Scolaris Content Display Scolaris Content Display

Cochrane Database of Systematic Reviews Protocol - Intervention

Immunomodulators and immunosuppressants for relapsing‐remitting multiple sclerosis: a network meta‐analysis

This is not the most recent version

Collapse all Expand all

Abstract

This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:

We aim to compare the efficacy and acceptability of immunomodulators and immunosuppressants to treat participants with RRMS and to generate a clinically useful hierarchy of available immunotherapies according to their efficacy and acceptability.

Background

Description of the condition

Multiple sclerosis (MS) is an inflammatory disorder of the brain and spinal cord resulting from interaction between unidentified environmental factors and susceptibility genes. Several pathological processes occur in MS involving the immune system, T‐cell‐mediated and B‐cell‐mediated mechanisms, demyelination, remyelination, microglial activation, and chronic neurodegeneration (Bennett 2009; Compston 2008). The sequential involvement of these processes influences the clinical course, which is characterized by attacks of neurological dysfunction with recovery, attacks leaving persistent deficits, and progression that causes permanent physical and cognitive disability. MS is among the most common causes of neurological disability in young people, with an annual incidence ranging from 2 to 10 cases per 100,000 persons per year and a north‐south gradient, lower incidence being closer to the equator. Its clinical manifestations typically occur between 20 and 40 years of age, with symptoms and signs involving different regions of the central nervous system: optic nerve, brainstem, cerebellum, cerebral hemispheres, spinal cord.

MS has a chronic course that evolves over 30 to 40 years. The clinical phenotypes include relapsing‐remitting MS (RRMS), secondary‐progressive MS (SPMS), primary‐progressive MS (PPMS), and progressive‐relapsing MS (PRMS) (Lublin 1996). The development of progression after a relapsing‐remitting course is responsible for permanent long‐term disability; it supervenes in about 80% of RRMS patients by 20 to 25 years from disease onset (Kremenchutzky 2006). Times to need assistance to walk, be confined to bed, or have died were 14, 24, and 45 median years from disease onset and 3, 12, and 30 median years from onset of secondary progression, respectively (Scalfari 2014).

Male sex, older age at onset, and high early relapse frequency (more than three attacks during the first three years) predicted significantly higher risk of conversion to SPMS and shorter latency to progression (Scalfari 2014). In RRMS patients, the onset of secondary progression is the determinant of long‐term prognosis, and its prevention is the key therapeutic goal.

According to the older Poser criteria (Poser 1983), MS can be clinically diagnosed by demonstrating two separate clinical attacks (dissemination in time) involving at least two different areas of the central nervous system (dissemination in space). The 2001 McDonald criteria and their 2005 and 2010 revisions incorporate magnetic resonance imaging (MRI) criteria for dissemination in space and time, allowing a MS diagnosis at the time of first symptoms (McDonald 2001; Polman 2005; Polman 2011). Dissemination in space is demonstrated by greater than or equal to oneT2 lesion in at least two MS typical central nervous system regions (periventricular, juxtacortical, infratentorial, spinal cord). Dissemination in time is demonstrated by: (i) simultaneous asymptomatic contrast‐enhancing and non‐enhancing lesions at any time; or (ii) a new T2 and/or contrast‐enhancing lesions(s) on follow‐up MRI, irrespective of its timing. The diagnostic criteria include exclusion of other possible diagnoses.

A declining trend in on‐study relapse rate (one of the most commonly used primary outcomes in MS trials) of placebo participants in trials has been observed (Inusah 2010; Nicholas 2012; Steinvorth 2013; Stellmann 2012). This decline is thought to result from decreasing pretrial relapse rates and a shorter time period over which pretrial relapse rates were calculated in newer trials (Steinvorth 2013; Stellmann 2012). Pre‐study relapse rate was found to be the best predictor for on‐study relapse rate. Other participant characteristics have changed in newer trials. Participants were older and with longer disease duration, whereas Expanded Disability Status Scale (EDSS) scores remained relatively stable. The introduction of the new McDonald criteria led to inclusion in newer trials of participants who had had earlier diagnosis and were later in their disease course, which was less severe compared to patients in older studies (Steinvorth 2013). These changes may explain the decrease in pretrial relapse rate and the associated decrease in on‐trial relapse rate. Unwelcome consequences of the expected decreased event rate were that the sample size of newer trials has been inflated and follow‐up periods shortened.

Another difference between older and newer studies is that the latter can have included participants who had made prior use of immunomodulators or immunosuppressants.

