Natalizumab and Tyruko (natalizumab biosimilar) for treating highly active relapsing-remitting multiple sclerosis after at least one disease-modifying therapy: a systematic review and economic model

Manzano, Catalina Lopez, Sadek, Ayman, Cooper, Chris, Tomlinson, Eve, Wang, Hanyu, Rice, Claire, Tallantyre, Emma ORCID: https://orcid.org/0000-0002-3760-6634, Rao-Middleton, Ananya, Whiting, Penny and Thom, Howard 2026. Natalizumab and Tyruko (natalizumab biosimilar) for treating highly active relapsing-remitting multiple sclerosis after at least one disease-modifying therapy: a systematic review and economic model. Health Technology Assessment 30 (60) , pp. 1-290. 10.3310/GJHT1811

PDF - Published Version Available under License Creative Commons Attribution. Download (3MB)

Abstract

Background Multiple sclerosis (MS) is a chronic autoimmune condition that affects the central nervous system, usually starting in early adulthood and often causing long-term disability in young adults. Symptoms can vary but commonly include fatigue, muscle weakness, vision problems and cognitive issues. In the UK, around 130 in every 100,000 people are affected. Most cases (85–90%) begin as relapsing–remitting multiple sclerosis (RRMS), with periods of relapses and remissions, which can later progress to secondary progressive multiple sclerosis (SPMS). A smaller group have primary progressive multiple sclerosis from the start. RRMS can be further categorised based on disease activity. Highly active relapsing–remitting multiple sclerosis (HARRMS), the focus of this appraisal, is broadly defined as MS with unchanged or increased disease activity – clinically or radiologically – despite prior treatment with at least one disease-modifying therapy (DMT). Management typically includes multidisciplinary care and DMTs to reduce relapses and slow progression. Objectives The overall aim was to appraise the clinical and cost-effectiveness of natalizumab (Tysabri) and natalizumab biosimilar (Tyruko) within their marketing authorisations for treating HARRMS after at least one DMT. Methods Clinical effectiveness review We conducted a systematic literature review (SLR) with network meta-analysis (NMA). As we did not expect to find many randomised controlled trials (RCTs) in people with HARRMS, we broadened inclusion to people with RRMS. We included RCTs that compared one of the interventions (natalizumab or natalizumab biosimilar) or comparators of interest [glatiramer acetate, interferon beta (IFNB)-1a, IFNB-1b, peginterferon beta-1a, alemtuzumab, cladribine tablets, fingolimod, ocrelizumab, ofatumumab, ponesimod and autologous haematopoietic stem cell treatment (AHSCT)] to each other or to placebo. We searched MEDical Literature Analysis and Retrieval System (MEDLINE), Excerpta Medica dataBASE (EMBASE) and trial registries from inception to April 2024. We screened existing relevant technology appraisals (TAs), SLRs and submissions from manufacturers of natalizumab and natalizumab biosimilar. Title and abstract screening and assessment of full-text papers were conducted by two reviewers independently. Data extraction and risk-of-bias (ROB) assessment were performed by one reviewer and checked by a second. ROB was assessed with the RoB-2 tool at the outcome level. We extracted and synthesised data on the following outcomes: annualised relapse rate (ARR) disability progression confirmed at 3 and 6 months (CDP3 and CDP6) MRI measurements (proportion of participants with gadolinium enhancing (Gd+) or new or enlarging T2 lesions) adverse effects (AEs) of treatment (any AEs, treatment-related AEs, serious adverse events (SAEs), AEs leading to treatment discontinuation) health-related quality of life assessed using the EuroQol-5 Dimensions (EQ-5D) or Short Form questionnaire-36 items (SF-36) scales. For each outcome, we provided a narrative summary of study details, ROB, and results. Bayesian random- and fixed-effects NMA was performed to compare the efficacy and safety of treatment options using the available trial information. Most treatments were not compared in head-to-head RCTs, and NMA allowed for the use of indirect information to make that comparison. We selected the model (random vs. fixed-effects) that provided the best fit to the data. We presented results as comparisons of each intervention