Stick or twist? Cost-effectiveness of siponimod compared with continuing existing disease-modifying therapies in the treatment of active secondary progressive multiple sclerosis in the UK

Abstract

Objective

Identification of the phenotypic transition from relapsing-remitting multiple sclerosis (RRMS) to secondary progressive multiple sclerosis (SPMS) is often delayed due to disease complexity and an unwillingness to withdraw RRMS disease-modifying therapies (DMTs), driven by limited SPMS treatment options. Despite the paucity of clinical evidence for efficacy in patients with SPMS, DMTs licensed for RRMS are frequently continued into the early stages of SPMS. The cost-effectiveness of oral siponimod, an active SPMS DMT, versus continued oral or infused RRMS DMTs for patients with active SPMS, was evaluated.

Methods

A cohort Markov model based on disease progression through Expanded Disability Status Scale health states, with annual cycles and lifetime horizon, was employed to determine the cost-effectiveness of siponimod from a UK National Health Service (NHS) perspective for patients with active SPMS. Baseline characteristics, health state utility values, hazard ratios for time to 6-month confirmed disability progression, annualized relapse rate ratios and adverse events for siponimod were obtained from the phase 3 EXPAND clinical trial, supplemented by published literature. Published costs, resource use data and comparator efficacy data were obtained from the literature and, in the absence of data, reasonable assumptions were made.

Results

Quality-adjusted life years (QALYs) were greater for siponimod versus all comparators (3.45 versus 2.69–2.83). Incremental cost-effectiveness ratios (ICERs), calculated as cost per QALY, for siponimod versus natalizumab (dominant), ocrelizumab (£4,760), fingolimod (£10,033) and dimethyl fumarate (£15,837) indicated that siponimod was cost-effective at the commonly accepted willingness-to-pay threshold of £30,000/QALY.

Conclusions

Recognition of active SPMS and treatment of this phenotype with siponimod offers a cost-effective and clinically beneficial treatment approach compared with the continuation of oral or infused RRMS DMTs.

Keywords:

• Secondary progressive multiple sclerosis
• siponimod
• cost-effectiveness
• disease-modifying therapy
• National Health Service

JEL Classification Codes:

• C63
• C6
• C
• I19
• I1
• I

Introduction

Multiple sclerosis (MS) is a chronic, neurodegenerative, autoimmune disorder of the central nervous system^1,2. In the United Kingdom (UK), approximately 130,000 people have a diagnosis of MS, with around 7,000 new diagnoses each year³.

Approximately 85% of patients with MS begin with relapsing-remitting multiple sclerosis (RRMS), characterized by relapses, followed by complete or partial recovery and separated by stable periods of remission^1,4. For the majority with RRMS, progressive disability eventually becomes apparent, independent of relapse activity. This marks the transition to secondary progressive multiple sclerosis (SPMS), characterized by cumulatively increasing disability with or without superimposed relapses⁵.

In untreated MS, the transition from RRMS to SPMS occurs in approximately half of patients with RRMS within 15 years; however, it is insidious and difficult to recognize, with a typical delay in recognition of three years retrospectively^1,5,6. Limited disease-modifying therapies (DMTs) are approved for use in SPMS, having been unable to demonstrate efficacy in clinical trials. This, and an unwillingness to withdraw DMTs initiated in the RRMS stage, has led to a reluctance to change the phenotypic label^7,8. Therefore, treatment with RRMS DMTs, which have not demonstrated efficacy in patients with SPMS, is often continued into the early stages of SPMS (Figure 1)^7,8.

Figure 1. A schematic representation of the transition between RRMS and SPMS over time along with the current scope of DMT use and that anticipated for siponimod. Abbreviations. DMT, disease modifying therapy; RMS, relapsing multiple sclerosis; RRMS, relapsing-remitting multiple sclerosis; SPMS, secondary progressive multiple sclerosis.

Siponimod, a sphingosine-1-phosphate receptor modulator with a central and peripheral mechanism of action⁹, is an oral DMT that has been shown to be clinically effective at slowing disability progression in SPMS, measured on the Expanded Disability Status Scale (EDSS) – a scale dominated by physical function, particularly ambulation¹⁰. Data from the EXPAND clinical trial demonstrated that siponimod was also effective at reducing the annualized relapse rate (ARR) and maintaining cognitive function in patients with SPMS, compared with placebo^11,12.

Post hoc subgroup analyses of patients with active SPMS (defined as having had relapses in the two years prior to screening and/or ≥1 gadolinium-enhancing lesion at baseline) highlighted more pronounced clinical efficacy, with respect to disability progression and ARR reduction than in the general SPMS population^13–15. Furthermore, siponimod treatment was associated with maintenance of cognitive functions at Months 12 and 24, as measured by the symbol digit modalities test (SDMT; a measure of attention, concentration, and processing speed), versus placebo in the active subgroup^14,15.

