Treating Relapsing Multiple Sclerosis with Subcutaneous versus Intramuscular Interferon-Beta-1a

Treating Relapsing Multiple Sclerosiswith Subcutaneous versus IntramuscularInterferon-Beta-1aModelling the Clinical and Economic Implications

Shien Guo,1,2 Duygu Bozkaya,1 Alexandra Ward,1 Judith A. O’Brien,1 Khajak Ishak,3

Randy Bennett,4 Ahmad Al-Sabbagh4 and Dennis M. Meletiche4

1 United BioSource Corporation, Lexington, Massachusetts, USA2 Department of Health Management and Policy, School of Public Health, Saint Louis University,

St Louis, Missouri, USA3 United BioSource Corporation, Montreal, Quebec, Canada4 EMD Serono, Inc., Rockland, Massachusetts, USA

Abstract Objective: The EVIDENCE trial concluded that administering high-

dose/high-frequency subcutaneous (SC) interferon-beta-1a (IFNb1a) was

more effective in preventing relapses among patients with relapsing multiple

sclerosis (MS) than low-dose weekly intramuscular (IM) IFNb1a after

64 weeks. This analysis utilized discrete-event simulation (DES) to model the

potential longer-term clinical and economic implications of this trial.

Methods: A DES predicting the course of relapsing MS and incorporating

the effect of IFNb1a therapy was developed. The model began by randomly

reading in actual patient data from the trial to create 1000 patients.

Each simulated patient was replicated – one was assigned to receive SC

IFNb1a three times a week and the other to receive IM IFNb1a once a week.

During the simulation, patients may (i) experience relapses, with associated

short- and long-term impacts on costs and disability; (ii) develop new T2

lesions detected by a magnetic resonance imaging scan; (iii) discontinue

treatment because of adverse events or lack of response; (iv) advance to sec-

ondary progressive MS; or (v) die.

Model inputs were mainly obtained from the EVIDENCE trial, but were

taken from published literature if they could not be obtained from the trial.

Direct medical costs ($US, year 2006 values) to the US payers were primarily

obtained by updating a published cost analysis. Costs and benefits were dis-

counted at 3% per annum. Extensive sensitivity analyses were conducted to

test the robustness of the model results.

Results: Based on 100 replications of 1000 patient pairs over 4 years, SC

IFNb1a was predicted to enable more patients to avoid relapse (216 vs 147).

Total mean costs per patient (discounted) were $US79 890 with SC IFNb1a

ORIGINAL RESEARCH ARTICLEPharmacoeconomics 2009; 27 (1): 39-53

1170-7690/09/0001-0039/$49.95/0

ª 2009 Adis Data Information BV. All rights reserved.

versus $US74 485 with IM administration, a net increase of $US5405 per

patient. However, SC IFNb1a was estimated to prevent 0.50 relapses and

save 23 relapse-free days per patient, yielding incremental cost-effectiveness

ratios of $US10 755 per relapse prevented and $US232 per relapse-free day

gained. Sensitivity analyses revealed that the result was most sensitive to the

treatment efficacy, model time horizon and cost of IFNb1a treatment.

Conclusion: Based on the results observed in the EVIDENCE trial, the

model predicted that SC IFNb1a would yield greater health benefits

over 4 years than IM IFNb1a, at a cost that would seem to be a reasonable

trade-off.

Background

Multiple sclerosis (MS) is a chronic neuro-logical disorder caused by damage to the nervoussystem resulting from an immunologic processcalled demyelination.[1] In the US, about 191 per100 000 among the White population were diag-nosed with MS in 2000 and the incidence wasestimated at about 7.3 per 100 000 person-years.[2] This disease is more likely to occuramong women and onset before the age of 40years.[3] Patients are often diagnosed with a re-lapsing form of the disease at onset, characterizedby a lack of disease progression between relapses,and then gradually advance to a progressivestage, during which disability increases with orwithout occasional relapses.[1] MS is a debilitat-ing and costly illness. Approximately 50% of pa-tients require walking aids within 15 years ofonset,[4,5] and only 10% of patients remain free ofmajor disability after 25 years.[6] The annual costof treating MS (both direct and indirect costs)was about $US47 000 per patient in 2004 in theUS.[7]

Disease-modifying drugs (DMDs) only be-came available in the 1990s. Since then, they havebeen increasingly used to manage MS. The an-nual cost of DMDs per patient in the US was esti-mated at about $US16 000 in 2004, accountingfor more than half of the total annual directmedical cost.[7] Interferon beta-1a (IFNb1a) is acommonly used DMD. Studies[8-10] have shownthat IFNb1a treatment delayed disease progres-sion and reduced the relapse rate compared with

placebo among patients with relapsing MS.The EVIDENCE (EVidence of IFNb1a Dose-response: EuropeanNorthAmericanComparativeEfficacy) trial, a head-to-head comparison of therelative efficacy of two different administrationregimens of IFNb1a, demonstrated that high-dose, high-frequency subcutaneous (SC) INFb1a(Rebif�, 44 mg, three times a week) was more ef-fective in reducing relapses and magnetic re-sonance imaging (MRI) lesions among patientswith relapsing MS than low-dose intramuscular(IM) IFNb1a (Avonex�, 30 mg, once a week) at24, 48 and 64 weeks.[11,12]

Despite the fact that the evidence supportingthe clinical benefits of SC over IM IFNb1a hasbeen based on rigorously designed clinical trials,longer-term clinical outcomes in actual practicesettings are still not fully understood. In addition,several economic studies have been conducted toassess the cost effectiveness of IFNb1a.[13-15]

However, they were never based on clinical trialswith a head-to-head comparison of two activeIFNb1a treatments and most of them had majormodelling flaws, such as failure to model thenatural history of the disease and incorporateboth mortality and transition to secondary pro-gressive MS in long-term treatment models.[13]

The goal of this model was to assess the potentiallonger-term clinical and economic consequencesof IFNb1a treatments in patients with relapsingMS based on the data from the EVIDENCE trial.Findings from this model are expected to assistdecision makers in determining whether higher-dose SC IFNb1a is a cost-effective treatment for

40 Guo et al.

ª 2009 Adis Data Information BV. All rights reserved. Pharmacoeconomics 2009; 27 (1)

patients with relapsing MS compared with lower-dose IM IFNb1a.

