A new study published in Acta Neuropathologica by Antonia Neubauer, Paul Feyen, Jochen Herms and colleagues at Ludwig-Maximilians-Universität München and the German Center for Neurodegenerative Diseases tackles that gap at unprecedented scale. The team systematically documented the abundance and morphological variety of amyloid-beta deposits in up to 23 brain regions of 84 cases of advanced Alzheimer’s disease, drawn from the Neurobiobank Munich. The cohort included both familial and sporadic cases, with an average age at clinical onset of 61.6 years and an average age at death of 73.9 years. All cases had died with dementia, and most showed the highest neuropathological stages of both tau and amyloid pathology. In total, the researchers extracted 57,233 individual amyloid-beta deposits from 1,179 annotated brain regions.

The technical achievement at the heart of the study is a machine-learning pipeline built for digital pathology. Formalin-fixed, paraffin-embedded tissue was cut into five-micrometer sections and stained with the monoclonal antibody 4G8 to visualize amyloid-beta, with diaminobenzidine producing the brown signal. Digitized slides were preprocessed by color deconvolution, and a random forest pixel classifier trained in ilastik detected candidate deposits. Each deposit was then classified by a convolutional neural network, an extension of an earlier architecture, into six classes: cored plaques, diffuse light plaques, diffuse dense plaques, compact plaques, small dense plaques, and cerebral amyloid angiopathy. The network was trained on 2,135 manually classified deposit images, validated on 533, and tested on a held-out set of 471, achieving a recall of 81.5 percent and a precision of 82.4 percent. Because cored plaques and vascular amyloid were rare and harder to call, every prediction in those two categories was manually verified and reclassified where necessary, a step that proved essential given the low precision the model initially achieved for cored plaques.

The regional map that emerged confirms and refines the classical staging of amyloid spread. Cortical regions carried by far the heaviest plaque loads, with median counts between roughly 81 and 104 deposits per square millimeter in areas such as the insula and frontal sulcus, while the occipital, parahippocampal and entorhinal cortex showed significantly lower densities. Subcortical regions, the hippocampus and the cerebellum were less affected still, with the dentate nucleus of the cerebellum essentially spared. Plaque counts correlated strongly within the cortical group, within the hippocampal group, and between cortex and hippocampus, hinting at shared mechanisms or connected pathways driving accumulation. Intriguingly, regions that are typically devastated by tau, alpha-synuclein and TDP-43 pathology, such as the amygdala and entorhinal cortex, were comparatively lightly affected by amyloid, a dissociation the authors suggest may reflect amyloid released from tau-bearing projection neurons accumulating at distant cortical targets rather than locally.

Across the brain, the plaque landscape was dominated by poorly differentiated deposits. Diffuse light plaques were the most abundant class in nearly every region, followed by small dense plaques and then diffuse dense plaques, with compact plaques in fourth place. Cored plaques and cerebral amyloid angiopathy played a statistically minor role, with median proportions near zero. Yet the correlation analyses revealed that these numerically rare classes behave differently from everything else. Densities of diffuse dense, small dense and compact plaques correlated positively with one another and with the total plaque load, suggesting shared origins or transitions. Cored plaques and vascular amyloid, by contrast, did not correlate with the overall amyloid burden at all, pointing to distinct mechanisms of formation that do not simply scale with the amount of amyloid in a region.

Perhaps the most striking finding concerns tau. When the researchers correlated each plaque class with the tau-covered area in the entorhinal cortex, only cored plaques showed a significant positive relationship. Every other class except vascular amyloid was negatively correlated with tau load. This is a provocative result, because cored plaques are known to overlap heavily with neuritic plaques, the variety surrounded by dystrophic neurites, activated microglia and reactive astrocytes, and the variety most consistently linked to synaptic loss and cognitive decline. The data suggest that while diffuse amyloid accumulates broadly and largely silently, the emergence of cored plaques may mark a closer pathological partnership with the tau process that ultimately drives neurodegeneration.

The study also tied plaque composition to patient characteristics in ways that varied by class. Women carried significantly higher densities of diffuse light, diffuse dense, compact and small dense plaques than men, but not of cored plaques or vascular amyloid, meaning that the well-documented sex difference in amyloid burden does not extend uniformly across plaque types. Carriers of at least one ApoE4 allele, the strongest genetic risk factor for late-onset Alzheimer’s disease, showed higher densities of diffuse dense and small dense plaques, consistent with experimental evidence that ApoE4 promotes denser, more condensed plaque morphologies. Among familial cases, PSEN1 mutation carriers had more cored plaques than sporadic cases. Younger age at clinical onset and younger age at death were both associated with higher densities of diffuse light plaques, and younger age at death specifically with more cored plaques, findings partly attributable to the inclusion of familial cases in the cohort.

Disease duration added another layer of nuance. Longer illness was associated with accumulating diffuse light and diffuse dense plaques, while cases with short disease durations showed proportionally more cored plaques and vascular amyloid. The authors interpret this cautiously but suggest that diffuse deposits can accumulate over many years without immediately limiting life, whereas cored plaques and vessel-wall amyloid may be more tightly linked to the pace of clinical progression. An integrative linear mixed-effects model combining sex, ApoE status, age at onset and disease duration confirmed the association between female sex and diffuse light plaques and between younger onset and cored plaques, but also revealed how much remains unexplained: these factors accounted for only a modest share of the variance in plaque densities, leaving room for lifestyle, comorbidities and vascular factors not captured in the analysis.

The authors are candid about the limitations. The cohort reflects voluntary brain donation in a Western country, with limited ethnic diversity and an overrepresentation of familial disease. Sampling was uneven across subjects, though control analyses in a fully sampled subset suggested this did not distort the main conclusions. The two-dimensional nature of histological sections likely underestimates cored plaques, whose dense cores may simply fall outside the cutting plane, and deposits smaller than 100 square micrometers, along with soluble oligomeric amyloid, were excluded entirely. Causality, moreover, cannot be established from post-mortem correlations. Even so, the study delivers something the field has lacked: a cross-regional, class-resolved atlas of amyloid pathology in the human Alzheimer’s brain, with code and models publicly available for the community. Its central message is that amyloid is not one target but many, and that cored plaques and cerebral amyloid angiopathy, despite their rarity, stand apart statistically from the diffuse background in ways that may prove crucial for understanding, and one day treating, the disease.

