Economic Evaluation of Leflunomide in Patients with Rheumatoid Arthritis Andreas Maetzel A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy, Institute of Medical Science, University of Toronto O Copyright by Andreas Maetzel, 200 1
87
Embed
Evaluation Leflunomide Patients with Rheumatoid Arthritis · 2020. 4. 8. · Economic Evaluation of Leflunomide in the Treatment of Rheumatoid Arthritis. Andreas Maetzel, Doctor of
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Economic Evaluation of Leflunomide in Patients with Rheumatoid Arthritis
Andreas Maetzel
A thesis submitted in conformity with the requirements
for the degree of Doctor of Philosophy,
Institute of Medical Science,
University of Toronto
O Copyright by Andreas Maetzel, 200 1
National Library of Canada
Bibliotheque nationale du Canada
Acquisitions and Acquisitions et Bibliographic Services sewices bibliographiques
395 Wellington Street 395, rue Wellington OttawaON KlAON4 Ottawa ON K1 A ON4 Canada Canada
The author has granted a non- L'auteur a accorde me licence non exclusive licence allowing the exclusive pennettant a la National Library of Canada to Bibliotheque nationale du Canada de reproduce, loan, distribute or sell reproduire, prster, distribuer ou copies of this thesis in microform, vendre des copies de cette these sous paper or electronic formats. la forme de microfiche/filrn, de
reproduction sur papier ou sur format electronique.
The author retains ownership of the L'auteur conserve la propriete du copyright in this thesis. Neither the droit d'auteur qui protege cette these. thesis nor substantial extracts &om it Ni la these ni des extraits substantiels may be printed or otherwise de celle-ci ne doivent Stre imprimes reproduced without the author's ou autrement reproduits sans son permission. autorisation.
Economic Evaluation of Leflunomide in the Treatment of Rheumatoid Arthritis. Andreas Maetzel,
Doctor of Philosophy, Institute of Medical Science. University of Toronto, 200 1
Abstract
Objective: To comprehensively evaluate the pharmacoeconomic properties of leflunomide, a novel
disease-modifymg anti-rheumatic drug (DMARD) to support reimbursement decisions by Canadian
provincial decisionmakers .
Methods: The evaluation was performed using two different analytical approaches. The first was based
on economic data collected prospectively in a Nonh American trial in patients with rheumatoid arthritis
(RA) randomized to placebo, leflunomide or methotrexate, and the second was based on a 5-year
simulation model comparing two treatment strategies, one including leflunomide. The model was
supported by an independently conducted survey of US and Canadian rheumatologists and a systematic
review of treatment withdrawals among patients on DMARDs.
Results: The survey confirmed methotrexate to be the treatment of choice without any clear alternative.
Analysis of the literature confirmed that patients stay significantly longest on methotrexate. Although the
economic consequences between Ieflunomide and methotrexate were similar, the higher drug costs of
leflunomide imply an extra 5 147,437 per QALY gained compared to placebo (methotrexate: 9 lZ,t26).
Methotrexate is both less expensive than leflunomide and produces higher utilities (non-significant), but
more patients on leflunomide achieve an American College of Rheumatology 20% response status at a
cost of $66,000 per responder gained compared to methotrexate. The model also showed that adding
leflunomide as a new option within a realistic sequence of DMARDs extends the time patients may
benefit from DMARD therapy at a reasonable cost-effectiveness of $10,682 per responder gained and
cost-utility of S18,474 per QALY gained.
Conclusion: This thesis shows that leflunomide has a place in the management of patients with RA and
that it comes with a reasonable economic profile.
Acknowledgments
I would like to express my deepest gratitude to my supervisor Dr. Claire Bombardier for her
continuing support, expert guidance and patience fkom the inception to the completion of this
project. I would also like to thank the thesis committee members Dr. Allan Detsky and Dr.
Murray Krahn for providing insightful and timely input throughout this research.
I would like to express my gratitude to Dr. Peter Tugwell, Dr. Vibeke Strand, Dr. George Wells,
Louise Lafortune, Dr. Anita Wong, Helena Sustackova, Nancy Ricard and Alison Baker for their
valuable assistance in part or all the projects that constitute this thesis.
The research was conducted within the Arthritis & Autoimmunity Research Centre at the
University Health Network Research Institute, which provided necessary resources and support
to make this research possible. As well, I am indebted to the Canadian Institute of Health
Research (formerly Medical Research Council) for providing the financial means through a PhD
fellowship for the last 3 years of this research.
Finally, I am indebted to my wife Simone, who has shown tremendous patience and tolerance by
bearing with me through three thesis-based degrees, and to my parents who worked hard to
provide their children with endless opportunities.
This research was supported by an unrestricted grant to the Arthritis and Autoimmunity Research
Centre om Aventis Canada hc. The terms of the contract stipulated that the authors should
retain the right to absolute control of the methods, conclusions and means of publication of the
antibodies (1 8), are some of the agents currently under investigation and showing promising
results. These agents, however, are very expensive and, according to results presented so far,
etanercept as well as IL- l RA are only marginally superior to rnethotrexate (1 9).
1.5 Economic Impact of RA in Canada
The total economic impact of musculoskeletal conditions including arthritis and
rheumatism has been studied through analyses of Canadian national and provincial health survey
databases. An earlier 1986 estimate of the total economic impact of musculoskeletal conditions
measured the total costs of all musculoskeletal conditions to be approximately 8.2 billion
Canadian dollars in 1986 (10.8 billion in 1994 Cdn. 6) , which at the time represented 2% of GIW
and 10% of all health care expenditures (20). A more recent study of the total healthcare costs of
arthritis and rheumatism puts this figure to approximately 5.9 billion dollars (1994 Cdn.S) (21).
However, only one study provides information on the direct and indirect costs of RA collected
prospectively during the years 1983 to 1994 in a Quebec and Saskatchewan population of RA
patients (22). In this study, the total annual costs incurred by RA patients amounted to 95,953
dollars (1983 - 1989) and 56,253 (1990 - 1994). The direct costs were responsible for 64%
(53,788) and 74% ($4,656) of the total for the periods From 1983 to 1989 and 1990 to 1994
respectively. Roughly multiplying the $6,253 with the 1% of the Canadian population estimated
to be affected by RA would put the total healthcare costs of RA patients in Canada to
approximately 1.75 billion dollan (1 994 Cdn.). Therefore, RA constitutes a substantial problem
in Canada in terms of burden of illness, requiring concerted efforts to find new treatments that
lead to economic savings and to evaluate the economic benefit of existing therapies.
1.6 Efficacy of Leflunomide
Leflunornide is an antirheumatic drug whose metabolite inhibits a critical enzyme in
the synthesis of pyrimidine and impedes T-cell proliferation which plays a pivotal role in the
pathogenesis of RA. It has demonstrated an anti-inflammatory effect in both in vivo and in vitro
animal models. Following a loading dose (100 mg tablet per day for 3 days), a daily dosing of 20
mg is recommended for treatment of adult patients with RA.
