Иммуномодуляторы и иммунопрессанты для релапсирующего многократного склероза: сетевой метаанализ.

Summary of findings

Background

Description of the condition

Multiple sclerosis (MS) is a chronic inflammatory and neurodegenerative disease of the brain and spinal cord resulting from a complex interaction between genetic background and environmental factors. Its pathophysiology encompasses several pathological processes, including immune system dysregulation, demyelination, remyelination, microglial activation, and chronic neuroaxonal loss ( Filippi 2018 ). In most patients, the clinical course is initially characterised by attacks of neurological dysfunction, with or without residual deficits. MS has been traditionally viewed as a two‐stage disease with early inflammation responsible for the initial relapsing‐remitting (RR) course and delayed neurodegeneration causing secondary disability progression. However, more recent evidence has pointed out a pathological continuum between the relapsing and progressive phases, with active ongoing inflammation and demyelination which can also be seen in end‐stage MS ( Lucchetti 2018 ). On the other hand, neurodegeneration and neuronal loss leading to brain atrophy have also been observed in the earlier stages of the disease, at least in some cases, as demonstrated by the high prevalence of a certain degree of cognitive impairment even at disease onset ( Haider 2021 ).

MS is amongst the most common causes of neurological disability in young people and currently affects about 2.8 million people worldwide (Atlas of MS [https://www.atlasofms.org/map/global/epidemiology/number‐of‐people‐with‐ms, last accessed on 01/13/23]). Since the 1950s, MS incidence has gradually increased and currently ranges from 2 to 10 per 100,000 persons per year with a latitudinal gradient and a lower incidence closer to the equator ( Koch‐Henriksen 2021 ). Similarly, the female to male ratio has enlarged over time, settling at over 3:1 during the 2000s ( Alonso 2008 ).

RRMS has a typical onset between 20 and 40 years of age, with a heterogeneous clinical presentation depending on the involvement of various regions of the central nervous system (CNS) (e.g. optic nerve, brainstem‐cerebellum, cerebral hemispheres, and spinal cord). The chronic course of MS typically evolves over 30 to 40 years, involving distinct clinical phenotypes. RRMS accounts for ~85% of patients and ~2–3% of patients per year will develop secondary progressive (SP) MS which is characterised by increasing and irreversible disability that occurs independently of the presence of relapses ( Lublin 1996 ). Male sex, older age at onset and high early relapse frequency predict higher risk of unremitting disability worsening ( Scalfari 2014 ). A minority of patients (~10‐15%) have a progressive disease course from onset, which is referred to as primary progressive MS (PPMS). These phenotypes have been subsequently revised based on the presence of inflammatory activity related to the disease which could be potentially targeted by disease‐modifying treatments (DMTs) ( Lublin 2014 ). Such classification includes the "clinically isolated syndrome" (CIS), when the first clinical attack does not completely fulfil MS diagnostic criteria. Furthermore, within each subtype, MS can be classified as active or not active, depending on the occurrence of relapses and/or new, enlarging or contrast‐enhancing, lesions detected with magnetic resonance imaging (MRI).

The diagnosis of MS relies on the demonstration of two subsequent clinical attacks (dissemination in time ‐ DIT) with the involvement of at least two different CNS areas (dissemination in space ‐ DIS) ( Korteweg 2006 ). Although MS diagnosis remains primarily based on clinical criteria, the 2001 McDonald criteria and their 2005 and 2010 revisions have incorporated criteria for DIT and DIS, thus allowing diagnosis at the time of its first symptoms ( McDonald 2001 ; Polman 2005 ; Polman 2011 ). Specifically, DIS can be demonstrated by greater than or equal to one MRI lesion in at least two CNS regions including: periventricular, juxtacortical, infratentorial and spinal cord. DIT is demonstrated by: (i) simultaneous asymptomatic contrast‐enhancing and non‐enhancing MRI lesions at any time; or (ii) a new lesion and/or contrast‐enhancing lesion(s) on follow‐up MRI, irrespective of its timing. Additionally, a more recent revision of the McDonald criteria ( Thompson 2018 ) allows diagnosing MS when DIS criteria are fulfilled and cerebrospinal fluid‐specific oligoclonal bands are detected, without detection in the serum. Other possible diagnoses must be excluded before confirming MS, as "no better explanation" should be demonstrated according to diagnostic criteria.

Pharmacological therapies for MS include MS‐specific DMTs and symptomatic treatments, the latter aimed at relieving symptoms resulting from neurological impairment. As the number of effective DMTs has constantly increased during the last decades, interest in early MS treatment has grown in order to prevent long‐term disability. Additionally, growing evidence suggests that early intervention with high‐efficacy DMTs is associated with a significantly greater reduction of either inflammatory activity and long‐term disease progression compared to escalating from lower efficacy drugs ( Daruwalla 2023 ; He 2020 ). Whilst historically DMTs have been mostly immunosuppressant or immunomodulatory, requiring ongoing administration to keep disease activity suppressed, immune reconstitution therapies that can be given as short courses have recently emerged. This enlarged treatment scenario raises the question whether a DMT should be initiated early, or even in pre‐symptomatic MS (e.g. when only MRI lesions are accidentally detected, in absence of clinical manifestations, a condition described as "radiologically isolated syndrome") and there are demands for an updated network meta‐analysis of randomised clinical trials (RCTs) assessing the efficacy and safety profile of the available DMTs.

Description of the intervention

Several DMTs are available for people with RRMS. Given the broad range of DMTs currently available, many factors related to context and to patients' preferences, expectations and values are usually taken into account when clinicians and patients make shared treatment decisions. For this review update, we considered all immunomodulators and immunosuppressants that, up to September 2021, have been studied in people with RRMS in randomised clinical trials (RCTs) with at least 12 months' follow‐up.

Interferon beta‐1b ( EMEA 2002 ; FDA 1993 ), interferon beta‐1a (Rebif) ( EMEA 1998 ; FDA 2002 ), interferon beta‐1a (Avonex) ( EMEA 1997 ; FDA 2003 ), and glatiramer acetate ( FDA 1996 ) were the first agents approved by national regulatory agencies. Interferon beta‐1b, interferon beta‐1a (Rebif), and glatiramer acetate are administered by subcutaneous injection; interferon beta‐1a (Avonex) by intramuscular injection. The main adverse effects of interferon beta are local injection site reactions and flu‐like symptoms with hyperthermia.

Natalizumab was initially approved by the US Food and Drug Administration (FDA) in November 2004 ( FDA 2004 ), but was withdrawn by the manufacturer in February 2005, after three participants in the drug's clinical trials developed progressive multifocal leukoencephalopathy (PML), a rare and serious viral infection of the brain. Two of the participants died. Following a re‐examination of the participants in the previous clinical trials, the FDA allowed a clinical trial of natalizumab to proceed in February 2006. No additional cases of PML were reported and marketing of the drug for severe RRMS resumed ( EMA 2006 ; FDA 2006 ; Yousry 2006 ). Natalizumab is administered by intravenous infusion, as a dose of 300 mg every four weeks.

Mitoxantrone was approved in 2000 under the indication "for reducing neurological disability and/or the frequency of clinical relapses in people with worsening RRMS, SPMS or PRMS" ( FDA 2000 ). Safety issues of concern for people treated with mitoxantrone are cardiotoxicity and acute leukaemia.

Fingolimod was the first oral treatment approved for people with RRMS to reduce the frequency of relapses and delay the accumulation of physical disability ( EMA 2011 ; FDA 2010 ). Even at the recommended low dose of 0.5 mg once daily, the FDA and European Medicines Agency (EMA) warned about decrease in heart rate following initiation of fingolimod treatment, recommending that all patients be monitored for at least six hours for signs and symptoms of bradycardia, considering that, in some patients, the nadir of heart rate can be observed up to 24 hours after the first dose.

Teriflunomide was the second oral agent approved for people with RRMS ( EMA 2013a ; FDA 2012 ). It is taken orally as a 7 mg or 14 mg tablet once daily. Warnings issued about this drug were hepatotoxicity and risk of teratogenicity.

Two other oral drugs, both with a mainly immunomodulatory mode of action, are available for the treatment of RRMS: teriflunomide is the active metabolite of leflunomide ( Oh 2013 ), inhibiting pyrimidine de novo synthesis, and dimethyl fumarate ( Linker 2011 ), the methyl ester of fumaric acid, is converted after administration into the active metabolite monomethyl fumarate. They were both approved for RRMS in the US in 2012 and in 2013, respectively.

Dimethyl fumarate has been approved as a first‐line oral treatment for people with RRMS ( EMA 2014a ; FDA 2013 ). The recommended dose is 240 mg twice a day. The most commonly reported adverse events leading to discontinuation in clinical trials were flushing and gastrointestinal events.

Alemtuzumab has been approved for treatment of people with RRMS who have had an inadequate response to two or more drugs indicated for the treatment of MS ( EMA 2013b ; FDA 2014a ). The drug is administered by intravenous infusion, as a dose of 12 mg/day for five consecutive days (60 mg total dose) followed by 12 mg/day for three consecutive days (36 mg total dose) administered 12 months after the initial treatment course. Particular warnings and precautions have to be taken into account for the treatment with alemtuzumab, since serious and sometimes fatal autoimmune conditions, life‐threatening infusion reactions, and increased risk of malignancies were observed in people treated with alemtuzumab.

Peg‐interferon beta‐1a, which has been designed to maintain the effects of interferon beta in the body for a longer period of time, was approved by the FDA and EMA for people with RRMS ( EMA 2014b ; FDA 2014b ). It is administered by subcutaneous injection at a dose of 125 µg every 14 days. The most common adverse reactions are injection site erythema, influenza‐like illness, pyrexia, headache, myalgia, chills, injection site pain, asthenia, injection site pruritus, and arthralgia.

Daclizumab is a monoclonal antibody licenced in 2016 for the treatment of RRMS but, due to safety concerns, it was withdrawn worldwide from the market by its manufacturer in 2018 ( EMA 2018a ; EMA 2018 ; FDA 2018 )

Ocrelizumab was approved as a treatment for relapsing MS and PPMS ( EMA 2018b ; FDA 2017 ). It is administered by intravenous injection.

Laquinimod is an oral immunomodulator investigated in two phase 3 trials for the treatment of people with RRMS. It is taken orally as a 0.6 mg tablet once daily. Its use in treating people with RRMS was approved in Russia but not in the EU, since in 2014 EMA refused authorisation. The EMA recommended refusal of the marketing authorisation for laquinimod as a treatment for RRMS due to concerns about potentially increased risks of cancer and teratogenicity in humans, especially given that the drug's mechanism of action is unclear ( EMA 2014c ).

Azathioprine is a purine analogue exerting its immunosuppressive action by affecting DNA replication through inhibition of the synthesis of nucleic acids. It has been used for the treatment of people with MS in many countries on the basis of favourable results reported by placebo‐controlled RCTs ( Laurson‐Doube 2021 ). It is taken orally as a 2 mg/kg or 3 mg/kg tablet daily. It was reported that chronic immunosuppression with azathioprine increases the risk of malignancy in humans ( FDA 2014c ).

Intravenous immunoglobulins may have a role for people with severe and frequent relapses for whom other treatments are contraindicated ( Association of British Neurologists 2005 ). Severe adverse events, including thrombosis of the jugular vein and allergic reaction leading to treatment discontinuation, were noted in 4% of 84 treatment courses with a total of 341 infusions under routine clinical conditions ( Elovaara 2008 ).

The current update includes the following additional interventions compared to the previous version of the review. Compounds with similar mechanism of action to fingolimod have been developed, in order to increase efficacy and improve safety, such as siponimod ( EMA 2020 ; FDA 2019b ), as well as ozanimod and ponesimod, licenced in 2020 and 2021, respectively ( EMA 2021a ; EMA 2021b ; FDA 2020 ; FDA 2021 ). Cladribine is a synthetic chlorinated deoxyadenosine analog approved for the treatment of RRMS in Russia and Australia in 2010, while in the EU and the US it was licenced in 2017 and 2019, respectively, for highly active RRMS and active SPMS ( EMA 2017 ; FDA 2019a ; Leist 2011 ). Recently, diroximel fumarate, a compound similar to dimethyl fumarate, was approved in 2019 in the US and EU for the treatment of RRMS ( EMA 2021c ). The anti‐CD20 monoclonal antibody ofatumumab was approved as a treatment for relapsing MS ( EMA 2021d ).

Given the limited efficacy of currently available DMTs in delaying the progression of RRMS, many clinicians commonly prescribe immunosuppressant drugs with registered indications for conditions other than MS (mainly in rheumatological or autoimmune diseases, or in people undergoing transplant). As such, in addition to those included previously (azathioprine, intravenous immunoglobulins), we decided to include in our review the following interventions used in MS as off‐label treatments: rituximab, methotrexate, cyclophosphamide and long‐term corticosteroids. Rituximab is an anti‐CD20 monoclonal antibody similar to ocrelizumab and ofatumumab, commonly used to treat malignant blood cell neoplasms and several autoimmune diseases, such as rheumatoid arthritis, idiopathic thrombocytopenic purpura, and pemphigus vulgaris. Its efficacy and safety have also been studied in MS and in several countries, since rituximab is frequently prescribed off‐label ( Berntsson 2018 ; Brancati 2021 ; Laurson‐Doube 2021 ). Methotrexate, cyclophosphamide, and long‐term corticosteroids are systemic immunosuppressors. Methotrexate is a common treatment for autoimmune diseases. Cyclophosphamide, a DNA‐alkylating agent used for the treatment of people with autoimmune disorders, has also been administered to people with MS ( Awad 2009 ). Long‐term corticosteroids, given their anti‐inflammatory properties, have been proposed for the treatment of patients with MS since 1961 with mixed results. They have been administered by different schedules as pulsed periodic high‐dose methylprednisolone or oral continuous low‐dose prednisolone ( Ciccone 2008 ). Fludarabine is a cytotoxic agent effectively used as a treatment for patients with lymphoproliferative disorders and haematologic malignancies. It has also been used as an add‐on treatment in patients with MS experiencing breakthrough disease with an increase in the frequency of relapses and active MRI disease during treatment with other DMTs ( Greenberg 2016 ). The antibiotic minocycline has been studied as a potential treatment for MS due to its anti‐inflammatory properties ( Love 2002 ; Metz 2017 ). Mycophenolate mofetil is an immunosuppressor mainly used as an anti‐rejection therapy in organ transplant recipients, that has also been considered as a potential treatment for MS ( Frohman 2004 ).

