Medical News
Research of the Year 2025
Jennifer Abbasi; Samantha Anderer; Rita Rubin, et al
JAMA Published Online: December 12, 2025
doi: 10.1001/jama.2025.20503
For this inaugural Research of the Year roundup, JAMA Medical News asked the journal’s top editors to nominate their favorite studies published in JAMA over the past year—the articles that they thought were the most impactful, the most newsworthy, and the most novel—and why they chose what they did. From these nominations, Editor in Chief Kirsten Bibbins-Domingo, PhD, MD, MAS, and Executive Editor Gregory Curfman, MD, handpicked the final selections.
The list spans 9 studies and clinical trials first published online between October 2024 through September of this year across a range of health care topics, from heart disease to dementia to artificial intelligence (AI).
“They speak to really hot areas in research right now and to clinical conditions that are of great importance to patients, to clinicians, and to the public health community,” Bibbins-Domingo said of the collection of articles. “They also speak to the range of methods that investigators are using to answer important questions.”
We present JAMA’s Research of the Year.
GLP-1 Drugs Cut Heart Failure Deaths
Heart failure with preserved ejection fraction (HFpEF) has become the most common type of heart failure, driven in part by the global rise in obesity and type 2 diabetes. Although it’s associated with high mortality, “few therapies have been identified that prevent hospitalization or death” from HFpEF, according to Deputy Editor Mary McDermott, MD.
Now, an investigation exploring the role of glucagon-like peptide-1 (GLP-1) receptor agonists in HFpEF risk reduction offers hope and a new clinical path forward.
In their analysis of several cohort studies using US health care claims data, researchers found that patients with obesity-related HFpEF and type 2 diabetes who had initiated the GLP-1 drugs semaglutide or tirzepatide had a more than 40% lower risk of hospitalization for heart failure or all-cause mortality than those who used sitagliptin, a glucose-lowering drug that served as a placebo proxy.
Although previous studies have favored tirzepatide in terms of weight reduction, a head-to-head comparison of the 2 GLP-1 medications in this study found the dual glucose-dependent insulinotropic polypeptide/GLP-1 receptor agonist offered no meaningful benefit over semaglutide in minimizing heart failure risks.
McDermott said that GLP-1 drugs are “dramatically influencing medical practice and public health” beyond weight loss.
The observational study “shows that the medications are also associated with reduced rates of hospitalization and death,” she said. “Results from this study will have wide-ranging implications for the millions of patients with HFpEF.”
With more than 58 000 participants, the study’s design allowed for a much larger and more diverse group of patients than a clinical trial might allow, and it exemplified how real-world evidence can meaningfully complement findings from randomized trials. In this case, it also reinforced the role of GLP-1–based therapies “beyond previously established benefits in the evolving management of cardiometabolic” diseases, the authors noted.
A Bonus Benefit of the Shingles Vaccine
The notion that shingles vaccination could reduce the risk of dementia got a shot in the arm in April with the publication of an observational study involving more than 100 000 people in Australia.
By the beginning of December, the article had received more than 70 000 views, and its findings had been reported by approximately 100 news outlets. The study “is timely and topical,” especially in light of growing skepticism about vaccines, McDermott noted.
Observational trials of vaccines are tricky because of healthy vaccinee bias, which refers to the fact that people who opt to receive vaccinations tend to be healthier than those who don’t. If individuals who get a shingles vaccine are found to have a lower dementia risk, it’s difficult to determine outside of a clinical trial whether the vaccine itself or other habits linked to dementia protection, such as limiting alcohol or maintaining a healthful weight, deserve the credit.
The researchers behind this study took advantage of natural randomization. Australia’s National Immunisation Programme began offering people aged 70 to 79 years a free live-attenuated herpes zoster vaccine (Zostavax) against shingles on November 1, 2016. People born on or after November 2, 1936, were eligible, whereas those born before November 2, 1936, were not.
Instead of comparing individuals who opted to get vaccinated with those who didn’t, the researchers based their comparison on birthdays, specifically whether that date made them eligible for a shingles vaccine. Except for a large difference in their probability of receiving a shingles vaccine, the health behaviors and other characteristics of those born shortly before the eligibility cutoff date weren’t expected to differ from those born shortly afterward.
