Introduction
From March 2021 to March 2026, adeno-associated virus (AAV) gene therapy has achieved remarkable clinical milestones. By March 2026, seven AAV-based gene therapies had FDA approval in the United States, including two products approved before this review window—Luxturna (2017) and Zolgensma (2019)—and five additional products approved from 2022 to 2024: Hemgenix, Roctavian, Elevidys, Kebilidi, and Beqvez 620. Despite this progress, broad clinical translation remains constrained by immunologic barriers, manufacturing complexity, safety signals at high doses, and access limitations. This review synthesizes pivotal regulatory, clinical, and translational advances while analyzing root causes of limited adoption across modalities and target classes.
Clinical Landscape and Regulatory Approvals (2021–2026)
Approved Therapies and Label Dynamics
Five AAV-based therapies received FDA approval between 2021 and 2026, joining previously approved Luxturna (voretigene neparvovec-rzyl, AAV2, 2017) and Zolgensma (onasemnogene abeparvovec-xioi, AAV9, 2019). Key approvals include Hemgenix (etranacogene dezaparvovec, AAV5) for hemophilia B in November 2022 at a dose of 2×10¹³ vector genomes per kilogram (vg/kg)21, Roctavian (valoctocogene roxaparvovec, AAV5) for hemophilia A in June 2023 at 6×10¹³ vg/kg23, Elevidys (delandistrogene moxeparvovec, AAVrh74) for ambulatory Duchenne muscular dystrophy (DMD) in 2023 at 1.33×10¹⁴ vg/kg22, Kebilidi (eladocagene exuparvovec, modified AAV2) for aromatic L-amino acid decarboxylase deficiency via intracranial delivery in 2024, and Beqvez (fidanacogene elaparvovec, AAVrh74var) for hemophilia B in April 202420.
However, post-approval label restrictions signal emerging safety concerns. In November 2025, FDA revised Elevidys labeling to exclude non-ambulatory DMD patients following two fatal acute liver failure cases, adding a boxed warning for hepatotoxicity and mandating a 200-patient prospective observational study with 12-month hepatotoxicity monitoring22. This action underscores dose-dependent risks in systemically administered, high-dose AAV therapies.
Durability and Real-World Evidence
Long-term durability data from pivotal trials demonstrate sustained transgene expression in select indications. Hemgenix's HOPE-B trial reported mean Factor IX activity of 37.4 IU/dL at 4 years (n=47), with 94% of patients remaining free of continuous prophylaxis and a 90% reduction in annualized bleeding rate (ABR)16. Similarly, Roctavian's GENEr8-1 trial showed mean Factor VIII levels of 27.1 IU/dL (one-stage assay) at 4 years, with 85% ABR reduction and 99% reduction in Factor VIII usage16. In a separate Phase I/II cohort, 7-year median Factor VIII levels plateaued at 10.3 IU/dL in the 6×10¹³ vg/kg group, with 96% ABR reduction maintained1610.
For Zolgensma, presymptomatic treatment of spinal muscular atrophy (SMA) Type 1 in the SPR1NT trial achieved 100% event-free survival at 14 months, with 100% independent sitting, 79% standing, and 64% independent walking in the 2-copy SMN2 cohort—outcomes never observed in natural history16. Long-term follow-up (mean 6.86 years, n=10) demonstrated 100% survival and ventilation-free status, with 70% achieving ventilatory independence1611.
In contrast, Elevidys's EMBARK Phase III trial failed its primary endpoint at 52 weeks (North Star Ambulatory Assessment [NSAA] change +2.6 vs. +1.9 placebo, p=0.24), though 2-year secondary endpoints showed statistical significance (+2.63 vs. -0.25 external control, p<0.01)16. This discordance highlights challenges in functional endpoint selection and disease heterogeneity in neuromuscular trials.
Routes of Administration and Biodistribution
Systemic intravenous delivery remains dominant for liver-targeted (hemophilia) and multi-systemic (neuromuscular) indications, though dose intensity correlates with safety risks. High-dose regimens (≥5×10¹³ vg/kg) for SMA and DMD trigger both innate and adaptive immune responses, including hepatotoxicity (~90% incidence in onasemnogene abeparvovec recipients, managed with prednisolone), thrombotic microangiopathy (TMA; 9 cases in SMA, 4 in DMD at doses ≥5×10¹³ vg/kg), and myocarditis (reported in DMD trials)12.
Local delivery routes offer dose-sparing advantages. Intrathecal AAV9 administration for SMA Type 2 (STEER trial, 1.2×10¹⁴ vg total dose) achieved primary endpoint improvement in Hammersmith Functional Motor Scale-Expanded (HFMSE) versus sham (+2.39 vs. +0.51, p=0.0074)16. Subretinal injection for Luxturna (1.5×10¹¹ vg/eye) minimizes systemic exposure, though procedure-related adverse events (conjunctival hyperemia, cataract, retinal tears) occur in 57% of injected eyes2625. Biodistribution studies demonstrate rapid CNS and systemic dissemination following intrathecal AAV5, with preferential dorsal root ganglia (DRG) targeting and possible macrophage-mediated clearance7. Notably, intra-nerve AAV6/AAV9 delivery in nonhuman primates produces no DRG toxicity, contrasting with systemic/intrathecal routes9.
