The research team, which included investigators from Erasmus MC in Rotterdam, the Institute of Agrochemistry and Food Technology in Valencia, and Helmholtz Munich, among other institutions, set out to answer a deceptively simple question: which exposures occurring before conception, during pregnancy and birth, and in infancy up to two years of age are associated with cardiometabolic outcomes in later childhood and adolescence? The outcomes they examined were the five canonical precursors of adult cardiovascular disease: glucose intolerance, dyslipidemia, hypertension, metabolic syndrome, and type 2 diabetes. The review was prospectively registered with PROSPERO under registration number CRD42022355152, and the authors followed PRISMA reporting guidelines to ensure methodological transparency. The 70 studies that met the inclusion criteria comprised 65 longitudinal observational studies, four case-control studies, and one randomized controlled trial, a distribution that reflects the ethical and practical constraints of studying developmental exposures in human populations.

What distinguishes this meta-analysis from previous efforts is its rigorous handling of statistical dependency. Because many large birth cohorts contribute multiple overlapping publications, naive pooling can artificially inflate the precision of an estimate by counting the same children more than once. The authors addressed this by revising their meta-analyses to retain only one independent estimate per source cohort, a correction that materially changed the precision of the pooled results. The gestational diabetes finding survived this scrutiny with remarkably low heterogeneity, an I-squared value of just 12.7 percent, indicating that the underlying studies were unusually consistent in the direction and magnitude of the association. The team also classified any meta-analysis based on fewer than five independent studies as exploratory, a conservative decision that separates robust conclusions from preliminary signals and provides a template for how future reviews in this field should handle sparse evidence.

Gestational diabetes mellitus emerged as the most consequential early-life exposure in the entire analysis. It was positively associated with every cardiometabolic risk outcome examined, with glucose intolerance being the most frequently reported. This finding carries particular biological plausibility. During a pregnancy complicated by gestational diabetes, the fetus is exposed to elevated maternal glucose, which crosses the placenta and stimulates fetal insulin production. Chronic hyperinsulinemia in the developing fetus can alter pancreatic beta-cell function, adipose tissue development, and hypothalamic programming of appetite and metabolism. Animal models have shown that such developmental exposures can permanently reprogram metabolic homeostasis, a phenomenon known as metabolic imprinting or fetal programming. The human evidence now assembled in this meta-analysis suggests that the same mechanisms operate in real-world populations, producing measurable cardiometabolic consequences that persist into adolescence.

Maternal body mass index during pregnancy was another exposure with broad reach across the cardiometabolic spectrum. Higher maternal pregnancy BMI was positively associated with glucose intolerance, dyslipidemia, hypertension, and metabolic syndrome in offspring. The biological pathways here are multifaceted. Maternal obesity is accompanied by elevated circulating free fatty acids, inflammatory cytokines, and leptin, all of which can cross the placenta and influence fetal development. Maternal insulin resistance, a hallmark of obesity, can also alter nutrient delivery to the fetus and modify the intrauterine metabolic environment. The consistency of the association across four distinct cardiometabolic outcomes suggests that maternal obesity acts through a general developmental perturbation rather than through a single organ-specific pathway, a conclusion that aligns with the developmental origins of health and disease framework that has gained prominence over the past three decades.

The analysis also examined birth characteristics and found that being born small for gestational age was associated with higher odds of hypertension in later childhood and adolescence, though this finding was classified as exploratory because it was based on fewer than five independent studies. The corresponding exploratory analysis for preterm birth and elevated systolic blood pressure did not reach statistical significance and is presented for methodological illustration only. The authors also found that low or high birth weight, and being born large for gestational age, were associated with cardiometabolic outcomes in childhood and adolescence. The link between restricted fetal growth and later hypertension is biologically coherent. Fetuses that experience intrauterine growth restriction often develop a preferential blood flow pattern that shunts oxygen and nutrients toward the brain at the expense of other organs, including the kidneys. Reduced nephron number in the developing kidney is a well-established mechanism for later hypertension, because a smaller nephron complement increases sodium retention and raises blood pressure throughout life.

Infant BMI and weight changes during the first two years of life were positively associated with dyslipidemia, hypertension, and metabolic syndrome in later childhood and adolescence. This finding is notable because it extends the window of cardiometabolic vulnerability beyond the intrauterine period and into postnatal development. Rapid weight gain in infancy, sometimes called catch-up growth, has been proposed as a mechanism that can alter body composition, insulin sensitivity, and lipid metabolism. The developmental logic is that infants who experience early growth restriction and then rapid postnatal weight gain may accumulate visceral adipose tissue preferentially, a pattern that is strongly associated with insulin resistance and dyslipidemia. The finding that infant BMI and weight trajectories are linked to three distinct cardiometabolic outcomes suggests that the first two years of life represent a critical period for metabolic programming, one that may be as consequential as the intrauterine environment itself.

The analysis also examined breastfeeding and found limited evidence suggesting a possible inverse association between breastfeeding and cardiometabolic risk in childhood and adolescence, though the authors characterized this evidence as limited and inconclusive because it was based on few studies per outcome. This finding is important because breastfeeding is a modifiable exposure with enormous public health implications. Human milk contains bioactive compounds, including long-chain polyunsaturated fatty acids, hormones, and growth factors, that could theoretically influence metabolic development. Breastfed infants also tend to have different growth trajectories than formula-fed infants, with slower early weight gain that may reduce the risk of later adiposity. However, the observational nature of the evidence means that confounding by socioeconomic status, maternal education, and other factors cannot be excluded, and the authors were appropriately cautious in their interpretation, noting that the evidence was limited and inconclusive rather than definitive.

The authors emphasize that their findings may support closer cardiometabolic monitoring of children with relevant early-life exposures and may inform future prevention research. This is a significant shift in clinical thinking. Traditionally, cardiometabolic risk assessment has been reserved for adults with established risk factors, but this meta-analysis suggests that children and adolescents with specific developmental exposures, such as maternal gestational diabetes, maternal obesity during pregnancy, or being born small or large for gestational age, may warrant earlier surveillance. The identification of modifiable early-life risk factors also opens the door to targeted interventions during pregnancy and infancy that could reduce the burden of cardiometabolic disease in future generations. The research was conducted by an expert group of the European branch of the International Life Sciences Institute and coordinated by the Early Nutrition and Long-term Health Task Force, with the scientific work carried out solely by the authors and ILSI Europe having no role in the definition of research questions, study design, data collection, statistical analysis, interpretation of results, or approval of the submitted manuscript.

The implications of this meta-analysis extend far beyond the clinic and into public health policy. With gestational diabetes affecting a growing proportion of pregnancies worldwide, and maternal obesity reaching epidemic levels, the developmental exposures identified in this study are becoming increasingly common. The finding that these exposures are associated with cardiometabolic outcomes in childhood and adolescence, rather than only in adulthood, suggests that the window for prevention may need to shift earlier. The authors’ rigorous methodology, including the correction for overlapping cohorts and the conservative classification of sparse evidence, provides a model for how future reviews in this field should be conducted. As the global burden of type 2 diabetes and cardiovascular disease continues to rise, understanding the developmental origins of these conditions has never been more urgent, and this meta-analysis represents a major step forward in that effort, offering a comprehensive map of the early-life risk factors that shape cardiometabolic health across the lifespan.

Subject of Research: Early-life risk factors and cardiometabolic outcomes in children and adolescents: a systematic review and meta-analysis

Article Title: Early-life risk factors and cardiometabolic outcomes in children and adolescents: a systematic review and meta-analysis

Article References: Brandimonte-Hernández, M., Ruiz-Ojeda, F. J., Blaauwendraad, S. M., Kamphuis, A. S., Flores-Ventura, E., Abrahamse-Berkeveld, M., Collado, M. C., van Diepen, J. A., Iozzo, P., Knipping, K., van Loo-Bouwman, C. A., Gaillard, R., & Gil, Á. (2026). Early-life risk factors and cardiometabolic outcomes in children and adolescents: a systematic review and meta-analysis. BMC Medicine. https://doi.org/10.1186/s12916-026-05316-7

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05316-7

Keywords: gestational diabetes, glucose intolerance, cardiometabolic risk, early-life risk factors, meta-analysis, BMC Medicine, pediatrics, developmental origins, maternal BMI, hypertension, Early-life, risk

The study, led by Luca Ünlü of the University Hospital Basel and University Medical Center Freiburg, together with Christian H. Nickel and a transdisciplinary steering committee spanning emergency medicine and psychiatry on four continents, used a modified Delphi process—a structured method for building consensus among experts who rarely see eye to eye. Between May and November 2025, the team ran four rounds of online surveys and steering committee discussions. The participants were not a random sample of clinicians: they were 25 official representatives from seven emergency medicine and psychiatry societies, including the European Society for Emergency Medicine, the American Association of Emergency Psychiatry, the Australasian College for Emergency Medicine, the Brazilian Association for Emergency Medicine, the Dutch Acute Medicine Society, the German Association for Psychiatry, Psychotherapy and Psychosomatics, and the Coalition on Psychiatric Emergencies. Crucially, the panel also included six patient representatives with prior experience of emergency department medical screening, either as patients or as relatives of patients, recruited with help from the National Alliance on Mental Illness and the Italian advocacy network E.N.E.A.

