and Related Disorders 10 (2016) 204–212

Contents lists available at ScienceDirect

Multiple Sclerosis and Related Disorders

journal homepage: www.elsevier.com/locate/msard

Comparing outcomes from clinical studies of oral disease-modifying therapies (dimethyl fumarate, fingolimod, and teriflunomide) in relapsingcrossmark MS: Assessing absolute differences using a number needed to treat analysis ⁎ Mark S. Freedmana, , Xavier Montalbanb, Aaron E. Millerc, Catherine Dive-Poulettyd, Steven Hasse, Karthinathan Thangavelue, Thomas P. Leistf a University of Ottawa and the Ottawa Hospital Research Institute, The Ottawa Hospital – General Campus, 501 Smyth Road, Ottawa, ON, Canada, K1H 8L6 b Department of Neurology-Neuroimmunology, Cemcat, Vall d’Hebron University Hospital, P. de la Vall d'Hebron, 119-129, Barcelona 08035, Spain c Icahn School of Medicine at Mount Sinai, The Corinne Goldsmith Dickinson Center for Multiple Sclerosis, 5 East 98th Street, New York 10029, USA d Sanofi Genzyme, 1 Avenue Pierre Brossolette, Chilly-Mazarin 91385, France e Sanofi Genzyme, 500 Kendall Street, Cambridge, MA 02142, USA f Comprehensive Multiple Sclerosis Center, Thomas Jefferson University Hospital, 900 Walnut Street, Ste. 200, Philadelphia, PA 19107, USA

ARTICLE INFO ABSTRACT

Keywords: Dimethyl fumarate (DMF), fingolimod, and teriflunomide are oral disease-modifying therapies (DMTs) Multiple sclerosis indicated for the treatment of relapsing-remitting multiple sclerosis. Despite well-established limitations of Dimethyl fumarate cross-trial comparisons, DMTs are still frequently compared in terms of relative reductions in specific endpoints, most commonly annualized relapse rate. Consideration of absolute risk reduction and number Teriflunomide needed to treat (NNT) provides an alternative approach to assess the magnitude of treatment effect and can Absolute risk provide valuable additional information on therapeutic gain. Using data from pivotal studies of DMF (DEFINE, Number needed to treat NCT00420212; CONFIRM, NCT00451451), fingolimod (FREEDOMS, NCT00289978; FREEDOMS II, NCT00355134), and teriflunomide (TEMSO, NCT00134563; TOWER, NCT00751881), we calculated NNTs to prevent any relapse, more severe relapses (such as those leading to hospitalization or requiring intravenous corticosteroids), and disability worsening. Higher relative reductions were reported for DMF and fingolimod vs placebo on overall relapse and relapses requiring intravenous corticosteroids in both individual and pooled studies (pooled data unavailable for fingolimod). However, NNTs for each outcome were similar for DMF and teriflunomide, with marginally lower NNTs observed with fingolimod. By contrast, for relapses requiring hospitalization, relative reductions were higher and NNTs were substantially lower for teriflunomide compared with DMF. For fingolimod, there were inconsistent outcomes between the two studies for relapses requiring hospitalization; thus, comparative conclusions against DMF or teriflunomide cannot be clearly established. The risk of disability worsening was significantly reduced in both teriflunomide studies, but only in a single study for DMF (DEFINE) and fingolimod (FREEDOMS). NNTs to prevent one patient from experiencing disability worsening were similar in DEFINE, FREEDOMS, and TEMSO and TOWER but were higher in CONFIRM and FREEDOMS II. This NNT analysis demonstrates broadly comparable effects for DMF, fingolimod, and teriflunomide across key clinical outcomes. These observations are clinically relevant and may help to inform treatment decisions by providing additional information on therapeutic gain beyond informal assessments of relative reductions alone.

1. Introduction nificant efficacy on various markers of multiple sclerosis (MS) disease activity in large, randomized controlled trials in patients with relap- The oral disease-modifying therapies (DMTs), dimethyl fumarate sing-remitting MS (Gold, 2012; Fox, 2012; Kappos, 2010; Calabresi, (DMF), fingolimod, and teriflunomide, have each demonstrated sig- 2014; O'Connor, 2011; Confavreux, 2014). In most countries, DMF

