This Week in General Medicine — May 21, 2026

Welcome to This Week in General Medicine. This week we're covering 9 notable papers spanning chronic disease management in COPD, diabetes, and cancer; innovations in emergency care for hemorrhage and cardiac arrest; and the growing role of artificial intelligence in both the clinic and the lab. Let's dive in.

We begin this week with several important papers on chronic disease, starting with new data on chronic obstructive pulmonary disease.

Published in *The Lancet*, two parallel trials investigated astegolimab, an antibody targeting the ST2 receptor, for patients with COPD and a history of frequent exacerbations [4]. The rationale is that the IL-33/ST2 pathway is involved in both neutrophilic and eosinophilic inflammation, offering a potential target regardless of baseline eosinophil counts. The results, however, were mixed. In the phase 2b ALIENTO trial, the every-2-week dosing schedule was associated with a statistically significant 15% reduction in the annual rate of moderate or severe exacerbations compared to placebo. But the every-4-week dose was not. Conversely, in the phase 3 ARNASA trial, it was the every-4-week dose that showed a significant 18% reduction in exacerbations, while the every-2-week dose failed to meet statistical significance. Adverse events were balanced across groups. These inconsistent findings across two large trials suggest a potential role for targeting this pathway, but leave questions about the reliability of the effect and optimal dosing, highlighting the challenges in finding new treatments for this difficult disease.

While the search for new COPD therapies continues, a separate study in *JAMA Internal Medicine* this week quantifies the profound impact of the disease on longevity [8]. Using pooled data from eight large United States cohorts including over 45,000 participants, investigators estimated life expectancy based on GOLD stage. For a 65-year-old, having no COPD meant an average life expectancy of 21.5 years. This dropped progressively with disease severity: to 20 years for GOLD stage 1, 16.4 for stage 2, 13.1 for stage 3, and just 10.7 years for stage 4. The study also calculated years of life lost, or YLL. For patients with GOLD stage 2 to 4 COPD, the years of life lost were similar to or even greater than the YLL associated with hypertension, diabetes, and cigarette smoking. Importantly, these reductions in life expectancy were also seen in patients with COPD who had never smoked. These stark figures provide powerful data for patient counseling and underscore the urgent need for effective interventions.

Moving to another major chronic condition, a large population-based study in *JAMA* explores screening for early-stage Type 1 Diabetes in children [2]. Researchers in Bavaria, Germany, screened over 220,000 children for islet autoantibodies. They found an adjusted population frequency of early-stage, presymptomatic type 1 diabetes of 0.3%. During a median follow-up of nearly 6 years, about 36% of these children with early-stage disease progressed to clinical, or stage 3, diabetes. The most critical finding was that the rate of progression to clinical diabetes was not significantly different between children with a first-degree family history and those without. This suggests that family history is not a reliable way to stratify risk among children who are already autoantibody-positive, and supports the idea that population-wide screening, rather than just screening genetically selected groups, may be necessary to identify children who could benefit from future disease-modifying therapies.

Finally in our chronic disease segment, two papers in *JAMA* highlight advances in both cancer screening and treatment. First, a study validates a new protein-based risk model for lung cancer screening called INTEGRAL-Risk [6]. Current screening eligibility relies on age and smoking history, but many cancers occur in people who don't meet these criteria. This new model combines age, smoking history, and a panel of 13 plasma proteins to estimate risk. In an independent testing set, the model's performance was impressive. At one year, it was significantly better at discriminating who would develop lung cancer compared to the standard questionnaire-based model. When calibrated to the same specificity as current United States Preventive Services Task Force guidelines, the protein model identified 85% of incident lung cancer cases, compared to just 63% captured by the current guidelines. This biomarker-based approach could substantially improve how we select high-risk individuals for lung cancer screening.

From better screening to better treatment, a second paper in *JAMA* reports on a phase 3 randomized trial for a high-risk form of diffuse large B-cell lymphoma [9]. Patients with MYC/BCL2 double-expressor lymphoma, or DEL, have poor outcomes with standard R-CHOP therapy. This trial tested the addition of tucidinostat, an oral histone deacetylase inhibitor, to R-CHOP as a first-line treatment. Among 423 patients in China, the group receiving tucidinostat had a 28% lower risk of an event, which included disease progression, relapse, or death. The two-year event-free survival rate was 60% in the tucidinostat group versus 50% in the placebo group. Complete response rates were also higher. While toxicity was increased, it was deemed manageable. This is the first trial to show a benefit for an epigenetic modulator in DLBCL, offering a new first-line therapeutic approach for this high-risk population.

