Abstract
Purpose: :
To evaluate the sensitivity and specificity of the three glaucoma classification programs, FSM discriminant function (FSM), Moorfields Regression Analysis (MRA) and the Glaucoma Probability Score (GPS) of the version 3.0 of Heidelberg Retina Tomograph (HRT) II in a population-based study.
Methods: :
Examination with HRT II was attempted on all subjects from the Tajimi Study, a population-based study of 3021 subjects (participation rate 78.1%) in Japan. We were able to obtain reliable HRT II measurements (SD <50µm) in 4029 eyes of 2229 normal subjects,103 eyes of 84 glaucoma subjects and 63 eyes of 56 glaucoma suspect subjects (diagnosed according to the International Society Geographical & Epidemiological Ophthalmology criteria). All HRT images were obtained by HRT II ver 1.0.7 during the Tajimi study and was converted to ver 3.0 for our current analysis.
Results: :
Sensitivity varied significantly among the three programs (65.0%, 40.8% and 68.0% with FMS, MRA and GPS respectively for glaucoma eyes, P<0.0001, Fisher’s exact test and 50.8%, 22.2% and 54.0% with FMS, MRA and GPS respectively for glaucoma suspect eyes, P=0.0004). For the normal eyes, specificity also varied significantly among the three programs with the highest specificity by MRA. (86.7%, 95.9% and 80.2% with FMS, MRA and GPS respectively, P<0.0001). Eighty-six normal eyes (2.1%) were judged as out of normal limits by all three programs. These eyes had significantly larger disc area, cup area and cup to disc area ratio than the remainder of normal eyes judged as within normal limits by at least one classification program. (2.63mm2 vs. 2.06mm2, 1.43mm2 vs. 0.48mm2, 0.54 vs. 0.22 respectively. All P<0.0001, unpaired t-test).
Conclusions: :
In a population-based setting, the sensitivity of the HRT II was not satisfactory with all three classification programs, especially with the MRA. Sensitivity for glaucoma suspect eyes were lower than that for definite glaucoma eyes despite the apparent optic disc changes detected by the graders. (P=0.02~0.04, chi-square test) However, specificity was satisfactory, especially with MRA. A significant percentage of the normal discs were labeled as abnormal by all three programs and eyes with larger discs or larger cup to disc area appear to be more likely to be misdiagnosed. HRT II sensitivity in detecting glaucoma in a population-based setting was much lower than that reported in hospital-based subjects.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound)