Herein, 121 patients with successful recanalization after endovascular treatment of acute cerebral infarction were divided into futile recanalization group and recanalization group (69 men and 52 women; age, 28–91 years; average age, 65 years). Table 1 lists all data of patients in both groups, including their demographics, disease details, treatment characteristics, and complications.
Table 1 Baseline characteristics of patients with acute large vessel occlusion treated with endovascular treatment
|
Futile Recanalization
(n = 31)
|
Effective recanalization
(n = 90)
|
Inspection value
|
P value
|
Demographics
|
|
|
|
|
Age (years)
|
72 ± 11
|
63 ± 12
|
3.84c
|
<0.01
|
Sex, male
|
15 (48.0)
|
54 (60.0)
|
1.27b
|
0.26
|
Hypertension
|
23 (74.0)
|
50 (55.6)
|
3.35b
|
0.09
|
Diabetes mellitus
|
8 (25.8)
|
24 (26.7)
|
0.009
|
0.925
|
Cardiovascular diseases
|
13 (41.9)
|
23 (25.6)
|
2.960
|
0.085
|
Atrial fibrillation
|
14 (45.1)
|
28 (31.1)
|
2.009
|
0.156
|
Dyslipidemia
|
25 (80.6)
|
69 (76.7)
|
0.211
|
0.646
|
Homocysteine levels
|
12 (10, 25)
|
12 (10,17)
|
-0.52 a
|
0.601
|
Glycosylated hemoglobin
|
6.1 (5.7, 6.5)
|
6.0 (5.7, 7.3)
|
-0.21 a
|
0.837
|
D-dimer
|
730 (311, 2774)
|
344 (177, 965)
|
-2.98 a
|
<0.01
|
Stroke etiology
|
|
|
|
|
Large-artery atherosclerosis
|
12 (38.7)
|
47 (52.2)
|
1.69b
|
0.19
|
Cardioembolism
|
18 (58.1)
|
36 (40.0)
|
3.64b
|
0.06
|
Others
|
1 (3.2)
|
7 (7.8)
|
0.77b
|
0.38
|
Initial NIHSS score
|
23 (18, 26)
|
12 (7, 16)
|
-5.57 a
|
<0.01
|
Discharged-NIHSS
|
23 (17, 36)
|
4 (2, 11)
|
-7.28 a
|
<0.01
|
Pre-stroke mRS 0 or 1
|
31 (100)
|
84 (93.3)
|
-
|
0.336d
|
Time from onset to admission
|
250 (120, 387)
|
273(167, 473)
|
-0.95 a
|
0.342
|
Time from puncture to recanalization
|
120 (75, 167)
|
135 (90, 200)
|
-1.44 a
|
0.149
|
Infection
|
30 (96.8)
|
47 (52.2)
|
19.78b
|
<0.01
|
sICH
|
11 (35.5)
|
18 (20.0)
|
3.03b
|
0.082
|
HT
|
12 (38.7)
|
18 (20.0)
|
4.33b
|
0.037
|
Stroke recurrence
|
1 (3.2)
|
1 (1.1)
|
-
|
0.448d
|
Death
|
10 (32.2)
|
0 (0)
|
-
|
<0.01d
|
Note: a is the Z value of the Mann–Whitney U test; b is χ2 Inspection χ2 value; c is the t-value of the t-test; d is Fisher's exact probability test, and there is no corresponding test value.
3.2 Selection of variable characteristics
Univariate analysis showed that in the futile recanalization group and effective recanalization group, the age was 72 ± 11 years and 63 ± 12 years (t = 3.84, P ≤ 0.01), D-dimer levels were 730 (311, 2774) and 344 (177, 965) (Z = -2.98, P = 0.012), the baseline NIHSS score was 11 (5, 19) and 16 (11, 21) (Z = -0.359, P < 0.01), the NIHSS score at discharge was 23 (17, 36) and 4 (2,11) (Z = -7.28, P < 0.01], infection was noted in 30 cases (96.8%) and 47 cases (52.2%) (χ2 = 19.78, P < 0.01); intravenous tirofiban was administered in 9 cases (29.0) and 50 cases (55.6) (χ2 = 6.49, P = 0.011), and deaths occurred in 10 cases (32.2%) and 0 cases (0%) (P < 0.01), respectively. There was no statistically significant difference in the characteristics of the covariates between the groups (Table 1). Based on above results of univariate analysis and previous literature reports, the known variables that may be related to futile recanalization were included in lasso regression analysis. The 16 variables include age, the TOAST classification, homocysteine levels, D-dimer levels, post-infarction-related infection, intravenous tirofiban use, glycosylated hemoglobin, the NIHSS score at discharge, the baseline NIHSS score, symptomatic intracranial hemorrhage, time from symptom onset to treatment, anticoagulant therapy, pre-stroke mRS 0 or 1, bleeding transformation, postoperative mTICI score, and recanalization time. Finally, based on the 121 patients in the cohort (Figures 2A and 2B), these 16 variables were reduced to 12 potential predictors by the lasso algorithm, and there were non-zero coefficients in the lasso regression model. These characteristics included age, intravenous tirofiban use, homocysteine levels, the TOAST classification, glycosylated hemoglobin levels, D-dimer levels, infection prevalence, sICH, the NIHSS score at discharge, anticoagulation, the pre-stroke mRS score, and bleeding transformation.
3.3 Development of a prediction model
Table 2 enlists the characteristic variables finally included in the risk prediction model of futile recanalization of endovascular treatment for patients with acute cerebral infarction. These variables were age, the TOAST classification, the NIHSS score at discharge, infection prevalence, and logistic regression analysis results for intravenous tirofiban use. The model containing these independent predictors was depicted as a nomogram (Figure 3).
Table 2. Characteristics of logistic regression model for recanalization of acute cerebral infarction after futile recanalization endovascular treatment
|
B
|
S.E.
|
Exp (B)
|
Wald
|
P value
|
95% CI
|
|
-12.740
|
3.970
|
0.000
|
10.297
|
0.001
|
-
|
Age
|
0.085
|
0.043
|
1.089
|
3.896
|
0.048
|
1.001–1.185
|
TOAST
|
-2.266
|
1.019
|
0.104
|
4.947
|
0.026
|
0.014–0.764
|
Infection
|
5.967
|
2.115
|
390.184
|
7.955
|
0.005
|
6.175–24656.320
|
Discharged-NIHSS
|
0.339
|
0.088
|
1.404
|
14.865
|
0.000
|
1.182–1.669
|
Tirofiban use
|
-2.020
|
0.998
|
0.133
|
4.097
|
0.043
|
0.019–0.938
|
3.4 Model Verification
The C-index of the nomogram used to predict the risk of futile recanalization after endovascular treatment of acute cerebral infarction was 0.975 (95% CI: 0.951–0.998, P < 0.001); the discrimination was good (Figure 3), and the absolute error of the calibration curve was 0.027, showing good calibration/consistency in this cohort (Figure 4). The queue prediction model is confirmed to be 0.975 through the internal verification of 500 resampling bootstrap. This implies that the nomogram of invalid recanalization risk has good generalizability.
3.5 Evaluation of clinical utility of prediction model
The decision curve analysis of the invalid recanalization risk prediction nomogram is shown in figures 5 and 6. Simple, the simple model includes age, TOAST classification, NIHSS score after discharge, and intravenous tirofiban. Complex, complex models include age, TOAST classification, NIHSS score after discharge, tirofiban use, and infection. The decision curve shows that in the range of 0.02–0.80, the net benefit of the complex model is higher than that of the simple model.