Electrodiagnostic Tests as Potential Efficacy Endpoints in Clinical Trials of Novel Pharmacological Therapies for Acquired Retinal Disorders

Electrodiagnostic tests (EDTs) provide objective, non-invasive, and measurable indications of retinal and visual pathway function and are used routinely to aid accurate or early diagnosis and to monitor many retinal and visual pathway pathologies [1, 2]. EDTs may also be used to guide or inform titration of treatment, highlighting the potential to evaluate novel pharmacotherapies to prevent irreversible vision loss or disease progression with greater objectivity and specificity than traditional “end-stage” outcome measures such as visual acuity [1].

Historically, the use of EDTs has been limited for various reasons, including greater availability of retinal imaging and tests of retinal structure, such as optical coherence tomography (OCT), and it is generally acknowledged that there may be reduced access to specialist electrodiagnostic centres and expertise [3, 4]. The technical quality of EDT recordings is an important consideration, and robust and large reference ranges (“normative” data) may be difficult to acquire and depend on factors such as the type of electrodes used and patient variables such as age [5‒7]. The International Society for Clinical Electrophysiology of Vision (ISCEV) has worked toward standardization of the main EDT techniques and has defined and published rigorous standards, guidelines, and extended protocols [2, 8‒22], facilitating consistency of recordings and multicentre studies [23, 24]. Notably, since 2012, the number of published papers describing the use of EDTs as specific outcome measures has increased by 60% compared with the previous decade (more than 250 published papers listed on PubMed since 2012). Furthermore, although ISCEV standard testing is often needed for diagnostic, phenotyping, or safety purposes, in some studies of acquired disease, less detailed or abbreviated non-standard electroretinography (ERG) protocols have been used (e.g., to monitor treatment efficacy according to limited or specific electrophysiological parameters) in patients with an established diagnosis. The purpose of this review is to highlight the use of EDTs in assessments of treatment efficacy and to assess their potential value for the assessment of novel drug safety and efficacy in clinical trials in acquired retinal disease.

In full-field ERG (ffERG), an active electrode touching the cornea or lower eyelid is referred to a reference electrode on the outer canthus; alternatively, a contact lens electrode touching the bulbar conjunctiva can be used [9, 25‒27]. ffERG records the global retinal response to light flashes or stimuli. The ISCEV and International Federation of Clinical Neurophysiology standard ffERG protocol includes dark-adapted (DA) and light-adapted (LA) testing and may be used to assess generalized rod and cone system function at the level of the outer and inner retina [2, 9, 28, 29]. ffERG is dominated by contributions from the retinal periphery and is normal if dysfunction is confined to the macula. Other EDTs, such as pattern ERG (PERG) or multifocal ERG (mfERG), are required to assess macular function.

The ISCEV standard ffERG protocol involves six main types of recording [9]. Under DA conditions, the DA 0.01 ERG (flash strength 0.01 cd·s·m⁻²) is a dim flash response that is generated by the rod bipolar cells with no cone system contribution. The DA 3 and DA 10 ERG (flash strengths 3 and 10 cd·s·m⁻², respectively) a- and b-waves (shown in Fig. 1) are mixed rod and cone system responses. The a-wave arises mainly in the rod photoreceptors and the b-wave mainly in the rod bipolar cells, but there is a small DA cone system contribution to both components. The DA oscillatory potentials are wavelets that appear on the ascending limb of the b-wave and are thought to originate from amacrine cells or amacrine cell signalling [2, 30]. Under LA conditions, the 30-Hz flicker ERG (LA 30 Hz) and single-flash cone (LA 3) ERGs (both responses obtained to a flash strength of 3 cd·s·m⁻²) are used to assess generalized (mainly peripheral) cone system function. The LA 30-Hz ERG is “driven” by L- and M-cones, whereas the LA 3 ERG has additional but minor input from the S-cone system; both LA 30-Hz and LA 3 ERGs are dominated by inner retinal contributions. The LA 3 ERG a-wave (shown in Fig. 1) is a cornea-negative potential that depends on cone photoreceptor function but is dominated by the activity of the inner retinal cone Off-bipolar cells [13, 31]. The LA 3 ERG b-wave (shown in Fig. 2) is a summation of activity generated by cone Off- and On-bipolar cells (mostly Off-bipolar cells) downstream of photoreceptors. DA and LA ERG a-wave amplitudes are measured from the pre-stimulus baseline and b-waves from the trough of the a-waves; peak time (implicit time) is the time between the onset of the stimulus and trough or peak being measured [27].

ffERG does not require accurate fixation or refraction, and portable (monocular) ffERG systems are available [32, 33], enabling wider accessibility and convenience. Flicker ERGs have received particular attention in recent studies of cone system function, as no dark adaptation is required, and many responses can be averaged and recorded in a short period. Some systems can partly compensate for small pupils or a lack of mydriasis [32, 34‒36], but with limitations and disadvantages (e.g., monocular testing of both eyes may be time consuming and there are significant caveats regarding compliance with international standards) [9]. Stimuli and background illumination must be adequately adjusted for the non-dilated pupil with ERG waveforms similar to those recorded after mydriasis, and compensation may be required for the Stiles-Crawford effect on cone photoreceptors (e.g., when pupil diameter is >6.5 mm) [36]. Reference ranges must be obtained with the same methods and care taken to maintain consistency of methods in monitoring studies [9].