Subject of Research: Machine-learning-based classification of amyloid-beta plaque diversity across brain regions in Alzheimer's disease and its clinicopathological associations

Article Title: Amyloid-β plaque diversity across brain regions in Alzheimer’s disease and its relation to major clinicopathological traits

Article References: Neubauer, A., Feyen, P., Pekrun, S., Roeber, S., Ruf, V., Strübing, F. L., & Herms, J. (2026). Amyloid-β plaque diversity across brain regions in Alzheimer’s disease and its relation to major clinicopathological traits. Acta Neuropathologica, 152(1), Article 46. https://doi.org/10.1007/s00401-026-03096-1

Image Credits: AI Generated

DOI: 10.1007/s00401-026-03096-1

Keywords: Alzheimer's disease, amyloid-beta, plaques, tau, cerebral amyloid angiopathy, ApoE4, convolutional neural network, neuropathology, digital pathology, machine learning, brain regions, plaque morphology

The core problem the authors identify is phenotypic heterogeneity. Sporadic Alzheimer’s disease accounts for roughly 90 percent of cases and is diagnosed clinically using cognitive testing, supported by amyloid and tau biomarkers in cerebrospinal fluid, blood plasma and PET imaging. But on the neuropathological level, the disease is far more complex than a simple case-versus-control division suggests. Amyloid plaques of extracellular beta-amyloid and intracellular neurofibrillary tangles of hyperphosphorylated tau are the hallmark lesions, yet comorbid pathology is the norm rather than the exception in the aging brain. Cerebral amyloid angiopathy, alpha-synuclein Lewy bodies, TDP-43 inclusions and granulovacuolar degeneration bodies frequently co-occur with hallmark Alzheimer lesions, sometimes at striking frequencies: TDP-43 inclusions can be found in up to 57 percent of Alzheimer’s patients and Lewy bodies in over 60 percent.

This complexity has direct consequences for genetic research. Twin studies estimate the heritability of sporadic Alzheimer’s disease at around 60 to 80 percent, and SNP-based heritability estimates rise from roughly 0.25 to 0.34 in clinical cohorts to 0.36 to 0.59 when autopsy or biomarker confirmation is used. More than 75 risk loci have now been associated with Alzheimer’s disease through GWAS, implicating pathways ranging from APP and tau processing to endocytosis, lipid metabolism and innate immunity. Yet clinical misclassification introduces type I and type II errors into large-scale studies, and some discovered signals may reflect co-pathologies or misdiagnosed subgroups rather than pure Alzheimer’s pathogenesis. The authors argue that smaller, deeply phenotyped cohorts can outperform ever-larger clinical samples because phenotypic accuracy is what limits statistical power, not sample size alone.

The review traces how this shift has already paid off for the hallmark lesions. Studies on the Religious Orders Study and Rush Memory and Aging Project (ROSMAP) cohorts, which follow cognitively normal individuals with longitudinal evaluation and brain donation, linked a CR1 intronic variant to increased neuritic plaque burden. Later work testing 83 risk variants from the largest clinical Alzheimer’s GWAS in a cohort of over 4,000 individuals found that BIN1, CR1 and COX7 variants increased neurofibrillary tangle burden while MME decreased it, and that BIN1, APH1B and FERMT2 variants raised neuritic plaque risk while MME, EED/PICALM and RBCK1 were protective. A genome-wide significant association between APOE and tangle pathology was repeatedly replicated, and an intronic PTPRD signal emerged in larger ROSMAP GWAS and was independently replicated.

Larger multicenter efforts extended these findings. The Alzheimer’s Disease Genetics Consortium performed GWAS of neuropathological phenotypes in nearly 5,000 individuals, confirming genome-wide significant APOE associations with all investigated lesions and identifying protective signals for GALNT7, ABCG1 and IER5I with neuritic plaques. A study of more than 7,000 autopsied participants from ROSMAP, NACC and ACT cohorts found APOE significantly associated with amyloid score, Braak stage and CERAD neuritic plaque score, plus a genome-wide significant BIN1 effect on higher Braak stages, with ABCA7, PTK2B, CR1, SNX1, FERMT2 and CELF/SPI1 raising plaque burden and PICALM, INPP5D, ZCWPW1 and SORL1 lowering it. Sex-stratified analyses added nuance, revealing a male-specific protective signal near TSPAN13 for tangle presence, though the authors caution that reduced statistical power in stratified analyses means such findings require independent replication.

Perhaps the most striking example of the approach’s value comes from primary age-related tauopathy, or PART, a condition in which tau lesions indistinguishable from Alzheimer’s occur without amyloid plaques and with limited cognitive consequences. A GWAS confined to neuropathologically confirmed PART cases uncovered a genome-wide significant locus on chromosome 4, with functional validation pointing to JADE1 as the plausible risk gene for tau pathology. Notably, the effect alleles in this locus were independent of earlier suggestive signals, illustrating how restricting cohorts to precisely defined pathological entities can reveal biology that overlapping, heterogeneously composed cohorts dilute or obscure.

Co-pathologies show equally illuminating genetic patterns. For cerebral amyloid angiopathy, found in up to 90 percent of Alzheimer’s cases, the APOE epsilon4 allele is specifically linked to capillary-involving Type I disease, while epsilon2 is more frequent in Type II. Beyond APOE, CR1 and UNC5C were implicated in hypothesis-driven studies, and a GWAS in over 800 individuals identified the long noncoding RNA locus LINC-PINT, associated with decreased CAA burden in APOE epsilon4-negative individuals and supported by in vitro evidence of neuroprotection. A protective APOC2 polymorphism in a large postmortem GWAS was later shown by colocalization to reflect reduced APOE expression in brain microglia, demonstrating how post-GWAS fine-mapping can redirect interpretation from the nearest gene to the true molecular driver.

