Variance components in confocal scanning laser tomography measurements of neuro-retinal rim area and the effect of repeated measurements on the power to detect loss over time.

Abstract

To estimate the variation in measurements of neuro-retinal rim area (NRA) determined by confocal scanning laser tomography and consequences for clinical follow-up. Altogether, 24 healthy subjects were randomized on -320μm, Moorfields and Standard NRA plane strategies. Additionally, NRA was measured in 32 glaucoma subjects. Variance components for subjects, visits and measurements were estimated with analysis of variance. Sample sizes required to detect a 6.0×10-2 mm2 NRA change were estimated assuming a significance level of 0.05 and a power of 0.8. Consequences for independent group, and paired comparison design, respectively, were analysed. Further, precision in estimates within subjects over time was investigated. The variation of NRA among subjects was considerably larger than the variation among visits and measurements. For glaucoma subjects, the variation among visits and measurements were of the same order but larger than in healthy subjects. It was found that independent group comparisons require inconveniently large sample sizes. Within-subject paired comparisons over time require sample sizes of below 15 subjects. The estimated variations for glaucoma subjects imply that 54months of follow-up is required for detection of change from baseline. The variance for subjects is substantial in relation to those for visits and measurements. Cross-sectional independent group comparisons of levels of NRA are unsuitable, due to considerable subject variation. Levels of NRA differences within subjects between visits can be estimated with acceptable precision. Neuro-retinal rim area (NRA) measurement can be used for long-term follow-up of glaucoma progression.