There are several clinically useful ffERG techniques that have yet to be standardized, including those described in the ISCEV extended protocols [13‒19]. These techniques include the photopic negative response (PhNR) and the photopic On-Off ERG.

The PhNR is a late negative potential manifesting after the LA single flash ERG b-wave and may be enhanced using a red flash on a short wavelength background [13]. The PhNR originates primarily in retinal ganglion cells (RGCs) [25] and may facilitate the detection or monitoring of a wide range of optic nerve pathologies, including common disorders such as glaucoma [25, 37, 38].

The photopic On-Off ERG is evoked by a relatively long-duration light stimulus (typically 150–200 ms) and reflects the activity of the inner retinal bipolar cell pathways. At light onset, there is a corneal positive b-wave, reflecting depolarizing cone On-bipolar cell function, and at stimulus offset, there is a d-wave, shaped mainly by the hyperpolarizing cone Off-bipolar cell pathway [15].

The ISCEV standard PERG is a response to an alternating high-contrast checkerboard [2, 22]. The response is highly sensitive to optical degradation and contact lens corneal electrodes cannot be used. The PERG has two major components. The positive P50 component arises largely (70%) in RGCs, but also partly in more anterior retinal structures [22]. Because this component is “driven” by macular cones, it is widely used to assess macular function, manifesting as a delay and/or reduction in P50 [2, 39]. The negative polarity N95 component arises wholly in RGCs and is widely used to assess RGC function [2] (e.g., in optic neuropathies [40], glaucoma, and, potentially, diabetic macular ischaemia) [41]. In severe cases of optic neuropathy, P50 can additionally be attenuated, resulting from a reduction in the RGC contribution to the P50 peak, but typically associated with abnormal shortening of P50 peak time. A detailed description of the clinical value of PERG P50 and N95 components in visual pathway diagnosis is beyond the scope of this paper but has been reviewed extensively elsewhere [39].

mfERG is used to assess cone system function over multiple discrete areas of the posterior pole [12, 29] and macula and is recorded with corneal electrodes. The stimulus comprises multiple hexagonal elements, each of which is illuminated in a rapid, irregular but predetermined sequence known as an “m-sequence.” Waveforms associated with each hexagon are extracted from an ongoing recording to the modulated stimulus array, according to the time that each hexagon is illuminated, using a cross-correlation algorithm. The ISCEV standard mfERG technique uses a stimulus array of 61 or 103 hexagons, subtending a visual angle of 40–50°.

The characteristic mfERG waveform comprises an early, negative deviation (N1), followed by a positive deviation (P1), and then a second, negative deviation (N2) [12]. The waveform is considered to derive from mixed On- and Off-bipolar cell contributions together with a smaller input from photoreceptors [12, 42]. The mfERG components are of low amplitude (typically measured in nanovolts) primarily derived from two retinal layers – the outer plexiform-bipolar bed and photoreceptors [29]. International standard mfERG data are presented as waveform trace arrays, illustrating responses associated with each hexagonal stimulus element, but may also be shown as group or “ring” averages for concentric hexagonal rings of different eccentricities (useful if abnormalities show radial symmetry), as quadrants, or as 3-dimensional scalar plots of response density [12]. Importantly, the trace array is essential and must always be included, as technical quality can be judged, localized abnormalities can be more easily visualized, and the other display modes may be misleading in isolation [12].

The visual-evoked potential (VEP) records the electrical signals produced by the occipital cortex after visual stimulation and is used in the assessment of optic neuropathy and visual pathway disease [29, 40, 43]. The ISCEV standard specifies three possible modes of stimulation suitable for eliciting VEPs: pattern-reversal VEPs (reversal rate 2/s); pattern-onset/-offset VEPs, in which the checkerboard stimulus is typically “on” for 200 ms and “off” for 400 ms (alternating between zero and a finite level of contrast with mean luminance constant); and flash VEPs [11]. Pattern-reversal VEPs are generally more sensitive than other forms of VEP to most causes of optic nerve dysfunction.

The ISCEV standard electro-oculogram (EOG) records the standing electrical potential between the cornea and back of the eye [10]. The standing potential across the retinal pigment epithelium (RPE) results in a dipole across the globe, with the cornea being positive with respect to the back of the eye. The dipole is quantified by recording 30-degree horizontal eye movements under conditions of dark and then light adaptation. The horizontal eye movements result in positive and negative polarity recordings, the amplitudes of which are minimized in the dark (the dark trough) and increased in the light (the light peak). The light peak to dark trough ratio, historically known as “the Arden ratio,” is used to provide a measure of generalized RPE function [10, 44]. The EOG has several clinical applications, including a role in helping to confirm or exclude BEST1-related RPE dysfunction. The EOG has also been used in assessments of a number of acquired retinal disorders, including acute zonal occult outer retinopathy (AZOOR), retinal toxicity, changes related to diabetes and diffuse chronic chorioretinal inflammation, and retinal detachment [45, 46], and was used in a Phase I clinical study of age-related macular degeneration (AMD) [47]. It is highlighted that the EOG is most useful in assessing generalized RPE function when ERGs are normal or relatively well preserved; the EOG depends on the function of rod photoreceptors, and if rod-mediated ERG components are reduced, there is typically proportionate reduction in the light peak to dark trough ratio.

