FormalPara Key Points

Untreated spinal muscular atrophy (SMA) is the leading genetic cause of death in children younger than 2 years of age; however, disease-modifying treatments are changing the clinical course of SMA.

Early recognition and prompt intervention through newborn screening are known to improve clinical outcomes for patients with SMA.

Newborn screening for SMA in Italy may increase life expectancy and health-related quality of life, and could also reduce National Health Service expenditures.

1 Introduction

Spinal muscular atrophy (SMA) is an inherited, autosomal, recessive neuromuscular disorder characterized by progressive muscle weakness and atrophy. This condition affects approximately 1 in 10,000 live births [1]. Although a rare disease, untreated SMA is the leading genetic cause of death in children younger than 2 years of age [2]. SMA is caused by a deficiency in the survival motor neuron (SMN) protein related to either homozygous deletion or mutations in the SMN1 gene. Homozygous SMN1 deletion occurs in approximately 95% of patients with SMA, whereas the remaining 5% of patients have mutations in SMN1 [2]. An almost identical SMN2 gene can produce a small amount of functional SMN protein. SMN2 copy numbers vary between individuals, and the severity of SMA is largely, although not solely, inversely related to the number of copies of SMN2 [3, 4]. With more SMN2 gene copies present, a milder phenotype is expected [3, 4].

Newborn screening (NBS) is critical for identifying a variety of conditions that, if untreated, pose significant risks of long-term disability and/or death. Since 1992, mandatory NBS has been provided free of charge in Italy for phenylketonuria, congenital hypothyroidism, and cystic fibrosis. The NBS program was updated in 2016, when Italian law expanded to include NBS for approximately 40 inherited metabolic diseases. At the same time, the Coordination Center on NBS was established to ensure consistency in the application of early neonatal diagnostic criteria [5]. To date, SMA is not yet universally included as part of the NBS program in Italy.

Recently, a survey was distributed to SMA and NBS key opinion leaders in several countries around the world to investigate current global screening practices [6]. According to the survey, as of 2021, approximately 2% of newborns worldwide were being screened for SMA. Taiwan was the only country where all newborns were screened for SMA [6].

In 2021, the European Alliance for Newborn Screening in Spinal Muscular Atrophy published a white paper outlining the need to include SMA in NBS programs in all European countries by 2025 [7]. In Italy, a 2-year pilot study of NBS for SMA conducted in the Lazio and Tuscany regions [8] was recently completed, and other regions currently have active pilot projects.

In recent years, the field of SMA has changed remarkably. To date, three disease-modifying treatments (DMTs) have received marketing authorization by the European Medicines Agency (EMA): nusinersen in May 2017 [9], onasemnogene abeparvovec in May 2020 [10], and risdiplam in March 2021 [11].

An economic evaluation of NBS for SMA in the era of DMTs is important for guiding decision making regarding its overall value and serves as an aid for its incorporation into diagnostic practice as part of a sustainable health policy for this rare neurogenetic disease. This economic evaluation aimed to assess, from the Italian National Health Service (SSN) perspective, the cost effectiveness of early identification of SMA by NBS and immediate treatment with DMTs compared with diagnosis and treatment without NBS in Italy.

2 Methods

2.1 Modelling Approach

We used Microsoft® Excel 2016 (Microsoft Corporation, Redmond, WA, USA) to develop a decision-analytic model with a structure that can be split into two submodels: a decision tree and a Markov model. In our analysis, the following treatment strategies were considered: (1) symptomatic diagnosis and treatment (SDT), in which no NBS was conducted, symptomatic diagnosis was achieved by clinical referral, and treatment of patients with symptomatic SMA was initiated; and (2) NBS and presymptomatic treatment (NBS+PST), in which NBS enabled early identification and presymptomatic treatment of infants at risk for SMA.

A decision tree described the screening arms of the cost-effectiveness model (Fig. 1). Each path had different outcomes in terms of health-related quality of life and corresponding costs. The decision tree accounted for the different outcomes a patient can receive as a diagnosis when undergoing NBS. The same structure was also used in a recently published analysis [12].

