Early haemodynamic predictors of poor functional outcomes in patients with acute ischaemic stroke receiving endovascular therapy: a single-centre retrospective study in China

Ethical approval

This study was approved by the Ethics Committee of the Second Affiliated Hospital of Shandong First Medical University (approval number: 2021-098) and conducted in accordance with the Declaration of Helsinki as revised in 2013. A written informed consent was obtained from conscious patients or the legal representatives of unconscious patients.

Study population and therapeutic process

We identified all consecutive hospitalised patients who underwent EVT for acute large-artery occlusion (LAO) in the anterior circulation (intracranial internal carotid artery (ICA) or middle cerebral artery (MCA)) between October 2021 and September 2022 at our stroke centre (the Second Affiliated Hospital of Shandong First Medical University). All patients underwent a CT and/or CTP or an MRI examination before treatment to confirm the diagnosis.

Interventional neurologists performed mechanical thrombectomy using stent retrievers or clot aspiration systems. The treatment results were evaluated using the modified treatment in cerebral infarction (mTICI) score at the end of the EVT. mTICI grades 2b and 3 indicated complete recanalization, and mTICI grade 2 indicated partial recanalization (Wang et al., 2015). After undergoing EVT, all patients were treated in the neurointensive care unit. The target blood pressure level was <180/105 mmHg according to international guideline recommendations (Jauch et al., 2013).

We excluded patients with ≥ 70% stenosis of the extracranial ICA or with ≥ 50% stenosis of the contralateral MCA. Moreover, patients with minimal recanalization (mTICI: 1) or no recanalization (mTICI: 0) of the lesion-side MCA were excluded because cerebral blood flow indicators could not be obtained. Additionally, we excluded patients who did not complete the TCCS assessment within 12 h after undergoing EVT and those without available temporal windows or complete clinical data during the study period.

Clinical data collection

Patient demographics and clinical variables were collected and analysed. The clinical variables included major vascular risk factors, stroke subtype, stroke severity, and treatment. Major vascular risk factors included hypertension, hyperlipidaemia, diabetes mellitus, coronary artery disease, atrial fibrillation, smoking, alcohol use, previous stroke, and medication use. Stroke subtypes were classified according to the Trial of Org 10172 in Acute Stroke Treatment criteria (Adams Jr et al., 1993). Stroke severity was assessed using National Institutes of Health Stroke Scale (NIHSS) scores at admission and Alberta Stroke Program Early CT (ASPECT) scores of the initial brain CT images. Regarding treatment, the use of intravenous recombinant tissue plasminogen activators and EVT parameters such as time from onset to treatment, time from onset to reperfusion, and type of EVT were recorded.

TCCS

All patients were evaluated using TCCS (devices: Mindray M9, handheld transducers SP5-1S, 1–5 MHz; Mindray Biomedical Electronics Co., Ltd., Shenzhen, Guangdong, China) within 12 h of undergoing EVT by three professional ultrasound specialists blinded to the clinical data. To ensure an accurate comparison of the bilateral MCA blood flow velocities (BFVs), TCCS monitoring was conducted at a depth of 45–60 mm. Gray and colour gains were adjusted for each patient to obtain optimal image quality. Most patients opted not to use ultrasound contrast agents due to concerns about cost and potential negative consequences. Therefore, we did not use ultrasound contrast agents during TCCS monitoring.

During examination, we recorded haemodynamic data, including peak systolic velocity (PSV), end-diastolic velocity (EDV), mean blood flow velocity (MFV), and pulsatility index (PI), from the MCA to assess the vessel status and cerebral haemodynamics. The highest BFV value was recorded after correcting for the insonation angle. The contralateral artery was examined along with the treated vessel. Moreover, we recorded the patients’ blood pressure at the time of ultrasonography and the time interval between undergoing EVT and TCCS.

To account for inter-individual differences (Kneihsl et al., 2022), the MCA-MFV ratio (ipsilateral MFV (iMFV)/contralateral MFV (cMFV)) and PSV ratio (ipsilateral PSV (iPSV)/contralateral PSV (cPSV)) were calculated. To evaluate the impact of differences in BFV in the bilateral MCA on prognosis, we introduced the adjusted PSV ratio, which was calculated as follows: If iPSV ≥ cPSV, the adjusted PSV ratio was equal to the PSV ratio. In contrast, if iPSV <cPSV, then the adjusted PSV ratio was equal to the inverse of the PSV ratio. The adjusted MFV ratio was calculated similarly.