The clinical use of EDTs has evolved and expanded over several decades, facilitated by the standardization of methods and interpretation, enabling meaningful inter-laboratory comparisons and multicentre studies. Visual acuity is often the main surrogate marker for clinical trials in ophthalmology, and although convenient and easy to measure, it has long been recognized that it may not be the most appropriate outcome [48], given that it provides a subjective and limited assessment related to central visual resolution and is dependent on the function of the whole visual pathway. It is also highlighted that central vision may be preserved or normal in the presence of severe (paracentral or peripheral) retinopathy (i.e., it cannot be used to monitor efficacy of treatment if normal or near normal at baseline). Inflammatory and other retinopathies may also be associated with conditions such as cystoid macular oedema, affecting visual acuity but responding differently to treatment [49]. The type of index selected (e.g., ffERG for assessment of the entire retina, PERG for the macula and macular RGCs, mfERG for focal areas within the macula or posterior pole) partly depends on the visual pathway disorder and goal of testing [48]. The following sections highlight some of the more common acquired retinal disorders in which visual electrophysiology has proved informative and pertinent to the monitoring of therapeutic interventions, including some studies that have employed unconventional or non-standard methods or protocols.

The value of ffERG in the assessment of central retinal vein occlusion (CRVO) is long established [50‒52]. Reduction in the scotopic ERG b:a ratio (with preservation of the largely rod photoreceptor-driven ERG a-wave), increased b-wave peak time, the ERG stimulus response function [53], and cone system-mediated ERG peak times have been used to assess the severity of inner retinal dysfunction, which is worse in ischaemic than in non-ischaemic CRVO. The risk of neovascular complications and visual prognosis is related to the severity of ischaemia, and studies have shown that flicker ERG [33, 54, 55] correlates with the vascular endothelial growth factor (VEGF) concentration in the aqueous of CRVO [56], and that flicker ERG peak times may predict the development of rubeosis (neovascularization of the iris) and neovascular glaucoma. ffERG amplitudes and cone-mediated ERG b-wave peak times have been shown to vary in the acute phase of CRVO, with some reports that the optimal predictive value for the detection of rubeosis is after 3 weeks from onset [57].

ERG may also be used to assess recovery from inner retinal ischaemia and to monitor treatments, with several recent investigations focussing only on flicker ERGs and recorded using handheld or portable contact lens devices. In patients with CRVO and macular oedema undergoing intravitreal injection of anti-VEGF therapy, flicker ERG peak time was shown to be substantially prolonged (worsened) 2–24 h after treatment [35], highlighting a role in the acute assessment of drug safety in both diseased and unaffected fellow eyes [34]. Flicker ERG delays were also associated with enlarged responses (≥117% of unaffected fellow eyes) in some cases [58], with reduction in amplitude demonstrated following anti-VEGF therapy, further highlighting how ERG may be used to assess treatment efficacy and restoration of retinal function.

Photopic ERG b-wave and 30-Hz flicker ERGs to light stimuli delivered using a portable contact lens device have been shown to correlate with retinal thickness and retinal volume in CRVO [59]. Moreover, in patients with CRVO who received intravitreal injections of anti-VEGF therapy for 1 year, improved retinal structure and function were evident; for example, central retinal thickness and total macular volume were significantly reduced, and significant enlargement was observed in the PhNR and ERG b:a ratio [60].

Diabetic retinopathy (DR) is one of the leading causes of blindness in the adult population. The diagnosis is usually made on fundal appearance and fluorescein angiography, distinguishing the stages of non-proliferative from proliferative disease. Patients are often asymptomatic in the early stages and evidence of retinal disease may be detected on routine diabetic screening.

The clinical management of DR is aided by the use of PERG, mfERG, and ffERG testing. A recent review on the use of ERG in the evaluation of DR [3] describes methods of early detection, disease monitoring, and assessment of neural dysfunction. Historically, it was thought that electrophysiological evaluation of DR was only useful in advanced stages of proliferative disease, focussing on selective reduction of the oscillatory potentials [61‒63], first described by Cobb and Morton [64] and suggested to indicate early deterioration of the neuronal synaptic activity of the amacrine cells [30]. Studies of larger patient groups, with both non-proliferative and proliferative DR, revealed that in addition to oscillatory potentials, flicker ERGs could be early indicators of disease [65].

Prior to the onset of clinical signs, mfERG peak times have been reported to predict, with 86% sensitivity and 84% specificity, the development of new retinopathy in regions of the posterior pole [66]. Several studies have described how changes in mfERG features such as response density, peak times, and amplitudes may be precursors to the development of proliferative DR [67‒69]. A model based on the mfERG findings has been proposed for predicting the development of DR and for identifying those “at risk,” including potential candidates for clinical trials of new treatments.

