Introduction

Mechanical thrombectomy (MT) is the standard of care for patients with large vessel occlusions (LVO) of the proximal anterior circulation1. An essential criterion for this treatment is that cerebral infarction is not established, and that salvageable brain tissue (so called penumbra), remains, to produce neurological improvement and avoid post-reperfusion neurological complications (hemorrhagic transformation/malignant cerebral edema)2. To this end, many clinical trials with different time windows have been conducted in recent years3,4.

To indicate MT within the first 6 h of an acute ischemic stroke (AIS) with a LVO in the anterior circulation is sufficient to estimate the salvageable brain tissue through a clinical-radiological mismatch, that is, an NIHSS (National Institutes of Health Stroke Scale) score >6 and an ASPECTS (Alberta Stroke Program Early Computed Tomography Score) score ≥ 6 on non-contrast computed tomography (NCCT)1. For time windows between 6 and 24 h, the “neuroimaging” criterion predominates, requiring advanced neuroimaging techniques, such as CT perfusion (CTP) and Diffusion-weighted magnetic resonance imaging (DW-MRI), with or without MRI perfusion1. However, the access and utilization of these modalities face several limitations such as heterogeneity in acquisition and postprocessing parameters, potentially hindering comparisons between studies6, as well as geographical variations in imaging protocols and resources7. A recent clinical trials and observational studies also supports the benefit of mechanical thrombectomy in patients with large core infarcts, showing improved functional outcomes and reduced mortality, even in expanded populations beyond traditional imaging-based criteria8.

It would be of interest to have estimates of salvageable brain tissue based on clinical parameters and simple neuroimaging9,10,11,12, which would also help to estimate which patients are most likely to haver early neurological improvement (ENI), independently of the time window and the procedural factors associated to MT13. Recent studies show that the key to patient selection may lie in the clinical-radiological mismatch, thus is, a NIHSS score on admission higher than expected based on the ASPECT score14,15. Indeed, after 24 h, when brain infarction is fully established, there is a strong negative lineal correlation between NIHSS and ASPECTS and, on average, an increase of 10 points on NIHSS corresponds to a decrease of ≈ 3 points on ASPECTS16. However, no studies are available that measure the discordance between the NIHSS on admission and the theoretical NIHSS that would be expected based on clinical and neuroimaging parameters.

The hypothesis of the present study is that it is feasible to develop a predictive model that measures the discrepancy between the “real” and “expected” NIHSS score on admission (referred to as NIHSSmismatch%) in patients with AIS caused by a LVO in the anterior circulation, based on clinical data and the ASPECTS score on NCCT; and this NIHSSmismatch% has a short-term prognostic value in patients undergoing MT. Thus, the objectives were: (1) To develop a model to estimate the “expected” NIHSS of a patient, considering the extension of the stroke, measured by the ASPECTS score on the concurrent cranial CT, as well as clinical parameters; (2) To calculate the NIHSSmismatch% as the percentage discrepancy between the “expected” NIHSS and the “real” NIHSS, and (3) to perform an internal validation of the predictive capacity of the NIHSSmismatch% for ENI in patients undergoing successful reperfusion.

Methods

Study design

A retrospective cross-sectional modeling study was conducted using a prospective database including patients aged ≥ 18 years with acute ischemic stroke (AIS) and large vessel occlusion (LVO) who were evaluated at Torrecárdenas University Hospital (TUH) in Almería, Spain, between 2017 and 2022. This hospital serves as the reference Stroke Center for a population of ≈ 700,000 inhabitants. Modeling was performed to estimate the “expected” NIHSSexpected of a patient on admission, followed by modeling to establish the value of NIHSSmismatch% that best predicts the development of ENI at 24 h.

The TRIPOD guidelines (Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis) were followed for study execution and reporting17.

Patient inclusion/exclusion criteria

The inclusion criteria were patients with ischemic stroke aged ≥ 18 years who were hospitalized in the Stroke Unit between 2017 and 2022, with a symptom duration window of < 24 h, and who had a LVO detected by CT angiography upon admission in the anterior territory (internal carotid artery and/or middle cerebral artery), and who underwent MT. The exclusion criteria were to present a transient ischemic attack or cerebral hemorrhage, patients with incomplete data, patients with single anterior cerebral artery occlusion, and patients with LVO in the posterior territory.

