Abstract
Purpose.:
To assess the amount of structural loss (retinal nerve fiber layer [RNFL] thickness loss, macular thickness [MT] and volume [MV] measured by optical coherence tomography [OCT]) and functional loss (visual acuity [VA], visual field mean deviation [MD], brightness sensitivity, and red perception) necessary for a relative afferent pupillary defect (RAPD) to manifest in patients with glaucoma.
Methods.:
In this case–control study, 50 glaucoma patients were prospectively enrolled: 25 with RAPD and 25 without. The presence of an RAPD was determined and quantified using the swinging-flashlight test, with neutral-density filters. A separate examiner, masked to the pupillary findings, assessed participants for brightness sense, red perception, VA, MD, RNFL thickness, MT, and MV.
Results.:
Differences in RNFL thickness (P < 0.0001), brightness sense (P = 0.0007), red perception (P = 0.030), and MD (P < 0.0001) were found between control and RAPD patients, but not in visual acuity or macular OCT parameters. An absolute difference in RNFL thickness of 14.6 μm or greater, intereye difference of 9.5 dB or greater, and brightness of less than 64% in the weaker eye, were all associated with 100% specificity of RAPD presence. When RNFL thickness was reduced to 83% of the less advanced eye, the sensitivity and specificity of RAPD presence were 72% (95% confidence interval [CI], 0.51–0.88) and 100% (95% CI, 0.86–1.00), respectively.
Conclusions.:
An RAPD was clinically detected in all participants in whom RNFL thickness decreased to 83% of that in the less advanced eye. Subjective brightness is the most accurate clinical surrogate for detecting an RAPD in patients with primary open-angle glaucoma.
Glaucoma is characterized by progressive loss of retinal ganglion cells and their optic nerve axons, with resultant visual field defects. Although glaucoma is usually a bilateral disease, structural and functional loss can be asymmetric between eyes. The relative afferent pupillary defect (RAPD) is an important objective sign of unilateral or asymmetric bilateral impairment of the anterior afferent visual pathways. Once a certain level of asymmetric neuronal damage and function is reached in glaucoma, an RAPD theoretically should be clinically detectable.
Glaucoma provides a useful model for evaluating the relationship between the amount of structural and functional loss necessary to identify an RAPD because, unlike other optic neuropathies (such as optic neuritis and compressive optic neuropathies), the damage is irreversible. Many studies have found multiple variables that show intereye asymmetry in glaucoma patients, including contrast sensitivity,
1,2 intraocular pressure,
3,4 visual field sensitivities,
5,6 visual evoked potentials,
7 and retinal nerve fiber layer (RNFL) thickness.
4 However, the amount of structural and functional loss that occurs in patients with glaucoma before an RAPD is clinically detectable must be investigated further.
The purpose of this study was to investigate the degree of structural loss of RNFL thickness, macular thickness, and macular volume, as measured by optical coherence tomography (OCT, Carl Zeiss Meditec, Dublin, CA); of functional loss of visual field sensitivity, as measured by automated perimetry (Humphrey Field Analyzer; SITA Standard 24-2; Carl Zeiss Meditec); and of the brightness sensitivity and red perception that occur before an RAPD is detectable.
All participants were assessed at the Eye Institute (Auckland, New Zealand). Patients with primary open angles were eligible to participate as cases if they had an RAPD, no evidence of other posterior segment, retinal or macular disease, media opacity causing suboptimal OCT scans, anisocoria greater than 0.5 mm, congenital dyschromatopsia, or visual acuity worse than 6/12 in the unaffected eye. Patients (without the exclusion criteria) were eligible to participate as control subjects if they had primary open angles without RAPD. Twenty-five consecutive patients who met the criteria with an RAPD and 25 consecutive patients who met the control criteria were enrolled.
The patients underwent full clinical ophthalmic evaluation including Snellen visual acuity testing converted to logMAR; automated perimetry (Humphrey Field Analyzer, SITA-standard, stimulus size III; Carl Zeiss Meditec); intraocular pressure measurement; slit lamp biomicroscopy of the anterior segment; funduscopy; assessment of pupillary function, brightness sense, and red perception; and OCT.
They also underwent pupillary assessment with the swinging-flashlight test and grading of the RAPD with neutral-density filters by an experienced neuro-ophthalmologist (HDM) masked to the results of the subjective assessment of optic nerve function and OCT. The validity of RAPD grading by a single observer (HDM) has been established.
8 Brightness sense, red perception, and OCT results were assessed by a separate examiner (SSLC) masked to the results of pupillary testing. All tests, except OCT, were performed through undilated pupils, and all were performed on the same day.
