Minimizing Pupil Size Dependence in Flicker ERG Using Stiles–Crawford Compensation

Abstract

Purpose: This study determined the impact of the Stiles–Crawford effect (SCE) on electroretinograms (ERGs). Compensating for the SCE can improve the diagnostic reliability of ERGs by providing a stimulus minimally affected by pupil size.

Methods: Flicker ERGs were recorded from 10 healthy subjects at 3-minute intervals over 21 minutes after mydriasis. The RETeval system adjusted retinal illuminance in real time based on pupil size measurements, using a Troland stimulus and preset SCE compensation factors (ρ_device) of 0, 0.05, 0.085, and 0.12 mm⁻².

Results: Larger pupil areas led to prolonged implicit times with ρ_device = 0 and 0.05 mm⁻², whereas ρ_device = 0.12 mm⁻² reduced implicit time. Amplitudes were lower with ρ_device = 0 mm⁻² but increased with ρ_device = 0.085 and 0.12 mm⁻². The values that minimized pupil size dependence were ρ_device = 0.086 mm⁻² for the implicit time of the fundamental component of the ERG and ρ_device = 0.05 mm⁻² for all other measures. Variability in ERGs based on pupil size is predicted to be ≤7% of the associated 95% reference interval for Troland stimuli over the range of nonmydriatic pupil sizes, compared to ≤43% for luminance stimuli over the range of mydriatic pupil sizes.

Conclusions: Using Troland stimuli with ρ_device = 0.05 mm⁻² for all cone-mediated ERGs would minimize the impact of pupil size, although the improvement would be modest for ERGs performed with Troland stimulation without SCE compensation on non-dilated subjects.

Translational Relevance: Applying the appropriate SCE coefficient (ρ_device) enables more reliable ERG measurements, improving diagnostic accuracy despite pupil size variations in clinical settings.

Introduction

Electroretinograms (ERGs) elicited by fast flicker rates are used widely to assess the function of the retinal cone pathway in both clinical and experimental settings. Flicker ERGs are particularly valuable for monitoring residual cone function in degenerative retinal diseases and evaluating retinal ischemia, including diabetic retinopathy,^1–4 central retinal vein occlusion,^5–7 and retinal artery occlusion.⁸ However, the practical use of flicker ERGs to evaluate ischemia has been limited, in part, due to the assumption that mydriasis standardizes pupil diameter. Mydriasis is time consuming, and even small variations in pupil size with dilation introduce variability in ERG waveforms. 

Reproducible ERGs can be recorded without mydriasis by using stimuli that compensate for the pupil size. This approach is implemented in the RETeval system (LKC Technologies, Gaithersburg, MD), which continuously measures pupil area in real time and delivers full-field flash stimuli with constant retinal illuminance (Troland seconds [Td·s]) by adjusting the flash luminance (cd·s/m²) to compensate for the pupillary area variations (mm²). Numerous clinical studies have demonstrated its utility for diagnosis and prognosis of retinal diseases.^2,7,9–12 

An important factor influencing ERG recordings is the Stiles–Crawford effect (SCE), which describes the directional sensitivity of cone photoreceptors in the human eye with a coefficient ρ.^13–15 This effect reduces luminance efficiency for light entering the eye away from the pupil center. In addition to adjusting for the pupil area, compensating for the SCE in ERG stimuli could further minimize the effect of pupil size on ERG measurements. 

The experimental ability to provide a consistent stimulus across pupil sizes has been tested by repeated measurements while the pupil was being chemically dilated. For example, Kato et al.¹⁶ found the flicker implicit time measured with RETeval Troland stimulation had a residual dependency of 0.045 ms/mm² on an average time of 32 ms. Davis et al.¹⁷ found that Troland stimulation reduced the variability with pupil size by factors between 2.3× and 15× compared to constant luminance stimulation with undilated pupils. These findings are consistent with the SCE (lower effectiveness of light entering the peripheral pupil), albeit under clinically implausible conditions of measuring throughout the dilation process. A study of Troland-stimulation flicker ERGs across young healthy subjects without artificial dilation found no amplitude dependence on pupil area but did find a statistically significant dependence on flicker time of 0.096 ms/mm² over the ∼15-mm² range of natural pupil areas.¹⁸ 

