The high ICCs found for the OCT (>0.90;
Table 3 ) reflect the instrument’s high reliability, which we believe agrees with the values reported by both Paunescu et al. for the OCT (Stratus 3; Carl Zeiss Meditec) and Massin et al. for previous OCT technology (see also the second paragraph above).
10 11 In contrast, the ICC values found for the RTA are moderate (0.60–0.89) to poor (0.30–0.59) in comparison to the OCT, particularly for extrafoveal areas. Zou et al.
15 report an ICC for the RTA of 0.95 for average foveal thickness in 24 eyes of 24 healthy Chinese subjects, which is in contrast with our findings (0.50;
Table 3 ). This difference is mainly due to the smaller total variance (i.e., square SD) of our study in comparison to their report (SD of 14 μm vs. 26 μm, respectively). In addition, in their method for calculation of the ICC only two scans were used in a two-way random effects model. The lower number of measurements (i.e., two instead of three) could cause an underestimation of the within-patient variance (i.e., square σ
within-patient) with possible overestimation of the ICC. Furthermore, they report excluding images with poor quality in their general survey, and thus possibly also for their assessment of reliability. The influence of total variance on the ICC is even more clear when in our study the ICC of average foveal thickness (A1) in healthy eyes (0.50) is compared to that in eyes with macular edema (0.89,
Table 3 ). Although there is a difference in σ
within-patient values between both groups (12 and 19 μm, respectively) this difference was not found to be statistically significant (
P = 0.14;
Table 2 ). Hence, the difference in ICC is mainly due to a difference in total variance. This can be seen in
Table 1 , where the SD of average foveal thickness (A1) between these two groups shows a marked difference (14 μm vs. 65 μm,
P < 0.001), which can be expected if macular edema influences macular thickness. This, in our opinion, underlines the importance of providing measurement precision values as an absolute value of the within patient variance (or σ
within-patient) or a modification thereof (such as the 95% limits of agreement), because the ICC alone can be misleading. To our knowledge, no previous studies have assessed measurement reliability for extrafoveal areas in both instruments. Our results clearly show the OCT to be more reliable and precise for detecting the presence of macular edema before the foveal center is involved and visual acuity is affected, which in our opinion makes it the preferred choice for detection and follow-up of diabetic macular edema.