Abbreviations: AE, adverse event; ARR, annualized relapse rate; DMF, dimethyl fumarate; DMT, disease-modifying therapy; EDSS, Expanded Disability Status Scale; FS, Functional System; IV, intravenous; MS, multiple sclerosis; NNH, number needed to harm; NNT, number needed to treat ⁎ Corresponding author. E-mail address: [email protected] (M.S. Freedman). http://dx.doi.org/10.1016/j.msard.2016.10.010 Received 26 April 2016; Received in revised form 30 September 2016; Accepted 31 October 2016 2211-0348/ © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). M.S. Freedman et al. Multiple Sclerosis and Related Disorders 10 (2016) 204–212 and teriflunomide are typically indicated as first-line therapies in Key information from these studies is summarized in Table 1. patients with active disease (Scolding, 2015), while fingolimod is All clinical studies were conducted in accordance with the generally reserved as a second-line therapy in patients with active International Conference on Harmonisation Guidelines for Good disease (European Medicines Agency, 2015; Torkildsen et al., 2016), Clinical Practice and the Declaration of Helsinki. The protocols were although it is approved as first-line option in the USA (FDA, 2016). approved by central and local ethics committees, and patients gave Given the growing use of oral DMTs in the treatment of relapsing- written consent prior to the studies. remitting MS, it is of interest to gain insight into the comparative efficacy of these three oral DMTs. 2.2. Study evaluations In the absence of head-to-head clinical trials, clinicians often compare the efficacy of DMTs on the basis of relative reductions in 2.2.1. MS relapses, relapses with severity features, and disability specific study endpoints, commonly in terms of the annualized relapse worsening rate (ARR), in order to inform prescribing decisions. Relative risk Across the DMF, fingolimod, and teriflunomide studies, the defini- reductions are commonly used as they are simple to compute, can be tions used for MS relapses, relapses with sequelae, and disability interpreted easily, provide a useful means for summarizing the worsening differed. Table 2 provides a summary of the outcome evidence (Schechtman, 2002), and are often the measure of choice measure definitions used in these pivotal studies. The frequency of reported in regulatory documentation (FDA, 2016; FDA, 2016; FDA, all MS relapses, relapses leading to hospitalization, or relapses requir- 2016). However, such comparisons can be influenced by differences in ing treatment with intravenous (IV) corticosteroids in the DMF, disease severity between study populations, and relative reductions can fingolimod, and teriflunomide studies was documented in terms of inflate treatment effects when overall event rates are low. Furthermore, ARR, which represents the number of relapses per patient-year in each the MS clinical trial population has evolved over the last two decades study. with a downward trend in relapse rates, largely due to improvements in To the best of our knowledge, post hoc analyses of DMF treatment early diagnosis of MS and introduction of new DMTs (Inusah, 2010; effects on relapse-associated sequelae have only been performed in Zakaria, 2015), which may have considerable impact when comparing CONFIRM (Chan, 2014). This analysis evaluated incomplete recovery studies conducted at different points in time. The use, therefore, of a post relapse (or disability worsening linked to relapse within a specific single measure of the magnitude of efficacy can lead to erroneous time interval), defined as the proportion of patients with 12-week conclusions, as noted by Zakaria in which hypothetical examples and confirmed disability worsening occurring within 180 days of relapse MS clinical trial data were used to demonstrate that the “use of the start date (Table 2). As a consequence of the lack of available data from relative risk reduction as a measure of drug efficacy can be misleading” DEFINE, DMF data pertaining to treatment effects on relapse-asso- (Zakaria, 2015). For example, in clinical trial populations where there ciated sequelae in this article focus on the CONFIRM study only. For is a high risk of disease activity, treatment may result in large, clinically fingolimod, assessment of post-relapse recovery in FREEDOMS and significant reductions in absolute risk, driven by the high event rate in FREEDOMS II was based upon the investigators’ clinical judgement as the control arm. Conversely, in clinical trial populations with a low risk complete or incomplete (either partial or none); and, to the best of our of disease activity, absolute reductions may be small and of marginal knowledge, the duration over which post-relapse recovery was assessed clinical significance but associated with high relative risk reductions has not been reported (Vollmer, 2012; Haas, 2011). In TEMSO and (Schechtman, 2002; Zakaria, 2015). TOWER, the primary assessment of relapse-associated sustained Given these challenges, it has been recommended to utilize both sequelae was performed at least 30 days post relapse; but, as sequelae relative and absolute risk reduction (and its inverse, number needed to can persist for longer, additional assessments were performed at least treat [NNT]) in any cross-trial comparison to provide a more reliable 60, 90, 120, 150, and 180 days post relapse (O'Connor, 2013; Miller, measure of comparative efficacy and thus ‘therapeutic gain’ (Freedman, 2014). However, data used to derive the NNT for relapses with 2008). NNT is a statistically valid and increasingly used tool for sequelae for teriflunomide in our analysis reflect published data evaluating comparative efficacy of different therapies and can provide collected at least 30 days post relapse (O'Connor, 2013; Miller, 2014). physicians with a relatively simple assessment of potential treatment efficacy that may assist with treatment planning decisions. 2.2.2. Calculation of NNT Here, we report post hoc analyses of individual and pooled data NNT data reported in this article were calculated from individual (where published or presented) from pivotal trials of DMF (DEFINE, and pooled outcomes data (where published or presented) from the NCT00420212 (Gold, 2012); CONFIRM, NCT00451451 (Fox, 2012)), DMF (DEFINE/CONFIRM) (Gold, 2012; Fox, 2012), fingolimod fingolimod (FREEDOMS, NCT00289978 (Kappos, 2010); FREEDOMS (FREEDOMS/FREEDOMS II) (Kappos, 2010; Calabresi, 2014), and II, NCT00355134 (Calabresi, 2014)), and teriflunomide (TEMSO, teriflunomide (TEMSO/TOWER) (O'Connor, 2011; Confavreux, 2014) NCT00134563 (O'Connor, 2011); TOWER, NCT00751881 pivotal studies, with the exception of relapses with sequelae, which (Confavreux, 2014)) to determine the respective NNTs to prevent were derived from CONFIRM only (Chan, 2014). relapse and disability worsening. Despite no universal definition of a For analysis of each outcome, source data were used exactly as ‘severe’ relapse in the MS literature, hospitalizations, corticosteroid provided in the published references. Consequently, there may be use, and sequelae post relapse can be considered as surrogates for instances where differences exist between studies and outcomes in severe relapses (O'Connor, 2013; Miller, 2014); therefore, NNTs for terms of the number of decimal places reported, which can ultimately these more clinically meaningful relapses were also derived for each subtly influence the NNT calculations. DMT. NNTs were calculated using data from patients receiving placebo or DMF 240 mg twice daily, fingolimod 0.5 mg once daily, or terifluno- 2. Methods mide 14 mg once daily, as these are the commercially available doses in the majority of markets where these treatments are approved for use 2.1. Study designs (European Medicines Agency, 2015; European Medicines Agency, 2015; European Medicines Agency, 2015). Detailed methodology of the pivotal DMF (DEFINE and NNTs to prevent one relapse, one relapse leading to hospitalization, CONFIRM), fingolimod (FREEDOMS and FREEDOMS II), and teri- or one relapse requiring treatment with IV corticosteroids were flunomide (TEMSO and TOWER) clinical trials included in our NNT calculated as the inverse of the absolute risk difference between the analyses have been published previously (Gold, 2012; Fox, 2012; ARRs in the placebo group (ARRp) and the active treatment group Kappos, 2010; Calabresi, 2014; O'Connor, 2011; Confavreux, 2014). (ARRa): NNT =1/(ARRp−ARRa). For relapses with sequelae, data were

205 ..Feda tal. et Freedman M.S. Table 1 Overview of the dimethyl fumarate, fingolimod, and pivotal studies.

Dimethyl fumarate Fingolimod Teriflunomide

DEFINE(Gold, 2012) CONFIRM(Fox, 2012) FREEDOMS(Kappos, 2010) FREEDOMS II(Calabresi, TEMSO(O'Connor, 2011) TOWER(Confavreux, 2014) 2014)

Design Phase 3, randomized, Phase 3, randomized, placebo- Phase 3, randomized, placebo- Phase 3, randomized, Phase 3, randomized, placebo- Phase 3, randomized, placebo- placebo-controlled controlled controlled placebo-controlled controlled controlled Population RRMS RRMS RRMS RRMS RMS RMS Intervention DMF 240 mg bid, DMF DMF 240 mg bid, DMF 240 mg Fingolimod 0.5 mg qd, 1.25 mg qd, Fingolimod 0.5 mg qd, Teriflunomide 14 mg qd, 7 mg qd, Teriflunomide 14 mg qd, 7 mg qd, or 240 mg tid, or placebo tid, sc GA 20 mg/day, or placebo or placebo 1.25 mg qd, or placebo or placebo placebo Duration of 96 weeks 96 weeks 24 months 24 months 108 weeks 48–173 weeks (treatment ended 48 treatment weeks after last patient randomized) Primary endpoint Proportion of patients ARR ARR ARR ARR ARR relapsed at 2 years

