Applications of extended reality in education included surgical simulators (46/54), ophthalmoscopy simulators (6/54), and optometry training simulators (2/54), with medical students, optometry students, trainee, or trained ophthalmologists as participants.

Of 46 studies evaluating surgical simulators, the EyeSi surgical simulator (VR Magic) was most commonly used (n=38). Others included MicroVisTouch (ImmersiveTouch) (n=1), PixEye Ophthalmic Simulator (SimEdge SA) (n=1), and 6 self-designed simulators. The most common surgical procedure simulated in these studies was cataract surgery (n=36), followed by vitreoretinal procedures (n=9), laser trabeculoplasty (n=1), and corneal laceration repair (n=1).

Of the 46 studies, 24 were studies that evaluated the efficacy of surgical simulators using evaluations of surgical performance on real patients, objective assessments by subspecialty experts, or simulator-based metrics as outcome measures. There were 14 [28-47] studies that used surgical performance on real patients as outcome measures, of which 4 were randomized trials [48-51] (Table 1); these randomized trials compared the use of virtual reality ophthalmic surgical simulators with conventional methods of surgical training. Objective assessment metrics of participants’ surgical performance on real patients were evaluated by subspecialty experts. All studies [48,50,51], except one [49], showed that simulator-training resulted in surgical performance significantly superior to that of conventional training methods in terms of quality, time efficiency, efficacy, or complication rates. In particular, Deuchler et al [50] found that warm-up simulation training improved performance for surgeons who had not operated for a significant period of time. Daly et al [49] showed that residents who underwent EyeSi training were significantly slower at performing their first continuous curvilinear capsulorhexis than participants who underwent wet-lab training but achieved similar surgical performance scores.

There were 10 nonrandomized studies [28-37] that evaluated the EyeSi simulator for cataract surgery training (Table 2). Thomsen et al [28] and la Cour et al [29] demonstrated statistically significant improved Objective Structured Assessment of Cataract Surgical Skill Scores (OSACSS) in novice surgeons after training with the EyeSi cataract module. Roohipoor et al [30] found significant correlations between residents’ EyeSi simulator-based scores and their eventual surgery count and Global Rating Assessment of Skills in Intraocular Surgery scores. The other 7 studies [31-37] investigated the relationship between EyeSi simulator use and complication rates in cataract surgery on real patients, of which 6 studies [31-36] showed that the use of the EyeSi simulator was associated with reduced complication rates. McCannel et al [37] found that EyeSi capsulorhexis training was not associated with lower vitreous loss rates overall but was associated with higher nonerrant continuous curvilinear capsulorhexis-associated vitreous loss.

Seven studies [38-44] investigated the efficacy of the EyeSi surgical simulator (n=6) or a self-made augmented reality microsurgery simulator (n=1) by evaluating participants’ surgical performance on the same simulators using simulator-based metrics (Table 3 and Table 4). Selvander et al [40] used the OSACSS and Objective Structured Assessment of Technical Surgical Skills (OSATS).

Three studies were randomized trials [38-40] (Table 3). Thomsen et al [39] investigated if there could be interprocedural transfer of skills and found that residents with simulated cataract surgery training did not perform significantly better than those without (simulator score with training: mean 381, SD 129 vs simulator score without training: mean 455, SD 82, P=.262) at the vitreoretinal surgery module. Selvander et al [40] had a similar aim and found that training on the capsulorhexis or cataract navigation training module on the EyeSi did not significantly improve performance at the other (OSACSS—training: 8, no training: 8, P=.64; OSATS score—training: 7, no training: 10, P=.52); however, repeated practice with each module significantly improved simulator-based scoring for the respective modules. Bergqvist et al [38] found that medical students who trained with simulated cataract surgery had higher overall simulator scores and fewer complications than those who did not train on the simulator. These trials suggest that simulator training improves performance only for the specific procedure being trained. The other 4 studies [41-44] were prospective cohort studies or case series (Table 4) that demonstrated that extended reality surgical training resulted in significant improvements on subsequent simulator-based performance scores.