The technical logic of the Delphi method is what makes this result notable. In the first round, participants freely generated candidate screening elements: 91 statements came from the society representatives and 27 from the patient representatives. In the second round, the panel rated and refined these statements, merging overlapping items and sharpening ambiguous wording. The third round applied a predefined consensus threshold: a statement would only survive if at least 80 percent of participants agreed with it. That threshold proved demanding. Only 48 of the 91 society statements—52.8 percent—reached consensus, while 21 of the 27 patient statements—77.8 percent—cleared the bar. The surviving items were then synthesized into a single instrument, the Delphi-Informed Medical Evaluation, or DIME, tool.

Retention rates across the rounds were high for a Delphi study of this scale. Of 28 invited society representatives and steering committee members, 25—89.3 percent—participated in the first round, 23 in the second, 22 in the third, and 23 in the fourth. All six patient representatives completed every round, a detail the authors highlight as evidence that the patient perspective was not a token gesture but a sustained contribution throughout the process. The study was registered with ClinicalTrials.gov under identifier NCT06936826, and the consensus process followed the ACCORD—ACcurate COnsensus Reporting Document—guidance, a framework designed to make consensus methods transparent and reproducible rather than a black box of expert opinion.

The fourth round served as the tool’s first stress test. Rather than asking participants whether they liked the DIME tool in the abstract, the researchers confronted them with 37 standardized clinical vignettes—short, structured descriptions of hypothetical emergency department patients with psychiatric complaints. Each participant independently applied the draft tool to every vignette and decided which screening elements it called for. The researchers then measured how often the raters agreed with one another using Gwet’s agreement coefficient, a statistical measure of interrater agreement that is more robust than simple percentage agreement when agreement is high, because it corrects for agreement that would occur by chance alone.

The numbers tell a striking story. Among the society representatives, applying the DIME tool to the 37 vignettes produced a Gwet agreement coefficient of 0.86, with a 95 percent confidence interval of 0.78 to 0.93 and a p-value below 0.001—a level the authors characterize as almost perfect agreement. Among the patient representatives, the coefficient was 0.76, with a confidence interval of 0.66 to 0.86 and the same p-value below 0.001, corresponding to substantial agreement. In practical terms, this means that when clinicians and patients from different countries, cultures, and professional traditions applied the same tool to the same hypothetical patient, they overwhelmingly reached the same conclusions about what medical screening was warranted. For a process that has historically generated more conflict than consensus, that level of agreement is remarkable.

The background problem the DIME tool addresses is well documented in the emergency medicine literature. Medical screening in the emergency department is performed to identify underlying medical conditions that may contribute to a patient’s psychiatric presentation or require treatment in their own right—think of delirium caused by an infection, agitation driven by intoxication, or psychosis secondary to a brain lesion. But the objectives and appropriate scope of this screening have never been standardized. Emergency physicians and psychiatrists frequently disagree about which tests are necessary, and the consequences are measurable: potentially unnecessary testing, prolonged emergency department length of stay, and increased costs. Patients with psychiatric complaints often wait longest, undergo the most redundant workups, and receive the least consistent evaluation. The DIME tool is designed to identify adult emergency department patients with psychiatric complaints who are at low risk for acute medical illness—effectively giving both specialties a shared, explicit framework for deciding when enough is enough.

The inclusion of patient voices is one of the study’s most distinctive technical features. Patient and public involvement has become an increasingly formalized element of health research, with reporting frameworks such as GRIPP2-SF—the Guidance for Reporting Involvement of Patients and the Public 2 Short Form—used to document how lay perspectives shaped the work. In this study, the patient representatives did not merely comment on a finished product; they generated their own statements in the first round, and those statements survived the 80 percent consensus threshold at a higher rate than the professional statements did. The synthesis of both streams of statements into the final tool means the DIME instrument reflects what patients consider important about being medically screened, not only what specialists consider defensible. The authors note that the patient perspective was incorporated with assistance from the National Alliance on Mental Illness and E.N.E.A., an organization of experts by experience based in Rome.

The authorship itself illustrates the transdisciplinary ambition. The panel includes figures such as Christopher R. Carpenter of the Mayo Clinic, Philipp Sterzer of the University of Basel’s Department of Psychiatry, Judith Tintinalli of the University of North Carolina, a towering name in emergency medicine education, and Michael P. Wilson of Virginia Tech-Carilion, whose work spans both emergency medicine and psychiatry. The DIME study group extends the effort to clinicians in New Zealand, Brazil, Australia, Spain, France, the Netherlands, the United Kingdom, and the United States. The work was funded by scientific funds of the Emergency Department of the University Hospital Basel, with open access funding provided by the University of Basel, and the article is published under a Creative Commons Attribution 4.0 license, making the tool freely available to any department that wants to examine it.

The authors are careful about what this study does and does not establish. The DIME tool is a consensus-based instrument, evidence-informed but not yet validated: it has not been tested against real patients, real outcomes, or real emergency department workflows. The researchers state explicitly that prospective multicenter validation studies are required before clinical implementation. That caveat matters, because a tool that performs well on paper vignettes among motivated expert raters may behave differently in a crowded department at three in the morning, applied by a tired resident to a frightened and agitated patient. Still, the study delivers something the field has lacked: a single, internationally agreed framework, endorsed by emergency physicians, psychiatrists, and patients on four continents, for deciding which adult psychiatric emergency patients need what medical evaluation. If prospective validation confirms its safety, the DIME tool could shorten waits, cut unnecessary testing, and—perhaps most importantly—end the turf war that has too often stood between psychiatric patients and the care they came for.

Subject of Research: Emergency Medicine

Article Title: International Delphi on a medical screening tool for emergency department patients with psychiatric complaints

Article References: Ünlü, L., Carpenter, C. R., Sterzer, P., Christ, M., Azad, N., Griese, J. A., Nordstrom, K., Skinner, C., Alsma, J., Stiebel, V., Kurzhals, S., Pascoe, E., Brister, T., van der Eycken, E., Maia, I. W. A., Tintinalli, J., Bingisser, R., Wilson, M. P., Nickel, C. H., … Zun, L. S. (2026). International Delphi on a medical screening tool for emergency department patients with psychiatric complaints. BMC Medicine. https://doi.org/10.1186/s12916-026-05324-7

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05324-7

Keywords: emergency department, medical screening, psychiatric complaints, Delphi study, consensus, DIME tool, emergency medicine, psychiatry, patient involvement, BMC Medicine, International, Delphi

The research team, led by Sithum Munasinghe with colleagues including Kate McBride, Phuong Dung Nguyen, Matthew Spittal and Andrew Page, focused on nine cancer types previously linked to rising rates in younger adults: breast, colorectal, uterine, kidney, liver, pancreatic, thyroid, prostate and oesophageal cancers. To interpret those trends, they compared them against three comparator cancers without early-onset increases: lung cancer, cervical cancer and melanoma of the skin. The design matters because it separates genuine generational risk changes from artefacts of improved detection, which have complicated the debate over why young people appear to be developing certain cancers at higher rates than their parents and grandparents did.

Age-period-cohort analysis is a statistical framework designed to disentangle three forces that all shape disease rates over time. Age effects capture the well-known rise in cancer risk as people get older. Period effects reflect influences that affect everyone alive at a particular moment, such as the introduction of a national screening programme, a change in diagnostic practice or a new treatment that alters recorded mortality. Cohort effects, by contrast, track the experiences of people born in the same era who carry shared exposures through life, from dietary patterns and physical activity levels to antibiotic use, obesity prevalence and early-life microbiome changes. Because the three dimensions are mathematically interdependent, with each year of birth equal to a calendar year minus an age, sophisticated modelling is required to estimate each component without confounding the others.

