Sage Journals: Discover world-class research

Abstract

Background: Alteplase intravenous thrombolysis is effective for treating acute ischemic stroke (AIS) within 4.5 h. Nevertheless, the prognosis remains poor for some patients.

Objective: To investigate the risk factors for poor prognosis in patients undergoing intravenous thrombolysis with alteplase following AIS based on propensity score matching and to develop a predictive model.

Result: Multivariate logistic regression analysis showed that baseline blood glucose (OR = 1.20, 95%CI, 1.03-1.39), baseline NIH Stroke Scale score (OR = 1.23, 95%CI, 1.12-1.35), and hyperlipidemia (OR = 6.60, 95%CI 1.74-25.00) were risk factors for poor prognosis in patients with AIS undergoing alteplase intravenous thrombolysis. Using these factors, a nomogram model was constructed for predicting patient prognosis at 3 months. The areas under the receiver operating characteristic curve (AUCs) of the training and validation groups were 0.792 (95CI% 0.715-0.870) and 0.885 (95CI% 0.798-0.972), respectively, showing good differentiation. The Hosmer Lemeshow goodness-of-fit test showed that the model had good fit. The calibration curve fitted well with the ideal curve, and the decision curve analysis curve showed that the model had good clinical applicability when the threshold probability was between 10%-80%.

Conclusion: The established nomogram could successfully predict the 3-month prognosis of patients with AIS after undergoing alteplase intravenous thrombolysis. The model thus has clinical application value.

Keywords

propensity score matching acute ischemic stroke alteplase prediction model

Introduction

With the increase in China’s aging population, the prevalence and disability rates of stroke are increasing year by year. A survey showed that stroke was the leading cause of death in Chinese adults, with acute ischemic stroke (AIS) accounting for about 80% of cases, resulting in a significant burden on both society and the economy.¹ Recombinant tissue plasminogen activator (rt-PA) is an effective therapy for AIS, and intravenous thrombolysis within 4.5 h has been demonstrated to significantly reduce disability following AIS.^2,3 However, not all patients benefit from intravenous thrombolytic therapy. A significant number of patients, even after receiving thrombolytic therapy, are at increased risk of hemorrhage transformation and fatal intracranial hemorrhage, often leading to poor prognosis.⁴ The efficacy of intravenous thrombolytic therapy is significantly related to the time of administration. Before treatment, assessment of the risks and benefits of thrombolytic therapy may delay the timely vascular recanalization of patients and hinder the recovery of neurological function. Therefore, an accurate early assessment of the prognosis of patients with AIS within the time window of intravenous thrombolytic therapy would be helpful for decision-making regarding treatment, thereby reducing time delay. In recent years, there have been numerous studies describing predictive models for the prognosis of patients receiving intravenous thrombolysis for AIS. The early and accurate prediction of the long-term prognosis of AIS is not only helpful to guide the decision-making of clinicians but can also provide an objective reference for the expectations of patients and their families.

A nomogram is a statistical tool used to predict the probability or risk of a specific event. It provides more accurate and personalized predictions by the comprehensive consideration of multiple factors, assisting clinicians in understanding the potential risks of patients, thus leading to more informed clinical decisions. Nomograms are often used to predict disease risk, treatment effects, and survival rates.⁵ The model is presented in a graphical visualization that allows doctors and researchers to quickly estimate risk without the need for complex calculations. For example, a nomogram based on the baseline NIH Stroke Score (NIHSS), blood sugar, and blood cholesterol level for patients with partial-and full anterior circulation infarction (OCSP type) were used to predict patient outcomes following intravenous thrombolysis. It was found that the areas under the receiver operating characteristic (ROC) curve (AUCs) for the training and validation cohorts were 0.767 (95%CI, 0.653-0.772) and 0.836 (95%CI, 0.697-0.847), respectively, showing good predictive accuracy.⁶ For AIS, early rehabilitation has been shown to positively affect functional outcomes at 3 months.^7,8 However, there is a lack of research on the diagnosis, rehabilitation, and prognosis of stroke recovery, which are currently mainly based on the doctor’s experience.^9,10 Therefore, researchers have used the TOAST classification, antihypertensive drugs, 15-day Barthel Index score, and 15-day mRS score to predict adverse functional outcomes in the recovery of patients after AIS. A nomogram based on these factors was found to have good predictive ability in both the training (AUC = 0.950) and test (AUC = 0.942) groups.¹¹ At present, most prognostic nodes for AIS use the time point of 3 months, and there are few studies on longer prognosis. Multivariate regression analysis of 149 patients with acute ischemic stroke showed that physical inactivity, neutrophil count, cerebrovascular disease score, and length of hospital stay were independent risk factors for poor 6-month functional recovery. The AUC for the nomogram based on the above indicators was 0.936, and had a good internal validation consistency index.¹² However, these models tend to involve many variables, have complex calculation formulae, low predictive efficacies, or require the use of specialized software, and thus present challenges for their application in clinical practice. Ideally, the main indicators involved in a predictive model should be clinical data and imaging information that can be easily and rapidly obtained.^13–15

The randomized controlled trial design is regarded as the most scientific and rigorous research design method, but such trials are often difficult to conduct in the real world due to medical ethical issues and complex operational difficulties, and observational methods are thus used in most investigations.¹⁶ Observational studies have the disadvantage that the allocation of study groups is not random, and their allocation is affected by the characteristics of individual patients and patient selection by the investigators, which inevitably leads to differences in some baseline covariates between the study group and the validation group, potentially leading to bias and research results that do not accurately reflect the real situation. Statistical methods are needed to reduce confounding.¹⁷ Propensity Score Matching (PSM) computes a score from a baseline covariate that varies among subjects and rebalances it between the study and validation groups, so that measured and unmeasured confounders are evenly distributed in both groups, making the results more reliable.¹⁸

In this study, the PSM method was used for the accurate identification of factors related to poor prognosis after intravenous thrombolysis of alteplase in AIS. A nomogram was then constructed to predict the risk of poor prognosis in patients with AIS, to provide a more accurate basis for clinical decision-making.

