Misclassification of Eyes With Progressive Keratoconus Using the KISA% Index

J Refract Surg. 2024 Sep;40(9):e614-e624. doi: 10.3928/1081597X-20240726-01. Epub 2024 Sep 1.

Abstract

Purpose: To determine the misclassification rate of the keratoconus percentage (KISA%) index efficacy in eyes with progressive keratoconus.

Methods: This was a retrospective case-control study of consecutive patients with confirmed progressive keratoconus and a contemporaneous normal control group with 1.00 diopters or greater regular astigmatism. Scheimpflug imaging (Pentacam HR) was obtained for all patients. KISA% index and inferior-superior (IS) values were obtained from the Pentacam topometric/keratoconus staging map. Receiver operating characteristic curves were generated to determine the area under the receiver operating characteristic curve (AUROC), sensitivity, and specificity values.

Results: There were 160 eyes from 160 patients evaluated, including 80 eyes from 80 patients with progressive keratoconus and 80 eyes from 80 control patients. There were 20 eyes (25%) with progressive keratoconus misclassified by the KISA% index, with 16 eyes (20%) of the progressive keratoconus cohort classified as normal (ie, KISA% < 60). There were 4 eyes (5%) with progressive keratoconus that would classify as having "normal topography" using the published criteria for very asymmetric ectasia with normal topography of KISA% less than 60 and IS value less than 1.45. All controls had a KISA% index value of less than 15. The optimal cut-off value to distinguish cohorts was 15.31 (AUROC = 0.972, 93.75% sensitivity). KISA% index values of 60 and 100 achieved low sensitivity (80% and 73.75%, respectively).

Conclusions: The KISA% index misclassified a significant proportion of eyes with progressive keratoconus as normal. Although highly specific for clinical keratoconus, the KISA% index lacks sensitivity, does not effectively discriminate between normal and abnormal topography, and thus should not be used in large data analysis or artificial intelligence-based modeling. [J Refract Surg. 2024;40(9):e614-e624.].

MeSH terms

  • Adolescent
  • Adult
  • Area Under Curve
  • Case-Control Studies
  • Cornea / diagnostic imaging
  • Cornea / pathology
  • Corneal Topography* / methods
  • Diagnostic Errors
  • Disease Progression*
  • Female
  • Humans
  • Keratoconus* / classification
  • Keratoconus* / diagnosis
  • Male
  • Middle Aged
  • ROC Curve*
  • Retrospective Studies
  • Sensitivity and Specificity
  • Visual Acuity / physiology
  • Young Adult