This study aimed to provide experimental evidence that the residual dependence of flicker ERG is due to the SCE and to find the SCE coefficient, ρ_device, that best reduces the dependence of the flicker ERG waveform on pupil area. Further, this study aimed to assess the impact of the SCE and put it in the context of the pupil dependence of other ERG methods. Unlike the original studies by Stiles and Crawford¹⁵ (and most of the subsequent work), this study provides full-field illumination, rather than illuminating only the macula. 

Methods

Study Design

This was a prospective, single-center study conducted at the Mie University Hospital between December 2017 and March 2018. The Medical Ethics Committee of Mie University Hospital approved the procedures used (No. 2595), and the procedures conformed to the tenets of the Declaration of Helsinki. All participants signed a written informed consent form after they were provided with information on the procedures to be used. 

Subjects

Participants included 10 healthy volunteers (four males and six females; median age, 30 years; range, 20–52 years). All subjects had no known ocular disease, including media opacities, and no systemic diseases. All were phakic and had myopia of less than 6.0 diopters. 

Flicker ERG Recordings by the RETeval System

Flicker ERGs were recorded using the RETeval system, which provides a full-field stimulus through a 60-mm diameter dome. The stimuli were white lights created by a combination of three colored light-emitting diodes (LEDs), with a flicker frequency of 28.306 Hz and a pulse duration of less than 1 ms. A small red fixation spot was presented at the center of the dome to help subjects maintain steady fixation during the recording. The flash strength used was 8 photopic Td·s, where the RETeval system continuously measured the pupil area in real time using a built-in infrared camera and adjusted the pupil area for four different candidate SCE amounts (see next section). 

Flicker ERGs were performed every 3 minutes over a period of 21 minutes, beginning immediately after the instillation of mydriatic eye drops (0.5% tropicamide and 0.5% phenylephrine). The recording time ranged from 5 to 15 seconds, depending on the reliability of the result as assessed by the standard error of the mean estimate of the implicit time. Measurements were made of the peak-to-peak amplitude and time to the first peak from both the fundamental component of the averaged flicker ERG (best fit sinusoid at the stimulus frequency) and of the flicker ERG waveform (reconstructed from the first eight harmonics). 

Stiles–Crawford Effect Compensation

The SCE can impact the accuracy of ERG measurements, even when the stimulus is corrected for pupil area, because the effective pupil area is smaller than its physical area A:  

\begin{eqnarray*} {{A}_{{\rm{effective}}}} = {\rm{\ }}\frac{{\rm{\pi }}}{{\rm{\rho }}}\left( {1 - {{e}^{ - {\rm{\rho }}\frac{A}{{\rm{\pi }}}}}} \right) \end{eqnarray*}

To compensate for the SCE, the RETeval device was programmed with ρ_device coefficients based on the above formula (after substituting ρ_device for ρ) to compute the effective area used in the Troland calculation: Td = LA_effective, where L is the luminance in cd/m². 

In this study, we tested four different values of ρ_device: 0 mm⁻² (no SCE correction, and the RETeval default value), 0.05 mm⁻², 0.085 mm⁻², and 0.12 mm⁻². The ρ_device value was preset in the RETeval system prior to each recording session, and the system applied the compensation in real time during the ERG acquisition. These values were chosen to explore their effectiveness in minimizing the variation in ERG waveform measurements due to changes in pupil area. The selection includes ρ_device = 0.05 mm⁻², a value reported as physiologically appropriate,^13–15 as well as amounts expected to be too little compensation (0 mm⁻²) and too much compensation (0.085 mm⁻² and 0.12 mm⁻²) in order to bracket the expected physiological value. The relationship between pupil diameter and effective pupil area under these stimulus conditions is illustrated in the Table. This allows for quantitative comparison of the relative contribution of each correction term across pupil sizes. 

