Abstract
Purpose:
The aims of this study were to evaluate the intra-examiner and inter-examiner reliability of the Gestalt, meiboscore, and meibomian gland counting grading scales associated with meibography.
Methods:
Sixty subjects were recruited for two study visits separated by 7 ± 2 days. At each visit, the Oculus Keratograph 4 was used to obtain a non-invasive break up time (NIK-BUT) and infrared meibography pictures of each inferior lid’s meibomian glands, amongst other tests. The meibography images were duplicated, and each set was randomized and graded by two masked graders using the gestalt, meiboscore, and meibomian gland count grading scales. The results were analyzed using kappa and weighted kappa values for the categorical scales of gestalt and meiboscore. Intraclass correlation coefficient (ICC) was used to analyze the continuous scale of meibomian gland counting.
Results:
Kappa values for intra-examiner reliability of the same picture suggest fair to moderate agreement (K = 0.25 - 0.60, Kw = 0.40 - 0.68) when using gestalt grading and fair to substantial agreement (K = 0.32 - 0.63, Kw = 0.52 - 0.69) when using meiboscore grading. The ICC to assess intra-examiner reliability using meibomian gland counting suggests moderate to almost perfect agreement (ICC = 0.58-0.85). The kappa values of the intra-examiner reliability of two pictures taken of the same lid at different visits suggest slight to moderate agreement (K = 0.11 - 0.47, Kw = 0.21 - 0.57) when using gestalt grading and slight to substantial agreement (K = 0.15 - 0.56, Kw = 0.21 - 0.63) when using meiboscore grading. The ICCs to assess intra-examiner reliability across two pictures of the same lid using meibomian gland counting suggest fair to strong agreement (ICC = 0.31 - 0.65). The kappa values used to assess inter-examiner reliability suggest no better than chance to light agreement (K = -0.03 - 0.03, Kw = 0.07 - 0.09) when using gestalt grading. The ICC to assess inter-examiner reliability using meibomian gland counting suggests moderate agreement (ICC = 0.46).
Conclusions:
The Oculus Keratograph provides qualitative, gradable meibomian gland images. Meibomian gland counting may be the most reliable way to grade meibography when considering intra- and inter-examiner grading followed by meiboscore and gestalt. The individual examiner reliability suggests that experience is an important factor in reliability.
Keywords: 486 cornea: tears/tear film/dry eye •
465 clinical (human) or epidemiologic studies: systems/equipment/techniques •
550 imaging/image analysis: clinical