Data and methods

General information

The study retrospectively enrolled 254 patients with AIS who underwent rt-PA intravenous thrombolysis in the Department of Neurology of Deyang People’s Hospital from January 2018 to December 2020.

Inclusion criteria: (1) Symptoms of neurological impairment and diagnosis of AIS; (2) Alteplase intravenous thrombolytic indication¹⁹; (3) Symptom onset time <4.5 h; (4) Aged 18 or above.

Exclusion criteria: (1) Patients with intravenous thrombolysis contraindications in alteplase; (2) Patients with incomplete clinical data and those lost to follow-up; (3) An mRS score ≥2 points before intravenous thrombolysis; (4) Patients with severe heart, liver, and renal insufficiency; (5) Patients with infections, blood diseases, malignant tumors. or rheumatic immune system diseases.

Data collection

The clinical characteristics of each patient were collected, consisting of the following three types: (1) General data, including age, sex, weight, and smoking; (2) Comorbidities and history, including hypertension, hyperlipidemia, diabetes, atrial fibrillation, previous stroke history, congestive heart failure history, and coronary heart disease; (3) Other indicators, including DNT, ONT, baseline systolic and diastolic blood pressure, baseline blood glucose, white blood cell count, red blood cell count, neutrophil ratio, hemoglobin, platelet count, PT, APTT, INR, fibrinogen, D-dimer, Na, K, urea nitrogen, baseline mRS Score, baseline NIHSS score, and mRS Score at 3 months after the disease.²⁰ All blood samples were collected at the time of hospital admission but before alteplase intravenous thrombolytic therapy.

Intravenous thrombolysis

All patients were treated with intravenous thrombolysis with recombinant tissue plasminogen activator (rt-PA) manufactured by Boehringer Ingelheim (Germany). The therapeutic dose of rt-PA was 0.9 mg/kg, of which 10% of the dose was injected intravenously within 1 min, and the remaining 90% was pumped intravenously within 60 min. The maximum dose of rt-PA should not exceed 90 mg.

Grouping, observation indicators and related definitions

The main outcome index of the study was the mRS score at 3 months. This score was used as an indicator to evaluate the good/poor outcome after intravenous thrombolysis (mRS scores were 0, 1, 2, 3, 4, 5, and 6, with 6 representing death), and the patients were divided into the good prognosis group (mRS Score ≤2, assigned a value of 0) and the poor prognosis group (mRS Score >2, assigned a value of 1).²¹ The general data (potential confounding factors) of the two groups were matched using PSM to achieve a balance in the general data of the two groups. The two matched patient groups were compared, and statistically different risk factors were included in a multifactorial logistic regression model to further identify risk factors for poor prognosis following alteplase intravenous thrombolysis in AIS. The results of the matched logistic regression analyses were then used to construct a nomogram to predict poor prognosis in patients with AIS following alteplase intravenous thrombolysis, together with validation of the findings in the validation group, while receiver operating characteristic (ROC) curves, decision curve analysis (DCA), and calibration curves were used to evaluate the predictive performance of the model (Figure 1).

Figure 1.

Flowchart.

Statistical analysis

In this study, R version 4.3 and SPSS 25.0 were used for statistical analysis. Normally distributed data were analyzed using t-tests and are presented as means and standard deviations (x ± s). Non-parametric tests were used for the analysis of non-normally distributed data and the data are shown as medians M (P25,P75). Clinical features were compared using chi-square, Mann-Whitney U, Kruskal-Wallas, or Fisher’s exact tests, and are expressed as n(%). In the orientation score matching, Age, Weight, Sex and Smoke were used as matching variables, the matching ratio was set to 1:1, and the caliper value was set to 0.02. With outcome indicators as dependent variables, univariate analysis was performed on the matched data, variables with P < 0.05 were included in univariate and multivariate Logistic regression, and the risk factors for poor prognosis were screened by backward method based on Wald statistics. Based on the above risk factors, Rstudio software (Data processing and analysis were performed using R(4.3), along with Zstats V1.0) was used to establish the Normograph and complete the internal and external verification web page. Receiver operating curve (ROC curve), clinical decision curve (DCA) and calibration curve were used to evaluate the prediction effect of the model. An inspection level of α = 0.05 and P < 0.05 were considered statistically significant.

Results

Comparison of general information before PSM

A total of 254 patients were included in the study. Of these, 142 cases were allocated to the good prognosis group (mRS ≤2 points) and 112 to the poor prognosis group (mRS >2 points); the incidence rate of poor prognosis was 44.09% (112/254). There were significant differences between the two groups in terms of age, weight, and smoking history (P < 0.05), while the groups did not differ in terms of sex (P > 0.05), as shown in Table 1.

Table 1.

Comparison of general information between the two groups before PSM.

Variable	Total (n = 254)	Good prognosis group	Poor prognosis group	Z/x²	p
Variable	Total (n = 254)	n = 142	n = 112	Z/x²	p
Age	69.00 (60.00, 78.00)	66.00 (56.00, 75.00)	73.00 (64.00, 80.00)	Z = −4.108	<.001
Weight	60.00 (55.00, 70.00)	65.00 (58.12, 70.38)	60.00 (54.00, 68.25)	Z = −3.111	0.002
Sex				χ² = 3.191	0.074
No	100 (39.37)	49 (34.51)	51 (45.54)
Yes	154 (60.63)	93 (65.49)	61 (54.46)
Smoke				χ² = 11.154	<.001
No	174 (68.5)	85 (59.86)	89 (79.46)
Yes	80 (31.5)	57 (40.14)	23 (20.54)

Comparison of general information between the two groups after propensity score matching

The PSM of the general information of the two groups was carried out with a matching ratio of 1:1 using the nearest-neighbour matching method, and 92 patients in the good prognosis group and 92 patients in the poor prognosis group were obtained after matching. The unbalanced variables between the two groups were balanced after matching (P > 0.05) (Table 2).