How the intervention might work

Immunosuppressive or immunomodulatory effects are common to all treatments included in the review. Although they all target the immune system, their effects vary as follows:

1. Immunomodulation (IFNβ‐1b, IFNβ‐1a, glatiramer acetate, pegylated IFNβ‐1a, immunoglobulins, dimethyl fumarate and diroximel fumarate, laquinimod)

The mechanism of action of interferons beta in MS is incompletely understood. Interferons beta are naturally occurring cytokines possessing antiviral activity and a wide range of anti‐inflammatory properties. Recombinant forms of interferons beta are believed to directly increase expression and concentration of anti‐inflammatory agents, while down‐regulating the expression of pro‐inflammatory cytokines ( Kieseier 2011 ). Glatiramer acetate has an immunomodulatory action by inducing tolerance or anergy of myelin‐reactive lymphocytes ( Schmied 2003 ). It is furthermore believed to promote neuroprotective repair processes ( Aharoni 2014 ). Pegylated interferon beta‐1a has a polyethylene glycol group attached to the α‐amino group of the N terminus of interferon beta‐1a (Avonex). Pegylation of interferon beta‐1a may improve its pharmacokinetic and pharmacodynamic properties, allowing for reduced dosing frequency, while maintaining the clinical effectiveness and safety of intramuscular interferon beta‐1a ( Hu 2012 ). The mechanism of action of intravenous immunoglobulins in MS remains unclear, although remyelination of demyelinated axons may occur through the mediation of the effects of cytokines ( Stangel 1999 ). Dimethyl fumarate is a derivative of fumaric acid. It acts primarily by triggering the activation of a nuclear factor (Nrf2) transcriptional pathway, the primary cellular defence against the cytotoxic effects of oxidative stress. It promotes anti‐inflammatory activity and can inhibit expression of pro‐inflammatory cytokines and adhesion molecules ( Wilms 2010 ). Exactly how laquinimod works is unknown, but it is believed to have an immunomodulatory effect on the peripheral and central nervous systems. Data from animal studies indicate that laquinimod has a primary effect on innate immunity. The drug modulates the function of various myeloid antigen‐presenting cell populations, which then down regulate pro‐inflammatory T‐cell responses. Furthermore, data indicate that laquinimod acts directly on resident cells within the central nervous system to reduce demyelination and axonal damage ( Varrin‐Doyer 2014 ). Minocycline is a tetracycline antibiotic agent with immune‐modulating properties.

2. Systemic immunosuppression, inducing a reduction in the activation or efficacy of the immune system through cytostatic or cytotoxic effects (mitoxantrone, methotrexate, cyclophosphamide, long‐term corticosteroids, cladribine, azathioprine, teriflunomide, and leflunomide)

Mitoxantrone is a cytotoxic drug that intercalates with DNA and inhibits both DNA and RNA synthesis, thus reducing the number of lymphocytes ( Fox 2004 ). Methotrexate is commonly used in autoimmune diseases and, since 1996, has been used in the treatment of MS. By acting as an antagonist to folic acid, it interferes with DNA synthesis, repair, and cellular replication, limiting cellular reproduction. Cyclophosphamide, a DNA‐alkylating agent used for the treatment of people with autoimmune disorders, has also been administered to people with MS ( Awad 2009 ). Long‐term corticosteroids have been proposed (such as pulsed periodic high‐dose methylprednisolone or oral continuous low‐dose prednisolone) for the treatment of patients with MS with mixed results ( Ciccone 2008 ). Cladribine is a purine antimetabolite, generally recommended for patients who are unresponsive or intolerant to an alternate drug indicated for the treatment of MS and administered orally divided into two treatment courses separated by one year. It provides a reduction of circulating T (CD4+ and CD8+) and B lymphocytes with relative sparing of other immune cells ( Beutler 1992 ). Azathioprine is a classical cytotoxic immunosuppressive drug that acts as a prodrug for mercaptopurine, inhibiting an enzyme that is required for DNA synthesis. Thus, it most strongly affects proliferating cells, such as the T‐cells and B‐cells of the immune system ( Tiede 2003 ). Fludarabine is a purine nucleoside analog prodrug determining systemic immunosuppression through a cytotoxic action by interfering with DNA synthesis of dividing lymphocytes and monocytes ( Zinzani 1994 ). Mycophenolate mofetil is an inhibitor of purine synthesis with anti‐inflammatory properties exerted on T and B lymphocytes and macrophages ( Barten 2002 ).

Teriflunomide, the active metabolite of leflunomide, is an inhibitor of new pyrimidine synthesis for DNA replication. Consequently, the drug reduces T‐ and B‐lymphocyte activation, proliferation, and function in response to autoantigens. The exact mechanism of action in MS is not fully understood. The drug is thought to reduce the number of activated lymphocytes, which would cause inflammation and damage myelin in the central nervous system ( Claussen 2012 ).

3. Selective immunosuppression, as with monoclonal antibodies or biological agents directed towards specific antigenic targets (natalizumab, fingolimod, siponimod, ozanimod, ponesimod, alemtuzumab, ofatumumab, daclizumab, rituximab and ocrelizumab)

Natalizumab is a monoclonal antibody against the alfa4 integrin on the surface of lymphocytes. This integrin is essential in the process by which lymphocytes gain access to the brain by allowing the cells to penetrate the blood brain barrier. Natalizumab blocks the action of the alfa4 integrin so that lymphocytes are unable to enter the brain and attack myelin protein ( Yednock 1992 ). Fingolimod acts as a functional antagonist of sphingosine‐1‐phosphate (S1P) receptor on lymphocytes, resulting in a reduced egress of lymphocytes from the lymph nodes. In particular, auto‐aggressive T‐cells are prevented from recirculating to the central nervous system ( Mandala 2002 ). Siponimod is a sphingosine‐1‐phosphate (S1P) receptor modulator acting as a functional antagonist on S1P1 receptors on lymphocytes. By preventing egress of T cells from lymph nodes, it reduces their recirculation into the central nervous system, limiting inflammation ( Behrangi 2019 ). The mechanisms of action of ozanimod and ponesimod is similar to that of siponimod, with higher receptor selectivity for the latter and short half life, facilitating faster reversibility of its effects on the immune system after discontinuation ( Scott 2016 ; Ruggieri 2022 ). Alemtuzumab is a monoclonal antibody against the CD52 antigen expressed on lymphocytes and monocytes. Its effects in MS are thought to be mediated by an extended lymphocyte depletion and change in the composition of lymphocytes that accompanies lymphocyte reconstitution ( Hill‐Cawthorne 2012 ). Ofatumumab is a CD20+ B‐cell‐targeting recombinant human monoclonal antibody. It was initially approved for the treatment of chronic lymphocytic leukaemia and is now approved in several countries as a subcutaneous injection for the treatment of RRMS ( Kang 2021 ). Daclizumab is a monoclonal antibody against the CD25 antigen (interleukin 2 receptor) expressed on immune cells. The mechanisms by which the drug exerts effects in MS are not clear. Daclizumab leads to expansion of regulatory CD56 natural killer T lymphocytes, which may be an important mechanism of action in MS. Furthermore, daclizumab modulates the function of dendritic cells, resulting in decreased T‐cell activation ( Wuest 2011 ). Rituximab is a chimeric monoclonal B‐cell‐depleting anti‐CD20 antibody similar to ocrelizumab and ofatumumab, administered intravenously, commonly used in the treatment of malignant blood cell neoplasms and several autoimmune diseases, such as rheumatoid arthritis, idiopathic thrombocytopenic purpura, and pemphigus vulgaris. It has been used off‐label in the treatment of MS for more than two decades ( Salzer 2016 ; Laurson‐Doube 2021 ).

Ocrelizumab is a monoclonal antibody against the CD20 antigen expressed on B‐lymphocytes. The antibody depletes circulating B‐lymphocytes predominately through antibody‐mediated cytotoxicity ( Oh 2013 ).

Mechanisms of action must be considered while assessing the risk of adverse events associated with the use of a drug, since safety is usually a consequence of the drug's main pharmacological effect ( Compston 2008 ; Hauser 2020 ; Massacesi 2002 ; Meinl 2008 ).

Why it is important to do this review

Although there is consensus that immunotherapies reduce the frequency of relapses in MS, their relative benefit in delaying new relapses or disability worsening remains unclear. This uncertainty is due to the limited number of direct comparison trials, which provide the most rigorous and valid research evidence on the relative efficacy and safety of different, competing treatments. Since the previous version of the review ( Tramacere 2015 ), new DMTs have been approved by regulatory agencies, offering a broader spectrum of treatment options for people with MS. Evidence of efficacy in chronic autoimmune conditions, relatively good tolerability and reasonable cost has prompted the off‐label use of several immunosuppressants and immunomodulators for the treatment of MS in many countries, particularly in settings with budget constraints ( Zeineddine 2020 ). A summary of the results, including both direct and indirect comparisons, may help to clarify the stated uncertainty ( Caldwell 2005 ; Glenny 2005 ).

The data underlying the present review and NMA served as the evidence base for the development of a separate clinical practice guideline on the treatment of RRMS and PMS by an international, highly representative multi‐stakeholder panel (Multiple Sclerosis Essential Medicines Panel, MEMP) appointed by the Multiple Sclerosis International Federation (MSIF). The panel included people with MS and advocacy group representatives, clinicians from different speciality areas involved in the management of MS, pharmaco‐epidemiologists and health economists. The guidelines were developed with methodological guidance by the Department of Health Research Methods, Evidence and Impact (HEI), McMaster University, Hamilton, Ontario, Canada, according to the GRADE Working Group method for guideline development ( Alonso‐Coello 2016 ; Alonso‐Coello 2016b ). The MEMP recommendations were used as the evidence base for an application for the inclusion of disease‐modifying treatments in the 23rd WHO Model List of Essential Medicines. The nine critical outcomes identified by MEMP were differentiated into primary and secondary outcomes in this review (see Methods ).

This is an update of a Cochrane review published in 2015 ( Tramacere 2015 ).

Objectives

To compare, through network meta‐analysis, the efficacy and safety of interferon beta‐1b, interferon beta‐1a (Avonex, Rebif), glatiramer acetate, natalizumab, mitoxantrone, fingolimod, teriflunomide, dimethyl fumarate, alemtuzumab, pegylated interferon beta‐1a, daclizumab, laquinimod, azathioprine, immunoglobulins, cladribine, cyclophosphamide, diroximel fumarate, fludarabine, interferon beta 1‐a and beta 1‐b, leflunomide, methotrexate, minocycline, mycophenolate mofetil, ofatumumab, ozanimod, ponesimod, rituximab, siponimod and steroids for the treatment of people with RRMS.

Methods

Criteria for considering studies for this review

Types of studies

We included individually randomised parallel controlled clinical trials (RCTs). We considered studies published in abstracts whenever sufficient information was available on study design, characteristics of participants, interventions, and outcomes. Only studies with a follow‐up of 12 months or longer were included.

Types of participants

We included adult participants aged 18 years or older with a diagnosis of RRMS according to Poser ( Poser 1983 ) or McDonald ( McDonald 2001 ; Polman 2005 ; Polman 2011 ) diagnostic criteria. We included all participants regardless of sex, degree of disability, and disease duration.

We included studies primarily focused on RRMS but also including a subgroup of people with PMS only if the proportion of people with RRMS was ≥ 80%. Evidence from studies including 80% to 99% of people with RRMS were considered for downgrading because of indirectness while assessing the certainty of the evidence, according to GRADE methodology ( Guyatt 2011 ).

Types of interventions

We included all immunomodulators or immunosuppressants (even if they were not licenced in any country). We excluded: (i) combination treatments; (ii) trials in which a drug regimen was compared with a different regimen of the same drug without another active agent or placebo as a control arm; (iii) all non‐pharmacological treatments; and (iv) interventions with over‐the‐counter drugs.

We included RCTs that evaluated one or more of the following pharmacological interventions as monotherapy, compared to placebo or to another active agent:

interferon beta‐1b

interferon beta‐1a (Avonex, Rebif)

glatiramer acetate

natalizumab

mitoxantrone

fingolimod

teriflunomide

dimethyl fumarate

alemtuzumab

pegylated interferon beta‐1a

daclizumab

ocrelizumab

laquinimod

azathioprine

immunoglobulins

In this update, we also considered the following additional interventions for inclusion:

cladribine

cyclophosphamide

diroximel fumarate

fludarabine

interferon beta 1‐a and beta 1‐b (when both are used in the same experimental arm as ‘interferon beta products’).

leflunomide

methotrexate

minocycline

mycophenolate mofetil

ofatumumab

ozanimod

ponesimod

rituximab

siponimod

steroids

We included regimens as defined in primary studies, irrespective of their dose.

Types of outcome measures

Primary outcomes

We estimated the relative effects of the competing interventions according to the following primary outcomes:

Efficacy

Relapses: proportion of participants who experienced new relapses over 12, 24, or 36 months after randomisation or at the end of the study. A relapse is defined as newly developed or recently worsened symptoms of neurologic dysfunction that last for at least 24 hours, occurring in the absence of fever or other acute diseases and separated in time from any previous episode by more than 30 days ( McDonald 2001 ; Polman 2005 ). A more stringent 48‐hour criterion has been used in some RCTs. A relapse can resolve either partially or completely.

Disability worsening: proportion of participants who experienced disability worsening over 24 or 36 months after randomisation or at the end of the study. Worsening is defined as at least a 1‐point Expanded Disability Status Scale (EDSS) increase or a 0.5‐point increase if the baseline EDSS was greater than or equal to 5.5, confirmed during two subsequent neurological examinations separated by at least a six‐month interval free of attacks ( Kurtzke 1983 ). Disability worsening confirmed after only three months of follow‐up is considered a surrogate marker for unremitting disability. EDSS is a common measure of MS disability (where 0 is normal, 3 mild disability, 6 care requirement, 7 wheelchair use, and 10 is death from MS) and is used to measure disability worsening in clinical trials for MS.

Safety

Discontinuation due to adverse events: measured by the number of participants who withdrew due to any adverse event at the end of the study

Serious adverse events (SAEs): number of participants with any (one or more) SAEs, defined according to the authors of the study

Secondary outcomes

Cognitive decline: assessed as a continuous outcome considering the variation in the score of the Symbol Digit Modalities Test (SDMT) ( Benedict 2017 ) when available or, alternatively, the Paced Auditory Serial Addition Test (PASAT) ( Gronwall 1977 );

Quality‐of‐life impairment: assessed as a continuous outcome considering the variation in the score of scales reporting quality‐of‐life impairment. Any available scale was considered;

New or enlarging T2‐weighted magnetic resonance imaging (MRI) lesions: measured by the number of participants with new or enlarging T2‐weighted MRI lesions at 12, 24, and 36 months after randomisation;

New gadolinium‐enhancing positive T1‐weighted MRI lesions: measured by the number of participants with new gadolinium‐enhancing T1‐weighted MRI lesions at 12, 24, and 36 months after randomisation;

Mortality: overall number of MS‐related deaths.

Search methods for identification of studies

Electronic searches

The previous version of this review searched the Cochrane Multiple Sclerosis and Rare Diseases of the CNS Group Trials Register (on 30 September 2014). As this was not possible for this version of the review, a bespoke search was created by the Information Specialist of the Cochrane Multiple Sclerosis and Rare Diseases of the CNS Group. Searches were run in September 2021, then topped up on 8 August 2022:

Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library (Issue 8 2022)( Appendix 1 );

MEDLINE (PubMed) (January 2012 to 8 August 2022) ( Appendix 2 );

Embase (Embase.com) (January 2012 to 2022 week 32) ( Appendix 3 ).

We did not apply any search limitation with respect to study outcomes, methods of analysis, or language.

Searching other resources

In addition to considering all studies from Tramacere 2015 , to identify eligible studies prior to 2012, we consulted the identified studies in Filippini 2013 , a prior Cochrane NMA review concerning immunomodulators and immunosuppressants for multiple sclerosis, where the search was performed until February 2012.