Methods

Model

An economic model predicting the courseof relapsing MS and incorporating the effect ofIFNb1a therapy was developed and implementedas a discrete-event simulation (DES).[16] Al-though Markov modelling is the most commonlyused technique in pharmacoeconomics and is alsosuitable for this assessment, DES was chosen inthis analysis for the following reasons. First, itpermits realistic handling of the time-dependentaspects of this illness (e.g. relapse and transitionto progressive MS). Second, DES has the cap-ability to predict the risks of events based on apatient’s characteristics and past history, handlethe multiple competing risks of events faced by anindividual patient in a temporal order and dealwith the great variations in treatment responseand disability progression, which are also appar-ent in the occurrence of relapses, relapse severityand duration of relapse. More importantly, DESis more flexible and able to answer complicatedclinical and economic questions and test varioustreatment scenarios (e.g. various clinical criteriaused to stop treatment and treatment switch ifnot responding to treatment) with fewer un-realistic assumptions or oversimplifications ofthe disease model than Markov modelling, whichforces a disease into a few mutually exclusivestates within a fixed time. These advantages ofDES have also led other researchers to select thistechnique for health economic studies in a varietyof disease areas.[17-22]

Model Concept

Figure 1 shows key parameters and their re-lationships within this model. The key modeldriver is the use of IFNb1a treatment (either SCor IM), which influences the occurrence of re-lapses, treatment-related adverse events and newor enlarging MRI T2 lesions. When a relapseoccurs, a short-term cost of managing the relapseis incurred and long-term disability, measured

with the expanded disability status scale (EDSS),may be aggravated if the recovery is incomplete.This subsequently increases the cost of managingthe disease during remission as it is determinedbased on a patient’s EDSS level.

Both adverse events and treatment non-response lead to treatment discontinuation. Lackof response can be determined by the followingcriteria individually or in combination as a com-posite outcome, depending on the scenario that isselected by the analyst: (i) reduction in relapserate <35% compared with the pre-treatment rate;(ii) a relapse requiring hospitalization; (iii) a re-lapse with severe effect on activities of daily liv-ing; (iv) a relapse with severe short-term impacton EDSS (i.e. change in EDSS >2 points); or (v) anew or enlarging T2 lesion detected by an MRIscan.[23] When treatment is stopped, the risk ofrelapse is re-adjusted to pre-treatment risk and thecost of INFb1a treatment is no longer accrued inthe model. Finally, as the model focuses only onthe IFNb1a treatment during the relapsing-remitting phase of MS, transition to secondaryprogressiveMS is also simulated with the assump-tion that it is not affected by IFNb1a treatments.If this event occurs, treatment is discontinued andthe patient is forced to exit the model after ac-cruing a long-term management cost between theevent and the end of the simulation or death,whichever occurs first.

IFNb1a treatment(SC vs IM)

Treatmentnon-response

MRI T2lesion

Adverseevent

Relapse

Treatmentdiscontinuation

Secondaryprogressive MS

Long-termcost of

managing MS

Permanentchange in

EDSS

Cost ofmanagingrelapse

Fig. 1. Model concept. EDSS = expanded disability status scale;IFN = interferon; IM = intramuscular; MRI =magnetic resonance im-aging; MS =multiple sclerosis; SC = subcutaneous.

Modelling IFNb1a in Multiple Sclerosis 41


Model Flow

A schematic representation of how the modelis simulated from the start to the end is shown infigure 2. At the beginning of the simulation, 1000patients are created by reading in real patientprofiles directly from the EVIDENCE trial da-taset. Each patient profile includes age, sex,EDSS, duration of disease and number of re-lapses over the past 2 years before treatment in-itiation. Based on each patient’s characteristics,risks and times of events such as relapse, sec-ondary progressive MS and death are estimatedand assigned. Then a copy of each simulated pa-tient is created: one ‘clone’ is assigned to receiveSC IFNb1a, while the other is assigned to receiveIM IFNb1a. This cloning step resembles a perfectrandomization where the treatment cohortscomprise exactly the same patients. Duplicatingpatients before the treatment step ensures thatfactors other than treatment do not introduceadditional variance to the model. Relevant risks(i.e. relapse, T2 lesion and adverse events) arethen updated based on the treatment received.

During the simulation, the next event of each si-mulated patient is identified and its consequences,such as resource consumption and re-adjustmentof event risks, are processed accordingly. Beforethe event is processed, model outcomes (i.e. treat-ment costs and relapse-free days) accruing duringthe period between events are accumulated first.If the event is transition to progressive MS, deathor end of model time, patients are forced to exitthe model after recording relevant data on out-comes. Otherwise, patients continue the simula-tion by proceeding to the next event until any ofthese three events occur.

Model Parameters and Data Sources

Data used to populate this model were primarilyobtained from the EVIDENCE trial.[12,24,25] Itrandomized 677 patients with relapsing MS to re-ceive either SC (n= 339) or IM (n= 338) IFNb1aand followed them for a mean of 64 weeks. To beincluded in the study, patients needed to be agedbetween 18 and 55 years, have an EDSS score from0 to 5.5 inclusive, and have experienced two ormorerelapses within the preceding 24 months. Modelinputs that could not be covered by this trial wereeither derived from the PRISMS (Prevention ofRelapses andDisability by IFNb1a Subcutaneouslyin Multiple Sclerosis) study[8,9] or supplementedwith data from published literature. The risk equa-tions, model parameters and their correspondingvalues and sources for the base-case analysis aresummarized in tables I–IV. Detailed information isdescribed below.

Model Settings and Baseline Characteristics

Each replication simulated 1000 pairs of pa-tients over a 4-year timeframe and discounted thecosts that were incurred beyond the first year by3% annually.[26] This timeframe was chosen be-cause prediction beyond 4 years is less likely to besupported by projections based on the trial with amean follow-up of 64 weeks and it is more con-sistent with many US payers’ views of a realistictime horizon. In addition, new IFNb1a treat-ments (e.g. oral administration) are likely to be-come available during this time span, making the

Create patients· Read in patient profiles· Estimate and assign baseline risks

Y N

Clonepatients

SC IFNb1a

IM IFNb1a

Updaterisks andtimes toeventsbased ontreatment

Search event and accrueoutcomes:· Find the next event: i.e. relapse, death, SPMS, AE, T2 lesion, MRI scan, update of relapse risk, and end of model· Accrue treatment costs, relapse-free days

Process events:· Update relevant patient characteristics and risks· Re-estimate and assign time to next event(s)· Accrue associated costs

Death,SPMS or end of

model?

