TAB-TICI Score: Successful Recanalization Score After Endovascular Thrombectomy in Acute Stroke

Background and Purpose: Successful reperfusion therapy is supposed to be comprehensive and validated beyond the grade of recanalization. This study aimed to develop a novel scoring system for defining the successful recanalization after endovascular thrombectomy. Methods: We analyzed the data of consecutive acute stroke patients who were eligible to undergo reperfusion therapy within 24 h of onset and who underwent mechanical thrombectomy using a nationwide multicenter stroke registry. A new score was produced using the predictors which were directly linked to the procedure to evaluate the performance of the thrombectomy procedure. Results: In total, 446 patients in the training population and 222 patients in the validation population were analyzed. From the potential components of the score, four items were selected: Emergency Room-to-puncture time (T), adjuvant devices used (A), procedural intracranial bleeding (B), and post-thrombectomy reperfusion status [Thrombolysis in Cerebral Infarction (TICI)]. Using these items, the TAB-TICI score was developed, which showed good performance in terms of discriminating early neurological aggravation [AUC 0.73, 95% confidence interval (CI) 0.67–0.78, P < 0.01] and favorable outcomes (AUC 0.69, 95% CI 0.64–0.75, P < 0.01) in the training population. The stability of the TAB-TICI score was confirmed by external validation and sensitivity analyses. The TAB-TICI score and its derived grade of successful recanalization were significantly associated with the volume of thrombectomy cases at each site and in each admission year. Conclusion: The TAB-TICI score is a valid and easy-to-use tool to more comprehensively define successful recanalization after endovascular thrombectomy in acute stroke patients with large vessel occlusion.


INTRODUCTION
Endovascular thrombectomy (EVT) with substantial recanalization is a crucial determinant of the functional independence of acute stroke patients with large vessel occlusion (LVO) (1). The likelihood of achieving functional independence increases with improved recanalization in patients treated with intravenous thrombolysis (2) or EVT (3,4). Therefore, to obtain the maximum benefit, current guidelines recommend achieving modified Thrombolysis in Cerebral Infarction (mTICI) 2b or 3 grade angiographic results as a technical goal of EVT (1). This substantial reperfusion is often regarded as a successful recanalization therapy. However, the status of reperfusion alone is insufficient as a criterion for a successful procedure because approximately three-quarters of the patients undergo EVT to obtain substantial reperfusion (mTICI 2b or 3) (5) and a great proportion of patients with substantial reperfusion show poor functional outcomes (6). Moreover, performance of procedures in patients with failed reperfusion should be further stratified, suggesting the need for a comprehensive evaluation tool for successful thrombectomy. However, except for angiographic reperfusion, there is no validated tool to evaluate the successful recanalization.
We, therefore, developed a novel scoring system to assess successful recanalization and tested its performance using an external dataset. Additionally, we validated the successful recanalization scoring system with different parameters in a nationwide multicenter cohort.

Study Design and Data Source
We analyzed data from the SElection CRiteria in Endovascular Thrombectomy and Thrombolytic Therapy registry (Clinicaltrials.gov NCT02964052). A total of 17 sites across South Korea were included in this project. Eligibility criteria were as follows: (1) age ≥18 years; (2) initial National Institutes of Health Stroke Scale (NIHSS) score ≥4; (3) direct admission to each participating hospital without hospital-to-hospital transfer; (4) onset-to-puncture time ≤600 min; (5) modified Rankin Score (mRS) before the qualifying stroke of 0/1; (6) recanalization success assessable on catheter angiogram; (7) documented catheter-accessible cerebral artery occlusion (intracranial internal carotid artery, middle cerebral artery M1 or proximal M2 occlusion, anterior cerebral artery, posterior cerebral artery, basilar artery or vertebral artery) on computed tomography, magnetic resonance imaging (MRI), or catheter angiography; and (8) available functional outcomes (mRS = 0-6) at 3 months ± 2 weeks. Because the primary purpose of this study was to build a model for measuring the performance of thrombectomy, patients who did not undergo thrombectomy were excluded (7). This project constituted two parts: a retrospective part that analyzed pre-existing prospective registries from each site from 2012 to 2015 and a prospective registry from November 2016 to December 2017.
The protocol of the entire project was approved by the ethical review board of each participating site. For patients enrolled retrospectively, informed consent was waived by the institutional review board because of the retrospective design. All patients who participated in the prospective part of the project provided informed consent.

