Ertz-Archambault, Natalie

Abstract While the use of NCCN guidelines is essential in lymphoma care, adherence among oncologists remains highly variable. Real-world concordance ranges from 65–80% overall, dropping to as low as 11–46% for certain lymphoma subtypes. Although 88% of clinicians report using NCCN guidelines for Hodgkin lymphoma, adherence to PET-adapted approaches is inconsistent. This gap highlights a critical need for tools that support guideline-concordant care.Large language models (LLMs) such as GPT–4 promise decision support, yet hallucinations and divergence from evidence-based care remain obstacles. We assessed whether retrieval-augmented generation (RAG) with National Comprehensive Cancer Network (NCCN) guidelines improves GPT–4 performance in lymphoma treatment recommendation tasks. To address this, we developed a RAG-enhanced GPT-4 agent (RAG-GPT) by indexing the 2025 NCCN Lymphoma Guidelines into a vector database. Fifty clinical vignettes spanning diverse lymphoma subtypes were compiled from published case reports; this sample provides 80 % power (α = 0.05) to detect a mean score difference of ≥ 0.18 on the modified Generative AI Performance Score (mG-PS). Guideline concordance was assessed based on NCCN guidelines and clinical judgment reflecting real-world oncological decision-making. Scores were normalized between -1.0 and 1.0 for consistent evaluation. Readability and rationale clarity were rated on a 5-point Likert scale. Baseline GPT-4 and RAG-GPT generated treatment plans via standardized prompts. An oncologist scored all outputs with mG-PS, which rates NCCN concordance (0.0–1.0) and penalizes hallucinations (-0.25 to –1.0). RAG-GPT achieved a higher mean mG-PS (0.32) than baseline GPT–4 (0.10) (t(98)=-2.11, p=0.038; mean score difference ΔmG-PS=0.22). Importantly, RAG-GPT eliminated severe hallucinations, reinforcing confidence in the safety of its treatment reccomendations. Fully concordant “gold-standard” plans were produced in 46 % of vignettes with RAG-GPT versus 22 % with baseline. Readability scores favored RAG-GPT modestly without statistical significance. Lymphoma's bimodal age distribution, heterogeneity, and lack of standardized treatment algorithms, coupled with regimen toxicity and the need to account for patient frailty and comorbidities, create complex clinical scenarios that constrain LLM applicability for nuanced decision-making. Non-Hodgkin lymphoma, particularly relapsed/refractory, transformed disease, or rare subtypes, lacks a defined standard of care, leading reviewers to rely on clinical judgment or published case series beyond NCCN guidelines. This underscores a key limitation of LLM-supported tools: their performance is highly dependent on the presence of clear algorithmic consensus within a given disease. Linking GPT-4 to NCCN guidelines via RAG significantly improved treatment accuracy and reduced hallucination severity in lymphoma, though performance remained lower than prior RAG studies in cancers like breast and CNS suggesting a greater impact on hallucination reduction than on retrieval of a consistent gold standard regimen. These findings support continued development and prospective validation of guideline-anchored LLMs for clinical use.

