Eladocagene Exuparvovec for the Treatment of Aromatic l-Amino Acid Decarboxylase Deficiency (AADCd): An Economic Evaluation from a US Perspective

Abstract

Background

Recently, the gene therapy eladocagene exuparvovec received accelerated approval from the US Food and Drug Administration (as eladocagene exuparvovec-tneq) for treatment of aromatic l-amino acid decarboxylase deficiency (AADCd), a rare, infantile-onset disorder characterized by developmental delays.

Objectives

To conduct a US, modified societal perspective cost-utility analysis comparing eladocagene exuparvovec versus best supportive care (BSC).

Methods

Multistate survival modeling was implemented tracking disease progression from a “no motor function” health state to achievement of motor-function improvements, measured by: (1) multiples of the meaningful score difference (MSD) of the Peabody Developmental Motor Scales-Second Edition (PDMS-2) total score and (2) motor milestones. Eladocagene exuparvovec trials informed clinical inputs. Health-state utilities were from a US time-trade-off study that valued AADCd quality-of-life impacts. Outcomes were discounted (3%); costs were reported in 2024 US dollars. Scenario analyses, characterizing alternative approaches of the multistate survival model analyses and probabilistic sensitivity analysis to assess the impact of parameter uncertainty, were conducted.

Results

Discounted incremental quality-adjusted life-years (QALYs) for eladocagene exuparvovec were 20.83 (multiples of the MSD of total PDMS-2) and 18.44 (motor milestones). Incremental cost per QALY ranged from $199,007–$224,104. The scenario and sensitivity analyses results supported the validity of the base case analysis.

Conclusions

Eladocagene exuparvovec is associated with considerable QALY gains compared with BSC. Within the context of other ultra-rare and/or one-time treatments, eladocagene exuparvovec provides substantial clinical improvements at lower cost than many other rare-disease treatments. Findings from this study highlight that eladocagene exuparvovec is an important treatment option for patients with AADCd.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40273-025-01542-8.

Key Points for Decision Makers

Illustrated by the substantial life years (LY) and quality-adjusted life-years (QALY) gains estimated, this economic evaluation reflects the benefits of eladocagene exuparvovec compared with best supportive care for treatment of aromatic l-amino acid decarboxylase deficiency

The QALY gains estimated exceed those for other available ultra-rare disease and one-time therapies, emphasizing the value of eladocagene exuparvovec as an important treatment option for patients with AADCd.

Introduction

Aromatic l-amino acid decarboxylase deficiency (AADCd) is a rare, neurometabolic disorder caused by an autosomal recessive mutation in the dopa decarboxylase (DDC) gene [1]. Prevalence estimates vary across regions, with global estimates ranging from 1/32,000 to 1/1,300,000 [2, 3]. As DDC is the final enzyme in the biosynthesis pathway for dopamine and serotonin, patients with AADCd are unable to synthesize these neurotransmitters, resulting in the inhibition of motor development [4].

Disease onset generally occurs within the first few months of life and is characterized by motor dysfunction and developmental delays. Patients experience very severe clinical manifestations across a wide range of symptoms and functional issues, including movement disorders, muscle weakness, pain, autonomic dysfunction, gastrointestinal dysfunction, and behavioral problems, as well as mood and sleep disturbance [4–6]. In addition, impairments to motor function, communication, and cognitive function may occur [7]. In a systematic review focused on burden and disease severity in AADCd, across included studies, most individuals were nonambulatory with a broad spectrum of other disease-related disorders [8]. Patients’ ability to be independent is severely impacted, especially among those with severe developmental delays. Consequently, patients with AADCd are reliant on caregivers throughout their lifetime [4, 9]. A study of caregiver burden found that the primary caregiver spent, on average, 109 hours per week on direct and indirect caregiving activities, with additional support from their partner [10]. Moreover, the most severe patients are bedridden and have heightened risk of mortality in the first decade of life [4, 11, 12].

In addition to the humanistic burden of AADCd, a number of studies have explored the impact of AADCd on the healthcare system and have noted substantial healthcare resource use, particularly among nonambulatory patients. A study in Italy found hospitalizations up to twice a year among nonambulatory patients [13], while in France the number was three times a year [14].

Until recently, patients with AADCd were managed by standard of care therapies that aimed to provide symptom relief, such as dopamine agonists, monoamine oxidase inhibitors, pyridoxal phosphate/pyridoxine, anticholinergic agents, folic acid, l-dopa, 5-hydroxytryptophan, benzodiazepines, melatonin, and selective serotonin-reuptake inhibitors [4]. However, as these therapies do not address the underlying mechanism of disease, for the most part, they do not enable the achievement of developmental milestones or improvement in motor function.

In 2024, the gene therapy eladocagene exuparvovec-tneq received approval from the US Food and Drug Administration (FDA) [15] for the treatment of adult and pediatric patients with AADCd. Eladocagene exuparvovec is a recombinant adeno-associated virus serotype 2 (rAAV2) vector containing the human DDC gene that encodes the AADC enzyme and compendial excipients [12]. The therapy is administered as four intraputaminal infusions in one surgical procedure. Administration of eladocagene exuparvovec restores AADC enzyme activity, leading to increased dopamine production [16].

In clinical trials of eladocagene exuparvovec [17–19] patients demonstrated improvements in motor function as measured by Peabody Developmental Motor Scales-Second Edition (PDMS-2) total score [19], as well as other measures of motor function. Moreover, by the end of their follow-up, the majority of patients achieved ≥1 motor milestones including full head control, ability to sit unassisted, ability to stand, and in notable cases, ability to walk freely without assistance [19]. Given the substantial burden experienced by patients with AADCd and their caregivers, the improvements in function and development resulting from treatment with eladocagene exuparvovec serve to address an important unmet medical need.

