Riaz, Irbaz Bin

464 Background: Based on LITESPARK-005 trail, Belzutifan was FDA-approved to treat advanced RCC following ICI and a VEGF-TKI. We conducted a retrospective cohort study to report the real-world outcomes of Belzutifan in pre-treated RCC patients. Methods: A retrospective chart review of all patients who received Belzutifan at the Mayo Clinic Comprehensive Cancer Center between Dec 2023 and Sep 2025 was conducted. Patients treated for sporadic RCC were eligible; those with von Hippel–Lindau disease, non-RCC pathologies (e.g., Pheochromocytoma, ovarian cancer) were excluded. Radiologic response was assessed by trained curators using RECIST-defined criteria, approximated through systematic review of radiology reports with clinician adjudication when needed. Progression-free survival (PFS), defined as the time from treatment initiation to radiographic progression or death, was estimated using the Kaplan–Meier method. Non-progressors were censored at the last radiology assessment. Patients who discontinued therapy for non-progression reasons were censored at the last radiology assessment prior to discontinuation. Any-cause hypoxia and anemia were reported using CTCAEv5 definitions. Results: A total of 158 patients were identified; 83 (52.5%) who received Belzutifan for sporadic RCC were included. The remaining were excluded for VHL disease (68, 43%) or other pathologies (7, 4.4%). Median follow up was 12 months, 67% were male, median age was 70 (range 39, 94) with 57 patients (69%) were ≥ 65. 28 patients (34%) had a poor IMDC score at baseline. The cohort received a median of 3 previous lines of therapy (range 1,8). Median treatment duration was 3 months (IQR: 1-9.5). Median PFS was 5.7 months (95% CI, 3.8–8.7). At 12 and 18 months, 23.1% and 13.9% of patients, respectively, remained alive and progression-free. Corresponding OS rates at these time points were 60.2% and 51.5%. A confirmed ORR occurred in 12 patients (14.4%). Hypoxia occurred in 44 (53%) patients, median onset 31 days (range 1–270) and 29 (34.9%) required hospitalizations. Median hemoglobin (Hb) levels at baseline were 12.65 g/dl (range, 7.2-15.5), and 68 (82%) patients developed Hb decline on Belzutifan, 27 (32.5%) required transfusions. Conclusions: In an older, high-risk cohort with higher number of prior lines of therapy, Belzutifan yielded PFS/OS and ORR landmarks comparable to the trial supporting its effectiveness in the real world. Outcomes compared with LITESPARK-005 trial. Variable Real world cohort (Mayo Clinic) LITESPARK-005, Belzutifan arm Median PFS (months, 95% CI) 5.7 (3.8-8.7) 5.6 (3.9-7) 12-month PFS rate (%, 95% CI) 23.1% (14.4-33.4) 33.4% (28.4-33.5) 18-month PFS rate (%, 95% CI) 13.9% (7.7-23.9) 24% (19-29.4) 12-month OS rate (%, 95% CI) 60.2% (48.9-70.8) 67.9% (62.9-72.4) 18-month OS rate (%, 95% CI) 51.5% (40.6-62.9) 55.2% (50-60.1) ORR (%, 95% CI) 14.4% (7.7-23.9) 21.9% (17.8-26.5)

405 Background: Clinical trials are critical to advancing prostate cancer treatment, yet patient enrollment remains a major bottleneck. Despite strong interest in AI for trial matching, real-world use remains limited due to reasons such as complex infrastructure requirements. To address this, we developed a lightweight, scalable AI framework leveraging the Google Healthcare Search API—requiring minimal technical expertise—to automate trial eligibility screening. Methods: We used a pre-trained GPT-5 large language model (LLM) to decompose the eligibility criteria of 13 prostate cancer clinical trials into 269 YES/NO questions, such that for an eligible patient of a given trial, the chart review would yield YES for inclusion and NO for exclusion questions. From 348 patients enrolled in one of the trials, we built a dataset linking each eligibility question to each patient, resulting in 4,947 patient–question pairs (2,034 inclusion pairs with YES labels and 2,913 exclusion pairs with NO labels). The task was framed as binary classification. During evaluation, for each patient–question