Journal of Pathology Clinical Research最新文献

{"title":"Costs of biomarker testing in advanced non-small cell lung cancer: a global study comparing next-generation sequencing and single-gene testing","authors":"Umberto Malapelle,&nbsp;Chien-Chin Chen,&nbsp;Enrique de Álava,&nbsp;Paul Hofman,&nbsp;Daniel Kazdal,&nbsp;Tae-Jung Kim,&nbsp;Tony Kiat Hon Lim,&nbsp;Aleš Ryška,&nbsp;Angelica A Saetta,&nbsp;Ed Schuuring,&nbsp;Giancarlo Troncone,&nbsp;Michele Biscuola,&nbsp;Yi-Lin Chen,&nbsp;Gek San Tan,&nbsp;Charles Hugo Marquette,&nbsp;Maria Michelli,&nbsp;Arja ter Elst,&nbsp;Hana Vošmiková,&nbsp;Joshua Kapp,&nbsp;Sebastian Gonzalez-McQuire,&nbsp;Andromachi Giannopoulou,&nbsp;Jean Marie Franzini,&nbsp;Victoria Lucia Rabsiun Aramburu,&nbsp;Anna Baggi,&nbsp;Albrecht Stenzinger","doi":"10.1002/2056-4538.70018","DOIUrl":"https://doi.org/10.1002/2056-4538.70018","url":null,"abstract":"<p>Current European/US guidelines recommend that molecular testing in advanced non-small cell lung cancer (aNSCLC) be performed using next-generation sequencing (NGS). However, the global uptake of NGS is limited, largely owing to reimbursement constraints. We compared real-world costs of NGS and single-gene testing (SGT) in nonsquamous aNSCLC. This observational study was conducted across 10 pathology centers in 10 different countries worldwide. Biomarker data collected via structured questionnaires (1 January–31 December 2021) were used to feed micro-costing analyses for three scenarios [‘Starting Point’ (SP; 2021–2022), ‘Current Practice’ (CP; 2023–2024), and ‘Future Horizons’ (FH; 2025–2028)] in both a real-world model, comprising all biomarkers tested by each center, and a standardized model, comprising the same sets of biomarkers across centers. Testing costs (including retesting) encompassed personnel costs, consumables, equipment, and overheads. Overall, 4,491 patients with aNSCLC were evaluated. Mean per-patient costs decreased for NGS relative to SGT over time, with real-world model costs 18% lower for NGS than for SGT in the SP scenario, and 26% lower for NGS than for SGT in the CP scenario. Mean per-biomarker costs also decreased over time for NGS relative to SGT. In the standardized model, the tipping point for the minimum number of biomarkers required for NGS to result in cost savings (per patient) was 10 and 12 in the SP and CP scenarios, respectively. Retesting had a negligible impact on cost analyses, and results were robust to variation in cost parameters. This study provides robust real-world global evidence for cost savings with NGS-based panels over SGT to evaluate predictive biomarkers in nonsquamous aNSCLC when the number of biomarkers to be tested exceeds 10. Widespread adoption of NGS may enable more efficient use of limited healthcare resources.</p>","PeriodicalId":48612,"journal":{"name":"Journal of Pathology Clinical Research","volume":"11 2","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/2056-4538.70018","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143564743","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Oncocytic and spindle cell typical carcinoids of lung: different immunophenotype and biological behavior","authors":"Giovanna Sabella,&nbsp;Giovanni Centonze,&nbsp;Patrick Maisonneuve,&nbsp;Federica Grillo,&nbsp;Vincenzo Lagano,&nbsp;Giovanna Garzone,&nbsp;Carlotta Pardo,&nbsp;Martina Filugelli,&nbsp;Alessia Mietta,&nbsp;Michele Simbolo,&nbsp;Alessandra Fabbri,&nbsp;Alessandro Mangogna,&nbsp;Natalie Prinzi,&nbsp;Sara Pusceddu,&nbsp;Luigi Rolli,&nbsp;Luisa Bercich,&nbsp;Salvatore Grisanti,&nbsp;Mauro Roberto Benvenuti,&nbsp;Ugo Pastorino,&nbsp;Luca Roz,&nbsp;Aldo Scarpa,&nbsp;Alfredo Berruti,&nbsp;Carlo Capella,&nbsp;Massimo Milione","doi":"10.1002/2056-4538.70020","DOIUrl":"https://doi.org/10.1002/2056-4538.70020","url":null,"abstract":"<p>Pulmonary typical carcinoids (TCs) are uncommon, well-differentiated neuroendocrine tumors of the lung that do not exhibit necrosis and have fewer than two mitoses per 2 mm², as defined by the current World Health Organization classifications. Despite their