Description of the intervention

Several treatments are available for patients with RRMS. For this review we will consider all immunomodulators and immunosuppressants that, up to 30 April 2014, have been studied in patients with RRMS in randomised clinical trials (RCTs) with more than six months' follow‐up.

Interferon beta‐1b (EMEA 2002; FDA 1993), interferon beta‐1a (Rebif) (EMEA 1998; FDA 2002), interferon beta‐1a (Avonex) (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. Five to 30 percent of treated patients develop persistent neutralising antibodies, usually in the first year of treatment and more commonly in those receiving interferon beta‐1b. The presence of persistent neutralising antibodies is associated with reduction in the treatment effect on relapse activity (Sorensen 2003).

Natalizumab was initially approved by the U.S. 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 patients 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 patients with SPMS, PRMS or worsening RRMS" (FDA 2000). Safety issues of concern for patients treated with mitoxantrone are cardiotoxicity and acute leukaemia.

Fingolimod was the first oral treatment approved for RRMS patients to reduce the frequency of clinical exacerbations and to 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, and considering that in some patients the nadir of heart frequency can be observed up to 24 hours after the first dose.

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

Dimethyl fumarate has been approved as a first‐line oral treatment for RRMS patients (EMA 2014a; FDA 2013a). 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 by the EMA for treatment of RRMS patients with active disease defined by clinical or imaging features (EMA 2013b). The FDA rejected this drug due to concerns of serious risks that may outweigh its benefits, including the risk of autoimmune diseases, malignancies, serious infections, and infusion reactions (FDA 2013b). 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.

Pegylated interferon beta‐1a is at a relatively advanced stage of drug development, with results from clinical trials available. It is administered by subcutaneous or intramuscular injection.

Daclizumab is currently being investigated in clinical trials for RRMS, but it has not yet been approved for MS by regulatory agencies. It is administered by subcutaneous or intravenous injections. Risks of serious infections and autoimmune diseases are increased with daclizumab.

Ocrelizumab is in development for treatment of RRMS, with two active phase clinical trials ongoing. It is administered by intravenous injections.

Laquinimod is an immunomodulator that is currently under evaluation for the treatment of RRMS. It is taken orally as a 0.6 mg tablet once daily. 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 2014b). Further studies of laquinimod as a monotherapy and an add‐on therapy in RRMS patients are ongoing.

Azathioprine has been used for the treatment of MS in many countries on the basis of placebo‐controlled RCTs published more than two decades ago. However, since the approval of interferons beta, azathioprine has been dismissed as toxic and insufficiently effective (Casetta 2007). It is taken orally as a 2 or 3 mg/kg tablet daily.

Intravenous immunoglobulins may have a role for patients 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 341 infusions under routine clinical conditions (Elovaara 2008).

How the intervention might work

Immunosuppressive or immunomodulatory effects are common to all treatments included in the review.

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).

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).

Mitoxantrone is an anthracenedione cytotoxic drug that intercalates with DNA and inhibits both DNA and RNA synthesis, thus reducing the number of lymphocytes (Fox 2004).

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).

Teriflunomide is an inhibitor of dihydroorotate dehydrogenase (DHODH), a mitochondrial enzyme involved in new pyrimidine synthesis for DNA replication. Consequently, the drug reduces T‐ and B‐lymphocytes 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).

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 defense against the cytotoxic effects of oxidative stress. It is an anti‐inflammatory that promotes anti‐inflammatory activity and can inhibit expression of pro‐inflammatory cytokines and adhesion molecules (Wilms 2010).

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).

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 the intramuscular interferon beta‐1a (Hu 2012).

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).

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).

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).

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).

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).

Why it is important to do this review

Although there is consensus that immunotherapies reduce the frequency of relapses in MS, their relative effectiveness in delaying new attacks or disability progression 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. A summary of the results, including both direct and indirect comparisons, may help to clarify the stated uncertainty (Caldwell 2005; Glenny 2005).

Objectives

We aim to compare the efficacy and acceptability of immunomodulators and immunosuppressants to treat participants with RRMS and to generate a clinically useful hierarchy of available immunotherapies according to their efficacy and acceptability.

Methods

Criteria for considering studies for this review

Types of studies

We will evaluate for inclusion RCTs that studied one or more of the agents for use in RRMS that were compared to placebo or to another active agent. We will also include trials for which it is unclear whether the method of randomisation provided adequate allocation concealment or open‐label studies, but the quality of these studies will be taken into account. We will exclude RCTs with follow‐up less than or equal to six months because we will include selected time points (12, 24, and 36 months) of outcome measurement. We will exclude quasi‐randomised trials and non‐randomised studies.

