Histopathology 2012, 61, 562–575. DOI: 10.1111/j.1365-2559.2012.04270.x
Diagnostic reproducibility of tumour budding in colorectal cancer: a multicentre, multinational study using virtual microscopy Giacomo Puppa, Carlo Senore,1 Kieran Sheahan,2 Michael Vieth,3 Alessandro Lugli,4 Inti Zlobec,4 Sara Pecori,5 Lai Mun Wang,6 Cord Langner,7 Hiroyuki Mitomi,8 Takatoshi Nakamura,9 Masahiko Watanabe,9 Hideki Ueno,10 Jacques Chasle,11 Stephen A Conley,12 Paulette Herlin,13 Gregory Y Lauwers14 & Mauro Risio15 Division of Pathology, ‘G. Fracastoro’ City Hospital, Verona, Italy, 1AOUS Giovanni Battista, CPO Piemonte, SCDO Epidemiologia dei Tumori, Torino, Italy, 2Department of Histopathology and Centre for Colorectal Disease, St Vincent’s University Hospital School of Medicine and Medical Science, University College Dublin, Dublin, Ireland, 3Institute of Pathology, Klinikum Bayreuth, Bayreuth, Germany, 4Institute of Pathology, University of Bern, Bern, Switzerland, 5 Department of Pathology, Section of Anatomical Pathology, Policlinico G. B. Rossi, University of Verona, Verona, Italy, 6 Department of Cellular Pathology, John Radcliffe Hospital, Headington, Oxford, UK, 7Institute of Pathology, Medical University of Graz, Graz, Austria, 8Department of Human Pathology, Juntendo University School of Medicine, Tokyo, Japan, 9Department of Surgery, Kitasato University School of Medicine, Sagamihara, Kanagawa, Japan, 10Department of Surgery, National Defense Medical College, Namiki, Tokorozawa, Saitama, Japan, 11Department of Pathology, Franc¸ois Baclesse Comprehensive Cancer Center, Caen, France, 12Pathology Media Lab, Pathology Service, Massachusetts General Hospital, Boston, MA, USA, 13Groupe Re´gional d’Etudes sur le Cancer, Franc¸ois Baclesse Comprehensive Cancer Center, University of Caen, Caen, France, 14Gastrointestinal Pathology Service and Division of Surgical Pathology, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA, and 15Unit of Pathology, Institute for Cancer Research and Treatment-IRCC, Candiolo, Torino, Italy Date of submission 7 November 2011 Accepted for publication 7 February 2012
Puppa G, Senore C, Sheahan K, Vieth M, Lugli A, Zlobec I, Pecori S, Wang L M, Langner C, Mitomi H, Nakamura T, Watanabe M, Ueno H, Chasle J, Conley S A, Herlin P, Lauwers G Y & Risio M (2012) Histopathology 61, 562–575
Diagnostic reproducibility of tumour budding in colorectal cancer: a multicentre, multinational study using virtual microscopy Aims: Despite the established prognostic relevance of tumour budding in colorectal cancer, the reproducibility of the methods reported for its assessment has not yet been determined, limiting its use and reporting in routine pathology practice. Methods and results: A morphometric system within telepathology was devised to evaluate the reproducibility of the various methods published for the assessment of tumour budding in colorectal cancer. Five methods were selected to evaluate the diagnostic reproducibility among 10 investigators, using haematoxylin and eosin (H&E) and AE1-3 cytokeratin-immu-
nostained, whole-slide digital scans from 50 pT1–pT4 colorectal cancers. The overall interobserver agreement was fair for all methods, and increased to moderate for pT1 cancers. The intraobserver agreement was also fair for all methods and moderate for pT1 cancers. Agreement was dependent on the participants’ experience with tumour budding reporting and performance time. Cytokeratin immunohistochemistry detected a higher percentage of tumour budding-positive cases with all methods compared to H&E-stained slides, but did not influence agreement levels.
Address for correspondence: G Puppa, Division of Pathology, ‘G. Fracastoro’ City Hospital, Via Circonvallazione, 19, I-37047 San Bonifacio (VR), Italy. e-mail:
[email protected] 2012 Blackwell Publishing Limited.
Tumour budding diagnostic reproducibility
Conclusions: An overall fair level of diagnostic agreement for tumour budding in colorectal cancer was demonstrated, which was significantly higher in early cancer and among experienced gastrointestinal pathol-
563
ogists. Cytokeratin immunostaining facilitated detection of budding cancer cells, but did not result in improved interobserver agreement.