The swinging-flashlight test was performed in a darkened room, and the RAPD was quantified with neutral-density filters, as described by Thompson et al.
9 The light source was an indirect ophthalmoscope (Welch-Allyn, Skaneateles, NY) set to 6 V (maximum brightness) and positioned approximately 30 cm away from patient's eyes. The light source was moved back and forth between the eyes, to examine the pupillary response. The depth of the RAPD was quantified by successively placing neutral-density filters in front of the eye without the RAPD, to decrease the light stimulus by a known amount until the pupillary constriction between the two eyes was symmetric. The density of filter necessary to balance the pupillary responses was recorded in log units. Measurements of RAPD in this study were in steps of 0.3 log units.
Tests for subjective assessment of optic nerve function were performed in the following sequence: visual field mean deviations (MDs), red perception, and brightness sense. Visual field mean deviations were calculated by automated perimetry (Humphrey Field Analyser II; [HFA] set to 24-2 analysis, SITA-standard protocol, stimulus size III, Carl Zeiss Meditec). Right and left eye perimetry test order was randomized. Visual field results were excluded if they were unreliable after repeated attempts (reliability indices used were from the manufacturer's manual and the Bickler-Bluth et al.
10 guidelines of >33% fixation losses, false positives, or false negatives). Red perception was assessed with a red top on a tropicamide eye drop bottle (∼20 mm in diameter). The stimulus was presented to each eye separately, approximately 30 cm from the eye tested, and was placed in the center of the patient's visual axis. The patient was asked to look directly at the tip of the bottle top and report whether the bottle top was equally red in both eyes. An affirmative answer ended the test, and the result was recorded as 100% in both eyes. If the patient indicated that the redness of the bottle top differed between the eyes then the stimulus was re-presented to the eye with normal red perception and the patient was asked, “If the top of the bottle is 100% red in this eye, then what is the percentage of redness in the other eye.”
Brightness sense was assessed with the same technique as was used for red perception. An indirect ophthalmoscope set at 6 V was positioned in the center of the visual axis of one eye. The patient was asked to fixate on the light, and the light source was then swung to the other eye for an equal amount of time. The patient was asked whether the light was of equal brightness in both eyes. If not, the same steps as were used for red perception were repeated. With this technique the intraobserver variability was negligible at 2.5% when the test was repeated twice.
OCT was conducted after subjective tests of optic nerve function (Stratus OCT; Carl Zeiss Meditec). Three fast RNFL scans and three fast macular scans were performed on each eye. The results of three scans were averaged to limit the amount of intraobserver variation, which is usually low with OCT.
11 Pupil dilation with tropicamide eye drops was used if pupil size limited scan quality. Scans that had a signal strength less than seven or had low analysis confidence were repeated until three optimal scans were obtained. If adequate-quality scans could not be obtained, the patient was excluded from the study.
Continuous normally distributed variables were compared by using Student's t-test and the non-normally distributed variables were compared by using Wilcoxon test to determine whether the assessed parameters differed between the RAPD and control groups. Parameters included intereye differences in brightness sense, red perception, best corrected visual acuity, MD, RNFL thickness, macular thickness, and macular volume. All parameters in the RAPD group were recorded by subtracting the data of the affected eye from those of the unaffected one. All parameters in the control group were recorded in absolute values after the right eye data were subtracted from the left eye data.
Receiver operating characteristic (ROC) curves were constructed to determine the accuracy of the intereye differences that were significantly different by Student's t- or Wilcoxon tests. Sensitivities, specificities, likelihood ratios, and odds ratios were calculated for various cutoffs for each significant parameter. In an analysis by severity of RAPD score, Spearman correlation coefficients were calculated for measured parameters against RAPD score. A stratified analysis using multivariate models (logistic regression) was then performed for predictors of worse severity. All tests were two-tailed and P < 0.05 was statistically significant (SAS, ver. 9.1; SAS Institute, Cary, NC).
The research complied with the Declaration of Helsinki, and informed consent was obtained from all participants.
The mean age of the 25 RAPD patients (64.0 years; SD, 18.5 years) was similar to that of the 25 control subjects (58.0 years; SD, 13.1 years, P = 0.19). Of the RAPD group (n = 11) and control group (n = 11), 44% were women (P = 0.99). Mean RAPD severity in the RAPD group was 0.46 log units (SD, 0.21; range, 0.3–0.9). Fifteen patients had a 0.3-log-unit RAPD, seven had a 0.6-log-unit RAPD, and three had a 0.9-log-unit RAPD.
