Type 2 diabetes increases the risk of chronic kidney disease, making prevention of kidney failure a key part of treatment. Newer diabetes medications, including GLP-1s and SGLT2 inhibitors, can reduce the risk of kidney disease, but previous studies have included large proportions of patients who already had signs of kidney damage, making it unclear whether the findings applied to patients without kidney damage.
A new study led by Mass General Brigham researchers found that GLP-1 agonists and SGLT2 inhibitors reduced the risk of kidney deterioration in patients with diabetes who had protein in their urine, a sign of kidney damage, before treatment. However, the researchers found little evidence of kidney benefit among patients without protein in their urine. They also showed that sulfonylureas, an older class of diabetes medication, led to faster declines in kidney function in patients with diabetes who didn’t have protein in their urine. Results are published in the BMJ.
“Our study shows that we need to tailor diabetes treatment to the individual patient instead of using a one-size-fits-all approach,” said corresponding author Alexander Turchin, MD, MS, of the Division of Endocrinology in the Mass General Brigham Department of Medicine. “A simple urine test could help doctors assess whether a patient may receive kidney protection from GLP-1 receptor agonists or SGLT2 inhibitors and whether sulfonylureas could pose additional risk.”
Tracking kidney outcomes
The researchers analysed medical record data from 75 455 patients with type 2 diabetes across the U.S., including 13,872 who presented with protein in their urine. They followed the patients for up to five years and compared the risk of developing chronic kidney disease among patients prescribed GLP-1s and SGLT2 inhibitors with those prescribed sulfonylureas or DPP4i. All patients had moderate cardiovascular risk and were receiving metformin as their primary diabetes medication.
Among patients with protein in their urine, GLP-1s and SGLT2 inhibitors were associated with substantially better kidney outcomes than DPP4i. However, GLP-1s and SGLT2 inhibitors did not appear to provide a significant kidney benefit in patients without protein in their urine. Among patients without protein in their urine at baseline, treatment with sulfonylureas was associated with an increased risk of kidney deterioration compared with DPP4i.
Gap for lower-risk patients
Further research is needed to identify therapies to prevent kidney disease in the large population of patients with type 2 diabetes who don’t present with protein in their urine, the researchers say.
“These findings will allow us to individualise medication choices that maximally benefit the patient in front of us rather than for the hypothetical ‘average person,'” said Turchin. “Patients and their doctors should discuss how to balance these different risks and benefits in their individual circumstances when choosing their type 2 diabetes medications.”
New research from the University of St Andrews has found that women are less like than men with the same medical condition to be offered active management such as surgery, a stent or a strong painkiller.
The findings, published in PLOS One, come from a review that sifted 1112 published studies down to those that directly compared the care given to male and female patients. Of the 38 that analysed patient records, 33 reported a significant difference in the treatment men and women received.
Researchers from the University of St Andrews School of Medicine found that almost none of those studies pointed to any guideline recommending different treatment by sex, leaving open whether this reflects sound clinical judgement or unequal care.
Dr Andrew O’Malley, who co-led the study, said: “For clinicians, the findings are a prompt to check whether treatment is being offered on clinical grounds rather than assumption. For example, one study found that when the teams deciding who receives advanced heart failure therapy functioned poorly, women were less likely to be selected. Other work shows doctors more often attribute women’s symptoms to anxiety and make more diagnostic errors with female patients, even when test results are positive.
Dr Miriam Veenhuizen, Honorary Lecturer in the School of Medicine, said: “While the direction of the findings was not a surprise the consistency was. The same pattern appeared in cardiology, surgery, transplant medicine and emergency care, and it survived statistical adjustment in most studies. It’s not entirely clear why this is happening, but it is likely because women were under-represented in clinical trials until recent decades, so many guidelines rest on data from men.”
Dr Veenhuizen added: “What struck us most was the imbalance in attention. Over the same period, 551 studies examined sex inequality affecting doctors and other health professionals. Only 41 examined what happens to patients.”
The researchers now intend to test whether the same patterns appear in the outputs of generative AI systems, which are trained on this literature and on clinical records, and which could entrench these differences at scale if left unchecked.
Many women usually say the same thing during pregnancy: they walk into a room and forget why, misplace their keys or struggle to follow a conversation. This phenomenon, often called ‘pregnancy brain’ or ‘momnesia,’ has long lacked a clear biological explanation.