PERG has also been used in the assessment of macular function in patients with diabetes, with the P50 component being used to assess macular cone photoreceptor function and the N95 macular RGC function. In patients with diabetes without any clinical signs of DR, including macular oedema, P50 can be subnormal and delayed compared with age-matched controls [70]. In addition, subsequent delays and increased amplitude of the N95 component correlated with arterial flicker-induced responses, thereby suggesting that in patients with diabetes, early neural and neurovascular dysfunction occurs before the onset of clinically evident DR [70]. In patients with diabetic macular oedema (35 eyes) treated with a single intravitreal injection of bevacizumab, improvements were evident in both the PERG P50 wave amplitude and visual acuity at 1- and 3-month post-injection [71].

More recently, handheld devices for ERG assessment, alongside structural imaging, have been used in the screening of DR in patients with diabetes without retinopathy [72, 73]. Studies have identified strong correlations between DR severity and ffERG amplitude, similar to that seen with conventional devices, potentially facilitating the diagnostic use of ffERGs in diverse clinical settings. However, definitive determination of the full promise of all subtypes of ffERG in DR (and not just of photopic 30-Hz flicker ERG, for example) requires large trials involving patients with varying degrees of DR severity [74]. Motz et al. [75] reported that delays in dim-flash oscillatory potentials with a handheld ERG device could facilitate the early detection of preclinical DR in patients with diabetes; moreover, 2 weeks of levodopa treatment (levodopa is neuroprotective in experimental models of DR [76]) normalized oscillatory potential peak times and reversed incipient retinal dysfunction [75].

In the EUROCONDOR trial, the largest clinical trial assessing mfERG in patients with diabetes, mfERG peak times were used to evaluate the efficacy of eye-drop formulations of the neuroprotective agents brimonidine and somatostatin in >400 patients with type 2 diabetes. In the subgroup of patients with pre-existing retinal neurodysfunction (i.e., changes in peak time at ≥6 locations), the neuroprotective agents prevented retinal disease progression as assessed by mfERG [77]. There are several reports of VEPs being abnormal in DR, and although attributed to neuronal damage, it is not always clear how macular dysfunction or sub-clinical maculopathy was excluded as a cause of VEP abnormality.

Early electrophysiological assessment of RGC function in glaucoma focussed on the PERG [78], and in particular the N95 component, which is generated entirely by the RGCs in the central macular region [79]. In patients with established or suspected primary open-angle glaucoma (POAG), thickness of the retinal nerve fibre layer correlates with a reduction in PERG N95 [80], and there may be P50 amplitude reduction and shortening of the P50 peak time. A number of studies have shown that these measures can precede detectable visual field defects [81, 82].

Parisi [83] reported a statistically significant correlation between abnormal PERG parameters [78] (delayed implicit times and reduced amplitudes) and retinal nerve fibre layer thickness in patients with POAG, whereas a more recent study described marked deterioration of the PERG several years prior to significant retinal nerve fibre layer loss, thereby proposing it to be a potential prognostic indicator or index for early intervention. In early glaucoma, PERG N95 amplitude has been shown to correlate with macular vessel density, in contrast to OCT measures of retinal nerve fibre RGC layer thickness and static perimetry [84].

Steady-state PERGs to medium to high frequencies lack the P50 and N95 components that characterize transient responses, but have been shown to be more sensitive to glaucoma than the transient PERG [85]. Steady-state PERGs have been used to predict conversion from ocular hypertension to glaucoma and can detect RGC dysfunction prior to loss of the nerve fibre layer [86]. Diagnostic performance may be further enhanced by combining steady-state PERG with structural assessments such as OCT [87]. It is highlighted that steady-state PERGs may be recorded to large stimulus fields and may conveniently allow relatively rapid signal averaging.

The more recent emergence of the PhNR [88, 89] in clinical practice has provided a measure of global (generalized) RGC function [85, 86]. The clinical applicability of the PhNR and PERG differs and may be influenced by the disorder under investigation, the PERG being more appropriate for disorders predominantly affecting the papillomacular bundle, and practical considerations such as the PERG requiring refractive correction and good fixation. The PhNR requires mydriasis but is significantly larger than the PERG (higher signal:noise ratio reduces the time taken for signal averaging), does not depend on good fixation or refraction, and may be used in cases with mild media opacity. The amplitude of the PhNR has been shown to correlate with glaucoma-induced clinical severity, visual field loss, and thinning of the retinal nerve fibre layer [90, 91], and in patients with suspected glaucoma with reductions in peripapillary and macular nerve fibre layers [92]. Comparative studies of the PERG and PhNR have shown similar sensitivity in detecting RGC dysfunction in suspected and early glaucoma, with the PhNR ratio reported to have higher sensitivity in pre-perimetric patients [92, 93]. PhNR methods have also been modified to examine local macular and posterior pole areas (e.g., focal flash stimulation in patients with POAG) and showed correlation between spectral domain-OCT measures of RGC layer thickness, which was greater for a central focal flash than for annular paracentral stimuli [94].

The ISCEV extended protocol for the PhNR recommends the use of a red flash on a blue background and specifies recording parameters and an analysis convention. However, a recent study of a large heterogeneous cohort of 200 patients compared such responses with the PhNR component of the ISCEV standard LA 3 (single white flash) ERG and showed only slightly less sensitivity and almost identical specificity [95]. The study highlighted the value and potential convenience of using white-on-white stimuli, already used for routine (LA 3) ERG assessment. Additionally, the development of handheld devices has widened the scope of potential use of the PhNR in the clinical setting, offering the possibility of wider availability and early detection in more patients [96].