Mean RNFL thickness in both eyes in the RAPD group (67.8 μm; SD, 10.7; range, 50.3–86.6) was significantly thinner (P < 0.0001) than in the control group (92.3 μm; SD, 15.3; range, 51.5–112.8). The mean intereye difference in RNFL thickness was significantly larger (P < 0.0001) in the RAPD group (17.8 μm; SD, 9.3; range, 3.6–42.4) than in the control group (5.1 μm; SD, 3.5; range, 0.0–14.2). Intereye differences in logMAR visual acuity (P = 0.09), macular thickness (P = 0.46), and macular volume (P = 0.19) were not significantly different between the RAPD and control groups.
The functional parameters were significantly different between the RAPD and control groups, with intereye differences in red perception (
P = 0.03), brightness sense (
P = 0.0007), and MD (
P < 0.0001).
Table 1 shows a comparison of the various parameters between RAPD and control groups.
Table 1. Parameters Assessed in RAPD and Control Groups
Table 1. Parameters Assessed in RAPD and Control Groups
Average Intereye Difference | RAPD (n = 25) | Control (n = 25) | P |
RNFL thickness, μm | 17.83 ± 1.858 | 5.071 ± 0.7079 | <0.0001 |
Inferior | 26.97 ± 3.678 | 6.907 ± 1.023 | <0.0001 |
Superior | 20.66 ± 3.519 | 12.28 ± 1.586 | 0.0349 |
Nasal | 10.87 ± 3.090 | 9.133 ± 1.209 | 0.6038 |
Temporal | 13.24 ± 2.970 | 6.547 ± 1.121 | 0.0404 |
Macular thickness, μm | 6.836 ± 4.214 | 3.660 ± 0.5404 | 0.4583 |
Macular volume, mm3 | 0.2675 ± 0.1150 | 0.1148 ± 0.01488 | 0.1943 |
Red perception, % of better eye | 89.52 ± 4.492 | 99.60 ± 0.4000 | 0.0301 |
Brightness sense, % of better eye | 77.92 ± 5.011 | 97.08 ± 1.671 | 0.0007 |
Visual acuity, logMAR | 0.09867 ± 0.05050 | 0.01163 ± 0.006428 | 0.0938 |
Mean deviation, dB | 8.615 ± 1.070 | 1.334 ± 0.3633 | <0.0001 |
Figure 1 shows the distribution of intereye differences in RNFL thickness, brightness sense, and MD in the RAPD and control groups. Intereye differences in quadrant RNFL thickness were also significantly different between RAPD and control groups for the inferior (
P < 0.0001), superior (
P = 0.03), and temporal (
P = 0.04), but not nasal (
P = 0.60) quadrants.
ROC curves were used to assess the accuracy of the different parameters as predictors of RAPD. Significant values for the area under the ROC (AUROC) curve were 0.94 (95% confidence interval [CI], 0.87–1.00, P < 0.0001) for intereye difference in RNFL thickness, 0.76 (95% CI, 0.63–0.90, P = 0.001) for difference in brightness sense, and 0.92 (95% CI, 0.83–1.02, P < 0.0001) for difference in MD. Looking again at quadrant RNFL thickness, inferior was the most predictive (AUROC, 0.88; 95% CI, 0.78–0.99; P < 0.0001), followed by superior (AUROC, 0.66; 95% CI, 0.50–0.82; P = 0.05). The AUROCs of the nasal (P = 0.98) and temporal (P = 0.12) quadrants and red perception (P = 0.13) were not statistically significant.
Table 2 shows the range of accuracy for intereye differences in brightness sense, MD, absolute RNFL thickness, and percentage change in RNFL thickness. When the intereye difference in brightness sense was greater than 36%, there was 100% (95% CI, 86.3%–100%) specificity for the presence of RAPD and 32% (95% CI, 15.0%–53.5%) sensitivity, with an odds ratio (OR) of 13.0 (95% CI, 1.5–111.8), meaning that those who had a difference in brightness sense of greater than 36% had 13 times the odds of having an RAPD than did those who had a difference in brightness sense of less than 36%. When the intereye difference in MD was greater than 9.5 dB, there was 100% (95% CI, 86.3%–100%) specificity for the presence of RAPD and 36% (95% CI, 18.0%–57.5%) sensitivity, with an OR of 15.2 (95% CI, 1.8–130.6). When the intereye difference in RNFL thickness was greater than 14.6 μm, there was 100% (95% CI, 86.3%–100%) specificity for the presence of RAPD and 60% (95% CI, 38.7%–78.9%) sensitivity, with an OR of 24.4 (95% CI, 3.9–149.1).
