Ghaffar, Umar

Abstract Clinical empathy is essential for patient care, but physicians need continually balance emotional warmth with factual precision under the cognitive and emotional constraints of clinical practice. This study investigates how large language models (LLMs) can function as empathy editors, refining physicians’ written responses to enhance empathetic tone while preserving underlying medical information. More importantly, we introduce novel quantitative metrics, an Empathy Ranking Score and a MedFactChecking Score to systematically assess both emotional and factual quality of the responses. Experimental results show that LLM-edited responses significantly increase perceived empathy while preserving factual accuracy compared with fully LLM-generated outputs. These findings suggest that using LLMs as editorial assistants, rather than autonomous generators , offers a safer, more effective pathway to empathetic and trustworthy AI-assisted healthcare communication.

Importance Prostate-specific membrane antigen positron emission tomography (PSMA-PET), a form of molecular imaging, increases detection of advanced and metastatic prostate cancer, but its potential to increase treatment costs due to increased detection is poorly defined. Objective To estimate lifetime health and cost outcomes and evaluate the cost-effectiveness associated with implementing PSMA-PET in the clinical evaluation of a patient with a biochemical recurrence (BCR) of prostate cancer, with evaluation of sources of uncertainty in the underlying evidence and assessment of the need and focus of additional data collection. Design, Setting, and Participants This economic evaluation developed a decision-analytic model consisting of a decision tree and a Markov model informed by published literature and a study of patients with BCR prostate cancer undergoing PSMA-PET after definitive surgery or radiation therapy at 2 high-volume US academic centers. The analysis was performed between August 1, 2024, and May 1, 2025. Interventions Patients underwent (1) PSMA-PET, (2) PSMA-PET imaging as a reflex test if computed tomography plus bone scan (CTBS) was negative or equivocal (CTBS + PSMA-PET), and (3) CTBS alone. Main Outcomes and Measures Lifetime quality-adjusted life-years (QALYs) and costs, incremental cost-effectiveness ratio (ICER), and value of information estimates were assessed, with 95% uncertainty intervals (UIs). We assumed a willingness-to-pay threshold of $150 000 per QALY. Results The model simulated 1000 patients with BCR (median age, 66 years) and estimated that up-front PSMA-PET is expected to have the highest mean QALYs (7.12 QALYs [95% UI, 6.71-7.51 QALYs]) compared with 6.55 QALYs (95% UI, 6.08-7.03 QALYs) for CTBS; it was also estimated to have higher mean costs ($451 000 [95% UI, $336 000-$577 000]) compared with $351 000 (95% UI, $263 000-$455 000]) for CTBS; consequently, PSMA-PET had mean increments of $99 000 (95% UI, $55 000-153,000) in costs and 0.58 QALYs (95% UI, 0.35-0.82 QALYs) compared with CTBS, leading to an ICER of $172 000 per QALY, which exceeded the assumed $150 000/QALY willingness-to-pay threshold. In patients with lower prostate-specific antigen levels (<2 ng/mL), PSMA had mean increments of $61 000 (95% UI, $21 000-$104 000) in costs and 0.54 QALYs (95% UI, 0.34-0.77 QALYs) compared with CTBS, leading to an ICER of $113 000/QALY, suggesting it had the potential to be cost-effective. Because of high decisional uncertainty, additional information regarding outcomes and diagnostic characteristics may be associated with reduced uncertainty and a gain of 15 747 QALYs. Conclusions and Relevance This study found that PSMA-PET may be associated with improved disease detection but may not be cost-effective due to high treatment costs offsetting moderate improvements in QALYs. Further research is needed to identify patient subsets and clinical scenarios in which molecular imaging provides the greatest value.

