JCO Clinical Cancer Informatics

Introduction: Manual data extraction from unstructured clinical notes is labor-intensive and impractical for large-scale clinical and research operations. Existing automated approaches typically require large language models, dedicated computational infrastructure, and/or task-specific fine-tuning that depends on curated data. The objective of this study is to enable accurate extraction with smaller locally deployed models using a disease-site specific pipeline and prompt configuration that are optimized and reusable. Materials/Methods: We developed OncoRAG, a four-phase pipeline that (1) generates feature-specific search terms via ontology enrichment, (2) constructs a clinical knowledge graph from notes using biomedical named entity recognition, (3) retrieves relevant context using graph-diffusion reranking, and (4) extracts features via structured prompts. We ran OncoRAG using Microsoft Phi-3-medium-instruct (14B parameters), a midsize language model deployed locally via Ollama. The pipeline was applied to three cohorts: triple-negative breast cancer (TNBC; npatients=104, nfeatures=42; primary development), recurrent high-grade glioma (RiCi; npatients=191, nfeatures=19; cross-lingual validation in German), and MIMIC-IV (npatients=100, nfeatures=10; external testing). Downstream task utility was assessed by comparing survival models for 3-year progression-free survival built from automatically extracted versus manually curated features. Results: The pipeline achieved mean F1 scores of 0.80 +/- 0.07 (TNBC; npatients=44, nfeatures=42), 0.79 +/- 0.12 (RiCi; npatients=61, nfeatures=19), and 0.84 +/- 0.06 (MIMIC-IV; npatients=100, nfeatures=10) on test sets under the automatic configuration. Compared to direct LLM prompting and naive RAG baselines, OncoRAG improved the mean F1-score by 0.19 to 0.22 and 0.17 to 0.19, respectively. Manual configuration refinement further improved the F1-score to 0.83 (TNBC) and 0.81 (RiCi), with no change in MIMIC-IV. Extraction time averaged 1.7-1.9 seconds per feature with the 14B model. Substituting a smaller 3.8B model reduced extraction time by 57%, with a decrease in F1-score (0.03-0.10). For TNBC, the extraction time was reduced from approximately two weeks of manual abstraction to under 2.5 hours. In an exploratory survival analysis, models using automatically extracted features showed a comparable C-index to those with manual curation (0.77 vs 0.76; 12 events). Conclusions: OncoRAG, deployed locally using a mid-size language model, achieved accurate feature extraction from multilingual oncology notes without fine-tuning. It was validated against manual extraction for both retrieval accuracy and survival model development. This locally deployable approach, which requires no external data sharing, addresses a critical bottleneck in scalable oncology research.

Medical oncology education faces a dual crisis: knowledge velocity that outpaces static curricula and large language model (LLM) risks--hallucination and automation bias--that threaten the fidelity of AI-assisted learning. We present Onco-Shikshak V7, an AI-native adaptive learning platform that addresses both challenges through a unified cognitive architecture grounded in learning science. The system replaces isolated educational modules with four authentic clinical workflows--Morning Report, Tumor Board, Clinic Day, and AI Textbook--each scaffolded by a nine-module pedagogy engine that integrates ACT-R activation dynamics (illness scripts), Item Response Theory (adaptive difficulty), the Free Spaced Repetition Scheduler (FSRS v4), Zone of Proximal Development (scaffolding), and metacognitive calibration training (Brier score). Six specialist AI agents--medical oncology, radiation oncology, surgical oncology, pathology, radiology, and oncology navigation--engage in multi-disciplinary deliberation with per-specialty retrieval-augmented generation (RAG) grounding across nine authoritative guideline sources including NCCN, ESMO, and ASTRO. The platform provides 18 clinical cases with decision trees across six cancer types, maps every interaction to 13 ACGME Hematology-Oncology milestones, and implements four closed-loop feedback mechanisms that connect session errors to targeted flashcards, weak domains to suggested cases, and all interactions to a persistent learner profile. Technical validation confirms algorithmic correctness across eight subsystems. To our knowledge, this is the first system to unify ACT-R, IRT, FSRS, ZPD, and metacognitive calibration in a single medical education platform. Formal learner evaluation via randomized controlled trial is planned.

