tumor budding in colorectal carcinoma: time to take notice

Tumor budding in colorectal carcinoma:time to take notice

Bojana Mitrovic, David F Schaeffer, Robert H Riddell and Richard Kirsch

Department of Pathology and Laboratory Medicine, Mount Sinai Hospital, Toronto, Ontario, Canada

Tumor ‘budding’, loosely defined by the presence of individual cells and small clusters of tumor cells at the

invasive front of carcinomas, has received much recent attention, particularly in the setting of colorectal

carcinoma. It has been postulated to represent an epithelial–mesenchymal transition. Tumor budding is a well-

established independent adverse prognostic factor in colorectal carcinoma that may allow for stratification of

patients into risk categories more meaningful than those defined by TNM staging, and also potentially guide

treatment decisions, especially in T1 and T3 N0 (Stage II, Dukes’ B) colorectal carcinoma. Unfortunately, its

universal acceptance as a reportable factor has been held back by a lack of definitional uniformity with respect

to both qualitative and quantitative aspects of tumor budding. The purpose of this review is fourfold: (1) to

describe the morphology of tumor budding and its relationship to other potentially important features of the

invasive front; (2) to summarize current knowledge regarding the prognostic significance and potential clinical

implications of this histomorphological feature; (3) to highlight the challenges posed by a lack of data to allow

standardization with respect to the qualitative and quantitative criteria used to define budding; and (4) to

present a practical approach to the assessment of tumor budding in everyday practice.Modern Pathology (2012) 25, 1315–1325; doi:10.1038/modpathol.2012.94; published online 13 July 2012

Keywords: colorectal carcinoma; epithelial–mesenchymal transition; prognostic factors; tumor budding

Colorectal carcinoma is one of the commonesthuman cancers, and one of the leading causesof cancer-related death.1 Prognosis and treatmentdecisions are based primarily on the extent of thedisease, as codified by the TNM staging system.Unfortunately, a substantial number of tumorsbehave poorly despite being categorized as low riskbased on their TNM stage.2 Thus, the search foradditional prognostic factors in the assessment ofcolorectal carcinoma has been a major researchfocus. Of the histopathological factors studied todate, the most promising include extramural venousinvasion, the nature of the advancing front (pushingvs infiltrative), an inflammatory infiltrate, micro-satellite instability, and tumor budding—the pre-sence of small discrete clusters of tumor cells at theinvasive edge (Figure 1). There is now overwhelm-ing evidence that tumor budding is an independentprognostic factor in colorectal carcinoma, particu-

larly in node-negative disease. Thus, its assessmenthas the potential to enhance prognostic accuracyand influence treatment algorithms. When exam-ined carefully, the majority of colorectal carcinomasdisplay some degree of budding; hence, attemptshave been made at developing scoring systemsto identify a prognostically significant degree ofbudding, commonly termed ‘high-grade’ budding.Definitions of high-grade budding vary substantiallyamong different authors and even among differentstudies by the same authors.

Although tumor budding only entered the main-stream pathology literature in the past decade, itwas first described in the 1950s by Imai,3 whopostulated that the presence of ‘sprouting’ at theinvasive edge of carcinomas reflected a more rapidtumor growth rate (or ‘Schub’ phase of growth).Budding has been described in association withcarcinomas at a number of different sites, but ithas been most extensively studied in colorectalcarcinoma.

The biology of tumor budding is not understoodbut the prevailing theory is that at least some typesof budding represent an example of epithelial–mesenchymal transition, a process thought to occurphysiologically during embryological development,

Received 15 December 2011; revised 30 March 2012; accepted 1April 2012; published online 13 July 2012

Correspondence: Dr B Mitrovic, MD, Department of Pathologyand Laboratory Medicine, Mount Sinai Hospital, 600 UniversityAvenue, Toronto, Ontario, Canada M5G 1X5.E-mail: [email protected]

Modern Pathology (2012) 25, 1315–1325

& 2012 USCAP, Inc. All rights reserved 0893-3952/12 $32.00 1315

www.modernpathology.org

http://dx.doi.org/10.1038/modpathol.2012.94

mailto:[email protected]

http://www.modernpathology.org

and pathologically during fibrosis and tumor inva-sion.4,5 The epithelial–mesenchymal transition ischaracterized by loss of cell adhesion molecules,cytoskeletal alterations, increased production ofextracellular matrix components, resistance to apop-tosis, and ability to degrade basement membrane,resulting in a phenotype with increased migratorycapacity and invasiveness.6,7 Current theoriesrelating to the epithelial–mesenchymal transitionand tumor budding have been comprehensivelyreviewed elsewhere, most recently by Zlobec et al,8

and will not be discussed here.The purpose of this review is to examine

practical issues related to tumor budding thatmay be of interest to the practicing pathologist.We will discuss the morphological featuresand prognostic significance of tumor buddingand summarize the various scoring systems for itsassessment in order to highlight problems relatedto definitional heterogeneity in tumor-buddingstudies.

Finally, we will present a practical approachto the assessment of tumor budding. Althoughperforming a tumor bud count is easy in theory,deciding what is or is not a bud can provesurprisingly difficult in practice. Buds are oftenobscured in the setting of a prominent inflammatoryreaction; fibroblasts and stromal cells may occasion-ally be mistaken for buds; and single-cell buds areusually inconspicuous and often missed on H&Eevaluation (Figure 2).

