Impact of Cerebral Small Vessel Disease on Functional Decline: Bleeding With Antithrombotic Therapy 2 Study

Clinical Question (PICO)

Population: Adults with cerebrovascular or cardiovascular disease receiving oral antiplatelet or anticoagulant therapy and undergoing standardized baseline MRI. Exposure/comparator: Total cerebral small vessel disease (SVD) score 0–4, based on cerebral microbleeds, confluent white-matter hyperintensity, enlarged basal-ganglia perivascular spaces, and lacunes; the primary contrast was severe SVD (scores 3–4) versus score 0. Outcome: Functional decline, defined as an increase of at least 1 point in modified Rankin Scale (ΔmRS ≥1) from registration to 24 months. Prespecified analyses examined major bleeding and ischemic events as time-dependent mediators/covariates, plus sex-stratified and sensitivity analyses.

Bottom Line

In this prospective Japanese registry, SVD burden was associated with functional decline beyond overt bleeding or ischemic events. Decline occurred in 13.2% of patients with SVD score 0 versus 28.1% with score 3 and 28.2% with score 4; after adjustment, scores 3 and 4 were associated with roughly 1.6-fold higher odds of ΔmRS ≥1 (aOR 1.65 and 1.62, respectively).

Design

Trial type: Investigator-initiated, prospective, multicenter observational cohort; open-label/noninterventional registry, with centrally adjudicated MRI features.
N: 4,765 analyzed of 5,378 enrolled; no randomization.
- SVD score 0: 857 (18.0%)
- SVD score 1: 1,302 (27.3%)
- SVD score 2: 1,182 (24.8%)
- SVD score 3: 934 (19.6%)
- SVD score 4: 490 (10.3%)
Randomization: None; exposure groups were defined by baseline MRI SVD score.
Setting: 52 hospitals in Japan within the Network for Clinical Stroke Trials.
Enrollment: 2016–2019; this analysis was published in 2026.
Mean follow-up: 24 months, with mRS assessments at registration and approximately 6, 12, 18, and 24 months.
Analysis: Multivariable logistic regression adjusted for demographics, premorbid mRS, vascular risk factors, antithrombotic therapy, MRI field strength, and bleeding/ischemic events in four predefined time periods; 4-way causal mediation analysis; prespecified sensitivity analyses. Missing data were not imputed because overall missingness was reported as <1% for the study outcome.
Primary outcome: Increase in mRS from baseline to final follow-up (ΔmRS ≥1). Secondary outcomes included mRS 3–6 at 24 months and all-cause death.

Population

Inclusion Criteria

Patients with symptomatic or asymptomatic cerebrovascular or cardiovascular disease who newly started or continued an oral antiplatelet agent or anticoagulant.
Baseline clinical data, standardized MRI obtained from 90 days before to 14 days after registration, and mRS at registration and final follow-up.

Exclusion Criteria

Contraindication to MRI or failure to undergo MRI, duplicate registration, incomplete baseline clinical data, absent baseline antithrombotic information, absent follow-up, or missing baseline/final mRS data.
Of 5,378 enrolled, 613 were excluded from this subanalysis, mainly because of missing clinical or imaging data.

Baseline Characteristics (overall or representative cohort)

Median age 73 years (IQR 66–79); 1,573 women (33.0%).
Baseline mRS 0 (IQR 0–1).
4,268 patients (89.6%) had prior ischemic stroke and were receiving antithrombotic therapy for secondary prevention; 497 (10.4%) were treated for primary stroke prevention or cardiovascular disease.
Higher SVD scores tracked with older age, higher premorbid mRS, hypertension, chronic kidney disease, and prior stroke/hemorrhage/acute coronary syndrome.

Interventions

SVD score 0–4: No assigned treatment; each point represented one MRI feature: at least one cerebral microbleed, confluent white-matter hyperintensity, at least 11 enlarged basal-ganglia perivascular spaces, or at least one lacune.
Background therapy: Clinically selected oral antiplatelet agents or anticoagulants; 15.1% of participants were taking cilostazol. Treatment was not randomized.

Outcomes

Primary Outcome (increase in mRS of ≥1 point by 24 months):

SVD score 0: 13.2%.
SVD score 1: 16.1%.
SVD score 2: 20.7%.
SVD score 3: 28.1%.
SVD score 4: 28.2% (trend across groups P<0.001).
Compared with score 0, score 3 was associated with ΔmRS ≥1 after adjustment (aOR 1.65, 95% CI 1.27–2.15), as was score 4 (aOR 1.62, 95% CI 1.19–2.19). The SVD score improved the model C-statistic only modestly, from 0.699 to 0.705 (ΔC=0.006; P=0.035).
When follow-up events were coded simply as present/absent, major bleeding was associated with ΔmRS ≥1 (aOR 4.86, 95% CI 3.05–7.75) and ischemic events with ΔmRS ≥1 (aOR 3.20, 95% CI 2.43–4.22). These were prespecified adjusted analyses, not randomized treatment effects.
In the 4-way mediation analysis of severe SVD (score ≥3), the controlled direct effect remained significant (coefficient 0.498, 95% CI 0.26–0.73) and accounted for 73.05% of the total effect; the pure natural indirect effect through major bleeding and ischemic events was also significant (coefficient 0.051, 95% CI 0.01–0.09).