In addition to clinical trials, other research has supported the benefit of eladocagene exuparvovec for the treatment of AADCd. A recent study estimated that the meaningful score difference (MSD) for total PDMS-2 score is 40 points, and that over 85% of patients treated with eladocagene exuparvovec in clinical trials achieved the MSD at 18 months [20]. Moreover, improvements in total PDMS-2 score were correlated with improvements in cognition and language domains on the Bayley Scale of Infant and Toddler Development, Third Edition (Bayley-III), suggesting that eladocagene exuparvovec can benefit nonmotor function as well [20]. Another study estimated the health-utility improvement in AADCd associated with attainment of motor milestone health states; consequently, improvements in quality of life associated with the achievement of motor milestones could inform the value of novel treatments such as eladocagene exuparvovec [21]. This study sought to assess the cost-effectiveness of eladocagene exuparvovec compared with best supportive care (BSC) for the treatment of patients with AADCd, from a US modified societal perspective, using the latest evidence and a multistate survival modeling (MSM) approach.

Methods

The present cost-effectiveness analysis used a cohort state-transition model with two different MSM approaches, informed by individual patient-level (IPD) data from three, single-arm, clinical studies of eladocagene exuparvovec [17–19] to compare with BSC. This study updates a previously published model [22] by using a new modeling approach and longer follow-up data and health state utilities for AADCd. The new modelling approach is presented here as the base case, and the approach used in the original study is presented as a scenario analysis.

Model structure

The cost-effectiveness model was structured as five motor function-related health states:

• i.No motor function
• ii.Full head control
• iii.Sitting unassisted
• iv.Standing with support
• v.Walking with assistance

along with death as an absorbing state (Fig. 1, Table 1). Outcomes were modeled over two phases. During the first phase (the developmental phase), comparative effectiveness of the treatment arms was modeled as differential achievement of motor milestones, with patients able to achieve improvement in (i.e., “higher”) motor milestones. In the second phase of the model (the long-term phase), patients were assumed to maintain the highest motor milestone that they achieved during the developmental phase. Mortality risk was modeled in both phases.

Fig. 1.

Model diagram. Bold lines reflect transitions between alive health states (without skipping motor milestones); dashed lines reflect transitions to mortality. In Scenarios 2 and 3, skipped milestone achievement was imputed at the midpoint between corresponding adjacent visits, and only transitions 1–4 were modeled in the developmental phase. In the base-case analysis (reflected above), transitions 1–10 were estimated allowing skipping in the MSM, for the N = 30 patients treated in clinical studies AADC-1601, AADC-010, and AADC-011(data on file). Motor milestone achievement with eladocagene exuparvovec was estimated based on multiples of the MSD of total PDMS-2 score, as it is expected to predict improvement in motor function with high certainty, and to be more sensitive to change than the five-level classification of motor milestones. FHC, full-head control; MSD, meaningful score difference; MSM, multistate survival modeling; no motor function, NMF; PDMS-2, Peabody Developmental Motor Scales-Second Edition sitting unassisted, SU; standing with support, SWS; WWA, walking with assistance

Table 1.

Health state definitions, multiples of meaningful score difference, utilities (described by Monteleone et al. [21]), and disutilities (described by Landfeldt et al. [33])

Health states	Definition	Multiples of MSD for total PDMS-2b	TTO (N = 114a) Mean (SD; 95% CI)	Mean disutility (SE) of caregiver burden
HS1: Bedridden	No motor function	0	− 0.258 (0.534; − 0.356, − 0.160)	0.170 (0.034)
HS2: Full head control	Patient can sit supported at his/her hips and holding his/her head aligned while rotating his/her head to follow a toy for 4 to 7 seconds.	1 (40 pts)	− 0.155 (0.569; − 0.259, − 0.050)	0.153 (0.031)
HS3: Able to sit unaided	Patient is required to sit without support and maintain balance while in sitting position for 30 to 59 seconds.	2 (80 pts)	0.452 (0.523; 0.356, 0.548)	0.055 (0.011)
HS4: Standing with support	Patient is able to take two to three alternating steps, either in place or in forward motion, with support around the trunk.	3 (120 pts)	0.775 (0.242; 0.731, 0.819)	0.003 (0.001)
HS5: Walking with assistance	Patient can walk 4–7 feet with alternating steps, with minimal support.	4 (160 pts)	0.796 (0.235; 0.753, 0.839)	0.000

All patients started from the “no motor function (NMF)” state

CI, confidence interval; HS, health state; MSD, meaningful score difference; PDMS-2, Peabody Developmental Motor Scales-Second Edition; SD, standard deviation; SE, standard error; TTO, time-trade-off; VAS, visual analog scale

^aSix participants were excluded as they did not understand the task [21]

^bMeaningful score difference was derived from an analysis of data from patients (n = 30) from three single-arm clinical studies of eladocagene exuparvovec. An MSD of 40 points yielded specificity >0.95 using the receiver operating characteristic approach, and generally aligned with the mean difference approach [20]

In the original model [22], motor milestone achievement was assessed based on scores for 4 items of the PDMS-2 (see Sect. 2.8) using data from clinical trials AADC-1601, AADC-010, and AADC-011 [17–19]. In the base case for this study, using updated PDMS-2 score data from those trials, MSM was conducted for the achievement of multiples of the MSD of total PDMS-2 score. MSD of total PDMS-2 score in AADCd was previously estimated at 40 points in an analysis anchoring improvement in motor milestones. The MSD of 40 points had specificity > 0.95 (false positive rate < 5%) for predicting improvement in motor milestones [20]. Multiples of the MSD were mapped to the corresponding motor milestone states (Table 1), and cumulative incidence of achieving each multiple of the MSD, conditional on the prior multiple, was modeled. In previously published research [20], using the MSD for Total PDMS-2 score to define health states has been shown to provide greater sensitivity in measuring improvements compared with motor milestone achievement because it captures a broader range of both gross and fine motor domains (reflexes, stationary performance, locomotion, object manipulation, grasping, visual-motor integration; Fig. A1). As a result, improvement can be ascertained earlier than using the motor milestone approach [23, 24]. Further, patient-level correlations between measures collected in clinical studies of eladocagene exuparvovec indicate a high level of association between cognitive development and motor function. In particular, correlations between change from baseline in total PDMS-2 and Bayley-III scores (cognition and language domains) were of large magnitude and statistically significant from month 6 onwards: r = 0.599 (p = 0.0032) at month 6, r = 0.796 (p = 0.0002) at month 18, and r = 0.861 (p = 0.0007) at month 60 [20]. Accordingly, capturing improvements in motor function may act as a proxy for impacts on other symptoms and consequences. Between the two approaches, the motor milestone approach (presented here as a scenario analyses) may be considered the more conservative analysis.

