The interpretation of Ki67 staining wasalso debated. Although it was agreed that the mostaccurate assessment would involve counting nuclei todetermine the true labeling percentage, this was consid-ered by some to be too time-consuming and cumbersomefor routine practice. Digital image analysis holds promisefor quantification of staining, but many pathologists donot currently have access to this technology. An ‘‘eye-balled’’ estimate was acknowledged to be inaccurateowing to have high interobserver variability, but thistechnique was agreed to be the only one that couldpractically be recommended.18 Nevertheless, given theissues with reproducibility of this technique and theinherent heterogeneity of staining within NETs (especiallymetastases, from which only limited biopsy samples are



often available15), it was not agreed that the Ki67 labelingindex is sufficiently reproducible to allow separation ofclinically relevant subsets of NETs. It is clear that muchof the debate about the use of Ki67 staining was related toa lack of data regarding intraobserver and interobserverreproducibility, correlation of automated quantificationwith clinical outcome, assessment of intratumoral hetero-geneity, and comparison with mitotic rate as the ultimatedeterminant of outcome and indication for specific pharma-cologic therapy. The resolution of these critical issues wasagreed upon as a priority in the future investigation andassessment of NETs.

The areas of consensus reached by this group weresufficient to allow the development of a Minimum Patho-logy Data Set for the reporting of NETs (Table 1). ThisMinimum Pathology Data Set includes recommenda-tions for reporting both biopsy and resection specimensfrom primary tumors and metastases, and it applies to allNETs of the gastrointestinal tract and pancreas. Ingeneral, the group recommended reporting only thosehistologic features for which abundant data exist todocument their prognostic significance. In each of thesesites, a specific diagnosis and grade are to be reported,and staging is to be carried out when the information isavailable (ie, for resections and biopsies of metastases).However, there was no mandate to use any specific systemof nomenclature, grading, or staging, because the infor-mation that is used for these purposes is to be includedelsewhere in the report. For example, the mitotic rate,with or without information about necrosis, is used tograde NETs in all major systems; and this informationshould be routinely reported. The extent of invasion is tobe reported, using similar landmarks to those used in theAJCC staging systems for carcinomas of correspondingorgans. In fact, this should translate easily into theupcoming AJCC staging system for NETs of the GI tractand pancreas, which in almost every instance is basedon the corresponding carcinoma staging system. It isimportant to indicate in the report which systems ofnomenclature, grading, and staging are being used toallow ready comparison among cases.

A number of items were regarded as ‘‘optional’’ forinclusion in reports, based upon lack of agreement amongthe participants about their importance. For example,routine staining for general neuroendocrine markers(chromogranin and synaptophysin) of resected NETswith classic histologic features was not agreed to benecessary, but can be carried out and reported if thepathologist finds it necessary or the clinical situationrequires it. Routine Ki67 staining was also deemedoptional, unless the specimen in question is a biopsyof a primary NET in which a high-grade neuroendocrinecarcinoma could not be excluded or a biopsy of a meta-stasis with inadequate tissue for accurate mitotic count-ing. Measurement of the distance from the tumor to theresection margins for cases grossly within less than 0.5 cmwas also considered optional. It was clear from thediscussions, however, that some of this optional informa-tion might be regarded as mandatory in certain practice

TABLE 1. Minimum Pathology Data Set: Information to beIncluded in Pathology Reports on NETs

For resection of primary tumors:

Anatomic site of tumorDiagnosis (functional status need not be included in pathologyreport)Size (3 dimensions)Presence of unusual histologic features (oncocytic, clear cell, gland-forming, etc.)Presence of multicentric disease(OPTIONAL: immunohistochemical staining for generalneuroendocrine markers)

ChromograninSynaptophysinPeptide hormones, IF a specific clinical situation suggests that the

correlation with a functional syndrome may be helpfulGrade (specify grading system used)Mitotic rate (number of mitoses per 10 high-power fields or 2mm2;

count 50 high-power fields in the most mitotically active regions, countmultiple regions)

[OPTIONAL: Ki67 labeling index (count multiple regions withhighest labeling density, report average percentage; ‘‘eyeballed’’estimate is adequate)]Presence of nonischemic tumor necrosisPresence of other pathologic components (eg, nonneuroendocrinecomponents)Extent of invasion (use anatomic landmarks for the AJCC T-stagingof analogous carcinomas of the same anatomic sites)

