Introduction

Spontaneous intracerebral hemorrhage (SICH) is a highly destructive disease with higher mortality and disability rates compared to ischemic stroke1. Despite substantial advances in treatment, SICH prognosis shows little improvement. The 30-day mortality rates are high, ranging from 35 to 52%, with merely about 20% achieving functional independence at 6 months2. Theoretically, surgical interventions can alleviate the mass effect, rapidly lower elevated intracranial pressure (ICP), enhance cerebral blood flow, and mitigate hematoma-induced cytotoxic reactions, potentially improving patient prognosis. Nevertheless, the conclusive advantage of surgical interventions, including craniotomy, endoscopic surgery, and minimal puncture drainage3,4,5,6,7, is still unclear.

Currently, many studies focus on hemorrhagic transformation after cerebral infarction as a main factor contributing to poor prognosis8. However, only a few studies have explored cerebral infarction complications in SICH. Recently, high rates of diffusion-restricted lesions on diffusion-weighted imaging (DWI) consistent with cerebral infarcts (11.1–41%) have been described in minor or moderate SICH patients9. These lesions are often small and associated with poor outcomes10. Given that severe SICH often necessitates surgical intervention, and the natural process of cerebral infarction complication in severe SICH is rarely observed, we focus on early postoperative cerebral infarction (ePCI) in such SICH patients. ePCI is a serious complication of SICH and may partly explain the difficulty in improving the outcomes through surgery. To the best of our knowledge, no study has specifically focused on ePCI in patients with SICH. In this study, we aimed to investigate the incidence and distribution of ePCI seen on computed tomography (CT), its predictors, and its association with six-month mortality in supratentorial SICH patients by retrospectively analyzing data from a large, single-center study of SICH.

Methods

Study design and participants

This is a retrospective cohort study and follows the Strengthening the Reporting of Observational Studies in Epidemiology guidelines. All SICH patients admitted to the Shengli Clinical Medical College of Fujian Medical University from May 2015 to September 2022 were enrolled. Inclusion criteria were: SICH involving the basal ganglia, lobar, or thalamic regions; Age ≥ 18 years; undergoing surgical treatment; follow-up period ≥ 6 months. Exclusion criteria included: hemorrhage due to trauma, intracranial tumors, aneurysms, arteriovenous malformations, arteriovenous fistulas, infarction, or other lesions; coagulation dysfunction from hereditary coagulopathy, cancer, liver or kidney dysfunction; infratentorial cerebellar or brainstem hemorrhage; primary ventricular hemorrhage; age < 18 years; absence of surgical treatment; incomplete follow-up; history of severe stroke, heart, lung, liver, or kidney diseases; cerebral infarction occurring before surgery or more than 72 h after surgery; and postoperative cerebral infarction directly caused by surgical procedures. Surgical indications, aligned with our prior reports11, included transtentorial herniation, supratentorial hematoma > 30 ml, midline shift > 1 cm, and obstructive hydrocephalus. At our center, the perioperative management of antiplatelet/anticoagulant agents involves immediate discontinuation upon the detection of intracerebral hemorrhage, with resumption of the therapy more than two weeks after surgery.

Data collection

Trained study staff collected demographic data and medical histories, including hypertension, diabetes, prior ischemic/hemorrhagic stroke, coronary heart disease, deep vein thrombosis, atrial fibrillation, and consumption of antiplatelet/anticoagulant medications, by reviewing electronic medical records, as previously described11. Clinical parameters before surgery, such as Glasgow Coma Scale (GCS), original intracerebral hemorrhage scale (OICH), modified intracerebral hemorrhage A score (MICH-A)12, and signs of cerebral herniation, were also retrieved. Surgical parameters, such as approach, time from ictus to surgery, duration of surgery, and intraoperative blood loss, were recorded. Clinical data on postoperative complications (infectious meningitis, pulmonary infection, epilepsy), as well as intensive care unit (ICU) and hospital stay lengths, were also collected. Surgical approaches included open surgery (decompressive craniectomy with hematoma evacuation, craniotomy microsurgery) and minimally invasive surgery (endoscopic surgery, minimal puncture drainage, and simple extra ventricular drainage (EVD)). Pneumonia, infectious meningitis13, and epilepsy14 were diagnosed according to established guidelines.

