The role of phosphatidylcholine transporters such as Stard7 in intestinal cancer development is unknown. To explore this issue, we generated a mouse model lacking Stard7 in intestinal epithelial cells (IECs). Loss of Stard7 impaired mitochondrial Complex I activity, led to a severe metabolic and lipid reprogramming, enhanced mitochondrial ROS production and potentiated an mTORC1/ATF4 signature. As a result, levels of enzymes involved in serine biosynthesis were enhanced in Stard7-deficient IECs. We next assessed the consequences of Stard7 deficiency in both Wnt-dependent tumor initiation and in inflammation-driven tumor development. Strikingly, despite generating similar molecular signatures, Stard7 deficiency inhibited tumor development in Azoxymethane (AOM)/Dextran Sulfate Sodium (DSS)-treated mice but promoted Wnt-driven cancer initiation in the intestine. Apc+/Min mice lacking Stard7 in IECs developed more tumors in the distal colon as well as a specific microbiota signature. Collectively, our results suggest that the genetic status critically controls the effects of Stard7 deficiency on intestinal tumor development.

MET exon 14 skipping mutations in non-small-cell lung cancer (NSCLC) are important biomarkers for targeted therapy, making accurate detection essential. This study analyzed 379 NSCLCs with mutations in MET exon 14 and adjacent splice sites using DNA- and RNA-based next-generation sequencing (NGS). The 379 samples contained 171 distinct mutations around MET exon 14, highlighting the diversity; 114 mutations were analyzed with a DNA- and an RNA-based NGS assay. Two large deletions and one synonymous splice variant causing exon 14 skipping were only detected by RNA-based NGS. Seven single-nucleotide variations in MET exon 14 did not induce skipping. A total of 57 variants could not be analyzed by RNA-based NGS because of insufficient material or RNA quality; among them, 18 were previously reported as skipping mutations, 30 were splice-site insertions/deletions, and 9 remained unclassified. Additionally, data from the first multinational external quality assessment schemes for MET exon 14 skipping mutation testing in formalin-fixed, paraffin-embedded tissue and liquid biopsies, organized by the lead panel institute and Quality in Pathology GmbH, are presented. High success rates (98%) for tissue are shown, whereas liquid biopsy tests had lower rates (37.5% in 2022, and 63% in 2024), primarily because of low allelic fractions and intronic deletions. This study highlights the importance of analyzing both DNA and RNA, urging improvements in sensitivity, coverage, and bioinformatics.

Objective: This study investigates how a polygenic risk score (PRS) influences colorectal cancer (CRC) risk across clinically and molecularly defined risk groups. Methods: In total, 1839 European-descendant individuals were stratified according to low (<20%), intermediate (20% to 80%), or high (>80%) PRS, based on 93 CRC-associated single-nucleotide polymorphisms (SNPs), for 4 high-risk groups: (i) Lynch syndrome (LS; with CRC: n = 679, CRC-free carriers: n =422); (ii) early-onset sporadic CRC (EOS-CRC; n = 518); (iii) positive family history for CRC (F-CRC; n = 220); and, in EOS-CRC and F-CRC patients, (iv) MSI/dMMR CRC (n =144) vs MSS/pMMR CRC (n =485). CRC risk was compared with population-based controls (n = 3119) and late-onset sporadic CRC patients from UK Biobank (LOS-CRC; n = 781) using multivariable logistic regression and Cox models. Results: Polygenic risk score (PRS) was significantly increased in all risk groups compared with population controls. Being in the high PRS category doubled CRC risk in EOS-CRC and F-CRC corresponding to cumulative incidences before 50 and 75 years of 24% and 13%, respectively. Polygenic risk score was significantly higher in EOS-CRC than LOS-CRC. In LS, PRS in non-CRC carriers lay between CRC-LS and population controls. Non-LS individuals with MSI/dMMR tumors showed significantly lower PRS than those with pMMR/MSS tumors, but no difference compared with LS CRC individuals. Conclusions: Polygenic risk score most strongly influences CRC risk in unexplained EOS-and F-CRC. The effect in LS individuals strongly depends on the penetrance of the altered gene and study design. Nonsignificant trends can partly be explained by the sample size of subgroups. Larger collaborative, prospective studies are needed to validate PRS for personalized CRC risk stratification.