When the researchers applied this framework to incidence data, a clear pattern emerged. For both males and females, there were increasing cohort effects in colorectal, liver, pancreatic and thyroid cancers, meaning that successive birth cohorts experienced higher rates of these cancers at equivalent ages than earlier generations did. Among women, uterine cancer showed the same generational rise, and among men, prostate cancer did as well. The steepest relative increases were recorded for pancreatic cancer in females and thyroid cancer in males, two sites that have featured prominently in international reports of early-onset cancer concern. These increases were evident after accounting for age and calendar-period influences, strengthening the interpretation that something about the lived experience of later-born generations is driving the trend.

Equally informative were the cancers that bucked the pattern. Cohort effects were decreasing for melanoma of the skin and for cervical cancer among women, a finding consistent with successful public health interventions: sun-protection campaigns that have reshaped behaviour since childhood for Australians born after the 1970s, and human papillomavirus vaccination and cervical screening that have driven down rates in younger cohorts. For lung cancer and oesophageal cancer, the analysis found no substantial differences across birth cohorts, suggesting that the early-onset increases reported for other cancer types are not a universal phenomenon but a site-specific one tied to particular exposures.

The period effects told a different and largely reassuring story. Once cohort effects were accounted for, period influences on incidence were relatively stable for most cancer types, with the exceptions of prostate cancer among men and cervical cancer among women. Stable period effects suggest that calendar-specific events, such as the gradual rollout of diagnostic technologies, have not been the dominant driver of the observed incidence trends for most of these cancers. This is a crucial point for the ongoing scientific debate, because critics of the early-onset cancer narrative have argued that increased screening and incidental detection could explain much of the apparent rise. The Australian data indicate that for the majority of the cancers studied, the generational signal, not the calendar signal, carries the weight.

Mortality trends painted a more optimistic picture overall. Decreasing period and cohort effects for cancer mortality were evident for most cancer types and for both males and females, reflecting decades of progress in early detection, surgical technique, systemic therapies and supportive care. However, the exceptions were notable: increasing period effects were observed for thyroid cancer and pancreatic cancer, and for cervical and uterine cancers among women. Rising period effects in mortality can reflect shifts in the underlying burden of disease arriving at the health system, or changes in how deaths are certified and classified, and for pancreatic cancer in particular they align with a cancer type that remains notoriously difficult to detect early and treat effectively.

The authors conclude that the higher population-level incidence of these cancer types among later-born cohorts may indicate cohort-specific differences in risk factors associated with these outcomes. That phrasing is deliberately cautious, and appropriately so. Age-period-cohort models are descriptive-analytic tools: they can establish that a generational pattern exists and quantify its magnitude, but they cannot identify which specific exposures caused it. Candidate explanations discussed in the broader literature include rising obesity rates, changes in diet and alcohol consumption, altered gut microbiota, earlier puberty, sedentary lifestyles and antibiotic exposure in early life, but the present study does not test these hypotheses directly. What it does provide is a rigorous temporal scaffold onto which future aetiological research can be mapped.

The Australian setting gives the findings particular weight. The country maintains comprehensive, population-based cancer registries with incidence data reaching back to 1982 and mortality data to 1971, allowing cohort comparisons that span people born across most of the twentieth century. Australia also offers instructive natural experiments in cancer control, from its world-leading sun-safety campaigns to its organised cervical screening and HPV vaccination programmes, and the decreasing cohort effects for melanoma and cervical cancer show those interventions registering clearly in the generational data. The contrast between these success stories and the rising cohort effects for colorectal, pancreatic, liver and thyroid cancers underscores that the early-onset phenomenon is not an inevitable feature of modern life but a modifiable one.

For clinicians and policymakers, the implications are twofold. First, health systems may need to reconsider age thresholds for awareness and, where evidence supports it, screening for cancers increasingly seen in younger adults, since guidelines historically calibrated to older populations can leave early-onset cases detected late. Second, prevention strategies must target exposures accumulated from childhood onward, because cohort effects by definition reflect risks carried across the life course rather than hazards encountered in middle age. The study used aggregate, publicly available tabulations of incidence and mortality data and required no individual-level ethics approval, and the authors declared no competing interests. As cohorts born in the 1980s and 1990s move further into middle age, continued surveillance will reveal whether the generational risk elevations documented here persist, plateau or, with effective intervention, follow the downward path already traced by melanoma and cervical cancer in the same population.

Subject of Research: Age-period-cohort analysis of early-onset cancer incidence and mortality trends in Australia

Article Title: Age, period, and cohort effects in the population-level incidence and mortality of cancer types associated with early onset increases in Australia

Article References: Munasinghe, S., McBride, K., Nguyen, P. D., Spittal, M., & Page, A. (2026). Age, period, and cohort effects in the population-level incidence and mortality of cancer types associated with early onset increases in Australia. BMC Medicine. https://doi.org/10.1186/s12916-026-05293-x

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05293-x

Keywords: early-onset cancer, age-period-cohort analysis, cancer epidemiology, Australia, incidence trends, mortality trends, colorectal cancer, pancreatic cancer, thyroid cancer, birth cohort effects, BMC Medicine, cancer prevention

Pancreatic portal hypertension is a distinct type of portal hypertension caused by various pancreatic diseases or by pancreatic surgery. Unlike the far more common portal hypertension driven by cirrhosis of the liver, this variant arises when inflammation, cysts, tumors, or scarring in the pancreas compress or obstruct the splenic vein and neighboring splanchnic vessels. Blood that cannot drain through its usual route finds alternative channels, engorging vessels around the stomach and esophagus and enlarging the spleen. The result is a condition that mimics liver disease on the surface but has an entirely different anatomy and, crucially, a different treatment logic.

The clinical significance of the new consensus lies partly in a quiet epidemiological shift. According to the guideline authors, the detection rate of pancreatic portal hypertension has steadily increased in recent years with the widespread application of advanced imaging techniques and endoscopy. Computed tomography and magnetic resonance imaging now routinely capture the pancreatic region in exquisite detail, while endoscopic ultrasound allows physicians to peer directly at the vessels surrounding the pancreas. Diseases that once went unrecognized are now being identified with growing frequency, making a standardized management framework increasingly urgent.

Yet the condition remains notoriously difficult to catch early. The consensus panel notes that the clinical manifestations of pancreatic portal hypertension are often insidious, compromising early identification. Patients may carry the vascular changes for years without symptoms, and the first warning signs frequently emerge only at an advanced stage. By that point, some complications, such as splenomegaly and upper gastrointestinal bleeding, have already taken hold. Bleeding from engorged gastric varices can be dramatic and life threatening, while an enlarged spleen can trap platelets and complicate the broader management of the underlying pancreatic disease.

To build the guideline, the Chinese Society of Gastroenterology of the Chinese Medical Association invited a panel of skilled experts to participate in a Delphi consensus process, systematically reviewing and discussing the available evidence on the causes, risk factors, clinical manifestations, diagnosis, therapy, and prognostic assessment of pancreatic portal hypertension. The Delphi method is a structured technique for extracting reliable group judgment from a large panel: experts answer questionnaires independently, the results are aggregated and fed back, and the process repeats until statements stabilize. After three rounds of questionnaire surveys and votes, the panel formulated 19 consensus statements to provide clinical guidance on the diagnosis and management of the condition.

The scope of the consensus is deliberately broad. By covering etiology, risk factors, clinical presentation, diagnostic strategy, therapeutic options, and prognostic assessment in a single document, the guideline fills a gap that existing publications on pancreatic diseases or portal hypertension had left open. Although the concept of pancreatic portal hypertension has been recognized in existing guidelines and consensus documents on the management of pancreatic diseases or portal hypertension, its management strategy had never been comprehensively summarized. Clinicians confronting a patient with pancreatic disease and unexplained gastric varices or splenic enlargement have had to piece together guidance from scattered sources, and the new consensus aims to end that improvisation.

Technically, the diagnostic challenge centers on distinguishing pancreatic portal hypertension from cirrhotic portal hypertension and from other causes of splanchnic vein thrombosis. Cross-sectional imaging plays the decisive role: contrast-enhanced computed tomography and magnetic resonance imaging can reveal the obstructing pancreatic lesion, the occluded or compressed splenic vein, and the collateral vessels that betray the redirected blood flow. Endoscopic evaluation identifies the varices that carry bleeding risk, while laboratory findings such as thrombocytopenia from hypersplenism can provide supporting clues. The consensus framework gives clinicians a structured pathway for assembling these pieces into a confident diagnosis, even when the liver itself is entirely healthy.