Secondary Outcomes:

Major bleeding, clinically relevant nonmajor bleeding, ischemic events, mRS 3–6 at 24 months, and all-cause death increased as SVD score increased in unadjusted comparisons.
The SVD score itself was not significantly associated with mRS 3–6 at 24 months or all-cause death in the adjusted analysis, although major bleeding and ischemic events were associated with both.
Individual-marker analyses identified cerebral microbleeds as associated with ΔmRS ≥1 (aOR 1.23, 95% CI 1.05–1.45) and confluent white-matter hyperintensity as associated with ΔmRS ≥1 (aOR 1.36, 95% CI 1.15–1.61) and mRS 3–6 (aOR 1.43, 95% CI 1.15–1.78).
Among patients without major bleeding or ischemic events, ΔmRS ≥1 occurred in 12.2% with SVD score 0, 14.8% with score 1, 19.2% with score 2, 24.8% with score 3, and 25.3% with score 4; adjusted ORs remained significant for score 3 (1.57, 95% CI 1.19–2.08) and score 4 (1.61, 95% CI 1.17–2.22).

Adverse Events / Safety:

This was not a safety trial and did not assign antithrombotic therapy. Major bleeding and ischemic events were recorded as outcomes/covariates; both increased with SVD burden and were strongly associated with later disability and death.
The article does not provide a randomized comparison of symptomatic intracranial hemorrhage, recurrent stroke, or mortality by a specific antithrombotic agent in this subanalysis.

Figures

*Figure 1.* Flow chart of patient selection. BAT2 indicates Bleeding with Antithrombotic Therapy 2; MRI, magnetic resonance imaging; mRS, modified Rankin Scale; and SVD, small vessel disease. *Tanaka et al.7 Licensed under CC-BY-NC-ND 4.0.

Change in modified Rankin Scale by SVD score — *Figure 2.* Change in modified Rankin Scale (mRS) scores from baseline to the end of follow-up by small vessel disease (SVD) score. Sankey diagrams show individual patient paths from baseline mRS to last recorded mRS at 24 months; red lines represent an mRS increase of at least 1 point and blue lines represent unchanged or decreased mRS.

Multivariable analysis of functional decline — *Figure 3.* Multivariable analysis for factors associated with an increase in mRS from baseline to the end of follow-up, adjusted for demographics, vascular risk factors, antithrombotic therapy, MRI field strength, bleeding, ischemic events, and clinically relevant nonmajor bleeding.

Causal mediation analysis of SVD burden — *Figure 4.* Causal mediation analysis of severe SVD burden (score ≥3) on ΔmRS ≥1 through major bleeding and ischemic events, partitioning controlled direct, natural indirect, reference interaction, and mediated interaction effects.

Criticisms

This is a prospective observational subanalysis, not a randomized test of an SVD-directed intervention or of one antithrombotic strategy; residual confounding and confounding by indication remain possible.
Nearly all participants were Asian, 89.6% had prior ischemic stroke, and all were receiving oral antithrombotic therapy, limiting external validity to other populations and to people without established vascular disease.
Stroke severity data were unavailable. SVD score rose alongside age, premorbid disability, hypertension, kidney disease, and prior vascular events; adjustment cannot guarantee that the observed association is independent of unmeasured baseline disability or frailty.
mRS is weighted toward mobility and does not isolate cognition, mood, dysphagia, gait, or parkinsonism—the mechanisms the authors propose for SVD-related decline. The study did not identify the specific causes of functional deterioration.
SVD markers were not assessable in 422 patients (8.9%) in the reported complete-case context, while 613 enrolled patients were excluded overall. For patients missing one imaging feature, the primary scoring convention assigned zero points, which may dilute or misclassify exposure.
The mediation estimate that 73.05% of the total effect was a controlled direct effect depends on causal assumptions that may not hold in an observational registry; it should not be read as proof that SVD directly caused 73% of individual patients’ decline.
The C-statistic improvement of 0.006 was statistically significant but small, so SVD score adds limited discrimination beyond standard clinical predictors.

Funding

The investigator-initiated BAT2 study was coordinated by the National Cerebral and Cardiovascular Center and supported by the Japan Agency for Medical Research and Development (JP18ek0210055, JP24lk0221171, and JP24lk0221186) and the Japan Society for the Promotion of Science (JP19K17023 and JP23K27522). The paper reports registration at ClinicalTrials.gov (NCT02889653) and the University Hospital Medical Information Network clinical trial registry (UMIN000023669).

