In this large population-based cohort study, patients with ICD-defined NMO had a clearly elevated risk of CBD, including ischemic stroke, hemorrhagic stroke, and TIA, among which hemorrhagic stroke posed the greatest risk. However, NMO did not increase the likelihood of CVD, including myocardial infarction, atrial fibrillation, coronary artery disease, and heart failure.

NMO has been reported worldwide and is associated with poor prognosis. NMO relapses typically worsen progressively over several days before reaching the peak clinical deficit and then gradually improving over the subsequent weeks or months. However, recovery is often incomplete, and many patients experience early cumulative disability due to frequent and severe relapses (20).

Although astrocytes maintain BBB integrity, polymorphonuclear leukocytes—not astrocyte loss—primarily drive BBB disruption. Consequently, leukocyte depletion restores BBB integrity, preventing astrocyte damage and permitting their repopulation (21). A recent study on neuromyelitis optica spectrum disorder (NMOSD) indicated that increased BBB permeability allows circulating AQP4 autoantibodies to infiltrate the central nervous system. The study also highlighted the role of glucose-regulated protein autoantibodies as BBB-reactive agents that contribute to antibody-induced BBB dysfunction (12).

The risk of CBD in patients with NMO remains insufficiently understood and investigated. Therefore, we examined whether CBD risk varied between individuals with ICD-defined NMO and a matched population without NMO in Taiwan. We observed an increased risk of various CBDs in patients with NMO, including ischemic stroke (aHR = 3.32), hemorrhagic stroke (aHR = 3.90), and TIA (aHR = 2.21), among which the highest risk was noted for hemorrhagic stroke. Previous pathological studies have demonstrated thickened and hyalinized small blood vessels and perivascular inflammation characterized by the deposition of IgG and complement within demyelinating lesions in NMOSD (22). Pathologies affecting small vessels can cause both ischemic and hemorrhagic outcomes. Moreover, inflammatory vascular markers associated with endothelial dysfunction are elevated in NMOSD during acute relapses, similar to observations reported in MS (23, 24). A case of NMOSD with recurrent intracranial hemorrhage was previously reported, suggesting a potential link between NMOSD and cerebellar vascular dysfunction (25). Several cases of acute ischemic or hemorrhagic stroke have also been reported during intravenous high-dose steroid treatment (14, 15). In a Korean cohort study, stroke risk was higher in patients with NMOSD than in matched controls (16).

From a mechanistic perspective, NMOSD is characterized by systemic inflammation and immune-mediated astrocytopathy, in which pathogenic AQP4-IgG binds to astrocytic AQP4 and promotes complement-dependent astrocytic injury, a central pathological feature of AQP4-IgG–positive NMOSD (26, 27). These inflammatory pathways, together with complement activation, may contribute to vascular pathology by promoting endothelial dysfunction, vascular inflammation, and subsequent vascular injury (28). In addition, proinflammatory cytokines and activated immune cells, including neutrophils, may promote a prothrombotic state through multiple mechanisms, including upregulation of tissue factor expression, impairment of endogenous anticoagulant pathways, and dysregulation of fibrinolysis (29). Neutrophil activation and the formation of neutrophil extracellular traps (NETs) may contribute to thrombosis by promoting platelet aggregation and activation of the coagulation cascade (30), thereby amplifying thromboinflammatory processes and facilitating thrombogenesis in NMOSD. These processes may collectively increase susceptibility to thrombotic events. Consistent with this biological plausibility, a nationwide cohort study from South Korea reported an increased risk of stroke among patients with NMOSD (16).

The relationship between CVD risk and NMO remains poorly understood and underexplored. To address this gap, we examined the association between NMO and CVD risk using data from Taiwan’s NHIRD. Patients with ICD-defined NMO in our cohort did not have a higher risk of myocardial infarction, atrial fibrillation, coronary artery disease, or heart failure. By contrast, in a previous cohort study conducted in Korea, the risk of myocardial infarction was elevated in both MS and NMOSD and remained similar between both conditions (17). Thus, the association between NMO and CVD risk remains unclear, highlighting the need for further investigation in large-scale studies.

In our sample, patients with certain comorbidities, including diabetes mellitus, hypertension, hyperlipidemia, and anxiety disorder, exhibited an increased risk of stroke. Additionally, stroke risk was higher among male and older patients. A study analyzing data from a U. S. insurance claims database revealed that several comorbidities, including hypertension and diabetes mellitus, were more common in patients with NMO than those without NMO (31). In another study of patients with NMO, elevated serum levels of low-density lipoprotein cholesterol were independently and positively associated with disease relapse (32). Patients exhibited a higher risk of developing diabetes mellitus than those with MS. However, this increased risk was primarily associated with prolonged or frequent steroid use, rendering the directness of the connection to NMO pathogenesis unclear. Severe disability in NMO has also been correlated with an elevated risk of diabetes mellitus (33). Previous studies indicated that autonomic dysfunction is commonly observed in both MS and NMOSD, involving cardiovascular, thermoregulation, and fatigue-related symptoms (34, 35). Moreover, NMOSD-related disability and decreased mobility may exacerbate risk factors and promote a sedentary lifestyle, thereby increasing stroke risk. This phenomenon may explain why a higher incidence of stroke and stroke-related complications has sometimes been observed in patients with NMOSD. Additionally, depression, anxiety, and sleep disturbances are highly prevalent among individuals with NMOSD (36, 37).

