Abstract
Abstract: :
Purpose: To formulate and test a decision model to predict the development of local non–proliferative diabetic retinopathy (NPDR) based primarily on multifocal electroretinogram (mfERG) implicit times. Methods: 28 eyes of 28 diabetics (12 eyes with NPDR and 16 without retinopathy) were studied. During the initial exam, mfERGs were recorded from the central 45 degrees using standard conditions. Fundus photos were taken within 2 months of the recording and at 12–month follow–up. The location and extent of retinopathy were determined relative to the mfERG stimulus array. At each of the 103 stimulated retinal locations, an mfERG implicit time was derived using a template stretching method (Hood & Li, 1997), and a Z–score was calculated based on results obtained from 20 age–matched normals. Thirty–five non–overlapping retinal zones were constructed in each diabetic eye by grouping adjacent stimulated locations and each zone was assigned the maximum Z–score within it. Zones already containing retinopathy at initial testing (n=61) were excluded from further analysis. The probability that the remaining zones developed new retinopathy at follow–up was modeled based on mfERG implicit time Z–score, duration of diabetes, eye status (NPDR or no retinopathy), diabetes type (I or II), and blood glucose level during mfERG recording. Results: At one–year follow–up, new retinopathy developed in 11 of the 12 NPDR eyes and 1 of the 16 eyes without initial retinopathy (i.e., within 64 of the 919 analyzed zones). After accounting for the correlation among zones within each eye, a GEE (generalized estimating equation) analysis showed that mfERG implicit time, duration of diabetes, and retinopathy stage at initial visit are significant factors in determining the probability of developing new retinopathy (p<0.05). Neither blood glucose level during mfERG recording, nor diabetes type, was a significant factor for new retinopathy development (p=0.18 and p=0.76, respectively). Holding the other variables fixed, a 1 unit Z–score increase in mfERG implicit time produced an odds ratio of 1.38 and a relative risk of developing new retinopathy of 1.24. The area under the ROC curve of this model is 0.90 (p<0.001), indicating the model performs well. The predictive model has a sensitivity of 86% and a specificity of 84%. Conclusions: The significant risk factors for the development of local diabetic retinopathy over a one–year period are abnormal mfERG implicit times, the duration of diabetes, and eye status. Local mfERG implicit times are the only factors in the model that identify the specific locations of future retinopathy.
Keywords: diabetic retinopathy • diabetes • electrophysiology: clinical