Abstract
Purpose.:
To investigate the topographic characteristics of optic disc hemorrhage (DH) and the factors associated with DH area in POAG.
Methods.:
We enrolled 128 POAG eyes (with DH) of 128 patients consecutively. Digital red-free retinal nerve fiber layer (RNFL) photographs and digital color stereo disc photographs of the enrolled eyes were reviewed, and the DH locations were measured. The DH area was calculated based on RNFL deviation map/RNFL photographs overlay image.
Results.:
Disc hemorrhages were most common in the inferotemporal inferior sector (58.0%) and the disc rim (40.6%) in terms of octant and proximal location, respectively. In a multivariate analysis, a larger DH area was associated with older age (P < 0.01), use of acetylsalicylic acid (P = 0.03), lower cup-to-disc ratio (P < 0.01), and lower baseline IOP (P = 0.01). In normal-baseline IOP-POAG eyes (n = 88), the overall DH area and length of maximum radial extent of DH were larger than in high-baseline IOP-POAG eyes (n = 40) (P = 0.04 and 0.03, respectively).
Conclusions.:
In POAG eyes, DH was larger in area and longer in length in cases of normal-baseline IOP than in cases of high-baseline IOP. This suggests the possibility that previous studies' findings of higher DH prevalence and incidence in normal-baseline IOP-POAG eyes were partially affected by these topographic characteristics, which make DH more easily detectable.
Optic disc hemorrhage (DH) has been confirmed as an important risk factor for glaucoma progression in numerous studies.
1–5 Recently, the association between DH and glaucoma progression was more strongly confirmed by adding the progression rate and its relation to the IOP-lowering effect.
6 The associations between DH and systemic diseases, glaucoma severity, and eye-related factors have yet to be fully elucidated, despite their central significance to glaucoma and its progression. The association between DH and IOP is particularly controversial. Adael et al.
7 reported that IOP at first DH detection was lower than the mean IOP of the three most recent visits. Miyake et al.
8 found that in POAG, after trabeculectomy, normal-baseline IOP POAG (normal-tension glaucoma [NTG]) showed decreased incidence of DH, but still higher incidences compared with the high-baseline IOP POAG (high-tension glaucoma [HTG]). However, most studies have demonstrated that DH development is not consistently related to mean IOP level.
9–12
Although the authors of the relevant previous studies have not concurred in their findings on the association of DH with IOP, it is interesting, with respect to reported DH prevalences and incidences, that NTG patients have always shown higher frequencies of DH than their HTG counterparts (1.89- to 5.00-fold in prevalence, 1.30- to 2.78-fold in incidence).
8,10,11,13–16 This suggests that there might be differences in DH characteristics relative to baseline IOP. Jonas et al.
14 had introduced the possibility that a high IOP stops bleeding earlier, thus, resulting in a small DH in eyes with HTG, and that, contrastingly, a low IOP leads to a large DH in eyes with NTG. Considering both the previous report and the aggregate clinical data, we were able to hypothesize that the DH area is larger in POAG patients with a lower-baseline IOP, and that this fact could influence the detection rate of DH because a larger DH can be detected more easily and may persist longer.
Therefore, the aims of the present study were to evaluate the topographic characteristics of DH, including area, extent, and location, and to identify patient-related and eye-related factors associated with DH area in POAG eyes.
This study followed the tenets of the Declaration of Helsinki and was approved by the institutional review board of Seoul National University Hospital. We performed a retrospective review of POAG patients' electronic medical records, which had been compiled by the author (KHP) in Glaucoma Service of Seoul National University Hospital between January 2004 and August 2012.
All of the patients underwent a comprehensive glaucoma evaluation, including a central 30-2 threshold test of the Humphrey visual field (HVF, HFA II; Humphrey Instruments, Inc., San Leandro, CA) and retinal nerve fiber layer (RNFL) and optic nerve head imaging by Cirrus spectral domain-optic coherence tomography SD-OCT; Carl Zeiss Meditec, Dublin, CA). The baseline IOP was measured, on three different visits without medication, by Goldmann applanation tonometry (GAT; Haag-Streit, Koniz, Switzerland). A digital color stereo disc photograph (SDP) and a digital red-free retinal nerve fiber layer photograph (RNFLP) were obtained for each patient. All of the images were saved in a digital 1600 × 1216-pixel format.