Table 2.

Comparison of general information between the two groups after propensity score matching.

Variable	Total (n = 184)	Good prognosis group	Poor prognosis group	Z/x²	p
Variable	Total (n = 184)	n = 92	n = 92	Z/x²	p
Age	70.00 (62.00, 77.00)	70.50 (62.00, 78.00)	70.00 (62.00, 77.00)	Z = −0.407	0.684
Weight	60.00 (55.00, 70.00)	60.00 (55.00, 70.00)	60.00 (55.00, 70.00)	Z = −0.307	0.759
Sex				χ² = 0.000	1.000
No	84 (45.65)	42 (45.65)	42 (45.65)
Yes	100 (54.35)	50 (54.35)	50 (54.35)
Smoke				χ² = 0.029	0.866
No	137 (74.46)	68 (73.91)	69 (75.00)
Yes	47 (25.54)	24 (26.09)	23 (25.00)

Propensity score matching equilibrium test

The standardized mean (SMD) was used for the balance test. As shown in Figure 2, the SMD of each continuous variable after PSM was less than 10%, and it could thus be considered that the two groups of patients reached equilibrium in terms of baseline clinical characteristics.

Figure 2.

Equilibrium test.

Common support test for propensity score matching

The co-support test is another important hypothesis test used in PSM.¹⁸ This is performed using kernel density plots, essentially observing that the two groups have the same propensity score value. As shown in Figure 3, in terms of the pre-matching kernel density (left part), there was a large difference in the distribution of the P score values between the good and poor prognosis groups, which if ignored (without PSM and direct analysis of the difference), is likely to lead to biased conclusions in the study. At the same time, it is also evident that there was an overlap between the two groups, indicating that the two groups were eligible to use the PSM model. From the nuclear density after matching (the right part), it can be seen that after matching, the two groups appeared to cover and approach each other, with a large common value interval, indicating that the two groups were very close to one another in terms of clinical characteristics, with good matching, and thus passed the co-support test.

Figure 3.

Co-support test.

One-way analysis of the clinical characteristics of the two groups after matching

The matched data were used for one-way analysis of the clinical characteristics of the two groups. This showed that there were significant differences between the two groups in terms of baseline blood glucose, leukocyte count, baseline NIHSS, hyperlipidemia, and atrial fibrillation (P < 0.05) (Table 3).

Table 3.

One-way analysis of the clinical characteristics of the two groups after matching.

Variables	Total (n = 184)	0 (n = 92)	1 (n = 92)	t/Z/x²	p
Baseline systolic blood pressure	161.61 ± 26.74	158.75 ± 25.07	164.47 ± 28.16	t = −1.45	0.148
Neutrophil ratio	63.03 ± 13.84	61.98 ± 12.15	64.08 ± 15.34	t = −1.03	0.304
Hemoglobin	136.45 ± 15.10	134.98 ± 15.52	137.92 ± 14.60	t = −1.33	0.186
APTT	31.52 ± 6.27	31.41 ± 6.55	31.63 ± 6.00	t = −0.24	0.814
DNT	62.00 (47.00, 82.50)	57.50 (45.75, 82.00)	65.00 (49.00, 84.00)	Z = −1.69	0.090
ONT	146.00 (120.00, 192.00)	146.50 (120.75, 192.00)	146.00 (119.25, 191.25)	Z = −0.61	0.544
Baseline diastolic blood pressure	87.00 (78.00, 100.00)	86.50 (78.00, 98.00)	88.00 (77.00, 103.25)	Z = −0.81	0.420
Baseline blood glucose	7.40 (6.40, 9.21)	7.00 (5.95, 8.55)	7.88 (6.80, 10.03)	Z = −3.14	0.002
Leukocyte count	7.10 (6.00, 8.62)	6.65 (5.40, 8.33)	7.40 (6.77, 9.25)	Z = −2.94	0.003
Erythrocyte count	4.52 (4.09, 4.82)	4.50 (3.97, 4.79)	4.55 (4.21, 4.88)	Z = −1.30	0.195
Platelet count	192.00 (154.75, 228.50)	193.00 (157.25, 219.00)	190.50 (154.00, 234.25)	Z = −0.01	0.992
PT	12.65 (12.00, 13.70)	12.70 (12.07, 13.70)	12.60 (11.88, 13.70)	Z = −0.40	0.687
INR	0.99 (0.93, 1.07)	0.99 (0.93, 1.07)	0.99 (0.93, 1.07)	Z = −0.33	0.745
Fibrinogen	3.29 (2.70, 3.90)	3.20 (2.68, 3.70)	3.39 (2.70, 4.10)	Z = −1.33	0.185
D dimer	0.97 (0.48, 2.98)	0.92 (0.47, 2.22)	1.07 (0.53, 3.29)	Z = −1.26	0.208
K	3.67 (3.33, 3.94)	3.70 (3.39, 3.97)	3.59 (3.29, 3.86)	Z = −1.45	0.146
Baseline NIHSS	9.00 (6.00, 14.00)	7.00 (5.00, 10.00)	12.50 (8.75, 17.25)	Z = −6.35	<.001
Baseline mRS	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)	Z = −1.14	0.252
Urea nitrogen	5.55 (4.40, 7.02)	5.05 (4.08, 6.93)	6.09 (4.72, 7.11)	Z = −1.70	0.089
History of strok				χ² = 3.56	0.059
No	169 (91.85)	88 (95.65)	81 (88.04)
Yes	15 (8.15)	4 (4.35)	11 (11.96)
Hypertension				χ² = 0.26	0.607
No	45 (24.46)	21 (22.83)	24 (26.09)
Yes	139 (75.54)	71 (77.17)	68 (73.91)
Hyperlipidemia				χ² = 7.81	0.005
No	154 (83.70)	84 (91.30)	70 (76.09)
Yes	30 (16.30)	8 (8.70)	22 (23.91)
Atrial fibrillation				χ² = 9.33	0.002
No	116 (63.04)	68 (73.91)	48 (52.17)
Yes	68 (36.96)	24 (26.09)	44 (47.83)
Coronary heart disease				χ² = 0.00	1.000
No	172 (93.48)	86 (93.48)	86 (93.48)
Yes	12 (6.52)	6 (6.52)	6 (6.52)
Diabetes				χ² = 0.03	0.864
No	139 (75.54)	70 (76.09)	69 (75.00)
Yes	45 (24.46)	22 (23.91)	23 (25.00)
Congestive heart failure				χ² = 1.87	0.172
No	175 (95.11)	90 (97.83)	85 (92.39)
Yes	9 (4.89)	2 (2.17)	7 (7.61)

