Retinal Nerve Fiber Layer Imaging with Spectral-Domain Optical Coherence Tomography: A Variability and Diagnostic Performance Study

To evaluate and compare the retinal nerve fiber layer (RNFL) measurement variability, diagnostic sensitivity and specificity for glaucoma detection, and strength of the structure-function association obtained with a spectral-domain optical coherence tomography (OCT) device (Cirrus HD-OCT; Carl Zeiss Meditec, Inc., Dublin, CA) and a time-domain OCT device (Stratus OCT; Carl Zeiss Meditec, Inc.). Prospective, cross-sectional study. Ninety-seven normal subjects and 83 glaucoma patients. One eye from each subject was imaged with Cirrus HD-OCT and Stratus OCT. Sixteen and 31 normal eyes were selected randomly to evaluate intravisit repeatability and intervisit reproducibility, respectively. The agreement of RNFL measurements was evaluated with Bland-Altman plots. The diagnostic sensitivity and specificity was examined with the area under the receiver operating characteristic curve (AUC). The association between average RNFL thickness and visual field sensitivity was evaluated with a second-order regression model. Retinal nerve fiber layer measurement variability, AUC, and coefficient of determination (R(2)). The intravisit repeatability of Cirrus HD-OCT ranged between 5.12 and 15.02 mum, and the intervisit reproducibility ranged between 4.31 and 22.01 mum. The intervisit variabilities of sectoral and average RNFL thicknesses were lower in Cirrus HD-OCT compared with Stratus OCT with significant differences at 1, 3, 4, and 8 to 11 o'clock (P< or =0.021). There were proportional biases of RNFL measurements between the 2 OCT devices. The difference of RNFL thicknesses increased with the means. The average (AUC, 0.962 for Cirrus HD-OCT and 0.956 Stratus OCT), superior (AUC, 0.963 and 0.950, respectively), and inferior (AUC, 0.949 and 0.931, respectively) RNFL thicknesses demonstrated the greatest AUCs in both OCT devices with no significant difference detected between the respective measurements (P> or =0.120). The strength of the structure-function association was comparable between Cirrus HD-OCT (R(2) = 0.580) and Stratus OCT devices (R(2) = 0.623; P = 0.918). Although the diagnostic performance and the strength of the structure-function association were comparable between Cirrus HD-OCT and Stratus OCT RNFL measurements, Cirrus HD-OCT demonstrated lower measurement variability compared with Stratus OCT with significant differences at 1, 3, 4, and 8 to 11 o'clock. The poor agreement was likely related to the different inherent characteristics of the 2 OCT systems.