Abstract Background Artificial intelligence has significantly advanced computational pathology by enabling high-resolution, clinical-grade tumor segmentation models with state-of-the-art diagnostic accuracy. Creating such models is resource-intensive, requiring substantial time and domain expertise. Additionally, deep learning models are typically restricted to single tumor types, making it challenging to develop separate models for each tumor type. Cross-cancer generalization of segmentation models could address this bottleneck and pave the way for pan-cancer segmentation models. Methods We evaluated the cross-tumor generalization capability of five tissue segmentation models (breast, colon, lung, kidney, prostate) using 21 cancer types from The Cancer Genome Atlas, totaling over 7,700 whole slide images. Representative large tumor and benign regions were manually selected, and segmentation accuracy was evaluated using a semiquantitative scale (0-10). Results Here we show that the lung model demonstrates excellent cross-cancer performance (overall mean score 7.9 ± 2.1), effectively segmenting tumor regions in many non-lung cancers with segmentation accuracy similar to its native domain in 11 of 19 other epithelial tumors and melanoma, achieving particularly strong results in ovarian cancer (9.2 ± 0.9). The breast and colon models also show strong cross-domain performance, while the kidney and prostate models exhibit more limited generalization. Overall, high-precision segmentation is achievable in most cancer types using existing models. Conclusions Existing segmentation models generalize across multiple cancer types, reducing the need to develop new, entity-specific models from scratch. This cross-domain generalization enables fast-track model development and supports future creation of robust pan-cancer segmentation models. Leveraging these capabilities could accelerate clinical integration of pathology artificial intelligence tools and enable reproducible biomarker discovery.

MTAP protein loss is a surrogate for homozygous CDKN2A deletion and marks a therapeutically actionable, PRMT5-dependent subgroup of pancreatic ductal adenocarcinoma (PDAC), but its tumor microenvironment (TME) remains uncharacterized. In 370 surgically resected PDAC from the prospective multicentric PANCALYZE cohort, MTAP status was determined by immunohistochemistry and confirmed by CDKN2A fluorescence in situ hybridization in MTAP-deficient cases. Immune and stromal cancer-associated fibroblast (CAF) markers were quantified by digital image analysis. Subtype-associated markers were scored by pathologists. MTAP loss occurred in 39.5% of PDAC and showed complete concordance with homozygous CDKN2A deletion. Patients with MTAP deficiency were significantly younger and at higher lymph-node stage. MTAP-loss PDAC displayed a distinct stromal and immune profile, with higher SMA + and FAP + and lower PDGFRβ + CAFs and reduced natural killer-cell and plasma-cell infiltration, not explained by basal-like subtype. MTAP status was not prognostic (median overall survival [OS] 22.0 months in both groups). Within MTAP-deficient PDAC, Claudin 18.2 positivity was associated with longer OS (p = 0.025), whereas high γH2AX (p = 0.003) and high M2-like macrophage infiltration (p = 0.007) were associated with shorter OS. Only high γH2AX remained independent on multivariable analysis (p = 0.028). MTAP loss thus marks a myofibroblastic, immune-attenuated TME, warranting consideration in translational studies of MTAP-directed therapies.

BackgroundSoft tissue sarcomas (STS) carry a high risk of relapse or metastasis even after complete resection. (Neo-)adjuvant chemotherapy benefits only a subset of patients, underscoring the need for predictive biomarkers. Schlafen 11 (SLFN11) has emerged as a marker of sensitivity to DNA-damaging agents. This study evaluated SLFN11 as a prognostic and predictive biomarker for (neo-)adjuvant chemotherapy in STS.Materials and methodsSLFN11 expression was assessed by immunohistochemistry in 242 patients with STS across different disease stages, using the H-score and percentage of positive tumor cells. Sub-cohorts included patients receiving neoadjuvant therapy (n = 33), primary resection (n = 193), palliative first-line chemotherapy (n = 26), or a palliative salvage therapy with trabectedin (n = 22). Associations between SLFN11 levels, clinicopathological features, and survival were analyzed.ResultsIn the neoadjuvant cohort, SLFN11 expression correlated with pathological tumor regression after chemotherapy alone (rho = 0.73, p = 0.016) and chemotherapy ± radiotherapy (rho = 0.62, p = 0.011). Among primarily resected STS treated with adjuvant chemotherapy ± radiotherapy, SLFN11-high tumors were associated with significantly longer overall survival (OS) (p = 0.007) and disease-free survival (DFS) (p = 0.022). SLFN11 was independently associated with improved outcome (OS: HR 0.06, p = 0.002; DFS: HR 0.08, p = 0.004). In the palliative first-line chemotherapy cohort, SLFN11-high tumors showed improved OS (p = 0.005) and progression-free survival (PFS) (p = 0.024), and SLFN11 remained independently predictive (OS: HR 0.11, p = 0.001). In the trabectedin cohort, SLFN11-high tumors demonstrated longer OS (p = 0.04) and PFS (p = 0.024).ConclusionSLFN11 is a prognostic and potentially predictive biomarker in STS in the context of chemotherapy. Our results support a prospective validation, standardization of SLFN11 assessment, and consecutive clinical implementation.