On the therapeutic side, the management logic differs fundamentally from that of cirrhotic portal disease. Because the obstruction is often localized to the splenic vein, treating the underlying pancreatic lesion can sometimes restore normal venous drainage, and interventions aimed at the vascular obstruction itself, including endovascular approaches, have been explored in the literature. The consensus statements address how to weigh bleeding risk against the risks of intervention, how to manage acute upper gastrointestinal bleeding when it occurs, and how to coordinate vascular care with the treatment of the primary pancreatic condition, whether that is acute pancreatitis, chronic pancreatitis, autoimmune pancreatitis, or a pancreatic tumor.

The panel was large and geographically diverse, drawing gastroenterologists, endoscopists, interventional radiologists, hepatobiliary surgeons, and intensive care specialists from major hospitals across China, from Shenyang in the northeast to Chengdu, Shanghai, Guangzhou, and Xi’an. Corresponding authors include Xingshun Qi of the General Hospital of Northern Theater Command, Hao Wu and Chengwei Tang of West China Hospital, Sichuan University, and Xiaozhong Guo of the General Hospital of Northern Theater Command. Four junior researchers, Xingshun Qi, Yuhang Yin, Zhiyin Huang, and Yuzheng Zhuge, contributed equally to the work, which was partially supported by the National Key R&D Program of China.

The methodological rigor of the consensus reflects modern guideline standards. The panel employed structured evidence frameworks, including PICO and PECO question formats for framing clinical questions, and used the GRADE system for grading the quality of evidence behind each recommendation. Reporting followed the RIGHT checklist for practice guidelines, and the document was appraised with the AGREE II instrument. Literature searches spanned Chinese databases such as CNKI and VIP alongside international sources, ensuring that both Chinese clinical experience and global evidence informed the final statements. The authors declare no competing interests, and the guideline is published under a Creative Commons license allowing free access for clinicians worldwide.

For the international medical community, the significance of the consensus extends beyond China’s borders. Pancreatic portal hypertension is a global condition, arising wherever pancreatitis, pseudocysts, tumors, or pancreatic surgery compromise the splenic vein, and clinicians on every continent face the same diagnostic and therapeutic dilemmas. A comprehensive, openly available framework for the causes, risk factors, clinical manifestations, diagnosis, therapy, and prognostic assessment of the disease gives physicians everywhere a reference point for building their own local protocols. As advanced imaging continues to reveal more cases that once went undetected, the demand for such standardized guidance will only grow, and the 19 statements now offer a foundation that future international guidelines can refine and extend.

Subject of Research: Medicine

Article Title: Chinese consensus on the management of pancreatic portal hypertension

Article References: Qi, X., Yin, Y., Huang, Z., Zhuge, Y., Cai, S., Chen, M., Chen, P., Chen, S., Chen, W., Ding, Z., Du, Y., Guo, G., Guo, W., He, F., He, W., Hu, S., Huang, L., Ke, L., Kong, D., … Tang, C. (2026). Chinese consensus on the management of pancreatic portal hypertension. BMC Medicine. https://doi.org/10.1186/s12916-026-05289-7

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05289-7

Keywords: pancreatic portal hypertension, portal hypertension, consensus, gastrointestinal bleeding, splenomegaly, BMC Medicine, Delphi, pancreatic disease, Chinese, management, pancreatic, portal

The clinical problem at the heart of the study is deceptively narrow but surprisingly stubborn. The parotid gland, the largest of the salivary glands, can develop two benign tumors that together account for the vast majority of non-cancerous parotid masses: pleomorphic adenoma and Warthin’s tumor. Both matter because their management differs, yet on ultrasound their appearances overlap enough that even experienced ultrasonographers can hesitate. Most AI tools pitched at this task have so far lived comfortably inside retrospective datasets, validated on images the algorithm had effectively already seen in development, and never tested in the messier environment where a real radiologist either accepts or ignores the machine’s suggestion. The Korean team, led from the Department of Biomedical Informatics at Korea University College of Medicine together with clinicians at Korea University Ansan and Anam Hospitals, set out to break that pattern.

Architecturally, the system is a hybrid pipeline built around a division of labor. The first stage uses a YOLO-family object detection model, which scans the ultrasound image to locate the suspicious lesion and delineate the region of interest. The second stage is a classification model, based on the ConvNeXt convolutional network architecture, that examines the localized region and renders a verdict: pleomorphic adenoma or Warthin’s tumor. Crucially, the pipeline is confidence-aware. When the detection model’s certainty about the region it has found falls below a pre-specified threshold, set at 0.80 in the final deployment configuration, the system falls back on the classification model to interpret the whole image instead. That fallback mechanism means the AI never has to commit to a localization it does not trust, a small piece of engineering humility with outsized consequences for performance.

The development data came from 949 ultrasound images collected at Korea University Ansan Hospital, with a completely separate external set of 427 images from Korea University Anam Hospital reserved for testing. Rather than assuming their architecture choices were optimal, the researchers explored 35 different combinations of detection and classification models to understand how the two components complement each other. The exploration revealed something subtle: the hybrid pipeline outperformed both of its standalone components in 22 of the 35 pairings on external data, and the benefit was most pronounced when the detection and classification models performed at a similar level. The best configuration achieved a macro-F1 score of 0.838, a balanced measure of precision and recall. When one component was dramatically stronger than the other, the hybrid’s advantage shrank, a finding that carries practical lessons for anyone assembling multi-stage medical AI systems.

The standalone results themselves underscore the diagnostic difficulty. On the external test set, the detection-first pipeline achieved an area under the receiver operating characteristic curve, or AUROC, of 0.811, while the classification model acting alone reached 0.919. AUROC is a standard gauge of a model’s ability to distinguish the two tumor types across all possible decision thresholds, with 1.0 representing perfect discrimination. But the authors argued that neither number alone answers the question that actually matters in a clinic: does the AI make radiologists better? That question demands a different kind of study entirely, one in which human observers read cases first unaided and then with the algorithm whispering over their shoulder, ideally without knowing which is which.

That is precisely what the two-phase observer-performance study delivered. Three board-certified radiologists read 107 external patient cases, three reads per case for a total of 321 interpretations, first on their own and then with AI assistance available through an interface the radiologists could edit. The statistical machinery, a generalized linear mixed model, was chosen because it can properly handle the fact that each radiologist read each case and the repeated measures are entangled. The verdict: pooled diagnostic accuracy climbed from 76.3 percent to 81.6 percent when AI was in the loop, an odds ratio of 1.54 with a 95 percent confidence interval of 1.12 to 2.13 and a p-value of 0.008, meaning the improvement is unlikely to be statistical noise. Just as strikingly, the three readers converged toward each other, with Fleiss’ kappa, a measure of inter-reader agreement that ranges from 0 to 1, rising from 0.475 to 0.611. In plain terms, the AI did not just make the radiologists more accurate; it made them more consistent with one another.

Perhaps the most revealing results concern how the human readers actually behaved, because selective trust is the difference between collaboration and automation bias. The radiologists were not rubber stamps. They followed the AI’s correct predictions 76.2 percent of the time but adopted its incorrect predictions only 51.1 percent of the time, a statistically significant gap that suggests genuine discrimination rather than blind deference. The benefits also concentrated where they were most needed: on harder cases, the gain from AI assistance was larger, which is exactly the profile one wants from a decision-support tool. The system’s explainability layer, using techniques such as Grad-CAM heatmaps to show which image regions drove the classification and D-RISE explanations for the detection model, gave readers visual evidence they could weigh against their own reading of the scan.

There is also an honest asymmetry in the data that the authors did not hide. Sensitivity for Warthin’s tumor jumped by 18.1 percentage points under AI assistance, while sensitivity for pleomorphic adenoma slipped by 9.3 percentage points. The pattern tracks the AI’s own prediction profile: the machine was better at spotting Warthin’s tumors, and readers absorbed that strength along with that weakness. It is a candid illustration that AI assistance reshapes the whole diagnostic landscape rather than uniformly lifting it, and it flags a priority for future iterations of the system, which should aim to rebalance per-class performance before wider deployment.

Why does this particular trial matter beyond the parotid gland? Because prospective, reader-based evaluations of medical AI remain rare, even as algorithms multiply in the radiology literature. A model can post a dazzling AUROC on a curated dataset and still fail to change a single clinical decision. By pre-specifying a single deployment configuration, testing on external cases from a different hospital, and measuring outcomes that track actual human decisions, this study models the kind of evidence regulators, clinicians, and patients should be demanding. The authors were careful to frame the result as support for the system’s potential as a decision-support tool that operates alongside the interpreting physician, not as a replacement, and to call for further evaluation in real-world clinical workflows where scanner variation, workflow pressure, and patient diversity all conspire to degrade performance.

