纵向杯盘比测量在识别青光眼进展中的无效性
该回顾性队列研究评估纵向杯盘比变化(ΔCDR)能否预测青光眼在OCT或标准自动视野计(SAP)上的进展,结果ΔCDR对OCT进展的AUC为0.55、对SAP进展的AUC为0.53,在90%特异性下灵敏度仅15%和16%,作者认为ΔCDR的识别能力类似随机选择,建议结合全面视神经乳头检查与辅助检查判断进展。
要点
- ΔCDR≥0.20在OCT进展者中占7.1%、在非进展者中占4.8%(P=.04);在SAP进展者中占7.7%、在非进展者中占5.7%(P=.23)。
- ΔCDR对OCT进展的AUC为0.55,在90%特异性下灵敏度为15%;对SAP进展的AUC为0.53,在90%特异性下灵敏度为16%。
- 作者建议临床医生结合全面的视神经乳头检查和辅助检查来评估青光眼进展。
核心结论
- 纵向CDR数据在识别青光眼结构和功能进展方面与随机选择的表现相似。
- 鉴于这种弱关联,作者建议临床医生使用全面的视神经乳头检查和辅助检查来评估青光眼进展。
研究设计与样本
- 回顾性队列研究,受试者来自Bascom Palmer青光眼数据库。
- 从电子健康记录中提取CDR,并匹配±6个月内完成的OCT和SAP检查;要求每只眼至少有5次检查且跨度至少2年。
- ΔCDR定义为最终CDR减去基线CDR,基线CDR和最终CDR分别取最接近基线检查和最终检查的CDR。
- 使用普通最小二乘法回归计算每只眼的OCT视网膜神经纤维层(RNFL)和SAP平均偏差(MD)变化率,将具有统计学显著负向眼特异性斜率(P<.05)的眼定义为进展眼。
- OCT分析纳入3313只眼,SAP分析纳入2174只眼。
主要发现
- OCT分析(N=3313只眼):ΔCDR≥0.20者在OCT进展眼中占7.1%,在非进展眼中占4.8%(P=.04);ΔCDR仅解释OCT RNFL斜率变异的2%。
- OCT分析中,ΔCDR预测OCT进展的AUC为0.55,在90%特异性下灵敏度为15%。
- SAP分析(N=2174只眼):ΔCDR≥0.20者在SAP进展眼中占7.7%,在非进展眼中占5.7%(P=.23);ΔCDR仅解释SAP MD斜率变异的1%。
- SAP分析中,ΔCDR预测SAP进展的AUC为0.53,在90%特异性下灵敏度为16%。
关键数值与推荐
- OCT进展眼的ΔCDR≥0.20比例为7.1%(非进展眼4.8%,P=.04)。
- SAP进展眼的ΔCDR≥0.20比例为7.7%(非进展眼5.7%,P=.23)。
- ΔCDR预测OCT进展的AUC为0.55,90%特异性下灵敏度为15%。
- ΔCDR预测SAP进展的AUC为0.53,90%特异性下灵敏度为16%。
临床怎么用
- 作者建议临床医生结合全面的视神经乳头检查和辅助检查来评估青光眼进展。
局限性
- 原文未报告局限性。
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
OBJECTIVE: To assess whether longitudinal change in cup-to-disc ratio (ΔCDR) is predictive of glaucoma progression on optical coherence tomography (OCT) or standard automated perimetry (SAP). DESIGN: Retrospective cohort study. SUBJECTS, PARTICIPANTS, AND/OR CONTROLS: Subjects in the Bascom Palmer Glaucoma Repository. METHODS, INTERVENTION, OR TESTING: We extracted CDRs from the electronic health record. We identified OCT and SAP tests performed within ±6 months of CDRs. Baseline and final CDRs were defined as CDRs closest to the baseline and final tests, respectively. Eyes were required to have ≥5 tests spanning ≥2 years. ΔCDR was calculated as CDR final - CDR baseline. Using ordinary least squares regression, we calculated eye-specific rates of change in OCT retinal nerve fiber layer (RNFL) and SAP mean deviation (MD). We classified eyes as progressors if they had statistically significant negative eye-specific slopes (P < .05). We then evaluated how well ΔCDR predicted progression on OCT or SAP. MAIN OUTCOME MEASURES: Area under receiver operating characteristic curve (AUC) predicting OCT or SAP progression using ΔCDR. RESULTS: In the OCT analysis (N = 3313 eyes), ΔCDR ≥0.20 was observed in only 7.1% of OCT progressors and 4.8% of non-progressors (P = .04). Only 2% of the variation in OCT RNFL slopes was explained by ΔCDR. AUC was 0.55. At 90% specificity, ΔCDR had a sensitivity of 15% in detecting OCT progressors. In the SAP analysis (N=2,174 eyes), ΔCDR ≥0.20 was observed in only 7.7% of SAP progressors and 5.7% of non-progressors (P = .23). Only 1% of the variation in SAP MD slopes was explained by ΔCDR. AUC was 0.53. At 90% specificity, ΔCDR had a sensitivity of 16% in detecting SAP progressors. CONCLUSIONS: Longitudinal CDR data performs similar to random selection in identifying glaucoma progression on both structural and functional testing. Given this weak association, clinicians should gauge glaucomatous progression using both a comprehensive optic nerve head examination and ancillary testing.
引用
- Kamdar A, Medeiros FA, Swaminathan SS, et al. Inefficacy of Longitudinal Cup-to-Disc Ratio Measurements in Identifying Glaucoma Progression. Am J Ophthalmol. 2025. PMID: 40653258. doi:10.1016/j.ajo.2025.07.005
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。