The standard mfERG (involving the first order kernel) and PERG P50 component may be used to assess macular function, but unlike the PERG, the standard mfERG has little or no direct value in the detection of glaucoma [85]. However, modified and non-standard methods have been suggested [93, 97‒100], and a study comparing the diagnostic performance of a slow sequence mfERG with PhNR and PERG in a non-human primate model of experimental glaucoma demonstrated that the mfERG yielded the highest diagnostic specificity and highest correlation with OCT measures of retinal nerve fibre [101]. In 2003, Hood et al. [102] reported that multifocal VEPs were linearly associated with RGC loss and glaucomatous damage; however, subsequent studies [103, 104] comparing multifocal VEP and standard automated perimetry in high-risk ocular hypertension and early glaucoma [104] concluded that although multifocal VEPs may be useful in monitoring disease, findings that parallel abnormalities already demonstrated on static perimetry.

The PERG and pattern-reversal VEP have been used in the monitoring of treatment for glaucomatous eyes in 30 patients with intramuscular citicoline 1 g/day (administered in 16 2-month “blocks” over 8 years). The PERG and pattern-reversal VEP P100 parameters showed significant improvement compared with pre-treatment measures and the placebo group [105]. In 2007, Kuba et al. [106] highlighted that motion-onset rather than pattern-reversal VEPs were more sensitive for the detection of glaucoma (sensitivity: 73–77% vs 33%), and in 2012, Kiszkielis et al. [107] outlined that focal rather than full-field PhNR of the flash ERG may be a more sensitive and specific method for glaucoma diagnosis.

ffERG abnormalities have been reported in some patients with AMD, suggesting generalized retinal involvement far beyond the area of visible maculopathy, but findings are contradictory [108‒110], and the importance of using closely age-matched reference values is highlighted [5, 6]. A study of inactivation of rod phototransduction using a double-flash ERG paradigm revealed no abnormalities in early AMD, but there was slowed inactivation kinetics in a group including the oldest individuals with late-stage disease [111]. Several disorders can also masquerade as AMD [112, 113], and the increasing availability of genetic testing may help resolve some of the controversies relating to the use of ffERGs in this patient group. There remains a role in safety monitoring of novel therapeutic interventions [47].

Tests that assess macular function have proven of particular importance, given the focal nature of age-related macular lesions. An early study of PERG to different check sizes showed correlation with visual acuity and AMD lesion size [114]. PERGs have also been shown to have a predictive value and to improve in patients treated with photodynamic therapy (PDT) [115] and have been used to demonstrate stability objectively in patients with neovascular AMD treated with monthly intravitreal injections of ranibizumab over 6 months [116].

mfERGs offer the advantage of higher spatial resolution and greater sensitivity than the PERG to paracentral macular dysfunction in AMD [115] and may detect dysfunction in the absence of visual acuity loss [117]. mfERGs have also been used widely to assess efficacy and safety of treatments, including PDT and anti-VEGF injections [118, 119], and may be of predictive value prior to PDT [120]. The value of mfERG was highlighted in early studies that revealed focal areas of response improvement and worsening following PDT, including parafoveal changes in patients with stable visual acuity [121]. Several other studies showed discrepancies between subjective and mfERG measures, including a recent 12-month study of 76 patients with wet AMD treated with ranibizumab, which revealed small but statistically significant reduction in mfERG response densities in spite of improvements in visual acuity and reductions in macular thickness [122]. In 15 patients with unilateral wet AMD, intravitreal ziv-aflibercept significantly improved mfERG amplitudes at 6 months within the central area, showing a linear association between P1 amplitude and visual acuity in bevacizumab-treated eyes with AMD [123].

The effects of dietary supplements on AMD have also been monitored with mfERG. A placebo-controlled study of 30 patients with intermediate AMD, involving supplementation with lutein, zeaxanthin, and antioxidants, resulted in significantly increased mfERG response density over central macular areas in the absence of structural changes [124]. Two similar supplementation studies with lutein and zeaxanthin revealed a significant increase in central mfERG response density in patients with early AMD [125, 126], highlighted by a recent meta-analysis involving both studies [127].

Although lacking the spatial resolution of mfERGs, various focal flash ERG techniques to assess only the macula or posterior pole have been used to assess the severity of macular dysfunction in AMD [128] and have shown improvements following treatment (e.g., with bevacizumab). Focal flash ERG improvements have been shown to be consistent with improvements in visual acuity and macular thickness. Structure-function correlations have also been established in a retrospective study of 74 patients with early or intermediate non-exudative age-related maculopathy; focal ERGs showed correlation with visual acuity, with micro-anatomic disruption of the outer retina-RPE complex and the area of RPE atrophy, with the latter predicting progression at 12 months [129].

AMD has been shown to have a considerably greater effect on macular rod than cone sensitivity [130], and a limitation of PERG, mfERG and focal flash ERGs is that all three types of recording are “driven” by and are dependent on cone system function. Rod-mediated mfERGs to assess macular rod system function have been developed [131] and have revealed abnormalities in AMD, including delays in the P1 component [132, 133]. However, such responses are relatively small, require longer recording periods than for photopic testing, and are prone to fixation errors and “drift” artefact [131, 132]. The applicability of such a technically demanding method to routine testing and monitoring is doubtful. It is also highlighted that neither rod-mfERGs nor photopic focal flash ERGs are standardized, restricting inter-laboratory comparisons and potentially limiting use as a future outcome measure.

