Safety and Efficacy of Adeno-Associated Viral Gene Therapy in Patients With Retinal Degeneration: A Systematic Review and Meta-Analysis

Abstract

Purpose: This systematic review evaluates the safety and efficacy of ocular gene therapy using adeno-associated virus (AAV).

Methods: MEDLINE, Embase, Cochrane Central Register of Controlled Trials, and ClinicalTrials.gov were searched systematically for controlled or non-controlled interventional gene therapy studies using key words related to retinal diseases, gene therapy, and AAV vectors. The primary outcome measure was safety, based on ocular severe adverse events (SAEs). Secondary outcome measures evaluated efficacy of the therapy based on best corrected visual acuity (BCVA) and improvements in visual sensitivity and systemic involvement following ocular delivery. Pooling was done using a DerSimonian Laird random effects model. Risk of bias was assessed using the Cochrane Risk of Bias Tool, version 1.

Results: Our search identified 3548 records. Of these, 80 publications met eligibility criteria, representing 28 registered clinical trials and 5 postmarket surveillance studies involving AAV gene therapy for Leber congenital amaurosis (LCA), choroideremia, Leber hereditary optic neuropathy (LHON), age-related macular degeneration (AMD), retinitis pigmentosa (RP), X-linked retinoschisis, and achromatopsia. Overall, AAV therapy vectors were associated with a cumulative incidence of at least one SAE of 8% (95% confidence intervals [CIs] of 5% to 12%). SAEs were often associated with the surgical procedure rather than the therapeutic vector itself. Poor or inconsistent reporting of adverse events (AEs) were a limitation for the meta-analysis. The proportion of patients with any improvement in BCVA and visual sensitivity was 41% (95% CIs of 31% to 51%) and 51% (95% CIs of 31% to 70%), respectively. Systemic immune involvement was associated with a cumulative incidence of 31% (95% CI = 21% to 42%).

Conclusions: AAV gene therapy vectors appear to be safe but the surgical procedure required to deliver them is associated with some risk. The large variability in efficacy can be attributed to the small number of patients treated, the heterogeneity of the population and the variability in dosage, volume, and follow-up.

Translational Relevance: This systematic review will help to inform and guide future clinical trials.

Introduction

Degenerative diseases of the retina and optic nerve are a heterogeneous group of incurable eye diseases that collectively account for a large proportion of visual impairment and blindness in developed countries.^1–3 Disease onset and progression are associated with retinal pigment epithelium (RPE), photoreceptor, or retinal ganglion cell dysfunction and loss. Age of clinical presentation varies from early childhood to late in life depending on the disease and causative factors involved. The etiology of these disorders is variable, with some having a clear genetic basis, and others involving both genetic and environmental components.^4–6 

Ocular gene therapy is a promising treatment approach for these conditions, owing to the ease of accessibility, isolation from systemic circulation, and relative immune privilege of the eyes.^7,8 Studies in Leber congenital amaurosis (LCA)^9–31 led to the development of Luxturna (voretigene neparvovec-rzyl) as the first approved human gene therapy for an inherited disease (reviewed in Refs. 32 and 33). Gene therapies for a number of other retinal degenerations have since entered clinical trials, with an increasing number of studies reporting visual improvements upon initial assessments and over long-term follow-up periods.^34,35 Adeno-associated virus (AAV) vectors are currently the most widely used vectors for ocular gene therapy; they are considered relatively non-toxic, purportedly associated with minimal immune response when delivered into the subretinal or intravitreal space, and they are thought to provide long-term expression.^7,36 The growing interest and investment in expanding AAV-based gene therapy programs and products for treating retinal diseases necessitates a synthesis of the knowledge gained from recent studies and findings to establish current best practices for optimal outcomes. Two recent systematic reviews examined the efficacy of RPE65-mediated gene therapy for the treatment of LCA,^37,38 each dealing with a small number of studies. A more recent review examined gene therapy across multiple retinal diseases,³⁹ but did not conduct a meta-analysis. Our systematic review and meta-analysis provide an objective assessment of safety and efficacy outcomes for intraocular delivery of existing AAV vectors for gene therapy in all retinal diseases treated to date to help inform and guide future clinical studies. 