APTT: activated partial thromboplastin time;DNT: Door to Needle Time; ONT: Onset to Needle Time;PT: ProthrombinTime;INR: international normalized ratio;K: potassium.

Logistic analysis of risk factors for poor prognosis

Factors that had been found to differ significantly between the two groups in the univariate analysis, namely, baseline blood glucose, leukocyte count, baseline NIHSS, hyperlipidemia, and atrial fibrillation, were included in the multivariate analysis, finding that baseline blood glucose, baseline NIHSS, and hyperlipidemia were risk factors for poor prognosis (P < 0.05) (Table 4).

Table 4.

Risk Factors for Poor Prognosis.

Variables	Univariate analysis			Multivariate analysis
Variables	B	OR (95%CI)	p	B	OR (95%CI)	p
Baseline blood glucose	0.13	1.14 (1.01 - 1.29)	0.040	0.18	1.20 (1.03 - 1.39)	0.019
Baseline NIHSS	0.17	1.18 (1.09 - 1.28)	<.001	0.21	1.23 (1.12 - 1.35)	<.001
Hyperlipidemia
No	Reference			Reference
Yes	1.44	4.20 (1.30 - 13.57)	0.016	1.89	6.60 (1.74 - 25.00)	0.006

Construction of the nomogram for the prediction of poor prognosis

Based on the results of the multivariate logistic regression analysis, a nomogram was constructed for the prediction of poor prognosis following intravenous thrombolysis with alteplase after AIS, as shown in Figure 4. When determining the risk of poor prognosis for a specific patient, it is first necessary to determine the values of each variable included in the study. Different values of variables can yield corresponding scores at the corresponding positions in the nomogram. The scores of all the variables included in the study are then summed to obtain the total score. Based on the total score, the risk axis (Risk axis) at the bottom of the column chart can be predicted and thus the corresponding probability of poor prognosis after intravenous thrombolysis with ateplase in AIS.

Figure 4.

Nomogram model.

Nomogram evaluation

Internal and external validation and calibration curves

A total of 184 patients were randomly assigned to either the training group (n = 128) or the validation group (n = 56) in a ratio of 7:3. There was no significant difference in clinical characteristics between the two groups (P > 0.05, Table 5). Bootstrap resampling (repeated sampling times = 1000 times) was used to verify and evaluate the degree of calibration of the model both internally (in the training group) and externally (in the validation group), allowing the determination of the calibration chart. The results showed that the calibration curves were consistent with the ideal curves, indicating a good consistency between the predicted and actual situations (Figure 5).

Table 5.

Comparison of clinical characteristics between the training and validation groups.

