The research team, led by Tomoko Nagamine and colleagues in the Department of Endocrinology, Metabolism and Nephrology, examined the records of 22 individuals diagnosed with thyroid storm between 2012 and 2024. That number may seem small, but it reflects the reality of a disease that strikes only a tiny fraction of patients with thyrotoxicosis, and it makes the study one of the more detailed single-center analyses of admission-time risk prediction in this population. The researchers classified outcomes into two groups: eighteen patients with favorable outcomes and four patients with poor outcomes, defined as in-hospital death or cardiopulmonary arrest requiring resuscitation, regardless of whether spontaneous circulation was subsequently restored.

To assess severity on arrival, the team applied four scoring systems that occupy different corners of acute medicine. The Acute Physiology and Chronic Health Evaluation II, or APACHE II, is a heavyweight intensive care score that combines twelve physiological measurements with age and chronic health status. The Sequential Organ Failure Assessment, or SOFA, tracks dysfunction across six organ systems with repeated measurements over time, while its abbreviated cousin, quick SOFA, distills the concept to three bedside variables: altered mentation, low systolic blood pressure, and rapid respiratory rate. The Burch–Wartofsky Point Scale, by contrast, was designed specifically for thyroid storm in the early 1990s and assigns points for features such as fever, tachycardia, atrial fibrillation, heart failure, delirium, and precipitating events.

The central finding was strikingly consistent: all four scores were significantly higher in the poor outcome group than in the favorable outcome group. In other words, every instrument captured some signal of impending catastrophe, whether it was built for general critical illness or tailored to thyroid derangement. This convergence matters because it suggests that the physiological chaos of thyroid storm leaves fingerprints across multiple domains of assessment, from respiratory and cardiovascular function to consciousness and renal performance, and that no single organ system tells the whole story.

When the researchers turned to exploratory receiver operating characteristic analysis, the standard statistical technique for judging how well a test discriminates between two groups, the Burch–Wartofsky Point Scale emerged with the highest discriminatory performance. A cutoff of 105 points or more identified poor outcomes with a reported sensitivity of 100 percent and a specificity of 94.4 percent, meaning that in this cohort every patient who went on to a catastrophic outcome scored above the threshold, while nearly all patients who recovered scored below it. The authors are careful to note that these estimates are constrained by the small number of poor outcomes, only four events, which inflates statistical uncertainty and demands validation in larger, independent cohorts before the threshold is adopted clinically.

Beyond the headline scores, the study mapped which clinical parameters traveled with adverse outcomes. Poor outcomes clustered with markers of spreading organ dysfunction: coagulopathy detected through abnormal prothrombin time, renal impairment, metabolic acidosis, impaired consciousness measured on the Glasgow Coma Scale, thrombocytopenia reflecting falling platelet counts, and atrial fibrillation, the chaotic upper-chamber rhythm that is both a classic feature of thyrotoxic cardiomyopathy and a harbinger of hemodynamic collapse. Each of these findings paints a picture of thyroid storm as a systemic disease in which excess circulating thyroid hormone drives a hypermetabolic state that then cascades into liver, kidney, blood, and brain injury.

The mechanistic story behind these associations is well understood in outline. Surging levels of triiodothyronine and thyroxine sensitize the heart to catecholamines, pushing cardiac output beyond sustainable limits while simultaneously impairing the heart’s ability to relax and fill. Fever and sweating cause fluid losses that compound hypotension. The resulting tissue hypoperfusion generates lactic acid, which appears as metabolic acidosis on arterial blood gas analysis. Meanwhile, the prothrombotic and proinflammatory milieu can tip coagulation pathways into dysfunction, and reduced hepatic clearance of clotting factors compounds the problem. Thrombocytopenia may signal both consumption and bone marrow suppression in the sickest patients. The study’s finding that these laboratory derangements distinguished survivors from nonsurvivors fits neatly into this physiological framework.

For emergency physicians and intensivists, the practical implication is that the Burch–Wartofsky Point Scale, despite being three decades old and never originally validated against hard outcomes, may carry prognostic information beyond its diagnostic role. The scale was conceived as a case-finding instrument, a way to decide whether a febrile, tachycardic patient with Graves’ disease has crossed into storm territory. The new data suggest that a markedly elevated score, well above the traditional diagnostic threshold, does not merely confirm the diagnosis but also flags a patient whose physiology is already failing in ways that predict death or arrest. In a disease where mortality historically approached 20 to 30 percent and hinges on rapid administration of thionamides, beta-blockers, iodine, and corticosteroids, any tool that accelerates triage toward intensive care could be consequential.

The study also carries methodological lessons for the field. Thyroid storm is so rare that no single center can assemble the hundreds of patients needed for definitive prognostic modeling, which is why the authors explicitly frame their receiver operating characteristic estimates as exploratory and call for validation in larger independent cohorts. Multicenter registries, such as those maintained by the Japan Endocrine Society and the Japanese Thyroid Association, whose diagnostic criteria informed the case definitions in this work, represent the most plausible path forward. Combining the disease-specific sensitivity of the Burch–Wartofsky scale with the organ-failure granularity of SOFA might ultimately yield a hybrid score that outperforms either instrument alone, a hypothesis the present data cannot test but clearly motivate.

Until such validation arrives, the message for clinicians is one of layered vigilance. Every patient with thyroid storm requires immediate intensive management, as the authors emphasize, but the admission Burch–Wartofsky Point Scale may provide additional information for early risk stratification, particularly when scores climb toward or beyond the 105-point threshold identified here. Coupled with close attention to coagulation parameters, renal function, acid–base status, mental status, platelet counts, and cardiac rhythm, the scale could help clinicians decide which patients need escalation to continuous hemodiafiltration, plasma exchange, or mechanical circulatory support before collapse occurs. In a condition where hours separate recovery from catastrophe, sharpening the first-hour assessment remains one of the most valuable interventions available, and this study adds a measured, cautiously optimistic data point to that effort.

Subject of Research: Prognostic value of admission severity scores, including the Burch–Wartofsky Point Scale, for predicting adverse in-hospital outcomes in thyroid storm

Article Title: Admission Burch–Wartofsky point scale and critical care scores as predictors of adverse in-hospital outcomes in thyroid storm: a retrospective observational study

Article References: Nagamine, T., Yada-Tanabe, T., Kobayashi, S., Nagao, M., Fukuda, I., Sugihara, H., & Iwabu, M. (2026). Admission Burch–Wartofsky point scale and critical care scores as predictors of adverse in-hospital outcomes in thyroid storm: a retrospective observational study. BMC Endocrine Disorders. https://doi.org/10.1186/s12902-026-02614-2

Image Credits: AI Generated

DOI: 10.1186/s12902-026-02614-2

Keywords: thyroid storm, Burch–Wartofsky Point Scale, APACHE II, SOFA, qSOFA, risk stratification, prognostic markers, critical care, hyperthyroidism, atrial fibrillation, organ failure, retrospective study

The case for looking beyond traditional biobanks rests on a structural weakness in how cardiovascular research is currently conducted. Large-scale databases such as UK Biobank, the All of Us Research Program, and the China Kadoorie Biobank contain cardiovascular data from hundreds of thousands of participants and have already informed work on predictive models and drug repurposing. Yet these resources typically rely on single-country data and participant pools that do not reflect the general population. That lack of representativeness limits how well findings generalise across different geographies, ethnicities, and socioeconomic groups, ultimately constraining the effectiveness of the research they support. A database built from smartwatches, by contrast, would draw on a device worn by nearly a quarter of the world’s population, spanning continents and demographics in a way no conventional cohort study could realistically achieve.

The clinical potential of smartwatch data is no longer speculative. Every day, these devices track heart rate, heart rate variability, blood pressure, and sleep quality, generating a longitudinal record of cardiovascular function that would have been unimaginable a generation ago. Case reports have documented smartwatch data leading to diagnoses of conditions such as cardiomyopathy and atrial fibrillation, giving patients the chance for early intervention before catastrophic events occur. In one observational clinical study, smartwatch-derived heart rate variability data showed high concordance with high-resolution electrocardiogram measurements in patients with established cardiovascular disease, demonstrating that consumer-grade sensors can approach the accuracy of clinical gold standards. As wearables become more accurate and their measurements more diverse, research institutions and technology companies have begun formal partnerships to harness this data at scale.

Several flagship initiatives illustrate what is already possible. The British Heart Foundation Data Science Centre has explored how to integrate smartphone and wearable information into a resource that can be linked to participants’ National Health Service records and to cardiovascular outcomes. Apple has partnered with the American Heart Association and Brigham and Women’s Hospital on the Apple Heart & Movement Study, which collects individual-level data to explore the relationships between activity, wellness, and health. Singapore has launched Health Insights Singapore, known as hiSG, in which participants are provided with a smartwatch to take part in a study assessing the health behaviours and lifestyles of residents. These programmes demonstrate that large-scale wearable-based cardiovascular research is feasible, but they also expose the barriers that must be resolved before such efforts can deliver equitable, globally representative science.

