Riaz, Irbaz Bin

Abstract Purpose: Inhibitors of the Ataxia telangiectasia and Rad3-related (ATR) protein kinase, a critical component of the DNA damage repair response, have synergistic anti-cancer activity with platinum compounds in preclinical models. We therefore conducted a Phase 2 study of the ATR inhibitor berzosertib+carboplatin vs. docetaxel+carboplatin in metastatic castration-resistant prostate cancer. Patients and Methods: Patients previously treated with at least one androgen receptor pathway inhibitor and taxane underwent mandatory biopsy and were randomized 1:1 to receive Arm A (docetaxel 60 mg/m2+carboplatin AUC4 day 1; or carboplatin AUC5 if not a docetaxel candidate) or Arm B (berzosertib 90 mg/m2 days 2,9+carboplatin AUC5 day 1) every 21 days. The primary endpoint was overall response rate (ORR; ≥50% PSA decline or radiographic response). Results: Of 73 randomized patients, 65 received protocol treatment: 34 on Arm A (26 docetaxel+carboplatin; 8 carboplatin alone) and 31 on Arm B. Thirteen patients (38%) in Arm A and 20 (65%) in Arm B had ≥Grade 3 treatment-related adverse events (TrAEs). ORR was 15% in Arm A (5/34; 5/26[19%] with docetaxel+carboplatin) and 0% in Arm B (0/31). At interim analysis after 65 of planned 130 patients were treated, enrollment was halted due to futility. Homologous recombination repair deficiency status from tissue and blood did not correlate with clinical outcomes, but ATM-deficiency or high levels of phoshpo-KAP1 detected in tissue was associated with clinical benefit across both arms of the trial. Conclusions: Berzosertib+carboplatin demonstrated lower ORR and more frequent ≥grade 3 TrAEs compared to docetaxel+carboplatin in this heavily pre-treated, biomarker-unselected population. NCT03517969

Abstract Background and Objective Management of von Hippel–Lindau (VHL) disease is challenging due to multisystem involvement and limited systemic treatment options. Belzutifan has shown promising efficacy in clinical trials; however, real-world data remain limited. We aimed to evaluate its effectiveness and safety in a real-world cohort. Methods We conducted a retrospective study of 59 patients with VHL disease treated with belzutifan between August 2021 and January 2026. Objective response rates (ORR) for renal cell carcinoma (RCC), central nervous system hemangioblastomas (CNS HB), and pancreatic neuroendocrine tumors (pNET) were assessed by an independent radiologist. Retinal hemangioblastoma outcomes were evaluated through ophthalmology records by chart review. Results Median follow-up was 29 months. ORR was 27.6%, 39.6%, and 85.7% for RCC, CNS HB, and pNET, respectively. Among 13 patients with retinal hemangioblastomas (20 eyes), 17 eyes (85%) showed improvement. At data cutoff, 47 patients (79.6%) remained on treatment, while 12 (20.3%) discontinued (7 due to adverse events, 3 to disease progression, and 2 to other reasons). The most common adverse event was anemia (86.4%). Grade ≥3 events occurred in 10 patients (16.9%). The procedure rate per patient-year decreased from 0.35 in the 3 years prior to treatment to 0.08 during follow-up (Incidence Rate Ratio, 0.23; 95% CI, 0.12–0.47; P < 0.001) Conclusions Belzutifan demonstrated efficacy across VHL-associated lesions, with favorable safety profile. Most adverse events were grade 1–2. A reduction in procedure rate was observed after treatment initiation.

This guideline has been updated. This ASCO Living Guideline is reviewed on an eight-week cycle. Please see the online dynamic version for the latest updates. Living guidelines are developed for selected topic areas with rapidly evolving evidence that drives frequent change in recommended clinical practice. Living guidelines are updated on a regular schedule by a standing expert panel that systematically reviews the health literature on a continuous basis, as described in the ASCO Guidelines Methodology Manual . ASCO Living Guidelines follow the ASCO Conflict of Interest Policy Implementation for Clinical Practice Guidelines . Living Guidelines and updates are not intended to substitute for independent professional judgment of the treating clinician and do not account for individual variation among patients. See appendix for disclaimers and other important information (Appendix I and Appendix II). Updates are published regularly and can be found on the ASCO Publications website .

