Citations

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 <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.

ABSTRACT Background Symptom management in myeloproliferative neoplasms (MPN) remains challenging despite advancements in disease‐directed therapies. This study assessed the impact of demographic, clinical, laboratory and treatment‐related variables on total symptom scores (TSS) and individual symptom scores in patients with polycythemia vera (PV), essential thrombocythemia (ET), and myelofibrosis (MF). Methods A cohort of 553 patients (PV: 200; ET: 190; MF: 163) was identified through a retrospective chart review. Symptom scores, demographic information and clinical variables were captured at the time of the first TSS assessment. Laboratory results within 1 month and treatment information within 90 days prior to the symptom assessment were captured. Univariable linear regression, followed by multivariable regression analyses using a robust variance estimator were performed, with a p value < 0.05 indicating a significant association. Results MF patients experienced the highest symptom burden and fatigue was the most reported symptom across all MPN subtypes. In PV, depression ( β : 10.53; p = 0.001) was associated with a higher TSS whereas older age (−0.17; 0.01) and higher hemoglobin (−1.24; 0.003) were associated with a lower TSS. In ET, depression ( β : 14.19; p < 0.001) was associated with a higher TSS, whereas in MF, depression (12.28; < 0.001) and higher WBC count (0.22; 0.02) were associated with a higher TSS. Depression and non‐White race in PV, depression in ET, and depression, low hemoglobin, and higher WBC count in MF were associated with multiple symptoms. Conclusions Integrating depression screening and management and optimizing hematologic parameters alongside disease‐directed therapy is crucial to improving patient outcomes.

Abstract Background Socioeconomic position (SEP) is a fundamental social determinant of health (SDoH) contributing to observed disparities in cancer and cardiovascular (CV) care among individuals with prostate cancer (PC). Understanding the influence of SEP on CV health is essential for addressing outcome inequities within this population. Methods We conducted a retrospective study utilizing Surveillance, Epidemiology, and End Results (SEER)-Medicare to evaluate CV outcomes in patients ≥65 years with PC from 2009 to 2017. We studied the impact of SEP on cardiovascular events (CVEs), cardiovascular mortality (CVm), PC-specific mortality (PCsm), and all-cause mortality. CVE, CVm, and PCsm were analyzed using competing risk models and all-cause mortality with Cox proportional hazards models. A subgroup analysis was performed in Non-Hispanic Black and Non-Hispanic White individuals. Results We included 141 242 patients of whom 76 844 were in high SEP areas (45.6%) and 64 398 in low SEP ones (54.4%). Both groups had a mean age of 72 years; patients in low SEP areas were 67.5% Non-Hispanic White, 34.4% having diabetes, 73.3% hypertension, and 67.7% dyslipidemia versus the high SEP group with 85.6% Non-Hispanic White, 30.4% diabetes, 69.9% hypertension, and 70.7% dyslipidemia. Patients in low SEP areas had higher risks of CVm (subdistribution hazard ratio [sHR] = 1.18, 95% CI = 1.12 to 1.25), PCsm (sHR = 1.12, 95% CI = 1.06 to 1.17), CVE (sHR = 1.07, 95% CI = 1.04 to 1.09), and all-cause mortality (HR 1.16, 95% CI = 1.13 to 1.20). Accentuated results were shown in the Non-Hispanic Black subgroup. Conclusion Adverse SEP plays a significant role in shaping outcomes among older patients with PC, contributing to elevated risks of adverse CV and survival outcomes.