The efficacy and safety of leflunomide in the treatment of RA was demonstrated in
two phase-ID, randomized and placebo-controlled clinical studies, one conducted in North
America (23) and one in Europe (24). The North American trial was conducted in 482 patients
with RA in the US and Canada. Patients in this trial were assigned to 1 of 3 treatment groups in a
3:2:3 randomization: leflunomide treatment (20mg/d), placebo or methotrexate (7.5mgwk).
Patients who were not classified as responders by week 16 or who othenvise did not tolerate
therapy up to this time were allowed to switch to alternate therapy, with methotrexate and
placebo patients being allowed to switch to leflunomide, and leflunornide patients to
methotrexate. Efficacy evaluation was performed in an intent-to treat analysis that included all
0 Leflunornide (n = 182) Placebo (n = 1 18)
A Metnotrexale (n= 180) . - - - - - .
-. -
0 3 6 9 1.2 No. of Months
Figure 1.1: Intent-to-treat last observation carried forward analysis of the percentage ofpatients who met the American CotIege of Rheumatology response criteria for improvement of 20% or greater at month 1 and quarterly thereafter by treatment group. (23)
Results of the North American study demonstrate that leflunomide is statistically
significantly superior to placebo in all efficacy measures including ACR response rate and all
individual components of the ACR response criteria (tender joint count, swollen joint count,
patient and physician global assessments, pain intensity assessment, HAQ or MHAQ, and ESR
or CRP) as well as morning stiffhess and rheumatoid factor levels (selection of results presented
in Table 1.3). The treatment effect of leflunomide was evident by 1 month, stabilizing by 3-6
months, and continuing throughout the course of treatment. A similar effect was seen for
methotrexate (Figure 1.1).
Both methotrexate and leflunomide achieved statistically equivalent ACR20 response
rates, with 35% and 41% respectively being considered successful responders, i.e. sustaining
response until study end at 52 weeks (Figure 1.1). Patients on leflunomide fared better than
methotrexate patients from a functional perspective, with statistically significant differences
favoring leflunomide in the HAQ disability index and 2 of 8 subscores of the SF-36 (Table 1.3).
Leflunomide is also statistically significantly superior to placebo in retarding disease progression
as measured by x-ray analysis of both erosions and joint space narrowing but no consistent
differences were noted between leflunomide and methotrexate on assessments of joint damage.
Table 13: Efficacy of leflunomide in comparison to methotrexate and placebo in 482 patients with RA followed for 12 months. Results are presented on an intent-to-treat basis
9 Leflunomide or methotrexate vs. placebo, p s= .001; fl Leflunomide vs. methoncxate, p = -05; $ Leflunomide vs. placebo, p r = .OO1: t Methotrexate vs. placebo , p = 0.02; 11 Leflunomide vs. methotrexate, p s= -001; * Leflunomide vs. methotrexate, p s= .O 1
The European trial was conducted in 358 patients with RA. Patients in this trial were
assigned to 1 of 3 treatment groups in a 3:2:3 randomization: leflunomide treatment (20mg/d),
placebo or sulfasalazine (2gId). Efficacy evaluation was performed at 24 weeks in an intent-to
treat analysis that included all patients who received drug and had at least one follow-up visit.
Results of the European Study are only briefly summarized here, as findings from this study will
not be fUrther utilized. ACR 20% response status was achieved by 55% of patients in the
leflunomide group (p = 0.0001 vs. placebo), 21% in the placebo group, and 56% in the
sulfasalazine group (p = 0.0001 vs. placebo). Leflunomide is thus a DMARD that is similar in
efficacy to methotrexate and sulfasalazine. Its increased drug acquisition costs require a
comprehensive assessment of its pharmacoeconomic properties to appropriately inform Canadian
decision makers about the value of this new DMARD.
1.7 Objective and Overview
?'he objective of the present thesis is to provide a comprehensive analysis of the
pharmacoeconomic characteristics of leflunomide versus potential comparators in the treatment
of RA for the purposes of reimbursement applications in Canadian jurisdictions.
This phmacoeconomic evaluation of leflunomide consists of five components. The
first component (Chapter 2) is a situational assessment, i.e. a study of the prescribing behaviour
of rheumatologists in Canada and the USA. The findings of this assessment assist in the planning
of the pharmacoeconomic evaluation by describing the type of choices physicians currently make
when treating patients with rheumatoid arthritis. The study allows to quantify the relative
importance of available S W s in the treatment of RA, when they are used, how often and in
what combinations.
The second component (Chapter 3) is a review of the clinical effectiveness of the
potential comparators of leflunomide: methotrexate, gold, hydroxychloroquine and sulfasalazine.
For economic purposes it is important to know how these drugs work in real life and also how
patients fare on these drugs over longer time periods. The review includes observational studies
and randomized clinical trials and has a special emphasis on summarizing how many patients are
able to maintain therapy over time, and their reasons for withdrawal fiom therapy. The findings
from all studies are presented as a combined "survival on therapy" curve for each comparator.
The summary withdrawal rates for specific time periods can thus be used in modelling
antirheumatic therapy over a longer time period in a simulation kamework.
The third component (Chapter 4) is a direct comparison of leflunomide to the "best"
alternative methotrexate and placebo. The data used for the direct comparison are taken from the
North-American randomized controlled trial comparing leflunomide to methotrexate and placebo
in 482 patients. Pharmacoeconomic data and outcomes relevant for pharmacoeconomic
evaluations were collected during the trial. All economic items collected as part of the trial were
costed in 1999 Canadian dollars to adapt the evaluation for a Canadian societal perspective (all
costs) and the perspective of the Ontario Health [nsurance Plan (direct medical costs). Cost-
effectiveness and cost-utility ratios are reported for leflunomide and methotrexate in comparison
to placebo and to each other.
The fourth component (Chapter 5) is an indirect evaluation of leflunomide within a
simulation framework and is modelled in a decision analysis tree with two treatment arms. One
arm consists of a sequence of DMARDs that includes leflunomide whereas the other arm
incorporates a sequence of DMARDs without leflunornide. The sequence of DMARDs
represents findings fiom the survey of rheumatologists (Chapter 2). Withdrawal rates and the
percentages of patients switching to new therapy are derived fiom the clinical review of the
effectiveness of DMARDs (Chapter 3). Costs for the management of adverse events are provided
by the results of a detailed adverse event questionnaire answered by five rheurnatologists.
Efficacy data are derived from meta-analyses of the literature. Extensive sensitivity analyses was
conducted over possible ranges for all variables used in the model. Results are reported as cost
per additional ACWO responder gained and per quality-adjusted life-year (QALY) gained.
