Multifocal Electroretinograms in Patients with Branch Retinal Artery Occlusion

Abstract

purpose. To investigate the usefulness of second-order multifocal electroretinograms (MERGs) for detecting inner retinal disorders.

methods. The MERG from 5 patients with branch retinal artery occlusion (BRAO) was recorded. Twelve eyes of 12 normal subjects were also tested. MERGs were recorded using 61 hexagons. Bright flash ERGs were also recorded to measure the oscillatory potentials (OP). Root mean square (RMS) measures of the local first- and second-order MERGs (fMERG and sMERG) were compared in the affected and unaffected areas. The first negative trough (N1) and first positive peak (P1) were also used for measuring the amplitudes and latencies of the fMERG.

results. The fMERG RMS-amplitudes decreased significantly (r = 0.56, P < 0.05) in the affected area compared with normal values. The fMERG latencies of N1 and P1 increased significantly (P < 0.05) in the affected area. Furthermore, the sMERG RMS-amplitudes decreased almost to the noise level (r = 0.28, P < 0.001) in the affected areas. The interocular ratio of the sMERG RMS-amplitudes (affected/normal) significantly correlated with that of the fMERG (r = 0.69, P < 0.001). The fMERG latencies significantly correlated with the sMERG RMS-amplitude (r = 0.37 ∼ 0.69, P < 0.05 ∼ 0.001), but only began to increase after a 30% to 50% loss of the sMERG amplitude. The summed OP amplitude decreased to the same extent as the sMERG in the affected eye (0.5 of the normal eye).

conclusions. Although the fMERG amplitude and latency were significantly changed, the sMERG was much more affected by BRAO. The marked reduction of the sMERG in the affected area strongly suggested its main source was from the more inner layers of the retina compared to the fMERG. The sMERG appeared to be a sensitive indicator of inner retinal dysfunction.

Although the usefulness of the first-order MERG (fMERG) ¹ has been widely accepted, that of the second-order MERG (sMERG) is still unclear. By definition, as described before, ² the sMERG represents focal temporal interactions of the retina between two consecutive flashes and is related to photograph-adaptive mechanisms. There have been few reports clearly showing the significance and usefulness of the sMERG. The sMERG was reported to be useful for the detection of subclinical diabetic retinopathy. ² Moreover, it was suggested that the sMERG has an inner retinal contribution, ^{2 3 4 5 6} and the sMERG response is decreased in glaucoma patients. ^{4 5} On the other hand, in a recent study no significant difference was found between the glaucoma and normal groups. ⁷ Thus, limited progress has been made in understanding the relationship between ganglion cell loss and sMERG. 

In patients with branch retinal artery occlusion (BRAO), it is known that the inner two-thirds of the retina is damaged ⁸ ; the structures involved are the inner nuclear layer, the inner plexiform layer, the ganglion cells, and the nerve fiber layers. To obtain information on the properties and origin of the second-order component that might allow distinction between inner and outer retinal contributions, we recorded MERGs from patients with BRAO. In our preliminary report, ⁹ averaged sMERGs were decreased in the affected areas of BRAO. In this study, the signal-to-noise of each local response was increased by reducing the number of stimulated areas (61 hexagons) and by increasing the length of the recording time. Here we present evidence that inner retina other than ganglion cells contribute to the second order component. 

Methods

Subjects

We used MERG to evaluate five eyes of patients (patients 1 to 5) with BRAO (Table 1) . All patients had unilateral damage in the upper and lower halves of the visual field. The counterpart eyes had no abnormality on ophthalmologic examination. In the affected eye, the affected area was defined as that with a perimetric sensitivity value lower than 10 dB. It was verified using fluorescein angiography that the affected area was identical with the nonperfusion area of artery occlusion. The unaffected area was defined as that with a mean sensitivity value greater than 24 dB, and was positioned horizontally opposite the affected quadrant. 

Three BRAO patients had a nonperfusion area of retinal artery occlusion in the superior retina, and two had such an area in the inferior soon after the artery occlusions occurred. Acute findings such as retinal edema due to artery occlusion had already disappeared by the time of recording. 

