Citations

Abstract Pain assessment plays a critical role in clinical decision-making, yet conventional methods rely heavily on subjective self-reports such as the Visual Analog Scale (VAS) or the Numeric Rating Scale (NRS). These approaches fail in scenarios involving infants, sedated patients, unconscious individuals, or those with cognitive or communication impairments. To address this gap, this research proposes PainNet, an AI-driven, real-time facial pain recognition system that integrates deep learning-based object detection with Action Unit (AU)–based pain intensity scoring. The system uses YOLOv8 for rapid facial expression detection and incorporates Py-Feat to extract pain-relevant AUs, enabling computation of the Prkachin and Solomon Pain Intensity (PSPI) score. A balanced dataset comprising 500 images sourced from Delaware, BioVid, and RaFD was curated, annotated, and preprocessed for training. The YOLOv8-based binary pain vs. no-pain classifier achieved a mean Average Precision (mAP) of 0.89 and an F1-score of 0.80 on the test dataset, demonstrating strong performance in diverse environments and facial variations. The design emphasizes computational efficiency by activating AU analysis only when YOLOv8 detects probable pain, making PainNet suitable for continuous monitoring in ICUs, elder-care facilities, and telemedicine systems. This study demonstrates that the fusion of real-time object detection and clinically validated AU scoring can significantly advance automated, objective, and scalable pain assessment.

ABSTRACT Cancer mortality is inequitably distributed across the U.S. and globally, with historically marginalized groups, particularly racial and ethnic minorities, being disproportionately affected. Disproportionate representation in clinical cancer research is a consequence of inequitable access as well as a central factor contributing to outcome disparities; it limits efficacy and safety data for underrepresented groups while directing drug development to be most beneficial for those enrolled. To tackle this problem and improve the efficacy and safety of commercially available therapies, a consensus conference on strategies to promote Diversity, Equity, Inclusion, and Accessibility (DEIA) in cancer research was empaneled in September 2022. The resulting “Indianapolis Black Paper” presented here is an international collaborative white paper with authors from industry, academia, and the government, as well as patient advocacy and community health groups. First, it outlines the roles and responsibilities of key groups involved in clinical cancer research, as they relate to representation and equity. The Black Paper then delineates five steps to promote diversity in clinical research. This actionable strategy is called “DRIVE,” and it involves: incorporating a Diversity officer into clinical research studies, Ranking clinical studies for diversity, developing an Individual diversity, equity, inclusion, and access plan, a process for Verifying study diversity, and Elevating and Enhancing training of minority investigators and research team members. Finally, we outline the timeline and milestones of implementation. Purposefully ambitious and crosscutting industry, academia, and governmental policy, the Black Paper aims to address “injustices in health” with enhanced safety, purpose, and humanity.

e19503 Background: The anti-B-cell maturation antigen (BCMA) antibody-drug conjugate (ADC) belantamab mafodotin is approved in the USA and EU for relapsed/refractory multiple myeloma (RRMM). Understanding treatment options and patient outcomes following ADC failure is clinically important. This study describes post-ADC treatment patterns and clinical outcomes in real-world patients with RRMM. Methods: Adult patients with MM who received an ADC and initiated ≥1 subsequent line of therapy (LOT) were included. The index date was the start of the first post-ADC LOT. Data were obtained from chart reviews at International Myeloma Foundation International Myeloma Working Group (IMF IMWG) academic sites, and US electronic health records from largely community sites (COTA and Guardian Research Network [GRN]). Patient characteristics and index regimens were summarized descriptively. Time to next treatment (TTNT) and overall survival (OS) were estimated with the Kaplan-Meier estimator. Results: The IMF IMWG cohort (n=40) had a mean age of 70.8 years, median of 7 prior LOTs and 82.5% were triple-class refractory (TCR). Overall, 85% received ADC monotherapy (70% belantamab mafodotin) and 15% combination therapy. Pre-ADC and not considering corticosteroids, patients were commonly exposed to lenalidomide (R; 100%), bortezomib (V; 92.5%), daratumumab (D) and pomalidomide (P; 90% each). Over a median follow-up of 26 months after index date, 29 unique index post-ADC regimens were reported, frequently selinexor (27.5%) and bispecific antibodies (15%). CAR T-cell therapy was administered to 5%, with 60% re-treated with a pre-ADC drug class (PI/IMiD/anti-CD38) after ADC. Median (95% CI) TTNT and OS were 10.8 (7.0–20.6) and 19.0 (13.4–25.1) months, respectively. The COTA-GRN cohort (n=39) had a mean age of 67.5 years, median of 8 prior LOTs, and 51.3% were TCR. Overall, 56% received ADC monotherapy (belantamab mafodotin) and 44% combination therapy. Pre-ADC and not considering corticosteroids, patients were commonly exposed to D and R (97.4% each), and carfilzomib, V and P (94.9% each). Over a median follow-up of 9.2 months, 32 unique index post-ADC regimens were reported, frequently selinexor (23.1%) and venetoclax (17.9%) based; 0% received CAR Ts and 54% were re-treated with a pre-ADC drug class (PI/IMiD/anti-CD38) after ADC. Median (95% CI) TTNT and OS were 4.9 (2.8–non-estimable [NE]) and 12.0 (4.9–NE) months, respectively. Conclusions: Findings highlight that most patients with heavily pretreated BCMA-ADC exposed MM were re-treated with similar classes. Academic centers used novel therapies (ie, CAR T, bispecific antibodies) more often and their patients had longer survival than those treated at community sites. However, clinical outcomes remained sub-optimal, emphasizing the unmet need for effective, accessible post-ADC treatments.