560 Background: Early diagnosis remains a critical challenge in oncology, particularly for the five deadliest cancers: lung, breast, prostate, colorectal, and pancreatic. These cancers often present with nonspecific symptoms, leading to repeated healthcare visits without a timely suspicion of malignancy. Artificial intelligence (AI) offers an opportunity to augment clinical decision-making and support earlier identification of cancer. C the Signs , an AI cancer prediction platform, has been widely adopted across the UK with substantial evidence supporting its ability to identify cancer at earlier stages. Methods: This retrospective study assessed the performance of C the Signs in intercepting cancer earlier than physician diagnosis. De-identified electronic medical records were accessed via the Mayo Data Platform, spanning a 20-year period from January 1, 2002 to December 31, 2021. The study cohort included 55,375 patients diagnosed with lung, breast, prostate, colorectal or pancreatic cancer. The primary outcome was the proportion of cases identified earlier by the AI model, measured in annual increments up to five years prior to physician diagnosis. Additional analyses were conducted across the five target cancer types. Results: The C the Signs platform identified 14,528 patients (26.2%) with cancer earlier than the original physician diagnosis. Early identification occurred across multiple time points: 8.9% in year 1, 4.7% in year 2, 4.4% in year 3, 4.3% in year 4, and 3.9% in year 5. Across cancer types, the highest early interception rate was observed in prostate cancer (31.7%), while breast cancer showed the lowest impact (21.0%). Conclusions: Use of the C the Signs AI cancer prediction platform demonstrated meaningful improvements in early cancer interception, particularly for lung and prostate cancers. These findings highlight the potential of AI tools to enhance early diagnosis and support timely clinical interventions, offering a scalable solution to improve cancer outcomes across populations.

Abstract BACKGROUND Von Hippel-Lindau (VHL) disease is a hereditary cancer syndrome characterized by development of multiple neoplasms, including central nervous system (CNS) hemangioblastomas, renal cell carcinoma (RCC), pancreatic neuroendocrine tumors (pNETs), and retinal hemangioblastomas. Belzutifan, a hypoxia-inducible factor 2-alpha (HIF-2α) inhibitor, has demonstrated efficacy in treating VHL-associated tumors. This study evaluates our institutional experience with belzutifan in patients with VHL-related tumors. MATERIAL AND METHODS We conducted a retrospective cohort study across three Mayo Clinic sites, including patients aged ≥18 years with genetically confirmed VHL disease who received belzutifan between October 2021 and March 2025. Clinical/radiological data, and adverse events (AEs) were extracted from electronic health records. Tumor responses were assessed using RECIST v1.1 criteria for RCC, pNET, CNS hemangioblastoma. Retinal hemangioblastoma responses were evaluated through ophthalmologic assessments. RESULTS Forty-five patients were included (median age 44 58% female). Sixty-seven percent had RCC, 91% had CNS hemangioblastomas, 53% had pNETs, and 62% had retinal hemangioblastomas. The median duration of belzutifan therapy was 22.6 months (range 1.0-41.4 months). Among patients with RCC (n=26), overall response rate (ORR) was 19% and median time to best response (MTBR) was 7.4 months. Five patients had partial response (PR), 15 had stable disease (SD), and 1 had progressive disease (PD). For CNS hemangioblastomas (n=41), ORR was 12%, MTBR was 6.2 months. 5 patients had PR, 30 had SD, and none had PD. For pNET (n=17), ORR was 41%, MTBR was 8.2 months. One patient had CR, 6 patients had PR, 6 had SD, none had PD. For retinal hemangioblastomas (n=28), 6 (21%) showed improvement, 9 (32%) remained stable, and 3 (11%) showed progression. MTBR was 6.7 months. Treatment-related AEs occurred in 98% of patients. Common AEs included anemia (89%), fatigue (80%), hypoxia (18%), and headache (13%). Grade 3 AEs were observed in 20% of patients, including anemia (7%), hypoxia (9%), fatigue (4%), pneumonitis (2%), loss of consciousness (2%). One patient experienced a grade 4 event (status epilepticus). Anemia management involved iron supplementation (n=11), RBC transfusion (n=3), dose reduction (n=2), treatment holiday (n=8), permanent discontinuation (n=1). Hypoxia management included dose reduction (n=1), treatment holiday (n=3), discontinuation (n=1). CONCLUSION Belzutifan shows clinical activity across various VHL-associated neoplasms in real-world practice, with a manageable safety profile. The lower ORR compared to the phase III trial may reflect differences in baseline disease, including lower tumor burden in our cohort. These findings support its continued use in VHL disease, with attention to monitoring and managing treatment-related AEs.