Twelve eyes of 12 normal subjects were also tested as controls. Mean ages and standard deviations were 47.0 ± 14.6 years for normal subjects, and 51.4 ± 19.9 years for BRAO patients. The time from initial onset of BRAO to the date of recording the MERGs was 46.8 ± 16.1 months. 

All patients had clear media and had no other eye diseases. Their corrected visual acuities were greater than 20/20 except for one eye of a patient (patient 2) with BRAO (20/200). A cross-line was used as a fixation target, and good fixation was maintained during recordings in this patient. Patients who had previous eye surgery were excluded from the study. Pupils were maximally dilated to >8 mm with tropicamide during recording. Refractive errors were within ±2 diopters. 

Recordings

VERIS Science 3.0.1 (Electro Diagnostic Imaging Inc., San Mateo, CA) was used to record the MERGs. The stimulus matrix consisted of 61 scaled hexagonal elements that were displayed on a monochrome monitor (MD-B1700; Chu-ou musen, Tokyo, Japan) driven at a 75 Hz frame rate. At a viewing distance of 27 cm, the radius of the stimulus array subtended 20 × 20°. Each element was independently alternated between black (<5 cd/m²) and white (200 cd/m²) according to a binary m-sequence. ¹ The cornea was anesthetized with oxybuprocaine hydrochloride, and then ERGs were recorded with a bipolar contact lens electrode (Kyoto Contact Lens, Kyoto, Japan). A ground electrode was attached to the earlobe. Mean stimulus luminance was adjusted to 110 cd/m². An individual recording session consisted of sixteen 30 segments separated by rests. The m-sequence had 2¹⁵-1 elements and required approximately 12 minutes to obtain a set of MERGs. The signals were fed into an amplifier (GRASS, Model 12; Quincy, MA) with band-pass filtered at 3 to100 Hz. The artifact removal procedure of VERIS was not used. 

Subjects’ eyes were refracted to achieve the best visual acuity for the viewing distance, which was adjusted to compensate for changes in the retinal image size caused by the corrective lenses. Humphrey static perimetry (program 30-2) was also used to measure perimetric sensitivity. 

Analysis

We examined the fMERG and sMERG components by using VERIS Science 3.0.1 (Electro Diagnostic Imaging Inc., San Mateo, CA). The second-order component (first slice) is defined as the difference between the responses to two successive light flashes (26.6-ms interval). The second-order component was calculated for each location as the mean difference of the responses to the stimuli that were preceded by stimulus of the same polarity and responses that were preceded by a stimulus of different polarity. The first-order component was calculated as the focal mean difference of responses to a white flash and responses to a black stimulus. In the fMERG, the first negative trough was labeled N1, the first positive peak P1, and the second negative trough N2. The root mean square (RMS) amplitude of the MERG were measured between 10 and 60ms. The RMS was calculated as:  

\[{\{}(1/n){\times}{\Sigma}{[}R(t){]}^{2}{\}}^{0.5},\]

where R(t) is response amplitude at time t, and n is the number of samples in the time period. We also measured implicit times (N1, P1) and peak-to-trough amplitudes of N1 (from the mean amplitude of 0 to 5 ms to N1) and P1 (from N1 to P1) in the fMERG. Statistical differences were examined with the Mann–Whitney U test. To know the limits of our measurment of the sMERG, noise levels were calculated as the RMS amplitude of the local MERGs to no flickering stimuli (white, 100 cd/m²; black, 100 cd/m²), recorded from 2 normals (n = 98). 

Our investigation followed the tenets of the Declaration of Helsinki, and informed consent was obtained from all participants once the nature of the study had been clearly explained. 

Results

Sensitivity Mapping of the Humphrey Static Perimetry from BRAO Patients

Figure 1A shows gray-scale mapping of the Humphrey static perimetry (program 30–2). Two BRAO patients had a perimetric sensitivity loss mainly in the upper area: that is, upper nasal quadrant for patient 1, and upper half and central area for patient 2. Three patients had a perimetric sensitivity loss mainly in the lower half area. 

