Modeling the Budget Impact of Disease-Modifying Therapies in Infantile-Onset Spinal Muscular Atrophy

Abstract

Background and Objective

Nusinersen, onasemnogene abeparvovec-xioi, and risdiplam are disease-modifying therapies that have demonstrated clinical benefits for patients with spinal muscular atrophy. However, their high costs pose significant challenges for healthcare payers. This study evaluated the potential budget impact of these treatments compared with best supportive care from a US healthcare payer perspective.

Methods

A budget impact analysis was conducted over a 5-year time horizon using a state-transition model to estimate direct healthcare costs associated with each disease-modifying therapy and best supportive care in a cohort of children with infantile-onset spinal muscular atrophy. A scenario analysis also assessed the budget impact over a lifetime horizon. Model inputs for transitions and costs were derived from clinical trials, observational registries, and administrative databases.

Results

Compared with best supportive care, the added 5-year per-patient budget impact was $2.18 million for onasemnogene abeparvovec-xioi, $1.51 million for nusinersen, and $1.17 million for risdiplam. When modeled over a lifetime horizon, the incremental per-patient budget impact versus best supportive care was $5.86 million for risdiplam, $3.16 million for onasemnogene abeparvovec-xioi, and $3.05 million for nusinersen.

Conclusions

Disease-modifying therapies for the management of infantile-onset spinal muscular atrophy have a substantial budget impact on US healthcare payers, primarily driven by high drug acquisition costs. The impact (per year) becomes less pronounced when evaluated over a longer time horizon, with the largest lifetime budget impact observed for risdiplam.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40261-026-01571-x.

Key Points

New disease-modifying treatments for infantile-onset spinal muscular atrophy improve survival and motor outcomes but also impose very high costs on healthcare payers.

The analysis in this study shows that the gene therapy (onasemnogene abeparvovec-xioi) produces the largest budget impact over a 5-year horizon compared with nusinersen and risdiplam.

Over a lifetime horizon, this pattern reverses: onasemnogene abeparvovec-xioi and nusinersen have the lowest cumulative budget impact, while risdiplam has the highest.

Introduction

Spinal muscular atrophy (SMA) is a rare progressive neuromuscular disorder that substantially reduces motor function and leads to disease-related complications [1]. Spinal muscular atrophy is caused by biallelic deletions or loss-of-function mutations in survival of motor neuron 2 (SMN1), which encodes the survival motor neuron protein and supplies most full-length survival motor neuron in humans. A second gene, survival of motor neuron 2 (SMN2), produces functional survival motor neuron protein but at a substantially lower level and cannot fully compensate for the loss of SMN1 [1]. Spinal muscular atrophy presents with progressive muscle weakness, atrophy, and hypotonia, leading to delayed or lost motor milestones (e.g., inability to crawl, sit or stand), scoliosis, and respiratory and swallowing dysfunction; severe forms are associated with early mortality [1]. Spinal muscular atrophy is subclassified into five types, with infantile-onset SMA (type 1 SMA) being the most common [2] and associated with the highest healthcare resource utilization and costs [3, 4]. Still, substantial healthcare resource utilization and economic burden are associated with all SMA types [3–6].

The disease course in patients with SMA is changing owing to the implementation of newborn screening programs that provide earlier screening and diagnosis of SMA [7], and the availability of three disease-modifying therapies (DMTs) [nusinersen, onasemnogene abeparvovec-xioi (OA), and risdiplam]. Nusinersen and risdiplam are US Food and Drug Administration approved for children and adults, while the gene therapy (OA) is approved for infants aged under 2 years in the USA [8–10]. Although DMTs have demonstrated clinical benefit in SMA, approvals were based primarily on randomized controlled or single-arm trials with small cohorts and limited follow-up [11–16], suggesting substantial uncertainty about their long-term effectiveness and safety. In addition, DMTs are expensive, further increasing the economic burden of SMA [17–19]. Uncertainty regarding long-term effectiveness, coupled with high prices, reduces payers’ willingness to cover these therapies and limits patient access [20, 21], thereby diminishing the health gains achievable through broad timely implementation.