ABSTRACT Introduction Community health workers (CHWs) help patients navigate community resources. CHW programs can improve health outcomes and reduce healthcare utilization, but identifying eligible patients is challenging. We developed an electronic health record (EHR)‐based algorithm trained to predict referrals to a CHW program. Our objective was to evaluate whether an algorithm trained on historical referrals identifies a subgroup of patients that differs from the subgroup identified through questionnaire‐based health‐related social needs (HRSN) screening alone. Methods This analysis used data from Mayo Clinic primary care patients in Southeast Minnesota. We developed a gradient boosted time‐to‐event model that incorporated demographics, diagnoses, HRSN questionnaire responses, and HRSNs derived from clinical notes using natural language processing. We compared the characteristics of algorithm‐identified patients with those identified through questionnaires and assessed the variable importance of model features. Results The algorithm had AUC: 0.87 in the training and 0.92 in the hold out set. Patients flagged by the algorithm tended to use interpreter services more commonly (12.6% vs. 1.3%), have a non‐English preferred language (14.1% vs. 2.3%), and have more instances of health literacy documented (5.36 vs. 1.29) compared to patients identified by the HRSN questionnaire. Only 47.0% of patients flagged by the algorithm had responded to the HRSN questionnaire within 2 years. Variable importance was highest for health literacy and financial strain from unstructured notes. Conclusions The algorithm is a chart‐review prioritization tool for referral to the CHW program for patients that may overcome challenges to identifying patients with HRSNs, especially when questionnaire data is incomplete.

9525 Background: Melanocortin-1 receptor (MC1R) is a novel target for radiopharmaceutical therapy and is highly expressed on melanoma tumor cells. VMT01 is an MC1R-targeted peptide. 203 Pb is used for patient selection via SPECT imaging. Peptide radiolabeling with 212 Pb delivers alpha-particle therapy. Here, we present data on the treatment of adult patients with MC1R-positive metastatic melanoma receiving [ 212 Pb]VMT01 either as a monotherapy or in combination with the programmed death-1 (PD-1) checkpoint inhibitor, nivolumab. Methods: This is an ongoing phase I/IIa, first-in-human, prospective, multicenter, open-label, radioactive dose-finding and dose-expansion trial. The objectives are to investigate safety, pharmacokinetics, dosimetry, and efficacy. Participants receive up to 3 treatment cycles with either [ 212 Pb]VMT01 3 mCi, 5 mCi, or 1.5 mCi monotherapy, or combination therapy with [ 212 Pb]VMT01 (1.5 mCi) + nivolumab (480 mg, Q4W) or [ 212 Pb]VMT01 (3 mCi) + nivolumab. Participants are evaluated for any DLTs for the first 6 weeks after cycle 1. Efficacy is assessed by RECIST v1.1 criteria by the investigator. Results: As of 24 December 2025, 26 participants (53.8% male; median age: 68 years [range: 27-81]) with MC1R-positive metastatic melanoma were enrolled. The median prior systemic therapies was 4. All 26 enrolled participants received ≥1 dose of [²¹²Pb]VMT01. No DLTs were reported. All SAEs (including one grade 5 SAE) were attributed to disease progression. Most TEAEs were grade 1 (19.2%) and grade 2 (34.6%). Grade 3 TEAEs occurred in 8 participants (30.8%), 5 on monotherapy and 3 on combination therapy (3 mCi + nivolumab). Of the 22 evaluable participants, 1 PR, 9 SD, and 12 PD were reported. Four participants had progression-free survival of ≥6 months. Conclusions: [ 212 Pb]VMT01 as a monotherapy and in combination with nivolumab was generally safe and well-tolerated in 26 enrolled participants, with antitumor activity observed at the 3 mCi dose level. These data may be supplemented with a more complete follow-up at the time of the meeting. Clinical trial information: NCT05655312 .