All patients were managed and treated according to the international guidelines1 and received general anesthesia with endotracheal intubation during the endovascular procedure.

Variables

Demographic data, comorbidities, vascular risk factors, neurological status (NIHSS score on admission and at 24 h), neuroimaging data on admission (NCCT plus multiphase CT angiography) and at 24 h (NCCT) including ASPECTS18, the most proximal location of arterial occlusion (M1, M2 or internal carotid artery), the affected hemisphere (right/left) and the presence of collaterals grade on CT-Angiography source images (CTA-SI) according to DAWN trial19 were collect. A single reader (M. F.-G.) determined the ASPECTS and collaterals grade on neuroimaging.

A NCCT were performed at 24+/−2 h hours according to our MT protocol and the presence of relevant neurological complications were recorded, such as parenchymal hemorrhage (PH1: hematoma within infarcted tissue, occupying < 30%, no substantive mass effect; PH2: hematoma occupying 30% or more of the infarcted tissue, with obvious mass effect)20 and severe cerebral edema (CED) (characterized by focal brain swelling with visible midline shift [grade 3]) according to SITS-MOST criteria21.

Early neurological improvement was defined as a reduction in the NIHSS score by ≥ 8 points or achieving an NIHSS score of 0/1 at 24 h, in patients undergoing MT22.

Sample size

It was a pilot study in which the NIHSSmismatch% was to be calculated for the first time, so there was no accurate estimate of the magnitude of the effect this parameter would have on ENI of patients with AIS undergoing MT. However, it was estimated that a sensitivity of at least 60% could be expected, with a 95% confidence interval and a lower limit of 50%, so at least 119 patients would be required23. To increase the study’s power, all available patients during the study period were included.

Statistical analysis

All statistical analyses were conducted with SPSS (IBM SPSS Statistics 28.0, SPSS Inc. Chicago, IL) and Python (version 3.11). Missing data were minimal (< 2%) across variables included in the main analyses. These missing values were assumed to be missing completely at random (MCAR) based on the clinical context and absence of systematic patterns. Therefore, no imputation was performed, and analyses were conducted using complete-case data. An exception was made for covariates in multivariable modeling, where appropriate imputation was applied to preserve sample size and avoid unnecessary bias. Comparisons between groups were analyzed with the χ 2 or Fisher’s exact test for dichotomous variables. Continuous variables were expressed as mean ± SD or median (interquartile range [IQR],) and compared with Student’s t test or the Mann-Whitney test, as appropriate. Comparison between two dependent continuous variables were performed with the Wilcoxon rank-sum test. The correlations between two quantitative variables or between two ordinal variables were performed with Spearman’s rho. Discrimination was assessed by calculating the area under the receiver operating characteristic (ROC) curve. Calibration was assessed by performing the R2 and the Hosmer–Lemeshow goodness-of-fit test.

Patients included in the analysis were divided into two cohorts: the derivation and the validation cohort. The derivation cohort comprised first 40% of patients treated with MT. The validation cohort included the remaining 60% of patients treated with MT (Fig. 1).

Fig. 1
figure 1

Flowchart. Between January 2017 and December 2022, 416 patients were attended for large vessel occlusion-related ischemic stroke. Reasons for non-inclusion and patients included in the derivation and validations cohort are listed in the figure. ACA, anterior cerebral artery; LVO, large vessel occlusion; MT, mechanical thrombectomy. * mTICI 2bc3.

Full size image

Calculation of NIHSSexpected and NIHSmismatch% on admission

This calculation was developed in the derivation cohort. Multivariate generalized linear models were performed to identify independent factors explaining the NIHSS at 24 h, when the cerebral infarction was assumed to be established, adjusted based on the ASPECT score from the 24 h-NCCT scan, as well potential confounders (age, sex and variables with a p-value < 0.2 in the bivariate analysis). With the β-coefficients and the intercept from this linear correlation, the regression line equation was obtained. Then, the formula was used to calculate the “expected” NIHSS upon the patient’s arrival at the Emergency Department, using neuroimaging data at admission (instead of at 24 h). After that, the following calculation was performed to determine the percentage discrepancy, or mismatch, between the “real” admission NIHSS (NIHSSactual) and the “expected” admission NIHSS (NIHSSexpected):

NIHSSmismatch% = (NIHSSactual – NIHSSexpectedl)/NIHSSactuall x 100.

