Multi-Institutional Evaluation of Pathologists' Assessment Compared to Immunoscore.

BACKGROUND: The Immunoscore (IS) is a quantitative digital pathology assay that evaluates the immune response in cancer patients. This study reports on the reproducibility of pathologists' visual assessment of CD3+- and CD8+-stained colon tumors, compared to IS quantification. METHODS: An international group of expert pathologists evaluated 540 images from 270 randomly selected colon cancer (CC) cases. Concordance between pathologists' T-score, corresponding hematoxylin-eosin (H&E) slides, and the digital IS was evaluated for two- and three-category IS. RESULTS: Non-concordant T-scores were reported in more than 92% of cases. Disagreement between semi-quantitative visual assessment of T-score and the reference IS was observed in 91% and 96% of cases before and after training, respectively. Statistical analyses showed that the concordance index between pathologists and the digital IS was weak in two- and three-category IS, respectively. After training, 42% of cases had a change in T-score, but no improvement was observed with a Kappa of 0.465 and 0.374. For the 20% of patients around the cut points, no concordance was observed between pathologists and digital pathology analysis in both two- and three-category IS, before or after training (all Kappa < 0.12). CONCLUSIONS: The standardized IS assay outperformed expert pathologists' T-score evaluation in the clinical setting. This study demonstrates that digital pathology, in particular digital IS, represents a novel generation of immune pathology tools for reproducible and quantitative assessment of tumor-infiltrated immune cell subtypes.

Fully Automated Tumor Bud Assessment in Hematoxylin and Eosin-Stained Whole Slide Images of Colorectal Cancer.

Tumor budding (TB), the presence of single cells or small clusters of up to 4 tumor cells at the invasive front of colorectal cancer (CRC), is a proven risk factor for adverse outcomes. International definitions are necessary to reduce interobserver variability. According to the current international guidelines, hotspots at the invasive front should be counted in hematoxylin and eosin (H&E)-stained slides. This is time-consuming and prone to interobserver variability; therefore, there is a need for computer-aided diagnosis solutions. In this study, we report an artificial intelligence-based method for detecting TB in H&E-stained whole slide images. We propose a fully automated pipeline to identify the tumor border, detect tumor buds, characterize them based on the number of tumor cells, and produce a TB density map to identify the TB hotspot. The method outputs the TB count in the hotspot as a computational biomarker. We show that the proposed automated TB detection workflow performs on par with a panel of 5 pathologists at detecting tumor buds and that the hotspot-based TB count is an independent prognosticator in both the univariate and the multivariate analysis, validated on a cohort of n = 981 patients with CRC. Computer-aided detection of tumor buds based on deep learning can perform on par with expert pathologists for the detection and quantification of tumor buds in H&E-stained CRC histopathology slides, strongly facilitating the introduction of budding as an independent prognosticator in clinical routine and clinical trials.

Image-based assessment of extracellular mucin-to-tumor area predicts consensus molecular subtypes (CMS) in colorectal cancer.

The backbone of all colorectal cancer classifications including the consensus molecular subtypes (CMS) highlights microsatellite instability (MSI) as a key molecular pathway. Although mucinous histology (generally defined as >50% extracellular mucin-to-tumor area) is a "typical" feature of MSI, it is not limited to this subgroup. Here, we investigate the association of CMS classification and mucin-to-tumor area quantified using a deep learning algorithm, and  the expression of specific mucins in predicting CMS groups and clinical outcome. A weakly supervised segmentation method was developed to quantify extracellular mucin-to-tumor area in H&E images. Performance was compared to two pathologists' scores, then applied to two cohorts: (1) TCGA (n = 871 slides/412 patients) used for mucin-CMS group correlation and (2) Bern (n = 775 slides/517 patients) for histopathological correlations and next-generation Tissue Microarray construction. TCGA and CPTAC (n = 85 patients) were used to further validate mucin detection and CMS classification by gene and protein expression analysis for MUC2, MUC4, MUC5AC and MUC5B. An excellent inter-observer agreement between pathologists' scores and the algorithm was obtained (ICC = 0.92). In TCGA, mucinous tumors were predominantly CMS1 (25.7%), CMS3 (24.6%) and CMS4 (16.2%). Average mucin in CMS2 was 1.8%, indicating negligible amounts. RNA and protein expression of MUC2, MUC4, MUC5AC and MUC5B were low-to-absent in CMS2. MUC5AC protein expression correlated with aggressive tumor features (e.g., distant metastases (p = 0.0334), BRAF mutation (p < 0.0001), mismatch repair-deficiency (p < 0.0001), and unfavorable 5-year overall survival (44% versus 65% for positive/negative staining). MUC2 expression showed the opposite trend, correlating with less lymphatic (p = 0.0096) and venous vessel invasion (p = 0.0023), no impact on survival.The absence of mucin-expressing tumors in CMS2 provides an important phenotype-genotype correlation. Together with MSI, mucinous histology may help predict CMS classification using only histopathology and should be considered in future image classifiers of molecular subtypes.