Table 2. Accuracy Parameters for Brightness Sense, Difference in Mean Deviation, Difference in RNFL Thickness, and Percentage RNFL Thickness, as Predictors for RAPD in Patients with Glaucoma
Table 2. Accuracy Parameters for Brightness Sense, Difference in Mean Deviation, Difference in RNFL Thickness, and Percentage RNFL Thickness, as Predictors for RAPD in Patients with Glaucoma
Cutoff | Sensitivity | Specificity | Likelihood Ratio (+ve test) | OR |
Brightness sense, % of unaffected eye | | | | |
<95 | 60 (39–79) | 88 (69–97) | 5.0 (1.6–15.2) | 11.0 (2.6–46.8) |
<80 | 32 (15–54) | 96 (80–99) | 8.0 (1.1–59.3) | 11.3 (1.3–98.9) |
<65 | 33 (17–54) | 96 (81–99) | 9.0 (1.2–66.2) | 13.0 (1.5–111.8) |
Difference in mean deviation, dB | | | | |
>3.5 | 88 (69–97) | 96 (80–99) | 22.0 (3.2–150.9) | 176.0 (17.0–1819.7) |
>6.5 | 64 (42–82) | 96 (80–99) | 16.0 (2.3–111.7) | 42.7 (4.9–370.2) |
>9.5 | 38 (20–59) | 96 (81–99) | 10.4 (1.4–75.5) | 16.3 (1.9–139.2) |
Difference in RNFL thickness, μm | | | | |
>5 | 96 (80–99) | 56 (35–76) | 2.2 (1.4–3.4) | 30.5 (3.6–262.4) |
>10 | 80 (59–93) | 96 (80–99) | 20 (2.9–137.8) | 96.0 (10.3–890.6) |
>15 | 59 (39–78) | 96 (81–99) | 16.0 (2.3–112.3) | 37.8 (4.5–321.4) |
RNFL thickness thinner/thicker (%) | | | | |
<90 | 92 (74–99) | 88 (69–97) | 7.7 (2.6–22.3) | 84.3 (12.8–553.9) |
<85 | 72 (51–88) | 96 (80–99) | 18.0 (2.6–124.7) | 61.7 (7.0–547.4) |
<80 | 59 (39–78) | 96 (81–99) | 16.0 (2.3–112.3) | 37.8 (4.5–321.4) |
Assessment of the intereye percentage decrease in RNFL thickness by ROC curves produced an even greater AUROC of 0.97 (95% CI, 0.93–1.01;
P < 0.0001).
Figure 2 shows the ROC curves for intereye percentage decrease in RNFL and the intereye difference in MD as predictors of the presence of RAPD. When the RNFL thickness in one eye declined to 83.3% of the other eye, there was 100% (95% CI, 86.3%–100%) specificity for the presence of RAPD in the thinner eye and 72% (95% CI, 50.6%–87.9%) sensitivity, with an OR of 40.5 (95% CI 6.5–253.1). Perhaps the best compromise between acceptable sensitivity and specificity in our cohort was when RNFL thickness decreased to 89.0% of the other eye, with a sensitivity and specificity for the presence of RAPD of 92% (95% CI, 74.0%–99.0%) and 92% (95% CI, 74.0%–99.0%), respectively.
Intereye differences in brightness sense, MD, absolute RNFL thickness and percentage decline in RNFL thickness correlated significantly (Spearman, all P < 0.0001) with increasing RAPD severity (r = −0.51, r = +0.80, r = 0.72, and r = −0.79, respectively). When a stratified analysis using multivariate models (linear regression) predicting RAPD scores of 0.6 or 0.9 from 0 or 0.3 was performed, the difference in MD between eyes alone was a significant independent predictor (P = 0.034). The linear combination of these parameters offered no additionally discriminative ability above that of the individual parameters.
The present study is the first in which both structural and functional test results were evaluated and their association with the presence of an RAPD determined. The findings established that a decrease in RNFL thickness to 83% of that in the other eye, which equates to a 17% loss in RNFL thickness, produced an RAPD in the thinner eye that was detectable by using the swinging-flashlight test. The functional test that may serve as a clinical surrogate for the presence of an RAPD is the brightness sense differential between the two eyes. When sense of brightness in one eye was reported to be less than 64% of that in the other eye, the specificity of this test was 100% for the presence of an RAPD. An intereye difference in MD of greater than 9.5 dB was also 100% specific for RAPD presence in the more affected eye. However, macular thickness, macular volume, and red perception did not correlate strongly with the presence of an RAPD in glaucoma.
Our findings support the results of Tatsumi et al.