The model was developed in Microsoft^® Excel^® for Microsoft 365 MSO 64-bit and R statistical software, version 4.2.3 (R Foundation for Statistical Computing) was used for the MSM.

Study Population

The modeled population reflected N = 30 patients from clinical studies AADC-1601 (n = 8), AADC-010 (n = 10), and AADC-011 (n = 12) (as well as the long-term extension study AADC-1602). Patients had a diagnosis of AADC deficiency with a classical presentation of the disease (i.e., with complications such as developmental retardation, hypotonia and oculogyric crises). Mean (SD) age was 1.1 (0.7) years at diagnosis and 3.8 (2.2) years at treatment; 14 patients (46.7%) were female. All patients started the model from the no motor function health state, at baseline age of 4 years (reflecting the mean in clinical studies) [17–19].

Comparator

Eladocagene exuparvovec efficacy and safety have been studied in the AADC-1601, AADC-010, and AADC-011 clinical trials [17–19], as well as the extension study AADC-1602, which evaluated long-term outcomes of patients with AADCd without head control at baseline.

Perspective of Anaslysis

The analysis was conducted from a modified US societal perspective. The modified perspective reflects quality-of-life impacts to parents and caregivers, beyond impacts on patients.

Time Horizon

The model had a lifetime horizon and cycle length of 3 months (the shortest interval between visits in the clinical studies).

Model Parameters

An overview of the model parameters can be found in Table 2.

Table 2.

Model input sources

Parameter	Value	Source
Model population
Eladocagene exuparvovec patient population	N = 30 patients (from clinical studies of eladocagene exuparvovec) Mean (SE) age: 4.0 (0.40) years Proportion female: 52.4%	Clinical trials AADC-1601, AADC-010, AADC-011 and AADC-1602 [17–19] [data on file]
Motor milestone achievement
BSC	At end of 12-year developmental phase: NMF: 97.10%, FHC: 0.00%, SU: 1.45%, SWS: 0.00%, WWA: 1.45%	Systematic review of natural-history data: Bergkvist et al (2022) [12]
Eladocagene exuparvovec	At end of 12-year developmental phase: Base case (MSM using MSD of total PDMS-2, with skipping, N = 30): NMF: 0.00%, FHC: 12.86%, SU: 21.44%, SWS: 15.78%, WWA: 49.91%	Clinical trials AADC-1601, AADC-010, AADC-011 and AADC-1602 [17–19] [data on file]
Mortality
AADCd mortality	On the basis of cerebral palsy as a proxy—see Table B1, Fig. B1	Advisory board meetings Brooks et al. (2014) [26]
Background mortality	On the basis of US life tables	Arias et al. (2023) [25]
Adverse events
Eladocagene exuparvovec	Moderate and serious/severe AEs modeled on the basis of the clinical trials (assumed to be administration related and modeled in the first cycle) Table C2 (incidence rates and duration).	Clinical trials AADC-1601, AADC-010, and AADC-011, AADC-011 and AADC-1602 [17–19] [data on file]
BSC	Included in the utility and healthcare management costs by health state	Model design/assumption
Health utilities
AADCd health state utilities	See Table 1—Mean (SE, derived from SD and N) utility by health state: NMF: − 0.258 (0.050), FHC: − 0.155 (0.053), SU: 0.452 (0.049), SWS: 0.775 (0.022), WWA: 0.796 (0.022)	TTO valuation of vignettes in a sample of N = 120 US general population respondents [21]
Caregiver disutility	On the basis of Duchenne muscular dystrophy as a proxy. See Table 1—Mean (SE) disutility by health state: NMF: − 0.170 (0.034), FHC: − 0.153 (0.031), SU: − − 0.055 (0.011), SWS: − 0.003 (0.001), WWA: 0.000 (NA) Modeled as 2.2 in the NMF health state, and reduced linearly to 1.2 in the WWA health state, reflecting reduced burden with improved patient motor function.[34]	Landfeldt et al. (2016) [33] NICE evaluation of TA755[34]
Bereavement disutility	An annual utility decrement of 0.04 for both parents until age 80 years.	Song et al. (2010) [35]
Disutilities of TEAEs	On the basis of EQ-5D-3L index scores reported from responses to the Medical Expenditure Panel Survey in the USA. See Tabe C2—mean (SE) disutility by AE type: Dyskinesia: − 0.067 (0.013) Pneumonia: − 0.034 (0.007) Gastrointestinal disorders: − 0.051 (0.010) Gastroenteritis: − 0.073 (0.015)	Sullivan et al. (2011) [28]
Healthcare resource use
BSC therapies	Use of BSC therapies, as described by the International Working Group on Neurotransmitter Related Disorders, is modeled as including dopamine agonists, MAO inhibitors, pyridoxal phosphate/pyridoxine, anticholinergic agents, folic acid, L-dopa (with or without carbidopa), benzodiazepines, melatonin, dietary supplements, and vitamin D. See Table C6.	Wassenberg et al. (2017)[4] Saberian et al. (2022)[30]
Medical management	Follow-up healthcare visits (see Table C3), medical procedures (see Table C4), and technical procedures and resources (see Table C5).	Saberian et al. (2022) [30] Clinical-expert opinion
BSC drug acquisition costs	Wholesale acquisition cost per pack, see Table C1	Merative Micromedex Red Book 2024
Eladocagene exuparvovec drug acquisition cost	Wholesale acquisition cost of $3,950,000 per 2.8×1011 vg vial	Merative Micromedex Red Book 2024
Eladocagene exuparvovec administration	Cost of the stereotactic procedure is modeled as $106,030, reflecting mean total hospital charges for selective dorsal rhizotomy at age ≤18 years, assumed as proxy.	Optum360 [29]
Medical management resource costs	Follow up visits: see Table C1 Technical procedures: see Table C1 Medical procedures: see Table C1	2024 Current Procedural Terminology (CPT®) codes listed by the Centers for Medicare and Medicaid Services (CMS)
Adverse events management costs	Moderate and severe AEs, cost per event: See Table C1	Costed from a US perspective, based on estimates sourced from Optum360 by diagnosis-related group (DRG) codes [29]
Discounting
Effects	Annual rate of 3.0%, as recommended in US guidelines for cost–utility analysis	Sanders et al. (2016)[48]
Costs

