Multifocal pattern electroretinography for the detection of neural loss in eyes with permanent temporal hemianopia or quadrantanopia from chiasmal compression
- Division of Ophthalmology, Hospital das Clínicas of the University of São Paulo Medical School, São Paulo, Brazil
- Correspondence to Dr Mário Luiz Ribeiro Monteiro, Av Angélica 1757 conj 61, 01227-200, São Paulo, SP, Brazil;
- Accepted 24 February 2011
- Published Online First 17 March 2011
Aims To evaluate the ability of multifocal transient pattern electroretinography (mfPERG) to detect neural loss and assess the relationship between mfPERG and visual-field (VF) loss in eyes with chiasmal compression.
Methods 23 eyes from 23 patients with temporal VF defects and band atrophy of the optic nerve and 21 controls underwent standard automated perimetry and mfPERG using a stimulus pattern of 19 rectangles, each consisting of 12 squares. The response was determined for the central rectangle, for the nasal and temporal hemifields (eight rectangles each) and for each quadrant (three rectangles) in both patients and controls. Comparisons were made using variance analysis. Correlations between VF and mfPERG measurements were verified by linear regression analysis.
Results Mean±SD mfPERG amplitudes from the temporal hemifield (0.50±0.17 and 0.62±0.32) and temporal quadrants (superior 0.42±0.21 and 0.52±0.35, inferior 0.51±0.23 and 0.74±0.40) were significantly lower in eyes with band atrophy than in controls (0.78±0.24, 0.89±0.28, 0.73±0.26, 0.96±0.36, 0.79±0.26 and 0.91±0.31, respectively). No significant difference was observed in nasal hemifield measurements. Significant correlations (0.36–0.73) were found between VF relative sensitivity and mfPERG amplitude in different VF sectors.
Conclusions mfPERG amplitude measurements clearly differentiate eyes with temporal VF defect from controls. The good correlation between mfPERG amplitudes and the severity of VF defect suggests that mfPERG may be used as an indicator of ganglion cell dysfunction.
Clinical trial registration number ClinicalTrial.gov identifier number NCT00553761.
- Multifocal pattern electroretinography
- temporal hemianopia
- chiasmal compression
- band atrophy of the optic nerve
- optic nerve
- visual pathway
- diagnostic tests/investigation
- field of vision
The transient pattern electroretinogram (PERG) is an electrical potential thought to be derived from retinal ganglion cells (RGC) and neighbouring inner retinal structures when a temporally modulated patterned stimulus of constant mean luminance is viewed.1 Primary PERG features are a prominent positive peak at 50 ms and a slow, broad trough at approximately 95 ms,2 most likely generated by the RGC.2–5 Several clinical studies have indicated that diffuse RGC damage can be revealed by PERG recordings.2 4 6–11 However, one major drawback of the use of PERG in the assessment of anterior visual pathway disease is its apparent inability to detect localised neural loss.3 12 13
The development of techniques for recording multiple local PERG responses using multi-inputs has made it possible to record simultaneous localised PERG response from multiple areas of the retina in order to detect localised retinal damage, making it possible to obtain multifocal PERG (mfPERG) responses. However, few studies have investigated the ability of mfPERG to detect localised neural loss in glaucoma patients with no evident correlation between mfPERG measurements and VF defects.14 15 Neural loss patterns in patients with longstanding bitemporal heamianopia, previous optic chiasmal compression and normal nasal hemifields may provide a clearer model for RGC deficits that are isolated to specific regions of the retina. In such patients, the crossed nerve fibres are lost, with relative preservation of the uncrossed fibres, which originate in the temporal hemiretina and penetrate the optic nerve through the superior and inferior arcuate fibre bundles. Therefore, retinal nerve fibre loss occurs predominantly on the nasal and temporal sides of the optic disc, a pattern identified on ophthalmoscopy as band atrophy (BA) or ‘bow tie’ atrophy of the optic nerve. Distinctive changes may also be observed in the RGC layer, with severe loss in the hemiretina nasally to the macula and relative preservation on the temporal side.16 17
The purpose of this study was to evaluate the ability of mfPERG for detection and quantification of neural loss in eyes with permanent temporal VF defects owing to previous chiasmal compression. We also evaluated the relationship between mfPERG findings and the severity of VF loss in these patients.
Materials and methods
Twenty-three eyes from 23 patients (12 male) with temporal VF defect, and 21 eyes from 21 normal (10 male) controls were studied. All patients had already been treated for a suprasellar tumour and had stable VF defects at least 1 year prior to study entry. All patients were scanned using MRI to confirm the diagnosis of pituitary adenoma compressing the optic chiasm and to document effective optic pathway decompression after treatment.
