NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Comparative Clinical and Cost-Effectiveness of Drug Therapies for Relapsing-Remitting Multiple Sclerosis [Internet]. Ottawa (ON): Canadian Agency for Drugs and Technologies in Health; 2013 Oct. (CADTH Therapeutic Review, No. 1.2B.)

Cover of Comparative Clinical and Cost-Effectiveness of Drug Therapies for Relapsing-Remitting Multiple Sclerosis

Comparative Clinical and Cost-Effectiveness of Drug Therapies for Relapsing-Remitting Multiple Sclerosis [Internet].

Show details


5.1. Summary of Evidence

The systematic review included 30 individual RCTs that reported the efficacy and safety of disease-modifying agents in patients with RRMS. There were 27 studies935 that provided comparisons of monotherapies and four3538 that provided comparisons between combination therapy and monotherapy. Evaluated interventions included alemtuzumab, natalizumab, interferon beta-1b, interferon beta-1a subcutaneous, interferon beta-1a intramuscular, glatiramer acetate, dimethyl fumarate, fingolimod, and teriflunomide. One monotherapy trial15 was restricted to treatment-experienced patients, and a number of studies10,11,13,14,23,2530 were restricted to treatment-naïve patients. However, the majority of monotherapy trials either did not specify whether patients had previously received disease-modifying treatments, or they included a mixed patient population (treatment-naive and treatment-experienced). None of the monotherapy trials specifically enrolled patients who had failed or were intolerant to previous treatments.

Of the four combination therapy trials, one (CombiRx) included treatment-naïve patients. The other three combination trials enrolled patients who had been previously treated with monotherapy; however, the extent to which patients could be considered to have failed previous treatment was unclear.

Data available for efficacy and safety outcomes were analyzed by direct pairwise meta-analyses. NMAs were conducted for two efficacy outcomes (ARR and sustained disability progression) to estimate comparative efficacies between all interventions.

5.2. Interpretation of the Results

5.2.1. Comparisons among treatment strategies

Comparisons were made for diverse treatment strategies that differed in chemical structure, mechanism of action, mode of administration, dosage, and treatment-related adverse events. The different modes of administration included intravenous infusion (alemtuzumab, natalizumab), subcutaneous (interferon beta-1b, interferon beta-1a, glatiramer acetate), intramuscular (interferon beta-1a), and oral (fingolimod, dimethyl fumarate, teriflunomide). Adverse events were treatment-specific, as expected because of the diversity of chemical structure and mechanism of action. Such diversity precluded the conduct of NMA on safety outcomes, which makes between-treatment comparisons of benefit and risk more challenging.

a. Monotherapy

For ARR, results from the NMA suggest that, compared with all other treatments, natalizumab and alemtuzumab result in statistically lower ARR, reducing the ARR by approximately 70% compared with no treatment (placebo). Two oral agents (fingolimod and dimethyl fumarate) appear to have similar activity to each other, reducing the ARR by approximately 50% compared with no treatment (placebo). Subcutaneous interferons, glatiramer acetate, and teriflunomide appear to have similar activity to each other, reducing the ARR by approximately 30% compared with no treatment (placebo). Intramuscular interferons appear to have the lowest activity of all agents.

Results from the NMA for ARR are generally consistent with the results from the available head-to-head trials that report greater efficacy for alemtuzumab compared with interferon beta-1a, greater efficacy of fingolimod and dimethyl fumarate compared with older agents (intramuscular interferon beta 1a and glatiramer acetate, respectively), and similar efficacy between glatiramer acetate and interferons.