Morphological features

Although tumor budding (‘sprouting’) was origin-ally described by Imai in the 1950s, the first detaileddescription of tumor buds in the English languageliterature was put forth by Gabbert et al9 who usedlight and electron microscopy to characterize whatthey termed ‘tumor dedifferentiation’ at the invasiveedge of colorectal carcinomas. They identified a

Figure 1 Tumor budding in colorectal carcinoma (a) Illustration of a colorectal carcinoma with an infiltrative border but no budding.Tumor buds, defined as individual cells or small clusters (o5 cells) of tumor cells at the invasive front (arrows), are illustrated in theseexamples from (b) a malignant polyp and (c and d) a colorectal cancer resection specimen. Note the blurring of the tumor–stromainterface (c), which corresponds with tumor budding at higher magnification (d). (Original magnification: a—�100; b—�200; c—� 20;d—�400).

Tumor budding

1316 B Mitrovic et al


subset of colorectal carcinomas with invasivefronts that displayed a strikingly different architec-ture compared with the central area of the tumor,

with well-developed glandular structures givingway to isolated tumor cells and small cell groups,as illustrated in Figure 1. At the ultrastructural level,

Figure 2 ‘Challenging scenarios’. (a) Peritumoral inflammatory cells, including histiocytes, can be difficult to differentiate from tumorbuds, and may sometimes obscure the underlying budding. (b) Immunohistochemistry for anti-cytokeratin may help to highlight tumorbuds in this setting. (c) Given that tumor budding may represent an epithelial–mesenchymal transformation, their separation fromstromal cells can also be challenging and (d) anti-cytokeratin stains may be very useful in identifying tumor buds. (e) Fragmentation ofneoplastic glandular structures (arrow) can sometimes give the impression of extensive tumor budding on H&E and onimmunohistochemistry with (f) anti-cytokeratin. In this example, the small fragments/individual tumor cells highlighted by anti-cytokeratin in the region of the fragmented gland (arrow between fragments) may be a consequence of fragmentation rather than truebudding. (Original magnification: a—�40; b—� 40; c—�100; d—� 100; e—�200; f—� 200).

Tumor budding

B Mitrovic et al 1317


these small clusters and isolated tumor cellsappeared poorly differentiated, with large nucleiand a loss of microvilli and basement membrane,and poorly developed or absent junctional com-plexes and desmosomes.

When seen in a single section, tumor buds appearas clusters of cells that have broken off from themain tumor mass. By examining serial sections ofhigh budding tumors stained with anti-cytokeratinantibodies, Prall et al10 demonstrated that most budsthat appear to represent discrete clusters of cells arein fact connected to adjacent larger glands. Indeed,Morodomi et al11 coined the term ‘budding’ becausethe undifferentiated single cells and tubular tumornests that they counted as buds both appeared to bebudding from larger neoplastic glands.

Shinto et al12,13 used immunohistochemistry withanti-cytokeratin antibodies to highlight anotherfeature of buds, termed cytoplasmic pseudofragmen-tation, whereby cytoplasmic fragments, visible onlyby immunohistochemistry, are seen in the immedi-ate vicinity of tumor buds. When examined on serialsections, some of the fragments were shown to beconnected to the buds (hence the term ‘pseudofrag-ments’). The presence of pseudofragments is asso-ciated with high-grade budding, but was shown tobe an independent prognostic factor on multivariateanalysis, suggesting that its presence signifies anaggressive budding phenotype.

Although tumor buds are usually most prominentat the invasive front, intratumoral budding has beendescribed as well. Only one study has attempted toassess the significance of this finding: Lugli et al14

reported that the presence of intratumoral buddingwas strongly correlated with peritumoral budding,but was found to be an independent prognosticfactor on multivariate analysis.

There is a strong association between budding andthe presence of lymph node metastases and lym-phovascular invasion,11,12,15–32 defined by the pre-sence of tumor cells within an endothelium-linedspace, and it has been suggested that buds representthe part of the tumor that has gained the ability toinvade lymphatics and vascular channels. This ideais supported by two intriguing morphologicalstudies: Morodomi et al33 examined serial sectionsof high-budding areas to demonstrate that buddingnests are often found adjacent to areas of lympho-vascular space invasion, and, in a more recentstudy, Ohtsuki et al31 performed double stainingfor anti-cytokeratin antibodies and anti-lymphaticantibodies, finding that a number of ‘buds’ at theinvasive edge of a tumor are in fact located in smalllymphatic spaces. Similarly, the presence of bud-ding has been associated with increased risk ofdistant metastases,34–36 suggesting that budding mayalso be associated with vascular invasion. A fewtumor-budding studies have used vascular mar-kers and/or elastic stains to assess vascular inva-sion,11,12,19,22,26,32,37,38 but only four have analyzedthe relationship between budding and vascular

invasion: Kazama et al22 found no relationshipbetween budding and vascular invasion, whereasthree other studies have reported a statisticallysignificant correlation between budding and venousinvasion, though the association was not as pro-nounced as the relationship with lymphatic inva-sion and lymph node metastases.11,12,32

The tumor–host interaction at the invasive frontmay be of prognostic importance in the setting oftumor budding. Several studies have shown peri-tumoral lymphocytic infiltration to be an indepen-dent prognostic factor in colorectal carcinoma.Lugli et al23 have demonstrated that a peritumorallymphocytic reaction is associated with improvedprognosis in the setting of tumor budding, suggest-ing that the immune response might target the tumorbuds. The nature of the stroma at the invasivemargin may provide further prognostic information,with myxoid stroma being associated with worseprognosis,20,39 but further studies are needed toconfirm the reproducibility and prognostic impor-tance of these findings.

Clinical significance of tumor budding

Tumor budding is associated with other histopatho-logical factors known to portend a worse prognosis,namely higher tumor grade, infiltrating tumor border,the presence of lymphovascular and perineuralinvasion, and lymph node metastases.11,12,15–32 Onmultivariate analysis, budding has consistentlyemerged as an independent adverse prognostic factor,associated with local tumor recurrence and distantmetastases, and significantly worse overall anddisease-free survival.12,15,16,21,27,28,31,32,34–36,38,40,41

The adverse prognostic impact of high-gradetumor budding is seen in both early and advancedcolorectal carcinoma, and there are several scenariosin which this feature might influence clinicaldecision making, particularly in early colorectalcarcinoma.