Now, a new study by researchers at Baylor College of Medicine and collaborating institutions and published in Science Bulletin, identifies a specific brain circuit in an animal model that becomes disrupted under the sustained high oestrogen levels present during pregnancy. The findings offer the first biological explanation of how exposure to high-level circulating oestrogen can temporarily impair memory.
A novel brain circuit links high oestrogen levels with memory problems
“We worked with mouse models designed to mimic the sustained, high blood-oestrogen levels of pregnancy. These models showed that elevated oestrogen caused reversible memory impairment without affecting mood or motivation, suggesting a specific cognitive effect rather than a general change in well-being,” said senior author Dr Zheng Sun, associate professor of medicine – endocrinology, diabetes and metabolism and of molecular and cellular biology at Baylor.
Digging into the underlying biology, the team found that oestrogen receptor alpha, the protein that transmits oestrogen’s signals into cells, is the dominant oestrogen receptor in the brain region called the lateral hypothalamus. Furthermore, this region has abundant GABAergic neurons – brain cells that normally send calming, inhibitory signals to other parts of the brain. Using single-nucleus RNA sequencing, the researchers discovered that high oestrogen levels suppress signaling in these neurons, leading them to fire more frequently. “When we genetically removed oestrogen receptors from these hypothalamic neurons, both oestrogen-induced and pregnancy-induced memory problems in mice were reversed,” said Sun, a member of Baylor’s Dan L Duncan Comprehensive Cancer Center.
The team also found that these overactive hypothalamic neurons project directly into a region of the hippocampus that is a hub for memory formation. Using chemogenetics, a technique that allows researchers to turn specific neurons on or off, the team showed that silencing this hypothalamus-to-hippocampus pathway protected mice from estrogen-induced memory problems, whereas artificially activating the same pathway was sufficient to impair memory on its own, even without elevated estrogen.
Reconciling mixed evidence
Oestrogen’s relationship with memory has puzzled researchers for decades. For instance, hormone replacement therapy after menopause has been linked to cognitive benefits in some studies, while high oestrogen during pregnancy or with oral contraceptive use has been linked to memory complaints in others. The new findings suggest a possible explanation – it may not simply be a matter of ‘more oestrogen is better’ or ‘worse,’ but rather where in the brain that oestrogen acts, and at what levels.
“Low-level, cyclical oestrogen exposure appears to support cognitive function, which is part of why hormone therapy can help postmenopausal women,” said senior author Dr Yanlin He, associate professor at Pennington Biomedical Research Center. “But sustained, high-level oestrogen exposure seems to engage a different pathway altogether, one centred in the hypothalamus rather than the hippocampus itself. That distinction may help reconcile a lot of conflicting data in the field.”
Confirming the link in pregnant women
To determine whether these findings translate to humans, the researchers assessed memory performance in women across different stages of pregnancy. They found task-specific memory impairments that emerged during late pregnancy, and that correlated with circulating oestrogen levels, even after accounting for other factors that might influence cognition. This human data supports the idea that the hormone-driven circuit identified in mice may underlie the memory changes many pregnant women experience.
“Momnesia is real, it has a defined biological basis and is temporary,” said senior author Dr Xianghua Zhuang, professor at the Second Qilu Hospital of Shandong University. “We hope this work helps validate what many women have described anecdotally for years, and gives researchers a concrete target for future study.”
“The memory changes observed in both mice and women were temporary and task-specific, not a sign of broader cognitive decline,” said senior author Dr Xinguo Hou, professor at the Qilu Hospital of Shandong University. “Nonetheless, understanding the underlying circuit could eventually inform how clinicians counsel patients about the cognitive side effects of pregnancy or hormonal contraceptives, and could open avenues for therapies targeting this specific pathway without disrupting oestrogen’s broader, beneficial roles in the body.”
Investigation reignites questions about what steps were taken to safeguard public health from any unintended consequences of the vaccination rollout
Photo by Mufid Majnun on Unsplash
US health officials knowingly relied on a compromised algorithm to detect signals of harm from mRNA covid-19 vaccines and silenced efforts to fix it, finds an investigation published by The BMJ.
The findings are based on government emails released under the Freedom of Information Act and in response to US Senate investigators, as well as exclusive interviews with public health officials and other scientists by investigative journalist David Willman.