AADCd, aromatic l-amino acid decarboxylase deficiency; AE, adverse event; BSC, best supportive care; CP, cerebral palsy; FHC, full head control; MAO, monoamine oxidase; MSD, meaningful score difference; MSM, multistate survival modeling; NMF, no motor function; PDMS-2, Peabody Developmental Motor Scales-Second Edition; QoL, quality of life; SD, standard deviation; SE, standard error; SU, sitting unassisted; SWS, standing with support; TEAE, treatment-emergent adverse events; TTO, time-trade-off; WWA, walking with assistance

Progression of Disease Under BSC

As previously described [22], the composition of BSC was informed by the International Working Group on Neurotransmitter Related Disorders guidelines [4] and the results of a clinician survey. BSC efficacy was based on published natural history data [12]. In the natural history study, patients who were aged < 4 years on average, had not achieved full head control (or a higher motor milestone) at diagnosis, and were treated with BSC, provide a comparable sample of patients to those treated with eladocagene exuparvovec in the three aforementioned clinical studies. Within approximately 1 year of BSC treatment, the highest motor milestone achieved was walking with assistance for 1.45% (n = 1/69) of patients, and sitting unassisted for another 1.45% [12]. Accordingly, during the developmental phase, the highest motor milestone achieved with BSC was modeled as 1.45% walking with assistance, 1.45% sitting unassisted, and 97.10% no motor function.

Treatment Efficacy

Longer-term clinical trial data [17–19] than were included in the original CEA analysis [22] were incorporated in this model, for the 30 patients treated with eladocagene exuparvovec. Data up to June 2023 included motor milestone achievement observed over a follow-up of ≥ 2 years for 26 (87%), and ≥ 5 years for 19 (63%) patients (follow-up range 0.75–10 years). On the basis of these data, in the base case, at the end of the 12-year developmental phase 0.00% of patients were assumed to be in the NMF health state; 12.86% in the FHC health state, 21.44% in sitting unassisted (SU); 15.78%, SWS; and 49.91%, WWA. Among patients with 5 + years of data, a sustained treatment effect has been supported by emission tomography data that demonstrated the durability of gene transduction effect and were consistent with the durability of motor milestone development [19]. Over the study period, the majority of patients achieved ≥1 motor milestones [19]. Of further note, over the course of the clinical trial follow-up, few severe, treatment-related adverse events were reported [17, 19].

Survival

Mortality was modeled as including disease-related risk varying by motor milestone state and background mortality (i.e., of the general population), which was consistent across health states. Background mortality was modeled based on US life tables [25]. As there is currently no published survival data for patients with AADCd, disease-related mortality risk associated with motor milestone health states was informed by an analog condition, cerebral palsy [26], which was previously confirmed by clinical experts as an appropriate proxy [22, 26]. Survival data for those with cerebral palsy were adapted for patients with AADCd by mapping the estimates based on motor function states and feeding ability in cerebral palsy to the health states used in the present model for AADCd (Supplementary Fig. B1). Parametric models (exponential, Weibull, Gompertz, log-normal, log-logistic, and gamma) were estimated, and the best fitting models were identified based on goodness-of-fit statistics and clinical plausibility of the extrapolations (see Supplementary Information A). Modeled survival curves for full head control, sitting unassisted, standing with support, and walking with assistance health states closely aligned with the estimates for cerebral palsy. For the no motor function health state, the modeled survival was slightly lower than the estimates for cerebral palsy, with a median survival of 15 years modeled for AADCd versus approximately 15–20 years for cerebral palsy. According to clinical expert input, the median survival of 15–20 years would be longer than expected for a patient with AADCd and no motor function [27].

Adverse Events

The model incorporated adverse events (AEs; moderate, serious and severe) from the three clinical trials (Supplementary Tables C1, C2). The AEs incorporated were eladocagene exuparvovec treatment-related and thus applied to that treatment arm and were assumed to occur only in the cycle immediately after treatment administration [28], AEs owing to treatment with BSC were expected to be included into the disease management cost estimates for the management of AEs sourced from Optum360 (2012) by diagnosis-related group (DRG) codes, as well as health-state disutilities [29].

Healthcare Resource Use

Modeling of healthcare resource use (HCRU) included follow-up healthcare visits, medical procedures, technical procedures and resources, and BSC pharmacotherapies, and was costed based on unit costs (see Supplementary Tables C1, C2, C3, C4, C5). HCRU varied by motor-milestone health state, as well as by treatment arm (i.e., BSC versus eladocagene exuparvovec). Estimates of HCRU modeled were based on Saberian et al. [30], who report HCRU estimates by disease severity (NMF or FHC, n = 8; SU or SWS, n = 2; WWA, n = 10). These estimates were mapped to the motor-milestone states. As a result of limited data for the SU and SWS states, HCRU for SWS is assumed to be the same as that for WWA, and HCRU for SU is assumed to be the average of the NMF/FHC and WWA estimates.

Costs

Unit costs can be found in Supplementary Table C1.

State Costs

Only direct healthcare-related costs are considered due to data limitations regarding indirect costs (i.e., nonhealthcare related costs including work productivity of patients, work productivity of parents/informal caregivers, formal caregiving costs, home-modification costs, transportation costs). Cost categories reflected in the model included those that are relevant to the management of AADCd: follow-up healthcare visits, medical procedures, and technical procedures and resources (see Supplementary Tables C3–C5). Cost estimates were based on 2024 Current Procedural Terminology (CPT®) codes listed by the Centers for Medicare and Medicaid Services (CMS) [31]. For each cost category, unit costs were multiplied by the frequency of healthcare resource use within each cycle. The cost was applied to the distribution of patients in each motor milestone state per cycle, to calculate total costs. Cost estimates for the management of AEs were sourced from Optum360 (2021) by diagnosis-related group (DRG) codes [29]. A summary of estimated costs by health states in the base case can be found in supplementary table D4.

BSC Costs

The proportion of patients using BSC was specified by motor milestone health state and treatment arm (Supplementary Table C6). All patients treated with eladocagene exuparvovec were assumed to continue BSC treatment thereafter. List prices of BSC pharmacotherapies in the US were sourced from Merative Micromedex Red Book 2024 [32].