One of the thorniest issues is the tension between commercial and research ethics. Research using patient data is traditionally conducted under strict ethical guidelines, but data collected by smartwatch companies is typically geared toward commercial and profit-oriented goals. If that data were repurposed for research, companies would likely need to adhere to more stringent ethical requirements, particularly around confidentiality, which could prove difficult when commercial and research interests conflict. The Apple Heart & Movement Study offers a cautionary example: although the study allows users to share their health records through a smartphone app, only about 10 percent of users were able to do so, owing to difficulties in interoperability between health records systems and smartwatch apps. The episode highlights persistent problems of data standardisation and sharing, complications that multiply when researchers attempt to integrate data across multiple smartwatch manufacturers to maximise the number of participants.

Beneath these organisational challenges lie genuine technical problems rooted in how the devices themselves work. Modern smartwatches generally rely on photoplethysmography, a low-cost optical technique that detects blood volume changes in the skin’s microvasculature to obtain direct heart rate measurements. Manufacturers then apply proprietary sensors and algorithms to derive additional metrics such as resting heart rate, meaning that identical physiological states can produce different readings on different devices. Standardisation between devices is therefore essential if data from millions of heterogeneous wearables is to be pooled meaningfully. More troubling still, research indicates that photoplethysmography may perform less accurately on darker skin tones due to melanin’s effect on light absorption, a disparity that demands correction algorithms and additional validation to ensure the resulting datasets remain accurate and representative of the full diversity of the populations they claim to describe.

Equity concerns extend well beyond sensor physics. A scoping review found that while wearables can help reduce cardiovascular disease burden in low- and middle-income countries, adoption is hindered by technological literacy, cost, and cultural considerations. As smartwatches grow more sophisticated and potentially less user-friendly, ownership among older adults, a population at the highest cardiovascular risk, may actually decline. Affordability compounds the problem: in 2019, 31 percent of US households that purchased a smartwatch earned more than 75,000 dollars, while only 12 percent of households earning under 30,000 dollars owned one. Individuals with lower incomes who do own devices often hold older or cheaper models that track heart-related biometrics less comprehensively and less accurately. Ethnic representation is equally problematic. In a public survey by the British Heart Foundation on which smartwatch data would be most useful for cardiovascular research, only 6 percent of the 194 respondents identified as non-white, a figure that, while not reflecting the demographics of smartwatch owners overall, signals the risk of research being shaped by and tailored toward particular groups, thereby entrenching the very health inequities such research aims to reduce.

Potential remedies are already being tested. Researchers and manufacturers could provide subsidised smartwatches, as the hiSG initiative in Singapore is doing, paired with targeted subsidy schemes and educational campaigns in low- and middle-income countries, and hands-on training for individuals willing to participate but facing practical limitations. Data governance presents a parallel challenge. Ownership of smartwatch data is dictated by terms and conditions that consumers agree to when purchasing a device or using its apps. While the most popular manufacturers maintain that consumers own their data and can delete it at any time, the same terms usually specify that the company can control and share that data. In the Apple Heart & Movement Study, data access is granted to Apple, Brigham and Women’s Hospital, the American Heart Association, and the Research Studies Support Center. Any international database would need to comply with research policies from multiple countries and with consumer disclosure laws governing how the data is used by both manufacturers and research stakeholders. Consumers will need genuine reassurance about safety, particularly regarding involuntary surveillance and medicalisation, communicated through concise and accessible means rather than buried in excessively lengthy legal documents.

The scientific payoff, if these hurdles can be cleared, would be substantial. Mapping longitudinal cardiovascular data from smartwatches could reveal the early physiological changes that precede the onset of cardiovascular disease, improving disease prediction and opening windows for early intervention. Connecting biometric data to cardiovascular outcomes and integrating it into researcher-accessible databases could accelerate insights into arrhythmia, hypertension, heart failure, and ischaemic events, ultimately supporting the optimisation of treatment. The hope articulated in the editorial is that such efforts, by sharpening prediction, diagnosis, and early intervention, could reverse the stagnated progress in reducing global cardiovascular mortality. But the authors are clear that this outcome is not automatic. It depends on smartwatch manufacturers and researchers working collaboratively to build globally representative databases, maximise their scientific utility, and place equitable research goals at the centre of the enterprise rather than treating them as an afterthought.

The smartwatch on a commuter’s wrist is, in effect, an unfinished clinical instrument: rich in potential, uneven in accuracy, and governed by rules written for commerce rather than science. Whether the billions of heartbeats it records each day become the foundation of a truly global cardiovascular database or merely another source of commercial data will depend on decisions now being made about standardisation, privacy, and access. The editorial’s message is ultimately one of conditional optimism. The technology exists, the partnerships have begun, and the clinical evidence is accumulating. What remains is the harder work of ensuring that the benefits of wearable-driven cardiovascular research reach the populations where the disease burden is greatest, not just those who can afford the newest device.

Subject of Research: Using smartwatch-derived cardiovascular data for equitable cardiovascular disease research

Article Title: From wrist to research: Harnessing smartwatch data for equitable cardiovascular research

Article References: Kaur, M., & on behalf of the PLOS Medicine Staff Editors (2026). From wrist to research: Harnessing smartwatch data for equitable cardiovascular research. PLOS Medicine, 23(9), e1005260. https://doi.org/10.1371/journal.pmed.1005260

Image Credits: AI Generated

DOI: 10.1371/journal.pmed.1005260

Keywords: smartwatches, cardiovascular disease, wearable technology, photoplethysmography, health equity, biobanks, data privacy, heart rate variability, low- and middle-income countries, digital health, atrial fibrillation, data standardisation

The numbers behind the headline finding are sobering. Across the studies that compared patients with central retinal artery occlusion against matched controls free of ocular events and stroke at baseline, 14.26 percent of eye-occlusion patients developed atrial fibrillation during follow-up, compared with 8.32 percent of controls. Expressed as a pooled odds ratio using a random-effects model, that translates to a 47 percent increase in the odds of developing the arrhythmia (OR 1.47, 95 percent confidence interval 1.18 to 1.83; p = 0.0006), with remarkably low heterogeneity between studies (I squared of just 1 percent). The secondary outcome was even more alarming: new-onset ischemic stroke occurred in 13.8 percent of eye-occlusion patients versus 6.7 percent of controls, a more than twofold elevation in risk (OR 2.10, 95 percent CI 1.57 to 2.82; p < 0.00001, with zero heterogeneity). In total, 115 cases of new-onset atrial fibrillation were recorded during a median follow-up of 34 months.

Perhaps the most provocative result is the comparison with cerebral stroke itself. When the researchers contrasted new-onset atrial fibrillation after central retinal artery occlusion with new-onset atrial fibrillation after ischemic stroke in other brain territories, the difference vanished: the pooled odds ratio of 0.80 (95 percent CI 0.40 to 1.61; p = 0.54) indicates that the eye event carries essentially the same arrhythmic risk as a stroke. Continuous monitoring with implantable loop recorders revealed cumulative atrial fibrillation incidence of 33.4 percent at 12 months and 49.6 percent at 24 months after the ocular event, figures that closely mirror the 33.6 percent and 43.2 percent seen at the same time points in the ischemic stroke cohort. On this evidence, the authors argue, central retinal artery occlusion should be regarded as a stroke equivalent, a red flag that demands the same intensive cardiac work-up that neurologists already apply after cryptogenic stroke or transient ischemic attack.

The pathophysiological logic supporting this association is straightforward. Central retinal artery occlusion occurs when a sudden vascular blockage cuts off perfusion to the retina, triggering rapid ischemic cellular damage and typically an acute, painless decrease or complete loss of monocular vision. In its non-arteritic form, which accounted for all patients in the included studies, the occlusion is usually embolic in origin, with clots travelling from the ipsilateral internal carotid artery, the aortic arch or, crucially, the heart itself, where atrial fibrillation can generate thrombi in the fibrillating atria. The eye and the brain share the same upstream risk factors, including hypertension, hyperlipidemia, diabetes mellitus, coronary artery disease and smoking, and the CRAO cohort in this analysis carried a heavy comorbidity burden: 54 percent had hypertension, 29 percent hyperlipidemia and 26 percent diabetes. Adding weight to a shared thromboembolic mechanism, prior MRI studies cited in the paper found silent cerebral ischemia in 24 to 34.2 percent of patients at the time of acute presentation with monocular visual loss.

One of the study’s most consequential findings concerns timing, and it exposes what the authors describe as a substantial monitoring gap in current clinical practice. New-onset atrial fibrillation after the ocular event follows a bimodal pattern. Early-onset cases appear within days, detected at a mean of just 0.1 months, or about three days, after the event by short-term Holter monitoring. But the bulk of cases emerge far later: under continuous surveillance with implantable loop recorders, cumulative incidence climbs steadily to nearly half of all patients by two years. Standard 30-day monitoring protocols, the analysis concludes, would miss approximately 70 percent of the atrial fibrillation cases eventually diagnosed within two years, and only about one in seven patients ultimately found to have the arrhythmia is captured within the first month. In other words, the conventional short window of cardiac surveillance is profoundly insufficient for this population, leaving the majority of cardiovascular threats effectively off-camera.