Immune-related adverse events (irAEs) affect up to 40% of patients receiving immune checkpoint inhibitors, yet their identification depends on laborious and inconsistent manual chart review. Here we developed and evaluated an agentic large language model system to extract the presence, temporality, severity grade, attribution, and certainty of six irAE types from clinical notes. Retrospectively (263 notes), the system achieved macro-averaged F1 of 0.92 for detection and 0.66 for multi-class severity grading; self-consistency improved F1 by 0.14. The best-performing configuration cost approximately $0.02 per note. In prospective silent deployment over three months (884 notes), detection F1 was 0.72-0.79. In a randomized crossover study of clinical trial staff (17 participants, 316 observations), agentic assistance reduced annotation time by 40% (P < 0.001), increased complete-match accuracy (OR 1.45; 95% CI 1.01-2.09; P = 0.045), and improved inter-annotator agreement (Krippendorffs from 0.22-0.51 to 0.82-0.85). These results demonstrate that agentic AI coupled with human verification could enhance efficiency, performance, and consistency for irAE assessment.

Purpose: Tumor genomic testing (TGT) is standard-of-care for most patients with advanced/metastatic cancer. Despite established guidelines, patient education prior to TGT is frequently omitted. The purpose of this study was to evaluate the impact and durability of a concise 3-4 minute video for patient education prior to TGT in community versus academic sites and across cancer types. Patients and Methods: Patients undergoing standard-of-care TGT were enrolled at a tertiary academic institution in three cohorts: Cohort 1-breast cancer; Cohort 2-lung cancer; Cohort 3-other cancers. Cohort 4 consisted of patients with any cancer type similarly undergoing SOC TGT at one of three community cancer centers. Participants completed survey measures prior to video viewing (T1), immediately post-viewing (T2), and after return of TGT results (T3). Outcome measures included: 1) 10-question objective genomic knowledge/understanding (GKU); 2) 10-question video message-specific knowledge (VMSK); 3) 11-question Trust in Physician/Provider (TIPP); 4) perceptions regarding TGT. Results: A total of 203 participants completed all survey timepoints. Higher baseline GKU and VMSK scores were significantly associated with higher income and greater years of education. For the primary objective, there was a significant and sustained improvement in VMSK from T1:T2:T3 (Poverall p<0.0001), with no significant change in GKU (p=0.41) or TIPP (p=0.73). This trend was consistent within each cohort (all p[&le;]0.0001). Results for four VMSK questions significantly improved, including impact on treatment decisions, incidental germline findings, and insurance coverage of testing. Conclusions: A concise, 3-4 minute, broadly applicable educational video administered prior to TGT significantly and sustainably improved video message-specific knowledge in diverse cancer types and in academic and community settings. This resource is publicly available at http://www.tumor-testing.com, with a goal to efficiently educate and empower patients regarding TGT while addressing guidelines within the flow of clinical practice.

Multi-cancer early detection (MCED) tests can detect several cancer types and stages. We previously developed a methylation and protein (MP V1) MCED classifier. In this study, we present a refined MP V2 classifier, developed by evaluating model architectures that improved performance in prospectively enrolled case-control cohorts under standard testing conditions. The newly developed MP V2 classifier was trained to be more generalizable and achieve increased early-stage sensitivity at a target specificity of [&ge;]97.0%. MP V1 and MP V2 classifier performances were compared using a previously described test set, and MP V2 performance was also evaluated in a new independent clinical validation set. Compared to MP V1, the MP V2 classifier demonstrated a 7.3% increase in overall sensitivity, with sensitivity increases of 7.6%, 9.2%, and 8.3% for stages I, II, and stages I/II, respectively, in the intended use (breast and prostate cancers excluded) test set. In an independent validation intended use set, the MP V2 classifier showed an overall sensitivity of 55.6%, with sensitivities of 26.8%, 42.9%, and 34.8% for stages I, II, and stages I/II, respectively. In a case-control setting, the MP V2 classifier offered improved sensitivity for early-stage cancers at a lower specificity target.