We searched for ongoing studies on the following databases (see Appendix 4 for search strings): • World Health Organisation (WHO) International Clinical Trials Registry Platform (ICTRP) (apps.who.int/trialsearch). Search terms: relapsing multiple sclerosis, filtered for "Phase 2" "Phase 3" trials; • US National Institutes of Health clinical trial register (www.clinicaltrials.gov). Search term: "relapsing multiple sclerosis".

We checked reference lists of all included studies and any relevant systematic reviews identified for additional references to studies. We examined any relevant retraction statements and errata for included studies.

Data collection and analysis

Selection of studies

For this version of the review, study selection was conducted with the Rayyan platform (https://rayyan.ai/) in accordance with the methods described in the Cochrane Handbook for Systematic Reviews of Interventions ( Higgins 2021 ). Six review authors in pairs (BR, EB, FN, GP, IT, MF) independently screened the titles and abstracts. Potentially relevant articles were acquired in full text and assessed for eligibility by the same six authors in pairs.

Data extraction and management

Two review authors (SM, MGL) independently extracted data from the studies using a predefined data extraction form in an Excel spreadsheet and piloting the data extraction form on at least five studies in the review. We resolved disagreements by discussion with a third author (FN). Whenever data were available from both peer‐reviewed journals as full publications as well as from trials registries (such as ClinicalTrials.gov or WHO ICTRP), we extracted them from the former. We extracted results from trials registers when these were the only available data.

We extracted from each included study data the following information:

Study: first author or acronym, number of centres, year of publication, years that the study was conducted (recruitment and follow‐up), publication (full‐text publication, abstract publication, unpublished data);

Study design: inclusion criteria, number of randomised participants, duration of follow‐up (12, 24, or 36 months);

Population: baseline mean age, gender, definition of relapse;

Potential effect modifiers: diagnostic criteria (Poser or McDonald criteria), previous treatments with DMTs, by structuring four categories: “no previous treatment with DMTs”, “previous treatment with DMTs”, “uncertain information on previous treatment with DMTs” and “mixed population of patients, previously treated and previously untreated with DMTs”;

Intervention: active agent, dose, frequency, or duration of treatment;

Funding source.

Outcome data

For dichotomous outcomes, we extracted the number of participants experiencing the event of interest over the number of randomised participants.

For continuous outcomes relative to the outcomes 'cognitive decline' and 'quality‐of‐life impairment', we extracted the mean and standard deviation of the comparison groups, where possible. We extracted data at baseline and end point, as well as change scores. We used change scores in case end point scores were not reported ( Da Costa 2013 ). We extracted data at the authors' defined timing points.

When outcomes were not reported at our predefined time points, we extracted data as close as possible to that time point. We extracted the authors' definition of relapses and disability worsening. We extracted arm‐level data, when available.

We further extracted the characteristics associated with the monitoring and reporting of adverse events, considering specific factors that may have a large influence on adverse event data. We evaluated methods of monitoring and detecting adverse events in each primary study: Did the researchers actively monitor for adverse events, or did they simply provide spontaneous reporting of adverse events that arose? Did the authors define adverse events according to an accepted international classification and report the number of SAEs? We reported this information in an additional table called 'Assessment of Adverse Events Monitoring' ( Table 7 ).

Assessment of risk of bias in included studies

We assessed the risk of bias of each included study using the Cochrane Collaboration’s tool for assessing the risk of bias ( Higgins 2011 ). These include: random sequence generation, allocation concealment, blinding of participants and providers, blinding of outcome assessor(s), incomplete outcome data, and selective outcome reporting. Other potential risks of bias included the role of the sponsor: we judged a study as being at high risk of bias if it was funded by industry, and it was stated that the funder was involved in data management, analysis and interpretation, in writing of the study report, or where it was reported that the funders approved the final version of the paper; we judged studies as being at high risk of bias also if the first or last author and authors who performed the statistical analysis were employed by industry. We explicitly judged the risk of bias of each study on each criterion and classified it as being at 'low', 'high', or 'unclear' risk of bias. We judged incomplete outcome data as at low risk of bias when numbers and causes of dropouts were balanced (i.e. absence of a significant difference) between arms and appeared to be unrelated to the studied outcomes. We assessed selective outcome reporting bias by comparing outcomes reported in the study protocol along with published outcome results. If a study protocol was not available, we assigned a judgement of unclear risk of bias.

Two authors (SM, MGL) assessed the risk of bias of each study independently and resolved any disagreement by discussion to reach consensus.

Measures of treatment effect

Relative treatment effects

For dichotomous outcomes (i.e. disability and relapses), we reported risk ratios (RRs) and 95% confidence intervals (CIs). If the number of observed events had been small (less than 5% of the sample per group), and if studies had balanced treatment groups, we reported the Peto odds ratios (ORs) with 95% CIs. For continuous outcomes, we calculated the mean difference (MD) or standardised mean difference (SMD) if the same continuous outcome was measured with different metrics. To interpret SMD, we used the guiding principles ( Guyatt 2013 ) of thresholds for small (SMD = ±0.2), moderate (SMD = ±0.5), and large effects (SMD = ±0.8). We presented results from network meta‐analysis as summary relative effect sizes (RR, MD, or SMD) for each possible pair of treatments.

Relative treatment ranking

We estimated the ranking probabilities for all treatments at each possible rank for each intervention. We obtained a treatment hierarchy using the surface under the cumulative ranking curve (SUCRA) and mean ranks. SUCRA was expressed as the percentage representing the relative probability of a treatment being amongst the best options without uncertainty ( Salanti 2011 ).

Unit of analysis issues

Studies with multiple treatment groups

For pairwise meta‐analysis, we considered the multi‐arm studies as multiple independent two‐arm studies. For network meta‐analysis (NMA), we accounted for the correlation between the effect sizes from multi‐arm studies ( Salanti 2012 ). For studies with multi‐arm trials involving the same agent at different doses compared to a control treatment, we converted the treatment arms into a single arm by merging the different doses, summing the number of events, and calculating the sample size.

Studies with multiple outcome scales

MS‐specific scales (e.g. MSQOL‐54, MSIS 29) were not combined with non‐MS‐specific scales (e.g. SF‐36 or EQ‐5D index). Where several scales were used in one RCT, we selected the scale that provided lower heterogeneity in combination (via SMD) with the others across studies.

Dealing with missing data

We used data that reflected the intention‐to‐treat (ITT) analysis for each included outcome. Primary analysis was performed considering the number of patients with the event in relation to the number of randomised individuals. In the case of participants with missing data, primary analysis was performed without any imputation. For adverse events, we used data from participants who received at least one dose of the study medication. Where standard deviations were missing for continuous outcomes, we calculated them according to the Cochrane Handbook for Systematic Reviews of Interventions ( Higgins 2021 ).

Assessment of heterogeneity

Assessment of clinical heterogeneity within treatment comparisons

To evaluate the presence of heterogeneity deriving from different characteristics of study participants, we assessed differences in age, disease duration, and baseline EDSS scores across the trials using information reported in the table ' Characteristics of included studies '.

Assessment of transitivity across treatment comparisons

We expected that the transitivity assumption held, assuming that all pairwise comparisons did not differ with respect to the distribution of effect modifiers. We evaluated the assumption of transitivity by comparing potential effect modifiers, which are reported in the ' Data extraction and management ' section, across the different pairwise comparisons.

Assessment of reporting biases

We planned to evaluate the possibility of reporting bias by means of contour‐enhanced funnel plots ( Peters 2008 ). Contour‐enhanced funnel plots show areas of statistical significance, and they can help in distinguishing reporting bias from other possible reasons for asymmetry. In a network of interventions, each study estimates the relative effect of different interventions, so asymmetry in the funnel plot cannot be judged. To account for this, we used an adaptation of the funnel plot by subtracting from each study‐specific effect size the mean of meta‐analysis of the study‐specific comparison and plotted it against the study's standard error ( Chaimani 2012 ; Chaimani 2013 ). We employed the comparison‐adjusted funnel plot for all placebo‐controlled trials. Note that any asymmetry in the plot indicates the presence of small study effects and not necessarily reporting bias.

Data synthesis

Methods for direct treatment comparisons

We performed conventional pairwise meta‐analyses for each primary outcome using a random‐effects model for each treatment comparison with at least two studies ( DerSimonian 1986 ).

Methods for indirect and mixed comparisons

We performed network meta‐analysis using a random‐effects model within a frequentist setting assuming equal heterogeneity across all comparisons, and we accounted for correlations induced by multi‐arm studies ( Miladinovic 2014 ; Salanti 2012 ). The models enabled us to estimate the probability of each intervention being at each possible rank for each outcome, given the relative effect sizes as estimated in network meta‐analysis. We summarised the probabilities of a treatment being at each possible rank using SUCRAs and mean rank. We performed network meta‐analysis in Stata 13 using the 'mvmeta' command and self‐programmed Stata routines available at http://www.mtm.uoi.gr ( Chaimani 2013 ; White 2011 ; White 2012 ).

Assessment of statistical heterogeneity

We statistically assessed the presence of heterogeneity for all direct pairwise comparisons using the common τ 2 and I 2 statistics. The assessment of statistical heterogeneity in the entire network was based on the magnitude of the heterogeneity variance parameter (τ 2 ) estimated from the network meta‐analysis models ( Jackson 2014 ).

Assessment of statistical inconsistency

Consistency in a network of treatments refers to the agreement between direct and indirect estimates. Joint analysis of treatments can be misleading if the network is substantially inconsistent. Inconsistency can be present if the trials in the network have different protocols and their inclusion/exclusion criteria are not comparable or may result in an uneven distribution of the effect modifiers across groups of trials that compare different treatments. To evaluate the presence of inconsistency locally, we used both the loop‐specific approach and the node splitting approach by using the software STATA ( Chaimani 2013 ; Veroniki 2013 ). We used the 'design‐by‐treatment' model to evaluate the assumption of consistency in the entire network ( Higgins 2012 ). This method accounts for different sources of inconsistency that can occur when studies with different designs (two‐arm trials versus three‐arm trials) give different results, as well as disagreement between direct and indirect evidence. Using this approach, we inferred the presence of inconsistencies from any source in the entire network based on a Chi 2 test. We performed the design‐by‐treatment model in Stata using the 'mvmeta' command.

Subgroup analysis and investigation of heterogeneity

We planned to perform subgroup analyses for efficacy outcomes at 12, 24, and 36 months' follow‐up by using the following effect modifiers as possible sources of inconsistency or heterogeneity, or both:

Diagnostic criteria (Poser or McDonald criteria);

Previous treatment with immunomodulators or immunosuppressants (no or yes), i.e. first‐ or second‐line treatments.

Sensitivity analysis

We planned to perform the following sensitivity analyses:

Including only trials with low risk of selection bias and attrition bias;

Excluding trials with a total sample size of fewer than 50 randomised participants to detect potential small study effects.

Summary of findings and assessment of the certainty of the evidence

We presented the main results of the review in a Summary of findings (SoF) table, according to recommendations described in Chapter 14 of the Cochrane Handbook for Systematic Reviews of Interventions ( Schünemann 2022a ). We assessed the certainty of the NMA estimates for the outcomes below using the GRADE approach ( Brignardello‐Petersen 2018 ):

Proportion of people who experienced new relapses over 12, 24 and 36 months;

Proportion of people who experienced disability worsening over 24 months;

Proportion of people who withdrew due to any adverse event;

Number of participants with any (one or more) SAEs.

Since the results of this review and NMA will serve as the evidence base for guidance on the use of DMTs in people with RRMS, the certainty of the evidence for this review was assessed using a fully contextualised approach. A fully contextualised approach is important in NMAs to incorporate the value of individual outcomes in the overall interpretation of the results ( Schünemann 2022b ). This involved predefining quantitative thresholds to determine the magnitude of each health effect (desirable or undesirable) measured by means of each outcome. The magnitudes of desirable and undesirable health effects were defined according to the GRADE wording as 'trivial', 'small', 'moderate', and 'large'.

For this NMA, we used outcomes assessed by the MSIF Essential Medicines Panel – which was convened to make recommendations on essential medicines for MS. The value of the outcomes was assessed by the guideline panel to judge both the priority of outcomes (not important/important/critical) and a health state utility value (HSUV) corresponding to the outcome in question. The panel identified nine critical outcomes. In this review, the authors further differentiated and reported those outcomes as four primary and five secondary outcomes. The HSUV was derived from a review of reviews or panel judgement if not identified from the literature. The HSUV is utilised to calculate thresholds for the magnitude of effects. The thresholds between trivial/small (T1), small/moderate (T2), and moderate/large (T3) were predefined through calculation informed by the health state utility value (HSUV) of each outcome ( Appendix 5 ).

Thresholds were based on absolute anticipated effects of each intervention's outcome estimate using the formula: Threshold = (threshold coefficient/(1 ‐ utility) x 1000). The threshold coefficient was derived from an interim analysis of an ongoing global survey on decision thresholds by assessing respondent judgements across varied disease category examples ( Morgano 2022 ).

We followed GRADE guidance for assessing imprecision using a fully contextualised approach ( Schünemann 2022b ). We graded the certainty of evidence for each outcome considering study limitations, indirectness, inconsistency, imprecision of effect estimates, and risk of reporting bias. According to the software GRADEpro 2008 , we assigned four levels of certainty of evidence: high, moderate, low, and very low.

In order to determine the imprecision of estimates, and therefore make imprecision judgements including downgrading by 1, 2, or 3 levels for certainty, point estimates of observed effects and their 95% CIs were contextualised in relation to the predefined thresholds ( Hultcrantz 2017 ). In accordance with the GRADE guidance on imprecision, the overall imprecision of interventions was assessed across all outcomes with guideline panel input. If most outcomes were not downgraded for imprecision, the overall certainty was not necessarily downgraded to the lowest certainty ( Schünemann 2022b ).

Results

Description of studies

Results of the search

We identified 16,926 reports through an electronic database search on 21 September 2021. A top‐up search conducted on 8 August 2022 identified a further 842 records (See Figure 1 ). A total of 59 reports were carried over from the previous version of this review ( Tramacere 2015 ), and 44 reports from another relevant NMA review which included studies with people with both RRMS and PMS ( Filippini 2013 ). We screened a total of 12,507 de‐duplicated records, of which 12,328 were excluded based on the titles and abstracts. We evaluated 179 full texts as potentially meeting inclusion criteria. We excluded 41 studies with reasons ( Characteristics of excluded studies ), and identified 21 ongoing studies ( Characteristics of ongoing studies ). After full‐text review, we included 50 studies. A total of 13 studies were identified as potentially meeting inclusion criteria. These studies are summarised in Characteristics of studies awaiting classification and will undergo full‐text review in a future update of this review.

On the following DMTs, we found no studies satisfying our inclusion criteria: cyclophosphamide, diroximel fumarate, fludarabine, leflunomide, methotrexate, minocycline, mycophenolate mofetil, rituximab, siponimod, and steroids.

Included studies

We included 50 studies involving 36,541 participants (68.6% female and 31.4% male) and published between 1987 and 2021 in this review.