Patients were deemed to have POAG if they met the following criteria: (1) presence of typical glaucomatous optic neuropathy with compatible visual-field loss, and (2) normal anterior chamber angles. Glaucomatous visual-field loss was defined as (1) a visual field with at least three adjacent test points having a deviation of equal to or greater than 5 dB and one test point having a deviation of more than 10 dB; (2) at least two adjacent test points with a deviation equal to or greater than 10 dB; (3) at least three adjacent test points with a deviation equal to or greater than 5-dB abutting the nasal horizontal meridian; or (4) a mean visual field defect of more than 2 dB. The rates of false-positive and false-negative answers each had to be equal to or less than 15%.
17 Patients were excluded from analyses if any of the following was true: (1) the presence of a secondary cause of glaucomatous optic neuropathy; (2) a history of intraocular surgery or retinal laser photocoagulation; (3) poor quality SDP or RNFLP that would render DH identification difficult; (4) the presence of high (< −6.0 diopters [D]) myopia; or (5) referral from other clinics. According to this latter criterion, only patients who developed DH during interval testing were included in the study. Disc hemorrhage was considered to be unrelated to glaucoma if (1) the disc was swollen or otherwise obviously abnormal due to nonglaucomatous optic neuropathy; (2) there were multiple nearby retinal hemorrhages suggestive of diabetic retinopathy or retinal vascular abnormality
3 ; or (3) there was acute posterior vitreous detachment that could have caused DH. Recurrent DH was defined as the condition under which another DH developed in a different location after detection of the initial DH.
Throughout the study, DH evaluation, based on SDP and RNFLP, were performed independently by two ophthalmologists (YKK, KHP). Discrepancies between the two observers' findings were resolved by consensus. The DH positions were defined, with reference to the RNFLPs, according to the proximal location
18 or the octant location. For the purpose of calculating the duration of DH, the start and end points were defined, respectively, as follows: (1) the initial detection of DH, and (2) the midpoint between the final detection of DH and the first observation of its disappearance.
For the purposes of angle analysis, a reference line (R) was set, as drawn from the center of the optic disc to the center of the macula on RNFLP.
19 Angular extent of DH was defined as the circumferential angle formed by two lines (r and r′) drawn from the disc center to the two points farthest apart from each other circumferentially. Angular location of DH was defined as the angle between the reference line and the median line (R′) between the two lines that had been drawn to measure the angular extent. Angular location for the right eye was measured counterclockwise from the reference line, and for the left eye, clockwise
19 (
Fig. 1B).
As the SD-OCT–computed value of optic disc area showed a discrepancy relative to the actual optic disc area, several corrections were needed.
20 The corrected optic disc area was calculated in consideration of the magnification factors related to the SD-OCT camera and the eye by substituting the subject's axial length (AL).
21–23 Thus, the formula for optic disc area correction was
An SD-OCT RNFL deviation map overlay onto RNFLP, which were aligned in Photoshop software (Version 9.0; Adobe, San Jose, CA) based on vascular landmarks, was employed to standardize the optic disc boundary (
Figs. 1C,
1D). Then, a method of comparing the optic disc area and DH-area pixel numbers, with ImageJ software (version 1.45s; National Institutes of Health, Bethesda, MD), was used to estimate the corrected DH area, which was calculated by the formula
The straight-distances from the optic disc center to the point of maximum radial extent of DH and to the RNFL calculation circle (3.46-mm diameter; provided by SD-OCT RNFL deviation map;
Fig. 1C) were measured. While compensating for the magnification factor,
22 the corrected length of maximum radial extent (LMRE) of DH was calculated by the formula
Four concentric circles were set as the standards for the proximal location of DH. Specifically, they were defined, from the innermost circle to the outermost, as four types of DH starting point: cup wall, cup margin, mid disc rim, and disc margin.
18 On a DH diagram, each DH started from one of the proximal location of DH, with a directionality heading to the outer side. The DH was conceived as a cone shape starting from the center of the disc, with the inner part of the proximal location of DH cut out. The differences in the number of subjects among the three diagrams (POAG, NTG, and HTG) were corrected by weighting the color intensities.
The correlations between the corrected DH area and the IOP were analyzed using Pearson's coefficient of correlation. Fisher's exact test assessed differences in the proximal and octant locations of DH, and the independent-samples t-test assessed differences in the angular extent of DH, corrected DH area, and DH area/disc area ratio. Univariate and multivariate linear regression analyses with forward stepwise selection were used to evaluate the factors associated with corrected DH area. A P value less than 0.05 was considered significant. All statistical analyses were performed using SPSS software version 12.0 (SPSS, Inc., Chicago, IL).