PurposeLarge language models (LLMs) are used for biomedical text processing, but individual decisions are often hard to audit. We evaluated whether enforcing a mechanically checkable "show your work" quote affects accuracy, stability, and verifiability for trial eligibility-scope classification from abstracts. MethodsWe used 200 oncology randomized controlled trials (2005 - 2023) and provided models with only the title and abstract. Trials were labeled with whether they allowed for the inclusion of patients with localized and/or metastatic disease. Three flagship models (GPT-5.2, Gemini 3 Flash, Claude Opus 4.5) were queried with default settings in two independent conditions: label-only and label plus a verbatim supporting quote. Models could abstain if they deemed the abstract to not contain sufficient information. Each condition was repeated three times per abstract. Quotes were mechanically validated as exact substrings after whitespace normalization, and a separate judge step used an LLM to rate whether each quote supported the assigned label. ResultsEvidence requirements modestly reduced coverage (GPT-5.2 86.2% to 84.3%, Gemini 98.3% to 92.8%, Claude 96.0% to 94.5%) by increasing abstentions and, for Gemini, invalid outputs. Conditional macro-F1 remained high but changed by model (slight gains for GPT-5.2 and Gemini, decrease for Claude). Labels were stable across repetitions (Fleiss kappa 0.829 to 0.969). Mechanically valid quotes occurred in 83.3% to 91.2% of runs, yet only 48.0% to 78.8% of evidence-bearing predictions were judged semantically supported. Restricting to supported predictions increased macro-F1 at the cost of lower coverage. ConclusionSubstring-verifiable quotes provide an automated audit trail and enable selective, higher-trust automation when applying LLMs to biomedical text processing. However, this approach introduces new failure modes and trades coverage for verifiability in a model-dependent way.

Guideline-based recommendations for metastatic lines of therapy (mLoTs), especially second lines and beyond, are comparatively sparse due to challenges in later-line treatment efficacy quantification. Scalable real-world evidence that captures the interaction between treatment and disease progression is therefore especially valuable, as regimens become increasingly individualized, confounding intensifies, and progression is rarely recorded as a structured EHR endpoint. We present a framework to (i) reconstruct clinically coherent mLoTs from longitudinal EHR using radiology-anchored progression evidence and (ii) generate individualized progression-free survival (PFS) estimates from a line-start multimodal snapshot in a highly heterogeneous cohort. In 2,881 patients contributing 8,791 metastatic mLoTs, the selected model shows strong discrimination over a 2-year horizon (Antolinis C = 0.680 {+/-} 0.006; cumulative/dynamic AUC at 1 year = 0.824 {+/-} 0.006). Predicted risk strata closely track Kaplan-Meier trends across line number and tumor subtypes, enabling calibrated risk stratification even in smaller sub-cohorts. Model prediction primarily relies on clinically plausible signals of recent metastatic burden and tumor markers, with limited dependence on surveillance cadence or subtype labels, and is robust to missingness. Together, this framework supports scalable evidence generation and interpretable, calibrated prognostication to inform risk assessment and care planning in heterogeneous metastatic practice.

IntroductionThe precise determination of diagnostic cut-off points is essential for the development of multimarker panels in oncology. In previous work on pulmonary nodules, it was observed that the standard two-parameter logistic fit could be insufficient for biomarkers with asymmetric distributions. Furthermore, the calculation of empirical cut-off points based on graphical visualization presented limitations in precision and reproducibility. ObjectiveThis study presents a methodological advancement in the data analysis phase (Stage 1), introducing new Python algorithms for the direct analytical calculation of empirical intersections and robust mathematical modeling using Dual Annealing with both two-parameter and four-parameter logistic functions. This improved methodology feeds into the ThresholdXpert 1.0 software tool for combinatorial optimization of biomarker panels (Stage 2), and is applied here to the diagnostic challenge of hepatocellular carcinoma (HCC). MethodsThe methodology was first validated by re-analyzing a dataset of patients with pulmonary nodules (N=895). It was subsequently applied to an HCC dataset derived from the cohort of Jang et al. (208 HCC, 193 cirrhosis, 401 total), randomly divided into a training set (280) and an independent test set (121). Scripts were developed to compare the previous two-parameter logistic fit with the new two- and four-parameter logistic models. Finally, ThresholdXpert 1.0 was used for multimarker panel optimization. ResultsThe integration of empirical calculation, logistic modeling, and combinatorial optimization through ThresholdXpert 1.0 provides a robust and coherent framework for the development of multimarker diagnostic panels. The four-parameter logistic model provided additional validation without substantially modifying cut-off values for most biomarkers, confirming the stability of the approach while offering greater flexibility for complex distributions. When applied to hepatocellular carcinoma, the framework identified a molecular panel composed of AFP, PIVKA-II, OPN, and DKK-1 with sensitivity of 0.77 and specificity of 0.72, and an optimized panel incorporating inverse MELD that achieved the best overall balance (sensitivity 0.73, specificity 0.75) in independent external validation. These results demonstrate the potential of this approach as a generalizable tool for the optimized design of binary diagnostic systems in oncology. ConclusionThe integration of complementary mathematical modeling enhances the capability of ThresholdXpert 1.0 to identify robust diagnostic panels, as in some cases a single biomarker may outperform biomarker combinations, and vice versa. This approach enabled the integration of molecular biomarkers and clinical variables under a unified mathematical framework. Contactroberto117343@gmail.com