Keywords: colorectal cancer, reproducibility, tumour budding, virtual microscopy Abbreviations: CRC, colorectal cancer; IHC, immunohistochemistry; TB, tumour budding; TNM, tumour–node– metastasis
Introduction Colorectal cancer (CRC) is one of the most common malignancies, ranking third in frequency in men and second in women worldwide. It accounts for 8% of all cancer deaths, making it the fourth most common cause of death from cancer worldwide.1 Tumour stage as defined by the American Joint Committee on Cancer ⁄ International Union Against Cancer (AJCC ⁄ UICC) tumour–node–metastasis (TNM) system is currently regarded as the strongest prognostic parameter for CRC patients. The failure of the AJCC ⁄ UICC staging system in predicting prognosis in intermediatestage tumours2,3 may be overcome by adding additional factors, either morphological, molecular or treatment-related, that could stratify patients more precisely into different risk categories.4 Among the morphological tumour-related factors, tumour budding (TB), defined as the presence of isolated single cells or small clusters of cells scattered in the stroma at the invasive tumour margin, has been shown to be a strong prognostic factor in colorectal,5 ampullary6 and lung cancers7 in particular. TB reflects a process of dedifferentiation, and represents the histological hallmark of epithelial–mesenchymal transition. In CRC the presence of TB is associated significantly with several tumour characteristics which are thought to indicate tumour aggressiveness, such as tumour differentiation,8,9 growth pattern,5,9 lymphatic invasion,5,8,10–16 level of invasion (T category),8,9,13,16–18 nodal involvement8–10,12–28 and distant metastasis.25,29 Importantly, TB has to be distinguished, histologically and prognostically, from other features present at the invasive tumour margin, usually assessed and reported separately (e.g. the growth pattern and the grade of tumour differentiation).30 Several studies have shown that TB is associated with shorter disease-free10,14,15,17,31–33 and overall5,10,11,14,15,23,31–33 survival in stages I–III CRC, and a worse progression-free survival in stage IV CRC.34 Therefore, TB appears to be a promising prognostic factor identifying those high-risk patients35 who need further therapy, including surgery (for 2012 Blackwell Publishing Ltd, Histopathology, 61, 562–575.
malignant polyps) and adjuvant therapy (for stage II disease).4 Furthermore, TB has been correlated recently with response to anti-epidermal growth factor receptor (EGFR) therapies in metastatic CRC patients, similar to K-ras mutation status.34 Despite its prognostic value there is great heterogeneity in the reporting of TB, and no consensus on assessment methods and cut-off values, limiting its clinical use. In particular, it is unclear whether TB should be assessed in the entire tumour (all sections ⁄ blocks)5 or in the area in which the highest level of TB is identified.9,36 Also, some authors have assessed TB on routine, haematoxylin & eosin (H&E)-stained slides alone,5,9,15,16,32 whereas others use broad-spectrum cytokeratin immunohistochemistry (IHC) to highlight foci of budding cells.21,22,33,37–39 In this regard, TB assessment may be impaired by the difficulty in discriminating neoplastic buds from active fibroblasts in the desmoplastic stroma37 and by the presence of a dense inflammatory infiltrate that may obscure TB.35 As expected, the literature shows that there is clearly a tendency to underestimate TB on H&E-stained sections compared to sections labelled using IHC.17,37 Also, variation exists among authors regarding the choice of field at the advancing edge. For example, Ueno’s method does not assess TB in fields where the tumour is fragmented because of the aggregation of inflammatory cells or because of a technical artefact.26 It has also not been established whether each method can be applied to all stages because, for example, the invasive front of cancer in a malignant polyp is significantly narrower than in more advanced tumours (T2, T3). A further problem is the lack of standardization of the field-of-view depending on the microscope used. Several methods have been introduced so far, the more commonly used being based on the evaluation of H&E-stained slides (Table 1). Hase advanced an approach based on subjective impression, identifying a predominant pattern of TB: either none or mild (termed BD-1), or moderate or severe (BD-2).8 This method was conceived for stages A–D of the Duke classification. Ueno devised a new approach based on the premise that TB could be
564
G Puppa et al.
Table 1. The five methods used in the interobserver study Author
Approach
Tumour field
Area (mm2)
Two-tiered classification
Stage(s)
Hase (1993)
Subjective impression, predominant pattern
Overall tumour
Not applicable
None or mild (BD-1)
Moderate or severe (BD-2)
Early and advanced
Nakamura (2005)
Semiquantitative, percentage of involvement by TB
Entire advancing edge
Not applicable
Low grade (none or mild: 1 ⁄ 3)
High-grade (moderate: 1 ⁄ 3–2 ⁄ 3 or marked: >2 ⁄ 3)
Advanced
Ueno (2002)
Count with cut-off
Field with maximum TB
0.385
Low grade (