low-grade status and favorable prognostic impact, the protein expression profile and morphological characteristics associated with tumor progression and metastatic spread remain largely unidentified. Oncocytic and spindle cell histological variants are acknowledged for their role in differential diagnosis, though their clinical significance remains a topic of debate. We centrally reviewed a multicenter series of 297 TCs to identify cases of oncocytic and spindle cell variants. We examined associations with clinicopathological features and immunohistochemical markers (orthopedia homeobox protein, thyroid transcription factor 1, mammalian achaete-scute homologue 1, somatostatin receptor 2A, Ki-67, anti-mitochondria, and S100); these data were further related to disease-free survival (DFS), overall survival, and cancer-specific survival (CSS). Our analysis identified oncocytic TCs (<i>n</i> = 36, 12.1%), spindle cell TCs (<i>n</i> = 55, 18.5%), and ordinary TCs, defined as those without either variant or with variants that were not prominent (<i>n</i> = 206, 69.4%). Interestingly, ordinary tumors were associated with a higher number of tumor-related deaths (<i>p</i> = 0.01) compared to the other histological variants. Additionally, patients with spindle cell morphology had longer CSS compared to those with ordinary morphology (<i>p</i> = 0.04). Parameters such as histological variant, age, tumor stage, and Ki-67 were significantly linked to DFS on multivariable analysis, even after accounting for differences between centers. In conclusion, oncocytic, spindle cell, and ordinary TCs are linked to distinct clinicopathological characteristics and exhibit varying clinical outcomes.</p>","PeriodicalId":48612,"journal":{"name":"Journal of Pathology Clinical Research","volume":"11 2","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/2056-4538.70020","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143554574","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhancing HER2 testing in breast cancer: predicting fluorescence in situ hybridization (FISH) scores from immunohistochemistry images via deep learning","authors":"Daniel O Macaulay,&nbsp;Wenchao Han,&nbsp;Mark D Zarella,&nbsp;Chris A Garcia,&nbsp;Thomas E Tavolara","doi":"10.1002/2056-4538.70024","DOIUrl":"https://doi.org/10.1002/2056-4538.70024","url":null,"abstract":"<p>Breast cancer affects millions globally, necessitating precise biomarker testing for effective treatment. HER2 testing is crucial for guiding therapy, particularly with novel antibody-drug conjugates (ADCs) like trastuzumab deruxtecan, which shows promise for breast cancers with low HER2 expression. Current HER2 testing methods, including immunohistochemistry (IHC) and <i>in situ</i> hybridization (ISH), have limitations. IHC, a semi-quantitative assay, is prone to interobserver variability. While ISH provides higher precision than IHC, it remains more resource-intensive in terms of cost and workflow. However, turnaround time is typically faster than that of other advanced molecular methods such as next-generation sequencing. We adapted the clustering-constrained-attention multiple-instance deep learning model to improve IHC testing and reduce dependence on reflex fluorescence ISH (FISH) tests. Using 5,731 HER2 IHC images, including 592 cases with FISH testing, we trained two models: one for predicting HER2 scores from IHC images and another for predicting FISH scores from equivocal cases. The HER2 IHC score prediction model achieved 91% ± 0.01 overall accuracy and a receiver operating characteristic (ROC) area under the curve (AUC) of 0.98 ± 0.01. The FISH score prediction model had an ROC AUC of 0.84 ± 0.07, with sensitivity at 0.37 ± 0.13 and specificity at 0.96 ± 0.03. External validation on cases from 203 institutions showed similar performance. The HER2 IHC model maintained a 91% ± 0.01 accuracy and an ROC AUC of 0.98 ± 0.01, while the FISH model had an ROC AUC of 0.75 ± 0.03, with sensitivity at 0.28 ± 0.04 and specificity at 0.93 ± 0.01. Our model advances HER2 scoring by reducing subjectivity and variability in current scoring methods. Despite lower accuracy and sensitivity in the FISH prediction model, it may be a beneficial option for settings where reflex FISH testing is unavailable or prohibitive. With high specificity, our model can serve as an effective screening tool, enhancing breast cancer diagnosis and treatment selection.