Stomach: depth of invasion into/through gastric wallSmall bowel: depth of invasion into/through bowel wallLarge bowel: depth of invasion into/through bowel wallAppendix: depth of invasion into/through appendiceal wall;

presence and extent of mesoappendiceal invasionPancreas: the presence of extrapancreatic invasion or invasion of

bile duct, duodenum, or ampullaAll sites: involvement of serosal/peritoneal surfaces; invasion of

adjacent organs or structuresPresence of vascular invasion (OPTIONAL: performimmunohistochemical stains for endothelial markers if needed)Presence of perineural invasionLymph node metastasesNumber of positive nodesTotal number of nodes examined

TNM Staging (specify staging system utilized)Resection margins (positive/negative/close) (OPTIONAL measuredistance from margin if within 0.5 cm)Proliferative changes or other abnormalities in nonneoplasticneuroendocrine cells

For Biopsy of Primary Tumors:

Anatomic site of tumorDiagnosis (functional status need not be included in pathology report)Presence of unusual histologic features (oncocytic, clear cell, gland-forming, etc.)(OPTIONAL: immunohistochemical staining for generalneuroendocrine markers)

ChromograninSynaptophysinPeptide hormones, IF a specific clinical situation suggests that the

correlation with a functional syndrome may be helpfulGrade (specify grading system used)Mitotic rate (number of mitoses per 10 high-power fields or 2mm2;

count up to 50 high-power fields)Ki67 labeling index, for biopsies in which a diagnosis of high-grade

neuroendocrine carcinoma cannot be excluded (count multiple regionswith highest labeling density, report average percentage; ‘‘eyeballed’’estimate is adequate)Presence of nonischemic tumor necrosisPresence of other pathologic components (eg, nonneuroendocrinecomponents)



situations, and close collaboration with treating cliniciansis suggested to ensure the necessary data are included inthe reports.

Although there was some disagreement about theimportance of various pathologic findings, and nouniversal grading or staging system arose through thisconsensus meeting, there was general agreement that itwould be preferable to develop such a system. Ideally, thegrading parameters for well-differentiated NETs wouldbe the same irrespective of the site of origin, and theterminology would also be uniform. One of the reasonsfor the variety of different systems is that most areproposals based on a reasonable assessment of potentially

significant factors, rather than data-driven subgroup-ings based on extensive study of actual cases. The grouprecognized the need for improved clinicopathologicstudies of NETs based on large numbers of cases withlong-term follow-up, which almost certainly will need toinvolve multiple institutions. The existing data actuallysuggest that different grading parameters may optimallyseparate prognostic groups from different anatomicsites.39,62 For example, some studies have shown that amitotic rate of 2/10 high-power fields optimally separateslow-grade NETs from intermediate-grade NETs in thelung,88 whereas other investigators suggest that a muchlower rate of 2/50 high-power fields should be usedfor NETs in the pancreas and rectum.22,30 Many studieshave used a mitotic rate of 10/10 high-power fieldsto distinguish well-differentiated (low and intermediategrade) NETs from poorly-differentiated (high grade)neuroendocrine carcinomas in sites such as the lung,thymus, ampulla of Vater, and pancreas28,57,85,88; how-ever, according to the ENETS grading scheme, a rateof 20/10 high-power fields is preferable for this distinc-tion.75,76 Unification of grading criteria among anatomicsites would have the advantage of allowing grading ofmetastatic NETs for which the primary site may not beknown. Until adequate data are obtained to determinewhether a unified grading system can be developed, how-ever, it is critical to continue to collect the specific datathat underlie the grading schema being used. Namely, theproliferative rate (mitotic rate, and/or Ki67 labelingindex) should be specifically recorded for all NETs. Thiswill allow translation between different systems and willfacilitate comparison of data from studies that may usedifferent cut-points for grading.

It must also be recognized that clinical or radio-graphic information is useful in predicting the biologyof NETs. For patients with metastases, a decision totreat can also be based on presence of symptoms,whether hormone-mediated or based on the site andburden of disease. Longitudinal radiographic studies canbe used to determine the growth rate of metastases, whichmay vary among different sites of metastatic disease.73

This clinical and radiographic information must beintegrated with the pathology data for optimal thera-peutic decision-making.