Types of participants

We will include participants 18 years of age or older with a diagnosis of RRMS. Only RCTs adopting Poser (Poser 1983) or McDonald diagnostic criteria (McDonald 2001; Polman 2005; Polman 2011) will be selected. All patients regardless of sex, degree of disability, and duration of the disease will be included.

Types of interventions

We will include all immunomodulators or immunodepressants (even if they are not licensed in any country). We will exclude: (i) combination treatments; (ii) trials in which a drug regimen is compared with 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 will include RCTs that evaluate one or more of the following pharmacological interventions as monotherapy, compared to placebo or to another active agent:

  • interferon beta‐1b

  • interferon beta‐1a (Rebif, Avonex)

  • glatiramer acetate

  • natalizumab

  • mitoxantrone

  • fingolimod

  • teriflunomide

  • dimethyl fumarate

  • alemtuzumab

  • pegylated interferon beta‐1a

  • daclizumab

  • ocrelizumab

  • laquinimod

  • azathioprine

  • immunoglobulins

Regimens as defined in primary study will be included irrespective of their dose. If we identify in the included studies any immunomodulators or immunodepressants of which we were not aware, we will consider them as eligible and will include them in the network of treatments after assessing their comparability with the pre‐specified set of competing interventions. We will report the findings for these interventions in the results and the conclusions of the review. Figure 1 shows the network of all possible pairwise comparisons between the eligible interventions. We assume that any patient who meets the inclusion criteria is, in principle, equally likely to be randomised to any of the eligible interventions.


Network of all possible pairwise comparisons of treatments and placebo. The placebo node is white as it is expected without any activity. Interferons include interferon beta‐1a (Avonex and Rebif) and interferon beta‐1b.

Network of all possible pairwise comparisons of treatments and placebo. The placebo node is white as it is expected without any activity. Interferons include interferon beta‐1a (Avonex and Rebif) and interferon beta‐1b.

Types of outcome measures

Primary outcomes

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

Efficacy

1. 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. Relapse will be defined as reported in the primary study.

2. Disability progression: proportion of participants who experienced disability progression over 24 or 36 months after randomisation or at the end of the study. Progression is defined as at least 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 six months' interval free of attacks (Kurtzke 1983). Disability progression measured after only three months’ follow‐up is considered a surrogate marker for unremitting disability. The 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 progression in clinical trials.

Acceptability

Treatment discontinuation due to adverse events (AEs) will be used to assess acceptability and will be measured by the number of participants who withdrew due to any AE throughout the study out of the total number of participants randomly assigned to each treatment arm.

Secondary outcomes

The total number of serious adverse events (SAEs). If the total number of SAEs was not presented, we will take the number of participants with any (greater or equal to 1) SAEs in a study (as defined in the study) for this outcome.

Search methods for identification of studies

We will search all possible comparisons formed by the interventions of interest. No language restrictions will be applied to the search.

Electronic searches

The Trials Search Co‐ordinator will search the Cochrane Multiple Sclerosis and Rare Diseases of the Central Nervous System Group Trials Register which, among other sources, contains the Cochrane Central Register of Controlled Trials (CENTRAL recent issue), MEDLINE (1966 to date) , EMBASE (1974 to date), CINAHL (1981 to date), LILACS (1982 to date), as well as clinical trials registries (http://clinicaltrials.gov/) and the World Health Organization International Clinical Trials Registry Portal (http://apps.who.int/trialsearch/).

We will also search FDA reports on all the treatments included in this review (www.fda.gov).

We will redo the search if we identify in the included studies any immunomodulator or immunodepressant not in the present list of interventions.

Information on the Trials Register or the Review Group and details of the search strategies used to identify trials can be found in the 'Specialised Register' section within the Cochrane Multiple Sclerosis and Rare Diseases of the Central Nervous System Group module.

The keywords used to search for trials for this review are listed in (Appendix 1).

Searching other resources

1. Principal authors of the included trials will be contacted for additional information.

2. Handsearching of the references quoted in the identified trials, symposia reports (1990 to 2014) from the most important neurological associations and MS societies in Europe and America.