The paper

Authors. Arakaki et al. for the BAT2 Investigators.
Title. Impact of Cerebral Small Vessel Disease on Functional Decline: Bleeding With Antithrombotic Therapy 2 Study.
Journal. Stroke.
Year. 2026.
DOI. 10.1161/STROKEAHA.125.054738
PMCID. PMC13281974

Deep Dive — click to expand

What this is

This is a large, prospective registry asking whether MRI-visible cerebral small vessel disease is merely a marker of future bleeding and recurrent ischemia, or whether it also tracks a slower loss of function between overt events. Among 4,765 Japanese adults on oral antithrombotic therapy followed for 24 months, functional decline rose from 13.2% at SVD score 0 to 28.1% at score 3 and 28.2% at score 4. The association persisted after time-dependent adjustment for major bleeding and ischemic events, but the study remains prognostic rather than interventional.

1. Shadow Audit

The headline is that severe SVD independently predicts decline, but the quieter result is that SVD adds little predictive discrimination: the adjusted model C-statistic moved only from 0.699 to 0.705. The study also does not show a corresponding adjusted association with mRS 3–6 at 24 months or all-cause death, even though major bleeding and ischemic events were associated with those outcomes. In other words, SVD appears useful as a vulnerability signal for a one-point mRS worsening, not yet as a stand-alone predictor of severe disability or mortality. The absolute gradient is clinically visible—28.1% versus 13.2%—but the score is not a magic threshold that explains which individual patient will deteriorate.

2. Inversion Engine

For the main interpretation to invert, the apparent SVD gradient would have to be largely explained by unmeasured baseline differences—especially stroke severity, cognition, frailty, or premorbid function—rather than by SVD-related brain disconnection. The score-0 versus score-3 unadjusted difference is 14.9 percentage points, and the adjusted OR is 1.65. The authors’ event-free sensitivity analysis still found 24.8% versus 12.2% decline for scores 3 versus 0, with aOR 1.57, so bleeding and ischemia alone would need to account for most of the residual gradient. That sensitivity result makes a complete inversion less plausible, but it cannot eliminate confounding because the registry did not measure all relevant functional and neurological determinants.

3. Second-Order Catalyst

The first practice change should not be to stop antithrombotic therapy. It should be to treat a high SVD score as a rehabilitation and prevention flag: review blood-pressure control, reconcile antithrombotic indication and intensity, screen for gait, cognition, dysphagia, mood, and falls, and involve rehabilitation earlier when score is 3–4. The institution that moves first is a stroke clinic with standardized MRI reporting, because the score can be recorded at the same visit as secondary-prevention planning. The paper supports triage and monitoring now; it does not establish that cilostazol, intensive blood-pressure treatment, or any other therapy improves outcomes in this specific high-SVD subgroup.

4. Asymmetric Leverage

The leverage is in the large denominator: roughly one in five patients overall declined by at least one mRS point over two years, and the study included 4,765 people across 52 hospitals. A modestly better prevention or rehabilitation pathway applied to every patient with severe SVD could matter more than a dramatic intervention limited to a small subgroup. The opposite asymmetry is equally important: a one-point mRS change is broad and may capture subtle or nonspecific deterioration. The signal is therefore valuable for population-level risk stratification, but its bedside effect size for any single patient is smaller than the OR suggests.

5. Paradigm Destroyer

This paper challenges the reflex that a patient with extensive SVD is only dangerous because the next event might be a hemorrhage or ischemic stroke. Even without those overt events, the event-free sensitivity analysis showed a graded functional decline signal. A practical protocol update for tomorrow morning would be: “For SVD score 3–4, keep the indicated antithrombotic plan unless a separate bleeding-risk review changes it, but automatically add structured gait/cognition/falls screening, blood-pressure review, and rehabilitation follow-up.”

MVP — Minimum Viable Proof

The minimum practice-changing statement is: in patients on oral antithrombotic therapy, severe MRI-defined SVD (score ≥3) is associated with approximately 25% functional decline over two years even when no major bleeding or ischemic event occurs, compared with approximately 12% at score 0. A randomized or rigorously prospective intervention study would then need to show that SVD-triggered rehabilitation or vascular-risk management reduces that decline.

Best Combination

Combine this study with the established literature in two layers. First, SVD burden already identifies vulnerability to hemorrhagic and ischemic complications during antithrombotic therapy; this study adds a slower functional pathway beyond those events. Second, prior work suggests that white-matter hyperintensity and disrupted subcortical networks can affect gait, cognition, and executive function. The appropriate synthesis is not “SVD means avoid antithrombotics,” but “SVD is a multidomain vulnerability phenotype requiring both event prevention and preservation-of-function strategies.”

Overvalue Warning

First, do not convert the 73.05% controlled direct-effect proportion from a mediation model into a claim that 73% of an individual patient’s decline was directly caused by SVD; that estimate depends on untestable causal assumptions in an observational registry. Second, do not treat an SVD score of 3 as a validated treatment threshold: the score had only a 0.006 incremental C-statistic gain, the study did not measure specific causes of decline, and it did not test an SVD-directed intervention.

Educational use notice. The Clotbust Review is dedicated to medical education. Figures from published articles are embedded for teaching purposes, with attribution to the original publication. We do not claim copyright over any embedded figure; all rights remain with the original publisher and authors under the article's published license. If you are a rights holder and would like a figure removed, please contact the editorial team. See our editorial notice for the full policy.