CHAPTER 2
How US and Canadian rheumatologists treat moderate or aggressive
rheumatoid arthritis: A survey
2.1 Preface
The survey described in this chapter was published in 1998 (9). The purpose of this
survey was to identify, which primary DMARDs rheurnatologists would choose in the treatment
of patients with mild, moderate and severe RA and what the alternative would be in case of
efficacy failure. It was planned to use the findings to support the order of treatments in the
strategies for the model-based economic comparison of leflunomide.
2.2 Introduction
Current guidelines recommend treating most patients with active rheumatoid arthritis
(RA) aggressively and early, possibly before joint destruction begins (25;26). As a result,
slow-acting antirheumatic drugs once reserved for the later stages of disease are now being
prescribed in the early years. However, although many patients benefit from the use of these
often toxic agents, others do not. The selection of therapeutic strategies has therefore become a
challenge, particularly as patients with milder manifestations of RA generally require a different
approach to that taken in individuals with aggressive disease.
Hydroxychloroquine, gold sodium aurothiomalate (gold), methotrexate, and
sulfasalazine are the most frequently used slow-acting antirheumatic agents. Preferences between
them have shifted slowly over recent years, with s w e y s showing an increasing trend towards
methotrexate administration effectively replacing gold as the treatment of first choice. As
recently as 1985, gold was the initial drug of choice for patients with RA recruited in
rheumatology practices in Alberta (27). Methotrexate was seldom prescribed, either as the drug
of first choice or as consecutive treatment. By 1992, a survey mailed to Canadian physicians
revealed that gold and methotrexate were equally preferable in moderate RA, but that
methotrexate was preferable to gold in aggressive RA (Table 2.1) (28). An identical survey in
Australia showed that 44% of physicians favoured sulfasalazine for the treatment of moderate
RA, and 6 1% preferred methotrexate for severely aggressive disease (29).
Table 2.1: Tabulation of results tiom surveys of rheumatologists' treatment preferences
Canada - 1985 (27) Canada 1992 (28) Australia 1992 (29)
Many physicians in Canada or the US may consider methotrexate to be the treatment
of choice for both moderate and aggressive disease, although individual factors may affect the
strength of preference (30). Several observational studies support the superiority of methotrexate,
reporting that approximately 36% of patients continue drug treatment over a period of five years
(3 1-34). Unfortunately, however, all presently available antirheumatic agents, including
methotrexate, start to fail from day 1 and will eventually need to be discontinued, whether due to
immediate toxicity in the early phases or to lack of efficacy or some other reason later on. That
fact, and the paucity of alternatives, has led to considerable attention being paid to the
investigation of novel combination treatments and the development of new slow-acting
antirheumatic drugs.
Many of the latter are now under investigation in phase II and III trials and will
require economic assessment for formulary approval. Current guidelines for the economic
evaluation of health care interventions require new approaches to be compared to current
practice. In order to appropriately position a new agent it is therefore necessary to ascertain
existing treatment preferences among rheumatologists. The present paper reports two separate
s w e y s in which the views of Canadian and US rheumatologists, respectively, were elicited on
the treatment of moderate RA, aggressive RA, and aggressive RA that fails to improve in
response to high doses of methotrexate. As well as being necessary for the purposes of economic
evaluation, the infomation gathered may also be of interest to clinicians and policy makers.
2.3 Materials and Methods
The Canadian survey was initiated in September 1996. The Canadian Rheumatology
Association supplied the names of all 295 members; 32 who were not living in Canada or who
were known not to practice adult rheumatology were excluded, leaving 263 to whom the survey
was sent by fax (n=191) or mail (n=72).
The US survey was initiated in November 1996, at which time 3594 physicians were
registered with the American College of Rheumatology (ACR) and indicated that patient care
was part of their daily routine. A 10% random sample, weighted by region, was obtained from
the ACR mailing list. It consisted of 351 physicians: 102 from the Southeastern region, 1 16 From
the Central region, 58 From the Northern region, and 75 From the Western region. Thirty-one
physicians identified as paediatric rheumatologists or out of practice, or who had moved to an
unknown location, were excluded, giving a total of 320 who were sweyed either by fau (n=268)
or by mail (n=52) during April and May 1 997.
In both studies, the questionnaire consisted of a one-page document featuring an introductory
statement, three case scenarios and some questions (Figure 2.1).
SCLVARIO 1: [Iist second line agent(s) you would prescribe, either alone or in combination] Aggressive RA, ? 38 y 1st choice: DMARD-naive 26 activeIy inflamed joints 6 erosions 2nd choice:
ESR & RF markedly elevated
SCENARIO t : [list second line agent(s) you would prescribe, either alone or in combination] Moderately progressive RA, ? 32 y 1st Choice: DMARD-naive, NSALDs not helpful 6 act~vely inflamed joints -- - a - - .. - - - -- - - - -- -
no erosions 2nd Choice:
ESR & RF moderately elevated
-- - -- - . - . - - -. - --
What percentage of your population of RA patients similar to the ? Scenario 1 Scenario t
SCESARIO 0 3: Patient from scenario 1 1st Choice: initially bcner on MTX 25 rnglwcck now loss of efficacy - - - - - - - - - -- -- - - -- - - . . - --
a 22 actively inflamed joints 2nd Choice:
7 erosions
What percentage of your RA patients are in managed care?
What percentage have their drug prescriptions restricted by formularies?
What percentage require a refernl from their primary care physician to see you?
Was your choice of agents in this survey influenced by formulary restrictions?
LLLi "/. LLL1"/0 L U I 1% CI yes 0 no
Figure 2.1: Scenarios and questions presented in both surveys. Items unique to the Canadian survey are dark shaded; items unique to the US survey are light shaded.
The fint scenario described a patient with aggressive RA, 26 actively inflamed joints
and three erosions. The patient in the second scenario had moderately progressive RA, six
actively inflamed joints and no erosions. Finally, the third example described the same patient as
in scenario 1, but with an increase in inflamed joints and erosions despite treatment with
methotrexate 25 mg. Respondents were asked for their fint and second choice of second-line
agent in each case. They were invited to be f?ee in their therapeutic choices and to consider both
single and combination therapy. Respondents were also asked about the percentages of their
patients with moderate and aggressive RA who were receiving systemic corticosteroids.
The Canadian survey differed slightly from the US one in that it omitted a specific
invitation to consider combinatior? therapy in the three patient scenarios. Instead, Canadian
rheurnatologists were asked to indicate which therapy they would use if they had been asked to
combine agents, rather than to select a single second-iine treatment. In addition, three questions
addressed exclusively to US rheumatologists inquired about the proportions of patients who were
in managed care, who had their drug prescriptions restricted by formularies, and who had
required referral from a primary care physician. Finally, US physicians were asked whether their
choice of agents as reported in the survey was influenced by formulary restrictions.