Table.

 

Pupil Diameter, Area, Effective Area, and Correction Ratios for Troland and SCE Compensation, Assuming ρ = 0.05 mm⁻²

 

Statistical Analysis

The ERG measurements were modeled using  

\begin{eqnarray*} {\rm{ERG}} = {\rm{\alpha }}\log {\rm{(}}L{{A}_{{\rm{effective}}}}) + {\rm{\beta }} \end{eqnarray*}

where α is a constant of proportionality between the log stimulus and the ERG measurement, and β is an offset, useful when the ERG measurement is a time or when examining small changes in luminance when the curve in nonlinear (e.g., near the photopic hill).¹⁹ 

As more fully explained in the Supplemental Material, the equation for the ERG can be rewritten as  

\begin{eqnarray*} {\rm{ERG}} = {\rm{\alpha }}\log \left(\frac{\pi }{{A\left( {{\rm{\rho }} - {{{\rm{\rho }}}_{{\rm{device}}}}} \right)}}\left( {1 - {{{\rm{e}}}^{ - \frac{{\left( {{\rm{\rho }} - {{{\rm{\rho }}}_{{\rm{device}}}}} \right)A}}{{\rm{\pi }}}}}} \right)\right) + {{{\rm{\beta }}}_{{\rm{Td}}}} \end{eqnarray*}

where the value ρ_device is used by the device to compensate for the human's SCE ρ, and the term β_Td = αlog(LA) + β is a constant assuming that Troland stimulation (without SCE compensation) is used. This equation has three known values: 

• • ERG, the measurement recorded
• • A, the pupil area measured
• • ρ_device, the SCE entered in the device

It also has three unknown values: 

• • α, the dependency of the ERG measurement on log stimulus
• • β_Td, the value of the ERG measurement with no SCE
• • ρ, the SCE for a human

To improve intuition, the first terms of the Taylor series, applicable for small values of A(ρ – ρ_device), is \({\rm{ERG\ }} \approx \ {{\beta }_{{\rm{Td}}}} - \alpha \frac{{( {\rho - {{\rho }_{{\rm{device}}}}} )A}}{{2\pi }}\). Underestimating the SCE (e.g., ρ_device = 0), causes the ERG to move in the opposite direction (–α) with increasing pupil area than in the direction of increasing luminance (α). 

Because each subject may have different values for these three unknowns, we applied a nonlinear fit separately for each subject. To ensure convergence during numerical fitting, we constrained ρ_device to be between 0 and 0.3 mm⁻², and α to be positive for amplitude measurements and negative for time measurements. (For luminance in the range of clinical flicker ERGs, an increasing pupil area for the same luminance increases the stimulation, leading to higher amplitude, and increasing the pupil area for the same luminance shortens the response time).²⁰ The SCE coefficients (ρ) for individual subjects were then combined using a weighted average based on the standard error of each fit. 

In addition to the nonlinear model fitting described above, we performed repeated-measures one-way analysis of variance (ANOVA) to compare ERG amplitudes and implicit times across different pupil sizes and compensation conditions (ρ_device = 0, 0.05, 0.085, and 0.12 mm⁻²). When significant main effects were identified, post hoc pairwise comparisons were conducted using Bonferroni correction. These analyses were applied to the data shown in Figure 1. Statistical analyses were performed using Mathematica 14.1²¹ and SPSS Statistics 29.0 (IBM, Chicago, IL).²² Results were considered statistically significant for P < 0.05. 

Figure 1.

 

Means across subjects for four different ρ_device values. Plots of means across subjects for the implicit time and amplitude values of flicker ERGs recorded during dilation. The left panels show the fundamental component, and the right panels show the waveform. The upper panels show the implicit times, and the lower panel shows the amplitudes. Stimuli were corrected by four different SCE coefficients: blue, ρ_device = 0 mm⁻²; orange, ρ_device = 0.05 mm⁻²; black, ρ_device = 0.085 mm⁻²; and red, ρ_device = 0.12 mm⁻². *P < 0.05, **P < 0.01, when comparing that time point to the baseline (time = 0) for that test.