Variable	Total (n = 184)	Training section (n = 128)	Validation group (n = 56)	t/Z/X²	p
Baseline systolic blood pressure	161.61 ± 26.74	162.02 ± 26.31	160.68 ± 27.91	t = 0.311	0.756
Neutrophil ratio	63.03 ± 13.84	63.09 ± 13.49	62.89 ± 14.72	t = 0.090	0.929
Hemoglobin	136.45 ± 15.10	136.66 ± 15.87	135.96 ± 13.28	t = 0.289	0.773
APTT	31.52 ± 6.27	32.17 ± 5.79	30.02 ± 7.07	t = 2.004	0.048
Age	70.00 (62.00 - 77.00)	70.00 (62.75 - 77.00)	69.00 (62.00 - 78.25)	Z = 0.003	0.998
DNT	62.00 (47.00 - 82.50)	62.00 (48.75 - 82.00)	62.50 (45.00 - 84.25)	Z = 0.516	0.606
ONT	146.00 (120.00 - 192.00)	145.00 (120.00 - 193.50)	152.00 (119.50 - 188.50)	Z = 0.376	0.707
Baseline diastolic blood pressure	87.00 (78.00 - 100.00)	87.50 (78.00 - 101.00)	85.00 (77.00 - 99.00)	Z = 0.998	0.318
Baseline blood glucose	7.40 (6.40 - 9.21)	7.27 (6.30 - 8.83)	7.45 (6.57 - 9.54)	Z = 0.921	0.357
Weight	60.00 (55.00 - 70.00)	60.00 (55.00 - 68.50)	60.00 (54.75 - 70.25)	Z = 0.606	0.546
Leukocyte count	7.10 (6.00 - 8.62)	7.10 (6.00 - 8.62)	7.10 (6.00 - 8.67)	Z = 0.226	0.821
Erythrocyte count	4.52 (4.09 - 4.82)	4.54 (4.11 - 4.87)	4.49 (4.04 - 4.77)	Z = 0.481	0.630
Platelet count	192.00 (154.75 - 228.50)	192.00 (148.75 - 228.50)	191.00 (159.75 - 227.25)	Z = 0.445	0.656
PT	12.65 (12.00 - 13.70)	12.70 (12.10 - 13.72)	12.50 (11.70 - 13.60)	Z = 1.512	0.131
INR	0.99 (0.93 - 1.07)	0.99 (0.93 - 1.07)	0.99 (0.94 - 1.05)	Z = 0.249	0.804
Fibrinogen	3.29 (2.70 - 3.90)	3.31 (2.71 - 3.86)	3.28 (2.57 - 3.98)	Z = 0.199	0.843
D dimer	0.97 (0.48 - 2.98)	1.01 (0.48 - 2.98)	0.91 (0.48 - 2.79)	Z = 0.126	0.899
Na	138.80 (137.10 - 140.30)	139.00 (137.40 - 140.43)	138.35 (136.17 - 139.83)	Z = 1.965	0.052
K	3.67 (3.33 - 3.94)	3.70 (3.35 - 3.95)	3.59 (3.30 - 3.93)	Z = 0.605	0.545
Urea nitrogen	5.45 (4.41, 7.22)	5.01 (4.07, 6.88)	6.03 (4.76, 7.31)	Z = −1.82	0.086
Baseline mRS	0.00 (0.00 - 0.00)	0.00 (0.00 - 0.00)	0.00 (0.00 - 0.00)	Z = 2.150	0.476
Baseline NIHSS	9.00 (6.00 - 14.00)	9.00 (6.00 - 14.00)	10.00 (6.00 - 15.25)	Z = 0.758	0.449
Sex				χ² = 0.033	0.856
No	84 (45.65)	59 (46.09)	25 (44.64)
Yes	100 (54.35)	69 (53.91)	31 (55.36)
History of stroke				χ² = 3.221	0.073
No	169 (91.85)	114 (89.06)	55 (98.21)
Yes	15 (8.15)	14 (10.94)	1 (1.79)
Smoke				χ² = 0.065	0.798
No	137 (74.46)	96 (75.00)	41 (73.21)
Yes	47 (25.54)	32 (25.00)	15 (26.79)
Hypertension				χ² = 1.010	0.315
No	45 (24.46)	34 (26.56)	11 (19.64)
Yes	139 (75.54)	94 (73.44)	45 (80.36)
Hyperlipidemia				χ² = 1.549	0.213
No	154 (83.7)	110 (85.94)	44 (78.57)
Yes	30 (16.3)	18 (14.06)	12 (21.43)
Atrial fibrillation				χ² = 0.053	0.817
No	116 (63.04)	80 (62.50)	36 (64.29)
Yes	68 (36.96)	48 (37.50)	20 (35.71)
Coronary heart disease				χ² = 1.438	0.231
No	172 (93.48)	122 (95.31)	50 (89.29)
Yes	12 (6.52)	6 (4.69)	6 (10.71)
Diabetes				χ² = 0.067	0.795
No	139 (75.54)	96 (75.00)	43 (76.79)
Yes	45 (24.46)	32 (25.00)	13 (23.21)
Congestive heart failure				χ2 = 0.032	0.859
No	175 (95.11)	121 (94.53)	54 (96.43)
Yes	9 (4.89)	7 (5.47)	2 (3.57)

APTT: activated partial thromboplastin time;DNT: Door to Needle Time; ONT: Onset to Needle Time;PT: ProthrombinTime;INR: international normalized ratio;K: potassium.

Figure 5.

Calibration curves for internal (A, training group) and external (B, validation group) calibration.

ROC curves

The model differentiation was evaluated by ROC curves. The areas under the ROC curves (AUC) for the training group and the validation group were 0.792 (0.715-0.870) and 0.885 (0.798-0.972), respectively. The Hosmer-Lemeshow goodness-of-fit test was used to evaluate the degree of fit, and the results showed that the model had good accuracy in predicting poor prognosis (Figure 6).

Figure 6.

ROC curves for the training group (A) and the validation group (B).

DCA curves

The DCA curves were compiled to evaluate the clinical practicability of the nomogram(Figure 7). In the DCA curve shown in Figure 7, when the threshold probability was between 10 and 80%, the net benefit of using the nomogram to predict poor prognosis after intravenous thrombolysis of alteplase in AIS was higher, and the inverse model showed good clinical practicability.

Figure 7.

DCA curves of the training group (A) and the validation group (B).

Discussion

As primary prevention based on traditional risk factors has become more widespread, the incidence of AIS has been greatly reduced.²² However, AIS can still lead to significant disability, reducing the patient’s ability to perform tasks associated with daily life and imposing major burdens on their families and society.^23,24 Reperfusion therapy is one of the main treatment options for AIS and can significantly improve the prognosis of patients.²⁵ Intravenous thrombolytic therapy with alteplase performed during the early stage of acute cerebral infarction is both safe and reliable. However, due to differences in age, sex, body weight and other factors, some patients may experience complications such as symptomatic intracranial hemorrhage and vasogenic edema, and the recovery of neurological function may even be reduced after thrombolytic therapy, indicating that not all patients within the time window should be treated with rt-PA intravenous thrombolytic therapy.²⁶ Therefore, the early accurate identification and prediction of groups at high risk of poor prognosis, together with early intervention and treatment, are of great importance in improving the quality of life of patients after AIS.