All subjects underwent a complete ophthalmological examination including VF evaluation using standard automated perimetry. Among the inclusion criteria for the study were a best-corrected VA of 0.8 or better in the study eye, within ±5 dioptres for the more ametropic meridian and intraocular pressure less than 22 mm Hg. In 15 patients, only one eye met the inclusion criteria. For the eight patients in whom both eyes fulfilled the inclusion criteria, one eye was randomly selected for analysis. Controls consisted of healthy volunteers with normal ophthalmic examination and normal VF.
VF testing was performed using the 24-2 SITA-Standard strategy (Humphrey Field Analyzer, Carl-Zeiss Meditec, Dublin, California) and a Goldmann size III stimulus on a 100 cd/m2 (31.5-apostilb) background. Reliability criteria were false positives, false negatives or fixation losses less than 30%. Patients with BA were required to have complete or partial temporal hemianopia and a nasal hemifield within normal limits. Only one eye of each patient and control was selected for analysis.
mfPERG measurements were made with the RETIScan System (version 3.20.30, Roland Consult, Wiesbaden, Germany) using a stimulus pattern of 19 patterned rectangles, each consisting of 12 alternately black and white squares in a checkerboard pattern. The stimulus was displayed on a 21-inch black-and-white rectangular flat screen (CRT colour monitor, Roland Consult) subtending a visual angle of 24° when viewed at 26 cm in a semidark (illuminance 0.34 cd/m2), acoustically isolated room. The size of the rectangles and the checks within them increased with eccentricity at a scaling factor of 1.4 (figure 1). The black/white alternation of the squares was induced by a modified binary m-sequence (four cycles, total duration of 6 min) with an on/off probability of 0.5 and roughly 98% contrast. The effective reversal rate of the squares was 4.65 Hz. The mean luminance was 180 cd/cm2 for each stimulus.
Each patient's refraction was optimally corrected, and stimulation was monocular after occlusion of the other eye. The subject was instructed to look at a 2 cm red crosshair fixation target at the centre of the screen. Fixation was monitored by the examiner. DTL electrodes placed in the lower bulbar conjunctiva 1 to 2 mm below the limbus (fastened to the outer and inner lid cantus) were used for all recordings, while gold cup electrodes and ground electrodes served as reference on the temples and forehead, respectively. Lighting conditions were the same for all examinations, and the pupil measured between 3 and 4 mm in all subjects. The duration of data acquisition was approximately 6 min divided into four sessions of 90 s each. Results are given as the average of these four sessions. Amplifier channels setting were: 1 Hz (low-cut); 50 Hz (high-cut) and 100 μV (range).
The waveform of transient mfPERG obtained in our study is characterised by a small initial negative component (N1) followed by a large positive component at ∼50 ms (P1) and a large negative component at ∼95 ms (N2). P1 and N2 amplitudes were measured from the preceding deflection to the peak of the wave. Peak times of both waves were also measured.
In order to evaluate the correlation between stimulation patterns and responses, the 19 focal mfPERG responses were displayed in their respective topographical positions, corresponding to the stimulated area in the VF. Responses from the central rectangles were averaged and analysed for eyes of patients and controls. The same was done with responses from two hemianopic subsets (nasal and temporal) of eight rectangles each. To avoid information in duplicate, we excluded data from the three rectangles located along the vertical meridian. The responses from each of the four quadrants (three rectangles each) were also averaged (figure 1).
mfPERG parameters of eyes with BA and normal controls were compared using variance analysis (ANOVA). Receiver operating characteristic curves were used to describe the ability of mfPERG parameters to distinguish BA from control eyes.
VF sensitivity loss was assessed for several of the 24-2 standard test locations, roughly the equivalent of the area tested by mfPERG. The quantification of VF sensitivity loss in patients and controls was based on the global mean defect (MD) and on the quadrant-specific results of the VF analysis: the temporal mean defect (22 temporal points, excluding two points immediately above and below the blind spot), the nasal mean defect (26 nasal points, excluding the two peripheral nasal points), the superonasal and inferonasal quadrants (13 points each) and the superotemporal and inferotemporal quadrants (11 points each). Each number represents the difference (dB) between the threshold luminance for that participant and that of a group of age-matched normals. For each calculation, the deviation from normal at each test location was converted from dB to 1/Lambert (1/L) units by dividing the dB value by 10 and unlogging the quotient. In this, we followed suggestions in the literature to the effect that the correlation between VF and anatomical RGC data should also be evaluated in a linear scale.18 Relative sensitivity was calculated globally and in each VF sector evaluated in patients and controls.
To determine the degree of association between measurements, we calculated the Pearson correlation (r) between mfPERG amplitude parameters and VF sensitivity loss parameters. Subsequently, the relationship between VF sensitivity loss expressed in 1/L was described with linear regression analysis. Since both multiple comparisons and correlations were performed, the level of statistical significance was set at a conservative <0.01.