Siponimod has been approved by the National Institute for Health and Care Excellence (NICE) and the Scottish Medicines Consortium (SMC) for the subgroup of adult patients with active SPMS evidenced by relapses or imaging features of inflammatory activity^16,17.

It is likely that the availability of a DMT licensed and reimbursed for SPMS will prompt earlier recognition and diagnosis of the phenotype, leading to earlier initiation of, or switch to, an effective specific therapy⁶. Interferon beta-1b is the only alternative DMT currently reimbursed for SPMS with continuing relapses, and during the NICE appraisal of siponimod, it was confirmed by the National Health Service (NHS) commissioning expert that the use of this treatment is negligible¹⁶.

During the NICE appraisal, comparisons to no treatment and treatment with injectable interferon beta-1b found siponimod to be cost-effective; however, the true alternative scenario for many patients will be continuation of their current RRMS DMT (oral or infused) into the secondary progressive phase^7,8,16,18.

Here, we evaluate the cost-effectiveness of siponimod in this scenario, using DMTs established for the treatment of RRMS in the UK as comparators (natalizumab, ocrelizumab [both infused], and fingolimod, dimethyl fumarate and teriflunomide [all oral])^19–23. These were considered relevant comparators for this assessment based on prescribing patterns at the time of study conception. Interferon beta and glatiramer acetate were not considered relevant to this analysis given the paucity of use at this disease stage within the NHS¹⁶.

Methods

Patient population

Patients considered in the model were those in the active SPMS subgroup of the EXPAND phase 3 clinical trial¹⁴. EDSS scores for these patients were in the range of 3.0–6.5 and active SPMS was evidenced by relapses in the two years prior to the study and/or the use of magnetic resonance imaging (MRI) to identify gadolinium-enhancing lesions at trial entry. The data cut associated with this definition of active SPMS was assessed by the European Medicines Agency (EMA) and informed the marketing authorization of siponimod²⁴. Baseline characteristics of patients with active SPMS provided a cohort comprised of 64% female patients with a mean age of 46.6 years (standard deviation: 8.26) and a median EDSS score of 6.0¹⁴.

Structure of the economic model

The cost-effectiveness of siponimod was assessed using a cohort Markov model based on disease progression through EDSS health states, as has previously been described (Figure 2)²⁵. The model took the perspective of the UK NHS and Personal Social Services with 1-year cycles and a lifetime time horizon.

Figure 2. The structure of the cohort Markov model shows the EDSS health states (0–9) through which patients could transition, or remain, during each cycle. An EDSS score of 10 is considered a death state. At any time, patients could discontinue treatment and move to the off-treatment arm, thus losing any treatment benefit. Upon reaching an EDSS health state of 7, treatment is discontinued. This figure has been redrawn from Schur et al.²⁵. Abbreviation. EDSS, Expanded Disability Status Scale.

The EDSS is a composite score (0–10) combining clinical signs of neurological impairment and assessment of function, predominantly driven by the loss of ambulation for the range 3.0–6.5. A lower score corresponds to reduced disability (or no disability for EDSS 0) and an EDSS score of 10 is indicative of death occurring due to MS; this was treated separately from the other health states.

In alignment with recent NICE appraisals, a half-point EDSS score was rounded down to the nearest whole number to give 10 EDSS states (0–9). During each cycle of the model, patients could remain at the same EDSS state or transition to a higher/lower EDSS state, as well as experience a relapse. As per the NHS England treatment algorithm for MS, it was assumed that patients reaching an EDSS health state of 7 or higher discontinued their DMT, receiving best supportive care (BSC) until death²⁶. These guidelines provide a conservative assumption of treatment regimes in SPMS and have not yet been updated to reflect the introduction of an SPMS DMT. Therefore, this stopping rule was applied as a modeling assumption to provide a direct comparison between siponimod and RRMS DMTs in current clinical practice.

Patients could be moved to the off-treatment cohort due to discontinuation for any reason, as per the EXPAND clinical trial¹¹. At the point of treatment discontinuation, any treatment benefit associated with the intervention was no longer applied and it was assumed that patients would not commence or revert to any other DMT; patients who discontinued were modeled to receive BSC until death.

Death was the absorbing state for the model and patients could transition to this state from any EDSS state. Mortality was incorporated based on gender-averaged mortality rates derived from the Office for National Statistics (2017–2019) and modified based on the EDSS state using a modifier derived from Pokorski et al. in the base case^27,28. Treatment with DMTs was not assumed to have a direct survival benefit.