Repeat faxes or mailings of both surveys were sent twice, with Cweek intervals. The
sample size was selected in order to obtain at least 200 usable responses From each country and
to permit estimation of percentages with a maximum error of 7%. Responses to all questions in
the survey were analysed using descriptive statistics, statistical tests were not performed as there
was no plan to test specific hypotheses.
2.4 Results
Choices of therapies
The Canadian survey was returned by 23 1 physicians (response rate, 87.8%), 17 who indicated
they were not practising adult rheumatology were excluded, leaving a total of214 eligible
respondents. Of 230 US physicians who returned the survey (response rate, 7 1.7%), 16 indicated
they were not practising adult rheumatology, again leaving 2 14 eligible respondents. The
majority of Canadians worked in Ontario or Quebec and most responding US rheurnatologists
had practices in the Central or southeastern regions (Table 2.2).
Table 2.2: Number of respondents by province or region * - Canada USA
Province Frequency Percent Region Frequency Percent - - - - - -- - - --
Ontario 96 44.9 Southeastern 5 8 27. I
Quebec 48 22.4 Central 7 8 36.4
British Columbia 22 10.3 Northern 35 16.4 Alberta 19 8.9 Western 43 20.1
No answer provided: * numbers may not add to 100% due to rounding
Looking only at the first
choices of single or combination
therapy (Figure 2.2 ) both Canadian
and US rheumatoIogists reported
relying primarily on methotrexate
when treating patients with
aggressive RA and
hydroxychloroquine or methotrexate
in patients with moderate RA.
Canadian rheumatologists also
favoured gold in aggressive or
moderate RA, thereby slightly
reducing the strength of their
preference for methotrexate. In
particular, Canadians said they liked
to prescribe gold to patients who fail
adequate doses of methotrexate,
whereas no leader emerged from the
choices of US rheumatologists. Most
combinations chosen by either group
of rheumatoIogists included
methotrexate plus
hydroxychloroquine with or without
sulfasalazine; 2 1.5% of US
rheumatologists preferred a triple
combination of all three, compared
to only 3.3% of the Canadian
sample.
MTX GST HCQ SSZ Chhcn 5 ( m GST HCQ 552 Olhcn
C3n3& US
10006 Scenario 2 - Moderate RA
MTX GST HCQ SSZ Wrn 5tTS GST HCQ SSZ Othm Cam& US
Scenario 3 - Aggressive RA failing 25mg MTX
W R G!X HCQ SSZ Ochcn M I 3 GSf HCQ SSZ Men Canada US
Figure 2.2: Percentages of Canadian and US physicians choosing methotrexate (MTX), parented gold (GST), hydroxychloroquine (HCQ), sulfasalazine (SSZ) or other agents either done or in combination (hatched parts of bargraph).
US rheumatologists indicated that, on average, 66.7% of patients with RA similar to that
described in scenario 1 would receive steroids (SD: 30.5%; median: 75%; inter-quartile range:
50% - 90%; missing: 11). In comparison, the Canadian figure was 44.5% (SD: 32.2%; median:
40%; inter-quartile range: 1575%; missing: 13). For patients with moderate RA similar to
scenario 2, US rheumatologists estimated that 38.9% would receive systemic steroids (SD:
29.1 %; median: 30%; inter-quartile range: 1 O-SO%; missing: 13), compared to 16.1 % in the
Figure 3.1: Survival curves representing the percentage of patients withdrawing From each agent because of inefficacy, toxicity or other reasons. Six-month interval data were generated from studies providing withdrawal information for intervals larger than 12 months. The data are from observational studies and randomized controlled trials.
Percentage of Patients remaining on treatment Withdrawals are for inefficacy
- -- -- -- - A A -- - -- - - - --- - - - - PA - - - - - - -- -
Figure 3.2: Survival curves representing the percentage of patients withdrawing from each agent because of inefficacy. Six-month interval data were generated from studies providing withdrawal information for intervals larger than 12 months. The data are from observational studies and randomized controlled trials.
Percentage of Patients remaining on treatment Withdrawals are for toxicity
Figure 3.3: Survival curves representing the percentage of patients withdrawing from each agent because toxicity. Six-month interval data were generated from studies providing withdrawal information for intervals larger than 12 months. The data are from observational studies and randomized controlled trials.
Combined data from all published studies indicate that, after adjusting for study type and
year of publication, patients treated with SSZ are up to 1.9 times more likely to fail therapy than
patients treated with MTX (Table 3.2). Patients treated with GST are up to 2.3 times more likely
to fail therapy compared to patients treated with MTX, however when looking at failures for lack
of efficacy only, it turns out that patients treated with GST are significantly less likely to
discontinue therapy compared to patients on MTX. This finding is significant only after adjusting
for type of study and year of publication. Overall, more patients withdrew from SSZ than GST
and, compared to GST, twice as many patients withdrew from SSZ due to lack of efficacy.
However, less patients withdrew from SSZ due to toxicity.
Table 3.2: Statistical comparisons of treatment withdrawal rates between drugs by
means of the Log-Rank, Wilcoxon, and Cox's proportional hazards regression. -
- - . . . . - - - - - - - .
All reasons Log-Rank Wilcoxon
Cox's Proportional Hazard
Toxicityflnefficacy Log-Rank Wilcoxon
Cox's Proportions[ Hazard
inefficacy Log- Rank Wilcoxon
Cox's Proportional Hazard
Toxicity Log-Rank
Wilcoxon Cox's Proportional Hazard
GST vs. hITX
P .- .- - -
< .0001 c .0001
< .0001 1.40
< .0001 < -000 I < .000f 1.68
0.36
0.02
0.0077 0.73
< .ooo 1
< ,000 1
< .0001 2.28
SSZvs. MTX SSZ vs. GST
HR = h a r d ratio (see text for explanation)
The comparison of withdrawal rates between observational studies and RCTs involving
MTX, GST and SSZ revealed no statistically significant differences for up to 24 months (Figure
3.4).
GST
Obs. Studies ~ Obs. Studies . - . . ._ - _.
RCTs RCTs - Figure 3.4: Percentage of patients withdrawing due to inefficacy, toxicity or other reasons, separated by observational studies and randomized controlled trials. The comparison was limited to 24 months duration and showed no statistically significant difference in withdrawal rates between observational studies and randomized controlled trials for all three agents.
3.5 Discussion
The present study summarises rates of treatment withdrawal in published observational
studies and RCTs of MTX, GST, SSZ and HCQ. Analysis of the combined data shows that
patients stayed longest on single DMARD therapy with MTX. However, nearly two thirds of all
subjects started another therapy within 5 yean because of lack of efficacy, treatment-associated
adverse events, or other factors. When withdrawals due to lack of efficacy alone were considered,
there were more patients withdrawing £kom MTX than GST, and it was noted that a large
proportion of patients who withdrew from MTX did so for reasons other than inefficacy or
toxicity.