Recordoutcomes andexit model

Fig. 2. Schematic representation of model flow.AE=adverse event;IFN= interferon; IM= intramuscular; MRI=magnetic resonance im-aging; MS=multiple sclerosis; N=no; SC= subcutaneous; SPMS=secondary progressive multiple sclerosis; Y= yes.

42 Guo et al.


findings of this analysis less relevant for decisionmaking if the model timeframe is too long.

Patient characteristics were obtained from allpatients participating in the EVIDENCE trial.These patients were mostly female (75%) with amean age of 38.3 and 37.6 years for males and fe-males, respectively. Patients had amean EDSS levelof 2.3 (median= 2.0), disease duration of 6.6 years(median= 4.0) and 2.6 (median= 2.0) relapses overthe 2 years leading up to enrolment in the trial.

Relapses and Relapse Profiles

The pre-treatment relapse hazard was esti-mated based on the number of relapses occurringover the preceding 2 years. The post-treatmentrelapse hazard was obtained from analysis of theEVIDENCE trial data using a Poisson regressionapproach to account for the effects of baselinecharacteristics and treatment. The coefficients ofthese predictors and the arithmetic expression forthe post-treatment annual relapse hazard areprovided in table I. Because the risk of relapsedecreases as disease duration increases,[27] thepost-treatment relapse hazard was assumed to

start decreasing by 13% each year at the end ofthe 16th month (i.e. the mean follow-up of theEVIDENCE trial). This rate was estimated basedon the 5-year relapse data from the placebo groupin a clinical trial.[28]

In order to simulate the course of a relapse,two libraries of actual relapse profiles were cre-ated: one included 477 relapses recorded inthe EVIDENCE trial and the other consisted of1174 relapses recorded in the PRISMS trial. Eachlibrary contains a unique set of attributes re-garding a relapse. Table II shows these attributeswithin a profile and their descriptive statistics.When a relapse occurs, a relapse profile fromeach of the libraries is randomly selected based onthe patient’s EDSS prior to the occurrence of therelapse and assigned to the patient. The modelprocesses the consequences of the relapse ac-cording to the relapse profiles assigned.

Magnetic Resonance Imaging (MRI) T2 Lesionsand MRI Scans

The risk of developing a new or enlarging T2lesion during a specific time interval was also

Table I. Risk functions used in the model

Parameters Risk functions Source

Annual post-treatment relapse hazard Exp ð�7:448þ 0:231� femaleþ 0:100

�EDSS level� 0:019� disease durationþ 0:226

� prior relapses in 2 yrsþ 0:296� IM IFNb1aÞa

EVIDENCE

Annual hazard of transition to secondary

progressive MS

Male: 0.064 transitions per patient year [30]

Female: 0.056 transitions per patient year

Time to next eventb t ¼ �ln ðmÞl

Coefficients for estimating

time to death

Male (aged 18–50 y)c: l= -8.37256; g =0.06343 [31]

Male (aged ‡51 y): l =-9.60288; g = 0.09080

Female (aged 18–50 y): l =-8.81371; g =0.05690

Female (aged ‡50 y): l= -10.54173; g =0.09898

Time to death t ¼ lnð½Kage � lnðmÞ�gÞ� ll

d

a Female =1 and male = 0; IM IFNb1a =1 and SC IFNb1a = 0.

b The equation is applied to obtain the time to event (i.e. relapse and transition to progressive MS) assuming a constant hazard, where u is a

random number between 0 and 1; l is the annual hazard.

c l is the natural logarithm of the ‘base hazard’ or hazard at time zero and g is the acceleration parameter.

d Definitions: t + (18 or 51) gives the age at death; age is at the time the prediction is made; u is a random number between 0 and 1; and Kage

is a constant reflecting survival over the interval to the age of prediction. Kage was calculated as follows: Kage¼ egðage� 18 or 51Þþ l=g:

EDSS = expanded disability status scale; Exp = exponential; IFN = interferon; IM = intramuscular; MS =multiple sclerosis; SC = subcutaneous.



estimated using the EVIDENCE trial data(table III). To simulate the occurrence of a newlesion beyond the trial period, the risk based onthe last 6 months of the trial was calculated andassumed to remain constant throughout the modeltime horizon. In order to detect the occurrenceof a T2 lesion, anMRI scan is needed. In the basecase, a scan was assumed to be performed imme-diately after the occurrence of a relapse.

Discontinuation Due to Non-Responseor Adverse Events

In the base case, patients were assumed to benon-responders if (i) the reduction in the annualpost-treatment relapse rate was <35% compared

with the pre-treatment rate; or (ii) a new or en-larging T2 lesion was detected. To be assessed fortreatment response, patients also needed to con-tinue with treatment for at least 18 months. Thisgives patients sufficient time for treatment beforebeing assessed for treatment response.

The EVIDENCE trial showed that the rate ofnon-persistence due to adverse events was slightlyhigher with SC than with IM IFNb1a. Theserates were used to simulate treatment disconti-nuation due to adverse events. In the first 16months of the simulation (equivalent to the meanfollow-up of the trial), differential risks wereapplied in different time intervals for differenttreatment groups (table III). After that, an an-nual risk of 1.2% based on the risk estimatedfrom the data collected in the last 6 months of thetrial was used for both treatments and assumed toremain constant until the end of simulation. Itshould be noted that adverse events that did notinterfere with a patient’s ability to continue treat-ment were not considered in this analysis.

Non-persistence due to lack of response or toadverse events leads to a re-adjustment of therelapse hazard to the pre-treatment rate. Becauseit has been suggested that there may be a re-fractory period of at least 6 months before theremaining effect of IFNb1a fully disappears,[29]

Table II. Summary statistics of relapse profiles

Attributes of relapse profilesa Value

(%)bSource

Impact on EDSS during relapse EVIDENCEc

mild (EDSS change 0–0.5) 29.6

moderate (EDSS change 1–2) 52.9

severe (EDSS change >2) 18.2

Impact on activities of daily living during

relapse

EVIDENCE

mild (little or no effect) 62.1

moderate (significant impact) 30.4

severe (hospitalized) 7.5

Level of medical management required EVIDENCE

mild (no treatment needed) 56.8

moderate (outpatient care) 35.6

severe (hospitalized) 7.5

Mean length of hospital stay (days) 4.39 EVIDENCE

Mean duration of a relapse (days) 46.9 PRISMSd

Increase by EDSS level after a relapse PRISMS

0 65.6

0.5–1.5 28.3

‡2 6.1

a Please note that although summary statistics are given here, the

model, in fact, uses patient-level data from trials to randomly

assign a relapse profile consisting of these attributes when

patients experience a relapse.

b Unless otherwise stated.

c Relapse profiles consist of 477 relapses from SC and IM IFNb1a

treatment groups.

d Relapse profile includes 1174 relapses from all treatment groups.