A challenge for payers is projecting and planning the economic impact of covering DMTs. Evidence on the budget impact of routine DMT use in infants with SMA, together with annual cost projections, can assist healthcare payers in their planning. These data inform resource-allocation decisions and support appropriate DMT uptake to improve health outcomes for affected infants. To the best of our knowledge, only one previous budget impact estimation for DMTs in SMA has been conducted in the US healthcare context [22], and it was published before the approval of risdiplam and the publication of more recent clinical follow-up data. Our study aims to use the most recent clinical and economic data to estimate the budget impact of nusinersen, OA, and risdiplam relative to best supportive care from a US third-party payer perspective.

Methods

Perspective and Time Horizon

We performed a budget impact analysis (BIA) to predict the impact of adopting nusinersen, OA, or risdiplam compared with best supportive care (BSC) in managing infants with type 1 SMA from a US (third-party) healthcare payer perspective. The study design was a state-transition (Markov) cohort model that followed a single incident cohort of infants with type 1 SMA. A 5-year time horizon was applied in the base-case analysis, a commonly used time horizon for BIA [23]. In an alternative scenario, we used a lifetime model horizon. For our primary analysis, we assumed a 100% market diffusion rate for each of the DMTs versus BSC.

Model Description

We used a five-state transition model (non-sitting, sitting, walking, permanent ventilation [PV], and death) to simulate the cost of the three DMTs and BSC in infants with type 1 SMA (Fig. 1). The model was structured based on previous early-stage cost-effectiveness analysis models conducted in the USA [22, 24–26] and globally [27, 28]. In the model, all infants with SMA 1 start in a non-sitting health state, which is the typical clinical presentation of SMA 1. During the first year, it was assumed that infants could transition between states every 3 months; after the first year, it was assumed that transitions could occur yearly.

Fig. 1.

State transition model for infantile spinal muscular atrophy

As seen by the model representation (Fig. 1), children could stay in the initial non-sitting health state until death, or progress to a state needing permanent respiratory (ventilation) support followed by death. The pathway from non-sitting to respiratory support to death was the typical natural history of type 1 SMA before the introduction of the DMTs. With DMTs, it is possible to alter this pathway and achieve higher motor milestones, such as sitting and walking. Thus, from the non-sitting health state, it is also possible that the child transitions to a sitting state (which is assumed to be equivalent to type 2 SMA) and to walking (which is assumed to be comparable to type 3 SMA) in a stepwise manner. In addition to reaching higher motor milestones, the DMTs could also affect the disease pathway by delaying the transition to PV and death. Children who achieve a sitting state may stay there until death or transition to a walking state and then to death. Those who achieve a walking state can remain in that state until death. These assumptions are based on the natural history of late-onset SMA [29] (i.e., type 2 SMA and type 3 SMA), where children do not fit the need for PV definition (usually defined as the need for a tracheostomy or ventilation support for ≥ 16 h/day for ≥ 14 [14] or ≥ 21 [15] days). Clinical trials of the DMTs in late-onset SMA populations [30, 31] did not include PV as an outcome measure, further supporting our assumption regarding the model structure and possible transitions.

Description of the Simulated Cohort

The target population characteristics in our model were constructed from the demographic and clinical characteristics in the infantile SMA clinical trials [13, 15, 32]. Our simulated hypothetical type 1 SMA cohort has an age of ~ 5.2 months at treatment initiation, and 57% of participants are girls. The model cohort is assumed to be symptomatic but not completely paralyzed or permanently ventilation dependent at the time of initial treatment Given newborn screening, there may be an increasing share of children treated presymptomatically, but we use a base-case analysis of symptomatic children in line with the populations enrolled in the pivotal clinical trials from which the main transition probabilities are derived; the potential budget impact implications of presymptomatic treatment are explored in an exploratory analysis (Sect. 2.7).

The cohort size used in the model is based on real-world incidence estimates from Cure SMA, an international registry of individuals with SMA. In 2023, the estimated incidence rate of SMA in the USA was 1 in 15,000 [33]. We apply the incidence rate to the US 2023 national live births estimate of 3,596,017 from the National Center for Health Statistics of the Centers for Disease Control and Prevention [34]. We therefore assumed an expected 240 new SMA cases per year in the USA. Of these 240 annual incident cases, we applied an incident rate estimate of 60% for type 1 SMA [2], yielding an expected annual incidence of 144 type 1 SMA cases in the USA.