The trajectory from here is familiar but demanding. The system was built and validated within a single national health system, with consent waived for the retrospective components and institutional review board approval at both hospitals, and the deployment threshold of 0.80 was fixed in advance rather than tuned on the test set. Broader multi-center validation, monitoring for performance drift as imaging hardware and patient populations change, and studies of how radiologists interact with editable AI interfaces over months rather than one sitting are the logical next steps. But the core finding stands on its own: when an AI system is designed to know the limits of its confidence and to keep the radiologist firmly in command, the combination reads salivary gland ultrasound better than either the machine or the human alone. That may be the most viral idea in medical AI right now, not that algorithms will replace doctors, but that carefully engineered humility in software can make human judgment measurably, and provably, sharper.

Subject of Research: Human–AI collaboration for AI-assisted differentiation of benign parotid tumors on ultrasound

Article Title: Human–AI interaction in parotid ultrasound diagnosis: a prospective two-phase observer-performance study

Article References: Kim, J., Lee, S., Hwang, H., Ha, S., Kim, Y., Jeong, I., Kim, M.-S., Kang, B., Kim, J. H., Hong, S. J., Lee, Y. H., Kim, Y. S., Baek, S.-K., You, S.-H., Kim, H., Park, H. S., Yoo, Y., Kwack, T. J., Kwon, S. Y., & Lee, H. (2026). Human–AI interaction in parotid ultrasound diagnosis: a prospective two-phase observer-performance study. BMC Medicine. https://doi.org/10.1186/s12916-026-05301-0

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05301-0

Keywords: artificial intelligence, parotid ultrasound, pleomorphic adenoma, Warthin's tumor, deep learning, radiology, observer-performance study, decision support, confidence-aware AI, ConvNeXt, YOLO, BMC Medicine

The trial, registered at ClinicalTrials.gov under identifier NCT05166239, enrolled 66 patients with previously untreated advanced hepatocellular carcinoma complicated by portal vein tumor thrombosis between December 2021 and December 2023. Participants were assigned to one of two treatment strategies. The control group received lenvatinib, an oral multikinase inhibitor that blocks several receptors driving tumor blood vessel growth and cell proliferation, combined with a programmed death receptor 1 inhibitor, an immunotherapy that releases the molecular brakes cancer places on T cells. The experimental group received the same drug combination plus hepatic arterial infusion chemotherapy, a technique in which chemotherapeutic agents are delivered directly into the hepatic artery, the vessel that supplies liver tumors with most of their blood.

The logic behind this triple combination rests on the distinctive vascular anatomy of liver tumors. Hepatocellular carcinoma is overwhelmingly fed by the hepatic artery, while the surrounding healthy liver parenchyma draws most of its blood from the portal vein. By infusing chemotherapy through the hepatic artery, clinicians can achieve very high local drug concentrations inside the tumor while limiting systemic exposure and the toxicity that follows it. This regional approach also matters for tumor thrombi in the portal vein itself, because arterial blood can supply the thrombus through tiny vascular networks that form within the clot. Delivering chemotherapy through the artery therefore reaches both the primary tumor mass and the malignant invasion of the portal vasculature that makes this stage of disease so lethal.

The immunotherapy component adds another layer of biological rationale. Programmed death receptor 1 inhibitors work by blocking the interaction between the PD-1 receptor on T cells and its ligand, which tumors exploit to render attacking immune cells dysfunctional. Lenvatinib has been shown in preclinical work to remodel the tumor microenvironment, reducing immunosuppressive cells and increasing the infiltration of activated T cells, which may sensitize tumors to immune checkpoint blockade. The open question was whether adding hepatic arterial infusion chemotherapy to this systemic duo would amplify the benefit or simply pile on toxicity. The new results provide a clear answer on the efficacy side: the triple regimen was associated with markedly better outcomes across every efficacy endpoint the investigators measured.

The primary endpoint of the study was the six-month progression-free survival rate, a measure of how many patients remain alive without their cancer worsening at the half-year mark. In the triple-therapy group, 78.8 percent of patients were progression-free at six months, compared with 46.9 percent in the group receiving only lenvatinib and the PD-1 inhibitor, a difference that reached statistical significance. Median progression-free survival, the point at which half of the patients had experienced disease progression, was 13.7 months in the triple-therapy arm versus 5.9 months in the control arm, corresponding to a hazard ratio of 0.422. In plain terms, patients receiving the added regional chemotherapy had roughly 58 percent lower risk of progression at any given time during follow-up.

The survival advantage extended to overall survival, the most consequential endpoint in oncology. Median overall survival was 25.4 months in the triple-therapy group compared with 12.6 months in the control group, with a hazard ratio of 0.502, meaning the risk of death was approximately halved. Time to progression followed the same pattern, at 15.1 months versus 6.2 months and a hazard ratio of 0.345. Perhaps most striking was the objective response rate assessed under the modified Response Evaluation Criteria in Solid Tumors, the imaging standard designed specifically for hepatocellular carcinoma that measures tumor viability rather than size alone. In the triple-therapy arm, 78.8 percent of patients achieved an objective tumor response, nearly double the 42.4 percent observed in the systemic-therapy-only arm.

Because this was a prospective controlled study rather than a fully randomized trial, the investigators took additional statistical precautions to guard against bias from imbalances between the two groups. They performed a sensitivity analysis using propensity score overlap weighting, a method that reweights the comparison groups to resemble a randomized experiment by balancing measured baseline characteristics. The treatment effect estimates from this analysis were generally similar in both direction and magnitude to the primary findings, strengthening confidence that the observed benefits were attributable to the therapy itself rather than to differences in patient characteristics between the groups.

Safety data told a more nuanced story. The incidence of grade 3 or 4 treatment-related adverse events, the severe toxicities that can interrupt or force discontinuation of treatment, was 57.6 percent in the triple-therapy group compared with 42.4 percent in the control group. This increase reflects the added burden of intra-arterial chemotherapy layered on top of two systemic agents, each of which carries its own toxicity profile. The investigators characterized the safety profile of the triple regimen as acceptable, and the substantial gains in survival and tumor response appear to justify the additional toxicity for many patients, though the finding underscores that this intensified approach requires careful patient selection, experienced procedural teams, and vigilant monitoring for liver function deterioration, cytopenias, and other serious events graded under the Common Terminology Criteria for Adverse Events.

The clinical significance of these numbers becomes clear when placed in context. Advanced hepatocellular carcinoma with macrovascular invasion has historically carried a grim prognosis, with median survival often measured in months rather than years. The control arm of this study, receiving lenvatinib plus a PD-1 inhibitor, represents the current standard of care and achieved a median overall survival of 12.6 months, consistent with what systemic therapy alone typically delivers in this setting. Doubling that figure to 25.4 months by adding a catheter-based regional chemotherapy procedure represents one of the larger survival improvements reported for this patient population, and the near-doubling of objective response rates suggests that shrinking the tumor and its portal vein invasion may open doors to further local treatments in some patients.

The authors and independent commentators alike emphasize that confirmation in larger randomized studies is warranted before the triple regimen can be considered a new standard of care. This study was conducted at a single center, which raises questions about generalizability across different patient populations, treatment protocols, and levels of interventional expertise, and the open-label design means neither patients nor treating physicians were blinded to the assignment. Nevertheless, the consistency of the benefit across progression-free survival, overall survival, time to progression, and objective response rate, together with supportive sensitivity analyses, marks this trial as one of the most compelling pieces of evidence to date that regional hepatic arterial infusion chemotherapy can meaningfully augment systemic immunotherapy-based treatment. For patients facing a diagnosis that combines advanced liver cancer with portal vein tumor thrombosis, a combination once considered a near-uniform death sentence, the study offers something that has been in short supply: a credible path toward two additional years of life.