Künstliche Intelligenz (KI) wird in der digitalen Pathologie zunehmend eingesetzt, um die Gewebeklassifizierung, Zelldetektion, Biomarkerquantifizierung, das Grading und die Vorhersage klinisch relevanter molekularer Veränderungen zu unterstützen. Aktuelle KI-Anwendungen in der Pathologie bei internistischen und onkologischen Erkrankungen werden zusammengefasst, mit Schwerpunkt auf den wichtigsten algorithmischen Ansätzen, repräsentativen diagnostischen Anwendungsfällen und derzeitigen Grenzen der klinischen Implementierung. Die vorliegende narrative Übersicht über jüngste Fortschritte beschreibt wesentliche Modelltypen, darunter Gewebesegmentierungssysteme, Algorithmen zur Erkennung einzelner Zellen, schwach überwachtes Lernen und Foundation-Modelle, Tools zur Auswertung der Immunhistochemie sowie multimodale oder Sprache-basierte basierte Anwendungen. Repräsentative Studien aus der Nieren‑, Leber‑, Lungen-, gastrointestinalen, hämatologischen und Schilddrüsenpathologie werden besprochen. In vielen Situationen verbessert KI die Reproduzierbarkeit, Objektivität und Effizienz. Einige Systeme erreichen eine Genauigkeit, die mit der von Experten vergleichbar ist. Die meisten KI-Tools befinden sich noch in der Validierungsphase, und nur wenige sind in die routinemäßige klinische Nutzung übergegangen. Künstliche Intelligenz könnte die diagnostische und prädiktive Histopathologie weiter optimieren. Ihre Rolle bleibt assistierend. Eine breitere Implementierung wird durch verschiedene Hürden und Engpässe begrenzt. Der zukünftige Fortschritt hängt von prospektiver Validierung und der Integration in routinemäßige digitale Arbeitsabläufe ab.

ABSTRACT:Diffuse large B-cell lymphoma (DLBCL) is a highly heterogeneous malignant disease that remains a major clinical challenge, as relapsed and refractory disease is difficult to treat. Apoptosis evasion is a major feature of DLBCL. However, while the suppression of intrinsic apoptosis has long been recognized as a lymphoma-promoting event, the role of extrinsic apoptosis has remained poorly defined. In this study, we demonstrated at the genetic level that expression of cellular Fas-associated death domain protein-like IL-1β-converting enzyme-inhibitory protein (cFLIP), the most crucial, non-redundant inhibitor of extrinsic apoptosis, in B cells is necessary for the development of DLBCL in an autochthonous murine model. Indeed, B-cell-specific deletion of Cflar, the gene encoding for cFLIP, prevented lymphomagenesis mediated by oncogenic Myd88 and overexpression of BCL2. In human lymphoma cells, we showed that the absence of cFLIP sensitized activated B-cell-like (ABC)- but not germinal center B-cell-like (GCB)-DLBCL subtype cells to TRAIL- or lipopolysaccharide-induced, caspase-8-mediated apoptosis. Furthermore, we unveiled a cell death-independent role of cFLIP in the suppression of proinflammatory cytokines at the transcriptional level, selectively in the ABC subtype. These results indicate that the suppression of intrinsic apoptosis can support lymphomagenesis only if extrinsic apoptosis is properly controlled. Moreover, licensing extrinsic apoptosis through CFLAR deletion can efficiently promote death in DLBCL cells, despite the suppression of the intrinsic pathway. Overall, these data provide a rationale for the development of cFLIP inhibitors for the treatment of ABC-DLBCL and possibly other hematological cancers.