Internal validation

To assess the discriminative performance of NIHSS mismatch% for predicting ENI, we performed receiver operating characteristic (ROC) curve analysis in both the derivation and validation cohorts. The area under the curve (AUC) was calculated for each cohort, and 95% confidence intervals were estimated using 1000 bootstrap iterations to account for sampling variability. To formally test whether the AUCs differed significantly between cohorts, we applied a z-test for independent AUCs, which evaluates whether the observed difference between AUCs exceeds what would be expected by chance, given their respective standard errors.

On the other hand, the relationship between NIHSSmismatch%, and the development of ENI was analyzed in the group of patients with successful MT (eTICI grades 2b, 2c or 3)24 in the validation cohort in two steps. First, another receiver ROC curve analysis was conducted determining the AUC as well as a cutoff point for NIHSSmismatch%. We considered the point at which the sum of the specificity and sensitivity was highest, giving the same weight to false-positives and false-negatives. Second, the relationship between this NIHSSmismatch% cutoff point and the developments of ENI in the validation cohort was assessed using multivariate logistic regression models. Variables with a value of P ≤ 0.2 in the bivariate testing were included. A forward stepwise logistic regression analysis was followed as the modelling strategy, using the log-likelihood ratio test to assess the correctness of fit and compare nested models. Two versions of the regression model were performed, one introducing NIHSSmismatch% as continuous variable and the other with it dichotomized according to the results of the ROC curve. Models were validated by a ROC curve analysis. 95% confidence intervals (CI) are presented. All tests were two-sided and P-values ≤ 0.05 were considered statistically significant.

Ethical considerations

All methods were carried out in accordance with relevant guidelines and regulations. All experimental protocols were approved by the Ethics Committee of the Torrecárdenas University Hospital. An informed consent was obtained from all subjects and/or their legal guardians.

Results

A total of 416 AIS patients with LVO were attended during the study period, of them 308 patients met de inclusion criteria, 123 in the derivation cohort and 185 in the validation cohort (Fig. 1).

Calculation of expected NIHSS (NIHSSexpected) on admission

In the derivation cohort, the mean (SD) age was 68.3 (14.9) and 58.5% were male. Median (IQR) of NIHSS on admission was 179 and at 24 h was 715. The medina (IQR) of NCCT-ASPECTS was 82 on admission and 74 at 24 h. Furthermore, at 24 h the NCCT showed parenchymal hemorrhagic transformation (PH1/PH2) in 11.4% whereas grade 3 CED was present in 16.3% of patients. Basal data of the derivation cohort are showed in Table S1.

Table 1 summarizes NIHSS scores at 24 h according to baseline characteristics in the derivation cohort (n = 123). Patients with diabetes mellitus had higher NIHSS scores compared with those without diabetes (13.5 [IQR 14] vs. 5 [IQR 14]; P = 0.009). Smokers showed a tendency toward lower NIHSS compared with non-smokers (3.5 [IQR 11] vs. 8 [IQR 18]; P = 0.091). Scores differed significantly by baseline NCCT-ASPECTS: 1 [IQR 4] for 9–10, 5 [IQR 11] for 6–8, and 19 [IQR 10] for 0–5 (P < 0.001). Thrombus location showed a statistical trend, with higher NIHSS in ICA occlusions (13 [IQR 15]) compared with M1 (6 [IQR 14]) and M2 (5 [IQR 11]) occlusions (P = 0.062). Left-sided occlusions were associated with higher scores than right-sided (11 [IQR 16] vs. 5 [IQR 10]; P = 0.012). Finally, parenchymal hemorrhagic transformation (PH1/PH2) and severe cerebral edema (grade 3 CED) were strongly associated with higher NIHSS scores (19 [IQR 6] and 21 [IQR 5], respectively) compared with those without these complications (5 [IQR 13] and 5 [IQR 12]; both P < 0.001).

Table 1 NIHSS scale score distribution at 24 h, according to baseline data, in the derivation cohort (n = 123).
Full size table

Multivariate generalized linear models were performed to identify independent factors explaining the NIHSS at 24 h based on age, sex, NCCT-ASPECTS at 24 h and other potential confounders showed in Table S2. With the approximated coefficients of this linear correlation, the following equation of the regression line was obtained:

NIHSS24 h = 10.8 + 0.05xAge + 2.5xDM + 2.6xHemisphere + 4.1xHemorrhagic transformation + 6.5xEdema – 1.5x ASPECTS24h.