PURPOSE To evaluate open-source large language models (LLMs) for extracting cancer-specific phenotypic data, benchmark their performance against GPT4 models, and assess the impact of fine-tuning with training data sizes. METHODS Open-source LLMs (Mistral, LLaMa, MAMBA, BioMistral) were evaluated in zero-/one-shot and fine-tuned setups against GPT4-turbo/GPT4o to extract the cancer presence, progression, response, and metastatic sites from radiology impressions of patients with solid tumors treated at Dana-Farber Cancer Institute. Performance metrics (accuracy, precision, recall, F1-score) were computed. McNemar's odds ratio (OR), measuring which model is more likely to be correct when they disagree, was computed with 95% CI. Statistical significance was assessed using the alpha of .000139. RESULTS This study included 2,623 patients (25,273 radiology impressions). In zero-/one-shot settings, GPT4-turbo/GPT4o outperformed open-source LLMs. However, fine-tuned open-source LLMs achieved higher F1-scores than GPT4 models. Compared with the best-performing GPT4 model, fine-tuned Mistral0.2-7.3B (OR, 0.27 [95% CI, 0.20 to 0.36]; P < .00001), Mistral0.3-7.3B (OR, 0.26 [95% CI, 0.19 to 0.36]; P < .00001), LLaMa2-6.7B (OR, 0.30 [95% CI, 0.22 to 0.40]; P < .00001), LLaMa3.1-8B (OR, 0.37 [95% CI, 0.28 to 0.48]; P < .00001), and MAMBA-2.8B (OR, 0.32 [95% CI, 0.24 to 0.42]; P < .00001) showed significantly better performance in ascertaining disease progression. Performance was consistently better for inferring overall response, any evidence of cancer, and sites of metastases, with no significant differences among fine-tuned open-source LLMs. Fine-tuning gains plateaued at 25% of training data (5,718 impressions) and remained comparable at 5% (1,144 impressions). CONCLUSION Open-source LLMs, when fine-tuned using labeled data, can effectively automate the ascertainment of key radiophenotypic variables using only the impression section of radiology reports, without the full report text. Their consistent performance in small training sets suggests that these models may provide a scalable approach for phenotypic characterization of patients with cancer in real-world clinical settings.