pair, the LLM extracted search phrases from the question to retrieve relevant EHR information via the Google Healthcare Search API, then used the question and retrieved information to predict YES or NO responses. Performance was measured using accuracy, precision, recall, specificity, and F1 score. Results: Across 4,947 question–patient pairs, our framework achieved 0.81 accuracy, 0.85 precision, 0.65 recall, 0.92 specificity and a F1 score of 0.74. Results for different phases of trials is reported in Table 1, which demonstrates the approach’s robustness across trial phases. Predictions for exclusion criteria were highly precise, while inclusion criteria were more challenging. Most errors were observed in criteria involving complex nested logical conditions, interpretable clinical definitions and data not explicitly measurable or recorded in the EHR (e.g., planned treatments or clinician-assessed scores). Conclusions: This work demonstrates a practical AI framework to support clinical trial matching. The system can identify complex disease states (biochemical recurrence vs radiographic metastasis), castration status (sensitive, resistant) and genomics. Future work should focus on developing human-in-the-loop supervision and improving question decomposition to reduce errors in complex eligibility criteria. Screening performance of the proposed framework across different phases of trials. Trial Phase #Patient-Question Pairs #YES/NO labels Accuracy Precision Recall Specificity F1 score I, II, I/II 4,227 Yes: 1,735 No: 2,492 0.80 0.84 0.66 0.92 0.74 III 720 Yes: 299 No: 421 0.82 0.90 0.63 0.95 0.74 All 4,947 Yes: 2,034 No: 2,913 0.81 0.85 0.65 0.92 0.74 #YES/NO labels: The number of patient-question pairs that are labeled YES/NO among total patient-question pairs of the same row.

Abstract Background The ability of large language models (LLMs) to work collaboratively and screen studies in a systematic review (SR) is under-explored. Hence, we aimed to evaluate the effectiveness of LLMs in automating the process of screening in systematic reviews. Methods This is an observational study which included labeled data (title and abstracts) for five SRs. Originally, two reviewers screened the citations independently for eligibility. A third reviewer cross-checked each citation for quality assurance. GPT-4, Claude-3-Sonnet, and Gemini-Pro-1.0 were used using zero-shot chain-of-thought prompting. Collaborative approaches included (i): conflict resolution using benefit of the doubt, (ii) majority voting using an independent third LLM and (iii) conflict resolution using an informed third LLM. Performance was assessed using accuracy, precision for exclusion, and recall for inclusion. Work saved over samples (WSS) was computed to estimate the reduction in manual human effort. Results A total of 11300 articles were included in this study. The individual models, GPT-4, Claude-3-Sonnet, and Gemini-Pro-1.0 exhibited a high precision for exclusion, achieving 99.7%, 99.7%, and 99.2% and high recall for inclusion achieving 95.5%, 96.6% and 85.7%, respectively. However, the collaborative approach utilizing the two best-performing models (GPT-4 and Claude-3S) achieved an average precision of 99.9% and a recall of 98.5% (across all collaborative approaches). Furthermore, the proposed collaborative approach resulted in an average WSS of 63.5%, compared to the average WSS of 45.2% for individual models. Conversational LLM interactions showed a consistent pattern of results. Limitations This study was limited due to reliance on proprietary models, and evaluation on oncology datasets. Conclusion Evidence shows that collaborative LLMs enable efficient, high-performing screening in systematic reviews, supporting continuous evidence updates. Primary funding source NIH (U24CA265879-01-1) and Carolyn-Ann-Kennedy-Bacon Fund.

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 1 and Appendix 2 ). Updates are published regularly and can be found at www.asco.org/genitourinary-cancer-guidelines .