Estimates of relative treatment effect from the NMA for disability (regarding the proportion of patients achieving sustained disability progression) show less precision than for ARR. Natalizumab and alemtuzumab again appear to have the greatest activity compared with all other agents; however, credible intervals are wide and demonstrate considerable overlap between agents, resulting in uncertainty in the relative efficacy of the remaining treatments. In addition, there is less consistency between direct and indirect estimates than was observed for ARR. Lesser between-treatment differences in disability compared with ARR may be an indication that relapse frequency is not directly related to disability progression. Alternatively, the disparity between the ARR and disability results may be related to the short duration of trials and the insensitivity of the measure of disability. Because RRMS is a slow-progressing disease, an accurate assessment of disability would require follow-up for longer than the two to three years used in many of the included studies. In addition, EDSS is an ordinal scale that focuses on mobility, and therefore does not capture all key components of disability in MS.

There were insufficient data to conduct an NMA for the remainder of the protocol-defined outcomes. However, findings from head-to-head trials report that MRI findings were more favourable for alemtuzumab compared with interferon beta-1a 44 mcg, and more favourable for all three of fingolimod, interferon beta-1b 250 mcg, and interferon beta-1a 44 mcg compared with interferon beta-1a 30 mcg. Compared with glatiramer, dimethyl fumarate resulted in a statistically lower mean number of T2 lesions, but the mean number of GdE lesions was not statistically different between these two treatments. However, it should be noted that only one trial contributed evidence for this comparison, the study was not powered for the active comparison, and MRI findings were not the planned primary outcome.

Evidence on HRQoL was limited to two treatments (interferon beta-1a 30 mcg and natalizumab) compared with placebo. For the comparison between interferon beta-1a 30 mcg and placebo in the MSCRG study, the difference in physical SIP score between interferon and placebo was 7.35, which was lower than the 12.5 points considered to represent a minimal clinically important difference in the physical domain. For the comparison between natalizumab and placebo in the AFFIRM study, a change of 5.0 points for PCS and MCS scales was considered to be a clinically meaningful difference. However, the differences between natalizumab and placebo in the proportion of patients achieving a clinically important change on the PCS improvement (8%) and PCS worsening (7%) were small. Thus, both treatments appear to have a small improvement in HRQoL on the physical domain, but not on the mental or psychosocial domain.

Adverse events of note were treatment-specific. Of the two treatments that appear to have the greatest activity (based on the NMA), alemtuzumab was associated with a high incidence of thyroid disorders compared with interferon beta-1a (17% versus 4%), while there is concern regarding the association between natalizumab and the risk of PML. Post-marketing data have estimated the risk of developing PML to be 1 in 500 patients treated with natalizumab.48 The risk of developing PML was increased with increasing treatment duration, history of previous exposure to immunosuppressive therapy, and presence of anti-JC Virus antibodies.48 Other adverse events that were associated with alemtuzumab included fatigue, infection, and skin disorders, and those associated with natalizumab included infusion reactions, hypersensitivity, and skin disorders.

Patient input provided specifically for this Therapeutic Review suggests persons with MS prefer oral agents to injectable agents. Their preference stems from a variety of reasons — including anxiety associated with needles, issues with rotation of sites, cannot use a needle because of coordination issues, side effects (injection site reactions, lipoatrophy, and bruising on the skin), and inconvenience with refrigeration/travel. The findings in this review suggest that the three oral agents had similar activity in sustained disability progression, while results were more favourable for fingolimod and dimethyl fumarate compared with teriflunomide regarding ARR. These oral drugs also showed improved MRI outcomes compared with placebo. Common adverse events among the three oral agents were gastrointestinal disorders (nausea, vomiting, and diarrhea) and liver enzyme elevation. An adverse event specifically associated with fingolimod was cardiovascular disorder, typically bradycardia and atrioventricular block, while for dimethyl fumarate it was flushing (warmth and redness), and for teriflunomide it was alopecia (hair loss). Adverse events that were commonly observed with interferons were injection site reactions and influenza-like symptoms, whereas injection site reactions and hypersensitivity were commonly reported for glatiramer acetate. Liver enzyme elevation was reported more frequently for interferon beta-1a 44 mcg compared with interferon beta-1a 30 mcg and glatiramer acetate.

