Abstract
Purpose :
To compare variables for early glaucoma detection.
Methods :
Both eyes from altogether 89 patients with at least one open angle glaucoma suspect eye were consecutively included in a 5-year prospective study with examinations every 4th month. Glaucoma suspect was defined as either of; ophthalmoscopically suspect optic nerve head, IOP above 22 mmHg, pigment dispersion syndrome, exfoliation syndrome, Humphrey 24-2 visual field horizontal hemifield asymmetry or any combination in at least one eye. IOP (Goldman applanation, 1), contrast sensitivity (Mean deviation. 1), ONH nerve fibre layer mass (confocal scanning laser tomography: CD-linear, NRA global, 3), ONH juxtaposed nerve fibre layer mass (OCT Rnfl, 3, Laser polarimetry, 3). Number indicates number of iterated measurements averaged at each occasion. For the current analysis, specificity for each of the variables measured was estimated based on a confidence limit for 0.95 tolerance as the critical limit in eyes clinically judged as not glaucoma suspect at the first occasion. The confidence coefficient was set to 0.95. Eyes expressing a loss of MD beyond the critical limit at the end of the 5-year observation period were considered to express manifest glaucoma. The measurent at the first occasion in a subsample of one manifest glaucoma eye was evaluated for estimation of sensitivity for early detection of glaucoma. Specificity and sensitivity estimates were based on measurements in one eye from each subject and on different subsamples of subjects.
Results :
The sensitivity for glaucoma detection at the first visit was estimated to; MD 0.92 (36/35), Stratus-Rnfl 0.47 (14/19), Rnfl Cirrus 0.18 (18/10), Gdx-Tsnit 0.29 (10/27), HRT-CD-linear 0.26 (24/29), HRT-NRA 0.19 (30/33), IOP 0.18 (39:34). Numbers in brackets are d.f.: not pathological/pathological. Losses were due to subject death, device malfunction or subject withdrawal. The sensitivities based on estimated means and standard deviations without considering sample size normalized to MD was: MD 1.0, Stratus-NFL 0.7, Cirrus-Rnfl 0.5, Gdx-tSnit 0.4, HRT-CD-linear 0.4, HRT-NRA 0.3, IOP 0.2.
Conclusions :
Estimates of sensitivity depends on sample sizes for not pathological and pathological. Our measurements indicate that the most efficient single variable for detection of glaucoma in a glaucoma suspect eye is MD. The most efficient morphometric variable is Rnfl. IOP is of limited value in glaucoma suspect patients.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.