Follow-up

Whenever possible, follow-up information was obtained at 90-day personal clinical visits. For the remaining patients, the clinical outcomes were assessed via telephone calls with the patient or the relatives. The modified Rankin Scale (mRS) score was used to estimate patient outcomes; a poor functional outcome was defined as an mRS score of 3–6 (Kneihsl et al., 2018).

Statistical analysis

Paired continuous variables were analysed using the Wilcoxon matched-pairs signed-rank test, and unpaired groups were analysed using the Student’s t-test or Mann–Whitney U test. The categorical variables were analysed using the χ²test or Fisher’s exact test. Multivariate logistic regression analysis was used to evaluate the association with poor functional outcomes by calculating the corresponding odds ratios (ORs) and 95% confidence intervals (CIs), and only those factors that had statistical significance with univariate logistic regression analysis (p < 0.05) were employed. Statistical significance was set at a p-value of <0.05.

The optimum cut-off values of haemodynamic predictors, which were significantly associated with poor functional outcomes, were calculated using receiver operating characteristic (ROC) analysis. Furthermore, by comparing the ROC curves, we investigated whether adding haemodynamic predictors could improve the prediction ability of the regression model for poor functional outcomes at 90 days in AIS patients underging EVT. A two-tailed p-value of <0.05 was considered statistically significant. Statistical analyses were performed using the SPSS software (version 22.0; SPSS Inc., Chicago, IL, USA) and Graphpad Prism (version 8.3.0; Graphpad Software, San Diego, CA, USA).

Population characteristics

During the study period, 156 consecutive patients with AIS caused by LAO of the anterior circulation were hospitalised. However, 48 patients were excluded as they did not meet the inclusion criteria. A detailed flowchart outlining the inclusion and exclusion criteria is shown in Fig. 1. The analysis was conducted on 108 patients (median age: 66 years; 69.4% men). Of these patients, 63 (58.3%) had an MCA occlusion, 31 (28.7%) had an intracranial ICA occlusion, and 14 (13.0%) had occlusions in both the intracranial ICA and MCA. The median NIHSS score at admission was 18 (interquartile range (IQR): 14–22). Before undergoing EVT, 41 patients (38.0%) underwent intravenous thrombolysis. Complete recanalization (mTICI: grade 3) was achieved in 93 patients (86.1%), and partial recanalization (mTICI: grade 2) was achieved in 15 (13.9%). After undergoing EVT, 18 patients experienced intracranial haemorrhage, 12 showed an increase in the infarct size, and two experienced diffuse cerebral oedema. Of the total 108 patients, 60 (55.6%) had poor 90-day outcomes, and 90-day all-cause mortality was 13.0%. Deaths were attributed to tentorial herniation caused by cerebral haemorrhage or diffuse oedema (in nine patients) and pulmonary infection (in five patients). A complete summary of the demographic and clinical characteristics of the study population is presented in Table 1.

TCCS indicators of the ipsilateral MCA (iMCA) and the contralateral MCA (cMCA)

TCCS was performed at a median time interval of 3.8 h after undergoing EVT. The median of iPSV, iEDV, iMFV, and iPI were 164.4 (IQR: 110.4–215.5) cm/s, 60.2 (IQR: 45.0–86.1) cm/s, 98.1 (IQR: 68.7–121.2) cm/s, and 1.01 (IQR 0.84–1.18), respectively; the median of cPSV, cEDV, cMFV, and cPI were 107.4 (IQR: 95.5–128.0) cm/s, 40.7 (IQR: 34.8–49.9) cm/s, 68.5 (IQR: 60.1–80.8) cm/s, and 0.92 (IQR: 0.80–1.06), respectively. Our results showed that PSV, EDV, MFV, and PI were significantly higher in the iMCA than in the cMCA, and the differences were statistically significant (Z = −7.53, −7.20, −6.53, and −2.29, respectively; p < 0.001, 0.001, 0.001, and = 0.022, respectively) (Fig. 2).