Before rolling out covid-19 vaccines in December 2020, the US Centers for Disease Control and Prevention (CDC) assured healthcare professionals and the public that its Vaccine Adverse Event Reporting System (VAERS) could quickly spot any potentially harmful reactions following vaccination.
The CDC planned to use two “data mining” techniques: “proportional reporting ratios” (PRRs) and an “empirical bayesian” method provided by the Food and Drug Administration (FDA), which operated VAERS jointly with the CDC. The two agencies planned to share and discuss results.
But according to a letter by CDC Director Rochelle Walensky, the agency did not perform PRR analyses until 2022, and both CDC and FDA “chose to rely” entirely on FDA’s bayesian method.
The BMJ can reveal that the FDA’s algorithm failed to signal a potential relationship between mRNA covid vaccination and myocarditis (inflammation of the heart muscle), pericarditis (inflammation of the fluid-filled sac around the heart), Bell’s palsy, tinnitus, and other reported disorders owing to a flaw in the detection methodology.
The problem arose because over 90% of initial VAERS reports were for the new mRNA covid vaccines made by Pfizer and Moderna. If both vaccines elevated the risk of an adverse event like myocarditis in roughly equal amounts, however, the “observed” frequency and “expected” frequency would be similar, resulting in no automated alert.
Internal records show that FDA officials were aware of this limitation before and during the pandemic. Government documents also show that CDC officials were informed during rollout of the vaccines.
In early 2021, FDA medical officer Dr Ana Szarfman, working with statistician William DuMouchel, who had developed the bayesian algorithm, warned top officials about the flaw and proposed an updated algorithm that flagged signals. But Szarfman was asked to “cease and desist.” Meanwhile, officials continued to cite the lack of system alerts while reassuring clinicians and the public of the vaccines’ safety.
When the CDC finally ran PRR analyses in 2022, CDC director Walensky said results revealed “no additional unexpected safety signals.” Yet the analyses – examined by The BMJ – show hundreds of adverse events that met the agency’s alert criteria, including myocarditis, pericarditis, Bell’s palsy, and tinnitus, which the FDA’s bayesian method had not triggered.
In October 2023, the FDA’s pharmacovigilance chief acknowledged in an email to colleagues that the agency knew – as the vaccines had rolled out more than two years earlier – of the detection deficiency, but did not respond to requests for comment.
Approached by The BMJ, Szarfman insisted that she had not sought to undermine public support for the covid vaccines. Expressing frustration, she noted, “Very few people understand the statistics. That’s the problem.”
DuMouchel said he could not explain officials’ resistance to switch to the updated method, stating, “I think that they were wrong.” He also regretted that FDA officials rejected Szarfman’s proposed fixes for VAERS, adding, “If they had paid attention to Ana, they would have done better.”
Eminent cardiologist involved in treatment guidelines received £50m from drug research contracts Case shows what’s at stake in the debate around transparency of doctor-industry relations
Source: CC0
A British cardiologist and former president of the European Society of Cardiology (ESC) has been judged by the ESC to have committed severe misconduct after a Danish TV documentary reported that he had been involved in drawing up guidelines for the use of the heart drug ivabradine (Corlanor, Procoralan) while profiting from lucrative contracts with the drug’s maker.
An investigation published by The BMJ describes how between 2006 and 2015, Kim Fox, emeritus professor of clinical cardiology at the National Heart and Lung Institute, Imperial College London, who reportedly served as cardiologist to the late Queen Elizabeth II, and his wife Karen Summers, a former drug industry executive, received more than £50m as co-directors of the UK based contract research organisation Heart Research.
This company was involved in running at least three clinical trials of ivabradine, made by French drug company Servier, that were published in the same period, reports freelance journalist, Laura Spinney.
In 2006, when Fox became president of the ESC, he chaired an ESC taskforce that published a guideline recommending ivabradine as an alternative treatment for angina in patients who couldn’t tolerate beta blockers. Authors were asked to disclose conflicts of interests, but the guideline did not publicly identify what Fox had disclosed.
As outgoing ESC president in 2008, Fox championed the drug publicly though it had failed to meet the primary endpoint in the first of the three trials Heart Research was involved in.
Ivabradine remains approved for angina and heart failure in Europe, and for heart failure in the US, but persistent doubts have been expressed over its efficacy.
It’s rare for alleged conflicts of interest to involve such large sums of money. The case remains unreported in most of Europe and beyond, even though the ESC’s guidelines are influential worldwide.
The Danish documentary, which aired in September 2024, triggered an internal review by the ESC which found that the 2006 guideline recommendations were “appropriate” and reported “no evidence to suggest a bias towards ivabradine.”
But in March 2025, the ESC ethics committee found that Fox “had failed to meet his ethical obligations and considered his behaviour as a severe misconduct.”
Pulmonologist Irène Frachon, described the sums involved in the case as “monstrous” while Rita Redberg, a cardiologist and former editor in chief of JAMA Internal Medicine, said: “It totally goes against the grain of the profession.”
Fox rejected the ethics committee’s findings and resigned his ESC membership, claiming that he had always declared any conflicts of interest and that “he was not prepared to be judged on the basis of rules and regulations described in 2024 retrospectively for activities in 2006 to 2008.”
The ESC admitted to The BMJ that its declarations process was relatively lax in the early 2000s and said that it had been substantially strengthened since. Meanwhile, Servier said that it “strictly complies” with transparency guidelines established by the European Federation of Pharmaceutical Industries and Associations, a trade body.
The Fox story has emerged at a time when some in Europe are pushing for reforms that would end voluntary declaration of conflicts of interest and enshrine greater transparency in law.
For former ESC vice president John Martin, the ESC has not done enough to restore public trust, and the recent revelations risk damaging the doctor-patient relationship while also leaving the volunteers who run the society feeling betrayed. He urged further action. “The ESC board might be seen as tacitly complicit unless there is a thorough public investigation,” he said. “Many questions remain.”
A conversation with Qingyu Chen, PhD, about what medical artificial intelligence learns, what it memorises, and what it takes to use it responsibly.
Qingyu Chen, PhD, and his team set out to study how artificial intelligence language models are adapted for medicine and found that what these models memorise can be both useful and risky. A model may retain valuable medical knowledge, but in a controlled study using real hospital records, the same fine-tuning – the added training that adapts a model to a specific task – that improved diagnostic performance also made it more likely to reproduce material it had seen during training, including sensitive patient information.
The study, published recently in Nature Communications, reflects a question at the center of Chen’s research: How can medical AI become not only more capable but also more reliable and safer? The study was led by its first author, Anran Li, PhD, who conducted the research as a postdoctoral researcher in Yale’s Department of Biomedical Informatics and Data Science.
Chen is an assistant professor of biomedical informatics and data science at Yale School of Medicine, with a secondary appointment in ophthalmology. He leads research on the accuracy and reasoning of medical language models and on multimodal AI-assisted disease diagnosis, which draws on both text and medical images.
The following conversation with Chen discusses what medical AI learns, what it memorises, and what it takes to use it responsibly.
What is your lab’s research focus?
Our lab does two things that are usually treated as separate: We build medical AI, and we study where it fails.
On the building side, we work with two main kinds of information medicine runs on. We develop models that read clinical records and the medical literature, models that analyse medical images to help diagnose disease and predict its course, and systems that combine both, so an AI can weigh a patient’s written history alongside their scans, the way a physician would.
But a model that performs well on a test is not the same as a model you can trust with a patient. So, we also study how these systems fail. They can state falsehoods with complete confidence. They can reach a right answer through faulty reasoning. And, as our recent work shows, they can memorise sensitive information from the data they were trained on.
Our goal is to develop medical AI that is genuinely useful, understand where it breaks down, and produce the evidence needed to know when it can – and cannot – be trusted.
Why work across both text and images?
Because medicine is inherently multimodal. A patient cannot be understood through a single paragraph or a single image. Clinical decisions often require bringing together a patient’s history, laboratory results, medical notes and imaging findings.
Our work spans both sides of that. On the language side, we develop methods to help computers understand clinical records and biomedical literature. On the imaging side, much of our research focuses on medical images and specialties that depend heavily on them—ophthalmology in particular, where we work on diagnosing eye disease and predicting how it will progress. This is also why I hold a secondary appointment in ophthalmology.
Progress here requires more than developing new models. One of the biggest barriers is the limited availability of medical data that is large enough, reliable enough and free for researchers to share.
To help address this, we recently developed MedPMC, a system that has assembled 11 million medical images paired with their accompanying text, drawn from research literature that is openly licensed for reuse—and that is designed to keep growing as new research is published. We’ve made the data, the tools used to assemble it, the tests for measuring performance and the resulting models publicly available so that other institutions can develop, evaluate, reproduce and adapt these systems rather than starting from scratch.