Clinical inputs

Natural history probabilities for transitioning between EDSS health states 3–6 were obtained from the placebo arm of the EXPAND clinical trial; the values of which are reported in previous cost-effectiveness analyses^11,25. Transition probabilities between EDSS health states 0–2 and 7–10 were extracted from the London Ontario database, to give the complete transition matrix (Supplementary Table 1) which was normalized and calculated, as previously described²⁵. In general, concordance was observed in the overlapping regions of the two data sets.

The hazard ratio (HR) for time to 6-month confirmed disability progression (CDP) for siponimod was applied as a treatment effect on the transition probabilities of disability worsening, but not applied to disability improvement. 6-Month CDP was chosen as the base case for the model since it is less likely to be impacted by incomplete recovery from relapses than 3-month CDP and is the preferred outcome for NICE appraisals in this disease area.

Natalizumab was the only comparator with clinical trial data for patients with SPMS, reporting an odds ratio of 1.060 (standard error: 0.74–1.53) versus placebo for the proportion of patients with 6-month CDP at Week 96 of ASCEND²⁹. The ASCEND trial did not report time to 6-month CDP (as measured by EDSS scores), thus requiring data for the proportion of patients with 6-month CDP at 96 weeks to be used. The results suggested that, while a highly effective DMT for RRMS, natalizumab was not observed to reduce 6-month CDP (measured by EDSS scores), versus placebo, in patients with SPMS²⁹.

In the absence of clinical efficacy data for other comparators in patients with SPMS, it was considered reasonable to assume that no effect on disability progression in SPMS would apply, in line with the ASCEND trial results. Further support for the validity of this assumption can be found in an observational study published by Roos et al.³⁰, reporting similar efficacy of both high- and low-efficacy RRMS DMTs on CDP in patients with active SPMS. Taken with the lack of efficacy reported for natalizumab with respect to 6-month CDP (as measured by EDSS scores), a treatment effect was not applied to the natural history of EDSS disability progression for comparators.

The effect of each intervention on the frequency of relapses was also considered, and for siponimod the ARR ratio (versus placebo) was sourced from the active SPMS subgroup of the EXPAND clinical trial¹⁴. The ARR ratio used for natalizumab was reported in the ASCEND clinical trial for patients with SPMS²⁹. For the remaining interventions without available SPMS-specific data, the ARR ratios used in the model were those reported by Samjoo et al.³¹ in a recent network meta-analysis (NMA) conducted for DMTs in patients with relapsing forms of MS. As a result, it was assumed that RRMS DMTs continue to have a positive effect on relapses in active SPMS, as observed by Roos et al.³⁰, but disability progression is unaffected, in line with results from the ASCEND trial²⁹.

Treatment benefits were applied to the natural history of ARRs by EDSS health state, obtained from the placebo arm of the EXPAND trial (for EDSS 3–7) and a study by Patzold et al. from a UK population of patients with MS (for EDSS 0–2)^29,32. These values are summarized in Supplementary Table 2. Values for EDSS 8 and 9 were assumed to be the same as for EDSS 7, aligning with Patzold et al. and recent NICE appraisals³². The values for 6-month CDP HRs and ARR ratios for each intervention are summarized in Table 1.

Table 1. Clinical efficacy inputs for the DMTs, versus placebo.

DMT	Hazard ratio (CI)	Source	ARR ratio (CI or CrI)	Source
Siponimod	0.630 (0.470–0.860)	EXPAND^14	0.540 (0.390–0.770)	EXPAND^14
Natalizumab	1.060^a (0.740–1.530)	ASCEND^29	0.453 (0.323–0.634)	ASCEND^29
Ocrelizumab	1.000	Assumption	0.330 (0.250–0.440)	NMA^31
Fingolimod	1.000	Assumption	0.460 (0.370–0.550)	NMA^31
Dimethyl fumarate	1.000	Assumption	0.500 (0.400–0.620)	NMA^31
Teriflunomide^b	1.000	Assumption	0.660 (0.560–0.810)	NMA^31

Hazard ratios for DMTs other than siponimod and natalizumab were assumed to be 1.000 given the paucity of clinical evidence for efficacy in delaying disability progression. ARR ratios were calculated from clinical trial data or from an NMA; data obtained from the NMA use CrIs. ^aEffect estimate is an odds ratio and the outcome is the proportion of patients with 6-month CDP at 96 weeks. ^bARR ratio for 14 mg teriflunomide. Abbreviations. ARR, annualized relapse rate; CI, confidence interval; CrI, credible interval; DMT, disease-modifying therapy; NMA, network meta-analysis.