The development of a standardized approach tothe analysis of the pathology of neuroendocrine tumordisease is a necessary step to provide a secure platformupon which the evolving therapeutic landscape can bebased, and it is vital to establish clear parameters withinwhich the disease can be uniformly evaluated and rationaltherapeutic strategies developed. A clear advance wasthe development of a Minimum Pathology Data Set asproposed by this group. The routine application and useof such information will result in the accumulation ofconsistent and globally applicable information about thepathology of NETs. With time, this will translate into animproved capability to define the disease of individualpatients and facilitate provision of a uniform basis toprognosticate and treat patients. Widespread adoption of

TABLE 1. (continued)

For Resection of Metastatic Tumors:

Location of metastasis(es)Diagnosis (functional status need not be included in pathology report)Number of metastases resectedExtent of involvement of resected tissue (percentage)Greastest dimension of largest metastasisPresence of unusual histologic features (oncocytic, clear cell, gland-forming, etc.)(OPTIONAL: immunohistochemical staining for generalneuroendocrine markers)

ChromograninSynaptophysinPeptide hormones, IF a specific clinical situation suggests the

correlation with a functional syndrome may be usefulGrade (specify grading system used)

Mitotic rate (number of mitoses per 10 high-power fields or 2mm2;count 50 high-power fields in the most mitotically active regions,provide separate mitotic rate for each major separate site of disease)

[OPTIONAL: Ki67 labeling index (count multiple regions withhighest labeling density, report average percentage; ‘‘eyeballed’’estimate is adequate)]Presence of nonischemic tumor necrosisPresence of other pathologic componentsResection margins (positive/negative/close) (OPTIONAL measuredistance from margin if within 0.5 cm)Identification of primary site

Immunohistochemistry for CDX2, TTF1

For Biopsy of Metastatic Tumors:

Location of metastasisDiagnosis (functional status need not be included in pathology report)Presence of unusual histologic features (oncocytic, clear cell, gland-forming, etc.)Immunohistochemical staining for general neuroendocrine markers

ChromograninSynaptophysin(OPTIONAL: peptide hormones, IF a specific clinical situation

suggests the correlation with a functional syndrome may be useful)Grade for adequate biopsy specimens; FNA specimens may not beadequate (specify grading system used)

Mitotic rate (number of mitoses per 10 high-power fields or 2mm2;count up to 50 high-power fields)

Ki67 labeling index (count multiple regions with highest labelingdensity, report average percentage; ‘‘eyeballed’’ estimate is adequate)Presence of nonischemic tumor necrosisPresence of other pathologic components (eg, nonneuroendocrinecomponents)Identification of primary site

Immunohistochemistry for CDX2, TTF1



this strategy will assist the more fluid translation of noveltherapeutic strategies from the research arena into routinepractice.

REFERENCES1. Ahmed A, Turner G, King B, et al. Midgut neuroendocrine

tumours with liver metastases. Results of the UKI NETS study.Endocr Relat Cancer. 2009;16:885–894.

2. Anlauf M, Perren A, Kloppel G. Endocrine precursor lesions andmicroadenomas of the duodenum and pancreas with and withoutMEN1: criteria, molecular concepts and clinical significance.Pathobiology. 2007;74:279–284.

3. Anlauf M, Schlenger R, Perren A, et al. Microadenomatosis of theendocrine pancreas in patients with and without the multipleendocrine neoplasia type 1 syndrome. Am J Surg Pathol. 2006;30:560–574.

4. Arnold R. Endocrine tumours of the gastrointestinal tract.Introduction: definition, historical aspects, classification, staging,prognosis and therapeutic options. Best Pract Res Clin Gastro-enterol. 2005;19:491–505.

5. Beasley MB, Thunnissen FB, Brambilla E, et al. Pulmonaryatypical carcinoid: predictors of survival in 106 cases. Hum Pathol.2000;31:1255–1265.

6. Bordi C, D’Adda T, Azzoni C, et al. Criteria for malignancy ingastrointestinal endocrine tumors. Endocr Pathol. 2006;17:119–129.