Reproducibility of tumor budding assessment in pancreatic cancer based on a multicenter interobserver study.

Tumor budding has been reported to be an independent prognostic factor in pancreatic ductal adenocarcinoma (PDAC). Its use in daily diagnostics would improve the prognostic stratification of patients. We performed a multicenter interobserver study to test various budding assessment methods for their reproducibility. Two serial sections of 50 resected, treatment-naïve PDACs were stained for Hematoxylin and Eosin (H&E) and pancytokeratin. Tumor budding was scored by independent observers at five participating centers in Switzerland, Germany, and Canada. Pathologists assessed tumor budding on a digital platform comparing H&E with pancytokeratin staining in 10 high-power fields (10HPF) and one HPF hotspot (1HPF). Additionally, tumor budding was assessed in one H&E hotspot at × 20 magnification, as suggested by the International Tumor Budding Consensus Conference (ITBCC). Correlation coefficients for bud counts between centers ranged from r = 0.58648 to r = 0.78641 for H&E and from r = 0.69288 to r = 0.81764 for pancytokeratin. The highest interobserver agreement across all centers was observed for pancytokeratin 10HPFs (ICC = 0.6). ICC values were 0.49, 0.48, 0.41, and 0.4 for H&E in 1HPF hotspot, H&E in 10HPFs, pancytokeratin in 1HPF, and H&E in one hotspot at ×20, respectively (ITBCC method). This interobserver study reveals a range between moderately poor to moderate agreement levels between pathologists for the different tumor budding assessment methods in PDAC. Acceptable levels of agreement were reached with the pancytokeratin 10HPF method, which can thus be recommended for the assessment of tumor budding in PDAC resection specimens. To improve the levels of interobserver agreement, the implementation of machine learning applications should be considered.

Cytokeratin-based assessment of tumour budding in colorectal cancer: analysis in stage II patients and prospective diagnostic experience.

Tumour budding in colorectal cancer is an important prognostic factor. A recent consensus conference elaborated recommendations and key issues for future studies, among those the use of pan-cytokeratin stains, especially in stage II patients. We report the first prospective diagnostic experience using pan-cytokeratin for tumour budding assessment. Moreover, we evaluate tumour budding using pan-cytokeratin stains and disease-free survival (DFS) in stage II patients. To this end, tumour budding on pan-cytokeratin-stained sections was evaluated by counting the number of tumour buds in 10 high-power fields (0.238 mm2), then categorizing counts as low/high-grade at a cut-off of 10 buds, in two cohorts. Cohort 1: prospective setting with 236 unselected primary resected colorectal cancers analysed by 17 pathologists during diagnostic routine. Cohort 2: retrospective cohort of 150 stage II patients with information on DFS. In prospective analysis of cohort 1, tumour budding counts correlated with advanced pT, lymph node metastasis, lymphovascular invasion, perineural invasion (all p < 0.0001), and distant metastasis (p = 0.0128). In cohort 2, tumour budding was an independent predictor of worse DFS using counts [p = 0.037, HR (95% CI): 1.007 (1.0-1.014)] and the low-grade/high-grade scoring approach [p = 0.02; HR (95% CI): 3.04 (1.2-7.77), 90.7 versus 73%, respectively]. In conclusion, tumour budding assessed on pan-cytokeratin slides is feasible in a large pathology institute and leads to expected associations with clinicopathological features. Additionally, it is an independent predictor of poor prognosis in stage II patients and should be considered for risk stratification in future clinical studies.