Subject of Research: Hepatic arterial infusion chemotherapy combined with lenvatinib and PD-1 inhibitor therapy for hepatocellular carcinoma with portal vein tumor thrombosis

Article Title: Hepatic arterial infusion chemotherapy plus lenvatinib and programmed death receptor 1 inhibitor for hepatocellular carcinoma with portal vein tumor thrombosis: a prospective controlled study

Article References: Zheng, K., Liu, M., Zhu, X., Cao, G., Xu, L., Fu, S., Dai, Y., Liu, P., Gao, S., Kou, F., Gao, Q., Wang, K., Yan, X., Xu, D., Bao, Q., Wang, L., Wu, J., Hao, C., Xing, B., & Wang, X. (2026). Hepatic arterial infusion chemotherapy plus lenvatinib and programmed death receptor 1 inhibitor for hepatocellular carcinoma with portal vein tumor thrombosis: a prospective controlled study. BMC Medicine. https://doi.org/10.1186/s12916-026-05309-6

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05309-6

Keywords: hepatocellular carcinoma, portal vein tumor thrombosis, hepatic arterial infusion chemotherapy, lenvatinib, PD-1 inhibitor, immunotherapy, liver cancer, progression-free survival, overall survival, clinical trial, interventional oncology, BMC Medicine

The significance of this consensus lies in what it replaces: a patchwork of institutional habits, fragmented society guidelines, and anecdotal practice. As interventional ultrasound has expanded from simple core-needle biopsies to complex tumor ablations, percutaneous transhepatic biliary drainage, and sclerotherapy of vascular malformations, the range of bleeding risks has widened dramatically. Severe hemorrhage, though uncommon, can be life-threatening and can undo the very benefits that minimally invasive approaches promise. Until now, no international consensus existed to guide clinicians through risk assessment, prevention, monitoring, and rescue management across this rapidly growing procedural landscape.

The rigor behind the document is notable. The development process followed the World Health Organization Handbook for Guideline Development, was reported in accordance with the Reporting Items for Practice Guidelines in Healthcare (RIGHT) statement, and was independently appraised using the AGREE II instrument, a validated tool for judging guideline quality. Evidence quality was graded with the GRADE approach, the international standard for rating the certainty of evidence behind clinical recommendations. To distill disagreement into agreement, the panel employed a two-round modified Delphi process, an iterative voting technique that allows large groups of experts to converge on consensus statements while minimizing the influence of any single dominant voice. The result is a set of 20 finalized recommendations spanning the entire arc of patient care.

At the heart of the consensus is a dual classification system for bleeding complications, organized simultaneously by time of onset and by severity. This dual axis is more than taxonomy; it is designed to standardize clinical communication and decision-making. A hemorrhage that declares itself on the table demands a different response than one that smolders and presents hours later, and a grading scheme that separates minor oozing from catastrophic intra-abdominal bleeding allows teams to escalate care along a shared mental model. Standardized classification also enables meaningful comparison across studies and institutions, something the field has historically lacked because bleeding events were defined inconsistently from one trial to the next.

Equally central is a multidimensional pre-procedural risk assessment that moves far beyond the traditional single-number approach of checking a platelet count and an international normalized ratio. The consensus directs clinicians to weigh procedural risk, meaning the inherent hemorrhagic danger of the specific intervention; patient-related factors, including coagulation status, comorbidities, and antithrombotic medications; lesion characteristics, such as vascularity and depth; device-related factors, including needle caliber and ablation modality; and operator experience, an often-overlooked variable that robust evidence links to complication rates. By forcing a structured evaluation across all five domains, the framework aims to identify high-risk scenarios before the first incision rather than after the first complication.

The recommendations then extend into individualized periprocedural management, addressing one of the thorniest questions in interventional practice: how to handle anticoagulant and antiplatelet therapy. Patients on these medications are increasingly common, and the balance between withholding drugs and risking thromboembolism versus continuing them and risking hemorrhage is genuinely difficult. The consensus emphasizes tailoring periprocedural strategies to the individual patient and procedure rather than applying rigid rules, reflecting the reality that a liver biopsy in a patient on dual antiplatelet therapy is a very different proposition from a superficial thyroid fine-needle aspiration.

After the procedure, the statement introduces risk-adapted ultrasound surveillance, another departure from one-size-fits-all practice. Low-risk patients may require little more than brief observation, while those who underwent high-risk procedures or carry patient-specific risk factors warrant closer monitoring, potentially with color Doppler flow imaging and contrast-enhanced ultrasound, techniques that can detect active extravasation and expanding hematomas at the bedside without ionizing radiation. The principle is that vigilance should be proportional to risk, concentrating resources where they are most likely to change outcomes and catching delayed bleeding before it becomes hemodynamic collapse.

Perhaps the most clinically actionable portion of the consensus is its stepwise hemostatic management algorithm, which escalates from the simplest measures to the most invasive. Local compression remains the first response to accessible bleeding, followed by tract embolization, in which the needle track itself is occluded to stop persistent oozing along the puncture route. When these measures fail, the framework contemplates selected ultrasound-guided interventions, then transarterial embolization, in which interventional radiologists catheterize the feeding vessel under digital subtraction angiography and occlude it, and finally surgery as the definitive rescue option. This graduated pathway gives clinicians a clear sequence of escalation, reducing hesitation and delay, the two factors most consistently associated with poor outcomes in major hemorrhage.

The authors are candid about the limits of their work. The available evidence across interventional ultrasound procedures remains heterogeneous, with robust randomized trial data abundant for some interventions and sparse for others, and some recommendations rest partly on expert opinion where high-quality direct evidence is limited. The panel also stresses that the framework is intended to guide rather than dictate, allowing individualized application according to procedural characteristics, patient-related risk factors, operator experience, and institutional resources. A small hospital without 24-hour angiography suites will apply the escalation pathway differently from a tertiary referral center, and the consensus explicitly accommodates that reality rather than pretending it away.

What emerges is a document with the potential to reshape daily practice across radiology suites, ultrasound departments, and surgical theaters worldwide. By uniting classification, risk assessment, prevention, surveillance, and rescue management under one evidence-graded umbrella, the consensus gives clinicians something they have never had: a shared language and a shared playbook for the complication they fear most. The authors call for further multicenter prospective studies to validate and refine the recommendations, and such research is likely to follow now that standardized definitions exist to make it possible. For patients, the practical meaning is straightforward: fewer surprises, faster responses, and safer minimally invasive care, delivered with the confidence that comes from global agreement rather than local guesswork.

Subject of Research: International expert consensus on prevention and management of bleeding complications in interventional ultrasound procedures

Article Title: Prevention and management of bleeding complications in interventional ultrasound: an international expert consensus statement

Article References: Xu, M., Wang, Y., Ye, Z., Dietrich, C. F., Lee, J. M., Yan, K., Zhao, Q., Shen, Y., Hu, X., Solbiati, L. A., Kuo, J., Mauri, G., Cho, J. Y., Lee, J. Y., Huang, K., Rangel, L. G., Huang, M., Jing, X., Cheng, W., … Jiang, T. (2026). Prevention and management of bleeding complications in interventional ultrasound: an international expert consensus statement. BMC Medicine. https://doi.org/10.1186/s12916-026-05315-8

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05315-8

Keywords: interventional ultrasound, bleeding complications, hemostasis, expert consensus, patient safety, GRADE, Delphi process, transarterial embolization, contrast-enhanced ultrasound, anticoagulation, tumor ablation, BMC Medicine

The study, published in BMC Medicine, enrolled 1,202 adults with suspected infection and systematically compared the LIP score against the existing arsenal of bedside and laboratory-based assessments, including SIRS criteria, qSOFA, SOFA, NEWS, and APACHE-II. The results were striking. At a cutoff of 3 points or more, the LIP score achieved an area under the receiver operating characteristic curve (AUROC) of 0.932, with a 95 percent confidence interval of 0.916 to 0.945. In practical terms, that means the score distinguished septic from non-septic patients with near-excellent discrimination. Its sensitivity reached 85.37 percent—meaning it caught the vast majority of true sepsis cases—while its specificity climbed to 94.83 percent, meaning it rarely raised false alarms. Statistically, it significantly outperformed every comparator score tested, with all comparisons yielding P values below 0.001.

What makes this performance notable is not just the numbers but the philosophy behind them. Rather than being mined from a database by machine learning and reverse-engineered afterward, the LIP score was constructed from pathophysiology—each component chosen because it captures a distinct biological axis of sepsis. The lymphocyte count reflects the immune system’s state: in sepsis, lymphocytes become depleted, a phenomenon linked to immunosuppression and the expansion of myeloid-derived suppressor cells that blunt the body’s defenses. The international normalized ratio, or INR, a standard measure of blood clotting tendency, captures the coagulation disturbances that accompany sepsis, from sepsis-induced coagulopathy to the life-threatening cascade of disseminated intravascular coagulation. Procalcitonin, a peptide precursor of the hormone calcitonin, rises sharply in bacterial infection and has long served as one of the most trusted biomarkers for distinguishing bacterial sepsis from non-infectious inflammation.