</p>","PeriodicalId":48612,"journal":{"name":"Journal of Pathology Clinical Research","volume":"11 2","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/2056-4538.70024","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143564993","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Association between individual Warburg-related proteins and prognosis in colorectal cancer","authors":"Kelly Offermans,&nbsp;Josien CA Jenniskens,&nbsp;Colinda CJM Simons,&nbsp;Iryna Samarska,&nbsp;Gregorio E Fazzi,&nbsp;Kim M Smits,&nbsp;Leo J Schouten,&nbsp;Matty P Weijenberg,&nbsp;Heike I Grabsch,&nbsp;Piet A van den Brandt","doi":"10.1002/2056-4538.70016","DOIUrl":"https://doi.org/10.1002/2056-4538.70016","url":null,"abstract":"<p>We previously showed that Warburg subtyping (low/moderate/high), based on the expression of six glycolytic proteins and transcriptional regulators [glucose transporter 1 (GLUT1), pyruvate kinase M2 (PKM2), lactate dehydrogenase A (LDHA), monocarboxylate transporter 4 (MCT4), p53, and PTEN], holds independent prognostic value in colorectal cancer (CRC) patients. The present study aimed to investigate whether the expression level of one of the proteins (GLUT1, PKM2, LDHA, MCT4, p53, and PTEN) can act as a proxy for our previously identified six protein-based Warburg subtypes. Protein expression levels for individual Warburg-related proteins were available for 2,251 CRC patients from the Netherlands Cohort Study. Kaplan–Meier curves and Cox regression were used to explore associations between individual Warburg-related proteins and CRC-specific and overall survival. Previously identified associations between Warburg subtypes and CRC-specific and overall survival were adjusted for individual proteins, showing a significant association with survival in the current study. Multivariable-adjusted analyses showed that the expression of GLUT1, LDHA, MCT4, PKM2, or p53 was associated with neither CRC-specific nor overall survival. Decreasing PTEN expression was associated with significantly poorer overall survival (<i>p-</i>trend_categories = 0.026). Additional adjustment for PTEN expression had minimal impact on the previously identified association between Warburg subtypes and survival, and the six protein-based Warburg-high subtype remained a statistically significant predictor of overall survival (hazard ratio 1.15; 95% CI 1.01–1.32). In conclusion, our results emphasise that individual Warburg-related proteins cannot serve as a proxy or surrogate marker for Warburg subtyping, thereby highlighting the importance of combining the expression levels of multiple Warburg-related proteins when examining the prognostic significance of a complex biological pathway such as the Warburg effect.</p>","PeriodicalId":48612,"journal":{"name":"Journal of Pathology Clinical Research","volume":"11 2","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/2056-4538.70016","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143513479","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Comprehensive transcriptomic profiling reveals molecular characteristics and biomarkers associated with risk stratification in papillary thyroid carcinoma","authors":"Congcong Yan,&nbsp;Chen Zheng,&nbsp;Jiaxing Luo,&nbsp;Xue Wu,&nbsp;Xinyu Meng,&nbsp;Chaoyue Lv,&nbsp;Shurong Shen,&nbsp;Meng Zhou,&nbsp;Ouchen Wang","doi":"10.1002/2056-4538.70022","DOIUrl":"https://doi.org/10.1002/2056-4538.70022","url":null,"abstract":"<p>Papillary thyroid carcinoma (PTC) is one of the most common endocrine malignancies, with varying levels of risk and clinical behavior. A better understanding of the molecular characteristics could improve molecular diagnosis and risk assessment. In this