Abstract INTRODUCTION: Prostate cancer (PC) shows marked heterogeneity in clinical outcomes, from indolent to highly aggressive disease. Alterations in TP53, PTEN, RB1, and DNA damage repair genes are linked to poor patient outcomes, yet how the interaction and co-occurrence of these alterations impact prognosis remains unclear. We leveraged the AACR Project GENIE prostate cancer cohort to assess the prognostic impact of individual and concomitant gene alterations, focusing on key PC drivers (BRCA2, TP53, PTEN, RB1). METHODS: the study was conducted in two stages: (i) identification of candidate alterations at individual gene level through a systematic literature review, followed by analyzing the prevalence and co-occurrence of genes in the AACR Project GENIE Public Release (PR) dataset. (ii) Prognostic analysis of the pre-selected biomarkers in the AACR Project GENIE Biopharma Collaborative (BPC) cohort. Analyses were performed in three clinical settings (entire BPC cohort, n=1114; subgroup of patients with localized disease at diagnosis, n=837; and subgroup of patients with metastatic PC with a biopsy collected before or at the time of mPC, n=455) using two statistical approaches: univariate analysis (UA) and inverse probability weighting using propensity scores (PS). Cox models were applied to estimate overall survival hazard ratios (HRs). RESULTS: after integrating the findings from the systematic review with the analysis of the GENIE PR (n=5684), 107 individual genes and 122 co-occurring alterations were selected for study. In the GENIE BPC (n=1114 patients) 75% were M0 at diagnosis, and 67% presented metastatic PC at any time. Alteration prevalence was: TP53, 25%; PTEN, 15%; BRCA2, 4%; RB1, 4%; co-alterations included, PTEN-TP53 (6.4%), RB1-TP53 (2.1%), PTEN-RB1 (1.3%), and BRCA2-TP53 (1.2%). All four key genes showed a statistically significant prognostic impact in the entire cohort with UA (BRCA2, HR:1.9; TP53, HR:2.6; PTEN, HR:2.4; RB1, HR:4.1, all p<0.001). Results were consistent in the subgroup analysis and when using PS. Additional genes such as FANCA, PIK3CA, and MYC were also associated with poor outcomes. Among co-alterations, BRCA2-TP53 (HR:2.0, p=0.05), RB1-TP53 (HR:3.6, p<0.001), PTEN-TP53 (HR:2.4, p<0.001); PTEN-RB1 (HR:5.5, p<0.001), and PTEN-TP53-RB1 (HR:9.0, p<0.001) were the strongest predictors of poor survival, with consistent effects across subgroups and analytical methods. Other combinations including MYC-TP53 (HR 2.2, p<0.001) and PIK3CA-TP53 (HR 3.2, p<0.001), also showed independent prognostic value. CONCLUSIONS: Genomic alterations in TP53, PTEN, RB1, and BRCA2, alone and in combination, independently associate with poor prognosis. Co-alteration patterns such as PTEN-TP53-RB1 define aggressive molecular subsets. This evidence supports integrating genomic interaction patterns for enhancing clinical risk models and guiding precision patient management. Furthermore, this study highlights the potential of AACR Project GENIE to enable data-driven development of biomarker-informed patient care strategies. Citation Format: Pablo Cresta Morgado, Victor Navarro, Guillermo Vilacampa, Haitham Alatoom, Manuel Ramos del Rio, Irbaz Riaz, Shawn Sweeney, Nikolaus Schultz, Ken Kehl, Gregory Riely, Wasim Abida, Katherine Panageas, Deborah Schrag, Philippe Bedard, Christine Micheel, Xindi Guo, Chelsea Nayan, Rodrigo Dientsmann, Joaquin Mateo. Genomic prognostic biomarkers in prostate cancer from the AACR GENIE cohort [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Innovations in Prostate Cancer Research and Treatment; 2026 Jan 20-22; Philadelphia PA. Philadelphia (PA): AACR; Cancer Res 2026;86(2_Suppl):Abstract nr A013.