Model Transition Probabilities

The transition probability from any health state (except non-sitting) to death was assumed to be the same for all alternatives and BSC. We used the US age-based mortality rate [35] to identify the time-dependent transition probability from walking to death. Children in the walking health state are assumed to be equivalent to type 3 SMA, and the predicted life duration for the type 3 SMA population is comparable to that of the general population [36]. For the transition from sitting to death, the sitting health state is assumed to be equivalent to type 2 SMA. To estimate the transition probability from sitting to death, we digitized the published Kaplan–Meier curve for individuals with type 2 SMA from Zerres et al. [29] using WebPlotDigitizer Version 4 [37]. We reconstructed the individual patient data [38, 39] and conducted parametric survival regressions (testing Weibull, exponential, log-normal, log-logistic, and generalized gamma distributions) using SAS, Version 9.4 (SAS Institute Inc., Cary, NC, USA). We selected the best-fitted parametric survival model based on the lowest Akaike Information Criterion/Bayesian Information Criterion (Fig. S1 of the Electronic Supplementary Material [ESM]). We applied the same steps to estimate the transition probability from PV to death; we used the published Kaplan–Meier curve for type 1 SMA individuals receiving non-invasive respiratory support by Gregoretti et al. [40] (Fig. S2 of the ESM).

Head-to-head trials that compare the relative effectiveness of all alternatives, as well as randomized controlled trials that evaluate each treatment relative to a common comparator (e.g., BSC) are not available. To parameterize the model transition probabilities, we relied on a systematic review and indirect comparison by Ribero et al. that estimated the effectiveness of risdiplam versus OA, nusinersen, and BSC in type 1, 2, and 3 SMA [41]. We initially populated the transition probability matrix (from non-sitting to sitting, PV, or death) for risdiplam, as the indirect comparisons in the previously mentioned study were based on risdiplam [41]. We retrieved the relative effectiveness estimates from this indirect comparison [41] and incorporated them into the risdiplam probability matrix to estimate the transition from non-sitting to sitting for nusinersen and OA (see ESM for technical details and calculation). Transition probability from non-sitting to sitting with BSC was zero [15]. We also retrieved the relative effectiveness estimates (i.e., hazard ratio) of overall survival and event-free survival from this indirect comparison [41] and incorporated these within the risdiplam probability matrix to estimate the transition from non-sitting to death or PV for nusinersen and BSC (see ESM for technical details and calculation). The transition probabilities from non-sitting to death or PV for OA were estimated from the STR1VE-US trial [32].

The probability of transitioning from sitting to walking was zero for BSC [15], nusinersen [15], and risdiplam [13, 16]. While for OA, one child out of 22 could walk independently in the STR1VE-US trial [32]. These assumptions were based on the fact that no patients achieved independent walking in the ENDEAR or FIREFISH trials; the sensitivity of this assumption was tested by varying the sitting-to-walking probability from zero to 20% of the OA estimate (Table 1, Fig. 4). All transition probabilities were adjusted to the model cycle length (see ESM for technical details).

Table 1.

Model input parameters related to transition probabilities

Parameter	Valuea	Variationa	Source
From non-sitting to sitting
Risdiplamb	28%	11–40%	[41]
Risdiplam vs OA: odds ratio	0.750	0.149–5.264	[41]
Risdiplam vs nusinersen: odds ratio	4.292	1.142–14.90	[41]
From sitting to walking
Risdiplam	0	0–20%d	[13, 16]
OAc	0.045	0–20%f	[32]
Nusinersen	0	0–20%d	[15]
From non-sitting to PV or death
OS, risdiplamh	2.34/44.42	±20%f	[41]
EFS, risdiplamh	5.12/44.42	±20%f	[41]
OS risdiplam vs nusinersenh: hazard ratio	0.261	0.028–0.665e	[41]
EFS risdiplam vs nusinersenh: hazard ratio	0.197	0.056–0.415e	[41]
OS risdiplam vs BSCh: hazard ratio	0.095	0.012–0.237e	[41]
EFS risdiplam vs BSCh: hazard ratio	0.091	0.024–0.193e	[41]
OAg	0.045	±20%d	[32]
From other health states (sitting, walking, or PV) to death
Sitting to death, all alternatives	Constant = 3.77 Gamma = 2.01 Weibull	Time dependent (survival function)	[29]
Walking to death, all alternatives	US population mortality ratei	–	[35]
PV to death, all alternatives	Constant = 4.19 Gamma = 1 Exponential	Survival function	[40]

BSC best supportive care, EFS event-free survival, OA onasemnogene abeparvovec-xioi, OS overall survival, PV permanent ventilation, TP transition probability

^aRaw values before applying a mathematical transformation to transition probability or adjusting for the cycle length

^bAt a median of 9 months

^cAt around 14 months from non-sitting

^d20% of OA TP to walking

^eOnly in probabilistic sensitivity, while it was ±20% in deterministic sensitivity

^fIn probabilistic sensitivity, the standard error assumed to be 0.05 of mean

^gAt the age of 18 months

^hAt 12 months

ⁱRate per 100,000 standard population at each age identified from the cited source and transformed to TP

Fig. 4.