In the approximately 7 years after the shingles vaccine became available, the probability of a new dementia diagnosis in people who were eligible for it was nearly 2 percentage points lower than of those who were ineligible, and no protective association was found for other conditions. Along with similar findings in other studies, the results suggest that shingles vaccination “is a low-cost, high-reward intervention to reduce the burden of dementia,” the authors concluded.
Deputy Editor Preeti Malani, MD, said the study “raises all kinds of questions about the role of viral infection”—and that its findings are “one more reason to get vaccinated.”
Studies Evaluating Health Care AI Tools Fall Short
Studies of large language model (LLM) applications in medicine have exploded in recent years, but are they asking the right questions about AI tools? This systematic review found that evaluations of LLMs in health care settings are “fragmented and insufficient,” with only 5% using real patient data.
The researchers looked at more than 500 studies published between 2022 and early 2024 to summarize the characteristics of LLM evaluations. They found that the testing approaches may not focus on the most useful health care applications.
The majority of the studies centered around assessing medical knowledge, like answering licensing examination questions and making clinical diagnoses. There was much less focus on testing the performance of LLM applications on administrative tasks—such as writing prescriptions and referrals, assigning billing codes, and taking clinical notes—that require manual labor and contribute to physician burnout.
Most LLM evaluations focused primarily on accuracy, but other key features including deployment considerations and fairness, bias, and toxicity were measured much less frequently. And instead of real patient care data, most studies relied on a mix of medical examination questions, expert-generated questions, and clinician-designed vignettes.
Deputy Editor Linda Brubaker, MD, called the analysis a “very timely assessment of LLM evaluation in clinical workflows” and said that future research should “prioritize real patient care data in evaluations to ensure alignment with real-world conditions.”
Based on the findings, the authors concluded that there’s a need for consensus-based methods to evaluate health care LLMs. They highlighted 6 major shortcomings of the existing studies and included recommendations to address them in the future, including using real patient data, quantifying biases, and prioritizing high-value administrative tasks. Moving forward, the guidance could help shape the critical work of testing LLMs in health care settings.
Newborn Genome Sequencing Gets Closer to the Clinic
About 1 in every 300 infants born in the US is diagnosed with a health condition first detected by a standard newborn screening. Although approximately 60 conditions have been recommended for this screening, genome sequencing can identify hundreds of disorders, particularly genetic conditions lacking the blood biomarkers required by standard screening methods.
Early results from a prospective study demonstrated that a targeted genome sequencing analysis that looks for variants in a predefined set of genes is not only feasible but can improve health outcomes by enabling the early detection of a wider range of treatable health conditions and disorders than standard screenings.
The Genomic Uniform-screening Against Rare Disease in All Newborns study, or GUARDIAN, was one of the first large-scale observational investigations to use DNA analysis as a method, and interim findings from the initial 4000 patients showed that it “can identify newborns with conditions that may otherwise be clinically undetected until symptom onset,” Brubaker said.
The study included a racially and ethnically diverse population of infants at 6 New York City hospitals who received expanded screening with genomic sequencing between September 2022 and July 2023. There was high acceptance for screening: 72% of those who were approached consented to participate in the study and 91% of enrollees agreed to an additional optional screening for neurodevelopmental disorders.
Nearly 4% received positive screening results, and of the 120 newborns with true positive findings, 92% had a confirmed diagnosis for a condition not included in a standard newborn screening. Such treatable conditions only picked up by genome sequencing included long QT syndrome, Wilson disease, and severe combined immunodeficiency disorder, which, in one case, resulted in a successful bone marrow transplant.
This approach to screening, Brubaker noted, could help children get earlier access to life-saving treatments. The study’s authors acknowledged that subsequent research is needed to assess if their results apply to other populations. Still, given declining costs of DNA sequencing and growing acceptance by parents, Brubaker believes that newborn genome sequencing is “nearing potential clinical utility.”
The ongoing GUARDIAN study aims to enroll 100 000 infants in the next few years.