365 Background: Automated scalp cooling (ASC) is utilized in breast cancer patients to reduce chemotherapy-induced alopecia. Nursing staff and patient perspectives on ASC are limited regarding impact on care delivery and outcomes. Methods: A survey-based study of patient and chemotherapy nurse perspectives of ASC was conducted at three multi-regional Mayo Clinic sites. One-time surveys were sent to breast cancer patients and nurses for self-reporting of the ASC experience. Patients were identified via ICD-10 codes for breast cancer with subsequent chart review to confirm relevant chemotherapy regimen and ASC use. Chemotherapy nurses were identified by their respective chemotherapy unit nurse managers. Patient survey responses were compared between groups that did or did not use ASC by use of Pearson chi-square test for frequency data and the Kruskal-Wallis rank sum test for continuous measures. P values <0.05 were considered statistically significant. Results: Surveys were sent to 181 nurses with 61 (33.7%) completing them. Nurses reported ASC administration to 10-50 patients (n=43, 63.9%) or ≥50 patients (n=18, 29.5%). Most felt they had little or no technical support available (67.2%), witnessed scheduling difficulties (52.5%), were distressed by workload intensity (52.5%), and did not receive adequate training (67.2%). Most felt that ASC demanded greater nursing time compared to other duties (57.4%). Most would not recommend ASC to a friend or family member (88.6%). Surveys were sent to chemotherapy patients with 93 of 465 (20.0%) ASC patients and 56 of 341 (16.4%) without ASC (control) completing them. Controls were significantly more likely to experience hair loss, shave their head, and wear a wig (all p<0.001). There was no significant difference between groups for the EORTC QLQC30, EORTC BR23, and Body Image Scale results or frequency of headaches. Patients who used ASC for less than half of their chemotherapy sessions reported significantly more cold discomfort ( p=0.040) and dizziness ( p=0.003). Time investment (33.3%) and side effects (31.2%) were the most frequent burdens cited. Conclusions: Most nurses indicated that ASC increased nursing workload compared with other duties, and they lacked adequate training and technical support. Increasing formal nurse education and access to technical support documents and live technical support may be options for improving nursing satisfaction with ASC. Patient responses suggest qualitative effectiveness of ASC in reducing hair loss. While those with ASC were less likely to shave or use a wig, this did not translate to significant differences in quality of life or psychosocial body image compared to controls. Cold discomfort and dizziness were the most frequent side effects among those with lower utilization of ASC.

Abstract This manuscript describes a novel approach for treating patients with long‐term sequelae from hemoglobin Evans (Hb Evans). After instituting conservative therapies for approximately 2 years, our patient's symptoms continually worsened. Therefore, we performed red blood cell exchange (RBCx) to reduce his Hb Evans percentage and his co‐existing elevation of methemoglobin. Our assumptions of clinical benefit were based on our collective experience performing RBCx for patients with sickle cell disease. After the first exchange, pre‐ and post‐laboratory results supported our approach and the patient experienced marked improvement in his clinical signs and symptoms. This report provides preliminary proof of principle for the use of RBCx to treat Hb Evans and other non‐Hb S hemoglobinopathies.

132 Background: Enhancing equitable oncologic care is an increasingly emphasized priority. Our study aims to identify aspects of breast cancer (BC) care in which differences exist based on insurance coverage. Methods: We performed a retrospective, case control study, (from 2014-2020); 39 Hispanic ethnicity uninsured patients (UP) from underserved communities with newly diagnosed BC and 119 insured patients (IP) diagnosed at Mayo Clinic Arizona (MCA). Patients were matched 3:1 for age, stage, year of diagnosis, estrogen receptors and HER-2 status. Demographic information, clinical variables, and zip code level specific socioeconomic information were compared. Continuous variables were compared by Wilcoxon rank-sum test and categorical variables by chi-square test. All patients were treated at MCA. Results: Similar treatment patterns with radiotherapy, chemotherapy and surgery were observed between groups. Primary language was Spanish for 94% of UP and English for 97.5% of IP. The majority of UP were of Hispanic ethnicity (97.4%); IP were 83.2% non-Hispanic White, 9.2% Hispanic, 3.4% African American. Zip code level information reflected more unemployment with a median of 10.6% versus 6.9% p < 0.001, percent of high school or lower (53.0 % v 23.2 %, p < 0.001), and lower income for UP (33733.5 v 64728.0 p values < 0.001).BMI was significantly higher for UP (30.6 V 24.7, p = 0.005), with presence of more co-morbidities; diabetes (28.2% v 5.0%, p < 0.001), hypertension (35.9 % v 20.2%, p = 0.046), dyslipidemia (28.2% v 12.6%, p = 0.023), metabolic syndrome (p 23.7% v 8.5, p = 0.013), and tobacco use (17.9% v 2.5%, p < 0.001). Genetics consultation was performed for 62.2% IP versus 35.9% UP (p = 0.004), lower acceptance of nutrition consultation for UP (29.4% vs 7.4%, p = 0.024). Median time from mammogram to biopsy (25.5 days vs. 14 days, p = 0.056), and interval from diagnosis to treatment (62 days vs. 39 days) (p = 0.001) were less favorable for UP compared to IP. Conclusions: In comparing the status of UP and IP with newly diagnosed BC, we identified greater prevalence of co-morbidities and adverse social determinants of health in the former group. We identified access to genetic counseling, nutrition consultation, and timeliness of diagnostic biopsy and initiation of treatment as disparate features in the care pathway. These observations allowed development of tailored interventions to achieve greater equity in delivery of BC care at Mayo Clinic.

