Abstract
Abstract: :
Purpose: To examine and compare the first (K1) and second (K2) order kernels of the standard mfERG in diabetics with early non–proliferative diabetic retinopathy (NPDR). The mfERG kernels were examined in two ways: (1) their local correlations with retinal lesions; and (2) their abilities to predict local NPDR that developed one year later. Methods: At baseline, standard mfERGs and 50 deg fundus photos were obtained from one eye of each of 16 diabetics with early NPDR. K1 and K2 waveforms were recorded using 10–300 Hz filtering and combined into 35 responses in each eye by averaging adjacent retinal areas. The 35 K1 and K2 waveforms were then digitally filtered 10–100 Hz. The most sensitive and/or reliable local measure currently available for each kernel was utilized: implicit time of K1 using the template stretching technique (Hood & Li, 1997) and signal–to–noise ratio (SNR) amplitude of K2. Abnormalities were defined for both kernels as values falling below the fifth percentile of a group of 30 normal subjects (P < 0.05). First, the spatial correlation of baseline mfERG abnormalities and the baseline fundus gradings was examined. Second, 12 of the 16 diabetic subjects were followed–up one year later, and the prediction of new NPDR based on the baseline K1 and K2 abnormalities was examined. Results: K1 and K2 abnormalities occurred in 58% and 42%, respectively, of a total of 560 (=35x16) retinal areas. Both mfERG kernels were spatially associated with fundus gradings (P=0.02 for K1 and P=0.002 for K2) at baseline, with a larger number of mfERG abnormalities associated with more severe lesions. Sensitivity of detecting NPDR was comparable for the two kernels (58% for K1 and 54% for K2; P = 0.2), but K2 had higher specificity for identifying healthy retinal areas than K1 (61% for K2 vs. 48% for K1; P = 0.002). In the 12 diabetics who were followed–up one year later, 28 retinal patches developed new retinopathy, majority of which were microaneurysms. Both of the kernels recorded at baseline predicted the new retinopathy (Odds Ratio (OR) = 2.3 for K2, P = 0.03; OR = 6.0 for K1, P < 0.001). K1 abnormalities identified 82% of the locations of future NPDR while K2 abnormalities identified 54% of them (P < 0.001). Conclusions: Abnormalities of second order kernel SNR have better spatial correlation with current NPDR than first order kernel implicit time abnormalities. However, if future diabetic retinopathy is of interest, the implicit time of the first order mfERG kernel has greater predictive power than second order kernel SNR.
Keywords: electroretinography: clinical • diabetic retinopathy • diabetes