Univariate sensitivity analysis of budget impact versus best supportive care (BSC) over a 5-year horizon per patient per year for nusinersen (A), onasemnogene abeparvovec (OA) [B], and risdiplam (C). Each row in a panel shows how the budget-impact estimate changes when one input parameter is varied across its plausible range while all other parameters are held at their base-case values. Bars extend left and right from the base-case budget impact (vertical reference line; values shown in italic above each panel). CI confidence interval, EFS event-free survival, OS overall survival, PV transition probability, SE standard error, TP transition probability

Costs

We accounted for three types of direct medical costs: (1) health state costs; (2) drug and administration costs; and (3) monitoring and follow-up costs (Table 2). Health state costs represent the disease-related costs incurred by the SMA population in each health state (i.e., non-sitting, sitting, walking, and PV), excluding the costs of pharmacological interventions. Cost estimates for non-sitting, sitting, and walking health states were retrieved from a study that utilized the HealthCare Integrated Research Database^® to estimate inpatient, outpatient, emergency room, and pharmacy prescription costs in a commercial healthcare plan [4]. These health state cost estimates were also used in previous type 1 SMA models [22, 25]. The PV health-state cost was assumed to equal the non-sitting health-state cost, plus the ventilation cost. The ventilation cost was based on Centers for Medicare & Medicaid Services (CMS) durable medical equipment costs [42] (Table 2).

Table 2.

Model input parameters related to direct medical costs

Parameter	Valuea	Description or CPT/HCPCS	Variation	Source
Supportive care-related costs
Type 1 SMA (non-sitting)	$25,517	Total medical cost PMPM	SEb	[4]
Type 2 SMA (sitting)	$6357	Total medical cost PMPM	SEb	[4]
Type 3 SMA (walking)	$2499	Total medical cost PMPM	SEb	[4]
Home ventilator, invasive/non-invasive	$1302.4	E0465 E0466 Per month	± 20%	[42]
PV	$(25,517+1302.4)	Total medical cost PMPM	±20%	[4, 42]
Nusinersen-related costs (per dose/episode/event)
Nusinersen (per dose)	$138,039.41	First year (6 doses) Subsequent years (3 doses)	± 20%	[46]
IT (administration)	$75.23	96,450	± 20%	[47]
IT (CSF drainage)	$91.54	62,272	± 20%	[47]
Anesthesia	133.13	HCPCS 00635	± 20%	[22]
Specialist for IT	$65.24	99,213	± 20%	[47]
Inpatient anesthesia (15 minutes)	$2520	To be adjusted by hospital mark and assumed one third need inpatient services	± 20%	[58]
Medical/surgical services (inpatient cost/diem)	$5200	To be adjusted by hospital mark and assumed one third need inpatient services	± 20%	[58]
Hospital mark-up	33%	Assumed one third need inpatient services		Assumed
Imaging cost (i.e., U/S, fluoroscopic, or CT)	$29.63, $28.29, $68.24	76,942, 77,003, 77,012	± 20%	[47]
Outpatient hospital observational	$96.53, $157.45, $205.72	99,234–99,236	± 20%	[47]
Platelet count PT PTT Total protein level, urine	$4.48 $4.29 $6.01 $3.67	85,049 85,610 85,730 84,156 (at baseline and prior to each dose)	± 20%	[48]
Onasemnogene abeparvovec-related costs
Onasemnogene abeparvovec (single dose)	$2,220,425.89	Single dose	± 20%	[46]
IV administration	$62.58	96,365	± 20%	[47]
Pre-medication cost (prednisolone) oral solution, 15 mg/5 mL)	$29.32	1 mg/kg 1 day prior to and for a total of 30 days	± 20%	[46]
Anti-AAV diagnostic test	$12.87	86,603 (prior administration)	± 20%	[48]
Observational care after onasemnogene abeparvovec	$96.53, $157.45, $205.72	99,234–99,236	± 20%	[47]
AST ALT Bilirubin, total	$5.18 $5.3 $5.02	84,450 84,460 82,247 (at baseline, weekly in the first month, at the end of CS tapering, and EOW for the second and third months)	± 20%	[48]
Troponin-I	$12.47	84,484 (before infusion, weekly in the first month, monthly in the second and third months)	± 20%	[48]
CBC test (includes differential and platelets)	$7.77	85,025 (before infusion, weekly for the first months, EOW in the second and third months)	± 20%	[48]
Prothrombin time	$4.29	85,610 (at baseline, weekly in the first month, at the end of CS tapering, and EOW for the second and third months)	± 20%	[48]
Creatinine	$5.12	82,565 (before infusion)	± 20%	[48]
Risdiplam-related costs
Risdiplam	$12,260.51	Per package 60 mg/80 mL	± 20%	[46]