Three randomized trials investigated the effect of extended reality surgical simulator training on surgical performance in the wet lab (Table 5). Surgical performance was assessed using objective outcomes. Feldman et al [46] and Feudner et al [47] showed that training on the EyeSi simulator significantly improved wet-lab performance for corneal laceration repair and capsulorhexis, respectively, while Jonas et al [45] showed that training on a self-made virtual reality simulator improved wet-lab performance for pars plana vitrectomies.

Of 46 studies, 20 studies evaluated the validity of surgical simulator-based assessments (Multimedia Appendix 3). Most validity studies achieved an OCEBM level of evidence of 2b, corresponding to exploratory cohort studies with good reference standards. The most common source of validity evidence was Relationship with other variables, addressed in 19 of 20 studies (95%). Studies achieved this by statistically evaluating the relationship between surgical performance on the simulator and participants’ levels of expertise. Content validity was addressed in 18 studies (90%), Response process was addressed in 9 studies (45%), Internal structure was addressed in 5 studies (25%), and Consequences was addressed in 2 studies (10%). Only 2 of 20 (10%) studies addressed all 5 sources.

Of these 20 studies, 12 assessed surgical performance using simulator-based scoring only [39,43,52-61], 6 studies compared simulator-based scores with video-based scoring (OSACSS, OSATS, or motion-tracking software) [62-67], and 2 studies used video-based scoring only [40,68].

For the EyeSi surgical simulator, most studies found that the surgical performance of experienced participants was significantly better than that of less-experienced participants. Sikder et al [52] found that intervening surgical experience significantly improved capsulorhexis performance on the MicroVisTouch cataract surgery simulator. Lam et al [62] showed that in a self-made phacoemulsification simulator, more experienced participants attained significantly higher scores in all main procedures and completed tasks significantly faster.

Five studies [69-73] assessed the perception of ophthalmologists and medical students toward surgical simulators using user-reported outcome measures. These studies achieved OCEBM evidence levels of 4 (n=4) and 2b (n=1). Users found the EyeSi and a novel virtual reality continuous curvilinear capsulorhexis simulator to be useful in improving surgical skill, confidence, and understanding, while providing a safe and realistic alternative for training.

Six studies [74-79] evaluated the use of extended reality as a tool for education in ophthalmoscopy. Simulators used were the EyeSi Augmented Reality Direct (n=1) and Binocular Indirect (n=3) ophthalmoscopy simulators, and 2 novel self-made direct ophthalmoscopy simulators comprising the HTC Vive Virtual Reality-Head-Mounted Display (n=1) and the RITECH II Virtual Reality-Head-Mounted Display (n=1).

Two randomized trials [74,75], with OCEBM evidence levels 2b, assessed the efficacy of the EyeSi Binocular Indirect Ophthalmoscopy simulator. Both studies showed that participants who trained with the EyeSi Binocular Indirect Ophthalmoscopy simulator performed significantly better than participants who underwent conventional training.

Three studies [75-77], with OCEBM evidence levels of 2b, assessed the validity of the EyeSi Binocular Indirect Ophthalmoscopy simulator (n=2) and the EyeSi Binocular Direct Ophthalmoscopy simulator (n=1) for training and assessment. All studies demonstrated Relationships with other variables as a source of validity evidence and found that participants with more experience had significantly higher ophthalmoscopy evaluation scores. Content validity was addressed in all studies. Only 1 study [77] addressed Internal structure by evaluating internal consistency between simulator modules and evaluated Consequences by calculating a pass or fail score.

Two user perception studies [78,79] found that medical students felt that self-assembled virtual reality direct ophthalmoscopy simulators were usable and useful in improving ophthalmoscopy skills.

Two studies [80,81] evaluated the preliminary user experience of an augmented reality optometry simulator comprising a head-mounted display, a slit-lamp instrument, and a simulated eye, which allowed the simulation of optometry training tasks. User studies involving undergraduate optometry students showed that the simulator was feasible in simulating foreign body removal as a training task with a high level of user satisfaction.

