Leber hereditary optic neuropathy (LHON) is a rare mitochondrial disorder characterised by acute/subacute bilateral vision loss, which in most cases persists throughout life.^{1 2} This maternally inherited condition is caused by variants in mitochondrial DNA, most commonly affecting the genes encoding the subunits of complex I of the mitochondrial respiratory chain.³ The three primary disease-causing variants─11 778G>A in MT-ND4, 3460G>A in MT-ND1 and 14 484T>C in MT-ND6─account for approximately 90% of LHON cases.³ A growing number of nuclear genes have also been reported to cause autosomal recessive LHON.⁴

LHON selectively targets retinal ganglion cells, leading to their degeneration and subsequent optic nerve atrophy.⁵ Hence, the phenotype of the disease is largely monosymptomatic with visual loss as its defining clinical feature.¹ Not all patients carrying a mitochondrial DNA variant associated with LHON will develop visual loss.¹ The variable penetrance and male predominance of the disease suggest the involvement of additional genetic, hormonal and/or environmental factors in its expression.68

The only approved treatment to date for LHON is idebenone (Raxone^{, Chiesi Farmaceutici), a synthetic analogue of coenzyme Q₁₀, which acts as an electron shuttle able to bypass dysfunctional complex I of the electron transport chain. Idebenone has been approved in Europe since 2015 at a dose of 900 mg/day for the treatment of LHON, regardless of the causative variant.9 Data from a retrospective observational study and an expanded access programme (EAP) showed that patients were more likely to recover if idebenone was initiated within 12 months of vision loss and treatment provided for at least 24 months.10 11 Recently, the results of the non-randomised controlled LEROS study demonstrated that idebenone was effective in improving visual function or stabilising vision at 12 and 24 months in 181 patients treated within 5 years of symptom onset, compared with a matched external natural history cohort.12 It has also been recently clarified that two polymorphic variants in the gene encoding the enzyme NQO1 affect idebenone efficacy, providing a potential explanation for the lack of response in ~50% of treated patients.13 NQO1 is key to presenting idebenone in its therapeutically effective reduced form.}

Differently, lenadogene nolparvovec is an intravitreal gene therapy product designed for patients with LHON due to the m.11778G>A MT-ND4 variant, which intends to permanently correct the underlying causative genetic defect. In two randomised phase 3 studies, 76 patients with vision loss ≤6 months (RESCUE¹⁴) or between 6 and 12 months (REVERSE¹⁵) were treated with a single unilateral intravitreal injection of gene therapy versus a sham injection. In another phase 3 randomised, placebo-controlled study (REFLECT), 98 participants with vision loss duration ≤12 months received a single injection of lenadogene nolparvovec in their first-affected eye and either a second injection of lenadogene nolparvovec or a placebo injection in their second-affected eye.¹⁶ These three studies revealed an unexpected contralateral effect, in which the eye treated with sham or placebo showed an improvement in best-corrected visual acuity (BCVA) alongside the treated eye.1416 Persistence of the benefit from lenadogene nolparvovec was demonstrated until 5 years post-treatment.¹⁷ To provide further evidence of the efficacy of lenadogene nolparvovec, visual acuity data from 174 MT-ND4 patients from the phase 3 studies were pooled and compared with an external control group of 208 untreated patients, showing a statistically significant improvement in final BCVA in favour of gene therapy over natural history.¹⁸

To determine the relative treatment effects of lenadogene nolparvovec and idebenone, we performed two matching adjusted indirect comparisons (MAICs), namely, LEROS versus REFLECT and EAP versus REFLECT.

The MAIC methodology enabled the study populations to be weighed on four key covariates that may impact the response to treatment: age, sex, baseline BCVA and time from vision loss to treatment injection (lenadogene nolparvovec) or initiation (idebenone). The first MAIC comparing lenadogene nolparvovec data from REFLECT to idebenone data from LEROS demonstrated that the responder rate for CRR from nadir at 24 months was doubled for patients receiving one-time bilateral injections of lenadogene nolparvovec compared with patients treated with idebenone over 24 months.²³ The large overlap of the two weighted study populations made the results of the comparison highly relevant. Importantly, all sensitivity analyses showed comparable results to the main analysis, confirming its robustness.

In the second MAIC comparing idebenone data from the EAP with lenadogene nolparvovec data from REFLECT, there was a large reduction (~67%) in sample size due to adjustment for time from vision loss to treatment. The low overlap of the two study populations limited the value of the results of the base case model. Removing time from vision loss to treatment as a covariate in the leave-one-out analysis revealed a significantly higher CRR in favour of lenadogene nolparvovec versus idebenone (76.7% vs 39.0%; p<0.001). The post hoc sensitivity analysis comparing patients from the EAP with patients from RESCUE revealed a CRR from nadir at 24 months almost doubled with lenadogene nolparvovec compared with idebenone.²³ The results of both MAICs are therefore consistent.

Comparison of the efficacy of idebenone and lenadogene nolparvovec in the MT-ND4 population was already conducted through a meta-analysis of CRR from nadir and final BCVA.²⁴ A better recovery rate was estimated with lenadogene nolparvovec than with idebenone, with 59% (54% to 64%) of eyes injected with gene therapy achieving a CRR from nadir versus 31% (24% to 40%) of eyes treated with idebenone.²⁴ The rates of visual recovery from nadir obtained by this method are in line with those obtained in our MAICs.