AAV adeno associated virus, ALT alanine transaminase, AST aspartate transaminase, CBC complete blood count, CPT current procedural terminology, CS corticosteroid, CSF cerebrospinal fluid, CT computed tomography, EOW every other week, HCPCS healthcare common procedural coding system, IT intrathecal, IV intravenous, PMPM, PT prothrombin time, PTT partial prothrombin time, SE standard error, SMA spinal muscular atrophy, U/S ultrasound

^aValues before adjustment to 2024 US$

^bCalculated from the reported standard deviation in the original source and the sample size

To identify the itemized costs to be included in drug administration, monitoring, and follow-up, we used the billing and reimbursement codes [43–45] and the leaflets [8–10] for nusinersen, onasemnogene abeparvovec, and risdiplam as guidance. For drug costs, we used the Federal Supply Schedule (FSS) pharmaceutical price data of the US Department of Veterans Affairs [46]. Other costs are derived from CMS payment for physician services [47] and laboratory tests [48] (Table 2). See the ESM (Sect. S.1.5) for details about assumed dosing schemes for nusinersen and risdiplam.

All cost values were inflated to 2024 US$ using the US consumer price index for all items [49]. No discounting was applied to the short-term BIA to project nominal cost estimates (5-year horizon). In the life-time modeling scenario, a discount rate of 3% [50] was applied.

Outcome Measures

For each DMT alternative and BSC, we estimated the annual cost per cohort, annual cost per child, cumulative cost per cohort, cumulative cost per child, mean per patient per year (PPPY), and potential budget impact compared with BSC per cohort and child. The annual cost is the total cost (i.e., drug related and disease related) of each alternative in each year. The annual cost was divided by the cohort size (n = 144) to calculate the annual cost per child. The cumulative cost is the sum of the total associated cost for each alternative over 5 years (the lifetime in the scenario analysis) and is divided by the cohort size to estimate the cumulative cost per child. The mean PPPY is the average yearly cost over a time horizon divided by the cohort size. The potential budget impact is the difference between the annual cost/cumulative cost for each alternative and the annual cost/cumulative cost for BSC.

Sensitivity and Scenario Analyses

To test the uncertainty of our BIA model input values, we performed a deterministic sensitivity analysis. We modified the model parameters input by the 95% confidence interval when such information was available from the primary studies, or by a plausible range of 20% for parameters where no such data were available. Additionally, a number of scenario analyses were performed: (i) a lifetime horizon was considered in a scenario analysis to illustrate potential longer-term cost impacts (ii) budget impact estimates with DMT market shares across different market segments based on real-world evidence of market uptake. The Cure SMA 2023 report showed that approximately 60–70% of the SMA population in the USA received one of the Food and Drug Administration-approved therapies [33]. As of late 2023, approximately 25% of children with SMA received nusinersen, 25% received risdiplam, and 20% received OA [33]; (iii) given the increasing use of newborn screening, we also included an exploratory scenario where a fraction of the cohort (40%) was assumed to receive presymptomatic treatment with associated transition probabilities giving better motor outcomes and near-zero PV risk; and (iv) an analysis that assumes a waning effect of OA where varying shares of children in the OA treatment will receive supplementary nusinersen or risdiplam.