Birdshot chorioretinopathy is a chronic posterior uveitis with unpredictable periods of exacerbation and remission, and ERGs are a valuable monitoring tool and aid to clinical management [134, 135]. ffERG reveals a high incidence of generalized retinal dysfunction that cannot be inferred reliably from clinical signs and symptoms [49, 136, 137]. Visual acuity may be normal or minimally affected in the presence of severe worsening of retinal function and may be affected by macular oedema, irrespective of whether there is ffERG abnormality [49, 136]. DA ffERGs often show evidence of inner retinal involvement, but LA 30-Hz peak times are generally the most sensitive ERG indicators of inflammatory retinal dysfunction [49, 137‒139] and have been used to monitor efficacy and to inform initiation, increment, and tapering of immunomodulation and steroids [49, 137]. A recent study on the use of tacrolimus in 25 patients with birdshot chorioretinopathy monitored over approximately 6 months revealed a good safety profile and a stable or improved ERG in most cases [140], and a recent evaluation of intravitreal implants of fluocinolone acetonide in 15 eyes showed improved ffERG in most patients over an average follow-up of 31 months, also highlighting the value of the LA 30 Hz as a robust measure of inflammatory retinal dysfunction [141]. A recent study of 32 patients with birdshot chorioretinopathy compared conventional LA 30 Hz recorded with corneal electrodes with flicker ERGs obtained using a portable device and skin electrodes and showed strong correlation between methods, highlighting the potential of wider accessibility for ERG assessment and screening [142].

There is a high incidence of macular involvement and cystoid macular oedema in birdshot chorioretinopathy [134]. PERGs are generally recognized as being more sensitive to the effects of cystoid macular oedema than mfERGs [2], but both tests have been used to assess and monitor macular dysfunction in birdshot chorioretinopathy [140, 141, 143], which is important given that macular function can worsen or improve with or without corresponding changes in subjective vision assessments [49, 141, 144].

The potential value of visual electrophysiology in the management of other forms of posterior uveitis has long been recognized [135, 145], and recent studies have emphasized the sensitivity of LA ERGs and additional insight provided by ERG in a range of inflammatory retinal disorders in both children and adults, including cases in which inflammation may appear to be quiescent [146].

Autoimmune retinopathies (AIRs) are defined as retinal disease caused by anti-retinal antibodies [147] and fall into two main categories: paraneoplastic and non-paraneoplastic. AZOOR shares features with the latter (see below) and is often associated with or precipitated by “white-dot” retinopathies, including multiple evanescent white-dot syndrome, punctate inner choroiditis, and multifocal choroiditis [148, 149]. There is significant overlap in the presentation, proposed pathogenesis, and treatment of these disorders, and diagnosis is made based on a careful history and clinical examination, including electrophysiology, multi-modal imaging, and retinal antibody testing [150‒153]; however, it has been highlighted that many individuals without AIRs have circulating anti-retinal antibodies [154]. Electrophysiological investigation has proven to be invaluable in the diagnosis and monitoring of AIR progression [155‒157], with varying degrees of cone and rod dysfunction at presentation, which progresses on long-term evaluation.

A prospective non-randomized clinical trial for the treatment of non-paraneoplastic AIR with intravenous rituximab was carried out by Armbrust et al. [158]. Three out of 5 patients, following treatment, showed disease stabilization on ffERG testing and visual fields, whereas the remaining 2 patients showed progressive deterioration in one or both outcome measures, with deterioration of the ffERG confined to the cone system. The trial concluded that there was no clinically significant improvement with rituximab treatment, consistent with a previous case series of 5 patients showing ERGs that stabilized or worsened [159]. Maleki et al. [160] retrospectively analysed the long-term follow-up of 6 patients with non-paraneoplastic AIR following treatment with rituximab as a monotherapy (2 patients) or combination therapy (4 patients). In total, 75% of patients treated showed improvement in visual acuity, visual fields, or ffERG, with most (66%) demonstrating stability or improvement in ffERG parameters; the 2 patients who showed deterioration in ffERG had treatment withdrawn due to adverse events and personal reasons.