The stakes of missing that window are high. Within an acute window of roughly six weeks, 7.4 percent of patients in one included study experienced a symptomatic new-onset ischemic stroke after the ocular event, and registry data spanning 10 to 12 years of follow-up indicate that long-term stroke risk peaks during the first four years. Recurrent ischemic events may even occur before atrial fibrillation is ever identified, which means that a patient can suffer a second, potentially devastating embolic event while the underlying rhythm disorder remains undetected. Because there is currently no evidence-based acute treatment that consistently improves visual outcomes in central retinal artery occlusion, with options such as ocular massage, hyperbaric oxygen, thrombolysis and antithrombotic therapy showing inconsistent results, the authors emphasize that the most valuable intervention may lie not in rescuing the eye but in protecting the brain through structured cardiovascular evaluation and secondary prevention.

Translating these findings into practice, the researchers propose that central retinal artery occlusion without an identified cause should be managed much like cryptogenic stroke, the diagnostic paradigm established by landmark trials such as CRYSTAL AF and EMBRACE. That means urgent, ideally same-day cardiovascular evaluation, serial electrocardiograms, and extended rhythm surveillance using implantable loop recorders or insertable cardiac monitors capable of capturing the delayed-onset, paroxysmal episodes that intermittent monitoring misses. Ophthalmological work-up with fundoscopy, dilated examination, optical coherence tomography and fluorescein angiography remains essential to characterize the retinal ischemia, but the authors stress that interdisciplinary cooperation between eye specialists, neurologists and electrophysiologists is vital. Where atrial fibrillation is confirmed, anticoagulation should be strongly considered to mitigate the risk of future embolic events, although the authors caution that blanket anticoagulation for all eye-occlusion patients remains premature in the absence of prospective randomized evidence and must be individualized through careful risk-benefit assessment.

The study is not without limitations, and the authors are candid about them. The pooled estimate for new-onset atrial fibrillation rests on only two studies, and the seven included investigations differed substantially in follow-up duration, which ranged from seven-day clinical trial surveillance to 15-year nationwide registry follow-up, and in detection methodology, spanning administrative ICD-9 and ICD-10 codes, intermittent Holter monitoring of 24 hours to 7 days, and continuous implantable loop recorders that flag episodes of two minutes or longer. Because implantable devices detect far higher atrial fibrillation rates than claims data or short-term electrocardiography, methodological differences likely explain much of the variation in reported prevalence, and the limited number of studies precluded subgroup analyses. The predominance of retrospective and registry-based designs heightens susceptibility to selection bias, residual confounding and coding inaccuracies, and subclinical or previously undiagnosed arrhythmia could not be fully excluded. Whether atrial fibrillation identified after the ocular event is truly new, previously silent, or an incidental finding remains uncertain, and the authors frame their cardioembolic interpretation as hypothesis generating.

Even with those caveats, the central message is difficult to ignore. Quality assessment using the Newcastle-Ottawa Scale scored most included studies at seven points or higher, funnel plot analysis revealed no evidence of publication bias, and the consistency of the stroke findings across cohorts with zero heterogeneity lends credibility to the core conclusion. For clinicians, the implication is that a patient presenting with a pale retina and a cherry-red spot should trigger not only an urgent eye evaluation but a cardiac investigation equivalent to that performed after a cerebral stroke. For the public, the message is starker still: sudden, painless vision loss in one eye is never just an eye problem. It may be the first and only warning that the heart’s upper chambers are silently throwing clots, and that a disabling stroke could be next. Recognizing the eye event as a cardiovascular emergency, and monitoring the heart long after the initial crisis has passed, could mean the difference between a patient who loses sight in one eye and one who loses much more.

Subject of Research: Association of central retinal artery occlusion with new-onset atrial fibrillation and ischemic stroke risk

Article Title: Central retinal artery occlusion: meta-analysis of the risk of new-onset atrial fibrillation and ischemic stroke

Article References: Gupta, P., Anjos, R., Miguel, A., Bekoju, P., Panthi, R. C., Salangsang, J., Lioutas, V. A., Hilbert, S., Seewöster, T., Bode, K., Bollmann, A., & Nedios, S. (2026). Central retinal artery occlusion: meta-analysis of the risk of new-onset atrial fibrillation and ischemic stroke. Clinical Research in Cardiology. https://doi.org/10.1007/s00392-026-03012-3

Image Credits: AI Generated

DOI: 10.1007/s00392-026-03012-3

Keywords: central retinal artery occlusion, atrial fibrillation, ischemic stroke, meta-analysis, cardiac monitoring, implantable loop recorder, anticoagulation, cryptogenic stroke, retinal ischemia, stroke prevention, cardiology, ophthalmology

The patient’s story began with two years of troubling palpitations. When his heart raced, an electrocardiogram captured an irregular narrow-complex tachycardia at 111 beats per minute, complete with right bundle branch block morphology and a leftward axis deviation of minus 21 degrees. Yet his heart was structurally sound: transthoracic echocardiography showed preserved left ventricular function and no abnormalities, and his thyroid function was normal. High-resolution, contrast-enhanced computed tomography revealed a crucial anatomical detail that would later prove decisive. His left superior and left inferior pulmonary veins did not drain into the left atrium separately, as they do in most people. Instead, they formed a single left common pulmonary vein, a variant present in a minority of the population that creates a wider, more complex target for any ablation strategy.

His first procedure was, by every conventional measure, a textbook success. Under general anesthesia with propofol and fentanyl, ventilated through a supraglottic i-gel airway, he underwent pulsed field ablation using the FARAPULSE system from Boston Scientific. The team delivered twenty-two applications to the left common pulmonary vein, ten to the right superior vein, and eight to the right inferior vein, alternating between the device’s basket and flower electrode configurations. Each application used a 2.0 kilovolt biphasic waveform with four pulse trains, and catheter contact was verified before every delivery using fluoroscopy and intracardiac echocardiography. When the team finished, electro-anatomical mapping with the EnSite NavX system confirmed that all the pulmonary veins had been acutely isolated from the left atrium. The electrical triggers responsible for his arrhythmia appeared, on the maps, to be sealed off for good.

Three months later, an external event monitor told a different story. Atrial fibrillation had returned. The team brought him back for a second procedure, this time with an esophageal temperature probe in place and access gained through the right internal jugular and right femoral veins. Using the OctaRay mapping catheter and the CARTO3 v8 system from Biosense Webster, they began reconstructing the arrhythmia in three dimensions. Almost immediately, they caught something valuable in the act: spontaneous premature atrial contractions, the stray beats that ignite atrial fibrillation, firing on the monitor in real time. Mapping placed their origin just beneath the ostium of the left common pulmonary vein on the posterior wall of the left atrium, slightly apart from the isolation line created during the first procedure.

What followed was a genuine detective story played out with millimeter precision and millisecond timing. Radiofrequency applications with a QDOT Micro catheter at 35 watts and a target ablation index of 400 transiently suppressed the premature beats but could not eliminate them permanently. Meanwhile, voltage mapping during right atrial pacing revealed that the left pulmonary vein had electrically reconnected to the atrium, with ripple mapping pinpointing the anterior carina, the ridge between the vein branches, as the earliest site of reconduction. The puzzle was this: the recurring trigger sat on the posterior wall, far from the isolation line, yet the reconnected vein was waking up first at the anterior carina. The two findings only made sense together if electricity was traveling along a concealed route outside the endocardial surface.

To test that hypothesis, the team paced from four distinct sites inside the left common pulmonary vein and measured how long each impulse took to reach the distal coronary sinus electrode. The numbers were striking. Pacing from the posterior bottom of the vein produced a conduction time of 140 milliseconds, and from the posterior roof 138 milliseconds, while pacing from the anterior ostium produced just 30 milliseconds. When the map was reconstructed with pacing from the vein, the earliest activation site matched the uneliminated trigger exactly. The pattern pointed to an epicardial connection, a subepicardial muscular bundle linking the anterior carina of the left pulmonary vein to the posterior atrial wall near the vein’s ostium, rather than a simple endocardial gap in the ablation line. The team targeted this presumed pathway at the inferior portion of the anterior carina with 35 watts and a target ablation index of 500, completed the pulmonary vein isolation, and rendered the arrhythmia non-inducible.

The anatomical suspect in this case has a name that cardiologists know well: the ligament of Marshall. This vestigial structure, a remnant of an embryonic vein, carries the Marshall bundle of myocardial fibers from the coronary sinus up the epicardial surface between the left atrial appendage and the left superior pulmonary vein. Its fibers can link the coronary sinus, the posterior and lateral left atrial wall, the ridge of the appendage, and the pulmonary vein antrum into a subepicardial conduction network that no endocardial ablation line can interrupt until lesions become fully transmural. Epicardial connections of this kind are not exotic curiosities. Studies estimate they occur in roughly 11 to 14 percent of patients with atrial fibrillation, and they are more numerous in patients with advanced disease and enlarged left atria. High-density mapping studies have shown that most cluster at the pulmonary vein carina and can extend well beyond the antral ablation line, explaining residual vein potentials and preserved atrial capture even when entrance and exit block appear complete.