Our second major theme this week focuses on innovations in acute and emergency medicine, from hemorrhage control to cardiac arrest. We start with a major pragmatic trial from *The New England Journal of Medicine* that challenges a growing practice in prehospital care [5].

The Study

The TOWAR trial was a multicenter, cluster-randomized trial that assigned air medical bases to use either up to two units of type O whole blood or standard blood components—that is, plasma and red cells—for prehospital transfusion in trauma patients with hemorrhagic shock.

Results

The primary outcome was 30-day mortality. The study found no significant difference between the groups. Mortality was 25.9% in the whole-blood group compared to 20.5% in the component group. The adjusted odds ratio was 1.24, but the 95% confidence interval ranged from 0.87 to 1.76, crossing 1.0 and indicating the result was not statistically significant. There were also no substantial differences in adverse events. The conclusion is clear: in this large trial, prehospital resuscitation with whole blood did not result in lower 30-day mortality than resuscitation with blood components.

While that trial questions a current strategy, a preclinical study in *Science Translational Medicine* introduces a potential future one [1]. Researchers developed a topical hemostatic agent made from what they call reprogrammed apoptotic platelets. By inducing apoptosis in isolated platelets, they created a derivative enriched in surface phosphatidylserine, which is highly procoagulant. These reprogrammed platelets also increased production of prostaglandin E2, further boosting clot formation. In mouse and pig models of liver injury and gastric ulcer bleeding, this agent demonstrated superior hemostatic efficacy compared to clinical thrombin and other commercial materials, even in the presence of antiplatelet therapy. This represents a novel, topical approach with potential for managing complex bleeding, particularly in endoscopic settings.

Rounding out our emergency medicine section, two papers explore the role of artificial intelligence, one at the point of care and the other in the research lab. In *JAMA Internal Medicine*, investigators evaluated the potential for AI to assist bystanders in delivering cardiopulmonary resuscitation, or CPR [7]. While 911 dispatchers provide telecommunicator-CPR, there can be delays and variability. This study first tested several widely available AI models on their ability to give guideline-concordant CPR instructions in simulated scenarios; performance was good but imperfect. The researchers then developed a purpose-built AI agent called ChatCPR. In the same simulated scenarios, ChatCPR achieved 100% adherence to both minimally viable and maximally effective CPR criteria. When tested on recordings of real 911 calls where CPR was indicated, the AI agent's performance represented an absolute improvement of 36 percentage points over human dispatchers for delivering maximally effective, nuanced instructions. These findings suggest AI holds real promise as a scalable public health tool to improve bystander response in out-of-hospital cardiac arrest.

And from the front lines to the laboratory, a paper in *Nature* showcases how AI can accelerate the very process of scientific discovery [3]. Researchers introduced 'Co-Scientist,' a multi-agent AI system built on Gemini, designed for structured scientific thinking. The system generates, critiques, and refines novel research hypotheses. To validate its utility, the authors focused on biomedical applications. Notably, Co-Scientist helped identify new drug repurposing candidates and synergistic combination therapies for acute myeloid leukemia. These AI-generated hypotheses were then successfully validated through in vitro experiments. This work demonstrates the potential for AI not just to analyze data, but to act as a creative partner in generating new knowledge and accelerating the path to discovery.

If you only have time for one paper this week, make it the randomized trial in *JAMA* on adding tucidinostat to R-CHOP for double-expressor DLBCL [9]. It offers a new, effective, first-line treatment for a high-risk population with previously poor outcomes, marking the first successful trial of an epigenetic modulator in this disease.

Here are the key takeaways from this week in General Medicine.

First: For patients with high-risk, newly diagnosed double-expressor DLBCL, adding the oral HDAC inhibitor tucidinostat to first-line R-CHOP significantly improves event-free survival and should be considered a new standard of care [9].

Second: In patients with traumatic hemorrhage, a large pragmatic trial found that prehospital transfusion with whole blood was not superior to blood components for reducing 30-day mortality [5].

Third: COPD significantly shortens life expectancy, with years of life lost from moderate-to-severe disease exceeding that from diabetes or smoking. This underscores the disease's burden, even as a new biologic, astegolimab, showed mixed results in reducing exacerbations [8, 4].

Fourth: Biomarker-based risk models are on the horizon for improving cancer screening. A 13-protein panel outperformed current questionnaire-based criteria in identifying individuals who will develop lung cancer within one year, potentially refining selection for screening [6].

And finally: Population-based screening for early-stage Type 1 Diabetes can identify children at risk of progression, and this risk is independent of family history, suggesting screening could be considered for the general pediatric population [2].

That's your roundup for This Week in General Medicine. The full transcript and references are available on the episode page in your AudioScholar library. This is an AI-curated summary — for clinical decisions, always consult primary sources and current guidelines. See you next week.