We evaluated global and localized rates of progression before and after DH and investigated whether there was a correlation between past VF loss, DH, and future progression. We found that rapid, localized VF loss not only occurs after DH, but importantly, also before it. Significant progression tended to remain within the same VF sector before and after the DH and the location and speed of progression were strong predictors of a future DH. Eyes with DH progressed faster than fellow non-DH eyes throughout the study period, and there was an accelerated VF loss after their onset. Structurally, most DH occurred within or adjacent to the area of localized neuroretinal rim or RNFL loss, which corresponded to the previously progressing VF sector.
To the best of our knowledge, this is the only series of DH to focus on localized VF progression both before and after their occurrence. Among the major glaucoma clinical trials, the Ocular Hypertension Treatment Study (OHTS)
8 found 128 DH eyes and the Early Manifest Glaucoma Trial (EMGT)
26 reported 140 patients with ever-observed DH based on photograph review. Progressive VF loss has been reported in ocular hypertensive and glaucomatous eyes after DH.
1 2 3 4 5 6 7 8 Most of the classification systems used were based on event analysis—that is, progression was determined by comparing a set of baseline VF tests with a final test resulting in a binary outcome (progression: yes/no). Our study used PLR to determine rates of progression which, despite requiring a greater number of VF tests, may be more specific and less subjective method to determine progression.
18 19 This greater specificity is particularly true when confirmatory methods are used.
19 By using a less conservative method for definition of progression, we may have encountered greater sensitivity and thus higher false-positive rates of progressing points. Even though not recommended for a clinical application, this method allowed us to increase the detection of progressing points that could ultimately be compared with the location of DH. Chauhan et al.
25 recently revised the general recommendations to evaluate rates of progression in glaucoma and reinforced the fact that the minimum number of VFs may vary according the velocity of progression. That is, more tests may be required for slow progressors and fewer tests for fast progressors. In our sample, which included eyes with expected fast rates of progression, a minimum number of 5 VF had a power of 80% to detect a significant rate of change in this population. The prevalence of significant VF progression after an average of 45 months after DH in our study (78%–82%) was similar to that in previous reports.
1 2 3 4 5 6 7
Several research groups have addressed topographic relationships between structural and functional loss and the onset of DH.
27 28 29 30 31 32 33 It has been reported that localized neuroretinal rim loss is a significant predictor of the occurrence of DH.
27 This observation supports our finding that 90% of DH occurred in areas of localized rim notching. This characteristic was found in several other studies
28 29 30 31 and supports the premise that ultrastructural features within areas of notching may ultimately lead to hemorrhage. We confirmed previous reports of the association between DH and future localized progression,
32 33 34 and assessed the local and global rates of progression for these eyes.
Our data demonstrate that rapid, progressive, localized functional loss occurs most often in the area in which the future DH will occur. Law et al.
29 evaluated a series of disc photographs preceding the onset of DH and found that among eyes that later developed DH, 100% had preexisting neural rim notches that antedated the occurrence of the DH by a median of 21.5 months. The authors also speculated that a neural rim notch and DH may be different presentations of an underlying degenerative process corresponding to a localized geographic area of the optic disc and its surroundings. Jeoung et al.
31 showed that significant RNFL loss using optical coherence tomography was already present in DH eyes with apparently normal RNFL configuration by fundus photography, indicating that a damaged RNFL, despite normal achromatic perimetry, was already present.
Given these prior reports and our findings of spatially congruous, rapid, localized, functional progression before DH, it appears likely that more rapid disease progression (i.e., a relatively rapid progressive neurodegeneration) leads to DH. In our study, each 1 dB/y increase in the progression rate within a VF sector increased the chance of developing a DH by 47%, whereas a localized rate of progression ≥3.0 dB/y increased the risk of DH by 65%. Our data support the hypothesis that local DH develops because of an ongoing degeneration of rim architecture (neural tissue+extracellular matrix+vessels), causing an abrupt rupture of the microvasculature. VF loss continues because of the ongoing structural damage at or adjacent to this location and is consistent with the usual clinical appearance of a DH at the edge of an existing RNFL or neuroretinal defect
(Fig. 5) . Slower disease progression (and therefore slower functional and structural injury) is less likely to be associated with DH because of the slower rate of localized progressive neurodegeneration.
The rates of VF loss in DH eyes were, on average, twice as fast as their fellow, non-DH eyes. This finding is particularly relevant, as it shows that these eyes are not only likely to progress in the future, but should also alert the clinician that treatment until that moment may have been suboptimal. The Early Manifest Glaucoma Trial (EMGT)
26 suggested that DH cannot be considered an indication of insufficient IOP-lowering treatment, and that glaucoma progression in eyes with DH cannot be totally halted by IOP reduction. The fact that the authors found that IOP-reducing treatment was unrelated to the presence or frequency of DH could indicate that treatment goal in that study may have been insufficient to slow VF progression before the onset of the DH. On the other hand, Miyake et al.