Abstract Background Diagnosis and risk stratification in rare hematologic malignancies such as myeloproliferative neoplasms (MPNs) – polycythemia vera (PV), essential thrombocythemia (ET), and myelofibrosis (MF) – require expert review of longitudinal, heterogeneous clinical records. This process is cognitively demanding, inconsistently applied, and difficult to scale beyond tertiary centers. No automated phenotyping workflow currently exists for hematologic malignancies. Methods A HIPAA-compliant large language model (LLM) framework for phenotyping MPN was developed to integrate (i) rule-based retrieval of bone marrow biopsy reports, clinical notes, and structured laboratory results from the electronic health record (EHR); (ii) zero-shot extraction of diagnostic and prognostic variables from unstructured text using GPT-4 Turbo; (iii) a clinician-informed source-prioritization algorithm to reconcile conflicting multi-source data; (iv) WHO/ICC-criteria–based diagnostic classification; and (v) NCCN-based risk stratification using the conventional risk model for PV, IPSET-thrombosis for ET, and DIPSS, DIPSS-plus, and MIPSS70/MIPSS70+ v2 for MF. Patients were identified via MPN-related ICD-9/10 codes; cases met 2017 WHO criteria or had a hematologist-documented diagnosis, and controls did not. The cohort was split into a prompt-development set (n = 60) and a held-out test set (n = 450; 75 cases and 75 controls per disease). Ground truth was established by independent dual-clinician chart review with consensus adjudication. LLM performance was evaluated against the ground truth: variable-level extraction using accuracy, F1 score, and Cohen’s kappa; patient-level diagnostic classification using sensitivity, specificity, and Cohen’s kappa; and prognostic risk stratification (among confirmed cases) using accuracy, weighted F1 score, and quadratic-weighted Cohen’s kappa. Wilson 95% confidence intervals (CIs) were used for proportions and bootstrap 95% CIs with 500 resamples for F1 scores. Results The held-out test set included 450 patients (PV: 150; ET: 150; MF: 150) with pathology reports and structured laboratory results, and 172 patients (PV: 52; ET: 55; MF: 65) with clinical notes. From pathology reports, overall variable extraction accuracy and F1 score were 99% (95% CI, 98–100) and 1.00 (0.99–1.00) for PV, 100% (99–100) and 0.99 (0.96–1.00) for ET, and 100% (99–100) and 0.99 (0.97–1.00) for MF. From clinical notes, overall accuracy and F1 score were 96% (91–100) and 0.94 (0.85–1.00) for PV, 100% (100–100) and 1.00 (1.00–1.00) for ET, and 100% (99–100) and 0.98 (0.95–1.00) for MF. Diagnostic sensitivity was 100% (95% CI, 95.1–100.0) for PV, ET, and MF; specificity was 98.7% (92.8–99.8) for PV and 100% (95.1–100.0) for both ET and MF, with Cohen’s kappa of 0.99 for PV and 1.00 for ET and MF. Risk stratification accuracy was 100% with weighted F1 score of 1.00 and quadratic-weighted Cohen’s kappa of 1.00 across all three diseases. A pre-specified source-ablation analysis showed that pathology reports alone were sufficient for diagnosis (sensitivity 98.7% for PV, 100% for ET, 96.0% for MF; specificity 100% across all three subtypes) but inadequate for prognostication (accuracy 69.3% for PV, 93.3% for ET, 77.3% for MF). Adding clinical notes to pathology reports recovered full prognostic accuracy of 100% across all three diseases. Conclusions This first-in-class automated framework achieved expert-level performance for MPN diagnosis and risk stratification from real-world EHR data, establishing a foundation for scalable, standardized phenotyping in rare hematologic malignancies. Prospective, multi-site validation is warranted before clinical deployment.

PURPOSE Incorporation of structured data elements within the electronic health record generates real-time data for many potential uses. However, this requires a change in documentation practices, which may increase documentation effort. The best strategies to support structured documentation interventions have not been thoroughly investigated. Herein, we report the results of extending a data capture tool across multiple disease sites and practice locations. METHODS A note-based data capture tool was developed to collect five data elements using minimum common oncology data elements language. Soft-stop close visit validation (CVV), offloading of data entry tasks to allied health staff, group education, individualized e-mails, and one-on-one assistance were used to support utilization. Baseline and follow-up surveys were collected. For each provider, we compared data capture in the 6 months before and after their CVV go-live. RESULTS Forty-five providers participated (31 physicians, 14 nurse pracationers and physician assistants [NPPAs]), resulting in 23,274 encounters pre-CVV go-live and 23,146 post-CVV go-live visits. Median percentage of all data elements captured increased from 69.5% to 82.6% across all providers. The effect of CVV was immediate and sustained. When surveyed, most respondents agreed with the value propositions of understanding practice population, outcomes, selecting trials to open, and identifying patients for trial enrollment. Most respondents (15/20, 75%) felt the soft stop was a helpful reminder. CONCLUSION Structured data can be successfully collected in provider notes with >80% data capture. Personalized assistance incorporating the structured data elements into users' preexisting templates and soft-stop CVV, along with monitoring the results of implementation and personalized follow-up, was used to support the structured data capture intervention.