Few statistically significant between-treatment differences in adverse events, and no statistically significant between-treatment differences in serious adverse events, were identified. This is unsurprising given that clinical trials are frequently underpowered to identify infrequent or rare adverse events and that the identification of important safety issues may not occur until the post-market period. It should be noted that older agents such as the interferons and glatiramer have the benefit of a longer post-market period. Further, given the differences in the adverse event profiles of the available treatments, it is desirable that patient specific factors be considered in treatment selection, as suggested by patient-group input.

b. Combination therapy

Three of four combination studies assessed the efficacy and safety of adding a second disease-modifying agent to ongoing treatment. However, it was not clear that patients would be considered to have failed prior treatment. Two studies (GLANCE37 and Freedman et al.36) did not report improvements in measures of relapse or disability with the addition of a second agent, likely because of their small size and short duration; both studies did report more favourable MRI findings with add-on therapy. The SENTINEL study38 reported that the addition of natalizumab to an ongoing regimen of interferon beta-1a 30 mcg IM for at least 12 months provided additional clinical benefit in lower relapse rate, reduced risk of disease progression, and favourable MRI findings. However, the lack of a natalizumab-only arm in the SENTINEL study precluded a definite conclusion that the observed effects of the combination therapy were the result of the additive effects of two active treatments; potentially, a switch to natalizumab may have produced similar benefits to the add-on strategy. Two patients treated with natalizumab in SENTINEL developed PML, and the development of neutralizing antibodies to natalizumab with reduced efficacy was noted.

The fourth combination trial (CombiRx35) was designed to compare both glatiramer acetate and interferon beta-1a monotherapy with the combination of the two agents as initial treatment (enrolled patients were treatment-naive). As an initial treatment, the combination did not appear to be superior to either agent alone for a number of outcomes, although the ARR was 25% lower for patients treated with the combination compared with interferon beta-1a alone.

5.2.2. Pharmacoeconomic Considerations

a. Monotherapy

The results of the base case show that treatment with any of the interferon therapies, glatiramer acetate, or dimethyl fumarate dominates no treatment; i.e., treatment is less costly and more effective than no treatment. The ICUR of fingolimod versus no treatment is $18,234, and the ICUR of natalizumab versus no treatment is $121,456.

In the base case analysis, glatiramer acetate was likely to be the cost-effective treatment choice, assuming decision-maker willingness-to-pay threshold is lower than $118,242 per QALY. For willingness to pay between $118,242 and $425,655, interferon beta-1b 250 mcg (Extavia) is the cost-effective treatment. If willingness to pay is between $425,625 and $872,972, dimethyl fumarate is a cost-effective treatment. If willingness to pay is higher than $872,972, then natalizumab is the cost-effective treatment. Interferon beta-1a 30 mcg and interferon beta-1a 22 mcg are dominated by glatiramer acetate and interferon beta-1b 250 mcg (Extavia). Interferon beta-1b 250 mcg (Betaseron) is dominated by interferon beta-1b 250 mcg (Extavia). Interferon beta-1a 44 mcg is dominated by glatiramer acetate, interferon beta-1b 250 mcg (Extavia) and dimethyl fumarate. Fingolimod is dominated by dimethyl fumarate.

Probabilistic sensitivity analysis showed that there is some degree of uncertainty regarding these results, especially related to the treatment efficacy. Cost-effectiveness acceptability curves were constructed, and at a willingness to pay of $50,000, glatiramer acetate was the cost-effective treatment in 70% of replications, followed by interferon beta-1b (Extavia) in 26% of replications and beta-1b (Betaseron) in 3% of replications.