We included 34 studies already included in previous reviews ( Achiron 1998 ; ADVANCE 2014 ; AFFIRM 2006 ; ALLEGRO 2012 ; BECOME 2009 ; BEYOND 2009 ; Bornstein 1987 ; BRAVO 2014 ; CAMMS223 2008 ; CARE‐MS I 2012 ; CARE‐MS II 2012 ; CombiRx 2013 ; CONFIRM 2012 ; DEFINE 2012 ; Etemadifar 2007 ; Fazekas 1997 ; FREEDOMS 2010 ; FREEDOMS II 2014 ; GALA 2013 ; Goodkin 1991 ; IFNB MS Group 1993 ; INCOMIN 2002 ; Johnson 1995 ; Koch‐Henriksen 2006 ; Lewanska 2002 ; MAIN 2014 ; Millefiorini 1997 ; MSCRG 1996 ; PRISMS 1998 ; REGARD 2008 ; SELECT 2013 ; TEMSO 2011 ; TOWER 2014 ; TRANSFORMS 2010 ). Three studies that were previously identified as ongoing in Tramacere 2015 were also included: DECIDE 2015 ; OPERA I 2017 ; OPERA II 2017 . We also included two studies that were previously excluded in Tramacere 2015 as we obtained further information and found that they met our pre‐defined inclusion criteria ( Etemadifar 2006 ; Knobler 1993 ).

The updated search identified an additional 11 new studies meeting our inclusion criteria, which were not previously included ( ASCLEPIOS I 2020 ; ASCLEPIOS II 2020 ; ASSESS 2020 ; CLARITY 2010 ; CONCERTO 2021 ; Gobbi 2013 ; GOLDEN 2017 ; Mokhber 2014 ; OPTIMUM 2021 ; RADIANCE 2019 ; SUNBEAM 2019 ).

Of the 50 included studies, seven studies ( ASCLEPIOS I 2020 ; ASCLEPIOS II 2020 ; BECOME 2009 ; RADIANCE 2019 ; SUNBEAM 2019 ; TEMSO 2011 ; TOWER 2014 ) included a mixed sample of participants with not only relapsing but also other forms of MS. As per the methods, only those with more than 80% of the sampled population affected by relapsing forms were included in analyses.

Median follow‐up was 24 months (including 12‐month follow‐up (11 studies); 18 months (1); 24 months (32), 25 months (1); 30 months (2) and 36 months (4)). Twenty‐five studies were placebo‐controlled and 25 were head‐to‐head studies. Funding came from industry in 40 studies, from public sources in six cases ( Bornstein 1987 ; CombiRx 2013 ; Gobbi 2013 ; INCOMIN 2002 ; Lewanska 2002 ; MAIN 2014 ) and funding sources were not reported in four cases ( Etemadifar 2006 ; Etemadifar 2007 ; Millefiorini 1997 ; Mokhber 2014 ).

Of the 50 included studies, one ( CARE‐MS II 2012 ) included only people with MS previously treated with DMTs; five ( BEYOND 2009 ; CAMMS223 2008 ; CARE‐MS I 2012 ; INCOMIN 2002 ; Mokhber 2014 ) included only people with MS previously untreated with DMTs, 17 ( Achiron 1998 ; AFFIRM 2006 ; BECOME 2009 ; Bornstein 1987 ; CombiRx 2013 ; Etemadifar 2006 ; Etemadifar 2007 ; Fazekas 1997 ; Goodkin 1991 ; IFNB MS Group 1993 ; Johnson 1995 ; Knobler 1993 ; Koch‐Henriksen 2006 ; Lewanska 2002 ; Millefiorini 1997 ; MSCRG 1996 ; REGARD 2008 ) did not report data about previous treatments with DMTs, and 27 ( ADVANCE 2014 ; ALLEGRO 2012 ; ASCLEPIOS I 2020 ; ASCLEPIOS II 2020 ; ASSESS 2020 ; BRAVO 2014 ; CLARITY 2010 ; CONCERTO 2021 ; CONFIRM 2012 ; DECIDE 2015 ; DEFINE 2012 ; FREEDOMS 2010 ; FREEDOMS II 2014 ; GALA 2013 ; Gobbi 2013 ; GOLDEN 2017 ; MAIN 2014 ; OPERA I 2017 ; OPERA II 2017 ; OPTIMUM 2021 ; PRISMS 1998 ; RADIANCE 2019 ; SELECT 2013 ; SUNBEAM 2019 ; TEMSO 2011 ; TOWER 2014 ; TRANSFORMS 2010 ) included a mixed population of patients with and without previous treatment with DMTs, but did not report separate outcome data for the two subgroups.

The table, Characteristics of included studies , provides details of the included studies. Table 7 provides the characteristics associated with the monitoring and reporting of adverse events, considering specific factors.

We identified 21 ongoing trials ( EUCTR2013‐003884‐71‐BE ; EUCTR2012‐003647‐30‐SK ; EUCTR2014‐001012‐19‐NL ; EUCTR2012‐000540‐10‐PL ; EUCTR2018‐000284‐93‐BG ; EUCTR2019‐001505‐24‐NO ; EUCTR2020‐002981‐15‐DK ; EUCTR2018‐005038‐39‐GB ; IRCT20130812014333N ; IRCT201404195280N ; NCT04121221 ; NCT04056897 ; NCT04121403 ; NCT04688788 ; NCT04578639 ; WHO‐ICTRP 002519 ; WHO‐ICTRP PER‐024‐14 ). Of these, four studies have been withdrawn, and results are unlikely to be available ( EUCTR2013‐002082‐19‐SE ; NCT01404117 ; NCT01975298 ; NCT01941004 ). The remaining 17 studies will be assessed and included, where appropriate, in a future update of this review. Please see Characteristics of ongoing studies for further details.

Excluded studies

After full‐text review, we excluded 41 studies with reasons (see Characteristics of excluded studies ):

Fifteen were excluded where the randomised, blinded portion of the trial was less than one year ( Bar‐Or 2017 ; Boyko 2016 ; Cohen 2015 ; Cohen 2016 ; Comi 2001 ; Fazekas 2008 ; IMPROVE 2010 ; Le Page 2015 ; Newsome 2015 ; Ochi 2018 ; OWIMS 1999 ; Saida 2017 ; Simaniv 2019 ; TENERE 2014 ; Ziemssen 2017 ); one study was non‐randomised ( Boiko 2018 ); one was a pooled post hoc analysis ( Agius 2014 ) and five used a comparator that was not admissible under our inclusion criteria ( Cascione 2018 ; Coyle 2017 ; Cree 2018 ; Fox 2014 ; Lampl 2013 ).

Studies excluded at full‐text review included five that were previously included in Tramacere 2015 for insufficient follow‐up duration ( Comi 2001 ; Fazekas 2008 ; OWIMS 1999 ; TENERE 2014 ) and incorrect comparator ( EVIDENCE 2007 ), as well as four included in Filippini 2013 for wrong intervention ( SENTINEL 2006 ), wrong publication type ( Ghezzi 1989 ) and mixed populations where < 80% were people with relapsing MS ( British and Dutch 1988 ; Milanese 1993 ).

The previous version of this review excluded the following studies: six studies for insufficient duration ( CHOICE 2010 ; Kappos 2006 ; Kappos 2008 ; Kappos 2011 ; Saida 2012 ; Sorensen 2014 ), five studies evaluating combination therapies ( ACT 2009 ; Freedman 2012 ; Havrdova 2009 ; Khoury 2010 ; SENTINEL 2006 ), two studies evaluating treatments that are not included in this review ( Ashtari 2011 ; ATAMS 2014 ), a study that was non‐randomised ( Calabrese 2012 ), and one dose‐finding study without a control group ( FORTE 2011 ). Etemadifar 2006 and Knobler 1993 were included in this version of the review, though they were excluded in the previous version.

Risk of bias in included studies

The risks of bias in the included studies for individual domains are summarised in Figure 2 and Figure 3 .

Allocation

Sequence generation

Ten studies (20%) did not provide enough information to assess sequence generation (unclear risk) ( Bornstein 1987 ; CONCERTO 2021 ; Etemadifar 2006 ; Gobbi 2013 ; GOLDEN 2017 ; IFNB MS Group 1993 ; Johnson 1995 ; Knobler 1993 ; MSCRG 1996 ; SELECT 2013 ), and the remaining 40 (80%) reported adequate methods (low risk).

Allocation concealment

One trial (2%) used an unconcealed procedure (high risk) ( Bornstein 1987 ). Twenty‐one (42%) did not provide sufficient information to enable a risk of bias judgement (unclear risk) ( Achiron 1998 ; BECOME 2009 ; BEYOND 2009 ; BRAVO 2014 ; CLARITY 2010 ; CONCERTO 2021 ; Etemadifar 2006 ; Etemadifar 2007 ; FREEDOMS 2010 ; FREEDOMS II 2014 ; GALA 2013 ; Gobbi 2013 ; GOLDEN 2017 ; Goodkin 1991 ; IFNB MS Group 1993 ; Johnson 1995 ; Knobler 1993 ; Lewanska 2002 ; Mokhber 2014 ; MSCRG 1996 ; REGARD 2008 ). The remaining 38 (76%) studies reported adequate methods of allocation concealment (low risk).

Blinding

Blinding of participants and personnel (performance bias )

Seventeen (34%) studies were not blinded (high risk)( ASSESS 2020 ; BECOME 2009 ; BEYOND 2009 ; Bornstein 1987 ; BRAVO 2014 ; CAMMS223 2008 ; CARE‐MS I 2012 ; CARE‐MS II 2012 ; CONFIRM 2012 ; Etemadifar 2006 ; Etemadifar 2007 ; Gobbi 2013 ; GOLDEN 2017 ; INCOMIN 2002 ; Koch‐Henriksen 2006 ; MAIN 2014 ; REGARD 2008 ); nine (18%) studies did not provide sufficient information to enable assessment (unclear risk) ( DEFINE 2012 ; Fazekas 1997 ; FREEDOMS 2010 ; IFNB MS Group 1993 ; Johnson 1995 ; Lewanska 2002 ; Millefiorini 1997 ; MSCRG 1996 ; PRISMS 1998 ); the remaining 24 (48%) studies reported that participants and investigators were blinded (low risk).

Blinding of outcome assessor (detection bias)

Six studies (12%) were at high risk ( INCOMIN 2002 ; Koch‐Henriksen 2006 ; MAIN 2014 ; Millefiorini 1997 ; TEMSO 2011 ; TRANSFORMS 2010 ), eleven studies (22%) did not provide sufficient information to enable assessment (unclear risk), and the remaining 33 studies (66%) were at low risk of detection bias (i.e. they reported that outcome assessors were blinded).

Incomplete outcome data

We judged 16 of 50 (32%) included studies as meeting the criteria for high risk of bias due to incomplete outcome data (balanced numbers across intervention groups with similar reasons for loss to follow‐up) ( ADVANCE 2014 ; ALLEGRO 2012 ; CAMMS223 2008 ; CARE‐MS I 2012 ; CARE‐MS II 2012 ; CombiRx 2013 ; CONCERTO 2021 ; CONFIRM 2012 ; DEFINE 2012 ; FREEDOMS 2010 ; FREEDOMS II 2014 ; GOLDEN 2017 ; Koch‐Henriksen 2006 ; OPERA I 2017 ; OPERA II 2017 ; TOWER 2014 ) and seven (14%) ( Etemadifar 2006 ; IFNB MS Group 1993 ; INCOMIN 2002 ; Knobler 1993 ; REGARD 2008 ; SELECT 2013 ; TEMSO 2011 ) did not provide sufficient information to assess risk of incomplete outcome data (unclear risk). The remaining 27 (54%) studies were judged to be low risk.

Selective reporting

Five studies (10%) were judged as being at high risk of bias for selective reporting because of non‐reporting all prespecified primary benefit outcomes ( CONFIRM 2012 ; DEFINE 2012 ; Gobbi 2013 ; Knobler 1993 ; TEMSO 2011 ). Two studies (4%) were judged as being at unclear risk of bias in this domain because of the lack of a protocol ( Etemadifar 2006 ) or because primary and secondary outcomes were submitted to Clinical Trials.gov after study completion ( CLARITY 2010 ). The remaining 43 (86%) reported all prespecified primary benefit outcomes and were judged as having low risk of bias.

Other potential sources of bias

We judged 34 studies (68%) as being at high risk of other bias; this includes the role of the sponsor in authorship of the study report or in data management or analysis ( ADVANCE 2014 ; AFFIRM 2006 ; ALLEGRO 2012 ; ASCLEPIOS I 2020 ; ASCLEPIOS II 2020 ; ASSESS 2020 ; BECOME 2009 ; BEYOND 2009 ; BRAVO 2014 ; CAMMS223 2008 ; CARE‐MS I 2012 ; CARE‐MS II 2012 ; CLARITY 2010 ; CONFIRM 2012 ; DECIDE 2015 ; DEFINE 2012 ; FREEDOMS 2010 ; FREEDOMS II 2014 ; GALA 2013 ; GOLDEN 2017 ; Johnson 1995 ; Knobler 1993 ; MSCRG 1996 ; OPERA I 2017 ; OPERA II 2017 ; OPTIMUM 2021 ; PRISMS 1998 ; RADIANCE 2019 ; REGARD 2008 ; SELECT 2013 ; SUNBEAM 2019 ; TEMSO 2011 ; TOWER 2014 ; TRANSFORMS 2010 ); eight studies (16%) were judged as being at unclear risk of bias for this domain because the role of the study sponsor was unclear ( Achiron 1998 ; Etemadifar 2006 ; Etemadifar 2007 ; Fazekas 1997 ; Goodkin 1991 ; IFNB MS Group 1993 ; Koch‐Henriksen 2006 ; Millefiorini 1997 ). The remaining 8 (16%) were judged as being at low risk of bias.

Effects of interventions

See: Table 1 ; Table 2 ; Table 3 ; Table 4 ; Table 5 ; Table 6

Table 1 , Table 2 , Table 3 , Table 4 , Table 5 , and Table 6 provide overall estimates of treatment effects compared with placebo and the certainty of the available evidence obtained through network meta‐analyses for the four efficacy outcomes (chance of experiencing one or more relapses over 12 months, chance of experiencing one or more relapses over 24 months, chance of experiencing one or more relapses over 36 months, chance of disability getting worse over 24 months), and for the two safety outcomes (discontinuation due to adverse events and SAEs).

Figure 4 , Figure 5 , and Figure 6 show the network geometries for the efficacy and safety outcomes of immunomodulators and immunosuppressants included in the review ( Primary outcomes ; Secondary outcomes ). Each line links the treatments that have been directly compared in studies. The thickness of the line is proportional to the number of participants included in the comparison and the width of each circle is proportional to the number of studies included in the comparison.

Figure 7 shows network estimates of primary efficacy and safety outcomes of each treatment against placebo within the networks.

Table 8 , Table 9 , Table 10 , Table 11 , Table 12 , Table 13 , and Table 14 show network estimates of primary efficacy and safety outcomes of each treatment against placebo and against each other treatment included in the network.

Appendix 6 , Appendix 7 , Appendix 8 , Appendix 9 , Appendix 10 , Appendix 11 , Appendix 12 , and Appendix 13 show estimates of secondary efficacy and safety outcomes of each treatment against placebo and against each other treatment included in the network.

1. Primary outcomes

1.1 Efficacy

Relapses (over 12, 24 and 36 months) and disability worsening (over 24 and 36 months)

Pairwise meta‐analysis (direct comparisons)

Treatment estimates from pairwise meta‐analyses for each efficacy outcome are reported in Analysis 1.1 , Analysis 1.2 , Analysis 1.3 , Analysis 1.4 , and Analysis 1.5 .