Tumor Budding in Early Colorectal Carcinoma

Submucosally invasive colorectal carcinoma isassociated with excellent outcomes and low ratesof lymph node metastases, and a subset of patientscan be successfully managed with endoscopicmucosal resection or polypectomy,42 thus avoidingthe risks associated with segmental resection.Identification of candidates for endoscopic resectionrests mainly on the absence of certain high-riskhistopathological features, namely high tumorgrade, lymphovascular invasion, and tumor bud-ding.26,42 Tumor extending to the cauterized margincan often be managed by endoscopic follow up.

Several large studies have shown tumor buddingin submucosally invasive colorectal carcinoma tobe an independent prognostic factor associatedwith lymph node metastases, local recurrence, and

Tumor budding



cancer-related death.16,19,22,25,26,29,30,37,43,44 In thelargest of these studies, Ueno et al26 demonstratedthat, in submucosally invasive colorectal carcinoma,high tumor grade, lymphovascular invasion, andtumor budding are the three factors independentlyassociated with lymph node metastases. Patientswithout any of these three features showed excep-tionally low rates of lymph node metastases (1%,1/138); in the presence of one risk factor, the rate ofnodal metastases increased substantially to 21%(12/58), and when two or three factors were present,the risk was 36% (20/55), suggesting that in theabsence of these factors, polypectomy alone issufficient treatment for early colorectal carcinomaprovided the resection margins are clear. In thisstudy, Ueno et al further showed that in the absenceof extensive submucosal invasion (defined asr4mm wide and r2mm deep), the risk of nodalmetastases (including isolated tumor cells, identifi-able only with anti-cytokeratin antibodies) intumors lacking the three above-mentioned featureswas 0; thus, extent of submucosal invasion mayneed to be included in the histopathological assess-ment of T1 tumors if strict criteria are to be appliedfor the identification of patients that can safely bemanaged without surgical resection.

Tumor Budding in Stage II (T3–4 N0) ColorectalCarcinoma

As patients with Stage II colorectal carcinomahave highly variable outcomes, tumor buddingmay be particularly useful in identifying high-risksubgroups within this population. In one of theearlier studies of tumor budding, Hase et al18

demonstrated that 5-year survival rates of Dukes B(stage II, T3–4 N0) patients with high-grade buddingare significantly worse than those of Dukes C (Nþ )patients without budding (29 percent vs 66 percent;Po0.001). A number of more recent studies haveconfirmed that patients with Stage II colorectalcarcinoma do significantly worse when high-gradebudding is present,34–36,38,41,45 and several studieshave shown that survival rates of patients with StageII colorectal carcinoma with high-grade budding areequivalent to survival rates of patients with Stage IIIcolorectal carcinoma.34–36 In their studies of Stage IIand III pT3 tumors, Okuyama et al34,35 found thattumor budding was the only factor on multivariateanalysis to be associated with decreased survivaland was more prognostically significant than lymphnode metastases.

Currently, the indications for chemotherapy inStage II colorectal carcinoma are unclear, and adju-vant therapy is considered only in subgroups withadverse prognostic features, eg, high-histologicalgrade, pT4 disease, and extramural venous invasion.The QUASAR trial demonstrated that chemotherapydoes result in increased survival in Stage II patients,but the improvement was small and may not justify

the costs and side effects of chemotherapy in allpatients.46 The significantly worse outcome experi-enced by Stage II patients with tumor buddinghas prompted some authors to suggest that adju-vant chemotherapy should be considered in thesepatients,18,34–36,41 but there are no published dataassessing the effectiveness of chemotherapy in StageII colorectal carcinoma with tumor budding.

Tumor Budding in Stage III (T1–4 Nþ ) ColorectalCarcinoma

Data regarding budding in Stage III colorectalcarcinoma are limited, and the two studies reportingon the prognostic significance of budding in thissubgroup of patients have produced conflictingresults. Choi et al15 studied tumor budding in 103patients with Stage III rectal carcinoma; buddingwas an independent poor prognostic factor: whenadded to N stage as a prognostic variable, there wasbetter prognostic stratification with respect to 5-yeardisease-free survival compared with the AmericalJoint Comittee on Cancer nodal staging alone. On theother hand, in a series of 477 Stage III colorectalcarcinomas reported on by Sy et al,47 although therewas a survival difference between high- and low-budding groups, on multivariate analysis buddingwas not found to be an independent prognosticfactor. Although more studies are needed to definethe prognostic significance of tumor budding in thisgroup, it is not likely that the reporting of this factorwill impact the clinical management of thesepatients.

Tumor Budding in Metastatic Colorectal Carcinoma

The significance of tumor budding in metastaticcolorectal carcinoma has received little attention.However, one study showed that the presence oftumor budding predicts poor response to anti-EGFRtherapy in patients with metastatic colorectal carci-noma: Zlobec et al48 studied a series of 43 patientswith metastatic colorectal carcinoma treated withcetuximab or penitumumab, and found all patientswith a KRAS mutation (n¼ 7) and/or high-gradetumor budding (n¼ 11, including 4 with KRASmutation) to be nonresponsive to anti-EGFR therapy.Although this finding needs to be confirmed inlarger cohorts, it raises the interesting possibilitythat tumor budding in addition to K-RAS analysismay more accurately determine eligibility for anti-EGFR therapy.