3. Contact with relevant pharmaceutical companies to identify unpublished trials or data missing from articles.

Data collection and analysis

Selection of studies

We will use the search strategy described above to obtain titles and abstracts of studies that may be relevant to the review. Two review authors (IT, IP) will independently screen the titles and abstracts and discard studies that are not applicable; however, studies and reviews that might include relevant data or information on trials will be retained initially. Two review authors (IT, IP) will independently assess retrieved abstracts and, if necessary, the full text of these studies to determine which studies satisfy the inclusion criteria. Multiple reports of the same study will be compared, and the most comprehensive report will be used. Multiple publications will then be linked together as companion reports, but true duplicates will be excluded. Discrepancies in judgement will be resolved by discussion with a third author (GF).

Data extraction and management

Two review authors (IT, IP) will independently extract data using a predefined data extraction form in an Excel spreadsheet. Disagreements will be resolved by discussion with a third review author (GF).

Outcome data

We will extract from each included study the number of participants who:

  1. had relapses or disability progression at 12, 24, and 36 months;

  2. were withdrawn for any AE at 12, 24, and 36 months;

  3. dropped out at each time point;

  4. were randomised.

We will extract the authors' definition of relapses and disability progression and the rate of SAE and/or the proportion of participants with at least one SAE. Arm‐level data will be extracted when possible. If arm‐level data are not available we will extract effect sizes.

If outcomes are not reported at the predefined time points, we will extract data as close as possible to that time point. When numbers of dropouts are not reported or unclear in the primary studies, we will consult reports from the FDA or ask the trial author to supply data. We will state for each included study and for each outcome the accuracy of reported numbers of dropouts, i.e. identifying studies that provided (or did not provide) complete data to enable a likely scenario analysis to be done.

Data on potential effect modifiers

We will extract from each included study data that may act as effect modifiers:

  1. Population: diagnostic criteria (Poser or McDonald criteria), baseline mean age, prior immunomodulator or immunosuppressant treatments (yes, no), definition of relapse, pre‐trial relapse rate and number of years over which the pretrial relapse rate was calculated;

  2. Intervention: administration route, dose, frequency or duration of treatment;

  3. Risk of bias: allocation concealment, blinding of outcome assessors, incomplete outcome data;

  4. Funding source.

Other data

We will extract from each included study data on the following additional information:

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

  2. Intervention: drug, dose, route;

  3. Comparison: active drug or placebo, dose, route;

  4. Study design: inclusion criteria, duration of follow‐up (12, 24, or 36 months), sequence generation, blinding of participants, selective outcome reporting, early termination of trial.

Assessment of risk of bias in included studies

We will assess the risk of bias (RoB) of each included study using The Cochrane Collaboration criteria (Higgins 2011). These include: random sequence generation, allocation concealment, blinding of participants, blinding of outcome assessor, incomplete outcome data, and selective outcome reporting. Other potential RoB includes the role of the sponsor. The RoB of each study will be explicitly judged on each criterion classified as at 'low', 'high', or 'unclear' risk of bias. Complete outcome data will be judged at low risk of bias when numbers and causes of dropouts were balanced (i.e. in the absence of a significant difference) between arms. We will assess selective outcome reporting bias by comparing outcomes intended to be analysed using published study protocol along with published study results. If study protocol is not available, we will assign high risk of bias when the study results do not include the two primary outcomes, i.e. relapse and disability worsening.

To summarize the quality of the evidence we will consider allocation concealment, blinding of outcome assessor, and incomplete outcome data in order to classify each study as at: low risk of bias when all of the three criteria are judged as at low risk of bias; high risk of bias when at least one criterion is judged as at high risk of bias; unclear risk of bias when all of the three criteria are judged as at unclear risk of bias; and moderate risk of bias in the remaining cases. Allocation concealment, blinding of outcome assessor, and incomplete outcome data are not expected to vary in importance across the two primary efficacy outcomes (relapses and progression), and therefore we will summarize RoB of each study considering the two outcomes together.

We will assess RoB for AEs considering specific factors that may have a large influence on AE data. We will evaluate methods of monitoring and detecting AE in each primary study: Did the researchers actively monitor for AEs (low risk of bias), or did they simply provide spontaneous reporting of AEs that arose (high risk of bias)? Did the authors define AEs according to an accepted international classification and report the number of SAEs? We will report RoB for AEs in an additional table called 'Assessment of Adverse Events Monitoring'.

The RoB of each study will be assessed independently by the three review authors (IT, IP, and GF), and any disagreement will be resolved by discussion to reach consensus.

Measures of treatment effect

Relative treatment effects

We will estimate, through pairwise meta‐analysis, the treatment effects of the competing interventions using risk ratio (RR) with 95% confidence intervals (95% CI) for each outcome at each time point. We will present results from network meta‐analysis (NMA) as summary relative effect sizes (RR) for each possible pair of treatments.