By combining these three markers, the score essentially triangulates sepsis from three directions at once: immune collapse, clotting dysfunction, and bacterial burden. Each component contributes points, and the total—ranging from 0 upward—signals the probability that a patient meets the Sepsis-3 definition, the current international consensus standard. This design choice matters clinically because all three tests are already ordered routinely in any hospital ward evaluating a febrile patient. No new assay, no specialized equipment, and no additional blood draw is required; the score can be computed from data that already exists in the patient’s chart, making it extraordinarily cheap and fast to deploy.

The study’s analytical rigor went well beyond a simple head-to-head comparison. The researchers used restricted cubic splines, a flexible statistical technique, to model the relationship between LIP scores and intensive care unit length of stay, uncovering a biphasic association—patients at both low and high ends of the score spectrum showed distinct patterns of ICU utilization. They also deployed advanced visualization methods, including Sankey and Chord diagrams, to map how patients flowed between risk strata, and SHAP (SHapley Additive exPlanations) analysis, a machine learning interpretability technique, to quantify how much each of the three components contributed to the score’s predictions. This blend of classical epidemiology and modern interpretability tools reflects a growing trend in critical care research: scores must not only work, but clinicians must understand why they work before trusting them with lives.

Prognostically, the picture was more nuanced. In predicting mortality, the LIP score surpassed the old SIRS criteria decisively (P < 0.001), confirming that the outdated systemic inflammatory response syndrome criteria—which many experts have long argued are too sensitive and too nonspecific—are inferior for risk stratification. However, LIP was outperformed by the National Early Warning Score (P = 0.031) and by SOFA (P < 0.001), while performing comparably to qSOFA (P = 0.28). This is an honest and important caveat: SOFA, which requires a fuller panel of physiological measurements, remains the gold standard for assessing organ dysfunction and predicting death. The LIP score is not positioned to replace it, but rather to serve as a rapid, early screening gate—flagging patients who then warrant full SOFA evaluation and aggressive intervention.

The study also explored the score’s role as an early-warning tool, and here the researchers offered a clever refinement. Using a simple threshold of 2 points or more, the LIP score achieved very high sensitivity of 94.58 percent—catching nearly every patient who would go on to develop sepsis—but specificity dropped to a limited 62.72 percent, meaning many flagged patients would be false positives. To solve this, the team tested a modified rule: instead of summing the total score, clinicians flag a patient if any two of the three components score at least 1 point each. This modification pushed specificity up to 83.14 percent, dramatically reducing false alarms while preserving the score’s ability to catch early cases. This kind of dual-threshold architecture—high sensitivity for initial screening, higher specificity for confirmation—mirrors how successful screening programs operate in other areas of medicine.

Beyond diagnosis and mortality, the researchers probed whether the LIP score could predict microbiological confirmation—whether blood cultures and other tests would actually identify a pathogen. Here the score performed modestly, on par with qSOFA, APACHE-II, and NEWS. That is a realistic limitation worth acknowledging: sepsis is a clinical syndrome, not a microbiological diagnosis, and many septic patients never yield a positive culture. A score that predicts clinical deterioration is arguably more valuable at the bedside than one that predicts culture results, but the modesty of this finding underscores that LIP is a clinical screening tool, not a substitute for microbiological workup.

The multicenter, prospective design strengthens the findings considerably. Prospective studies, in which data are collected according to a pre-specified protocol as patients are enrolled, are far less prone to the biases that plague retrospective analyses of electronic health records. Spanning ten tertiary hospitals across China, the study captured a geographically and demographically diverse patient population, and the authors report that all participants or their legally authorized representatives provided written informed consent, with ethics approval from the Medical Ethics Committee of Chongqing Emergency Medical Center. The research was supported by funding from the National Natural Science Foundation of China and several Chongqing regional science and health programs, and the authors declare no competing interests—an important transparency note given that sepsis biomarkers often intersect with commercial diagnostic interests.

Where does this leave the field? The authors conclude that the LIP score is a simple and efficient screening tool for sepsis in hospitalized patients, with good diagnostic performance at a cutoff of 3 or more and a close association with clinical outcomes. They also note preliminary evidence suggesting promise for emergency department use, though they are careful to state that further validation is needed before extending the score beyond hospitalized wards. Sepsis kills millions worldwide each year, and the Surviving Sepsis Campaign has long emphasized that early recognition is the single most impactful intervention. A tool that requires nothing more than three numbers already sitting in a patient’s chart—and that outperforms the scores clinicians have relied on for decades—could, if validated in broader and more diverse populations, become a genuinely transformative addition to the sepsis detection toolkit. For now, the LIP score stands as a compelling example of how pathophysiology-driven design, tested with modern statistical rigor, can yield clinical tools that are both scientifically elegant and immediately practical.

Subject of Research: Validation of the LIP score, a three-component blood-test tool for sepsis screening and prognosis

Article Title: The LIP score: a pathophysiology-driven tool for sepsis screening and prognosis—a multicenter prospective study

Article References: Liu, B., Zhang, F., Yan, Y., Jiang, J., Wang, J., Wen, K., Wang, Z., Zhou, R., Li, T., Qin, J., Cheng, P., Chen, M., Fu, X., Han, J., Yan, B., Zhang, M., Hu, Q., Gong, J., Zhang, A., … Niu, B. (2026). The LIP score: a pathophysiology-driven tool for sepsis screening and prognosis—a multicenter prospective study. BMC Medicine. https://doi.org/10.1186/s12916-026-05312-x

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05312-x

Keywords: sepsis, LIP score, procalcitonin, INR, lymphocyte count, diagnostic markers, screening, prognosis, SOFA, early warning, critical care, BMC Medicine

The research, led by Gabrielle Y. Park, A. James O’Malley, Jamaica R. M. Robinson, and Inas S. Khayal, drew on data from Medicare fee-for-service beneficiaries who died with poor-prognosis cancers in the United States during 2018. The team examined the relationship between hospice availability and hospice utilization, using a logistic mixed-effect regression model to account for a battery of individual and contextual variables: age, sex, race and ethnicity, driving distance to the nearest hospice, rurality, neighborhood income, and the racial composition of the patient’s neighborhood. By modeling these factors simultaneously, the researchers could isolate which ones truly moved the needle on enrollment decisions and, crucially, whether the effect of one factor depended on the levels of others.

The results painted a stark demographic picture. Beneficiaries who were female, older, and living in urban ZIP codes had higher odds of enrolling in hospice than their male, younger, and rural counterparts. These gradients align with patterns seen in earlier work on end-of-life care, where rural isolation and the greater intensity of treatment among younger patients have both been linked to lower hospice uptake. But the most consequential finding concerned race. Black, Hispanic, and Asian beneficiaries all had lower odds of hospice use than White beneficiaries, even after the model adjusted for distance to care, rurality, income, and the other measured characteristics.

What makes the study particularly revealing is its treatment of context. The researchers found statistically significant interactions between beneficiary race and two neighborhood-level factors: income and racial composition. In practical terms, this means the effect of living in a wealthier or demographically different neighborhood was not uniform across racial groups. The same environmental condition that might facilitate hospice enrollment for one population could fail to do so, or even work differently, for another. This heterogeneity is exactly the kind of signal that single-factor analyses miss, and it suggests that the pathways leading patients toward or away from hospice are shaped by how individual identity and community context combine, not by either alone.

Distance, by contrast, behaved differently than many would expect. When the team stratified their analyses by race and ethnic subgroups and by ranges of driving distance to a hospice, proximity did not meaningfully modify the effects of the other predictors. In other words, being physically close to hospice services did not erase or substantially reshape the disparities observed across racial and ethnic groups. The study’s title captures this paradox bluntly: patients can be close but unreachable. A hospice a few miles away is of little use if referral practices, communication patterns, cultural expectations, trust, or insurance and administrative hurdles stand between the patient and the service.

The technical machinery behind these conclusions is worth unpacking. A logistic mixed-effect regression is well suited to this kind of question because it models a binary outcome, here hospice enrollment versus non-enrollment, as a function of multiple predictors while incorporating random effects that account for clustering in the data, such as patients nested within regions or health care markets. Stratifying the analysis by distance ranges allowed the investigators to test whether the familiar drivers of hospice use operated differently at ten miles versus fifty. The finding that they largely did not is a methodologically grounded rebuke to the intuition that access is primarily a matter of miles.