This cohort study examines patterns of bone-modifying agent use from 2013 to 2024 among men with metastatic castration-resistant prostate cancer.

21 Background: Large language models (LLMs) are becoming an increasing source of medical information for both patients and providers. Clinical decision making is occasionally complex and may use factors outside of traditional guidelines. We conducted a comparative performance evaluation of frontier models on standardized genitourinary (GU) oncology knowledge assessments derived from the American Society of Clinical Oncology Self-Evaluation Program 2025 (ASCO SEP 2025) in addition to challenging questions developed by GU oncology experts. Methods: Three frontier LLMs were selected: ChatGPT-5.2 (OpenAI, San Francisco, CA, USA), Claude Opus 4.6 (Anthropic, San Francisco, CA, USA), and Gemini 3 Pro (Google DeepMind, Mountain View, CA, USA). Model selection was based on publicly reported performance on “Humanity’s Last Exam,” a benchmark of advanced scientific reasoning (Scale AI leaderboard Jan-Feb 2026). The OpenEvidence (OpenEvidence Inc., MA, USA) search engine was also selected for evaluation. All models were asked 145 ASCO SEP 2025 GU oncology questions and 12 real world challenging scenarios using an uniform prompt. Subject matter experts peer reviewed LLM generated responses to challenging questions and credit was given only for correct rationale and option selection. Results: All models demonstrated high and comparable accuracy on Standardized knowledge assessment, with Gemini 3 Pro leading numerically at 97.24% (95% CI, 93.1% – 99.0%), OpenEvidence followed with an accuracy of 94.48% (95% CI, 89.5% – 97.3%), while Chat GPT -5.2 and Claude OPUS 4.6 scored 92.41% (95%, CI 86.9% – 95.7%) and 79.31% (95% CI, 71.9% – 85.1%) respectively. On challenging real world scenarios, ChatGPT-5.2 demonstrated the highest numerical accuracy at 100% (95% CI, 75.8% – 100%), Gemini 3 Pro and Open Evidence scored 75.0% (95% CI, 46.8%–91.1%) each, followed by Claude Opus 4.6 at 66.7% (39.1–86.3). Conclusions: All current frontier AI models appear to provide accurate medical information and rationale on a large majority of guideline based medical knowledge assessments for GU oncology. Frontier AI models also appear to have emerging competence in highly challenging real-world scenarios. AI based tools may become a valuable and reliable tool for patients and providers. Comparative LLM performance. Model ASCO SEP 2025 % (95% CI) Real World Scenarios % (95% CI) Gemini 3 Pro 97.24% (93.1–99.0) 75.0% (46.8–91.1) OpenEvidence 94.48% (89.5–97.3) 75.0% (46.8–91.1) ChatGPT-5.2 92.41% (86.9–95.7) 100% (75.8–100) Claude Opus 4.6 79.31% (71.9–85.1) 66.67% (39.1–86.3)

As machine learning (ML)-based decision support tools proliferate in clinical practice, understanding how clinicians integrate personalized ML predictions alongside randomized controlled trial (RCT) evidence is critical. We designed a web-based clinical decision support system (CDSS) presenting survival and adverse event data from a simulated RCT and ML model across 12 synthetic multiple myeloma scenarios. In a within-subjects study with 32 physicians, we evaluated how clinicians synthesize competing evidence sources to make treatment decisions. When ML and RCT outputs were concordant, physicians reported greater confidence than with RCT data alone. When results were discordant, most physicians shifted toward the ML-supported treatment, often before reviewing any information about model training or validation, suggesting a tendency toward automation bias rather than algorithm avoidance. Despite reporting higher perceived reliability after viewing model quality disclosures, physicians were largely unable to describe the validation procedures they had reviewed. Taken together, these findings reveal that clinicians may over-rely on ML recommendations even when equipped with tools designed to support critical appraisal. We discuss implications for CDSS design, clinician training, and the institutional safeguards needed before ML-based systems are deployed in high-stakes oncology settings.