EDSS = expanded disability status scale; IFN = interferon; IM = intra-muscular; SC = subcutaneous.

Table III. Model inputs[12]

Model parameter Base-case values

SC IFNb1a IM IFNb1a

Percentage developing at least

one T2 lesion during the trial

21 62.2

Among them, percentage with a

new or enlarging lesion in

each interval

0–6 mo 69.3 73.3

7–12 mo 54.0 55.5

13–16 mo 20.4 25.3

Annual risk post-16 mo 14.3 22.9

Percentage discontinuation due

to adverse events by time interval

0–6 mo 3.2 0.9

7–16 mo 1.8 2.3

‡16 mo (annually) 1.2 1.2

IFN = interferon; IM = intramuscular; SC = subcutaneous.

44 Guo et al.


this re-adjustment was assumed to take place6 months after treatment discontinuation.

Secondary Progressive Multiple Sclerosis and Death

The hazards of transition to secondary pro-gressive MS were derived from a published studythat prospectively followed a newly diagnosedcohort of 210 patients with relapsing MS (137females, 73 males).[30] The study reported that atleast three-quarters (76% of females and 80% ofmales) of the patients advanced to progressiveMS within 25 years of disease; these rates wereconverted into constant annual hazards for bothsexes (table I). Transition to progressive MS wasassumed to be independent of IFNb1a treatment.Life expectancy was predicted conditional on thepatient’s age and sex. Death hazards based on the2003 US Life Table were fitted to Gompertzfunctions for each sex and used to project time todeath[31] (table I). Life expectancy was assumedto be independent of IFNb1a treatment and MSitself, as suggested by Poser et al.[32]

Costs

The analyses were conducted from the per-spective of healthcare payers in the US. Thus,only direct medical costs ($US, year 2006 values),were considered. Indirect costs, such as loss ofproductivity or caregiver time, were excluded.Cost items included in the analysis are shown intable IV. The costs of IFNb1a treatment werebased on the average wholesale price obtained fromthe 2006 Red Book,[33] but an additional cost dueto drug administration was added to the cost ofIM IFNb1a treatment by assuming 12% of patientsrequired clinician-administered injections. Thecosts of managing the disease during relapseand remission were obtained by updating a pub-lished database analysis.[34] As management costduring remission is likely to increase by disabilitylevel, the estimated amount of $US367 per yearderived from the analysis was applied only topatients with EDSS scores between 0 and 1. Costratios obtained from a published study[35] wereused to estimate the cost for patients with in-creased EDSS levels. An average annual cost ofmanaging patients with progressive MS wasobtained from a cross-sectional study that esti-mated the direct medical cost based on 1909patients with MS.[7]

Analyses

Model outcomes, including the proportion ofpatients remaining relapse free, relapse-free days,EDSS, non-persistence with treatment due toadverse events and lack of response, T2 lesions,transition to progressive MS, survival and costs,were estimated for each treatment group. Incre-mental cost-effectiveness ratios (ICERs) basedon relapse-free days gained and relapse avoidedwere calculated. Extensive one-way sensitivityanalyses were performed to examine the effectsof changes in key model parameters. Several cri-tical scenario analyses, including clinical crite-ria used to determine suboptimal response andswitch to other DMDs (i.e. natalizumab) if notresponding to the treatment, were also conduc-ted. Furthermore, probabilistic sensitivity ana-lyses taking into account statistical uncertainty

Table IV. Model costs (year 2006 values)

Costs $US

IFNb1a treatment[33]

SC IFNb1a (per wk) 401

IM IFNb1a (per wk) 371

natalizumaba (per mo) 2 696

Management of relapses by severity (per event)[34]

severe (hospitalized) 24 610

moderate (outpatient care) 2 732

mild (medications) 305

Routine management during remission by

EDSS (per y)[34,35]

0–1 367

1.5–2 503

2.5–3 82

3.5–4 826

4.5–5 1472

‡5.5 2 547

Average cost of managing progressiveMS (per y)[7] 31 626

MRI scan (per scan)[34] 987

a The cost included the cost of administration. As natalizumab was

not available in the market until late 2006, its cost was actually

based on the 2007 price.

EDSS = expanded disability status scale; IFN = interferon; IM = intra-muscular; MRI =magnetic resonance imaging; SC = subcutaneous.



were performed by allowing several key suchparameters to simultaneously vary within a rangeof plausible values defined by a specific distri-bution (table V). To facilitate interpretation ofthese ICERs, a cost-effectiveness acceptabilitycurve was constructed using the results ofthe probabilistic sensitivity analyses. It displaysthe cumulative frequency of the ICER for SCIFNb1a treatment being below a given ICERthreshold.

Validation

Face validity of the model was evaluated bypresenting the model logic to expert cliniciansinvolved in the treatment of MS in the US. Tech-nical validation was performed by checking theresults for logical consistency after setting modelinputs to extreme values and comparing the re-sults with those from the EVIDENCE trial toensure the model faithfully replicated the trialoutcomes. Predictions at 64 weeks, the meanfollow-up of the EVIDENCE trial, showed re-sults similar to those of the trial: 54% versus56% (model vs trial) of SC IFNb1a, and 45%versus 48% of IM IFNb1a users remained relapse

free, and the annual relapse rate was 0.51 versus0.54 with SC IFNb1a and 0.66 versus 0.65 withIM IFNb1a users. It should be noted that theseminor differences arise because the model popu-lations are randomly sampled from the patientsparticipating in the trial, thus both groups arenot identical, and although the mean follow-upof the EVIDENCE is 64 weeks, the follow-uptime is different between treatment groups (64.3weeks with SC vs 63.3 weeks with IM). Withregards to relapses, 57% required no treatment,36% an outpatient visit and 7% a hospital stay,almost identical to the trial results (57%, 36% and8%, respectively).

The simulation was carried out using ARE-NA�,[36] a software package that facilitates DEScoded in SIMAN language.