Materials and Methods

Our protocol was registered on PROSPERO, the international prospective registry for systematic review protocols (CRD42020182842). We followed the reporting guidelines set by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement⁴⁰ (completed checklist in Supplementary Table S1). Protocol development and review conduct were guided by best practice principles outlined by the Cochrane Handbook for Systematic Reviews of Interventions. There were no deviations from the protocol. 

Search Strategy

We conducted a systematic literature search through MEDLINE (OVID interface, including In-Process and Epub Ahead of Print), Embase (OVID interface), Cochrane Central Register of Controlled Trials (Wiley interface), and ClinicalTrials.gov. The search strategies were created in collaboration with an information specialist and clinical content experts in the field (authors C.T. and P.S.L.); the strategies were developed using keywords related to retinal diseases and gene therapy. The initial search was performed on May 8, 2020, and updated on May 24, 2022. Only full-text articles in English were retained. Reference lists from included trials and pertinent reviews discerned from the primary search were also examined. Our search strategies can be found in Supplementary Appendix 1. 

Eligibility Criteria

Non-controlled (i.e. single-arm) or controlled interventional studies that investigated AAV-mediated gene therapy in patients with retinal diseases were considered. Controlled studies were either randomized controlled trials (RCTs) or nonrandomized studies that incorporated a comparator. Studies reporting at least one of the outcomes of interest (described below) were included. Grey literature, abstracts, conference abstracts, commentaries, letters, reviews, and editorials were excluded. 

Outcomes

The primary outcome was safety, which included the following ocular adverse events: 

• 1. Direct potential adverse events from intraocular/subretinal administration: intraocular inflammation (including endophthalmitis), intraocular infection, cataract formation, vitreous hemorrhage, intra-retinal or subretinal hemorrhage, retinal tear, retinal detachment, macular hole formation, reduced visual acuity, change in central sub-field foveal thickness, and elevated intraocular pressure.
• 2. Secondary effects: eye pain or irritation, color vision changes, diplopia, and changes in pupillary response (pupillometry).

Secondary outcomes included the following: 

• 1. Gene therapy efficacy as determined by best-corrected visual acuity (BCVA) using the Early Treatment Diabetic Retinopathy Study (ETDRS) scale and visual sensitivity testing (based on full-field stimulus threshold [FST], visual fields, or contrast sensitivity). Due to heterogeneity in reporting, we treated efficacy as a dichotomous outcome of whether any improvement in a patient was reported or not. We could not assess other objective and subjective measurements of visual structure or function because they were reported in only a small percentage of eligible clinical trials.
• 2. Systemic involvement: detection of disseminated recombinant AAV in the systemic circulation and/or evidence of the development of an immunologic response to the virus.

Study Selection and Data Extraction

Management of screening and data extraction was supported by DistillerSR software (Evidence Partners, Ottawa, Canada). All stages of screening, study selection, and data extraction were performed by two trained reviewers (authors M.S. and P.S.L.) independently and in duplicate. Data extraction forms were pilot tested on three studies to ensure consensus between reviewers. Any disagreements were resolved by further discussion, or with a third team member. Data items extracted included the following domains: publication details, study population characteristics, intervention characteristics, study design, outcomes of interest, and risk of bias. 

Risk of Bias Assessment

Risk of bias was assessed using the Cochrane Risk of Bias tool (version 1)⁴¹ and reported separately by trial type (RCTs, nonrandomized controlled trials, and single-arm studies). The risk of bias assessment figure was produced with the Risk-of-Bias VISualization (robvis) online tool.⁴² 

Data Analysis

Studies were pooled using Comprehensive Meta-Analyst (version 3; Biostat Inc., USA). Dichotomous outcomes were analyzed using a random effects meta-analysis based on the DerSimonian Laird model and reported as proportions (cumulative incidence) with accompanying 95% confidence intervals (CIs). 

