Current stroke guidelines recommend intravenous thrombolysis (IVT) using recombinant tissue-type plasminogen activator (rt-PA) as the first-line treatment for acute ischemic stroke (Chinese Society of Neurology and Chinese Stroke Society, 2018; Powers et al., 2018; Berge et al., 2021). However, approximately 7.0–27.0% of patients treated with IVT using rt-PA experienced intracerebral hemorrhage (ICH) (Hacke et al., 2008; Dorňák et al., 2018; Anderson et al., 2019), which might be related to unfavorable clinical outcomes. Previous efforts have been made to identify several risk factors for developing ICH after IVT, including aging, hypertension, atrial fibrillation, and stroke severity (Mazya et al., 2012; Whiteley et al., 2012). Cerebral small vessel disease (cSVD) refers to a group of age-related (but not clearly age-derived) pathological processes with various etiologies that affect the microvasculature in the brain and might relate to intracranial bleeding risk (Pantoni et al., 2014; Charidimou et al., 2016). Several studies showed that cSVD markers including cerebral microbleeds (CMBs) and white matter hyperintensities (WMH) were predictors for any ICH after IVT (Prats-Sánchez et al., 2016; Charidimou et al., 2017; Capuana et al., 2020). However, a retrospective study of 434 patients with ischemic stroke failed to demonstrate the association of white matter disease with symptomatic ICH (sICH) after IVT (Capuana et al., 2020). Recently, the role of enlarged perivascular spaces at basal ganglia (BG-EPVS) or centrum semiovale (CSO-EPVS) in ICH risk has been explored (Arba et al., 2017; Duperron et al., 2019). To date, the effect of total cSVD burden on the risk of ICH after IVT in patients with acute ischemic stroke undergoing rt-PA thrombolysis remains unclear. Therefore, we aimed to investigate the relationship between total cSVD score and the risk of ICH after IVT in a retrospective observational study.

A total of 785 patients with acute ischemic stroke underwent IVT between 1 January and 31 December 2020. After excluding 514 patients [63.5% men, median age 69 (IQR 60–76)] who did not have baseline MRI data or had insufficient MRI data, we included 271 eligible patients [67 years (IQR 59–75); 67.1% men] in the final analysis.

Supplementary Table 1 summarizes the prevalence of individual SVD markers. Regarding the total cSVD score, 81 (29.9%) patients had a score of 0, 99 (36.5%) patients had a score of 1, 71 (26.2%) patients had a score of 2, and 20 patients (7.4%) had a score of 3–4. Supplementary Table 2 shows that the intra-rater and inter-rater reliability for individual cSVD markers were satisfactory.

Notably, 55 (20.3%) out of 271 patients experienced any ICH after IVT, 16 (5.9%) patients experienced a sICH, and 5 (1.9%) patients had rPH. Table 1 summarizes the baseline clinical and imaging characteristics among patients with and without any ICH after IVT. Patients with any ICH after IVT were older [70 (63–76) vs. 66 (58–740), P = 0.017], more likely to have atrial fibrillation [29 (52.7%) vs. 42 (19.4%), P < 0.001], a higher NIHSS score [11 (7–14) vs. 4 (2–8), P < 0.001], and undergo endovascular therapy [16 (29.0%) vs. 8 (3.7%), P < 0.001]. Regarding cSVD markers, patients with any ICH after IVT were more likely to have ≥ 1 CMB presence [35 (63.6%) vs. 60 (27.7%)], ≥ 11 BG-EPVS presence [38 (69.0%) vs. 108 (50.0%)], moderate to severe WMH presence [10 (18.1%) vs. 17 (7.8%)], and a higher cSVD score (P < 0.001). We found similar ≥ 1 lacune presence between patients with and without any ICH after IVT [8 (14.5%) vs. 31 (14.3%)]. Table 2 shows that patients with rPH were more likely to have lacunes [3 (60.0%) vs. 36 (13.5%)], moderate to severe WMH [2 (40.0%) vs. 25 (9.3%)], and a higher cSVD score. There were no significant differences in individual SVD markers or total cSVD score among those with and without sICH (Table 3).

The LASSO model filtered out four variables (i.e., NIHSS, endovascular therapy, AF, and total cSVD score) from 17 risk characteristic individual factors for any ICH in the final multivariable regression model. Figure 1 shows how the regression coefficients change as a function of the penalization parameter [lambda (λ)]. On the far left, we saw the coefficients from standard logistic regression. As we go from left to right, the coefficients are “shrunk” toward zero. The coefficient estimated from the optimal model is indicated by the dashed line, which shows that only NIHSS, endovascular therapy, AF, and the total cSVD score were kept in the model. Multivariable logistic regression analysis showed that the total cSVD score (per unit increase, OR 2.03 95% 1.22–3.41, P = 0.007) was significantly associated with a higher risk of any ICH after IVT. The risk of any ICH increased with increasing cSVD score (cSVD score 1, OR 1.29, 95% CI 0.44–3.94, cSVD score 2, OR 4.90, 95% CI 1.81–14.81, cSVD 3–4, OR 12.40, 95% CI 3.40–48.77, compared with a score of 0, P < 0.001, AUC 0.698, 95% CI 0.620–0.777, Table 4). The very small number of sICH (n = 16) and rPH events (n = 5) did not allow us to perform the logistic regression analysis using the LASSO approach. Sensitivity analyses using the E-value revealed that it would need moderately robust unobserved confounding to render the exposure-outcome (cSVD-any ICH) association null (Figure 2).