Network meta‐analysis estimates (combination of direct and indirect comparisons) of treatment effects

See: Table 1 ; Table 2 ; Table 3 ; Table 4 .

a) Relapses over 12 months were provided in 18 studies involving 9310 participants with RRMS (25.49% of the participants in this review) ( Achiron 1998 ; ADVANCE 2014 ; AFFIRM 2006 ; ASSESS 2020 ; BECOME 2009 ; Bornstein 1987 ; Etemadifar 2007 ; GALA 2013 ; Gobbi 2013 ; GOLDEN 2017 ; Goodkin 1991 ; Lewanska 2002 ; MAIN 2014 ; Millefiorini 1997 ; PRISMS 1998 ; SELECT 2013 ; TOWER 2014 ; TRANSFORMS 2010 ) and assessing 13 treatments. The network geometry for relapses over 12 months is shown in Figure 4 .

Nine treatments assessed in 18 studies were compared to placebo, of which seven treatments were evaluated in head‐to‐head comparisons in seven studies. Using placebo as a common comparator (see Figure 7 ), treatment with natalizumab results in a large reduction of people with relapses (RR 0.52, 95% CI 0.43 to 0.63; high‐certainty evidence). Treatment with fingolimod (RR 0.48, 95% CI 0.39 to 0.57; moderate‐certainty evidence), daclizumab (RR 0.55, 95% CI 0.42 to 0.73; moderate‐certainty evidence) or immunoglobulins (RR 0.60, 95% CI 0.47 to 0.79; moderate‐certainty evidence) probably results in a large reduction of people with relapses. Treatment with mitoxantrone (RR 0.40, 95% CI 0.21 to 0.74; low‐certainty evidence), teriflunomide (RR 0.66, 95% CI 0.55 to 0.78; low‐certainty evidence), pegylated interferon beta‐1a (RR 0.68, 95% CI 0.56 to 0.82; low‐certainty evidence) or glatiramer acetate (RR 0.64, 95% CI 0.55 to 0.75; low‐certainty evidence) may result in a large reduction of people with relapses. Treatment with interferon beta‐1a (Avonex, Rebif) (RR 0.76, 95% CI 0.68 to 0.85; low‐certainty evidence) may moderately reduce the number of people with relapses. Treatment with interferon beta‐1b (RR 0.82, 95% CI 0.50 to 1.33; very low‐certainty evidence) may moderately reduce the number of people with relapses, but the evidence is very uncertain. When interferon beta 1‐a and beta 1‐b (RR 1.42, 95% CI 0.78 to 2.60; very low‐certainty evidence) were both used in the same intervention, there may have been a large increase in the number of people with relapses, but the evidence is very uncertain. Differences across treatments were found and are shown in Table 8 .

b) Relapses over 24 months were provided in 28 studies involving 19,869 participants with RRMS (54.37% of those included in this review) ( Achiron 1998 ; AFFIRM 2006 ; ALLEGRO 2012 ; BECOME 2009 ; BEYOND 2009 ; Bornstein 1987 ; BRAVO 2014 ; CAMMS223 2008 ; CARE‐MS I 2012 ; CARE‐MS II 2012 ; CLARITY 2010 ; CONCERTO 2021 ; CONFIRM 2012 ; DEFINE 2012 ; Etemadifar 2006 ; Fazekas 1997 ; FREEDOMS 2010 ; FREEDOMS II 2014 ; Goodkin 1991 ; IFNB MS Group 1993 ; INCOMIN 2002 ; Johnson 1995 ; Koch‐Henriksen 2006 ; MAIN 2014 ; Millefiorini 1997 ; MSCRG 1996 ; OPTIMUM 2021 ; PRISMS 1998 ; REGARD 2008 ; TEMSO 2011 ) and assessing 15 treatments. The network geometry for relapses over 24 months is shown in Figure 4 .

Twelve treatments, assessed in 21 studies, were compared to placebo, ten treatments were evaluated in head‐to‐head comparisons in 11 studies and two studies involving 2751 participants had both placebo and active treatment arms. Using placebo as a common comparator, treatment with cladribine (RR 0.53, 95% CI 0.44 to 0.64; high‐certainty evidence), alemtuzumab (RR 0.57, 95% CI 0.47 to 0.68; high‐certainty evidence) and natalizumab (RR 0.56, 95% CI 0.48 to 0.65; high‐certainty evidence) resulted in a large decrease in people with relapses. Treatment with fingolimod (RR 0.54, 95% CI 0.48 to 0.60; moderate‐certainty evidence), dimethyl fumarate (RR 0.62, 95% CI 0.55 to 0.70; moderate‐certainty evidence), and ponesimod (RR 0.58, 95% CI 0.48 to 0.70; moderate‐certainty evidence) probably results in a large decrease of people with relapses. Treatment with glatiramer acetate (RR 0.84, 95%, CI 0.76 to 0.93; moderate‐certainty evidence) and interferon beta‐1a (Avonex, Rebif) (RR 0.84, 95% CI 0.78 to 0.91; moderate‐certainty evidence) probably moderately decreases people with relapses. Treatment with mitoxantrone (RR 0.47, 95% CI 0.27 to 0.80; low‐certainty evidence) may result in a large decrease in people with relapses. Treatment with immunoglobulins (RR 0.73, 95% CI 0.59 to 0.90; low‐certainty evidence) may result in a large reduction of people with relapses. Treatment with interferon beta‐1b (Betaferon) (RR 0.85, 95% CI 0.76 to 0.94; low‐certainty evidence) and laquinimod (RR 0.83, 95% CI 0.76 to 0.91; low‐certainty evidence) may moderately decrease the number of people with relapses. Treatment with teriflunomide (RR 0.82, 95% CI 0.71 to 0.94; very low‐certainty evidence) and azathioprine (RR 0.77, 95% CI 0.51 to 1.18; very low‐certainty evidence) may moderately decrease people with relapses but the evidence is very uncertain. When interferon beta 1‐a and beta 1‐b (RR 1.21, 95% CI 0.66 to 2.19; very low‐certainty evidence) were both used in the same intervention arm, there may have been a moderate increase in the number of people with relapses, but the evidence is very uncertain. Differences across treatments were found and are shown in Table 9 .

c) Relapses over 36 months were provided in five studies involving 3087 participants with RRMS (8.44% of the participants in this review) ( CAMMS223 2008 ; CombiRx 2013 ; DECIDE 2015 ; IFNB MS Group 1993 ; Knobler 1993 ) and assessing five treatments. The network geometry for relapses over 36 months is shown in Figure 4 .

Only one treatment, assessed in two studies, was compared to placebo. Three treatments were evaluated in head‐to‐head comparisons in three studies. Treatment with interferon beta‐1b (Betaferon) (RR 0.86, 95% CI 0.67 to 1.11; very low‐certainty evidence) may moderately decrease people with relapses, but the evidence is very uncertain. Differences across treatments were found and are shown in Table 10 .

d) Disability worsening over 24 months was available from 31 studies and 24,303 participants with RRMS (66.51% of those included in this review) ( Achiron 1998 ; AFFIRM 2006 ; ALLEGRO 2012 ; ASCLEPIOS I 2020 ; ASCLEPIOS II 2020 ; BECOME 2009 ; BEYOND 2009 ; Bornstein 1987 ; BRAVO 2014 ; CARE‐MS I 2012 ; CARE‐MS II 2012 ; CLARITY 2010 ; CONCERTO 2021 ; CONFIRM 2012 ; DEFINE 2012 ; Fazekas 1997 ; FREEDOMS 2010 ; FREEDOMS II 2014 ; Goodkin 1991 ; IFNB MS Group 1993 ; INCOMIN 2002 ; Johnson 1995 ; Koch‐Henriksen 2006 ; MAIN 2014 ; Millefiorini 1997 ; MSCRG 1996 ; OPERA I 2017 ; OPERA II 2017 ; OPTIMUM 2021 ; RADIANCE 2019 ; REGARD 2008 ; TEMSO 2011 ) and assessing 18 treatments. The network geometry for disability worsening over 24 months is shown in Figure 4 .

Eleven treatments, assessed in 20 studies, were compared to placebo (see Figure 7 ), thirteen treatments were evaluated in head‐to‐head comparisons in 15 studies and two studies involving 2751 participants had both placebo and active treatment arms. Using placebo as a common comparator, treatment with natalizumab (RR 0.59, 95% CI 0.46 to 0.75; moderate‐certainty evidence) probably results in a moderate reduction of people with disability worsening. Treatment with dimethylfumarate (RR 0.65, 95% CI 0.55 to 0.77; low‐certainty evidence), alemtuzumab (RR 0.67, 95% CI 0.46 to 0.99; low‐certainty evidence), and fingolimod (RR 0.68, 95% CI 0.56 to 0.83; low‐certainty evidence) may moderately reduce people with disability worsening. Treatment with cladribine (RR 0.72, 95% CI 0.56 to 0.91; low‐certainty evidence) and interferon beta‐1b (Betaferon) (RR 0.77 95% CI 0.62 to 0.94; low‐certainty evidence) may result in a small reduction of people with disability worsening. Treatment with mitoxantrone (RR 0.20, 95% CI 0.05 to 0.83; very low‐certainty evidence) may result in a large reduction of people who experience disability worsening, but the evidence is very uncertain. The evidence is very uncertain about the following effects on numbers of people experiencing disability worsening: moderate reductions with ofatumumab (RR 0.54, 95% CI 0.38 to 0.77; very low‐certainty evidence), ocrelizumab (RR 0.61, 95% CI 0.41 to 0.90; very low‐certainty evidence), azathioprine (RR 0.60, 95% CI 0.22 to 1.63; very low‐certainty evidence) and ponesimod (RR 0.63, 95% CI 0.41 to 0.96; very low‐certainty evidence). The evidence is very uncertain about the following effects on numbers of people experiencing disability worsening: small reductions with glatiramer acetate (RR 0.74, 95% CI 0.61 to 0.89; very low‐certainty evidence), teriflunomide (RR 0.76, 95% CI 0.61 to 0.95; very low‐certainty evidence) and laquinimod (RR 0.78, 95% CI 0.63 to 0.96; very low‐certainty evidence). Treatment with interferon beta‐1a (RR 0.92, 95% CI 0.73 to 1.16; low‐certainty evidence) may result in a trivial reduction in the number people who experience disability worsening. Treatment with immunoglobulins (RR 0.75, 95% CI 0.41 to 1.37; very low‐certainty evidence) may result in a small reduction in the number of people with disability worsening, but the evidence is very uncertain. When interferon beta 1‐a and beta 1‐b (RR 3.19, 95% CI 0.31 to 33.21; very low‐certainty evidence) were both used in the same intervention arm, there may have been a large increase in the number of people with disability progression, but the evidence is very uncertain. Treatment with ozanimod (RR 1.19, 95% CI 0.74 to 1.91; very low‐certainty evidence) may result in a trivial increase in the number of people with disability progression, but the evidence is very uncertain. Differences across treatments were found and are shown in Table 11 .

e) Disability worsening over 36 months was available from three studies and 2684 participants with RRMS (7.35% of those included in this review)( CAMMS223 2008 ; CombiRx 2013 ; DECIDE 2015 ) and assessing four treatments. No studies were compared to placebo. Differences across treatments were found and are shown in Table 12 . The network geometry for disability worsening over 36 months is shown in Figure 4 .

1.2 Safety

Treatment discontinuation due to adverse events and serious adverse events (SAEs)

Pairwise meta‐analysis (direct comparisons)

Treatment estimates for pairwise meta‐analyses are reported in Analysis 2.1 ; Analysis 2.2 .

Network meta‐analysis estimates (combination of direct and indirect comparisons) of treatment effects

See: Table 5 ; Table 6 .

a) Treatment discontinuation due to adverse events was available from 43 studies and 35,410 participants with RRMS (96.9% of those included in this review) ( Achiron 1998 ; ADVANCE 2014 ; AFFIRM 2006 ; ALLEGRO 2012 ; ASCLEPIOS I 2020 ; ASCLEPIOS II 2020 ; ASSESS 2020 ; BEYOND 2009 ; Bornstein 1987 ; BRAVO 2014 ; CAMMS223 2008 ; CARE‐MS I 2012 ; CARE‐MS II 2012 ; CLARITY 2010 ; CombiRx 2013 ; CONCERTO 2021 ; CONFIRM 2012 ; DEFINE 2012 ; DECIDE 2015 ; Etemadifar 2007 ; Fazekas 1997 ; FREEDOMS 2010 ; FREEDOMS II 2014 ; GALA 2013 ; Gobbi 2013 ; GOLDEN 2017 ; Goodkin 1991 ; IFNB MS Group 1993 ; INCOMIN 2002 ; Johnson 1995 ; Lewanska 2002 ; MAIN 2014 ; OPERA II 2017 ; OPERA II 2017 ; OPTIMUM 2021 ; PRISMS 1998 ; RADIANCE 2019 ; REGARD 2008 ; SELECT 2013 ; SUNBEAM 2019 ; TEMSO 2011 ; TOWER 2014 ; TRANSFORMS 2010 ) and assessing 19 treatments. The network geometry for treatment discontinuation due to adverse events is shown in Figure 4 .

Thirteen treatments, assessed in 24 studies, were compared to placebo, 16 treatments were evaluated in head‐to‐head comparisons in 21 studies and two studies involving 2751 participants had both placebo and active treatment arms. Using placebo as a common comparator, treatment with alemtuzumab probably results in a trivial reduction of people who discontinued due to adverse events (OR 0.39, 95% CI 0.19 to 0.79; moderate‐certainty evidence). Treatment with ocrelizumab may result in a trivial reduction in people who discontinue due to adverse events (OR 0.82, 95% CI 0.42 to 1.60; low‐certainty evidence). Treatment with daclizumab (OR 2.55, 95% CI 1.40 to 4.63; moderate‐certainty evidence), fingolimod (OR 1.84, 95% CI 1.31 to 2.57; moderate‐certainty evidence), teriflunomide (OR 1.82, 95% CI 1.19 to 2.79; moderate‐certainty evidence), interferon beta‐1a (OR 1.48, 95% CI 0.99 to 2.20; moderate‐certainty evidence), laquinimod (OR 1.49, 95 % CI 1.00 to 2.15; moderate‐certainty evidence), natalizumab (OR 1.57, 95% CI 0.81 to 3.05), and glatiramer acetate (OR 1.48, 95% CI 1.02 to 2.14; moderate‐certainty evidence) probably result in a trivial increase in people who discontinued due to adverse events. Treatment with interferon beta‐1b (Betaferon) (OR 2.27, 95% CI 1.05 to 4.91; low‐certainty evidence), ofatumumab (OR 2.00, 95% CI 1.05 to 23.81; low‐certainty evidence), cladribine (OR 1.38, 95% CI 0.46 to 4.15; low‐certainty evidence) and dimethyl fumerate (OR 1.35, 95% CI 0.94 to 1.95; low‐certainty evidence) may result in a trivial increase in people who discontinued due to adverse events. Treatment with ozanimod (OR 1.01, 95% CI 0.52 to 1.95; low‐certainty evidence) may have little to no effect on people who discontinued due to adverse effects. Treatment with pegylated interferon beta‐1a (OR 3.58, 95% CI 1.47 to 8.73; very low‐certainty evidence), azathioprine (OR 6.26, 95% CI 0.67 to 58.05) and ponesimod (OR 5.04, 95% CI 2.15 to 11.82; very low‐certainty evidence) may result in a small increase in people who discontinued due to adverse events, but the evidence is very uncertain (see Figure 7 ). Treatment with immunoglobulins (OR 2.49, 95% CI 0.37 to 16.50; very low‐certainty evidence) and interferon beta 1a‐1b (OR 3.02, 95% Ci 0.27 to 33.65; very low‐certainty evidence) may result in a trivial increase in people who discontinued due to adverse effects, but the evidence is very uncertain. Differences across treatments were found and are shown in Table 13 .

b) SAEs were available from 35 studies and 33,998 participants with RRMS (93% of those included in this review) ( ADVANCE 2014 ; AFFIRM 2006 ; ALLEGRO 2012 ; ASCLEPIOS I 2020 ; ASCLEPIOS II 2020 ; ASSESS 2020 ; BEYOND 2009 ; BECOME 2009 ; BRAVO 2014 ; CAMMS223 2008 ; CARE‐MS I 2012 ; CARE‐MS II 2012 ; CLARITY 2010 ; CombiRx 2013 ; CONCERTO 2021 ; CONFIRM 2012 ; DEFINE 2012 ; DECIDE 2015 ; FREEDOMS 2010 ; FREEDOMS II 2014 ; GALA 2013 ; Gobbi 2013 ; GOLDEN 2017 ; Johnson 1995 ; Millefiorini 1997 ; OPERA I 2017 ; OPERA II 2017 ; OPTIMUM 2021 ; RADIANCE 2019 ; REGARD 2008 ; SELECT 2013 ; SUNBEAM 2019 ; TEMSO 2011 ; TOWER 2014 ; TRANSFORMS 2010 ) and assessing 17 treatments. The network geometry for SAEs is shown in Figure 4 .