in the network with placebo, mean ranking of each intervention, probability that each intervention would rank first or in specific positions, and a pairwise comparison of each intervention included in the network. Bayesian 95% credible intervals (CrI) were used to represent uncertainty. We used the R (The R Foundation for Statistical Computing, Vienna, Austria) package ‘multinma’ for all analyses. Cost-effectiveness We undertook an independent economic assessment using a discrete event simulation (DES) individual patient model. Previous NICE TAs have been criticised as they did not capture treatment sequencing and that they were unable to accurately reflect the course of the condition. Our DES aimed to overcome these limitations by using by modelling of treatment sequences. To design the model, we reviewed models used in previous NICE TAs. These used very similar Markov multistate models based on Expanded Disability Status Scale (EDSS) severity with transition rates informed by the British Columbia Multiple Sclerosis (BCMS) registry and London Ontario MS databases and treatment effects by individual trials and NMA. Our DES modelled EDSS as an individual attribute, aligning with the structure of the prior models. We also included attributes for age, sex, SPMS status and current treatment. Simulated events were EDSS increase, EDSS decrease, SPMS progression, relapse, SAEs, treatment discontinuation, and death. Patients could switch treatment twice, meaning that up to fourth line therapy was modelled. Patients who progressed SPMS could experience the events EDSS increase, relapse, SAEs and death. Event rates were informed by a combination of new analyses conducted by the UK MS registry and treatment effects of ARR and CDP6 estimated by the NMA. Baseline SAEs and discontinuation came from AFFIRM and ANTELOPE with treatment effects from the NMA. Rates in the SPMS population were informed by the MS registry analyses as no treatment effects were assumed. Our approach to costs and utilities were aligned with previous TAs. The cost of John Cunningham human polyomavirus (JCV) testing was included for both natalizumab and natalizumab biosimilar as clinical advice was that the manufacturer scheme of paying for JCV testing is not widely available. The economic model was implemented in the R programming language using the discrete event simulation for cost-effectiveness modelling (DESCEM) package and the code was validated by an independent analyst at the consultancy Evidera. The model precited EDSS severity over time was validated by comparison to a Markov model prediction. The selected base-case analysis used the HARRMS population from the MS registry for baseline rates and the base-case selection from the NMA results. Treatment class effects was assumed where relative treatment effects not estimated by NMA. Sensitivity analyses were conducted using the All RRMS estimates from the MS registry, switching to alternative NMA sensitivities, excluding the price of JCV testing for natalizumab intravenous (i.v.) and natalizumab subcutaneous (s.c.) (not the biosimilar), reducing the natalizumab s.c. treatment administration costs, and using mortality rates that vary with EDSS. Value of information analysis was used to assess the impact on parameter uncertainty and identify the most influential parameters. The expected value of partial perfect information (EVPPI) was estimated for each of the NMA treatment effects, all costs, all utilities, the MS registry baseline rates, the baseline discontinuation rate, and the baseline SAE rate. Results We included 42 studies (22,409 participants): 40 reported data for a general RRMS population and 2 were conducted in HARRMS. Six studies reported data separately for those with HARRMS. Only four studies evaluated natalizumab or natalizumab biosimilar, the technologies of interest for this appraisal; none provided data on those with HARRMS. AHSCT was only evaluated in people with HARRMS. General relapsing–remitting multiple sclerosis population All studies were considered to be sufficiently similar for inclusion in the NMAs. The fixed-effects model gave the best fit to the data with little evidence of heterogeneity for all outcomes. Annualised relapse rate (39 studies, 20,718 participants; 17 interventions) Follow-up ranged from 4 to 36 (median 