ImportancePatients with metastatic castration–resistant prostate cancer (mCRPC) harboring BRCA1/2 alterations are eligible to receive poly(adenosine diphosphate-ribose) polymerase (PARP) inhibitors as single agents or in combination with an androgen receptor pathway inhibitor after these agents showed survival improvement in their landmark clinical trials. However, data are limited regarding the uptake of PARP inhibitors in these patients.ObjectiveTo investigate the use of PARP inhibitors in patients with mCRPC harboring BRCA1/2 alterations.Design, Setting, and ParticipantsThis retrospective cohort study used the deidentified Flatiron-Health electronic health record–derived database of US community and academic practices to extract patient-level data. The data cutoff date was May 31, 2024. Patients with mCRPC with evidence of harboring BRCA1/2 alterations and alive after August 15, 2020 (ie, 3 months after the approval of the first PARP inhibitor, rucaparib, in mCRPC) were included. Statistical analysis was performed from September 2024 to May 2025.ExposuresAge, race and ethnicity, insurance status, and practice type at the time of mCRPC diagnosis.Main Outcomes and MeasuresThe receipt of PARP inhibitors. Multivariable logistic regression was conducted to assess the association between the exposures and the main outcome.ResultsOf 24 105 patients with metastatic prostate cancer, 443 male patients (median [IQR] age, 72 [65-79] years) had mCRPC with BRCA1/2 alterations and were eligible and included in our analysis. Of these patients, 227 (51.2%) received a PARP inhibitor, whereas 216 (48.8%) did not. Compared with patients covered by a commercial health plan, those covered by Medicare were significantly more likely to receive a PARP inhibitor (odds ratio, 1.91; 95% CI, 1.02-3.66; P = .047). The odds of receiving a PARP inhibitor were not significantly higher in patients treated in community practice compared with those treated in academic centers (odds ratio, 1.64; 95% CI, 1.00-2.70; P = .05).Conclusions and RelevanceThis retrospective cohort study of patients with mCRPC and evidence of BRCA1/2 alterations found that approximately half of patients did not receive PARP inhibitors despite evidence of survival improvement in this population. These findings highlight the need to increase awareness of the survival data and access to life-prolonging therapies in patients with mCRPC.

Background: Longitudinal genomic profiling through serial circulating tumor DNA (ctDNA) testing offers a noninvasive method to monitor clonal evolution in advanced prostate cancer. This study evaluated the frequency and nature of newly emergent and potentially actionable genomic alterations detected through serial testing in a real-world setting. Methods: We conducted a retrospective analysis of advanced prostate cancer patients who underwent multiple Guardant360 ctDNA tests between October 2020 and March 2023. The study focused on identifying new genomic alterations absent in the initial test, with particular attention to alterations relevant for on-label therapies, therapies approved in other oncologic indications (i.e., off-label), or a clinical trial. Results: Among 479 patients with at least two ctDNA tests, the median interval between the first and second evaluable tests was 207 days. New and potentially actionable alterations emerged in 57.8% of patients, including potential targets for on-label therapies (16.7%), off-label therapies (16.5%), and clinical trials (55.7%). Tumor mutational burden (TMB) increased from “low” to “high” in 11% of patients, although none had microsatellite instability or mismatch repair deficiency. In a Mayo Clinic subset, ten patients received olaparib based on treatment-emergent alterations, but none achieved a prostate-specific antigen (PSA) response. Two patients who transitioned from low to high TMB received pembrolizumab, both with progressive disease as best response. Conclusions: In a large real-world cohort, serial ctDNA testing frequently identified new alterations that were not detected at baseline and are potentially actionable therapeutic targets, highlighting the value of serial genomic profiling for capturing clonal dynamics. Additional research is needed to better establish a framework for retesting and to clarify how these results should influence subsequent treatment decisions.