For TDP-43 pathology and the related entity LATE (limbic-predominant age-related TDP-43 encephalopathy), much of the genetic insight has come from frontotemporal lobar degeneration research, where neuropathology-centered GWAS revealed TMEM106B, DPP6, UNC13A, HLA-DQA2 and TNIP1 as risk factors, with distinct loci emerging for specific FTLD-TDP subtypes. In the Alzheimer’s context, APOE epsilon4 allele count correlates with TDP-43 burden, and TMEM106B was confirmed as associated with LATE neuropathological change, alongside suggestive signals for GRN, SORL1 and TPCN1. Both TMEM106B, a lysosomal transmembrane protein whose C-terminal domain can form amyloid fibrils, and GRN, encoding the lysosomal growth factor progranulin, implicate lysosomal dysfunction as a shared mechanism spanning Alzheimer’s and TDP-43 proteinopathies. For Lewy body pathology, APOE and BIN1 are firmly established Alzheimer’s risk genes whose links to alpha-synuclein aggregates remain contested; one stratified GWAS concluded they raise Alzheimer’s risk regardless of Lewy body status, while BLMH and ZNF365 signals emerged for Lewy body presence in Alzheimer’s-negative individuals. Granulovacuolar degeneration, an underexplored lesion strongly tied to tau pathology, endocytosis, autophagy and necroptosis, showed APOE and APH1B associations that network modeling suggested are largely indirect, mediated through amyloid and tangle pathology.

The authors are candid about the caveats. Neuropathological cohorts are necessarily small because donated brain tissue is limited, forcing reliance on multi-center compilations that introduce inter-observer variability, differing staining protocols and harmonization challenges, though initiatives like BrainNet Europe have shown that immunohistochemistry with standardized antibodies and dichotomous diagnostic approaches yield the most robust results. Overlapping study populations across published GWAS also limit independence of findings, underscoring the need for new, standardized cohorts. Moreover, GWAS lead variants are often not the causal variants, and the nearest protein-coding gene is not necessarily the affected gene, since most hits sit in noncoding regulatory regions; gene prioritization and functional validation remain essential before mechanistic claims can be made.

Even so, the message of the review is unambiguous: bigger is not always better. Combining genetic data with detailed neuropathology enables refinement of clinical GWAS signals, discovery of variants tied to the burden of specific lesions rather than a diagnostic label, network analyses that position APOE’s effects on tau as largely indirect through amyloid and vascular pathology, and identification of overlapping genetic risk across disease boundaries. As the first disease-modifying Alzheimer’s therapies arrive, understanding which genes drive which lesions, and how co-pathologies shape disease trajectory, could enable the patient stratification that clinical trials increasingly demand, and point toward treatments addressing the full neuropathological spectrum rather than a single pathological feature.

Subject of Research: Neuropathology-centered genome-wide association studies of Alzheimer's disease and overlapping co-pathologies

Article Title: A new era of genome-wide association studies in the field of Alzheimer’s disease and overlapping co-pathologies: lessons learned from a neuropathology-centered approach

Article References: A new era of genome-wide association studies in the field of Alzheimer’s disease and overlapping co-pathologies: lessons learned from a neuropathology-centered approach. (n.d.). https://doi.org/10.1007/s00401-026-03085-4

Image Credits: AI Generated

DOI: 10.1007/s00401-026-03085-4

Keywords: Alzheimer's disease, genome-wide association studies, neuropathology, amyloid plaques, neurofibrillary tangles, cerebral amyloid angiopathy, TDP-43, Lewy bodies, APOE, TMEM106B, co-pathologies, LATE

The new research, conducted at the Unidade Corino de Andrade in Porto, Portugal, one of the world’s principal referral centers for the condition, followed 151 patients carrying the V30M mutation, the variant responsible for the large endemic focus of the disease in northern Portugal. All participants were systematically evaluated for a spectrum of central nervous system manifestations, including transient focal neurological episodes, headaches, seizures, stroke, and symptoms of cranial nerve dysfunction. The results, published in the Journal of Neurology, reveal that nearly one in three patients had experienced transient focal neurological episodes, brief spells of cortical dysfunction that can mimic small strokes, while close to a quarter reported recurrent headaches and nearly seven percent had suffered seizures.

The mechanism behind these symptoms is a form of cerebral amyloid angiopathy specific to transthyretin. Although current disease-modifying therapies, from liver transplantation to tafamidis and gene-silencing drugs, act on the liver to suppress or stabilize the circulating form of the protein, the choroid plexus inside the brain continues to produce its own transthyretin, which is secreted directly into the cerebrospinal fluid. This intrathecal production is essentially untouched by liver-directed treatments. Over years, the protein deposits in the leptomeningeal membranes and arteries, particularly around the brainstem and spinal cord, gradually stiffening and damaging the vessels that bathe the brain’s surface.

The study’s most striking finding concerns the relationship between disease duration and brain symptoms. Transient focal neurological episodes were virtually absent in patients within their first five years of illness, but their prevalence climbed steadily with time, reaching more than seventy percent among those living with the disease for over two decades. When researchers dichotomized the cohort at fifteen years of disease duration, the difference was stark: fewer than eight percent of patients with shorter disease duration reported a history of these episodes, compared with just over half of those with longer disease. In statistical models, disease duration greater than fifteen years carried nearly a sixfold increase in the odds of having experienced such episodes, an association that remained robust after adjustment for age, sex, treatment category, and disease severity.

These episodes themselves are clinically distinctive and troubling. Most patients reported recurrent attacks, with a median of ten episodes per person, lasting around fifteen minutes. Language dysfunction was the most common manifestation, followed by sensory disturbances, and the majority of patients experienced more than one type of symptom. Crucially, most episodes involved a temporary loss of function, such as weakness or numbness, rather than positive phenomena like tingling or jerking. This makes them nearly impossible to distinguish from transient ischemic attacks on clinical grounds alone, a diagnostic dilemma the researchers emphasize repeatedly. Because recurrence is the main differentiating feature, the team advises that any first episode should prompt a full stroke work-up, including brain magnetic resonance imaging whenever feasible.