Selected secondary ARR Proportion of patients relapsed Time to disability worseninga Percent brain volume Time to disability worseninga Time to disability worseninga endpoints at 2 years (confirmed after 3 months) change from baseline (confirmed for 12 weeks) (confirmed for 12 weeks) Time to disability worseninga Time to disability worseninga MRI measures Time to disability MRI measures (confirmed for 12 weeks) (confirmed for 12 weeks) worseninga (confirmed at 3 months) MRI measures MRI measures Other MRI measures

ARR=annualized relapse rate; bid=twice daily; DMF=dimethyl fumarate; GA=; MRI=magnetic resonance imaging; qd=once daily; RMS=relapsing forms of MS; RRMS=relapsing-remitting MS; sc=subcutaneous; tid=three times daily. a Disability worsening duration presented as reported in each study. The terminology used here differs from the primary publications and has been aligned with the recommendations of the US National Multiple Sclerosis Society Advisory Committee on Clinical Trials in Multiple Sclerosis (Lublin, 2014). 206

Table 2 Outcome measure definitions used in the dimethyl fumarate, fingolimod, and teriflunomide pivotal studies.

Dimethyl fumarate Fingolimod Teriflunomide

DEFINE(Gold, 2012) CONFIRM(Fox, 2012; Chan, 2014) FREEDOMS(Kappos, 2010; FREEDOMS TEMSO(O'Connor, 2011; TOWER(Confavreux, 2014; Miller, Haas, 2011) II(Calabresi, 2014; O'Connor, 2013) 2014) Vollmer, 2012)

Confirmed relapses New or recurrent neurologic symptoms not associated with fever or Symptoms accompanied by an increase of: New clinical symptom or clinical worsening of a previous symptom that had Multiple SclerosisandRelatedDisorders10(2016)204–212 infection, that lasted ≥24 h, that were accompanied by new objective ≥0.5 EDSS points or been stable for ≥30 days and that persisted for ≥24 h in the absence of fever neurologic findings according to the examining neurologist's 1 point in 2 EDSS FS scores or and was accompanied by an increase from baseline of: evaluation and separated from the onset of other confirmed relapses 2 points in 1 EDSS FS scorea ≥0.5 EDSS pointsb or by at least 30 days 1 EDSS point in each of 2 EDSS FS scores or 2 EDSS points in 1 EDSS FS scorea Relapses with N/A Disability worsening confirmed for 12 Relapse recovery based upon the investigators’ clinical Relapse (as defined above) leading to increases in EDSS/FS score ≥30 daysd sequelae weeks (defined below) occurring within judgement as complete or incomplete (either partial or none)c post relapse or relapses with incomplete neurological recovery as defined by the 180 days of relapse start date investigator Disability worsening Increase from baseline of: Increase from baseline of: Increase from baseline of: confirmed for 12 ≥1 EDSS point if baseline score is ≥1.0 or ≥1 EDSS point or ≥1 EDSS point or weeks/ 3 monthse ≥1.5 EDSS points if baseline EDSS score is 0 0.5 EDSS points if baseline EDSS score ≥5.5 0.5 EDSS points if baseline EDSS score > 5.5