Results

Clinical Impact

The modeled clinical outcomes of adopting nusinersen, OA, or risdiplam by the healthcare system for infants with type 1 SMA over a 5-year horizon are presented in Fig. 2. The majority of children in the modeled cohort receiving BSC were on PV or dead after 1 year. With the modeled effectiveness assumptions for the DMTs, the proportion of children who were on PV or dead at year 5 reduced to 82%, 15%, and 22% with nusinersen, OA, and risdiplam, respectively, compared with 100% in BSC. Further, health benefits in the form of motor milestone achievement were observed with all DMTs, with the most pronounced effects observed with OA and risdiplam. At year 5, 81% and 73% of children receiving OA and risdiplam, respectively, were in the sitting health state, compared with 17% for nusinersen and 0% for BSC.

Fig. 2.

Distribution of health states over a 5-year horizon for a simulated cohort of 144 infants with infantile-onset spinal muscular atrophy receiving best supportive care (BSC; A) or one of three disease-modifying therapies: nusinersen (B), onasemnogene abeparvovec (OA; C), or risdiplam (D). Each vertical bar represents one model year; bar segments show the proportion of the cohort occupying each health state at the end of that year, ordered from best to worst outcome (top to bottom in the legend): walking, sitting, non-sitting, permanent ventilation (PV), and death. Health states are mutually exclusive and sum to 100% within each year

Annual Projected Total Costs and Budget Impact

The annual modeled medical costs (i.e., total of disease-related and drug-related costs) for BSC and DMTs over a 5-year horizon per a single cohort of 144 infants with type 1 SMA and per child are presented in Fig. 3. Our analysis indicates that treating a single cohort of infants with type 1 SMA with nusinersen, OA, or risdiplam will require an additional budget, compared with a situation where children only receive BSC, of ~$218 million (M) and $1.513M per child, $314M ($2.183M per child), and $168M ($1.167M per child), respectively, from the US third-party healthcare payers (Fig. 3C, D). The spread of the budget impact over the 5 years differs between the DMTs; with the budget impact for OA mostly occurring in the first year, whereas the budget impact for nusinersen and risdiplam is spread more evenly over the years (Fig. 3C, D).

Fig. 3 .

Predicted 5-year disease-related and treatment-related costs and the resulting budget impact under 100% uptake of best supportive care (BSC) or each disease-modifying therapy. A shows total costs incurred by the annual incident cohort of infants with infantile-onset spinal muscular atrophy in the USA, expressed in millions of US dollars. B shows the same costs expressed per child (US dollars). C and D show the budget impact, defined as the cost difference versus BSC, at the cohort level (C) and per child (D). OA onasemnogene abeparvovec

Disease-Related and Drug-Related Costs

For nusinersen and risdiplam, the cumulative 5-year projected costs per child were evenly distributed between drug-related and disease-related costs. It was $1.319M and $1.298M for nusinersen and $1.124M and $1.147M for risdiplam. In contrast, the drug-related cost of OA contributed to two thirds of its total cost—the drug-related costs were $2.221M, and disease-related costs were $1.066M (Table 3). Further, we estimated PPPY costs over a 5-year horizon by averaging costs across years and members. The highest PPPY was seen for OA ($657,417 per child; additional $436,626 per child compared with BSC), and the lowest PPPY was seen for risdiplam ($454,180 per child; additional $233,389 per child compared with BSC) [Table 3].

Table 3.

Disease-related and drug-related costs

Cost ($)	BSC	Nusinersen			OA			Risdiplam
	Total (disease related only)	Disease related	Drug related	Total	Disease related	Drug related	Total	Disease related	Drug related	Total
Cumulative cost (per cohort)a	158,969,881	186,955,761	189,893,263	376,849,024	153,461,102	319,879,468	473,340,570	165,110,752	161,898,934	327,009,685
Average annual cost (per cohort)b	31,793,976	37,391,152	37,978,653	75,369,805	30,692,220	63,975,894	94,668,114	33,022,150	32,379,787	65,401,937
Cumulative cost (per child)c	1,103,958	1,298,304	1,318,703	2,617,007	1,065,702	2,221,385	3,287,087	1,146,602	1,124,298	2,270,901
Potential budget impact vs BSC (per child)d	Reference			1,513,050			2,183,130			1,166,943
PPPYe	220,792	259,661	263,741	523,401	213,140	444,277	657,417	229,320	224,860	454,180
Potential budget impact vs BSC (PPPY)f	Reference			302,610			436,626			233,389

BSC best supportive care, OA onasemnogene abeparvovec-xioi, PPPY per patient per year