This is not the first investigation to make a formal attempt to review withdrawals from
treatment with selected DMARDs (48;55;56), but it is the first to combine data from
observational studies and RCTs, and to specifically compare rates of withdrawal due to lack of
efficacy, toxicity and other reasons. Wolfe noted in a review that estimates of 5-year treatment
withdrawal vary considerably, with those for MTX ranging kom 30% to 75% (57). Wolfe
combined the 5-year estimates of MTX therapy, weighting them according to the sample size of
patients at risk at baseline; however, this may not be appropriate as the precision of the survival
estimates is determined by the number of patients at risk not only at baseline but throughout the
study period. Felson et al. compared withdrawal rates between DMARDs but restricted their
review to RCTs (56). Furthermore, they compared overall withdrawal rates and those due to
toxicity, but did not calculate combined survival curves.
Some investigations that were not included in the present meta-analysis - because they
did not disclose the actual number of patients withdrawing - documented higher rates of MTX
maintenance at the 5-year interval. For example, Wolfe et al. estimated that -45% of subjects
remained on MTX at 5 years (3 l), and Pincus et al. put the figure at -65% beyond 5 years (32).
Inclusion of these data would have slightly increased the combined percentage of patients
continuing MTX for up to 5 years. In fact, the year of publication independently contributed to
survival differences for MTX and SSZ. Withdrawal rates on MTX and SSZ have therefore
declined in recent years. Year of publication was insignificant for HCQ and GST, documenting
no change in survival on therapy For these drugs over the years. This is confirmed by the
similarity of the present 5-year treatment survival estimate of 23% to those by Wolfe et al. (3 1)
and Pincus et al. (32) who noted, respectively, that 20% and 28% of patients continued GST
therapy for up to 5 years.
Even though the combined rates of survival on therapy here are obtained from several
studies, only two treatment arms each for GST (58;59) and SSZ (59;60), and three for MTX (6 1 -
63), contributed information for up to 60 months. However, survival-on-therapy curves for the
combined studies reveal a consistent pattern within the first two years and most are parallel,
suggesting that additional studies of the same agents would be unlikely to make a considerable
difference. More detailed information concerning the disease status and demographic
characteristics of patient populations would probably facilitate analysis of subgroups exhibiting
different rates of survival on therapy, but that level of detail could not, unfortunately, be
achieved. Thus, this meta-analysis can offer only a rough guide to physicians who want to know
how long a particular patient is likely to continue treatment with an individual DMARD.
A M e r limitation of the present study is the lack of adjustment for meaninghl
covariables in the statistical comparisons. Although information was sought on common
variables such as joint count, rheumatoid factor status, disease duration, and the number of prior
DMARDs, it had too seldom been collected by investigators to allow for a more detailed
analysis. Earlier studies might have included patients with more severe disease, leading to lower
treatment survival rates - particular with GST.
Examination of the underlying reasons for treatment withdrawal in the present meta-
analysis indicated that, after adjustment for year of publication, patients on GST were
significantly less likely to withdraw from therapy due to lack of efficacy than MTX patients.
These data do not reveal how many patients achieved remission or significant clinical
improvement during treatment with these two agents, but they do encourage us to take a closer
look. GST is one of the least favoured DMARDs (9), despite efficacy comparable to MTX
having been demonstrated by RCTs (64;65), on large databases (66;67), and in the meta-analysis
by Felson et al. (56) The present results suggest that toxicity is the main reason for GST being
out of favour. It is likely that most of the problems relate to the cutaneous rash that, it has been
speculated, occurs more often in patients slated for remission (55). However, GST-associated
adverse events can be managed in such a way as to avoid discontinuation of therapy (68) or
promote initiation of a second course (68-70). Furthermore, withdrawal rates reported in the
literature may obscure the true efficacy of GST, as it has been observed that patients who
withdraw from GST are more likely to experience a sustained improvement compared to patients
withdrawing fkom MTX (71). This finding may be a direct consequence of using implicit
judgment to classify patients as toxicity or inefficacy withdrawals. h this meta-analysis we had
to rely on the original author's classification of withdrawals which is often performed without
clear guidance by specific criteria, which in turn leaves the possibility for classification biases
that may artificially favour one drug over others.
It is encouraging that, contrary to current opinion, there is no apparent difference in
withdrawal rates between observational studies and RCTs. Theoretically, withdrawal rates in
RCTs could be either higher (due to protocol-mandated withdrawals or patients fearing that they
are taking placebo) or lower (because subjects participating in RCTs receive more care and
attention than is usual in a real world setting). However, although the present results show that
equal trust can be placed in reported withdrawal rates, whether they originate from RCTs or
observational studies, the relevant comparison was possible for only up to 24 months of follow-
UP*
The product limit calculation of treatment survival probabilities is a routine method
employed in the reporting of treatment success of various DMARDs. However, it is possible that
the method is often applied to data which violate some basic assumptions for its use (72). For
example, survival estimates in many older studies may be biased downwards, as these patients
generally received their first DMARD late in the disease course, but also because survival on a
specific treatment may improve as rheumatologists gain experience with it (72). This better
survival for patients will not be given its due respect when all patients are grouped together for
the purpose of the survival analysis. The results presented here will only repeat the same
potential biases that may afflict the underlying studies, unless these biases cancel each other out.
The decision to withdraw therapy is oflen implicitly made in observational studies. Those
that adopt explicit definitions of clinical response and follow common methodological reporting
standards will enhance the short-term results provided by RCTs and supply valuable information
on the long-term performance of antirheumatic therapies to physicians, patients and researchers.
CHAPTER 4
A 1-year randomized controlled triabbased economic evaluation of Leflunomide
and Methotrexate in Canadian patients with rheumatoid arthritis.
4.1 Preface
This chapter reports the economic comparison of leflunomide and methotrexate based on
data that were prospectively collected alongside the North-American phase ID randomized
controlled trial comparing leflunomide to methotrexate and placebo during a 52-week period.
4.2 introduction
The last decade has seen an intense debate about appropriate treatment strategies in the
management of patients with rheumatoid arthritis (RA). A consensus seems to have emerged that
patients with moderate or aggressive RA should be treated early and aggressively, with some
experts arguing for combination of disease modifying anti-rheumatic drugs @MARDs) as initial
therapy. However, these recommendations were made in the absence of potent therapeutic
alternatives, thus necessitating a recurrence to combination of presently available, but less potent
drugs such as sulfasalazine or hydroxychloroquine. Fortunately, in the last few years several new
medications were approved for the treatment of patients with RA. In 1998, leflunomide was the
first disease modi fylng drug approved in over a decade by the US Food and Drug Administration
(FDA). Soluble TNFa receptor and monoclonal anti-TNFa antibodies followed. These drugs add
new and powefil options to the management of RA. Compared to available therapies which may
cost as low as USS300 per year, these new agents cost between USS3,500 and USS 13,000 per
year. Thus, their therapeutic potential needs to be measured not only by their ability to slow or
halt disease progression, but also by their share in decreasing the healthcare costs incurred by
patients with RA. Reimbursement approval for these new agents will likely depend on their
demonstrated cost-effectiveness.