Where age was a quantitative variable (years), DM was a dichotomous variable (1 = yes, 0 = no), hemisphere was a dichotomous variable (0 = right, 1 = left), Hemorrhagic transformation (PH1/PH2) was a dichotomous variable (1 = yes, 0 = no), edema (grade 3 CED) was a dichotomous variable (1 = yes, 0 = no), and ASPECTS24h (NCCT-ASPECTS at 24 h) was a quantitative variable (with scores ranging from 0 to 10).

That formula was then used to calculate the theoretical NIHSS that a patient should have on admission, based on the ASPECT score on the first NCCT performed on arrival to the Emergency Department. As patients eligible for MT should not present with hemorrhagic transformation or grade 3 CED prior to treatment, the formula was simplified as follow:

NIHSSexpected = 10.8 + 0.05xAge + 2.5xDM + 2.6xHemisphere – 1.5x ASPECTSadmission.

Then, the NIHSSmismatch% was calculated, showing a median (IQR) of 72.528. Those cases with negative NIHSS (n = 7) were considered as having a value of 0.

Internal validation

The validation cohort comprised 185 patients (Fig. 1), being 56,8% males. The mean age (SD) was 71.4 (13.1) years (Table 2).

On admission, the median (IQR) of NIHSSexpected was lower than NIHSSactual (4 [4] vs. 17 [7], P < 0.001). There was a weak and negative correlation between the NIHSSactual and the NCCT-ASPECTS on admission (Figure S1). However, the correlation between the NIHSSexpected and the NCCT-ASPECTS on admission was as strong as that one between NIHSS and NCCT-ASPECTS at 24 h (Figure S1).

The NIHSSmismatch% showed a median (IQR) of 77.4 (22.3). Those cases with negative NIHSSmismatch% (n = 1) were considered as having a value of 0.

Table 2 shows the basal data of the validation cohort according with ENI. Sixty-four experienced ENI (34.6%). There was a higher prevalence of coronary heart disease in patients without ENI (21.5% vs. 9.4%, p = 0.038) as well as a higher median NIHSS on admission (18 vs. 16, p = 0.022). The median (IQR) of NIHSSmismatch% was a greater in the ENI group (83.1 [17.6] vs. 72 [21.3], P < 0.001). Additionally, the median (IQR) of NCCT-ASPECTS on admission was higher in the ENI group (9 [2] vs. 8 [3], P < 0.001), whereas the median (IQR) onset-to-groin puncture time was shorter in the ENI (210 [168] min vs. 298.5 [311] min, P = 0.004).

Figure 2 shows the boxplots of the NIHSSmismatch% distribution according to ENI in the derivation (n = 123) and the validation cohort, both in the whole cohort (n = 185) and in the cohort in which thrombectomy was successful (n = 164). Patients with ENI showed significantly greater NIHSSmismatch% scores than patients without it, with minimal overlap in their distribution (P < 0.001).

In the derivation cohort, the ROC analysis for NIHSS mismatch% yielded an AUC of 0.754 with a 95% confidence interval of 0.659 to 0.838. In the validation cohort, the AUC was 0.689 (95% CI: 0.610 to 0.767) (Figure S2). Although there was a slight difference in point estimates, the overlap in confidence intervals and a non-significant z-test (z = 1.07, p = 0.28) indicate no meaningful difference in discriminative performance between the two cohorts.

The ROC curve for NIHSS mismatch% in the subgroup of patients with successful recanalization in the validation cohort (n = 164) is shown in Fig. 3. The analysis identified 75% as the optimal NIHSS mismatch% cutoff for predicting ENI, yielding a sensitivity of 80% and specificity of 58%.

The multivariate analyses (Table 3) showed that NIHSSmismatch% and NIHSSmismatch% >75 were predictors of ENI (OR 1.062; 95% CI 1.033–1.092 and OR 5.687; 95% CI 2.562–12.623, respectively), adjusted for potential confounders. Both models showed a good discriminatory ability assessed by the c-statistic (AUC = 0.777; 95% CI 0.705–0.848 and AUC = 0.791; 95% CI 0.722–0.859, respectively) (Fig. 3) as well as a good calibration (Cox and Snell R2 = 0.212 and 0.231, respectively) (Table 3). Furthermore, the discriminatory ability of NIHSSmismatch% and models was superior to that of the ASPECTS (Fig. 3).