Tumor Budding in Microsatellite Unstable Tumors

The association between tumor budding and poorprognosis in sporadic microsatellite stable colorectalcarcinoma is well established, but whether this canbe extrapolated to sporadic and Lynch syndrome-related high frequency microsatellite unstable

Tumor budding



colorectal carcinoma is unclear. Many studies oftumor budding have excluded Lynch syndrome-related cases, while others have not attempted toanalyze the groups separately. Only a few studieshave compared the frequency of tumor budding inthese different types of colorectal carcinomas: inthese studies, the frequency of tumor budding wasapproximately 50% in sporadic microsatellite stableand low-frequency microsatellite unstable tumors;lower rates of tumor budding were seen in Lynchsyndrome-associated tumors (B20%); whereastumor budding was found to be virtually absentin sporadic high-frequency microsatellite unstablecolorectal carcinoma.23,49,50 It has been postulatedthat the relative infrequency of budding in micro-satellite unstable tumors may, at least in part,explain the relatively better prognosis of thesetumors.49 Interestingly, compared with mismatchrepair-proficient tumors, mismatch repair-deficienttumors have been shown to display less cytoplasmicpseudofragmentation in the context of high-gradebudding,51 and to be associated with less intratu-moral budding,14 both of which are associated with aworse prognosis. Given the differing biologicalpathways of microsatellite stable, sporadic high-frequency microsatellite unstable and Lynch syn-drome-related tumors, it seems likely that thefrequency and clinical significance of tumor buddingwill vary significantly among these three groups, butawaits confirmation in controlled studies.

Scoring systems

Although a large number of studies have showntumor budding to be an independent adverseprognostic factor in colorectal carcinoma, studymethodologies have varied widely. Different authorshave used different scoring systems as well asdiffering definitions as to what constitutes a bud.The most widely cited scoring system, developed byUeno et al,27 defines ‘high-grade budding’ as 410buds composed of fewer than 5 cells in a 25� field(field area¼ 0.385mm2). Other authors have definedbuds as being composed of r5 cells,38 while somehave not set an upper cell limit and instead countboth single cells and bundles of five or more cells asbuds, as proposed by Morodomi et al11 andcommonly applied in the Japanese literature. Thisheterogeneity in defining budding is compoundedby the use of varying field diameters and cutoffs fordefining ‘high-grade’ budding, especially when ascore is given as #/hpf rather than/mm2. Althoughmost authors have used Ueno’s definition of ‘high-grade budding’ (410 buds of fewer than 5 cells),all have used a 20� objective, rather than the25� objective used by Ueno et al, resulting insignificantly greater field areas than that originallyreported by Ueno. Furthermore, while the majorityof studies have evaluated only areas with thegreatest amount of budding, others have proposed

calculating an average bud count,11 counting thenumber of buds along the entire invasivefront19,24,52,53 or calculating the proportion of inva-sive front with tumor budding.21,36,40 Such metho-dological heterogeneity is illustrated in Table 1,which summarizes the bud counting methods usedin a sample of tumor-budding studies.

Establishing a Prognostically Significant Cutoff for‘High-Grade’ Budding

Although Ueno’s definition of ‘high-grade’ budding(ie, 10 buds in a 25� field) is the most widelyapplied in the literature, there is no evidence thatthis is the optimal cutoff for evaluating budding.Only a handful of studies have attempted to analyzetheir data so as to generate a clinically usefulthreshold for the assessment of high-grade budding.The results and methodologies of these studies aresummarized in Table 2. In addition, Masaki et al24,53

have proposed a formula based on the total numberof buds counted along the entire invasive margin topredict the probability of lymph node metastasesin T1 and T2 colorectal carcinoma: although thisis an interesting idea, it is not practical for use ineveryday practice.

Subjectivity in the Qualitative Assessment of TumourBudding

In addition to the quantitative issues of inconsistentcutoffs and differing field diameters, there are anumber of qualitative considerations contributing tosubjectivity in defining a tumour bud. Althoughonly a few authors have specifically commented onthis problem, the identification of tumor buds canprove very difficult in several circumstances, eg, (1)stromal cells or histiocytes masquerading as buds,(2) marked inflammation obscuring buds, (3) diffi-culties in determining whether a small cluster ofcells represents a true bud or mechanical fragmenta-tion of a larger gland, etc (Figure 2). Most studies oftumor budding have not provided much detailregarding qualitative criteria used to include orexclude a potential ‘bud’. In addition, there appearto be discrepancies in the qualitative assessment ofbudding, even among studies ostensibly using thesame objective criteria of clusters of o5 cells: eg,while Jass et al49 counted only ‘discrete clusters’ ofcells as buds, Ha et al17 defined buds as clusters‘appearing to bud from a larger gland’, suggestingthat they may have also counted buds that were notcompletely separate from adjacent glands. Althoughsuch variability may be reduced to some degree byestablishing consensus criteria, there will likelyremain a subset of cases in which budding will bedifficult to evaluate owing to the resemblance ofbuds to surrounding stromal cells and/or theconcealment of buds in the setting of a prominentperitumoral inflammatory reaction (Figure 2).