RA is a costly disease due to its chronic nature and sometimes severe progression that can
require hospitalization and intense medical and surgical treatment. Published estimates of the
yearly costs of RA were recently summarized by Cooper, who estimated that RA patients incur
yearly costs of US$8,416 per patient in 1996 dollan. This estimate was shared to equal parts by
direct medical and nonmedical costs and indirect costs, such as losses in productivity. The
authors documented wide variation around the cost-estimates translating into standard deviations
of nearly 53,000. Only one cost of illness study was conducted in Canada, so far. In this study the
annual costs incurred by RA patients were calculated to average CADS6,253 (1994 dollars) in
the period extending from 1990 to 1994 (22). Direct costs were responsible for 74% ($4,656) of
the total costs and prescription drugs for - 20% of the direct costs. Extrapolating these costs to
the 0.5% to 1% of the Canadian population estimated to be affected by RA leads to an estimate
of the total healthcare costs of RA patients in Canada between 0.9 and 1.75 billion dollan (1994
Cdn.).
There are few published economic evaluations of DMARDs (73-75) and only one
collected data prospectively alongside a clinical trial (75). Traditionally, economic evaluations
are made by combining information From various sources into a modelling Framework, however
most current ongoing trials incorporate economic outcomes such as healthcare and productivity
costs into their protocols. The present study reports the results of an economic comparison of
leflunornide, methotrexate and placebo based on economic outcomes measured prospectively
alongside a North American phase EI trial comparing leflunomide to methotrexate and placebo in
482 patients with RA followed for one year (76). Costs for all economic outcomes were
calculated in Camdim dollars from the perspective of the Ontario Health Insurance Plan.
4.3 Materials and Methods
The material analysed in the present study was collected as part of a 1-year randomized
double blind comparison of leflunomide, methotrexate and placebo in 482 US patients with
rheumatoid arthritis (76). Patients with active RA aged 18 years or older were eligible for
treatment if they met the American College of Rheumatology (ACR) criteria for having RA for 6
months or longer. Patients could not have previously received methotrexate, other DMARDs had
to be discontinued for at least 30 days. Patients were assigned to 1 of 3 treatment groups in a
3:2:3 randomization: leflunomide treatment (20 mg/d), placebo, or methobexate treatment (up to
15 mglwk). Patients identified on or after week 16 as non-responders according to ACR response
criteria (2) were allowed to switch to alternate therapy: methotrexate and placebo patients
switched to leflunornide and leflunomide patients to methotrexate.
Clinical data and information on costs were collected biweekly from weeks 4 to 12 and
monthly thereafter. Questionnaires examining health related quality of life and interviews
eliciting patients' utilities were applied at baseline, week 24 and 52, or at study exit. The intent-
to treat analysis included all patients who received drug and had at least one follow-up visit.
Clinical study results have been published elsewhere (76).
Utilities
Patients' utilities for their current health status at baseline, week 24, 52 or at study exit
were elicited through rating scale and standard gamble (SG) measurements. Patients were
initially asked to rate their health on a rating scale anchored by perfect health (1 00) and
immediate death (0). In the SG interview, patients were asked to make a choice between their
current health as a certain outcome or a gamble involving death and perfect health as the
uncertain outcomes. Patients were then told to imagine a treatment leading to either death or
perfect health. The chances of treatment success, i.e. perfect health, were varied systematically
from 100% downwards, and the chance of perfect health that made patients insecure as to
whether they preferred their current health state or the gamble was then recorded as the SG utility
for the patient's current health state. Utilities for patients who rated their own health worse than
death were considered 0 for the purpose of this analysis.
Costs
Treatment or disease related costs were quantified by asking the patient about: 1) any
health resources used that were not specified by the clinical protocol and that were associated
with an adverse event; 2) any expenses incurred by the patient out of her own pocket because of
the disease or treatment related adverse events; and 3) any loss in the patient or her caregiver's
productivity or the performance of social or leisure activities including costs resulting as a
consequence of a loss in productivity.
Health resource utilization questions inquired about 1) the number and duration of
hospital, intensive care unit (ICU) or nursing home stays, 2) the number of office or emergency
room visits to health professionals, 3) the number and type of outpatient diagnostic tests and
laboratory tests received and 4) the number and type of outpatient procedures performed.
Personal expense questions asked for any out of pocket costs for health services, transportation
costs, costs for meals on wheels and the type and costs of equipment purchased in order to cope
with the disease or adverse events related to the study drug. Finally, patients were asked about
the amount of work time they or their caregivers lost, the extra time incurred through performing
additional chores, the amount of lost leisure time and the costs incurred by hiring extra help to
perform daily chores.
Canadian costs were estimated for all resource utilization and cost items. Hospitalizations
were coded into case mix group (CMG) codes according to the C M G ~ 1998 Directory for use
with ~lx". Average 1998 daily hospitalization costs by CMG were then derived From a mid-size
inner-city teaching hospital's database of hospitalization costs for over 18,000 patient encounters.
These were multiplied by the patient's length of stay in order to obtain the full hospitalization
costs. Average daily hospitalization costs across all CMGs were used for three patients whose
CMG codes did not appear in the hospital's database. The physician component of the
hospitalization costs were obtained from the Ontario Health Insurance Plan (OHIP) Schedule of
Benefits (SOB), 1999 version. All hospitalizations were classified by the study physicians on
whether they were not, possibly or probably related to the drugs under investigation. All
hospitalizations rated by the study physicians as possibly or probably related were considered
related for the purpose of this analysis.
Costs of office or emergency room visits to health professionals and costs for imaging
procedures were obtained from the OHIP SOB. Costs for laboratory tests were derived &om the
Ontario Schedule of Benefits for Laboratory Services, 1999 version. A standard request including
CBC, transaminases, creatinine, BUN was used to cost performance of laboratory tests
independent of the number indicated by the patient. The OHIP SOB was also used to cost
outpatient procedures such as endoscopy, skin biopsy or pulmonary function test, assuming the
assessment conducted most frequently under routine conditions. Personal expenses incurred by
patients during the trial were converted from US to Canadian dollars using 1997 Purchasing
Power Parities published for 1997 by the Organization for Economic Cooperation and
Development. Costs for productivity time lost to the patient by age and gender were derived from
1996 Canadian Census data using the average 1995 yearly income for each age and gender group.
Costs for productivity time lost to the caregiver were calculated using the average 1995 yearly
income of the total population, as age and gender information was not provided for the caregiver.
Costs for lost leisure time were not calculated.