A Lifestyle Intervention for People With Dementia Risk
Finland had FINGER, and now the US has POINTER—both clinical trials assessing the benefits of lifestyle interventions such as exercise and social engagement on the trajectory of cognition in older adults at risk of dementia.
The US POINTER trial enrolled about 2100 participants at 5 clinical sites and randomly assigned them to 1 of 2 groups. People in both groups were encouraged to increase their physical and cognitive activity, social engagement, and cardiovascular health monitoring and to eat a more healthful diet.
But 1 group was pretty much left on its own to make those changes, receiving only 6 facilitated peer team meetings over 2 years as well as publicly available educational materials. The other group’s intervention was more structured and intense—with 38 team meetings—and participants were held to a higher level of accountability.
Both groups’ cognitive function improved over the 2-year study, but the structured intervention group’s enhancement was significantly greater than that of the self-guided intervention group. Participants who started the trial with lower cognition benefitted more from the intervention, but the cognitive benefits were the same for carriers and noncarriers of APOE ε4, a major genetic risk factor for Alzheimer disease.
The trial wasn’t big enough to assess whether participants eventually developed cognitive impairment or dementia, and the authors concluded that further research is needed to understand the clinical significance of their findings. They also noted that the absence of a no-intervention control group limited their ability to rule out alternative explanations for participants’ cognitive improvement, such as practice effects from repetitive cognitive testing and the general benefits of trial participation.
Still, “this trial is important because it confirmed the benefits of a structured multidomain lifestyle intervention in a US population of older adults at risk of cognitive decline or dementia, including those from racial and ethnic groups often underrepresented in dementia clinical trials,” said Deputy Editor Christopher Muth, MD.
Pinpointing a Blood Transfusion Threshold in Brain Injuries
Patients with acute brain injury often have anemia, which is associated with increased morbidity and mortality. Although red blood cell transfusions may alleviate this issue, they have also been associated with complications such as secondary infection and lung injury. Previous studies on blood transfusions in the absence of life-threatening bleeding have shown mixed results, with no clear optimal hemoglobin threshold for transfusion in patients with acute brain injury.
Findings from the Transfusion Strategies in Acute Brain Injured Patients (TRAIN) randomized clinical trial support a more liberal red blood cell transfusion threshold to improve neurological outcomes.
“Prior to TRAIN, transfusion thresholds in neurocritical care were based on limited or extrapolated data, and the optimal approach for brain-injured patients was controversial due to conflicting results from smaller studies and meta-analyses,” said Senior Editor Philip Greenland, MD.
The multicenter, phase 3 trial included about 800 patients with acute brain injury and a hemoglobin level lower than 9 g/dL within the first 10 days after injury. The researchers assigned patients to either a liberal transfusion strategy—triggered by hemoglobin lower than 9 g/dL—or a restrictive strategy, where patients only received transfusion if hemoglobin dropped to lower than 7 g/dL. After 6 months, 63% of people in the liberal treatment group had unfavorable neurological outcomes, compared with 73% in the restrictive group. About 14% of patients in the restrictive group had at least 1 cerebral ischemic event vs 9% in the liberal transfusion group.
The results suggest there’s a lower likelihood of poor long-term outcomes for people with acute brain injury when transfusion is initiated at 9 g/dL hemoglobin vs 7 g/dL. Considering the study’s findings and its high number of citations, Greenland predicts that TRAIN may “actually change practice.”
Sparing Patients With Low-Risk Breast Cancer From Surgery
Ductal carcinoma in situ (DCIS) is known to be a major contributor to breast cancer overdiagnosis—it may never cause harm if left untreated. Despite evidence that not all DCIS progresses to invasive cancer, treatment typically involves surgery, which may be combined with adjuvant radiation or endocrine therapy.
Considering the adverse effects of unnecessary cancer treatment, researchers behind the COMET trial conducted a noninferiority study comparing active monitoring with guideline-concordant care. The trial included nearly 1000 women aged 40 years or older with newly diagnosed hormone receptor (HR)–positive, ERBB2-receptor negative, low-risk DCIS. Women who were randomly assigned to standard medical care received the usual treatment, including breast-conserving surgery and mastectomy, whereas those in the active monitoring group regularly underwent physical examination and imaging, with surgical intervention conducted as needed.