^aSum of cost over 5 years for cohort

^bSum of cost over 5 years/5

^cSum of cost over 5 years/144

^dThe difference between the cumulative cost per child for each drug and BSC

^eAverage annual cost/144

^fThe difference between PPPY for each drug and BSC. For BSC, no disease-modifying therapy is administered, the reported total cost reflects disease-related costs only. For nusinersen, OA, and risdiplam, the total cost is the sum of disease-related and drug-related costs

Sensitivity and Scenario Analyses

Figure 4 displays the univariate sensitivity analysis of the modeled budget impact (mean PPPY vs BSC) for each DMT. The relative ranking of treatments by budget impact (OA > nusinersen > risdiplam) was consistent across all 22 deterministic sensitivity analysis scenarios examined. The analyses reveal that the budget impact results were highly sensitive to the price of nusinersen ($250,413–$354,807), OA ($347,809–$525,443), and risdiplam ($188,417–$278,361). Results were also highly sensitive to the transition probability from the non-sitting to the sitting health state. The costs of PV and non-sitting health states had a modest impact.

Lifetime horizon and market share scenario analyses results are presented in Figs. S3 and S4 of the ESM. The modeled lifetime budget impact PPPY versus BSC was lowest with nusinersen ($3.046M), followed by OA ($3.163M), while highest with risdiplam ($5.856M). The exploratory scenario analysis, assuming that 40% of the cohort are treated presymptomatically, is presented in Table S1 of the ESM, and the results show a 10% reduction in the 5-year budget impact for OA and risdiplam compared with the base-case analysis, whereas the budget impact increases for nusinersen. A scenario analysis exploring the potential budget impact of concomitant DMT use alongside OA is presented in Table S2 of the ESM, showing that co-administration of nusinersen or risdiplam increases the total 5-year per-patient cost by 2–13% relative to OA alone, depending on the proportion of patients receiving a second DMT.

Discussion

This study aimed to estimate the potential budget impact of adopting nusinersen, OA, or risdiplam compared with BSC for the management of infantile-onset SMA from a US third-party payer perspective. Our results show that coverage of each DMT is associated with a considerable budget impact, with cumulative 5-year per-patient costs exceeding those of BSC by $1.17M for risdiplam, $1.51M for nusinersen, and $2.18M for OA, driven primarily by high drug acquisition costs. When interpreting the budget-impact analysis, it should be noted that DMTs are associated with significantly greater health gains compared with BSC. For context, the projected median age at time of death or requiring PV with BSC in our analysis was ~11.2 months, which is consistent with a previous observation of 13.5 months of age (interquartile range 8.1–22) [51], relative to approximately 32.5 years and 28.5 years when receiving OA or risdiplam, respectively.

The primary analysis assumed 100% uptake of each DMT. At the same time, two market-share scenarios based on real-world Cure SMA 2023 uptake data were also assessed, showing that total cohort-level budget impact scales proportionally with market penetration. The per-patient estimates reported in our study can be directly applied by individual payers to project budget impact under any assumed uptake scenario. In another scenario analysis, assuming that 40% of the cohort of children are treated while still presymptomatic, showed a slightly reduced budget impact for OA and risdiplam. With expanding newborn screening programs, a larger share of children will likely be treated earlier. Still, our modeling results do not indicate that this will shift budget impact outcomes in any substantial direction.

Despite the longer lifespan that children gain with the DMTs compared with BSC only, the total disease-related costs do not differ much between treatment alternatives. One explanation for this finding is that the DMTs lead to a sizeable share of children achieving higher motor milestones, which are associated with lower costs than being in a non-sitting or a PV state. Instead, the added budget impact with the DMTs is driven by drug costs. For OA and nusinersen, these costs add to a payer’s budget almost exclusively (OA) and mostly (nusinersen) in the first year, reflecting the one-time nature of OA and the higher loading doses required with nusinersen. The budget impact for risdiplam is more evenly distributed across years.

A previous US study on the budget impact of OA in the SMA population showed that the average annual potential budget impact PPPY of OA at a placeholder price of $2M versus BSC over a 5-year horizon was $946,300 [22]. Our results predicted a lower potential budget impact of $436,606 with OA compared with BSC, and the difference could be explained by differences in treatment effectiveness inputs [32, 41] in our analysis, which assumed the possibility of additional motor milestone gains beyond the time period examined in the clinical trials.