Based on a previous meta-analysis and our LASSO selection results, the NIHSS score was considered a robust predictor for any ICH after IVT (Whiteley et al., 2012). Therefore, we compared the predictive performance of a combination of NIHSS and cSVD score with that of the NIHSS score alone. The results showed that combining cSVD and NIHSS score had a better predictive performance than NIHSS score alone (AUC 0.811, 95% CI 0.756–0.866 vs. 0.784, 95% CI 0.723–0.846, P = 0.0004, Table 5 and Figure 3).

This study shows that a higher SVD burden is associated with increased risk of any ICH after IVT, suggesting that the risk of any ICH in patients with acute ischemic stroke who received IVT with rt-PA might be related to SVD severity. Moreover, the cSVD score adds to the NIHSS score alone in predicting any ICH after IVT.

In our cohort, patients with any ICH were more likely to have ≥ 1 CMB presence. However, CMB presence was not screened out using the LASSO model. The limited sample size of our study does not enable us to perform the further analysis stratified by CMB burden. Our findings were supported by several studies that showed the CMB presence was not associated with an increased risk of any ICH (Kakuda et al., 2005; Kim et al., 2006; Fiehler et al., 2007) or sICH (Charidimou et al., 2017; Capuana et al., 2020) after IVT. Current stroke guidelines do not recommend IVT in patients with > 10 CMBs at baseline (Chinese Society of Neurology and Chinese Stroke Society, 2018; Powers et al., 2018; Berge et al., 2021). A recent study using a multistep algorithm populated with data showed that in patients with > 10 CMBs, IVT is associated with higher mortality and net harm defined by age, stroke symptom severity, and treatment delay (Schlemm et al., 2020). To date, there are no studies reporting directly on the risk of sICH in patients without IVT stratified by CMB burden. Future prospective studies are needed to define the CMB burden threshold in individual risk stratification predicting the bleeding risk after IVT and functional outcome.

Evaluating the total cSVD burden may allow more accurate estimates of size effects on ICH risk after IVT than individual cSVD markers. In our cohort, the total SVD score showed a dose-effect response, reinforcing the validity of the findings. Similarly, a small sample-sized retrospective study showed that severe cSVD burden (cSVD score of 3–4) was associated with an increased risk of hemorrhagic transformation after IVT (OR = 8.429, 95% CI 1.643–43.227, P = 0.011) (Liu et al., 2019). Our study adds to the aforementioned study by investigating the association between total cSVD score (0 vs. 1, 2, 3, 4) and the bleeding risk after IVT, using variables that were selected by LASSO. Moreover, our data supported the possible relationship between the risk of rPH and the total cSVD score. However, the small number of rPH events did not enable us to perform the regression analysis. A previous single-center observational study showed that patients with stroke with r-PH after IVT were more likely to have strictly lobar CMBs, suggesting amyloid angiopathy (Drelon et al., 2020). Since the occurrence of r-PH may be related to worse outcomes, the association between r-PH and total cSVD score needs further investigation. Possible mechanisms that account for the role of cSVD in ICH after IVT include endothelial dysfunction, arterial stiffening (arterial stiffening might increase ICH risk), and increased blood-brain barrier permeability (Campbell and Khatri, 2020).

Consistent with a previous meta-analysis (Whiteley et al., 2012), our data support that the NIHSS score is a robust predictor for ICH risk after IVT. The larger AUC of a combined NIHSS and cSVD score than an NIHSS score alone suggests that adding cSVD as neuroimaging biomarkers to a widely validated factor (NIHSS) might improve specificity and sensitivity in predicting ICH after IVT. Since most of our cohort had mild to moderate acute ischemic stroke [median NIHSS score 5 (IQR 3–10)], our results might not be generalized to the patients with severe stroke. Future large-scale studies are required to refine and validate robust risk prediction tools.

There are some limitations of this study. First, this is a retrospective study with a limited sample size. Second, excluding patients who did not undergo MRI within 48 h after onset or insufficient MR image might introduce selection bias since most severe patients were not included. Third, we did not analyze the functional outcome data after discharge in this study; a longer follow-up period data will better determine the significance of our findings. Fourth, very few patients underwent MR before IVT; therefore, we could not exclude the new cSVD markers (particularly new CMBs) after IVT (Braemswig et al., 2019). However, the cSVD findings did not influence the IVT decisions. Fifth, due to the small number of outcome events, we did not assess the predicting performance of cSVD for sICH and rPH, which is the most clinically relevant. Finally, due to the data unavailability, we were unable to calculate the estimated effect size adjusted for some unmeasured confounders. However, we used the E-value sensitivity analysis to quantify the potential implications of unmeasured confounders and found that it was unlikely to explain the entire outcome.

This study shows that the total cSVD score is associated with a higher risk of any ICH after IVT. However, our study was not powered to assess the association between cSVD score and sICH risk. Further studies are warranted to determine whether rt-PA can be safely initiated in patients with acute ischemic stroke who have coexistent cSVD.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

The studies involving human participants were reviewed and approved by the Fujian Medical University Union Hospital Ethics Committees (2019KY076). Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

HD and NL: concept, design, and full access to all of the data in this study and take responsibility for the integrity of the data and the accuracy of the data analysis. HD, SW, HhL, and NL: acquisition, analysis, and interpretation of data. HD, SW, and HfL: drafting of the manuscript. GA, DW, HfL, YC, DC, QY, SF, RC, YZ, JW, GC, and JC: critical revision of the manuscript for important intellectual content. HD, HhL, and GA: statistical analysis. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.