Extensive sensitivity analyses were conducted around the model input parameters and the structural uncertainty was tested. Although ICURs did vary, none of these analyses, with an exception of cost per treatment, changed the conclusions of the analysis. The emerging treatments in RRMS for which regulatory approval has not been granted (alemtuzumab and teriflunomide) were included in an exploratory analysis. Because the cost of these treatments is unknown, international prices were used as a guide, where available. The price of teriflunomide was available for the US market, and therefore the ratio between the US price and the US price of fingolimod was applied to estimate a Canadian cost for teriflunomide.

The price of alemtuzumab was not publicly available from international sources at the time of the review, and therefore it was assumed that alemtuzumab would be priced in line with the highest-cost treatment (natalizumab). Under these assumptions, teriflunomide 7 mg and teriflunomide 14 mg are dominated by interferon beta-1b (Extavia); alemtuzumab 12 mg is dominated by alemtuzumab 24 mg; and the ICUR of alemtuzumab 24 mg versus interferon beta-1b (Extavia) is $295,793. With inclusion of the emerging treatments, natalizumab, dimethyl fumarate and fingolimod are extendedly dominated by interferon beta-1b (Extavia) and alemtuzumab 24mg. Therefore, if the willingness to pay is $295,793 or higher, then alemtuzumab 24 mg is the cost-effective treatment.

Due to the differences in treatment-stopping rules across the Canadian provincial plans, a sensitivity analysis was conducted varying the EDSS score that would lead to treatment discontinuation. The base case scenario assumed a stopping rule at EDSS = 7 or progression to SPMS. The sensitivity analysis showed earlier treatment discontinuation (at EDSS = 5), as well as late discontinuation (at EDSS > 7), increased the ICURs; i.e., the optimal stopping rule for the ICUR is at EDSS = 6 to 7. As the model assumes no treatment benefit for patients who progressed to SPMS, a stopping rule without considering SPMS progression resulted in much higher ICURs.

PML has been identified by physicians and decision-makers as an important concern, and consequently the risk of PML associated with natalizumab was included in the model. The results of the analysis show that, because of the low rate of PML associated with natalizumab, the scenario assuming no PML associated with natalizumab has similar results to the base case when there is a risk of 0.15% for patients on natalizumab developing PML.70

To measure the impact of the starting EDSS score, this parameter was varied in the economic model. As there was no subgroup analysis available per EDSS score, the treatment efficacy was assumed to be the same, regardless of choice of baseline EDSS score. The results showed that early treatment with the more expensive treatments leads to significantly higher ICURs.

b. Combination therapy

As there is not enough clinical evidence to support the inclusion of combination therapy in the health economics model, the cost-effectiveness of combination therapy in RRMS remains unknown.

5.3. Strengths and Limitations of the Systematic Review

5.3.1. Strengths

This systematic review was conducted according to a pre-specified protocol, using standard approaches for collecting evidence, performing data extraction, quality assessment, and analysis. This review included currently available and emerging treatment agents of different classes for RRMS. The evidence was analyzed and presented using both direct pairwise meta-analyses and an NMA. The robustness of the NMA was supported by the similarity between the results of the indirect comparison and those of the pairwise comparison. Selected meta-regression and subgroup analyses were conducted to explore heterogeneity, and demonstrated the robustness of the findings in the reference case analysis. A comprehensive economic evaluation was conducted using available cost data and the results of the NMAs.

5.3.2. Limitations

a. Clinical limitations

In addition to the aforementioned strengths, key limitations of the review are related to the availability of data and the suitability of available data for pooling. As previously noted, we identified no trials specifically designed to assess comparative efficacy and safety of disease-modifying treatments in patients who had failed, or were intolerant to, previous treatments. In addition, a number of outcomes of particular interest to patients were not widely captured in the included trials. These include fatigue, difficulty walking, memory or attention problems, and impact on work life. Fatigue was captured as an adverse event in a number of trials, rather than assessed with a valid sleep scale. Difficulty walking and memory or attention problems may be captured as components of the MSFC (as the timed 25-foot walk test and the paced auditory serial-addition test, respectively). However, only five trials included the MSFC as an outcome, and only a global score for the MSFC was reported; the trials did not report the components separately. Patient-group input suggests that there is considerable inter-patient variability in MS symptoms and thus patients desire to have numerous treatment options available.