Results

Base Case

Table VI summarizes the results of the base-case analysis. Based on 100 replications of 1000patient pairs over 4 years, the model predictedthat SC IFNb1a would reduce the occurrence of

Table V. Ranges for parameters varied in probabilistic sensitivity analyses (PSA)

Parameter Base case Range for PSA

Decrease in relapse hazard each year after trial time 13.4% per year TRIA (4.4%, 13.4%, 22%)a

Duration of residual treatment effect on relapse hazard after discontinuation 6 mo UNIF (6,12)b

Treatment efficacy (IM vs SC IFNb1a) 0.296 NORM (0.296, 0.091)c

Costs ($US; year 2006 values)

Management of relapse (per relapse)

severe 24 610 NORM (24610, 2 511)d

moderate 2 732 NORM (2732, 279)d

mild 305 NORM (305, 31)d

Routine MS management during remission by EDSS (per y)

0–1 367 NORM (367, 37)d

MRI scan 987 NORM (987, 101)d

a The lowest and highest rates for a specific year during the study were used.

b The maximal value was assumed.

c The SD was derived from the regression model.

d The SD was derived from an assumption that the upper and lower bounds of 95% CI are equivalent to –20% of the mean.

EDSS = expanded disability status scale; IFN = interferon; IM = intramuscular; MRI =magnetic resonance imaging; MS =multiple sclerosis;

NORM =Normal distribution (mean, SD); SC = subcutaneous; SD = standard deviation; TRIA = triangular distribution (minimum, mean,

maximum); UNIF = uniform distribution (minimum, maximum).

46 Guo et al.


relapses by 21% and increase the number ofpatients remaining relapse free by 47%. Theseimprovements resulted in a smaller proportion ofpatients with disability progression, measuredwith EDSS, and a greater gain in relapse-freedays. In addition, SC IFNb1a would decreasethe number of non-responders by 34% and thenumber of patients with new or enlarging T2lesions by 34% to 47%, depending on the timeinterval. However, SC IFNb1a was predicted to

increase the number of patients stopping treat-ment due to adverse events (70 vs 50). Becausetransition to progressive MS and death were as-sumed to be independent of treatment, there wasno difference in these two endpoints.

The model estimated that the total directmedical cost per patient (discounted) over 4 yearswas $US79 890 with SC IFNb1a versus$US74 485 with IM IFNb1a (table VI). Theuse of SC IFNb1a over 4 years was predicted to

Table VI. Model outcomes predicted over 4 years (mean of 100 replications)

Outcomes Per 1000 patients Improvement (SC vs IM)

SC IFNb1a IM IFNb1a absolute relative (%)

Number of relapses 1 882 2 385 -503 -21

Relapse-free patients 216 147 69 47

Relapse-free days 1 210000 1 187000 23 000 1.96

Treatment discontinuation 367 498 131 -26.3

non-response to treatment 297 448 -151 -34

adverse events 70 50 20 40

MRI T2 lesions

0–6 mo 283 446 -163 -37

6–12 mo 210 324 -114 -35

12–15 mo 76 143 -66 -47

post-trial 270 407 -137 -34

Disability, increase in EDSSa by

0 550 470 80 17

0.5–1 220 240 -20 -8

1.5–2 130 160 -30 -16

‡2.5 100 130 -30 -23

Progression to SPMS 207 207 0 0

Death 9 9 0 0

Total cost ($US per patient; 2006 values)

Undiscounted 84 514 78667 5847 7.4

IFNb1a 61 047 53149 7898 14.9

relapse 5 649 7 234 -1585 -21.9

remission and MRI 4 185 4651 -466 -10.0

SPMS 13632 13632 0 0

Discounted 79890 74485 5405 7.2

IFNb1a 58005 50658 7347 14.5

relapse 5 328 6 825 -1 497 21.9

remission and MRI 3 956 4401 -445 10.1

SPMS 12601 12 601 0 0

a EDSS is measured at the time when patients exit the model (i.e. end of model time, death and transition to progressive MS). Please also

note that the model simulates change in EDSS only due to incomplete recovery of relapses. Disability progression during remission was not

considered in this model. The higher the EDSS, the greater the disability.

EDSS = expanded disability status scale; IFN = interferon; IM = intramuscular; MRI =magnetic resonance imaging; MS =multiple sclerosis;

SC = subcutaneous; SPMS = secondary progressive multiple sclerosis.



increase the cost of IFNb1a treatment by$US7347 per patient. Such an increase was par-tially offset by avoiding utilization of medicalresources for managing the disease during bothrelapse and remission periods, resulting in a netincrease of $US5405 per patient. In return, theuse of SC IFNb1a was predicted to avoid 0.5 re-lapses and increase relapse-free days by 23. Theseyielded ICERs of $US10 755 per relapse avoidedand $US232 per relapse-free day gained.

Sensitivity Analyses

Univariate Sensitivity Analyses

Results of the one-way sensitivity analyses arepresented in figure 3. The ICER based on relapseavoided was most sensitive to treatment efficacy,model time horizon, cost of IFNb1 treatment andtreatment duration before assessing response. Itwas less sensitive to the changes in baseline chara-cteristics, cost items (except cost of IFNb1a), riskof discontinuation due to adverse events, percentdecrease in annual relapse hazard (due to naturalprogression of the disease), discount rate andduration of the residual effect of IFNb1 treat-ment. Results of the sensitivity analyses based onICER per relapse-free day (not shown) were verysimilar to those based on relapse avoided.

Scenario Analyses

The model outcomes were also greatly influ-enced by the changes in the criteria used to assesstreatment response. The ICERs were the lowest,with $US3913 per relapse avoided and $US84per relapse-free day when none of the five criteriawas used (i.e. no treatment discontinuation dueto lack of response), and were the highest whenthe base-case scenario was used among all thepossible combinations being tested. If switchingto natalizumab was allowed for all patients withMS that did not respond to the initial treatment,as defined by the criteria used in the base-caseanalysis, the ICERs decreased substantially to$US5746 per relapse avoided and $US128 perrelapse-free day gained, assuming that relapsehazards prior to switching were maintained bythe new treatment. The ICERs was further re-duced to $US4960 per relapse avoided and$US110 per relapse-free day gained if patients notresponding to IM IFNb1a were assumed to beswitched to SC IFNb1a and vice versa.