EDSS=Expanded Disability Status Scale; FS=Functional System; N/A=not available. a Excluding scores for bowel/bladder and cerebral functional systems; b from a previous clinically stable assessment; c for FREEDOMS, the proportion of patients experiencing relapses with sequelae has not been reported and is not included in this analysis; d additional assessments performed ≥60, 90, 120, 150, and 180 days post relapse; e disability worsening duration reported as reported in each study; terminology used here differs from the primary publications and has been aligned with the recommendations of the US National Multiple Sclerosis Society Advisory Committee on Clinical Trials in Multiple Sclerosis (Lublin, 2014). M.S. Freedman et al. Multiple Sclerosis and Related Disorders 10 (2016) 204–212 only available for the crude rate of relapses in CONFIRM and mod compared with teriflunomide. Fingolimod was associated with the FREEDOMS II (Chan, 2014; Vollmer, 2012). Therefore, to ensure largest absolute risk reductions in ARR and, consequently, marginally consistency across studies, the crude rate of these relapses rather than lower NNTs compared with DMF or teriflunomide. NNTs were similar the ARR was used as the basis for the NNT calculation. In this case, the between DMF and teriflunomide, suggesting a similar therapeutic gain NNT was calculated as the inverse of the absolute risk difference with respect to the prevention of one relapse (Table 4a). between the proportion of patients with such relapses in the placebo group (Pp) and the active treatment group (Pa): NNT =1/(Pp–Pa). 3.2.2. NNT to prevent one relapse leading to hospitalization Analyses of disability worsening in all studies were reported using a The risk of relapse leading to hospitalization was significantly time-to-event method. Therefore, the NNT to prevent one patient from reduced with active treatment vs placebo in FREEDOMS (Haas, experiencing disability worsening was calculated using the estimated 2011), TEMSO, TOWER, and the pooled TEMSO/TOWER data set proportion of worsening events from Kaplan–Meier curves at 2 years (O'Connor, 2013; Miller, 2014). By contrast, relative reductions on this with a hazard ratio calculated from a Cox model using an Altman relapse subset were not significant in DEFINE, CONFIRM, or derivation (Altman and Andersen, 1999). This approach takes into FREEDOMS II (European Medicines Agency, 2013; Havrdova, 2012). h account censoring and is calculated as: NNT =1/{[Sc(t)] – Sc(t)}, However, a significant relative reduction was reported for DMF when where the survival probability in the placebo group, Sc(t), is equal to 1 relapse data were pooled, although the overall magnitude of this – the probability of disability worsening confirmed for 12 weeks and h reduction was not as large as in the teriflunomide pooled data set is the hazard ratio of the active treatment group vs the placebo group. (Miller, 2014; Havrdova, 2012)(Table 4b). Compared with both DMF In all NNT analyses, lower NNTs reflect greater benefit for the studies and FREEDOMS II, absolute risk reductions were greater and respective therapeutic intervention. NNTs lower in the teriflunomide individual studies and pooled data sets, as well as in FREEDOMS (Table 4b). 3. Results 3.2.3. NNT to prevent one relapse requiring IV corticosteroids 3.1. Baseline characteristics The risk of relapse requiring treatment with IV corticosteroids was significantly reduced with active treatment vs placebo in all individual Patient demographics and disease characteristics were generally studies and the respective pooled data sets (where published or similar across all DMF, fingolimod, and teriflunomide studies, with presented) although higher relative risk reductions were observed with some notable differences (Table 3)(Gold, 2012; Fox, 2012; Kappos, both DMF and fingolimod compared with teriflunomide, with fingoli- 2010; Calabresi, 2014; O'Connor, 2011; Confavreux, 2014; European mod being associated with the highest relative risk reductions of the Medicines Agency, 2015; European Medicines Agency, 2015; European three DMTs (O'Connor, 2013; Miller, 2014; Vollmer, 2012; Haas, Medicines Agency, 2013; Hutchinson, 2013; Bar-Or, 2013; European 2011; Havrdova, 2012)(Table 4c). However, similar absolute risk Medicines Agency, 2011; Kremenchutzky, 2014; European Medicines reductions and thus NNTs were observed across individual and pooled Agency, 2013). A proportion of patients with progressive disease DMF and teriflunomide data sets, with marginally greater absolute risk (secondary progressive MS; progressive relapsing MS) were included reductions and lower NNTs observed for both fingolimod studies in the teriflunomide studies, but not in the DMF or fingolimod studies. compared with DMF and teriflunomide (Table 4c). Patients in DEFINE, FREEDOMS, and FREEDOMS II had slightly lower EDSS scores, and a greater percentage, particularly in 3.2.4. NNT to prevent one relapse with sequelae/incomplete recovery FREEDOMS II, had prior exposure to other DMTs compared with NNTs to prevent one patient from experiencing a relapse with those in TEMSO, TOWER, and CONFIRM. Patients in FREEDOMS II sequelae were as follows: DMF, 24.4 (CONFIRM) (Chan, 2014); also had a longer time since first symptoms of MS than the other fingolimod, 26.6 (FREEDOMS II) (Vollmer, 2012); and teriflunomide, studies. In addition, there were differences in treatment duration 7.9 (TEMSO) (O'Connor, 2013; Miller, 2014), 12.7 (TOWER) (Sanofi within the respective studies. DEFINE and CONFIRM had a fixed Genzyme), and 9.8 (pooled TEMSO and TOWER) (Sanofi Genzyme). treatment duration of 96 weeks, whereas the fixed treatment duration For teriflunomide, NNTs to prevent relapses with sequelae slightly in FREEDOMS and FREEDOMS II was 104 weeks. In TEMSO, increased when evaluated over 90 days post-relapse (similar to the treatment duration was 108 weeks, but was variable in TOWER, with analysis performed in CONFIRM) but remained lower than with DMF the study ending 48 weeks after the last patient was randomized. Mean (13.7 vs 24.4, respectively). However, it is important to note that and median treatment exposure for teriflunomide 14 mg in the pooled differences exist in how relapses with sequelae were evaluated in the TEMSO/TOWER data set was 85 and 106 weeks, respectively, corre- respective studies (Table 2). Furthermore, to the best of our knowledge, sponding to 1188 patient-years (Sanofi Genzyme). To the best of our the duration over which recovery was assessed in FREEDOMS II has knowledge, mean and median treatment exposure in the pooled not been reported. Given the non-comparable nature of the endpoint CONFIRM/DEFINE data set has not been reported; however, treat- assessments used in the above studies, it is not possible to draw any ment exposure in the pool was reported as 1129 patient-years (Center firm conclusions with regard to relative efficacy between the different for Drug Evaluation and Research, 2013). As far as we are aware, oral DMTs with respect to relapses with sequelae/incomplete recovery. treatment exposure data in the two fingolimod studies have not been For the purposes of these analyses, NNTs were calculated based on the reported. The DMF, fingolimod, and teriflunomide studies were con- data available, and NNTs to prevent one patient from experiencing a temporary, with patient recruitment occurring at broadly similar times relapse with sequelae/incomplete recovery were derived from the (Table 3). proportion of patients with such relapses, rather than the ARR (see Methods section). NNTs to prevent one patient from experiencing a 3.2. Efficacy outcomes relapse with sequelae were lower with teriflunomide (in both the individual studies and pooled data set) compared with DMF or 3.2.1. NNT to prevent one relapse fingolimod. NNTs to prevent one patient from experiencing a relapse DMF, fingolimod, and teriflunomide all significantly reduced the with sequelae were comparable between DMF and fingolimod. As risk of MS relapse vs placebo in each of the individual studies (Gold, information regarding the duration over which recovery was assessed 2012; Fox, 2012; Kappos, 2010; Calabresi, 2014; O'Connor, 2011; in the fingolimod trials has, to our knowledge, not been reported, it is Confavreux, 2014) and in the pooled data sets for DMF (Fox, 2013) and challenging to draw firm conclusions based on these observations. teriflunomide (Kappos, 2013). However, higher relative reductions When relapses from the teriflunomide data set were assessed for were observed with DMF (individual and pooled studies) and fingoli- sequelae at a later time point – at least 90 days post relapse (i.e.,

207 ..Feda tal. et Freedman M.S.

Table 3 Baseline demographic and disease characteristics in the dimethyl fumarate, fingolimod, and teriflunomide pivotal studies (total study populationsa).

Dimethyl fumarate Fingolimod Teriflunomide

DEFINE (N=1234)a (Gold, CONFIRM (N=1417)a (Fox, FREEDOMS FREEDOMS II TEMSO (N=1086)a (O'Connor, TOWER 2012; European Medicines 2012; European Medicines (N=1272)a (Kappos, 2010; (N=1083)a, b (Calabresi, 2014; 2011; European Medicines (N=1165)a (Confavreux, 2014; Agency, 2015; European Agency, 2015; European European Medicines Agency, European Medicines Agency, 2015) Agency, 2013) Sanofi Genzyme) Medicines Agency, 2013; Bar-Or, Medicines Agency, 2013; 2011; Kremenchutzky, 2014) 2013) Hutchinson, 2013)

Study initiation Jan 2007 Mar 2007 Jan 2006 Jun 2006 Sep 2004 Sep 2008

Age, median 39 (18:56) 37 (18:56) 37 (17:55) 41 (18:57) 38 (18:55) 38 (18:56) (min:max), years Female, % 74 70 69.9 77.9 72.1 71.2 Time since 5.5 (5.49) 4.7 (5.01) 6.5 (4.8) NR 5.33 (5.48) 5.16 (5.67) diagnosis, mean (SD), years Time since first 8.3 (6.65) 7.1 (6.39) 8.2 (6.60) 10.6 (8.0) 8.68 (6.90) 8.00 (6.73) symptoms of MS, mean (SD), years

