Quantitative Electroretinogram Measures of Phototransduction in Cone and Rod Photoreceptors: Title and subTitle BreakNormal Aging, Progression With Disease, and Test-Retest Variability FREE

THE ELECTRORETINOGRAM (ERG) is widely used to help diagnose and follow patients with genetic eye disease.^{1 - 2} The International Society for Clinical Electrophysiology of Vision (ISCEV) established standards for the full-field ERG.³ The ERG has also become an important outcome measure in clinical trials.⁴ Historically, the focus has been on the amplitude and implicit time of the b-wave, and much is known about the decline of b-wave amplitude with age in control subjects,^{5 - 6} its inherent test-retest variability,^{7 - 9} and the natural history of b-wave decline in retinitis pigmentosa (RP).^{7 - 8} To our knowledge, this kind of information is not available for the a-wave of the ERG.

The a-wave of the ERG has long been known to reflect primarily photoreceptor activity.^{10 - 13} More specifically, the leading edge of the rod-only response has been shown to behave similarly to isolated rod photoreceptor recordings in response to changes in intensity, wavelength, and adaptation state.^{14 - 16} After the elaboration of a computational model that relates the response to events in the phototransduction cascade,¹⁷ we showed that an equivalent model could be fit to the leading edge of the rod-only,¹⁸ and subsequently cone,^{19 - 20} a-wave.

More recently, we²¹ presented a rapid protocol for the clinical assessment of photoreceptor activity. During the past few years, we used this protocol in 100 control subjects. We also attempted to use this protocol in several hundred patients with RP (R.T.T., K.G.L., D.C.H., et al, unpublished data, 2002). The purpose of the present study is to analyze reference cone and rod phototransduction variables and their variation with age; to evaluate progression in cone and rod variables over time in a subset of patients with XLRP; and to compare the test-retest variability of these a-wave variables with the variability in cone and rod b-wave measures.

Cone and rod a-waves to high-intensity stimuli were obtained from 100 controls aged 5 to 75 years (mean ± SD age, 41 ± 19 years) with no findings on eye examinations. The sample included 55 males and 45 females. To obtain test-retest measures, a-waves were obtained from a subset of 20 controls aged 9 to 73 years (mean ± SD age, 48 ± 18 years) in whom one eye was tested twice within 1 year.

To evaluate the feasibility of monitoring progression in phototransduction variables, high-intensity a-wave recordings were obtained yearly for 4 years(5 visits) from 24 patients with X-linked RP (XLRP) aged 5 to 38 years (mean± SD age, 16 ± 8 years). All 24 patients were participating in a masked, placebo-controlled clinical trial to determine possible benefits of nutritional supplementation on the rate of disease progression. All patients were genotyped and represented supplemented and placebo groups. They were selected for analysis simply because they retained sufficient cone and rod photoreceptor function for computational analysis (all had cone maximum amplitude[Rm_P3] >6 µV; 20 of 24 had rod Rm_P3 >9 µV). The purpose was to explore the feasibility and utility of repeated measures of photoreceptor function rather than to provide novel data on progression in what is admittedly a heterogeneous group of patients.

To compare test-retest variability in a-wave variables to that in b-wave variables, ISCEV standard responses³ were obtained twice in 2 months from 46 young patients (mean age, 16 years) with XLRP and 17 older patients (mean age, 26 years) with isolated RP. To obtain test-retest measures for controls representative of those participating in drug safety trials, ISCEV test-retest variability was assessed in a subset of 20 controls aged 50 to 65 years. Each was tested twice in 1 year.

All patients were recruited from the data files of the Retina Foundation of the Southwest. The tenets of the Declaration of Helsinki were followed, and written consent was obtained after all procedures were fully explained.

