Search

The spread of early-stage (T1 and T2) adenocarcinomas to locoregional lymph nodes is a key event in disease progression of colorectal cancer (CRC). The cellular mechanisms behind this event are not completely understood and existing predictive biomarkers are imperfect. Here, we used an end-to-end deep learning algorithm to identify risk factors for lymph node metastasis (LNM) status in digitized histopathology slides of the primary CRC and its surrounding tissue. In two large population-based cohorts, we show that this system can predict the presence of more than one LNM in pT2 CRC patients with an area under the receiver operating curve (AUROC) of 0.733 (0.67-0.758) and patients with any LNM with an AUROC of 0.711 (0.597-0.797). Similarly, in pT1 CRC patients, the presence of more than one LNM or any LNM was predictable with an AUROC of 0.733 (0.644-0.778) and 0.567 (0.542-0.597), respectively. Based on these findings, we used the deep learning system to guide human pathology experts towards highly predictive regions for LNM in the whole slide images. This hybrid human observer and deep learning approach identified inflamed adipose tissue as the highest predictive feature for LNM presence. Our study is a first proof of concept that artificial intelligence (AI) systems may be able to discover potentially new biological mechanisms in cancer progression. Our deep learning algorithm is publicly available and can be used for biomarker discovery in any disease setting. © 2021 The Pathological Society of Great Britain and Ireland. Published by John WileySons, Ltd.

BackgroundSince implementing the NHS bowel cancer screening programme, the rate of early colorectal cancer (eCRC; pT1) has increased threefold to 17%, but how these lesions should be managed is currently unclear.AimTo improve risk stratification of eCRC by developing reproducible quantitative markers to build a multivariate model to predict lymph node metastasis (LNM).MethodsOur retrospective cohort of 207 symptomatic pT1 eCRC was assessed for quantitative markers. Associations between categorical data and LNM were performed using χ2 test and Fisher’s exact test. Multivariable modelling was performed using logistic regression. Youden’s rule gave the cut-point for LNM.ResultsAll significant parameters in the univariate analysis were included in a multivariate model; tumour stroma (95% CI 2.3 to 41.0; p=0.002), area of submucosal invasion (95% CI 2.1 to 284.6; p=0.011), poor tumour differentiation (95% CI 2.0 to 358.3; p=0.003) and lymphatic invasion (95% CI 1.3 to 192.6; p=0.028) were predictive of LNM. Youden’s rule gave a cut-off of p>5%, capturing 18/19 LNM (94.7%) cases and leading to a resection recommendation for 34% of cases. The model that only included quantitative factors were also significant, capturing 17/19 LNM cases (90%) and leading to resection rate of 35% of cases (72/206).ConclusionsIn this study, we were able to reduce the potential resection rate of pT1 with the multivariate qualitative and/or quantitative model to 34% or 35% while detecting 95% or 90% of all LNM cases, respectively. While these findings need to be validated, this model could lead to a reduction of the major resection rate in eCRC.

AIMS Screening all patients newly diagnosed with colorectal cancer (CRC) for possible Lynch syndrome (LS) has been recommended in the United Kingdom since the National Institute for Health and Care Excellence (NICE) released new diagnostics guidance in February 2017. We sought to validate the NICE screening pathway through a prospective regional programme throughout a 5.2-million population during a 2-year period. METHODS AND RESULTS Pathology departments at 14 hospital trusts in the Yorkshire and Humber region of the United Kingdom were invited to refer material from patients with newly diagnosed CRC aged 50 years or over between 1 April 2017 and 31 March 2019 for LS screening. Testing consisted of immunohistochemistry for MLH1, PMS2, MSH2 and MSH6 followed by BRAF mutation analysis ± MLH1 promoter methylation testing in cases showing MLH1 loss. A total of 3141 individual specimens were submitted for testing from 12 departments consisting of 3061 unique tumours and 2791 prospectively acquired patients with CRC. Defective mismatch repair (dMMR) was observed in 15% of cases. In cases showing MLH1 loss, 76% contained a detectable BRAF mutation and, of the remainder, 77% showed MLH1 promoter hypermethylation. Of the patients included in the final analysis, 81 (2.9%) had an indication for germline testing. CONCLUSION LS screening using the NICE diagnostics guidance pathway is deliverable at scale identifying significant numbers of patients with dMMR. This information is used to refer patients to regional clinical genetics services in addition to informing treatment pathways including the use of adjuvant/neoadjuvant chemotherapy and immunotherapy.

