Deceptive learning in histopathology.

AIMS: Deep learning holds immense potential for histopathology, automating tasks that are simple for expert pathologists and revealing novel biology for tasks that were previously considered difficult or impossible to solve by eye alone. However, the extent to which the visual strategies learned by deep learning models in histopathological analysis are trustworthy or not has yet to be systematically analysed. Here, we systematically evaluate deep neural networks (DNNs) trained for histopathological analysis in order to understand if their learned strategies are trustworthy or deceptive. METHODS AND RESULTS: We trained a variety of DNNs on a novel data set of 221 whole-slide images (WSIs) from lung adenocarcinoma patients, and evaluated their effectiveness at (1) molecular profiling of KRAS versus EGFR mutations, (2) determining the primary tissue of a tumour and (3) tumour detection. While DNNs achieved above-chance performance on molecular profiling, they did so by exploiting correlations between histological subtypes and mutations, and failed to generalise to a challenging test set obtained through laser capture microdissection (LCM). In contrast, DNNs learned robust and trustworthy strategies for determining the primary tissue of a tumour as well as detecting and localising tumours in tissue. CONCLUSIONS: Our work demonstrates that DNNs hold immense promise for aiding pathologists in analysing tissue. However, they are also capable of achieving seemingly strong performance by learning deceptive strategies that leverage spurious correlations, and are ultimately unsuitable for research or clinical work. The framework we propose for model evaluation and interpretation is an important step towards developing reliable automated systems for histopathological analysis.

Optimizing tissue stewardship in non-small cell lung cancer to support molecular characterization and treatment selection: statement from a working group of thoracic pathologists.

Many patients with non-small cell lung cancer do not receive guideline-recommended, biomarker-directed therapy, despite the potential for improved clinical outcomes. Access to timely, accurate, and comprehensive molecular profiling, including targetable protein overexpression, is essential to allow fully informed treatment decisions to be taken. In turn, this requires optimal tissue management to protect and maximize the use of this precious finite resource. Here, a group of leading thoracic pathologists recommend factors to consider for optimal tissue management. Starting from when lung cancer is first suspected, keeping predictive biomarker testing in the front of the mind should drive the development of practices and procedures that conserve tissue appropriately to support molecular characterization and treatment selection.

Artificial Intelligence-Powered Assessment of Pathologic Response to Neoadjuvant Atezolizumab in Patients With NSCLC: Results From the LCMC3 Study.

INTRODUCTION: Pathologic response (PathR) by histopathologic assessment of resected specimens may be an early clinical end point associated with long-term outcomes with neoadjuvant therapy. Digital pathology may improve the efficiency and precision of PathR assessment. LCMC3 (NCT02927301) evaluated neoadjuvant atezolizumab in patients with resectable NSCLC and reported a 20% major PathR rate. METHODS: We determined PathR in primary tumor resection specimens using guidelines-based visual techniques and developed a convolutional neural network model using the same criteria to digitally measure the percent viable tumor on whole-slide images. Concordance was evaluated between visual determination of percent viable tumor (n = 151) performed by one of the 47 local pathologists and three central pathologists. RESULTS: For concordance among visual determination of percent viable tumor, the interclass correlation coefficient was 0.87 (95% confidence interval [CI]: 0.84-0.90). Agreement for visually assessed 10% or less viable tumor (major PathR [MPR]) in the primary tumor was 92.1% (Fleiss kappa = 0.83). Digitally assessed percent viable tumor (n = 136) correlated with visual assessment (Pearson r = 0.73; digital/visual slope = 0.28). Digitally assessed MPR predicted visually assessed MPR with outstanding discrimination (area under receiver operating characteristic curve, 0.98) and was associated with longer disease-free survival (hazard ratio [HR] = 0.30; 95% CI: 0.09-0.97, p = 0.033) and overall survival (HR = 0.14, 95% CI: 0.02-1.06, p = 0.027) versus no MPR. Digitally assessed PathR strongly correlated with visual measurements. CONCLUSIONS: Artificial intelligence-powered digital pathology exhibits promise in assisting pathologic assessments in neoadjuvant NSCLC clinical trials. The development of artificial intelligence-powered approaches in clinical settings may aid pathologists in clinical operations, including routine PathR assessments, and subsequently support improved patient care and long-term outcomes.