Integrated diagnostics is an emerging multidisciplinary model of care that integrates traditionally siloed domains, including pathology, radiology, genomics, and laboratory medicine, into a coordinated framework designed to improve diagnostic accuracy, treatment selection, and clinical outcomes. In this position statement, the EFLM Integrated Diagnostics Committee proposes a concise and operational definition of integrated diagnostics and outlines the relevance for healthcare systems. Integrated diagnostics has the potential to advance precision medicine through improved data integration, more coordinated interpretation, and enhanced clinical decision-making. Drawing on multidisciplinary expert consensus, we recommend adoption of a standardized conceptual framework to support research, clinical implementation, interoperability, and policy development across healthcare systems with implementation adapted to local needs, scale, and digital maturity.

Maintaining DNA damage response (DDR) protein homeostasis is vital for genome integrity. This requires coordinated crosstalk between repair factors, such as RAD51 and MRE11, and the ubiquitin-proteasome system. Here, we identified a new node in the DDR network – ubiquilin-1 (UBQLN1), a member of the ubiquilin protein family that act as shuttles to transport ubiquitinated substrates to the proteasome. We observed that the loss of UBQLN1 exposes RAD51 to ubiquitination and degradation, thus functionally repressing homologous recombination (HR). Mechanistically, we showed that UBQLN1 exhibits a non-canonical function whereby it stabilizes RAD51 localized to the sites of DNA damage by preventing its ubiquitination. Intriguingly, overexpression of UBQLN1 repressed HR-mediated DSB repair by engaging ubiquitinated MRE11, shifting repair towards mutagenic single-strand annealing. We also noted frequent overexpression of UBQLN1 in human tumors, and its association with an aggressive phenotype. Using a Ubqln1 overexpression allele, we demonstrated an oncogenic role of Ubqln1 in lung adenocarcinoma resulting in increased tumor mutational burden, in vivo. Lastly, we showed that Ubqln1 overexpression causes lung adenocarcinomas to sensitize to PARP1 inhibition and immune-checkpoint blockade in vivo. We also identified human patients with a deleterious mutation in UBQLN1 that results in a phenotype consistent with the role of UBQLN1 in HR. Taken together, our study of UBQLN1 reveals that this ubiquilin constitutes an important node in the DDR network, in part by inducing ubiquitin-independent RAD51 stabilization. Furthermore, UBQLN1 overexpression can serve as a predictor of clinical sensitivity to PARP1 inhibition and immune-checkpoint blockade, which has important implications for cancer patient treatment.

Abstract Esophageal adenocarcinoma belong to the cancer entities with a sharply increasing incidence. Though multimodal therapies have improved outcomes, prognoses remain poor in advanced stages and the 5-year survival rate is only between 20 to 25%. For this purpose, new therapeutic options are urgently necessary. In this context, this project investigated a SIX1-specific PROTAC (ELX19) in EAC - including potential adverse effects on the immune system. SIX1 is an embryonic transcription factor, which is required during embryogenesis for the development of various organs and tissues. In contrast, it is no longer or only weakly expressed in adult tissue. However, many cancers show a reactivation of cancer, which leads to different pro-tumorigenic features, making SIX1 a potentially therapeutic target. We first tested the efficiency of ELX19 in degrading SIX1 via WB and ELISA. Thereby, a higher SIX1 expression correlated with a better degradation efficiency, leading to a DC50 of below 1 µM in FLO-1 cells. ELX19 also reduced proliferation and increased apoptosis in FLO-1 cells. Importantly, ELX19 did not affect proliferation and apoptosis in a cell line with low SIX1 expression (OE19) suggesting that the effects caused by ELX19 are SIX1-specific. As adverse effects on the immune system can be a major impediment for further clinical development, we tested the effect of ELX19 on primary T helper and cytotoxic T cells. Thereby, unspecific effects only started 3 µM which is common for VHL-based PROTACs. In conclusion, these preliminary results show the potential of a SIX1-targeting PROTAC for the treatment of esophageal adenocarcinoma. Citation Format: Asad Faili, Vic Scharfenberger, Ellen Weber, Samaneh Heydarzadeh, Reinhard Buettner, Tristan Lerbs. Development of a SIX1-specific PROTAC for the treatment of esophageal adenocarcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 3076.