After 2 years, the cumulative rate of ipsilateral invasive cancer was 4.2% in the active monitoring group, compared with 5.9% in the guideline-concordant care group, achieving noninferiority. The characteristics of the invasive tumors didn’t significantly differ between the groups.
COMET is the first randomized clinical trial to report outcomes comparing guideline-concordant care with active monitoring for low-risk DCIS management, with results indicating that monitoring isn’t inferior to usual practice. About 10% of patients with standard medical care underwent mastectomy vs less than 2% in the active monitoring group.
“Over the years, practitioners and patients have had concerns that DCIS is overtreated, resulting in unnecessary morbidity and cost,” said Deputy Editor Mary L. Disis, MD, who is also editor of JAMA Oncology. The new data “suggest that active monitoring could be an option for some women with DCIS to avoid more extensive surgical and medical treatments.”
A Better Hepatitis B Vaccine for People With HIV
People with immunocompromising conditions such as HIV may not mount an adequate immune response to the conventional vaccine for hepatitis B virus (HBV), which remains a global health problem and a leading cause of liver-related death. But in a randomized clinical trial, researchers found that replacing the traditional aluminum hydroxide adjuvant with a cytosine phosphoguanine adjuvant provided superior seroprotection in people with HIV who hadn’t responded to the vaccine initially.
The BEe-HIVe trial included nearly 600 adults with HIV across 41 sites in 10 countries who were nonresponsive to prior hepatitis B vaccination and were receiving antiretroviral therapy. Participants were randomly assigned to receive either 2 or 3 doses of the HBV vaccine with a cytosine phosphoguanine adjuvant (HepB-CpG vaccine) or 3 doses of the conventional HBV vaccine with an aluminum hydroxide adjuvant (HepB-alum).
The researchers found that 93% of participants receiving 2 doses of HepB-CpG and 99% of those receiving 3 doses generated protective antibody levels against the virus. Both of these regimens were superior to 3 doses of HepB-alum, which provided seroprotection to only 81% of people. The newer vaccine also elicited a more rapid response than the conventional version.
Although the phase 3 trial focused on people living with HIV, according to Malani it “provides evidence to try this approach in other patients who do not respond to a series of HepB vaccine.” If successful, this could extend protection to additional immunocompromised groups, as well as to older persons, current smokers, and people with diabetes or obesity, among others who may not develop an adequate response to conventional vaccine.
New Hope for Treatment-Resistant Hypertension
For the millions of people around the world whose blood pressure remains dangerously elevated despite the use of multiple medications, the results of a phase 3 randomized clinical trial could offer a new treatment option. Lorundrostat, a first-in-class aldosterone synthase inhibitor, was shown to significantly reduce blood pressure when added to existing medication regimens.
The international study enrolled more than 1000 adults with uncontrolled hypertension, many of whom were treatment resistant, who were already taking between 2 and 5 prescription antihypertensive drugs. Participants received either lorundrostat—at 50 mg daily, with some escalating to 100 mg based on their response—or a placebo for 12 weeks.
The findings of the Launch-HTN trial were promising: at 6 weeks, lorundrostat reduced systolic blood pressure by a clinically meaningful 16.9 mm Hg compared with 7.9 mm Hg in the placebo group.
The drug was also found to be well tolerated with a favorable safety profile. Although hyponatremia, hyperkalemia, and reduced kidney function occurred more frequently with lorundrostat than the placebo, discontinuation rates due to these adverse events were less than 1%.
For Curfman, what makes this study particularly important is its focus on this subset of patients with hypertension.
“This agent opens a new approach to the treatment of uncontrolled hypertension, which may affect up to 40% of patients,” he said.
It may also serve an unmet need for those with treatment-resistant hypertension who have exhausted conventional treatment options.
Up until now, these patients have had limited options despite facing heightened cardiovascular risks, including myocardial infarction, stroke, or chronic kidney disease.
Lorundrostat targets excess aldosterone production, a root cause of hypertension. Unlike existing aldosterone blockers that obstruct the hormone’s receptor, lorundrostat inhibits the enzyme that produces aldosterone itself, offering a novel mechanism of action.
