Previous studies have identified various factors associated with the outcomes of acute ischemic stroke (AIS) but considered only 1 or 2 predictive factors. The present study aimed to use outcome-related factors derived from biochemical, imaging and clinical data to establish a logistic regression model that can predict the outcome of patients with AIS following endovascular treatment (EVT). The data of 118 patients with anterior circulation AIS (ACAIS) who underwent EVT between October 2014 and August 2018 were retrospectively analyzed. The patients were divided into 2 groups based on the modified Rankin Scale score at three months after surgery, where 0–2 points were considered to indicate a favorable outcome and 3–6 points were considered a poor outcome. Non-conditional logistic stepwise regression was used to identify independent variables that were significantly associated with patient outcome, which were subsequently used to establish a predictive statistical model, receiver operating characteristic (ROC) curve was used to show the performance of statistical model and analyze the specific association between each factor and outcome. Among the 118 patients, 47 (39.83%) exhibited a good and 71 (60.17%) exhibited a poor outcome. Multivariate analysis revealed that the predictive model was statistically significant (χ2=78.92; P<0.001), and that the predictive accuracy of the model was 83.1%, which was higher compared with that obtained using only a single factor. ROC curve analysis shows the area under curve of the statistical model was 0.823, the analysis of diagnostic threshold for prognostic factors indicated that age, diffusion-weighted imaging lesion volume, glucose on admission, National Institutes of Health Stroke Scale score on admission and hypersensitive C-reactive protein were valuable predictive factors for the outcome of EVT (P<0.05). In conclusion, a predictive model based on non-conditional logistic stepwise regression analysis was able to predict the outcome of EVT for patients with ACAIS.
Acute ischemic stroke (AIS) is the leading cause of death in China and one of the primary mortality-associated factors worldwide (
A number of biochemical, clinical and imaging factors have been identified to be associated with the prognosis of patients with AIS. In terms of biochemical factors, the levels of glucose on admission (AG), high-sensitivity C-reactive protein (hs-CRP) and N-terminal pro-brain natriuretic peptide (NT-proBNP) on admission all associated with the outcome of patients with AIS (
Of the imaging characteristics, diffusion-weighted imaging (DWI) lesion volume, cerebral blood flow, cerebral blood volume and mean transit time have been associated with patient outcome (
At present, the number of studies predicting the outcome of EVT for AIS is relatively low, and the outcomes for patients who received EVT have only been predicted using single factors belonging to a single category; they have not been comprehensively predicted using a combination of biochemical, imaging features and clinical indicators. The present study aimed to use widely recognized prognostic indictors, including biochemical, imaging and clinical factors to predict the outcome of EVT, and to provide a reference for the establishment of personalized treatment plans.
A retrospective analysis was performed using the clinical and imaging data of 169 patients with anterior circulation acute ischemic stroke (ACAIS) who received EVT between October 2014 and August 2018 at the Second Clinical College of Guangzhou University of Chinese Medicine (Guangzhou, China). All patients accepted mechanical thrombectomy using a solitaire stent retriever; if the vessel remained narrow or the distal circulation remained poor following thrombectomy, angioplasty was subsequently performed. A total of 10 patients with malignant tumors or neurological diseases were excluded, 12 patients were excluded due to a history of stroke, 6 due to having received unknown pre-treatment at another hospital and 23 patients dropped out during the follow-up period. Ultimately, 118 patients were included in the present study. The clinicopathological and demographic data of all patients were obtained from the hospital information management system and image examination center, and included age, sex, risk factors, DWI images, AG, hs-CRP, high-density lipoprotein (HDL), low-density lipoprotein (LDL) and total cholesterol (TC) levels, in addition to systolic blood pressure (SBP), diastolic blood pressure (DBP), ORT, NIHSS score on admission and clinical outcome (
All patients met the following inclusion criteria: i) Diagnosed with AIS according to the 2018 Guidelines for the Early Management of Patients with Acute Ischemic Stroke (
