Prosthetic Visual Acuity with the PRIMA Subretinal Microchip in Patients with Atrophic Age-Related Macular Degeneration at 4 Years Follow-up

Patients

This feasibility study of the PRIMA implant (NCT03333954) aimed to test safety and functionality in 5 patients with atrophic AMD. The study adhered to the Declaration of Helsinki and received ethical approval from the Comité de Protection des Personnes Ile de France II and the Agence Nationale de Sécurité du Médicament et des Produits de Santé. Study participants were aged > 60 years and had advanced dry AMD with an atrophic zone of at least 3 optic disc diameters and best-corrected VA of ≤ 20/400 in the worse-seeing study eye; no foveal light perception (absolute scotoma) but visual perception in the periphery, with preferred retinal locus determined by microperimetry; absence of photoreceptors and presence of the inner retina in the atrophic area, as confirmed by OCT; absence of MNV verified by retinal angiography. All other ocular and general pathologies that could contribute to low VA were excluded. Patients provided written informed consent to participate in the study. Visual acuity in the nonimplanted fellow eye was measured throughout the study as a control. Surgical procedures were performed in the Fondation Ophtalmologique A. de Rothschild (Paris, France), with the first subject implanted in December 2017 and the fifth in June 2018.¹⁸ The patients’ rehabilitation and visual function assessment was carried out at the clinical Investigation Center of the Quinze-Vingts National Eye Hospital (Paris, France).

PRIMA Implantation

In a preplanning phase of the implantation, the surgeon (Y.L.M.) examined the fundus images and OCT scans. PRIMA implantation site was determined based on the preferred retinal locus position and atrophy size. The implantation surgery was performed under local anesthesia in 1 case, and under general anesthesia in 4 patients. Complete 23-gauge trans pars plana vitrectomy was performed with peripheral vitreous base shaving. A posterior vitreous detachment was confirmed using intravitreal triamcinolone (Kenalog-40, Bristol Myers Squibb). A 27-gauge chandelier and 23-gauge Alcon light pipe were used for intraocular illumination. The retina was detached between the macula and the planned position of the retinotomy by injecting a balanced salt solution with controlled visco fluid injection to form a bleb using a 41-gauge straight needle, at a maximal pressure of 12 pounds per square inch. The location of the sclerotomy was chosen to obtain a straight alignment between the sclerotomy, retinotomy, and the fovea. The detached retina was marked with diathermy. Retinotomy of 2.4 mm in size was created in the temporal area, about 1 mm outside the atrophic retina, using 23-gauge vertical scissors, and then incised on a length of 3 mm at 4 to 5 mm distance from the fovea. Retinotomy location was selected based on preoperative fundus image and microperimetry to reduce the risk of jeopardizing the useful residual vision, including the preferred retinal locus. In 1 case, a localized endodiathermy was needed to stop a retinal vessel bleeding.

A subretinal spatula was used to detach the macula, with care taken to avoid button-holing any thinned areas of the macula and avoid the formation of a full-thickness macular hole. During surgery, the PRIMA implant was delivered through the retinotomy and into the subretinal space at the macula using silicone-tipped forceps. A bubble of perfluorocarbon liquid was injected over the macula to flatten the fluid bleb and stabilize the implant at the submacular docking site during the forceps removal and to close the retinotomy. The chip was then shifted to its final location near the fovea by careful transretinal manipulation using the blunt side of the subretinal pick to push the edge of the chip slowly under the attached retina. When the chip location was considered to be optimal, intraoperative OCT was used to confirm the correct placement and the contact of the macular retina with the anterior surface of the chip. After moving the implant at this location, perfluorocarbon liquid was fully exchanged with air. The absence of subretinal fluid was confirmed using intraoperative OCT in all cases. An indented peripheral retina examination was performed to check for retinal breaks, and any residual perfluorocarbon liquid was removed with a soft tip cannula. A final tamponade with silicone oil (3 cases), 1 case hexafluoroethane (C₂F₆), and 1 case with sulfur hexafluoride (SF₆) was achieved. All subjects were instructed to posture face-down after surgery to stabilize the retina for 24 hours. Postoperatively, silicone oil removal was carried out 4 weeks after the first operation using the 25G pars plana vitrectomy technique in 2 subjects.