The mean age±SD was 48.9±9.4 years (range 28–64) in patients, and 43.0±12.6 years (range 28–65) in controls (p=0.08). All patients had pituitary adenoma. On standard automated perimetry, 10 eyes presented complete temporal hemianopia, seven had a defect affecting approximately one quadrant, and six had a defect involving less than one quadrant. The mean±SD standard automated perimetry MD, temporal mean defect and nasal mean defect were −9.3±4.8, −20.2±10.3 and −1.8±1.5, respectively. The corresponding values for the superotemporal, superonasal, inferotemporal and inferonasal VF quadrants were −21.7±9.1, −1.7±2.0, −18.3±12.0 and −1.9±1.3, respectively.
Table 1 shows the mfPERG amplitude measurements in eyes with BA and control eyes. The records from one control and three patients are given as an example in figure 2. The average P1 amplitudes of the central rectangle, the temporal hemifield and both the superotemporal and inferotemporal quadrants were significantly lower in eyes with BA than in normal eyes. There was no significant difference between the two groups with regard to nasal field measurements. Similar findings were observed for N2 amplitudes except for the inferotemporal quadrant. The best area under the receiver operating characteristic curves for distinguishing between BA eyes and controls were those of the temporal hemifield (0.83) and the superotemporal (0.83) and inferotemporal (0.79) quadrants, which do not differ significantly.
In the eyes with BA, the average peak time measurements (mean±SD) for P1 in the central rectangle, the eight-rectangle nasal and temporal subsets and the three-rectangle superotemporal, inferotemporal, superonasal and inferonasal subsets were: 42.4±2.0; 40.1±1.3; 40.4±1.6; 40.2±1.8; 40.1±1.6; 40.7±2.3; 39.6±2.4, respectively. The corresponding values for control eyes were: 42.4±3.0; 39.7±2.5; 40.7±2.3; 40.5±1.5; 40.4±1.4; 40.7±2.2; 40.6±2.1. No significant difference was found between the two groups.
The average peak time measurements (±SD) for N2 in the central rectangle, the nasal and temporal subsets, and the superotemporal, inferotemporal, superonasal and inferonasal quadrantic subsets were: 100.7±11.8; 109.4±9.3; 113.2±19.4; 107.2±11.2; 106.7±10.7; 105.9±10.8; 107.0±7.5, respectively. The corresponding values for controls were: 103.1±6.2; 101.1±14.1; 100.4±13.7; 101.4±10.7; 96.9±13.1; 99.7±15.1; 101.1±12.2. No significant difference was found in N2 peak time values between control eyes and eyes of patients with BA in any comparison, with the exception of the inferotemporal quadrant (p=0.009).
mfPERG amplitude normal values were defined when above the lower 10th percentile value of the normal controls. The lower limit amplitude measurements (μV) for P1 in the central rectangle, the eight-rectangle nasal and temporal subsets and the three-rectangle superotemporal, inferotemporal, superonasal and inferonasal subsets were: 0.81; 0.59; 0.47; 0.44; 0.48; 0.37 and 0.49, respectively. Corresponding values for N2 amplitude measurements (μV) were: 1.18; 0.63; 0.53; 0.54; 0.58; 0.62 and 0.55. Under these conditions, the P1 value for the eight-rectangles temporal average, the superotemporal and the inferotemporal quadrantic subsets was abnormal in 13, 14 and 12 eyes, respectively. The corresponding value for the N2 amplitude was abnormal in 10, 14 and nine eyes, respectively.
Table 2 shows the associations between mfPERG amplitude measurements and VF parameters. Correlations were significant for several mfPERG and VF parameters, especially between temporal hemianopic or quadrantic mfPERG average amplitude measurements and temporal VF deviations from normal in both dB (range 0.51–0.66) and 1/L (range 0.60–0.73). The correlations between the average P1 amplitude of the temporal subset and the VF measurements of the inferotemporal quadrant, between the temporal subset and the MD (ρ=0.73; p<0.001) and between the temporal subset and the temporal mean defect (ρ=0.72; p<0.001) were the three most significant when VF loss was assessed in 1/L units. Correlations were also strong between the average P1 amplitude of the superotemporal quadrant and the MD (ρ=0.73; p<0.001), between the superotemporal quadrant and the TMD and between the superotemporal quadrant and the VF deviation from normal in the inferotemporal quadrant (ρ=0.69; p<0.001).
Figure 3 shows the results from the linear regression analysis of the best-performing mfPERG parameters and respective VF findings. The greatest R2 was observed between the P1 amplitude of the temporal subset and the global VF relative sensitivity (R2=54%; p<0.001), the inferotemporal VF relative sensitivity (R2=52%; p<0.001) or the TMD (R2=51%; p<0.001), and between the P1 of the superotemporal quadrant and the global VF relative sensitivity (R2=52%; p<0.001) or the inferotemporal quadrant VF relative sensitivity (R2=48%; p<0.001).