Data Presentation

Studies are identified and reported according to their clinical trial registration number (National Clinical Trial [NCT] #) as many of the trials followed the same patients over several years and thus had multiple reports with different publication dates and different first authors. The five postmarket surveillance studies describing the use of Luxturna in patients after its regulatory approval by the US Food and Drug Administration (FDA) are grouped and presented collectively. 

Results

We conducted a literature search of non-controlled (single-arm) or controlled interventional studies that investigated AAV-mediated gene therapy in patients with retinal diseases that reported on at least one of the outcomes of interest: ocular adverse events; therapeutic efficacy using BCVA or visual sensitivity testing (FST, visual fields, or contrast sensitivity); and systemic immune involvement. Studies reporting at least one of the outcomes of interest were included (Fig. 1). The search yielded a total of 3548 citations for title and abstract screening. Following full-text screening, 80 articles representing 28 clinical trials and 5 postmarket surveillance studies met our eligibility criteria. The postmarket surveillance studies involved patients that were treated with Luxturna for RPE65-mediated LCA. These 5 studies are grouped together in Table 1 and for the meta-analysis. Two of the clinical trials (NCT03406104 and NCT03326336) are discussed but were not included in the meta-analysis because one had only averaged data of the patient population⁴³ and the other had a single patient enrolled.⁴⁴ 

Figure 1.

 

PRISMA flow diagram describing the study selection process for the systematic review.

Table 1.

 

Study Characteristics for Registered Trials and Post-Market Surveillance Studies

 

Study Characteristics

Eligible studies were published between 2008 and 2022 and were conducted in 11 countries, with some registered trials conducted in multiple countries. The highest number of trials were conducted in the United States of America (n = 15), followed by the United Kingdom (n = 6), Germany (n = 5), France (n = 5), Italy (n = 5), China (n = 2), and Israel (n = 2). The other countries involved were Australia, Canada, Saudi Arabia, and Brazil. 

A total of 595 patients were enrolled across all the studies (see Table 1). Sixteen of the 28 registered clinical trials (63%) involved dose escalation studies. The majority of trials (23 of 28) were early phase (I/II) clinical trials, and 20 had safety as the primary outcome, whereas 9 had treatment efficacy as the primary outcome (one trial identified both safety and efficacy as primary outcomes). The targeted gene/protein was RPE65 in 7 trials and 5 postmarket surveillance studies, ND4 in 7 studies, CHM/REP1 in 5 studies, sFLT-1 in 2 studies, CNGA3 in 2 studies, and one trial each for GUCY2D, MERTK, RPGR, RS1, and ChrimsonR. Virtually all of these gene therapy trials replaced genes that, when mutated, cause recessive or X-linked diseases. The notable exceptions were the age-related macular degeneration (AMD) trials,^45–50 which sequester vascular endothelial growth factor (VEGF) to inhibit neovascularization, and the ChrimsonR trial,⁴⁴ which used optogenetics to make neurons downstream of the degenerated photoreceptors sensitive to light in advanced retinitis pigmentosa (RP). The majority of studies (n = 24) used the AAV2 serotype, which can target photoreceptors, RPE, and retinal ganglion cells. The AAV2 studies enrolled 547 patients in total and treated LCA (autosomal recessive; RPE65),^{9–31,51–62} Leber hereditary optic neuropathy (LHON; mitochondrial; ND4),^43,63–79 AMD (multifactorial; sFLT-1),^45–50 choroideremia (X-linked; CHM/REP1),^80–89 RP (autosomal recessive; MERTK),⁹⁰ advanced RP (autosomal recessive treated with ChrimsonR),⁴⁴ and achromatopsia (autosomal recessive; CNGA3).⁹¹ Three studies used the AAV8 serotype to treat 36 patients with retinoschisis (X-linked; RS1),^92,93 RP (X-linked; RPGR),⁹⁴ and achromatopsia (autosomal recessive; CNGA3).^95–98 One study used the AAV4 serotype to treat 9 patients with LCA (RPE65).^56,57 One study used AAV5 to treat a patient with LCA caused by the GUCY2D mutation.⁹⁹ Of the gene therapy-treated diseases that we examined in the selected reports, only LHON involves retinal ganglion cell pathology. All the others involve RPE and photoreceptor dysfunction and/or loss. 