Fig. 1
figure 1

Decision tree used to evaluate newborn screening for SMA versus no newborn screening for SMA (adapted from the literature) [12]. Reprinted from Velikanova et al. [12], International Society for Pharmacoeconomics and Outcomes Research, Inc., published by Elsevier Inc., and licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0). SMA spinal muscular atrophy, SMN1 survival motor neuron 1 gene, SMN2 survival motor neuron 2 gene

Full size image

Briefly, in this model, all newborns underwent a general heel-prick test for NBS. For the NBS+PST strategy, the heel-prick test for newborns also screened for SMA. Presymptomatically detected SMA was assumed to be caused by SMN1 deletion only, which reflects real-world practice whereby 5% of patients with SMA will remain undetected due to deletion of only one copy of SMN1 and mutation of the second SMN1 copy (assumed in 3.8% of patients in the model) [2]. Therefore, SMA caused by SMN1 point mutations was assumed to be undetectable because of test limitations; this condition would not be detected by NBS, and those patients will later develop symptomatic SMA. Initial screening tests were assumed to have 100% sensitivity and 100% specificity, which is in line with what has been reported in studies across a range of countries [13,14,15]. Patients identified through the screening were assumed to be treated presymptomatically, except for patients with two SMN2 gene copies (see Table 1 for the choice of treatment). Presymptomatic treatment would occur immediately after a positive genetic test, whereas symptomatic treatment would occur only after the appearance of symptoms and a subsequent diagnosis by genetic test. The model also accounted for patients identified by NBS but treated symptomatically (only patients with two copies of SMN2).

Table 1 Model parameters and expected values
Full size table

The Markov model focused on motor milestone achievement during the disease course, and transition probabilities between health states were estimated based on observed clinical trial data. Regression to lower health states was assumed in patients with SMA who were not treated. Higher functioning health states were assumed to be associated with increased survival, higher utilities, and lower costs. A visual representation of the model is presented in Fig. 2. Clinical aspects of the health states are described in Online Resource Table 1.

Fig. 2
figure 2

Markov model [12]. Reprinted from Velikanova et al. [12], International Society for Pharmacoeconomics and Outcomes Research, Inc., published by Elsevier Inc., and licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0). BRND within broad range of normal development, PAV permanent assisted ventilation

Full size image

Patients with SMA type 1, two copies of SMN2, three copies of SMN2, or four copies of SMN2 entered the model in the ‘not sitting’ health state, and, at the end of each model cycle, patients could either remain in the same health state or they could transition to a new health state: ‘death’, ‘permanent assisted ventilation’, or ‘sitting’. Patients in the ‘sitting’ health state could stay in that health state or transition to the ‘walking’ health state. Patients with SMA type 2 entered the model in the ‘sitting’ health state, whereas patients with SMA type 3 entered the model in the ‘walking’ health state, and, at the end of each model cycle, patients could either remain in the same health state or transition to a new health state as described above. It was assumed in the model that at the end of the clinical trial follow-up, all patients in the ‘walking’ health state would transition to the ‘within broad range of normal development’ health state. The sources for the clinical data inputs are described in the following section. For patients treated with a DMT, it was assumed in the analysis that the motor milestones acquired at the end of the clinical trial follow-up were sustained until death. Therefore, regression from a higher functioning health state to a worse functioning health state was not considered in this model. To estimate what would happen in terms of costs and effects over the long term, it was necessary to extrapolate survival data obtained from the literature.

To populate the decision tree, epidemiological data were retrieved from published literature. We considered an SMA birth prevalence of 9.4 per 100,000 live births [16], and it was assumed that 3.8% of SMA cases were caused by SMN1 point mutations [17]. For the genotype distribution of NBS-detected SMA, it was assumed that 45% of patients had two copies of SMN2, 33% had three copies, and 22% had four copies [18]. For the phenotype distribution of symptomatically diagnosed SMA, it was assumed that 58% of patients had SMA type 1, 29% had SMA type 2, and 13% had SMA type 3 [19]. To assess the impact of these assumptions, a scenario with different epidemiological data [20] was investigated.

