Effect of the electrode array-retina gap distance on visual function in patients with the Argus II retinal prosthesis

Patients enrolled in the Argus II Post-Approval Study (PAS) were invited to join this study. Eligibility criteria included a diagnosis of outer retinal dystrophy, some residual light perception, ability to undergo surgery under anesthesia, and ability and willingness to comply with post-operative follow-up and testing. Exclusion criteria included co-morbid ocular conditions (e.g. advanced glaucoma, optic glioma) that could compromise the functional optic nerves and tracts. Pretesting at the baseline visit included visual function testing, general slit lamp examination, Optical Coherence Tomography imaging (Cirrus HD-OCT Machine, Carl Zeiss Meditec, Dublin, Calif.) and fundus photography. A total of 5 patients with implantation of the Argus II device were enrolled in the study (Table 1). All patient recruitment, study investigations and surgeries were conducted at Stony Brook University Hospital and its affiliated sites. Surgeries were performed by one vitreoretinal surgeon under general anesthesia at the Stony Brook University Hospital.

Data was collected from patient visits at baseline prior to implantation, post-operative month 1 (M1), month 3 (M3), month 6 (M6), and year 1 (Y1). At M1, activation and programming of the device took place and visual function testing was deferred. Visual function testing was performed at M3, M6 and Y1. Sensitivity detection thresholds (to be defined later) were collected when possible at M1, M3, M6 and Y1. Gap distances were measured from Optical Coherence Tomography (OCT) images at M1, M3, M6, and Y1. All data was collected by two authors, cross-checked for accuracy and thoroughness, and independently reviewed by two other authors. Microsoft Excel was utilized for data collection.

The visual function of patients with the Argus II Retinal Prosthesis was assessed using Square Localization (SL) and Direction of Motion (DOM) measures which were developed by Second Sight. SL involves locating and touching a white square set against a black background on a touchscreen monitor with the device ON and OFF. The distance (measured in number of pixels) between the true center of the square and the location at which the patient touches the monitor where they perceive the center of the square to be is measured, with smaller pixel distances indicating higher accuracy [2]. DOM testing utilizes a white bar that moves across a black screen in a particular direction, and the patient attempts to determine the direction of motion of the white bar. The angle between the true direction of motion and the patient’s perceived direction of motion is measured, with smaller angles indicating higher accuracy [2] All testing was performed under controlled conditions and overseen by trained personnel. Visual function data was extracted retrospectively from the Argus II Post-Approval Study.

The Argus II device is programmed at M1 for initial programming prior to turning on the camera in order to find stimulation levels that are suitable for the patient to be set on their Video Processing Unit (VPU) [4, 5]. During the first session, the electrodes whose resistances are too high are disabled and the electrodes that are functional and usable are determined. A quick scan of the array is done using various stimulation amplitudes to determine which electrodes are able to produce a percept or phosphene. The minimum current that is needed for the patient to see a phosphene 50 % of the time is defined for each electrode [1, 4]. The video signal from the camera is mapped to the electrical signal for individual or groups of electrodes, determining which electrodes are being stimulated and at what respective frequency. These values are set to the patients VPU to be used in different conditions [1]. Programming for this study included use of the “Programming Assistant” software that is designed to streamline the programming process to measure general sensitivity [1, 4]. Through use of this software, the camera alignment process is simplified and comfort is determined under more real-world stimulation conditions [1]. Depending on the patient experience, the device can be reprogrammed on a regular basis for a readjustment in response to changes to the array or the patient’s responses to the electrode stimulation [5, 6]. The brightness of perception and the number of electrodes that give the patient a perception may decrease over time. Patients may need several reprogramming sessions in order to fully optimize the use of the device [5]. The sensitivity detection from each programming session for each patient was provided by from the programming technician from Second Sight where the values for sensitivity (μA) of each electrode for each session and the maximum current (μA) of each electrode per stimulation were collected and analyzed. Unfortunately, sensitivity detection thresholds were not collected for every patient at the scheduled sessions due to various logistical issues.