399 Background: The standard of care for resectable GEA adenocarcinomas is perioperative chemotherapy with the FLOT (fluorouracil, leucovorin, oxaliplatin, docetaxel) regimen. Recent studies investigating the addition of checkpoint inhibitors (CPIs) to perioperative chemotherapy have shown mixed results. An NMA can leverage both direct and indirect comparisons to evaluate the relative efficacies of available regimens in this setting. Methods: MEDLINE, EMBASE, Scopus, Web of Science and CENTRAL were searched until August 27th, 2025. Eligible studies included phase 3 randomized trials evaluating perioperative or neoadjuvant chemotherapy +/- CPIs ± radiation in resectable GEA. The primary outcomes were disease-free survival (DFS) and overall survival (OS). Hazard ratios and 95% confidence intervals (CIs) were estimated using a frequentist NMA framework. Results: Fifteen trials including 8072 patients were analyzed. For DFS, perioperative durvalumab + FLOT (dFLOT) ranked highest (P-score = 0.99), followed by FLOT (P-score = 0.84) (HR for dFLOT vs. FLOT was 0.71, 95% CI 0.58-0.86). dFLOT also ranked higher than perioperative pembrolizumab + cisplatin and fluorouracil/capecitabine (KN-585), though this comparison was not statistically significant (HR 0.65, 95% CI 0.40-1.04). Compared with CROSS, dFLOT showed superior DFS (HR of 0.48, 95% CI 0.36-0.63). For OS, dFLOT ranked the highest (P-score=0.95) and was superior to CROSS (HR 0.57, 95% CI 0.42-0.77). While dFLOT appeared favorable versus FLOT (HR 0.78, 95% CI 0.62-0.98), this finding was primarily driven by MATTERHORN, which did not meet its prespecified OS significance boundary (P = 0.03 vs threshold P < 0.0001), and the proportional hazards assumption was violated. Similarly, no significant OS difference was observed between dFLOT and the KN-585 arms (HR 0.75, 95% CI 0.45-1.23). Conclusions: In this NMA, perioperative FLOT remained the benchmark regimen for resectable GEA. dFLOT has shown promising early results, ranking the highest for DFS and showing favorable, but not statistically definitive OS trends in comparison to FLOT. Importantly, the network enabled indirect comparisons not tested in head-to-head trials, demonstrating that dFLOT ranked above pembrolizumab+CF/CX (KN-585) without reaching statistical significance, and also outperformed CROSS. However, these findings should be interpreted with caution, as they are primarily driven MATTERHORN data in which OS did not cross the prespecified threshold and proportional hazards assumptions were violated. Perioperative immunotherapy combinations appear promising, but final MATTERHORN results and further confirmatory studies are needed before dFLOT can be considered as a new standard of care.

103 Background: Clinicians spend significant time extracting and synthesizing information to risk-stratify and make adjuvant chemotherapy (AC) recommendations for CRC. By automating parts of this process, LLMs could help reduce documentation burden and variability in decision-making. Methods: We selected 144 stage III, pMMR CRC patients treated at Mayo Clinic (2017–2021). HIPAA-compliant Gemini-2.0-Flash-001 and GPT-4o extracted clinical variables (age, neuropathy grade, pathological stage) from EHRs and generated AC recommendations. Two approaches were compared: (1) a rule-based algorithm with nested if/else statements translating variables into NCCN-based recommendations, and (2) a dynamic LLM agent generating recommendations using a structured prompt informed by NCCN guidelines. Concordance with treating physicians was assessed, with disagreements reviewed by a blinded GI oncologist. Results: The dataset included 144 patients with median age 67 (IQR 66–68), 48.6% female. The LLM pipeline extracted neuropathy grade with 100% accuracy and pathological T/N stages with 99.3% accuracy. Concordance with physician recommendations was 93.1% for the rule-based algorithm and 62.5% for the LLM agent. In discordant cases, blinded adjudication by a GI oncologist favored the algorithm and LLM agent over physicians in most cases (Table 1). Conclusions: This exploratory work shows LLM agents can reliably extract variables and generate evidence-based treatment recommendations for stage III CRC. The high rate of expert preference for LLM recommendations in discordant cases is intriguing but limited by small sample size and possible adjudicator bias. Discordant physician recommendations reflect the complexity of decision-making, including patient preferences, performance status, comorbidities, and subtle