Immunology and Redosing Barriers
Pre-Existing Neutralizing Antibody Prevalence
A global prospective study (n=546 hemophilia A patients across 9 countries) reported AAV5 neutralizing antibody (NAb) seroprevalence of 34.8% (day 1), the lowest among tested serotypes (AAV2: 58.5%; AAV6: 48.7%; AAV8: 45.6%; AAVrh10: 46.0%)3. Geographic variability was substantial: United Kingdom 5.9%, United States 26.8%, Brazil 26.9%, Germany 28.1%, Japan 29.8%, France 37.2%, Italy 40%, Russia 46.2%, South Africa 51.8%3. AAV5 antibody titers among seropositive individuals were 1–2 orders of magnitude lower than other serotypes, and serostatus remained stable over 6 months (R²=0.808 at 3 months, R²=0.592 at 6 months)3.
Post-Dosing Immune Responses and Redosing Obstacles
Post-infusion NAb levels remain elevated for years, precluding redosing with the same serotype. In a 13-year hemophilia B follow-up, AAV8 NAb levels remained persistently high10. T-cell responses to AAV capsid and transgene product contribute to loss of efficacy; prednisolone administered at dosing or upon transaminase elevation abrogated T-cell responses in hemophilia trials, though some patients failed immunosuppression at very high doses1.
Emerging immune-modulation strategies under investigation include IgG-degrading enzymes (IdeS/IdeZ), FcRn inhibitors, plasmapheresis, capsid switching, and tolerogenic nanoparticle regimens, though clinical validation remains limited in the 2021–2026 timeframe615. Prophylactic corticosteroids showed no benefit over reactive approaches in Roctavian's NCT04323098 trial, with 20/22 patients experiencing ALT elevations despite prophylaxis16.
Safety and Toxicology
Dose-Dependent Adverse Events
Thrombotic microangiopathy (TMA) emerged as a critical dose-dependent complication at ≥5×10¹³ vg/kg, with 9 SMA cases (ages 4 months–4 years, all female) and 4 DMD cases (ages 7–12 years, male) reported. Symptom onset occurred 6–12 days post-infusion, presenting with vomiting, hypertension, oliguria, elevated creatinine, proteinuria, hemolytic anemia, and thrombocytopenia. Most patients responded to plasmapheresis, steroids, hemodialysis, and eculizumab (complement inhibitor), though one patient in each trial died from TMA complications1. Postmarketing surveillance of onasemnogene abeparvovec identified 4 additional TMA cases among 665 adverse event reports, all occurring 6–11 days post-dosing2.
Hepatotoxicity manifested as ALT/AST elevations in ~90% of onasemnogene abeparvovec recipients (90/102 patients), with initial elevations at day 7, near-resolution by day 14, and transient increases at month 12. All clinical trial hepatotoxicity events resolved with prednisolone. However, postmarketing data revealed 14 cases with clinical signs (jaundice, ascites, coagulopathy) and 4 acute liver failure cases2. Elevidys's fatal hepatotoxicity cases led to restricted labeling and enhanced monitoring requirements22.
Dorsal root ganglia toxicity was observed in nonhuman primates after intrathecal delivery and in one ALS trial patient receiving intrathecal AAV with prednisolone, who developed ganglionopathy with neurological symptoms and MRI changes1. Route-specific risk profiles suggest intrathecal/systemic administration poses higher DRG toxicity risk than localized intra-nerve delivery9.
Emerging Toxicity Signals
AAV-driven nephrotoxicity via NFκB signaling pathways represents an emerging safety consideration for high-dose systemic delivery8. Myocarditis was reported in 2 Pfizer DMD trial patients (1 fatal) and 1 Sarepta patient (resolved with steroids)1. Postmarketing Luxturna data identified chorioretinal atrophy in approximately 28% of treated eyes in some real-world series, occurring within weeks to months post-treatment, though causation (vector-related, inflammation, or mechanical) remains unsettled2526.
Manufacturing and CMC Landscape
Production Platforms and Scalability
Three primary platforms dominate clinical AAV production: transient transfection in HEK293 cells (69% of trials), baculovirus expression vector systems (BEVS) in Sf9 insect cells (20%), and mammalian viral infection-based methods19. HEK293 suspension systems in serum-free media achieve titers >10⁵ vg/cell and >10¹⁴ vg/L, though batch-to-batch plasmid variability and large plasmid quantities raise costs19. BEVS platforms yield up to 10¹² vg/mL with improved full/empty ratios but face baculovirus instability and post-translational modification deviations (e.g., insect-cell-derived capsid glycosylation differs from human patterns)1913.