Eladocagene Exuparvovec-Tneq Costs

For costing of eladocagene exuparvovec-tneq treatment, the list price (wholesale acquisition cost, WAC) in the USA of $3,950,000 (per 2.8 × 10¹¹ vg vial) was used [32], as well as the cost of administration to the putamen of the brain through a stereotactic neurosurgical procedure. The cost of administration, $106,030, is a proxy value for the cost of the stereotactic procedure (reflecting mean total hospital charges for selective dorsal rhizotomy at age ≤ 18 years, Optum360) [29].

Utilities

Health state utilities were from a US time-trade-off (TTO) study valuing the quality-of-life impact of AADCd by motor milestone state [21] (Table 1). The impacts of variation in motor function on patients with AADCd were described across five motor milestone state “vignettes.” TTO assessment of the quality-of-life variation between the health states was then assessed by a general population sample of N = 120 individuals in the USA [21].

Caregiving for a patient with AADCd has a profound impact on caregivers’ quality of life, affecting physical health and emotional wellbeing, social/leisure activities, relationships, work, and finances [9]. Thus, disutility estimates reflecting caregiver burden were considered from a proxy condition, Duchenne muscular dystrophy [33], and scaled according to the relative patient health utility modeled on the basis of the TTO study described above (Table 1), on the basis that caregiver burden is expected to vary with the health of their patients. The number of caregivers was based on a previous National Institute for Health and Care Excellence (NICE) evaluation of risdiplam in SMA (TA755) [34], and assumed to be reduced with improved motor function. The number of caregivers required per patient (and accordingly, experiencing disutility) was modeled as 2.2 in the NMF health state, and reduced linearly to 1.2 in the WWA health state, reflecting reduced burden with improved patient motor function. In addition, the quality-of-life impact of a patient’s death on parents was incorporated into the model as a bereavement disutility, which was obtained from Song et al. [35]. In that study the long-term effects of child death on bereaved parents was examined; both mothers and fathers experienced an ongoing utility decrement of 0.04, as observed after more than 35 years. Song et al. [35] reported the mean age of parents at their child’s death as 25.7 years for males and 28.1 years for females. For parents who experienced the death of a patient before age 55 years, it was assumed that both parents would experience an annual 0.04 disutility until age 80 years.

Disutilities of AEs for eladocagene exuparvovec were modeled on the basis of published estimates [28].

Model Outcomes

The model’s primary outcomes were overall quality-adjusted life years (QALYs), overall costs, incremental life years (LYs), incremental QALYs, and the incremental cost-utility ratio (ICUR) for eladocagene exuparvovec versus BSC. Treatment outcomes, namely expected LYs, QALYs, and costs, were calculated separately by treatment arm, as well as incrementally (i.e., eladocagene exuparvovec versus BSC). All analyses are presented as discounted (3% annually), with undiscounted analyses presented in the appendix. Overall survival was also calculated for each treatment. Note that eladocagene exuparvovec acquisition, administration, and adverse event costs are not discounted as these occur in the first cycle.

Scenario Analyses

Several scenario analyses were conducted, characterizing alternative versions of the MSM analyses. In the first, aligning with the outcome modeled for eladocagene exuparvovec in the previously published CEA [22], motor milestone achievement (no motor function, full head control, sitting unassisted, standing with support, walking with assistance) was modeled based on updated (longer-term) data from clinical studies [20]. Motor milestone achievement was assessed in study AADC-1601, and served as the primary efficacy endpoint in studies AADC-010 and AADC-011, based on five classification levels [17–19]. Classification levels included no motor function, as well as four determined based on the following four components of the PDMS-2: full head control (Stationary item 10), sitting unassisted (Stationary item 14), standing with support (Locomotion item 28), and walking with assistance (Locomotion item 34). Achievement of these items was defined as mastery (PDMS-2 item score of 2 points) or emerging/partial mastery (PDMS-2 item score of 1 point).

Additional scenarios were also considered (1) omitting two patients with follow-up <12 months in clinical studies, for whom the required “independence assumption” of the MSM may not hold, and (2) imputing achievement of “skipped” milestones at the midpoint of the time between corresponding adjacent visits in the clinical studies (Fig. 1). These scenarios were informed by appraisal of the clinical validity of the MSM approach by clinical-expert authors, with rationale as described further in the supplementary materials. Eladocagene exuparvovec efficacy inputs for these scenario analyses (scenarios 2–6) can be found in Supplementary Table C7.

Finally, a scenario analysis was run excluding caregiver and bereavement disutility.

Probabilistic Sensitivity Analysis

Probabilistic sensitivity analysis (PSA) was conducted to assess the combined impact of parameter uncertainty on the base-case results using the total PDMS-2 MSD approach. Variables for which parameter values were sampled in the PSA included: percentage female, baseline age, disease-related mortality, motor milestone achievement, health state utility values, disutility from caregiver burden, disutility from bereavement, disutility from AEs, AE incidence, AE duration, and healthcare resource use estimates. The PSA was conducted using Monte Carlo simulation over 1000 iterations, with parameters and associated distributions described further in Supplementary Table C8.

Validation

The use of motor function-related health states in the model was determined on the basis that impaired motor development is among the most important consequences of AADCd. This choice reflects available natural history evidence and was confirmed in advisory board meetings by clinical experts (n = 8) from six countries selected for their previous experience with economic models for rare diseases. Three participants also had previous experience with advanced therapy medicinal products (ATMPs). Qualitative modelling insights were collected during two separate meetings by using metric tools and questionnaires to measure the level of consensus, both in real-time and in between the discussions.[22, 27] The model’s five motor milestone health states were informed by five motor milestone state “vignettes,” which were validated by N = 5 healthcare professionals with experience managing AADCd and N = 4 caregivers of patients with AADCd [21]. Face validity of model inputs, results, and assumptions were reviewed with clinical-expert authors of the manuscript. Mortality risk associated with motor milestone health states was informed by an analog condition, cerebral palsy, which was previously confirmed by clinical experts as an appropriate proxy [22].