7. Burke AP, Thomas RM, Elsayed AM, et al. Carcinoids of thejejunum and ileum: an immunohistochemical and clinicopathologicstudy of 167 cases. Cancer. 1997;79:1086–1093.

8. Capella C, Solcia E, Sobin LH, et al. Endocrine tumours of thecolon and rectum. In: Hamilton SR, Aaltonen LA, eds. Pathologyand Genetics of Tumours of the Digestive System. Lyon: IARCPress; 2000:137–139.

9. Capella C, Solcia E, Sobin LH, et al. Endocrine tumours of thesmall intestine. In: Hamilton SR, Aaltonen LA, eds. Pathology andGenetics of Tumours of the Digestive System. Lyon: IARC Press;2000:77–82.

10. Capella C, Solcia E, Sobin LH, et al. Endocrine tumours of thestomach. In: Hamilton SR, Aaltonen LA, eds. Pathology andGenetics of Tumours of the Digestive System. Lyon: IARC Press;2000:53–57.

11. Chatzipantelis P, Konstantinou P, Kaklamanos M, et al. The roleof cytomorphology and proliferative activity in predicting biologicbehavior of pancreatic neuroendocrine tumors: a study byendoscopic ultrasound-guided fine-needle aspiration cytology.Cancer Cytopathol. 2009;117:211–216.

12. Chetty R. Requiem for the term ‘‘carcinoid tumour’’ in thegastrointestinal tract? Can J Gastroenterol. 2008;22:357–358.

13. Cheuk W, Kwan MY, Suster S, et al. Immunostaining for thyroidtranscription factor 1 and cytokeratin 20 aids the distinction ofsmall cell carcinoma from Merkel cell carcinoma, but notpulmonary from extrapulmonary small cell carcinomas. ArchPathol Lab Med. 2001;125:228–231.

14. Cho CS, Labow DM, Tang L, et al. Histologic grade is correlatedwith outcome after resection of hepatic neuroendocrine neoplasms.Cancer. 2008;113:126–134.

15. Couvelard A, Deschamps L, Ravaud P, et al. Heterogeneity oftumor prognostic markers: a reproducibility study applied to livermetastases of pancreatic endocrine tumors. Mod Pathol. 2009;22:273–281.

16. DeLellis RA. The neuroendocrine system and its tumors: anoverview. Am J Clin Pathol. 2001;115(suppl):S5–S16.

17. Deshpande V, Fernandez-del Castillo C, Muzikansky A, et al.Cytokeratin 19 is a powerful predictor of survival in pancreaticendocrine tumors. Am J Surg Pathol. 2004;28:1145–1153.

18. Dhall D, Frishberg DP, Galliano G, et al. Interobserver varia-bility in assessing Ki67 proliferative index in gastrointestinalwell-differentiated neuroendocrine neoplasms. Mod Pathol. 2009;22:116A.

19. Dixon E, Armstrong C, Maddern G, et al. Development of qualityindicators of care for patients undergoing hepatic resection for

metastatic colorectal cancer using a Delphi process. J Surg Res.2009;156:32–38.

20. Drozdov I, Kidd M, Nadler B, et al. Predicting neuroendocrinetumor (carcinoid) neoplasia using gene expression profiling andsupervised machine learning. Cancer. 2009;115:1638–1650.

21. Edge SE, Byrd DR, Carducci MA, et al. AJCC Cancer StagingManual. 7th ed. New York: Springer; 2009. (in press).

22. Fahy BN, Tang LH, Klimstra D, et al. Carcinoid of the rectum riskstratification (CaRRS): a strategy for preoperative outcomeassessment. Ann Surg Oncol. 2007;14:396–404.

23. Fenoglio-Preiser CM. Gastrointestinal neuroendocrine/neuroecto-dermal tumors. Am J Clin Pathol. 2001;115(suppl):S79–S93.

24. Ferrone CR, Tang LH, Tomlinson J, et al. Determining prognosisin patients with pancreatic endocrine neoplasms: can theWHO classification system be simplified? J Clin Oncol. 2007;25:5609–5615.

25. Fink A, Kosecoff J, Chassin M, et al. Consensus methods:characteristics and guidelines for use. Am J Public Health. 1984;74:979–983.