Tumor budding



Table 1 Summary of scoring systems for tumor budding evaluation in a selected sample of relevant studies

First author Patient population Evaluationmethod

Bud definition Cutoff for ‘high-grade’ budding Proportion of cases classified as highgrade

Morodomi11 40 advanced rectalcarcinomas

H&E Isolated cancer cellsand bundles of 5cells

Count performed at four locations(1.25mm2 field area), and average takenThree categories:0–4, negative;4–14, mildly positive;414, strongly positive

Mildly positive, 37.5%;Strongly positive, 27.5%

Hase18 663 colorectal carcinomas H&E Not specified(subjective evaluation)

N/A: classified according to subjectiveimpression

26%

Ueno27 638 rectal carcinomas H&E o5 cells 10 buds in 25� field (0.385mm2) 30.1%

Okuyama35 196 pT3 well to moderatelydifferentiated colorectalcarcinomas

H&E Up to 5 cells N/A: classified according to subjectiveimpression

43.3%

Jass49 95 colorectal carcinomas,stratified according to MSIstatus

H&E Up to 5 cells 5 buds in 40� field (area not specified) 32%

Guzinska-Ustymowicz16 24 T1 rectal carcinomas H&E Up to 5 cells Any budding considered positive 35%

Ha17 90 well to moderatelydifferentiated rectalcarcinomas

H&E o5cells 47 buds in 20� field (area not specified)Cutoff selected based on mean intensityof budding (7.5)

N/A (mean bud count selected ascutoff such that 50% of cases wouldbe classified as ‘high grade’)

Kanazawa22 159 colorectal carcinomas H&E ‘Small cluster’of cells

Divided into three groups based onproportion of invasive front with budding:0–1/3, mild;1/3–2/3, moderate;42/3, marked

29% marked37% moderate

Losi44 295 Stage I colorectalcarcinomas

Immunohisto-chemistry

o5 cells 45 buds in 20� field (area not specified) 6%

Ishikawa43 71 submucosally invasivecolorectal carcinomas



Yamauchi29 164 T1 colorectalcarcinomas



Wang41 128 pT3 N0 colorectalcarcinomas


o5 cells Budding was counted in 5 fields(20� 0.95mm2); a median countof 1 buds considered positive

45%

Tumorbudding

BMitro

vicetal

1321

ModernPathology(2012)25,1315

–1325

Role of Immunohistochemistry in the Evaluation ofBudding

Some authors have attempted to eliminate the sub-jectivity inherent in bud counting by using immuno-histochemistry with anti-cytokeratin antibodies tohighlight buds and distinguish them from surroundingstromal cells.12,13,22,23,31,32,38,40,43,44,48,51,54,55 The use ofimmunohistochemistry in the assessment of buddingis somewhat controversial: some authors argue that theevaluation of budding should be limited to H&E-stained slides because of the cost and impracticality ofperforming immunohistochemistry in routine cases,while others argue that immunohistochemistry shouldbe used routinely to improve the accuracy andreproducibility of bud counts. In addition to costconsiderations, the use of immunohistochemistry mayalso require the establishment of a separate cutoff fordefining positive budding. One study comparing thebud counts obtained on H&E to those obtained withimmunohistochemistry, not surprisingly, found signif-icantly higher counts with the latter.31 Indeed, the fewstudies that have attempted to define an optimalbudding threshold on anti-cytokeratin-stained slideshave reported higher cutoffs than those found to beprognostically significant on H&E (summarized inTable 2).

Currently, the role for immunohistochemistry in theevaluation of tumor budding is unclear, and furtherstudies are needed to assess the relationship betweenbud counts obtained on H&E vs those obtained withanti-cytokeratin antibodies, and to determine the mostprognostically useful cutoff for bud counting with bothH&E and anti-cytokeratin antibodies.

Interobserver Variability in the Assessment of TumorBudding

In light of these myriad sources of subjectivity,one would expect the evaluation of budding to

suffer from suboptimal interobserver variability.A few studies on tumor budding have incorporatedan assessment of interobserver variability, withreported kappa values ranging from 0.41 to0.938.15,27,32,34,38,41,48 Kappa values, a commonly usedmeasure of interobserver agreement, can be interpretedas follows: o0.20, poor; 0.21–0.40, fair; 0.41–0.60,moderate; 0.61–0.80, good; and 40.80, very good.56

Intuitively, one would expect the use ofimmunohistochemistry to improve interobservervariability, but studies assessing interobserver varia-bility on immunohistochemically stained slideshave reported kappa values ranging from 0.53 to0.874,32,38,48 similar to those reported by authorsassessing tumour budding on H&E.15,27,32,34,41 Onlyone study has directly compared interobservervariability between assessments performed on H&Eand immunohistochemically stained slides amongthe same group of pathologists: when Suzuki et al32

evaluated budding with both H&E and immuno-histochemistry, they showed only a modestimprovement in interobserver variability whenanti-cytokeratin antibodies were used (k¼ 0.41 withH&E, and k¼ 0.53 with immunohistochemistry).

Practical considerations in the assess-ment and reporting of tumor budding

Opinions are divided as to whether tumor buddingshould be recorded in the pathological assessmentof colorectal carcinoma. Although the College ofAmerican Pathologists has not included budding inits checklist of reportable features, both the Associa-tion of Directors of Anatomic and Surgical Pathologyand the Union for International Cancer Controlrecommend its inclusion in routine reporting.57,58

Our institutional policy is to report the presenceor absence of tumor budding in all malignantpolyps and colorectal cancer resection specimens.

Table 2 Summary of studies proposing evidence-based cutoffs for ‘high-grade’ budding

Firstauthor

Method ofevaluation

Buddefinition

Method for determination of optimal cutoff Resulting definition of‘high-grade’ budding

Choi15 H&E o5 cells Patients were divided semiquantitatively into four budding groupsbased on quartiles. The cutoff considered the best discriminator ofdisease-free survival was between groups 3 and 4.

410 buds in 20� field(area not specified)

Sy47 H&E o5 cells Patients were divided arbitrarily into five budding groups.As there were no statistically significant differences amongthe first four groups, while the highest budding group showedsignificantly worse survival, the cutoff between groups 4 and 5was deemed the most prognostically significant.

49 buds in 20� field(0.95mm2)

Prall38 Immunohisto-chemistry

Up to5 cells

ROC curves were constructed to define the budding cutoff withthe greatest sensitivity and specificity for predicting thedevelopment of metastases.

25 buds in 25� field(0.785mm2)

Lugli23 Immunohisto-chemistry

Up to5 cells

ROC curves were constructed to define the budding cutoff withthe greatest sensitivity and specificity for discriminatingbetween survivors and nonsurvivors.