The baseline assessments were performed within 1.5 h of admission and included evaluations of the NIHSS score, AG, blood pressure, demographic data, history of smoking or drinking and history of diabetes mellitus or hypertension. A history of smoking and drinking was defined as the patient admitting that they had previously smoked or consumed alcohol. A history of high blood pressure was defined as a systolic blood pressure >140 mmHg or diastolic blood pressure >90 mmHg at any point prior to admission. Diabetes was defined as a blood glucose level >11.1 mmol/l. A fasting venous blood sample was used to detect blood lipids and hs-CRP levels were determined in the early morning and within 24 h of admission. Neurological assessments were performed by evaluating the NIHSS and mRS scores at admission and during the follow-up examination (3 months after surgery). The 3-month mRS score was used to assess patient outcome; a score of 0–2 was considered as a favorable outcome and a score of 3–6 was considered to indicate a poor outcome (
MRI was performed using a MAGNETOM Verio 3.0T scanner (Siemens AG). The MRI protocol included DWI, apparent diffusion coefficient, T1-weighted imaging, T2-weighted imaging, fluid-attenuated inversion recovery imaging and MR angiography. The DWI scan parameters were as follows: B value, 0 and 1,000; echo time, 100 msec; repetition time, 6,600 msec; phrase encoding direction, AàP; field of view (FOV) read, 220 mm; FOV phase, 100%; slice thickness, 5.0 mm; slice spacing, 1.0 mm; slices, 20; bandwidth, 1,002 Hz/Pixel; echo spacing, 1.08 msec; fat saturation mode, weak; concatenation, 1; and scan time, 74 sec. Within the DWI images, the lesions were manually delineated using ImageJ Software (version 1.8.0; National Institutes of Health) by three associate chief physicians (initials, BL, GL and WZ) (
Continuous or numerical variables were expressed as the mean ± standard deviation and median (interquartile range). Shapiro-Wilk test was used to test the normality of variances and Levene test was used to test the homogeneity of variances. Normally distributed variables with assumed homogeneity of variance were compared using the independent-samples t-test, and non-normally distributed variables or those without homogeneity of variance were compared using the Mann-Whitney U-test. Categorical variables were expressed as the composition ratio or rate and comparisons were performed using Pearson's χ2 or Fisher's exact test. Non-conditional logistic stepwise regression was used to identify factors that were significantly associated with patient outcome, which can establish the optimal statistical model. The specificity and sensitivity of the threshold value for each predictive factor were analyzed by receiver operating characteristic (ROC) curve analysis. Statistical analyses were performed using SPSS 19.0 for Windows (IBM Corp.) and P<0.05 was considered to indicate a statistically significant difference.
Patient outcome was predicted using non-conditional logistic stepwise regression analysis, and the follow-up result at 3 months after EVT was used as the dependent variable (Favorable outcome, Y=0; Poor outcome, Y=1). The results indicated that the model was able to effectively predict patient outcome (χ2=78.916; P<0.001) and that the coincidence rate (accuracy) of the model was calculated to be 83.1% (
ROC curve analyses showed that the Area Under the ROC Curve (AUC) of the statistical model was calculated to be 0.823 (
EVT is one of the main methods to treating AIS at present. In order to determine the treatment plan after EVT, clinicians need to predict the outcome of patients with AIS who underwent EVT. However, as the outcome of AIS is under the influence of a variety of factors, the accuracy of in predicting the outcome of EVT by using only one or two factors is relatively low. In the present study, non-conditional logistic stepwise regression was used to identify the factors significantly associated with patient outcome, including biochemical, clinical and imaging parameters, which were then applied to establish an optimal predictive statistical model. In addition, multivariate logistic regression model is convenient and stable, and is commonly applied in many outcome prediction models of some diseases (
Previous studies have predicted the outcome of AIS patients using only one or two factors, particularly the NIHSS score and age, which have been recognized as the most direct contributors to patient outcome (