BackgroundHealthcare utilization forecasting systems are often derived from static, annualized market share assumptions that fail to represent real-world treatment dynamics. Such approaches systematically misestimate future utilization by ignoring longitudinal treatment sequencing, discontinuation with surveillance, recurrence-driven re-entry, and provider adoption dynamics. ObjectiveThis study proposes a reusable, governance-driven health informatics forecasting framework designed to generate realistic utilization forecasts using real-world data by integrating longitudinal patient-flow modeling, persistence-based exposure estimation, provider behavioral adoption, and multi-source calibration into a single architecture. MethodsLongitudinal U.S. administrative claims data representing oncology treatment populations (approximately 80,000 treated patients annually across therapy lines) were curated through a governance layer that refines diagnosis and treatment pools, infers clinically valid lines of therapy, and corrects for lookback-limited recurrence bias. Patients were modeled as transitioning across explicit clinical states, including treatment initiation, sequential therapy lines, discontinuation, surveillance, and recurrence-driven re-entry. Forecast outputs were calibrated using volume-weighted and behaviorally dampened provider adoption dynamics derived from primary research and claims-revealed utilization and evaluated against static share-based forecasts under identical peak-share assumptions. ResultsAcross multiple oncology contexts, longitudinal patient-flow-based forecasting recovered approximately 50-70% more cumulative treated months than static approaches. Underestimation in traditional models was driven primarily by failure to capture later-line persistence, surveillance exit, and re-treatment dynamics. Setting-specific calibration revealed earlier adoption in academic centers and slower, payer-constrained uptake in community practices. ConclusionsThe proposed framework demonstrates a forecast-oriented health informatics architecture that improves utilization estimation and decision support in complex, longitudinal care ecosystems. The methodology generalizes across tumor types and chronic conditions characterized by treatment sequencing, persistence variability, and relapse-driven re-entry.

BackgroundOnce the treatment starts, early prediction of treatment benefit and its correlation with overall survival (OS) remains challenging in metastatic castration-resistant prostate cancer (mCRPC). Existing prognostic models require long-term follow-up, limiting their ability to inform timely treatment decisions. To address this gap, we evaluated tumor growth rate (g-rate)-based survival models across multiple treatment lines to assess their ability to predict OS and support early clinical decision-making. MethodsWe developed GxSurv, a Random Survival Forest (RSF)-based framework that incorporates baseline clinical variables and g-rate calculated from serial on-treatment PSA, to construct line-specific prediction models of OS, a direct measure of treatment outcome. Three variants were developed: G3Surv, using the 3-month g-rate; G6Surv, using the 6-month g-rate; and GfSurv, using the final observed g-rate. Model performance was evaluated using Harrells C-index, Unos C-index, Integrated Brier Score (IBS), time-dependent area under the curve (tAUC). Model interpretability was assessed using permutation importance to quantify predictor contributions within the GxSurv framework. FindingsThe study included 15912 treatment records from 11014 patients with mCPRC across four lines of therapy. We found that incorporation of g-rate consistently improved model performance across all treatment lines, with all GxSurv models outperforming Cox proportional hazards (CoxPH). As the earliest prognostic model, our G3Surv demonstrated strong early predictive performance, with Harrells C-index values ranging from 0{middle dot}700 to 0{middle dot}746 and tAUC values of 0{middle dot}766 to 0{middle dot}822 across all lines, representing 5-8% and 4-5% improvements over CoxPH, respectively. These results indicate that G3Surv accurately predicts individual treatment outcomes at 3 months after treatment initiation. Feature importance analyses consistently identified g-rate as a top predictor, followed by baseline PSA and hemoglobin, with relative variation across treatment lines. InterpretationIntegrating g-rate calculated from on-treatment PSA values enables accurate, line-specific prediction of treatment outcomes in mCRPC, with the 3-month g-rate providing robust early prognostic information to support timely, personalized clinical decision-making. FundingU.S. National Science Foundation, National Institutes of Health, American Cancer Society.