Vector dose varied from 2 × 10⁸ to 1 × 10¹² viral genomes (vg) per eye. The average dose injected was 1.1 × 10¹¹ vg/eye. The volume injected varied from 0.03 mL to 1 mL. There was a wide range in patient age and in stage of disease across the 29 studies. Participants were 70.8% men, aside from the 7 trials that targeted X-linked disease, which had only male participants (see Table 1). 

Ocular Adverse Events

For the purpose of this review, we defined ocular adverse events (AEs) as any AEs associated with the AAV vector or its delivery, and severe ocular adverse events (SAEs) as ocular AEs that resulted in long-term and treatment-refractive structural or functional damage. Overall, it is difficult to estimate the exact frequency of AEs as they were not systematically reported. Some studies identified all AEs, both mild and severe, whereas others only reported SAEs, and only if they were related to the therapeutic vector (and not the surgical procedure). Supplementary Table S2 provides a list and prevalence of the AEs reported in each of the studies. Many of the trials reported mild to moderate AEs associated with the surgical procedure for delivery of the therapy. The most common AEs were conjunctival hyperemia or hemorrhage, eye pain or irritation, blurred vision or diplopia, and transient changes in intraocular pressure (see Supplementary Table S2). These events would be expected in anyone undergoing subretinal surgery. Most of these were considered mild and resolved soon after the surgery. Another common AE was the development of cataract, a common and expected side effect of vitrectomy (which was performed in all gene therapy studies involving subretinal injections). Many of the studies used pre- or postoperative prophylactic immunosuppression to limit intraocular inflammation (see Supplementary Tables S3 and S4 for dosing regimens in subretinal and intravitreal studies, respectively). Despite this immunosuppression regimen, some studies still reported increased ocular inflammation that resolved on treatment with additional topical or systemic corticosteroids (see Supplementary Table S2). Ocular inflammation was most often associated with higher viral doses. 

Based on the reported occurrences, the pooled cumulative incidence for SAEs was 8% (95% CI of 5% to 12%, I² = 0.00; Fig. 2). Although the studies that reported SAEs ranged in follow-up from 1 month to 5 years, virtually all SAEs that were reported were identified within 30 days of gene therapy delivery and were associated with surgical complications. The subgroup analysis suggests that RCTs are associated with a slightly lower incidence of SAEs (3%, 95% CI of 1% to 7%), but the numbers are too small to make definitive conclusions. Supplementary Figure S1 groups the data based on disease (rather than trial type) and includes information on route of administration. Bainbridge et al.²⁶ reported a reduction in macular thickness in eyes in which the injection involved the fovea. This study also delivered the largest volume of therapeutic vector (up to 1 mL in the subretinal space); consequently, macular thinning may have been influenced by the duration and height of the detachment induced by the injection. Dimopoulos et al.⁸⁵ identified one patient with an intra-retinal immune response resulting in long-term decline in visual function and loss of outer retinal structures documented by optical coherence tomography (OCT). Russell et al.⁵⁴ reported one case of foveal dysfunction, most likely due to the surgical procedure, and one case of steroid-induced glaucoma leading to optic neuropathy after treatment for endophthalmitis. Xue et al.⁸³ identified retinal stretch (thinning) as a surgical complication. Morgan et al.⁸⁹ reported cone sensitivity loss in one patient following subfoveal vector delivery. 

Figure 2.

 

Cumulative incidence of severe ocular adverse events in published clinical trials.

Best Corrected Visual Acuity Improvement

With the exception of three trials (NCT02341807, NCT03116113, and NCT02317887), all studies had at least one patient (and often many more) that exhibited improvements in BCVA. The pooled proportion of patients with any improvement in vision was 41% (95% CI of 31% to 51%, I² = 65.62%; Fig. 3, Supplementary Fig. S2). These results were seen over follow-up timelines ranging from 1 month to 7 years. The results appear to be similar for all trial types. This proportion identifies the potential to see any improvement during the course of follow-up, including improvements that did not meet the clinically significant cut-off of 0.3 logMAR (15 letters). The proportions do not consider subsequent losses in BCVA that may have occurred with time after treatment. 