findings not fully captured in EHRs. Rather than replacing judgment, we envision LLMs as decision support tools—analogous to an advanced trainee reviewing charts and presenting evidence-based recommendations to the attending physician with ultimate authority. Larger multi-institutional studies with multiple adjudicators are needed to validate these findings and clarify AI’s role in oncology. Patient characteristics, LLM extraction accuracy, and concordance with physician recommendations. Metric Result Total patients 144 Median age (IQR) 67 (66-68) Female (%) 48.6% LLM extraction accuracy – Neuropathy grade 100% LLM extraction accuracy – Pathological T and N stages 99.3% Primary Concordance – Rule-based algorithm vs. physician 93.1% Adjudicator preference of rule-based algorithm over treating physician for discordant cases 100% Concordance – LLM agent vs. physician 62.5% Adjudicator preference of LLM agent over treating physician for discordant cases 82.0%

Abstract Introduction Persistent symptom burden in classic myeloproliferative neoplasms (MPNs) i.e., polycythemia vera (PV), essential thrombocythemia (ET) and myelofibrosis (MF) is linked to shorter survival, more thromboembolic events and impaired quality of life. Identifying demographic, clinical, laboratory, and treatment-related predictors of symptom improvement can inform risk-adapted care and guide supportive interventions. Methods Patients with PV, ET, and MF were identified through a retrospective chart review. Total symptom scores (TSS) and scores of individual symptoms were recorded at the time of first TSS assessment. Demographic (age, sex, race, ethnicity), seasonal (summer: April–September; winter: October–March), clinical (time since diagnosis, depression, obesity, spleen size), treatment-related (phlebotomy, hydroxyurea, interferon-alpha, anagrelide, JAK inhibitors, antidepressants) and laboratory (complete blood counts, serum chemistry) factors were captured at the baseline. Symptom improvement was defined as a ≥50% TSS reduction from baseline. Cumulative incidence analyses, accounting for death as a competing risk, were performed to estimate the probability and timing of symptom improvement. Multicollinearity was assessed using variance inflation factors (VIFs). Highly correlated (VIF ≥5) features and features with ≥15% missing observations were excluded. Multivariable Fine-Gray hazards regression was performed to identify predictors of symptom improvement. Sub-distribution hazard ratios (sHR) with 95% confidence intervals (CI) were computed with p-values &lt;0.05, indicating a statistically significant association with symptom improvement. Results A total of 532 patients (PV: 195, ET: 180, MF: 157) were identified between 2019 and 2024. The median age was 65 (IQR: 57-74) for PV, 65 (52-75) for ET, and 69 (63-75) for MF. Most patients were White (PV: 178, ET: 164, MF: 145) and female (PV: 99, ET: 134, MF: 84). Fatigue was the most frequently reported symptom (PV: 161, ET: 155, MF: 136) whereas weight loss was the least reported (PV: 9, ET: 12, MF: 26). Baseline TSS was highest in MF (median: 15; IQR: 7-27) followed by PV (13; 6-25) and ET (13; 5-23). Symptom improvement was observed in 89 (46%) PV, 71 (39%) ET, and 44 (28%) MF patients. Among those who improved, the cumulative incidence of symptom improvement reached 100% by 55 months in PV, 53 months in ET, and 58 months in MF. At 6, 12, and 24 months, cumulative symptom improvement was 22%, 34%, and 57% in PV; 15%, 25%, and 43% in ET; and 19%, 28%, and 39% in MF. Median time to improvement was not reached in any group. Death before symptom improvement was observed in 5 (3%) PV, 6 (3%) ET, and 37 (24%) MF patients. Multivariable regression analyses in PV identified higher baseline TSS (sHR: 1.02; 95%CI: 1.01-1.03) as associated with symptom improvement (p = 0.02). In ET, lower white blood cell count (sHR: 0.89; 95%CI: 0.81-0.98) and less time since diagnosis (0.99; 0.98-0.99) were associated with symptom improvement (p = 0.02). In MF, more months since diagnosis (sHR: 1.01 95%CI: 1.01-1.01) was associated with symptom improvement (p = 0.01). Conclusions Symptom trajectories differed across MPN subtypes. In PV, a higher baseline symptom burden; in ET, a lower white blood cell count and more recent diagnosis and in MF, a longer disease duration were associated with symptom improvement. These readily available parameters may help inform supportive strategies, tailor response expectations, and identify patients for symptom-directed interventions or clinical trials.