PurposeTo describe the rationale, methods, and baseline sample descriptives of the Adolescent and Young Adult Tracking Engagement and Management Skills (AYA TEAMS) cohort. The AYA TEAMS study is a longitudinal observational cohort study that aims to identify determinants and patterns of self-management and engagement in cancer-related long-term follow-up (LTFU) care and validate a novel transition readiness assessment among adolescent and young adult (AYA) survivors of childhood cancer. ParticipantsAYA survivors of childhood cancer (ages 16-25) and their caregivers were enrolled from 3 large pediatric oncology centers across the United States from 2020-2022 and followed for 2 years (minimum) to 3 years and 3 months (if transferred to adult care). AYA inclusion criteria were: past childhood cancer diagnosis, at least 2 years off-treatment, 5 years since diagnosis, engaged with the participating pediatric health care system within the last 18 months, cognitively able to complete study procedures, and English speaking. AYA completed a comprehensive battery of measures including assessments of self-management and transition readiness at baseline and annually for 2 years. For AYA transferred to adult care, separate measures were administered at the time of transfer (following last pediatric visit) and 15 months post transfer. Caregivers (English or Spanish-speaking) completed a single survey at baseline to capture family functioning, psychosocial risk, and transition readiness. Cancer diagnosis, treatment modalities, treatment-related late effects, and engagement in LTFU care were captured via electronic medical record review. In total, 709 AYA were enrolled and 587 were included in the final cohort [Mage=19.7 years, 52.5% female, 38.2% from racial and/or ethnic minoritized groups, (REMG)]. The cohort was on average 7.3 years old at the time of diagnosis and 10.5 years off treatment. Half (52.5%) were survivors of leukemia/lymphoma, 38.0% solid tumors, and 9.5% central nervous system tumors. Three hundred and ninety-nine caregivers participated (90% mothers). Findings to DateEnrolled AYA excluded from the baseline cohort were more likely to be male, from REMG, and/or to enroll without a caregiver. Baseline cohort differences between sites emerged for age, race and ethnicity, socioeconomic status, and treatment modalities and intensity. Future PlansData collection was completed in April 2025. Findings from this cohort will elucidate important predictors of self-management and engagement in recommended annual LTFU and inform the design of interventions to reduce disengagement in LTFU. Strengths and LimitationsO_LIThis study is the first known prospective cohort of AYA-only long-term survivors of childhood cancer in the United States recruited from pediatric cancer centers. C_LIO_LIThis study achieved high enrollment and retention rates across a medically and demographically diverse sample. C_LIO_LIInformed by multiple theoretical self-management models, this study will be able to examine predictors and transactional relationships of AYA survivor self-management, including engagement in pediatric and adult cancer-related long-term follow-up care. C_LIO_LIReliance on English-speaking AYA and those currently engaged with the health care system are limitations. C_LI

PurposeCastration-resistant prostate cancer (CRPC) is characterized by marked clinical heterogeneity and poor long-term survival, underscoring the need for tools that can rapidly and reliably individualize patient risk. While several prognostic models exist, their complexity has limited routine clinical use. Here, we developed and validated PROGRESS (PROstate cancer Global Risk Evaluation and Stratification Score), a simplified prognostic score, derived through machine learning-guided feature selection, to enhance risk stratification and support individualized, risk-informed clinical decision-making. MethodsPROGRESS was developed using baseline data from 2,035 metastatic CRPC patients enrolled in four different phase III trials. An unsupervised machine-learning approach was applied to identify latent patient subgroups with distinct survival outcomes irrespectively of allocated treatment arm, followed by classical multivariable modelling to derive a simple and straight-forward prognostic score based on routinely available objective laboratory variables. External validation was performed in three independent datasets comprising metastatic CRPC patients treated across different therapeutic settings (n=1,239) and non-metastatic CRPC patients managed with standard care (n=660). Overall survival was assessed using Kaplan-Meier and Cox regression analyses. ResultsUnsupervised modelling identified two patient risk subpopulations with significantly different overall survival rates (median 27.4 vs 17.7 months; hazard ratio [HR] 2.20, 95% CI 1.91-2.54; p<.00001). Feature contribution analysis yielded three independent predictors -PSA, ALP, and AST-used to build PROGRESS. In the training cohort, PROGRESS demonstrated strong discrimination (AUC 0.89). Using a prespecified cut-off, patients classified as increased risk had significantly shorter survival than low-risk patients (median 18.3 vs 25.6 months; HR 1.72, 95% CI 1.50-1.97; p<.0001). PROGRESS prognostic performance was consistent across all validation cohorts, including metastatic and non-metastatic disease, with HRs ranging from 1.74 to 3.46 (all p<.0001). ConclusionsBy integrating machine-learning-based pattern discovery with classical statistical modelling, PROGRESS provides a simple, objective, and clinically accessible approach for individual risk stratification in CRPC. Its reliance on three inexpensive, routinely measured laboratory parameters would facilitate practical implementation in clinical settings, enhancing visibility of underlying disease aggressiveness for individual clinical decision-making. PROGRESS could represent a pragmatic first step toward improving patient selection for clinical trials while identifying regulatory meaningful endpoints achievable in a sizeable patient population; further validation in prospective clinical studies and real-world datasets would allow to confirm its clinical utility and generalizability. PROGRESS can be freely accessed for research use only at the following link: https://dev.ai.topazium.com.