e18518 Background: Initiatives enhancing equitable oncologic care are an increasingly emphasized priority. Our study aims to identify aspects of breast cancer (BC) care in which differences exist based on insurance coverage status. Methods: We performed a retrospective, case control study consisting of 39 Hispanic ethnicity uninsured patients (UP) with newly diagnosed BC at federally qualified health centers and 119 insured patients (IP) diagnosed at Mayo Clinic Arizona (MCA). Patients were matched 3:1 for age, stage, year of diagnosis, ER and HER2 status. Demographic information, clinical variables, and zip code level specific socioeconomic information were compared. Continuous variables were compared by Wilcoxon rank-sum test and categorical variables by chi-square test. All patients ultimately received their cancer treatment at MCA. Results: Similar treatment patterns with chemotherapy, surgery, and radiation treatment were observed between groups. Primary language was Spanish for 94% of UP and English for 97.5% of IP. The majority of UP were of Hispanic ethnicity (97.4%); IP were 83.2% non-Hispanic White, 9.2% Hispanic, 3.4% African American. Zip code level information reflected more unemployment with a median of 10.6% versus 6.9% p ˂ 0.001, percent of high school or lower (53.0 % v 23.2 %, p ˂ 0.001), and lower income for UP (33733.5 v 64728.0 p values ˂ 0.001). UP BMI was significantly higher (30.6 V 24.7, p=0.005), with presence of more co-morbidities; diabetes (28.2% v 5.0%, p ˂ 0.001), hypertension (35.9 % v 20.2%, p= 0.046), dyslipidemia (28.2% v 12.6%, p = 0.023), metabolic syndrome (p 23.7% v 8.5, p= 0.013), and tobacco use (17.9% v 2.5%, p ˂ 0.001). IP had higher alcohol use (52.9% v 5.3%, p ˂ 0.001). Genetics consultation was performed for 62.2% IP versus 35.9% UP (p=0.004), lower acceptance of oncology nutrition consultation for UP (29.4% vs 7.4%, p= 0.024) Median time from abnormal mammogram to biopsy (25.5 days vs. 14 days, p=0.056), and interval from diagnosis to treatment (62 days vs. 39 days) (p=0.001) were less favorable for UP compared to IP. Conclusions: In comparing the status of UP (primarily Hispanic, Spanish-speaking) and IP (primarily non-Hispanic White, English-speaking) with newly diagnosed BC we identified greater prevalence of co-morbidities and adverse social determinants of health in the former group. We identified access to genetic counseling services, access to oncology nutrition consultation, and timeliness of diagnostic biopsy and initiation of treatment as disparate features in the care pathway. These observations can allow development of tailored interventions to achieve greater equity in delivery of BC care.