Comparison of the Importance of Troland Corrections and SCE Corrections in the ERG

To investigate the SCE on ERGs made with Troland-based and luminance-based stimuli, we can examine the fitted ERG models. For Troland-based stimuli, ERG = αlog(LA) + β, incorporating both the luminance (L) and pupil area (A). Luminance-based ERG is modeled as αlog(L) + β′, where the stimulus is independent of pupil size and β′ equals β plus an offset to make the two ERG curves intersect at a 6-mm diameter pupil. Reference data built in the RETeval device for the 8-Td∙s flicker ERG is used to provide a relevant scale for whether a change is large or small. Pupil measurements from the reference data trial for the RETeval device (REACT trial, clinicaltrials.gov NCT03065881) for chemically dilated (all tests) and natural pupils (8-Td∙s flicker test) are used to provide the 90% reference interval for these measurements. 

Results

Theoretical Comparison of the Importance of Troland Corrections and SCE Corrections in ERGs

We showed earlier that the ERG model can be rewritten as  

\begin{eqnarray*} {\rm{ERG}} = {\rm{\alpha }}\log {\rm{(}}L{{A}_{{\rm{effective}}}}) + {\rm{\beta }} \end{eqnarray*}

 

After substituting definition of A_effective and multiplying by A/A, we get  

\begin{eqnarray*} {\rm{ERG}}\ = {\rm{\alpha }}\log \left[ LA\frac{\pi }{{A{\rm{\rho }}}}\left( {1 - {{e}^{ - {\rm{\rho }}\frac{A}{{\rm{\pi }}}}}} \right) \right] + {\rm{\beta }} \end{eqnarray*}

 

Expansion of the natural logarithm of the products yields  

\begin{eqnarray*} {\rm{ERG}} &=& {\rm{\alpha }}\log {\rm{(}}L{\rm{)\ + \ \alpha log\ (}}A)\ \\ && + \ \alpha log\left[ {\frac{{\rm{\pi }}}{{A{\rm{\rho }}}}\left( {1 - {{e}^{ - {\rm{\rho }}\frac{A}{{\rm{\pi }}}}}} \right)} \right] + {\rm{\beta }} \end{eqnarray*}

where the first three addends representing the stimulus can be considered as 

• • log (L), the light stimulus
• • log (A), the Troland correction
• • \({\rm{log}}[ {\frac{{\rm{\pi }}}{{A{\rm{\rho }}}}( {1 - {{e}^{ - {\rm{\rho }}\frac{A}{{\rm{\pi }}}}}} )} ]\), the SCE correction

Because the Troland correction is added to the light stimulus, increasing the pupil area increases the effective stimulus, and, as shown in the Table, the SCE correction decreases the effective stimulus. The Table shows the Troland term is 9× to 46× more important than the SCE for all pupil sizes, using ρ = 0.05 mm⁻². 

Flicker Electroretinography During Pupillary Dilation

As the pupil area increased during dilation, the implicit time of the fundamental component of the flicker ERG was prolonged when the coefficient (ρ_device) was set to 0 and 0.05 mm⁻² but became faster when ρ_device was set to 0.12 mm⁻² (Fig. 1). Similarly, the amplitudes of the fundamental component decreased significantly when ρ_device was 0 mm⁻² but increased significantly when ρ_device was set to 0.085 mm⁻² and 0.12 mm⁻². For the flicker waveform, timing was faster with ρ_device = 0.12 mm⁻², and amplitudes were larger for ρ_device = 0.085 mm⁻² and ρ_device = 0.12 mm⁻² and smaller for ρ_device = 0 mm⁻². These findings suggest that the coefficient ρ_device can reduce the ERG dependency on pupil size. 