study, we performed whole transcriptomic sequencing on 113 PTC cases, including 70 high-risk and 43 low-risk Chinese patients. Comparative transcriptional profiling analysis revealed two functionally distinct patterns of gene dysregulation between the risk subtypes. Low-risk PTCs showed significant upregulation of immune-related genes and increased immune cell infiltration, whereas high-risk PTCs presented extensive alterations in gene expression and activation of oncogenic signaling pathways. Additionally, we developed a 31-gene transcriptomic signature (PTCrisk) for differentiating high-risk from low-risk PTCs, which was validated across both in-house and external multicenter cohorts. PTCrisk scores were positively correlated with key clinicopathological features, including tumor size, lymph node metastasis, TNM stage, and <i>BRAF</i> mutation status. Overall, our study provides further molecular insights into PTC risk stratification and may contribute to the development of personalized therapeutic strategies for PTC patients.</p>","PeriodicalId":48612,"journal":{"name":"Journal of Pathology Clinical Research","volume":"11 2","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-02-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/2056-4538.70022","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143489789","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Subtype-associated complexity and prognostic significance of the NLRP3 inflammasome landscape in pancreatic neoplasms","authors":"Kristóf Németh,&nbsp;Eszter Mezei,&nbsp;Justína Vörös,&nbsp;Katalin Borka,&nbsp;Adrián Pesti,&nbsp;István Kenessey,&nbsp;András Kiss,&nbsp;András Budai","doi":"10.1002/2056-4538.70019","DOIUrl":"https://doi.org/10.1002/2056-4538.70019","url":null,"abstract":"<p>Intraductal papillary mucinous neoplasm (IPMN) can progress into malignant pancreatic cancer, posing challenges in accurately assessing the risk of malignancy. While the nucleotide-binding oligomerization domain (NOD)-like receptor pyrin domain containing 3 (NLRP3) inflammasome pathway's role in pancreatic ductal adenocarcinoma (PDAC) has been extensively studied, its implications in IPMN remain unexplored. This study aimed to investigate the prognostic significance of NLRP3 inflammasome-related proteins across IPMN subtypes and their associations with tumor characteristics, with a secondary focus on comparing expression patterns in IPMN and PDAC. A cohort of 187 patients (100 IPMN and 87 PDAC) underwent high-dimensional histopathological imaging using the multiplexed immunohistochemical consecutive staining on single slide method and a semi-automated image analysis workflow. Expression levels of NLRP3, apoptosis-associated speck-like protein containing a caspase-recruitment domain (ASC), caspase-1, interleukin-1 beta, interleukin-18 (IL-18), interleukin-1 receptor antagonist, and interleukin-18 binding protein (IL-18BP) were evaluated and compared between IPMN and PDAC samples. The relationships between protein expression and tumor characteristics were examined. Principal component analysis distinguished between intestinal and nonintestinal clusters based on NLRP3-associated proteins. Lower IL-18 expression was linked to the intestinal subtype, while higher caspase-1 was linked to the pancreatobiliary subtype. Elevated caspase-1 and ASC expression were associated with invasiveness in IPMN. No significant correlation was found between the examined proteins and later-stage tumor characteristics in invasive cases. The IL-18/IL-18BP ratio was an independent prognostic factor in invasive IPMN. Our findings highlight the prognostic significance of IL-18 and the IL-18/IL-18BP ratio in invasive IPMNs. These results point to a complex regulation of NLRP3 inflammasome proteins, especially effector cytokines, in pancreatic neoplasms, which are strongly linked to subtype and prognosis.