For siponimod, the all-cause discontinuation rate was obtained from the EXPAND clinical trial¹¹. An exponential function was chosen as the most appropriate fit for the data for the long-term extrapolation of discontinuation rate. Discontinuations for natalizumab were obtained by applying the relative risk from a discontinuation indirect treatment comparison (ITC), between the EXPAND and ASCEND trials, to the discontinuation rate of siponimod obtained from the parametric curve for the respective year, as in a previous cost-effectiveness analysis of siponimod²⁵. In the absence of treatment discontinuation data in SPMS patients for comparators other than natalizumab, assumptions were required. Given the lack of established efficacy on disability progression, combined with guidelines to discontinue such DMTs once SPMS is formally recognized, it was assumed that the discontinuation rate would be 30% higher for such comparators relative to siponimod; this assumption of higher discontinuation was found to be conservative in the model. The exponential plot for siponimod and relative risk of discontinuation for comparators, versus siponimod, are given in Supplementary Figure 1 and Supplementary Table 3, respectively.

Adverse events (AEs) for siponimod included those reported in ≥5% of siponimod-treated patients, by preferred term, during the placebo-controlled treatment phase of the EXPAND clinical trial, as with the NICE technology appraisal of siponimod¹⁶. The nature and probabilities of AEs associated with comparator DMTs, accepted by NICE¹⁶, are shown in Supplementary Table 4.

Cost inputs

Treatment costs considered in the model included acquisition (list prices obtained from the British National Formulary [BNF]) and administration, using the most recently published NHS references costs (for the year 2019–2020) and Personal Social Services Research Unit (PSSRU) costs (for the year 2020) (Supplementary Table 5)^33,34. Monitoring resources and costs were sourced from the relevant NICE appraisals of each DMT^16,19–23.

All relapses were modelled identically with average costs and utility decrements applied; the proportion of these requiring hospitalization was taken from the EXPAND clinical trial and the costs were obtained from Tyas et al., inflated to the year 2020 (Supplementary Table 6a)³⁵. AE costs were estimated based on the resources required to manage each AE and were dependent on whether the AE was classified as serious or non-serious (the proportions of which were taken from the EXPAND clinical trial and were 0.858 and 0.142, respectively). Resource use was obtained from recent RRMS NICE appraisals and used NHS reference and PSSRU costs (Supplementary Table 6b). EDSS health state costs were obtained by inflating those reported in the 2005 UK MS Survey to 2019–2020 costs, consistent with the NICE committee preference (Supplementary Table 7)³⁶.

Utility inputs

Utility inputs for EDSS health states 3–7 were calculated from the active SPMS subgroup of the EXPAND clinical trial while those for EDSS health states 0–2 and 8–9 were extracted from data published by Orme et al. (Supplementary Table 8)³⁷. The values derived from the study by Orme et al. were obtained by the application of utility decrements, dependent on EDSS state and due to SPMS, to the coefficient calculated for patients with RRMS and EDSS 0. Data from EXPAND and those derived from Orme et al. showed concordance in the overlapping EDSS states.

Disutility values for relapse events were those calculated from the active SPMS subgroup of the EXPAND clinical trial (Supplementary Table 9). AE utility decrements were considered for each treatment and sourced from the relevant NICE appraisals (Supplementary Table 10). Caregiver utility decrements by EDSS health state were included in the model; data for which were obtained from the NICE appraisal for natalizumab (Supplementary Table 11)¹⁹.

Model outcomes

Results generated from the model are provided in terms of incremental cost-effectiveness ratios (ICERs), calculated as cost per quality-adjusted life year (QALY); ICERs are estimated for siponimod versus each comparator. Costs and benefits are discounted by 3.5% annually, aligned with the NICE reference case³⁸.

Sensitivity analyses

Probabilistic and deterministic sensitivity analyses (PSA and DSA) were undertaken. PSAs were generated by assigning distributions to all model inputs and randomly sampling from these over 1,000 simulations. These analyses allowed calculation of the probabilities of siponimod being cost effective at the willingness-to-pay (WTP) threshold of £30,000 versus each comparator.

One-way DSAs were also carried out for each comparator to determine the impact of individual model inputs on the outcomes. Where available, data for the upper and lower confidence intervals (or credible intervals for ARR ratios taken from the NMA) were used from the relevant source. In the absence of data, an assumed change of ±20% was applied.

Scenario analyses

Several scenarios were explored to investigate the effects of changes to input choices and assumptions in the model. Such scenarios included variations on the sources of inputs used for the natural history of disease progression and relapse, utility values, and relapse disutilities. Additional scenarios investigated the exclusion of treatment discontinuation, relapses and adverse events, as well as varying the model time horizon. Based on the assumptions made for the relative risk of discontinuation for comparators, the change to ICERs when using the upper and lower bounds of these values (based on an assumed change of ±20%) were explored.