Purpose: Prostate-specific membrane antigen positron emission tomography (PSMA-PET) improves prostate cancer (PCa) localization, particularly in biochemical recurrence (BCR). The primary objective was to assess management patterns after PSMA-PET for BCR at 2 academic centers. Methods: This study evaluated clinical management within 6 months after first PSMA-PET for patients with BCR after local therapy. Management categories included no treatment, local therapy, systemic therapy, and metastasis-directed therapy (MDT). We also evaluated clinical management decision variation across strata of PSMA-PET results PSA levels (<2.0, 2.0-4.99, and ≥5.0 ng/mL) at time of imaging. Results: We identified 581 patients who received PSMA-PET from February 2021 to June 2023. PSMA-PET findings were negative in 142 (24%) and demonstrated localized recurrence in 86 (15%), regional nodal disease in 60 (10%), and distant metastasis in 293 (50%). Distant metastasis was identified in 105 (35%) patients with PSA < 2.0, 67 (60%) with PSA 2.0 to 4.99, and 115 (71%) with PSA ≥ 5.0. In patients with negative PSMA-PET, 63 (44%) received treatment. Conversely, 92% with detected metastases were treated (systemic therapy in 189 [64%], MDT in 75 [26%], and local therapy in 7 [2%]). Treatments differed by the PSA level within strata of PSMA-PET findings. Among 293 patients with metastases, MDT was used more commonly in patients with PSA < 2.0 (37%) than PSA 2.0 to 4.99 (30%) or PSA ≥ 5.0 (13%). Conclusions: Most patients undergoing PSMA-PET for BCR had identifiable disease and initiated therapy within 6 months. Identification of distant metastases and treatment choices differed by the PSA level at imaging. Prospective research is needed to determine whether earlier detection and treatment initiation improve clinical outcomes.

520 Background: The annual incidence of renal cancer is 0.02% among the general United States population. End stage renal disease (ESRD) and exposure to dialysis are thought to be the primary drivers of increased risk of renal cancer in renal transplant patients rather than immunosuppression. This has been evidenced by how the incidence of RCC in heart transplant recipients is no different than the normal population. In this study, we aim to characterize RCC incidence among kidney transplant patients (usually exposed to ESRD/dialysis) versus other solid organ transplant patients (usually unexposed to ESRD/dialysis). Methods: We used CPT codes to search for patients who had received a solid organ transplant (kidney, liver, heart, lung, pancreas) at all Mayo Clinic sites from 1/1/2018 to 1/1/2024. Within this cohort, we identified those who had been diagnosed with kidney cancer after initiating dialysis or receiving a transplant, and retrospectively extracted data related to demographics, cancer diagnosis, and transplant details. Results: We identified 10873 patients who had received a solid organ transplant (Table 1). Of these, 178 (1.76%) were diagnosed with kidney cancer after either starting dialysis or receiving a transplant. These 178 patients were of median age 56 years at time of cancer diagnosis (STD 12), 71% male, 70% white, 20% black, 4.5% Asian, 2.8% American Native. Of 5792 patients who had received a kidney transplant (with or without another solid organ transplant), 172 (2.97%) were diagnosed with kidney cancer after having started dialysis or receiving a kidney transplant. Of these 172, 156 (91%) were exposed to dialysis (107 primarily hemodialysis and 48 peritoneal). Of 4392 patients who had received a solid organ transplant other than kidney (liver, heart, lung, and/or pancreas), 6 (0.14%) were diagnosed with kidney cancer after having received a kidney transplant. Of these 6, none underwent dialysis. Conclusions: Patients with kidney transplants are at increased risk of kidney cancer compared to patients with non-kidney solid organ transplants (2.97% vs 0.14%). This supports prior evidence showing ESRD plays a significant role in the pathogenesis of kidney cancer. Patients who received solid organ transplants at Mayo Clinic from 2018-2024 and were subsequently diagnosed with kidney cancer. n % With RCC after TXP 178 1.7587% Cancer Diagnosis after Kidney HD/TXP 172 2.9696% Cancer Diagnosis after Other TXP 6 0.1386% Patients who received solid organ transplants at Mayo Clinic from 2018-2024 n % Unique Patients 10121 100% Kidney Transplant 5792 57.23% Alone 5088 50.27% &Liver 332 3.28% &Heart 119 1.18% &Lung 8 0.08% &Pancreas 245 2.42% Non-kidney Transplant (Liver, Heart, Lung, or Pancreas) 4329 42.77% TXP: Transplant, HD: Hemodialysis.