Yet the picture may differ from the cerebral amyloid angiopathy caused by amyloid-beta, the protein implicated in Alzheimer’s disease and sporadic cerebral amyloid angiopathy. In that condition, transient focal neurological episodes are strongly linked to superficial siderosis, a chronic accumulation of blood products on the brain’s surface, and to a high risk of devastating lobar hemorrhages. In transthyretin amyloidosis, neuropathological studies from Portugal have found superficial siderosis in only a small minority of cases, and brain imaging studies of patients with these episodes, including those from the present cohort, have frequently shown no hemorrhagic or ischemic lesions despite clear evidence of amyloid deposition on PET scans. This raises the possibility that the transient episodes arise from a different mechanism, perhaps cortical spreading depression or focal seizure activity, rather than from the vessel rupture seen in other amyloid conditions.

The study also broadened the recognized symptom spectrum beyond transient episodes. Nearly forty percent of patients reported symptoms attributable to cranial nerve dysfunction, most commonly reduced smell, followed by hearing loss and tinnitus. On bedside examination, objective hearing loss was detected in almost a quarter of participants and was associated with both older age and longer disease duration, findings that align with the dense amyloid deposition these patients often show around the brainstem, where cranial nerves exit. Headaches were also frequent, though the overall prevalence, including migraine-type headache, appeared similar to that of the general population. Symptomatic stroke, by contrast, was rare, occurring in only two patients, both with alternative explanations that may or may not have involved the underlying disease.

Therapeutic questions loom large over these findings. The researchers found that patients on tafamidis appeared less likely to report transient episodes in unadjusted analyses, but this association was heavily confounded by disease duration and treatment era, and it lost statistical significance after adjustment. Similarly, isolated case reports have described stabilization of central nervous system symptoms and improvement in leptomeningeal enhancement with gene-silencing therapies such as vutrisiran, despite the theoretical limitation of poor blood-brain barrier penetration. Notably, tafamidis has been detected in the cerebrospinal fluid at concentrations capable of partially stabilizing transthyretin, and higher doses approved for cardiac disease show stronger effects in laboratory assays. But no clinical trial has ever included central nervous system outcomes as endpoints, leaving the true effect of every available therapy on the brain essentially unknown.

The study also tested whether plasma neurofilament light chain, a widely used blood marker of axonal injury, could serve as a surrogate for brain involvement. The answer, at least for now, appears to be no. Baseline levels did not differ between patients with and without a history of transient episodes, and longitudinal measurements over twelve months showed no divergence in trajectory between the two groups. The researchers suggest that the substantial peripheral nerve damage characteristic of the disease may drown out any small central signal in the bloodstream, and they propose that cerebrospinal fluid neurofilament measurements might prove more informative. Alternatively, axonal injury may simply not be the central mechanism driving the brain complications.

The authors caution that their cohort consisted almost entirely of patients with early-onset V30M disease seen at a single center, and that survival bias cannot be excluded, since patients who died of catastrophic strokes would not have been recruited. Recall bias in self-reported transient symptoms is another limitation, though the researchers cross-checked patient accounts against medical records. Still, the consistency of the transient episode frequency with earlier studies lends credibility to the central conclusion: central nervous system assessment should become a routine component of follow-up care for patients with longstanding hereditary amyloidosis. As disease-modifying therapies extend survival, more patients are living long enough to enter the window in which brain involvement emerges, and the field urgently needs longitudinal studies, validated biomarkers, and clinical trials designed to answer whether current drugs can protect the brain as effectively as they protect the heart and the peripheral nerves.

Subject of Research: Central nervous system manifestations in hereditary transthyretin amyloidosis

Article Title: Central nervous system manifestations in hereditary transthyretin amyloidosis: a cross-sectional study with 151 patients

Article References: Sousa, L., Coelho, T., Fernandes, J., Cardoso, M., Alves, C., da Silva, A. M., Sousa, A. P., Pinto, M. M., Pereira, R. M., Ruano, L., Lemos, C., Baldeiras, I., Reis, C., & Taipa, R. (2026). Central nervous system manifestations in hereditary transthyretin amyloidosis: a cross-sectional study with 151 patients. Journal of Neurology, 273(10), Article 611. https://doi.org/10.1007/s00415-026-14146-9

Image Credits: AI Generated

DOI: 10.1007/s00415-026-14146-9

Keywords: hereditary transthyretin amyloidosis, ATTRv amyloidosis, central nervous system, cerebral amyloid angiopathy, transient focal neurological episodes, V30M mutation, tafamidis, gene silencing therapy, neurofilament light chain, transthyretin, Journal of Neurology, Porto

The study under scrutiny, led by Banker and colleagues and published online in December 2025, examined patients with lobar intracerebral hemorrhage, a form of bleeding stroke that occurs in the lobes of the brain rather than deep structures such as the basal ganglia or thalamus. Lobar hemorrhages are strongly associated with cerebral amyloid angiopathy, an age-related condition in which amyloid protein deposits weaken vessel walls, and they carry a particularly high risk of subsequent seizures. Because clinical seizures can worsen outcomes after hemorrhagic stroke, physicians have long debated whether to prescribe preventive antiseizure drugs even in patients who have never seized, despite growing concern that these medications may themselves be harmful.

That concern was echoed in 2025 clinical practice guidelines from the Neurocritical Care Society, authored by Frontera and colleagues, which recommended against routine prophylactic antiseizure medication for most patients with nontraumatic intracerebral hemorrhage, citing trials suggesting possible harm, particularly with older agents such as phenytoin. Against this backdrop, observational studies comparing treated and untreated patients carry enormous weight. If such studies are methodologically sound, they can suggest when cautious prophylaxis might still help selected patients; if they are flawed, they can mislead clinicians and guideline developers alike.

The letter by Kobayashi and colleagues centers on the mechanics of immortal time bias, a problem first articulated in detail by epidemiologist Samy Suissa in a landmark 2008 paper in the American Journal of Epidemiology. The bias arises whenever a study classifies patients by a treatment that can only be started after they have survived for some period, yet counts that survival period as if it belonged to the treated group. Patients who die early, before ever receiving the drug, are automatically assigned to the untreated comparison group, even if they would have received the medication had they lived longer. The treated group thus appears to be populated by people who, by construction, could not die during the time window in question. They are, in a literal sense, immortal within that interval, and any apparent survival advantage may reflect nothing more than the way time was allocated.

