Abstract Objective: Subjective cognitive complaints (SCC) can precede cognitive decline and are associated with demographic, exposure, lifestyle, and psychological factors. Prevalences of SCC and their correlates in individuals with repetitive head impacts (RHI) are poorly understood. This study characterized SCC in former elite American football players by frequency, mood and behavioral correlates, concordance with informant reports, and associations with neuropsychological test performance, cerebrospinal fluid (CSF), and magnetic resonance imaging (MRI) markers of neurodegeneration. Method: Former American football players ( n = 180) completed measures of global and domain-specific SCC, neuropsychiatric symptom questionnaires, neuropsychological testing, lumbar puncture, and MRI. Elastic net regression evaluated the relative importance of potential SCC correlates. Intraclass correlation coefficients measured concordance between self and informant reports. Multiple linear regressions tested associations between SCC and verbal memory and executive functioning scores. CSF Aβ 1-42 , p-tau 181 , t-tau, neurofilament light (NfL), hippocampal volume, and regional cortical thickness were examined for their potential associations with SCC. Results: Rates of SCC ranged from 43 to 77% depending on the domain. Symptoms of depression, impulsivity, and anxiety were strongly associated with SCC. Self- and informant-reported SCC showed moderate inter-rater agreement. Adjusting for age, race, education, APOE ϵ4 carrier status, and depressive symptoms, SCC were associated with lower objective verbal memory and executive functioning performance. SCC were associated with lower parahippocampal cortical thickness but not with hippocampal volume or any of the measured CSF tests. Conclusions: SCC are strongly associated with neuropsychiatric factors in former American football players. SCC may also be a marker of cognitive decline and neurodegeneration.
Characterization of adrenal masses in young adults at a tertiary care academic medical center
Abstract Context Evidence on the etiology and function of adrenal masses in younger adults is scarce. Objective This work aimed to characterize the clinical, imaging, and biochemical presentation of adrenal masses in 18- to 40-year-olds. Methods A retrospective cohort study was conducted January 2016 to December 2021 at a tertiary academic medical center. Participants included young adults (age 18-40 years) with adrenal masses identified through radiologic registry. Exclusion criteria included preexisting adrenal masses before 2016, hyperplasia, or masses smaller than 1 cm. Outcomes included adrenal mass diagnosis, as determined by reference standard of histopathology, radiologic surveillance, and biochemical evaluation. Results A total of 255 patients (women 142 [56%]; White 156 [61%]; median age 35 years [interquartile range 30-39 years]) were included. Comorbidities included genetic predisposition syndromes (16, 6.3%), preexisting malignancy (70, 27%), obesity (100, 48%), hypertension (89, 35%), and diabetes (25, 10%). The majority of adrenal masses were incidentally found (184, 72%) and median size was 1.9 cm (IQR 1.4-2.7 cm) on detection. Among adrenal masses in young adults, prevalence of malignancy and pheochromocytoma was 16% and 4%, respectively; among adrenal incidentalomas, 5% were malignant. Factors associated with malignancy or pheochromocytoma in multivariable analysis included nonincidental mechanism of discovery (odds ratio [OR] 15.5; 95% CI, 7.2-35.7; P < .001), and larger size (OR 1.5; 95% CI, 1.2-2.0; P < .001). Hormonal evaluation was performed in only 57 (22%) patients and was positive for hormone hypersecretion in 16 (28%) of tested individuals. Conclusion We demonstrated a high prevalence of malignancy and pheochromocytoma in our cohort of young adults with adrenal masses. Completion of appropriate hormonal work-up was infrequent, potentially missing patients with functioning adrenal adenomas.
Evidence that extra copies of chromosome 1q play a role in the early phases of pancreatic neoplasia
We searched for oncogenes activated by copy number increases using whole-genome sequencing data of 535 pancreatic ductal adenocarcinomas (PDACs). We found that gains of 1q were the second most common gain, occurring in 213 (39.8%) of PDACs. Single-cell analysis via fluorescence in situ hybridization on 33 cancers confirmed these results. A portion of 1q, rather than the entire 1q arm, was gained in 75 (14.0%) PDACs, allowing us to pinpoint two ~3-megabase regions of 1q that were nearly always gained. These two regions contained NCSTN and PSEN2 , genes that code two subunits of the γ-secretase complex. Evaluation of 267 precancerous lesions revealed that extra copies of NCSTN and PSEN2 were common (49%) in noninvasive neoplasms (high-grade pancreatic intraepithelial neoplasms), which are at relatively high risk for progression to PDACs, but uncommon (6%) in low-grade pancreatic intraepithelial neoplasia lesions, which have low malignant potential. We hypothesize that γ-secretase genes are genetically activated oncogenes in the early phases of pancreatic neoplasia.