Molecular subtyping of cancer is traditionally defined in transcriptomic space, yet routine clinical deployment is limited by the availability and cost of sequencing. Meanwhile, histopathology captures rich morphological information that is known to correlate with molecular state but lacks a principled, mechanistic bridge to gene-level representations. We propose a graph-constrained learning framework that aligns morphology-derived signals with a fixed, data-driven gene network discovered via hierarchical Monte Carlo screening. We can derive new gene sets for classification using random sampling, and use the coexpression network of that graph to enforce the learning of a pure morphology model without using gene expression. The resulting model performs subtype prediction using morphology alone, while being explicitly forced to operate through a gene-structured latent space. Structural alignment is enforced during training. For Moffitt classification in pancreatic cancer using PANCAN and TCGA datasets, the model has a reported 85% AUC using an alternative gene set network structure, while the alternate gene set itself has an 84% AUC in all patients that were classified with subtyping with pancreatic cancer in the dataset. This demonstrates that virtual transcriptomics can provide biologically grounded molecular insights using only routine histopathology slides, potentially expanding access to precision oncology in resource-limited settings.

BackgroundPancreatic cancer is a leading cause of cancer mortality, and early recognition is challenging. To achieve early diagnosis using symptoms alone, we examined patterns across different stages using network analysis to derive clinically useful insights. MethodsSymptom variables from a de-identified dataset of 50,000 pancreatic cancer patients were analyzed. Stratification by stage was done, followed by bootstrap resampling to address imbalances across strata. Symptom networks were then constructed with nodes representing symptoms and edges representing conditional dependencies estimated via an Ising-style neighborhood selection approach implemented through L1-regularized logistic regression. Strength, betweenness, and closeness centrality indices were then calculated, and their stability was analyzed using the case-dropping bootstrap. Network comparison tests were done, and difference networks were analyzed. Spring-layout algorithm was used for visualization, with node size being the predictability (pseudo-R{superscript 2}), and the edge weight being the mean partial correlation magnitude. ResultsOn average, symptoms were present in about one out of four patients (M = 0.26). Weight loss and abdominal discomfort were the most prevalent of the symptoms, followed by jaundice and back pain. Network structures became sparser across stages with a decreasing number of edges and centrality indices. Jaundice emerged as the dominant hub in Stage I, but shared dominance with Weight Loss in Stage II. Node predictability (pseudo-R2) was effectively zero across all disease stages. ConclusionOur network analysis of pancreatic cancer symptomatology across stages revealed distinct patterns that may improve understanding of its clinical presentation and support earlier recognition.

BackgroundQuality-of-life (QoL) assessment is essential in breast cancer care, yet limited evidence describes how interrelated QoL domains change during pharmacotherapy. This study aimed to evaluate correlations among functional and symptom scales using the EORTC QLQ-C30 and QLQ-BR23, highlighting their ability to reveal multidimensional QoL patterns. MethodsA prospective observational study was conducted in two second-referral hospitals in Indonesia, enrolling 106 female breast cancer patients. QoL was assessed before and after pharmacotherapy using QLQ-C30 and QLQ-BR23. Changes in scores ({Delta}) were computed, and interdomain relationships were analyzed using Spearmans rho. ResultsPhysical functioning correlated with role functioning ({rho} = 0.55, p <0.001), emotial functioning ({rho} = 0.33, p <0.001), and social functioning ({rho} = 0.31, p = 0.002). Role and social functioning were likewise correlated ({rho} = 0.32, p = 0.001), indicating that improvements across functional domains tended to occur in parallel. Symptom scales showed strong positive clustering, including fatigue with pain ({rho} = 0.37, p <0.001), insomnia ({rho} = 0.35, p <0.001), and systemic side effects ({rho} = 0.48, p <0.001). Functional and symptom domains generally exhibited inverse relationships: physical functioning negatively correlated with fatigue ({rho} = -0.40), pain ({rho} = -0.43), both p <0.001, and systemic side effects ({rho} = -0.26; p = 0.01). ConclusionThe QLQ-C30 and QLQ-BR23 instruments effectively captured structured, clinically meaningful interdependencies. Functional improvements consistently aligned with symptom reductions, revealing coherent functional-symptom clustering. These findings underscore the sensitivity of QoL instruments to detect multidimensional patient-reported changes during breast cancer pharmacotherapy.