One hundred sixty-two POAG eyes with DH, representing 162 patients, were initially included in the study. Of these, 34 patients were excluded: (1) 19 patients lacked data such as the SD-OCT or AL, (2) nine patients had a poor quality SDP or RNFLP, and (3) eight patients could not match the retinal vessel on the RNFL deviation map and RNFLP. The remaining 128 eyes of 128 patients were examined further in the ensuing analysis.
The study population's clinical characteristics are listed in
Table 1. The mean age on first visit was 59.2 ± 10.8 years. The mean follow-up was 6.0 ± 2.9 years at intervals of 4.5 ± 3.3 months; DH was detected 1.5 ± 0.7 times over the course of 3.4 ± 2.9 months. The most common proximal location of DH was the disc rim (52/128), and the largest number of DHs were detected in the inferotemporal inferior sector (74.3/128). The overall angular extent of DH was 12.02 ± 7.13°, and the overall DH area and DH area/disc area ratio were 0.072 ± 0.04 mm
2 and 0.038 ± 0.026, respectively.
Univariate and stepwise multivariate linear regression analyses were carried out to determine which variables were associated with corrected DH area. A larger corrected DH area was associated with older age (
P < 0.01, standardized
β = 0.29), use of acetylsalicylic acid (ASA;
P = 0.03, standardized
β = 0.18), lower baseline IOP (
P = 0.01, standardized
β = −0.19), and lower cup-to-disc ratio at time of DH detection (
P < 0.01, standardized
β = −0.31) (
Table 2). When the effect of IOP on corrected DH area was separately analyzed, DH area was found to be negatively correlated with the baseline IOP (
R = −0.20,
P = 0.04) (
Fig. 2).
For further evaluation, 88 eyes of 88 patients were assigned to the NTG group (baseline IOP ≤ 21 mm Hg) and 40 eyes of 40 patients to the HTG group (baseline IOP ≥ 22 mm Hg). As comparing clinical characteristics between the two groups, they did not differ in systemic parameters including age (P = 0.52), sex (P = 0.06), diabetes mellitus (P = 0.78), systemic hypertension (P = 0.29), or use of ASA (P = 0.70). On average, the NTG group had 5.9 ± 3.1 years of follow-up at intervals of 4.3 ± 3.1 months; DH was detected 1.5 ± 0.8 times/eye, and lasted for 3.8 ± 3.1 months. The HTG group had 6.1 ± 2.9 years of follow-up at intervals of 4.8 ± 2.3 months; DH was detected 1.4 ± 0.7 times/eye, and lasted for 3.0 ± 2.6 months. There were no significant differences between the two groups in the total follow-up periods (P = 0.42), follow-up intervals (P = 0.68), SDP- and RNFLP-intervals (P = 0.60), number of DH detections (P = 0.49), or duration of DH (P = 0.18). In the baseline phase, there was neither any significant difference in the ocular parameters, including best-corrected visual acuity (P = 0.27), mean deviation of HVF (P = 0.47), pattern standard deviation (PSD) of HVF (P = 0.58), spherical equivalent (P = 0.39), central corneal thickness (P = 0.23), and corrected optic disc area (P = 0.57).
The proximal location, octant location, and angular extent of DH did not show significant differences between the two groups (
P = 0.32, 0.45, and 0.20, respectively) (
Table 3). However, the NTG group had a significantly larger corrected DH area (0.078 ± 0.050 mm
2 vs. 0.060 ± 0.036 mm
2,
P = 0.04) and a larger corrected LMRE of DH (1.162 ± 0.324 mm vs. 0.972 ± 0.251 mm,
P = 0.03). Overall, the DH area/disc area ratio was also larger in the NTG group (0.041 ± 0.028 vs. 0.032 ± 0.022,
P = 0.07) (
Table 4).
On the DH diagrams, the majority of first-detected DHs were located in the inferotemporal inferior sector and started from the mid disc rim. The NTG group showed a more widespread distribution of DHs, which is to say, a distribution farther from the optic disc center, than did the HTG group (
Fig. 3).
This study demonstrated both that DH area is negatively correlated with the baseline IOP and that the NTG group evaluated had a larger DH area than the HTG group.
In the present study, DH recurred in 41.5% of subjects, and the overall number of DH detections per subject was 1.5 ± 0.7. Previous studies have been reported DH recurrence rates ranging from 12% to 73% of DH patients.
9–11,24–26 The broad variance in these results is largely attributable to variances in follow-up periods, the diagnostic parameters employed, and study populations. Disc hemorrhage is detectable only within a certain time window: Kitazawa et al.