Another data limitation was the general lack of data stratified by subgroups of interest. Only four trials (CAMMS223,13 CONFIRM,18 FREEDOMS,22 and TEMSO32) reported subgroup analyses of clinical efficacy outcomes including ARR and sustained disability progression. All four studies reported consistent effects across subgroups (defined by age, gender, EDSS score, prior relapse, and GdE lesions), with no evidence of effect modification.

NMAs could not be conducted on MRI and safety outcomes. The evidence networks of MRI outcomes were relatively unstable because of a sparse connection between treatments, and the MRI populations in many studies were subsets of patients with unclear selection criteria for MRI scans. For safety outcomes including death, serious adverse events, and treatment discontinuation because of adverse events, the low frequency of events precluded NMA.

NMA involves the pooling of trials. To avoid the introduction of bias, it is important that clinical and methodological variation across studies is minimized. If variability does exist, the assessment of its effects on NMA results is required. We observed between-trial variability in both study characteristics (treatment duration, year of publication) and baseline patient characteristics (EDSS score, prior relapses, time since symptom onset, treatment history). The included studies were conducted over a 20-year time period, over which the diagnosis and treatment of MS evolved. The resultant between-trial differences in-patient characteristics may be important predictors of treatment effect. To address this heterogeneity, we performed meta-regression and subgroup analyses using patient characteristics as covariates. However, the small number of studies in relation to the number of treatment strategies may not allow for adequate control of confounding.

For a fixed-effects model to be applicable, an assumption that all the studies included in the analysis are functionally identical must be met, therefore enabling the computation of the common effects size for the identified population but restricting extrapolation to other populations. The results of the systematic review indicated variation in the characteristics of included patient populations: RRMS versus SPMS and treatment-naïve versus other. Typically, when the subjects or interventions in studies differ in ways that would affect the results, a common effects size cannot be assumed. Therefore, in these cases, the random-effects model is more easily justified than the fixed-effects model.

For analyses of ARR, there were some limitations in the reporting of data required for pooling in meta-analyses and NMA. Specifically, not all studies reported the total relapses and/or total observed person-years; several studies reported only mean ARR values as an outcome. As described in the NMA methodology, imputations were required to circumvent this issue in order to derive a best estimate of the total relapses and person-years to conduct the NMA using a Poisson distribution model. Incorporation of a mean value to impute the model inputs raises uncertainty as to the accuracy and reliability of the resulting model inputs, as it overlooks the issue of skewness of mean values in the presence of outliers. In order to address this heterogeneity, sensitivity analyses (results not shown) using only studies that reported total relapses and total observed person-years were performed; results were consistent with those obtained for the base case.

The definition of sustained disability progression differed between the trials, with the main difference being that, in some trials, the reduction in EDSS needed to be maintained for six months, whereas in other trials, a reduction sustained for three months was sufficient. For that reason, the proportion of patients experiencing disability progression was expected to vary across trials. We combined data for this outcome across all trials despite the difference in definition, based on the expectation that the relative differences between treatments would be unaffected. However, this is a potential source of heterogeneity. To examine the effect of this potential source of heterogeneity, a meta-regression analysis (results not shown) using three- and six-month time intervals for measuring sustained disease progression was conducted, which failed to detect any significant changes in the results. Nevertheless, given the between-trial differences in the definition of sustained disability and the wide credible intervals observed in the NMA, small between-treatment differences observed in the NMA should be interpreted with caution.

