advertisement

Editors Selection IGR 16-4

Clinical Examination Methods: Perimetry in early glaucoma

Comment by Gustavo de Moraes on:

55312 Performance of the visual field index in glaucoma patients with moderately advanced visual field loss, Lee JM; Cirineo N; Ramanathan M et al., American Journal of Ophthalmology, 2014; 157: 39-43

Find related abstracts

Lee et al. investigated the relationship between changes in the visual field index (VFI) and mean deviation (MD) among patients with moderately advanced visual field damage whose sequence of 24-2 SITA tests crossed the -20 dB value at some point during follow-up. They included 148 eyes with six or more visual field tests that had a change in MD values of -16 dB to -24 dB, with at least one episode of crossing the -20 dB MD value. Absolute differences of MD and VFI between two consecutive tests (ΔMD, ΔVFI) were compared among three subsets of paired visual field tests: those with serial MDs all better than -20 dB (i.e., between -16 and -20 dB), those with serial MDs all worse than -20 dB (i.e., between -20 and -24 dB), and those with serial MDs crossing -20 dB. With regard to MD changes, the authors found that the mean ΔMD for serial MDs better than -20 dB was 1.1 (± 0.8) dB, the mean ΔMD for two consecutive MDs crossing -20 dB was 2.4 (± 1.2) dB, and the mean ΔMD for serial MDs worse than -20 dB was 1.0 (± 0.8) dB. On the other hand, the mean ΔVFI for serial MDs better than -20 dB was 4.6% (± 3.6%), the mean ΔVFI for two consecutive MDs crossing -20 dB was 15.8% (± 8.4%), and the mean ΔVFI for serial MDs worse than -20 dB was 3.6% (± 2.7%). When the visual field pairs crossed the -20 dB MD value, both the MD and VFI revealed greater change than when the serial MDs were better or worse than -20 dB, although this change was statistically significant only for ΔVFI (P < 0.001). To minimize the effect of differences in scales of measurement and their ranges, the authors then compared the ΔVFI/ΔMD ratio among the three scenarios. For ΔVFI/ΔMD ratios, the mean values were 6.8% (± 10.4%)/dB when the range of MD fell on either side of -20 dB and 7.9% (± 6.2%)/dB when the range of MD crossed the -20 dB values. This mean 1.1%/dB difference reached borderline statistical significance (P = 0 .042).

The authors' conclusion was that the value of VFI becomes highly variable in serial visual fields of eyes with MD crossing -20 dB, in comparison to those VFIs associated with MDs on either side of -20 dB. This finding was most likely due to the fact that the VFI is calculated from the pattern deviation probability value in eyes with MD better than -20 dB and from the total deviation probability values in eyes with MD worse than -20 dB.¹ They also stated that, given the conclusion above, the 'usefulness and the predictive ability of VFI are essentially lost.' They also suggested that the development of indices to measure visual field rates that are free from this effect in moderate to severe glaucoma is warranted and that using MD in this range of advanced field loss would be better than VFI. The conclusion driven by their data, as well as their recommendations needs to be interpreted with caution due to clinical and statistical issues not addressed as potential limitations in the article. One should be reminded that standard automated perimetry (SAP) threshold sensitivity values become highly variable in advanced glaucoma.² Recently, Gardiner et al.³ showed that SAP results may be unreliable when visual field locations have sensitivity below 15 to 19 dB, possibly due to saturation of retinal ganglion cells (RGC) with high-contrast stimuli. Since the MD is a weighted-average of point-wise sensitivities, at MD levels around -20 dB and worse any unit change could be solely due to randomness of RGC response in severely damaged parts of the field. Therefore, it remains questionable whether MD changes could provide estimates of rate of progression in severe glaucoma any better than the VFI or any other available visual field index.

It would be interesting to show the actual distribution of ΔVFI and ΔMD values between the second and first tests in this sample, instead of their absolute difference and their means and standard deviation. The example shown in Figure 1 provides a lot of information (note: the horizontal axes should have been be adjusted to include the entire range of both VFI and MD values. Also, the MD range should have been converted to percentage, as Bengtsson and Heijl did in their paper¹). Five of the seven (71%) times the MD crossed the -20 dB line corresponded to visual field 'improvement' as opposed to eight of 17 (47%) when looking at either side of the -20 dB. Those lines of MD 'improvement' match the lines with largest VFI change. If such observation was seen among other study eyes as well, the mean VFI variation could have been overestimated and could reflect test variability rather than true glaucomatous progression. Despite the valid argument provided for not excluding tests based on falsenegative responses, it would be worth knowing what would happen even with a loose cut-off value (e.g. 33%).

Both the VFI and MD calculation allot greater weight to central field locations ‐ which have smaller variability³ ‐ even though the VFI is more heavily weighted in central areas of the visual field due to their larger cortical representation.² Also, note that even though the mean ΔVFI/ΔMD ratio was larger when the range of MD crossed the 20 dB values (7.9% vs. 6.8%) than when the range of MD fell on either side of -20 dB, the standard deviation was smaller (6.2% vs. 10.4%), in addition to the marginal statistical significance (P = 0.042). Another point is that while the VFI is essentially a linear scale, MD is a non-linear one. VFI differences can be calculated as the authors did, but MD values need to be converted to a linear scale (differential light sensitivity, DLS) before their subtraction. For instance, a ΔMD of two below -20 dB is different from the same ΔMD above -20 dB.

Linear regression of the VFI in the GPA report should be interpreted with caution when a patient's MD crosses -20 dB

Such clinically and statistically small effect size (~1%) does not support the conclusion that the VFI becomes highly variable in serial visual fields of eyes with MD crossing -20 dB, in comparison to those VFIs associated with MDs on either side of -20 dB. In addition, one cannot conclude that using MD in this range of advanced field loss would be better than VFI. A more conservative interpretation would be that linear regression of the VFI in the GPA report should be interpreted with caution when a patient's MD crosses -20 dB. Given the difference in how the VFI is calculated on either side of a -20 dB MD, a 'broken-stick' regression, keeping in mind the limitations of SAP in cases of severe perimetric damage. Alternatively, switching to 10-2 strategy could be more appropriate in severely damage fields to monitor progression. Notwithstanding the points discussed above, Lee et al. highlighted the need to interpret the relationship between MD and VFI longitudinal changes critically before driving definite conclusions about glaucoma progression.

• ← Previous
• Next →