Validation of PRKCB Immunohistochemistry as a Biomarker for the Diagnosis of Ewing Sarcoma.

Background: Ewing sarcoma (ES) can be confirmed by identifying the EWSR1-FLI1 fusion transcript. This study is to investigate whether immunostaining (IHC) of PRKCB-a protein directly regulated by EWSR1-FLI1 is a surrogate maker for diagnosing ES in routine practice. Methods: Microarray gene expression analyses were conducted. RKCB IHC was applied to 69 ES confirmed by morphology and molecular methods, and 41 non-Ewing small round cell tumors. EWSR1 rearrangement, EWSR1-FLI1 fusion or t(11;22)(q24;q12) were identified by fluorescence in situ hybridization, reverse transcriptase polymerase chain reaction, or cytogenetic analysis, respectively. Results: Gene array analyses showed significant overexpression of the PRKCB in ES. PRKCB IHC was positive in 19 cases of ES with EWSR1-FLI1 fusion, 3 cases with cytogenetic 11:22 translocation and 59 cases with EWSR1 rearrangement while negative in only one EWSR1 rearranged case. PRKCB IHC is sensitive (98%) and specific (96%) in detecting EWSR1 rearranged ES. Conclusions: PRKCB is a reliable antibody for diagnosing ES in routine practice.

Frontline pembrolizumab for the treatment of RET-rearranged non-small cell lung cancer: A case report.

RET-rearrangement occurs in 1 to 2% of patients with non-small cell lung cancer (NSCLC), typically among non-smokers, and it is associated with an increased incidence of brain metastasis. While selpercatinib and pralsetinib are active for RET-rearranged NSCLC, the optimal standard frontline regimen for this clinical setting remains undefined. Here, we report on a patient with RET-rearranged NSCLC who received frontline pembrolizumab and achieved a complete tumor response lasting for 29 months and ongoing. Single-agent ICI can be an effective frontline treatment for RET-rearranged NSCLC.

Round Robin Evaluation of MET Protein Expression in Lung Adenocarcinomas Improves Interobserver Concordance.

INTRODUCTION: Overexpression of the mesenchymal-epithelial transition (MET) receptor, a receptor tyrosine kinase, can propel the growth of cancer cells and portends poor prognoses for patients with lung cancer. Evaluation of MET by immunohistochemistry is challenging, with MET protein overexpression varying from 20% to 80% between lung cancer cohorts. Clinical trials using MET protein expression to select patients have also reported a wide range of positivity rates and outcomes. MATERIALS AND METHODS: To overcome this variability, the Lung Cancer Mutation Consortium Pathologist Panel endeavored to standardize the evaluation of MET protein expression with "Round Robin" conferences. This panel used randomly selected Aperio-scanned formalin-fixed paraffin-embedded lung cancer specimens stained by MET immunohistochemistry for the Lung Cancer Mutation Consortium 2.0 study (N=838). Seven pathologists in separate laboratories scored images of 5 initial cases and 2 subsequent rounds of 39 cases. The pathologists' scores were compared for consistency using the intraclass correlation coefficient. Issues affecting reproducibility were discussed in Round Robin conferences between rounds, and steps were taken to improve scoring consistency, such as sharing reference materials and example images. RESULTS: The overall group intraclass correlation coefficient comparing the consistency of scoring improved from 0.50 (95% confidence interval, 0.37-0.64) for the first scoring round to 0.74 (95% confidence interval, 0.64-0.83) for the second round. DISCUSSION: We found that the consistency of MET immunohistochemistry scoring is improved by continuous training and communication between pathologists.

