Abstract Body: Objective: Mechanical thrombectomy (MT) for large vessel occlusion (LVO) has become a standard treatment, but outcomes vary depending on patient characteristics and treatment approaches. This study aims to develop a scoring model to quantify the likelihood of good outcomes using data from the K-NET Registry, a prospective study on MT. Additionally, we analyzed the impact of age on these scores and the factors contributing to good outcomes.
Methods: Among 3,187 patients in the K-NET Registry who underwent MT, 2,381 met the following criteria: 1) pre-mRS 0-2, and 2) occlusion of the internal carotid artery or middle cerebral artery (M1, M2). Patients were randomly divided into 80% training data (n=1921) and 20% test data (n=460). Predictive factors for good outcomes (mRS 0-2 at 90 days) were selected from the training data using a stepwise method, and these factors were utilized as predictor variables. Logistic regression analysis was performed using 10-fold cross-validation, and the K-value minimizing log-loss was applied to the test data. Exponential weights were determined using the model's odds ratio, and discriminative performance was assessed via ROC analysis to estimate the area under the curve (AUC).
Results: In 2,381 cases, the median values were: age 77 years, NIHSS 18, ASPECTS 8, and 142 minutes (min) from onset to puncture. Predictive factors and their coefficients were as follows: Age (<60: 8, 60-79: 3, ≥80: 0), recanalization (TICI 3: 8, 2b: 5, 0-2a: 0), NIHSS (≤10: 4, 11-20: 2, ≥21: 0), ASPECTS (≥9: 4, 6-8: 2, ≤5: 0), number of passes (1 pass: 2, 2 passes: 1, ≥3 passes: 0), time from onset to puncture (≤140 min: 1, ≥141 min: 0), diabetes (no: 1, yes: 0), intravenous tPA (yes: 1, no: 0). The model, validated with the test data, achieved an AUC of 0.768. Scores ranged from 0 to 29, with higher scores indicating better outcomes. A score of 17 or more was necessary to achieve a likelihood of good outcomes exceeding 50%. The age-specific thresholds were 9 for those under 60, 14 for those aged 60-79, and 17 for those aged 80 or older. Among elderly patients, a low NIHSS (≤10), high ASPECTS (≥9), and complete recanalization were crucial for achieving good outcomes.
Conclusion: A predictive model for MT in LVO patients was developed, demonstrating good discriminative ability. The cutoff points for predicting good outcomes varied by age, with lower NIHSS, higher ASPECTS, and complete recanalization being particularly important for elderly patients.