10 reported that 92% could be found between 4 weeks and 2 months after first presentation. In the present study, the mean duration of DH was 3.4 ± 2.9 months.
Among the other results of this study, the largest number of DHs was detected in the inferotemporal inferior sector, which finding corresponds with those of previous studies describing DH as most commonly located inferotemporally, followed by superotemporally, and usually near the poles.
25,27–30 The most common proximal location of DH in the present study was the disc rim, followed by, in order, the cup margin, the disc margin, and the cup wall. This result was contrary to that of a previous study in which cup margin was the most common proximal location of DH in nonmyopic DHs.
18 It is possible that the papillary part of DH is resorbed more rapidly than that outside the disc boundary,
30 which would explain why peripapillary retina–type DH is more frequently detected in cross-sectional studies. However, in the present, long-term follow-up study, more DHs showed proximal, the disc rim or the cup margin locations.
This study used the corrected optic disc area and the DH area/disc area ratio to determine the actual DH area. The mean corrected optic disc area (2.01 ± 0.63) was similar to the range of values reported in a previous study (1.87 ± 0.46 to 2.02 ± 0.51 mm
2).
20 Further, a novel, SD-OCT RNFL map/RNFLP overlay method was applied to the measurement of the DH area/disc area ratio. This method minimized the risk of inconsistency between the pixel numbers derived by RNFLP-based optic disc area and SD-OCT–computed optic disc area, respectively, by applying the standardized optic disc boundary.
This study found that the NTG group had a significantly larger DH area and length of maximum radial extent of DH than the HTG group. The larger DH area in NTG group can be explained in two ways. First, NTG, having a relatively lower IOP, can produce a larger area of DH owing to the lesser tamponade effect. That is, a high IOP stops bleeding earlier; thus, resulting in a small DH in eyes with HTG, and that, contrastingly, a low IOP leads to a large DH in eyes with NTG.
14 Second, NTG may have a relatively more vulnerable blood vessel than HTG even at a lower IOP.
The association between NTG and longer DH length seemed to be related to the difference in the characteristics of RNFL defects. Generally, RNFL defects are more localized in NTG.
31,32 In NTG patients, this allows DH to migrate more easily along the boundary of the localized RNFL defect and thus to be longer in length.
As for angular extent of DH, the mean value was smaller in the NTG group (11.32 ± 7.36°) than in the HTG group (13.07 ± 6.18°), though without statistical significance, while DH area was significantly larger in the NTG group. We could predict, therefore, that DH would be longer in length in the NTG group than in the HTG group: This was in fact confirmed by the result for the corrected length of maximum radial extent of DH, which was longer in the NTG group by approximately 19.5% (1.162 ± 0.324 mm vs. 0.972 ± 0.251 mm) (
Table 4). Moreover, in the NTG group, the duration of DH was longer than in the HTG group by roughly 0.8 months (3.8 ± 3.1 months vs. 3.0 ± 2.6 months). All of these results, namely the larger DH area, the longer DH length, and the longer DH duration, made DH more detectable in the NTG group.
In the present study, a DH diagram was developed for provision of information on DH clock-hour location, proximal location and relative angular extent. The diagram, expressing varied color intensity according to DH frequency, whereby higher intensity indicates higher frequency, makes possible visualization of numeric data and, thereby, enhanced understanding of DH distribution and accumulative frequency in POAG (
Fig. 3).
It should be noted that the present study has some limitations. First, the patients were recruited from a single tertiary referral hospital (cases of DH referred from other clinics were excluded). That is, only patients who developed DH during interval testing were included (which minimized the possibility of selection bias). Second, this study proceeded retrospectively. Third, systemic vascular disorders or blood disorders, which are associated with the occurrence of DH, were not compared in their clinical characteristics between the NTG and HTG groups. Fourth, the manual method of image alignment employed might incur error.
In conclusion, DH was larger in area and longer in length in the NTG group than in the HTG group, which suggests the possibility that previous studies' findings of higher DH prevalence and incidence in the NTG were partially affected by these topographic characteristic of DH. Additionally, it was found that the DH area in POAG had an inverse relationship with the baseline IOP. These results suggest that NTG and HTG form a continuum, that they are not separate entities. Clarification of this relationship and its mechanisms will require a long-term prospective follow-up study with a larger patient cohort.
Disclosure: Y.K. Kim, None; K.H. Park, None; B.W. Yoo, None; H.C. Kim, None