Professor Hood and colleagues evaluated the sensitivity and specificity of 24-2 guided progression analysis (GPA) in 29 healthy controls and 70 glaucoma suspects or patients with early glaucoma (baseline MD > −6 dB) with at least four serial VF and OCT studies over a period of at least one year. The reference standard was expert assessment of OCT imaging data and Humphrey visual fields (VF).1 Subjects prospectively underwent a mean of 9.2 tests over 31 months in the context of the Macular Damage in Early Glaucoma and Progression Study. Among the glaucoma and glaucoma suspect subjects, almost half had normal baseline 24-2 and 10-2 VFs.The study's reference standard for progression required agreement among three glaucoma specialists on three findings: (1) worsening of the circumpapillary RNFL profile; (2) corresponding change of the RNFL and GCL on probability maps; and (3) worsening of corresponding locations on the 10-2 or 24-2 VFs. The graders used the investigators' custom-made OCT progression report (which only relies on the first and final OCT studies) and commercially available VF reports.Ten eyes were defined as having progressed by the reference standard criteria and 15 eyes had likely (n = 9) or possible progression (n = 6) on GPA. Two of the healthy control eyes had likely progression on GPA (false positives). The severity of the VF abnormality in these cases is striking, particularly for VFs performed in a research setting, and goes unexplained.Six of the ten eyes defined as having progressed by the reference standard were not detected as having progressed by GPA (false negatives). All six of these had progression in the central 8° of the VF. There are only 12 locations in the central 10° of the VF represented in the 24-2 test, and macular glaucomatous damage may only affect one or two locations in the 24-2 VF. Since progression by GPA requires the presence of at least three locations of significant deterioration, early progression of focal central VF damage can go undetected by the software algorithm. Glaucoma experts reading this already know a clinician should not accept the GPA progression classification result without a critical review. It has been reported that the central 12 locations in the 24-2 field can be used to characterize central visual field damage.2 Indeed, in another study, more than 90% of cases of disagreement between GPA and expert graders (who were aware of the GPA result but had no other information) were due to experts calling progression when GPA did not.3
Early glaucoma sometimes results in central visual field damage that can be missed by the 24-2 GPA algorithm
Nine eyes that did not meet the reference standard criteria for progression had likely (n = 4) or possible progression (n = 5) with GPA (possible false positives). Most of these eyes had normal and unchanged cpRNFL profiles, suggesting these were indeed false positive progression calls by GPA. This is not a surprise for the five eyes with the GPA classification of possible progression, since it is established that false positive progression calls are common in the absence of three confirmatory tests. The criteria for likely progression with GPA are essentially the same as those for defining VF progression in the Early Manifest Glaucoma Trial (EMGT). Since the subjects in the control group in EMGT were untreated, the progression detection algorithm was designed to be very sensitive for the detection of progression in eyes with glaucomatous VF defects at baseline.The investigators did not report rates of change in either the visual field or structural parameters. Although some information can be gleaned from viewing the GPA reports in the figures, the duration of observation and the number of visual fields and OCT tests that were performed in the cases in which GPA resulted in false negative or false positive progression calls is not formally reported. Optic disc photographs were not evaluated as part of the study nor were OCT parameters of optic disc morphology. Nevertheless, this study nicely highlights the fact that early glaucoma sometimes results in central visual field damage that can be missed by the 24-2 GPA algorithm.