Here the left common pulmonary vein becomes more than an anatomical footnote. Research by Kueffer and colleagues has found that this vein variant shows the lowest durability after pulsed field ablation, and the reasoning is anatomically elegant. In patients with a left common trunk, the ligament of Marshall lies closer to the ablation target, increasing the opportunity for epicardial fibers to bypass standard lesion sets. Intriguingly, the picture reverses with radiofrequency energy: lower recurrence rates have been reported for left common pulmonary vein patients treated with point-by-point radiofrequency, while the cryoballoon, another one-shot device like FARAPULSE, has shown reduced durability in the same anatomy. The implication is uncomfortable for the newest technology. One-shot ablation devices, which deliver energy in a fixed geometric pattern, may be inherently more vulnerable to anatomical variants than flexible, point-by-point techniques that an operator can adapt lesion by lesion.

The outcome for the patient, at least so far, is encouraging. After the second procedure he took no anti-arrhythmic drugs or beta-blockers, only a single anticoagulant for six months. Over a full year of follow-up, serial 24-hour Holter monitoring and 12-lead electrocardiography found no recurrence of atrial fibrillation, and he remained entirely asymptomatic, with annual Holter monitoring planned thereafter. The authors are candid about the limits of their evidence. This is a single case that cannot be generalized to everyone with a left common pulmonary vein, and because no electrode catheter was placed within the epicardial pathway and no ethanol was infused into the vein of Marshall, the exact route of the connection was inferred from post-pacing intervals rather than directly proven. The endocardial latency gap, a phenomenon in which conduction recovers with delayed timing, could not be definitively excluded, though the successful ablation at the anterior carina strongly supports the epicardial explanation.

Even with those caveats, the case lands at a consequential moment. Pulsed field ablation is expanding rapidly across the world on the strength of high acute isolation rates and a favorable early safety profile, sparing the esophagus and phrenic nerve from the collateral injury that thermal energy sources can cause. But durability, not acute success, is what determines whether a patient is truly cured, and long-term data relative to conventional treatment remain uncertain. The message from this case is that a perfect-looking lesion set on the inside of the heart can be quietly defeated by conduction running along the outside. For patients with a left common pulmonary vein, the authors argue, careful evaluation and explicit consideration of epicardial reconnection should be part of every ablation plan. As pulsed field technology matures, the hearts it fails to fix may teach clinicians as much as the ones it heals, and the answers may lie in structures no catheter can see from within.

Subject of Research: Epicardial connection as a mechanism of atrial fibrillation recurrence after pulsed field ablation

Article Title: Epicardial Connection as a Resource of Atrial Fibrillation Recurrence After Pulsed Field Ablation: A Case Report

Article References: Lee, K. H., Kitamura, T., Sahashi, S., Sugiyama, H., Izumi, C., & Hayashi, K. (2026). Epicardial Connection as a Resource of Atrial Fibrillation Recurrence After Pulsed Field Ablation: A Case Report. Clinical Case Reports, 14(10), Article e73282. https://doi.org/10.1002/ccr3.73282

Image Credits: AI Generated

DOI: 10.1002/ccr3.73282

Keywords: atrial fibrillation, pulsed field ablation, epicardial connection, pulmonary vein isolation, ligament of Marshall, left common pulmonary vein, electroanatomic mapping, cardiac electrophysiology, FARAPULSE, arrhythmia recurrence, case report, ablation durability

The research, led by Mengyuan Jing, Haoxiang Lu, and Qing Liu with colleagues at Lanzhou University Second Hospital, the Guangdong Cardiovascular Institute, and partner institutions, set out to build an explainable machine learning model that could predict functional improvement after ablation in patients with atrial fibrillation combined with heart failure. Rather than relying on a single measurement, the team fused three complementary streams of information: the geometry of the left atrium and pulmonary veins as seen on cardiac CT angiography, quantitative radiomics features extracted from the left atrial wall, and a handful of routine clinical variables. The result was a combined model, dubbed COMB, that achieved an area under the receiver operating characteristic curve of 0.866 in the training set, 0.803 in the validation set, and 0.845 in an independent testing set drawn from other hospitals.

Those numbers deserve unpacking, because the area under the curve, or AUC, is the workhorse metric for judging how well a model separates patients who will improve from those who will not. A value of 0.5 would mean the model performs no better than a coin flip, while 1.0 would indicate perfect discrimination. Scores consistently above 0.80 across three separate patient cohorts, including an external test set the model had never encountered during development, represent solid, clinically meaningful performance. Equally important is how the team arrived at that performance. Patients from the primary institution were randomly divided into training and validation sets in a seven-to-three ratio, while two other organizations contributed entirely separate testing cohorts, a design that guards against the model simply memorizing the quirks of one hospital’s scanner or patient population.

The technical heart of the study lies in what the researchers chose to measure. Cardiac CT angiography, already a routine part of pre-ablation planning because it maps the pulmonary veins before catheter insertion, provides exquisitely detailed three-dimensional images of the left atrium, the chamber where fibrillation typically originates. From these images the team extracted morphological features describing the shape of the left atrium and the pulmonary veins, capturing subtle geometric signatures such as chamber distortion and remodeling that a human reader might overlook. In parallel, they applied radiomics, a computational approach that converts medical images into hundreds of quantitative descriptors of texture, intensity, and spatial pattern, to the left atrial wall itself, the thin muscular sleeve where ablation lesions are created and where fibrotic change often determines whether the procedure succeeds.

From this high-dimensional feature space, the researchers distilled two focused models. Two shape features and three left atrial wall radiomics features survived rigorous screening and were used to construct what they called the Shape model and the Wall model, respectively. Each model produced a continuous score, the Shape_score and the Wall_score, which quantified how strongly a given patient’s cardiac anatomy resembled that of patients who went on to improve. These scores were then integrated with the clinical variables that proved most informative, namely gender, hyperlipidemia, blood urea level, and the type of atrial fibrillation, to yield the final combined model. The parsimony is striking: out of the vast number of features that could have been included, the final predictor rests on just a handful of imaging and clinical inputs, all obtainable from examinations and blood tests that are already standard practice.

What elevates the study beyond a typical prediction exercise is its commitment to explainability. Black-box algorithms have long been a stumbling block for clinical adoption, because physicians are understandably reluctant to act on a probability they cannot interrogate. The researchers addressed this by applying SHAP, or Shapley additive explanations, a technique borrowed from cooperative game theory that assigns each input feature a precise contribution to every individual prediction. In effect, SHAP reveals which factors pushed a particular patient’s predicted probability up or down, and by aggregating these contributions across the cohort, the team identified the features with the greatest impact on outcomes in both the Shape and Wall models. This transparency allows a cardiologist to see, for example, whether an unfavorable prediction stems from pronounced atrial remodeling, abnormal wall texture suggesting fibrosis, or a clinical factor such as persistent rather than paroxysmal fibrillation.

The clinical stakes of this kind of tool are considerable. Patients with atrial fibrillation and concomitant heart failure represent a particularly vulnerable group, and ablation in these individuals carries procedural risk, substantial cost, and a demanding recovery. If a model can reliably flag patients unlikely to experience functional improvement, clinicians could counsel them more honestly, weigh alternative management strategies such as optimized medical therapy, or reserve ablation for those with the greatest expected benefit. Conversely, identifying patients with a high predicted probability of improvement could support earlier referral and shared decision-making grounded in quantitative evidence rather than clinical intuition alone. Because the model’s inputs come from CT angiography and routine laboratory work, it could in principle be deployed without any additional testing beyond what pre-ablation workups already include.

The multicenter architecture of the study strengthens its claims in another important way. Models trained and tested within a single institution often flatter themselves, absorbing center-specific artifacts in scanner protocol, image reconstruction, and patient mix. By including 240 patients in the training set, 101 in the validation set, and 75 in external testing sets from two other organizations, the team demonstrated that the model’s discrimination held up across different hospitals and imaging environments. The ethics committees of all three participating institutions approved the retrospective study, and the requirement for individual informed consent was waived because of its retrospective design, with the work conducted in accordance with the Declaration of Helsinki.

The authors themselves are careful about the limits of what they have shown. The study was retrospective, meaning it looked backward at patients who had already undergone ablation rather than prospectively assigning the model to guide care in real time. The team explicitly states that further prospective evaluation is required before the model can be implemented clinically, a caveat that applies to nearly every prediction algorithm now emerging in cardiovascular medicine. Prospective validation would involve using the COMB model to generate predictions before ablation and then tracking whether those predictions match observed outcomes, ideally across diverse populations and healthcare systems. Questions about calibration, the agreement between predicted probabilities and actual event rates, and about how the model behaves in subgroups underrepresented in the data, would also need answers.

Even with those caveats, the study offers a compelling glimpse of where cardiac imaging and artificial intelligence are converging. The same CT scan obtained to map a patient’s pulmonary veins becomes, through shape analysis and radiomics, a quantitative portrait of atrial health, and machine learning converts that portrait, together with a few clinical facts, into an individualized forecast of recovery. The SHAP framework keeps the forecast auditable, showing clinicians exactly which anatomical and biological signals drove the conclusion. For the millions of patients whose fibrillation and heart failure travel together, and for the physicians deciding whether to offer them ablation, a tool that turns pre-procedural images into evidence-based expectations could reshape the conversation, replacing uncertainty with a number that both doctor and patient can understand, question, and trust.