Abstract Background: Hereditary factors play a key role in the risk of developing breast cancers. Identification of a germline predisposition can have important implications for treatment decisions, risk-reducing interventions, cancer screening, and testing for family members. Aim: To determine the prevalence of pathogenic or likely pathogenic germline mutations (P/LP) using a “universal” testing approach and uptake of no-cost cascade family testing in patients with breast cancer. Methods: We undertook a prospective multi-site study of germline genetic alterations among breast cancer patients receiving care at Mayo Clinic cancer centers in Rochester, MN; Eau Claire, WI; Jacksonville, FL and Phoenix, AZ between April 1, 2018 and March 31, 2020. Patients with a new or active breast cancer diagnosis (all stages) irrespective of cancer family history were tested with a &gt;80-gene next generation sequencing panel.Results: Of 390 patients, the median age was 58 years (SD 12.3), 1% were male, 85% were white and 28% had advanced (stage 3-4) disease. P/LP were found in 12.1% (n=47) of patients, including 29 in moderate and high penetrance cancer susceptibility genes. 13 (3.3%) patients had mutations in BRCA1 or 2, while 33 (8.4%) had mutations in BRCAness (ATM, BAP1, BARD1, BLM, BRCA1, BRCA2, BRIP1, CHEK2, NBN, PALB2, RAD50, RAD51C, RAD51D, WRN) associated genes. Of the P/LP findings the most frequent aberrations were in BRCA2 (13.5%), BRCA1 (11.5%), CHEK2 (11.5%), MUTYH (11.5%), and WRN (9.6%). Variants of uncertain significance were found in 209 (53.6%) including 26 (6.6%) with concurrent P/LP and VUS. 37 (9.4%) patients had mutations associated with published management recommendations, precision therapies and/or clinical trial eligibility. 10 (2.6%) patients had P/LP that would not have met current screening guidelines, including 4 with moderate or high penetrance mutations. Patients with younger age of diagnosis were less likely than patients with older age of diagnosis to have a P/LP mutation (OR= 0.47, 95%CI: 0.25-0.90 p = 0.020). Only 10 (21%) patients with P/LP had family members undergo familial site- specific testing at no cost.Conclusions: In this prospective multi-center study of unselected breast cancer patients, universal multi-gene panel testing found that 1 in 8 patients harbor P/LP germline variants. Current guidelines were able to identify the majority of patients with P/LP mutations. Familial site specific testing is greatly under-utilized even when cost is not a barrier. Multigene panels impact cancer patient care by identifying precision medicine treatment interventions, and guiding long-term medical management and preventive surveillance. Citation Format: Brenda Ernst, Natalie Ertz-Archambault, Deborah J Rhodes, Donald W Northfelt, Myra Wick, Douglas L Riegert-Johnson, Scott Okuno, Katie Kunze, Michael Golafshar, Cindy M Azevedo, PLS Uson Junior, ED Esplin, Robert Nussbaum, Margaret Klint, Sarah Mantia, Megan Hager, Keith Stewart, Niloy Jewel Samadder. Universal genetic testing in breast cancer patients: A multi-center prospective study [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PD10-05.

Objective: To evaluate the incidence of and characterize the presentation of neuropsychiatric symptoms and/or Parkinsonism as a presentation of central nervous system lymphoma (CNSL) in either its primary CNSL form or when it spreads to the brain in systemic diffuse large B-cell lymphoma (secondary CNSL). Patients and Methods: With Institutional Review Board approval we identified patients who had been treated at Mayo Clinic from 1998 to 2018 and were recorded to have a combination of ICD 9/10 codes for CNSL and various psychiatric diagnoses. Results: A total of 20 of the 232 patients (9%) were noted to have neuropsychiatric symptoms preceding diagnosis. The average age at diagnosis was 62, with even split for sex. The majority (85%) of patients had primary CNSL. The average duration of symptoms before the diagnosis was 4.8 months. Confusion (80%), depression (40%), apathy (30%), anxiety (30%), and agitation (30%) were the most common symptoms identified. The majority (65%) of patients had subcortical lesions followed by the frontal lobe (50%). Parkinsonism was identified in 5 of the 20 patients with 4 demonstrating resolution of symptoms with treatment of the lymphoma. Conclusions: Neuropsychiatric symptoms are a rare but notable symptom before the presentation of CNSL. There is an increasing awareness of neurological illness presenting as pure psychiatric disturbance, prompting the need to exclude organic and treatable diseases, particularly in elderly patients. Acknowledgment and diagnosis are important for an appropriate management as there is a significant impact on patient and caregiver quality of life.