In the context of lobar intracerebral hemorrhage, the problem is especially acute because the timing of antiseizure medication initiation is tightly linked to disease trajectory. Sicker patients, those with larger hemorrhages, deeper depressions of consciousness, or impending withdrawal of life-sustaining therapy, often die or transition to comfort care before prophylaxis is ever considered. Survivors who stabilize, by contrast, are the ones who live long enough to be started on medication. If the original analysis grouped patients from the moment of admission according to whether they ultimately received prophylaxis, while allowing the days before treatment to count toward the treated group’s follow-up time, the comparison would systematically favor the treated cohort for reasons unrelated to any pharmacological benefit.

Recent work has clarified just how pervasive and structurally varied this bias can be. In 2025, Hernán, Sterne, Higgins, Shrier, and Hernández-Díaz published a formal structural description in the journal Epidemiology showing how different causal structures generate immortal time, emphasizing that the bias is not a statistical nuisance to be adjusted away but a design-level misalignment between the question being asked and the analysis being performed. Their companion concept, the target trial, introduced by Hernán and colleagues in the Journal of Clinical Epidemiology in 2016, urges observational researchers to first specify the randomized trial they would ideally conduct, including a clearly defined time zero at which eligibility and treatment assignment both begin, and then emulate that design as faithfully as the data allow.

Applying that framework to seizure prophylaxis after lobar hemorrhage reveals the core difficulty: treatment is not assigned at a single moment but initiated at heterogeneous times, often days into the hospital course. Modern methods exist to handle this. The clone-censor-weight approach, increasingly used in pharmacoepidemiology and described in accessible detail by Gaber and colleagues in Cancer Medicine in 2024, creates artificial copies of each patient at time zero, one assigned to each treatment strategy, and then censors the copies when they deviate from their assigned strategy, using inverse probability weighting to correct for the informative censoring that results. This preserves the integrity of time zero and prevents the spurious survival advantage that plagues naive comparisons.

The exchange between the letter writers and the authors of the original study, with a formal response also published in Neurocritical Care, illustrates a broader and increasingly visible shift in how the medical literature polices itself. Methodological letters were once relegated to statistical footnotes; today, as large observational datasets drive clinical guidelines in neurocritical care, they sit at the center of practice-defining debates. The authors of the letter report no funding and no conflicts of interest and disclose that language-model tools were used only to refine phrasing and check citation consistency, with all content verified by the authors themselves. Their intervention does not claim that prophylactic antiseizure medication is definitively harmful or beneficial after lobar hemorrhage; it claims something more fundamental, that the existing estimate cannot be trusted until the time-zero problem is resolved.

For clinicians, the practical message is one of humility: observational evidence suggesting that prophylaxis is safe or effective in lobar hemorrhage should be weighed carefully against the randomized-trial signals of harm that shaped current guidelines. For researchers, the letter is a template for how to interrogate pharmacoepidemiologic analyses in acute neurology, where treatment timing, mortality, and disease severity are deeply entangled. As big-data studies continue to fill the evidence gaps left by trials that are difficult to conduct in critically ill stroke patients, the battle against immortal time bias, fought in the pages of methodological letters like this one, may prove as consequential for patient care as any new drug.

Subject of Research: Immortal time bias in observational studies of prophylactic antiseizure medication after lobar intracerebral hemorrhage

Article Title: Addressing Immortal Time Bias in Estimating the Effect of Prophylactic Antiseizure Medication after Lobar Intracerebral Hemorrhage

Article References: Kobayashi, G., Shibata, T., & Shiraishi, A. (2026). Addressing Immortal Time Bias in Estimating the Effect of Prophylactic Antiseizure Medication after Lobar Intracerebral Hemorrhage. Neurocritical Care. https://doi.org/10.1007/s12028-026-02643-8

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02643-8

Keywords: immortal time bias, lobar intracerebral hemorrhage, antiseizure medication, seizure prophylaxis, neurocritical care, observational study design, target trial emulation, clone-censor-weight method, epidemiology, stroke, cerebral amyloid angiopathy, clinical guidelines

The blood–brain barrier is not a simple wall but a dynamic interface built from specialized endothelial cells, pericytes, astrocyte endfeet, and a dense basement membrane. Together these components tightly regulate which molecules and cells can enter the brain, maintaining the precise ionic and metabolic environment that neurons require. In Alzheimer’s disease, imaging and post-mortem studies have repeatedly shown that this barrier becomes leaky decades before overt dementia, allowing blood-borne proteins, fibrin, thrombin, and immune cells to infiltrate brain tissue. This vascular leakage is thought to amplify neuroinflammation, impair amyloid clearance, and accelerate neuronal injury. Yet the molecular switches that flip the barrier from protective to permissive in Alzheimer’s have remained poorly defined, particularly with respect to how genetic risk translates into physical barrier breakdown.

The new work focuses on apolipoprotein E, or APOE, a lipid-transport protein produced abundantly in the brain by astrocytes and microglia. The ε4 variant of APOE is carried by roughly one in four people and multiplies Alzheimer’s risk several-fold compared with the common ε3 variant. Carriers of ε4 show earlier and more pronounced blood–brain barrier leakage, pericyte loss, and cerebral amyloid angiopathy, but the intermediate steps connecting APOE4 expression to vascular failure have been elusive. Bhattarai, Yilmaz and their colleagues reasoned that APOE4 might reprogram astrocytes, the star-shaped glial cells that normally cradle blood vessels and help maintain barrier integrity, into a state that actively undermines the very interface they are supposed to support.

Using a combination of human brain tissue analysis, Alzheimer’s disease mouse models carrying humanized APOE variants, and single-cell molecular profiling, the team discovered that astrocytes in APOE4-bearing brains dramatically upregulate the production of fibronectin, encoded by the FN1 gene. Fibronectin is an extracellular matrix glycoprotein that is normally present at very low levels in the adult brain’s vascular basement membrane. In developing tissue and in peripheral wounds, fibronectin provides a scaffold for repair, but its accumulation in the mature brain vasculature appears to be anything but helpful. The researchers found that fibronectin deposits accumulate around brain capillaries and arterioles in Alzheimer’s disease, and that this accumulation is markedly greater in individuals and animals carrying the ε4 allele.

