The prediction of glaucoma from ocular biometric data: part 2 a evaluation

A method of glaucoma prediction from ocular biometric data has been described previously. A study was undertaken to evaluate the performance of the existing multiple regression equations (prediction systems) on data obtained from an independent sample consisting of 22 angle-closure glaucoma, 29 open angle glaucoma and 44 normal subjects. This performance, found by comparing the predicted and actual classification for this sample, was such that between 2 and 7% false positives and 12 and 27% false negatives were found on the equations differentiating glaucoma from normal subjects; and between 14 and 27% false positives, with 10 to 14% false negatives on the equations classifying the glaucoma subjects as angle-closure or open angle. From these results the efficiency of glaucoma prediction from ocular biometric data would appear to be equal to that of the combined tonography and provocative tests, provocation with corticosteroids and visual field screening.