Clinical and TCCS characteristics of patients with poor functional outcomes at 90 days

In the univariate logistic regression analysis, the group with poor functional outcomes was of older age (median [IQR]: 69 [58.8–73.8] years vs. 64 [58–69.5] years; p = 0.035), had higher admission NIHSS scores (median [IQR]: 20 [16–23] vs. 17 [9.3–19.8], p < 0.001), and lower ASPECT scores (median [IQR]: 8 [8–9] vs. 9 [8−9.8], p < 0.001). Multivariate logistic regression analysis showed that, after adjustment, the initial NIHSS scores (adjusted OR (aOR): 1.07; 95% CI [1.01–1.13]; p = 0.023) and ASPECT scores (aOR: 0.53; 95% CI [0.32–0.86]; p = 0.010) were still independently associated with functional outcomes (Table 2).

To investigate the effect of haemodynamics on patient outcomes after undergoing EVT, we analysed the correlation between TCCS variables and 90-day outcomes. Patients with poor outcomes had significantly higher PSV ratio (mean ± standard deviation (SD): 1.60 ± 0.64 vs. 1.35 ± 0.31; p = 0.016), higher adjusted PSV ratio (median [IQR]: 1.75 [1.38–1.99] vs. 1.37 [1.09–1.61]; p <  0.001), higher MFV ratio (mean ± SD: 1.48 ± 0.64 vs. 1.22 ± 0.25; p = 0.010), and higher adjusted MFV ratio (median [IQR]: 1.60 [1.27–1.86] vs. 1.24 [1.11–1.41]; p < 0.001) than those of the good outcome group. Furthermore, after adjusting for the initial NIHSS and ASPECT score, an increased PSV ratio (aOR: 1.11 for every 0.1 increase; 95% CI [1.02–1.22]; p = 0.022), an increased adjusted PSV ratio (aOR: 1.31 for every 0.1 increase; 95% CI [1.13–1.50]; p < 0.001), an increased MFV ratio (aOR: 1.15 for every 0.1 increase; 95% CI [1.04–1.28]; p = 0.007), and an increased adjusted MFV ratio (aOR: 1.48 for every 0.1 increase; 95% CI [1.23–1.78]; p < 0.001) remained independently associated with poor outcomes at 90 days (Table 3). During TCCS monitoring, no significant differences in the reperfusion-to-TCCS time and mean blood pressure were detected between two groups (p = 0.058 and 0.057, respectively) (Table 2).

ROC curves for prediction of poor functional outcomes

Table 4 displays the results of the univariate ROC analysis for PSV ratio, adjusted PSV ratio, MFV ratio and adjusted MFV ratio. The adjusted PSV ratio and adjusted MFV ratio achieved higher area under the curve (AUC) values than the others (p < 0.05). The adjusted PSV ratio had a sensitivity of 56.7%, a specificity of 81.3%, a positive predictive value of 79.1%, and a negative predictive value of 60.0%. The adjusted MFV ratio had a sensitivity of 58.3%, a specificity of 93.8%, a positive predictive value of 92.1%, and a negative predictive value of 64.3%. No statistically significant difference was observed in the AUC values between the adjusted PSV ratio and the adjusted MFV ratio (p = 0.672). Therefore, both ratios were further evaluated in the predictive models for their ability to predict poor functional outcomes in patients with AIS undergoing EVT. We used ROC curves to assess the effectiveness of the prediction models. Model A included age, admission NIHSS score, ASPECT score, and onset-to-reperfusion time and achieved an AUC of 0.75 (95% CI [0.66–0.84]) (Fig. 3). Adding the adjusted MFV ratio (Model C), but not the adjusted PSV ratio (Model B), to Model A significantly improved the predictability of 90-day poor functional outcomes compared with that in Model A (p = 0.013). The ROC of Model C had an AUC of 0.85 (95% CI [0.79–0.92]; Hosmer-Lemeshow goodness-of-fit test: χ² = 11.67; p = 0.167), which indicated that the predictive model achieved good discrimination and calibration.