24) months. Most interventions were associated with a greater reduction in the risk of relapses compared to placebo [i.e. rate ratio (RR) < 1 and 95% CrI excluding 1.00]. There was no evidence of a difference between natalizumab and natalizumab biosimilar [RR 0.65 (95% CrI 0.34 to 1.26)]. Seventeen (44%) studies were at low ROB, 15 (38%) had some concerns regarding ROB, and 7 (18%) were at high ROB. Sensitivity analysis restricted to studies at low ROB showed similar results. Disease progression (23 studies; 12 interventions) Studies on teriflunomide, ponesimod and ofatumumab did not connect to the network and studies of natalizumab biosimilar and glatiramer acetate s.c. 40 did not report on disease progression, and those on IFNB-1a s.c. 22 only reported data on CDP3. Fifteen studies (10,824 participants; 11 interventions) reported CDP3 and 14 studies (9006 participants; 10 interventions) reported CDP6. Alemtuzumab, ocrelizumab, natalizumab, fingolimod, cladribine and peginterferon beta-1a were associated with a lower risk of both CDP3 and CDP6. Six studies were judged at low ROB, nine at some concerns and five at high ROB. Magnetic resonance imaging outcomes (20 studies; 12 interventions) Follow-up ranged from 4 to 24 (median 24) months. There were no data on MRI outcomes for studies of ofatumumab, glatiramer acetate (s.c. 40), ponesimod, teriflunomide and peginterferon beta 1a. Data were only available for T2 lesions for IFNB-1a (s.c. 22). Nineteen studies (9, 471 participants; 11 interventions) reported data on Gd + lesions and 17 studies (8883 participants; 12 interventions) on T2-weighted lesions. All interventions were associated with a greater reduction in the risk of developing MRI lesions compared to placebo, with the exception of IFNB-1a s.c. 44 for T2-weighted lesions. There was no evidence of a difference between natalizumab and natalizumab biosimilar [hazard ratio (HR) 1.29 (0.69, 2.37)] for Gd + lesions or for T2-weighted lesions [HR 1.07 (0.73, 1.57)]. Adverse events (36 studies) Follow-up ranged from 6 to 24 months (median 18 months) follow-up. Twenty-four studies (14,513 participants; 16 interventions) reported data on any AEs – data were not available for IFNB-1a (s.c. 22). Thirty-one studies (18,149 participants; 15 interventions) reported data on SAEs – data were not available for IFNB-1a (s.c. 22), andnatalizumab biosimilar. There was no evidence of a difference in the risk of developing any AEs or SAEs between any of the interventions and placebo. There was no evidence of a difference between natalizumab and natalizumab biosimilar 1.06 (0.79 to 1.45) in the risk of any AEs; data were not available for SAEs. Only eight studies (n = 3361) reported data on treatment-related AEs. These did not create a connected network and so a NMA was not possible. There was no evidence of a difference in AEs within any of the studies. Twenty nine studies (17,892 participants) reported data on AEs leading to treatment discontinuation. These did not create a completely connected network – teriflunomide, ponesimod and ofatumumab did not connect to the network and data were not available for IFNB-1a (s.c. 22). Fingolimod, glatiramer acetate, IFNB-1a, IFNB-1b and peginterferon beta-1a were associated with an increased risk of treatment discontinuation compared with placebo. There was no evidence of a difference between natalizumab and natalizumab biosimilar [HR 0.48 (0.13, 1.76)]. Twenty studies were judged at low ROB for AEs, 11 at some concerns and 5 at high ROB. Quality of life Only eight studies reported quality of life (QoL) assessed using the EQ-5D or SF-36 tools. Interventions evaluated were cladribine, fingolimod, peginterferon beta and glatiramer acetate versus placebo and alemtuzumab versus IFNB-1a. There was little evidence for a difference in QoL in any of these studies. Highly active relapsing–remitting multiple sclerosis population We had data for six studies that evaluated fingolimod, ocrelizumab, alemtuzumab, cladribine, IFNB, AHSCT and placebo in people with HARRMS. Three studies were at high ROB, one had some concerns and two were low risk. Five studies reported data on ARR. As there were no studies on natalizumab in people with HARRMS, we included one study that compared natalizumab with placebo in a population where participants were