A further scenario, whereby patients with SPMS were treated up to an EDSS score of 8 (as opposed to 7 in the NHS England treatment algorithm²⁶), was also evaluated for siponimod versus all comparators, based on the possibility of a change in clinical practice due to the introduction of a DMT licensed and reimbursed for active SPMS.

Results

Base case

The results of the base case scenarios for siponimod versus each comparator are presented in Table 2. The ICERs estimated by the model predicted that siponimod was estimated to be cost-effective at the commonly accepted WTP threshold of £30,000/QALY when compared with natalizumab, ocrelizumab, fingolimod and dimethyl fumarate.

Table 2. Base case outcomes of the model for each DMT.

				Siponimod vs comparator
DMT	Total LYs	Total costs, £	Total QALYs	Incremental LYs	Incremental costs, £	Incremental QALY	ICER, £/QALY	NMB at WTP of £30,000, £
Siponimod	16.39	335,714	3.45	–	–	–	–	–
Natalizumab	16.25	347,805	2.69	0.14	−12,092	0.76	Dominant	35,000
Ocrelizumab	16.26	332,591	2.79	0.12	3,122	0.66	4,760	16,558
Fingolimod	16.26	329,334	2.81	0.12	6,380	0.64	10,033	12,697
Dimethyl fumarate	16.26	325,725	2.82	0.12	9,988	0.63	15,837	8,933
Teriflunomide	16.26	314,786	2.83	0.12	20,928	0.62	33,689	−2,291

Abbreviations. DMT, disease-modifying therapy; ICER, incremental cost-effectiveness ratio; LYs, life years; NMB, net monetary benefit; QALY, quality-adjusted life year; WTP, willingness-to-pay. The key outcomes used to assess the cost-effectiveness of siponimod are represented in bold.

When comparing siponimod with teriflunomide, the model predicted an ICER of £33,689, marginally above the commonly accepted WTP threshold of £30,000/QALY. Despite this, the QALYs predicted for siponimod treatment (3.45) were greater than those for the comparators in all cases, including teriflunomide (2.69–2.83).

Sensitivity analysis

The results from the PSAs for each comparator are shown in Table 3 and were aligned to the deterministic results reported in Table 2. The results show that siponimod has a high probability of being cost-effective at a WTP of £30,000 versus each comparator, with the exception of teriflunomide. Scatter plots containing further results from the PSAs for each comparator are given in Supplementary Figure 2 and cost-effectiveness acceptability curves are provided in Supplementary Figure 3.

Table 3. Average probabilistic results obtained from the PSAs.

						Siponimod vs comparator			Probability of being cost-effective at WTP of £30,000
PSA run	Comparator	Siponimod total costs, £	Siponimod total QALYs	Comparator total costs, £	Comparator total QALYs	Incremental costs, £	Incremental QALY	ICER, £/QALY
1	Natalizumab	330,473	3.33	341,820	2.64	−11,452	0.71	Dominant	0.94
2	Ocrelizumab	331,584	3.41	328,636	2.77	3,100	0.65	4,792	0.84
3	Fingolimod	336,881	3.35	330,983	2.75	6,259	0.61	10,217	0.74
4	Dimethyl fumarate	330,032	3.36	319,829	2.76	10,142	0.62	16,447	0.66
6	Teriflunomide	331,758	3.37	310,600	2.77	21,064	0.60	35,185	0.40

Abbreviations. ICER, incremental cost-effectiveness ratio; PSA, probabilistic sensitivity analysis; QALY, quality-adjusted life year; WTP, willingness-to-pay. The key outcomes used to assess the cost-effectiveness of siponimod are represented in bold.

Tornado plots of the DSAs for each comparator are given in Figure 3 showing the parameters with the largest impact on the net monetary benefit (NMB). In all cases, the greatest uncertainty was found in HRs of disability progression for both siponimod and/or the comparator.

Figure 3. Tornado plots of deterministic sensitivity analyses results versus each RRMS DMT comparator. Abbreviations. CDP, confirmed disability progression; DMF, dimethyl fumarate; EDSS, Expanded Disability Status Scale; NMB, net monetary benefit.

Scenario analysis

The scenario analyses and resulting ICERs are summarized in Table 4. Exploration of the use of literature data in place of those from the EXPAND trial allowed evaluation of the effect on the resulting ICERs for each comparator. The natural history of disability progression (from the London Ontario and British Columbia databases), relapse (from Patzold et al.³²) and utility values (from Orme et al.³⁷) were all explored as scenarios.

Table 4. Scenario analyses performed for siponimod versus each comparator.