Your team recently published a study in Nature Communications on how AI models ‘memorise’ medical data. What does memorisation mean here?
Memorisation means that a model can recall or reproduce content it encountered during training. If a model has been trained on clinical guidelines, it may reproduce part of a guideline when you give it the opening of that passage. If it has had additional training on a set of medical exam questions, it may produce an answer choice that appeared in that training data – even after we removed that choice from the question.
That is different from simply saying that a model performs well. When a model answers a question correctly, there are several possible explanations. It may have acquired genuine medical knowledge. It may have learned a pattern of reasoning it can apply to problems it hasn’t seen before. Or it may simply recognise the question and reproduce something it previously saw during training. If all we check is whether the final answer is correct, we cannot tell these apart.
So, our goal was to look beyond accuracy and ask a different set of questions: How often does memorisation occur? What types of content get memorised? How much can a model reproduce? Does what it memorised earlier survive further training? And what does all of this mean for using these systems in medicine?
What did you find?
We looked at the main stages a general-purpose model goes through on its way to becoming a medical one.
First, we examined models that had already undergone continued pretraining, in which a general-purpose model is trained further on large collections of medical text, including biomedical literature, clinical guidelines and clinical notes. Second, we evaluated models that had been fine-tuned on the standard question-and-answer datasets that the field uses to test medical models. Third, we conducted a privacy-protected, HIPAA-compliant study using more than 13 000 medical records to fine-tune models for disease diagnosis.
This was done in an isolated and secure computing environment. These records had already been collected in the course of care; the study did not recruit patients or change anyone’s treatment.
Across those settings, we examined both general-purpose models and models trained on medical data, 10 different datasets containing hundreds of thousands of records and thousands of model responses that we manually reviewed.
The patterns differed depending on the training stage. Continued pretraining was more likely to produce long, word-for-word matches to source documents. Fine-tuning produced less long-form copying in some settings, but more memorisation tied to the specific task the model was trained on. For example, after fine-tuning on medical question-and-answer datasets, models reproduced roughly 14% to 21% of the answer choices that had been removed from the question the model was shown.
We also found that memorisation was persistent. Depending on the setting, as much as 87% of what a model memorised during continued pretraining was still present after it had been fine-tuned on a new medical task. Fine-tuning does not necessarily erase what a model previously memorised. It may preserve that content while adding new memorisation specific to the task it was just trained on.
What did the clinical case study show?
The clinical case study showed both the potential benefit and the risk of adapting these models to real-world medical data. Fine-tuning improved diagnostic performance; for one model, the correct diagnosis came up as its first choice 54.8% of the time, up from 48.6%. In some specialties, the gains were larger than that – more than 10 percentage points in areas such as cardiology and nephrology, which deal with heart and kidney conditions.
At the same time, the study showed a real privacy risk. In a controlled test conducted in a secure research setting, we found that the model could sometimes reproduce sensitive information from the records used to train it. This was not something that would happen during patient care, but it shows that privacy risks should be evaluated before models trained on clinical data are shared or deployed.
Is memorisation always harmful?
No. One of the central findings of our study is that memorisation in medicine is not a single behaviour.
We identified three broad types. The first is beneficial memorisation. A model may accurately retain biomedical concepts, clinical guidelines, the medical literature it has read or specific medical knowledge tied to its task. That kind of memory may support factual accuracy and help the model perform medical tasks more effectively.
The second is uninformative memorisation. Models sometimes reproduce document disclaimers, section headings, formatting instructions or other boilerplate language. This adds little medical value and may indicate that the model is learning surface-level patterns rather than deeper medical understanding.
The third is harmful memorisation. This includes reproducing quirks specific to a particular dataset, word-for-word passages from patient notes, protected health information or other sensitive patient content. This form of memorisation may create privacy risks and may also indicate that the model is leaning too heavily on its training data rather than generalising to new cases.
The important question, then, is not simply whether a model memorises; it is what the model memorises, why it memorises it, and whether that memory supports or undermines the medical use it is intended for.
Did anything about the way memorisation develops surprise you?
One thing that stood out: Memorisation did not only show up late, after a model had been trained too long. It began early.
