家族性渗出性玻璃体视网膜病变合并孔源性视网膜脱离风险模型的建立与验证

Establishment and validation of a risk model for familial exudative vitreoretinopathy combined with rhegmatogenous retinal detachment

摘要:

目的 探讨家族性渗出性玻璃体视网膜病变（FEVR）合并孔源性视网膜脱离（RRD）的风险因素，建立预测模型并验证。方法 选取2021年1月—2023年1月济南市妇幼保健院收治的103例FEVR患者为研究对象，根据是否合并RRD分为RRD组（41例）与非RRD组（62例），收集患者的临床资料。应用LASSO-Logistic回归模型筛选FEVR合并RRD的影响因素，构建临床预测模型。采用赤池信息准则（AIC）和贝叶斯信息准则（BIC）评价模型拟合效果，将传统Logistic与LASSO-Logistic两组回归模型对比，通过校准曲线进行验证。结果 RRD组中男性青少年更多，且合并眼部外伤史、高度近视、角膜、视网膜及颞侧玻璃体增生、嵴样改变和基因突变的概率较非RRD组更高（P <0.05）。RRD组红细胞计数、血小板计数、等效球镜度及整体血流密度均低于非RRD组，眼压高于非RRD组（P <0.05）。LASSO-Logistic回归分析结果显示，男性［O^R=5.257（95% CI：2.012，15.828）］、青少年［O^R=0.887（95% CI：0.848，0.923）］、眼部外伤史［O^R=4.750（95% CI：1.762，13.781）］、高度近视［O^R=3.867（95% CI：1.330，11.678）］、眼压高［O^R=1.171（95% CI：1.065，1.298）］、整体血流密度低［O^R=0.684（95% CI：0.579，0.790）］、视网膜增生［O^R=2.977（95% CI：1.214，7.750）］、颞侧玻璃体增生［O^R=3.379（95% CI：1.522，7.861）］、嵴样改变［O^R=6.715（95% CI：2.824，17.789）］、FZD4突变［O^R=5.430（95% CI：1.814，18.183）］、LRP5突变［O^R=5.057（95% CI：1.749，15.756）］的FEVR患者合并RRD的风险更高（P <0.05）。LASSO-Logistic回归模型的AIC及BIC分别为212.380和263.308，精确度高于传统Logistic回归模型。结论 基于LASSO法筛选出的变量组成的LASSO-Logistic回归模型拟合效果较好，在评估FEVR患者合并RRD方面有较好的拟合精度和预测效果。

Abstract:

Objective To investigate the risk factors for familial exudative vitreoretinopathy (FEVR) combined with rhegmatogenous retinal detachment (RRD) and to establish a predictive model.Methods A total of 103 FEVR patients admitted to Jinan Maternal and Child Health Hospital from January 2021 to January 2023 were selected as the study subjects. They were divided into an RRD group (n = 41) and a non-RRD group (n = 62) based on the presence of RRD, and their clinical data were collected. The LASSO-Logistic regression model was applied to select the influencing factors for FEVR combined with RRD and to construct a clinical prediction model. The model's fit was evaluated using the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), comparing traditional Logistic and LASSO-Logistic regression models, and validated through calibration curves.Results In the RRD group, male adolescents were more prevalent, and there was a higher likelihood of coexisting with a history of eye trauma, high myopia, corneal, retinal and temporal side vitreous proliferation, ridgelike changes, and genetic mutations. Their red blood cell count, platelet count, equivalent sphere degree, and overall blood flow density were lower than those in the non-RRD group, while intraocular pressure was higher than in the non-RRD group, with these differences being statistically significant (P < 0.05).LASSO-Logistic regression analysis showed that the risk of FEVR patients with RRD was higher in males [O^R = 5.257 (95% CI: 2.012, 15.828)], adolescents [O^R = 0.887 (95% CI: 0.848, 0.923) ], history of eye trauma [O^R = 4.750 (95% CI: 1.762, 13.781)], high myopia [O^R = 3.867 (95% CI: 1.330, 11.678) ], high intraocular pressure [O^R = 1.171 (95% CI: 1.065, 1.298)], low overall blood flow density [O^R = 0.684 (95% CI: 0.579, 0.790) ], retinal proliferation [O^R =2.977 (95% CI: 1.214, 7.750) ], temporal side vitreous proliferation [O^R = 3.379 (95% CI: 1.522, 7.861) ], ridgelike changes [O^R = 6.715 (95% CI: 2.824, 17.789) ], FZD4 mutation [O^R = 5.430 (95% CI: 1.814, 18.183) ], and LRP5 mutation [O^R = 5.057 (95% CI: 1.749, 15.756) ]. The LASSO-Logistic regression model had an AIC and BIC of 212.380 and 263.308, respectively, showing better precision than the traditional Logistic regression model.Conclusion The LASSO-Logistic regression model, composed of variables selected by the LASSO method, showed a good fit and predictive accuracy in assessing the risk of FEVR patients combined with RRD.