Data Collection and Assessment
Each participating site entered the data for demographic characteristics, vascular risk factors, neurological status, time metrics for reperfusion therapy, procedural details, and functional outcomes using a web-based case report form. All imaging data were sent to the central laboratory in an anonymized DICOM format. Reperfusion grade was independently assessed by two neuroradiologists according to the mTICI scale, and decisions were made by consensus in case of discrepancies. Early neurological aggravation was defined as any increase in the NIHSS score at 24 h from baseline. A favorable outcome was defined as an mRS of 0-2 at 90 days.

Predictive Model Construction, Validation, and Statistical Analysis
The primary purpose of this study was to produce a scale for evaluating successful recanalization. For this purpose, we established three principles: (1) the items in the score should be confined to those directly associated with thrombectomy (several well-known outcome predictors were omitted if the items could not be manipulated by the interventional operators or acute stroke management team); (2) the scoring system should be valid for evaluating functional neurological outcomes and risk of procedure-related neurological worsening; and (3) the scale should be related to the metrics with a known association with successful recanalization, such as the volume of thrombectomy at each site (7) or on each admission date.
We divided the dataset into training (retrospective) and validation (prospective) datasets and used only the training dataset for producing the score. All available variables in the database were screened and selected when the items were potentially associated with favorable outcomes or early neurological aggravation. Especially, the adjuvant device use was defined when the second thrombectomy device system replaced the first thrombectomy device system. To construct a predictive model for outcomes using the selected items, we performed univariate and multivariate logistic regression analyses with each item and selected items that were significantly associated with either outcome. Item selection and score rating were performed according to the probability obtained with a logistic model incorporating receiver operating curve (ROC) analyses. On the basis of the results of these analyses, we developed a new score.
The stability of the developed score was internally validated with k-fold cross-validation using the area under the curve (AUC). The model was tested using prospectively collected validation datasets and presented with sensitivity, specificity, and AUC.
We divided the population into three groups according to performance: excellent, fair, and poor. In addition, we validated the clinical implications of the TAB-TICI score in terms of volume of cases at each site and in each admission year as well as changes in trends with each admission year. The independent associations among these clinical implications were proved using ordinal logit regression analysis with covariates.
Thrombectomy was performed using a retrievable stent in most cases (83.8%), and a contact aspiration device was used as the first or adjuvant device in 19%. During the procedure, 8.8% of patients experienced procedure-related intracranial bleeding, and 8.2% suffered from procedural adverse events other than intracranial bleeding.

Model Development
Using univariate analysis, we selected the major categories of each item for developing the new score ( Table 2). Final items were selected using multivariate analyses ( Table 3). Because the model was required to predict both early neurological aggravation and favorable outcomes at 90 days, all items from the models for early neurological aggravation and favorable outcome were selected. Finally, ER-to-puncture time (T), adjuvant devices used (A), procedural intracranial bleeding (B), and postthrombectomy reperfusion status (TICI) were selected as items for the final model. ROC analyses to find the best model using differently weighted items showed that the model with a 0-2 range for reperfusion status score showed the highest AUC for both early neurological  aggravation and favorable outcomes (Supplementary Table 1). Finally, we developed a single score ranging 0-5 and including four items (TAB-TICI score; Table 4).

Association of Performance Grade With Thrombectomy Case Volume at Each Participating Site and in Each Admission Year
In ordinal logistic regression analyses, the TAB-TICI score was negatively associated with the volume of thrombectomy cases at each site [adjusted odds ratio (aOR) 0.992, 95% CI 0.990-0.994, P < 0.01] and decreased over time (aOR 0.76, 95% CI 0.69-0.84, P < 0.01). Figure 2 demonstrates the association of performance grade (excellent, good, and poor) with admission year (Figure 2A) and the volume of thrombectomy cases at each site ( Figure 2B). The proportion of excellent performance was 23.9% in 2012, which increased gradually and became 47.9% in 2017 (P-value for Kendall tau <0.01).  The volume of thrombectomy cases at each site was also significantly associated with the performance grade, with a higher proportion of excellent performance observed in the group with a higher volume of thrombectomy cases (P-value for Kendall tau <0.01).