Age is a clinical indicator of the aging of cells of the nervous system. Aging results in molecular damage, organelle dysfunction and cellular injury within the neurovascular unit, which leads to structural and functional impairments (
Since brain tissues have strict requirements regarding recanalization time, the outcome of patients with AIS has been indicated to correlate with the ORT (
In addition, the present study indicated that males had a more favorable outcome compared with females; this supported the results of a study by White
The DWI infarct lesion volume indicates the size of the area of damaged brain tissue. Numerous studies have indicated that the DWI infarct lesion volume was associated with the outcome of patients with AIS and may serve as an independent predictor for outcome (
Among the biochemical factors used to predict patient outcome, AG is the most representative. It is generally accepted that AG represents an independent predictor of patient outcome following AIS (
In addition, various studies have indicated that LDL and blood pressure may be considered to be risk factors for AIS, which may be associated with the prognosis of AIS (
As clinical, imaging and biochemical factors indicate the severity of AIS from different aspects (and the abnormalities of these factors may occur at different times), the accuracy of predicting the outcome of AIS by combining three different types of factor may be higher than that achieved with single factors alone. Regarding the prediction of the outcome of EVT using a combination of clinical, biochemical and imaging factors, the present study achieved an accuracy of 83.1%, which is higher than that obtained with any single factor alone (as determined by ROC curve analysis) in the present and in previous studies (
The present study had several limitations. First, due to the limitation of the condition, not all of the indicators associated with patient outcome were investigated (e.g., homocysteine and NT-proBNP levels) and abnormal alterations to these variables may still cause injury to the brain. Investigation of additional factors in future studies may further enhance the predictive accuracy. Furthermore, the present study had a relatively small sample size and was a single-center study lacking the validation of multicenter results.
In conclusion, a predictive model based on non-conditional logistic stepwise regression analysis was able to predict the outcome of EVT following AIS. Patients belonging to different outcome groups displayed significant differences in sex, age, DWI lesion volume, hs-CRP, NIHSS score on admission and/or AG. The present study provided novel insight into the predictive capacity of these factors for patients with AIS.
We would like to acknowledge the valuable cooperation of The Department of Neurosurgery, The Second Clinical College of Guangzhou University of Chinese Medicine (Guangzhou, China) in evaluating patient NIHSS score and mRS as part of their daily clinical work.
No funding was received.
The datasets used and/or analyzed during the present study are available from the corresponding author on reasonable request.
BL and XW conceived and designed the study. XW, XL, YW, SZ and WL collected and organized the data. BL, GL and WZ manually delineated the DWI infarct lesion data. XW and AO analyzed and interpreted the data, and XW wrote the manuscript.
The present study was approved by the Ethics Committee of The Second Clinical College of Guangzhou University of Chinese Medicine (Guangzhou, China). The present study is a retrospective clinical analysis without any additional intervention for the patients. The requirement for informed consent was waived due to the retrospective nature of this analysis.
Not applicable.
The authors declare that they have no competing interests.
anterior circulation acute ischemic stroke
glucose on admission
acute ischemic stroke
brain natriuretic peptide
diffusion-weighted imaging
endovascular treatment
high-density lipoprotein
high-sensitivity C-reactive protein
low-density lipoprotein
modified Rankin Scale
National Institutes of Health Stroke Scale
onset-to-reperfusion time
receiver operating characteristic
Manual delineation of the infarct lesion in a DWI map (B=1,000) by referring to the ADC map of the same cross section. (A) Outline of the DWI lesion. (B) ADC map corresponding to DWI slice. DWI, diffusion-weighted imaging; ADC, apparent diffusion coefficient.
Receiver operating characteristic curves of factors associated with outcome. DWI, diffusion-weighted imaging; AG, glucose on admission; NIHSS, National Institute of Health Stroke Scale score at admission; hs-CRP, hypersensitive C-reactive protein; ORT, onset-to-reperfusion time.
Demographic and baseline data of the patients (categorical variables).