Comprehensive assessment of tumour budding by cytokeratin staining in colorectal cancer.

AIMS: Tumour budding in colorectal cancer (CRC) is a recognized prognostic parameter. The aim of this study was to address the use of cytokeratin immunostaining for the visualization and scoring of tumour buds. METHODS AND RESULTS: Ten hotspots (0.238 mm2 ) of peritumoural budding (PTB) and intratumoural budding (ITB) were evaluated in surgical resections from 215 patients. The budding counts in the 10 densest regions anywhere in the tumour were combined into an overall tumour budding (OTB) score. The PTB, ITB and OTB hotspot with the maximum budding count was then evaluated. Finally, continuous and cut-off values of 10 buds per high-power field (HPF) (PTB10HPF ), five buds per HPF (ITB10HPF ) and eight buds per HPF (OTB10HPF ) were used to categorize budding counts into low-grade and high-grade scores. All budding scores were highly correlated. PTB and ITB counts were associated with many clinicopathological features, including tumour stage, lymph node and distant metastasis, venous and lymphovascular invasion, and disease-free survival (DFS) (all P < 0.05). Analyses of OTB counts recapitulated these associations, including a lower DFS with a greater number of tumour buds (P = 0.0309; hazard ratio 1.0332, 95% confidence interval 1.003-1.062). One OTB hotspot performed similarly to 10 OTB hotspots in terms of relationship with outcome. These statistical significances were largely lost when cut-offs were applied to PTB, ITB or OTB counts. CONCLUSIONS: An OTB count in a single hotspot on cytokeratin-stained CRC tissue sections is a fast and reliable prognostic scoring system for the assessment of tumour budding. This approach should be considered in future studies.

Assessment of tumor regression of esophageal adenocarcinomas after neoadjuvant chemotherapy: comparison of 2 commonly used scoring approaches.

Histopathologic determination of tumor regression provides important prognostic information for locally advanced gastroesophageal carcinomas after neoadjuvant treatment. Regression grading systems mostly refer to the amount of therapy-induced fibrosis in relation to residual tumor or the estimated percentage of residual tumor in relation to the former tumor site. Although these methods are generally accepted, currently there is no common standard for reporting tumor regression in gastroesophageal cancers. We compared the application of these 2 major principles for assessment of tumor regression: hematoxylin and eosin-stained slides from 89 resection specimens of esophageal adenocarcinomas following neoadjuvant chemotherapy were independently reviewed by 3 pathologists from different institutions. Tumor regression was determined by the 5-tiered Mandard system (fibrosis/tumor relation) and the 4-tiered Becker system (residual tumor in %). Interobserver agreement for the Becker system showed better weighted κ values compared with the Mandard system (0.78 vs. 0.62). Evaluation of the whole embedded tumor site showed improved results (Becker: 0.83; Mandard: 0.73) as compared with only 1 representative slide (Becker: 0.68; Mandard: 0.71). Modification into simplified 3-tiered systems showed comparable interobserver agreement but better prognostic stratification for both systems (log rank Becker: P=0.015; Mandard P=0.03), with independent prognostic impact for overall survival (modified Becker: P=0.011, hazard ratio=3.07; modified Mandard: P=0.023, hazard ratio=2.72). In conclusion, both systems provide substantial to excellent interobserver agreement for estimation of tumor regression after neoadjuvant chemotherapy in esophageal adenocarcinomas. A simple 3-tiered system with the estimation of residual tumor in % (complete regression/1% to 50% residual tumor/>50% residual tumor) maintains the highest reproducibility and prognostic value.

Proposal for a 10-high-power-fields scoring method for the assessment of tumor budding in colorectal cancer.