In this study, PSM was used to achieve a balance between the good prognosis group and the poor prognosis group in terms of age, weight, sex, and smoking, among other factors. Logistic regression analysis found that baseline NIHSS, baseline blood glucose, and hyperlipidemia were risk factors for poor prognosis in patients with AIS after alteplase intravenous thrombolysis. The NIHSS score is widely used in clinical neurology, mainly for evaluating the severity of stroke. Some studies have also found that the NIHSS score has a specific relationship with the prognosis of AIS and the transformation of symptomatic intracranial hemorrhage.^27,28 In recent years, the prediction of AIS based on the NIHSS score has been gradually developed, such as the use of the NIHSS score combined with time of onset. A Japanese study investigated the use of a modified rating scale of time and OTT, using the data of patients who had received rt-PA thrombolysis at 10 stroke centers to obtain a simple formula for calculating prognosis: NIHSS-time score = initial NIHSS score × OTT (h). A score ≤20 indicated good prognosis, while a score ≥40 indicates poor prognosis.²⁹ In addition, the model was subsequently validated externally, and it was found that NIHSS-time score predicted the prognosis of patients after receiving endovascular therapy with good accuracy (OR 0.372,95%CI 0.175-0.789).³⁰ In addition, studies have found that compared with the current NIHSS score at the onset of stroke, the NIHSS score on day 7 is more accurate for predicting prognosis, possibly because the condition of the patient is not stable for 7 days after stroke onset and is prone to change, so that a correlation between the NIHSS score on the day of the stroke and the outcome on day 90 is not ideal.^27,31 The Lausanne Stroke Scale (Acute Stroke Registry and Analysis of Stroke) Lausanne, ASTRAL score, is a tool used to predict the outcome of stroke at 3 months; this includes the patient’s age, NIHSS score, time to hospital, visual field, acute stage blood glucose, and level of consciousness. External validation in the European population showed that it had good predictive efficacy. The advantage of this score is that the calculation process is relatively simple, and includes skull imaging results.^32,33 Elevated blood glucose is closely related to the prognosis of alteplase intravenous thrombolysis. Current guidelines recommend that patients with random blood glucose levels higher than 22.2 mmol/l should undergo hypoglycemic therapy before thrombolysis. Hyperglycemia can affect the prognosis of patients with AIS by aggravating neurotoxicity in the ischemic penumbral zone, increasing the permeability of the blood-brain barrier, and thus exacerbating intracranial hemorrhage transformation and brain edema, increasing the risk of post-stroke infection, increasing platelet aggregation, and thus increasing thrombus volume, shown by the therapeutic effects of anti-platelet aggregation.^34–37 Blood lipid levels are closely related to the prognosis of patients with AIS. Studies have shown that the probability of recurrence after AIS is increased by 25% for every 1 mmol/L increase in total cholesterol level of patients following AIS, with the risk of recurrence increasing by 15% for every 1 mmo/L increase in the triglyceride level. Abnormal levels of low-density lipoprotein (LDL) in serum are closely related to stroke recurrence and elevated levels are associated with the formation of atherosclerotic plaque, which when enlarged or detached, can lead to ischemic stroke. At present, in terms of the prevention and treatment of stroke, reductions in LDL levels are used as an important indicator to determine the status of blood lipid levels, and reducing the levels of LDL can effectively reduce the risk of poor prognosis of AIS. A number of studies have shown that statins can significantly reduce LDL levels and have a clear effect on improving the prognosis and recurrence of patients with ischemic stroke.^38–40

At present, while there are a small number of studies on predictive models related to the prognosis of patients with AIS, none of these models can be widely used in clinical practice, mainly for the following reasons: (1) Few studies have provided simple and rapid predictive models, and some studies have included small sample sizes⁴¹; (2) At present, most studies have focused on simple univariate analysis and logistic regression, without analyzing confounding factors and bias risks⁴²; (3) Few existing studies have undergone rigorous internal and external validation¹⁵; (4) There are few prognostic models for patients receiving alteplase intravenous thrombolysis in China.⁴³ Nomograms are an important part of modern medical decision-making, and represent a graphical statistical tool that estimates the probability of an event occurring by integrating a number of variables to derive a continuous scoring system, essentially using a line graph as a visualization of the results of a regression equation.⁴⁴ For these reasons, the present study adopted a retrospective cohort study approach using PSM to balance between-group confounders, and then constructed a nomogram prediction model based on the risk factors for poor prognosis identified by logistic regression analysis to evaluate the risk of poor prognosis in patients with AIS after alteplase intravenous thrombolysis. The constructed nomogram could predict a patient’s prognosis using only three pieces of information that can be quickly and easily accessed in the clinic, namely, the baseline NIHSS score, baseline glucose level, and hyperlipidemia. For example, a patient with baseline blood glucose of 8 mmol/l, baseline NIHSS score of 10, and a history of comorbid hyperlipidemia, corresponding to the point values of 18, 28, and 32.5, respectively, and yielding a total score of 78.5, would be predicted to have projected to have a 0.80 likelihood of poor prognosis at three months. In the DCA curve, the Treat All line represents the net benefit when patients undergo clinical interventions, while the Treat None line represents the net benefit when patients do not undergo clinical interventions. When the net benefit received corresponding to the threshold probability is at the upper right of both the Treat All and Treat None lines (i.e., the Treat model line in this study, representing the nomogram), this indicates that the model has good clinical utility for making predictions. The results of ROC and calibration curves also showed that the prediction and actual occurrence associated with the model were better.

However, this study has the following limitations: First, Subject to strict inclusion conditions and regional factors, the sample size was relatively small, and the sample size was not calculated. This limitation may introduce statistical errors, cautioning against broad generalizations of our findings. More data need to be collected in the future to verify the prediction accuracy of the model. Second, Patients treated with intravenous thrombolytic urokinase were not included in this study. Moreover, the retrospective nature of our research design represents an inherent limitation.