A Cirrus HD-OCT Machine was utilized by the PAS study to capture images during each patient visit. Parameters for capture at each visit included a 6 mm × 6 mm macular cube image (acquiring 512 nasotemporal scans × 128 superoinferior scans) which were utilized for study purposes. Images were located using unique patient identifiers and analyzed using Zeiss OCT analytical software. Eligible scans included a majority of the 60 electrodes and a surrounding portion of retina, the ability to visualize or estimate the precise location of each electrode, and a signal strength of at least 5 out of 10 (determined by the Zeiss software). While in the macular cube view of the software, the perpendicular nasotemporal and superoinferior raster lines were placed directly over the center of each visualized electrodes (Fig. 1). On the associated horizontal and vertical tomograms, the built-in caliper measurement tool was used to measure the vertical distance (in μm) between the electrode array and retina (Fig. 2). These measurements were taken for each patient at M3, M6 and Y1. The electrodes were designated a label (A-F, 1–10) to ensure consistency of measurements at each visit. Out of 1200 electrodes, 1124 (93.7%) were able to be measured.

At each visit for SL and DOM testing, results were documented with the device OFF and ON. The difference in pixel distance (SL testing) or degrees (DOM testing) measured as the testing performance with the device OFF minus ON (Δ^OFF-ON) was used to determine visual function changes with the Argus II Retinal Prosthesis. A positive Δ^OFF-ON indicated an improvement in functional visual testing with the device ON. The unit of analysis was the patient with the assumption that visual function performance was patient-dependent. As such, Wilcoxon Signed-Rank tests were used with paired samples for each individual patient as well as for patients that improved or worsened with the device ON to determine whether Δ^OFF-ON was significant. Non-parametric testing was chosen due to the small sample size and because no assumptions were made about the normality of the distribution in testing performance. To correlate visual function with gap distance, generalized linear models (GLMs) using Δ^OFF-ON for SL and DOM testing (dependent variables), and gap distances (covariates) were used. A separate model examining the interaction between patient and gap distance as predictor variables was performed to determine the patient-dependent effects of gap distance on visual function performance. GLMs were also used to determine the effect of sensitivity detection threshold on average Δ^OFF-ON. Statistical analyses were performed using IBM SPSS Version 27 (IBM Corp., Armonk, N.Y., USA).

The sensitivity detection threshold measured for a patient at M1, M3, M6 or Y1 corresponded to the average electrode array-retina gap distance at the same visit, as the OCT image was captured on the same day. Using a GLM with pooled data, gap distances were found to have no significant effect on sensitivity detection thresholds (p > 0.05). A separate interaction model found no significant effect of gap distance on sensitivity detection threshold when taking patient-dependent effects into account (p > 0.05) (Table 2).

Using Wilcoxon Signed-Rank tests to compare testing performance with the device OFF and ON for each patient individually, no statistically significant difference was found for any patient in both SL and DOM testing. We then analyzed testing performance for patients that had an average positive Δ^OFF-ON, average negative Δ^OFF-ON, and with pooled data to determine if these differences were significant. In SL testing, 3 out of 5 patients demonstrated an improvement in visual function at M3, M6 and Y1 with the device ON (Table 3, Fig. 3), indicated by an average positive Δ^OFF-ON. These improvements were found to be statistically significant (p < 0.05). In patients with an average negative Δ^OFF-ON, this difference was also found to be statistically significant (p < 0.05). Using pooled data from all patients, Δ^OFF-ON for SL testing was not statistically significant (p > 0.05). In DOM testing, 2 patients demonstrated a positive Δ^OFF-ON when averaged over the three visits, with 1 patient having an improvement at all testing visits (Table 3, Fig. 4). However, paired Wilcoxon Signed-Rank Tests for patients with an average positive Δ^OFF-ON and for a pooled sample with all patients were not statistically significant (p > 0.05). The difference in patients with a negative Δ^OFF-ON was found to be statistically significant (p < 0.05).

Using pooled data, no significant effect was found between Δ^OFF-ON in SL and DOM testing and electrode array-retina gap distance (Fig. 5). When accounting for patient-dependent effects using an interaction model, a significant effect was seen between electrode array-retina gap distances and Δ^OFF-ON in SL testing (p < 0.05). The β coefficients for patients 3 and 4 where the most statistically significant effects were seen were − 0.240 and 0.237, respectively, indicating that the directionality of a patient-dependent effect of gap distance of SL testing performance may be positive or negative. This patient-dependent effect was not seen between electrode array-retina gap distances and Δ^OFF-ON in DOM testing (p > 0.05).