Five studies evaluated the use of extended reality for the production of immersive and interactive content for diagnostic applications. Two studies evaluated the use of extended reality to display ocular imaging data [82,83], and 3 studies [84-86] evaluated the validity of extended reality as a simulation tool for the functional assessment of patients with ophthalmic diseases.

Two case series, which achieved OCEBM evidence levels of 4, evaluated the presentation of ocular imaging modalities in virtual reality and augmented reality environments.

Maloca et al [82] tested the feasibility of displaying optical coherence tomography images in a virtual reality environment with a virtual reality-head-mounted display. A user perception survey involving 57 participants found it to be well tolerated with minimal side-effects. Berger et al [83] demonstrated feasibility for a method of direct overlay of photographic and angiographic fundus images onto a real-time slit lamp fundus view in 5 participants.

Three studies evaluated the use of extended reality simulators for the functional assessment of patients with ophthalmic diseases. The studies achieved OCEBM levels of evidence of 4.

Goh et al [84] trialed the use of the Virtual Reality Glaucoma Visual Function Test with a smartphone paired with the Google Cardboard head-mounted display to assess the visual function of glaucoma patients and found that stationary test person scores demonstrated criterion and convergent validity, corresponding to Relationship with other variables.

Ungewiss et al [85] compared the assessment of driving performance in a driving simulator with that in a real vehicle in patients with glaucoma (n=10), hemianopia (n=10), and normal controls (n=20) and found that patients with hemianopic and glaucoma performed worse than healthy controls on the driving simulator, demonstrating Relationship with other variables as a source of validity evidence.

Jones et al [86] evaluated the use of a head-mounted display to simulate visual impairment in glaucoma using virtual reality and augmented reality and found it able to replicate and objectively quantify functional impairments associated with visual impairments. When the simulated visual field loss was inferior, impairments were noted to be significantly greater than those noted when the simulated visual field loss was superior, which was consistent with previous experiences of real patients with glaucoma [87-89].

A total of 28 studies evaluated the use of extended reality in therapeutics. These studies evaluated heads-up surgery (n=21), binocular treatment of amblyopia (n=2), functional improvement for the visually impaired (n=4), and an aid for achromatopsia (n=1).

Heads-up surgery involves the use of a 3D camera to capture images from a stereomicroscope for presentation on a 3D display. Of 21 studies [90-110], the most common heads-up surgical system evaluated was NGENUITY 3D (Alcon Laboratories) (n=12), followed by TrueVision 3D HD System (TrueVision Systems Inc) (n=2), TRENION 3D HD (Carl Zeiss Meditec) (n=1), MKC-700HD and CFA-3DL1 (Ikegami) (n=1), Digital Microsurgical Workstation (3D Vision Systems) (n=1), TIPCAM 1S 3D ORL endoscope (Karl Storz) (n=1). Surgical procedures included vitreoretinal procedures (n=17), cataract surgery (n=5), scleral buckle (n=1), and endoscopic lacrimal surgery (n=1).

Six studies had OCEBM evidence levels of 2b, corresponding to randomized trials (n=4) and cohort studies (n=2). There were 15 studies with OCEBM evidence levels of 4, corresponding to case series, case-control studies, or poor-quality cohort studies.

The 4 randomized trials [90-93] demonstrated noninferiority of heads-up surgery in comparison with conventional microscope surgery in postoperative outcomes and complications. Qian et al [90] performed phacoemulsification and intraocular lens implantation and reported no significant difference in mean surgery time, postoperative mean endothelial cell density between conventional surgery (n=10) and heads-up surgery (n=10). Talcott et al [91] performed pars plana vitrectomies and showed that compared with conventional surgery (n=16), heads-up surgery (n=23) significantly increased macular peel time (14.76 minutes vs 11.87 minutes, P=.004) but not overall operative time. There was no significant difference in visual acuity (logarithm of the minimum angle of resolution) or change from baseline, and no clinically significant intraoperative complications. Romano et al [92] randomized 50 eyes to the use of an unspecified heads-up surgical system (n=25) and conventional surgery (n=25) for 25-gauge pars plana vitrectomy; there was no significant difference in mean operation duration. Surgeons and observers were significantly more satisfied (P<.001) using the heads-up system. Kumar et al [93] randomized 50 patients to macular hole surgery with an unspecified heads-up system (n=25) and conventional macular hole surgery (n=25); there were no significant differences in postoperative visual acuity, macule hole indices, surgical time, total internal limiting membrane peel time, number of flap initiations, and macular hole closure rates. Microscope illumination intensity (heads-up: 100%; conventional: 45%) and endoillumination was significantly lower in heads-up surgery (heads-up: 40%; conventional: 13%).