Three patients had a nonperfusion area of retinal occlusion in the superior retina, and two in the inferior retina. The macula of the patient 2 was also damaged by retinal artery occlusion. 

fMERGs and sMEGs from the BRAO Patients

Figures 1B and 1C show the mappings of the fMERGs and sMERGs from the five BRAO patients. The fMERG local response is composed of the first negative trough, N1, followed by the first major positive peak, P1. The second negative trough, N2, is small and somewhat difficult to measure. The sMERG local response is much smaller than the fMERG, and is composed of a number of peaks and troughs. The scale for sMERG in the figure is 5× that for the fMERG. Note that the fMERG local responses from the pathologic area are decreased compared with those from the unaffected area (Fig. 1B) . Although the perimetric sensitivity decreased to less than 5 dB, the fMERG responses from the area did not decrease to the noise level. On the contrary, almost normal response with clear peaks and troughs was observed, even from the area in which psychophysical data exhibit no perimetric sensitivity (Fig. 1B) . 

Surprisingly, the sMERG local responses from the pathologic area were totally extinguished to the noise level (Fig. 2C ) in all patients. It was almost impossible to measure any peaks or troughs of the sMERG from the affected area. The area and degree of the abnormal sMERG correlated well with those of the perimetric sensitivity. Thus, there are distinct discrepancies between the sMERG and fMERG in the affected area. 

Waveform Changes of the fMERG and sMERG in BRAO

Figure 2A shows 61 fMERG and sMERG responses from the normal eye of patient (Pt) 3. The locations of the U1 to 4 and L1 to 4 are also shown in the figure. Figure 2B shows the waveforms of the local MERGs from the affected eyes of patients 1 to 5. MERGs responses from the contralateral normal eye of Pt 5 are also shown (gray waveforms). The waveforms at the same latitude (U1 to 4 and L1 to 4) are superimposed (n = 2 to 4) in each of four quadrants. Nine responses on the central latitude and four responses on the central meridian are not included in the figure (Figs. 2A 2B) . Although perimetric sensitivity is near 0, relatively large fMERG responses can be seen in the affected areas. It is obvious that the fMERG latencies are increased in the affected area compared with those in the unaffected area. The sMERG responses cannot be detected from the pathologic quadrants (arrows in Fig. 2B ). It is easily recognized that the mapping of the sMERG correlates better with the perimetry than does that of the fMERG. 

RMS-Amplitudes of fMERG and sMERG from Affected and Unaffected Areas in All BRAO Patients

Naso-temporal asymmetry is seen in the fMERG. Mean RMS-amplitude values of local fMERGs from temporal areas (ST or IT) are significantly (P < 0.03 to 0.001) smaller than those from nasal areas (SN or IN). On the other hand, the sMERG amplitude for the nasal and temporal areas were not significantly different. 

RMS-amplitudes of both fMERG and sMERG from the unaffected area (n = 92) are slightly decreased (P < 0.05 ∼ 0.01) compared with those (n = 276) from normal eyes (Figs. 3A B ). The RMS amplitude of the fMERG from the affected area is decreased significantly (P < 0.01) compared with that from normal eyes (Fig. 3A) . Moreover, marked reduction (P < 0.001) is found in the RMS amplitude of the sMERG from affected area compared with that from normal eyes (Fig. 3B) . 

In the affected eyes, the mean value of the fMERG RMS-amplitudes from the affected area is approximately two-thirds of that from the unaffected area, and the difference is statistically significant (P < 0.01, n = 92). Moreover, the mean value of the sMERG RMS-amplitudes from the affected area is one-third of that from the unaffected area, and the difference is significant (P < 0.001, n = 92). The RMS-amplitude of the sMERG decreased almost to noise level. 

Interocular Ratio of the Affected to Unaffected Eye in the Amplitudes of the MERGs and of the OPs

We calculated the interocular ratio of the RMS-amplitude as (RMS-amplitude from affected eye)/(RMS-amplitude from unaffected eye) for each symmetrically positioned location (Fig. 3C) . 

Although the mean ratios are slightly decreased for the fMERG (0.7, n = 92) and sMERG (0.81, n = 92) in the unaffected area, there is no significant difference between these ratios. In the affected area, the ratio is 0.48 for the fMERG and 0.23 for the sMERG. The ratio for the sMERG is significantly decreased to the noise level (0.23, n = 92, P < 0.001). 