We determined hematoma lateralization, hematoma location, hematoma volume, presence of intraventricular hemorrhage, Graeb, s scores, postoperative hematoma volume, rebleeding, and hematoma evacuation rate (%) on brain CT scans obtained within 24 h before surgery and within 24 h after surgery according to previously validated methodology11. Hematoma locations, including lobar, thalamus, and basal ganglia, were dichotomized into lobe and deep (thalamus and basal ganglia). Neuroimaging was analyzed blindly, without clinical information.

Early postoperative cerebral infarction (ePCI) definition

ePCI was defined as a new identifiable low-density lesion in the non-surgical areas observed on CT within 72 h after surgery, similar to previous reports15,16. The mere presence of newly added low-density regions in the surgical area was indicative of postoperative edema or surgical tissue injury, rather than cerebral infarction (Supplementary Fig. 1a,b). Lesions aligning with damaged cerebral arteries during surgery were defined as ePCI directly caused by surgical procedures (Supplementary Fig. 1c,d). New low-density lesions in non-surgical areas, apart from those directly caused by the surgical procedures, were defined as ePCI not directly caused by surgical procedures (Supplementary Fig. 1e,f). Cases with both types of ePCI were classified under ePCI not directly caused by surgical procedures (Supplementary Fig. 1g,h). ePCI territories were documented based on vascular regions: anterior cerebral artery (ACA), middle cerebral artery (MCA), posterior cerebral artery (PCA), vertebrobasilar (VB), and anterior choroidal artery (AChA) territories. ePCI lateralization was categorized as ipsilateral, contralateral, or bilateral.

Follow up data

Follow-up was conducted by independent, clinically data-blinded study staff through outpatient visits, telephone interviews, or WeChat communications. The primary outcome was mortality at 6 months. Secondary outcomes included functional outcomes at 6 months, lengths of ICU and hospital stay, and postoperative complications such as infectious meningitis, pulmonary infection, epilepsy, and rebleeding. Neurological functional status at 6 months was evaluated using the Extended Glasgow Outcome Scale (GOS-E), categorizing results into favorable (GOS-E score, 5–8) and unfavorable (GOS-E score, 1–4).

Statistical methods

Baseline characteristics were categorized by ePCI or no postoperative cerebral infarction (nPCI). Continuous and categorical variables were reported as mean ± standard deviation (SD) or median (interquartile range, IQR), and counts (percentages), respectively. Normality was assessed using the Shapiro–Wilk test, and t-tests or Mann–Whitney U-tests were used for intergroup comparisons depending on variable distribution. Categorical variables were compared using the chi-square tests or Fisher's exact probability tests. Fisher`s exact probability test was used when the theoretical frequency less than 5.

Logistic regression models were used to investigate each risk factor. In the univariable analysis, variables with p < 0.1 were selected for multivariable logistic regression analysis in a backward stepwise model. Before entering candidate predictors into the multivariable analysis, multicollinearity was assessed using variance inflation factors (VIFs).

Cox regression analysis evaluated the ePCI-mortality relationship, calculating hazard ratios (HRs) and 95% confidence intervals (CIs). Multivariable models were built, cumulatively adjusting for confounders such as age, gender (Model 1), diabetes, hypertension (Model 2), GCS score, bleeding location (Model 3), preoperative hematoma volume, brain herniation presence, Graeb’s score (Model 4), OICH, and MICH-A (Model 5). Selection of confounding factors was guided by their clinical significance. Kaplan–Meier curves for mortality, based on ePCI or nPCI status, were generated and compared using the log-rank test. To assess result robustness, subgroup analyses were also conducted for different patient groups, such as gender, age (< 60 years, ≥ 60 years), hypertension status, diabetes status, bleeding location (lobe, deep), and ventricular hemorrhage status. Interaction terms were added to the adjusted model to evaluate potential interactions. E-values were calculated to explore the impact of unmeasured confounders on the robustness of the association between ePCI and mortality.