Scenario	Scenario ICER, £
Base case result, £	Natalizumab (dominant)	Ocrelizumab (4,760)	Fingolimod (10,033)	Dimethyl fumarate (15,837)	Teriflunomide (33,689)
Societal perspective	−17,690	2,836	8,049	13,836	31,658
Time horizon: 5 years	−5,724	51,537	80,161	108,514	198,046
Time horizon: 10 years	−13,034	22,031	32,977	44,355	79,526
Time horizon: 20 years	−17,009	5,820	11,747	18,252	38,287
Time horizon: 50 years	−15,835	4,760	10,033	15,837	33,689
NH disability progression: London Ontario	−14,063	7,290	13,283	19,861	40,050
NH disability progression: British Columbia	−10,671	13,045	19,471	26,342	47,397
NH relapse: Patzold et al.	−14,952	5,842	10,834	16,558	33,998
Utilities: Orme et al.	−17,809	5,361	11,347	17,930	38,218
Relapse disutility: Orme et al.	−15,838	4,760	10,037	15,844	33,715
Exclude relapses	−15,667	4,907	10,408	16,296	34,506
Exclude treatment discontinuation	1,986	6,495	11,909	17,198	34,215
Upper bound of assumed RR of discontinuation (1.560)	N/A	15,709	20,518	25,664	41,218
Lower bound of assumed RR of discontinuation (1.040)	N/A	Dominant	Dominant	3,204	23,961
Exclude AEs	−15,283	5,530	10,343	16,025	33,232
Exclude caregiver disutility	−17,503	5,272	11,151	17,617	37,540

Abbreviations. AEs, adverse events; ICER, incremental cost-effectiveness ratio; NH, natural history; RR, relative risk.

The use of the London Ontario database, data from which were not specific to patients with active SPMS, for natural history of disease progression through all EDSS health states resulted in increased ICERs versus all comparators. A more pronounced increase was observed when implementing the British Columbia database; however, these data are predominantly from patients with RRMS. The effect of changing the source of relapse data was minimal, and minor differences resulted from the use of literature utility values. These results reflect the concordance between the EXPAND data and those reported in the literature for these parameters.

Interestingly, the exclusion of treatment discontinuation resulted in a significant increase in the ICER (to £1,986 from dominant in the base case) versus natalizumab. This increase was due to siponimod incurring higher total costs than natalizumab in this scenario (siponimod: £454,055; natalizumab: £449,342 [incremental costs: £4,714]) and an incremental QALY of 2.37, resulting from total QALYs of 4.82 and 2.45 for siponimod and natalizumab, respectively. This change in outcome was primarily driven by the change from negative to positive incremental costs, resulting in a positive ICER.

Continuation of siponimod and comparators to EDSS 8 was also explored, with the results from this analysis provided in Table 5. This scenario found siponimod to be slightly more cost-effective versus natalizumab, ocrelizumab, fingolimod and dimethyl fumarate when compared with stopping treatment at EDSS 7 as in the base case analysis.

Table 5. Model outcomes with treatment discontinuation at EDSS health state 8.

				Siponimod vs comparator
DMT	Total LYs	Total costs, £	Total QALYs	Incremental LYs	Incremental costs, £	Incremental QALY	ICER, £/QALY	NMB at WTP of £30,000, £
Siponimod	16.39	339,739	3.47	–	–	–	–	–
Natalizumab	16.25	354,062	2.68	0.15	−14,323	0.80	Dominant	38,289
Ocrelizumab	16.26	336,769	2.79	0.13	2,970	0.68	4,340	17,561
Fingolimod	16.26	333,234	2.81	0.13	6,504	0.66	9,812	13,383
Dimethyl fumarate	16.26	329,351	2.82	0.13	10,387	0.66	15,800	9,335
Teriflunomide	16.26	317,535	2.83	0.13	22,203	0.65	34,300	−2,783

Abbreviations. DMT, disease-modifying therapy; ICER, incremental cost-effectiveness ratio; LYs, life years; NMB, net monetary benefit; QALY, quality-adjusted life year; WTP, willingness-to-pay. The key outcomes used to assess the cost-effectiveness of siponimod are represented in bold.

Discussion

The analysis reported here examines the cost-effectiveness of siponimod in an important real-world decision problem, taking into account current clinical practice in the NHS, as reported by healthcare professionals during the NICE technology appraisal¹⁶. Within the NHS it is common for DMTs, reimbursed for use in patients with RRMS but not recommended for those with SPMS, to be continued during the transition to, and initial stages of SPMS^7,16. This continuation may be a conscious decision so as not to disadvantage the patient; however, may also be a result of difficulties identifying the phenotypic transition to SPMS^7,8.