Figure 3.

 

Vision improvement using best corrected visual acuity (BCVA) for each of the published clinical trials.

Visual Sensitivity Improvement

Visual sensitivity improvements were assessed using full-field light sensitivity (FST), visual fields, or contrast sensitivity outcomes. Among 18 studies reporting sensitivity results, 12 studies reported improvements in at least 2 patients. Ten of these 12 studies showed improvements in at least half of their patient cohorts. The meta-analysis demonstrated that the overall proportion of patients with visual sensitivity improvement was 51% (95% CI of 31% to 70%, I² = 71.50%; Fig. 4, Supplementary Fig. S3). Visual sensitivity was assessed in these studies with follow-up periods ranging from 1 month to 6 years. RCTs are associated with the highest rates of visual sensitivity improvement (92%; with 95% CI of 70% to 98%), but with n = 2, it is difficult to make any general conclusions. 

Figure 4.

 

Visual sensitivity improvement in published trials.

Figure 5.

 

Incidence of immune involvement for each of the registered clinical trials with published findings.

Systemic Immune Involvement

Systemic immune involvement was most often determined by an increase in AAV antibodies or neutralizing antibodies, or by a cellular immune response. The pooled event proportion was 31% (95% CI of 21% to 42%, I² = 54.97; see Fig. 5, Supplementary Fig. S4). In the majority of cases, increases in any of these parameters were transient and reduced or returned to baseline levels during the follow-up period. Studies that reported on immune effects had follow-up times of 1 to 7 years. 

A summary of the results for all outcomes is provided in Table 2. 

Table 2.

 

Incidence and Confidence Intervals (CI) for Published Trials

 

Risk of Bias

Twenty-three of the 28 trials were early phase safety studies and only 6 of the 28 trials involved randomization. Most of the trials were associated with a high risk of bias. Only three of the studies were considered to have a low risk of bias across most of the categories (Supplementary Fig. S5). These three trials (NCT02652780, NCT02652767, and NCT03406104) were randomized, double-masked, sham-controlled phase III trials for LHON.^43,69,75 

Discussion

AAVs are limited in their payload capacity, preventing their use in gene therapies involving genes larger than 4.7 kb. Nevertheless, the relative success of the Luxturna studies for the treatment of LCA, followed by clinical trials for diseases such as RP, LHON, choroideremia, and achromatopsia, all of which have used various serotypes showing good tropism for RPE, photoreceptors, and retinal ganglion cells, have secured AAV vectors as the vectors of choice for ocular gene therapy. From our synthesis of the published literature, we can provide the following conclusions and some potential recommendations for future trials. 

Safety of AAV Therapy

Wild-type AAV integrates into the genome (primarily at a locus on chromosome 19), but the recombinant AAV vectors used for gene therapy appear to remain mostly episomal, significantly reducing the risk of insertional mutagenesis. Some insertional events have been recorded in mouse studies that have resulted in hepatocellular carcinomas following systemic delivery of the virus; however, genotoxicity has not been seen in larger animal studies (dogs and nonhuman primates), nor in any of the human clinical trials involving AAV therapy (reviewed in Ref. 100). The localized (rather than systemic) delivery of AAVs for ocular gene therapy further supports their safety in the eye. 

Overall, the current study shows that AAV ocular gene therapy is associated with an 8% cumulative incidence of SAEs. Whether this is an “acceptable” level of risk will depend on the particular circumstances of the individual patient and should be determined after extensive consultations with their health care provider. Most SAEs were likely caused by the surgical procedure used to deliver the therapeutic vector. Many of the trials were phase I safety trials that targeted patients with advanced disease. The retinas in these patients had limited structural and functional integrity, and were fragile and easily damaged. One of the most serious AEs following subretinal delivery was macular thinning.^26,83,89 A retrospective study by Gange et al.¹⁰¹ also identified perifoveal atrophy in some younger patients that had received Luxturna. This complication was found within and outside the site of vector delivery. Although the authors state that this is potentially due to vector toxicity, the possibility that this was associated with normal disease progression or high myopia cannot be excluded, especially because a good proportion of the fellow (untreated) eyes developed similar atrophic retinal changes. Other AEs related to the subretinal surgical procedure were subconjunctival hemorrhage, eye pain and irritation, changes in intraocular pressure, and cataract formation. All of these were treatable and transient and were not considered serious. 