All analyses were conducted using R version 4.2.2 (R Foundation for Statistical Computing) (http://www.R-project.org, The R Foundation) and the Free Statistics analysis platform. Two-sided tests were used, with a significance threshold set at p < 0.05.

Ethical approval

The study was approved by the Institutional Review Board of Fujian Provincial Hospital, Fujian Medical University (K2022-11-004), and was registered in the Chinese Clinical Trial Registry (ChiCTR2300074977). Due to the retrospective cohort study design, the requirement to obtain informed consent was waived by the Institutional Review Board.

Results

Study participants

During the study enrollment period, 3968 consecutive SICH patients were admitted to Fujian Medical University Provincial Clinical Medical College. Of these, 843 patients (21.2%) underwent surgery. Following the inclusion and exclusion criteria, 637 patients (454 males, 71.3%), with a mean age of 57.3 ± 12.5 years, were included in the analysis. Patient selection is illustrated in flowchart Fig. 1. Baseline characteristics of the study cohort are described in Table 1.

Figure 1
figure 1

Flowchart for selection of cases in study. ICH, intracerebral hemorrhage.

Full size image
Table 1 Baseline characteristics of participants.
Full size table

Incidence, distribution and predictors of ePCI

In our cohort, the incidence of ePCI was 11.1% (71/637). Of these, ePCI was present on the bleeding side in 40 cases (56.3%), the opposite side in 5 cases (7.0%), and bilaterally in 26 cases (36.6%). The MCA territory was the most common site for ePCI (63.4%, 45/71). ePCI also occurred in the PCA territory (28 cases, 39.4%), ACA territory (18 cases, 25.4%), VB territory (10 cases, 14.1%), and AChA territory (8 cases, 11.3%). The lesions were confined to one vascular territory in 40 cases and to multiple territories in 31 cases.

The results of univariable analyses of ePCI predictors are presented in Supplementary Table 1. Candidate predictors for multivariable analysis, selected based on a p-value < 0.10, included hypertension, history of antiplatelet/anticoagulation medication use, GCS, Graeb score, brain herniation status, surgical method, operation time, bleeding volume (per 100 ml), history of myocardial infarction, atrial fibrillation, and hematoma volume (per 10 ml). OICH and MICH-A were not included in the model because of their severe multicollinearity (VIF > 5). Following backward-stepwise selection, independent predictors of ePCI were identified as GCS (OR 0.62, 95%CI 0.48–0.80, p < 0.001), bleeding volume (per 100 ml) (OR 1.17, 95%CI 1.03–1.32, p = 0.016), hematoma volume (per 10 ml) (OR 1.14, 95%CI 1.02–1.28, p = 0.023) and bilateral brain herniation (OR 6.48, 95%CI 1.71–24.48, p = 0.006) (Table 2).

Table 2 Multivariable model of independent predictors of early postoperative cerebral infarction.
Full size table

ePCI associated with mortality

In the unadjusted model, ePCI was significantly associated with an increased risk of six-month mortality (HR 5.87; 95% CI 4.05–8.52; p < 0.001). Extended multivariable models consistently revealed significant HRs for ePCI across all five models (HRs ranged 3.30–5.97, p < 0.001 for all; see Table 3 for details ). Subgroup analysis did not show any significant interactions (Fig. 2). Kaplan–Meier survival curves also showed higher six-month mortality in ePCI (log-rank test: p < 0.001, Fig. 3). The E-value for the association between ePCI and six-month mortality ranged from 3.82 to 6.66.