208 Number of 1.3 (0.70) 1.4 (0.70) 1.5 (0.77) 1.5 (0.9) 1.4 (0.7) 1.4 (0.7) relapses in previous year, mean (SD) MS subtype, % RRMS 100 100 100 100 91.4 97.5 SPMS 0 0 0 0 4.7 0.8 PRMS 0 0 0 0 3.9 1.7 EDSS score Mean (SD) 2.42 (1.240) 2.56 (1.194) 2.40 (1.32) 2.43 (1.30) 2.68 (1.30) 2.71 (1.37) Median 2.0 (0.0:6.5) 2.5 (0.0:5.5) 2.0 (0.0:5.5) 2.5 (0.0:6.5) 2.5 (0.0:6.0) 2.5 (0.0:6.5) (min:max) Multiple SclerosisandRelatedDisorders10(2016)204–212 Previous DMT 40 30 40.9 74.8 27.1c 33b,c use, % Patients with Gd- 36d 45e 38.1f 35.3g 36.3 No MRI enhancing T1 lesions, %

DMT=disease-modifying therapy; EDSS=Expanded Disability Status Scale; Gd=gadolinium; MRI=magnetic resonance imaging; NNT=number needed to treat; NR=not reported; PRMS=progressive relapsing MS; RRMS=relapsing-remitting MS; SD=standard deviation; SPMS=secondary progressive MS. a Randomized and treated patients (including active treatment groups not included in the NNT analyses); b total population data have not been published (data presented are derived from individual treatment groups); c within the last 2 years; d 540 patients in the MRI cohort; e 679 patients in the MRI cohort; f 1264 patients with data; g 1078 patients in the MRI cohort. ..Feda tal. et Freedman M.S.

Table 4 Number needed to treat to prevent (a) one relapse, (b) one relapse leading to hospitalization, or (c) one relapse requiring IV corticosteroids in DEFINE, CONFIRM, FREEDOMS, FREEDOMS II, TEMSO, and TOWER and respective pooled data sets (dimethyl fumarate and teriflunomide only).

(a) Dimethyl fumarate 240 mg bid Fingolimod 0.5 mg qd Teriflunomide 14 mg qd

DEFINE CONFIRM DEFINE + CONFIRM FREEDOMS FREEDOMS II TEMSO TOWER TEMSO + TOWER pooled (N=818)a (Gold, (N=722)a (Fox, pooled (N=843)(Kappos, (N=713) (N=721)a (O'Connor, (N=758)a (Confavreux, (N=1479)a (Miller, 2014; 2012) 2012) (N=1540)a (Fox, 2013) 2010) (Calabresi, 2014) 2011) 2014) Kappos, 2013)

ARR Placebo 0.36 0.40 0.37 0.40 0.40 0.54 0.50 0.534 Intervention 0.17 0.22 0.19 0.18 0.21 0.37 0.32 0.354 Relative reduction 53 44 49 54 48 31.5 36.3 33.7 vs placebo, % P value vs placebo p < 0.001 p < 0.001 p < 0.0001 p < 0.001 p < 0.0001 p < 0.001 p=0.0001 p < 0.001 Absolute 0.19 0.18 0.18 0.22 0.19 0.17 0.18 0.18 reductionb NNT 5.3 5.6 5.6 4.5 5.3 5.9 5.6 5.6

(b) Dimethyl fumarate 240 mg bid Fingolimod 0.5 mg qd Teriflunomide 14 mg qd

209 DEFINE CONFIRM DEFINE + CONFIRM FREEDOMS FREEDOMS II TEMSO TOWER TEMSO + TOWER (N=818)a (Havrdova, (N=722)a (Havrdova, pooled (N=843)(Kappos, (N=713) (N=721)a (O'Connor, (N=758)a (Miller, pooled 2012) 2012) (N=1540)a (Havrdova, 2010; Haas, 2011) (Calabresi, 2014; 2013) 2014) (N=1479)a (Miller, 2012) Vollmer, 2012) 2014)

ARR Placebo 0.056 0.055 0.048 0.18 0.023 0.139 0.15 0.16 Intervention 0.036 0.038 0.031 0.07 0.011 0.057 0.10 0.09 Relative 35 32 34 61c 52c 59 33.6 45.5 reduction vs placebo, % P value vs p=0.071 p=0.109 p < 0.05 p≤0.001 NS p < 0.0001 p=0.0155 p < 0.0001 Multiple SclerosisandRelatedDisorders10(2016)204–212 placebo Absolute 0.020 0.017 0.017 0.110 0.012 0.082 0.05 0.07 reductionb NNT 50.0 58.8 58.8 9.1 83.3 12.2 20.0 14.3

(c) Dimethyl fumarate 240 mg bid Fingolimod 0.5 mg qd Teriflunomide 14 mg qd

DEFINE CONFIRM DEFINE + CONFIRM FREEDOMS FREEDOMS II TEMSO TOWER TEMSO + TOWER (N=818)a (Havrdova, (N=722)a (Havrdova, pooled (N=843)(Kappos, (N=713) (N=721)a (O'Connor, 2013) (N=758)a (Miller, pooled 2012) 2012) (N=1540)a (Havrdova, 2010; Haas, 2011) (Calabresi, 2014; 2014) (N=1479)a (Miller, 2012) Vollmer, 2012) 2014)

ARR Placebo 0.31 0.34 0.32 0.38 0.303 0.428 0.43 0.43 Intervention 0.15 0.19 0.16 0.17 0.125 0.284 0.27 0.28 Relative 52 44 48 55c 59c 34 35.7 34.5 reduction vs placebo, % (continued on next page) M.S. Freedman et al. Multiple Sclerosis and Related Disorders 10 (2016) 204–212

similar to the CONFIRM study; sequelae sustained for 12 weeks/84 days post relapse) – NNTs remained lower in teriflunomide-treated patients in the pooled TEMSO/TOWER data set than in those receiving Miller, ( DMF (13.7 vs 24.4, respectively). a

) 3.2.5. NNT to prevent confirmed disability worsening The risk of disability worsening confirmed for 12 weeks/3 months 0.15 2014 TEMSO + TOWER pooled (N=1479) 6.7 was significantly reduced vs placebo in DEFINE (Gold, 2012) and FREEDOMS (Kappos, 2010) but was significantly reduced in both the TEMSO (30%) (O'Connor, 2011) and TOWER (32%) studies Miller,

( (Confavreux, 2014). Corresponding absolute risk reductions and a NNTs were similar in DEFINE, FREEDOMS, TEMSO, and TOWER ) but were higher in CONFIRM and FREEDOMS II (Table 5). NNTs in fl 0.16 2014 TOWER (N=758) 6.3 the pooled data sets were similar for both DMF and teri unomide