required to have had at least one relapse in the previous year and a very high proportion of participants (88%) had previously been treated with a DMT. A connected network for ARR was formed by combining two IFNB-1a comparators. The network included six studies (2162 participants) of seven interventions. All interventions except IFNB-1a, were associated with a reduced ARR compared to placebo, with natalizumab and ocrelizumab ranking highest. As we only had data on a limited number of interventions in HARRMS, to allow direct comparisons between RRMS and highly active populations, we conducted a sensitivity analysis in RRMS where we restricted the network to the eight interventions in the network for ARR in the highly active population. Results were very similar, although 95% CrI were wider in the highly active population. CDP data were limited and disconnected, but all evaluated interventions reduced progression risk. MRI, QoL and AEs outcomes were only evaluated in one or two studies and so there was insufficient information on these outcomes to draw conclusions. Cost-effectiveness The clinical review found no evidence on AHSCT so this was not included in the economic model. The NMA estimates in all RRMS were used for treatment effects on CDP6, ARR, SAEs, and discontinuation due to AEs, as only limited data were found for HARRMS. Base-case results used 1000 patients and 1000 samples while sensitivities used 100 patients and 100 samples; the lower number were found sufficient for stable results by convergence checks. Validation of EDSS severity over time found less severe trend that was explained by the comparator model mixing RRMS and SPMS patients and not using the latest DMT sequences. Compared with natalizumab i.v., natalizumab biosimilar- i.v. and natalizumab s.c., all treatments had greater net benefit at £20,000–30,000/quality-adjusted life-year (QALY), with the only exception being ocrelizumab. The natalizumabs had close to 0% chance of having highest net benefit at £20,000–30,000/QALY. Costs were generally higher on natalizumab than other treatments, though there was no difference in QALYs with 95% CrI overlapping. Natalizumab i.v. has lower mean net benefit at £20,000–30,000/QALY than natalizumab biosimilar i.v., although the 95% CrI overlap. Natalizumab s.c. has very similar mean net benefit to natalizumab i.v. The 95% CrI for costs and QALYs on natalizumab biosimilar i.v. also overlapped with those for natalizumab i.v. suggesting no difference. Natalizumab s.c. has very similar costs and QALYs to natalizumab i.v., again with no evidence of a difference. Conclusions were unchanged under all sensitivities. EVPPI estimates indicated the parameters with greatest impact were the NMA treatment effects on ARR, CDP6, SAEs and discontinuation. However, costs, utilities and MS registry rates also had substantial impact on the results, indicating high parameter uncertainty. Conclusions There is no direct evidence on the effectiveness of natalizumab or its biosimilar in patients with highly active disease. Limited data indicate that both treatments show similar effectiveness in patients with RRMS. Comparisons of DMT effectiveness in people with highly active disease and general RRMS suggest that DMTs are at least as effective in the highly active population, although this is based on sparse data. Assuming natalizumab and its biosimilar follow this trend, they may also be effective in this group. However, trials specifically targeting this population are needed to confirm these assumptions. The economic model used evidence on treatment effects in the general RRMS population and baseline rates in HARRMS. Natalizumab i.v., natalizumab biosimilar i.v. and natalizumab s.c. were not cost-effective compared to any of the included comparators in HARRMS, with similar QALYs but higher costs, with the only exception being ocrelizumab. The greatest decision uncertainty was found in the treatment effects, again supporting the need for trials targeting this population.

Item Type:	Article
Date Type:	Publication
Status:	Published
Schools:	Schools > Medicine
Publisher:	NIHR Journals Library
ISSN:	2046-4924
Date of First Compliant Deposit:	18 August 2026
Last Modified:	18 Aug 2026 15:45
URI:	https://orca.cardiff.ac.uk/id/eprint/189070

Actions (repository staff only)

Edit Item