Table 3 Cox Regression for ePCI on 6 month Mortality.
Full size table
Figure 2
figure 2

Stratified multivariable analysis of the association between early postoperative cerebral infarction and 6-month mortality. ePCI, early postoperative cerebral infarction; nPCI, no postoperative cerebral infarction.

Full size image
Figure 3
figure 3

Kaplan–Meier survival curves for days 180 of spontaneous supratentorial intracerebral hemorrhage with early postoperative cerebral infarction. The shaded areas in the Kaplan–Meier curves represent 95% confidence intervals. ePCI, early postoperative cerebral infarction; nPCI, no postoperative cerebral infarction.

Full size image

Secondary outcome analysis

The proportion of favorable clinical outcomes (GOSE 5–8) at 6 months in the ePCI group was merely 4.2% (3/71), significantly lower than in the nPCI group (49.6%, 281/566) (p < 0.001). The distribution of GOSE scores at 6 months for both the ePCI and nPCI groups is shown in Supplementary Fig. 2. Other secondary outcomes indicated that ePCI did not increase the risk of postoperative infectious meningitis, pulmonary infection, epilepsy, or rebleeding. It also did not contribute to an increase in the length of hospital stay. However, ePCI significantly prolonged ICU stay duration (see Supplementary Table 2).

Discussion

Our study revealed that ePCI seen on CT within 72 h after surgery (1) occurred in 11.1% of SICH patients, (2) predominantly affected the bleeding side and the MCA territory, (3) was predicted by GCS score, bleeding volume, hematoma volume, and brain herniation, and (4) was associated with higher postoperative 6 month mortality.

In our cohort, ePCI frequently occurred at the bleeding site, potentially due to the more severe compression of ipsilateral blood vessels. Unlike posttraumatic cerebral infarction (PTCI), typically located in the PCA territory17,18,19, ePCI in supratentorial SICH primarily occurred in the MCA territory in our study. This distinction may stem from PTCI often being caused by transtentorial herniation due to subdural or epidural hematomas compressing the PCA17,18,19. Conversely, SICH predominantly occurs in the basal ganglia or thalamic regions, exerting a significant mass effect on the MCA. Furthermore, the trans-sylvian surgical approach to SICH may increase the likelihood of ePCI in the MCA territory due to surgical disruptions.

Low GCS, large hematoma volume, and brain herniation, closely associated with SICH severity20,21, emerged as independent risk factors for ePCI, underscoring that the more severe the condition, the higher the incidence of ePCI. Hence, we speculate that the primary pathophysiological mechanisms behind ePCI appear to involve elevated ICP reducing cerebral perfusion pressure and blood flow, alongside mechanical compression and occlusion of intracranial vessels by the mass effect of the hematoma. In brain herniation, compression of the intracranial artery by displaced brain tissue can cause infarction. These findings indicate that immediate surgical intervention to lower ICP and alleviate vascular compression is essential in reducing ePCI in severe SICH cases.

Our study also identified intraoperative bleeding as an independent predictor for ePCI. Significant intraoperative bleeding may lead to hypotension and hypoperfusion, inducing ischemic brain injury. While our univariable analysis identified preoperative antiplatelet/anticoagulant therapy (with perioperative cessation) as a risk factor for ePCI, the multivariable analysis did not demonstrate statistical significance. Nonetheless, caution is advised in interpreting these results due to potential patient selection biases and the lack of detailed disease stratification. Notably, patients on preoperative antiplatelet or anticoagulant therapy may be more inclined to refuse surgery, given a higher likelihood of adverse outcomes.