Abstract Background Accurate diagnosis and risk stratification in myeloproliferative neoplasms (MPNs) rely on expert interpretation of complex clinical records. Outside major academic centers, specialist expertise is scarce, and time-consuming chart reviews extend clinic visits and delay treatment decisions, underscoring the need for automated, point-of-care phenotyping for MPN patients. Methods Patients with the ICD codes of PV, ET and MF were identified from the electronic health record. Cases were defined as patients meeting the 2017 WHO diagnostic criteria for PV, ET or MF or those with a definitive MPN diagnosis determined by a hematologist. Controls were the patients who met neither the diagnostic criteria nor had a hematologist-assigned diagnosis. Two clinicians extracted demographic, clinical (symptoms, thrombotic history), laboratory (complete blood count, serum chemistry), genetic (mutations, karyotype), bone marrow morphology, diagnostic mentions and prognostic score information from unstructured clinical notes and pathology reports. Prognostic scores included the conventional risk model for PV, the International Prognostic Score of Thrombosis (IPSET) for ET and the Dynamic International Prognostic Scoring System (DIPSS), DIPSS-plus or Mutation-enhanced International Prognostic Scoring System (MIPSS-70) for MF. Discrepancies in data extraction were resolved by a consensus. A random sample of 510 patients (PV: 85, ET: 85, MF: 85, controls: 255) was split into prompt development (n=60) and test (n=450) sets for iterative workflow engineering and evaluation. The workflow included: (i) a rule-based algorithm for automated retrieval of demographic details, the first bone marrow biopsy report, the first clinical note after biopsy and the latest laboratory results within 90 days preceding biopsy (ii) a HIPAA-compliant LLM (GPT-4 turbo) for data extraction from the clinical notes and pathology reports (iii) a regular expressions-based algorithm for post-processing extracted data into structured tables, (iv) a clinician-informed source prioritization algorithm to select data from preferred sources, (v) a modified WHO-informed diagnostic characterization tool incorporating hematologist-assigned diagnoses to distinguish cases from controls and; (vi) an NCCN guidelines-based algorithm for prognostic tool selection and risk-stratification based on available data. Structured prompts were used to extract demographic, clinical, laboratory, genetic and pathological data in a zero-shot deterministic setting. The performance of automated data extraction, diagnostic characterization and prognostication was compared against the manually curated data. Accuracy and F1 scores were computed for data extraction and prognostication tasks while sensitivity and specificity were calculated to assess diagnostic performance. Results A total of 450 (PV: 150, ET: 150, MF: 150) pathology reports and 50 (PV: 52, ET: 55; MF: 65) clinical notes were identified for 450 test set patients. The test set data extraction accuracy was 95% (reports: 97%, notes: 92%) for PV, 93% (reports: 92%, notes: 94%) for ET and 98% (reports: 99%, notes 97%) for MF. The test set F1 score was 0.95 (reports: 0.96, notes: 0.93) for PV, 0.95 (reports: 0.95, notes: 0.96) for ET and 0.98 (reports: 0.99, notes: 0.97) for MF. PV patients were diagnosed with 100% sensitivity and 98% specificity whereas ET and MF patients were diagnosed with 100% sensitivity and specificity. The accuracy of risk stratification was 100% for PV and ET and 94% for MF. The F1 score for risk stratification was 1.00 for PV and ET and 0.94 for MF. Conclusions The first-in-class automated framework successfully identified pertinent clinical information, performed accurate data extraction from unstructured records and enabled clinician-guided source prioritization for accurate diagnosis and risk stratification of MPN patients. Prospective validation of the framework is required for widespread clinical implementation.

The emergence of state-of-the-art large language models (LLMs), which hold the ability to generalize to diverse natural language processing tasks, has led to new opportunities in health care. Oncology is especially well-suited to leverage these resources as the journeys of patients with cancer inherently yield extensive, longitudinal data sets comprising clinical narratives, pathology and radiology reports, and genomic sequencing reports. This review begins with an overview of the fundamental concepts behind LLMs, including the definitions, architecture, training paradigm, and performance optimization through prompt engineering and retrieval-augmented generation. We also take a moment to explore the newly emerging paradigm of LLMs in a multiagentic framework. We then synthesize current research on how LLMs may benefit stakeholders within the practice of oncology, including patients, oncologists, researchers, and learners. Finally, we address the limitations and risks of LLMs, including hallucinations, inherent biases, patient privacy, and clinician deskilling. While research thus far shows significant potential for LLMs to transform cancer care, necessary future directions include studies emphasizing patient stakeholder perspectives on LLM incorporation in clinical workflows, the development of relevant clinical benchmarks for LLM evaluation, a greater focus on real-world prospective testing, and deeper exploration of LLM reasoning capabilities.