The mechanistic story that emerges from the study is one of a vicious cycle. Amyloid-beta peptides, which accumulate in Alzheimer’s disease and deposit in cerebral blood vessels as cerebral amyloid angiopathy, stimulate astrocytes to secrete fibronectin, and APOE4 amplifies this response. The excess fibronectin then remodels the vascular basement membrane, disrupting the molecular interactions that keep endothelial tight junctions sealed and pericytes anchored to their vessels. The result is a barrier that becomes progressively more permeable, allowing plasma proteins to leak into the brain parenchyma. Some of these leaked proteins, including fibrinogen, are themselves pro-inflammatory and can further activate astrocytes and microglia, perpetuating the cycle of vascular damage and neuroinflammation that characterizes the Alzheimer’s brain.

Crucially, the researchers did not stop at correlation. In experimental models, reducing fibronectin production or interfering with its deposition produced striking protective effects. Animals with diminished fibronectin signaling maintained better barrier integrity despite the presence of APOE4 and amyloid pathology, showing less vascular leakage, reduced inflammatory activation, and improved structural preservation of the neurovascular unit. These findings position FN1 not merely as a biomarker of vascular damage but as an active, druggable participant in the disease process. The authors highlight FN1 as a potential therapeutic target, a designation that carries real weight because fibronectin biology is already well understood pharmacologically, with existing tools and compounds capable of modulating fibronectin assembly and its interactions with integrin receptors.

The implications for Alzheimer’s drug development are significant. Most therapeutic efforts to date have targeted amyloid-beta and tau, the canonical protein pathologies of the disease, with recent anti-amyloid antibodies demonstrating the ability to slow cognitive decline, albeit modestly. Vascular-targeted approaches have attracted growing interest precisely because blood–brain barrier breakdown appears so early in the disease course and correlates strongly with cognitive impairment independent of plaque burden. If fibronectin sits at the junction of genetic risk, amyloid pathology, and vascular failure, then therapies aimed at curbing fibronectin deposition could protect the barrier in ε4 carriers, a population that represents a large fraction of Alzheimer’s patients and that responds differently to some existing treatments.

The study also reframes the role of astrocytes in neurodegeneration. Long viewed primarily as support cells, astrocytes are increasingly recognized as active regulators of brain health whose dysfunction can drive disease. The finding that APOE4 pushes astrocytes toward a fibronectin-secreting, matrix-remodeling state adds to a growing body of evidence that reactive astrocyte phenotypes are not uniform and that specific astrocyte outputs, in this case a single extracellular matrix protein, can have outsized consequences for the entire neurovascular unit. It also helps explain a long-standing clinical observation: ε4 carriers tend to show more hemorrhagic and vascular contributions to their dementia, and therapies that address only amyloid may leave this vascular component untreated.

As with any study, important questions remain. The precise molecular pathway by which APOE4 sensitizes astrocytes to amyloid-driven fibronectin expression will need to be mapped in detail, and the safety of chronically inhibiting a protein that also participates in normal tissue repair must be carefully evaluated. Translating findings from mouse models and human tissue into a therapy that can be tested in patients will take years. Nevertheless, the identification of astrocyte-derived fibronectin as a mediator of APOE4-driven blood–brain barrier dysfunction provides the field with a concrete, mechanistically grounded target that connects the strongest genetic risk factor for Alzheimer’s disease to one of its earliest and most consequential pathological events. For the millions of people carrying the ε4 allele, that connection may prove to be one of the most important discoveries in the vascular dimension of Alzheimer’s research.

Subject of Research: Astrocyte-derived fibronectin mediates APOE4-driven blood–brain barrier dysfunction in Alzheimer's disease

Article Title: Fibronectin mediates APOE4-driven blood–brain barrier dysfunction in Alzheimer’s disease

Article References: Fibronectin mediates APOE4-driven blood–brain barrier dysfunction in Alzheimer’s disease. (n.d.). https://doi.org/10.1038/s43587-026-01204-0

Image Credits: AI Generated

DOI: 10.1038/s43587-026-01204-0

Keywords: Alzheimer's disease, APOE4, blood–brain barrier, fibronectin, astrocytes, FN1, amyloid-beta, cerebral amyloid angiopathy, neurovascular unit, Nature Aging, therapeutic target, neuroinflammation

The Flemish mutation is an A692G substitution in the gene encoding the amyloid precursor protein, or APP. When APP is processed, the change corresponds to an A21G substitution within amyloid-β, the short peptide that aggregates into the plaques characteristic of Alzheimer’s disease. Carriers of this mutation develop a rare, early-onset form of the disease with two unusual pathological hallmarks: pronounced cerebral amyloid angiopathy, in which amyloid accumulates in the walls of cerebral blood vessels, and senile plaque cores that are unusually large. Affected individuals suffer from both progressive dementia and cerebral hemorrhages, a combination that distinguishes the Flemish variant from more common presentations of familial Alzheimer’s disease. Until now, the molecular reason for this vascular tropism remained obscure.

To understand the structural consequences of the mutation, the team turned to cryo-electron microscopy, a technique that images flash-frozen biological molecules at near-atomic resolution and has in recent years transformed the study of amyloid diseases. Filaments were extracted from parietal lobe tissue of the two affected individuals and their structures determined. Although tau paired helical filaments, the tangles that accompany most forms of Alzheimer’s disease, were present in the samples, the predominant filaments were built not from tau but from the amyloid-β peptide itself, specifically the Aβ40 species carrying the A21G substitution. The Flemish fold, it turned out, is written directly into the peptide sequence.

Structurally, the filaments are composed of two identical protofilaments, each spanning residues D1 through V40 of the amyloid-β peptide, packed together with two-start helical symmetry. This means that two protofilaments run in parallel along the length of the filament, twisting around each other in a helical arrangement. Comparative analysis revealed a striking continuity with previously known amyloid structures: the region of the peptide preceding the mutation site, residues Y10 through F19, adopts essentially the same substructure in both the Flemish fold and in wild-type Aβ42 filaments. The mutation’s effect is therefore not a wholesale remodeling of the peptide but a subtle yet consequential local change. The glycine substitution removes a single methyl group from residue 21, and the loss of that small hydrophobic moiety is enough to redirect the packing of the entire filament.