Through the implementation of national bowel cancer screening programmes we have seen a three-fold increase in early pT1 colorectal cancers, but how these lesions should be managed is currently unclear. Local excision can be an attractive option, especially for fragile patients with multiple comorbidities, but it is only safe from an oncological point of view in the absence of lymph node metastasis. Patient risk stratification through careful analysis of histopathological features in local excision or polypectomy specimens should be performed according to national guidelines to avoid under- or over-treatment. Currently national guidelines vary in their recommendations as to which factors should be routinely reported and there is no established multivariate risk stratification model to determine which patients should be offered major resectional surgery. Conventional histopathological parameters such as tumour grading or lymphovascular invasion have been shown to be predictive of lymph node metastasis in a number of studies but the inter- and intra-observer variation in reporting is high. Newer parameters including tumour budding and poorly differentiated clusters have been shown to have great potential, but again some improvement in the inter-observer variation is required. With the implementation of digital pathology into clinical practice, quantitative parameters like depth/area of submucosal invasion and proportion of stroma can be routinely assessed. In this review we present the various histopathological risk factors for predicting systemic spread in pT1 colorectal cancer and introduce potential novel quantitative variables and multivariable risk models that could be used to better define the optimal treatment of this increasingly common disease.

Tumor budding is a long-established independent adverse prognostic marker in colorectal cancer, yet methods for its assessment have varied widely. In an effort to standardize its reporting, a group of experts met in Bern, Switzerland, in 2016 to reach consensus on a single, international, evidence-based method for tumor budding assessment and reporting (International Tumor Budding Consensus Conference [ITBCC]). Tumor budding assessment using the ITBCC criteria has been validated in large cohorts of cancer patients and incorporated into several international colorectal cancer pathology and clinical guidelines. With the wider reporting of tumor budding, new issues have emerged that require further clarification. To better inform researchers and health-care professionals on these issues, an international group of experts in gastrointestinal pathology participated in a modified Delphi process to generate consensus and highlight areas requiring further research. This effort serves to re-affirm the importance of tumor budding in colorectal cancer and support its continued use in routine clinical practice. Supplementary Information The online version contains supplementary material available at 10.1007/s00428-021-03059-9.

International audience; Tumor budding is a promising and cost-effective biomarker with strong prognostic value in colorectal cancer. However, challenges related to interobserver variability persist. Such variability may be reduced by immunohistochemistry and computer-aided tumor bud selection. Development of computer algorithms for this purpose requires unequivocal examples of individual tumor buds. As such, we undertook a large-scale, international, and digital observer study on individual tumor bud assessment. From a pool of 46 colorectal cancer cases with tumor budding, 3000 tumor bud candidates were selected, largely based on digital image analysis algorithms. For each candidate bud, an image patch (size 256 × 256 µm) was extracted from a pan cytokeratin-stained whole-slide image. Members of an International Tumor Budding Consortium (n = 7) were asked to categorize each candidate as either (1) tumor bud, (2) poorly differentiated cluster, or (3) neither, based on current definitions. Agreement was assessed with Cohen's and Fleiss Kappa statistics. Fleiss Kappa showed moderate overall agreement between observers (0.42 and 0.51), while Cohen's Kappas ranged from 0.25 to 0.63. Complete agreement by all seven observers was present for only 34% of the 3000 tumor bud candidates, while 59% of the candidates were agreed on by at least five of the seven observers. Despite reports of moderate-to-substantial agreement with respect to tumor budding grade, agreement with respect to individual pan cytokeratin-stained tumor buds is moderate at most. A machine learning approach may prove especially useful for a more robust assessment of individual tumor buds.