In the base case, a lifetime time horizon was considered, and costs and outcomes were discounted at an annual rate of 3%, according to Italian national guidelines [16]. In the model, a cycle length of 6 months was used for the first 3 years, followed by a cycle length of 12 months for the remainder of the model to adequately capture transitions between health states during clinical trial follow-up. A half-cycle correction was applied in the model.

Health outcomes were reported as life-years (LYs) and quality-adjusted life-years (QALYs). The incremental cost-effectiveness ratio (ICER) was estimated by comparing incremental costs and incremental QALYs of the two model strategies. Model parameters, values, and their sources of valuation are presented in Table 1.

2.2 Data Sources to Inform the Markov Model Transition Probabilities

Clinical evidence used to inform the efficacy of onasemnogene abeparvovec was taken from pooled START, STRIVE-EU, and STRIVE-US [21,22,23,24] clinical trials for patients with symptomatic SMA type 1 and from the SPRINT clinical trial for presymptomatic infants at risk for SMA [25, 26].

Clinical evidence used to inform the efficacy of nusinersen was taken from the SHINE extension study (ENDEAR) [27] for patients with symptomatic SMA type 1, from CS2/CS12 clinical trials [28] for patients with symptomatic SMA types 2 and 3, and from the NURTURE clinical trial [29] for presymptomatic infants at risk for SMA.

Clinical evidence used to inform the efficacy of risdiplam was taken from the pooled FIREFISH clinical trial [30,31,32] for patients with symptomatic SMA type 1 and from the RAINBOWFISH clinical trial [33] for presymptomatic infants at risk for SMA.

Data from the clinical trials described above were incorporated into the model. Transition probabilities between health states were estimated using extracted data on the number of patients able to sit independently and the number of patients able to walk at a specific age. For patients with two copies of SMN2 identified presymptomatically (via NBS) but who became symptomatic by the time they received treatment, it was assumed they would follow a clinical trajectory between patients detected and treated symptomatically and those identified and treated presymptomatically. Therefore, an average of motor milestones gained was estimated. It is important to emphasize that the inputs were unadjusted data and no adjustments for differences in patient characteristics between studies were conducted. Because the aim of this analysis was to compare two strategies (NBS+PST vs. SDT) and not to compare the results of the clinical trials, the differences in patient characteristics were not deemed critical.

In the absence of observed data for the long term, health state-specific survival curves were extrapolated from published literature on the natural history of SMA for the ‘permanent assisted ventilation’ [34], ‘not sitting’ [35], and ‘sitting’ [36] health states. For the ‘walking’ and ‘within broad range of normal development’ health states, survival was assumed to be the same as the Italian general population life expectancy [37]. Parametric survival curves were fitted to the empirical data to extrapolate survival and calculate transition probabilities using published methods [38]. All reconstructions of individual patient data and fitting of parametric curves were conducted using the R software (R Core Team). Parametric curves fitted to the survival data included exponential, log-normal, log-logistic, Weibull, generalized gamma, and Gompertz curves. Survival was projected over the entire lifetime time horizon to ensure all potential benefits and costs were captured. To avoid long curve tails leading to clinically implausible results, curves were terminated based on limits informed by expert opinion of observed life expectancy. The probability of transitioning to death during each model cycle was calculated from the survival function using published methods [39].

2.3 Health-Related Quality of Life

The utilities used in the base-case analyses were derived from the report by the Institute for Clinical and Economic Review [40] and from expert opinion. Health states were defined as follows.

  • ‘Permanent assisted ventilation’: a value of 0.000 was assumed, in line with guidance from the National Institute for Health and Care Excellence [41]. According to expert clinicians, this health state should have a lower value than the ‘not sitting’ health state.

  • ‘Not sitting’: a value of 0.190 was assumed, in line with what has been reported by the Institute for Clinical and Economic Review [40].

  • ‘Sitting’: a value of 0.600 was assumed, in line with what has been reported by the Institute for Clinical and Economic Review [40].

  • ‘Walking’ and ‘within broad range of normal development’: patients were attributed with general population age-dependent utilities. An ordinary least square regression on the Italian general population data [42] was performed to estimate model parameters.