</p>","PeriodicalId":48612,"journal":{"name":"Journal of Pathology Clinical Research","volume":"11 2","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/2056-4538.70019","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143447109","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Sarcomatoid areas of urothelial carcinoma are enriched for CD163-positive antigen-presenting cells","authors":"Burles A Johnson III,&nbsp;Vamsi Parimi,&nbsp;Sonia Kamanda,&nbsp;David C Corney,&nbsp;Woonyoung Choi,&nbsp;Jean Hoffman-Censits,&nbsp;Max Kates,&nbsp;David J McConkey,&nbsp;Noah M Hahn,&nbsp;Andres Matoso","doi":"10.1002/2056-4538.70021","DOIUrl":"https://doi.org/10.1002/2056-4538.70021","url":null,"abstract":"<p>Sarcomatoid urothelial carcinoma (SUC) is a rare histologic subtype with poor prognosis. While there is known intra-tumoral heterogeneity between individual SUC tumors, the relationship between sarcomatoid and conventional urothelial carcinoma (CUC) within the same patient is poorly understood. The objective of this study was to identify differences between the sarcomatoid and CUC tumor microenvironment components that may drive this aggressive phenotype. Using tissue microarrays from eight patient tumors with mixed CUC and SUC, we examined paired CUC, mixed urothelial carcinoma (UC) regions, and SUC using the Nanostring Digital Spatial Profiling platform. We found SUC and mixed UC had higher levels of stromal cells, predominately macrophages and fibroblasts, when compared with CUC within the same tumor. CD14, CD163, and transforming growth factor-beta levels were significantly higher in SUC than in CUC. Immunohistochemical analysis revealed consistently moderate to strong expression of CD163-positive antigen-presenting cells (APCs) in SUC regions, whereas CD68-positive APC expression was generally absent. Thus, in mixed histology SUC, the SUC component preferentially expressed CD163-positive APCs and fibroblasts compared to the CUC component. As CD163-positive APCs and fibroblasts are known to be tumor-promoting and immune-suppressive, this infiltration may contribute to epithelial to mesenchymal transition and other aggressive properties of SUC.</p>","PeriodicalId":48612,"journal":{"name":"Journal of Pathology Clinical Research","volume":"11 2","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/2056-4538.70021","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143447112","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Platinum drugs upregulate CXCR4 and PD-L1 expression via ROS-dependent pathways, with implications for novel combined treatment in gastric cancer","authors":"Xiaoyu Kang,&nbsp;Lin Zhang,&nbsp;Shushang Liu,&nbsp;Fei Wang,&nbsp;Haiming Liu,&nbsp;Fenli Zhou,&nbsp;Fei Wu,&nbsp;Haohao Zhang,&nbsp;Daiming Fan,&nbsp;Yongzhan Nie,&nbsp;Zhangqian Chen","doi":"10.1002/2056-4538.70015","DOIUrl":"10.1002/2056-4538.70015","url":null,"abstract":"<p>CXC chemokine receptor 4 (CXCR4) and programmed cell death-ligand 1 (PD-L1) are two critical molecules involved in the tumor immune microenvironment. However, the impact of platinum drugs, such as cisplatin, on CXCR4 or PD-L1 expression and the underlying mechanisms in gastric cancer (GC) remain unknown. Moreover, the correlation between their expression levels in GC remains elusive. Immunohistochemistry, western blot, and RT-qPCR were performed to determine the expression pattern of CXCR4 and PD-L1 in GC. Clinical samples, patient-derived xenografts, and cell-derived xenografts were utilized to investigate the effects of platinum drugs on the expression levels of CXCR4 and PD-L1. Postchemotherapy resected GC tumor tissues showed higher CXCR4 and PD-L1 expression levels than pretreatment biopsies (<i>p</i> &lt; 0.05). Similarly, GC xenografts treated with platinum-based chemotherapy exhibited increased CXCR4 and PD-L1 expression levels compared to saline-treated controls (<i>p</i> &lt; 0.05). A positive correlation was detected between the expression levels of CXCR4 and PD-L1 in GC tumor tissues. Increased levels of CXCR4 and PD-L1 expression, in a dose- and time-dependent manner upon cisplatin treatment, were observed in GC cells (<i>p</i> &lt; 0.05). Cisplatin-induced CXCR4 upregulation relies on ROS/HIF-1α and ROS/NF-κB pathways, while cisplatin-induced PD-L1 upregulation is cyclic GMP-AMP synthase/stimulator of IFN genes-dependent and associated with elevated ROS levels in GC cells. CXCR4 expression was found to be positively correlated with PD-L1 expression in GC. Platinum drugs upregulated the levels of CXCR4 and PD-L1 expression in GC. A combined strategy targeting CXCR-4 and PD-L1 might have clinical prospects for GC patients.