Finally, the subgroup analyses to examine the effect of prior treatment experience as a potential source of heterogeneity was complicated by a lack of clarity within the published reports. Several trials had clearly stated inclusion or exclusion criteria that established patients as either treatment-naive or treatment-experienced. In many trials, the included patients were a mixture of treatment-naive and treatment-experienced, which could be determined from the baseline characteristics reported. However, in numerous trials, prior history was unclear; in several instances, assumptions were made regarding treatment history based on a mixture of inclusion and exclusion criteria, the year(s) the study was conducted, and clinical expert input, although it remained unclear in many trials. Thus, our subgroup analyses by prior-treatment history were based on categorization as treatment-naive or “other.” The results of the subgroup analyses did show minor changes to the treatment effects on both ARR and sustained disease progression outcomes; however, we may not have precisely captured the effects of treatment history because of the data limitations. In line with our findings, the subgroup analyses from three studies comparing active agents (fingolimod, teriflunomide, dimethyl fumate, and glatiramer acetate) with placebo showed slight changes in ARR and sustained disability progression between patients who did and did not receive previous MS treatment. However, there were no statistically significant between-group differences for those two outcomes when categorizing based on treatment history.

b. Economic limitations

As with all economic models, a simplification of reality was necessary, and numbers of assumptions were made in this economic evaluation. It was assumed that the adverse events, except PML, were transient in nature and not associated with significant health costs or implications to quality (did not affect the ICUR), and were not included in the economic evaluation. Fixed discontinuation rate across all treatments for the first two years was assumed, followed by no discontinuation thereafter. Neutralizing antibodies were not included in the analysis because of lack of data, and confirmation from clinical experts that results are still controversial.

The pricing of the emerging treatments that are not marketed in Canada yet (alemtuzumab and teriflunomide) was not available at the time the analyses were conducted, and therefore it was assumed to be in line with international pricing. Where international pricing was not available, the price was assumed to be in line with the highest-priced drug.

Ideally, the model would use transitional probabilities derived directly from one of the large Canadian database studies, such as the London, Ontario study on natural history47 or the British Columbia study.46 However, none of these data were directly available or easily accessible, and therefore the transitional probabilities were based on published literature estimates. The data on natural history of disease were based on the published ScHARR report to NICE, which in turn was based on data from the London, Ontario longitudinal study.78

With respect to the efficacy data inputs (sustained disability progression and relapse rate) used in the model, the CADTH systematic review combined data from trials with differences in study populations, primarily as a means of allowing comparison across as many treatments as possible. The observed variability in both study characteristics (treatment duration, year of publication) and baseline patient characteristics (EDSS score, prior relapses, time since symptom onset, treatment history) may be important predictors of treatment effect. To address this heterogeneity, meta-regression and subgroup analyses using patient characteristics as covariates were performed; however, the small number of studies in relation to the number of treatment strategies may not allow for adequate control of confounding.

There is also limited clinical evidence relating to the sequential use of treatments after failure of first-line treatment or switching across treatments. The CADTH systematic review identified no trials specifically designed to assess the comparative efficacy and safety of disease-modifying treatments in patients who had failed, or were intolerant to, previous treatments. Therefore, the economic model does not assess separately the comparative cost-effectiveness between individual disease-modifying agents in RRMS in treatment-naive and experienced patients.

Finally, based on the available data, there is not enough clinical evidence to support the inclusion of combination therapy in the health economics model. Three out of four clinical studies identified in the systematic review resulted with no proven improvements in measures of relapse or disability. The forth one, the SENTINEL study,38 reported that the addition of natalizumab to interferon beta-1a 30 mcg provided improvements in relapse and disease progression; however, the lack of the natalizumab-only arm precluded a definite conclusion that the observed effects of the combination therapy were the result of additive effects of two active treatments. Therefore, the cost-effectiveness of combination therapy in RRMS remains unknown.

Copyright © CADTH March 2013.

Except where otherwise noted, this work is distributed under the terms of a Creative Commons Attribution-NonCommercial- NoDerivatives 4.0 International licence (CC BY-NC-ND), a copy of which is available at

Bookshelf ID: NBK169771


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (9.8M)

Other titles in this collection

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...