Abstract PS3-06-29: Retrieval-augmented GPT-4 improves NCCN-concordant breast cancer treatment recommendations
Abstract Guideline-concordant breast cancer treatment selection is increasingly complex due to rapid evolution in biomarker-driven therapies. Large language models (LLMs), such as GPT-4, offer a potential tool to support clinical decision-making. However, hallucinations and deviation from evidence-based guidelines remain significant concerns. This study evaluated whether retrieval-augmented generation (RAG) using National Comprehensive Cancer Network (NCCN) guidelines improves GPT-4 performance in treatment recommendation tasks. We developed a retrieval-augmented GPT-4 agent (RAG-GPT) by indexing 2025 NCCN Breast Cancer Guidelines into a vector database using sentence embeddings. Forty clinical vignettes were created from published case reports to reflect diverse scenarios, including HER2-low, triple-negative, and BRCA-mutated disease across all stages. Two models—a baseline GPT-4 (without augmentation) and the RAG-GPT—were prompted with structured templates requesting treatment plans, rationale, clinical trial options, and citations.Outputs were independently scored by two blinded oncology-trained reviewers using the modified Generative AI Performance Score (mG-PS). The mG-PS scored outputs for 2 categories, guideline concordance (Gold standard = 1.0, Acceptable = 0.5, Non-concordant = 0.0) and hallucination errors (Severe = -1.0, Moderate = -0.5, Mild = -0.25, None = 0.0). Guideline Concordance was deemed based on NCCN guidelines and what an oncologist would due based on using the guidelines for each case. Hallucinations are a term used for falsehoods in Artificial Intelligence, like made-up facts or studies, incorrect information, improper doses, or inappropriate or wrong treatments. To ensure consistency and minimize inter-rater variability, reviewers were provided with comprehensive scoring guidelines and detailed instructions for each evaluation metric. Scores were normalized between -1.0 and 1.0 for improved evaluation of tools. Readability and rationale clarity were rated on a 5-point Likert scale. Inter-rater reliability was assessed using Cohen’s kappa. The RAG-GPT demonstrated a significantly higher mean mG-PS score (0.599 ± 0.102, 95% CI: 0.498-0.702) compared to the baseline GPT-4, which not only scored lower but had a negative mean (-0.21 ± 0.143, 95% CI: -0.353 to -0.067). For the mG-PS Cohen’s Kappa test resulted in a score of .69 with substantial aggreament. An independent sample t-test showed a difference (t(78) = 29.1, p < 0.0001). Severe hallucinations were also absent in RAG-GPT outputs (0/40) but occurred in 8/40 baseline GPT-4 outputs. The evaluation of readability and rationality also showed notable differences, with the augmented LLM scoring significantly higher (M = 4.35, SD = 0.66) compared to the naïve LLM (M = 3.05, SD = 1.06). An independent samples t-test confirmed this difference t(65.3) = 6.58, p < 0.0001, further supporting the benefit of augmented Large Language models. Cohen’s Kappa test resulted in a score of .76 for a substantial agreement for readability and rational clarity scale. Limitations include small sample size; further fine-tuning and larger real-world validation studies remain necessary. Linking GPT-4 to NCCN guidelines via retrieval-augmented generation significantly improves treatment accuracy and reduces hallucination severity in breast cancer care scenarios. While further refinement is needed, the reduced incidence of severe hallucinations, increase in guideline-concordant recommendations, and enhanced readability findings support further development and prospective validation of guideline-anchored LLMs for clinical use. Citation Format: C. Yost, A. Aseem, B. Callas, S. Monick, L. Bonilla, R. Nguyen, R. Galamaga, S. Osborn, Y. Kumar, N. Ertz-Archambault. Retrieval-augmented GPT-4 improves NCCN-concordant breast cancer treatment recommendations [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-06-29.