Primary bronchial Ewing sarcoma.

INTRODUCTION: Primary bronchial Ewing sarcoma (ES) is a rare endobronchial tumor. PRESENTATION OF CASE: A 65-year-old male presented with six-month history of progressive shortness of breath. Flexible bronchoscopy showed an endobronchial polypoid tumor in the left main stem bronchus about 2 cm from the carina. The tumor was resected by a left bronchial sleeve resection using a right postero-lateral thoracotomy approach. Pathology showed complete tumor resection with negative margins. The morphological and immune-phenotypical features of the resected specimen were compatible with ES. He had an uneventful post-operative recovery. He did not receive adjuvant radiation or chemotherapy and remains disease free at 9 months follow up. DISCUSSION: A review of the literature identified six other cases of primary bronchial ES. In addition, there were three reported cases of primary ES involving the trachea and thirteen involving the lung parenchyma. Bronchial ES appeared to have a relatively better prognosis than ES involving the trachea or the lung. Our case demonstrates that primary bronchial ES may be treated safely with limited resection, lung preservation and without the need for adjuvant therapy if negative margins can be achieved. CONCLUSION: Sleeve resection without adjuvant therapy may be a safe treatment option for primary bronchial ES.

Clinical Activity of Nivolumab for Human Papilloma Virus-Related Juvenile-Onset Recurrent Respiratory Papillomatosis.

BACKGROUND: Juvenile-onset recurrent respiratory papillomatosis (JO-RRP) is a human papilloma virus-mediated progressive benign neoplasm that affects children and young adults. Primary management consists of regular surgical debulking to maintain airway patency and vocal function. Like condyloma acuminata, JO-RRP is associated with immune dysregulation, and T cells isolated from papillomas express an anergic phenotype. Therefore, we hypothesized that programmed death protein 1 axis inhibition could stabilize tumor growth. MATERIALS AND METHODS: We treated two patients with refractory JO-RRP using nivolumab, with the primary objective of assessing clinical activity. We explored baseline papilloma features using immunohistochemistry and comprehensive genomic profiling. RESULTS: Both patients experienced symptomatic improvement, and interval laryngoscopies revealed a reduction in papillomatosis burden. One patient has not required subsequent surgical debridement for almost 2 years. On pathologic examination of pretreatment papillomas from both cases, infiltrating T cells were evident in the papilloma stroma, and papilloma programmed death ligand 1 expression was absent. Papilloma mutational load ranged between three and six mutations per megabase for each case. From on-treatment biopsy tissue, a higher amount of intraepithelial T cells and programmed death ligand 1 expression were detected in the papilloma. CONCLUSION: Nivolumab appears to have promising activity in JO-RRP, and further clinical investigation with more patients in clinical trials is warranted. IMPLICATIONS FOR PRACTICE: To the authors' knowledge, this article is the first report describing clinical activity with a programed cell death-1 (PD-1) inhibitor to treat a rare but detrimental type of respiratory tract epithelial neoplasm that afflicts young adults. Two patients were treated, and tumor features, such as mutational load, were examined with the intent to stimulate future hypotheses for translational research. The safety and activity of PD-1 inhibitors in this population still need to be corroborated in clinical trials and should not yet be adopted into clinical practice.

Clinical associations of mucin 1 in human lung cancer and precancerous lesions.