DISCUSSION
In this study, using a nationwide multicenter registry for acute stroke patients who underwent EVT, we developed a new scoring system to assess the successful recanalization of endovascular thrombectomy comprising four items: ER-to-puncture time, adjuvant devices used, procedural intracranial bleeding, and post-thrombectomy reperfusion status. Using these items, we developed the TAB-TICI score (with a higher score indicating a worse performance) and tested its internal and external validity. Decreased TAB-TICI score or shift from poor to excellent performance grade derived from the TAB-TICI score was associated with an increased volume of thrombectomy cases at each site and in later admission years. This association reflects the feasibility of using the TAB-TICI score as a metric for evaluating EVT performance. We selected four items to develop the new score on the basis of favorable functional outcomes and risk of early neurological aggravation. Reperfusion, not simply angiographic recanalization, was the most critical predictor of EVT outcome, and this finding is consistent with those of prior investigations (8,9). However, further stratification according to the risk of outcome events is required. In this regard, the TAB-TICI score exhibited potential usefulness in stratifying the performance of patients with or without substantial reperfusion in sensitivity analysis.
Workflow time metrics are well-known predictors of outcomes (10) and are the primary targets for quality assurance and control in acute stroke management at the same time. In this study, we selected the interval from emergency room (ER) arrival to groin puncture on the basis of analytical results. Previous studies support the importance of ER-to-puncture time (11,12). Saver and colleagues addressed the relative importance of ER-to-puncture time over onset-to-puncture time using analysis of the Highly Effective Reperfusion Using Multiple Endovascular Devices study (11). They suggested that the vague definition of symptom onset time and the possibility for eliminating the "onset-to-ER effect" through deselection of fast or very slow progressors in the clinical trials have resulted in the observed trend. The importance of the ER-to-puncture time effect was doubled by the fact that the rate of achievement of substantial reperfusion decreased with an increase in ER-to-groin puncture time, although this was not associated with the total onset-topuncture time (12). Additionally, because the collaboration of multiple domains and parties in a hospital is essential to effectively reduce the workflow time metrics in an acute stroke management setting, ER-to-puncture time as an item related to procedural performance is advantageous as it represents the capability of the acute stroke care team.
We considered another time metric-the puncture-toreperfusion time-as an item related to procedural performance. However, we discarded it because puncture-to-reperfusion time was not associated with early neurological aggravation and the significance of its association with favorable outcomes in univariate analysis was eliminated in multivariate analysis.
In terms of technical aspects of thrombectomy, we screened for the following items as a measure of performance: a number of passes for stent retrieval or aspiration (13), firstpass recanalization (14), adjuvant chemical thrombolysis (15), and adjuvant intracranial angioplasty or stenting in cases of intracranial atherosclerotic disease (16). Among these, only adjuvant thrombectomy device use was selected as an item in the developed scoring system via covariate selection using a multivariate logistic regression model. In contrast, the firstpass effect was denied a significant item although the firstpass effect was a reputed predictor for the favorable outcome after endovascular thrombectomy. In this study, the firstpass effect was not significantly associated with the early neurological aggravation nor a favorable outcome. In terms of procedural complications, non-intracranial bleeding was not significantly associated with neurological outcomes, whereas procedural intracranial bleeding reduced the probability of favorable outcomes.
Because accurate and standardized performance measurement is crucial in acute stroke management, several scoring systems have been developed to predict clinical outcomes in an acute stroke setting (Supplementary Table 2). The iSCORE has been designed to predict 30-day and 1-year mortalities of patients with acute stroke after hospitalization (17). Although the iSCORE performed well in predicting mortality reflected by high C-statistics, the tool was not appropriate in an acute stroke with reperfusion therapy setting. In contrast, Pittsburgh Outcomes After Stroke Thrombectomy (POST) score targets acute stroke patients within 8 h of being last seen well and undergoing EVT (18). The POST score comprises three items showing high predictability for functional independence at 90 days. However, the POST score has