As we trained the models, we tracked their progress and compared three things: how much they were memorising, how well they were learning by the usual measure, and how accurate their diagnoses were. Memorisation began increasing relatively early, even while the standard measures still showed the model improving, and before its diagnoses had reached their peak accuracy.
That means traditional indicators researchers watch during training – such as whether the model keeps improving on held-out data, or the point at which they would normally stop training – are useful but may not be sufficient on their own. A model can appear to be learning effectively according to standard performance measures while simultaneously increasing its retention of training-specific content.
We also found two other patterns. Larger models and longer inputs were generally associated with more memorisation. By contrast, changing common generation settings such as temperature – which controls how varied the model’s answers are – had relatively limited effects. This suggests that memorisation is fundamentally connected to how a model is trained and what data it is exposed to, rather than being something that can simply be solved by adjusting how the model generates its answers after training.
What do you most want readers to take away from this work?
Adapting an AI model to medicine does not simply make it “more medical.” It changes what the model knows, what it remembers and what it may reproduce.
Some of that memory is valuable. We want models to retain accurate medical knowledge and clinical guidance. But we do not want them to rely on meaningless repetition, echo back the answers to test questions or expose sensitive information from patient records.
Trustworthy medical AI therefore requires more than measuring whether a model gets the answer right. We need to understand how it got there, what it retained from training, and whether it will stay safe and reliable when used in a new setting.
Scientists have shown that a naturally occurring virus can dismantle complex bacterial communities responsible for severe gum disease, offering a potential alternative to traditional antibiotics.
In the first study to demonstrate this effect, La Trobe University researchers found the FNU1 virus selectively attacked Fusobacterium polymorphum, a bacterial species that helps other disease-causing bacteria to stick together and thrive.
The virus reduced total bacterial levels by 94 per cent, leaving only about six per cent of the original disease-causing plaque intact.
Despite targeting just one bacterial species, the treatment dramatically reduced the size and density of the plaque, with levels of three other bacteria linked to gum disease falling by more than 85 per cent.
Dr Mwila Kabwe, lead author and Post Doctoral Research Fellow at La Trobe’s Holsworth Biomedical Research Centre, said the findings were important given gum disease affects more than half the world’s population and remains a leading cause of tooth loss.
The research focused on bacteriophages or phages, viruses that infect bacteria but are harmless to humans.
“Current treatments for gum disease don’t directly target the bacterial imbalance that drives the disease, and the dense layers of bacteria that form around teeth can be resistant to antibiotics,” Dr Kabwe said.
“By targeting one of the key bacteria that holds these disease-causing communities together, we were able to destabilise the entire plaque community.
“The findings suggest phages could offer a more precise way to treat gum disease while preserving healthy bacteria.”
While the laboratory findings are promising, further research is needed before phage-based treatments could be used in dental practice.
The study forms part of Dr Kabwe’s early career-research investigating how phages could be used to target harmful bacteria linked to oral disease and other chronic health conditions.
A study by DZNE finds that healthy eating is associated with slower biological aging – and that there is more than one way to eat healthily. The researchers examined ten dietary patterns considered healthy, including the Mediterranean, Nordic, and a plant-based diet, as well as the DASH diet, which is designed to help lower blood pressure. All were linked with slower biological aging, with age-related changes in the DNA serving as markers. These results, based on data from DZNE’s Rhineland Study and confirmed with data from an independent study, are published in the journal Nature Communications.
“Our study suggests that healthy eating goes hand in hand with slower biological aging. The effects are not massive, but they are measurable and relevant for prevention. By slowing the aging process, the risk of age-related diseases such as dementia or cardiovascular disorders can be reduced. Thus, healthy eating contributes to healthy aging,” says Prof. Monique Breteler, Director of Population Health Sciences at DZNE and head of the Rhineland Study. “Importantly, there is probably no single ‘correct’ diet. According to our data, various dietary patterns are associated with slower aging, some more, some less. That is an encouraging finding, because it leaves room to tailor healthy eating to personal and cultural preferences, budget and taste.”
Molecular Aging Measures
For their analysis, the researchers evaluated blood samples and dietary habits from about 7500 women and men: This included 6470 participants of DZNE’s Rhineland Study in Bonn – where the findings were initially made – and 1034 participants of the EPIC-Potsdam Study of the German Institute of Human Nutrition Potsdam-Rehbrücke (DIfE), where the key results were independently confirmed.