Mucin 1 (MUC1) is a cell membrane glycoprotein overexpressed in non-small cell lung cancer (NSCLC) and has been implicated in carcinogenesis of premalignant lung lesions. Thus, MUC1 has been a target of interest for vaccine strategies for lung cancer treatment and prevention. Here, we assessed MUC1 expression by immunohistochemistry using tumor samples from patients with biopsy-proven NSCLC. Levels of expression in areas of dysplasia, metaplasia, adenocarcinoma in situ, and carcinoma within the same tissue sample were characterized independently on a scale of 0-3 for paired comparison. We also assessed clinical data for correlations with MUC1 expression. Our analysis included 16 samples from patients with squamous lesions and 19 from patients with adenocarcinoma lesions. Among squamous lesions, MUC1 expression score was significantly increased in dysplastic compared with metaplastic areas (mean difference = 0.83, 95% confidence interval [CI], 0.21-infinity; P = 0.021). MUC1 expression was also increased among areas of squamous cell carcinoma versus dysplastic areas (mean difference = 0.44, 95% CI, -0.006-infinity; P = 0.052). In the adenocarcinoma lesions, MUC1 expression was increased in adenocarcinoma versus adenocarcinoma in situ, although not significantly (mean difference = 0.20, 95% CI, -0.055-infinity; P = 0.094). The increase in MUC1 expression with the progression of premalignant lung lesions to invasive carcinoma in patients with NSCLC supports MUC1 as a possible therapeutic target for the prevention and treatment of lung cancer.

A phase I/randomized phase II study of GM.CD40L vaccine in combination with CCL21 in patients with advanced lung adenocarcinoma.

The GM.CD40L vaccine, which recruits and activates dendritic cells, migrates to lymph nodes, activating T cells and leading to systemic tumor cell killing. When combined with the CCL21 chemokine, which recruits T cells and enhances T-cell responses, additive effects have been demonstrated in non-small cell lung cancer mouse models. Here, we compared GM.CD40L versus GM.CD40L plus CCL21 (GM.CD40L.CCL21) in lung adenocarcinoma patients with ≥ 1 line of treatment. In this phase I/II randomized trial (NCT01433172), patients received intradermal vaccines every 14 days (3 doses) and then monthly (3 doses). A two-stage minimax design was used. During phase I, no dose-limiting toxicities were shown in three patients who received GM.CD40L.CCL21. During phase II, of evaluable patients, 5/33 patients (15.2%) randomized for GM.DCD40L (p = .023) and 3/32 patients (9.4%) randomized for GM.DCD40L.CCL21 (p = .20) showed 6-month progression-free survival. Median overall survival was 9.3 versus 9.5 months with GM.DCD40L versus GM.DCD40L.CCL21 (95% CI 0.70-2.25; p = .44). For GM.CD40L versus GM.CD40L.CCL21, the most common treatment-related adverse events (TRAEs) were grade 1/2 injection site reaction (51.4% versus 61.1%) and grade 1/2 fatigue (35.1% versus 47.2%). Grade 1 immune-mediated TRAEs were isolated to skin. No patients showed evidence of pseudo-progression or immune-related TRAEs of grade 1 or greater of pneumonitis, endocrinopathy, or colitis, and none discontinued treatment due to toxicity. Although we found no significant associations between vaccine immunogenicity and outcomes, in limited biopsies, one patient treated with GMCD40L.CCL21 displayed abundant tumor-infiltrating lymphocytes. This possible effectiveness warrants further investigation of GM.CD40L in combination approaches.

Evaluation of invasive breast cancer samples using a 12-chemokine gene expression score: correlation with clinical outcomes.