Costs for the routine monitoring of patients on leflunomide or methotrexate were
calculated according to monitoring guidelines published in the product monograph of each drug
in the Compendium of Pharmaceuticals and Specialties. Wholesale costs of drugs were increased
by the 10% maximum allowable markup and a prescription fee of 56.1 1 was added for each 6-
month prescription. All costs were updated to 1999 levels using the the respective portions of the
Canadian Consumer Price Index.
Analysis
Patient's utilities derived from the Feeling Thermometer and Standard Gamble
questionnaire for the baseline, week 24 and week 52 or study exit measurement were weighted
by the time period covered, according to the trapezoidal rule, and divided by the total time of
observation in order to obtain a utility per person-year of observation. Utilities between study
groups were compared using analysis of variance, with Tukey'studentized range test for painvise
comparisons.
Costs for medical procedures and visits unrelated to the disease, such as dental visit were
excluded from hrther analysis. Costs for services, tests, procedures and hospital visits (medical
costs) and patients' out of pocket costs (non-medical costs) formed the direct cost component,
patient time and productivity costs, excluding leisure time, the indirect medical cost component.
Each individual patient's direct medical costs and indirect costs were then added over the study
period and further adjusted by the total time of follow-up in order to obtain an annualized cost
estimate.
The statistical analysis of cost data is being debated in the literature, with some experts
arguing for non-parametric tests on often transformed costs, due to the heavily skewed
distributions of cost values (77), and others for parametric tests, which are thought to be
sufficiently robust with larger sample sizes (78). Several statistical comparisons were made thus
in the absence of a consensus on the appropriate statistical tests to be used in cost-comparisons
(79): 1) Student's t-test on untransformed costs to compare means between two groups; 2)
Student's t-test on untransformed costs with reference to 200 bootstrap samples; 3) Wilcoxon test
on log-transformed cost data; 4) Student's t-test on log-transformed costs; 5) a 2-Score test
statistic, proposed by Zhou et al. (80) was calculated to adjust for possible inequalities of
variances; and 6) an analysis of log-transformed costs in an ordinary least squares regression
model with adjustment for age and gender. Expected values, still on a log-transformed scale,
were then back transformed, using a nonparametric smearing estimate that takes into account the
residuals remaining from the least squares estimation (8 1).
Statistical comparisons were performed on: 1) all costs, to form a societal perspective, 2)
direct healthcare costs only and 3) direct healthcare costs excluding, arbitrarily, the top 1%
extreme outliers to perform the cost comparison
4.4 Results
01485 patients enrolled, 482 received at least one dose of a study drug or placebo and
were evaluated for safety; 480 had at least one follow-up visit to evaluate efficacy (1 82 patients
received leflunomide therapy, 180 received methotrexate therapy and 1 18 received placebo). All
patients were similar in the measured demographic and prognostic variables (Table 4.1). The
efficacy results, published elsewhere (76), are briefly recaptured here: Both leflunomide and
methotrexate achieved statistically equivalent response rates with 52% and 46%, respectively
being considered successful responders as defined by the American College of Rheumatology
preliminary criteria for 20% improvement (ACR20) (2). Furthermore, 41 % and 35% of
leflunomide and methotrexate patients sustained ACR20 response status until study end at 52
weeks. Patients on leflunomide fared better than methotrexate patients from a functional
perspective, with statistically significant differences favouring leflunomide in the HAQ disability
index and 2 of 8 subscores of the SF-36. Leflunomide is also statistically significantly superior to
placebo in retarding disease progression as measured by x-ray analysis of both erosions and joint
space narrowing but no consistent differences were noted between leflunomide and methotrexate
on assessments of joint damage.
Table 4.1: Demographic and disease chiuacteristics of the study patients at baseline (76).
Leflunomide Placebo Methotrexate
-- --- - (n= 182) (n=118) (n= 1 82) -- -
Female, % 72.5 70.3 75.3
Age, years ' 54.1 k 12.0 54.6 10.7 53.3 k 1 1.8
Rheumatoid Arthritis, duration, y ' 7.0 5 8.6 6.9 k 8.0 6.5 * 8.1
Rheumatoid Anhritis, s 2 years, % 3 9 33.3 40.1
Rheumatoid factor positive, % 64.8 60.2 59.4
No. of DMARDs that failed ' 0.8 5 1.0 0.9 i 0.9 0.9 -C 1 .O
No prior DMARD treatment, O/o 44.5 39.8 44
Taking concomitant NSAIDs, % 75.2 65.2 69.7
Taking concomitant steroids, % 53.8 55.1 52.7
No. of tender joints (range, 0-28) 15.5 k 6.4 16.5 = 6.3 15.8 i 6.9
Patients' global assessment of 5.6 * 2.2 5.8 i 2.2 5.4 i 2.3 disease activity (10 point VAS)
Modified health assessment 0.8 k 0.6 0.9 = 0.5 0.8 * 0.5 questionnaire scores
Rating scale values were missing for 19, 1 1, and 1 1 patients in the leflunomide,
methotrexate and placebo group respectively, and SG utilities were missing for 2 1, 12, and 1 1
patients, respectively. Missing values exclusively occurred in patients who were considered non-
responders, with the exception of l patient in the leflunomide group and 1 in the rnethotrexate
group who were responders. Analysis of variance showed no statistically significant differences
between leflunomide and methotrexate (Table 4.2). Rating scale values for both leflunomide
(67.7) and methotrexate (64.8) were significantly different fiom placebo (57.5) [p < .05]. In
addition, standard gamble results were statistically significantly different between methotrexate
(83.2) and placebo (77.0) [p < .05], but not between leflunomide (80.2) and placebo (Table 4.2).
Table 4.2: Standard gamble and rating scale values for 480 patients participating in a 12-month randomized controlled trial comparing leflunornide to methotrexate and placebo.
Leflunomide Placebo Methotrexate
n = I63 (RS) and 16 1 (SG) N = 107 (RS and SG) N = 169 (RS) and 168 (SG)
Resource utilization episodes, such as the number of hospitalizations (all), outpatient physician
visits and the number of separately ordered laboratory tests, occurred more frequently in the
leflunomide group, while the number of outpatient diagnostic imaging tests and the number of
outpatient procedures were similar in the leflunomide, methotrexate and placebo groups (Table
4.4). These observations also hold when resource utilization episodes are annualized. The
annualized direct costs and indirect costs exhibit markedly skewed distributions: almost 50% of
patients incurred no costs, and a few patients incurred extremely high costs (Table 4.5).