Electroretinographic testing was conducted after pupil dilation (using 1% tropicamide and 2.5% phenylephrine hydrochloride) and 45 minutes of dark adaptation. Pupil size was measured immediately after testing. Bipolar contact lens electrodes were used in all tests. All lenses were tested at least monthly for integrity, but no attempt was made to use the same lens for repeated measures. Dark-adapted ISCEV standard responses were followed by a-waves to high-intensity achromatic stimuli. As described previously,²¹ responses to several achromatic flashes for each of 4 increasing intensities were obtained in the dark. The intensities ranged from approximately 3.2 to 4.4 log scotopic troland-seconds (log sc td-s), with precise values depending on the measured pupil size for each patient. For each intensity, 3 replications were selected for averaging by computer. The time between flashes was at least 30 seconds and was long enough for complete recovery before a subsequent flash. Light-adapted ISCEV responses and the same 4 high intensities (5 replications with 5 seconds between flashes) were later presented against a rod-saturating background (3.2 log td). The resulting cone a-waves were subtracted from the dark-adapted responses to produce rod-only a-waves. The rod-only a-wave ensemble was fit with a computational model describing the response (R) as a function of time (t) and intensity (i):

where Rm_P3 is the maximum amplitude, S is a sensitivity variable, and t_d is a brief delay before the response onset.¹⁸ The cone model contained an additional time constant to reflect the capacitance of the cone outer segment.²⁰ The delay (t_d) was held constant at 1.7 milliseconds for cones and 3.2 milliseconds for rods.

All computer fits satisfied a least-squares goodness-of-fit criterion that was used to exclude "noisy" data. Electroretinographic variables were transformed to log values, and the normality of the data was verified using the Shapiro-Wilk test. The Bland-Altman analysis was used to determine repeated(test-retest) variability, with the repeatability coefficient defined as 2 SDs of the test-retest difference distribution.²² Threshold criteria were determined for change to be significant at the 95% confidence limit.

A representative set of data from an elderly control subject is shown in Figure 1. Figure 1A shows computer-averaged responses recorded in the dark to 4 intensities of achromatic flash (3.2, 3.7, 4.0, and 4.4 log sc td-s). These same intensities are presented again in the presence of a background of 3.2 log td (Figure 1B). This background is sufficient to completely eliminate any rod contribution to the response, but it does not measurably light adapt the cones.²⁰ To isolate the rod photoresponses, the light-adapted responses are subtracted from the dark-adapted responses (Figure 1C). Dashed lines in C show the fit of the equation shown in the"Methods" subsection to the leading edge. Finally, the intensities presented in the presence of the background range from 2.9 to 4.0 log photopic troland-seconds(log ph td-s) (Figure 1D) and are ideal for modeling the cone a-wave (dashed lines).

A comparable analysis is shown in Figure 2 for a patient with XLRP. The dark-adapted responses (Figure 2A) are approximately 15% of the reference range, whereas the light-adapted responses (Figure 2B) are approximately 33% of the reference range. Thus, the rod-only (Figure 2C) and cone-only(Figure 2D) responses differ in amplitude by a factor of 2 rather than the typical factor of 4 to 5 in controls.

Cone and rod phototransduction variables from all 100 controls are summarized in Table 1. The distribution of log values for each variable was not significantly different from a normal distribution. For cone and rod a-waves, the SD for Rm_P3 was lower than for S. For convenience, the mean and lower limits (P<.05) have been converted to linear units in the last 2 columns of Table 1.

Cone phototransduction variables from controls are shown as a function of age in Figure 3. As shown in Figure 3A, there is virtually no relationship between log cone Rm_P3 and age (r² = 0.01; P = .85) There is, however, a significant dependence of log S on age (r² = 0.29; P<.001) such that cone log S declines by approximately 0.04 log units per decade (Figure 3B).

Variation in normal rod phototransduction variables with age is shown in Figure 4. Log Rm_P3 shows a slight but significant decline with age (r² = 0.06; P = .01)(Figure 4A). The slope of the best-fit regression equation is 0.0013 log Rm_P3/y, or approximately 0.01 log unit per decade. The relationship between log S and age is considerably stronger (Figure 4B). The correlation is highly significant (r² = 0.58; P<.001), and the slope is clinically meaningful in that log S declines 0.006 log unit/y, or 0.06 log units per decade. Thus, the rod log S value for a control at age 70 years is likely to be approximately 0.3 log units lower than that of a teenager. Because cone and rod phototransduction variables are age dependent, values for controls younger and older than the median age are also given separately in Table 1.