Supplementary Figure S1 uses the first-order Taylor expansion \({\rm{ERG\ }} \approx \ {{{\rm{\beta }}}_{{\rm{Td}}}} - {\rm{\alpha }}\frac{{( {{\rm{\rho }} - {{{\rm{\rho }}}_{{\rm{device}}}}} )A}}{{2{\rm{\pi }}}}\) to plot the ERG as a function of pupil area A to arrive at the same result—that a nonzero coefficient ρ_device can reduce the ERG dependency on pupil size. 

Estimating the SCE Coefficient ρ

Each subject had measurements at eight time points and four values of ρ_device, providing 32 values to estimate the three unknowns (β_Td, α, and ρ). Supplementary Figure S2 shows representative nonlinear fits for one of the subjects. In this example, the SCE coefficient ρ was highly statistically different from 0 for fundamental time, fundamental amplitude, and waveform amplitude (P < 0.0001), but not for waveform time (P = 0.5). 

Figure 2 shows the computed values for the SCE, along with the other fit parameters for the flicker ERG amplitude and timing. All four ERG measurements demonstrated a nonzero SCE (P < 0.0001). The waveform amplitude, fundamental amplitude, and waveform time were all consistent (ρ = 0.05 mm⁻²), and the fundamental time had a larger SCE (ρ = 0.086 mm⁻²). More detailed analyses on the values and uncertainty for ρ for each subject and measurement are shown in Supplementary Figure S3, where the between-subject differences are not convincingly related to the subject. 

Figure 2.

 

Fitted parameters ρ, α, and β_Td with 95% confidence intervals. Fitted parameters for ρ, the SCE (top); α, the dependency of the ERG measurement on log stimulus (middle); and β_Td, the value of the ERG measurement with no SCE (bottom) for the four measurements. Uncertainty is represented by the 95% confidence interval. Time indicates implicit time.

The fit values for α were consistent with expectations, where an increase in luminance (near the 8-Td∙s stimulus used) increases amplitudes and decreases times. The values for β_Td showed the average (SCE-free) values for the four ERG measurements, which were all inside their respective reference intervals built into the RETeval device (Fig. 3). 

Figure 3.

 

Expected variation in ERG with pupil area with the SCE. Troland-based (blue) and luminance-based (gold) ERGs, where shaded areas represent the uncertainty (standard error of the mean) in the model parameters. The green area indicates the center 90% of the reference distribution taken from the RETeval user manual. Yellow is the next 2.5% on each side, and red areas represent the outermost 2.5%. The regions for the middle 90% of pupil areas from the RETeval reference data are shown for natural and dilated pupils. The tables show the expected variations of the given measurements for Troland tests over the center 90% range of natural pupil sizes (for the 8-Td∙s stimulus), Troland tests over pupil sizes spanning the fifth percentile natural pupil to the 95th percentile dilated pupil, and luminance tests over the center 90% range of dilated pupil sizes. The change for times is the maximum–minimum time; for amplitudes, the change is (maximum – minimum)/minimum. Also shown is the size of the variation compared to the 95% RI.

Comparison of the Importance of Troland Corrections and SCE Corrections in ERGs

Troland-based and luminance-based flicker ERGs change with pupil size (Fig. 3). The slopes are in opposite directions, as increasing pupil size for luminance-based tests increases the retinal stimulation, but the SCE causes Troland-based tests to overestimate the retinal stimulation at larger pupil sizes. For Troland-based tests on eyes without chemical dilation (Td undilated, percent of 95% reference interval [RI]) (Fig. 3), the changes in the ERGs were 1% to 7% of the 95% RI, a minor effect. In absolute terms, the changes in the ERGs for Troland stimulation without dilation were 0.1 ms, 0.4 ms, 4%, and 5%. depending on the measurement. If stronger Troland stimuli were used, pupil diameters without chemical dilation were smaller, leading to even less variability caused by the SCE (Supplementary Fig. S5). For luminance-based tests on dilated eyes, the changes in ERGs were 15% to 43% of the reference range (candela (cd) dilated, percent of 95% RI) (Fig. 3), a much larger effect. In absolute terms, the changes in the ERGs for luminance-based tests on dilated subjects were 0.8 ms, 1.9 ms, 63%, and 54% depending on the measurement (cd dilated, change) (Fig. 3). If both dilated and natural pupil sizes are considered, Troland-based ERGs varied by 7% to 34% of the 95% RI—in absolute terms, 0.4 ms, 1.9 ms, 28%, and 32% (Td all pupil) (Fig. 3). For rod-based ERG recordings, the SCE does not exist.^14,22 These same Troland and luminance ERG models can be used to analyze these tests by setting ρ = 0, as shown in Supplementary Figure S4. In these cases, Troland-based ERGs had no dependence on pupil size, whereas the luminance-based ERGs had an even greater dependency on pupil size, because the SCE reduced the effectiveness of increasing pupil size. Considering only dilated pupils, the variations in the ERG measurement were 1 ms, 3.1 ms, 73%, and 66% for times and amplitudes, respectively. 

