Citations

Abstract Large Language Models (LLMs) demonstrate strong capabilities in healthcare applications. However, recruiting human evaluators to assess usability and effectiveness becomes challenging in the large-scale LLM-generated outcomes. Existing metrics are often outdated for evaluating LLM-based healthcare technologies. This paper proposes a Research through Evaluation (RtE) approach that can refine metrics and improve assessments of LLM-generated outputs. We deploy RtE in a retrospective comparative study of prostate cancer patient inquiries, comparing responses from human clinical care teams with those generated by a GPT-4-based in-basket bot. Through three rounds of co-evaluations with clinical professionals and an LLM-as-graders study with four LLMs, the RtE method shows that starting with small, iterative evaluations helps human graders adjust metrics to suit their daily practices better. Additionally, in the LLM-as-graders study, LLMs exhibit similar behavior, preferring the in-basket bot's responses over the clinical care team. Our methodology and findings demonstrate that the RtE method can enhance the transparency and robustness of evaluation metrics for LLM-generated content, with the potential to generalize to evaluations in other clinical practices.

Abstract Background Sleep disturbances are common in Alzheimer's disease and have been reported in its preclinical stage. A detailed characterization of sleep disruption across the Alzheimer’s disease continuum is required to clarify its role as a potential therapeutic target and to evaluate the utility of sleep-related biomarkers. Methods Sleep architecture was characterized through detailed spectral and temporal analysis of the slow-wave activity using polysomnography across the Alzheimer’s disease continuum . Participants included: cognitively unimpaired individuals without abnormal Alzheimer’s disease biomarkers (n = 30), cognitively unimpaired individuals with abnormal Alzheimer’s disease biomarkers (n = 45), and individuals with mild cognitive impairment or mild dementia due to Alzheimer’s disease (n = 41). Associations between slow-wave features and longitudinal cognitive performance were tested in cognitively unimpaired participants. Sex differences were also explored. Results A stepwise increased wake after sleep onset and latency to rapid eye movement sleep was observed, alongside reduced spectral power of slow-wave activity and a progressive flattening of slow-waves across the Alzheimer’s disease continuum in both sexes. In cognitively unimpaired individuals, higher slow-wave slope and greater peak-to-peak amplitude were associated with better cognitive trajectories over time. Conclusions Sleep macro- and microarchitecture are altered early and progressively deteriorate across the Alzheimer’s disease continuum . Slow-wave features relate to cognitive trajectories in preclinical Alzheimer’s disease, with exploratory analyses suggesting stronger associations in men. These findings support slow-wave morphology as a candidate non-invasive marker for future sleep-based intervention studies.

Abstract In 2024, the National Academies of Sciences, Engineering, and Medicine (NASEM) published a report entitled Advancing Research on Chronic Conditions in Women. A major purpose of this report was to summarize the current state of research into a series of 21 chronic conditions that affect women differently or that are more common in women. The American Society for Bone and Mineral Research (ASBMR) established a working group to evaluate the NASEM report’s implications for women’s musculoskeletal (MSK) health. This perspective summarizes key research gaps identified in the NASEM report, including the lack of sex-specific data, underrepresentation of women in clinical studies, and minimal integration of MSK health with other chronic conditions. The NASEM report underscores the need for more comprehensive study designs that consider hormonal, genetic, and social determinants of health across the lifespan. In addition to osteoporosis, in this perspective, the ASBMR working group highlights emerging but poorly understood evidence of detrimental effects on musculoskeletal biology in non-MSK, chronic diseases that were identified in the NASEM report to disproportionately affect women. To advance the field, we advocate for better diagnostics, increased investment, and stronger representation of women in research studies. Addressing these gaps is critical for improving prevention, care, and outcomes for chronic conditions that uniquely or disproportionately affect women.

Abstract Drug-resistant epilepsy affects tens of millions of people worldwide and is associated with considerable morbidity and mortality. Thalamic deep brain stimulation and cortical responsive neurostimulation are proven treatments for focal epilepsy. Both have been used to target a range of thalamic nuclei; yet, the roles of these thalamic nuclei in focal seizure generation remain incompletely understood. Thirteen patients with drug-resistant focal epilepsy undergoing intracranial EEG were consented to undergo investigation of thalamocortical networks. Sampled regions included cortical, mesial temporal, and thalamic brain regions. Visual and spectral analyses were performed to identify seizure onset patterns and correlate thalamic and cortical seizure activity. Thalamic ictal discharges were observed in 89% of seizures. Of these, 56% demonstrated synchronous thalamocortical activity with distinct patterns. These onset patterns included hypersynchronous spiking, low-voltage fast activity, ictal baseline shifts, and broadband suppression. Multiple thalamic nuclei were involved in ictal organization and propagation, with the specific nuclei depending on the cortical seizure network. The thalamus plays a crucial role in focal onset seizure generation and propagation, with distinct seizure onset patterns and nuclei involved. These findings support exploring a broader range of thalamic nuclei in epilepsy neurostimulation and have implications for seizure detection settings in intracranial sensing devices.