Zlobec48 Immunohisto-chemistry

o5 cells ROC curves were constructed to define the budding cutoffwith the best predictive ability for anti-EGFR therapynonresponse in a population of metastatic CRC.


Tumor budding



Although we recognize the limitations posed by alack of definitional uniformity, we feel that itsassessment is important both for the purposes ofdata gathering, and, in the setting of malignantpolyps, for guiding treatment decisions.

Until universally accepted criteria are established,we apply those of Ueno et al which are not only ofproven prognostic significance, but are also easilyapplied and are the most widely used criteria in thebudding literature. We report tumor budding aspresent if Z10 groups of o5 cells are counted ina 20� objective field (ie, Ueno’s so-called ‘high-grade budding’).27 In borderline cases, where only5–10 definite buds/20� field are identified but thereare additional possible tumor cells that cannot beconfidently differentiated from stroma or histio-cytes, a cytokeratin stain is performed. If thisconfirms the impression of additional tumor cells,bringing the count to Z10 buds, we report aspositive for tumor budding. However, we cautionagainst the routine use of cytokeratin stains in caseswhere bud counts on H&E do not approach 10/20� objective. In our experience, counts by cyto-keratin immunohistochemistry are substantiallyhigher than those on H&E and the limited datasuggest that much higher cutoffs are needed to reachprognostic significance.23,38,48 As tumor buddingis most prominent at the invasive tumor front, weconcentrate on, but do not restrict our assessment to,this area.

At scanning power, clues to the presence of tumorbudding include:

(a) infiltrating growth pattern;(b) marked irregularity at invasive front; and(c) blurring of the interface between tumor and

underlying stroma.

The presence of these features prompts closeevaluation at medium and high magnification.

In our experience, there are several scenarios inwhich tumor-budding assessment is challenging:

(a) abundance of histiocytes and activated lympho-cytes at invasive front (Figure 2a);

(b) prominent stromal reaction (Figure 2c);(c) glandular fragmentation associated with a

marked acute inflammatory infiltrate;(d) tumour fragments floating in mucin pools; and(e) fragmented glands with surrounding retraction

artifact (Figure 2e)

In scenarios (a) and (b), a cytokeratin stain is usedto distinguish between tumor cells and stromal/inflammatory cells (Figures 2b and d). In scenarios(c) to (e), the fragmented cells are excluded from budcounts.

We believe it is particularly important to reporttumour budding in the setting of malignant polyps,as this finding increases the risk of nodal metastasisto around 15–20%,26 and should prompt considera-tion of a segmental resection with lymphadenec-tomy. As such, we make a point of commenting not

only on its presence, but on its associated risk ofnodal metastasis (as we do for lymphovascularinvasion and a poorly differentiated component).

Our approach will undoubtedly need to be refinedonce optimal evidence-based guidelines are estab-lished. However, based on the available evidence,and in an area lacking standardization, this strategyprovides institutional guidelines and consistencyfor reporting of tumour budding.

Conclusions and future directions

There is strong evidence to suggest that tumor bud-ding is an adverse prognostic factor that can help tostratify patients into more meaningful risk groupsthan TNM staging alone, and, even more impor-tantly, has the potential to guide decision making.Its presence in T1 colorectal carcinoma, whenevaluated in conjunction with other prognosticallysignificant clinical and histopathological features,can identify a subset of high-risk patients requiringsegmental resection including nodes. In the settingof Stage II colorectal carcinoma, the presence ofbudding has been associated with significantlyworse outcomes, and if this subset of patients isshown to derive a survival benefit from chemo-therapy, the evaluation of budding could potentiallyassume an important role in treatment algorithms inStage II colorectal carcinoma.

In order for the considerable prognostic powerof tumor budding to be fully realized, consensuscriteria for its evaluation must be established, bothto guide further research in this area and to providethe practicing pathologist with guidelines for itsreporting. With respect to setting these criteria,studies focusing on budding should be designed todetermine, if possible, the optimal quantitativecutoff for defining clinically significant tumorbudding. Tumor budding is a continuous variable;thus, any cutoff is bound to be at least somewhatarbitrary, but an attempt must be made at definingthe budding threshold that will provide the mostclinically relevant prognostic and predictive infor-mation. Furthermore, authors reporting on tumorbudding should provide more detailed informationregarding the qualitative and quantitative criteriaused to evaluate budding in order to allow formeaningful comparisons among different studies.Finally, the role of immunohistochemistry in theevaluation of budding needs to be clarified.Although it is probably unnecessary and impracticalto perform immunohistochemistry on all colorectalcarcinoma cases, there may be certain situations,particularly in the context of a prominent peritu-moral inflammatory reaction, where a cytokeratinstain may reveal buds that are obscured on H&E.Criteria for immunohistochemical evaluation there-fore also need to be established.

The greatest obstacle to the adoption of buddingas a routinely reportable feature is the lack of

Tumor budding



well-defined, evidence-based criteria for its assess-ment. Given the overwhelming evidence that tumorbudding is one of the most powerful prognosticfactors at our disposal, it is incumbent on the GIpathology community to move swiftly to addressand remove the barriers to its universal reporting.

Acknowledgement

We thank Dr Masanori Tanaka (Department ofPathology, City Hospital, Hirosaki, Japan) for assis-tance in obtaining and translating Japanese articles.

Disclosure/conflict of interest

The authors declare no conflict of interest.

References

1 Greenlee RT, Hill-Harmon MB, Murray T, et al. Cancerstatistics, 2001. CA Cancer J Clin 2001;51:15–36.

2 Puppa G, Sonzogni A, Colombari R, et al. TNM stagingsystem of colorectal carcinoma: a critical appraisalof challenging issues. Arch Pathol Lab Med 2010;134:837–852.