Eleven treatments, assessed in 18 studies, were compared to placebo, 14 treatments were evaluated in head‐to‐head comparisons in 21 studies, and two studies involving 2751 participants had both placebo and active treatment arms. Compared to placebo (see Figure 7 ), treatment with interferon beta‐1b (OR 0.92, 95% CI 0.55 to 1.54; moderate‐certainty evidence) probably results in a trivial reduction in people who experience SAEs. Treatment with fingolimod (OR 0.86, 95% CI 0.64 to 1.13; low‐certainty evidence) and glatiramer acetate (OR 0.94, 95% CI 0.68 to 1.28; low‐certainty evidence) may result in a trivial reduction in people who experience SAEs. Treatment with mitoxantrone (OR 0.89, 95% CI 0.02 to 47.22; very low‐certainty evidence) may result in a trivial reduction in people who experience SAEs, but the evidence is very uncertain. Treatment with daclizumab may result in a small increase in people with SAEs (OR 1.90, 95% CI 1.21 to 2.99; low‐certainty evidence). Treatment with dimethyl fumerate (OR 1.04, 95% CI 0.71 to 1.52; low‐certainty evidence) and natalizumab (OR 1.24, 95% CI 0.73 to 2.09; low‐certainty evidence) may result in a trivial increase in people who experience SAEs. Treatment with ocrelizumab (OR 1.00, 95% CIs 0.58 to 1.72; very low‐certainty evidence) may result in little or no effect on people who experince SAEs, but the evidence is very uncertain. Treatment with the remaining DMTS may result in a trivial increase in the people experiencing SAEs, but the evidence is very uncertain: alemtuzumab (OR 1.52, 95% CI 0.94 to 2.48; very low‐certainty evidence), cladribine (OR 1.39, 95% CI 0.80 to 2.40; very low‐certainty evidence), interferon beta‐1a (OR 1.21, 95% CI 0.88 to 1.67; very low‐certainty evidence), laquinimod (OR 1.25, 95% CI 0.92 to 1.70; very low‐certainty evidence), pegylated beta‐1a (OR 1.08, 95% CI 0.59 to 1.96; very low‐certainty evidence), ofatumumab (OR 1.52, 95% CI 0.89 to 2.57; very low‐certainty evidence), ozanimod (OR 1.50, 95% CI 0.85 to 2.64; very low‐certainty evidence), ponesimod (OR 1.24, 95% CI 0.66 to 2.35; very low‐certainty evidence) and teriflunomide (OR 1.16, 95% CI 0.81 to 1.64); very low‐certainty evidence). Differences across treatments were found and are shown in Table 14 .

2. Secondary outcomes

2.1 Efficacy

Cognitive decline, quality‐of‐life impairment, new or enlarging T2‐weighted MRI lesions, new gadolinium‐enhancing positive T1‐weighted MRI lesions (at 12, 24, and 36 months)

Pairwise meta‐analysis (direct comparisons)

Treatment estimates for pairwise meta‐analyses for each outcome are reported in Analysis 3.1 ; Analysis 3.2 ; Analysis 3.3 ; Analysis 3.4 ; Analysis 3.5 ; Analysis 3.6 ; Analysis 3.7 ; Analysis 3.8 ; Analysis 3.9 ; Analysis 3.10 .

Network meta‐analysis estimates (combination of direct and indirect comparisons) of treatment effects

The network geometry for each outcome is presented in Figure 5 and Figure 6 , and Appendix 6 ; Appendix 7 ; Appendix 8 ; Appendix 9 ; Appendix 10 ; Appendix 11 ; and Appendix 12 show the network estimates for each treatment against placebo or against another treatment within the network of each outcome.

a) New gadolinium‐enhancing positive T1‐weighted MRI lesions at 12 months were reported in six studies involving 5212 participants with RRMS (14.3% of those included in this review) ( AFFIRM 2006 ; ASSESS 2020 ; INCOMIN 2002 ; SELECT 2013 ; SUNBEAM 2019 ; TRANSFORMS 2010 ) and assessing seven treatments. Two treatments assessed in two studies were compared to placebo. Using placebo as a common comparator, treatment with natalizumab results in a large reduction of new gadolinium‐enhancing positive T1‐weighted MRI lesions (RR 0.11, 95% CI 0.07 to 0.17). A trivial increase in new gadolinium‐enhancing positive T1‐weighted MRI lesions probably resulted in people treated with daclizumab (RR 1.00, 95% CI 0.97 to 1.04). Differences across treatments were found and are shown in Appendix 6 .

b) Data on new gadolinium‐enhancing positive T1‐weighted MRI lesions at 24 months was available from 11 studies involving 7935 participants with RRMS (21% of those included in this review) ( AFFIRM 2006 ; CARE‐MS I 2012 ; CARE‐MS II 2012 ; CONFIRM 2012 ; FREEDOMS 2010 ; INCOMIN 2002 ; MAIN 2014 ; OPERA I 2017 ; OPERA II 2017 ; RADIANCE 2019 ; REGARD 2008 ) and assessing eleven treatments. Four treatments assessed in three studies were compared to placebo, seven treatments were evaluated in head‐to‐head comparisons in nine studies and one study involving 1417 participants had both placebo and active treatment arms.

Using placebo as a common comparator, natalizumab (RR 0.11, 95% CI 0.07 to 0.17) resulted in a large reduction in new gadolinium‐enhancing positive T1‐weighted MRI lesions. Treatment with ocrelizumab (RR 0.09, 95% CI 0.05 to 0.16) and alemtuzumab (RR 0.13, 95% CI 0.08 to 0.23) probably resulted in a large reduction in new gadolinium‐enhancing positive T1‐weighted MRI lesions. Treatment with interferon beta‐1b (Betaferon) (RR 0.16, 95% CI 0.08 to 0.31) may have resulted in a large reduction in new gadolinium‐enhancing positive T1‐weighted MRI lesions. Treatment with interferon beta‐1a (Avonex, Rebif) (RR 0.34, 95% CI 0.21 to 0.55), fingolimod (RR 0.29, 95% CI 0.23 to 0.38), dimethyl fumarate (RR 0.50, 95% CI 0.37 to 0.69), ozanimod (RR 0.30, 95% CI 0.18 to 0.49), and glatiramer acetate (RR 0.59, 95% CI 0.42 to 0.84) may have resulted in a moderate reduction in new gadolinium‐enhancing positive T1‐weighted MRI lesions. Differences across treatments were found and are shown in Appendix 7 .

c) No studies assessed new gadolinium‐enhancing positive T1‐weighted MRI lesions at 36 months.

d) New or enlarging T2‐weighted MRI lesions at 12 months was reported in seven studies involving 5234 participants with RRMS (14.32% of those included in this review) ( AFFIRM 2006 ; ASSESS 2020 ; Gobbi 2013 ; INCOMIN 2002 ; SELECT 2013 ; SUNBEAM 2019 ; TRANSFORMS 2010 ) and assessing seven treatments. Two treatments, assessed in two studies, were compared to placebo, and six treatments were evaluated in head‐to‐head comparisons in five studies. Using placebo as a common comparator, treatment with interferon beta‐1a (Avonex, Rebif) resulted in a large reduction of new or enlarging T2‐weighted MRI lesions (RR 0.51, 95% CI 0.45 to 0.57). Treatment with immunoglobulins may have resulted in a trivial reduction in new or enlarging T2‐weighted MRI lesions (RR 0.98, 95% CI 0.30 to 3.28). There was probably little or no effect on new or enlarging T2‐weighted MRI lesions with daclizumab (RR 1.00, 95% CI 0.97 to 1.04). A large increase in new or enlarging T2‐weighted MRI lesions may have resulted in people treated with glatiramer acetate (RR 2.37, 95% CI 0.50 to 11.30), interferon beta‐1b (RR 2.12, 95% CI 0.45 to 9.97), natalizumab (RR 2.01, 95% CI 0.43 to 9.51) and fingolimod (RR 1.89, 95% CI 0.40 to 8.99). Differences across treatments were found and are shown in Appendix 8 .

e) New or enlarging T2‐weighted MRI lesions at 24 months were reported in ten studies involving 6893 participants with RRMS (19% of those included in this review) ( AFFIRM 2006 ; CARE‐MS I 2012 ; CARE‐MS II 2012 ; FREEDOMS 2010 ; INCOMIN 2002 ; MAIN 2014 ; OPERA I 2017 ; OPERA II 2017 ; RADIANCE 2019 ; REGARD 2008 ) and assessing ten treatments. Two treatments, assessed in two studies, were compared to placebo. Using placebo as a common comparator, treatment with natalizumab (RR 0.50, 95% CI 0.45 to 0.55) resulted in a large reduction of new or enlarging T2‐weighted MRI lesions. Treatment with fingolimod (RR 0.62, 95% CI 0.56 to 0.68) may have resulted in a large reduction of new or enlarging T2‐weighted MRI lesions. Differences across treatments were found and are shown in Appendix 9 .

f) No studies assessed new or enlarging T2‐weighted MRI at 36 months.

g) Cognitive decline was available from six studies involving 4243 participants with RRMS (11.6% of those included in this review) ( ASSESS 2020 ; CombiRx 2013 ; DECIDE 2015 ; Gobbi 2013 ; Mokhber 2014 ; SUNBEAM 2019 ) and assessing seven treatments. No studies were compared to placebo, and six treatments were evaluated in head‐to‐head comparisons in seven studies. Differences across treatments were found and are shown in Appendix 10 .

h) Data on quality‐of‐life impairment were reported in studies with different scales (non‐MS related quality of health questionnaires and MS related questionnaires) and subscales (physical and mental).

Quality‐of‐life impairment (non‐MS related; total) was available from two studies involving 1061 participants with RRMS (3% of those included in this review) and assessing three treatments ( CLARITY 2010 ; Gobbi 2013 ). One study was compared to placebo. Compared to placebo, cladribine (SMD 0.19 SD, 95% CI 0.06 to 0.32) probably resulted in a trivial improvement in quality of life.

Quality of life impairment (non‐MS related; index) was available from one study involving 1042 participants with RRMS (3% of those included in this review) and assessing one treatment ( CLARITY 2010 ). Compared to placebo, cladribine probably resulted in a trivial improvement in quality of life (SMD 0.24 SD, 95% CI 0.11 to 0.37).

Quality‐of‐life impairment (non‐MS related; physical subscale) was available from five studies involving 4073 participants with RRMS (11% of those included in this review) and assessing four treatments ( CARE‐MS I 2012 ; CARE‐MS II 2012 ; OPERA I 2017 ; OPERA II 2017 ; TOWER 2014 ). Compared to placebo, teriflunomide may have resulted in a trivial improvement in quality of life (non‐MS related; physical subscale) (SMD 0.1 SD, 95% CI ‐0.02 to 0.22).

Quality of life impairment (MS related; physical) was available from six studies involving 6261 participants with RRMS (17% of those included in this review) ( ASSESS 2020 ; DECIDE 2015 ; Mokhber 2014 ; RADIANCE 2019 ; SELECT 2013 ; SUNBEAM 2019 ) and assessing six treatments. One treatment, assessed in one study, was compared to placebo, and six treatments were evaluated in head‐to‐head comparisons in five studies. Using placebo as a common comparator, treatment with daclizumab (SMD 0.22 SD, 95% CI 0.05 to 0.38) probably resulted in a trivial improvement in quality of life on this scale. Ozanimod (SMD 0.5 SD, 95% CI 0.29 to 0.71), interferon beta‐1a (Avonex, Rebif) (SMD 0.36 SD, 95% CI 0.16 to 0.55) and interferon beta‐1b (SMD 0.41 SD, 95% CI ‐0.13 to 0.95) may have resulted in a trivial improvement in quality of life on this scale. Differences across treatments were found and are shown in Appendix 11 .

Quality of life impairment (non‐MS related; mental subscale) was available from three studies involving 2417 participants with RRMS (6.61% of those included in this review) ( CARE‐MS I 2012 ; CARE‐MS II 2012 ; TOWER 2014 ) and assessing three treatments. One treatment, assessed in one study, was compared to placebo, and two treatments were evaluated in head‐to‐head comparisons in two studies. Compared to placebo, teriflunomide may have resulted in a trivial improvement in quality of life (non‐MS related; mental subscale) (SMD 0.1 SD, 95% CI ‐0.02 to 0.22).

Quality‐of‐life impairment (MS related; mental) was available from six studies involving 6261 participants with RRMS (17% of those included in this review) ( ASSESS 2020 ; DECIDE 2015 ; Mokhber 2014 ; RADIANCE 2019 ; SELECT 2013 ; SUNBEAM 2019 ) and assessing six treatments. One treatment, assessed in one study, was compared to placebo, and six treatments were evaluated in head‐to‐head comparisons in five studies. Using placebo as a common comparator, interferon beta‐1b (SMD 0.30 SD, 95% CI ‐0.24 to 0.84) may have resulted in a small improvement in quality of life (MS related; mental subscale). Treatment with daclizumab (SMD 0.12 SD, 95% CI ‐0.05 to 0.28), iozanimod (SMD 0.05 SD, 95% CI ‐0.15 to 0.26) and interferon beta‐1a(SMD 0.03 SD, 95% CI ‐0.17 to 0.22) may have resulted in a trivial improvement in quality of life (MS related; mental subscale). Differences across treatments were found and are shown in Appendix 12 .

2.2 Safety

MS‐related mortality

Pairwise meta‐analysis (direct comparisons)

Treatment estimates for pairwise meta‐analyses are reported in Analysis 3.11 .

Network meta‐analysis estimates (combination of direct and indirect comparisons) of treatment effects

The network geometry is presented in Figure 5 , and Appendix 13 shows the network estimates of each treatment against placebo or against another treatment within the network.

