Avaliação da concordância interobservador na classificação ASPECTS no acidente vascular cerebral isquêmico agudo

Abstract

Introduction: The Alberta Stroke Program Early CT Score (ASPECTS) is a semiquantitative scoring system developed to estimate the extent of ischemic changes in the middle cerebral artery territory. A score of ≥ 6 is used by international guidelines as one of the inclusion criteria for patients with ischemic stroke patients undergoing endovascular therapy. However, this tool is subject to interobserver variability, which can impact therapeutic decisions. Objective: To evaluate the interobserver reliability of ASPECTS when applied by a neurologist and a neuroradiologist. Methods: A cross sectional interobserver reproducibility study was conducted at HUL. 148 non-contrast head CT scans were included, obtained within 24 hours of symptom onset in patients with suspected stroke. The ASPECTS for each scan was independently assigned by a neurologist and a neuroradiologist. The analysis included three approaches: total ASPECTS score, individual assessment of the ten regions, and dichotomous classification (≥6 or <6). Intraclass correlation coefficient (ICC), Cohen's kappa (κ), Gwet's AC1 (AC1), and Bland-Altman analysis were used. Results: The reliability for the total score was good (ICC = 0.78; 95% CI: 0.72 to 0.86), with no evidence of fixed bias between the evaluators. The linear regression between the score differences and means revealed a β coefficient of -0.17 (95% CI: -0.28 to -0.06). For the dichotomized ASPECTS, the results were κ = 0.56 (95% CI: 0.39 to 0.72) and AC1 = 0.79 (95% CI: 0.70 to 0.88). The caudate nucleus (AC1 = 0.89) and M5 region (AC1 = 0.84) showed the highest reliability, while the lentiform nucleus had the lowest (AC1 = 0.46 and κ = 0.33). Conclusion: The interobserver reliability of the total ASPECTS score is good. When used dichotomously, reliability ranged from moderate to substantial. Furthermore, the binary categorization of the score may penalize small differences and impact the selection of candidates for thrombectomy. Interobserver variability is not homogeneous across the score's regions.