</p>","PeriodicalId":48612,"journal":{"name":"Journal of Pathology Clinical Research","volume":"11 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11772088/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143053834","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Authors' reply: Re: Koga et al. Retrieval-augmented generation versus document-grounded generation: a key distinction in large language models","authors":"Katherine J Hewitt,&nbsp;Isabella C Wiest,&nbsp;Jakob N Kather","doi":"10.1002/2056-4538.70013","DOIUrl":"10.1002/2056-4538.70013","url":null,"abstract":"&lt;p&gt;We thank Koga &lt;i&gt;et al&lt;/i&gt; for their knowledgeable comments on our work. Their letter highlights a valid question that requires clarification [&lt;span&gt;1&lt;/span&gt;].&lt;/p&gt;&lt;p&gt;Our study assessed the ability of three large language models (LLMs) to diagnose neuropathology cases from free-text descriptions of adult-type diffuse gliomas, for which we compared two methodologies. The first method provided each model with the free-text tumor descriptions alone, while the second approach additionally provided the models with a Word document of the WHO CNS5. We termed these approaches zero-shot and retrieval-augmented generation (RAG), respectively [&lt;span&gt;2&lt;/span&gt;]. Koga &lt;i&gt;et al&lt;/i&gt; point out that the methodology we describe in our paper as RAG, may be better described as document-grounded generation, or in-context learning.&lt;/p&gt;&lt;p&gt;While we agree with the definition of RAG provided in the letter as it was initially defined [&lt;span&gt;3&lt;/span&gt;], the field has evolved significantly since the approach was first proposed by Lewis &lt;i&gt;et al&lt;/i&gt; in 2020. Three paradigms of RAG are now increasingly recognized: naive RAG, advanced RAG, and modular RAG [&lt;span&gt;4&lt;/span&gt;]. Naive RAG is an approach where the data for indexing are generally obtained offline and converted into a format such as PDF or Word, and uploaded with the query via the context window. Advanced RAG and modular RAG offer specific improvements to address the limitations of naive RAG; however, to achieve this, they utilize more technical approaches.&lt;/p&gt;&lt;p&gt;The intention for our paper was to use naive RAG. We chose this approach as it leverages the easiest possible way for improving an LLM response that would be reproducible by doctors, considering that most doctors would be unable to utilize the application programming interface and programmatically build a RAG pipeline. As discussed by Koga &lt;i&gt;et al&lt;/i&gt;, the key difference between naive RAG and document-grounding lies in how the document is utilized when the model retrieves its response [&lt;span&gt;5&lt;/span&gt;]. Document-grounding submits the document with the user query and is equivalent to inserting the entire document text into the context window [&lt;span&gt;5&lt;/span&gt;]. Whereas with naive RAG, relevant parts of the document are identified by the model and used with the query to dynamically search its database [&lt;span&gt;4&lt;/span&gt;]. Both approaches are examples of in-context learning as they acquire additional knowledge from the prompt without requiring parameter updates [&lt;span&gt;6&lt;/span&gt;].