Data on mortality were available from 33 studies involving 34,500 participants with RRMS (94% of those included in this review) ( ADVANCE 2014 ; AFFIRM 2006 ; ALLEGRO 2012 ; ASCLEPIOS I 2020 ; ASCLEPIOS II 2020 ; ASSESS 2020 ; BEYOND 2009 ; BRAVO 2014 ; CAMMS223 2008 ; CARE‐MS I 2012 ; CARE‐MS II 2012 ; CLARITY 2010 ; CombiRx 2013 ; CONCERTO 2021 ; CONFIRM 2012 ; DEFINE 2012 ; DECIDE 2015 ; FREEDOMS 2010 ; FREEDOMS II 2014 ; GALA 2013 ; GOLDEN 2017 ; MSCRG 1996 ; PRISMS 1998 ; OPERA I 2017 ; OPERA II 2017 ; OPTIMUM 2021 ; RADIANCE 2019 ; REGARD 2008 ; SELECT 2013 ; SUNBEAM 2019 ; TOWER 2014 ; TEMSO 2011 ; TRANSFORMS 2010 ) and assessing 16 treatments.

Ten treatments assessed in 15 studies were compared to placebo, thirteen treatments were evaluated in head‐to‐head comparisons in 15 studies and two studies involving 2751 participants had both a placebo and active treatment arms. Using placebo as a common comparator, a trivial reduction in the number of deaths probably occurred in people treated with daclizumab (OR 0.32, 95% CI 0.04 to 2.37) and laquinimod (OR 0.51, 95% CI 0.12 to 2.18). A trivial reduction in the number of deaths may have occurred in people treated with ponesimod (OR 0.30, 95% CI 0.01 to 13.20), interferon beta‐1b (OR 0.37, 95% CI 0.02 to 5.81), fingolimod (OR 0.38, 95% CI 0.07 to 1.98), ocrelizumab (OR 0.39, 95% CI 0.03 to 4.50), ofatumumab (OR 0.49, 95% CI 0.01 to 24.84), glatiramer acetate (OR 0.49, 95% CI 0.10 to 2.38), pegylated interferon (OR 0.49, 95% CI 0.07 to 3.51), interferon beta‐1a (OR 0.63, 95% CI 0.18 to 2.17), and dimethyl fumerate (OR 0.75, 95% CI 0.11 to 5.24). A trivial increase in the number of deaths probably occurred in the people treated with cladribine (OR 0.98, 95% CI 0.18 to 5.39). A trivial increase in the number of deaths may have occurred in the people treated with ozanimod (OR 0.96, 95% CI 0.03 to 29.63), alemtuzumab (OR 1.38, 95% CI 0.16 to 12.14), teriflunomide (OR 1.50, 95% CI 0.16 to 14.45) and natalizumab (OR 2.52, 95% CI 0.12 to 52.69). Differences across treatments were found and are shown in Appendix 13 .

Relative treatment rankings (SUCRA and mean rank) for each primary and secondary outcome are presented in Appendix 14 . Due to the small number of studies for comparisons and large number of treatments, interpretation of these results should be considered with caution.

3. Assessment of heterogeneity and incoherence within the network analyses

We performed an assessment of heterogeneity and incoherence within the network analyses for all analyses whenever possible. The values for common heterogeneity (τ2) for the network for each outcome seem to show no evidence of heterogeneity ( Appendix 15 ). Assessment of incoherence was possible for disability worsening over 24 months, relapses over 12 and 24 months, number of patients with any serious adverse events, number of patients who discontinued due to adverse events, and mortality ( Appendix 16 ). We observed evidence of local statistical incoherence, estimated as a difference between direct and indirect treatment estimates in networks, for one loop for relapses over 24 months, for the number of patients with any serious adverse events, and for treatment discontinuation due to adverse events.

4. Subgroup and sensitivity analyses

Subgroup analysis

In the subgroup analysis by diagnostic criteria, for the efficacy outcomes, we didn't find any relevant difference in the dimension or direction of the effect for those drugs that were represented in both subgroups. For safety outcomes, we found that, in the studies using Poser criteria, the incidence of SAEs was higher for some drugs compared to studies using Mc Donald criteria; however, the estimates were highly imprecise ( Appendix 17 ).

We could not perform a subanalysis based on previous treatment with immunomodulators or immunosuppressants because the vast majority of the included studies did not report data about previous treatments with DMTs or included a mixed population of patients with and without previous treatment with DMTs, but did not report separate outcome data for the two subgroups. Moreover, amongst the studies including people with MS previously treated with DMTs, we found notable heterogeneity in the definition of “previous treatment”, with different washout criteria adopted for different DMTs.

Sensitivity analysis

In the analysis including only studies at low risk of selection bias (allocation concealment), we did not find any important difference in the direction or dimension of the effect estimate for all outcomes except "discontinuation due to adverse events". For this outcome, the direction of the effect remained the same and the dimension increased for some drugs while decreased for others. The direction of the effect changed only for pegylated interferon beta‐1a.

In the analysis, including only studies at low risk of attrition bias, the estimates of the effect of the interventions changed in the dimension of the effect, increasing for some drugs and decreasing for others, but not in the direction for all the outcomes for several drugs. For disability at 24 months, the estimate of effect of fingolimod changed direction ( Appendix 18 ).

We did not perform sensitivity analyses excluding studies with a sample size smaller than 50 randomised participants because only two studies satisfied this criterion.

5. Reporting bias

We did not produce a contour‐enhanced funnel plot for each pairwise comparison due to the low number of studies. We employed a comparison‐adjusted funnel plot for all placebo‐controlled trials for relapses over 12 and 24 months and disability worsening over 24 months. Small study effects (not necessarily due to reporting bias) appeared to be present for relapses over 12 and 24 months, but not for disability worsening over 24 months (data not shown).

Discussion

Summary of main results

This updated review of the effects of treatments for RRMS included 50 studies involving 36,541 randomised adult participants. Twenty‐five (50%) studies were placebo‐controlled and 25 (50%) were head‐to‐head comparisons. The majority of studies were short‐term trials, with the median treatment duration being 24 months. Four studies reported a 36‐month follow‐up; therefore, the effects of these treatments beyond two years remain uncertain.

1. Recurrence of relapses

Considering placebo as a common comparator, 12‐month follow‐up treatment with natalizumab resulted in a large reduction of people with relapses. Treatment with fingolimod, daclizumab, and immunoglobulins probably resulted in a large reduction of people with relapses. Treatment with mitoxantrone, teriflunomide, glatiramer acetate and pegylated interferon beta‐1a may result in a large reduction of people with relapses. Treatment with interferon beta‐1a (Avonex, Rebif) may moderately reduce people with relapses. Treatment with interferon beta‐1b may moderately reduce the number of people with relapses, but the evidence is very uncertain. When interferon beta 1‐a and beta 1‐b were both used in the same intervention, there may have been a large increase in the number of people with relapses, but the evidence is very uncertain.

At 24 months follow‐up, treatment with cladribine, alemtuzumab, and natalizumab resulted in a large decrease in people with relapses. Treatment with fingolimod, dimethyl fumarate and ponesimod probably resulted in a large decrease in people with relapses. Treatment with glatiramer acetate and interferon beta‐1a (Avonex, Rebif), probably moderately decreases people with relapses. Treatment with mitoxantrone and immunoglobulins may result in a large decrease in people with relapses. Treatment with interferon beta‐1b (Betaferon) and laquinimod may moderately decrease people with relapses. Treatment with teriflunomide and azathioprine may moderately decrease people with relapses, but the evidence is very uncertain. When interferon beta 1‐a and beta 1‐b were both used in the same intervention arm, there may have been a moderate increase in the number of people with relapses, but the evidence is very uncertain.

At 36 months follow‐up, we found that interferon beta‐1b (Betaferon) may moderately decrease the number of people with relapses, but the evidence is very uncertain.

2. Disability worsening

At 24 months follow‐up, treatment with natalizumab probably results in a moderate reduction of people with disability worsening. Treatment with dimethyl fumarate, alemtuzumab, and fingolimod may moderately reduce people with disability worsening. Treatment with cladribine and interferon beta‐1b (Betaferon) may result in a small reduction of people with disability worsening. Treatment with mitoxantrone may result in a large reduction of people who experience disability worsening, but the evidence is very uncertain. Treatment with interferon beta‐1a may result in a trivial reduction in the number of people who experience disability worsening. Treatment with mitoxantrone may result in a large reduction of people who experience disability worsening, but the evidence is very uncertain. Treatment with ofatumumab, ocrelizumab, azathioprine and ponesimod may moderately reduce the number of people who experience disability worsening, but the evidence is very uncertain. Treatment with glatiramer acetate, teriflunomide, immunoglobulins and laquinimod may result in a small reduction of people who experience disability worsening, but the evidence is very uncertain. When interferon beta 1‐a and beta 1‐b were both used in the same intervention arm, there may have been a large increase in the number of people with disability progression, but the evidence is very uncertain. When treated with ozanimod, there may have been a trivial increase in the number of people with disability progression, but the evidence is very uncertain.

At 36 months follow‐up, we did not find any study that compared our drugs of interest with placebo.

3. Safety

Using placebo as the common comparator, treatment with alemtuzumab probably results in a trivial reduction of people who discontinued treatment due to adverse events. Treatment with ocrelizumab may result in a trivial reduction in people who discontinue treatment due to adverse events. Treatment with daclizumab, fingolimod, teriflunomide, interferon beta‐1a, laquinimod, natalizumab, and glatiramer acetate probably result in a trivial increase in people who discontinued due to adverse events. Treatment with interferon beta‐1b, ofatumumab, cladribine, and dimethyl fumarate may result in a trivial increase in people who discontinued due to adverse events. Treatment with ozanimod may have little to no effect on people who discontinued due to adverse effects. Treatment with pegylated interferon beta‐1a, azathioprine, and ponesimod may result in a small increase in people who discontinued due to adverse events, but the evidence is very uncertain. Treatment with immunoglobulins and interferon beta 1a‐1b may result in a trivial increase in people who discontinued due to adverse effects, but the evidence is very uncertain.

Compared to placebo, treatment with interferon beta‐1b probably results in a trivial reduction in people who experience serious adverse effects. Treatment with fingolimod and glatiramer acetate may result in a trivial reduction in people who experience SAEs. Treatment with mitoxantrone may result in a trivial reduction in people who experience SAEs, but the evidence is very uncertain. Treatment with daclizumab may result in a small increase in people with SAEs. Treatment with dimethyl fumarate and natalizumab may result in a trivial increase in people who experience SAEs. Treatment with ocrelizumab may result in little or no effect on people who experience SAEs, but the evidence is very uncertain. Treatment with the remaining DMTS may result in a trivial increase in the people experiencing SAEs, but the evidence is very uncertain: alemtuzumab, cladribine, interferon beta‐1a, laquinimod, pegylated beta‐1a; ofatumumab, ozanimod, ponesimod and teriflunomide (very low‐certainty evidence).

Overall completeness and applicability of evidence

Nine critical outcomes were identified by the Multiple Sclerosis Essential Medicines Panel appointed by the Multiple Sclerosis International Federation. These informed the current review, but the outcomes were further differentiated into primary and secondary outcomes and assessed solely for certainty and efficacy/harm, in line with standard Cochrane methodology. The data underlying the review, from all nine outcomes, served as the evidence base for the MEMP guideline panel, where it was contextualised with the perspective of low‐resource settings by considering further evidence related to other domains, in line with GRADE Evidence to Decision Framework methodology ( Alonso‐Coello 2016 ; Alonso‐Coello 2016b ). The MEMP recommendations were used as the basis of an application for the inclusion of disease‐modifying treatments in the 23rd WHO Model List of Essential Medicines.

Many of the trials included in this review provided evidence on the proportion of participants who experienced new relapses, disability worsening, and adverse events over 12 or 24 months' follow‐up, but only five studies reported data on these outcomes over 36 months. Considering that MS is a chronic disease of 30 to 40 years' duration, such findings limit the applicability of the available evidence for both efficacy and safety, particularly for long‐term and uncommon adverse events.

Evidence relevant to 29 treatments considered in this review was derived from 50 studies involving 36,541 adult participants, half of which (25/50 studies; 17,294 participants, 50% of those included in this review) compared the intervention with placebo and not with another DMT. There is, therefore, uncertainty whether the results of the review fit into the context of current practice since about 50% of people with MS are treated with at least one DMT ( Carroll 2014 ).

The reasons why the 50 available randomised studies for RRMS are mostly placebo‐controlled and outcome data are reported mainly over 24 months, are probably due to the following reasons: i) comparison with placebo in one double‐blind, superiority RCT is sufficient for approval of DMTs for RRMS by many national regulatory agencies ; ii) the lack of interest by pharmaceutical companies in conducting longer and more expensive studies, given that it is only recently that some regulatory agencies have recommended a duration of three years for confirmatory trials ( EMA ‐ Guideline on clinical investigation of medicinal products for the treatment of Multiple Sclerosis, 2015 https://www.ema.europa.eu/documents/scientific‐guideline/guideline‐clinical‐investigation‐medicinal‐products‐treatment‐multiple‐sclerosis‐revision‐2_en.pdf) ; iii) the unlikely advantage of pharmaceutical companies in conducting head‐to‐head trials directly comparing active treatments.

In order to increase the comprehensiveness of this review, in addition to outcomes present in the previous version of the review, this update also included new/enlarging T1‐ and T2‐weighted MRI lesions, cognitive and quality of life as secondary outcomes, which are considered relevant by people with MS ( Secondary outcomes ).

Quality of the evidence

Considering the risk of bias, the most frequent concern was related to the role of the sponsor in the authorship of the study report or in data management and analysis, for which we judged that 68% of the studies were at high risk of bias. The other most frequent concerns were performance bias with 34% of the studies at high risk of bias and attrition bias with 32% of studies at high risk of bias.

The most frequent reasons for downgrading the certainty of evidence were study limitations and imprecision, across all the outcomes and comparisons.

For relapses at 12 and 24 months, we judged several treatments as having high (natalizumab, alemtuzumab, cladribine) and moderate (daclizumab, fingolimod, immunoglobulins, dimethylfumarate, glatiramer acetate, interferon beta‐1a, ponesimod) certainty of evidence. On the other hand, for the outcome disability at 24 months, we found moderate certainty of evidence only for natalizumab, while the certainty of evidence for all the other treatments was low or very low.

For treatment discontinuation due to adverse events, we judged several treatments as having moderate certainty of evidence (natalizumab, alemtuzumab, daclizumab, fingolimod, glatiramer acetate, interferon beta‐1a, laquinimod, and teriflunomide). For the outcome SAEs, only interferon beta‐1b provided a moderate certainty of evidence.

Considering all the outcomes, natalizumab was the drug for which we found the highest certainty of evidence.

Potential biases in the review process

1. Transitivity assumption

We assumed that any patient who met the inclusion criteria was, in principle, equally likely to have been randomised to any of the eligible interventions. However, as we discussed in the Background section, several participant characteristics have changed in newer trials, and thus a transitivity hypothesis may not have been a reasonable assumption to make, due to differences in patient or trial characteristics. Thus, we evaluated the assumption of transitivity by assessing differences in patient characteristics such as age, disease duration, and baseline Expanded Disability Status Scale (EDSS) scores across the trials and by comparing the predefined potential effect modifiers across the different comparisons in the networks. We did not find any evidence that important variables varied across comparisons or altered the effectiveness of the treatments; although some confounders may be hidden and unmeasured, it might be reasonable to analyse the network as a whole. Thus, we assumed that the transitivity held and a network meta‐analytical approach was reasonable. However, few studies per comparison were available and limitations in study reporting cannot exclude the possibility of intransitivity.