From blood, the team determined chemical modifications in the DNA, known as “DNA methylation patterns”, which influence gene activity and thus biological mechanisms. Since these molecular markers change systematically over the course of life, they are considered indicators of biological aging.
“We applied three different approaches, so-called epigenetic clocks, to read biological aging from DNA methylation,” explains Juliana Tavares, a doctoral researcher at DZNE and lead author of the current publication. “To this end, we leveraged state-of-the-art technology, which allowed us to cover about 850 000 sites in the DNA. This is roughly twice as many as in most previous studies.”
Healthy Eating: Different for Everyone
The researchers also matched every participant’s eating habits against the recommendations of each of the ten diets studied. Adherence was quantified and graded using a scoring system based on how closely individuals followed each dietary pattern. “One and the same person could therefore rank high on one diet score and low on another. In other words: People who ate healthily by one standard were often not the same people who ate healthily by another. According to our data, hardly anyone counts as a healthy eater by every standard at once,” says Tavares.
However, across the board, higher diet quality was associated with slower biological aging. This applied to all of the ten diets examined. “While there are measurably differences between the diets, they are modest in absolute terms. It’s the big picture what matters,” says Tavares. “In summary: When it comes to slowing down the aging process, there isn’t just one healthy diet, but indeed a whole repertoire of possible diets with a positive effect.”
One result was particularly noteworthy: the association between diet quality and slower biological aging was most pronounced in smokers, even though smokers were by no means eating healthier than non-smokers. The biological explanation is still open. “Smoking is very harmful to health,” says Tavares, “One explanation for our findings might be that there is just more health benefit to gain for smokers by eating healthy because of the enormous detrimental effects of smoking.”
Shared Biological Pathways
Although each dietary pattern was associated with its own distinct set of DNA methylation sites, the biological mechanisms involved showed substantial overlap. “More than 70 percent of the affected biological pathways were shared across all diets. These involve cell structure, cell signalling and metabolism – processes central to aging and chronic disease. Therefore, it is understandable that the dietary patterns we examined were all linked with a slowing of biological aging,” says Tavares. “A few pathways were diet-specific, and they matched each diet’s main objectives: heart-related pathways were unique to the DASH diet, and pathways related to cognition and memory were unique to the so-called MIND diet, which aims to support brain health. You don’t need a perfect diet – that may be the most practical message here. But moving forward, the diet-specific pathways point to future research on targeted dietary interventions. This could be helpful in designing targeted health recommendations.”
Study found distinct immune changes preceding multiple sclerosis relapse, providing new insight into how Epstein-Barr virus reactivation may trigger an MS attack in people with genetic risk factors
An electron micrograph showing three Epstein-Barr virus (EBV) particles colourised red-orange. Credit: NIAID
A new study illuminates the connection between the Epstein Barr virus (EBV) and multiple sclerosis (MS), pointing to a causal role for reactivation of the virus in triggering MS attacks in people with certain genetic risk factors. In a study of blood samples from more than 100 participants with MS, Mass General Brigham researchers detected increased EBV lytic activity in immune cells up to three months before MS relapse and discovered that these cells also showed elevated expression of genes linked to MS risk. The finding helps explain how a common virus and genetic risk may work together to trigger MS attacks. The work is published in Nature Medicine.
“These findings open a whole new avenue for targeted therapeutics,” said senior author Tanuja Chitnis, MD, director of the Translational Neuroimmunology Research Center and Chief of the Division of Neuroimmunology at Mass General Brigham. “Currently, most MS treatments work by broadly suppressing the immune system. Our results suggest there’s an opportunity to be more precise and develop approaches that target EBV or the immune cells involved in relapse.”
Researchers analysed blood samples from 114 patients with MS and 21 healthy participants in the Comprehensive Longitudinal Investigation of Multiple Sclerosis (CLIMB) study, a decades-long MS cohort based at Brigham and Women’s Hospital, to track immune changes before relapse. The study included samples collected up to 90 days before participants experienced a relapse, allowing researchers to compare the pre-relapse immune state with periods of remission in the same patients.