Assessment of Natural VA

The best-corrected natural VA of each patient was measured without the PRIMA glasses to assess any potential VA changes after implantation and over time. ETDRS letter charts were used at 4 m; if the patient could not correctly identify at least 20 letters at 4 m, the test continued at 1 m. The smallest font size for which the patient could read at least 4 letters in 1 line was recorded.

Assessment of Prosthetic Vision

A camera on the PRIMA-2 AR glasses captured a visual field of 50° × 40°, with a central third of the image being projected onto the retina to match the display’s 17° field of view without zoom. The projected images covered a field of 5.1 mm (17°) on the retina, with a resolution of 10.5 μm. Maximum peak retinal irradiance was 3.5 Mw/mm², with a maximum pulse duration of 10 ms and frame rate of 30 Hz, well within the thermal safety limits for chronic use of near-infrared light.¹⁹ Computer-generated images could also be projected on the ocular display directly, without the use of the camera. The angular magnification of the system without zoom was 1:1. In the measurements of letter recognition, patients were also allowed to use the electronic zoom at their preferred level of magnification (1×, 2×, 3×, or 8×), and this setting was recorded.

Visual acuity measurements were performed up to 3 times on different days near the 48-month follow-up, and results are reported as the median of these 3 measurements. The subject was asked to read as many letters as possible from top to bottom on the ETDRS chart placed at 1 meter. The baseline VA test without-PRIMA glasses was also performed at 1 meter but only once because at the time it was not defined as an efficacy end point.

Assessment of Safety

Investigators reported all the adverse events. These were categorized as serious or nonserious, and whether they were related to the procedure or a device. During follow-up, investigator assessments were reviewed by a clinical events committee.

Data Synthesis and Analysis

To estimate the mean VA within the study settings, a mixed regression model with random effect for subjects was used.²⁰ This corresponds to the regression model with logMAR and ETDRS VA as output variables for natural and prosthetic vision, respectively, study setting as input variable, and random effects for subjects. The estimates are reported as standard error (SE) for the mean, 95% confidence intervals (CIs) for the mean, and SE for the mean difference and 95% CI for the mean difference. The estimated means were then compared between study settings using the post hoc Tukey method. The adjusted P value of < 0.05 was considered statistically significant. The analysis was performed using R version 4.1.2.²¹

To test changes within the study and nonstudy eyes, as well as to compare the possible differences between them, a mixed regression model was developed with VA as output variable; time, eye status (study eye and nonstudy eye), and interaction of time with eye status as input variables; and, finally, with subject as a random effect.

Natural Vision

During the 48-month follow-up, all subjects experienced a slight reduction of VA in the nonstudy eye, albeit not significant. For all 5 patients, the mean baseline VA in the nonstudy eyes was a logMAR of 0.78 (SE, 0.13; 95% CI, 0.41–1.15) or 20/120, and, in the study eyes, it was a logMAR of 1.48 (SE, 0.13; 95% CI, 1.11–1.85) or 20/600 (P < 0.00)1. At the 48-month visit, in the nonstudy eyes, the estimated mean VA was a logMAR of 1.03 (SE, 0.14; 95% CI, 0.63–1.42) or 20/215; in the study eyes, the mean VA was logMAR 1.33 (SE, 0.14; 95% CI, 0.93–1.72) or 20/430 (P = 0.058). Residual natural VA in the treated eye of any patient did not decrease during the 48-month follow-up period. It even exhibited a temporary improvement, albeit not significant: P = 0.8 (Fig 5). By 48 months, the mean VA in the study eyes improved by 0.1 logMAR mainly because of 1 subject who improved by up to a logMAR of 0.4, whereas in other subjects it improved by no more than a logMAR of 0.1.