2. Heterogeneity and inconsistency

We did not find any strong evidence of the presence of heterogeneity either in direct pairwise comparisons or in the entire networks. Similarly, the loop‐specific approach and the 'design‐by‐treatment' model did not provide any clear indication of the presence of inconsistency either locally or in the entire networks. Thus, we believe that the consistency assumption is reasonable for this type of data. However, the power of these tests and approaches to detect inconsistency is low, particularly for networks with few included studies per comparison.

3. Subgroup and sensitivity analyses

In the subgroup analysis by diagnostic criteria, for the efficacy outcomes, we didn't find any relevant difference in the dimension or direction of the effect for those drugs that were represented in both subgroups. For safety outcomes, we found that, in the studies using Poser criteria, the incidence of SAEs was higher for some drugs compared to studies using Mc Donald criteria; however, the estimates were highly imprecise.

In sensitivity analysis including only studies at low risk of selection bias, we found that the outcome "discontinuation due to adverse effects" changed in the dimension of the effect, increasing for some treatments and decreasing for others. When we included only studies at low risk of attrition bias, this phenomenon happened also for other outcomes, though differences were not substantial. These findings indicate that attrition bias could have an impact on the dimension of the estimates, though it is not possible to identify a consistent direction across the interventions.

4. Reporting bias

The possible presence of reporting bias, partially supported by the comparison‐adjusted funnel plot for comparisons versus placebo for relapses over 12 and 24 months, could not be totally excluded.

5. Certainty of the evidence

As reported in the Data collection and analysis (Summary of findings and assessment of the certainty of the evidence) section, the certainty of the evidence for this review was assessed using a fully contextualised approach, involving the definition of quantitative thresholds to determine the magnitude ('trivial', 'small', 'moderate', and 'large') of each health effect measured by each outcome. Quantitative thresholds between magnitudes of health effects were considered when assessing imprecision, one of the domains contributing to the certainty of the evidence. Thresholds were calculated from outcome‐specific numerical health state utility values (HSUVs). Whenever HSUVs were not obtainable from published evidence, they were set through panel judgement, therefore reflecting the panel members' potentially biased views and expectations.

Agreements and disagreements with other studies or reviews

In this review, which included 50 RCTs on 29 pharmacological treatments for patients with RRMS, we found with high certainty that, compared to placebo, people with RRMS receiving natalizumab, cladribine or alemtuzumab for 24 months had a lower risk of experiencing new relapses at both 12 and 24 months compared to people receiving other DMTs. Moreover, people receiving natalizumab may also have a lower risk of disability progression at 24 months. Results from our previous review ( Tramacere 2015 ) on disability worsening at 24 months were confirmed and extended in this current review update.

In terms of safety, compared to those receiving placebo, we are moderately confident that a higher proportion of people treated with daclizumab, fingolimod, teriflunomide, interferon beta‐1a, laquinimod, natalizumab, and glatiramer or glatiramer acetate are at a higher risk of withdrawing due to AEs. Our certainty that the same conclusion could be drawn for the other DMTs except alemtuzumab (all with higher risk than placebo) was low or very low. Indirect comparisons show that alemtuzumab was the only DMT with a lower number of patients discontinuing for AEs than placebo (moderate certainty).

Daclizumab, dimethyl fumerate and natalizumab are associated with a higher proportion of severe adverse events than other DMTs, when taking placebo as the common comparator, although the certainty of such evidence is low. There is low‐certainty evidence that people treated with interferon beta‐1b, fingolimod and glatiramer acetate may experince a slightly lower number of SAEs. For the remaining DMTs, we found only very low‐certainty evidence.

According to an NMA ( Li 2020 ) including 23 RCTs on DMTs for RRMS with methodology similar to that of our review, in that the authors adopted the GRADE methodology to assess certainty in the evidence, alemtuzumab, natalizumab, ocrelizumab and fingolimod were the regimens associated with both the lowest risks of relapse rate and of treatment discontinuation due to AEs. However, in interpreting such similarities and differences in the results compared to those of our review, it has to be noted that this NMA considered only studies with a minimum follow‐up of 24 months assessing the 12 DMTs approved by the FDA, and did not include cladribine and daclizumab.

A recent NMA ( Liu 2021 ), including 21 studies involving 23 DMTs, showed that all of them, except interferon beta‐1b, were more effective than placebo in reducing the annualised relapse rate (ARR) over 24 months while, in a multi‐comparison analysis, ofatumumab ranked as the most effective, and alemtuzumab, dimethyl fumarate and ocrelizumab were ranked in second place. Considering the effectiveness in reducing relapse frequency and tolerability in terms of discontinuation rate due to AEs, alemtuzumab showed the best combination and, according to the authors, could be considered the optimal treatment, together with ofatumumab and natalizumab. In terms of undesirable effects, drug discontinuation due to AEs was more frequent with most DMTs compared to placebo, but such difference was not statistically significant, except for dimethyl fumarate. Although conclusions on alemtuzumab and natalizumab may be considered as in agreement with our review, and safety relative to discontinuation due to AEs is in disagreement, it should be noted that, in contrast to our review, in the review by Liu and colleagues, the risk of bias of the included studies was assessed by means of the Cochrane Risk of Bias 2 (RoB2) but outcome‐specific certainty of the reported estimates according to the GRADE methodology was not assessed.

Similar to our conclusions, in respect to placebo, a NMA on the safety of immunotherapies for MS ( Tramacere 2023 ) found an increased proportion of people discontinuing treatment for teriflunomide, interferon beta‐1a, interferon beta‐1b, glatiramer acetate, fingolimod, immunoglobulins, and daclizumab.

Favourable results relative to alemtuzumab and natalizumab in terms of efficacy were also reported by a NMA ( Lucchetta 2018 ), including most of the available licensed DMTs for RRMS, and azathioprine and rituximab as off‐label treatments. In contrast to our review, ocelizumab also, together with alemtuzumab and natalizumab, were reported as the most effective DMTs in reducing ARR and disability accumulation. This NMA, though, included studies with a minimum follow‐up of 12 weeks and used interferon beta‐1a as a common comparator in the NMA.

A NMA of RCTs on DMTs for RRMS ( Fogarty 2016 ), considering ARR and disability progression as primary efficacy outcomes, found that natalizumab and alemtuzumab ranked higher than other DMTs across outcomes. This review, however, did not include amongst the considered interventions most of the recent selective immunosuppressors (i.e. ponesimod, siponimod, ozanimod, ofatumumab, daclizumab), cladribine and off‐label treatments.

Authors' conclusions

Implications for practice

Conservative interpretation of these results is warranted, since most of the included treatments have been evaluated in few trials. Nevertheless, we used a comprehensive, transparent, and pragmatic approach for rating the certainty of the evidence (i.e. the GRADE approach), so the results of this review may provide guidance to clinicians and patients. According to the GRADE approach, implications for practice should be based on moderate to high certainty of evidence, since any estimate of effect based on low to very low certainty of evidence is very uncertain and further research is likely to change the estimate. The results of this updated review show that, for preventing clinical relapses in the short term (24 months), natalizumab, alemtuzumab, and cladribine may be preferable to several other treatments, based on high certainty of evidence. For preventing disability from worsening in the short term (24 months), natalizumab is superior to several other treatments based on moderate certainty of evidence. Moreover, alemtuzumab may also have a better treatment durability, being associated with less discontinuation due to AEs. The number of SAEs seems, from the available evidence, to be only slightly affected by treatment with DMTs amongst people with RRMS.

In addition to the available evidence for benefit provided above, there are two major concerns that have to be considered. First, the efficacy of all of these treatments beyond two years is uncertain and this is a relevant issue for a disease with a duration of 30 to 40 years. Second, short‐term trials provide limited safety data and do not provide useful evidence to obtain a reliable risk profile of treatments. In order to provide information on the long‐term safety of the treatments included in this review, non‐randomised studies should be considered.

Finally, more than 68% of the studies included in this review were sponsored by pharmaceutical companies and this may have influenced their results.

Implications for research

The research agenda on DMTs for MS should address the following needs: first, randomised trials of direct comparisons between active agents would be useful, avoiding further placebo‐controlled studies that in the current context do not comply with the principle of clinical equipoise for RRMS, given the broad availability of therapeutic alternatives. Second, follow‐up of at least 36 months should be mandatory for confirmatory RCTs on DMTs for MS. Third, more studies are needed to evaluate the medium and long‐term benefit and safety of immunotherapies and the comparative safety of the different agents. As the number of drugs, including biological agents, that are available for the treatment of RRMS increases, more options will become available to participants and clinicians. In the absence of comparative trials, national and international registries and other types of large non‐randomised studies might be relevant sources for providing complementary data regarding the long‐term benefit and safety of immunotherapies for RRMS. Fourth, future clinical research on MS may benefit from long‐term data on outcomes deemed as relevant by people with MS, such as cognitive status and quality of life, as well as definition and validation of health state utility values. Finally, methods of outcome assessment should be more consistent across studies. Although relapses and disability progression are commonly used in MS research as efficacy outcomes, comparison of the relative effectiveness and safety across DMTs is challenging due to the heterogeneity in the way such outcomes are measured (e.g. annualised relapse rate rather than number of people with relapse; mean EDSS scores rather than number of people with MS presenting disability progression). Therefore, consistency in the choice of recommended efficacy and safety outcome measures is highly desirable, particularly in industry‐initiated trials, in which the findings may inform regulatory approval of novel DMTs.

What's new

History

Protocol first published: Issue 11, 2014 Review first published: Issue 9, 2015

Acknowledgements

We thank Chiara Bassi, Information Specialist for Cochrane Multiple Sclerosis and Rare Diseases of the CNS Group, and Robin Featherstone, Information Specialist, former member of the Cochrane Central Executive Team, who developed the search strategy for this review.

This work was funded in part by grants from the MS International Federation. The MS International Federation does not endorse any publications arising from these grants unless stated explicitly.

The publication of this article was supported by the "Ricerca Corrente" funding from the Italian Ministry of Health.

We would also like to extend our thanks to Deanna Saylor, Department of Neurology, Johns Hopkins University School of Medicine, for their methodological and clinical input.

We thank all the members of the MS Essential Medicines Panel (MEMP) for providing valuable input during the formulation of the research question and the selection and prioritisation of outcomes.

Editorial and peer‐review contributions

Cochrane Multiple Sclerosis and Rare Diseases of the CNS supported the authors in the development of this review update. Ben Ridley, Graziella Filippini, Elisa Baldin, and Francesco Nonino are members of Cochrane Multiple Sclerosis and Rare Diseases of the CNS but were not involved in the editorial process or decision‐making for this update. The following people conducted the editorial process for this update:

Sign‐off Editor (final editorial decision)*: Toby Lasserson, Cochrane Evidence Production & Methods Directorate

Managing Editor (selected peer reviewers, collated peer‐reviewer comments, provided editorial comments/guidance to authors, edited the article): Joey Kwong, Cochrane Central Editorial Service

Editorial Assistant (conducted editorial policy checks, collated peer‐reviewer comments, supported editorial team): Lisa Wydrzynski, Cochrane Central Editorial Service

Copy Editor (copy‐editing and production): Anne Lethaby, c/o Cochrane Central Production Service

Peer‐reviewers (provided comments and recommended an editorial decision): Iván Pérez‐Neri, National Institute of Neurology and Neurosurgery, Manuel Velasco Suárez (consumer review), Jennifer Hilgart, Cochrane Evidence Production & Methods Directorate (methods review), Jo Platt, Cochrane GNOC (*closed in March 2023) (search review). Two additional peer reviewers provided clinical/content peer review but chose not to be publicly acknowledged.

*Robert Boyle, Imperial College London, acted as Sign‐off Editor in May 2023 and submitted the first editorial decision post‐peer review.

Appendices

Appendix 1. Search strategy ‐ CENTRAL

Appendix 2. Search strategy ‐ MEDLINE (PubMed)

Appendix 3. Search strategy ‐ Embase

Appendix 4. Search strategy ‐ Clinical trial registers

World Health Organisation (WHO) International Clinical Trials Registry Platform (ICTRP) (apps.who.int/trialsearch)

Search terms: relapsing multiple sclerosis, filtered for "Phase 2" "Phase 3" trials.

US National Institutes of Health clinical trial register (www.clinicaltrials.gov)

Search term: "relapsing multiple sclerosis".

Appendix 5. State utility values and effect thresholds for the primary outcomes in relapsing‐remitting multiple sclerosis

Appendix 6. Netleague table: T1 lesions (12 months)

Significant results are bolded and underlined

Appendix 7. Netleague table: T1 lesions (24 months)

Significant results are bolded and underlined

Appendix 8. Netleague table: T2 lesions (12 months)

Significant results are bolded and underlined

Appendix 9. Netleague table: T2 lesions (24 months)

Significant results are bolded and underlined

Appendix 10. Netleague table: Cognitive decline

Significant results are bolded and underlined

Appendix 11. Netleague Table: MS‐related ‐ QoL Physical

Significant results are bolded and underlined

Appendix 12. Netleague table: MS‐related ‐ QoL mental

Significant results are bolded and underlined

Appendix 13. Netleague table: Mortality

Significant results are bolded and underlined

Appendix 14. Relative treatment ranking (SUCRA and Mean Rank)

Appendix 15. Heterogeneity results within the network analyses

Appendix 16. Incoherence results within the network analyses

Relapses at 12 months

Loop‐specific heterogeneity approach

Node splitting approach

* Warning: all the evidence about these contrasts comes from the trials which directly compare them.

Global test: 'design‐by‐treatment' approach

chi2 (3) = 1.49

Prob > chi2 = 0.6847

Relapses at 24 months

Loop‐specific heterogeneity approach

Node splitting approach

Global test: 'design‐by‐treatment' approach

chi2 (7) = 9.66

Prob > chi2 = 0.2086

Disability at 24 months

Loop‐specific heterogeneity approach

Node splitting approach

Global test: 'design‐by‐treatment' approach

chi2 (7) = 7.65

Prob > chi2 = 0.3648

Number of patients who discontinued treatment due to adverse

Loop‐specific heterogeneity approach

Node‐splitting approach

Global test: 'design‐by‐treatment' approach

chi2 (12) = 17.79

Prob > chi2 = 0.1223

Number of patients with any serious adverse events

Loop‐specific heterogeneity approach

Node‐splitting approach

Global test: 'design‐by‐treatment' approach

chi2 (9) = 17.10

Prob > chi2 = 0.0472

T1 at 24 months

Not available because the loop is from a three‐arm trial

Mortality

Loop‐specific heterogeneity approach

Node‐splitting approach

Global test: 'design‐by‐treatment' approach

chi2 (7) = 5.94

Prob > chi2 = 0.5465

Appendix 17. Subgroup analyses

Relapse at 12 months

Mc Donald criteria

Poser criteria

Relapse at 24 months

Mc Donald criteria

Poser criteria

Disability at 24 months

Mc Donald criteria

Poser criteria