Accurate prediction of biochemical recurrence (BCR) after radical prostatectomy is critical for guiding adjuvant treatment and surveillance decisions in prostate cancer. However, existing clinicopathological risk models reduce complex morphology to relatively coarse descriptors, leaving substantial prognostic information embedded in routine histopathology underexplored. We present a deep learning-based biomarker that predicts continuous, patient-specific risk of BCR directly from H E-stained whole-slide prostatectomy specimens. Trained end-to-end on time-to-event outcomes and evaluated across four independent international cohorts, our model demonstrates robust generalization across institutions and patient populations. When integrated with the CAPRA-S clinical risk score, the deep learning risk score consistently improved discrimination for BCR, increasing concordance indices from 0.725-0.772 to 0.749-0.788 across cohorts. To support clinical interpretability, outcome-grounded analyses revealed subtle histomorphological patterns associated with recurrence risk that are not captured by conventional clinicopathological risk scores. This multicohort study demonstrates that deep learning applied to routine prostate histopathology can deliver reproducible and clinically generalizable biomarkers that augment postoperative risk stratification, with potential to support personalized management of prostate cancer in real-world clinical settings.

Artificial intelligence has advanced cancer pathology, but many systems still depend on hand-crafted features, are hard to explain and rely on fragmented workflows. We introduce SPARK (System of Pathology Agents for Research and Knowledge), a foundational agentic artificial intelligence approach that uses language as a universal interface to autonomously generate biologically driven concepts for tumor analysis. SPARK turns biological ideas into analytical tools and works directly with complex pathology data without extra model training. We evaluated SPARK across 18 patient cohorts spanning five cancer types (lung adenocarcinoma, lung squamous cell carcinoma, colorectal cancer, breast cancer and oropharyngeal squamous cell carcinoma) and more than 5,400 patients with available histopathology images and clinical/follow-up information, in both prognostic and predictive settings and on a well characterized spatial biology breast cancer dataset (patient n = 625). We found that SPARK produced clinically and biologically relevant concepts correlated with prognosis, known pathological variables and predictive biomarkers, including patterns of tumor progression and temporal change inferred from static images. A dedicated module allows for human interaction with SPARK. Further prospective validation is needed to evaluate the clinical utility of the tools created by SPARK. All code, parameters and results are openly released to help researchers and clinicians improve diagnostic precision and deepen tumor biology insights.

4083 Background: Anti-PD-(L)1 is effective in esophageal adenocarcinoma (EGA). The phase II RICE trial (NCT04159974) evaluated the addition of neoadjuvant and adjuvant durvalumab, with or without tremelimumab, to neoadjuvant chemoradiotherapy (CROSS). Methods: Patients with locally advanced (≥uT3/Nx or uT2/N+) non-metastatic EGA received two doses of neoadjuvant durvalumab in addition to CROSS, followed by surgery. Adjuvant therapy was randomized 1:1 to durvalumab (12 doses) alone (arm A) or durvalumab plus one dose of tremelimumab (arm B). Translational analyses included HLA-I genotyping, immunohistochemistry, whole-exome sequencing and RNA-sequencing. Clinical outcomes were analyzed in patients who received at least one neoadjuvant dose (mITT) and in those who were randomized and received at least one adjuvant dose (R-mITT). Results: In the mITT set (n=56), 95% completed neoadjuvant therapy and 93% underwent resection. The rate of grade ≥3 non-hematologic adverse events during neoadjuvant treatment was 25% (13/56 G3, 1/56 G4). Surgery was feasible with a rate of anastomotic leakage of 6%. Major pathological response (MPR, <10% viable tumor cells) was achieved in 54% (30/56) including a 23% (13/56) pathological complete response rate (pCR, ypT0/ypN0). The rate of MPR was higher than in propensity-based comparisons with FLOT (25%) and CROSS (37%). Two-year overall (OS) and progression-free (PFS) survival of the mITT set were 72.7% and 53.1%, respectively. After completion of Stage I of the trial (safety run in), 38 patients were included in the R-mITT set. 80% in arm A (durvalumab monotherapy, n=20) and 44% in arm B (durvalumab plus tremelimumab, n=18) completed at least 80% of adjuvant immunotherapy. Combined immunotherapy in the R-mITT set was associated with higher grade ≥3 toxicity (22% G3/ 22% G4 in Arm B vs. 15% G3 / 0% G4 in arm A) and did not improve survival (2-year OS/PFS 95.0%/80.0% [arm A] vs. 82.4%/54.2% [arm B]). Tumor mutational burden, heterozygosity of HLA-I, expression of genes related to antigen presentation and T-cell abundance were associated with pathological response. However, discordant cases across all biomarkers underscore the complexity of tumor-immune interactions underlying response to immunotherapy. Conclusions: RICE demonstrates safety and feasibility of adding anti-PD-L1 to neoadjuvant CROSS in EGA and shows promising response rates. While CROSS is no longer considered standard of care, our results support efficacy of durvalumab, consistent with results for FLOT plus durvalumab in the MATTERHORN trial. Addition of tremelimumab to durvalumab is not supported by our study. Clinical trial information: NCT04159974 .