&lt;/p&gt;&lt;p&gt;Bereft of transparency from the LLM providers regarding how they process the document once it has been submitted via the graphical user interface, it is difficult to know whether naive RAG or document-grounding was used to formulate a response. To our knowledge, details regarding how appended documents are utilized during a query are not freely available online by ChatGPT, Llama, or Claude. Furthermore, due to the speed of development in the field, technical aspects of how documents are u","PeriodicalId":48612,"journal":{"name":"Journal of Pathology Clinical Research","volume":"11 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11747990/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143014353","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Retrieval-augmented generation versus document-grounded generation: a key distinction in large language models","authors":"Shunsuke Koga,&nbsp;Daisuke Ono,&nbsp;Amrom Obstfeld","doi":"10.1002/2056-4538.70014","DOIUrl":"10.1002/2056-4538.70014","url":null,"abstract":"&lt;p&gt;We read with great interest the article by Hewitt &lt;i&gt;et al&lt;/i&gt;., ‘Large language models as a diagnostic support tool in neuropathology’ [&lt;span&gt;1&lt;/span&gt;]. The authors effectively applied large language models (LLMs) to interpreting the WHO classification of central nervous system tumors; however, we wish to address a technical aspect of their study that warrants clarification.&lt;/p&gt;&lt;p&gt;The authors described their approach as retrieval-augmented generation (RAG). Based on the methods described, the study involved attaching a Word document containing the WHO diagnostic criteria to the prompt to guide its responses. We believe that this approach is more accurately described as document-grounded generation rather than true RAG. Document-grounded generation refers to methods where the model generates outputs explicitly based on a preprovided document, which serves as a static reference [&lt;span&gt;2&lt;/span&gt;]. Unlike RAG, which retrieves information dynamically from external sources [&lt;span&gt;3&lt;/span&gt;], document-grounded generation relies entirely on data embedded in the input prompt at the time of execution. In this study, the WHO criteria were provided with the prompt, which allowed the model to use this information without real-time retrieval. This method is a type of in-context learning, relying on curated contextual data embedded in the input [&lt;span&gt;4&lt;/span&gt;].&lt;/p&gt;&lt;p&gt;Our own work provides an example of in-context learning in a different domain, namely image classification. We evaluated GPT-4 Vision (GPT-4V) for classifying histopathological images stained with tau immunohistochemistry, including neuritic plaques, astrocytic plaques, and tufted astrocytes [&lt;span&gt;5&lt;/span&gt;]. Although GPT-4V initially struggled, few-shot learning with annotated examples, which is a specific application of in-context learning, significantly improved its accuracy, matching that of a convolutional neural network model trained on a larger dataset. These findings demonstrate the utility of in-context learning for both text-based and image-based tasks, with the latter presenting unique challenges for LLMs [&lt;span&gt;6&lt;/span&gt;].&lt;/p&gt;&lt;p&gt;Although in-context learning is an effective approach, it has limitations worth considering. Since this method uses static data that are preloaded data into the prompt, errors can occur if the information is outdated or inaccurate. In-context learning may also lead to overfitting to the given context, limiting the model's ability to generalize to other scenarios. If the contextual data are overly complex, the model might misinterpret the information or fail to generate accurate outputs [&lt;span&gt;7&lt;/span&gt;]. To ensure reliability, it is important to carefully select the input data, update it regularly, and consider these limitations when designing tasks.&lt;/p&gt;&lt;p&gt;In summary, clarifying the differences between RAG, document-grounded generation, and in-context learning is essential, especially for readers less familiar with these concepts. Nonetheless, we support the au","PeriodicalId":48612,"journal":{"name":"Journal of Pathology Clinical Research","volume":"11 1","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11736412/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143014454","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}