Clinical Pharmacists and Industry: Guidelines for Professional Interactions
ABSTRACT Interactions between clinical pharmacists and industry partners continue to evolve as health care increases in complexity and technology assumes a more dominant role. To address the corresponding ethical challenges, the 2025 Industry Relations Committee of the American College of Clinical Pharmacy (ACCP) was charged with updating the 2008 ACCP position statement, Pharmacists and Industry: Guidelines for Ethical Interactions . This update expands its perspective, reflecting the broader and more influential roles pharmacists play across the health care ecosystem. All clinical pharmacist–industry interactions should be grounded in improving patient care. Through this lens, six recommendations are provided: (1) Pharmacists should not engage in industry partnerships or collaborations that might unduly influence objectivity, independence, or fairness in clinical and professional judgments; (2) Pharmacists should disclose and mitigate financial, consulting, or other relationships that constitute potential conflicts of interest and should not participate in activities where an existing conflict cannot be resolved; (3) Pharmacists who participate in industry‐sponsored research should follow accepted ethical, regulatory, and scientific standards; (4) Pharmacists providing continuing education programs or developing drug information materials should maintain control of content; ensure their content is fair, balanced, and free from bias; and disclose and mitigate any conflicts; (5) Formal instruction on ethical interactions with industry professionals should begin early and continue through all stages of pharmacist education, training, and professional development; (6) Pharmacists should maintain the confidentiality and privacy of patients and other health care professionals when working with industry partners. Questions to guide personal industry interactions are provided. This paper does not detail every emerging area (e.g., artificial intelligence, social media), but recommendations are applicable wherever clinical pharmacists leverage their professional expertise. By maintaining a critical, discerning, and ethical approach to industry collaborations, clinical pharmacists can uphold professional obligations while contributing to the advancement of evidence‐based practice, ultimately prioritizing patient welfare.
Toxicity of Ablative Radiation Therapy in the Management of Patients with Child-Pugh B/C Liver Function and Unresectable Hepatocellular Carcinoma (HCC)
Background/Objectives: Hypofractionated ablative radiation is an increasingly popular option for patients with hepatocellular carcinoma (HCC). However, concern remains about the risk for radiation-induced liver toxicity in patients with decompensated liver function. Methods: We retrospectively identified patients with underlying Child-Pugh (CP) B or C liver function treated at our University and Veterans Affairs (VA) departments from 2014 to 2019. Primary endpoints included treatment-related toxicity and dosimetric parameters. Results: 38 patients were included in the analysis. Most patients (98%) had CP B or Albumin-Bilirubin (ALBI) grade 2–3 (100%) liver disease. The median dose was 50 Gy (range 30–50) delivered in 5 or 10 fractions. Most patients had a single tumor treated (66%) with a median size of 3.1 cm (Interquartile Range (IQR) 2.3–4.1). The mean liver dose was 9.28 Gy (IQR 6.76–13.64) with a liver D800cc of 3.99 Gy (IQR 1.41–8.02). All patients completed their intended course with a median follow-up of 43 months. Four patients (10.3%) developed non-classical radiation-induced liver disease (RILD), comparable to the rate for patients with CP A function treated contemporaneously (8.3%). Otherwise, one patient (2.6%) experienced acute grade 3+ (non-RILD) hepatobiliary toxicity, while one patient (2.6%) experienced late grade 3+ hepatobiliary toxicity. Local control was promising with 2-year freedom from progression in the treated lesion of 73% (95% CI 38–91%). Median overall survival was 12 months (95% CI 5–25 months). Conclusions: Ablative radiation for patients with decompensated liver function and HCC appears well tolerated with low rates of RILD and encouraging local control. With careful selection, these patients should be considered for inclusion in future randomized trials.
Patient-specific Rods Reduce the Rate of Mechanical Complications Following Adult Thoracolumbar Spinal Deformity Surgery
BACKGROUND AND OBJECTIVES: The purpose of this study was to analyze the impact of patient-specific rods on mechanical complications after adult thoracolumbar spinal deformity surgery. METHODS: A consecutive series was analyzed of 200 adult thoracolumbar spinal deformity patients. Half of the patients (n = 100) had surgery just before implementing patient-specific rods, and the other 100 patients had surgery immediately after adoption of patient-specific rods. All patients had instrumented fusions from the thoracic spine to the pelvis. Mechanical complications were defined as rod fracture, proximal junctional kyphosis, proximal junctional failure (revision surgery requiring proximal fusion extension), and/or revision surgery for rod fracture or pseudoarthrosis. Minimum follow-up was 2 years. RESULTS: A total of 200 patients were included with a median (IQR) age of 68 (61, 73), body mass index of 28 (25, 32), and 73% were female. The patients with patient-specific rods compared with those without had less preoperative kyphosis (39° vs 44°, P = .01), a higher median number of rods (4 vs 2, P = .04), more often a combined anterior-posterior approach (61% vs 46%, P = .047), more commonly a upper instrumented vertebra vertebroplasty (67% vs 45% P = .003), and postoperatively had less thoracic kyphosis (45% vs 49%, P = .02). Patient-specific rod patients had a lower rate of mechanical complications (29% vs 47%, P = .01), a lower rate of rod fracture (3% vs 16%, P = .003), and shorter operative times (215 minutes vs 250 minutes, P = .04). On multivariable analysis, the only 2 independent predictors of reducing the rate of mechanical complications were the utilization of patient-specific rods (odds ratio 0.47, P = .013, CI 0.26-0.85) and lower postoperative thoracic kyphosis (odds ratio 1.02, P = .04, CI 1.00-1.05). CONCLUSION: In a consecutive series of adult thoracolumbar spinal deformity patients, patient-specific rods reduced the odds of mechanical complications by over 50%. Adoption of patient-specific rods should be considered to optimize patient outcomes after adult spinal deformity surgery.