Although tumor budding is linked to adverse prognosis in colorectal cancer, it remains largely unreported in daily diagnostic work due to the absence of a standardized scoring method. Our aim was to assess the inter-observer agreement of a novel 10-high-power-fields method for assessment of tumor budding at the invasive front and to confirm the prognostic value of tumor budding in our setting of colorectal cancers. Whole tissue sections of 215 colorectal cancers with full clinico-pathological and follow-up information were stained with cytokeratin AE1/AE3 antibody. Presence of buds was scored across 10-high-power fields at the invasive front by two pathologists and two additional observers were asked to score 50 cases of tumor budding randomly selected from the larger cohort. The measurements were correlated to the patient and tumor characteristics. Inter-observer agreement and correlation between observers' scores were excellent (P<0.0001; intraclass correlation coefficient=0.96). A test subgroup of 65 patients (30%) was used to define a valid cutoff score for high-grade tumor budding and the remaining 70% of the patients were entered into the analysis. High-grade budding was defined as an average of ≥10 buds across 10-high-power fields. High-grade budding was associated with a higher tumor grade (P<0.0001), higher TNM stage (P=0.0003), vascular invasion (P<0.0001), infiltrating tumor border configuration (P<0.0001) and reduced survival (P<0.0001). Multivariate analysis confirmed its independent prognostic effect (P=0.007) when adjusting for TNM stage and adjuvant therapy. Using 10-high-power fields for evaluating tumor budding has independent prognostic value and shows excellent inter-observer agreement. Like the BRE and Gleason scores in breast and prostate cancers, respectively, tumor budding could be a basis for a prognostic score in colorectal cancer.

Standardization of whole slide image morphologic assessment with definition of a new application: Digital slide dynamic morphometry.

BACKGROUND: In histopathology, the quantitative assessment of various morphologic features is based on methods originally conceived on specific areas observed through the microscope used. Failure to reproduce the same reference field of view using a different microscope will change the score assessed. Visualization of a digital slide on a screen through a dedicated viewer allows selection of the magnification. However, the field of view is rectangular, unlike the circular field of optical microscopy. In addition, the size of the selected area is not evident, and must be calculated. MATERIALS AND METHODS: A digital slide morphometric system was conceived to reproduce the various methods published for assessing tumor budding in colorectal cancer. Eighteen international experts in colorectal cancer were invited to participate in a web-based study by assessing tumor budding with five different methods in 100 digital slides. RESULTS: The specific areas to be tested by each method were marked by colored circles. The areas were grouped in a target-like pattern and then saved as an .xml file. When a digital slide was opened, the .xml file was imported in order to perform the measurements. Since the morphometric tool is composed of layers that can be freely moved on top of the digital slide, the technique was named digital slide dynamic morphometry. Twelve investigators completed the task, the majority of them performing the multiple evaluations of each of the cases in less than 12 minutes. CONCLUSIONS: Digital slide dynamic morphometry has various potential applications and might be a useful tool for the assessment of histologic parameters originally conceived for optical microscopy that need to be quantified.

Systematic assessment of protein phenotypes characterizing high-grade tumour budding in mismatch repair-proficient colorectal cancer.