To inform the appropriateness of MSM compared with a Markov transition model, the fit of nonconstant-hazard parametric models (i.e., Weibull, Gompertz, log-normal, log-logistic, and gamma) was compared with an exponential parametric model for each transition (i.e., no motor function to higher milestone, full head control to higher milestone, SU to higher milestone, SWS to WWA). Best fit was assessed by lowest Akaike information criterion (AIC) and Bayesian information criterion (BIC). The computerized CEA model was examined by study authors with experience conducting health-economic modeling and who were not primary developers of the model. Additional details on model validation can be found in Supplementary Information A, including the Assessment of the Validation Status of Health-Economic decision models (AdViSHE) tool (Table A1).

Ethics Statement

This study used de-identified secondary data and therefore did not require ethical approval or individual patient consent.

Results

Base Case Results

Health State Membership

Health state membership traces, reflecting the share of the modeled population by health state over time, are shown in Fig. 2. Over the horizon of the analyses, patients receiving BSC primarily remained in the no motor function health state. With eladocagene exuparvovec treatment, membership in higher health states (i.e., improved motor milestones), in particular walking with assistance, was observed.

Fig. 2.

Health state membership over the horizon of the cost-effectiveness analysis. BSC, best supportive care; FHC, full head control; NMF, no motor function; SU, sitting unassisted; SWS, standing with support; WWA, walking with assistance

Discounted Costs

Total costs (Table 3) were higher for eladocagene exuparvovec ($4,352,439) compared with BSC ($207,566), corresponding to an incremental cost of $4,144,873 for eladocagene exuparvovec over BSC. Drug acquisition costs were the largest component of total costs for eladocagene exuparvovec, while disease management costs were the largest component of total costs for BSC (for undiscounted costs see Supplementary Table D1; for estimated costs by health states in the base case see Supplementary Table D2).

Table 3.

Summary of estimated discounted costs and CUA results in the base case (MSD of total PDMS-2 score approach) and scenario analysis 1 (motor milestone achievement approach)

		MSM using MSD of total PDMS-2 (base case)			MSM using MM achievement (sensitivity analysis)
		EE	BSC	Incremental	EE	BSC	Incremental
Discounted (3%) costs, US$
Total costs		4,352,439	207,566	4,144,873	4,340,503	207,566	4,132,937
Breakdown of total costs
Drug acquisition		4,222,470	96,693	4,125,777	4,212,752	96,693	4,116,059
Disease management		121,604	110,873	10,731	119,387	110,873	8514
Follow-up visits		113,519	103,242	10,276	111,383	103,242	8141
Technical procedures		279	569	− 290	316	569	− 253
Medical procedures		7806	7061	745	7688	7,061	626
Adverse events		8365	0	8365	8365	0	8365
Dyskinesia		2394	0	2394	2394	0	2394
Pneumonia		2385	0	2385	2385	0	2385
Gastrointestinal disorders		1952	0	1952	1952	0	1952
Gastroenteritis		1634	0	1634	1634	0	1634
Discounted (3%) results
Total LY	24.60		12.75	11.85	22.98	12.75	10.22
Total QALY	12.52		− 8.31	20.83	10.13	− 8.31	18.44
Breakdown of total QALYs
Patients	14.39		− 2.67	17.06	12.39	− 2.67	15.06
Bereavement (parents)	− 0.30		− 1.10	0.80	− 0.40	− 1.10	0.70
Caregiver burden	− 1.56		− 4.54	2.98	− 1.84	− 4.54	2.70
Adverse events	− 0.01		0.00	− 0.01	− 0.01	0.00	− 0.01
Cost per QALY				$199,007			$224,104

Background mortality was modeled based on US life tables, and mortality risk associated with motor milestone health states informed by an analog condition, cerebral palsy [26]

BSC, best supportive care; CUA, cost-utility analysis; EE, eladocagene exuparvovec; LYs, life years; MM, motor milestone; MSD, meaningful score difference; MSM, multistate survival modeling; PDMS-2, Peabody Developmental Motor Scales-Second Edition; QALYs, quality-adjusted life-years

Overall Survival

Overall survival (Fig. 3) with eladocagene exuparvovec was a median of 61 years and 19 years with BSC.

Fig. 3.

Overall survival, eladocagene exuparvovec versus BSC. BSC, best supportive care; MSD, meaningful score difference; MSM, multistate survival model; PDMS-2, Peabody Developmental Motor Scale, Second Edition

Discounted Life-Years and Quality-Adjusted Life-Years

In the base case, 24.60 LYs were achieved with eladocagene exuparvovec, an incremental LY gain of 11.85 compared with BSC. QALYs were 12.52 for eladocagene exuparvovec, an incremental gain of 20.83 QALYs compared with BSC (Table 4). Caregiver disutility with BSC contributed approximately 10%-15% of the QALY gains achieved (for undiscounted QALYs see table C1).

Table 4.

Summary of CUA results in scenarios varying specifications of the MSM, discounted

	Measure of motor function improvement	Transitions modeledin the MSM	Patient sample	Incremental costs	Incremental LYs	Incremental QALYs	ICUR
Base case	Total PDMS-2 score MSD	With skipping	All patients (N = 30)	4,144,873	11.85	20.83	199,007
Scenario 2	Motormilestones	With skipping	All patients (N = 30)	4,132,937	10.22	18.44	224,104
Scenario 3	Total PDMS-2 score MSD	Without skipping	All patients (N = 30)	4,144,875	11.86	20.85	198,796
Scenario 4	Motormilestones	Without skipping	All patients (N = 30)	4,132,642	10.15	18.29	225,995
Scenario 5	Total PDMS-2 score MSD	With skipping	Excluding patients 011-313 and 1601-07	4,144,858	11.94	21.06	196,802
Scenario 6	Motor milestones	With skipping	Excluding patients011-313 and 1601-07	4,133,225	10.31	18.64	221,783

Scenarios 5 and 6 excluded two patients (011-313 and 1601-07) who were included in the base case, one lost to follow-up due to travel restrictions associated with coronavirus disease 2019, one with limited motor-functioning assessments owing to an injury deemed unrelated to treatment but impacting assessment of treatment effect (data on file)

ICUR, incremental cost–utility ratio; LY, life-year; MSD, meaningful score difference; MSM, multistate survival modeling; PDMS-2, Peabody Developmental Motor Scales-Second Edition; QALY, quality-adjusted life-year

Incremental Cost–Utility Ratio (ICUR)

The cost per QALY gained was $199,007 (for the undiscounted ICUR see Supplementary Table C1).