To identify which immune cells changed the most before a relapse, researchers used single-cell RNA sequencing and other molecular techniques to analyse hundreds of thousands of immune cells. They found that B cells (immune cells that can house dormant EBV) showed some of the strongest changes, activating genes associated with antiviral responses, inflammation and EBV activity. The researchers also observed an increase of ABC-like B cells, a subset of cells linked to viral infections and autoimmune disease. They found that EBV proteins (derived from EBV genes) triggered MS risk genes to be expressed in immune cells prior to relapse, but not during remission or in healthy controls.
If these findings are validated in larger, prospective studies, blood biomarkers of EBV activity could help identify patients at increased risk of relapse, complementing MRI scans and existing blood biomarkers that typically detect disease activity only after inflammation is already underway. The authors note that future studies are needed to determine whether these findings extend to early-stage and progressive forms of MS.
“We believe this work provides foundational insights into the cause of multiple sclerosis,” said Chitnis. “This study puts all the pieces together, showing a timeline of how the reactivated virus interacts with risk genes to unleash inflammation before relapse
A summer intervention programme for autistic children ages 4-6 with no intellectual disability significantly improved their social performance, according to a study by researchers at the Institute for Autism Research (IAR) at the University at Buffalo. It was published in Advances in Neurodevelopmental Disorders.
“Completion of our 5-week summerMAXyc programme was associated with significant social, behavioral and autism symptom benefits for the children,” says Christopher J. Lopata, PsyD, co-author on the study and co-founder/co-director of the IAR with Marcus L. Thomeer, PhD; both are professors of pediatrics in the Jacobs School of Medicine and Biomedical Sciences at UB. “Developing such foundational skills early in development sets the stage for development of more advanced skills later in childhood, adolescence and adulthood, thus leading to better long-term outcomes.”
Thomeer notes that while programmes for autistic children with no intellectual disability ages 4-6 do exist, few provide intensive social programming. This population also tends to engage in significantly fewer group recreational activities in general than their peers without autism, further limiting social development opportunities.
“This constitutes a significant gap in social intervention development and service provision, which prompted our interest in developing such a programme for this population,” he says.
Five weeks, five days a week
The intensive summerMAXyc programme was conducted five-days-a-week over five weeks in the summer. The programme follows a similar model to that of the IAR’s successful, evidence-based summerMAX programme for older children, which has helped improve the social skills of hundreds of children since it began 23 years ago.
In the summerMAXyc programme, each 6-hour day included 8 cycles of 30 minutes each, which began with 10 minutes of instruction in particular skills, such as having a conversation or accepting consequences, followed by a 20-minute cooperative activity.
According to the study, the children found that the programme was fun and helped them learn skills and make new friends. Average child satisfaction was 14.9 out of a possible 15 with parent satisfaction at 69.8 out of a possible 70. Parents reported significant improvement in the children’s social skills, behaviors and autism symptoms.
There is broad consensus that earlier intervention is associated with better outcomes for autistic children. And parents are often the first to recognize the need.
“Parents reach out to us because their child has difficulty navigating the social world,” says Thomeer. “They describe their autistic children as wanting to have friends but not knowing what to do in social situations.”
Families noticed the difference
Over the course of their child’s participation in summerMAXyc, Thomeer says, parents reported that other family members began to notice the child playing with others, not just playing alone. They also noticed that the child now makes eye contact, asks about others, plays games without getting upset and can recognise and understand different emotions.
According to Thomeer, a mother of one of the participants reported that her child not only noticed that she was frustrated about something, but also told her to “squeeze the orange,” a technique the children are taught where they squeeze an imaginary orange and take a deep breath in order to calm themselves down.
One of the study’s major strengths involved the use of objective observers to assess the children’s social performance. According to Lopata, parents are a critical source of information, however there is a risk of bias because they know that their children are in the intervention. Most social skills intervention studies for autistic children rely on parent ratings alone and therefore suffer from this limitation, he adds.
“Our use of masked observers – trained individuals – eliminated that risk because they were unaware that the children received an intervention,” says Lopata.
The authors also note that the intervention was similarly successful in improving the children’s social performance and autism features regardless of the child’s age, IQ level, communication ability or level of diagnostic symptoms.
“There is widespread recognition in the field that no intervention will be effective for all autistic children and there is a significant need to determine which children are most likely to benefit from a specific intervention,” says Lopata. “Answering this question can help ensure that scarce resources are efficiently allocated and that children receiving a given intervention are most likely to benefit.”
The next step is to test the intervention in a large-scale randomised clinical trial with a control group, which would be a first for such a social intervention in this population.
