早期人工耳蜗植入效果预测长期言语识别
这项回顾性队列研究纳入625例语后聋成人人工耳蜗使用者,构建逻辑回归模型,利用术后1或3个月的辅音-核-辅音(CNC)词识别改善情况预测12个月时的显著改善。结果显示早期改善与长期 outcomes 强相关,3个月模型预测准确性最高(验证队列AUC=0.92)。
要点
- 术后1个月或3个月CNC词识别改善可强预测12个月改善。
- 3个月模型预测准确性最高,验证队列AUC=0.92。
- 早期评估有助于实时估计长期获益概率。
核心结论
- 包含术后1个月CNC改善的模型具有良好预测区分度,包含3个月改善的模型在训练和验证队列中均表现出优秀区分度。
研究设计与样本
- 回顾性队列研究,纳入2个三级人工耳蜗中心的625例语后聋成人双侧听力损失患者,数据按3:1分为训练队列(n=469)和验证队列(n=156)。
主要发现
- 625例中513例(82%)在术后12个月时CNC词识别得分较术前改善。
- 术后1个月改善者12个月改善的比值比为45.72,术后3个月改善者为22.22。
- 验证队列中,3个月模型AUC=0.92,1个月模型AUC=0.88,基线术前模型AUC=0.77。
关键数值与推荐
- 改善定义为得分超过术前评分95%置信区间上限。
- 模型均包含性别、种族、植入年龄、耳聋时长和术前CNC评分。
临床怎么用
- 临床可利用术后1或3个月的CNC改善情况,通过回归方程实时估计患者12个月CNC改善概率。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
OBJECTIVES: Early identification of cochlear implant (CI) users at risk for poor audiological outcomes may allow for timely intervention to optimize long-term CI benefit. This longitudinal study aims to develop and evaluate logistic regression models that can predict patients' 12-months speech recognition performance from their early speech recognition scores and other patient factors. DESIGN: This retrospective cohort study included 625 postlingually deafened, adult CI users with bilateral hearing loss from 2 tertiary CI centers. Logistic regressions were fit to model the likelihood of a clinically significant improvement in consonant-nucleus-consonant (CNC) word scores at 12-months postimplantation. All models included sex, race, age at implantation, duration of deafness, and pre-CI CNC score. The model performance benefit of adding CNC improvement status (improved versus not improved, with improvement defined as scoring beyond the upper limit of the 95% confidence interval of the preoperative score) at 1- or 3-months postactivation to the baseline models was evaluated by comparing the area under the receiver operating characteristic curve (AUC). Data were combined across institutions and then separated into training and validation cohorts at a 3:1 ratio. Models were developed using the training cohort (n = 469) and then applied to the validation cohort (n = 156). RESULTS: Of the 625 patients included, 513 patients (82%) demonstrated improvement in CI-only CNC word recognition scores at 12-months postimplantation compared with their preimplantation scores. Early improvement was strongly associated with long-term outcomes. Patients were more likely to improve by 12 months if they improved by 1 month (odds ratio = 45.72) or 3 months (odds ratio = 22.22) postimplantation. Model performances were similar across the training and validation cohorts. In the validation cohort, the model using 3-months data had the highest predictive accuracy (AUC = 0.92), followed by the 1-month model (AUC = 0.88), and the baseline preoperative model (AUC = 0.77). Based on the regression coefficients obtained from the training cohort, equations can successfully estimate the probability of CNC improvement at 12 months postimplantation for CI patients. CONCLUSIONS: Models that included CNC score improvements at 1 month demonstrated good predictive discrimination, while those incorporating 3-months improvements showed excellent discrimination in both the training and validation cohorts. This study highlights the importance of assessing early post-CI speech recognition improvement and proposes a regression model that clinicians can use in real-time to provide early estimates of their patients' probability of 12-months CNC improvement.
引用
- Chau IJ, Kunnath AJ, Gothard A, et al. Early Cochlear Implant Outcomes Predict Long-Term Speech Recognition. Ear Hear. 2025. PMID: 41403015. doi:10.1097/AUD.0000000000001774
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。