12 who estimated that RAPD detection occurs when RNFL thickness drops to 73% of that in the thicker eye. The difference between our results and theirs may be attributable to the use of RNFL data from patients who had an RAPD severity of 0.6 log units or worse in the Tatsumi study, whereas our minimum cutoff for an RAPD detected by neutral-density filters was 0.3 log units. In addition, we used a control group of glaucoma patients, whereas Tatsumi et al. performed a linear regression analysis and extrapolated the line backward until a percentage could be calculated for a hypothetical subject with RAPD severity of 0 log units.
Of the functional tests, the intereye difference in MD was most predictive of the presence of RAPD, with an AUROC of 0.92. Those participants with an intereye difference of greater than 9.5 dB had a definite RAPD presence in the more affected eye. This result compared favorably with the 12 dB estimated by Tatsumi et al.,
12 who used a Humphrey Field Analyser set to 30-2 analysis, whereas Brown et al.
13 calculated a 13% sensitivity difference with an Octopus 2000 perimeter (Haag-Streit, Köniz, Switzerland). In terms of subjective tests of value in the clinical setting, brightness sense was a moderately predictive test with an AUROC of 0.76.
In our cohort, those who had a brightness sense of 64% or less compared with the other eye had definite RAPD presence. However, the sensitivity at this level is 32%. A brightness score of 95% would improve the sensitivity to 60% but the specificity would fall to 88%. This result is in contrast that in another study comparing nonglaucomatous optic neuropathies that reported a sensitivity of 99% and specificity of 95% when brightness sense was reported to be 90% or less of the other eye.
8 In that same study, no control patient reported a brightness sense lower than 85% compared with the fellow eye. In the present study among glaucoma patients, no control patient reported a brightness sense lower than 80% compared with the fellow eye. Red perception was not as useful in glaucoma as brightness sensitivity. Although the AUROC for brightness sense was 0.76 (95% CI, 0.63–0.90;
P = 0.001), the AUROC for red perception was not statistically significant as a discriminator for the presence of RAPD. This is uniquely different from other optic neuropathies in which both brightness sense and red perception are strongly discriminators for the presence of RAPD, with an AUROC of 0.99 (95% CI, 0.98–1.00;
P < 0.0001) and 0.93 (95% CI, 0.90–0.96;
P < 0.0001), respectively.
8
The poorer sensitivities and specificities for various predictors in glaucoma compared with other optic neuropathies
8 raises some interesting questions. One possibility is that there is a differing susceptibility of RGC cell populations to the glaucomatous process. The pupillary light reflex is mediated by the small proportion of non–image-forming (melanopsin) RGCs, whereas the visual RGC axons of the retino-geniculo-cortical pathway determine both perimetric sensitivity and most of the RNFL thickness.
14,15 The spatial distribution of these pupillomotor RGCs projecting to the pretectum is thought to be proportional to the distribution of visual RGCs projecting to the lateral geniculate nucleus.
16 The highest distribution is seen in the perifoveal region, especially temporally in the papillomacular bundle, decreasing with greater eccentricity.
15 For this reason, we also evaluated the macular OCT parameters of thickness and volume, but neither was significantly different between the RAPD and control groups. Similarly, the temporal RNFL quadrant thickness, corresponding to the papillomacular bundle, was not a significant predictor of RAPD presence. In fact, as expected, the inferior followed by the superior RNFL quadrant thicknesses were the most predictive of RAPD presence in this glaucoma cohort.
Although the selective susceptibility of melanopsin cells in glaucoma is a possibility, there is no current evidence to support this notion. In fact, despite the lower sensitivities and specificities in the present study, the ROC curves for intereye difference in RNFL thickness (AUROC = 0.94) and intereye difference in MD (AUROC = 0.92) suggest a strong relationship between the loss of RGCs of the retino-geniculo-cortical pathway and RAPD presence and thus, by inference, the loss of melanopsin RGCs. Furthermore, the magnitude of RAPD has been shown to correlate with the degree of visual field loss.
9,17 Another explanation for the difference in sensitivity and specificity between glaucoma and other optic neuropathies may well be the eccentricity of the visual field defects in glaucoma which may influence the perception of brightness less than the more centrally located defects.
In conclusion, an RAPD was always clinically detectable when RNFL thickness decreased to 83% of that in the thicker eye, or when the intereye difference in MD was 9.5 dB. A clinically useful surrogate for the presence of an RAPD in glaucoma patients is brightness sense, with an RAPD being present when brightness sense is 64% of that in the unaffected eye. However, red perception has been shown not to be clinically useful as an indicator of RAPD in patients with glaucoma.
Supported by unrestricted research funds provided by Allergan NZ.
Disclosure:
S.S.L. Chew, None;
W.J. Cunnningham, None;
G.D. Gamble, None;
H.V. Danesh-Meyer, None