Probabilistic Sensitivity Analyses

Based on 1000 replications of 1000 patientpairs, the probabilistic sensitivity analyses pro-jected ICERs ranging from $US6500 to

0 5 10 15 20 25 30

Cost per relapse avoided ($US × 1000; year 2006 values)

Treatment efficacy (±95% CI) 0.11770.474

Age All ≤40 y All > 40 y

Duration of residual IFNb1a 12 mo 0 mo

Cost of remission All = cost of EDSS 1 All levels increased 2-fold

Baseline disease duration All >2 All ≤2

Baseline EDSS All >2 All ≤2

Percentage drop in relapse rate annually 0% 20%

Cost of moderate relapses Increased 2-fold Decreased 2-fold

Adverse event rate All = IM IFNb1a rate All = 0%

Cost discount rate 6% 0%

Sex All female All male

Number of prior relapses All ≥3 All <3

Time horizon 2 y 10 y

24 mo 6 moTreatment duration for assessment

Cost of IFNb1a +20%–20%

All costs (except IFNb1a) +20% –20%

Fig. 3. Results of univariate sensitivity analyses. EDSS = expanded disability status scale; IFN = interferon; IM = intramuscular.

48 Guo et al.


$US32 000 per relapse avoided, with 95% of thereplications falling below an ICER threshold of$US20 000 (figure 4) and from $US130 to$US880 per relapse-free day gained, with 95% ofthem below $US420 (figure 5).

Discussion

The primary purpose of this model was toassist clinicians or other decision makers in deter-mining whether higher-dose SC IFNb1a wouldyield greater health benefits than lower-dose IMIFNb1a at a justifiable cost in the treatment ofpatients with relapsing MS. Although the EVI-DENCE trial has shown that high-dose, high-frequency SC IFNb1a is superior to low-doseIM IFNb1a in reducing relapse and T2 lesions,it is still not clear whether these clinical benefitswould be sustained over a longer term in actualpractice and what additional cost would be nee-ded to pay for these additional benefits. Thismodel predicts that the use of SC IFNb1a pre-vents more patients from experiencing relapse,delays disability progression and increases thenumber of relapse-free days over 4 years com-pared with IM IFNb1a. The overall persistencewith SC IFNb1a is predicted to be much higherbecause of the improved response rate, evenwith more treatment discontinuations due to ad-verse events.

These clinical benefits are expected to begained at an estimated ICER less than $US20 000per relapse avoided or $US420 per relapse-freeday gained after accounting for the statisticaluncertainties in several key parameters, such asdifference in treatment efficacy and cost of IFN-b1a treatment. Although there is no acceptablethreshold available to determine whether SCIFNb1a is cost effective relative to IM IFNb1abased on these ICERs, they seem to be a reason-able trade-off, given that a relapse may result inincreased short- and long-term disability, dimin-ishing quality of life (QOL), loss of productivityand increased utilization of healthcare resources.Such a trade-off would be even more attractive ifswitching to other DMDs, because of suboptimalresponse, is considered. In fact, the latter is morelikely to be the case in actual practice as patientswould be assigned a different treatment whenfailing to respond.

Criteria to determine whether patients re-spond to their IFNb1a treatment have been pro-posed, but consensus on this issue has not yetbeen reached.[23] This analysis also assessed theimpact of various clinical criteria used, indi-vidually and jointly, on the model outcomes. Theresults show that the base-case scenario wouldyield the highest ICERs among all the scenariosbeing tested. This is because the net total costwould increase significantly as a result of agreater decrease in the cost of treatment for the

010

20

30

40

50

60

70

80

90

100

0 200 400 600 800 1000

ICER threshold (cost/relapse-free days gained, $US)

Pro

port

ion

of r

eplic

atio

ns (

%)

Fig. 5. Cost-effectiveness acceptability curve based on relapse-free day gained (year 2006 values); results of 1000 replications.ICER = incremental cost-effectiveness ratio.

0

10

20

30

40

50

60

70

80

90

100

50 10 15 20 25 30 35ICER threshold (cost/relapse avoided, $US × 1000)

Pro

port

ion

of r

eplic

atio

ns (

%)

Fig. 4. Cost-effectiveness acceptability curve based on relapseavoided (year 2006 values); results of 1000 replications. ICER =incremental cost-effectiveness ratio.



IM IFNb1a group when more IM than SCIFNb1a users would discontinue treatment dueto lack of response. It should be noted that thisresult is very likely to be different if alternativeDMD treatments for switching are allowed inthis model, as the cost of treatment would notdecrease (or decrease that much) for the IMIFNb1a group. This is shown in the scenarioanalyses. This underlying reason also explainswhy key parameters, such as treatment durationrequired before assessing treatment responseand model time horizon, have a significant im-pact on the model outcomes as they all havea considerable influence on the number ofpatients discontinuing treatment due to lack ofresponse. The use of 4 years as the base-casemodel time horizon may appear somewhat arbi-trary. If the model time horizon is set at 5 years,the ICERs are expected to increase to $US13 057per relapse gained and $US277 per relapse-free day gained, still considered to be a reason-able trade-off.

This is the first economic model implementedas a DES to assess the cost effectiveness of DMDsin the treatment of relapsing MS. The strengthsof this model include, first, realistically modellingthe course of this illness and its management fromthe relapsing to progressive phase or death,properly handling the consequences of relapsesbased on their severity and level of medicalmanagement required, and considering treatmentdiscontinuation due to adverse events and lack ofresponse. Second, the model uses patient-leveldata to create simulated patients and assign at-tributes to a relapse, ensuring that all relevantcharacteristics borne by a patient or a relapse arenaturally correlated. Such use of patient-leveldata also provides a good solution for modellinga disease with great variations in its disabilityprogression and treatment outcomes. Third, thismodel has the ability to assess the impact ofvarious criteria for assessment of suboptimalresponse to treatment and inclusion of switchingto other DMDs for patients with suboptimalresponse on the model outcomes. Finally, thismodel is based on the results of a study with ahead-to-head comparison of two active treat-ments, which avoids the common flaw of indirect

comparison observed in many health economicmodels.

Studies utilizing Markov models to evaluatethe cost effectiveness of DMDs in relapsing MShave been published.[14,15] In addition to themodelling technique, the current study is quitedifferent from the previous studies in severalways, making comparisons among these resultsimpossible. First, this study directly comparedtwo active IFNb1a treatments, while the otherscompared DMDs versus no DMD treatment orstandard care (i.e. symptom management). Inaddition, the perspective of the US healthcarepayers used in this analysis is different from thesocietal perspective employed in the previousstudies. The timeframe used in this analysis is alsomuch shorter than that of the previous analyses(i.e. 10 years or lifetime). Finally, QALYs werenot included in the current model, but were in theprevious studies.