) (Table 5)(Fox, 2013; Kappos, 2013) and remained similar when applying an alternative method proposed by Altman (Altman and unomide 14 mg qd

fl Andersen, 1999) for time-to-event analyses using the inverse of the difference of estimated probabilities at 2 years for a worsening event (data not shown). O'Connor, 2013 ( a 4. Discussion 0.14 6.9 TEMSO (N=721) Despite the well-established limitations of cross-trial comparisons, clinicians frequently compare study outcomes on the basis of relative ) reductions in a specific endpoint, commonly in terms of the ARR in MS, to inform prescribing decisions. However, such comparisons do not give a full picture of the therapeutic benefit of the respective interven-

unomide 14 mg qd and the respective placebo groups in each study; tions. The use, therefore, of evidence-based approaches such as NNT Calabresi, 2014; fl 0.18 5.6 FREEDOMS II (N=713) ( Vollmer, 2012 can provide additional information on the relative benefit/risk profiles of specific treatments (Goodin, 2008). NNT is well supported in the ) literature with sufficient evidence to establish its scientific validity as a method for providing indirect comparisons of efficacy across studies, Kappos, ( which can provide critically important, clinically meaningful informa- tion in the absence of head‐to‐head trials (Freedman, 2008). FREEDOMS (N=843) 0.21 2010; Haas, 2011 4.8 Using this approach, we demonstrated broadly comparable NNTs fi fl ngolimod 0.5 mg qd, or teri for DMF, ngolimod, and teri unomide with respect to relapses and fi relapses requiring treatment with IV corticosteroids despite differences in relative reductions between the three treatments. However, NNTs were marginally lower for fingolimod across both of these relapse outcomes. NNTs to prevent one relapse requiring hospitalization were Havrdova, (

a higher for fingolimod compared with teriflunomide based on data from FREEDOMS II; although, the converse was observed when ) FREEDOMS data were analyzed. For DMF, NNTs to prevent one DEFINE + CONFIRM pooled (N=1540) 0.16 2012 6.3 relapse requiring hospitalization were considerably higher than for teriflunomide or fingolimod (FREEDOMS only) and were more in line with NNTs derived from FREEDOMS II. However, relapses leading to hospitalization in the DMF studies as well as in FREEDOMS II were low, which may, in part, account for these observations. Havrdova, (

a The variation in NNTs observed between FREEDOMS and FREEDOMS II may be related to differences in patient demographics ) and disease characteristics. In FREEDOMS II, on average, patients CONFIRM (N=722) 0.15 2012 6.7 were older and had a longer disease duration, with a greater proportion exposed to prior DMTs. In addition, the discontinuation rate was greater in FREEDOMS II compared with FREEDOMS, which may have been driven in part by additional safety assessments required by regulatory agencies that were not included in FREEDOMS (Kappos, Havrdova, (

a 2010; Calabresi, 2014). It should be noted that these differences in trial characteristics could impact any of the outcomes reported and are ) probably not exclusive to the variations observed in the NNT data. For p < 0.0001 p=0.0002 p < 0.0001 p < 0.001 p < 0.001 p=0.0003 p=0.0002 p < 0.0001 0.16 2012 DEFINE (N=818) 6.3 example, these differences may have led to the considerable difference in the relative risk reductions in disability worsening that was observed

b between the FREEDOMS and FREEDOMS II trials. Furthermore, relapse management, including indications for the (continued) use of IV corticosteroids, and the need to hospitalize for corticosteroid reduction placebo Absolute P value vs (c) Dimethyl fumarate 240 mg bid Fingolimod 0.5 mg qd Teri NNT infusion and/or relapses varies, and certain countries may not routi- Absolute reductions were calculated as ARR for placebo-treated patients minus ARR for patients treated with intervention; The total number of patients includes those randomized and treated with dimethyl fumarate 240 mg bid, Relative reduction vs placebo derived from absolute ARR values reported in cited sources. Table 4 a b c ARR=annualized relapse rate; bid=twice daily; NNT=number needed to treat; NS=not significant; qd=once daily. nely use steroids in relapse management or admit patients experien-

210 M.S. Freedman et al. Multiple Sclerosis and Related Disorders 10 (2016) 204–212

) cing relapses (O'Connor, 2013). Thus, the geographical location of patients included in these pivotal studies may also have an impact on NNT outcomes reported here. Of the three oral DMTs evaluated, only teriflunomide showed

Kappos, 2013 fi ( consistent and signi cant reductions in the risk of disability worsening a in each of the individual studies, which was reflected in lower NNTs in TEMSO and TOWER compared with CONFIRM and FREEDOMS II (in which significant effects of DMF and fingolimod were not demon- p=0.0029 15.1 0.695 0.695 0.179 p=0.003 0.240 TEMSO + TOWER pooled (N=1479) 30.5 strated) (Fox, 2012; Calabresi, 2014; O'Connor, 2011; Confavreux, 2014). However, NNTs for prevention of confirmed disability worsen- ing were broadly comparable between DMF and teriflunomide in the respective pooled study data (Fox, 2013; Kappos, 2013) and also between fingolimod (FREEDOMS data only) and teriflunomide

Confavreux, fi ( (Kappos, 2010). This nding was obtained using either method a proposed by Altman (Altman and Andersen, 1999), both of which are

) appropriate for calculating NNTs based on results from a time-to-event – p=0.0442 17.1 0.68 0.68 0.158 p=0.0442 0.197 TOWER (N=758) 2014 32 analysis (calculated using either the proportion of events from Kaplan Meier curves and hazard ratios derived from a Cox model or from the inverse of the difference in estimated probabilities for a worsening event). For an assessment of relative benefit/risk, number needed to harm