Subject of Research: Explainable machine learning on cardiac CT angiography to predict functional improvement after atrial fibrillation ablation in patients with heart failure

Article Title: Explainable machine learning model based on cardiac CT angiography for predicting functional improvement after atrial fibrillation ablation: a multicenter study

Article References: Jing, M., Lu, H., Liu, Q., Jing, Y., Lei, F., Yang, X., Chen, G., Xi, H., Xin, W., Zhu, H., Sun, Q., Zhang, Y., Ren, J., Ren, W., Liu, Z., Wang, G., & Zhou, J. (2026). Explainable machine learning model based on cardiac CT angiography for predicting functional improvement after atrial fibrillation ablation: a multicenter study. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02822-1

Image Credits: AI Generated

DOI: 10.1186/s12880-026-02822-1

Keywords: atrial fibrillation, heart failure, catheter ablation, cardiac CT angiography, machine learning, radiomics, left atrium, SHAP explainability, predictive modeling, multicenter study, BMC Medical Imaging, cardiac imaging

A new review published online in the Journal of Intensive Medicine on July 6, 2026, led by a team including researchers at Mayo Clinic in Arizona, with collaborators at the University of Pennsylvania in the United States and Mansoura University in Egypt, assembles the current evidence on sepsis-associated atrial fibrillation, abbreviated SAAF. The review examines how often the arrhythmia occurs, which patients are most vulnerable, how it is detected and diagnosed, and how it is currently managed, before laying out the research agenda needed to close the remaining gaps. The corresponding author, Sameh Hozayen, MB, BCh, of the Division of Hospital Internal Medicine at Mayo Clinic in Arizona, describes the condition as sitting in a clinical blind spot: common, dangerous, and treated without a map. The stated aim of the review is to bring existing evidence together in one place and to define precisely what still needs to be learned so that these patients can receive standardized, evidence-based care rather than ad hoc, institution-by-institution improvisation.

The clinical stakes are considerable. Unlike primary atrial fibrillation, which arises without a clear acute trigger and is supported by decades of randomized trial data and established treatment guidelines, SAAF develops in response to the acute stress of critical illness. In many patients it resolves once the underlying infection is brought under control, which has historically encouraged the assumption that it is a transient and relatively benign companion to sepsis. The review makes clear that this assumption is untenable. Sepsis-associated atrial fibrillation is associated with a 4.6-fold increase in stroke risk and a 1.7-fold increase in in-hospital mortality compared with sepsis patients who never develop the arrhythmia. Recurrence is common, affecting up to half of patients within five years, and survivors carry elevated risks of heart failure, stroke, and death that persist long after they have been discharged from the hospital and the acute infection has been cured.

Part of what makes SAAF so difficult to manage is that its causes are fundamentally multifactorial. The review traces a convergence of mechanisms: systemic inflammation driven by the septic response, autonomic dysfunction that disrupts the neural regulation of heart rate, direct myocardial injury, hemodynamic stress from fluctuating blood pressure and fluid shifts, and electrolyte disturbances that destabilize the electrical properties of the atria. Each of these forces alone can promote arrhythmia; together, in a critically ill patient, they create an environment in which the heart’s upper chambers are pushed toward disorganized electrical activity. This mechanistic complexity also explains why the arrhythmia’s course tracks so closely with the trajectory of the underlying illness, waxing and waning as the infection, the inflammatory response, and the patient’s hemodynamic status evolve hour by hour.

That volatility would be challenging enough on its own, but the therapeutic landscape is complicated by a more fundamental problem: no randomized studies have compared management strategies specifically in this population. Clinicians are therefore forced to extrapolate from trials conducted in patients with primary atrial fibrillation, an imperfect fit for critically ill patients whose physiology can shift dramatically within hours and whose organ dysfunction alters drug metabolism, tolerance, and risk. The review is explicit that this extrapolation, while unavoidable in current practice, leaves every major treatment decision shadowed by uncertainty, and that acknowledging that uncertainty is the first step toward designing the studies that will eventually resolve it.

The uncertainties begin with the simplest question: how fast should the heart be allowed to beat? For rate control, beta-blockers appear superior to alternative agents in observational data, but no randomized trial has confirmed which approach is genuinely best in septic patients, who may tolerate beta-blockade poorly when blood pressure is fragile. For rhythm control, electrical cardioversion can restore normal sinus rhythm, yet it frequently fails to hold while the underlying illness persists, because the same inflammatory, autonomic, and metabolic forces that triggered the arrhythmia remain active and quickly reassert themselves. Choosing between accepting a rapid rhythm and repeatedly attempting to restore sinus rhythm is, at present, a matter of clinical judgment rather than evidence.

The thorniest question of all is anticoagulation, the use of blood-thinning medication to prevent the clots that cause stroke in atrial fibrillation. Critically ill patients with sepsis face simultaneous and opposing risks: dangerous clotting on one side and dangerous bleeding on the other. Standard stroke-risk prediction tools used to guide anticoagulation in primary atrial fibrillation have not been validated in the sepsis setting, so the usual calculus for deciding who should be treated does not reliably apply. In observational studies, anticoagulation has not clearly reduced stroke risk in these patients—and in one multicenter cohort it actually increased bleeding—leading many clinicians to defer it. The review stresses that deferral or initiation must be individualized to each patient’s phase of illness, weighing the evolving clotting and bleeding risks as sepsis progresses or resolves, rather than applying a fixed rule.

Against this backdrop of uncertainty, the review distills the field’s needs into concrete research priorities. The authors call for a universally accepted definition of sepsis-associated atrial fibrillation, without which studies cannot be compared or pooled; validated prediction models to identify high-risk patients early in their illness; prospective randomized trials comparing rate control and rhythm control strategies; clearer, evidence-based anticoagulation strategies tailored to the phases of sepsis; and standardized rhythm monitoring after discharge to capture recurrence and long-term outcomes that are currently invisible to follow-up care. The authors also underscore the economic dimension of the problem: SAAF adds roughly 9,000 dollars in charges per patient, a substantial burden multiplied across the millions of sepsis admissions worldwide each year. Emerging technologies may help close the detection gap, and the review points to growing roles for artificial intelligence in predicting and detecting the arrhythmia in real time, potentially flagging high-risk patients before the rhythm deteriorates.

Until dedicated trials are completed and the research priorities are fulfilled, the authors conclude, the management of sepsis-associated atrial fibrillation will remain a matter of careful, individualized judgment. The central message for clinicians is to treat the whole critically ill patient rather than the arrhythmia in isolation: controlling the infection, correcting electrolyte disturbances, supporting hemodynamics, and reassessing rhythm and anticoagulation decisions continuously as the patient’s condition changes. For researchers, the message is equally direct. A condition that affects as many as one in five sepsis patients, multiplies stroke risk nearly fivefold, and drives measurable excess mortality and cost can no longer be managed on borrowed evidence from a different disease. The review’s roadmap—definition, prediction, randomized comparison of treatments, rational anticoagulation, and structured follow-up—offers the first coordinated plan for turning a clinical blind spot into standardized, evidence-based care.

Subject of Research: Sepsis-associated atrial fibrillation: incidence, risks, diagnosis, and management

Article Title: Atrial fibrillation during sepsis: what clinicians know—and what remains unclear

Article References: Atrial fibrillation during sepsis: what clinicians know—and what remains unclear. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: atrial fibrillation, sepsis, critical care, stroke risk, anticoagulation, rate control, rhythm control, intensive care medicine, arrhythmia, clinical review, artificial intelligence, research priorities

The study, led by Susana Sousa of CUF Tejo Hospital in Lisbon together with collaborators at Nox Medical in Iceland and the University of Porto, set out to test a cloud-based software called DeepRESP in a population where conventional home testing struggles most. The system had already earned FDA 510(k) clearance after validation on nearly 3,500 routine clinical recordings from sleep clinics across the United States, demonstrating non-inferiority or superiority to established predicate devices. But the atrial fibrillation population poses a unique challenge. Irregular heart rhythms and rate-controlling medications such as beta-blockers scramble the cardiac signals that many simplified sleep-staging algorithms depend on, potentially undermining their accuracy precisely in the patients who need reliable testing the most.

The researchers enrolled 88 consecutive patients with atrial fibrillation referred for sleep assessment at a cardiology outpatient unit, excluding only those with prior sleep apnea diagnoses or recent acute cardiac events. Two-thirds of the participants were men, and the women were significantly older than the men, averaging nearly 67 years compared with about 61. The cohort carried a heavy burden of comorbidity: half were obese, nearly half had hypertension, and roughly one in five had diabetes. Most telling of all, every single participant met criteria for sleep apnea, with an apnea-hypopnea index of at least five events per hour, and 80 percent had moderate to severe disease. Yet only about a quarter reported clinically significant daytime sleepiness, underscoring how easily the condition hides in this cardiac population.