ObjectiveAcademic pathology is facing a crisis; an ongoing decline in academic pathology posts, a paucity of academic pathologist’s in-training and unfilled posts at a time when cellular pathology departments are challenged to deliver increasing numbers of molecular tests. The National Cancer Research Institute initiative in Cellular & Molecular Pathology commissioned a survey to assess attitudes of cellular pathology consultants towards research in order to understand barriers and identify possible solutions to improve this situation. As cellular pathology is encompassing an increasing number of diagnostic molecular tests, we also surveyed the current approach to and extent of training in molecular pathology.MethodsThe survey was distributed to all UK-based consultant pathologists via the Pathological Society of Great Britain & Ireland and Royal College of Pathologist networks. Heads of Department were contacted separately to obtain figures for number of academic training and consultant posts.Results302 cellular pathologists completed the survey which represents approximately 21% of the total cellular histopathology workforce. Most respondents (89%) had been involved in research at some point; currently, 22% were undertaking research formally, and 41% on an informal basis. Of those previously involved in research, 57% stopped early in their consultant career. The majority of substantive academic posts were Professors of which 60% had been in post for >20 years. Most respondents (84%) used molecular pathology in diagnostic work, independent of where they worked or the length of time in post. Notably, 53% of consultants had not received molecular pathology training, particularly more senior consultants and consultants in district general hospitals.ConclusionsThe survey reveals that the academic workforce is skewed towards senior individuals, many of whom are approaching retirement, with a missing cohort of ‘junior consultant’ academic pathologists to replace them. Most pathologists stop formal research activity at the beginning of a consultant career. While molecular pathology is an increasing part of a pathologist’s workload, the majority of consultant cellular pathologists have not received any formal molecular training.

Colorectal cancer (CRC) is common with 3% of cases associated with germline mutations in the mismatch repair pathway characteristic of Lynch syndrome (LS). The UK National Institute for Health and Care Excellence recommends screening for LS in all patients newly diagnosed with CRC, irrespective of age. The Yorkshire Cancer Research Bowel Cancer Improvement Programme includes a regional LS screening service for all new diagnoses of CRC. In the first 829 cases screened, 80 cases showed deficient mismatch repair (dMMR) including four cases showing areas with loss of expression of all four mismatch repair proteins by immunohistochemistry. The cases demonstrated diffuse MLH1 loss associated with BRAF mutations and MLH1 promoter hypermethylation in keeping with sporadic dMMR, with presumed additional double hit mutations in MSH2+/−MSH6 rather than underlying LS. Recognition and accurate interpretation of this unusual phenotype is important to prevent unnecessary referrals to clinical genetics and associated patient anxiety.

Background Since 2006, detection of early colorectal cancer lesions through implementation of the National Health Service Bowel Cancer Screening programme has increased by a factor of three to about 17%, but how these lesions are managed is unclear. Therefore, new guidelines for predicting cancer spread to determine treatment are urgently needed. We aimed to build a multivariate model of quantitative reproducible markers to predict lymph node metastases (LNM). Methods Our cohort consisted of 206 symptomatic patients with pT1 colorectal cancer (19 with and 187 without LNM). Qualitative factors (tumour grade, lymphatic invasion, vascular invasion) were obtained from routine pathological reports. Quantitative markers of tumour stroma content, area of submucosal invasion of the tumour, tumour budding, and width of cancer were measured, and cut-off values were assessed as previously described. Tumour budding at the hot spot was counted. Associations between categorical data and LNM were performed with the χ 2 test and Fisher's exact test. Multivariate modelling was carried out with logistic regression to estimate predicted probabilities for LNM. Youden's rule was used to define the probability cut-off point for LNM. Findings All significant parameters in the univariate model were included in a multivariate model: tumour stroma content (odds ratio 9·7, 95% CI 2·3–41·0; p=0·002), area of submucosal invasion (24·5, 2·1–284·6; p=0·011), poor grade of differentiation (26·7, 2·0–358·3; p=0·003), and lymphatic invasion (16·1, 1·3–192·6; p=0·028) were predictive of LNM, but tumour budding and vascular invasion were not significant. Youden's rule gave a cut-off value of p>0·05 capturing 18 (95%) of 19 LNM cases, leading to resection in 70 (34%) of 206 cases. When the model only included quantitative factors all quantitative factors were significant. Youden's rule gave an optimum cut-off point at p>0·1 capturing 17 (90%) of 19 LNM cases and leading to resection of 61 (30%) of 206 cases. Interpretation In this small retrospective pilot study, we show that we could reduce the resection rate in pT1 colorectal cancer to 34% with the full multivariate model or 30% with the quantitative only model while detecting 95% or 90% of all LNM cases, respectively. This would be a major improvement to the current resection rate of around 50% with its known side-effects. At present, we are investigating our findings in a larger cohort, which could lead to improved treatment decisions for patients with pT1 colorectal cancer. Funding Yorkshire Cancer Research, Pelican Cancer Foundation, Pathological Society of Great Britain and Ireland.