Scenario Analyses

As in the base case, in scenario analysis 1 (the motor milestone achievement approach) total costs (Table 3) were higher for eladocagene exuparvovec ($4,340,503) compared with BSC ($207,566), corresponding to an incremental cost of $4,132,937 over BSC. Overall survival (Figure 3) with this approach was 52 years for eladocagene exuparvovec and 19 years with BSC. Modeling motor milestone achievement resulted in 22.98 LYs, an incremental LY gain of 10.22 over BSC; 10.13 QALYs were achieved for eladocagene exuparvovec, an incremental gain of 18.44 QALYs over BSC (Table 3). The cost per QALY gained using this approach was $224,104.

In additional scenario analyses characterizing alternative versions of the MSM analyses, results remained consistent with the base case analysis (Table 4). Using the total PDMS-2 MSD approach, in the scenario analyses discounted QALY gains ranged from 20.85 to 21.06 compared with 20.83 in the base case. Using the motor milestones approach, discounted QALY gains ranged from 18.29 to 18.64 in the scenario analyses compared with 18.44 in the base case. The (discounted) ICUR also remained consistent, varying from $196,802 to $198,796 compared with $199,007 in the base case using the total PDMS-2 MSD approach, and from $221,783 to $225,995 (discounted) compared with $224,104 to using the motor-milestones approach. Undiscounted results can be found in Supplementary Table D3. Discounted and undiscounted results from the scenario analysis excluding caregiver and bereavement disutility can be found in Supplementary Table D4. Their exclusion does not materially impact the results; the undiscounted ICUR is < 150k, and discounted ICUR is < 250K, consistent with the base case.

Probabilistic Sensitivity Analysis

Summary results (discounted) of the probabilistic sensitivity analysis assessing the combined impact of parameter uncertainty in the base case analysis using the total PDMS-2 MSD approach are provided in Supplementary Table D5. Mean incremental QALY gains for eladocagene exuparvovec compared with BSC were 20.69 (95% CI: 16.32–25.42), mean incremental costs were $4,145,438 (95% CI: $4,130,785–$4,159,343), and the mean ICUR was $202,907 per QALY (95% CI: $162,841–$253,913). Incremental QALYs and costs estimated in each iteration are presented in Fig. D1.

Model Validation Results

Findings from the review of the face validity of the model inputs, results, and assumptions were used to inform the scenario analyses reported in this manuscript. In model fit testing, nonexponential models had best fit for all transitions modeled. Accordingly, the use of the MSM approach to modeling motor-milestone achievement, rather than a traditional Markov health-state transition approach was deemed suitable. Additional details on model validation can be found in Supplementary Information A.

Discussion

Eladocagene exuparvovec is the first approved gene therapy for the treatment of AADCd and has been shown in clinical studies to provide significant and clinically meaningful improvements in motor function and development [17–19]. In the present study, two approaches (base case and scenario analysis 1) were used to model the efficacy of eladocagene exuparvovec compared with BSC for the treatment of patients with AADCd from a US perspective. In both cases, the findings demonstrated greater LY and QALY gains with eladocagene exuparvovec compared with BSC. LY and QALY gains were larger using the MSD for total PDMS-2 approach, which reflects a wider scope of motor skills development. Even with the less sensitive approach based on motor milestone achievement, eladocagene exuparvovec was shown to provide survival and QALY benefits, supporting the validity of these findings.

Of the two approaches, using multiples of MSD for total PDMS-2 score to define health states provides greater sensitivity in measuring improvements because it captures a broad range of both gross and fine motor domains. The PDMS-2 was used in the three clinical studies of eladocagene exuparvovec to assess motor skills development comprising various subtests, including Reflexes, Stationary, Locomotion, Object Manipulation, Grasping, and Visual-Motor Integration [36–39]. In contrast, motor milestones in the eladocagene exuparvovec clinical studies were based on items that mapped to the Stationary and Locomotion subtests only [40, 41]. As such, improvements in total PDMS-2 score may reflect improvements in a wider variety of skills that contribute to child development.

A previous modeling study in the United Kingdom (UK) estimated the long-term benefits of eladocagene exuparvovec in patients with AADCd compared with BSC, and reported lower LY and QALY gains than the present study [22]. The UK model was based on the same motor-milestone health states as modeled in the present analysis; however, motor milestone achievement was modeled using a Bayesian growth model to generate extrapolations of total PDMS-2 score by patient, based on parametric assumptions informed by patients with greater follow-up, followed by a cumulative ordered logit model, which predicted the probability of achieving a higher motor milestone conditional on projected total PDMS-2 scores. While the previous model accordingly depended on parametric extrapolations of outcomes, with associated uncertainty, the present analysis uses MSM of a patient’s probability of achieving a higher motor milestone based on cumulative-incidence analysis of observed outcomes in clinical studies. By using long-term clinical trial data evidencing sustained treatment effect, the MSM avoids the need for assumptions about long-term extrapolation. This approach also did not require extrapolation of PDMS-2 score, which would be associated with uncertainty. Moreover, MSM allows for the probabilities of transitioning between states to vary over time, which may allow for the modeling process to better fit observed health state membership over time. As for all health-state transition models, it does require an independence assumption that censored patients have similar outcomes as those who are still followed, and this was explored with clinical-expert authors and tested in scenario analyses, which supported the validity of the base case results. Consequently, the MSM approach offers greater simplicity and reduced uncertainty, relative to the previous study.

The health state utilities incorporated in this study used lead-time time trade-off (TTO), which allowed the utilities of health states considered “worse than death” to span between − 1 and 0 [21]. As such, the most severe health states (“bedridden” and “full head control”) are negative. This valuation method reflects the clinical disease presentation and burden, as participants in the utilities study perceived these severe health states as worse than dead. Additionally, the values are in line with those estimated for metachromatic leukodystrophy (MLD) [42], a similarly grave rare, neuromuscular disease.