Relapse avoided and relapse-free days, insteadof QALYs, were chosen as the main outcomes ofinterest for this analysis because the model focusesonly on two active IFNb1a treatments during therelapsing-remitting phase, which is often char-acterized by a lack of disability progressionbetween relapses.[1,37] During this phase, relapseis the primary cause of increased disability (ifrecovery is incomplete), diminishing QOL, andutilization of medical resources.[34,37,38] Thus, re-lapse-related measures would be more appro-priate than QALYs. Along with other clinicaloutcomes (i.e. T2 lesions, change in EDSS, treat-ment response), these outcome measures shouldprovide decision makers with relevant informa-tion to assess whether higher-dose SC or lower-dose IM IFNb1a should be used to treat patientswith relapsing MS. Furthermore, survival andtransition to progressive MS are not impacted byIFNb1a therapy, and relapses are not very com-mon (about one relapse per year at baseline) andonly a small proportion of them (34%) lead togreater permanent disability, as shown in table II.Therefore, QALYs will not be a sensitive enoughoutcome measure to capture the benefits ofIFNb1a therapies in patients with MS duringthe relapsing-remitting phase. This, in fact, issupported by the findings from a study assessing

50 Guo et al.


the impact of IFNb1a therapy on QOL of pa-tients with relapsingMS.[39] It indicates that QOLof patients did not change significantly over 1year of IFN1ba treatment. Finally, relapse is oneof the primary endpoints measured in the EVI-DENCE trial, directly showing the relative ben-efits between these two IFNb1a treatments.

The disadvantages of these clinical outcomemeasures are that the results of this analysis be-come difficult to compare across all the cost-effectiveness analyses of DMD treatments and tobe used in decision making for resource alloca-tion among different treatment interventions, asthese measures are disease specific. Moreover,unlike an ICER based on QALYs, the ICERsbased on relapse avoided and relapse-free daysgained do not have an explicit threshold for ac-ceptability and such a threshold is likely to varyamong stakeholders. Thus, to facilitate the inter-pretation of the results, the cost-effectivenessacceptability curve associated with each ICERwas provided – it gives the probability of theICER for SC IFNb1a being below a given ICERthreshold set by a decision maker.

Limitations

Several limitations must be addressed. First,the efficacy data were derived from a clinical trialwith a relatively short follow-up period. Project-ing long-term outcomes based on short-term datais prone to error and involves assumptions aboutthe long-term rates that have not been verified.Second, development of neutralizing antibodies(NAb+) was not considered in this analysis.However, as this model is primarily based on thedata from the EVIDENCE trial, the estimates fortreatment efficacy should reflect the influence ofNAb+. Third, the cost of managing adverseevents was not considered in the analysis. How-ever, this should not have a significant impact onthe results of this analysis since the most commonadverse events reported in the trial were injection-site reaction and flu-like symptoms, which re-quire minimal or no medical interventions.Fourth, since the analysis was performed from aUS payer’s perspective, indirect costs were notincluded. If a societal perspective is used, the

ICERs are more likely to be more favourable toSC IFNb1a as the net cost can be greatly offset bythe savings from the reduction in loss of pro-ductivity or work days due to relapses or disability.Fifth, although the majority of the model inputsare from the EVIDENCE trial, other data sourcesare also used to populate this model, increasing thelevel of uncertainty of the model results. Finally,because the patients in the EVIDENCE trial wereselected based on a set of inclusion and exclusioncriteria, the generalizability of the model results toall patients with relapsing MS is likely to be lim-ited. The findings from this analysis are more ap-plicable to patients with baseline characteristicssimilar to those participating in the trial, such asat least two relapses in the preceding 2 years andan EDSS score below 5.5, and to patients transi-tioning to progressive MS.

Conclusion

The EVIDENCE trial demonstrated that SCIFNb1a yielded greater clinical benefits than IMIFNb1a at 64 weeks. Based on the results ob-served from this trial, the model predicts thatsuch benefits are expected to be sustained over alonger term after accounting for treatment dis-continuation due to adverse events and lack ofresponse, and are expected to be gained at a costthat would seem to be a reasonable trade-off.

Acknowledgements

ShienGuo, Duygu Bozkaya, AlexandraWard andKhajakIshak are employees of United BioSource Corporation (UBC,formally Caro Research, Inc.), a consultancy that has alsoreceived grants for other unrelated research from variouspharmaceutical companies. Judith A. O’Brien was an em-ployee of UBC during the time when this study was con-ducted. This work was supported in part by a grant fromEMD Serono, Inc. (MA, USA), the maker of products po-tentially affected by this work. At the time of the study, RandyBennett, Ahmad Al-Sabbagh and Dennis Meletiche wereemployees of EMD Serono, Inc. All the authors contributedsignificantly to the intellectual content of the paper.

References1. Noseworthy JH, Lucchinetti C, RodriguezM, et al. Multiple

sclerosis. N Engl J Med 2000; 343: 938-52



2. Mayr WT, Pittock SJ, McClelland RL, et al. Incidence andprevalence of multiple sclerosis in Olmsted County, Min-nesota 1985–2000. Neurology 2003; 61: 1373-7

3. Kantarci O, Wingerchuk D. Epidemiology and natural his-tory of multiple sclerosis: new insights. Curr Opin Neurol2006; 19: 248-54

4. Weinshenker BG, Bass B, Rice GP, et al. The natural historyof multiple sclerosis: a geographically based study. I: clin-ical course and disability. Brain 1989; 112:1 33-46

5. Weinshenker BG, Bass B,RiceGP, et al. The natural history ofmultiple sclerosis: a geographically based study. 2: predictivevalue of the early clinical course. Brain 1989; 112: 1419-28

6. Kantarci O, Siva A, Eraksoy M, et al. Survival and pre-dictors of disability in Turkish MS patients. Turkish Mul-tiple Sclerosis Study Group (TUMSSG). Neurology 1998;51: 765-72

7. Kobelt G, Berg J, Atherly D, et al. Costs and quality of life inmultiple sclerosis: a cross-sectional study in the UnitedStates. Neurology 2006; 66: 1696-702

8. PRISMS Study Group. Randomised double-blind placebo-controlled study of interferon-1a in relapsing/remittingmultiple sclerosis. Lancet 1998; 352: 1498-504

9. PRISMS Study Group. PRISMS-4: long-term efficacy ofinterferon-beta-1a in relapsing MS. Neurology 2001; 26: 56(12): 1628-36