IntroductionBuilding on our previously published methodology for claims-based pathway mapping, we extended the analysis by incorporating disease staging. The aim of this study was to develop and evaluate quality indicators (QIs) in patients with non-small cell lung cancer (NSCLC). MethodsThis retrospective, longitudinal cohort study spanned 2017-2023, with follow-up data extending to September 2025. Data were obtained from the National Cancer Registry (NCR), the National Registry of Reimbursed Health Services (NRRHS), which is organized through seven health insurance funds providing nationwide coverage. The index date was defined as the date of the first biopsy (BX) followed by a histopathological examination (HP), along with the ICD-10 code C34. Incident patients aged [&ge;]18 years were included if no prior malignancy was reported, and the presence of PET/CT or CT examination was mandatory in the final verified cohort. The presence of multidisciplinary team (MDT) discussion, time to treatment, availability of care in a Complex Oncology Center (COC), and completeness of predictive biomarker testing were considered key QIs. ResultsWe analyzed the care pathways of 15,886 patients with NSCLC; 3,380 (21.3%) were not treated, and 1,837 (11.6%) were excluded due to the absence of (PET) CT prior to biopsy (BX). The final verified cohort included 10,669 patients with a median age of 69 years (interquartile range, 64-74). The incident stage distribution comprised of stage I/II (27.6%), stage III/IV (67.9%), and 4.5% unknown. Multidisciplinary team (MDT) review was reported in 53.9% of patients, with a median time to MDT discussion of 37 days. Surgery (SX) was performed in 81.0% of stage I and 68.4% of stage II patients. Fewer than 50% of patients initiated treatment within 8 weeks, regardless of disease stage. Centralization of care in COCs and implementation of MDT review showed a positive temporal trend, although disparities across disease stages and regions persisted. PD-L1 testing was documented in 70.0% of stage IV and 65.2% of stage III patients. ConclusionsAdministrative claims data linked with the NCR enabled stage-stratified monitoring of NSCLC care pathways and the identification of actionable QIs, which were implemented as a national tool for continuous quality evaluation of cancer care in the Czech Republic. KEY MESSAGESO_ST_ABSWhat is already known on this topicC_ST_ABSO_LIPatient pathway monitoring and quality indicators (QIs) for lung cancer care -- including timeliness of treatment, multidisciplinary team discussion (MDT), and centralization in specialized centers (COCs) -- have been established in several European countries. C_LIO_LIPopulation-level data integrating administrative claims with cancer registry staging data to evaluate QIs across disease stages remain limited. C_LI What this study addsO_LIStage-stratified analysis of 10,669 NSCLC patients revealed that fewer than 50% initiated treatment within 8 weeks, with a declining trend over time despite improvements in MDT utilization and care centralization in COCs. C_LIO_LIPD-L1 testing rates in stages III-IV increased over 2021-2023 but showed substantial regional variability, highlighting opportunities for improving equity of access to biomarker-guided therapy. C_LI How this study might affect research, practice or policyO_LIThe methodology has been implemented as a national tool for continuous quality evaluation of cancer care in the Czech Republic, with PD-L1 testing completeness proposed as an additional OI alongside MDT discussion, time to treatment, and COC centralization. C_LI

We propose a lesion-centric phenotype learning pipeline for interpretable breast ultrasound (BUS). Predicted lesion masks are used for mask-weighted pooling of segmentation-encoder latents, producing compact embeddings that suppress background influence; a lightweight calibration step improves cross-dataset consistency. We cluster embeddings to discover latent phenotypes and relate phenotype structure to morphology descriptors (compactness, boundary sharpness). On BUSI and BUS-UCLM with external testing on BUS-BRA, lesion-centric pooling and calibration improve separability and enable strong malignancy probing (AUC 0.982), outperforming radiomics and a standard CNN baseline. A simple rule-gated generator further improves BI-RADS-style descriptor consistency on difficult cases.