Response Assessment in Pediatric Neuro-Oncology Symposium, Pediatric Society of Neuro-Oncology proceedings: Current status and future initiatives
Abstract Response Assessment in Pediatric Neuro-Oncology (RAPNO) is an international working group dedicated to developing standardized recommendations for response assessment in children with central nervous system tumors enrolled on clinical trials. This standardization facilitates easier and more reliable comparison of responses across trials. Since the first RAPNO publication in 2013, response guidelines have been issued for medulloblastoma (MB), low-grade glioma (LGG), high-grade glioma (HGG), diffuse intrinsic pontine glioma, ependymoma, and craniopharyngioma. Many of these guidelines have already been incorporated into clinical trials to evaluate their effectiveness and accuracy. RAPNO has always been predicated on the idea that these consensus recommendations will require updates over time as they are applied in practice given technological advancements, evolving treatment approaches, and growing knowledge of disease pathology. A half-day RAPNO symposium sponsored by the Society of Neuro-Oncology was held in San Diego, California, United States, on May 15, 2025. This event coincided with the 8th Biennial Pediatric Society of Neuro-Oncology meeting. The main objectives were to identify ongoing RAPNO challenges, discuss potential updates, and plan future initiatives across 3 specific groups: (1) LGG/Craniopharyngioma, (2) HGG/diffuse midline glioma (DMG)/Ependymoma, and (3) MB and other seeding tumors. This manuscript reviews the symposium proceedings and highlights key topics and challenges discussed, as well as outlining future RAPNO initiatives.
Collaborative large language models (LLMs) are all you need for screening in systematic reviews
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.
A phase 1b/2 study of cabozantinib in combination with pembrolizumab in advanced cutaneous melanoma
Abstract Background Pembrolizumab is approved for advanced cutaneous melanoma (aCM). Cabozantinib, an oral multi‐tyrosine kinase inhibitor, has demonstrated antitumor activity as monotherapy or in combination with anti‐PD‐1 therapy in malignancies. The objective of this study was to determine the safety and efficacy of cabozantinib and pembrolizumab for patients with aCM. Methods This phase 1b/2 study enrolled 28 patients with unresectable aCM. In phase 1b, escalating dose levels of cabozantinib (20 mg, 40 mg, and 60 mg) were administered concurrently with pembrolizumab 200 mg intravenously every 3 weeks using a 3 + 3 design to determine recommended phase 2 dose (RP2D). In phase 2, patients received cabozantinib at RP2D in combination with pembrolizumab using a Simon 2 stage design. Primary end point of phase 2 was overall response rate (ORR). Secondary end points included disease control rate, progression‐free survival (PFS), and overall survival (OS). Results Phase 1b enrolled eight patients, cabozantinib 40 mg daily dose level was selected as RP2D. Response rate in stage 1 exceeded predefined criteria for proceeding to expansion phase; however, phase 2 was terminated after 20 patients due to low accrual in the setting of evolving standard of care. Among all treated patients, the most common grade ≥3 toxicities were hypertension ( n = 10, 36%), hypokalemia ( n = 5, 18%), hypophosphatemia ( n = 4, 14%), and ALT elevation ( n = 4, 14%). Among 20 patients treated in phase 2, ORR was 45%, median PFS was 6.6 months, and median OS was 29.5 months. Conclusions Cabozantinib and pembrolizumab combination was explored in treatment‐naive aCM. Toxicity was consistent with known profiles, but discontinuation rates were notable.