10. Jacobs LD, Cookfair DL, Rudick RA, et al. Intramuscularinterferon beta-1a for disease progression in relapsingmultiple sclerosis. The Multiple Sclerosis Collabora-tive Research Group (MSCRG). Ann Neurol 1996; 39:285-94

11. Panitch H, Goodin DS, Francis G, et al. Randomized, com-parative study of interferon beta-1a treatment regimens inMS: the EVIDENCE trial. Neurology 2002; 59 (10): 1480-1

12. Panitch H, Goodin D, Francis G, et al. Benefits of high-dose, high-frequency interferon beta-1a in relapsing-remitting multiple sclerosis are sustained to 16 months:final comparative results of the EVIDENCE trial. J NeurolSci 2005; 15: 239 (1): 67-74

13. Tappenden P, Chilcott J, O’Hagan T, et al. Cost-effective-ness of beta interferons and glatiramer acetate in the man-agement of multiple sclerosis: final report to the NationalInstitute for Clinical Excellence [online]. Available fromURL: http://www.nice.org.uk/nicemedia/pdf/msscharrreport.pdf [Accessed 2006 Sep 9]

14. Prosser LA, Kuntz KM, Bar-Or A, et al. Cost-effectivenessof interferon beta-1a, interferon beta-1b, and glatirameracetate in newly diagnosed non-primary progressive mul-tiple sclerosis. Value Health 2004; 7: 554-68

15. Bell C,Graham J, EarnshawS, et al. Cost-effectiveness of fourimmunomodulatory therapies for relapsing-remitting multi-ple sclerosis: a Markov model based on long-term clinicaldata. J Manag Care Pharm 2007; 13: 245-61

16. Caro JJ. Pharmacoeconomic analyses using discrete eventsimulation. Pharmacoeconomics 2005; 23(4): 323-32

17. Caro JJ, Guo S, Ward A, et al. Modelling the economicand health consequences of cardiac resynchronizationtherapy in the UK. Curr Med Res Opin 2006 Jun; 22 (6):1171-9

18. McEwan P, Baboolal K, Conway P, et al. Evaluation ofthe cost-effectiveness of sirolimus versus cyclosporin for

immunosuppression after renal transplantation in theUnited Kingdom. Clin Ther 2005; 27: 1834-46

19. Guest JF, Cookson RF. Cost of schizophrenia to UKSociety: an incidence-based cost-of-illness model for thefirst 5 years following diagnosis. Pharmacoeconomics1999; 15(6): 597-610

20. Cooper K, Davies R, Roderick P, et al. The development ofa simulation model of the treatment of coronary heartdisease. Health Care Manag Sci 2002; 5: 259-67

21. McEwan P, Peters JR, BergenheimK, et al. Evaluation of thecosts and outcomes from changes in risk factors in type 2diabetes using the Cardiff stochastic simulation cost-utilitymodel (DiabForecaster). CurrMedRes Opin 2006; 22: 121-9

22. Bozzette SA, Boer R, Bhatnagar V, et al. A model for asmallpox-vaccination policy.NEngl JMed 2003; 348: 416-25

23. Freedman MS, Patry DG, Grand’Maison F, et al. Treat-ment optimization in multiple sclerosis. Can J Neurol Sci2004; 31: 157-68

24. Schwid SR, Thorpe J, Sharief M, et al. Enhanced benefit ofincreasing interferon beta-1a dose and frequency in relap-sing multiple sclerosis: the EVIDENCE Study. ArchNeurol 2005; 62 (5): 785-92

25. Sandberg-Wollheim M, Bever C, Carter J, et al. Compara-tive tolerance of IFN beta-1a regimens in patients with re-lapsing multiple sclerosis. J Neurol 2005; 252 (1):8-13

26. Gold MR, Siegel JE, Russell LB, et al. Cost-effectiveness inhealth and medicine. New York: Oxford University Press,1996

27. Held U, Heigenhauser L, Shang C, et al. Predictors ofrelapse rate in MS clinical trials. Neurology 2005; 65:1769-73

28. The IFNB Multiple Sclerosis Study Group. Interferon beta-1b in the treatment of multiple sclerosis: final outcomeof the randomized controlled trial. Neurology 1995; 45:1277-85

29. Richert ND, Zierak MC, Bash CN, et al. MRI and clinicalactivity in MS patients after terminating treatment withinterferon beta-1b. Mult Scler 2000; 6: 86-90

30. Runmarker B, Andersen O. Prognostic factors in a multiplesclerosis incidence cohort with twenty-five years of follow-up. Brain 1993; 116: 117-34

31. Center for Disease Control. US life table 2003 [online].Available from URL: http://www.cdc.gov/nchs/data/nvsr/nvsr54/nvsr54_14.pdf [Accessed 2006 Nov 10]

32. Poser S, Kurtzke JF, Poser W, et al. Survival in multiplesclerosis. J Clin Epidemiol 1989; 42: 159-68

33. Drug topics red book. Montvale (NJ): Medical Economics,2007

34. O’Brien JA, Ward AJ, Patrick AR, et al. Cost of managingan episode of relapse in multiple sclerosis in the UnitedStates. BMC Health Serv Res 2003; 3 (1): 17-29

35. Grima DT, Torrance GW, Francis G, et al. Cost and healthrelated quality of life consequences of multiple sclerosis.Mult Scler 2000; 6: 91-8

36. Kelton WD, Sadowski RP, Sadowski DA. Simulation withARENA. 3rd ed. Boston (MA): McGraw-Hill, 2003

37. Lublin FD, Reingold SC. Defining the clinical courseof multiple sclerosis: results of an international survey.National Multiple Sclerosis Society (USA) Advisory

52 Guo et al.


Committee on Clinical Trials of New Agents in MultipleSclerosis. Neurology 1996; 46: 907-11

38. Lublin FD, Baier M, Cutter G. Effect of relapses on devel-opment of residual deficit in multiple sclerosis. Neurology2003; 61:1528-32

39. Zivadinov R, ZorzonM, TommasiMA, et al. A longitudinalstudy of quality of life and side effects in patients with

multiple sclerosis treated with interferon beta-1a. J NeurolSci 2003; 216: 113-8

Correspondence: Dr Shien Guo, United BioSource Corpora-tion, 430 Bedford Street, Suite 300, Lexington,MA02420,USAE-mail: [email protected]



Documents

Treating Relapsing Multiple Sclerosis with Subcutaneous versus Intramuscular Interferon-Beta-1a