It is noteworthy, however, that the reporting of AEs was quite variable across studies. Some studies identified all AEs in all patients (both mild and severe), including systemic events completely unrelated to the gene therapy. Other studies only reported on SAEs that did not resolve over time or with treatment, or resulted in permanent visual consequences. Some studies only reported events that were related to the therapeutic vector, ignoring AEs associated with the surgical procedure required to deliver the vector. Because the vector cannot be delivered without the surgical procedure, we can assume that the number of mild and transient AEs is higher than was reported. 

It is also difficult to assess inflammatory events related to the therapy as some of the study protocols used topical or systemic corticosteroid treatment before and after the vector delivery which would suppress the development of inflammatory side effects. The lengths of these treatments and dosages of the immunosuppressive drugs were variable (see Supplementary Tables S3, S4), again making it difficult to compare adverse inflammatory events between studies. Nevertheless, most inflammatory events were temporary and resolved with time or further treatment. Several studies reported ocular inflammation attributed to the delivery of higher viral doses rather than surgical delivery, but events were mostly mild and resolved with topical or systemic corticosteroids. In general, most studies reported that a dose of 1 to 1.5 × 10¹¹ viral genomes/eye was considered both effective and safe, and associated with minimal inflammation. 

To improve the safety profiles of ocular gene therapy delivery, higher doses should be avoided to prevent ocular inflammation. Subretinal injections should avoid the fovea to prevent iatrogenic macular thinning,¹⁴ except in cases of very reduced visual acuity or pre-existing foveal atrophy. In addition, intra-operative OCT can help guide the injection and avoid damaging the delicate retinas of patients with advanced disease.^{83,86,102,103} 

Efficacy of AAV Therapy

Among eligible trials, efficacy measures, as determined by BCVA and increases in visual sensitivity, revealed that 41% and 51% of patients, respectively, showed an improvement. These efficacy measures are greatly influenced by the success of the Luxturna trials, which, to date, have shown the most promising results. It remains to be seen if efficacy outcomes in the other gene therapy trials will match those of Luxturna, because many of the other trials are more recent and have shorter follow-up timelines. 

All of the clinical trials listed in Table 1 were not included in the meta-analysis to assess gene therapy efficacy. For example, NCT03406104, the phase III RESTORE trial for LHON, did not report any individual patient data but gave overall average gains across the entire cohort, making it difficult to determine if all patients responded to the treatment or if some patients had large gains that increased the average. For studies included in the meta-analysis, the proportions documented above represent any improvement throughout the treatment, even if the improvement was subsequently lost during follow-up. The BCVA improvements in many patients did not reach the 0.3 logMAR cutoff, which is generally considered clinically significant. Nevertheless, for patients with very limited visual capacity, small improvements in vision, or even maintenance of existing vision (i.e. halting or slowing disease progression), could translate into meaningful impacts on quality of life. 

The target population for our review was any patient with retinal degeneration eligible to receive adeno-associated viral gene therapy. The variability of patient characteristics across our included studies, including age, gender, and type and stage of retinal disease (see Table 1) help increase the external validity of our findings. However, the variability of patient characteristics also contributed to heterogeneous efficacy estimates across studies. It was difficult to generalize findings between diseases or between individuals with the same disease. The therapeutic outcome often depended on the stage of the disease and the physical state of the retina at the time of treatment. If vision was relatively good prior to treatment (as in some of the choroideremia trials⁸⁰), the therapeutic effects would not readily be evident until the disease advanced in the fellow eye. For example, one study found that 2 of 6 patients showed efficacy at the 6-month follow up,⁸¹ but 4 of 6 patients showed a therapeutic benefit at 3.5 years after treatment.⁸⁰ Additionally, some of the earliest studies assessed safety only, delivering the gene therapy to patients with advanced disease with little potential for a beneficial therapeutic outcome. On the one hand, it is possible that some of the eligibility criteria resulted in a selection bias where only patients deemed to be “good candidates” for a favorable therapeutic outcome were chosen. On the other hand, the multifactorial nature of diseases such as AMD necessitates a treatment that is not focused on a particular genetic mutation and makes treatment efficacy more elusive. 