Our results show that several mfPERG amplitude parameters were significantly lower in eyes with BA than in control eyes. Our results are in accordance with previous studies on glaucoma showing reduced mfPERG amplitudes,14 15 19 with our previous studies using full-field PERG stimulation in a similar setting of patients13 20 and with other studies demonstrating PERG abnormalities in chiasmal compression.11 21 While one should be careful when comparing full-field with mfPERG measurements, the fact that the abnormalities observed in the present study matched previous full-field PERG measurements suggests that the mfPERG technology provides valid findings and is probably related to the integrity of the RGCs.
In order to further evaluate the possibility of detecting sector-specific neural loss, we averaged and compared the responses of different subsets of rectangles (hemifields and quadrants). Subsets were preferred to individual rectangles, as the rectangles did not correspond exactly to sets of points in the visual-field test. In the temporal subset as well as in the superotemporal and inferotemporal subsets, the P1 and N2 amplitudes were significantly lower in eyes with BA than in controls, corresponding to the affected areas of the VF. Not surprisingly, findings did not differ significantly with regard to the nasal field, which remained unaffected in both groups. In fact, the strong correlation observed between mfPERG amplitudes and VF deviation from normal indicates that the mfPERG parameters used in this study provide sufficient information to identify sector-specific axonal damage.
However, some of our findings are at variance with findings previously published. Klistorner et al14 studied the higher-order kernels of the mfPERG of glaucoma patients with a well-established scotoma defect and observed an overall reduction in mfPERG amplitudes; however, since the reduction did not correspond to the area of scotoma, they concluded that mfPERG does not reflect localised RGC loss in such patients. The authors suggested that perhaps mfPERG responses are dependent not only on RCG but also on other retinal neurons in the outer retina, masking the small changes resulting from RGC damage. Alternatively, they suggested the stimulus configuration may not have been optimal to elicit responses from the RGC. Stiefelmeyer et al15 studied 23 eyes with glaucoma and found mfPERG amplitudes to be significantly reduced in relation to normals, particularly in the central area. Amplitudes decreased as the disease progressed, suggesting a correlation between mfPERG parameters and disease severity. However, the abnormalities were more evident in the central area and did not correlate directly with areas of VF defect.
Many variables should be considered when evaluating mfPERG responses and their relationship to VF defects, particularly differences in equipment, stimulus configuration, and stimulation size and rate.22 These differences, along with the aetiology of axonal involvement in each case, explain some of the discrepancies between present and previous findings. In glaucoma, even when patients with localised scotomas are selected for study, the amount of apparently normal VF compromised by the disease cannot be determined with precision, since glaucoma is known to affect RGC diffusely even before VF defects are detectable. We believe that patients with chiasmal lesions provide a better model for understanding the results of mfPERG parameters owing to the more clear-cut separation between affected and healthy areas of the retina. Accordingly, most of our patients had large areas with VF abnormalities (temporal hemianopia or quadrantanopia) which could be compared with corresponding areas in healthy eyes on standard automated perimetry.
The fact that mfPERG responses were averaged for relatively large sectors (eight or three out of 19 rectangles) probably made the relationship between mfPERG and VF more evident. Although, in this study, the mfPERG response could not be directly correlated with specific VF areas, we believe we have demonstrated the possibility of detecting localised neural loss with this technique, at least in eyes with temporal hemianopia due to chiasmal compression. P1 and N2 waves were found to have a good discrimination ability in this study, matching findings by other authors.14 15 As in previous studies, no significant abnormalities were observed in peak-time values, except for the N2 response corresponding to the inferotemporal quadrant. Although both P1 and N2 amplitude parameters were able to differentiate normal eyes from eyes with BA, the performance was slightly better for the P1, a finding that is somewhat different from the results obtained in full-field PERG in which N95 is more affected than P50 in ganglion cell loss.2 8 13 The exact reason for such a discrepancy is unknown but may be related to a possible difference in the origin of the wave in each test and the fact that mfPERG N2 wave is somewhat more variable than the second positive wave in full-field PERG.
In conclusion, mfPERG amplitude measurements were shown to be able to differentiate eyes with temporal hemianopia, owing to chiasmal lesions from normal controls, and to correlate with VF defect severity as assessed by standard automated perimetry. However, further studies are necessary to confirm these findings before mfPERG can be used in the diagnosis and management of patients with chiasmal compression.
Funding Supported by grants from Conselho Nacional de Desenvolvimento Científico e Tecnológico, CNPq (No 309709/2007-5 and 479716/2008-0), Brasília, Brazil.
Competing interests None.
Patient consent Obtained.
Ethics approval Ethics approval was provided by the Institutional Review Board Ethics Committee (Comissão de Ética para projetos de pesquisa).
Provenance and peer review Not commissioned; externally peer reviewed.