RRMS DMTs were modeled as comparators with the respective ARR ratios obtained from the ASCEND trial for natalizumab and the NMA published by Samjoo et al. for all other comparators^29,31. Induction therapies cladribine and alemtuzumab, although approved by NICE for use in some forms of RRMS, were not considered as comparators (as with the NICE appraisal of siponimod) as these are given as induction doses only and are not an ongoing treatment. Thus, if a patient begins the transition to SPMS having been treated with these drugs, retreatment with additional courses of the original drug would not be considered.

Ofatumumab, recently approved by NICE and the SMC, and ozanimod, approved by the SMC (but not recommended by NICE) for use in RRMS were not considered relevant comparators for this analysis as they are yet to be established in current clinical practice and were not approved at the time of study conception. Interferon beta and glatiramer acetate were also excluded from the model due to limited use within the NHS¹⁶.

The reported lack of clinical efficacy of natalizumab²⁹, a highly efficacious treatment for RRMS^31,39, in reducing 6-month CDP for patients with SPMS provided the basis of the key assumption that continuation of RRMS DMTs does not slow disability progression (an increase in EDSS score) in patients with SPMS. No clinical trial data are currently available for the other comparators, requiring an assumption to be made about the treatment effect in the modeled population. Since Roos et al.³⁰ found no difference between high-efficacy (including natalizumab, ocrelizumab and fingolimod) and low-efficacy (including interferons, glatiramer acetate and teriflunomide) RRMS DMTs on CDP, it was assumed that the lack of clinical efficacy observed for natalizumab in the ASCEND study was applicable to all comparators, thus no effect on disability progression was applied.

The results of the model estimated that siponimod is likely to be cost-effective for the treatment of patients with active SPMS at the usual NICE WTP thresholds when compared to the continuation of RRMS DMTs, driven by its clinical efficacy in slowing disability progression. The one exception was teriflunomide; however, considering the lack of evidence and usual positioning of the drug, expert clinical opinion deemed it to be an inappropriate and unlikely clinical choice in patients transitioning to or identified as having SPMS, despite the low list price which is likely to have influenced the ICER value.

While the results demonstrate the cost-effectiveness in this context, the true values are uncertain given that DMTs are known to be provided to the NHS with confidential discounts. Despite this uncertainty, the therapeutic benefit of siponimod is evident from the incremental QALY results which drive the predicted cost-effectiveness, despite higher total costs versus all comparators (excluding natalizumab). Treatment with natalizumab (which does not have a confidential discount and is commissioned at list price) was predicted to incur higher costs (total and incremental) with a reduced therapeutic benefit (reduced QALYs) when compared with siponimod, leading to the dominant cost-effectiveness result and large NMB at WTP of £30,000 (Table 2).

Additional clinical benefits reported with siponimod use, such as the improved cognitive measures versus placebo^14,15, were not included in the model. Such inclusion is not trivial and may not be consistently or accurately reported in cost-effectiveness models⁴⁰; however, it is likely that the conversion of these scores to utility values would improve the overall clinical benefit predicted for siponimod (and be reflected in the QALY values). This is, however, uncertain and further assumptions for comparator treatments would be required due to the limited comparative data and inconsistent use of the SDMT measure in clinical trials.

In common with many economic models, one limitation of this evaluation is the use of a clinical trial patient population, based on pre-specified and stringent inclusion criteria. The generalizability of the data and model outcomes to the general population of patients with active SPMS could be queried on these grounds; however, in practice, these data have been considered generalizable by the EMA²⁴.

The paucity of clinical trial data pertaining to the efficacy of RRMS DMTs in SPMS populations was also a limitation, leading to a lack of direct comparative evidence. Only the ASCEND trial of natalizumab evaluated the use of a comparator in patients with SPMS, with no reduction in 6-month CDP observed versus placebo²⁹; results that were further supported by Roos et al. ³⁰. Extrapolation of this lack of efficacy across all comparators was therefore required and assumed to be reasonable given that natalizumab is known to be a highly efficacious DMT in RRMS^31,39 and that there are differences in the mechanism of action compared with siponimod⁹. This assumption was further explored in the DSA, with intervention and comparator HRs for 6-month CDP found to be the most influential parameters on the model outcomes. Despite this, in cases where siponimod was predicted to be cost-effective, large deviations from the HRs used in the model would be required for negative NMBs to be observed, which would be inconsistent with the available evidence.

As a further limitation, no clinical trial data are available for comparator use in the active SPMS subgroup; in particular, the ASCEND trial of natalizumab does not explore efficacy with respect to 6-month CDP, as measured by EDSS scores, in this subgroup of patients. Therefore, a comparison in this sub-population cannot be made, requiring the assumption that the treatment effect (with respect to 6-month CDP) and ARR ratio observed in the ASCEND trial of the overall SPMS population are relevant to patients with active SPMS.