It also still remains to be seen if the conditions for gene therapy and the therapeutic vectors have been appropriately optimized. Bainbridge et al.²⁶ suggested that higher viral doses gave better outcomes but were associated with more ocular inflammation. They suggest the need for vectors that can promote higher expression levels of the therapeutic transgene, thus allowing lower doses of the virus to be administered, lowering the risk of side effects without compromising efficacy. Jacobson et al.¹⁴ suggest multiple injections to cover more of the retina. It is also clear that the injection volume is important; the volume should be large enough to cover a significant area of the retina but not so large as to cause a significant detachment that would take longer to resolve and further compromise photoreceptor health. The majority of clinical trials (23 of 28) used delivery volumes of 300 µL or less. 

Delivery of therapy to younger patients who have relatively good functional vision may help to preserve this vision over time,¹⁰⁴ even though the effects of the therapy would take longer to detect.²⁶ However, age is not the sole predictor of efficacy. Older patients with some residual functional vision may be equally good candidates for gene therapy. Staging the disease and identifying pockets of functional retina may help to target the therapy and lead to an improved outcome.¹⁷ One study¹⁶ identified the presence of pseudo-foveas in patients at 9 to 12 months after treatment that persisted for up to 6 years of follow-up, suggesting that gains may be seen in the long term as the part of the retina not covered by the therapy continues to degenerate, thereby allowing patients to develop alternative fixation loci centered on the treatment area. 

The variability in efficacy outcomes may have been influenced by the immune response. Aside from immediate innate immune responses to the injection (such as inflammation), which are controlled by immunosuppression medications given around the time of viral delivery, numerous studies suggest that adaptive immunity needs to be taken into consideration. A study in nonhuman primates, conducted alongside the clinical trials, showed immune responses following intravitreal delivery of AAVs.⁹⁷ This, and other studies, suggest that baseline antibody titer should be considered when determining inclusion criteria as an immune reaction to the virus may affect the efficacy of an intravitreal injection.^50,97 However, it is difficult to determine what levels of neutralizing antibodies will be clinically relevant and influence the therapeutic outcome. It is estimated that up to 95% of the general population has been infected by wildtype AAVs¹⁰⁵ and 30% to 60% of the population carry neutralizing antibodies.¹⁰⁶ Although this might be a consideration for intravitreal gene therapy, it has been suggested to be less relevant for injections into the subretinal space, which is considered to be more immune privileged. The NCT01208389^{28–30,51,52} trial administered AAV therapy to the second eye years after the first eye was treated and saw no decrease in efficacy, suggesting that, at least for subretinal injections, immune interference with viral efficacy is not a concern. Nevertheless, 9 of 13 trials using subretinal injections have shown increases in antibody titers and in neutralizing antibodies to the AAV vectors in at least a subset of the treated patients, suggesting that subretinal injections can also lead to an immune response. The increases were often transient and returned to baseline during the follow-up periods, making it difficult to say whether they had any clinical significance in relation to efficacy outcomes, or if they will compromise future treatments with the same vector. At this time, definitive conclusions are difficult to make, as most gene therapy studies have small sample sizes. 

Unusual efficacy outcomes were obtained in all of the LHON clinical trials (NCT02161380, NCT02652780, NCT01267422, NCT02064569, NCT02652767, NCT03153293, and NCT03406104), which revealed improvements in both treated and fellow (untreated) eyes. Whereas the natural course of LHON disease involves some recovery of vision in a limited number of cases, it does not fully explain the improvement in fellow eyes, because untreated cohorts did not consistently show similar gains. Yu-Wai-Man et al.⁷⁵ used quantitative polymerase chain reaction (qPCR) to detect viral DNA in the fellow eye and optic nerve in nonhuman primates following unilateral injection. They suggest that inter-eye transfer of AAV or mitochondria may account for bilateral gains in efficacy. However, because the qPCR assay does not require viable virus to work, and it is unlikely that sufficient amounts of viable virus could be transferred between the two eyes, the likely explanation for bilateral gains in vision in these patients remains to be determined. 