First described by Gass [148], AZOOR has been considered within the spectrum of non-paraneoplastic AIR [153] and is characterized by varied and often severe ERG abnormalities in seminal studies [148]. AZOOR has often been treated with systemic corticosteroids, although visual outcome can vary and in some cases the disorder may resolve or be self-limiting [161]. Saito et al. [162] showed improvement in final best-corrected visual acuity compared with pre-treatment best-corrected visual acuity in 21 eyes of 14 patients following oral and intravenous steroidal therapy. Electrophysiology measures correlated with best-corrected visual acuity at initial presentation; however, limited post-treatment analysis of mfERG and ffERG measures suggested no improvement or correlation with post-treatment best-corrected visual acuity. Similarly, Chen et al. [163] in the same year (2015) demonstrated marked improved in post-treatment visual acuity and visual fields in all of 9 patients treated with systemic steroids. Although mfERG and ffERG were performed at presentation and follow-up, the authors do not comment on these post-treatment outcomes. Other studies have concluded varying visual outcomes, and alternative forms of treatment including immunosuppressants, antiviral drugs, and intravitreal injections of anti-VEGF have been used with similarly varying effect [164‒167]. In 2020, Guijarro et al. [168] reported the case of a female patient aged 42 years presenting with acute photopsias and diagnosed with AZOOR following ffERG assessment. The patient was treated with pulse methylprednisolone therapy followed by azathioprine and the efficacy of treatment was monitored over 5 years using ffERG. The ffERG, particularly the 30-Hz flicker ERG, continued to deteriorate during this period, demonstrating significant disease progression. Subsequently, the patient was tested for anti-retinal antibodies and was found to be positive against the carbonic anhydrase II antigen. A revised and secondary diagnosis of AIR was given, and treatment commenced with intravenous immunoglobulin therapy. The ffERG continued to deteriorate despite the combined therapy.

Paraneoplastic AIRs include carcinoma- and melanoma-associated retinopathy (CAR and MAR). The ffERG abnormalities seen in CAR primarily reflect photoreceptor dysfunction, to varying degrees, although rare cases of inner retinal dysfunction have been described [155, 169]. The severity of ffERG abnormality is often not explained by clinical examination, making ffERG assessment essential for diagnosis and prognosis. CAR can be associated with many different cancers [156, 170, 171] and is often a diagnosis of exclusion. As several cancers may exhibit CAR, it is more commonly encountered than MAR. The heterogeneous nature of the disorder, including varying severity of the ffERG, can make management difficult, and currently, no standardized approach or protocol exists for the treatment of CAR. Systemic steroids can be used for treatment, but they have variable outcomes, and the vision loss may progress despite therapy. Safadi et al. [157] retrospectively analysed the long-term follow-up of 9 patients, five of whom were known to have cancer, following treatment with immunomodulatory therapy and plasmapheresis. All patients demonstrated ffERG abnormalities in both eyes at presentation, and these abnormalities stabilized in 64% of eyes at a mean follow-up interval of 63 months. A review by Brossard-Barbosa et al. [172] included a case report of a patient aged 70 years with CAR who showed significant subjective improvement in vision after treatment with high doses of systemic corticosteroids, followed by plasma exchange, although electrophysiology was not reported following treatment. The review also highlighted 3 anti-recoverin-positive patients who underwent ffERG testing prior to plasma exchange or steroid treatment; however, all 3 patients had poor final visual outcomes.

MAR is associated with malignant melanoma, and ffERGs are pathognomonic for generalized rod and cone On-bipolar cell dysfunction, resulting from anti-TRPM1 retinal antibodies [173, 174]. Identical ffERG features are associated with the inherited disorder complete congenital stationary night blindness, including an autosomal recessive form consequent upon biallelic variants in TRPM1. It is documented that patients with MAR may have antibodies that are protective against cancer spread, and systemic immunosuppression may risk development of metastases. This has prompted different approaches to treatment including cases studies involving the use of intravitreal slow-release corticosteroid implants, resulting in dramatic recovery of On-bipolar cell function, evident on ffERG testing, including On-Off ERGs, together with improvements in visual fields and visual acuity [175].

Krebs et al. [176] performed a study on 20 eyes treated with PDT for severe myopia. mfERG stabilization was reported in 55% of eyes after 1 year. In 2010, a pilot study of intravitreal diclofenac in 10 patients with macular oedema of various causes revealed no significant changes from baseline in ERG a- or b-wave amplitude [177]. In 2012, Kandel et al. [178] outlined that mfERG is a sensitive test for evaluating the ocular toxicity of ethambutol in patients with tuberculosis; in 88 eyes (44 patients), although N1 amplitude and latency were not significantly altered in any ring, P1 amplitude was significantly (p < 0.05) reduced and P1 latency significantly (p < 0.05) increased in all rings after ethambutol use. These isolated studies highlight the potential wide applicability of EDTs in the treatment of retinal and ocular disorders.

Advances in the understanding of retinal pathophysiology have led to the development of novel therapeutic interventions for retinal disorders, but accepted functional endpoints in clinical trials are often limited to psychophysical measures such as visual acuity or visual fields [179]. Important considerations are that visual acuity may be preserved in cases of severe retinal disease and may be of limited value if severely impaired due to severe macular dysfunction or macular atrophy. Psychophysical assessments may also be confounded by subject compliance difficulties and response variability [27, 180], and there is a well-recognized need to develop better and accessible outcome measures in clinical trials [181, 182].

Patient symptoms and retinal/fundus imaging studies do not reliably reflect the nature or severity of retinal dysfunction, and EDTs are often fundamental to diagnosis and phenotyping. EDTs have played an important role in the selection of candidates for gene therapy for inherited retinal diseases [183, 184] and have provided objective functional measures of safety and efficacy in pioneering therapeutic interventions [24, 182, 184]. A disadvantage of EDTs is that responses depend on the summed activity of hundreds of thousands of cells to suprathreshold stimuli, limiting sensitivity to small but potentially significant changes in retinal or macular function, detectable by only specialized psychophysical methods [185]. EDT methods used to assess acquired retinal disorders vary, and there is inter-session variability, but international standards developed over the past few decades have helped establish consistency of methods and interpretation. Furthermore, validation, training, and centralized ERG readings have demonstrated the value and suitability of EDT endpoints in large multicentre studies [23].

