Citations

Abstract Enfortumab vedotin (EV), pembrolizumab, and EV plus pembrolizumab (EV/P) have rapidly reshaped the treatment landscape of advanced urothelial carcinoma, but toxicity increasingly determines treatment delivery, treatment continuity, and real-world benefit. As EV/P moves into broader clinical use, clinicians must distinguish adverse events related to antibody-drug conjugate exposure from immune-mediated toxicities and recognize syndromes in which both mechanisms may contribute. In this practical, mechanism-informed review, we synthesize pivotal trials, mature follow-up reports, and representative real-world cohorts describing the safety of EV monotherapy, pembrolizumab monotherapy, and EV/P in urothelial carcinoma. Because adverse-event reporting and ascertainment are heterogeneous across studies, including differences in preferred terms, composite categories, grading frameworks, and treatment-related versus all-cause reporting, we emphasize clinically interpretable patterns rather than pooled estimates. EV monotherapy is characterized by cutaneous toxicity, peripheral neuropathy, hyperglycemia, dysgeusia, and ocular events, including early epithelial/metabolic events and cumulative neuropathy. Pembrolizumab monotherapy has a lower overall high-grade toxicity burden but is defined by immune-mediated endocrinopathies, pneumonitis, hepatitis, colitis, and nephritis. Combination therapy produces a predictable overlap of these toxicity domains, creating diagnostic and management challenges for syndromes such as skin toxicity, diarrhea, fatigue, and pulmonary symptoms. We provide comparative visual toxicity signatures, regimen-specific summary tables, temporal toxicity mapping, and a pragmatic attribution framework that classifies events as EV-predominant, immune-predominant, overlap/indeterminate, or high-risk requiring initial interruption of both agents. This review aims to support bedside recognition, provisional attribution, early management, and safer implementation of EV/P in contemporary urothelial carcinoma practice.

Abstract Purpose Large language models (LLMs) have proven performance for certain diagnostic tasks; however, limited studies have evaluated their consistency in recommending appropriate medication regimens for a given diagnosis. Medication management is a complex task that requires synthesis of drug formulation and complete order instructions for safe use. Here, the performance of GPT-4o, an LLM available with OpenAI’s ChatGPT, was tested on 3 medication management tasks. Methods GPT-4o performance was tested on 3 medication tasks: identifying available formulations for a given generic drug name, identifying drug-drug interactions (DDIs) for a given medication regimen, and preparing a medication order for a given generic drug name. For each experiment, the model’s raw text response was captured exactly as returned and evaluated using clinician evaluation in addition to standard LLM metrics, including Term Frequency-Inverse Document Frequency (TF-IDF) vectors, normalized Levenshtein similarity, and Recall-Oriented Understudy for Gisting Evaluation (ROUGE-1/ROUGE-L) F1 score between each response and its reference string. Results For the first task of drug-formulation matching, GPT-4o had 49% accuracy for generic medications being matched to all available formulations, with an average of 1.23 omissions per medication and 1.14 hallucinations per medication. For the second task of drug-drug interaction identification, the accuracy was 54.7% for identifying the DDI pair. For the third task, GPT-4o generated order sentences containing no medication or abbreviation errors in 65.8% of the cases. Conclusion Model performance for basic medication tasks was consistently poor. This evaluation highlights the need for domain-specific training through clinician-annotated datasets and a comprehensive evaluation framework for benchmarking performance.

Abstract Background At hospice enrollment, patients are asked preferences regarding antibiotic use. Limited evidence on how these medications impact survival makes it difficult for clinicians to offer clear guidance. To support decision-making and antimicrobial stewardship, we assessed the association between antibiotic use and survival in outpatient hospice patients. Methods The association between antibiotic use and survival was evaluated using a time-dependent Cox proportional hazards model, adjusted for age, sex, Charlson Comorbidity Index, and hospice-qualifying conditions (HQCs). The cohort included patients enrolled in a large health system’s outpatient hospice program between June 20, 2017, and January 1, 2023. Results Among 6169 patients with an HQC, 1233 received at least one antimicrobial prescription during the study period. The median time from hospice enrollment to first prescription was 16 days (IQR 3-83). Antimicrobial prescribing patterns varied by HQC category. In time-dependent Cox models, antibiotic use was associated with increased hazard of death (aHR 1.47; 95% confidence interval [CI], 1.37–1.57).Stratified analyses showed the strongest association with increased mortality in patients with cancer HQCs (aHR 1.67; 95% CI, 1.50–1.87) and a survival benefit in those with infectious HQCs (aHR 0.60; 95% CI, 0.40–0.89). Discussion Physicians, patients, and families often overestimate the benefits of antibiotics near the end of life. The observed increase in mortality shortly after antibiotic initiation suggests that symptoms prompting antibiotic use may reflect the dying process rather than treatable infection. Knowledge of the limited role of antibiotics in prolonging life in hospice settings may improve decision-making and enhance the quality of end-of-life care.

Abstract To characterize invasive neuromodulation for insular drug‐resistant epilepsy, we retrospectively evaluated 22 patients treated with responsive neurostimulation (RNS; n = 14) or chronic subthreshold stimulation (CSS; n = 8). Stereoelectroencephalography was performed in 20 cases (91%). Median baseline seizure frequency was 10.5/month. Forty‐one percent had prior epilepsy surgery. At year 3, median seizure reduction (MSR) was 83% ( p = .01, Wilcoxon signed‐rank test), with a responder rate (RR) of 77%; MSR was 92% for CSS and 75% for RNS. There were no significant differences in MSR ( p ≥ .4) or RR ( p > .9) when comparing RNS versus CSS. Three patients achieved seizure freedom by year 3. Patients undergoing combined insular–thalamic stimulation ( n = 5) demonstrated an MSR of 84%, 86%, and 88% at 1, 2, and 3 years; seizure reduction in thalamocortical versus cortical‐only stimulation was not different ( p ≥ .1). Parasagittal lead trajectories achieved closer insular proximity than transopercular trajectories ( p < .0001, Mann–Whitney U ‐test). Median charge density at last follow‐up was 1.3 μC/cm 2 . There were no acute surgical complications and all stimulation‐related side effects resolved with programming adjustments. This study expands the limited literature on insular neuromodulation by demonstrating effectiveness of RNS and CSS, suggesting potential benefit from combined thalamocortical targeting, and defining parasagittal trajectories as a surgical approach for optimal lead proximity.