BACKGROUND: A unique 12-chemokine gene expression score (CS) accurately predicted the presence of tumor-localized, ectopic lymph node-like structures (TL-ELNs) and improved overall survival (OS) in primary colorectal cancer and metastatic melanoma. We analyzed the correlation between CS, clinicopathological variables, molecular data, and 366 survival in Moffitt Cancer Center's Total Cancer Care (TCC) patients with non-metastatic breast cancer. METHODS: Affymetrix gene expression profiles were used to interrogate the CS by the principal component method. Breast tumors were classified as high or low score based on median split, and correlations between clinicopathologic variables, PAM50 molecular subtype, and ELN formation were analyzed using the TCC dataset. Differences in overall survival (OS) and recurrence-free survival (RFS) in the larger KM Plot breast cancer public datasets were compared using Kaplan-Meier curves. RESULTS: We divided the Total Cancer Care (TCC) breast cancer patients into two groups of high or low CS. Mean CS was 0.24 (range, 2.2-2.1). Patients with higher CS were more likely to be white (172 vs. 159; p = 0.03), had poorly differentiated tumors (112 vs. 59; p <0.0001), ER/PR negative (41 vs. 26) and HER2 positive (36 vs. 19; p = 0.001), and contain TL-ELNs. Higher CS scores were also seen in the basal and HER2+ molecular subtypes. In the KM Plot breast cancer datasets higher CS patients demonstrated superior OS (HR = 0.73, p = 0.008) and RFS (HR 0.76, p = <0.0001), especially in basal and HER2+ patients. CONCLUSIONS: High CS breast tumors tend to be higher grade, basal or HER2+, and present more frequently in Caucasians. However, this group of patients also shows the presence of TL-ELNs within the tumor microenvironment and has better survival outcomes. The CS is a novel tool that can identify breast cancer patients with tumors of a unique intratumoral immune composition and better prognosis. Whether or not the CS is a predictive response marker in breast cancer patients undergoing immunotherapy remains to be determined.

NUT Midline Carcinoma: A Rare Malignancy.

Nuclear protein of the testis (NUT) midline carcinoma can present in the head, neck, and mediastinum. In general, it presents in young adult men and has a poor prognosis. We report on a case of NUT midline carcinoma of the mediastinum in a man 27 years of age without any prior malignancy. Due to the location of the tumor, mediastinal lymphoma and germ cell tumor were initially considered; however, immunohistochemistry was performed using NUT antibody that revealed it to be NUT midline carcinoma. Although guidelines exist for squamous cell carcinoma of the head, neck, and mediastinum, no such specific guidelines are available for NUT midline carcinoma, which looks morphologically similar to squamous cell carcinoma but behaves more aggressively and carries a poor prognosis.

Cell-surface marker discovery for lung cancer.

Lung cancer is the leading cause of cancer deaths in the United States. Novel lung cancer targeted therapeutic and molecular imaging agents are needed to improve outcomes and enable personalized care. Since these agents typically cannot cross the plasma membrane while carrying cytotoxic payload or imaging contrast, discovery of cell-surface targets is a necessary initial step. Herein, we report the discovery and characterization of lung cancer cell-surface markers for use in development of targeted agents. To identify putative cell-surface markers, existing microarray gene expression data from patient specimens were analyzed to select markers with differential expression in lung cancer compared to normal lung. Greater than 200 putative cell-surface markers were identified as being overexpressed in lung cancers. Ten cell-surface markers (CA9, CA12, CXorf61, DSG3, FAT2, GPR87, KISS1R, LYPD3, SLC7A11 and TMPRSS4) were selected based on differential mRNA expression in lung tumors vs. non-neoplastic lung samples and other normal tissues, and other considerations involving known biology and targeting moieties. Protein expression was confirmed by immunohistochemistry (IHC) staining and scoring of patient tumor and normal tissue samples. As further validation, marker expression was determined in lung cancer cell lines using microarray data and Kaplan-Meier survival analyses were performed for each of the markers using patient clinical data. High expression for six of the markers (CA9, CA12, CXorf61, GPR87, LYPD3, and SLC7A11) was significantly associated with worse survival. These markers should be useful for the development of novel targeted imaging probes or therapeutics for use in personalized care of lung cancer patients.

Extensive plastic bronchitis: etiology of a rare condition.

We present the case of a 55-year-old man who developed extensive occlusive bronchial casts after trachea-esophageal (TE) fistula repair. The bronchial casts were treated by bronchoscopic extraction, high dose steroids, antibiotics, and antifungals. Despite this multi-modality treatment, the rapid formation of these occlusive bronchial casts was very aggressive and could not be controlled even with a series of five rigid bronchoscopic extractions within a 48-hour period. The patient quickly deteriorated and succumbed to the inflammatory state. The multiple factors that might have led to the patient's bronchial cast formation are discussed.