Figure 3D shows the mean interocular ratios in the amplitudes of the MERGs and OPs in all BRAO patients. To compare with the single response of the OP, all local responses of the MERGs were averaged using the VERIS grouping procedure (all trace, n = 61), and the ratios of affected to unaffected eye were calculated for the fMERG and sMERG. The amplitude of the OP wave was calculated as the algebraic sum of the 4 OPs as shown in the literature. ¹⁸ The mean interocular ratio of the sMERG is the same as that of the OP (0.5, n = 5) and smaller than that of the fMERG (0.7, n = 5). 

The fMERG Parameters in the BRAO Patients

In the normal controls, mean amplitude values of local fMERGs from the nasal areas (SN or IN) are significantly larger than those from the temporal areas (ST or IT) for N1 (P < 0.05 to 0.001) and for P1 (P < 0.003 to 0.001). Mean amplitude values were 332 to 435 nV for N1 and 647 to 852 nV for P1. In the affected area, mean amplitude values of local fMERGs are significantly decreased (P < 0.001) for N1 (231 nV) and for P1 (357 nV) compared with the normal values. In the unaffected area of the affected eye, the mean amplitude value of P1 was also significantly decreased (568 nV, P < 0.02) compared with that of normal controls. The fMERG amplitude ratios of P1/N1 in the affected area are significantly decreased (mean value 1.7, P < 0.01) compared with those in the normal eye (2.2). The fMERG latencies (N1, P1) are significantly increased (P < 0.01) in the affected area compared with those in the normal eye. 

In the unaffected area of the affected eye, although the P1/N1 ratio is also smaller than that in the normal eye, the difference is not statistically significant (Fig. 4A ). Similarly, the fMERG peak latencies of P1 and N1 are larger in the unaffected area than those in the normal eye, but the difference is not statistically significant (Fig. 4B) . The interocular ratio of these parameters reveals a significant decrease in the P1/N1 ratio (P < 0.05) and a significant increase in the N1 (P < 0.01) and P1 (P < 0.01) latencies in the affected area compared with those in the unaffected area (Fig. 4C) . 

It was sometimes difficult to determined the N2 trough because of the large variations in the waveform of N2 component. The N2 latencies have a large coefficient of variance (0.11 to 0.13) compared with the P1 latencies (0.04 to 0.06). Although sample size was smaller than those of N1 or P1, changes in the amplitude and latency of the N2 component were almost the same as for P1 (P < 0.05, not shown). 

Correlation of the RMS-Amplitude between the sMERG and fMERG

The RMS-amplitude of the sMERG was significantly correlated with that of the fMERG in all affected eyes (r = 0.50, P < 0.01, Fig. 5A ). However, the inter-ocular ratio of affected to normal eye reveals a much higher correlation between these ERGs (r = 0.69, P < 0.001, Fig. 5B ). As shown in the Figure 5B , the ratio of the sMERG decreases abruptly when the fMERG ratio is within the normal range (at around 0.8 ∼ 1.0). On the other hand, there are no significant correlations between the ratios of fMERG and sMERG in the affected area or in the normal controls. The slope and y-intercept of the regression line are 1.2 and -0.3, respectively. Almost all points in the affected area (black circle in Fig. 5B ) fall below the line with a slope of 1.0. Thus the RMS-amplitude of the sMERG is much more sensitive to arterial occlusion than that of the fMERG. 

Relationship of fMERG Parameters to RMS Amplitudes of the sMERG

Figures 6A B and Table 2 show the correlation between the RMS amplitude of the sMERG and fMERG parameters: N1 and P1 latency in patient 5. It is noteworthy that the fMERG latencies of N1 and P1 remain within the normal range after the sMERG amplitudes has begun to decrease; the latencies only begin to increase after the sMERG amplitude has decreased by approximately 15 nV (Figs. 6A 6B) . In the affected eyes of the other patients (patients 1 to 4), P1 and N1 latencies share the same trend in the relationship to the sMERG. It is clear that prolonged implicit times are associated with very small second order responses. 

Discussion

Sutter and colleagues suggested that the sMERG could be decomposed into a retinal component and an optic nerve head component (ONHC). ³ They hypothesized that the ONHC was generated by retinal ganglion cells. Bearse et al. reported an abnormal sMERG in glaucoma. ^{4 5} They suggested that the ONHC was either reduced in amplitude or completely disappeared in patients with glaucoma and optic atrophy. Palmowski and colleagues ² showed that early changes occurred in the waveform of the second-order responses that had contributions from sources in the inner retina and optic nerve head as well as from sources in the outer retina. 