Interestingly, initial CRR in both MAICs occurred within a comparable mean time from injection of the gene therapy and initiation of the idebenone treatment (approximately 9 to 11 months), with a range in time to response comparable between the two treatment options. These results indicate that MT-ND4 patients treated within 12 months from vision loss achieved a clinically significant recovery on average within the first year after gene therapy injection or idebenone initiation, and, for almost all cases, within the following 24 months. Other studies have corroborated this time window for recovery with idebenone treatment.^{10 25} Regarding gene therapy, other results confirm a recovery within a fixed time frame, ranging mainly from 3 to 24 months after injection of MT-ND4 LHON patients who received similar AAV2-ND4 gene therapies.^{26 27}

For the MAIC LEROS versus REFLECT, the change from baseline of BCVA at 24 months was assessed and no statistically significant difference was observed between the two treatment arms. These results are not surprising. In LHON disease, the initial (baseline) BCVA varies considerably during the subacute phase of the disease. In lenadogene nolparvovec studies, initial BCVA ranged from LogMAR ≤0 to hand motion, including eyes with preserved BCVA, which makes the baseline value less suitable for evaluating treatment efficacy. Indeed, a proportion of lenadogene nolparvovec patients had no capacity for improvement from their baseline BCVA, explaining the overall comparable improvement with idebenone when analysed based on LogMAR change from baseline at 24 months. The nadir (lowest visual acuity), which is usually reached more than 6 months after vision loss, is a more stable and relevant reference point for assessing treatment response^{2 28} and this parameter was used in the MAIC LEROS versus REFLECT.

Due to the risk of reduction in the ESS of the 48 patients bilaterally treated with gene therapy in REFLECT (96 eyes) and the inherent potential limitation of data interpretation, sensitivity analyses comparing the idebenone data from LEROS or EAP to the pooled lenadogene nolparvovec data (348 eyes) were provided. These sensitivity analyses demonstrated a statistically significant higher probability of responding to treatment with gene therapy compared with idebenone for the two MAICs (p<0.0025 for both base case models with a large overlap between the two study populations). This suggests that the improved efficacy of lenadogene nolparvovec over idebenone in the MT-ND4 patients was consistent across all lenadogene nolparvovec studies, whether the injection was unilateral or bilateral.

One of the strengths of this work is the use of the MAIC, a robust statistical methodology developed and applied recently in some therapeutic areas.^{21 22} Regulators have been favourable to this type of advanced statistical techniques that reduce bias when comparing clinical trial data to external comparator arms.²⁹ Furthermore, health technology assessment bodies have endorsed the MAIC methodology.^{22 30 31} To our knowledge, these are the first MAICs comparing the effects of lenadogene nolparvovec and idebenone in the MT-ND4 LHON population. Importantly, the similarity between the populations to be compared, which is illustrated by the large overlap after matching for the MAIC LEROS versus REFLECT and the post hoc analysis EAP versus RESCUE, makes the results of the comparisons highly relevant. Since LHON is a progressive disease, the time from disease onset to treatment injection/initiation could influence the final outcome. Therefore, to ensure comparability between the lenadogene nolparvovec and idebenone studies, only patients treated within 1 year from disease onset, which corresponds to the subacute/dynamic phase of the disease, were selected. The choice of the 24-month post-treatment time point for the outcomes was relevant for assessing the persistence of the treatment effect.

These two MAICs present some limitations. The low overlap between the EAP and REFLECT populations limits the robustness of the MAIC EAP versus REFLECT. However, as noted above, sensitivity analyses comparing EAP with RESCUE and the lenadogene nolparvovec pool revealed significant overlap and provided reliable results. The absence of ethnicity as a covariate may be a limitation. Patients from Asia carrying the MT-ND4 variant have less severe loss of visual acuity than those from Europe and North America.^{32 33} However, because of the absence of data on the number of Asian patients in EAP, and the inclusion of only one Asian patient in LEROS,^{11 12} a bias related to ethnicity is unlikely. Regarding the MAIC LEROS versus REFLECT, three of the four matching covariates (age, sex and time from vision loss to treatment) were available only for the whole LHON population and not for the subgroup of interest, that is, MT-ND4, in the LEROS study. Nevertheless, as baseline BCVA was available in the MT-ND4 population, we performed a sensitivity analysis using it as the only matching covariate, and the results were consistent with those of the main analysis. Some minor differences between lenadogene nolparvovec and idebenone studies in terms of LogMAR assessments may have favoured idebenone for the calculation of change in LogMAR BCVA, and lenadogene nolparvovec for the switch from off-chart to on-chart. However, sensibility analysis addressing this methodological difference yielded results consistent with those of the main analysis. It would have been more appropriate to account for the eye-level effect and within-patient correlation in the analyses, but this was not possible due to individual data from the LEROS study and EAP being unavailable. Lastly, the use of non-parametric methods for variables with a skewed distribution was not feasible due to the aggregated nature of the data from comparative studies (LEROS and EAP).

In conclusion, unanchored MAICs comparing idebenone and lenadogene nolparvovec clinical studies demonstrated a higher response rate at 24 months in MT-ND4 patients who received a single bilateral injection of lenadogene nolparvovec compared with MT-ND4 patients treated with a regimen of 900 mg/day of idebenone. However, to compare these two treatment options as rigorously as possible, a randomised controlled trial would be necessary. These results support lenadogene nolparvovec as a promising and effective therapeutic option for patients with LHON caused by the most common m.11778G>A mitochondrial DNA variant.