Discussion

ERGs are known to be dependent on pupil size, and the International Society for Clinical Electrophysiology of Vision (ISCEV) standard requires dilating for luminance-based testing.²⁰ The latest update of the ISCEV standard²¹ now accepts non-mydriatic testing with constant Trolands, although it cautions to be aware of the SCE on cone-mediated ERGs. Previous studies have reported that, despite compensation for pupil size, some residual dependency remains.^16–18 This study hypothesized that the residual dependency is attributed to the SCE, characterized by the coefficient ρ. To address this, we investigated the impact of different ρ_device values on flicker ERG measurements to determine the optimal coefficient for minimizing waveform variation caused by differences in pupil area. Although SCE correction in the RETeval device is not activated by default, it can be enabled via a custom protocol. Unlike most research on the SCE,^13,14,23,24 here we stimulated the whole retina rather than the macula. 

Using ρ_device = 0.05 mm⁻² reduced the dependence of the flicker ERG waveform implicit time, fundamental amplitude, and waveform amplitude to near zero, whereas ρ_device = 0.085 mm⁻² eliminated the pupil dependence for implicit times of the fundamental waveform (Fig. 1). Fitting the data to a model of the ERG where the measurements were dependent on the logarithm of the luminance multiplied by the effective pupil area found similar results (Fig. 2). Because values of ρ_device were found to eliminate the pupil dependence of the ERG, these results are consistent with humans having a full-field SCE (ρ), with a value similar to that reported for focal stimuli. The SCE of ρ = 0.05 mm⁻² is also the value typically reported in psychophysics experiments.^13,14,22,23 

The model developed here can be used to predict how ERGs using Troland and luminance stimuli for cone and rod responses are affected by pupil size variations. For cone-based tests with the SCE (Fig. 3) and rod-based tests without the SCE (Supplementary Fig. S4), luminance-based ERGs showed reduced pupil size dependency after chemical dilation. The residual dependence on pupil size was still very large, being 15% to 43% of the 95% RI for cone-based tests (cd dilated, percent of 95% RI) (Fig. 3), and caused variations of 1 ms to 3.1 ms in time and 66% to 73% in amplitude for rod-based tests (cd dilated) (Supplementary Fig. S4). In contrast, compensating for pupil size with Troland-based tests greatly reduced this pupil dependency. For rod-based tests, we would expect Troland-based testing to have no dependency on pupil size, based on a lack of the SCE in those photoreceptors.^14,23 For cone-based testing, we found the dependency to be 1% to 7% for natural pupils (Td undilated) (Fig. 3) and 7% to 34% for natural and dilated pupils of the associated 95% RI (Td all pupil) (Fig. 3). The stimulus used here (8 Td∙s) is the dimmest provided in the RETeval device. With natural pupils, stronger Troland stimuli further constrict the pupils and therefore result in an even smaller SCE (Supplementary Fig. S5). To minimize test–retest variability and reduce both inter- and intrasubject variability caused by pupil size differences, the most reliable stimulus is a Troland-based test with natural pupils. This consistency can be further improved by setting ρ_device = 0.05 mm⁻². 

