Purchase this article with an account.
C. Bowd, M. Leite, L. M. Zangwill, R. N. Weinreb, H. L. Rao, A. Tafreshi, P. A. Sample, F. A. Medeiros; Agreement Among SD-OCT Instruments for Assessing Retinal Nerve Fiber Layer Thickness in the Diagnostic Innovations in Glaucoma Study (DIGS). Invest. Ophthalmol. Vis. Sci. 2010;51(13):2973.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To assess the agreement of peripapillary retinal nerve fiber layer thickness (RNFLT) measurements among three SD-OCT instruments. We also examined the influence on agreement of age, axial length and disease severity (standard perimetry mean deviation).
330 eyes (88 with glaucoma, 206 OHT or glaucoma suspect by disc appearance, 36 healthy) from 208 individuals were imaged using RTVue (v188.8.131.52), Spectralis (v3.1) and Cirrus (v4.5) in a single visit. Agreement among RNFLT measurements (globally and by quadrant) was assessed using Bland-Altman plots and associations between measurements were assessed using linear regression. The influence of covariates on agreement was assessed by regression analysis between each covariate and the differences in RNFLT measurements between pairs of instruments.
Bland-Altman analyses showed a significant fixed bias (measurements are consistently different) for almost all of the pair-wise agreements, and a significant proportional bias (measurements differ more as measured RNFLT increases) for most of the pair-wise agreements. In general, RTVue measurements were somewhat thicker than Spectralis and Cirrus measurements (by up to 15 µm). Linear regression indicated that pairwise measurements from each instrument were significantly positively correlated and these correlations were moderate to strong (R2 ranged from 0.30 for Spectralis and Cirrus nasal thickness to 0.87 for RTVue and Cirrus global thickness). Global RNFLT agreement between RTVue and Spectralis was affected by age (p = 0.002), agreement between Spectralis and Cirrus was affected by axial length (p = 0.004) and agreement between Spectralis and both RTVue and Cirrus was affected by disease severity (p ≤ 0.001).
RNFL thickness measurements between several SD-OCT instruments are not entirely compatible and therefore these instruments likely cannot be used interchangeably. This may be due in part to difference in RNFL detection algorithms. Comparisons with histological measurements could determine which technique is most accurate.
This PDF is available to Subscribers Only