35 showed that the cumulative probability of detecting a DH significantly decreased after trabeculectomy. The more aggressive IOP reduction obtained surgically is likely to be more effective in slowing rates of progression,
36 which may ultimately prevent the degeneration of disc neuroretinal tissue. Moreover, the OHTS found an increased cumulative incidence of DH after a glaucoma end point was reached (0.5%–2.5% per year).
8 As we demonstrated, DH eyes already have faster rates of progression than fellow, non-DH eyes before the onset of the DH, which subsequently increased after its occurrence. Our results agree with the findings of the OHTS and are consistent with localized progression even before the DH were detected. Even though we did not evaluate the role of IOP reduction on the rates of progression after DH in this study, given that IOP is the main risk factor for glaucoma development and progression, there is strong evidence that eyes with lower IOP are more likely to progress more slowly than those with higher IOP.
37 38 39 40
We also observed that the correlation between DH and past or future progression depended on its location. Hemorrhages in nasal optic disc sectors were associated with slightly slower progression rates and showed poor topographic agreement with VF progression. Other studies investigating structure and function correlations in glaucoma have also shown weaker correlations between those sectors and VF loss.
22 23 A possible explanation is that the Garway-Heath map and a 24-2 strategy resulted in worse representation of peripheral retinal areas that are not tested with conventional static perimetry. Moreover, the nasal optic disc sector corresponds to 110°, or approximately 30% of its circumference. The corresponding VF sector contains only four points, or 7% of the points of a 24-2 test
(Fig. 1) . It is possible that if remote areas of the retina could be tested and plotted in a map, we could have found faster rates and better agreement for the nasal sectors. However, since standard achromatic perimetry remains the best method of assessing visual function in a clinical basis,
41 42 the finding of DH in the nasal sectors of the disc has low association with past or future field loss, as assessed by conventional perimetry.
The limitations of our study are its retrospective nature and the relatively low frequency of optic disc documentation (once per year), which may have lead to underestimation of the true incidence of DH in our sample. In the EMGT,
26 for instance, all patients were seen every 3 months and were photographed twice a year, which resulted in a much higher incidence of DH (55%) than our study. Hence, one could argue that the fast progression rates observed in our study before DH could be due to previously undetected hemorrhages, as most recurrences tend to occur close to previous ones.
1 7 We tried to minimize this effect by enrolling only the first detected DH of each eye in cases of recurrence. We maximized the likelihood of detecting DH by choosing a population undergoing repeated optic disc photography and perimetry and by performing masked review of disc photographs.
1 8 Similar to group B, eyes in group A were followed for a mean of 6 years before the onset of DH and underwent photography once per year, on average
(Table 1) . The use of the fellow, non-DH eyes as a control group allowed us to demonstrate that the progression rates before DH were significantly greater in the DH eyes. We used fellow eyes as controls to (1) match the groups for sex, race, age, type of glaucoma; (2) eliminate the bias that could be introduced by differences between the groups with respect to systemic factors that could be implicated in the development of DH; (3) maintain parity in the number of VF and disc photographs to keep the chance of finding a DH roughly the same; and (4) permit use of a paired statistical analysis. It is possible, or even likely, that we may have missed some DH that occurred during the intervals between photographs in our study or that a DH, before the beginning of our study, was followed by later, fast progression. These possibilities notwithstanding, the DH identified as the DH reference for groups A and B invariably occurred in the region of most rapid localized VF progression and that in group C this was true both before and after the hemorrhage. Ideally, a prospective longitudinal study with disc photographs performed at much more frequent intervals would give a better estimation of the true incidence of DH in a treated glaucoma population. The findings of the present study suggest that clinicians may want to increase surveillance of patients undergoing rapid localized progression and focus diagnostic resources (such as disc photography) on these individuals on a more frequent basis.
Our results suggest that rapid, localized, spatially consistent, VF progression occurs before and after DH. This supports the hypothesis that DH results from an ongoing structural degeneration of the neuroretinal rim, rather than a primary vasculopathic process
11 12 that subsequently leads, de novo, to further structural and functional injury. Restated, the glaucomatous damage to the optic nerve occurs before and after DH and appears to be independent of it. Eyes with slower rates of change may develop fewer DHs, because the degeneration of rim tissue is correspondingly slower and produces less frequent vascular rupture. The rapid structural and functional progression after DH continues at the same or adjacent location after its resolution and provides evidence that the microvascular disruption that results in bleeding is the result of a structural degeneration rather than a cause of glaucoma progression in and of itself, as there were very few eyes that developed a DH without evidence of prior, rapid, corresponding, localized functional injury. In addition, a rapid, localized rate of progression was a good predictor of the location of a future DH and continued field loss in glaucomatous eyes. Lastly, our results demonstrate that a DH should be viewed not only as a risk factor for future progression, but also as evidence and confirmation of past, localized progression. Consideration should be given to the inclusion of DH as a structural endpoint consistent with glaucoma progression in future glaucoma clinical trials.
In conclusion, spatially consistent, localized VF loss precedes the onset of DH in glaucomatous eyes. DH most often occurs in the region of most rapid glaucoma progression and is associated with future sustained and accelerated functional loss in these eyes.