Item | Poor outcome |
Favorable outcome |
Sum | χ2 | P-value |
---|---|---|---|---|---|
Sex | |||||
Males | 37 (52.1) | 35 (74.5) | 72 | 5.94 | 0.02 |
Females | 34 (47.9) | 12 (25.5) | 46 | ||
Smoking | |||||
Yes | 20 (28.2) | 19 (40.4) | 39 | 1.92 | 0.23 |
No | 51 (71.8) | 28 (59.6) | 79 | ||
Drinking | |||||
Yes | 7 (9.9) | 9 (19.1) | 16 | 2.08 | 0.18 |
No | 64 (90.1) | 38 (80.9) | 102 | ||
Hypertension | |||||
Yes | 45 (63.4) | 29 (61.7) | 74 | 0.03 | 1.00 |
No | 26 (36.6) | 18 (38.3) | 44 | ||
Diabetes | |||||
Yes | 18 (25.4) | 7 (14.9) | 25 | 1.85 | 0.25 |
No | 53 (74.6) | 40 (85.1) | 93 |
The mRS score at 3 months after EVT was used as the efficacy criterion, in which 0–2 was considered as a favorable outcome and 3–6 was considered as poor outcome.
Patient data in association with the outcome.
Poor outcome (n=71) |
Favorable outcome (n=47) |
|||||
---|---|---|---|---|---|---|
Item | Mean (SD) | Median (P75-P25) | Mean (SD) | Median (P75-P25) | t/Z | P-value |
Age (years) | 70.5 (10.2) | 72 (14) | 63.49 (11.21) | 62.0 (16.0) | 3.53 | <0.01 |
DWI lesion volume (mm3) | 30,914.6 (41489.7) | 12,921.9 (34015.8) | 7,724.8 (7606.2) | 5,840.3 (6852.0) | 3.60 | <0.01 |
AG (mmol/l) | 9.6 (4.6) | 8.1 (3.2) | 7.1 (1.9) | 6.6 (2.3) | 3.93 | <0.01 |
NIHSS score | 15.0 (5.8) | 14.0 (7.0) | 9.4 (3.4) | 9.0 (5.0) | 5.40 | <0.01 |
ORT (min) | 376.4 (232.2) | 300.0 (225.0) | 301.4 (141.4) | 270.0 (177.0) | 1.61 | 0.11 |
hs-CRP (mg/l) | 52.0 (36.3) | 45.2 (32.0) | 39.0 (25.8) | 33.1 (33.5) | 2.13 | 0.03 |
Operation time (min) | 142.1 (56.4) | 135 (80) | 133.6 (56.3) | 130 (46) | 0.82 | 0.41 |
SBP (mmHg) | 152.4 (25.1) | 152 (34) | 153.7 (24.9) | 149.0 (31.0) | 0.07 | 0.94 |
DBP (mmHg) | 84.9 (15.0) | 81 (25) | 88.8 (15.7) | 85 (21) | 1.33 | 0.18 |
HDL (mmol/l) | 1.2 (0.3) | 1.1 (0.2) | 1.1 (0.3) | 1.1 (0.3) | 1.04 | 0.30 |
LDL (mmol/l) | 2.9 (1.0) | 2.8 (1.2) | 2.9 (0.7) | 2.9 (1.1) | 0.26 | 0.79 |
TC (mmol/l) | 4.6 (1.1) | 4.4 (1.5) | 4.3 (0.8) | 4.2 (1.1) | 1.31 | 0.19 |
The mRS score at 3 months after EVT was used as the efficacy criterion, which 0–2 was considered as a favorable outcome and 3–6 was considered as poor outcome (
Rank-Sum test. DWI, diffusion-weighted imaging; AG, glucose on admission; NIHSS, National Institutes of Health Stroke Scale; ORT, onset-to-reperfusion time; hs-CRP, high-sensitivity C-reactive protein; SBP, systolic blood pressure; DBP, diastolic blood pressure; HDL, high-density lipoprotein; LDL, low-density lipoprotein; TC, total cholesterol.
