Comparison of subjective grading and objective assessment in meibography

Aim: To analyse repeatability of subjective grading and objective assessment in non-contact infra-red meibography.

Methods: Meibography photographs of 24 subjects (female 14; mean age=46; range=19-69 years, upper-lid images=12, lower-lid images=12) were classified in two sessions by three experienced observers (OI, OII, OIII). Relative area or portion affected by meibomian glands (MG) loss was classified applying three different grading scales in randomized order: a four-grade scale (4S) (degree 0=no partial glands; 1=<25% partial glands; 3=25-75% partial glands; 3=>75% partial glands), a pictorial five-grade scale (5S) (degree 0=no meibomian gland loss (MGL); 1=<25% MGL; 3=26-50% MGL; 3=51-75%; 4=>75% MGL) and objectively by a 100-grade scale (DA) applying ImageJ software.

Results: Observed MG loss ranged from 0% to 69%. Intra-observer agreement of the 5S (OI: κ=0.80, p<0.001; OII: κ=0.40, p=0.009; OIII κ=0.81, p<0.001) was better than of the 4S (OI: κ=0.79, p<0.001; OII: κ=0.15, p=0.342; OIII κ=0.50, p=0.0071). Intra-observer agreement of OI and OIII (±0.88 (95% confidence interval), ±1.305) was better than of OII (±2.21) in 4S and 5S (±0.99, ±2.00 and ±0.91; OI, OII and OIII, respectively) while it was relatively similar in DA (±18, ±17 and ±17). Inter-observer agreement was better in DA (OI-OII: ±13, OI-OII: ±19, OII-OIII: ±26) than in 4S (OI-OII: ±1.76; OI-OIII: ±1.29 and OII-OIII: ±1.31) or 5S (OI-OII: ±1.49; OI-OIII: ±0.91 and OII-OIII: ±1.20).

Conclusion: Intra-observer and inter-observer agreement was better in computerized grading followed by the subjective five-grade scale and four-grade scale.