26. Granberg D. Investigational drugs for neuroendocrine tumours.Expert Opin Investig Drugs. 2009;18:601–608.

27. Green FL, Page DL, Fleming ID, et al. AJCC Cancer StagingManual. 6th ed. New York: Springer; 2002.

28. Heitz PU, Komminoth P, Perren A, et al. Pancreatic endocrinetumours: introduction. In: DeLellis RA, Lloyd RV, Heitz PU, et al,eds. Pathology and Genetics of Tumours of Endocrine Organs. Lyon:IARC Press; 2004:177–182.

29. Heymann MF, Joubert M, Nemeth J, et al. Prognostic andimmunohistochemical validation of the capella classification ofpancreatic neuroendocrine tumours: an analysis of 82 sporadiccases. Histopathology. 2000;36:421–432.

30. Hochwald SN, Zee S, Conlon KC, et al. Prognostic factors inpancreatic endocrine neoplasms: an analysis of 136 cases witha proposal for low-grade and intermediate-grade groups. J ClinOncol. 2002;20:2633–2642.

31. Hofer MD, Chang MC, Hirko KA, et al. Immunohistochemicaland clinicopathological correlation of the metastasis-associatedgene 1 (MTA1) expression in benign and malignant pancreaticendocrine tumors. Mod Pathol. 2009;22:933–939.

32. Jaffee IM, Rahmani M, Singhal MG, et al. Expression of theintestinal transcription factor CDX2 in carcinoid tumors is a markerof midgut origin. Arch Pathol Lab Med. 2006;130:1522–1526.

33. Jamali M, Chetty R. Predicting prognosis in gastroentero-pancreatic neuroendocrine tumors: an overview and the value ofki-67 immunostaining. Endocr Pathol. 2008;19:282–288.

34. Janson ET. Somatostatin analogs in the treatment of neuro-endocrine gastroenteropancreatic and intrathoracic tumors.J Endocrinol Invest. 2005;28:137–140.

35. Klimstra D, Perren A, Oberg K, et al. Pancreatic endocrinetumours: non-functioning tumours and microadenomas. In:DeLellis RA, Lloyd RV, Heitz PU, et al, eds. Pathology andGenetics of Tumours of Endocrine Organs. Lyon: IARC Press;2004:201–204.

36. Kloppel G, Anlauf M. Epidemiology, tumour biology andhistopathological classification of neuroendocrine tumours of thegastrointestinal tract. Best Pract Res Clin Gastroenterol. 2005;19:507–517.

37. Kloppel G, Anlauf M, Perren A. Endocrine precursor lesionsof gastroenteropancreatic neuroendocrine tumors. Endocr Pathol.2007;18:150–155.

38. Kloppel G, Perren A, Heitz PU. The gastroenteropancreaticneuroendocrine cell system and its tumors: the WHO classification.Ann N Y Acad Sci. 2004;1014:13–27.

39. Kloppel G, Rindi G, Anlauf M, et al. Site-specific biology andpathology of gastroenteropancreatic neuroendocrine tumors.Virchows Arch. 2007;451(suppl 1):S9–S27.

40. Koura AN, Giacco GG, Curley SA, et al. Carcinoid tumors of therectum: effect of size, histopathology, and surgical treatment onmetastasis free survival. Cancer. 1997;79:1294–1298.

41. Kulke MH. New developments in the treatment of gastrointestinalneuroendocrine tumors. Curr Oncol Rep. 2007;9:177–183.



42. Kulke MH, Hornick JL, Frauenhoffer C, et al. O6-methylguanineDNA methyltransferase deficiency and response to temozolomide-based therapy in patients with neuroendocrine tumors. Clin CancerRes. 2009;15:338–345.

43. Kytola S, Hoog A, Nord B, et al. Comparative genomichybridization identifies loss of 18q22-qter as an early and specificevent in tumorigenesis of midgut carcinoids. Am J Pathol. 2001;158:1803–1808.

44. La Rosa S, Sessa F, Capella C, et al. Prognostic criteria innonfunctioning pancreatic endocrine tumours. Virchows Arch.1996;429:323–333.

45. Landry CS, Brock G, Scoggins CR, et al. A proposed stagingsystem for gastric carcinoid tumors based on an analysis of 1543patients. Ann Surg Oncol. 2009;16:51–60.