Abstract Dysregulation of chromatin remodeling is a key driver of malignant progression. The SWI/SNF ATPase subunits SMARCA2 and SMARCA4 are essential for chromatin dynamics, yet the clinicopathological and microenvironmental landscape of SMARCA2/4-deficient esophageal adenocarcinoma (EAC) remains insufficiently defined. We analyzed 722 resected EACs from a large Western cohort. SMARCA2 and SMARCA4 status were assessed by immunohistochemistry, and tumors were classified as SMARCA-intact or SMARCA-deficient (complete loss of nuclear expression in tumor cells with internal controls). Molecular co-alterations were evaluated by immunohistochemistry and fluorescence in situ hybridization, including amplifications of MET , ERBB2 (HER2) , EGFR , PIK3CA , MYC , MDM2 , TERT , and Y-chromosome loss (LOY). Several markers were available from prior works. Digital pathology workflows quantified CAF markers (SMA, PDGFRβ, FAP, Periostin, Tenascin) and immune infiltrates (including CD4, CD8, FOXP3, CD20, MUM1, mast cell tryptase). Overall survival was examined using Kaplan–Meier estimates and Cox regression models. SMARCA2/4-deficient tumors accounted for 11.2% of the cohort (81/722) and were enriched among patients aged ≥ 65 years. SMARCA-deficient EACs showed a significantly higher frequency of MET amplification (16.9%), particularly after neoadjuvant therapy. Within SMARCA-deficient tumors, PDGFRβ-positive CAFs, increased plasma cell (MUM1+) and mast cell infiltrates correlated with a favorable outcome, whereas loss of Y-chromosome (LOY) identified an adverse-risk subgroup with particularly poor prognosis. This largest-to-date study defines SMARCA-deficient EAC as a distinct subtype characterized by frequent MET amplification and high-risk interaction with LOY, alongside prognostically relevant stromal-immune features, supporting refined biomarker-based risk stratification and therapeutic exploration.

Purpose: KRAS G12V is among the most frequent KRAS mutations in non-small cell lung cancer (NSCLC), yet its clinical and molecular features remain poorly understood. Experimental Design: In this retrospective study, we analyzed 636 patients with KRAS G12V-mutated NSCLC diagnosed between 2018 and 2023. Clinical, pathological, and molecular characteristics, including co-mutations, smoking history, PD-L1 expression, CD8+ T-cell infiltration, and treatment outcomes, were assessed. Results: The majority of patients (94.2%) were current or former smokers, with a median tobacco exposure of 40 pack-years. Co-mutations were frequent, most commonly in TP53 (40.2%), STK11 (30.2%), and KEAP1 (29.3%). Heavy smokers exhibited significantly higher PD-L1 expression and more frequent TP53, KEAP1, and NTRK1–3mutations than light smokers. CD8+ T-cell infiltration showed a non-significant trend toward higher values in G12V compared to non-G12V KRAS subtypes. Among 151 patients with advanced disease, those treated with immune checkpoint blockade (ICB) alone or in combination with chemotherapy had significantly higher response rates and improved real-world progression-free (rwPFS) and overall survival (rwOS) compared to chemotherapy alone. In patients with PD-L1 TPS ≥50%, ICB-based treatment achieved a median rwOS of 30 months. Conclusions: KRAS G12V-mutated NSCLC is characterized by a strong association with tobacco use, high co-mutation rates in clinically relevant genes, and a favorable response to PD-L1-based immunotherapy. The observed mutation landscape supports the potential for dual checkpoint blockade in a significant subset.