Although subtle changes in the fMERG have been suggested, ^{7 10} it is generally difficult to confirm the ganglion cell contribution to those components. Hood and colleagues found that naso-temporal variation in the fMERG waveform was reduced after the intravitreal injection of tetrodotoxin in macaques. ¹¹ In the present study, such a change could not be found because the more outer layers of the retina were also involved in the BRAO patients. Recently, Hood and colleagues ⁷ concluded that there was no significant difference between the sMERG from glaucoma patients and from controls. Thus it is highly possible that the ganglion cells, which are lost as optic nerve damage progresses in glaucoma, ^{12 13} are not the main sources of the sMERG. 

On histologic evaluation a few weeks after acute BRAO, diffuse inner retinal atrophy is noted, involving the inner two-thirds of the inner nuclear layer, as well as the inner plexiform layer, the ganglion cells, and the nerve fiber layers. ⁹ The photoreceptor cells remain relatively intact in the early stages because of retention of choriocapillaris nourishment. Electroretinography characteristically discloses a diminution of the b-wave as a result of inner-layer retinal ischemia. The a-wave, which corresponds to photoreceptor function, is relatively unaffected. ^{15 16} The positive wave of the fMERG is a combination of the positive components of the full-field ERG. ¹⁷ The contribution to the N1 from photoreceptors is thought to be small at the intensities used here. However, from the fact that the later components (P1, N2) of the fMERG are more changed than the earlier one (N1) by artery occlusion, it is possible that these later components may contain more information regarding the inner retina than the earlier one. There is a large variation in the waveform of N2 component. This may be caused by adaptation lasting over the length of a stimulus base interval. 

In the present study, the fMERG responses were relatively preserved in the affected area, being 30% to 60% as large as in the normal controls. The fMERGs did not decrease to noise level in affected areas where the perimetric sensitivity showed 0 dB. Conversely, in these BRAO patients the sMERG showed a much greater reduction than the fMERG. There were no substantial second-order responses from the pathologic areas. Moreover, some of the sMERG responses decreased considerably even before waveform changes of the fMERG occurred: in the correlation between the sMERG RMS-amplitude and the fMERG parameters, the sMERG response decreased more abruptly even while the fMERG parameters remained at almost normal values. Thus, there was an obvious difference between the sMERG and fMERG in terms of their responsive properties. Regional difference in the susceptibility to ischemia among various cells of retina was shown histopathologically. ¹⁴ sMERG can detect more subtle changes in the inner retina than can the fMERG. 

Oscillatory potentials (OP) can be depressed in diabetic patients without retinopathy ^{18 19 20} and in patients with retinal artery occlusion. ²¹ The oscillatory potentials reflect neuronal synaptic activity in inhibitory feedback pathways initiated by the amacrine cells in the inner layer; the bipolar or the inner-plexiform cells are the probable generators of the response. ^{22 23} Recently, Wu and Sutter reported that the OPs could be mapped by recording the sMERG at a lower stimulating rate. ²⁴ We found that the averaged sMERG amplitudes decreased to the same extent as the OP amplitudes in the affected eye in BRAO patients. 

Even when perimetric sensitivity reveals that little information is reaching the visual cortex from a region of the field, a considerable proportion of the fMERG response can still be recorded from that region. This result is also supported by the experimental finding ¹⁴ that the flash ERG is a poor predictor of a visual function because it gives no information about the functional integrity of the innermost layer of the retina. The sMERG agree well with the perimetric sensitivity in terms of the area and extent of damage. The damage caused by the lesion must be to the more inner retina, preventing proximal transmission of information, rather than to the bipolar cells. Thus, we hypothesize that the reduction of sMERG is attributed to the dysfunction of the inner-plexiform layer, the integrity of which is needed for adaptive mechanisms and the oscillatory potentials (OP). This working hypothesis is also supported by an experimental study showing the OP and sMERG were reduced after intravitreal injection of glycine and GABA in pigmented rabbits. ⁶  

From these results, although changes of the fMERG in BRAO patients can be attributed mainly to damage to the inner nuclear layer, the extinction of the sMERG can be related to damage to the more proximal retina, probably the inner plexiform layer. This suggests that the second-order component of the MERG can be a very sensitive indicator of local inner retinal dysfunction. 