Pancreatic ductal adenocarcinoma (PDAC) is an aggressive malignancy characterized by profound molecular heterogeneity and inconsistent responses to gemcitabine-based therapy. Although KRAS mutations are nearly ubiquitous, the broader RTK-RAS and MAPK signaling networks, and their association with therapeutic response, remain insufficiently characterized. We performed an integrative clinical-genomic study of 184 PDAC tumors, stratified by age at diagnosis and gemcitabine exposure, systematically evaluating somatic alterations within curated RTK-RAS/MAPK gene panels. Conversational artificial intelligence agents (AI-HOPE-RTK-RAS and AI-HOPE-MAPK) were deployed to dynamically construct cohorts and conduct pathway-level analyses, with results subsequently confirmed using conventional statistical approaches. Among late-onset PDAC cases, ERBB2 and RET mutations were significantly enriched in gemcitabine-treated tumors. In early-onset disease, CACNA2D family alterations were more common in untreated tumors, whereas FLNB and TP53 mutations were observed at higher frequencies in treated cases. Notably, late-onset patients who did not receive gemcitabine and lacked RTK-RAS or MAPK pathway alterations demonstrated significantly improved overall survival. These findings identify age- and treatment-specific signaling dependencies extending beyond canonical KRAS alterations and reinforce a precision oncology framework in PDAC. Conversational AI enabled rapid, multidimensional integration of clinical and genomic data, facilitating the identification of clinically meaningful pathway architectures.

IntroductionDespite advancements in non-small cell lung cancer (NSCLC) management through the use of molecular biomarkers, the recently introduced 9th edition of the TNM staging system remains based exclusively on anatomic descriptors, with no consistently demonstrated improvement in risk stratification for early-stage disease. This study explores the integration of a molecular prognostic classifier into the conventional TNM staging system. MethodsWe analyzed 502 patients with stage I-III lung adenocarcinoma (LUAD) who underwent surgical resection with tumor-based gene expression profiling at the Quebec Heart and Lung Institute. A molecular prognostic classifier was developed and integrated into the 9th edition TNM staging system to generate a novel model (TNMEx). Prognostic performance was compared with the 8th and 9th TNM editions using prognostic discrimination and reclassification metrics. External validation of the molecular classifier was performed in 271 LUAD cases from The Cancer Genome Atlas (TCGA). An independent cohort of 606 resected LUAD patients from the National Cancer Center Hospital (Tokyo) was used to externally compare the prognostic performance of the 8th and 9th TNM staging systems in the absence of molecular data. ResultsThe molecular prognostic classifier was developed based on the expression levels of 26 prognosis-associated genes, weighted by their corresponding coefficients. The classifier was subsequently integrated into the 9th edition TNM staging to generate the TNMEx model. The TNMEx system demonstrated superior prognostic performance, achieving a higher concordance index (C-index = 0.72) compared to the 9th edition TNM (C-index = 0.65, p=0.006). Moreover, TNMEx significantly improved patient risk reclassification compared to both the 8th (net reclassification improvement [NRI] = 0.27, integrated discrimination improvement [IDI] = 0.04) and 9th editions (NRI = 0.40, IDI = 0.05), underscoring its superior ability to stratify outcomes. The 8th and 9th editions showed only limited improvement in overall prognostic accuracy and risk stratification, as reflected by their relatively modest C-index values (0.62 and 0.65, respectively) and minimal reclassification gains (NRI = -0.06, IDI = 0.003). ConclusionsIncorporating a molecular-based prognostic model significantly enhanced the ability to recognize patients at high risk and to predict their survival outcomes more accurately than traditional TNM staging systems.

AO_SCPLOWBSTRACTC_SCPLOWLung cancer is characterized by profound intratumoral and inter-patient heterogeneity, spanning histological subtypes, molecular landscapes, and the tumor microenvironment. While multi-omics integration is essential for capturing this complexity, leveraging these data to explicitly define survival-associated subpopulations remains a significant challenge. In this study, we developed NeuroMDAVIS-FS, an unsupervised deep learning framework designed to stratify lung cancer patients by survival risk, and identify molecular determinants underlying improved clinical outcomes. Using the CPTAC cohort, we integrated genomic (CNV), transcriptomic (RNA-seq), and proteomic profiles to extract modality-specific features. Candidate biomarkers were validated through Kaplan- Meier (KM) survival analysis and univariate Cox proportional hazards (CoxPH) regression. A final multivariate CoxPH model effectively stratified patients into high-risk and low-risk cohorts (Kaplan Meier p-value < 0.001). Notably, the integration of these molecular features with baseline clinical models significantly enhanced prognostic accuracy, improving the concordance index by 43.79% in LUAD, 31.05% in LSCC, and 23.76% across the pan-lung cancer cohort. These results demonstrate that NeuroMDAVIS-FS identifies robust, biologically relevant features that surpass traditional clinical variables in predicting patient outcomes, offering a scalable path for precision oncology.