[Repeatability and reliability of the visual acuity examination on logMAR ETDRS and Snellen chart]

The main goal of our study was to prove the statistical significant correlation between repeated measurements and test-retest variability TRV expressed with confidence interval CI according to Bland-Altman's method on 95% level of statistical confidence. The methods are threshold interpolation logMAR method on ETDRS chart, whole-line method on Snellen chart and modified threshold interpolation method on Snellen chart with Sloan letters.

We had 468 measurements measured with threshold interpolation log MAR method on ETDRS chart, whole-line method on Snellen chart and modified threshold interpolation method on Snellen chart with Sloan letters. The average value of the first sequence of measurements measured with whole-line method on Snellen chart was -0.043 logMAR (min. 1, max. -0.30. SD 0.25) and of the second sequence of measurements was -0.045 logMAR (min. 1, max. -0.30. SD 0.23). The average value of the first sequence of measurements measured with interpolation method on Snellen chart was -0.018 logMAR (min. 0.98, max. -0.30. SD 0.29) and of the second sequence of measurements was -0.024 logMAR (min. 1, max. -0.80. SD 0.29). The average value of the first sequence of measurements measured with interpolation method on ETDRS chart was -0.0612 logMAR (min. 0.72, max. -0.30. SD 0.21) and of the second sequence of measurements was -0.0610 log MAR (min. 0.8, max. -0.28, SD 0.21).

We have proved that all methods do not have statistical significant difference between repeated measurements (Wilcoxon paired test, whole-line method on Snellen chart p = 0.74, interpolation method on Snellen chart p = 0.33 and interpolation method on ETDRS p = 0.95) and they also have statistical significant correlations (Spearman correlation coefficient, whole-line method on Snellen chart r = 0.91, p < 0.0001, interpolation method on Snellen chart r = 0.89, p < 0.0001 and interpolation method on ETDRS chart r = 0.89, p < 0.0001). TRV expressed with CI on 95% statistical significance level according to method of Bland-Altman was with whole-line method on Snellen chart +/- 0.11 (i.e., 5 letters), with interpolation method on Snellen chart +/- 0.20 (i.e., 10 letters) and with interpolation method on ETDRS +/- 0.08 (i.e., 4 letters).

In our study we proved that there is no statistical significant difference between repeated measurements with all three methods. After that we proved statistical significant correlations between repeated measurements with all three methods. The biggest clinical importance has determination and confirmation of the CI value of TRV. With interpolation method on ETDRS it was CI +/- 0.08 (i.e., +/- 4 letters), with whole-line method on Snellen it was CI +/- 0.11 (i.e., +/- 5 letters) and with interpolation method on Snellen it was CI +/- 0.20 (i.e., +/- 10 letters). Our recommendation based on results of our study is that is not clinically appropriate to convert measured values from Snellen optotype to logMAR and perform interpolation method on Snellen chart.