46. Landry CS, Woodall C, Scoggins CR, et al. Analysis of 900appendiceal carcinoid tumors for a proposed predictive stagingsystem. Arch Surg. 2008;143:664–670. (discussion 670).

47. Leja J, Essaghir A, Essand M, et al. Novel markers forenterochromaffin cells and gastrointestinal neuroendocrine carci-nomas. Mod Pathol. 2009;22:261–272.

48. Lejeune M, Jaen J, Pons L, et al. Quantification of diversesubcellular immunohistochemical markers with clinicobiologicalrelevancies: validation of a new computer-assisted image analysisprocedure. J Anat. 2008;212:868–878.

49. Lin O, Olgac S, Green I, et al. Immunohistochemical staining ofcytologic smears with MIB-1 helps distinguish low-grade fromhigh-grade neuroendocrine neoplasms. Am J Clin Pathol. 2003;120:209–216.

50. Lloyd RV. Practical markers used in the diagnosis of neuro-endocrine tumors. Endocr Pathol. 2003;14:293–301.

51. Lu J, Xue LY, Lu N, et al. Superficial primary small cell carcinomaof the esophagus: clinicopathological and immunohistochemicalanalysis of 15 cases. Dis Esophagus. In press

52. Modlin IM, Kidd M, Latich I, et al. Current status ofgastrointestinal carcinoids. Gastroenterology. 2005;128:1717–1751.

53. Modlin IM, Kidd M, Drozdov I, et al. Pharmacotherapy ofneuroendocrine cancers. Expert Opin Pharmacother. 2008;9:2617–2626.

54. Modlin IM, Moss SF, Chung DC, et al. Priorities for improvingthe management of gastroenteropancreatic neuroendocrine tumors.J Natl Cancer Inst. 2008;100:1282–1289.

55. Modlin IM, Oberg K, Chung DC, et al. Gastroenteropancreaticneuroendocrine tumours. Lancet Oncol. 2008;9:61–72.

56. Modlin IM, Shapiro MD, Kidd M, et al. Siegfried oberndorfer andthe evolution of carcinoid disease. Arch Surg. 2007;142:187–197.

57. Nassar H, Albores-Saavedra J, Klimstra DS. High-grade neuroen-docrine carcinoma of the ampulla of vater: a clinicopathologic andimmunohistochemical analysis of 14 cases. Am J Surg Pathol. 2005;29:588–594.

58. Nikou GC, Lygidakis NJ, Toubanakis C, et al. Current diagnosisand treatment of gastrointestinal carcinoids in a series of 101patients: the significance of serum chromogranin-A, somatostatinreceptor scintigraphy and somatostatin analogues. Hepatogastro-enterology. 2005;52:731–741.

59. Oberg K. Somatostatin analog octreotide LAR in gastro-entero-pancreatic tumors. Expert Rev Anticancer Ther. 2009;9:557–566.

60. Oberg K, Eriksson B. Nuclear medicine in the detection, stagingand treatment of gastrointestinal carcinoid tumours. Best Pract ResClin Endocrinol Metab. 2005;19:265–276.

61. Oberg K, Jelic S. Neuroendocrine gastroenteropancreatic tumors:ESMO clinical recommendations for diagnosis, treatment andfollow-up. Ann Oncol. 2008;19(suppl 2):ii104–ii105.

62. Panzuto F, Nasoni S, Falconi M, et al. Prognostic factors and survivalin endocrine tumor patients: comparison between gastrointestinal andpancreatic localization. Endocr Relat Cancer. 2005;12:1083–1092.

63. Pape UF, Berndt U, Muller-Nordhorn J, et al. Prognostic factorsof long-term outcome in gastroenteropancreatic neuroendocrinetumours. Endocr Relat Cancer. 2008;15:1083–1097.

64. Pape UF, Jann H, Muller-Nordhorn J, et al. Prognostic relevanceof a novel TNM classification system for upper gastroentero-pancreatic neuroendocrine tumors. Cancer. 2008;113:256–265.

65. Papotti M, Bongiovanni M, Volante M, et al. Expression ofsomatostatin receptor types 1-5 in 81 cases of gastrointestinal andpancreatic endocrine tumors. A correlative immunohistochemicaland reverse-transcriptase polymerase chain reaction analysis.Virchows Arch. 2002;440:461–475.