Discussions of efficacy cannot avoid the lingering question in ocular gene therapy: how long does the therapy last? A single injection, delivered during a period of dysfunction but prior to the onset of degeneration in the canine model of LCA, was effective up to 11 years later.¹⁵ However, when the injection was delivered after the onset of degeneration, functional gains were seen but the structural degeneration continued. In patients with LCA, there is no dysfunction phase prior to degeneration, and thus patients showed a maintenance of functional gains at 3 years of follow-up, but the course of photoreceptor degeneration was not altered by the gene therapy.¹⁵ Examination of the same cohort at 6 years showed sustained functional improvements over baseline but progressive declines in the areas of improved vision, as would be expected if degeneration continued.¹⁷ Efficacy gains were lost in some LCA studies over a 3-year period,²⁶ but were retained at 4 years in the studies performed by the group that developed Luxturna.⁵⁸ In LHON, a 7-year follow-up showed persistent functional gains.⁷⁶ This variability between studies may reflect different virus manufacturing practices, different regulatory elements driving transgene expression,⁹ different patient populations, and different follow-up times for the various studies. In studies in which early gains were lost with time, it is also difficult to assess what contribution the immune system may have played in reversing these losses, because immune studies were not universally conducted across all trials. At this time, it is difficult to say how long the therapeutic effects will last in any of the ocular gene therapy trials, or in patients receiving Luxturna, as the length of follow-up is extremely variable between studies, and there are simply not enough long-term studies to determine the longevity of the functional gains. 

Limitations and Conclusions

Ocular gene therapy is in its relative infancy; thus, there were only a limited number of registered clinical trials to evaluate. The 28 trials and 5 postmarket surveillance studies described in this report represent 7 distinct diseases, affecting different parts of the retina, and with different modes of inheritance. They used four AAV serotypes, and assessed different promoters, dosages and stages of disease progression. They also used different criteria for evaluations of AEs. Some studies did not report any AEs that were unrelated to the therapeutic vector (i.e. surgical complications), whereas others reported all AEs, both ocular and systemic, in treated and untreated eyes. For each study, we determined the number of patients that showed a particular trait relative to the total number of patients, regardless of the magnitude or duration of the effect. Thus, our cumulative incidence values may overestimate the effect where one was present but of low magnitude, and may underestimate it where effect sizes were large but in only a subset of patients. Our analysis may also underestimate efficacy as some studies were strictly safety studies using patients with end-stage disease and limited potential for improvement. At this time, definitive conclusions are difficult to make, as most gene therapy studies have small sample sizes, lack a control arm, and potentially suffer from methodological bias. 

Overall, AAV ocular therapy is a promising approach for the treatment of ocular disease. Nevertheless, dosage, site of delivery, volume, and number of injections have to be carefully considered to promote efficacy but avoid AEs. 

Acknowledgments

The authors would like to thank Risa Shorr (Learning Services, The Ottawa Hospital) for assisting in the creation of the literature search and Zeinab Daham for help with data extraction. M.M.L. is supported by The Ottawa Hospital Anesthesia Alternate Funds Association and a University of Ottawa Junior Clinical Research Chair in Innovative Translational Research. CT is supported by the Don and Joy Maclaren Chair for Vision Research. 

Supported by Fighting Blindness Canada's Restore Vision 2020. 

Data Availability Statement: All data generated or analyzed during this study are included in this published article (and its supplemental information files). 

Disclosure: M. Sobh, None; P.S. Lagali, None; M. Ghiasi, None; J. Montroy, None; M. Dollin, None; B. Hurley, None; B.C. Leonard, None; I. Dimopoulos, None; M. Lafreniere, None; D.A. Fergusson, None; M.M. Lalu, None; C. Tsilfidis, None