In line with the Institute for Clinical and Economic Review assessment, we assumed further utility benefits for patients treated with a DMT for achieving interim motor milestones such as head control, rolling, standing, and crawling. This was implemented in the model as an additional utility of 0.1 for the ‘not sitting’ health state and another utility of 0.05 for the ‘sitting’ health state for those patients receiving an active treatment [40].

Given the nature of the disease, it is difficult to differentiate the adverse events owing to treatment from the complications associated with SMA itself, which are already accounted for in the health state utility values. Thus, specific disutilities for adverse events or administration procedures (such as intrathecal injection of nusinersen) were not included in the model.

2.4 Treatment in the Italian Context

In Italy at the time of the analysis, onasemnogene abeparvovec was reimbursed for the treatment of (1) patients with a body weight up to 13.5 kg with 5q SMA, a clinical diagnosis of SMA type 1, and onset of SMA symptoms in the first 6 months of life; and (2) patients with a genetic diagnosis of SMA type 1 (biallelic mutation in SMN1 and up to two copies of SMN2) [43]. However, in the base-case analysis, onasemnogene abeparvovec was also considered for presymptomatic patients with three copies of SMN2, in line with its EMA-approved indication [10]. Furthermore, risdiplam was considered for infants before 2 months of age, in line with its EMA-approved indication [11], even if, at the moment of the analysis, it was not reimbursed yet for these patients in Italy. These assumptions were made in alignment with the Italian SMA clinical expert opinion that soon both onasemnogene abeparvovec and risdiplam will likely also be reimbursed for these patients.

Consistent with Italian SMA clinical expert opinion, it was assumed that 40% of patients with two copies of SMN2 identified via NBS would become symptomatic before treatment initiation.

2.5 Costs

The analysis was conducted from the perspective of the Italian SSN and therefore only direct health care costs were included, with 2023 as the reference year for costing.

For estimation of the drug costs, an ex-factory price net of mandatory discounts was considered. The price of onasemnogene abeparvovec considered in the analysis was €1,945,000 per vial [43], the price of nusinersen was €63,173 per vial [44], and the price of risdiplam was €7477 per vial [45].

The posology in the analysis was modeled as described in the summary of product characteristics for each treatment. Onasemnogene abeparvovec is administered as a single-dose intravenous infusion of 1.1 × 1014 vector genomes per kilogram of bodyweight [10]. The recommended dosage of nusinersen is 12 mg per administration by intrathecal injection in the lower back. The model considered four loading doses and a maintenance dose administered once every 4 months thereafter [9]. For risdiplam, which is administered orally, the recommended once-daily dose is determined by age and body weight. Before 2 months of age, the recommended daily dose is 0.15 mg/kg; from 2 months of age up to 2 years, the recommended daily dose is 0.20 mg/kg; from 2 years of age and a body weight <20 kg, the recommended daily dose is 0.25 mg/kg; and from 2 years of age and a body weight ≥20 kg, the recommended daily dose is 5 mg [11]. We used the World Health Organization ‘Child growth standards’ and ‘Growth reference data for 5–19 years’ data set [46] in the model to define patients’ weight by age. Because patients with SMA tend to be underweight, we used 15th percentile data.

Regarding administration costs, it was assumed that all patients receiving onasemnogene abeparvovec required hospitalization, and the costs of health care resources were valued in line with national tariffs [47]. For nusinersen, in line with Italian SMA clinical expert opinion, it was assumed that 10% of patients required inpatient treatment, and the remaining 90% were treated as outpatients. Outpatient procedures were valued in line with the Lombardy region tariff for ‘Macroattività Ambulatoriale Complessa’ (MAC) [48]. Because risdiplam is taken orally, we did not include any administration costs for this treatment in the analysis.

In the model, all patients underwent a diagnostic genetic test, the cost of which was assumed to be €600. With the NBS+PST strategy, a heel-prick test for newborns also screened for SMA, at a cost of €6. These costs were confirmed by Italian SMA clinical experts.

The health state-specific cost for the management of patients was estimated from the results of a recent Italian study [49] that reported the average annual cost for disease management for patients with SMA from the SSN perspective. In particular, the following assumptions were made.

  • The estimated cost for the management of patients with SMA type 2 (€20,677) was used to value the health care resources consumed for patients in the ‘sitting’ health state.

