Densitometry marks delineating the affected area in keratoconus: clinical suitability of a new descriptive system...

To present a descriptive system for the elliptic demarcation area seen in keratoconus densitometry maps (obtained with a Scheimpflug tomographer) and to evaluate its suitability for clinical practice.

The subjects were 30 keratoconus patients at different stages and 20 healthy subjects. The averaged densitometry maps ('two-layers' scan, with fixed layers 120 µm and endothelium) were analysed using a system of four categories (termed 'Brightness', 'Contrast', 'Decentration' and 'Octants surrounded by a dark line') that we created to characterise the demarcation area. Four examiners (three corneal specialists and one junior resident) used the system to classify the maps. The inter-rater agreement was calculated for two subgroups: (1) clinical keratoconus patients and (2) both healthy patients and forme fruste keratoconus patients. Intra-rater agreement was also determined.

Inter-rater agreement on classification was higher when analysing clinical keratoconus, reaching levels of substantial agreement. Despite this, only low levels of agreement were found in 'Decentration', penalized due to the skewness in the distribution of this descriptor. Almost perfect intra-rater agreement was obtained for all descriptors in the first subgroup of clinical keratoconus, whereas the agreement was generally moderate within the second subgroup of normal and forme fruste eyes. Agreement was slightly lower with the less experienced observer. At least three observers agreed on four forme fruste keratoconus eyes presenting abnormalities in the images. The observers reported that the 'Brightness' descriptor was subjective and redundant with 'Contrast'.

The description of the area was repeatable and reproducible, and may be a valuable supplement when documenting clinical keratoconus stage and progression in daily practice. However, a minor learning curve was noticed and agreement was higher among the more experienced observers. Since the descriptor 'Brightness' was found to be subjective and redundant, it was excluded from the final proposed classification.