Necessity of Microdissecting Different Tumor Components in Pulmonary Tumor Pyrosequencing.

Microdissection is a useful method in tissue sampling prior to molecular testing. Tumor heterogeneity imposes new challenges for tissue sampling. Different microdissecting methods have been employed in face of such challenge. We improved our microdissection method by separately microdissecting the morphologically different tumor components. This improvement helped the pyrosequencing data analysis of two specimens. One specimen consisted of both adenocarcinoma and neuroendocrine components. When both tumor components were sequenced together for KRAS (Kirsten rat sarcoma viral oncogene homolog) gene mutations, the resulting pyrogram indicated that it was not a wild type, suggesting that it contained KRAS mutation. However, the pyrogram did not match any KRAS mutations and a conclusion could not be reached. After microdissecting and testing the adenocarcinoma and neuroendocrine components separately, it was found that the adenocarcinoma was positive for KRAS G12C mutation and the neuroendocrine component was positive for KRAS G12D mutation. The second specimen consisted of two morphologically different tumor nodules. When microdissected and sequenced separately, one nodule was positive for BRAF (v-raf murine sarcoma viral oncogene homolog B1) V600E and the other nodule was wild type at the BRAF codon 600. These examples demonstrate that it is necessary to microdissect morphologically different tumor components for pyrosequencing.

Assessment of cholecystokinin 2 receptor (CCK2R) in neoplastic tissue.

The expression of cholecystokinin 2 receptor (CCK2R, CCKBR or gastrin receptor) has been reported on a diverse range of cancers such as colorectal, liver, lung, pancreatic, ovarian, stomach, thyroid and numerous neuroendocrine/carcinoid tumors. Some cancers of the colorectum, lung, pancreas and thyroid have been shown to overexpress CCK2R in relation to normal matched tissues of the same organ. This reported overexpression has led to the development of a number of CCK2R-ligand targeted imaging and therapeutic agents. However, no comprehensive study comparing the expression of CCK2R in multiple cancers to multiple normal tissues has been performed. Herein, we report the immunohistochemical analysis of cancer samples from gastrointestinal stromal tumor (GIST), hepatocellular carcinoma (HCC), non-small cell lung cancer (NSCLC), pancreatic adenocarcinoma, and thyroid cancer against multiple normal tissue samples from esophagus, liver, lung, pancreas, stomach, spleen and thyroid. These results show that CCK2R expression is present in nearly all cancer and normal samples tested and that none of the cancer samples had expression that was statistically greater than that of all of the normal samples.

Advances in EGFR as a Predictive Marker in Lung Adenocarcinoma.

BACKGROUND: Worldwide, lung cancer is the most common cause of mortality. Toxins from tobacco smoke are known to increase the risk of lung cancer; however, up to 15% of lung cancer-related deaths in men and up to 50% of lung cancer-related deaths in women occur in people who do not smoke. Despite the fact that chemotherapy generally provides a survival benefit for non-small-cell lung cancer, not every patient will respond to therapy and many experience therapy-related adverse events. Thus, predictive markers are used to determine which patients are more likely to respond to a given regimen. METHODS: We reviewed the current medical literature in English relating to predictive markers that may be positive, such as the presence of an activating EGFR mutation. RESULTS: The advances in using EGFR as a molecular predictive marker were summarized. This biomarker influences therapeutic response in patients with lung adenocarcinoma. Clinical evidence supporting its value is also reviewed. CONCLUSIONS: The use of EGFR as a predictive factor in lung adenocarcinoma may help target therapy to individual tumors to achieve the best likelihood for long-term survival and to avoid adverse events from medications unlikely to be effective.