Conclusions

In conclusion, this study developed a predictive model for poor prognosis after intravenous thrombolysis with alteplase in patients with AIS based on three factors, namely, baseline blood glucose, baseline NIHSS, and hyperlipidemia.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Trial registration

This randomized clinical trial was not registered because this is an observational and anonymous study.

Ethical statement

ORCID iD

Pan Huang

References

Zhou

Wang

Zeng

, et al. (2019) Mortality, morbidity, and risk factors in China and its provinces, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet 394(10204): 1145–1158.

Hacke

Donnan

Fieschi

, et al. (2004) Association of outcome with early stroke treatment: pooled analysis of ATLANTIS, ECASS, and NINDS rt-PA stroke trials. Lancet 363(9411): 768–774.

Lees

Bluhmki

von Kummer

, et al. (2010) Time to treatment with intravenous alteplase and outcome in stroke: an updated pooled analysis of ECASS, ATLANTIS, NINDS, and EPITHET trials. Lancet 375(9727): 1695–1703.

Wahlgren

Ahmed

Dávalos

, et al. (2007) Thrombolysis with alteplase for acute ischaemic stroke in the safe implementation of thrombolysis in stroke monitoring study(SITS-MOST): an observational study. Lancet 369(9558): 275–282.

Battersby

Bouliotis

Emmertsen

, et al. (2018) Development and external validation of a nomogram and online tool to predict bowel dysfunction following restorative rectal cancer resection: the POLARS score. Gut 67(4):688–696.

Ping

Min

Qiuyun

Qingke

(2022) Prognostic nomogram for the outcomes in acute stroke patients with intravenous thrombolysis. Frontiers in Neuroscience 16: 1017883. DOI: 10.3389/fnins.2022.1017883.

Winstein

Stein

Arena

, et al. (2016) Guidelines for adult stroke rehabilitation and recovery: a guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke 47(6): e98–e169.

Ahmed

Mauad

VAQ

Gomez-Rojas

, et al. (2020) The impact of rehabilitation-oriented virtual reality device in patients with ischemic stroke in the early subacute recovery phase: study protocol for a phase iii, single-blinded, randomized, controlled clinical trial. Journal of Central Nervous System Disease 12: 1179573519899471.

Moradi

Mafi

Shariat

Cleland

Nakhostin Ansari

Savari

(2021) Neurorehabilitation, the practical method of returning to work after stroke. Iranian Journal of Public Health 50: 209–210.

10.

Preston

Ada

Stanton

Mahendran

Dean

(2021) Prediction of independent walking in people who are nonambulatory early after stroke: asystematic review. Stroke 52: 3217–3224.

11.

Yan

Zheng

Zhang

, et al. (2023) Development and validation of a nomogram model for predicting unfavorable functional outcomes in ischemic stroke patients after acute phase. Frontiers in Aging Neuroscience 15: 1161016.

12.

Huang

Luo

Liu

(2022) A novel nomogram for predicting poor 6-month function in patients with acute ischemic stroke receiving thrombolysis. The Journal of Cardiovascular Nursing 37(6): E206–E216.

13.

Kent

Ruthazer

Decker

, et al. (2015) Development and validation of a simplified stroke-thrombolytic predictive instrument. Neurology 85(11): 942–949.

14.

Bonkhoff

Grefkes

(2022) Precision medicine in stroke: towards personalized outcome predictions using artificial intelligence. Brain: A Journal of Neurology 145: 457–475.

15.

Strbian

Engelter

Michel

, et al. (2012) Symptomatic intracranial hemorrhage after stroke thrombolysis: the SEDAN score. Annals of Neurology 71(5): 634–641.

16.

Zabor

Kaizer

Hobbs

(2020) Randomized controlled trials. Chest 158(1s): S79–S87.

17.

Schober

Vetter

(2020) Propensity score matching in observational research. Anesthesia and Analgesia 130(6): 1616–1617.

18.

Mitchell

Williams

Atherton

Larvin

Lund

Narici

(2012) Sarcopenia, dynapenia, and the impact of advancing age on human skeletal muscle size and strength; a quantitative review. Frontiers in Physiology 3: 260.

19.

Powers

Rabinstein

Ackerson

, et al. (2019) Guidelines for the early management of patients with acute ischemic stroke: 2019 update to the 2018 guidelines for the early management of acute ischemic stroke: a guideline for healthcare professionals from the American heart association/American stroke association. Stroke 50: e344–418.

20.

Heldner

Zubler

Mattle

, et al. (2013) National institutes of health stroke scale score and vessel occlusion in 2152 patients with acute ischemic stroke. Stroke 44: 1153–1157.

21.

Huybrechts

Caro

Xenakis

Vemmos

(2008) The prognostic value of the modified Rankin Scale score for long-term survival after first-ever stroke. Results from the Athens Stroke Registry. Cerebrovascular Diseases 26(4): 381–387.

22.

Zhang

, et al. (2022) Nontraditional risk factors for residual recurrence risk in patients with ischemic stroke of different etiologies. Cerebrovascular Diseases (Basel, Switzerland) 51(5): 630–638.

23.

Liu

, et al. (2019) Stroke in China: advances and challenges in epidemiology, prevention, and management. The Lancet. Neurology 18(4): 394–405.

24.

Sennfalt

Norrving

Petersson

Ullberg

(2019) Long-term survival and function after stroke: a longitudinal observational study from the Swedish stroke register. Stroke 50: 53–61, STROKEAHA118022913.

25.