Relative treatment ranking

We will also estimate the ranking probabilities for all treatments of being at each possible rank for each intervention. We will then obtain a treatment hierarchy using the surface under the cumulative ranking curve (SUCRA) and mean ranks. SUCRA can also be expressed as a percentage interpreted as the percentage of efficacy/safety of a treatment that would be ranked first without uncertainty (Salanti 2011).

Unit of analysis issues

Cluster and cross‐over trials have not been done for evaluating MS treatments.

We will perform separate analyses for participants who had relapses at 12, 24, and 36 months and disability progression at 24 and 36 months.

Studies with multiple treatment groups

For multi‐arm trials, the intervention groups of relevance will be all those that could be included in a pairwise comparison of intervention groups that, if investigated alone, would meet the criteria for including studies in the review. For example, if we identify a study comparing ‘interferon beta versus natalizumab versus interferon beta plus natalizumab’, only one comparison (‘interferon beta versus natalizumab’) addresses the review objective, and no comparison involving combination therapy does. Thus, the 'interferon beta plus natalizumab' therapy group is not relevant to the review. However, if the study had compared ‘interferon beta‐1b versus interferon beta‐1a (Rebif) versus interferon beta‐1a (Avonex)', all three pairwise comparisons of interventions are relevant to the review. In this case we will treat the multi‐arm studies as multiple independent two‐arm studies in pairwise meta‐analysis; we will account for the correlation between the effect sizes from multi‐arm studies in NMA. Multi‐arm trials involving the same agent at different doses within the therapeutic range will be converted into a single arm by merging of doses and summing the number of events and the sample size.

Dealing with missing data

In order to assess the effect of missing outcome data, we will analyse data according to a likely scenario, i.e. we will assume that treated‐ and control‐group participants who contributed to missing data both had had an unfavourable outcome (relapse or disability progression).

Assessment of heterogeneity

Assessment of clinical and methodological heterogeneity within treatment comparisons

To evaluate the presence of heterogeneity deriving from different trial designs or different clinical characteristics of study participants, we will generate descriptive statistics for trial and study population characteristics across all eligible trials that compare each pair of interventions. We will assess the presence of clinical heterogeneity within each pairwise comparison by comparing these characteristics.

Assessment of transitivity across treatment comparisons

We expect that the transitivity assumption will hold assuming that all pairwise comparisons do not differ with respect to the distribution of effect modifiers (for example, the pre‐study relapse rates of participants in interferons versus placebo trials are similar to daclizumab versus placebo trials).

We will evaluate the assumption of transitivity by comparing the clinical and methodological characteristics (potential effect modifiers presented in the ‘Data extraction and management’ section) across the different pairwise comparisons.

Assessment of reporting biases

We will evaluate the possibility of reporting bias by means of contour‐enhanced funnel plots if enough studies per comparisons are available (Peters 2008). Contour‐enhanced funnel plots show areas of statistical significance, and they can help in distinguishing publication 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 will use 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 plot it against the study's standard error (Chaimani 2012, Chaimani 2013). We will draw funnel plots for all interventions versus placebo.

Data synthesis

Methods for direct treatment comparisons

We will perform conventional pairwise meta‐analyses for each primary outcome using a random‐effects model in Stata 13 for every treatment comparison with at least two studies (DerSimonian 1986).

Two different measures and meta‐analyses will be conducted for SAEs. One will consider the rate of SAEs and the other the proportion of patients with at least one SAE.

Methods for indirect and mixed comparisons

We will perform NMA for primary outcomes (relapses, progression, and acceptability). NMA is a method of synthesising information from a network of trials addressing the same question but involving different interventions. NMA combines direct and indirect evidence across a network of randomised trials into a single effect size, and under certain assumptions it can increase the precision of the estimates while respecting randomisation. We will perform NMA using a random‐effects model within a frequentist setting assuming equal heterogeneity across all comparisons, and we will account for correlations induced by multi‐arm studies (Lu 2006; Salanti 2009). The models will enable us to estimate the probability for each intervention to be the best for each outcome, given the relative effect sizes as estimated in NMA. We will perform NMA 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
Assumptions when estimating the heterogeneity

As we expect to have few studies (around two to four) in each direct comparison, in standard pairwise meta‐analysis we will assume a common heterogeneity variance for all direct comparisons. In NMA we will assume a common estimate for the heterogeneity variance across the different comparisons.

