This study was approved by the institutional ethics review boards of Saiseikai Shiga Hospital (No. 226) and Tokyo Saiseikai Central Hospital (No. 28-15), which were the core hospitals of this registry. The ethics committee waived the requirement for informed consent because of the anonymity of data.

The study utilized data from the Saiseikai Stroke Database. This database is a prospective registry of acute stroke patients collected within the first week after stroke onset. The Saiseikai Stroke Research Group is composed of 27 hospitals under the Social Welfare Organization Saiseikai Imperial Gift Foundation in Japan. The registry was operational from April 2013 to March 2021. Details and data from the Saiseikai Stroke Database can be found in Supplementary Material 1 and have been published previously [11].

Our study included cases of ischemic stroke due to BAD and LI registered in the Saiseikai Stroke Database between 2013 and 2021. We excluded patients younger than 16 years, those with incomplete discharge modified Rankin Scale (mRS) data, and patients diagnosed with stroke subtypes other than BAD or LI. Stroke diagnoses were categorized according to the TOAST (Trial of Org 10172 in Acute Stroke Treatment) criteria in this database [12]. In this database, BAD patients were recorded independently of atherothrombotic and LI, and they were diagnosed and enrolled by the stroke specialist at the institution. Generally, patients with intracerebral lesions with a diameter of ≥15 mm and more than three slices or lesions extending to the surface of the pontine base were defined as BAD, according to previous reports [3,13]. The diagnosis of LI generally followed the TOAST criteria, which includes the following: (1) lesion location in the subcortical hemisphere or brainstem; (2) size less than 1.5 cm as evaluated by computed tomography and/or magnetic resonance imaging (MRI); (3) no potential cardiac source of embolism; and (4) no stenosis greater than 50% in an ipsilateral artery.

The following clinical data were collected and analyzed from the Saiseikai Stroke Database: sex, age, medical history (hypertension, diabetes mellitus, dyslipidemia, chronic kidney disease, congestive heart failure, and previous stroke), smoking habits, ongoing antithrombotic therapy, level of consciousness on arrival (Japan Coma Scale [JCS]) [14,15], lesion vessel, acute phase medical treatment, intervention (mechanical thrombectomy), mRS at discharge, and in-hospital mortality. The JCS, a consciousness rating scale specific to Japan, is a reliable tool for assessing intracranial disease, and it can also be converted to the Glasgow Coma Scale (Supplementary Tables 1, 2) [16,17]. In this database, the diagnosis was classified according to the artery from which the dominant vessel of the lesion originated. Thus, the territories registered as BAD vascular lesions were determined to be the following: middle cerebral artery as lenticulostriate form artery, basilar artery as paramedian pontine artery, posterior cerebral artery as thalamoperforating artery, internal carotid artery as anterior choroidal artery, and anterior cerebral artery as Heubner artery [2,9].

The outcome of this study was poor functional outcome as assessed by mRS at discharge, as well as length of hospital stay. We defined good functional outcome as an mRS score of 0 to 2 and poor functional outcome as an mRS score of 3 to 6 [11].

In our study, we analyzed the characteristics and clinical data of patients in both groups (BAD group, LI group) utilizing the Wilcoxon rank sum test for numerical variables and Pearson chi-square test for categorical variables. We conducted logistic regression analyses for the groups to determine adjusted odds ratios (aORs) for BAD outcomes, accompanied by 95% confidence intervals (CIs), compared with LI outcomes. We identified potential confounders including sex, categorized age, previous medical history, categorized JCS scores (each digit), and vascular lesions. Multivariable logistic regression analyses were conducted to estimate aORs for BAD outcomes, including their 95% CIs. Point estimates with 95% CIs were used to present statistical results. In addition, a similar subgroup analysis was performed by classifying lesions into internal carotid artery, anterior cerebral artery, middle cerebral artery, basilar artery, and posterior cerebral artery vascular lesions. We determined statistical significance as a nonoverlap of the 95% CIs with the null effect value (OR, 1) or a two-sided P-value of <0.05. All statistical analyses were performed using JMP Pro 17 software (SAS Institute Inc).

Of the 27,691 patients registered in the Saiseikai Stroke Database, 5,966 patients were eligible for analysis: 1,549 (25.9%) in the BAD group and 4,434 (74.1%) in the LI group. The study flowchart is presented in Fig. 1. The background characteristics and clinical data of the patients are summarized in Tables 1 and 2. In the entire cohort, 3,579 (60.0%) were male, 1,349 (22.6%) were 64 years old or younger, 1,653 (27.7%) were 65 to 74 years old, and 2,964 (49.7%) were at least 75 years old. In the BAD group, 828 (53.6%) were at least 75 years old, and 1,108 (71.7%) had hypertension. Patient background information was generally similar in both groups.

Table 2 shows the clinical data including parent vascular territories in each group. Vascular territories were more common with middle cerebral artery (46.7%), 877 (56.8%) in the BAD group and 1,911 (43.2%) in the LI group. Table 3 shows medical treatment in the acute phase. Overall, tissue plasminogen activator was administered to 102 patients (1.7%), with 47 patients (3.0%) in the BAD group. Argatroban was administered in 790 cases (51.5%) in the BAD group and dual antiplatelet therapy (DAPT) was taken in 638 cases (41.3%) in the BAD group.

