Reproducibility of retardation measurements with the nerve fiber analyzer II

To evaluate the reproducibility of a scanning laser polarimeter, the Nerve Fiber Analyzer II (NFA II, Laser Diagnostic Technologies, San Diego, CA, U.S.A.).

Five independent retardation maps of the peripapillary retina of five normal eyes were acquired by three experienced operators (including V.S.G.) on each of three separate days for a total of 45 retardation maps per patient. Two methods of image processing, one using a baseline image and another using the individual scans, were used to compare the reproducibility of three summary measures, average retardation, integral, and retardation ratio.

The average standard deviation (and its 95% confidence interval) of average retardation within a 10-pixel-width-band of the 9 baseline images was 0.43 degree (0.36-0.51 degree) with a mean coefficient of variation of 4.2% (3.8-4.5%). In a random effects model, each of the three retinal nerve fiber layer (RNFL) summary measures varied significantly by patient (p < 0.016), but not by operator (p > 0.19), or operator by patient interaction (p > 0.524). In addition, there was small, but statistically significant day-by-operator-within-patient (intraobserver) variation in the random effects model.

These results suggest that the NFA II provides reproducible measurements and that, on average, measurements obtained by separate operators on different days are similar.