To evaluate progression in a-wave variables, responses to high intensities were obtained yearly for 4 years from 24 patients with XLRP. As shown in Figure 5A, there was a modest decline in cone log Rm_P3 (F = 2.18; P = .08) during the 4 years of follow-up and a more pronounced decline in rod log Rm_P3 (F = 5.08; P<.001). The maximum amplitude of the rod photoresponse declined by 0.25 log unit (45%) in 4 years. Yearly average values of cone and rod log S are shown in Figure 5B. Although the average values of cone and rod log S were significantly below the mean reference values of 1.67 and 1.10, respectively, for the younger age group (Table 1), there was no systematic variation in log S during 4 years of follow-up (F = 1.4; P= .25).

Repeatability coefficients for test-retest measures of Rm_P3 in 20 controls were determined for cone and rod-only a-waves. The repeatability coefficients for cone and rod-only responses were 0.243 and 0.137, respectively. Criteria for a change to be significant at the 95% confidence level are given in Table 2. For cone Rm_P3, the maximum amplitude had to decrease by 37% before the change could be considered significant. A decrease in rod Rm_P3 of 23% should be considered significant. Cone and rod-only repeatability coefficients for the a-wave are also compared with previously reported values for the b-wave in Table 2.

These measures of repeatability in "single-flash" a- and b-waves can also be compared with repeated variability measures for cone amplitude to 31-Hz flicker stimulation. Figure 6shows repeatability measures based on 3 groups of patients: (1) young males with XLRP (mean age, 16 years; n = 46) tested twice in 2 months, (2) older patients with RP (mean age, 26 years; n = 17) tested twice in 2 months, and(3) older controls (mean age, 54 years; n = 20) tested twice in 1 year. As shown by the dashed lines indicating 2 SDs, the repeatability coefficient for cone amplitude to 31-Hz flicker is 0.24 log unit. This coefficient does not seem to vary with the age of the patient (Figure 6A) or the mean amplitude of the response (Figure 6B). Cone flicker b-wave amplitude must decline by 37% for progression to be considered significant at the 95% confidence level.

The protocol described recently²¹ for isolating cone and rod-only a-waves from the ERG is highly efficient. Sufficient data for rigorous derivation of cone and rod phototransduction variables can be obtained from 8 light intensities. The first 4 intensities (3.2-4.4 log sc td-s) are obtained after dark adaptation. Even with averaging and after allowing for complete recovery after each flash, these dark-adapted responses can be obtained in 6 to 8 minutes. The same 4 intensities are then presented on a 3.2–log td background, which is sufficient to completely eliminate any rod contribution without measurably light adapting the cones.²⁰ Because less time is needed between the flashes under light-adapted conditions, these cone responses can be obtained in approximately 3 minutes. Thus, the entire high-intensity protocol adds approximately 10 minutes to the ISCEV protocol. Although sophisticated software is presently required for modeling the photoresponse and deriving sensitivity and amplitude variables, it should be possible to simplify data analysis for routinely deriving the true maximum photoresponse amplitude. Whereas the present ISCEV responses were developed to detect abnormal retinal function, the maximum photoresponses may provide better longitudinal measures for natural history and treatment trials.

Because it requires a fairly short time, the high-intensity protocol can be attempted in patients with RP. Of 589 consecutive patients with RP, we obtained cone a-wave variables on 174 (30%) (R.T.T., K.G.L., D.C.H., et al, unpublished data, 2002). Only 11 (6%) of these 174 patients had a cone response to 31-Hz flicker of less than 2 µV. Rod variables in addition to cone variables were analyzable in 144 (24%) of 589 consecutive patients. Of these 144 patients, rod b-waves to the ISCEV standard rod stimulus measured less than 2 µV in 19 (13%). Thus, cone and rod transduction variables can be followed in the approximately one quarter of patients with RP in our population who retain cone or rod responses of greater than 2 µV to ISCEV standard stimuli.