International standards for ffERG endorse the use of skin electrodes as an optional and less intrusive alternative to corneal electrodes, and the most recent 2022 standard accepts that comparable, consistent retinal illumination is possible without mydriasis, subject to important caveats [9], highlighting the potential and convenience of using portable devices with real-time compensation for undilated pupils. Such methods are advantageous in patients or studies in which mydriasis is contraindicated [12, 34‒36, 58, 75], and would increase convenience and accessibility. The ISCEV standard protocols remain of prime importance for routine diagnostic and clinical purposes, but ISCEV has also suggested a shorter ffERG protocol to encourage convergence of methods when standard testing is not possible. These methods may be less suitable for safety studies, but even limited ffERG protocols are likely to provide objective and added insight when used for monitoring purposes or as additional surrogate endpoints, provided the ERG parameters of interest are suitably verified as an appropriate biomarker for the disorder under investigation. For example, LA flicker ERGs may have a valuable role in evaluating drug efficacy in CRVO, diabetic macular oedema, diabetic macular ischaemia, and posterior uveitis [34, 35, 58, 59], and PERG and mfERG may be a future fundamental endpoint in clinical trials in patients with AMD [118, 124] or retinopathies with macular involvement [186].

It is highlighted that time-series data such as EDTs are amenable to interrogation using artificial intelligence, and machine learning approaches have already proved informative in ffERG studies [4]. Artificial intelligence offers the potential of finding implicit relationships within large serial datasets, with the future possibility to interrogate combined electrophysiological and multi-modal data, relevant to clinical trials.

As in many inherited retinal diseases, acquired retinopathies can vary in severity, but EDTs have shown that in some acquired and inflammatory retinopathies, there may be stabilization or even marked recovery of function following treatment, not predictable from routine psychophysical measures and retinal imaging assessments. This further highlights the potential value of EDT endpoints in clinical trials and drug development studies.

Visual EDTs provide an objective and potentially definitive index of the efficacy and safety of novel neuroprotective therapies in acquired retinal disorders, including CRVO, DR, glaucoma, AMD, and uveitis. Although EDTs require further assessment before they become universally accepted as standard outcome measures of drug efficacy by regulatory authorities, such tests provide an opportunity to objectively evaluate drug efficacy across short periods, in which conventional psychophysical and retinal imaging measures are less sensitive and less informative.

In the preparation of this manuscript, David Murdoch, BSc (Hons) and Tom Priddle, DPhil, of OPEN Health Communications (London, UK) and Tabasum Mughal, PhD, of HCG (London, UK) provided editorial support.

Dr. Magella Neveu has no conflicts of interest to declare. Dr. Hendrik Scholl is supported by the Swiss National Science Foundation (Project funding: “Developing novel outcomes for clinical trials in Stargardt disease using structure/function relationship and deep learning” #310030_201165), the Wellcome Trust (PINNACLE study), and the Foundation Fighting Blindness Clinical Research Institute (ProgStar study). Dr. Scholl is a chief medical officer of Belite Bio, but this work is performed outside of his activities for Belite Bio and not endorsed by Belite Bio. He is also a director of Bioptima AG. He is a member of the Scientific Advisory Board of Boehringer Ingelheim Pharma GmbH & Co, Janssen Research & Development, LLC (J&J Innovative Medicine), Kerna Ventures, Okuvision GmbH, and Tenpoint Therapeutics. Dr. Scholl is a member of the Investment Advisory Board of Droia NV. Dr. Scholl is a member of the Data Monitoring and Safety Board/Committee of ViGeneron (NCT06291935) and an adviser of the Steering Committee of Novo Nordisk (FOCUS trial; NCT03811561). Dr. Hendrik Scholl was a member of the journal’s Editorial Board at the time of submission. Dr. Theo Empeslidis is employed by Boehringer Ingelheim. Prof. Victor Chong is an employee of Johnson and Johnson (J&J), but this work is performed outside J&J employment and not endorsed by J&J. Prof. Chong is an employee of Clearside Biomedical Inc., but this work is performed outside Clearside employment and not endorsed by Clearside. Prof. Anthony Robson has no commercial disclosures but is supported by the National Institute for Health Research Biomedical Research Centre based at Moorfields Eye Hospital NHS Foundation Trust and UCL Institute of Ophthalmology (NIHR203322).

Editorial support in the preparation of this manuscript was provided by OPEN Health Communications and HCG. The study and manuscript editorial support were funded by Boehringer Ingelheim.

Magella M. Neveu and Anthony G. Robson: writing – review and editing, visualization, and supervision. Victor Chong: conceptualization, methodology, revising draft, and funding acquisition. Theo Empeslidis: conceptualization, writing – review and editing, and supervision. Hendrik P.N. Scholl: conceptualization, investigation, writing – review and editing, and supervision. All authors meet criteria for authorship as recommended by the International Committee of Medical Journal Editors (ICMJE) and made the decision to submit the manuscript for publication.