Measures and tests for heterogeneity

We will statistically assess the presence of heterogeneity for all direct pairwise comparisons using the common τ2.

The assessment of statistical heterogeneity in the entire network will be based on the magnitude of the heterogeneity variance parameter (τ2) estimated from the NMA models. We will compare the magnitude of the heterogeneity variance with the empirical distribution as derived by Turner (Turner 2012). We will also estimate a total I2 value for heterogeneity in the network as described elsewhere (Jackson 2014).

Assessment of statistical inconsistency

Consistency in a network of treatments refers to the agreement between direct and indirect evidence on the same comparisons. Joint analysis can be misleading if the network is substantially inconsistent. Inconsistency can be present if the trials in the network have very different protocols and their inclusion/exclusion criteria are not comparable or may result as an uneven distribution of the effect modifiers across groups of trials that compare different treatments.

Local approaches for evaluating inconsistency

To evaluate the presence of inconsistency locally we will use the loop‐specific approach. This method evaluates the consistency assumption in each closed loop of the network separately as the difference between direct and indirect estimates for a specific comparison in the loop (inconsistency factor) (Veroniki 2013). The magnitude of the inconsistency factors and their 95% CIs can then be used to infer about the presence of inconsistency in each loop. We will assume a common heterogeneity estimate within each loop. We will present the results of this approach graphically in a forest plot using the 'ifplot' command in Stata (Chaimani 2013).

Global approaches for evaluating inconsistency

We will use 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 will infer about the presence of inconsistency from any source in the entire network based on a chi2 test. We will perform the design‐by‐treatment model in Stata using the 'mvmeta' command. Inconsistency and heterogeneity are interwoven; to distinguish between these two sources of variability we will employ the I2 for inconsistency, which measures the percentage of variability that cannot be attributed to random error or heterogeneity (within comparison variability) (Jackson 2014).

Investigation of heterogeneity and inconsistency

If sufficient studies are available, we will perform network metaregression or subgroup analyses, or both, for the efficacy outcomes at each time point by using the following effect modifiers as possible sources of inconsistency or heterogeneity, or both:

  1. diagnostic criteria (Poser or McDonald criteria);

  2. previous treatment with immunomodulators or immunosuppressants (yes or no);

  3. definition of relapse (24hour definition or 48hour definition);

  4. pre‐trial relapse rate and number of years over which the pre‐trial relapse rate was calculated (1 or grater than 1 during the year before randomisation, 1 or grater than 1 during the 2 years before randomisation, 2 or grater than 2 during the two/three years before randomisation);

  5. treatment's administration route (oral, subcutaneous, intravenous).

Sensitivity analysis

If enough studies per comparison are identified, we plan to undertake a sensitivity analysis including only trials of low risk of bias. We will perform a sensitivity analysis excluding studies that did not provide complete data in order to enable a likely scenario analysis to be done (see 'Data extraction and management' section). We will also perform a sensitivity analysis excluding trials with a total sample size of less than 50 randomised patients to detect potential small‐study effects.

Summary of findings table

The main results of the review will be presented in 'Summary of findings' (SoF) tables, as recommended by The Cochrane Collaboration (Schünemann 2011). We will make the SoF tables for estimates from the network meta‐analysis based on the methodology developed from the GRADE Working Group (GRADE Working Group 2004). For more details, see Salanti 2014 (Salanti 2014). We will include an overall grading of the evidence for the following major outcomes:

  • proportion of patients who experienced new relapses over 12, 24, and 36 months;

  • proportion of patients who experienced disability progression over 24 and 36 months.

We will grade quality of evidence considering study limitations, indirectness, inconsistency, imprecision of effect estimates, and risk of publication bias. Since a likely scenario, accounting for incomplete outcome data, was chosen for the overall analyses, the grading of the evidence related to the study limitations will be based on allocation concealment and blinding of outcome assessor only, and not on incomplete outcome data. According to the software GRADEpro 2008, we will assign four levels of quality of evidence: high, moderate, low, very low. In the SoF tables, the control event rates that will be used in the calculation of absolute risks will be based on the number of events in the included studies.

Network of all possible pairwise comparisons of treatments and placebo. The placebo node is white as it is expected without any activity. Interferons include interferon beta‐1a (Avonex and Rebif) and interferon beta‐1b.
Figures and Tables -
Figure 1

Network of all possible pairwise comparisons of treatments and placebo. The placebo node is white as it is expected without any activity. Interferons include interferon beta‐1a (Avonex and Rebif) and interferon beta‐1b.