Omnibus tests of model coefficients.
Step 6 | χ2 | Degree of freedom | P-value |
---|---|---|---|
Step | 4.134 | 1 | 0.042 |
Block | 78.916 | 6 | <0.001 |
Model | 78.916 | 6 | <0.001 |
Classification Table.
Predicted | |||
---|---|---|---|
Groups | |||
Observed | Favorable outcome | Poor outcome | Percentage correct |
Step 6 | |||
Groups | |||
Favorable outcome | 37 | 10 | 78.7 |
Poor outcome | 10 | 61 | 85.9 |
Overall percentage | 83.1 |
The role of each factor for the prediction of AIS in the non-conditional logistic stepwise regression model
95% EXP (β) | |||||||
---|---|---|---|---|---|---|---|
Variable | β | S.E. | Wald | P-value | Exp (β) | Lower limit | Upper limit |
Sex | 1.221 | 0.611 | 3.984 | 0.046 | 3.389 | 1.022 | 11.235 |
Age | 0.090 | 0.031 | 8.537 | 0.003 | 1.094 | 1.030 | 1.163 |
DWI lesion volume | <0.001 | 0.001 | 11.208 | 0.001 | 1.000 | 1.000 | 1.000 |
AG | 0.421 | 0.153 | 7.602 | 0.006 | 1.524 | 1.129 | 2.055 |
NIHSS score | 0.193 | 0.072 | 7.108 | 0.008 | 1.213 | 1.052 | 1.397 |
ORT | 0.004 | 0.002 | 4.548 | 0.033 | 1.004 | 1.000 | 1.007 |
In the statistical mode, the mRS score at 3 months after EVT was used as the dependent variable (Favorable outcome, Y=0; Poor outcome, Y=1). DWI, diffusion weighted imaging; AG, admission glucose; NIHSS, National Institutes of Health Stroke Scale; ORT, onset-to-reperfusion time; β, regression coefficient; S.E., standard error; Exp (β), odds ratio; 95% EXP (β), 95% confidence interval of odds ratio.
Receiver operating characteristic curve analysis of statistical model for the outcome of patients with anterior circulation acute ischemic stroke
Asymptotic 95% CI | ||||
---|---|---|---|---|
AUC | S.E. | P-value | Lower limit | Upper limit |
0.823 | 0.042 | <0.001 | 0.740 | 0.906 |
Null hypothesis: The classification cutoff value for each patient is 0.5. S.E., standard error.
Receiver operating characteristic curve analysis of predictive factors for the outcome of patients with anterior circulation acute ischemic stroke.
Asymptotic 95% CI | |||||||
---|---|---|---|---|---|---|---|
Outcome measure | AUC | Cut-off value | P-value | Lower limit | Upper limit | Specificity (%) | Sensitivity (%) |
Age | 0.677 | 68.5 | 0.001 | 0.578 | 0.776 | 63.4 | 61.7 |
DWI lesion volume | 0.696 | 6494.5 | <0.001 | 0.603 | 0.790 | 70.4 | 61.7 |
AG | 0.714 | 7.2 | <0.001 | 0.620 | 0.809 | 73.2 | 61.7 |
NIHSS score | 0.794 | 11.5 | <0.001 | 0.713 | 0.874 | 76.1 | 70.2 |
hs-CRP | 0.616 | 39.5 | 0.034 | 0.512 | 0.720 | 60.6 | 63.8 |
ORT | 0.588 | 272.5 | 0.108 | 0.483 | 0.692 | 53.5 | 55.3 |
DWI, diffusion-weighted imaging; AG, glucose on admission; NIHSS, National Institutes of Health Stroke Scale; hs-CRP, high-sensitivity C-reactive protein; ORT, onset-to-reperfusion time; HDL, high-density lipoprotein; LDL, low-density lipoprotein; TC, total cholesterol; SBP, systolic blood pressure; DBP, diastolic blood pressure; CI, confidence interval; AUC, Area Under the ROC Curve.