Each patient underwent ambulatory level II polysomnography with a full complement of physiological channels, including six-lead electroencephalography, electrooculography, submental electromyography, nasal pressure airflow, thoracic and abdominal respiratory inductance plethysmography, oximetry, and body position sensing. Certified sleep physicians and technicians manually scored these recordings according to American Academy of Sleep Medicine criteria, establishing the gold-standard reference. DeepRESP then processed the same recordings, but crucially it was allowed to see only the subset of signals available in a home sleep apnea test: oximetry, nasal pressure airflow, and the respiratory inductance plethysmography belts. No electroencephalography, no eye movement channels, no muscle electrodes, and no human editing of the automated output.

The technical heart of the system is the Nox BodySleep 2.0 algorithm, which infers sleep states and arousals exclusively from the thoracic and abdominal breathing belts sampled at 25 hertz. The underlying physiology is elegant. Breathing is irregular and behaviorally influenced during wakefulness, becomes remarkably stable and metabolically regulated in non-REM sleep, and turns erratic again in REM sleep, when skeletal muscle atonia reshapes thoracoabdominal mechanics. These state-dependent signatures are imprinted on the plethysmography signals, and the neural network learns to decode them. Arousals add a further fingerprint: each one triggers a transient ventilatory response, a gasp-like perturbation in the breathing pattern that the algorithm can detect and quantify, enabling it to score hypopneas that end in arousal even when oxygen saturation barely dips.

The results were remarkable for a system flying blind without brain waves. At the epoch level, overall agreement with manual scoring reached 0.91 for wake, 0.95 for REM sleep, and 0.87 for non-REM sleep. Arousal detection achieved an overall percent agreement of 0.80, with a positive percent agreement of 0.72 and a negative percent agreement of 0.84. Respiratory event detection was even stronger, with overall agreement of 0.94 for apneas, 0.78 for hypopneas, and 0.81 for respiratory events combined. Bootstrapping with 10,000 iterations generated confidence intervals for every metric, and the estimates held tightly across resamples.

At the level of the clinical indices that actually drive treatment decisions, the concordance was equally persuasive. The apnea-hypopnea index, the single most important number in sleep medicine, showed an intraclass correlation coefficient of 0.92 against manual polysomnography, with Bland-Altman analysis revealing a mean bias of 6.36 events per hour and limits of agreement stretching from minus 6.83 to 19.56. That bias was driven almost entirely by the hypopnea index, which contributed a bias of 5.45 events per hour, while the apnea index remained highly stable at just 0.91 events per hour. Total sleep time correlated at 0.82 with a mean bias of about 18.5 minutes, and the arousal index reached an intraclass correlation of 0.83 with a low mean bias of 1.44, although individual variance widened at higher arousal frequencies.

Why does this matter beyond the sleep laboratory? Untreated sleep apnea is a recognized saboteur of atrial fibrillation management. It increases recurrence after cardioversion and catheter ablation, blunts the efficacy of antiarrhythmic drugs, and elevates cardiovascular mortality through intermittent hypoxemia, sympathetic activation, and structural cardiac remodeling. Yet conventional home sleep apnea tests, lacking electroencephalography, systematically underestimate disease in patients whose hypopneas terminate in arousals rather than significant desaturation, and in those whose recorded sleep duration is short. A recent evaluation of peripheral arterial tonometry-based home testing in atrial fibrillation patients found only slight to fair agreement with polysomnography for severity classification, with a tendency to overestimate disease. By contrast, the breathing-based approach sidesteps the confounding effects of arrhythmia and cardiac medication entirely, because respiratory effort signals are robust to the electrical chaos of a fibrillating heart.

The authors are candid about the caveats. The cohort was modest in size and drawn from a single tertiary cardiology center, raising the possibility of referral bias. Although recordings were ambulatory, they were acquired as level II polysomnography under technologist supervision, so respiratory belt quality may have exceeded what unattended home testing typically delivers, where sensor displacement and suboptimal positioning are common. The arousal index showed wider limits of agreement than the other parameters, suggesting that breathing-derived arousal estimates, while clinically informative, may not fully replace electroencephalography for fine-grained characterization of sleep fragmentation. Comorbidities such as heart failure, Cheyne-Stokes respiration, and hypoventilation syndromes can produce breathing patterns that mimic obstructive events, and the system has not been assessed in patients already using respiratory support. Nor does DeepRESP score respiratory effort-related arousals.

Even with these limitations, the study marks a genuine inflection point for precision medicine in sleep diagnostics. It challenges the long-standing assumption that neurophysiological channels are indispensable for reliable sleep architecture assessment, at least in the populations where home testing is needed most. A scalable, electroencephalography-independent pathway that reproduces gold-standard indices for apnea severity, total sleep time, and arousal burden could transform screening for the vast, underdiagnosed population of atrial fibrillation patients, many of whom are minimally symptomatic and would otherwise never reach a sleep laboratory. The authors call for validation in fully unattended home recordings and for studies of long-term impact on clinical management and cardiovascular outcomes. If those efforts succeed, the humble breathing belt, read by a neural network, may become the stethoscope of twenty-first-century sleep medicine.

Subject of Research: Validation of a deep learning system for EEG-independent sleep staging, arousal detection, and respiratory event scoring in atrial fibrillation patients

Article Title: Validation of a deep learning-based system for sleep staging, arousal detection, and respiratory event scoring in atrial fibrillation patients

Article References: Sousa, S., Teixeira, C., Sigmarsdóttir, T., Finnsson, E., Ágústsson, J., Sigþórsson, S., Bjarkason, S., Drummond, M., & Bugalho, A. (2026). Validation of a deep learning-based system for sleep staging, arousal detection, and respiratory event scoring in atrial fibrillation patients. Journal of Clinical Sleep Medicine, 22(1), Article 130. https://doi.org/10.1007/s44470-026-00145-0

Image Credits: AI Generated

DOI: 10.1007/s44470-026-00145-0

Keywords: deep learning, atrial fibrillation, obstructive sleep apnea, home sleep apnea testing, polysomnography, respiratory inductance plethysmography, sleep staging, arousal detection, apnea-hypopnea index, machine learning, sleep-disordered breathing, clinical validation

The research, published in eClinicalMedicine, drew on the Clinical Practice Research Datalink Aurum database, an anonymized repository of medical records covering more than 50 million patients registered with over 1800 general practices across England. The investigators linked these primary care records to Hospital Episode Statistics, which capture every admission to publicly funded hospitals in the country, allowing them to track both routine diagnoses and serious exacerbations of lung disease. From this vast data resource, they assembled a cohort of 444,668 adults aged 50 and over with a physician diagnosis of COPD, who together contributed nearly 2.3 million person-years of follow-up. Within this group, 37,645 people, or 8.5 percent, were newly diagnosed with atrial fibrillation during the study window.

The scale of the comorbidity problem is difficult to overstate. COPD affects more than 390 million people worldwide and is the third leading cause of death globally, while atrial fibrillation touches over 52 million lives. Previous meta-analyses have estimated that patients with COPD face roughly twice the risk of developing atrial fibrillation compared with those without the lung disease, and that risk climbs as lung function deteriorates. The new study reinforces this dose-response relationship with unusual granularity. Incidence rates rose consistently with age, were higher in men than women across every age band, and were markedly elevated among patients with more severe COPD, whether severity was measured by forced expiratory volume in one second, breathlessness scores, or recent exacerbations requiring hospitalization. Patients hospitalized for a severe COPD flare in the previous year showed incidence rates above 55 per 1000 person-years by 2022, several times higher than those with well-preserved lung function.

The temporal pattern also tells a story about modern healthcare. After a gradual year-on-year climb, incidence plateaued between 2016 and 2019, then dropped 22 percent in 2020 compared with 2019, a dip the authors attribute to reduced access to primary care during the COVID-19 pandemic. The rebound afterward suggests the underlying trend was never interrupted, merely masked. The researchers note that the rising trajectory mirrors patterns seen in the general UK population, implying that improved detection and better survival among people with atrial fibrillation risk factors, rather than something unique to COPD, may be driving much of the increase. Still, the consistency of the elevation across every marker of COPD severity points to a genuine biological contribution, likely involving chronic inflammation, hypoxia, and the structural cardiac changes that accompany long-standing lung disease.

Why does this matter so much? Because atrial fibrillation in a COPD patient is more dangerous than atrial fibrillation alone. Recent cohort evidence indicates that people with both conditions face approximately 1.6 times the risk of ischemic stroke compared with those who have atrial fibrillation without COPD. Oral anticoagulants are the cornerstone of stroke prevention in atrial fibrillation, and current guidelines recommend that treatment decisions be made independently of COPD status. Yet the real-world evidence on whether this actually happens has been sparse and contradictory, with some studies suggesting higher anticoagulation rates in COPD patients and others suggesting lower or equivalent rates.

The new findings go further than any previous work by tracking prescribing trends over thirteen years. Among the 34,851 patients newly diagnosed with atrial fibrillation during follow-up, 66.3 percent received an oral anticoagulant within a year of diagnosis. The rate of initiation nearly tripled over the study period, from 10.3 prescriptions per 100 person-months in 2010 to 30.1 per 100 person-months in 2022, a 2.9-fold increase. The most dramatic rise occurred among patients aged 90 and older, whose initiation rate was more than nine times higher in 2022 than in 2010, albeit from a very low baseline. This transformation was driven almost entirely by the direct oral anticoagulants, or DOACs, which received European approval for stroke prevention in 2011. Warfarin and other vitamin K antagonists, once the only option, collapsed to a 96 percent lower initiation rate by 2022, while DOAC use surged 64-fold between 2012 and 2022.

