Accuracy of ultrasonic pachymetry and videokeratography in detecting keratoconus.

PURPOSE: To compare the accuracy of ultrasonic pachymetry measurements and videokeratography-derived indices in distinguishing keratoconus patients from those with normal eyes.
SETTING: A subspecialty cornea practice (Los Angeles, California, USA) and the Keratoconus Genetics Research Project.
METHODS: Corneal thickness was measured by ultrasonic pachymetry at the center and inferior margins of the pupil of 142 normal and 99 keratoconus patients The corneal surface topography of patients was studied with the Topographic Modeling System (TMS-1). The videokeratographs obtained were analyzed with a computer program that automatically calculates two indices derived from data points in the central and paracentral cornea: central K and I-S values. Linear discriminant analysis was used to determine the correct classification percentages using pachymetry measurements and indices derived from videokeratography as the independent variables.
RESULTS: The range of corneal thickness in normal and keratoconic eyes overlapped considerably. In the discriminant analysis, videokeratography indices provided a 97.5% correct classification rate and pachymetry data, an 86.0% rate (P < .01, McNemar's test).
CONCLUSION: Keratoconus is more accurately distinguished from the normal population by videokeratography-derived indices than by ultrasonic pachymetry measurements. This may be due to the large variation in corneal thickness in the normal population or the inability of ultrasonic pachymetry to accurately detect the location of corneal thinning in keratoconus by measuring standard points on the cornea. Pachymetry should not be relied on to exclude or diagnose keratoconus because the false-negative and false-positive rates are unacceptably higher than those obtained by videokeratography.