列线图构建原发性闭角型青光眼患者术后发生恶性青光眼的模型及验证

Construction and validation of a nomogram model for postoperative malignant glaucoma in primary angle-closure glaucoma

摘要:

目的 探讨原发性闭角型青光眼患者术后恶性青光眼的发生率及其危险因素，建立列线图模型并进行验证。方法 选取2015年6月—2023年6月在西安交通大学第一附属医院眼科接受治疗的3 389例原发性闭角型青光眼患者2 100眼作为研究对象。术后6个月内出现恶性青光眼的97例（42眼）作为发生组，未出现恶性青光眼的3 292例（2 058眼）作为未发生组。收集两组患者的性别、年龄、合并糖尿病、高血压等一般资料；比较两组患者的术前晶状体厚度、前房深度、相对晶状体位置（LOWE系数）、高眼压情况、房角结构、眼轴长度、青光眼类型等；采用多因素一般Logistic回归模型分析术后发生恶性青光眼的危险因素；绘制列线图模型，并采用受试者工作特征（ROC）曲线评估模型的区分度；使用Bootstrap方法进行1 000次重复采样，以验证列线图模型的预测效能。结果 发生组年龄< 50岁、术前持续高眼压、房角结构完全关闭占比均高于未发生组（P <0.05），晶状体厚度低于未发生组（P <0.05），眼轴长度短于未发生组（P <0.05），前房深度浅于未发生组（P <0.05）。多因素一般Logistic回归分析结果显示，术前眼压高［O^R =24.179（95% CI：1.724，339.108）］、房角结构完全关闭［O^R =59.427（95% CI：4.173，846.224）］、急性青光眼：［O^R =15.507（95% CI：1.063，226.222）］是术后发生恶性青光眼的危险因素（P <0.05）；年龄大［O^R =0.022（95% CI：0.001，0.692）］、晶状体厚［O^R =0.037（95% CI：0.003，0.493）］、眼轴长度长［O^R =0.506（95% CI：0.264，0.971）］和前房深度深［O^R =0.0004（95% CI：0.000，0.963）］是术后发生恶性青光眼的保护因素（P <0.05）。模型评估显示，ROC曲线下的面积为0.979，约登指数为0.871，敏感性为89.3%（95% CI：0.790，0.960），特异性为97.8%（95% CI：0.850，0.980），显示出良好的区分能力。校准曲线进一步证实了模型的高精确度，Bootstrap方法显示均方误差为0.032。结论 本研究建立的原发性闭角型青光眼术后恶性青光眼是否发生的预测模型是可行的，在手术过程中应给予有这些危险因素的患者特别关注，以降低术后恶性青光眼的发生风险。

Abstract:

Objective To investigate the incidence and risk factors of malignant glaucoma in patients with primary angle-closure glaucoma after surgery, to establish and validate a nomogram model.Methods Select 3,389 patients (2,100 eyes) with primary angle closure glaucoma who received treatment in the ophthalmology department of a certain hospital from June 2015 to June 2023 as the research subjects. 42 eyes with malignant glaucoma within 6 months after surgery were selected as the occurrence group, and 2,058 eyes without malignant glaucoma were selected as the non occurrence group. Data on gender, age, comorbidity such as diabetes and hypertension, and preoperative factors like lens thickness, anterior chamber depth, LOWE coefficient, high intraocular pressure, chamber angle structure, axial length, and type of glaucoma were collected and compared. Univariate and multivariate logistic regression analyses were conducted to explore the risk factors for malignant glaucoma. A nomogram model was constructed using R software version 4.3.0 and its discrimination ability was assessed using the Receiver Operating Characteristic (ROC) curve. The model's predictive performance was validated with 1000 bootstrap resamples.Results The proportion of patients in the incidence group who were under 50 years old, had sustained preoperative high intraocular pressure, complete closure of the angle structure, thin lens thickness, short axial length, and shallow anterior chamber depth was higher than that in the non incidence group (P < 0.05). A multivariate general logistic regression analysis showed that, high preoperative intraocular pressure [O^R = 24.179 (95% CI: 1.724, 339.108) ], complete closure of angle structure [O^R = 59.427 (95% CI: 4.173, 846.224)], and acute glaucoma [O^R = 15.507 (95% CI: 1.063, 226.222) ] were the risk factors for postoperative malignant glaucoma (P < 0.05); Age [O^R = 0.022 (95% CI: 0.001, 0.692) ], lens thickness [O^R = 0.037 (95% CI: 0.003, 0.493)], long axial length [O^R = 0.506 (95% CI: 0.264, 0.971) ], and deep anterior chamber depth [O^R = 0.0004 (95% CI: 0.000, 0.963) ] are protective factors for postoperative malignant glaucoma (P < 0.05). Model evaluation shows that the area under the ROC curve is 0.979, the Jordan index is 0.871, the sensitivity is 89.3% (95% CI: 0.790, 0.960), and the specificity is 97.8% (95% CI: 0.850, 0.980), demonstrating good discrimination ability. The calibration curve further confirms the high accuracy of the model, and Bootstrap self validation shows a mean square error of 0.032 through 1000 resampling.Conclusion The predictive model for postoperative malignant glaucoma in primary angle-closure glaucoma developed in this study is feasible. Patients with these risk factors should be given special attention during surgery to reduce the risk of developing malignant glaucoma.