  • The estimated cost for the management of patients with SMA type 3 (€7694) was used to value the health care resources consumed for patients in the ‘walking’ health state.

  • The estimated cost for patients in the ‘within broad range of normal development’ health state was assumed to be the same as the cost for those in the ‘walking’ health state.

We then elaborated on the estimated cost for the management of patients with SMA type 1 to account for patients in both the ‘not sitting’ and ‘permanent assisted ventilation’ health states. According to Italian SMA clinical experts, 11.3% of patients with SMA type 1 required permanent assisted ventilation, and the cost for the management of these patients was 43% greater than the management cost for patients who did not require permanent assisted ventilation. The estimates were validated by Italian SMA clinical expert opinion.

2.6 Sensitivity Analysis

A univariate deterministic sensitivity analysis was performed to identify the parameters with the greatest impact on the results. All inputs were varied by ±20% of the base-case value, while holding all other parameters constant. The results of this analysis were presented as a tornado diagram. To account for the joint uncertainty of the underlying parameter estimates, a probabilistic sensitivity analysis was performed. For all inputs, the standard error was assumed to be 10% of the base-case value. The analysis was conducted using 1000 iterations, where parameter estimates were repeatedly sampled from probability distributions to determine an empirical distribution for costs, LYs, and QALYs. Time horizon, discount rates, and other structural assumptions related to data sources and treatment costs were excluded from the analysis because these were not subject to parameter uncertainty. A willingness-to-pay threshold of €40,000 per QALY was selected based on the Italian standard supported by the Italian Medicines Agency; a threshold of €20,000 per QALY was also applied as an additional exercise to demonstrate the analysis results [50]. A scenario analysis was conducted to assess the impact of a shorter time horizon and different data source for epidemiological data.

3 Results

3.1 Base-Case Analysis

The base-case analysis included a cohort of 400,000 newborns [51]. In the long term, the NBS+PST strategy brought about an additional 390 QALYs and reduced costs of €1,513,375 to the SSN over a lifetime time horizon. Thus, this strategy was considered dominant compared with the SDT strategy, as presented in Table 2.

Table 2 Results of cost-effectiveness analysis (for a cohort of 400,000 newborns)
Full size table

3.2 Sensitivity Analysis

The univariate deterministic sensitivity analysis demonstrated that NBS+PST remained more effective for every parameter variation. Figure 3 presents the 10 most influential parameters that affect the estimated costs. Influential factors impacting estimated costs included the cost of the treatments and the cost for the management of patients.

Fig. 3
figure 3

Tornado diagram of univariate deterministic analysis on total costs. Note: A state, within broad range of normal development; C state, sitting. SMA spinal muscular atrophy, SMN1 survival motor neuron 1 gene

Full size image

Results of the probabilistic sensitivity analysis are presented on an incremental cost-effectiveness plane in Fig. 4 and as a cost-effectiveness acceptability curve in Fig. 5. As presented in Fig. 4, more than half of the simulations (59.0%) were located in the southeast quadrant where NBS+PST is dominant (more effective and less costly), while the rest of the simulations (41.0%) were in the northeast quadrant (more effective and more costly). Using a willingness-to-pay threshold of €20,000 per QALY gained, the probability that NBS+PST is cost effective was 97.8%. At a willingness-to-pay threshold of €40,000, the probability of it being cost-effective was 100% (Fig. 5). The results of the additional scenarios investigated are presented in Table 3.

Fig. 4
figure 4

Incremental cost-effectiveness plane of the probabilistic sensitivity analysis. QALYs quality-adjusted life-years, WTP willingness-to-pay threshold

Full size image
Fig. 5
figure 5

Cost-effectiveness acceptability curve. NBS newborn screening, PST presymptomatic treatment

Full size image
Table 3 Results of scenario analysis
Full size table

4 Discussion

The clinical course of SMA is changing in the era of DMTs. What was once a fatal disease is becoming a treatable condition with improved functional status and outcomes. Early recognition and prompt intervention are now known to improve clinical outcomes, including survival and acquisition of motor milestones [52].