66. Pelosi G, Bresaola E, Bogina G, et al. Endocrine tumors of thepancreas: Ki-67 immunoreactivity on paraffin sections is anindependent predictor for malignancy: a comparative study withproliferating-cell nuclear antigen and progesterone receptor proteinimmunostaining, mitotic index, and other clinicopathologic vari-ables. Hum Pathol. 1996;27:1124–1134.

67. Pelosi G, Rodriguez J, Viale G, et al. Typical and atypicalpulmonary carcinoid tumor overdiagnosed as small-cell carcinomaon biopsy specimens: a major pitfall in the management of lungcancer patients. Am J Surg Pathol. 2005;29:179–187.

68. Perigny M, Hammel P, Corcos O, et al. Pancreatic endocrinemicroadenomatosis in patients with von Hippel-Lindau disease:characterization by VHL/HIF pathway proteins expression. AmJ Surg Pathol. 2009;33:739–748.

69. Perren A, Roth J, Muletta-Feurer S, et al. Clonal analysis ofsporadic pancreatic endocrine tumours. J Pathol. 1998;186:363–371.

70. Phan AT, Yao JC, Evans DB. Treatment options for metastaticneuroendocrine tumors. Surgery. 2008;144:895–898.

71. Plockinger U, Rindi G, Arnold R, et al. Guidelines for thediagnosis and treatment of neuroendocrine gastrointestinaltumours. A consensus statement on behalf of the Europeanneuroendocrine tumour society (ENETS). Neuroendocrinology.2004;80:394–424.

72. Plockinger U, Wiedenmann B. Neuroendocrine tumors. Biother-apy. Best Pract Res Clin Endocrinol Metab. 2007;21:145–162.

73. Reidy DL, Tang LH, Saltz LB. Treatment of advanced diseasein patients with well-differentiated neuroendocrine tumors. NatClin Pract Oncol. 2009;6:143–152.

74. Rigaud G, Missiaglia E, Moore PS, et al. High resolutionallelotype of nonfunctional pancreatic endocrine tumors: identifi-cation of two molecular subgroups with clinical implications.Cancer Res. 2001;61:285–292.

75. Rindi G, Kloppel G, Alhman H, et al. TNM staging of foregut(neuro)endocrine tumors: a consensus proposal including a gradingsystem. Virchows Arch. 2006;449:395–401.

76. Rindi G, Kloppel G, Couvelard A, et al. TNM staging of midgutand hindgut (neuro) endocrine tumors: a consensus proposalincluding a grading system. Virchows Arch. 2007;451:757–762.

77. Rowe G, Wright G. The Delphi technique as a forecasting tool:issues and analysis. Intl J Forecasting. 1999;15:353–375.

78. Saqi A, Alexis D, Remotti F, et al. Usefulness of CDX2 and TTF-1in differentiating gastrointestinal from pulmonary carcinoids. AmJ Clin Pathol. 2005;123:394–404.

79. Schmitt AM, Riniker F, Anlauf M, et al. Islet 1 (Isl1) expression isa reliable marker for pancreatic endocrine tumors and theirmetastases. Am J Surg Pathol. 2008;32:420–425.

80. Sen N, Calli Demirkan N, Aksoy Altinboga A, et al. Synchronousendocrine tumors of small intestine: report of a case. TurkJ Gastroenterol. 2008;19:193–196.

81. Shia J, Tang LH, Weiser MR, et al. Is nonsmall cell type high-grade neuroendocrine carcinoma of the tubular gastrointestinaltract a distinct disease entity? Am J Surg Pathol. 2008;32:719–731.

82. Soga J. The term ‘‘carcinoid’’ is a misnomer: the evidence based onlocal invasion. J Exp Clin Cancer Res. 2009;28:15.

83. Solcia E, Fiocca R, Villani L, et al. Hyperplastic, dysplastic, andneoplastic enterochromaffin-like-cell proliferations of the gastricmucosa. Classification and histogenesis. Am J Surg Pathol. 1995;19(suppl 1):S1–S7.