Bala

Singh

Buck

, et al. (2023) Safety and efficacy of tenecteplase compared with alteplase in patients with large vessel occlusion stroke: a prespecified secondary analysis of the ACT randomized clinical trial. JAMA Neurology 80(8): 824–832.

26.

IST-3 collaborative group Sandercock

Wardlaw

, et al. (2012) The benefits and harms of intravenous thrombolysis with recombinant tissue plasminogen activator within 6 h of acute ischaemic stroke (the third international stroke trial [IST-3]): a randomised controlled trial. Lancet 379(9834): 2352–2363.

27.

Adams

Davis

Leira

, et al. (1999) Baseline NIH stroke scale score strongly predicts outcome after stroke A report of the trial of org 10 172 in acute stroke treatment(TOAST). Neurology 53(1): 126–131.

28.

Blackburn

D'arco

Devito

, et al. (2021) Predictors of motor outcome after childhood arterial ischemic stroke. Developmental Medicine and Child Neurology 63(10): 1171–1179.

29.

Aoki

Kimura

Koga

, et al. (2013) NIHSS-time score easily predicts outcomes in rt-PA patients:the SAMURAI rt-PA registry. Journal of the Neurological Sciences 327(1-2): 6–11.

30.

Todo

Sakai

Kono

, et al. (2016) National institutes of health stroke scale-time score predicts outcome after endovascular therapy in acute ischemic stroke: a retrospective single-center study. Journal of Stroke and Cerebrovascular Diseases: The Official Journal of National Stroke Association 25(5): 1187–1191.

31.

Bang

Park

Yoon

, et al. (2005) Predicting the long-term outcome after subacute stroke within the middle cerebral artery territory. Journal of Clinical Neurology (Seoul, Korea) 1: 148–158.

32.

Michel

Odier

Rutgers

, et al. (2010) The Acute STroke Registry and Analysis of Lausanne (ASTRAL): design and baseline analysis of an ischemic stroke registry including acute multimodal imaging. Stroke 41(11): 2491–2498.

33.

Ntaios

Faouzi

Ferrari

Lang

Vemmos

Michel

(2012) An integer-based score to predict functional outcome in acute ischemic stroke: the ASTRAL score. Neurology 78: 1916–1922.

34.

Kersten

CJBA

Zandbergen

AAM

Berkhemer

, et al. (2022) Association of hyperglycemia and computed tomographic perfusion deficits in patients who underwent endovascular treatment for acute ischemic stroke caused by a proximal intracranial occlusion: a subgroup analysis of a randomized phase 3 trial (MR CLEAN). Journal of the Neurological Sciences 440: 120333.

35.

Sharma

Zhang

Akter

, et al. (2023) Permeability of metformin across an in vitro blood-brain barrier model during normoxia and oxygen-glucose deprivation conditions: role of organic cation transporters (octs). Pharmaceutics 15: 1357. DOI: 10.3390/pharmaceutics15051357.

36.

Dai

Cao

Wang

, et al. (2023) Impacts of stress hyperglycemia ratio on early neurological deterioration and functional outcome after endovascular treatment in patients with acute ischemic stroke. Frontiers in Endocrinology 14: 1094353.

37.

Schiattarella

Carrizzo

Ilardi

, et al. (2018) Rac1 modulates endothelial function and platelet aggregation in diabetes mellitus. Journal of the American Heart Association 7: e007322. DOI: 10.1161/JAHA.117.007322.

38.

Wang

Han

, et al. (2022) Triglyceride-glucose index and stroke recurrence inelderly patients with ischemic stroke. Frontiers in Endocrinology 13: 1005614.

39.

Wang

Dai

Zhang

, et al. (2020) Oxidized low-density lipoprotein (LDL) and LDL cholesterol are associated with outcomes of minor stroke and TIA. Atherosclerosis 297: 74–80.

40.

Prinz

Endres

(2011) Statins and stroke: prevention and beyond. Current Opinion in Neurology 24(1): 75–80.

41.

Ling

Shen

Yang

(2019) Development of a prediction model for 1-year poor prognosis in patients with acute ischemic stroke. J Invest Med : The Official Publication of the American Federation for Clinical Research 67: 957–963.

42.

Nisar

Hanumanthu

Khandelwal

(2019) Symptomatic intracerebral hemorrhage after intravenous thrombolysis:predictive factors and validation of prediction models. Journal of Stroke and Cerebrovascular Diseases: The Official Journal of National Stroke Association 28(11): 104360.

43.

Wang

Zeng

Wang

Zhuang

Chen

Miao

(2019) Risk factors of hemorrhagic transformation after intravenous thrombolysis with rt-PA in acute cerebral infarction. QJM: Monthly Journal of the Association of Physicians 112(5): 323–326.

44.

Zacharek

, et al. (2020) ABCA1/ApoE/HDL signaling pathway facilitates myelination and oligodendrogenesis after stroke. International Journal of Molecular Sciences 21(12): 4369.

Risk factors and a model for prognosis prediction after intravenous thrombolysis with alteplase in acute ischemic stroke based on propensity score matching

Abstract

Keywords

Introduction

Data and methods

General information

Data collection

Intravenous thrombolysis

Grouping, observation indicators and related definitions

Statistical analysis

Results

Comparison of general information before PSM

Comparison of general information between the two groups after propensity score matching

Propensity score matching equilibrium test

Common support test for propensity score matching

One-way analysis of the clinical characteristics of the two groups after matching

Logistic analysis of risk factors for poor prognosis

Construction of the nomogram for the prediction of poor prognosis

Nomogram evaluation

Internal and external validation and calibration curves

ROC curves

DCA curves

Discussion

Conclusions

Footnotes

Declaration of conflicting interests

Funding

Trial registration

Ethical statement

ORCID iD

References