AIMS: A tumour bud is defined as a single tumour cell or tumour cell cluster of up to five cells at the invasive tumour front. Significant differences in survival have been detected in colorectal cancer patients with low- compared to high-grade budding. The aim of this study was to identify potential multi-marker phenotypes characterizing low- and high-grade budding in mismatch repair (MMR)-proficient colorectal cancer. METHODS AND RESULTS: Established and promising prognostic proteins such as epidermal growth factor receptor (EGFR), pERK, RHAMM, RKIP, beta-catenin, E-cadherin, pAKT, p16, p21, Ki67, Bcl-2, vascular endothelial growth factor (VEGF), apoptotic protease activating factor-1 (APAF-1), MUC1, EphB2, matrix metalloproteinase 7, pSMAD2, CDX2, laminin5gamma2 and MST1 were analysed on 208 MMR-proficient colorectal cancers with complete clinicopathological data. The most accurate markers for predicting high-grade budding (more than six tumour buds) were EphB2 (P < 0.001), Bcl-2 (P < 0.001), RKIP (P < 0.001), E-cadherin (P = 0.004), laminin5gamma2 (P = 0.004) and APAF-1 (P = 0.005). On multivariable analysis, only loss of Bcl-2 (P < 0.001) and EphB2 (P < 0.001) were independent predictors of high-grade budding. Bcl-2-/EphB2- tumours were more frequently poorly differentiated (P < 0.001), of advanced pT stage (P = 0.002), lymph node positive (P = 0.023), presented vascular (P = 0.053) and lymphatic invasion (P = 0.005) and had a negative impact on patient survival (P = 0.012). CONCLUSIONS: The multi-marker phenotype EphB2-/Bcl-2- is an independent predictor of high-grade budding and implies increased aggressive behaviour in MMR-proficient colorectal cancer.

Assessment of mean EGFR gene copy number is a highly reproducible method for evaluating FISH in histological and cytological cancer specimens.

The aims of this study were first, to systematically assess the inter-observer reproducibility of mean Epidermal Growth Factor Receptor (EGFR) gene copy number (MCN) in histological and cytological specimens from lung and non-lung cancers, second to compare the performance of this quantitative approach to the current Colorado criteria for the assessment of fluorescence in situ hybridization (FISH) positivity and third to develop a model to convert cytology into histology MCN. EGFR FISH analysis was performed on 170 histological and 153 cytological specimens. The MCN and Colorado criteria were assessed by two independent observers and agreement evaluated using Bland-Altman plots and kappa values. Conversion of cytology into histology MCN was tested on two randomized subgroups of specimens. Applying the Colorado criteria, agreement between observers was moderate with histology (k=0.5) and excellent with cytology (k=0.94). Positivity in histology versus cytology led to fair agreement (k=0.33). MCN was significantly greater in cytology compared to histology (p<0.001) with excellent inter-observer agreement in both sample types (r=0.99 and 0.89, respectively). The average difference in MCN between observers was -0.003 (95%CI: -0.05 to 0.05) and 0.008 (95%CI: -0.09 to 0.11) for cytology and histology, respectively. A reliable conversion model with an R(2)-value of 0.91 in the validation subgroup (p<0.001) was obtained. The MCN is a reproducible scoring method for evaluating EGFR FISH in samples from lung and non-lung cancers. This quantitative scoring system may constitute an alternative to current methods of gene copy number assessment and will further help to optimize patient selection for targeted therapies.

Systematic assessment of the prognostic impact of membranous CD44v6 protein expression in colorectal cancer.

AIMS: To assess systematically the membranous expression of CD44v6 in colorectal cancer by immunohistochemistry to determine its prognostic impact, the differential expression between primary and metastatic tumours and expression differences between the tumour centre and invasive front. METHODS AND RESULTS: Immunohistochemistry was performed for CD44v6 on two tissue microarrays. The first included 1279 colorectal tumours with full clinicopathological data. The second consisted of 50 matched primary and metastatic tumours sampled from the tumour centre and the invasive margin. A scoring system was tested by multiple observers. Receiver-operating characteristic curve analysis was used for cut-off point determination. Loss of membranous CD44v6 was associated with pT stage (P = 0.016; sensitivity 85.8%, specificity 20.1%), lymph node metastasis (P = 0.015; sensitivity 52.8%, specificity 55%), an infiltrating tumour margin (P < 0.001; sensitivity 71.4%, specificity 40%) and adverse prognosis (P = 0.011; hazard ratio 0.79, 95% confidence interval 0.7, 0.9), but was not an independent prognostic factor on multivariable analysis. Loss of expression occurred at the invasive front in both primary and metastatic lesions (P < 0.001). CONCLUSIONS: This study outlines an approach to help standardize the immunohistochemical evaluation of CD44v6 and similar markers in colorectal cancer and highlights a significant role for loss of membranous CD44v6 expression in colorectal cancer progression and prognosis.