 Submitted for publication June 19, 2000; revised September 5, 2000; accepted September 15, 2000.

 Commercial relationships policy: N.

 

Table 1.

 

Five Patients with BRAO

Patient	Sex	Age	Affected	v.a.	Ref.	MD	VF Defect	Follow-Up Term
Pt 1	Male	18	Rt	20/20	+0	−7.0	Sup nasal	48 M
Pt 2	Male	53	Lt	20/200	−0.25	−20.2	Sup+Ct	30 M
Pt 3	Male	64	Lt	20/20	+1.0	−10.7	Inf	36 M
Pt 4	Female	69	Lt	20/20	−1.0	−17.3	Inf	72 M
Pt 5	Male	53	Lt	20/20	+0.25	−18.4	Inf	48 M

 Affected, Affected eye; Rt, right eye; Lt, left eye; v.a., visual acuity; Ref., refractive error; MD, mean deviation (dB); Sup, superior; Ct, center; Inf, inferior; M, month.

Figure 1.

 

Mappings of the Humphrey static perimetry (A) and MERGs (fMERG, B; sMERG, C) from the five patients with BRAO.

Figure 2.

 

(A) All local first- and second-order responses from the normal eye of Pt 3 are displayed. The location of U1 to 4, L1 to 4, ST, IT, SN and IN used in Figure 2A are also illustrated. ST, superior temporal; SN, superior nasal; IT, inferior temporal; IN, inferior nasal. MERGs (n = 13) in the vertical and horizontal meridian were excluded from analysis. (B) Local MERGs from affected eyes of Pt 1 to 5 and those from the normal eye of Pt 5 (Pt 5′, gray waveforms). The waveforms of local MERGs at the same latitude of the upper and lower quadrant areas (U1 to 4 and L1 to 4) are superimposed. More than two responses are superposed in each location. Arrows indicate marked reduction of the sMERG.

Figure 3.

 

RMS-amplitudes of the fMERG (A) and sMERG (B) from affected and unaffected area in all BRAO patients, and interocular ratios of affected to unaffected eyes for the amplitudes of the MERGs (C) and of the OPs (D). Normal, normal value; unaffected, unaffected area; Lesion, affected area (arterial occlusive area); Noise, noise levels of local MERG obtained from 2 normals (n = 98). * P < 0.05, ** P < 0.01, *** P < 0.001.

Figure 4.

 

fMERG parameters in the BRAO patients. (A) Mean and SD of the fMERG P1/N1 ratio. (B) Mean and SD of the fMERG N1 and P1 latencies. (C) Interocular ratios of affected to normal for the fMERG parameters. * P < 0.05, ** P < 0.01.

Figure 5.

 

Correlation of the RMS amplitude between sMERG and fMERG. (A) RMS-amplitude in all affected eyes. (B) Interocular ratio of affected to normal eye for the RMS-amplitude. Significant correlation is found between the sMERG and fMERG. The ratio of the sMERG decreases abruptly when the fMERG ratio is within the normal range.

Figure 6.

 

Correlation between fMERG parameters and sMERG RMS-amplitude for patient 5. It is noteworthy that the fMERG latencies of N1 and P1 remain within the normal range after the sMERG amplitudes have begun to decrease. The latencies only begin to increase after the sMERG amplitude has decreased by approximately 15 nV.

Table 2.

 

Correlation Coefficients Between the fMERG Parameters and RMS-Amplitude of the sMERG in Each of Five Patients with BRAO

Patient	fMERG Parameter
		P1/N1 Ratio	N1 Latency	P1 Latency
Pt 1	0.48, **	−0.50, ***	−0.61, ***
Pt 2	0.54, ***	−0.45, **	−0.66, ***
Pt 3	0.57, ***	−0.44, **	−0.51, ***
Pt 4	0.12	−0.45, **	−0.37*
Pt 5	0.28	−0.69, ***	−0.66, ***

 High correlations are found between the sMERG RMS-amplitude and each of the fMERG parameters.

*   P < 0.05,

**   P < 0.01,

***   P < 0.001.