Background: Targeted therapies have transformed advanced non-small-cell lung cancer (aNSCLC), but whether structured clinical networks deliver survival gains at population level remains unclear. In 2010, the Cologne Network Genomic Medicine (NGM) Lung Cancer implemented comprehensive sequencing, expert interpretation, and personalised treatment recommendations into routine care. We evaluated its long-term effect on overall survival (OS). Methods: In a regional matched historical cohort study, patients with aNSCLC diagnosed Jan 2014–March 2019 and receiving first-line (1L) systemic therapy were followed to Dec 2023. NGM participants were linked to German statutory health insurance (AOK) claims and matched by age, sex, and treatment year to patients treated outside the network. OS was compared by Kaplan–Meier and Cox regression. Findings: Among 666 NGM and 3,330 matched non-NGM patients, NGM participation conferred a significant survival benefit (median OS 1.02 vs 0.65 years; HR 0.74, 95% CI 0.68–0.81; p<0.001), with one-year and five-year OS of 50.6%/11.4% vs 35.9%/7.1%. NGM patients more often received 1L tyrosine kinase inhibitors (TKI; 5.9% vs 3.5%; p=0.004) and immune checkpoint inhibitor-based therapy (9.9% vs 6.4%; p=0.001). Among TKI-treated patients, median OS more than doubled (2.35 vs 1.07 years; p=0.008). Interpretation: Transferring precision oncology from a single specialised centre into an entire region delivered a substantial population-level survival benefit persisting over years, with higher use of targeted and immune therapies. As the earliest and longest-observed implementation, the Cologne NGM provided a blueprint for implementing precision oncology in routine care.

The rapid advancement of vision-language models (VLMs) has accelerated progress in computational pathology; however, whole-slide image (WSI)-based pathology report generation remains limited by the scarcity of large-scale WSI–report datasets and the complexity of mapping spatially distributed visual patterns to structured clinical text. To address this, we introduce a clinically curated Pan-Asia WSI–report dataset of approximately 10,500 pairs from five institutions and establish the REG 2025 benchmark through a MICCAI challenge for systematic evaluation of multimodal models. We analyze submitted methods spanning pretrained VLMs, multiple-instance learning frameworks, hierarchical expert models, retrieval-augmented generation, and cross-modal Transformers. Rather than indicating that VLM use alone was sufficient for superior performance, the results suggest that top-performing methods benefited from structured report representations, hierarchical diagnostic decomposition, and effective multimodal grounding. We identify key limitations, including instability in quantitative attribute estimation (e.g., numeric hallucination) and a tendency toward diagnostic overspecification, with some errors resembling known diagnostic pitfalls in routine pathology. These findings establish REG 2025 as a benchmark for evaluating WSI-based structured report generation and vision-language understanding in computational pathology, providing insights for the design of clinically grounded multimodal pathology models.

Die massive/massebildende duktuläre Reaktion (MDR) der Leber ist eine Form ausgedehnten parenchymatösen Remodellings unter der Bedingung fokaler relativer Ischämie. Diese pseudotumoröse Reaktion des Leberparenchyms wird durch eine lokal ausgedehnte Okklusion hepatovenöser, portovenöser und sinusoidaler Blutgefäße hervorgerufen. Oft ist eine angioinvasive Obliteration durch ein koinzidentes Malignom die Ursache. Aber auch ein Auftreten im nichtmalignen Kontext in einer Leberzirrhose oder nach ausgedehnten Lebernekrosen ist möglich. Nicht immer gelingt die sichere präinterventionelle Einordnung, sodass eine leberbioptische Abgrenzung notwendig ist.