O'Connor, (NNH) analyses are often conducted based on absolute risk differences ( a between treatment groups in adverse events (AEs), AEs of special unomide 14 mg qd ) fl interest, AEs leading to discontinuation, and serious AEs. However, because of distinct safety profiles for DMF, fingolimod, and terifluno- p=0.0279 13.7 0.70 0.70 0.202 p=0.03 0.273 Teri TEMSO (N=721) 2011 29.8 mide (as a consequence of different mechanisms of action), differences in collection of AE data (e.g., inclusion of MS relapses as an AE or unomide 14 mg qd, and the respective placebo groups in each study; fl serious AE in the DMF and fingolimod studies, respectively, but not in the teriflunomide studies), and differences between study protocols for fi Calabresi, mandated discontinuation for certain AEs (e.g., for con rmed alanine ( ≥ ) transaminase increase 3x upper limit of normal on two consecutive

c readings in the teriflunomide but not the DMF studies), NNH analyses p=0.227 23.5 0.83 0.83 0.253 p=0.227 0.290 17 FREEDOMS II N=713 2014 are challenging to conduct and have not been included in this study. There are a number of limitations of our analyses that should be highlighted. First, ARRs in the placebo group were typically higher in ngolimod 0.5 mg qd, or teri fi the teriflunomide studies compared with the DMF and fingolimod Kappos, ( studies; a common concern with NNT analyses is that treatment effects ) in trials with high placebo rates may be magnified when compared c 30 p=0.02 15.3 0.70 0.70 0.177 p=0.02 0.241 Fingolimod 0.5 mg qd FREEDOMS N=843 2010 against trials with low placebo rates (Goodin, 2008). This could

) potentially explain the variability observed for NNTs to prevent one relapse leading to hospitalization. Second, we appreciate there is variability in criteria for hospitalization for MS relapse, as well as Fox, 2013

( indications for corticosteroid treatment, between countries; however, a as this was a post hoc analysis of published data from clinical trials, it was not possible to adjust for such differences in study practices. Therefore, NNTs to prevent one relapse leading to hospitalization or 32 p=0.0034 15.4 0.68 0.680 0.146 p=0.0034 0.222 DEFINE+ CONFIRM pooled (N=1540) requiring IV corticosteroids should be interpreted with some caution. Third, for relapses with sequelae, data were available on the crude rate of such relapses from CONFIRM only, while the duration over which Fox, ( incomplete recovery was assessed has not been reported for a FREEDOMS II (Vollmer, 2012). In addition, definitions of post-relapse ) sequelae differed between the studies for each of the oral DMTs. 21 p=0.25 30.2 0.79 0.79 0.13 p=0.25 0.17 CONFIRM (N=722) 2012 Nevertheless, when a similar definition to that used in CONFIRM was applied to the teriflunomide data sets, lower NNTs with teriflunomide were still observed. Finally, for calculation of NNTs, we applied data exactly as published in the reference sources. Consequently, there are Gold, (

a differences between studies in terms of the number of decimal places reported for each respective outcome, which may subtly influence the ) overall size of the NNT calculated. 38 10.8 p=0.005 Dimethyl fumarate 240 mg bid 0.27 2012 DEFINE (N=818) Ideally, for a more robust assessment of comparative efficacy, head- to-head controlled clinical trials are required to establish firm conclu- sions. However, such studies are rarely conducted in practice for a

b variety of reasons, and therefore we are often reliant on indirect comparisons such as those reported herein. Other more complex reduction (%) patients with CDW % Intervention 0.16 Placebo 3-month CDW at 2 years; The total number of patients includes those randomized and treated with dimethyl fumarate 240 mg bid, relative reduction vs placebo derived from hazard ratios reported in cited source. approaches (e.g., propensity score matching on individual patient a b c Relative risk P value vs placeboNNT p=0.005 Hazard ratio 0.62 Hazard ratio 0.62 Proportion of P value vs placebo, bid=twice daily; CDW=confirmed disability worsening; NNT=number needed to treat; qd=once daily. Table 5 Number needed to treat to prevent one patient experiencing confirmed disability worsening in TEMSO, TOWER, DEFINE, CONFIRM, FREEDOMS, FREEDOMS II, TEMSO, TOWER, and the respective pooled data sets. level) have been recently applied in an attempt to compare the efficacy