The authors argue that these results demand a different conceptual framework altogether. Rather than treating hospice utilization as the product of isolated variables, they propose a pathway-based model grounded in systems thinking, an approach that captures how, when, and through what interactions multiple factors emerge and influence hospice decision-making over time. Under this view, a patient’s journey toward hospice is not a single decision at a single moment but a sequence of encounters: conversations with oncologists, family deliberations, prior experiences with the health care system, community norms, and the behavior of referring clinicians. Disparities can arise at any link in that chain, and intervening on distance alone leaves the rest of the pathway intact.

Such a framework also offers clearer targets for intervention. If the analysis shows that neighborhood racial composition interacts with patient race in shaping enrollment, then outreach programs, clinician training, and community partnerships may need to be tailored rather than uniform. If the effect of neighborhood income differs across groups, then policies that subsidize or expand hospice capacity in low-income areas may not automatically reach the patients who would benefit most. The systems perspective encourages researchers and policymakers to map the entire decision pathway and identify where the flow of patients toward comfort care is constricted, rather than assuming that widening the pipe at the geographic end will fix the flow overall.

The stakes are considerable. Hospice care is associated with reduced use of aggressive care at the end of life, including fewer intensive care unit admissions and invasive procedures in the final days, and it is widely regarded as a marker of high-quality end-of-life care. Persistent lower utilization among Black, Hispanic, and Asian patients therefore represents not just a statistical disparity but a difference in how Americans experience dying, in the burden placed on grieving families, and in the costs borne by the Medicare program. Because the study population consisted entirely of decedents, the researchers note that individual consent to participate was not applicable, and the work was approved by the Dartmouth Committee for the Protection of Human Subjects.

The study was supported by an American Cancer Society Award, and the authors report no competing interests. Its publication in an open-access journal means the full analysis is available to clinicians, health services researchers, and community advocates alike. What it offers them is a reframing: the question is no longer simply whether a hospice exists within driving distance, but why so many patients who live close to one never walk through its door. Answering that question, the authors contend, will require understanding hospice enrollment as a dynamic system of interacting factors, and designing interventions that reach patients wherever in that system the barrier actually lies.

Subject of Research: Racial and ethnic disparities in hospice utilization among older adults with poor-prognosis cancer in the United States

Article Title: Close but unreachable: racial disparities in hospice utilization beyond geographic proximity among older adult patients with cancer

Article References: Park, G. Y., O’Malley, A. J., Robinson, J. R. M., & Khayal, I. S. (2026). Close but unreachable: racial disparities in hospice utilization beyond geographic proximity among older adult patients with cancer. BMC Medicine. https://doi.org/10.1186/s12916-026-05249-1

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05249-1

Keywords: hospice care, end-of-life care, health disparities, cancer, Medicare, geographic access, racial disparities, palliative care, health services research, systems thinking, logistic regression, BMC Medicine

System dynamics is a modeling tradition that treats stubborn problems as products of feedback structures rather than single causes. In group model building, stakeholders gather in facilitated workshops and construct a map of the system together, naming the variables they believe matter and drawing arrows that show which factors influence which. The approach has a distinctive strength: it surfaces assumptions that participants may never have articulated, and it forces the group to confront how a factor that helps in one place can quietly undermine progress elsewhere. For a problem like cancer screening in people with intellectual disabilities, where general practitioners, screening organizations, family caregivers, and professional support workers all act at different moments, that relational view is essential.

The workshops brought together healthcare professionals and support workers involved in caring for people with intellectual disabilities in the Netherlands. Participants mapped the key variables influencing access to and participation in cancer screening, including the national programs for breast, cervical, and bowel cancer screening that rely on tools such as the faecal immunochemical test. The discussions during the sessions were then used to interpret and contextualize the emerging diagram. Crucially, the researchers did not stop at the workshop stage: an independent group of similar participants, healthcare professionals and support workers with comparable experience, validated the finished causal loop diagram during a structured assessment, giving the model a check on credibility that many qualitative maps lack.

The resulting model is organized around five thematic clusters that together trace the decision-making trajectory of a screening invitation. The first cluster concerns awareness of the screening invitation itself, because an invitation letter that arrives at a shared address, or that is opened by a caregiver rather than the person it names, may never become a topic of conversation. The second cluster addresses understanding of information, since standard screening materials are written at a level that assumes health literacy and abstract reasoning that many people with intellectual disabilities do not possess. The third cluster captures well-informed decision making, the point at which someone must weigh the benefits and discomforts of a test and arrive at a choice that is genuinely their own.

The fourth cluster covers access to participation, the practical machinery of making an appointment, arranging transport, preparing for a procedure, and physically getting through it. The fifth cluster is perhaps the most novel: system-level visibility and oversight, meaning whether anyone in the system can actually see who has been invited, who has decided, and who has fallen through the cracks. When no single actor holds oversight of the whole trajectory, a person who misses one step can disappear from view entirely, and no alarm sounds anywhere in the system.

A central finding of the study is the critical role of decision-making moments throughout the screening process. The model shows that these moments are not a single event but a chain of junctures, each one capable of ending participation. A person may understand an invitation but not the procedure; a caregiver may grasp the procedure but hesitate to influence a decision they believe belongs to the individual; a professional may support the decision but find that the screening organization cannot accommodate the extra time or adjusted communication the appointment requires. Each handoff between stakeholders is a point where momentum can be lost, and the diagram makes those losses visible as feedback loops rather than isolated failures.

This systems view has a direct implication for policy. Interventions that target only one variable, such as producing easier-to-read invitation letters, address a real need but leave the surrounding structure intact. If accessible information is not paired with support at the moment of decision, and with procedures that flex at the point of delivery, the loops that suppress participation will continue to operate through other variables. The authors argue that continuous support for decision making across the different stages of the screening process, and among all the stakeholders involved, is essential to improve access. That is a call for a multidisciplinary approach in which general practitioners, screening program staff, and disability support workers coordinate deliberately rather than incidentally.

The methodological contribution is as interesting as the substantive one. Group model building is still rare in disability health research, where the literature has largely accumulated lists of barriers through surveys and interviews. By converting those perspectives into a causal structure, the Dutch team produced something a list cannot: a testable hypothesis about how the system behaves. The 48 variables and their connections can now be examined, challenged, and refined, and the diagram can serve as a shared reference point for stakeholders who otherwise hold fragmented pieces of the same problem. The validation step, in which an independent group assessed the diagram, strengthens the claim that the model reflects a consensus understanding rather than the view of a single workshop.

The study also clarifies where responsibility currently falls. Support workers and family caregivers often absorb the invisible labor of screening, translating letters, explaining procedures, and accompanying people to appointments, yet they sit outside the formal screening system and receive little guidance or recognition for this role. Meanwhile, screening organizations typically see only whether an invitation was answered, not why it was not. Making the whole trajectory visible, the fifth cluster of the model, is therefore not an abstract ambition but a practical prerequisite: without oversight, no actor can tell whether an intervention worked, and without feedback, the system cannot learn.

For a problem that has persisted across decades of screening policy, the message of this research is quietly radical. People with intellectual disabilities are not failing to participate because they lack motivation or because any single gatekeeper blocks them. They are embedded in a system whose parts, each acting reasonably in isolation, collectively produce exclusion. The causal loop diagram built by Banda and colleagues gives clinicians, program designers, and caregivers a common map of that system for the first time, and it points to a concrete design principle: build support for decision making into every stage of the screening journey, and give the system eyes to see where people are lost. The study was funded by the Maarten van der Weijden Foundation, the Netherlands Organisation for Health Research and Development, and the Dutch Ministry of Health, Welfare and Sport, none of which influenced the design or interpretation of the work.

Subject of Research: System dynamics analysis of cancer screening access for people with intellectual disabilities

Article Title: Healthcare workers and caregivers’ perspectives on access to cancer screening for people with intellectual disabilities as revealed by system dynamics and group model building

Article References: Banda, A., Cuypers, M., Rouwette, E. A., Nägele, M., Timen, A., Leusink, G., & Naaldenberg, J. (2026). Healthcare workers and caregivers’ perspectives on access to cancer screening for people with intellectual disabilities as revealed by system dynamics and group model building. BMC Medicine. https://doi.org/10.1186/s12916-026-05294-w

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05294-w

Keywords: cancer screening, intellectual disability, system dynamics, group model building, causal loop diagram, decision making, health equity, caregivers, healthcare professionals, Netherlands, screening participation, BMC Medicine