3 Imai T. The growth of human carcinoma: a morpho-logical analysis. Fukuoka Igaku Zasshi 1954;45:30.

4 Chaffer CL, Weinberg RA. A perspective on cancer cellmetastasis. Science 2011;331:1559–1564.

5 Kalluri R. EMT: when epithelial cells decide to becomemesenchymal-like cells. J Clin Invest 2009;119:1417–1419.

6 Kalluri R, Weinberg RA. The basics of epithelial-mesenchymal transition. J Clin Invest 2009;119:1420–1428.

7 Acloque H, Adams MS, Fishwick K, et al. Epithelial-mesenchymal transitions: the importance of changingcell state in development and disease. J Clin Invest2009;119:1438–1449.

8 Zlobec I, Lugli A. Epithelial mesenchymal transitionand tumor budding in aggressive colorectal cancer:tumor budding as oncotarget. Oncotarget 2010;1:651–661.

9 Gabbert H, Wagner R, Moll R, et al. Tumor dedifferen-tiation: an important step in tumor invasion. Clin ExpMetastasis 1985;3:257–279.

10 Prall F, Ostwald C, Linnebacher M. Tubular invasionand the morphogenesis of tumor budding in colorectalcarcinoma. Hum Pathol 2009;40:1510–1512.

11 Morodomi T, Isomoto H, Shirouzu K, et al. An indexfor estimating the probability of lymph node metas-tasis in rectal cancers. Lymph node metastasis and thehistopathology of actively invasive regions of cancer.Cancer 1989;63:539–543.

12 Shinto E, Jass JR, Tsuda H, et al. Differential prognosticsignificance of morphologic invasive markers in color-ectal cancer: tumor budding and cytoplasmic podia.Dis Colon Rectum 2006;49:1422–1430.

13 Shinto E, Mochizuki H, Ueno H, et al. A novel classi-fication of tumour budding in colorectal cancer basedon the presence of cytoplasmic pseudo-fragmentsaround budding foci. Histopathology 2005;47:25–31.

14 Lugli A, Vlajnic T, Giger O, et al. Intratumoral buddingas a potential parameter of tumor progression inmismatch repair-proficient and mismatch repair-defi-cient colorectal cancer patients. Hum Pathol 2011;42:1833–1840.

15 Choi HJ, Park KJ, Shin JS, et al. Tumor budding as aprognostic marker in stage-III rectal carcinoma. Int JColorectal Dis 2007;22:863–868.

16 Guzinska-Ustymowicz K. The role of tumour buddingat the front of invasion and recurrence of rectalcarcinoma. Anticancer Res 2005;25:1269–1272.

17 Ha SS, Choi HJ, Park KJ, et al. Intensity of tumorbudding as an index for the malignant potentialin invasive rectal carcinoma. Cancer Res Treat 2005;37:177–182.

18 Hase K, Shatney C, Johnson D, et al. Prognostic valueof tumor ‘budding’ in patients with colorectal cancer.Dis Colon Rectum 1993;36:627–635.

19 Homma Y, Hamano T, Otsuki Y, et al. Severe tumorbudding is a risk factor for lateral lymph node meta-stasis in early rectal cancers. J Surg Oncol 2010;102:230–234.

20 Kajiwara Y, Ueno H, Hashiguchi Y, et al. Risk factors ofnodal involvement in T2 colorectal cancer. Dis ColonRectum 2010;53:1393–1399.

21 Kanazawa H, Mitomi H, Nishiyama Y, et al. Tumourbudding at invasive margins and outcome in colorectalcancer. Colorectal Dis 2008;10:41–47.

22 Kazama S, Watanabe T, Ajioka Y, et al. Tumourbudding at the deepest invasive margin correlates withlymph node metastasis in submucosal colorectalcancer detected by anticytokeratin antibody CAM5.2.Br J Cancer 2006;94:293–298.

23 Lugli A, Karamitopoulou E, Panayiotides I, et al. CD8+lymphocytes/tumour-budding index: an independentprognostic factor representing a ‘pro-/anti-tumour’approach to tumour host interaction in colorectalcancer. Br J Cancer 2009;101:1382–1392.

24 Masaki T, Matsuoka H, Sugiyama M, et al. Tumorbudding and evidence-based treatment of T2 rectalcarcinomas. J Surg Oncol 2005;92:59–63.

25 Sohn DK, Chang HJ, Park JW, et al. Histopathologicalrisk factors for lymph node metastasis in submucosalinvasive colorectal carcinoma of pedunculated orsemipedunculated type. J Clin Pathol 2007;60:912–915.

26 Ueno H, Mochizuki H, Hashiguchi Y, et al. Risk factorsfor an adverse outcome in early invasive colorectalcarcinoma. Gastroenterology 2004;127:385–394.

27 Ueno H, Murphy J, Jass JR, et al. Tumour ‘budding’ asan index to estimate the potential of aggressivenessin rectal cancer. Histopathology 2002;40:127–132.

28 Ueno H, Price AB, Wilkinson KH, et al. A newprognostic staging system for rectal cancer. Ann Surg2004;240:832–839.

29 Yamauchi H, Togashi K, Kawamura YJ, et al. Patholo-gical predictors for lymph node metastasis in T1colorectal cancer. Surg Today 2008;38:905–910.

30 Wang HS, Liang WY, Lin TC, et al. Curative resectionof T1 colorectal carcinoma: risk of lymph nodemetastasis and long-term prognosis. Dis Colon Rectum2005;48:1182–1192.

31 Ohtsuki K, Koyama F, Tamura T, et al. Prognostic valueof immunohistochemical analysis of tumor budding incolorectal carcinoma. Anticancer Res 2008;28:1831–1836.