Background: Thoracic aortic aneurysm (TAA) remains a high-risk vascular condition despite major advances in imaging surveillance, operative repair, and endovascular therapy. Medical management still relies largely on blood pressure control and global cardiovascular risk reduction. Renin–angiotensin–aldosterone system (RAAS) inhibitors are frequently used in TAA, but contemporary data evaluating survival and cardiovascular outcomes in broad TAA populations are limited. Glucagon-like peptide-1 receptor agonists (GLP-1 RAs) and sodium-glucose cotransporter-2 inhibitors (SGLT2 inhibitors) have established cardiometabolic benefits, yet their role in TAA has not been well defined. Methods: We performed a retrospective multicenter cohort study of adults with imaging-confirmed TAA diagnosed between 1 January 2018 and 1 January 2026 using a Mayo Clinic electronic data platform encompassing more than 15 million patient records. Primary exposures were documented use of RAAS inhibitors, GLP-1 RAs, and SGLT2 inhibitors, evaluated individually and in prespecified combination-therapy analyses. Propensity score matching was used to balance demographics, comorbidities, aortic procedural history, and concomitant cardiovascular medications. Primary outcomes were all-cause mortality and major adverse cardiovascular events (MACE) through 60 months. Results: The study included 162,126 patients with TAA. After matching, RAAS inhibitor use was associated with higher 60-month overall survival (88.3% vs. 85.5%; hazard ratio [HR], 0.79; 95% CI, 0.76–0.83; p < 0.001) and MACE-free survival (86.1% vs. 84.2%; HR, 0.87; 95% CI, 0.83–0.91; p < 0.001). GLP-1 RA therapy was associated with higher overall survival (97.5% vs. 92.5%; HR, 0.32; 95% CI, 0.27–0.38; p < 0.001) and MACE-free survival (93.2% vs. 89.3%; HR, 0.62; 95% CI, 0.56–0.70; p < 0.001). SGLT2 inhibitor therapy was similarly associated with higher overall survival (89.8% vs. 81.5%; HR, 0.51; 95% CI, 0.47–0.54; p < 0.001) and MACE-free survival (86.3% vs. 79.1%; HR, 0.62; 95% CI, 0.58–0.66; p < 0.001). Combination therapy with RAAS inhibitors plus either GLP-1 RAs or SGLT2 inhibitors was associated with incremental improvements in overall survival and MACE-free survival compared with GLP-1 RA or SGLT2 inhibitor monotherapy. Conclusions: In this large propensity-matched TAA cohort, RAAS inhibitors, GLP-1 RAs, and SGLT2 inhibitors were each associated with improved survival and fewer major cardiovascular events, with additional benefit observed for RAAS-based combination therapy. These findings support further prospective investigation of integrated cardiometabolic and vascular-targeted therapy in TAA, while underscoring that observational associations should not be interpreted as proof of aneurysm-specific disease modification.

Abstract Surgery continues to be a mainstay of treatment for patients with symptomatic fibroids. Hysterectomy is an option for those desiring definitive treatment and myomectomy is preferred for those that have not yet completed childbearing or wish to retain their uterus. While open laparotomy remains a viable approach to both hysterectomy and myomectomy, minimally invasive techniques for the surgical management of symptomatic fibroids are preferred options for many due to improved patient outcomes. These include hysteroscopic myomectomy, laparoscopic and robotically assisted myomectomy, and vaginal, laparoscopic, and robotically assisted hysterectomy. Patients have a faster recovery and less postoperative pain compared to a laparotomic approach. More recently, uterine‐sparing outpatient modalities such as ultrasound‐guided radiofrequency ablation and high‐intensity focused ultrasound have emerged as promising alternatives for selected patients. Benefits include clinically significant reduction in fibroid size and menstrual bleeding with small or no abdominal incisions and an expedited convalescence. This article reviews preoperative evaluation and imaging for fibroids, as well as the techniques, approach, indications, patient selection, procedural management, and surgical and fertility outcomes for each surgical and procedural management of uterine fibroids.

Artificial intelligence (AI) is now applied across oncology, including imaging, pathology, therapy selection, toxicity assessment, and survival modeling, and large language models (LLMs) have been adopted particularly quickly. Reported physician use of AI roughly doubled between 2023 and 2026. The frameworks used to evaluate these tools have not advanced at the same rate. Many are available, but each was designed for a single stage of the AI lifecycle, and none directly answers the question most relevant to a treating clinician: whether a given tool is appropriate for the individual patient under care. We reviewed the frameworks in current use, including TRIPOD+AI, PROBAST+AI, CLAIM, SPIRIT-AI, CONSORT-AI, DECIDE-AI, MINIMAR, and CREMLS, together with the oncology-specific ESMO EBAI and the newer LLM-specific guidance (TRIPOD-LLM, MI-CLEAR-LLM, CHART, and ESMO ELCAP). Each addresses part of the evaluation problem, but none is sufficient on its own for oncology, where the standard of care changes rapidly, assays drift, biomarker-defined subgroups are small, and a model validated in one period may perform poorly in the next. LLMs introduce additional challenges, including sensitivity to prompting, undisclosed vendor updates, behavioral drift, and hallucination rates approaching 50% on clinical quality-assurance tasks. The predictive-model frameworks were not designed to capture these failure modes. The central argument of this review is that the frameworks the field requires already exist; what is missing is their mandatory adoption. We propose that journals and regulators move from recommending these frameworks to requiring their use, and we outline a cancer-aware evaluation pathway together with the specific responsibilities of authors, reviewers, journals, and regulators in implementing it.