Comparative genome quality analysis revealed BEVS-produced vectors exhibit higher degrees of truncated and unresolved genome species versus HEK293 platforms, with "mutated" inverted terminal repeats (ITRs) correlating with inaccurate genome abundance13. Despite platform differences, no statistically significant variations in clinical safety, efficacy, or durability have been observed across manufacturing systems19.
Downstream Purification and Analytics
Affinity chromatography (e.g., AVB Sepharose) and ion-exchange chromatography enable scalable purification with high purity (goal >90% full particles), though full/empty capsid separation and impurity removal (host cell DNA <10 ng/dose, host cell proteins, residual plasmid) remain challenging1219. Analytical characterization employs digital droplet PCR for genome titer, ELISA for capsid titer, and mass spectrometry for capsid integrity; differential scanning fluorimetry assesses stability19.
FDA's January 2026 flexible CMC guidance permits process refinement across clinical phases, eliminates strict three-lot process performance qualification requirements, and allows concurrent commercial lot release during validation18. EMA guidelines mandate biodistribution studies (ICH S12), environmental risk assessment, germline transmission testing, and long-term patient follow-up for gene therapy medicinal products17.
Modality Comparison: AAV vs. Lentiviral vs. LNP Systems
AAV vectors dominate in vivo gene delivery due to episomal persistence, broad serotype-specific tropism, and proven clinical efficacy, with payload capacity limited to ~4.5 kb (or ~2.2 kb for self-complementary AAV)642. Lentiviral vectors support larger payloads, integrate durably into host genomes (enabling ex vivo hematopoietic stem cell therapies for hemophilia, SCID-X1, beta-thalassemia), and achieve polyclonal long-term engraftment, but carry insertional mutagenesis risk and require predominantly ex vivo application444651. Lipid nanoparticles (LNPs) deliver transient mRNA payloads with manufacturing simplicity, reduced immunogenicity versus AAV capsids, and systemic liver targeting, yet transient expression limits utility for chronic genetic diseases requiring durable correction434951.
For CRISPR delivery, AAV's size constraint necessitates dual-vector approaches for standard Cas9 or use of compact nucleases (e.g., sRGN3.1) to enable all-in-one packaging436. LNP-mRNA platforms deliver Cas9 transiently with superior storage stability versus AAV and enable systemic editing, though durability is limited43. Ex vivo HSPC editing employs AAV6 for homology-directed repair donor delivery (5,000–100,000 vg/cell) alongside Cas9 RNPs, achieving high editing rates but triggering DNA damage response pathways and potential long-term fitness deficits4248.
The field is converging on platform-specific niches: AAV for in vivo, tissue-specific delivery (retina, CNS, liver, muscle); lentiviral for ex vivo HSPC integration; LNP for transient mRNA-based therapies515255.
Barriers to Broad Clinical Practice and Solution Roadmap
Eligibility Constraints and Access Limitations
Pre-existing NAb prevalence excludes 30–60% of patients depending on serotype and geography, with no validated clinical protocol for antibody removal36. High upfront costs (Hemgenix list price $3.5 million, Zolgensma ~$2.1 million) and specialized treatment center infrastructure restrict access. Health technology assessment bodies demand long-term durability evidence and outcomes-based agreements, while lack of redosing options limits applicability in progressive diseases2023.
Manufacturing Cost and Scalability
Multi-systemic indications require doses >10¹⁴ vg total, raising production costs and posing scale challenges even with optimized yields619. Empty capsid content (typically 70–92% empty in some preparations) reduces payload efficiency and may contribute to immunogenicity19. Process optimization via machine learning and real-time monitoring aims to reduce empty capsids and improve consistency658.
Near-Term Innovations
Capsid engineering via directed evolution and machine learning has yielded variants with enhanced transduction, reduced immunogenicity, and altered tropism, with multiple engineered capsids entering clinical evaluation535850. Transient immune modulation (e.g., tolerogenic nanoparticles, sequential capsid dosing) is under preclinical investigation6. Dual-AAV and multi-vector strategies address payload limitations for large genes651. Policy changes to support outcomes-based reimbursement and expanded treatment center networks may improve access, though implementation timelines remain uncertain.
Conclusion
AAV gene therapy has transitioned from an investigational platform to an approved therapeutic modality for select monogenic diseases. However, translation across broader patient populations is constrained by dose-dependent toxicities (TMA, hepatotoxicity, DRG neurotoxicity), pre-existing immunity excluding 30–60% of candidates, manufacturing complexity limiting scalability, and access barriers related to cost and infrastructure. Advances in capsid engineering, immune modulation, manufacturing analytics, and regulatory flexibility provide a roadmap toward expanded applicability, yet fundamental challenges—redosing limitations, payload constraints, and safety at systemic high doses—require continued innovation before AAV can achieve truly broad clinical penetration. For clinicians, patient selection must balance measurable biomarker-driven eligibility (NAb titers, viable target tissue) against realistic expectations of durability and re-treatment options. For translational teams, prioritizing indications amenable to local delivery, lower doses, and immunologically privileged sites (retina, CNS) may accelerate near-term impact while next-generation platforms mature.