6588 Background: TET2 is the most frequently mutated gene in chronic myelomonocytic leukemia (CMML) and has been associated with favorable outcome, with recent studies suggesting a more pronounced prognostic impact in the presence of ≥2 TET2 mutations ( Csizmar et al. leukemia 2025;39:2030; Kynning et al. Br J Haematol. 2025 Nov 25. doi: 10.1111/bjh.70264 ). We sought to determine if this particular observation extended to patients with ≥3 TET2 mutations ( TET2 MUT ) and also held true in the context of recently reiterated genetic risk factors. Methods: The current study was conducted under institutional review board approved minimum risk protocols allowing retrospective patient data collection and analysis. Diagnostic criteria were according to the International Consensus Classification (Arber et al. Blood 2022. 140:1200). All statistical analyses were conducted using JMP 18 software. Results: 536 CMML patients who underwent CLIA-approved NGS testing were stratified according to the number of TET2 MUT : 221 (41%) wild-type, 148 (28%) with one, 153 (29%) 2, and 14 (2%) ≥3 TET2 MUT . Co-mutation frequencies with significant differences between one vs. two TET2 MUT included ASXL1 (46% vs. 29%; p<0.01), KRAS (7% vs. 14%; p=0.02) and SETBP1 (5% vs. 0.6%; p=0.02). Phenotypic comparisons revealed preponderance of prognostically favorable traits associated with wild-type TET2 but no significant differences between one vs. two TET2 MUT . In age-adjusted univariate analysis, transplant-censored overall survival (TCOS), calculated from the time of mutation detection, was superior in patients with one (p<0.01; HR 0.7) or 2 (p<0.01; HR 0.4) but not ≥3 (p=0.4) TET2 MUT , vs. those with wild-type TET2 . TCOS was also superior with 2 vs. 1 (p=0.02; HR 0.7) but not with ≥3 vs. 1 (p=0.7) TET2 MUT ; the results were similar when TCOS was calculated from the time of diagnosis. Multivariable analysis confirmed the survival advantage of exactly two TET2 MUT vs. wild-type TET2 (p<0.01; HR 0.5), one TET2 MUT (p<0.01; HR 0.6), or ≥3 TET2 MUT (p<0.01; HR 0.4), after adjusting for age, sex, anemia, circulating blasts ≥2%, and leukocyte count ≥13×10⁹/L. The favorable impact of exactly two TET2 MUT persisted in CMML-MD and CMML-MP and remained independent of TET2 VAF, mutation type, karyotype, and other prognostically relevant mutations. This association was sustained in the BLAST (p<0.01; HR 0.5) and CPSS-mol (p=0.02; HR 0.7) models, but not in BLAST-mol model (p=0.12). Conclusions: Mechanistic explanation for this novel observation includes the possibility that TET2 mutational categories based on mutation count reflect distinct biological states rather than having a simple linear effect. The persistence of this signal despite adjustment for established molecular and genetic risk factors supports its consideration in the development of future CMML prognostic models.