Prosthetic Vision

The PRIMA system inverted the contrast of the ETDRS chart so that black letters on a whiteboard are projected as bright stimulation patterns on a black background. As shown in Figure 6, VA improved in all 3 subjects compared with the baseline, and their prosthetic vision (with the PRIMA glasses) was better than their natural vision (without-PRIMA glasses). The means for each setting are as follows: baseline, 11.7 ETDRS letters (SE, 4.8; 95% CI, 0–32.1) or 20/584; without-PRIMA glasses at 48 months, 18.2 ETDRS letters (SE, 4.8; 95% CI, 0–38.6) or 20/433; and with-PRIMA glasses (prosthetic vision with preferred magnification) at 48 months, 43.3 ETDRS letters (SE, 4.8; 95% CI, 22.9–63.8) or 20/136.

The differences between the settings were as follows: natural vision at 48 months (without-PRIMA glasses) compared with the baseline, +6.5 ETDRS letters (SE 5.1; 95% CI, –11.7 to 24.7; P = 0.480); prosthetic vision (with-PRIMA glasses) compared with the baseline, +31.7 ETDRS letters (SE, 5.1; 95% CI, 13.4–49.9; P = 0.008); and prosthetic vision compared with natural vision at 48 months (with-PRIMA versus without-PRIMA glasses), +25.2 ETDRS letters (SE, 5.1; 95% CI, 6.9–43.4; P = 0.017). Both differences, between prosthetic vision and baseline, as well as between prosthetic vision and natural vision at 48 months, are significant. The magnification/zoom range for each participant was as follows: patient 1(1–8)×; patient 2 (1–8)×; and patient 3 (1–4)× magnification.

In addition to objective VA measurements, subjects were observed during training sessions. Given their ability to recognize letters, they were able to read sequences of letters and words. This was also observed during exercises, such as reading product names on food packages, panels, and signs outdoors and indoors or a train timetable (Video S1, available at www.ophthalmologyscience.org).

Safety

In total, 4 study-related serious adverse events were reported in the follow-up of all 5 patients, but none was related to the device (Table 2). Per the investigator assessment, 1 event was definitely procedure-related, 2 were probably procedure-related, and 1 was possibly procedure-related, as described below. The subject’s death because of cancer was neither procedure nor device-related.

The MNV event, which was deemed to be procedure-related, was observed in an area where Bruch’s membrane got damaged during implantation surgery under local anesthesia. An unexpected head movement of the subject at the time of retinal implant delivery under the macula led to damage of the Bruch membrane and choroidal hemorrhage. After resorption of blood, fibrotic tissue was observed around the implant which, on OCT, seemed to be located under the pigment epithelium. This patient had visual perception of stimulation but no spatial resolution using the implant, and hence this patient was excluded from the acuity measurements (Fig 4A). Thirty-one months after implantation, an asymptomatic MNV was discovered in the same eye of this subject, within an area that was not touched during surgery. Therefore, it was judged as unlikely to be procedure- or device-related. The MNV was treated with a course of intravitreal ranibizumab injections and resolved after a year without sequelae. At 42 months postimplantation, the same subject developed a second MNV in a different area of the posterior pole near the retinotomy site. This second MNV event was deemed as procedure related.

At 12-months postoperatively, a retinal detachment developed in the control eye of another patient. This was successfully treated with vitrectomy and gas injection, and the retina was reattached. This subject was also known to have underlying glaucoma at baseline and developed uncontrolled ocular hypertension in the implanted eye 2 years postoperatively, which was unresponsive to medical treatment. The patient underwent a successful trabeculectomy achieving eye pressure control, and treatment remains ongoing. Because vitrectomy risks exacerbation of the preexisting glaucoma, this adverse event was deemed procedure-related, although the patient already had glaucoma.

All the study-related nonserious adverse events are listed in Table 3. Microcysts were observed in 3 subjects, which resolved without sequelae. The microcysts did not create any symptoms and seemed not to affect the perception of the stimulation by the implant. Two patients developed significant ocular hypertension that required medical treatment.

Implant stability in the subretinal space has been previously reported.¹⁸ The authors have also published data that showed that, in the 3 patients reported in this study, PRIMA implants did not cause any obvious structural changes in the vicinity of the device nor in the outer retinal layers over a 3-year period.¹⁸