The other 17 studies [94-110] were case series or cohort studies (Multimedia Appendix 4). They reported experiences with different heads-up surgical systems with various surgical procedures and assessed a few common outcomes with the following findings.

Of studies comparing heads-up surgical systems with conventional surgery, most reported procedural success or success that was not significantly different from that of conventional surgery. There were no significant differences in mean procedural durations, postoperative visual acuity, or improvements in visual acuity. The minimum required endoillumination was lower in heads-up surgery than that in conventional surgery. Zhang et al [94] found a significantly lower mean minimum required endoillumination for heads-up surgery than that for conventional surgery (10% vs 35%, 598.7 vs 1913 lx, P<.001). Matsumoto et al [95] operated safely and successfully on 74 eyes with an endoillumination intensity of 3% using a heads-up system. No study reported major perioperative complications or significant differences in complication rates between heads-up surgery and conventional surgery. Users preferred heads-up surgery to conventional surgery.

Two studies [21,111] evaluated the efficacy of the use of extended reality for interactive and immersive binocular treatment for amblyopia. Lee et al [111] randomized 22 children with amblyopia (mean age 8.7 years, SD 1.3) to treatment with virtual reality videogaming on an unspecified virtual reality-head-mounted display (n=7), virtual reality videogaming and Bangerter foil (n=5), or Bangerter foil only (n=10), achieving an OCEBM evidence level of 2b. Two of 7 (29%) of patients in the virtual reality videogaming group, and 2 of 5 (40%) patients in the virtual reality videogaming with Bangerter foil group gained more than 0.2 in logarithm of the minimum angle of resolution of vision. Ziak et al [21] trialed the use of the Oculus Rift virtual reality head-mounted display for dichoptic virtual reality video gaming in treating 17 adults with amblyopia, achieving an OCEBM evidence level of 4. There was a significant improvement in mean amblyopic eye visual acuity (logarithm of the minimum angle of resolution: from mean 0.58, SD 0.35 to mean 0.43, SD 0.38; P<.01). Mean stereoacuity also improved significantly from 263.3 s of arc to 176.7 s of arc (P<.01).

Five case series studies [20,112-115] evaluated the use of extended reality technologies to improve the function of patients with visual impairment and disorders. All reached an OCEBM evidence level of 4.

Two studies evaluated the use of digital spectacles (DSpecs) to improve mobility in a total of 43 patients with peripheral visual field deficits [20,112]. DSpecs work by capturing, relocating, and resizing video signals to fit within a patient’s visual field in real time using augmented reality technology. DSpecs enabled patients with visual field defects to have improved object identification, hand-eye coordination, and walking mobility.

Sanchez et al [113] evaluated Augmented Reality Tags for Assisting the Blind, an augmented reality system which helps the user determine the position of indoor objects by generating an audio-based representation of space. Blind participants perceived Augmented Reality Tags for Assisting the Blind to be a useful tool for assisting orientation and mobility tasks.

Tobler-Ammann et al [114] evaluated the use of virtual reality exercise games to encourage exploration of neglected space in patients with visuospatial neglect after stroke. Cognitive and spatial exploration skills trended toward improvement after the use of virtual reality exercise games and continued improving at follow-up in 5 of 7 participants. Adherence rates were high, and there were no adverse events.

Melillo et al [115] evaluated the efficacy of an augmented reality wearable improved vision system for patients with color vision deficiency. The system captures and remaps colors from the environment and displays it to the user via a head-mounted display. The system significantly improved Ishihara Vision Test scores in participants with color vision deficiency (mean score 5.8 vs 14.8, P=.03).