632 Background: Prospective data on the risk of breast cancer (BC) recurrence among young BC survivors who have a subsequent pregnancy are limited. We sought to evaluate the long-term impact of pregnancy and live birth on BC outcomes in the Young Women’s Breast Cancer Study (NCT01468246), a prospective multicenter study of women aged ≤40 years at BC diagnosis. Methods: Women with stage 0-III BC without prior hysterectomy were included. The primary endpoint was breast cancer-free interval (BCFI) between patients with and without a (a) pregnancy and (b) live birth after BC. Secondary endpoints were distant recurrence-free interval (DRFI) and overall survival (OS). A time-varying Cox proportional hazards model was performed by estrogen receptor (ER) status, controlling for age at diagnosis, tumor stage, HER2 status, tumor grade, parity at diagnosis, and germline pathogenic variant status. Results: Among 1,016 BC survivors at a median follow-up of 12 (range, 0.5-18.5) years, 198 reported ≥1 pregnancy and 165 reported ≥1 live birth post-diagnosis. Among survivors who reported a post-diagnosis pregnancy, median age at diagnosis was 32 (range, 17-40) years; most had stage I (35%) or II (40%) BC, with 72% hormone receptor (HR)-positive and 25% HER2+; 42% were nulligravid and 60% were nulliparous at diagnosis. Among BC survivors who did not become pregnant post-diagnosis, median age at diagnosis was 37 (range, 21-40) years; most had stage I (35%) or II (43%) BC, with 75% HR-positive and 28% HER2+; 25% were nulligravid and 29% were nulliparous at diagnosis. Among BC survivors with ER-positive tumors (n = 737), pregnancy after BC was not associated with a difference in BCFI (adjusted hazard ratio [HR] 0.60, 95% CI 0.31-1.16, P = 0.130), DRFI (HR 0.73, 95% CI 0.35-1.55, P = 0.416), or OS (HR 0.58, 95% CI 0.24-1.40, P = 0.227). Similarly, live birth after BC did not impact BCFI (HR 0.83, 95% CI 0.45-1.55, P = 0.560), DRFI (HR 1.04, 95% CI 0.52-2.06, P = 0.923), or OS (HR 0.77, 95% CI 0.34-1.76, P = 0.541). Outcomes among BC survivors with ER-negative tumors (n = 278) were also not affected by post-diagnosis pregnancy (BCFI [HR 1.48, 95% CI 0.70-3.14, P = 0.304], DRFI [HR 1.59, 95% CI 0.68-3.75, P = 0.287], OS [HR 0.99, 95% CI 0.39-2.49, P = 0.975]) or live birth (BCFI [HR 1.02, 95% CI 0.41-2.53, P = 0.974], DRFI [HR 0.80, 95% CI 0.24-2.69, P = 0.722], OS [HR 0.39, 95% CI 0.09-1.69, P = 0.208]). Conclusions: In this multicenter, prospective study with 12 years of median follow-up, pregnancy and live birth after BC did not impact BC events or overall survival, irrespective of ER status. These long-term data from a modern cohort provide reassurance for young BC patients interested in future fertility. Clinical trial information: NCT01468246 .

