SCREENING STRATEGIES FOR DIABETIC RETINOPATHY

糖尿病视网膜病变仍然是20-60岁年龄组新发失明的主要原因，尽管许多治疗方法已被证明有效。 建议眼科医生每年对糖尿病患者进行一次监测，以尽早发现视网膜病变，防止严重的视力损害。 患者对该指南的依从性通常较差。 另一种方法是在初级保健医生的办公室与常规访问一起筛选患者。 然而，目前还没有有效的筛选程序。本研究的目的是开发一系列筛选方案或方案，通过眼底检查检测因存在或极有可能发展为威胁视力的视网膜病变[为此目的定义为增殖性糖尿病视网膜病变（PDR）或黄斑水肿（ME）]而需要眼科转诊的糖尿病患者。 筛选选项将根据早期治疗糖尿病视网膜病变研究（ETDRS）的眼底摄影数据集分析制定。其目的是确定视网膜病变的后眼底的有限部分的检查是否可以提供足够的信息来预测由七个标准摄影视野定义的眼底区域中的视力威胁视网膜病变。 具体目的是确定哪些糖尿病病变，严重程度如何，以及在哪个后摄影视野中是PDR或ME的最佳预测因子;统计回归技术将用于识别最显著相关的变量。 将确定包括最佳受试者操作特征（ROC曲线）的筛选标准，即具有最佳灵敏度（真阳性率）和特异性（真阴性率）的标准。

Diabetic retinopathy remains the leading cause of new blindness in the 20-60 year old age group, in spite of the proven efficacy of a number of therapies. Annual surveillance of patients with diabetes by an ophthalmologist is recommended to detect retinopathy sufficiently early to prevent serious visual impairment. Patient compliance with this guideline has generally been poor. An alternative approach is to screen patients in the offices of primary care physicians in conjunction with routine visits. However, no screening procedures of proven efficacy are yet available. The goal of this research is to develop a series of screening options or protocols to detect by fundus examination patients with diabetes requiring ophthalmological referral because of the presence of, or high likelihood of developing, vision threatening retinopathy [defined for this purpose as proliferative diabetic retinopathy (PDR) or macular edema (ME)]. The screening options will be developed from an analysis of the fundus photographic data set from the Early Treatment Diabetic Retinopathy Study (ETDRS). The objective is to determine whether the examination of a limited portion of the posterior fundus for retinopathy lesions can provide sufficient information to predict vision threatening retinopathy in the fundus region defined by the seven standard photographic fields. The specific aim is to determine which diabetic lesions, of what severity level, and in which of the posterior photographic fields are the best predictors of PDR or ME; statistical regression techniques will be used to identify the most significantly correlated variables. The screening criteria comprising the optimal Receiver Operator Characteristics (ROC curves), i.e. the criteria with the best sensitivity (true positive rate) and specificity (true negative rate) will be identified.