The fine-grained prescribing data reveal how completely the therapeutic landscape has changed. Apixaban emerged as the dominant agent, peaking at 208.1 prescriptions per 1000 person-months in 2021 and accounting for 61.6 percent of all oral anticoagulants started between 2020 and 2022. Edoxaban, introduced in 2016, overtook rivaroxaban by 2020 and then jumped from 49.4 to 95.4 prescriptions per 1000 person-months between 2021 and 2022. Rivaroxaban, an early favorite, peaked in 2016 and declined steadily thereafter. Dabigatran, the first DOAC approved, never gained significant traction. Notably, these patterns closely match prescribing trends reported in atrial fibrillation patients generally, suggesting that the presence of COPD did not materially alter which drug clinicians chose, only, in some cases, whether they prescribed one at all.

And that is where the treatment gap emerges. Despite the overall tripling of anticoagulation, several high-risk groups were persistently less likely to receive treatment. Patients aged 90 and older had the lowest initiation rates of any age group throughout the study, even though stroke risk climbs steeply with age and guidelines recommend anticoagulation for most patients with atrial fibrillation over 75. Non-White patients were prescribed anticoagulants less often than White patients, echoing documented disparities in atrial fibrillation care more broadly. Most strikingly, people with the most severe COPD, identified by low percent-predicted lung function, severe breathlessness, or recent hospitalization for an exacerbation, consistently received fewer prescriptions than those with milder disease, a pattern that held across all years, ages, and sexes. Patients with chronic kidney disease, liver disease, a history of bleeding, cancer, or dementia were also less likely to be treated.

The authors caution that these patterns may not reflect inappropriate care alone. Contraindications, frailty, polypharmacy, and patient preferences all legitimately shape prescribing decisions, and clinicians may reasonably hesitate before adding a blood thinner to a regimen already burdened by multiple respiratory medications. Yet the concern about physician hesitancy is hard to dismiss, because evidence on the safety and efficacy of anticoagulants specifically in patients with coexisting COPD and atrial fibrillation remains thin. Without dedicated trials or robust observational outcome data in this population, uncertainty itself becomes a barrier to treatment, and the patients facing the highest thromboembolic risk are left in a zone of clinical ambiguity.

The study has limitations worth noting. The database does not capture prescriptions issued by specialists or during hospitalizations, which may lead to modest underestimation of prescribing rates, though the central role of general practitioners in the UK system likely mitigates this. Restricting the cohort to adults over 50 reduced but did not eliminate the possibility of misclassifying some asthma patients as having COPD. The ethnicity-stratified analyses aggregated small numbers of non-White patients into a single category, limiting insight into specific minority subgroups, and the cohort included only individuals with COPD, so comparisons with the general population are indirect. Even so, the sheer size of the cohort, the representativeness of the data, and the consistency of results across sensitivity analyses using multiple imputation lend considerable weight to the conclusions. The message for clinicians and health systems is clear: as atrial fibrillation becomes increasingly common among people with COPD, stroke prevention must keep pace, and the oldest, sickest, and most ethnically marginalized patients, precisely those at greatest risk, should not be left behind. Integrated care pathways and routine bidirectional screening for both conditions, identified as research priorities by the American Thoracic Society, may offer a route forward, but the first step is recognizing that a growing treatment gap exists at all.

Subject of Research: Temporal trends in atrial fibrillation incidence and oral anticoagulant prescribing among adults with COPD in UK primary care

Article Title: Trends in the incidence of atrial fibrillation and the prescription of oral anticoagulants in adults with chronic obstructive pulmonary disease in UK primary care (2010–2022): a cohort study

Article References: Trends in the incidence of atrial fibrillation and the prescription of oral anticoagulants in adults with chronic obstructive pulmonary disease in UK primary care (2010–2022): a cohort study. (n.d.). https://doi.org/10.1016/j.eclinm.2026.104245

Image Credits: AI Generated

DOI: 10.1016/j.eclinm.2026.104245

Keywords: atrial fibrillation, COPD, oral anticoagulants, DOACs, stroke prevention, UK primary care, CPRD, cohort study, health disparities, warfarin, apixaban, epidemiology

The study, published in BMC Geriatrics by a team of hospital pharmacists and pharmacoepidemiologists affiliated with Assistance Publique–Hôpitaux de Paris and Paris-Saclay University, set out to measure how often DOAC prescriptions in hospitalized older adults deviated from European Society of Cardiology (ESC) guidelines, and whether the situation had improved since a comparable assessment in 2018. The answer, in short, is that it had not.

Between November 2022 and May 2023, the researchers consecutively enrolled older adults who were already taking a DOAC before admission and who stayed in one of three French university hospitals for more than 24 hours. The participating departments spanned cardiology, internal medicine, post-emergency care, and acute geriatrics, giving a broad picture of real-world prescribing across the acute care pathway. In total, 536 patients were eligible, with a mean age of 84.1 years and a slight majority of women; 421 patients formed the primary analysis cohort.

The pharmacological profile of the cohort reflected current practice. Apixaban dominated, accounting for 70.1 percent of prescriptions, and atrial fibrillation was the indication in 83.6 percent of cases. This matters because apixaban, like all DOACs, has a narrow therapeutic window whose appropriateness hinges on patient-specific factors: renal function, age, body weight, and the indication being treated. Unlike warfarin, which is monitored with routine blood tests, DOAC dosing is fixed at prescription time, so an error at the moment of prescribing persists silently until harm occurs.

Using ESC guidelines as the benchmark, the team classified prescriptions as appropriate or inappropriate at two critical moments: hospital admission and hospital discharge. On admission, 32.0 percent of prescriptions were judged inappropriate. By discharge, the figure had fallen to 24.9 percent, a statistically significant improvement (p = 0.004), suggesting that the hospital stay does correct some errors, but leaves a substantial fraction of patients heading home with potentially unsafe anticoagulation.

The consequences of such errors are not abstract. An excessive dose in a frail 85-year-old with declining kidney function increases the risk of major bleeding, including intracranial hemorrhage. An insufficient dose, conversely, can mean therapeutic failure, allowing a stroke or systemic embolism to occur despite the patient dutifully taking their medication. In older patients with polypharmacy, where DOACs interact with other drugs and renal clearance is often reduced, the margin for error shrinks further, which is why inappropriate prescribing is considered a major patient safety concern in geriatric care.

To understand who was most at risk, the researchers used logistic regression to identify predictive factors for inappropriate prescribing. Two variables stood out at admission: age over 80 years nearly doubled the odds of an inappropriate prescription (odds ratio 1.80, 95 percent confidence interval 1.20–2.72), and elevated serum creatinine above 133 µmol/L, a marker of impaired kidney function, similarly raised the odds (OR 1.84, 95 percent CI 1.13–2.98). At discharge, age over 80 remained a strong risk factor (OR 2.71, 95 percent CI 1.63–4.61), while a body weight below 60 kilograms was unexpectedly protective (OR 0.42, 95 percent CI 0.25–0.69), possibly because clinicians scrutinize low-weight patients more carefully when selecting doses.

The renal connection deserves particular emphasis. DOACs are cleared substantially by the kidneys, and estimates of glomerular filtration rate in older adults vary depending on the formula used, whether Cockcroft–Gault, MDRD, or CKD-EPI. The study’s authors assessed prescriptions against the Summary of Product Characteristics for each molecule, which specify dose reductions based on renal function, age, and weight. When kidney function is overestimated, or when a hospitalization changes a patient’s renal status without the prescription being adjusted, the fixed daily dose can quickly become dangerous. Advanced age and impaired renal function, the two risk factors identified here, are precisely the conditions that make this pharmacokinetic tightrope hardest to walk.

Perhaps the most sobering finding is the comparison with 2018. Using chi-squared tests, the team compared their results with data collected five years earlier and found no significant changes. Despite growing awareness of DOAC dosing pitfalls, updated guidelines, and the expansion of clinical pharmacy services in French hospitals, the rate of inappropriate prescribing in this vulnerable population has remained essentially flat. The authors conclude that further efforts to optimize DOAC prescribing are warranted, and they point toward concrete interventions: pharmacist-led medication reviews at admission and discharge, and integrated prescribing tools embedded in electronic medical records that would automatically flag doses inconsistent with renal function, age, and weight.

The study has the strengths and limits of its design. As a prospective observational study, it captures real prescribing behavior rather than idealized trial conditions, and its multicenter scope across four department types strengthens generalizability within the French hospital system. But it also reflects a single country’s practice in a single year, and appropriateness was defined by ESC guidelines rather than patient-specific outcomes, so the findings measure guideline concordance rather than directly measured harm. Even so, the message is clear and actionable: the transition points of hospital care, admission and discharge, remain moments where older patients on DOACs are at elevated risk of a prescribing error, and after years of stagnation, systematic medication reconciliation and decision support may be the tools that finally bend the curve.