In our sample, CBD risk was significantly associated with ICD-defined NMO severity, and the highest risk was observed in patients with severe NMO. CBD risk was elevated in both hospitalized and nonhospitalized patients with NMO, although hospitalized patients exhibited a higher risk. A previous NMOSD study pointed out that most NMOSD patients experience frequent relapses that cause more severe, longer-lasting deficits than MS. Managing these flare-ups typically requires hospitalization and immunomodulatory therapy. However, readmission rates peak shortly after discharge, with half occurring within 13 days—primarily driven by neurological complications (50.2%) (38, 39).

Our findings revealed a higher risk of various CBDs in patients with ICD-defined NMO. However, patients with ICD-defined NMO in our sample did not have an increased risk of CVDs, such as myocardial infarction, atrial fibrillation, coronary artery disease, or heart failure. Studies have suggested that AQP4 antibodies play a key role in NMOSD-related BBB dysfunction (12, 40, 41). In turn, antibody-mediated BBB impairment in NMOSD (14) may make patients with NMOSD more susceptible to CBD than CVD. The present study found that patients with ICD-defined NMO have an increased risk of CBD, including ischemic stroke, TIA, and hemorrhagic stroke, which presents the highest risk. However, ICD-defined NMO does not appear to elevate CVD risk, including the risk of myocardial infarction, atrial fibrillation, coronary artery disease, and heart failure. Our findings highlight the need for medical professionals to recognize the possibility of acute macrovascular events when patients with ICD-defined NMO experience the sudden onset or worsening of new or existing symptoms.

However, these mechanistic interpretations should be approached with caution. The present study was based on administrative claims data from the NHIRD, which does not include biomarker, imaging, or laboratory information (e.g., inflammatory markers, coagulation profiles, or neuroimaging findings) necessary to directly evaluate these pathways. Therefore, the proposed mechanisms remain hypothetical and should be considered hypothesis-generating rather than confirmatory. Importantly, without serological data regarding AQP4-IgG and MOG-IgG status, we could not distinguish between contemporary definitions of NMOSD and MOGAD, nor could we further categorize patients by their specific clinical phenotypes (such as the presence of optic neuritis, longitudinal extensive transverse myelitis, or specific cerebral lesions). Given that NMO and MOGAD are now recognized as distinct immunological entities with different target cells, our reliance on historical, ICD-defined cohorts inherently introduces population heterogeneity. Therefore, the proposed macrovascular mechanisms remain hypothetical and should be considered hypothesis-generating rather than confirmatory.

The lack of smoking and BMI data might lead to an overestimation of the risk (positive bias), as these factors are prevalent in patients with chronic inflammatory conditions. To address the potential detection bias, we utilized relevant clinical diagnoses as surrogate variables. For example, COPD was used as an indicator of heavy smoking, while hypertension, dyslipidemia, and diabetes were included to address metabolic risks. As highlighted in Tables 2, 3, these comorbidities emerged as significant predictors of CVD and CBD, demonstrating that our model effectively accounted for the influence of established risk factors. Furthermore, adjustments were made for the CCI score and key comorbidities, which reflect both the patients’ overall health status and their interaction frequency with the healthcare system.

This population-based cohort study has several notable strengths. First, we employed a nationwide, population-based design, selecting patients from the entire Taiwanese population and monitoring them over an extended follow-up period. This methodology provided a large, representative sample with substantial statistical reliability and reduced the selection bias prevalent in observational research. Additionally, the large sample size allowed for subgroup stratification during statistical analysis, enabling us to thoroughly evaluate the effects of NMOSD on the risk of macrovascular events. Second, in Taiwan, insurance claims for in-hospital treatments are stringently monitored and audited under the NHI system. This rigorous surveillance program improves the reliability of diagnoses derived from insurance claims. Third, all patients with NMOSD and age- and gender-matched controls were selected from a nationally representative dataset.

This study also has several limitations. Therefore, certain confounding variables may not have been properly controlled for in this study. First, we were unable to collect data on factors associated to the risk of macrovascular events, including smoking status, alcohol use, body mass index, physical activity, personal medical history, disease activity, and disease duration. Second, a major limitation inherent to the claims-based nature of the NHIRD is the lack of detailed laboratory, serological, and neuroimaging data. Consequently, we could not classify patients by their immunological subtypes (e.g., AQP4-IgG vs. MOG-IgG status) or further categorize them by clinical phenotypes, such as the presence or absence of optic neuritis, longitudinally extensive transverse myelitis (LETM), or specific cerebral lesions. Crucially, our study period spanned from 2003 to 2020, capturing patients based on ICD-9-CM and ICD-10-CM administrative codes before the formulation of the 2023 international diagnostic criteria for MOGAD. Therefore, our cohort reflects an administrative and historical definition of NMO/NMOSD, which likely contains individuals who would currently be diagnosed with MOGAD rather than NMOSD. Nevertheless, prior epidemiological studies in Asian populations indicate that AQP4-IgG seropositivity accounts for the vast majority (approximately 70–80%) of cases historically coded as NMO, whereas MOGAD represents a small fraction (42). Although this heterogeneity is likely non-differential for macrovascular outcomes, whether distinct clinical presentations carry varying risks remains unclear. Future prospective cohorts incorporating precise clinic-immunological, imaging, and molecular classifications are warranted to confirm these associations.