When a patient is tested with dilation on some occasions (perhaps after a detailed examination) and later without dilation, how the pupil size affects the ERG must be considered. Interestingly, increasing pupil size in luminance-based tests increases the retinal stimulation, whereas the SCE causes Troland-based tests to overestimate the effective pupil area, thereby reducing retinal stimulation. Thus, the effect of pupil size is in the opposite direction. Without the SCE (e.g., rod-mediated results), Troland stimulation has no dependence on pupil size but the luminance-based tests have a larger effective stimulus with larger pupil sizes. 

Cone-driven flicker ERGs are increasingly being used in clinical research and practice to monitor retinal function in diseases associated with inflammation or ischemia, such as uveitis or diabetic retinopathy. In such conditions, small but measurable changes in ERG waveforms over time can reflect disease activity or response to treatment. When ERG is used quantitatively, rather than qualitatively, pupil-size–related variability may confound interpretation, especially in longitudinal assessments. Therefore, reducing this variability, even modestly, helps improve the reliability of serial ERG recordings in real-world clinical settings. 

Surprisingly, the SCE we found differed between the implicit time of the fundamental component and the other three measurements from the same waveform: the implicit time of the reconstructed waveform and both amplitudes. Although this could be measurement artifact, our uncertainty measures in the mathematical fits generated non-overlapping 95% confidence intervals. Further research is necessary to determine whether this difference is reproducible and to understand the underlying biological mechanism. 

Several limitations should be acknowledged. First, the subjects were visually normal phakic adults, which may not be representative of patients with media opacities or outer-segment abnormalities. Second, although flicker ERGs were recorded under a fixed retinal illuminance, our model was also applied to luminance-based stimuli and rod-mediated ERGs, which may have greater uncertainty due to differing physiological mechanisms. 

Although we expect that the same SCE compensation would apply to all cone-pathway stimuli (e.g., light-adapted 3.0 ERGs), this could be confirmed in a future study. The effect of compensating for the SCE on existing reference data would also be of great interest. Finally, investigating ERGs in individuals with retinal pathologies such as glaucoma and diabetic retinopathy could confirm if these groups have a similar full-field SCE. 

Conclusions

This study shows that cone-mediated ERGs using Troland-based stimuli have a pupil dependency consistent with the SCE, although that dependence is small for the range of expected nonmydriatic pupils (≤7% of the 95% RI). The pupil dependency on dilated luminance-based stimuli is far larger (≤43% of the RI), even with the SCE reducing the effect for cone-based testing. Using Troland-based testing over all pupil sizes has a smaller variation (≤34% of the RI) in pupil size than luminance-based testing for dilated pupils. Using Troland stimuli with ρ_device = 0.05 mm⁻² for cone-mediated flicker ERGs minimizes the impact of pupil size, although the improvement would be modest in non-dilated subjects when using Troland stimulation without SCE compensation. 

Acknowledgments

The authors thank Eriko Uchiyama, Daisuke Kurose, Takumi Ohashi, Iroha Fujimoto, and Yuki Mizuno for performing ophthalmologic examinations on our subjects. We also thank Duco I. Hamasaki, PhD, of the Bascom Palmer Eye Institute of the University of Miami (Miami, FL) for critical discussion and final manuscript revisions. 

Supported by Grants-in-Aid for Scientific Research (C) (18H02954 and 17K19721 to MK) from the Ministry of Education, Culture, Sports, Science and Technology. 

Author Contributions: K.K. designed the study; A.S., M.S., H.M., Y.K., K.M. and K.I. collected the data; K.K., M.K., H.K., E.N., N.Z., and Q.D. analyzed the data; K.K., D.M., N.Z., Q.D. and M.K. wrote the manuscript; Q.D., D.M., K.K., and M.K. proofed the article. 

Disclosure: K. Kato, None; C.Q. Davis, LKC Technologies (E); N. Zeinali, LKC Technologies (E); A. Sugawara, None; H. Matsubara, None; M. Sugimoto, None; Y. Kashima, None; K. Mizumoto, None; H. Kudo, None; E. Nagasaka, None; D.L. McCulloch, None; M. Kondo, None