4173 Background: [ 212 Pb]VMT-α-NET is an alpha-particle radiation-delivering agent targeted to somatostatin receptor type 2 (SSTR2)-expressing tumors. Here, we present safety and efficacy results from the dose-finding phase 1/2a clinical trial (NCT05636618). Methods: This phase 1/2a dose-escalation study uses a Bayesian algorithm to identify an optimal dose of [ 212 Pb]VMT-α-NET for participants with advanced, well-differentiated NETs expressing SSTR2. Eligible participants must have unresectable or metastatic disease, documented radiologic progression following at least one prior systemic treatment, and evidence of SSTR2 expression (Krenning Score ³2) in ≥1 lesion confirmed on an FDA-approved somatostatin receptor PET scan. Patients previously treated with systemic peptide receptor radionuclide therapy (PRRT) are excluded from enrollment. During the dose-finding phase, participants may receive up to four administrations of [ 212 Pb]VMT-α-NET at their assigned dose level, and they are monitored for dose-limiting toxicities (DLTs) for 42 days and safety throughout the study. Efficacy is evaluated by investigators according to RECIST criteria v1.1. Dosimetry is included as a supportive analysis. Results: As of the 10-Dec-2025 data cut-off (DCO), 56 participants had been enrolled across Cohorts 1, 2, and 3 and received at least one dose of [ 212 Pb]VMT-α-NET: 2 participants in Cohort 1 (2.5 mCi), 46 in Cohort 2 (5 mCi), and 8 in Cohort 3 (6 mCi). No dose-limiting toxicities, grade 5 adverse events, treatment-related discontinuations, serious renal events, dysphagia, or clinically meaningful treatment-related myelosuppression were observed. The median follow-up for all treated participants was 40 weeks (range 6–97). Efficacy has been summarized in this abstract for 25 participants (2 in Cohort 1 and 23 in Cohort 2). Nineteen of these 25 participants remained progression-free at the time of the DCO, with a median efficacy follow-up of 49 weeks (range 6–97). In Cohort 2, investigators observed RECIST v1.1 objective responses in 9 of 23 participants (39%), including 8 confirmed responses. Both participants in Cohort 1 continued to exhibit stable disease after two years of follow-up. Updated safety for all participants and updated efficacy for participants with sufficient maturity beyond the 25 summarized here will be presented at the congress. Conclusions: Treatment with [ 212 Pb]VMT-α-NET continues to demonstrate a favorable safety profile across all treated participants (n = 56) and shows signs of sustained efficacy at the 2.5 mCi and 5 mCi dose levels. The therapy has been well-tolerated with no observed dose-limiting toxicities or serious treatment-related adverse outcomes. Based on these encouraging safety and efficacy findings, the study is ongoing, and enrollment into Cohort 3 (6 mCi) remains active. Clinical trial information: NCT05636618 .

1617 Background: Large language models (LLMs) are increasingly evaluated for oncology decision support, yet guideline concordance and hallucinations remain variable across cancer types and tasks. Retrieval-augmented generation (RAG) may improve safety by constraining the evidence base used during generation. Prior oncology LLM studies report discordance with guideline-based care, but few describe a reproducible health-system “vetting mechanism” that couples guideline-only retrieval with clinician-scored safety rubrics across multiple oncology domains. Methods: We assembled 316 de-identified oncology cases in a tumor-board template: medical oncology (breast, GI, GU, leukemia; n=216), gynecologic oncology (n=50), and neuro-oncology/CNS metastatic cancer (n=50). Each case was run on three systems: baseline GPT-4 (O1), NCCN-guideline anchored RAG (O2; guideline-constrained retrieval), and a literature-anchored system (OpenEvidence; O3). Oncologists independently scored outputs using a modified Generative Performance Score (mGPS; −1 to +1) combining guideline concordance with hallucination penalties; readability/rationality was scored on a 1–5 Likert scale. Inter-rater reliability was assessed (κ =.63). This workflow was designed as a pre-deployment model credentialing step for health-system governance (source policy → case bank → clinician-scored acceptance criteria). Results: Across all 316 cases (case-averaged), O2 achieved the highest mean total mGPS (0.75) compared with O1 (0.43) and O3 (0.59). In the medical oncology cohort (n=216), O2 outperformed O1 and O3 (mean mGPS 0.72 vs 0.35 and 0.54), with higher guideline concordance (0.85 vs 0.68 and 0.75) and lower hallucination penalties (−0.09 vs −0.25 and −0.19); readability was also highest for O2 (4.17 vs 3.22 and 3.86). In gynecologic oncology (n=50), O2 again performed best (mGPS 0.827) relative to O3 (0.697) and O1 (0.647). In the CNS cohort (n=50), mean composite mGPS favored O2 (0.82) over O3 (0.68) and O1 (0.55), while readability remained stable across models (3.55–3.73). Severe hallucination/unsafe recommendation rate: O1 [26%], O2 [5%], O3 [17%] (pre-specified definition; e.g., major fabricated citation or guideline-discordant harmful recommendation). Conclusions: In a 316-case, multi-domain clinician-scored evaluation, limiting retrieval to oncology guideline sources improved safety-aligned performance relative to unconstrained generation or literature-only anchoring. A guideline-constrained RAG configuration paired with clinician-scored rubrics and tail-risk reporting can serve as a practical health-system vetting mechanism for governance, re-validation, and change control of LLM decision support, including in resource-limited settings where reliable guideline-consistent recommendations are essential.