What emerges downstream of residue 21 is a completely different arrangement from all previously characterized Aβ folds. Since the first amyloid structures were determined, structural biologists have catalogued a growing repertoire of filament folds in Alzheimer’s disease and related disorders, each associated with different clinical presentations. The Flemish fold now joins this catalogue as a distinct entry, defined by its own hydrophobic interface, the interdigitating surface through which the two protofilaments and successive peptide layers stabilize one another. That an alteration of a single methyl group can generate a fold with no known counterpart underscores the exquisite sensitivity of amyloid assembly to sequence chemistry, and suggests that the growing library of familial APP mutations may each encode their own structural signatures.

The structural work alone could not explain why the Flemish fold associates with blood vessels. To address that question, the researchers employed a cell-based assay, and their results provide the study’s most intriguing mechanistic clue. The distinctive Flemish fold was found to be associated with the vascular tropism characteristic of this variant. In other words, the particular architecture of the filament, dictated by the A21G substitution, correlates with the propensity of amyloid to deposit in vessel walls rather than, or in addition to, the brain parenchyma. Cerebral amyloid angiopathy arises when amyloid-β, particularly the Aβ40 species, accumulates within the smooth muscle and basement membrane of leptomeningeal and cortical arteries, weakening them until they rupture. The Flemish findings now provide a structural vocabulary for that process: a fold that appears optimized, in some sense, for the vascular environment.

The implications extend beyond a single family. Dominantly inherited mutations in APP and in the presenilin genes have long served as guideposts for understanding Alzheimer’s disease more broadly, on the logic that if a mutation is sufficient to cause the disease, the pathway it disrupts must be central to the illness. The identification of a mutation-specific filament fold strengthens an increasingly influential idea in the field: that amyloid polymorphism, the existence of distinct filament structures, may correspond to distinct disease phenotypes. Previous cryo-EM studies have shown that different folds correlate with different dementias, including variants of Alzheimer’s disease and cerebral amyloid angiopathy with differing clinical courses. The Flemish fold, being tied to a specific pedigree and a specific clinical syndrome, offers one of the cleanest demonstrations yet that structure and phenotype are mechanistically linked.

The rarity of the Flemish pedigrees makes the achievement particularly notable. Only two such families are known worldwide, and obtaining postmortem brain tissue from affected members is exceptionally difficult. The fact that filaments from two independent individuals yielded consistent structures, with Aβ40-A21G as the predominant component, provides confidence that the Flemish fold is a reproducible biological entity rather than an artifact of a single brain. The presence of tau paired helical filaments alongside the amyloid filaments also confirms that the Flemish variant produces a bona fide Alzheimer-type pathology, complete with both plaques and tangles, even as its vascular features set it apart.

For the molecular understanding of Flemish-type dementia and cerebral hemorrhage, the study delivers what its authors describe as the definition of a familial Alzheimer-disease-associated amyloid fold. This framing matters because it establishes the Flemish fold as a structural unit of analysis in its own right. Future work can now ask mechanistic questions at atomic resolution: how the hydrophobic interface of the Flemish fold interacts with vascular basement membranes, whether the fold nucleates more readily on vascular surfaces, whether the faster clearance of Aβ40 from brain tissue funnels the peptide into vessels, and whether small molecules could be designed to destabilize the Flemish fold specifically. The loss of a single methyl group at residue 21, and the filament architecture it produces, has gone from a genetic curiosity to a defined structural target.

More broadly, the work exemplifies the maturing relationship between cryo-EM and neurodegenerative disease research. In the space of a decade, amyloid structures have progressed from generic models to fold-specific structures that distinguish diseases, mutations, and even individual patients. Each new fold added to the catalogue is a potential biomarker, a phenotypic explanation, and a drug-discovery template all at once. The Flemish fold demonstrates that even in a disease as intensively studied as Alzheimer’s, the structural landscape remains incompletely mapped, and that rare families, however small in number, can illuminate molecular mechanisms invisible in the common sporadic form of the illness. For the members of the two Flemish pedigrees and for the broader effort to understand cerebral hemorrhage in Alzheimer’s disease, the structure now provides a foundation on which diagnostic, mechanistic, and therapeutic advances can be built, one methyl group at a time.

The study, led by Yuhui Sha and Jun Ni, together with collaborators including Joanna M. Wardlaw of the University of Edinburgh, set out to systematically characterize symptomatic diffusion-weighted imaging lesions, known as sDWI lesions, and incidental DWI lesions, known as iDWI lesions, in patients with probable CAA. Diffusion-weighted imaging is an MRI sequence exquisitely sensitive to acute cellular injury: when blood flow to brain tissue fails and cells begin to swell with cytotoxic edema, water molecules lose their freedom to diffuse, and the affected region lights up brightly on the scan. In CAA, these bright spots have been the subject of a running debate. Autopsy work from previous groups, including studies by Ter Telgte and colleagues, has suggested that DWI-positive lesions in CAA correspond histologically to microinfarcts, confirming that they represent genuine ischemia rather than hemorrhage or artifact. Yet the clinical relevance, prevalence, and mechanisms of these lesions remained contested.

To resolve the question, the team drew on a prospective cohort of cerebral small vessel disease patients recruited at Peking Union Medical College Hospital between March 2017 and February 2026. Using the Boston criteria version 1.5, the standard clinicoradiological framework for diagnosing CAA without tissue biopsy, they enrolled 185 patients with probable CAA. Each participant underwent detailed baseline clinical assessment and multimodal MRI, including imaging markers of small vessel disease such as lacunes, cerebral microbleeds, and cortical superficial siderosis, the latter being a sign of chronic blood leaking over the surface of the brain. Ninety-nine of the patients returned for follow-up MRI, generating 163 scans in total, which allowed the researchers to estimate not only cross-sectional prevalence but also cumulative incidence over time.