The subgroups analyses showed that the prognostic value of NIHSSmismatch% remained constant in most cases (Figure S3).

Table 2 Basal data of the validation cohort (n = 185) according with early neurological improvement (ENI).
Full size table
Fig. 2
figure 2

NIHSSmismatch% distribution according to early neurological improvement in the derivation and the validation cohort. P < 0.001 for comparisons between early neurological improvement (ENI) and non-ENI patients in all groups.

Full size image
Fig. 3
figure 3

ROC curve for early neurological improvement in patient with successful recanalization (validation cohort, n = 164). ROC curves for early neurological improvement based on NIHSSmismatch% (blue curve) as well as on the logistic regression Model 1 (green curve) which include NIHSSmismatch%, the Model 2 (mauve curve) which include NIHSSmismatch% >75 and on the ASPECTS on admission (red curve). Models 1 and 2 are detailed in Table 3.

Full size image
Table 3 Multivariate analysis of factors associated with early major neurological improvement in patient with successful recanalization (validation cohort, n = 164).
Full size table

Representative examples of patients with similar baseline NIHSS scores and M1 occlusions but differing NIHSS mismatch% are shown in Fig. 4. Patient 1, with a mismatch of 88%, experienced early neurological improvement at 24 h, while Patient 2, with a mismatch of 33%, did not.

Fig. 4
figure 4

Representative cases of high and low NIHSSmismatch% illustrating its prognostic value for early neurological improvement. Comparative clinical and imaging data from two patients with left M1 middle cerebral artery occlusion and similar baseline NIHSS scores15. Patient 1 had a higher NCCT-ASPECTS score on admission (ASPECTS 10), resulting in a lower predicted NIHSS (NIHSSexpected = 2) and a higher NIHSSmismatch% (94%), and experienced early neurological improvement (NIHSS at 24 h = 1). Patient 2 had a lower ASPECTS score (ASPECTS 6, arrow and arrowhead), a higher predicted NIHSS (NIHSSexpected = 10), and lower NIHSSmismatch% (33%), and did not show early neurological improvement (NIHSS at 24 h = 18). This figure illustrates how NIHSSmismatch% reflects the discordance between clinical severity and infarct extent, and its potential value as a marker of salvageable tissue. NIHSS, National Institutes of Health Stroke Scale; NIHSSactual, NIHSS on admission; NCCT; Non-contrast computed tomography; ASPECTS, Alberta stroke program early CT score; CTA-SI, Computer Tomography Angiography source images. *NIHSSexpected = 10.8 + 0.05xAge + 2.5xDM + 2.6xHemisphere – 1.5x ASPECTSadmission. **NIHSSmismatch% = (NIHSSactual – NIHSSexpectedl)/NIHSSactuall x 100.

Full size image

Discussion

To the best of our knowledge, this is the first study that analyses the discordance or mismatch between the “real” NIHSS and the “expected” NIHSS (NIHSSmismatch%), as well as its influence on the short-term prognosis in patients with LVO in the anterior circulation reaching successful reperfusion. The results show that patients with an NIHSSmismatch% >75 had five times more the probability of experiencing ENI compared to those with a lower mismatch, adjusted for potential confounders.

In this study, it was observed that the factors related to the NIHSS score on admission included the ASPECTS score, the affected hemisphere, the thrombus location, age, DM and acute neurological complications (hemorrhage/edema). Previous studies have shown that these factors affect stroke severity, as measured by the NIHSS25,26,27,28. For example, several studies have demonstrated an inverse relationship between the NIHSS, and the ASPECTS score at stroke onset25,26,27,28, and it has been estimated that each 10-point increase in the initial NIHSS is associated with a 3-point decrease in the ASPECTS score16 as previously said, while an ASPECTS score of 6–10 points is associated with a higher likelihood of a favorable outcome29. Additionally, the NIHSS depends on the affected hemisphere, being greater if left one30, as well as the thrombus location, the age and the vascular risk factors28.