428 Background: Prostate cancer management presents a significant healthcare burden, with the need to efficiently triage patients for treatment. Our objective is to leverage large language models to predict physician-recommended treatment plans from unstructured clinical notes. By accurately predicting treatment plans, we aim to risk stratify and triage patients effectively, thereby optimizing the allocation of physician resources. Methods: 448 unstructured initial urology consultation patient notes following first positive prostate cancer biopsy were identified. The recommended and final treatments received were manually annotated to establish ground truth labels (Table 1). The dataset was split 80:20 for training and testing, preprocessed to remove plan sections and formatted into question-answer (QA) format. A domain-specific large language model (LLM) inspired by GPT and a specialized tokenizer (PCa- LLM) for prostate cancer terminology were developed. QA models were built using the PCa-LLM and compared with those using GPT-2 as the backbone to predict recommended and final treatments. Results: For the physician-recommended treatment plans, our LLM (PCa-LLM) showed superior performance with higher AUROC scores for curative vs. non-curative treatments (0.78 vs. 0.65), chemo-hormonal vs. other non-curative treatments (0.89 vs. 0.65), and surveillance vs. all other treatments (0.72 vs. 0.70), while both models achieved the same high AUROC of 0.99 for chemo-hormonal vs. all other treatments. For final treatments, PCa-LLM demonstrated better AUROC for curative vs. non-curative treatments (0.77 vs. 0.74) and chemo-hormonal vs. other non-curative treatments (0.71 vs. 0.66), while GPT2 outperformed PCa-LLM for surveillance vs. all other treatments (0.78 vs. 0.70). Both models achieved an AUROC of 0.99 for chemo-hormonal vs. all other treatments. Conclusions: PCa-LLM accurately predicted most treatment categories better than GPT2, with higher AUROC scores, and can be utilized to triage prostate cancer patients using initial consultation notes. Task Physician -Recommended Treatment Plan Final Treatment Received Curative Prostatectomy/Radiation 228 230 Non-Curative Focal Therapy 22 15 Active Surveillance 40 45 Chemo-hormonal 30 30 Model Predictions (AUROC) GPT2 PCa-LLM GPT2 PCa-LLM Curative vs. non-curative 0.65 0.78 0.74 0.77 Chemohormonal vs. other non-curative 0.65 0.89 0.66 0.71 Chemohormonal vs. all other 0.99 0.99 0.99 0.99 Surveillance vs. other non-curative 0.64 0.60 0.67 0.59 Surveillance vs. all other 0.70 0.72 0.78 0.70

ImportanceAlthough prostate-specific membrane antigen positron emission tomography (PSMA-PET) has shown improved sensitivity and specificity compared with conventional imaging for the detection of biochemical recurrent (BCR) prostate cancer, the long-term outcomes of a widespread shift in imaging are unknown.ObjectiveTo estimate long-term outcomes of integrating PSMA-PET into the staging pathway for recurrent prostate cancer.Design, Setting, and ParticipantsThis decision analytic modeling study simulated outcomes for patients with BCR following initial definitive local therapy. Inputs used were from the literature and a retrospective cohort study conducted at 2 institutions. The base case analysis assumed modest benefits of earlier detection and treatment, and scenario analyses considered prostate-specific antigen (PSA) level at imaging and different outcomes of earlier vs delayed treatment. The analysis was performed between April 1, 2023, and May 1, 2024.Exposures(1) Immediate PSMA-PET imaging, (2) conventional imaging (computed tomography and bone scan [CTBS]) followed by PSMA-PET if CTBS findings were negative or equivocal, and (3) CTBS alone.Main Outcomes and MeasuresThe main outcomes were detection of metastases, deaths from prostate cancer, and life-years and quality-adjusted life-years (QALYs) gained.ResultsThe model estimated that per 1000 simulated patients with BCR (assumed median age, 66 years), PSMA-PET is expected to diagnose 611 (95% uncertainty interval [UI], 565-656) patients with metastasis compared with 630 (95% UI, 586-675) patients diagnosed using CTBS followed by PSMA-PET and 297 (95% UI, 202-410) patients diagnosed using CTBS alone. Moreover, the estimated number of prostate cancer deaths was 512 (95% UI, 472-552 deaths) with PSMA-PET, 520 (95% UI, 480-559 deaths) with CTBS followed by PSMA-PET, and 587 (95% UI, 538-632 deaths) with CTBS alone. Imaging with PSMA-PET yielded the highest number of QALYs, which were 824 (95% UI, 698-885) higher than CTBS. These results differed by PSA level at the time of testing, with the highest incremental life-years and QALYs and lowest number of deaths from prostate cancer among patients with PSA levels of at least 5.0 ng/mL. Finally, the estimates were sensitive to the expected benefit of initiating therapy for recurrent prostate cancer earlier in the disease course.Conclusions and RelevanceThe results of this decision-analytic model suggest that upfront PSMA-PET imaging for the evaluation of BCR is expected to be associated with reduced cancer mortality and gains in life-years and QALYs compared with the conventional imaging strategy, assuming modest benefits of earlier detection and treatment.