32 Suzuki A, Togashi K, Nokubi M, et al. Evaluation ofvenous invasion by Elastica van Gieson stain and

Tumor budding



tumor budding predicts local and distant metastasesin patients with T1 stage colorectal cancer. Am J SurgPathol 2009;33:1601–1607.

33 Morodomi T. Clinicopathological studies of advancedrectal cancers—prediction of the degree of lymphnode metastasis from histopathological finding of pre-operative biopsy specimens. Nippon Geka GakkaiZasshi 1988;89:352–364.

34 Okuyama T, Oya M, Ishikawa H. Budding as a usefulprognostic marker in pT3 well- or moderately-differ-entiated rectal adenocarcinoma. J Surg Oncol 2003;83:42–47.

35 Okuyama T, Nakamura T, Yamaguchi M. Budding isuseful to select high-risk patients in stage II well-differentiated or moderately differentiated colon ade-nocarcinoma. Dis Colon Rectum 2003;46:1400–1406.

36 Nakamura T, Mitomi H, Kanazawa H, et al. Tumorbudding as an index to identify high-risk patients withstage II colon cancer. Dis Colon Rectum 2008;51:568–572.

37 Yasuda K, Inomata M, Shiromizu A, et al. Risk factorsfor occult lymph node metastasis of colorectal cancerinvading the submucosa and indications for endo-scopic mucosal resection. Dis Colon Rectum 2007;50:1370–1376.

38 Prall F, Nizze H, Barten M. Tumour budding asprognostic factor in stage I/II colorectal carcinoma.Histopathology 2005;47:17–24.

39 Ueno H, Jones AM, Wilkinson KH, et al. Histologicalcategorisation of fibrotic cancer stroma in advancedrectal cancer. Gut 2004;53:581–586.

40 Syk E, Lenander C, Nilsson PJ, et al. Tumour buddingcorrelates with local recurrence of rectal cancer.Colorectal Dis 2011;13:255–262.

41 Wang LM, Kevans D, Mulcahy H, et al. Tumor buddingis a strong and reproducible prognostic marker in T3N0colorectal cancer. Am J Surg Pathol 2009;33:134–141.

42 Tytherleigh MG, Warren BF, Mortensen NJ. Manage-ment of early rectal cancer. Br J Surg 2008;95:409–423.

43 Ishikawa Y, Akishima-Fukasawa Y, Ito K, et al. Histo-pathologic determinants of regional lymph node meta-stasis in early colorectal cancer. Cancer 2008;112:924–933.

44 Losi L, Ponti G, Gregorio CD, et al. Prognosticsignificance of histological features and biologicalparameters in stage I (pT1 and pT2) colorectal adeno-carcinoma. Pathol Res Pract 2006;202:663–670.

45 Tanaka M, Hashiguchi Y, Ueno H, et al. Tumorbudding at the invasive margin can predict patientsat high risk of recurrence after curative surgery forstage II, T3 colon cancer. Dis Colon Rectum 2003;46:1054–1059.

46 Gray R, Barnwell J, McConkey C, et al. Adjuvantchemotherapy versus observation in patients withcolorectal cancer: a randomised study. Lancet 2007;370:2020–2029.

47 Sy J, Fung CL, Dent OF, et al. Tumor budding andsurvival after potentially curative resection of node-positive colon cancer. Dis Colon Rectum 2010;53:301–307.

48 Zlobec I, Molinari F, Martin V, et al. Tumor buddingpredicts response to anti-EGFR therapies in metastaticcolorectal cancer patients. World J Gastroenterol2010;16:4823–4831.

49 Jass JR, Barker M, Fraser L, et al. APC mutation andtumour budding in colorectal cancer. J Clin Pathol2003;56:69–73.

50 Kevans D, Wang LM, Sheahan K, et al. Epithelial-Mesenchymal Transition (EMT) protein expression ina cohort of stage II colorectal cancer patients withcharacterized tumor budding and mismatch repairprotein status. Int J Surg Pathol 2011;19:751–760.

51 Shinto E, Baker K, Tsuda H, et al. Tumor buds showreduced expression of laminin-5 gamma 2 chain inDNA mismatch repair deficient colorectal cancer. DisColon Rectum 2006;49:1193–1202.

52 Mikami T, Yoshida T, Numata Y, et al. Invasivebehavior of ulcerative colitis-associated carcinoma isrelated to reduced expression of CD44 extracellulardomain: comparison with sporadic colon carcinoma.Diagn Pathol 2011;6:30.

53 Masaki T, Matsuoka H, Sugiyama M, et al. Actualnumber of tumor budding as a new tool for theindividualization of treatment of T1 colorectal carci-nomas. J Gastroenterol Hepatol 2006;21:1115–1121.

54 Ogawa T, Yoshida T, Tsuruta T, et al. Tumor buddingis predictive of lymphatic involvement and lymphnode metastases in submucosal invasive colorectaladenocarcinomas and in non-polypoid compared withpolypoid growths. Scand J Gastroenterol 2009;44:605–614.

55 Prall F, Ostwald C. High-degree tumor budding andpodia-formation in sporadic colorectal carcinomaswith K-ras gene mutations. Hum Pathol 2007;38:1696–1702.

56 Altman DG. Practical Statistics for Medical Research,Vol, Chapman and Hall: London, 1991.

57 Turner RR, Li C, Compton CC. Newer pathologicassessment techniques for colorectal carcinoma. ClinCancer Res 2007;13:6871s–6876s.

58 Jass JR, O’Brien MJ, Riddell RH, et al. Recommenda-tions for the reporting of surgically resected specimensof colorectal carcinoma. Virchows Arch 2007;450:1–13.

Tumor budding



tumor budding in colorectal carcinoma: time to take notice

Documents