84. Srivastava A, Hornick JL. Immunohistochemical staining forCDX-2, PDX-1, NESP-55, and TTF-1 can help distinguishgastrointestinal carcinoid tumors from pancreatic endocrine andpulmonary carcinoid tumors. Am J Surg Pathol. 2009;33:626–632.

85. Suster S, Moran CA. Neuroendocrine neoplasms of the media-stinum. Am J Clin Pathol. 2001;115(suppl):S17–S27.



86. Tang L, Shia J, Vakiani E, et al. High grade transformation ofdifferentiated neuroendocrine neoplasms (NENs) of the entero-pancreatic system-a unique entity distinct from de novo high gradeneuroendocrine carcinoma (HGNECa) in pathogenesis and clinicalbehavior. Mod Pathol. 2008;21:137A.

87. Travis WD. The concept of pulmonary neuroendocrine tumours.In: Travis WD, Brambilla E, Muller-Hermelink HK, et al, eds.Pathology and Genetics of Tumours of the Lung, Pleura, Thymusand Heart. Lyon: IARC Press; 2004:19–20.

88. Travis WD, Rush W, Flieder DB, et al. Survival analysis of 200pulmonary neuroendocrine tumors with clarification of criteriafor atypical carcinoid and its separation from typical carcinoid.Am J Surg Pathol. 1998;22:934–944.

89. Van Essen M, Krenning EP, De Jong M, et al. Peptide receptorradionuclide therapy with radiolabelled somatostatin analogues inpatients with somatostatin receptor positive tumours. Acta Oncol.2007;46:723–734.

90. Vilar E, Salazar R, Perez-Garcia J, et al. Chemotherapy and role ofthe proliferation marker Ki-67 in digestive neuroendocrine tumors.Endocr Relat Cancer. 2007;14:221–232.

91. Volante M, Brizzi MP, Faggiano A, et al. Somatostatin receptortype 2A immunohistochemistry in neuroendocrine tumors: aproposal of scoring system correlated with somatostatin receptorscintigraphy. Mod Pathol. 2007;20:1172–1182.

92. Wang W, Epstein JI. Small cell carcinoma of the prostate.A morphologic and immunohistochemical study of 95 cases.Am J Surg Pathol. 2008;32:65–71.

93. Wick MR, Graeme-Cook FM. Pancreatic neuroendocrine neo-plasms: a current summary of diagnostic, prognostic, anddifferential diagnostic information. Am J Clin Pathol. 2001;115(suppl):S28–S45.

94. Yao JC. Neuroendocrine tumors. Molecular targeted therapy forcarcinoid and islet-cell carcinoma. Best Pract Res Clin EndocrinolMetab. 2007;21:163–172.

95. Yao JC, Hassan M, Phan A, et al. One hundred years after‘‘carcinoid’’: epidemiology of and prognostic factors for neuro-endocrine tumors in 35,825 cases in the United States. J Clin Oncol.2008;26:3063–3072.

96. Yao JC, Hoff PM. Molecular targeted therapy for neuroendocrinetumors. Hematol Oncol Clin North Am. 2007;21:575–581; x.

97. Yao JC, Phan AT, Chang DZ, et al. Efficacy of RAD001(everolimus) and octreotide LAR in advanced low- to intermedi-ate-grade neuroendocrine tumors: results of a phase II study. J ClinOncol. 2008;26:4311–4318.

98. Zee SY, Hochwald SN, Conlon KC, et al. Pleomorphic pancreaticendocrine neoplasms: a variant commonly confused with adeno-carcinoma. Am J Surg Pathol. 2005;29:1194–1200.

99. Zhang L, Smyrk TC, Oliveira AM, et al. KIT is an independentprognostic marker for pancreatic endocrine tumors: a findingderived from analysis of Islet cell differentiation markers. AmJ Surg Pathol. 2009. In press

100. Zhao J, Moch H, Scheidweiler AF, et al. Genomic imbalances inthe progression of endocrine pancreatic tumors. Genes Chromo-somes Cancer. 2001;32:364–372.



pathology reporting of neuroendocrine tumors

Documents

pathology of nets

pathology zepidemiology

minimal pathology data

fundamental pathology

ga ydepartment of pathology

wi zzzdepartment of

ct zdepartment of pathology

surgical pathology service