The overall poor functional outcome at discharge was 2,130 (35.7%), and the median length of a hospital stay was 14 days (interquartile range [IQR], 9–23 days). The distribution of outcomes in each group is shown in Table 4. A comparison of the percentage of mRS 0 to 6 at discharge in each group is shown in Fig. 2. BAD was associated with worse functional outcomes, with an aOR of 2.77 (95% CI, 2.42–3.17) compared to LI, indicating a higher likelihood of poor functional outcome (mRS score of 3–6) at discharge. The median hospital stay for BAD patients was 19 days (IQR, 11–29 days) compared to 13 days (IQR, 9–21 days) for LI patients, with a P-value of <0.01.

In this nationwide multicenter registry analysis, a comparison between BAD and LI patients revealed that BAD was associated with poorer functional outcomes and longer hospital stays. A trend towards intensive treatment of BAD with argatroban and DAPT also was observed in Japan.

This study had several advantages over previous investigations. First, its novelty lies in characterizing the epidemiology of BAD and the association of BAD with functional prognosis. While BAD was previously recognized to cause progressive paralysis, its prognosis, even with detailed adjustment for confounders, has rarely been discussed [5,13]. Although progress has been made in recent years in large-scale data studies using insurance databases, BAD is not typically classified under the International Statistical Classification of Diseases and Related Health Problems code. In addition, regional differences in cases between Asia and Western countries also may be a reason for the lack of such comprehensive epidemiological information [2]. Therefore, the use of a specialized registry as in the current study may be important [18]. We consider the clear association between BAD and prognosis, demonstrated even after adjusting for potential confounders in this study’s sample size, to be a key epidemiological finding. Second, we elucidated the status quo of aggressive treatment of BAD in Japan. Although treatment guidelines for BAD have not been established definitively, due to the progressive nature of the disease, it has been reported that aggressive treatment with not only tissue plasminogen activator but also multiple drugs is beneficial [10,19,20]. In this database, multidrug therapy combining argatroban and DAPT was administered more frequently compared to the LI group, reflecting the above reports. BAD was considered to result from an atherothrombotic mechanism at the branch vessel. Given its pathogenesis, antiplatelet agents are the initial treatment, and early multimodal use of antiplatelet agents has been reported to be effective [10]. In summary, this study adds to the epidemiological information on BAD and clarifies the reality of treatment in Japan, but further research is needed to determine whether aggressive treatment is effective.

The prognostic impact observed in the results of this study may be reasonably attributed to the extent of early paralysis and early neurological deterioration, as previously reported [2,13]. In BAD, a relatively large-diameter (700–800 μm) perforating artery is stenosed or occluded by an atheromatous lesion near the orifice of the parent artery, resulting in infarction of the entire perforating artery. As an indicator of its pathology, it has been reported that hyperintense lesions identified on diffusion-weighted MRI gradually enlarge and show progressive motor deficits [21]. Early neurological deterioration is associated with a poor prognosis, and we postulate that the clinical nature of BAD as such contributes to its poor prognosis [22].

The findings of this study may be useful in detecting and diagnosing BAD and predicting prognosis. In particular, they emphasize the importance of identifying and suspecting LI at an early stage in the emergency department and linking the patient to a stroke specialist, with subsequent accurate diagnosis and prognosis. Emergency departments receive a large number of stroke patients, more than 70% of whom have ischemic stroke, suggesting that the number of potential BAD patients is large [23]. No clear trend was observed between BAD and LI in patient demographics, which is consistent with the findings of previous studies [7]. This emphasizes the need for greater emphasis on a radiographic approach in the identification of BAD, and early referral to a stroke specialist is desirable. Information on the prognosis of BAD may be useful in explaining the prognosis to patients and their families after diagnosis. The longer duration of treatment and poorer functional prognosis for BAD compared to LI may assist health care providers in treatment planning, including rehabilitation.

The present results also suggest the need for intensification of treatment strategies. Although there is no clear evidence that intensified antithrombotic therapy improves the prognosis of BAD, the progressive nature and high risk of BAD may justify early consideration of intensified antithrombotic therapy. Therefore, we believe that establishing a treatment protocol that includes multiple agents is critical to improving the quality of care in the emergency department and subsequent stroke specialist consultation.

The present study had several limitations, including the lack of initial US National Institutes of Health (NIH) Stroke Scale data, potential diagnostic discrepancies among institutions, and the limited generalizability due to the predominantly Asian population. First, because the NIH Stroke Scale was not included in this database, it is impossible to measure the extent to which initial neurological severity, apart from level of consciousness, influences prognosis. Second, there were concerns about the external validity of the study results for global application because BAD is thought to be more common in the Asian population. Third, there is the possibility of selection bias. Although there is some consensus on the diagnosis of BAD, as described in the Methods section, there was not a fully standardized and elaborated diagnostic protocol in this study. It is possible that some difficult-to-define cases were diagnosed differently between institutions. There may also have been a small number of cases in which the disease progressed to the point where the patient was retrospectively judged to have BAD. This raises concerns about selection bias. However, because Japan has the largest number of MRIs in the world and because diagnostic imaging is homogenized in Japan, we do not believe that there is a significant effect.

In conclusion, BAD presented worse functional outcomes and longer hospital stays than LI, indicating the need for early treatment and tailored rehabilitation plans to improve outcomes.