From a lifetime perspective, the results of this study are comparable to a previous estimate, which predicted that adopting OA would yield a cost offset in later years of life [52]. The lifetime mean cost PPPY with OA was $40,547 higher compared with BSC, which was lower than the model results for both risdiplam ($75,082) and nusinersen ($39,049). To achieve the relatively more modest budget impact with OA from a single health insurer’s perspective, a long retention time is required. This long retention time may be feasible for families with a child with SMA, as a 10-year analysis for retention time in the MarketScan commercial claims database showed that families of children with rare diseases kept their same commercial plan longer than other families; 36% (fully insured) and 60.3% (self-insured) retained their commercial plan at 10 years [53].

Several limitations should be considered when interpreting the results of our study. One key source of uncertainty surrounds the transition probabilities, which were estimated from multiple sources, including studies with indirect comparisons and from survival curve extrapolations. To address the impact of uncertainty around transition probabilities, we performed sensitivity analyses by varying key transition probabilities. Another source of uncertainty is driven by varied cost sources: drugs from Veterans Affairs FSS pricing [46], disease-related costs from a study using a large commercial plan [4], and other associated costs, such as physician fees and laboratory charges, from CMS data [47, 48]. The Veterans Affairs FSS drug costs may be conservative, as they apply to federal and government agencies. Still, FSS pricing is negotiated under most-favored commercial terms, and we believe it is a reasonable base case for US third-party payers [46, 54]. The health state costs were retrieved from a previous analysis of commercial health plan enrollees [4], which may represent a private payer. While the administration and monitoring costs were extracted from CMS [42, 47, 48], which reflected Medicare costs, these costs may differ from Medicaid costs, which are expected to be lower, or private payers’ costs, which could be higher. However, the impact of these uncertainties may not be substantial, as risdiplam has no administration costs, while OA administration and monitoring costs contribute only to the first cycle and represent less than 0.05% of OA drug-related costs, and nusinersen administration and monitoring costs represent 1% of its total costs.

Additionally, health-state costs for the PV state were approximated as the non-sitting health-state cost plus a ventilator rental fee, which may underestimate the full care intensity associated with home mechanical ventilation in pediatric patients. This likely results in a modest downward bias in disease-related costs, particularly for BSC, where a greater proportion of patients occupy the PV state. Finally, BIAs focus exclusively on costs to the payer and do not capture the health gains associated with treatment. Recent cost-effectiveness evidence, relating costs to health outcomes, has found OA to be cost effective compared with other DMTs and BSC, especially in the context of accompanying newborn screening that facilitates presymptomatic treatment [55–57].

Conclusions

Disease-modifying therapies for infantile-onset SMA are expected to generate a substantial 5-year budget impact for US third-party payers. High acquisition costs are the primary driver of this impact. In the short term, the one-time gene therapy (OA) produces the largest budget impact compared with nusinersen and risdiplam, reflecting its front-loaded expenditure profile. In contrast, continuous oral dosing of risdiplam distributes costs more evenly over time; however, over a lifetime horizon, model results indicate that its cumulative budget impact surpasses that of both nusinersen and OA.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

Open access funding provided by University of Gothenburg. No financial support was received for the conduct of this study, or for the preparation or publication of this article.

Declarations

Conflicts of Interest

Sumaya Abuloha, Amie Goodin, Tianze Jiao, Manuela Corti, and Mikael Svensson have no conflicts of interest that are directly relevant to the content of this article.

Ethics Approval

Ethics approval was not applicable for this study, as no human participants, identifiable patient data, or biological samples were involved. The analysis is based entirely on published aggregate data from clinical trials, observational registries, and publicly available administrative databases.

Consent to Participate

Not applicable.

Consent for Publication

Not applicable.

Availability of Data and Material

All data sources that were used in the analyses are cited in the article and supplementary material. The model file is available upon request to the corresponding author.

Code Availability

Not applicable.

Authors’ Contributions

Conceptualization: SA, MS; methodology: SA, MS; data acquisition: SA; formal analysis and investigation: SA; interpretation and implications: SA, MS, AG; writing (original draft preparation): SA; critical reviewing and editing: SA, MS, AG, TJ, MC; supervision: MS, AG. All authors read and approved the final version and agree to be accountable for the work.

Previous Publication

This study was submitted previously to the graduate school at the University of Florida as a part of the student/first author (Sumaya Abuloha) dissertation requirements for the PhD degree.