In recent decades, several studies have analyzed clinical-radiological mismatch in patients undergoing reperfusion treatments, especially those receiving intravenous thrombolysis (IVT) with alteplase, with conflicting results14,16. Kent et al. concluded that clinical-radiological mismatch between the NIHSS and the ASPECTS score did not reliably identify patients with higher or lower likelihoods of benefiting from IVT16. In contrast, Deng et al. showed that patients with a positive NIHSS-ASPECTS mismatch, defined as NIHSS ≥ 8 and ASPECTS ≥ 8, had a higher probability of favorable evolution, with better risk profiles for intracranial hemorrhage and mortality, after IVT14. A “clinical-radiological mismatch” means that the patient’s clinical symptoms may be milder or less pronounced than expected based on visible changes in brain imaging. This may be due to the presence of “ischemic penumbra”, an area of brain tissue where cerebral blood flow is insufficient, but cell death has not yet occurred31. The DAWN clinical trial was the first to demonstrate that patients eligible for MT could be selected based on a large penumbra area, considering the clinical-radiological mismatch from a high NIHSS score and the presence of a small infarct on advanced neuroimaging (DW-MRI or CTP)32.

On the other hand, ENI following MT for AIS is a strong predictor of favorable long-term outcomes. ENI, typically defined as a significant reduction in the NIHSS score within 24 h post-treatment, is associated with higher rates of good functional outcomes at 90 days, lower mortality, and reduced symptomatic intracranial hemorrhage33. Previous studies have shown that factors predicting ENI include younger age, lower baseline NIHSS score, absence of early CT hypodensity, arterial patency, and presence of collateral blood supply13,34. Additionally, shorter time from onset to admission and from groin-puncture to reperfusion, as well as higher baseline CT perfusion ASPECTS, are associated with ENI13. It has recently been reported that ischemic penumbra, estimated by perfusion CT, is present in at least 80% of patients with large vessel occlusion undergoing TM and that reperfusion is associated with higher rates of good functional outcomes at 3 months in those patients with more favorable mismatch patterns5. However, the effect of ischemic penumbra on ENI in these patients is unknown.

In the present study, the mismatch between the “real” and “expected” NIHSS was estimated using a model that included the ASPECTS and other clinical and neuroimaging data. The presence of a mismatch > 75% between the “real” and “expected” NIHSS on admission was associated with a higher probability of experiencing an ENI after MT. This simple method for estimating the penumbra area, using only clinical and basic neuroimaging data, has not been described in the scientific literature to date, opening a novel line of research that will need to be developed in multicentric studies.

This study has several limitations. First, it was conducted at a single center and is based on retrospective data, which may introduce selection bias and limits generalizability. Although the sample size is relatively large for a pilot study and exceeded the a priori estimation based on the expected effect size, the findings should be confirmed in larger, multicenter prospective cohorts. Second, our approach derives NIHSSexpected at admission from a multivariable model of NIHSS at 24 h; this operational choice assumes that determinants of NIHSS at 24 h approximate those at baseline. We recognize that acute ischemic stroke evolves non-linearly and that treatments administered between admission and 24 h (including MT) may shift NIHSS independently of baseline infarct burden, potentially biasing estimates of NIHSSexpected. Third, while site of occlusion and collateral grade were included among the candidate predictors and tested in the modelling process, they were not retained in the final parsimonious model after collinearity checks and stepwise selection. Moreover, other relevant covariates—particularly perfusion-based imaging parameters —were not consistently available in our dataset and therefore could not be incorporated. Finally, NIHSS provides a quantitative clinical measure of deficit, whereas ASPECTS is a semi-quantitative measure of infarct extent that does not encode eloquence or lateralization; this conceptual difference may limit the precision of NIHSSmismatch% and motivates further refinement of the model. Despite these limitations, the study provides proof-of-concept evidence that NIHSSmismatch% is a promising predictor of ENI after MT.

Conclusions

The “expected” NIHSS score at the emergency admission of a stroke patient can be calculated using a model based on clinical and conventional neuroimaging variables. The discordance between the “expected” and the “real” NIHSS, or NIHSSmismatch%, can be quantified. The ROC curve identifies a NIHSSmismatch% of 75% as the optimal cutoff for predicting ENI (80% sensitivity and 58% specificity). NIHSSmismatch% is a factor associated with ENI after MT, defined as a reduction in the NIHSS of ≥ 8 points or achieving a NIHSS score of 0/1 at 24 h. Moreover, the NIHSSmismatch%, could be an estimator of ischemic penumbra in patients with AIS, although this needs to be confirmed in specific studies.

In conclusion, the NIHSSmismatch% is a predictor of ENI after successful MT. Its value needs to be tested in large multi-center studies.