been designed to predict outcomes instead of the procedural performance, and the items included in this score (age, 24-72 h final infarct volume, and hemorrhagic transformation) are substantially unmodifiable outcome predictors. In particular, measurement of the final infarct volume is somewhat difficult because the infarct volume might not be consistent across different imaging modalities and analytical programs (10,19). Furthermore, the MT-DRAGON is designed to predict functional independence at 90 days using clinical and imaging parameters (20). Specifically, the MT-DRAGON score includes the DWI-ASPECT score instead of infarct volume. Moreover, the MT-DRAGON score has been developed to predict clinical outcomes following thrombectomy. However, the complexity of calculating the score has rendered its prompt use challenging.
Compared with the abovementioned scores, the TAB-TICI score has several distinctive features. The TAB-TICI score is targeted at measuring the successful recanalization rather than predicting the outcome of each patient. The determinants of neurological outcomes associated with thrombectomy can be categorized into unmodifiable factors such as demographics and stroke severity and modifiable factors such as in-hospital workflow time metrics and procedural complications. Because prior scores typically used unmodifiable predictors as items, these scores are inappropriate to evaluate performance (Supplementary Table 2). Contrary to these scores, the TAB-TICI score was developed by exclusively selecting modifiable predictors that are amenable toward improvement. Therefore, the TAB-TICI score seems to be suitable for quality assurance/control, longitudinal monitoring, comparisons among different organizations, and academic activities. Another remarkable peculiarity of the TAB-TICI score is that it is the most straightforward scoring system to remember and calculate, without requiring a calculating chart or an electronic calculator.
Notably, the TAB-TICI score and its derived procedural performance grade, which were used as surrogate markers for procedural performance, were significantly associated with the volume of thrombectomy cases at each site and in each admission year. Although second-generation EVT devices are easy to use, multiple studies have demonstrated the pattern of a learning curve in thrombectomy procedures (21,22) and the mode of the learning curve differed according to the technique (23). This practitioner-level learning curve effect can be presented as the center-level cumulative volume effect. Kim and colleagues analyzed the impact of accumulated volume of thrombectomy cases at multiple centers and reported that the accumulated volume of thrombectomy cases affected outcomes, safety, and workflow time (7). Decrease in the TAB-TICI score with the later admission years can be interpreted in the same context. Between 2012 and 2017, two critical changes regarding EVT occurred in Korea; in August 2014, retrievable stent reimbursement was approved for the use of Solitaire FR and TREVO in acute stroke patients, and in February 2015, the results of three successful randomized clinical trials were reported, which entailed the updating of clinical practice guidelines to recommend EVT for eligible acute stroke patients (24). These changes could improve the overall performance of EVT, reflected as an increase in the proportion of excellent performance grades.
This study has limitations. First, the study population was entirely Korean, and the validity of this score should be confirmed in different populations. Second, this study included both anterior and posterior circulation stroke patients. However, sensitivity analysis revealed comparable outcome predictions between anterior and posterior circulation stroke patients. Third, the clinical usefulness of the TAB-TICI score should be validated to be accepted as a useful score.
Despite these limitations, the present study provides a novel tool to measure the procedural performance of thrombectomy. Performance of the TAB-TICI score was stable an independent populations and their subgroups, and the excellent, good, and poor performance grades derived from the TAB-TICI score were significantly associated with the surrogate markers for the successful recanalization. Therefore, the TAB-TICIE score, being simple and easy to use, is expected to become a useful tool for measuring and improving thrombectomy performance in acute stroke patients with LVO.

DATA AVAILABILITY STATEMENT
Anonymized data will be provided upon request to the corresponding author.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Samsung Medical Center IRB SMC-2016-08-119. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

AUTHOR CONTRIBUTIONS
W-KS performed statistical analysis, provided the tables and figures, and wrote the first draft of the manuscript. HN supervised the project. All authors were involved in the design and interpretation of the data, edited the manuscript, and approved the final draft.