While discounted outcomes are the primary focus of the manuscript, the authors argue that consideration of the undiscounted results is also important so as not to bias against near-term intervention to achieve long-term outcomes. As eladocagene exuparvovec is a one-time treatment, its costs are incurred at the start of the model, while its benefits extend over the model’s time horizon. Accordingly, in discounted analyses, drug costs of eladocagene exuparvovec are not discounted (as they incur in the first cycle), while the benefits are discounted. As noted by Severens and Milne [43], discounting thus discriminates against early investment to achieve long-term health outcomes.

Another aspect of this analysis that should be considered in interpreting results is the inclusion of only direct costs. Given the substantial disease burden of AADCd, the inclusion of indirect costs including caregiving costs and loss of productivity would likely have a material impact on the costs per QALY.

The QALY gains and ICUR results herein compare favorably with results from 15 evaluations (for 25 treatments) conducted by the Institute for Clinical and Economic Review for (1) treatments for ultra-rare diseases and (2) “single and short-term therapies (SSTs)” [44]. Among these analyses, the highest estimates of QALY gains were for atidarsagene autotemcel versus supportive care for metachromatic leukodystrophy [45], consisting of 43.59 undiscounted and 19.26 discounted annually at 3.0% over a lifetime horizon. In the present analysis of eladocagene exuparvovec compared with BSC, base-case estimated QALY gains were 43.87 undiscounted and 20.83 discounted annually at 3.0%. Across the 15 evaluations, the median ICUR was $911,000 per QALY. The ICUR for eladocagene exuparvovec falls well below the median ICUR of other ultra-rare and/or single and short-term therapies. As such, within the context of these innovative treatments, eladocagene exuparvovec provides substantial clinical improvements at a lower cost. Notably, the WAC of the intervention was not known at the time of release of ICER’s final report in five analyses, such that placeholder values were used instead, which were considerably lower than the current WAC [32] values ($4,250,000 WAC versus $2,800,240 placeholder used for atidarsagene autotemcel, $2,800,000 WAC versus $2,100,000 placeholder used for betibeglogene autotemcel, $2,391,705.69 WAC versus $2,000,000 placeholder used for onasemnogene abeparvovec, $2,900,000 WAC [46] versus $2,500,000 placeholder used for valoctocogene roxaparvovec), and accordingly the ICURs would have been higher had the WAC been used.

Similar to the modeling of other ultra-rare diseases, the limited availability of long-term natural history data (including survival data) in AADCd is a challenge for economic modeling. On the basis of expert recommendation [22], cerebral palsy survival data were considered as an analog in this study, and the incremental LYs modeled reflect the differences in mortality risk modeled by motor-milestone defined health states. Moreover, the substantial improvements in overall survival modeled with eladocagene exuparvovec compared with BSC may be conservative, as clinical input on this research indicated that the cerebral palsy data might represent an optimistic prognosis for AADCd; for example, median survival modeled for the no motor function health state was 19 years, which clinical experts suggested may exceed survival expected for patients with AADCd and no motor function at 4 years of age [22]. In addition, the health state utilities informing these analyses reflect an accurate representation of the burden and impact of AADCd on quality of life. The utilities were derived from methods that demonstrated face validity and included obtaining feedback from healthcare providers and caregivers on the impact of AADCd on quality of life [21]. While the inclusion of caregiver disutility in CUA is not without its challenges [47], and the utility gain resulting from their inclusion in the model is substantial, the condition’s infantile onset and the rapid and profound decline patient’s experience result in an immense caregiver burden that should be acknowledged. It forms a core part of the overall disease burden. The importance of including caregiver burden in CUA has been recognized by regulatory and advisory bodies [48, 49] and the use of proxy conditions, similar to the use of Duchene muscular dystrophy caregiver utilities in this analysis, has previously been accepted by HTAs in the analyses in rare diseases [47]. The inclusion of the bereavement disutility in the model reflects the family impact of the significant mortality risk faced by patients with AADCd and has previously been used in HTAs for treatments of pediatric conditions [47, 50].

As is frequently the case for disease modifying treatment in rare disease, the lack of long-term efficacy data for eladocagene exuparvovec are a limitation of the model that must be acknowledged. However, given the substantial differences in trajectory between the eladocagene exuparvovec-treated and natural history cohorts, assumptions of long-term efficacy are considered justified (note that data from patients with up to 10 years of follow-up were incorporated in this model). Taken together, the estimated survival benefit of 30–40 years provided by eladocagene exuparvovec over BSC, combined with the substantial increase in QALYs, suggests that eladocagene exuparvovec is an important treatment option that has the potential to dramatically improve patients’ lives in the long-term. Finally, it should be noted that while incorporation of the WAC in the ICUR is useful to contextualize eladocagene exuparvovec cost-per-QALY in relation to other products evaluated by ICER using the WAC, it is likely higher than actual cost. As such, this analysis could be reviewed when further pricing data are available.

Conclusions

The economic evaluation conducted in this research reflects the meaningful benefits of eladocagene exuparvovec compared with BSC for treatment of AADCd. Modeling of overall survival showed marked improvements in median survival with eladocagene exuparvovec over BSC, with improvements exceeding four decades in the base case. Moreover, substantial QALY gains were achieved with eladocagene exuparvovec at a lower cost relative to many other rare and/or one-time treatments, emphasizing the value of eladocagene exuparvovec as an important treatment option for patients with AADCd.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

This study was funded by PTC Therapeutics Inc. Medical writing/editorial support was provided by Broadstreet HEOR and funded by the study sponsor.

Declarations

Conflicts of interest

RZ, PC, and IT are employees of PTC Therapeutics Inc. and may be shareholders in the company. TOC and YT are employees of Medicus Economics which was contracted by PTC Therapeutics Inc. for the conduct of this study. PW-HL and BM received consulting fees from PTC Therapeutics Inc.

Availability of data and material

The datasets produced during this study are available from the corresponding author upon reasonable request.

Ethics approval

This study used de-identified secondary data, and therefore did not require ethical approval or individual patient consent.

Authors’ contributions

All authors contributed to the study conception and design. Data analysis was performed by YT and TOC. All authors contributed to the writing of the manuscript. All authors read and approved the final manuscript.

Consent to participate

As the data used in this study were anonymized, participant consent to participate was not necessary.

Consent for publication

As the data used in this study were anonymized, participant consent for publication was not necessary.