The ARR ratios reported in the RRMS population are used in the model under the assumption that the treatment benefit, with respect to relapse reduction, in the RRMS population is identical in patients transitioning to, or identified as having, active SPMS³¹. Roos et al.³⁰ support this assumption, highlighting that a reduction in relapses is observed in patients with active SPMS when treated with high-efficacy RRMS DMTs compared with low-efficacy RRMS DMTs. Equal effectiveness on relapses in SPMS and RRMS was considered to be a conservative assumption. The requirement for the abovementioned assumptions relating to the efficacy of DMTs in this patient cohort highlights a key data gap pertaining to a prospective comparison of RRMS DMTs in the transition to, and early stages of, SPMS. This may reflect the difficulty in identifying the transition point, which remains a challenge in clinical practice and cannot be addressed by this study.

The relative risk of discontinuation for RRMS DMTs, other than natalizumab, versus siponimod required assumptions to be made as no clinical trial data in an SPMS population was available. The assumption of a relative rate 30% above the discontinuation rate for siponimod was conservative in the model, as was reflected in the ICER values obtained in the scenario analyses using the upper and lower bounds of this value (±20% of the relative risk value of 1.3). Despite an increase in the values of the predicted ICERs when using the upper limit of the relative risk of discontinuation (a relative rate 56% above that of siponimod), siponimod remained cost-effective versus all comparators except for teriflunomide, as in the base case. The use of the lower limit, assuming a similar rate of discontinuation, gave predicted ICERs well below the WTP threshold of £30,000/QALY, however, appeared to be an unfair representation of the real-world scenario. The use of this assumption represents an additional limitation of this model, as well as highlighting the sparsity of clinical data for RRMS DMTs in a secondary progressive patient cohort.

Finally, results from the scenario analysis exploring the continuation of treatment to EDSS 8 showed very little deviation from the base case results (within the limitations outlined, extrapolating clinical efficacy data from patients in EDSS health states 3–7). This suggests that cost-effectiveness considerations may not present a significant barrier to a potential future policy change in which NHS patients are allowed to continue on treatment for longer, if the assumptions underpinning the analysis are valid.

The consideration of continued RRMS DMT use versus the earlier identification of the transition to SPMS, and treatment with siponimod is an important real-world decision problem in which to assess the cost-effectiveness of siponimod. This approach, focusing on continued RRMS treatment, has predicted that siponimod provides a cost-effective and therapeutically beneficial treatment option for patients with active SPMS (demonstrated by incremental QALY benefits of 0.62–0.76), treated in the NHS. Earlier identification of the transition to SPMS and delaying disability progression as a result of siponimod treatment are key influences in the outcome of this model and offer a distinction between this treatment strategy and the continuation of ineffective RRMS DMTs.

Conclusions

The introduction of siponimod, a clinically effective DMT licensed for active SPMS, may result in earlier recognition and acknowledgement of the transition from RRMS to SPMS, thus displacing continued RRMS DMT use. The analysis presented here shows that siponimod provides a cost-effective alternative to current treatments (continued RRMS DMTs) used within the NHS in the UK. Siponimod is estimated to be cost-effective in most cases and was predicted to show improved QALYs versus all comparators.

Transparency

Declaration of funding

This research was funded by Novartis Pharmaceuticals UK Ltd.

Declaration of financial/other interests

SM and FW are employees of Costello Medical, who were contracted by Novartis Pharmaceuticals UK Ltd to undertake the work. UV is an employee of Novartis Ireland Limited. KG is an employee of Novartis Healthcare Private Limited. MD has received honoraria for advisory boards, speaker’s fees, research funding and expenses to attend educational events from Novartis, Biogen, Sanofi Genzyme, Teva, Mylan, Roche and TG Therapeutics. MK was an employee of Novartis Pharmaceuticals UK Ltd when this analysis was conducted and is now an employee of Novartis Pharma AG and a Novartis shareholder.

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Author contributions

Substantial contributions to study conception and design: SM, FW, UV, KG, MK; substantial contributions to analysis and interpretation of the data: SM, FW, UV, KG, MD, MK; drafting the article or revising it critically for important intellectual content: SM, FW, UV, KG, MD, MK; final approval of the version of the article to be published: SM, FW, UV, KG, MD, MK.

Previous presentations

An interim analysis and the corresponding results were presented at virtual ISPOR Europe 2021.

Acknowledgements

The authors would like to acknowledge Nicholas Adlard, Novartis Pharma AG, Basel, Switzerland, for support with model development. Medical writing assistance was provided by Luke Green, Costello Medical, Cambridge, UK.