We found virtually no decline in the maximum cone a-wave amplitude (log Rm_P3) with age. This finding is consistent with histologic studies^{24 - 26} reporting minor age-related decline in numbers of cone photoreceptors outside the macula. There was, however, a significant decrease in cone log S with age. The decrease is such that the difference in cone log S between controls at age 20 years and at age 70 years is approximately 0.2 log units. This decrease in cone photoreceptor transduction efficiency is consistent with the decline in cone b-wave amplitude with age.^{5 - 6}

We also found little decline in rod Rm_P3 with age. This too is consistent with available histologic findings, which suggests that some rods die with age but that the remaining rods swell to fill the vacant areas.²⁷ Presumably, this results in a fairly constant number of cyclic G-gated channels with age. The significant decrease in rod log S with age was such that sensitivity decreased 0.3 log units between the ages of 20 and 70 years. This decrease in log S completely accounts for the 50% reduction in rod b-wave amplitude by age 69 years in controls.⁶ The decrease has been noted previously and is much greater than expected from preretinal absorption.^{28 - 29} Cideciyan and Jacobson²⁹ suggested that one possible cause could be a buildup of cholesterol in outer segments, with age leading to decreased membrane fluidity and a subsequent slowing of activation stages.^{30 - 31}

Prospective measures of a-wave function have not previously been reported in RP. In a cross-sectional analysis,^{32 - 33} it was reported that patients of all ages with autosomal dominant RP associated with the pro-23-his rhodopsin mutation show low values of log S. Patients with RP associated with other mutations have normal values of log S.^{34 - 36} In this study, we followed a-wave variables in 20 patients with XLRP for 4 years. There was no significant variation in log S over time. Log Rm_P3, the maximal photoreceptor response amplitude, clearly depends on the stage of the disease. Both cone and rod Rm_P3 values showed progression in XLRP during the 4 years, with rod amplitudes declining faster than cone amplitudes. The Rm_P3 is proportional to the total number of cyclic guanosine monophosphate gated channels³⁷ and decreases as a result of either outer segment shortening/disarray or the death of entire photoreceptors.³³

Test-retest variability for rod and cone a-wave measures has not been reported previously, to our knowledge. Published data for computer-averaged rod and cone b-wave amplitudes, however, are in line with the values reported herein,^{8 ,23} and the values tend to be higher than those for a-wave Rm_P3values, perhaps because of the choice of intensities for ISCEV rod and cone responses. Because they are derived from regions where amplitude is linearly related to intensity, ISCEV responses are sensitive to disease but are also affected by variations in pupil size, flash variations, and so on. In a prospective natural history study of patients with RP or cone-rod dystrophy, Birch et al⁸ reported that the cone and rod b-waves had to decrease by 35% and 46%, respectively, to reach the threshold for significant decrease at P<.05. Similarly, values for elderly patients participating in a drug treatment trial were 55% and 46% for cone and rod b-waves, respectively (S. Grover, MD, G. A. Fishman, MD, D. G. Birch, PhD, K. G. Locke, RN, and B. Rosner, PhD, unpublished data, 2002). Published values for 31-Hz flicker amplitude tend to be lower than for single-flash cone b-waves, ranging from 35% to 55%.^{7 - 9}

One conclusion to be drawn from the present results is that Rm_P3, rather than log S, should be the outcome measure of choice when using the a-wave to follow photoreceptor function in prospective studies. Log S, which reflects the efficiency of the transduction steps between light absorption and channel closure, usually declines with age, which complicates its interpretation in any long-term study. The Rm_P3, on the other hand, is remarkably stable with age for cones and rods. Thus, any decline observed in a prospective study should reflect progression of the disease process.

Submitted for publication October 11, 2001; final revision received April 8; accepted April 24, 2002.

This study was supported in part by grants EY-05235 and EY-09076 from the National Eye Institute, Bethesda, Md; the Foundation Fighting Blindness, Owings Mills, Md; grant FD-R-001232 from the US Food and Drug Administration, Rockville, Md; and generous gifts from Steve and Nancy Rogers and Richard F. McCarthy.

Corresponding author and reprints: David G. Birch, PhD, Retina Foundation of the Southwest, 9900 North Central Expressway, Suite 400, Dallas, TX 75231(e-mail: dbirch@retinafoundation.org).