211 M.S. Freedman et al. Multiple Sclerosis and Related Disorders 10 (2016) 204–212 of two treatments indirectly (Koch-Henriksen, 2015; Barbin, 2016). remitting multiple sclerosis (FREEDOMS II): a double-blind, randomised, placebo- controlled, phase 3 trial. Lancet Neurol. 13 (6), 545–556. Relative risk reduction and NNT analyses have advantages and Center for Drug Evaluation and Research, 2013. NDA 204063 - Tecfidera (Dimethyl disadvantages, and each measure captures a different aspect (relative Fumarate) Medical Review. and absolute) of the trial. Therefore, it is perhaps important to consider Chan, A., et al., 2014. Differential Recovery From Relapse Between Treatment Groups in the CONFIRM Study of Delayed-Release Dimethyl Fumarate. Mult. Scler. J. 20 (S1), both NNT and relative risk reductions when making cross-trial 67–284, (P096). comparisons so that a complete picture of treatment effectiveness can Confavreux, C., et al., 2014. Oral teriflunomide for patients with relapsing multiple be formed (Schechtman, 2002; Goodin, 2008). sclerosis (TOWER): a randomised, double-blind, placebo-controlled, phase 3 trial. – In summary, these post hoc analyses of data from individual clinical Lancet Neurol. 13 (3), 247 256. European Medicines Agency, 2011. Gilyena Assesment Report. trials of DMF, fingolimod, and teriflunomide and the respective pooled European Medicines Agency, 2013. Tecfidera Assesment Report. data (DMF and teriflunomide only) demonstrate comparable effects for European Medicines Agency, 2013. Aubagio Assesment Report. DMF, fingolimod, and teriflunomide across key clinical outcomes as European Medicines Agency, 2015. Aubagio SmPC. European Medicines Agency, 2015. Gilenya SmPC. determined by NNT evaluations. These observations have clinical European Medicines Agency, 2015. Tecfidera SmPC. relevance and may help to inform treatment decisions by providing FDA, 2016. Tecfidera (dimethyl fumerate) Prescribing Information. additional information on therapeutic gain beyond informal assess- FDA, 2016. Aubagio (teriflunomide) Prescribing Information. FDA, 2016. Gilenya (fingolimod) Prescribing Information. ments of relative reductions alone. Fox, D.A., et al., 2013. Clinical efficacy of BG-12 (dimethyl fumarate) in relapsing– remitting multiple sclerosis (RRMS): an integrated analysis of the Phase 3 DEFINE Conflicts of interest and CONFIRM studies. In: Proceedings of the 65th Annual Meeting of the American Academy of Neurology (AAN). San Diego, CA, USA: Neurology Fox, R.J., et al., 2012. Placebo-controlled phase 3 study of oral BG-12 or glatiramer in MSF: Research/educational grant support ( HealthCare, multiple sclerosis. N. Engl. J. Med. 367 (12), 1087–1097. Genzyme); honoraria/consulting fees (Bayer HealthCare, Idec, Freedman, M.S., et al., 2008. Efficacy of disease-modifying therapies in relapsing fi remitting multiple sclerosis: a systematic comparison. Eur. Neurol. 60 (1), 1–11. EMD Canada, Novartis, Sano , Teva Canada Innovation); member of Gold, R., et al., 2012. Placebo-controlled phase 3 study of oral BG-12 for relapsing company advisory boards/board of directors/other similar group multiple sclerosis. N. Engl. J. Med. 367 (12), 1098–1107. (Bayer HealthCare, Biogen Idec, Chugai, Merck Serono, Novartis, Goodin, D.S., 2008. Disease-modifying therapy in multiple sclerosis: update and clinical – Opexa Therapeutics, Sanofi, Teva Canada Innovation). implications. Neurology 71 (24 Suppl 3), S8 13. Haas, J., et al., 2011. Fingolimod reduces the number of severe relapses in patients with XM: Speaking honoraria/travel expenses for scientific meetings, relapsing multiple sclerosis: results from phase III TRANSFORMS and FREEDOMS steering committee member of clinical trials, participation in advisory studies. European Neurological Society (ENS), Lisbon, Portugal, P902. boards, research grants (Almirall, Bayer Schering, Biogen Idec, Havrdova, E., et al., 2012. Relapses requiring intravenous steroid use and MS-related hospitalizations: findings from the Phase 3 DEFINE and CONFIRM studies. In: 28th ECTRIMS, Genentech, Genzyme, GSK, Merck Serono, MSIF, NMSS, Congress of the European Committee for Treatment and Research in Multiple Novartis, Roche, Sanofi, Teva Pharmaceuticals). Sclerosis. Lyon, France AEM: Consulting fees (Accordant Health Services, Acorda Hutchinson, M., et al., 2013. Clinical efficacy of BG-12 (dimethyl fumarate) in patients with relapsing-remitting multiple sclerosis: subgroup analyses of the CONFIRM Therapeutics, Alkermes, Biogen Idec, EMD Serono, Genentech, study. J. Neurol. 260 (9), 2286–2296. Genzyme, GSK, Mallinckrodt Pharmaceuticals [Questcor], Novartis, Inusah, S., et al., 2010. Assessing changes in relapse rates in multiple sclerosis. Mult. Roche, Teva); contracted research (Biogen Idec, Genentech, Novartis, Scler. 16 (12), 1414–1421. fi fi Kappos, L., et al., 2010. A placebo-controlled trial of oral ngolimod in relapsing multiple Questcor, Roche, Sano ). Member of Mult Scler Rel Dis Editorial Board. sclerosis. N. Engl. J. Med. 362 (5), 387–401. CD-P: Employee of Sanofi Genzyme at the time study was Kappos, L., et al., 2013. Pooled efficacy data from two phase 3 placebo-controlled trials of conducted. oral, once-daily teriflunomide. In: 29th Congress of the European Committee for fi Treatment and Research in Multiple Sclerosis, Copenhagen, Denmark. SH: Employee of Sano Genzyme. Koch-Henriksen, N., Magyari, M., Sorensen, P.S., 2015. A comparison of MS disease KT: Employee of Sanofi Genzyme. activity between patients treated with and fingolimod. In: 31st Congress TPL: Consulting fees (AbbVie, Biogen Idec, EMD Serono, of the European Committee for Treatment and Research in Multiple Sclerosis. Genentech, Genzyme, Novartis, Teva Neuroscience); fees from non- Barcelona, Spain. Kremenchutzky, M., et al., 2014. Impact of prior treatment status and reasons for CME services (Novartis, Teva Neuroscience); contracted research discontinuation on the efficacy and safety of fingolimod: Subgroup analyses of the (Biogen Idec, Genzyme, GSK, Novartis, Roche, Sun Pharma). Fingolimod Research Evaluating Effects of Daily Oral Therapy in Multiple Sclerosis (FREEDOMS) study. Mult. Scler. Relat. Disord. 3 (3), 341–349. Lublin, F.D., et al., 2014. Defining the clinical course of multiple sclerosis: the 2013 Funding revisions. Neurology 83 (3), 278–286. Miller, A.E., et al., 2014. Teriflunomide reduces relapses with sequelae and relapses This study was supported by Sanofi Genzyme. leading to hospitalizations: results from the TOWER study. J. Neurol. 261 (9), 1781–1788. O'Connor, P., et al., 2011. Randomized trial of oral teriflunomide for relapsing multiple Acknowledgments sclerosis. N. Engl. J. Med 365 (14), 1293–1303. O'Connor, P.W., et al., 2013. Teriflunomide reduces relapse-related neurological sequelae, hospitalizations and steroid use. J. Neurol. 260 (10), 2472–2480. The manuscript was reviewed by Larisa Miller, PharmD. Editorial Sanofi Genzyme Data on File. assistance was provided by Scott Chambers and Steve Banner, of Schechtman, E., 2002. Odds ratio, relative risk, absolute risk reduction, and the number Fishawack Communications Ltd, also funded by Sanofi Genzyme. needed to treat–which of these should we use? Value Health 5 (5), 431–436. Scolding, N., et al., 2015. Association of British Neurologists: revised (2015) guidelines for prescribing disease-modifying treatments in multiple sclerosis. Pr. Neurol. 15 (4), References 273–279. Torkildsen, O., Myhr, K.M., Bo, L., 2016. Disease-modifying treatments for multiple sclerosis - a review of approved medications. Eur. J. Neurol. 23 (Suppl 1), 18–27. Altman, D.G., Andersen, P.K., 1999. Calculating the number needed to treat for trials Vollmer, T., et al., 2012. Effect of fingolimod on severe relapses, healthcare utilization where the outcome is time to an event. BMJ 319 (7223), 1492–1495. and relapse recovery in patients with relapsing-remitting multiple sclerosis: Results Barbin, L., et al., 2016. Comparative efficacy of fingolimod vs natalizumab: a French from the phase 3 FREEDOMS II study. In: 28th Congress of the European multicenter observational study. Neurology 86 (8), 771–778. Committee for Treatment and Research in Multiple Sclerosis. Lyon, France Bar-Or, A., et al., 2013. Clinical efficacy of BG-12 (dimethyl fumarate) in patients with Zakaria, M., 2015. Smoke and mirrors: limited value of relative risk reductions for relapsing-remitting multiple sclerosis: subgroup analyses of the DEFINE study. J. assessing the benefits of disease-modifying therapies for multiple sclerosis. Mult. Neurol. 260 (9), 2297–2305. Scler. Relat. Disord. 4 (3), 187–191. Calabresi, P.A., et al., 2014. Safety and efficacy of fingolimod in patients with relapsing-

212