PURPOSE: To determine the interobserver and intraobserver repeatability in using the OCULUS Keratograph 4 (K4) and 5M (K5M) to grade meibomian gland (MG) dropout using meibography grading scales.
METHODS: The inferior and superior eyelids of 40 participants (35 women, 5 men; mean age = 32 years) were imaged three times each on both instruments. The images were split into one training and two study sets; the latter were graded (four-point meibography scale) by two observers on two separate occasions (24 hours apart) to determine repeatability. Semiobjective quantification of percentage MG dropout was conducted using ImageJ on K4 and K5M images. A finer seven-point meibography scale was used to grade a separate set of K5M images.
RESULTS: For the four-point scale, interobserver mean difference (MD) (±SD) was 0.08 (±0.55) on day 1 and 0.13 (±0.50) on day 2, and the concordance correlation coefficient (CCC) was 0.79 and 0.81 on days 1 and 2, respectively. Intraobserver MD (±SD) was 0.04 (±0.54), CCC = 0.79 for observer 1; intraobserver MD (±SD) was -0.09 (±0.60), CCC = 0.74 for observer 2. For the seven-point scale, interobserver MD (±SD) was 0.05 (±0.45), CCC = 0.89 on day 1, and interobserver MD (±SD) was 0.01 (±0.41), CCC = 0.91 on day 2. Intraobserver MD (±SD) was -0.10 (±0.35), CCC = 0.93 for observer 1, and intraobserver MD (±SD) was -0.06 (±0.30), CCC = 0.95 for observer 2. Percentage dropout measured between the K4 and K5M images showed lack of agreement, with 21.8% coefficient of repeatability. There was no significant correlation (r < 0.2; p > 0.05) between meibography score and clinical signs (corneal staining, gland expressibility, telangiectasia, vascularity, lash loss); however, there was a high correlation (r = 0.77; p < 0.05) between meibography score with percentage dropout.
CONCLUSIONS: Observers graded from -1 to +1 grade units between and within themselves for a four-point scale, 95% of the time. Although the interobserver and intraobserver repeatability of the K4 and K5M were very similar, a high rate of disagreement in percentage dropout between K4 and K5M images suggests that the two instruments cannot be interchanged. Meibomian gland dropout scores did not correlate significantly with clinical signs. Using a finer scale may be beneficial for detecting change.