The results were revealing in two directions. First, ischemia turned out to be anything but rare. Symptomatic DWI lesions were detected in 29 patients, while incidental DWI lesions, discovered on scans performed for other reasons, appeared in 49 patients, more than a quarter of the entire cohort. Kaplan–Meier analysis of the followed subgroup showed that over three years, the cumulative incidence of new symptomatic DWI lesions was 9.45 percent, with a 95 percent confidence interval of 0.67 to 17.46 percent, while the cumulative incidence of incidental lesions was considerably higher at 25.74 percent, with a confidence interval spanning 15.2 to 34.97 percent. In other words, roughly one in four CAA patients followed longitudinally was destined to develop a small ischemic lesion within three years, most of them without any clinical fanfare.

Second, and perhaps more striking, the two lesion types behaved like two different diseases. When the researchers constructed lesion probability maps to visualize topography, symptomatic DWI lesions clustered predominantly in deep brain regions, accounting for 72.7 percent of their locations. Incidental lesions, by contrast, were mainly found in cortical and juxtacortical regions, the ribbon of gray matter and the immediately underlying white matter, in 62.5 percent of cases. This anatomical split is not trivial. Deep and superficial cerebral territories are supplied by different vascular trees, with deep perforating arterioles and leptomeningeal cortical vessels subject to different hemodynamic stresses and pathological processes. A preferential distribution suggests a preferential mechanism.

That suspicion was confirmed by prospective Cox regression analysis, which adjusted for age and sex and, for the incidental lesions, also for traditional vascular risk factors and focal superficial siderosis. For symptomatic DWI lesions, the independent predictors were the number of traditional vascular risk factors, including hypertension, diabetes mellitus, coronary artery disease, and smoking, with a hazard ratio of 2.968 per additional risk factor, and the presence of lacunes, small old infarcts typical of hypertensive arteriopathy, with a hazard ratio of 1.187. Both reached conventional statistical significance, with P values of 0.013 and 0.016 respectively. This profile points toward the familiar machinery of atherosclerotic and hypertensive small vessel disease, superimposed on the amyloid-laden vasculature of CAA patients.

Incidental DWI lesions told a different story. Their independent predictors were lacunes, with a hazard ratio of 1.158, disseminated cortical superficial siderosis, with a hazard ratio of 2.994, and lobar cerebral microbleed grade, with a hazard ratio of 2.059. Notably, traditional vascular risk factors did not independently predict these silent lesions once adjustments were made. Disseminated siderosis and a heavy burden of lobar microbleeds are both hallmarks of advanced amyloid pathology, marking vessels that leak blood products into the subarachnoid space and brain parenchyma. Their association with cortical and juxtacortical ischemic lesions fits a mechanistic picture in which amyloid-laden leptomeningeal and cortical vessels, damaged by both amyloid deposition and hemosiderin-related injury, suffer episodic failure of perfusion, producing tiny cortical microinfarcts that the patient never notices.

The implications ripple outward in several directions. Clinically, the findings suggest that when a CAA patient presents with an acute neurological event, a DWI-positive lesion in deep structures is likely to be associated with conventional vascular risk factors and may warrant aggressive management of those factors. Conversely, incidental cortical lesions in a patient with disseminated siderosis may signal active amyloid-related vascular injury, which has been linked in prior work to increased risks of both future hemorrhage and cognitive decline. Earlier studies have shown that silent ischemic infarcts are associated with hemorrhage burden in CAA, and that DWI lesions after intracerebral hemorrhage predict recurrent stroke, so the new prospective risk factor data give clinicians a sharper tool for risk stratification.

Scientifically, the study reframes DWI-positive lesions as a dynamic window on small vessel disease activity. Traditional imaging markers such as microbleeds and siderosis are cumulative, recording injury that may have accumulated over years. DWI lesions, by contrast, are acute, appearing and resolving within days to weeks, and their signal characteristics evolve in a characteristic sequence as microinfarcts mature. Their detection rate in a longitudinal cohort therefore reflects ongoing disease activity, in the same way that incident troponin release reflects active myocardial injury. The authors propose that DWI-positive lesions could serve as a useful imaging marker of SVD activity and injury in future longitudinal studies and clinical trials of CAA. This matters because the CAA therapeutic pipeline is, for the first time, showing real signs of life. Immunotherapy trials with anti-amyloid antibodies such as ponezumab have been conducted, and a Phase 2 study of the RNA interference therapeutic ALN-APP, designed to reduce production of amyloid precursor protein, is currently recruiting patients with CAA. Trials like these need sensitive, quantifiable markers of target engagement and disease progression, and DWI lesion rates could fill that role.

There are, as always, caveats. The cohort of 185 patients, and the 99 with follow-up imaging, is modest by the standards of national registry studies, and the confidence interval around the three-year incidence of symptomatic lesions is wide, reflecting the relative rarity of these events. Probable CAA by Boston criteria is not histologically confirmed disease, although the criteria perform well in autopsy-validated settings. The observational design cannot definitively separate correlation from causation: lacunes may predict DWI lesions because they share upstream causes rather than because old infarcts directly provoke new ones. And MRI at typical clinical field strengths can miss very small cortical microinfarcts, meaning the true incidence of ischemia in CAA may be even higher than these figures suggest. Nevertheless, the consistency of the topographical and risk factor dissociation across analyses lends credibility to the central conclusion.

The broader message is a conceptual one. CAA, the study concludes, is not simply a hemorrhagic disease with an occasional ischemic afterthought. Cerebral ischemia is common, mechanistically patterned, and readable on standard clinical MRI. Symptomatic deep lesions track the burden of systemic vascular risk, while silent cortical lesions track the severity of amyloid vasculopathy itself, a dissociation that echoes recent work in CAA and CADASIL cohorts showing high prevalences of incidental DWI lesions with distinct clinical associations. For a disease whose diagnosis has historically hinged on documenting bleeding, the growing recognition that acute ischemic lesions can be counted, mapped, and followed over time offers both a warning and an opportunity: the amyloid-diseased vessel fails in more ways than one, and each failure mode now has its own signature that clinicians and trialists can learn to read.