In confirmation of this, a recently published analysis demonstrated that NBS for SMA, coupled with early access to DMTs, is effective in boosting the attainment of motor skills, enhancing functional independence, and reducing comorbidities associated with SMA [53]. These results underline how crucial the implementation of NBS for SMA as a secondary prevention strategy is to reduce health burdens for affected children.

Although there is consensus that the screening is much needed and evidence of the efficacy of treatments for presymptomatic patients exists, there is less agreement on which patients, once identified, should be treated [54]. In early 2018, a group of US expert clinicians and scientists recommended immediate treatment for patients with two or three copies of SMN2 [55], but recently these recommendations were revised to also include patients with four copies [56]. In addition, Blaschek and colleagues [57] reported that almost 70% of patients with four copies of SMN2 develop irreversible symptoms within the first 4 years of life if a wait-and-see strategy is followed. However, a group of screening and pediatric neuromuscular experts in Ontario, Canada, recommended deferral of treatment for infants with four SMN2 copies pending clinical or electrophysiological evidence of disease onset [58].

The advances in genetic treatments for SMA have changed the perspective of affected patients and their families. Clinical data indicate that earlier treatment yields greater efficacy, and if treatment is initiated in the presymptomatic phase of the disease, patients will likely follow normal developmental trajectories with no delays in the achievement of milestones with respect to the general population. The choice between the three available DMTs for early treatment depends on the age of the child, the severity and the form of SMA, and the professional judgment of the clinician and the personal beliefs of the patient and their family.

In this analysis, in line with Italian SMA clinical expert opinion, we assumed that onasemnogene abeparvovec was the preferred treatment for both presymptomatically and symptomatically diagnosed patients with SMA. Moreover, for symptomatically diagnosed patients and presymptomatically diagnosed patients with four copies of SMN2, we assumed that nusinersen and risdiplam were used with equal frequency. This assumption stems from the consideration of these two treatments as viable and comparable options for managing symptomatic SMA. According to Italian SMA clinical expert opinion, this provides a balanced representation of the treatment landscape and of the evolving nature of therapeutic preferences within the medical community. For presymptomatically diagnosed patients with two or three copies of SMN2, we assumed that nusinersen would be utilized more compared with risdiplam. We acknowledge that treatment preferences may evolve over time based on emerging clinical findings and real-world experiences.

In the present study, an economic evaluation was conducted from an Italian payer perspective to estimate the long-term cost effectiveness of NBS to enable early identification and presymptomatic treatment of patients with SMA compared with no NBS with symptomatic diagnosis by clinical referral and symptomatic treatment of patients with SMA. Based on long-term projections of clinical and cost outcomes, NBS+PST was both more effective and less costly. The robustness of the base-case findings was further confirmed by the results of sensitivity analyses.

Our results are consistent with other economic evaluations reporting on health benefits and reduced costs with NBS for SMA in other countries (e.g., The Netherlands, Australia, England, and the United States) [12, 59,60,61]. Our analysis differs from these in location and other aspects. For example, in contrast with The Netherlands analysis, we also considered risdiplam and did not assume that the treatments are equally effective. In addition, we assumed that a percentage of patients with two copies of SMN2 identified presymptomatically (via NBS) would become symptomatic by the time they received treatment.

Several potential limitations to this study should be considered. The primary limitation of this model is in the clinical trial data used. Because SMA is a rare pediatric disease, clinical trials have small sample sizes, short follow-up periods, and no control groups. The model relies on extrapolations of survival and sustained benefits of motor milestones acquired for all treatments considered. In addition, it is a naïve comparison between clinical trials, because we did not make any adjustment for differences in patient characteristics between studies. However, a matching-adjusted indirect comparison study supports our assumption that there would be no impact on the results based on potential differences in patient characteristics between studies [62]. Lastly, the sensitivity analysis model was limited to varying inputs by ±20% of the base-case value rather than observed or reported ranges (e.g., 95% CI). This may be a simplified approach that did not consider variation around reported ranges.

5 Conclusions

The evidence from this study demonstrates the cost effectiveness of NBS followed by presymptomatic SMA treatment when compared with SDT from the perspective of the Italian SSN. Therefore, our findings support the inclusion of screening for SMA in the NBS program in Italy.