e16531 Background: Hypoxia-inducible Factor-2α (HIF-2α) is a transcription factor that accumulates in clear cell renal cell carcinoma (ccRCC) and is responsible for genes involved in oncogenesis. Belzutifan is a HIF-2α inhibitor that has shown promising anti-tumor effects in patients with ccRCC. Our objective is to summarize the available literature on the safety and efficacy profile of the drug. Methods: An electronic search was conducted on PubMed, Embase, and Cochrane from inception till January 31st, 2024. We identified patients aged ≥ 18 years with ccRCC who received Belzutifan. We extracted data on objective response rate, complete response rate, any grade adverse effect (AE) and grade ≥3 AE. Data was analyzed using R (v.4.3) and reported as proportions with respective 95% confidence intervals (CI) on forest plots. Results: Three studies (n=168 patients) were identified that reported the use of Belzutifan in ccRCC. Our analysis showed that pooled proportion of objective response rate and complete response rate was 35% (95% CI: 24% – 47%) and 1% (95% CI: 0% – 5%) respectively. Pooled proportion of patients experiencing any grade AE and grade ≥3 AE was 98% (95% CI: 95% – 99%) and 49% (95% CI: 29% – 69%) respectively. Conclusions: Belzutifan is a novel drug that inhibits HIF-2α which in turn results in anti-tumor effects of the drug against ccRCC. The drug has an acceptable safety and efficacy profile from limited data but more studies are needed to establish a complete picture of drug’s adverse effects and clinical benefits. Pooled proportion of patients developing (A) Objective Response, (B) Complete response (C) Any adverse effect and (D) Grade ≥3 adverse effects. [Table: see text]

Abstract Objectives ANCA-associated vasculitis (AAV) is currently categorized under the small vessel vasculitides. There is limited knowledge about large vessel involvement in AAV (L-AAV), mainly described in case reports and small series. L-AAV can involve temporal arteries (TA-AAV), aorta (A-AAV), and periaortic soft tissue (PA-AAV). We sought to characterize the features of patients with L-AAV. Methods Patients older than 18 years at diagnosis of TA-AAV, A-AAV and PA-AAV seen at the Mayo Clinic, Rochester between 1 January 2000 and 31 December 2021 were identified through a proprietary medical text search algorithm. Patients were included if diagnosed with L-AAV, fulfilled 2022 ACR/EULAR classification criteria for GPA, MPA or EGPA, had positive ANCA test results, and had more than one outpatient or inpatient visit. Results The study cohort consists of 36 patients with L-AAV. Of those, 23 had p-ANCA and/or MPO-ANCA, and 13 had c-ANCA and/or PR3-ANCA. Mean (s.d.) age at AAV diagnosis was 63.4 (12.79) years; 20 (56%) were male. Seventeen patients had TA-AAV, 10 had A-AAV and 9 had PA-AAV. Most patients (n = 25, 69%) were diagnosed with large vessel vasculitis and AAV within a 1-year timespan. Twenty-five (69%) patients had histopathological confirmation of AAV diagnosis in a location other than temporal artery, aorta or periaortic soft tissue. Glucocorticoids (36/36), rituximab (19/36) and methotrexate (18/36) were the most frequent treatments. Conclusion This is the largest single-centre cohort of patients with L-AAV to date. AAV can involve large arteries, albeit infrequent. AAV-targeted therapy should be considered in patients with L-AAV.