Univariable analysis suggested minimally invasive surgery could potentially reduce ePCI incidence, yet multivariable analysis did not confirm this difference as statistically significant. This may reflect a bias wherein patients with more severe conditions, often selected for open surgery, inherently possess a higher risk of ePCI. Unexpectedly, our study did not find an increase in ePCI among patients with a history of certain conditions, such as cerebral infarction, ICH, or myocardial infarction. At present, this outcome remains unexplained. Of note, patients with a history of severe cerebral infarction or ICH were not included in the study. Literature indicates that 11.1–41% of minor or moderate ICH patients show remote diffusion-restricted infarct lesions on DWI, potentially due to microangiopathy9. Further research is needed to ascertain whether microangiopathy is responsible for ePCI seen on CT after severe ICH. Another potential mechanism of ePCI is vasospasm, which frequently occurs in cases of subarachnoid hemorrhage and intraventricular hemorrhage22,23,24. Our data indicate that the ePCI group exhibited a higher proportion of intraventricular hemorrhage and elevated Graeb scores; however, these differences did not achieve statistical significance. The relationship between vasospasm and ePCI warrants further investigation.

Similar to previous reports on PTCI in traumatic brain injury and early cerebral infarction following aneurysmal subarachnoid hemorrhage16,17,18,19,25,26, our data showed a strong association between ePCI after SICH and poorer clinical outcomes, as well as higher six-month mortality, attributable to several factors. First, the acute phase of SICH renders brain tissue fragile, increasing vulnerability to secondary ischemic damage. Second, the predominance of ePCI in the functional areas of MCA territory exacerbates the negative impact on prognosis. Third, currently, no effective treatment is available for ePCI after SICH. Although intravenous thrombolysis (IVT) and mechanical thrombectomy (MT) are established treatments for acute ischemic stroke27,28, IVT is contraindicated in ePCI cases due to bleeding risks. MT may be a potential treatment option for ePCI after SICH, with favorable outcomes in cases of SICH complicated by acute large vessel occlusion reported in literature29. The etiology of ePCI after SICH is multifactorial, including vascular compression, reduced cerebral perfusion, surgery-related damage, and cerebral thrombus formation. Identifying thrombotic infarction within the time window for MT after SICH is challenging, as patients are often in a comatose or sedated state postoperatively. Our study suggests ePCI is a cornerstone complication of ICH and may partly explain why it seems so difficult to improve the outcome by surgery. In instances where ePCI is inevitable, surgical intervention should be approached with caution.

Our study has several strengths. This is the first study in severe ICH patients with surgery reporting on incidence, distribution, predictors, and outcomes of ePCI seen on CT within 72 h after surgery. Moreover, we found a significant increase in mortality among ePCIs, as evidenced by subgroup analysis and an E-value of 3.09–6.84, suggesting the robustness of our findings against potential unmeasured confounders.

Some limitations of our study warrant discussion. The use of CT over Magnetic Resonance Imaging (MRI) for ePCI evaluation may underestimate its true incidence by missing small lesions. Early postoperative ICH patients, often in the critical stage, render MRI examinations impractical. CT scanning, with its wide availability, ease of repetition, and short scanning duration, stands out. Hence, CT emerges as the more practical choice for real-world clinical applications in these patients. Our CT scan protocol, involving routine check within 24 h after surgery and periodic follow-up scans, could introduce detection bias, as critically ill patients receive more frequent scans, increasing the likelihood of identifying cerebral infarctions. At the same time, asymptomatic cerebral infarctions might be missed and underestimated. Additionally, excluding ePCI cases directly caused by surgical manipulation also could underestimate its true occurrence. Surgeon experience and skill significantly influence the occurrence of ePCI directly caused by surgical manipulation. Excluding these cases helps make the remaining ePCI cases more representative and generalizable. Furthermore, the inability to perform vascular status examinations due to poor clinical conditions presents another limitation.

Conclusions

Our results indicate ePCI is a significant complication in patients with supratentorial SICH, and predominantly affects the bleeding side and the MCA territory. It markedly increases postoperative mortality and can be independently predicted by low GCS score, brain herniation, increased intraoperative bleeding, and large preoperative hematoma volume. This discovery enriches our understanding of the impact of ePCI and will help identify high-risk patients in advance. Future studies are needed to establish whether ePCI recognition and prevention after ICH could improve outcomes.