Subject of Research: Appropriateness of direct oral anticoagulant prescribing in hospitalized older adults

Article Title: Appropriateness prescriptions of direct oral anticoagulants in hospital in 2023 in the older population: a prospective observational multicenter study

Article References: Rodier, T., Gibert, A., Lepors, A., Fernandez, C., Hindlet, P., Savoldelli, V., & Schwab, C. (2026). Appropriateness prescriptions of direct oral anticoagulants in hospital in 2023 in the older population: a prospective observational multicenter study. BMC Geriatrics. https://doi.org/10.1186/s12877-026-08330-9

Image Credits: AI Generated

DOI: 10.1186/s12877-026-08330-9

Keywords: DOACs, anticoagulants, older adults, prescribing appropriateness, medication safety, atrial fibrillation, apixaban, renal function, polypharmacy, hospital discharge, pharmacist-led medication review, geriatrics

Atrial fibrillation, the most common sustained cardiac arrhythmia, becomes increasingly prevalent with age, and contemporary patients are typically older, frailer and burdened by multiple chronic conditions. This complexity raises the risk not only of stroke and bleeding but also of non-cardiovascular events and hospital admissions, driving substantial healthcare costs. The ABC pathway was proposed as a structured framework to streamline integrated management: ‘A’ for anticoagulation and stroke risk avoidance, ‘B’ for better symptom control, and ‘C’ for cardiovascular and comorbidity optimisation. Observational studies repeatedly linked adherence to the pathway with lower mortality, stroke and bleeding, and two cluster-randomised trials in Asia—one app-based in largely urban China and one delivered by village doctors with telehealth support in rural settings—had shown that ABC implementation improved guideline adherence and reduced adverse outcomes, including hospitalisations.

What remained unknown was whether the same approach would work in European healthcare systems, specifically among patients aged 65 or older with atrial fibrillation and at least one additional chronic condition. The AFFIRMO investigators, led by Gregory Y.H. Lip and Marco Proietti, designed a multicentre, open-label cluster-randomised trial across Bulgaria, Denmark, Italy, Romania, Serbia and Spain. Clinical centres, not individual patients, were randomised to deliver either the integrated mHealth intervention—dubbed the iABC system—or routine guideline-based care. The intervention combined a patient-facing mobile application with a clinician dashboard, and every intervention patient underwent a comprehensive geriatric assessment at baseline to identify functional domains requiring management.

The AFFIRMO Mobile App collected daily data on vital signs, oral anticoagulant adherence, arrhythmia symptoms and chronic conditions, while dispensing tailored health tips for lifestyle and comorbidity management. Each patient also completed the Patient Health Engagement scale at activation, which determined the level of personalised educational material they received about atrial fibrillation. On the clinical side, the AFFIRMO Clinician Dashboard summarised app usage and patient-entered data, organised by the three ABC pillars, to inform follow-up consultations. Between April 2024 and January 2025, the trial enrolled 1,260 patients, with 634 assigned to the iABC system and 626 to usual care. Follow-up visits occurred at 3, 6 and 12 months, and the primary endpoint was unplanned all-cause hospitalisation during one year of follow-up, adjudicated by a blinded event validation committee.

The enrolled population reflected the modern reality of atrial fibrillation care: a mean age of 73 to 74 years, a median CHA2DS2-VASc stroke risk score of 4, and a heavy burden of comorbidity. Roughly 40 percent had more than two chronic conditions, hypertension treated with multiple drugs was nearly universal, and polypharmacy affected 72 percent of participants. Yet the trial’s most consequential baseline finding was how well these patients were already being treated: 97.3 percent were on oral anticoagulation, overwhelmingly direct oral anticoagulants, and use of beta-blockers, anti-arrhythmic drugs and other guideline-directed therapies was uniformly high. Comprehensive geriatric assessment in the intervention arm revealed a largely functionally preserved cohort, with 95.7 percent showing normal cognition, 83.9 percent normal nutritional status, and only about 20 percent reporting meaningful mobility impairment.

After a mean follow-up of 367 days, unplanned all-cause hospitalisation occurred in 17.1 percent of the iABC group versus 18.2 percent of the usual care group—an adjusted odds ratio of 0.95 with a 95 percent confidence interval of 0.61 to 1.49 and a p-value of 0.84. Sensitivity analyses, including a Cox proportional hazards model and a composite endpoint of hospitalisation or death, told the same neutral story. No secondary endpoint differed between groups: all-cause death, any hospitalisation, stroke or cardiovascular death, heart failure events, renal worsening, and major bleeding, which occurred in only 0.7 percent of all participants, were statistically indistinguishable across arms. Subgroup analyses by age, sex, comorbidity count, medication number and country revealed no hidden pockets of benefit.

The investigators attribute the null result to a convergence of factors, each instructive in its own right. First, the trial’s power calculation assumed a 30 percent event rate in usual care and a 25 percent relative reduction with the intervention; instead, usual care produced only an 18.2 percent hospitalisation rate. With such low residual risk, several thousand patients would have been required to detect a statistically significant difference. Second, the baseline quality of care starkly contrasts with the prior Asian trials: in the mAFA trial, oral anticoagulant use at baseline was just 48.4 percent in usual care, and in the MIRACLE-AF rural trial it was around 11 percent. In those settings, the ABC intervention raised anticoagulation dramatically and reduced hospitalisations. In AFFIRMO, with anticoagulation already at 97 percent, there was simply little therapeutic ground left for a digital nudge to reclaim.

Third, and perhaps most telling, patients barely used the app. The median percentage of days on which patients accessed the AFFIRMO Mobile App was 15.1 percent, and 68.5 percent of patients fell into the lowest tertile of use, opening the app on only about one-third or fewer of their study days. The authors point to well-documented implementation barriers—digital literacy in an older population, workflow incompatibility, and difficulty integrating digital tools into complex health systems. Clinical decision support, they note, only works when it is actually used; in the O’CAFÉ trial, modest overall effects sharpened into significant anticoagulation improvements only among clinicians who actively engaged with the tool. Because the mobile app is half of the iABC system, its companion physician dashboard inheriting the consequences of underuse, suboptimal engagement directly undermined the intervention’s theoretical effectiveness.

Fourth, comprehensive geriatric assessment itself ran into a ceiling effect. Although more than half of the intervention patients showed some degree of frailty on the FRAIL scale, most were robust on objective measures of daily functioning, cognition, nutrition and mood. In such high-functioning individuals, geriatric assessment may serve a descriptive rather than an interventional role, and prior evidence suggests little clinical benefit from CGA in robust patients. The trial’s design assumed more functional impairment than it found, constraining the scope for assessment-driven management changes. Not all signals were negative, however: patients in the intervention arm showed numerically higher uptake of several guideline-directed treatments at the final visit, including mineralocorticoid receptor antagonists, SGLT2 inhibitors and anti-arrhythmic drugs, and a significantly lower rate of uncontrolled systolic blood pressure at 12 months (1.5 versus 4.0 percent), hinting at secondary improvements in care quality even without hard outcome benefits.

The broader lesson echoes the recent STEEER-AF trial, in which an electronic education programme for clinicians also yielded only marginal improvements because guideline adherence was already high in both arms. The authors emphasise that the results do not disqualify integrated care or the ABC pathway, which remains embedded in European Society of Cardiology guidance through the 2020 ABC framework and the 2024 AF-CARE scheme. Rather, they argue, integrated care interventions and digital health tools must be targeted at settings with genuine unmet clinical need—populations with low baseline guideline adherence, higher residual risk, or greater functional impairment. Future studies may also need easier-to-use apps, co-designed with older patients, and more selected populations in which the theoretical benefit of holistic optimisation has room to translate into fewer hospitalisations. For now, AFFIRMO stands as a rigorous, well-conducted demonstration that in medicine, as in engineering, the marginal return on optimisation shrinks as the system approaches its ceiling—and that digital tools, however elegantly engineered, cannot multiply benefit that better baseline care has already claimed.

Subject of Research: A cluster-randomised trial of mHealth-based integrated care and comprehensive geriatric assessment in older multimorbid patients with atrial fibrillation.

Article Title: Integrated care management and comprehensive geriatric assessment using a mHealth-based approach in older multimorbid patients with atrial fibrillation: the AFFIRMO cluster-randomised trial

Article References: Lip, G. Y., Proietti, M., Ainsworth, J., Dan, G.-A., Frost, L., Graffigna, G., Lane, D. A., Lucci, D., Fabbri, G., Marin, F., O'Flaherty, M., Petrovic, M., Potpara, T. S., Proietti, R., Sanaullah, A., Tokmakova, M., Vetrano, D. L., Johnsen, S. P., Maggioni, A. P., … Tokmakova, M. (2026). Integrated care management and comprehensive geriatric assessment using a mHealth-based approach in older multimorbid patients with atrial fibrillation: the AFFIRMO cluster-randomised trial. The Lancet Regional Health – Europe, 70, Article 101832. https://doi.org/10.1016/j.lanepe.2026.101832

Image Credits: AI Generated

DOI: 10.1016/j.lanepe.2026.101832

Keywords: atrial fibrillation, AFFIRMO trial, mHealth, integrated care, ABC pathway, comprehensive geriatric assessment, multimorbidity, cluster-randomised trial, unplanned hospitalisation, digital health, older adults, guideline adherence