Table 4.4: Health Care Resource Utilization Items Incurred by patients over the 52 week study period, unadjusted, and adjusted by total time of follow up as number of events per person-year
- - - - - -
Leflunomide .Methotrexate PIacebo Items Unadj. per person- Unadj. per person- Unadj. per person-
-- -- - -- year year year Total person-years of observation 126.1 135.1 65.8
Hospital or Nursing Home Stays 3 5 0.28 6 0.04 10 0.15
Statistical analyses of the annualized total costs, representing the societal perspective, revealed
no statistically significant differences between leflunomide and methotrexate, or in comparison
of these two agents with placebo (Table 4.6 - Following page). Analysis of direct medical costs
only, representing the perspective of the Provincial Health Insurance Plan, also revealed an
absence of a statistically significant difference between leflunomide and methotrexate. However,
direct annualized medical costs associated with leflunornide were statistically significantly higher
than placebo in the statistical comparisons of the log-transformed costs: Wilcoxon Rank Sum, t-
test and regression on log-transformed costs. Direct annualized medical costs associated with
methotrexate were significantly higher than placebo in the regression analysis on log transformed
cost data. The differences in direct costs between leflunomide and placebo persisted after
arbitrarily trimming all costs by the top 1%.
Table 4.6: Statistical Evaluation of total costs, direct costs and direct costs trimmed by the highest 1%. Results of statistical tests on untransformed and Iog- transfomed costs.
Leflunomide vs. Methotrexate $66,000 MTX dominates
Payer's Perspective
Methotrexate vs. Placebo S6,788 S 17.5 16
Leflunomide vs. Placebo $21,355 $146,812
Leflunomide vs. Methotrexate --
$60,200 MTX dominates -
ACR20: Response status by American College of Rheumatology Criteria for Improvement of at least 20%
4.5 Discussion
[n this investigation we compared, from the perspective of a Canadian health insurance
plan, the economic consequences and health utilities of RA patients treated with leflunornide,
methotrexate or placebo. All data were collected during the first 12 months of a North American
randomized controlled trial which was conducted in 482 patients with rheumatoid arthritis. The
results of this analysis demonstrate that, when excluding drug monitoring and acquisition costs,
leflunomide has an othenvise similar economic profile compared to the best alternative
methotrexate. In comparison to placebo, leflunomide costs an extra $147,437 per QALY gained
and methotrexate S12,226. Although no statistically significant differences were found in health
state utilities between leflunomide and methotrexate, methotrexate is both less expensive than
leflunomide and produces higher utilities. However, more patients on leflunomide achieved an
ACMO response status at a cost of $66,000 per responder gained in comparison to methotrexate.
This is one of the few economic evaluations in rheumatology, where economic data were
collected concurrently to a randomized controlled trial. Both patient and investigator were
blinded to treatment assignment and data collection procedures were identical in each treatment
arm, which lends support to the conclusion that the economic consequences were similar
between the two active treatments. It is unclear, however, how high the economic consequences
would have been, had this study been performed under routine treatment conditions similar to the
daily practice of the rheumatologists in the study.
Because a portion of medical resource consumption was mandated by the protocol, it was
not possible to determine the full costs that would have occurred in the absence of protocol-
related monitoring. Treatment or disease-related resource consumption was collected in a fairly
detailed way, but no attempt was made at collecting resource utilization data at a micro level.
Additionally, the use of healthcare resources unrelated to the treatment of RA was not
documented and thus not included in the costing process. Even though some unrelated costs
might have been documented due to judgmental errors on part of the investigator, these are
probably distributed fairly evenly and unlikely to influence the final comparison.
Consumption of medical resources, in the present study, occurred to US patients in a US
healthcare setting but these resources were costed in Canadian dollars. There is sufficient
evidence from other diseases that US utilization of healthcare may be up to twice as intense as
that in Canada (82). Overall direct costs are thus inflated compared to potential Canadian costs
and the already non-significant difference between leflunomide and methotrexate would have
been smaller than the one observed had this study been conducted in a Canadian setting. A
differential effect on the direct costs in both treatment arms is unlikely given the fact that these
data were collected in a randomized controlled trial.
Methotrexate is considered the standard therapy in the management of RA and, among
DMARD monotherapies, patients on methotrexate achieve the highest relative response. Thus
the requirement of regulatory agencies to use methotrexate as the active comparator in clinical
trials of new therapies. In Canada, methotrexate is also one of the least expensive DMARDs,
which makes it one appropriate comparator for a direct economic comparison. Other comparators
could have been conceivable as well, particularly those DMARDs that are chosen if a patient
fails or presents adverse events to methotrexate. Because methotrexate is considered both the
least expensive and most effective DMARD, the cost-effectiveness and cost-utility ratios of
leflunomide in comparison to other DMARDs are likely to be more favourable than the present
values in comparison to methotrexate.
There are only few studies to which we can compare the results of the present study. The
closest, and most recent study is the one by Verhoeven and associates (83) who compared the
cost-effectiveness and cost-utility of combined step-down prednisolone, methotrexate and
sulphasalazine, to sulphasalazine alone in a Dutch setting. The authors found a not statistically
significantly better economic profile and a significantly better clinical profile in favour of
combination therapy. Measurement of costs in this study were comprehensive, including costs
unrelated to the disease. This explains the higher total costs oFUS$5,500 - US$6,500 incurred by
the study patients. Even though the combination therapy arm was clinically superior to the
sulfasalazine arm, the economic consequences were not statistically significantly different.
Althoug it is unknown, how much clinically superior a new therapy has to be, in order to produce
more favourable economic consequences in comparison to methotrexate, it is likely that this is
difficult to achieve given the low costs of this drug and its high effectiveness.
Before the approval of leflunomide by the FDA in 1998, there was a decade without new
treatments that physicians could offer to patients with RA. Recently, several new drugs have
been approved and rheumatologists face better treatment choices than before. Our study showed
that leflunomide has an economic profile that is similar to methotrexate and that the extra costs
faced by policymakers are fixed treatment costs that, however, are higher than the costs of
generic drugs, such as methotrexate. From an economic perspective, leflunomide imposes itself
as an alternative once methotrexate fails, due to its equally high efficacy. But leflunomide might
also be a drug of first choice, provided that drug acquisition costs are covered.
CHAPTER 5
An estimation of the i y e a r cost effectiveness of adding Leflunomide to a strategy of
conventional disease modifying antirheumatic drugs in patients with RA
5.1 Preface
This chapter reports the results of the model-based economic evaluation of leflunomide as
an element of a typical treatment strategy for patients with more aggressive rheumatoid arthritis.
Data that were collected in the surveys (Chapter 2), the meta-analysis (Chapter 3) and the
randomized clinical trial (Chapter 4) were used to support a decision-analysis-model. The model
runs over 5 years with cycles of 6 months duration and evaluates the cost-effectiveness and cost-
utility of adding leflunomide to a typical DMARD-strategy.
5.2 Introduction
The last decade has seen an intense debate about appropriate treatment strategies in the
management of patients with RA. A consensus seems to have emerged that patients with
moderate or aggressive RA should be treated early and aggressively, if possible by combining
several disease modifying anti-rheumatic drugs (DMARDs). However, these recommendations
were made in the absence of potent therapeutic alternatives. Fortunately, several new therapies
were recently approved for the treatment of patients with RA, which now add powerfbl options
to the management of RA. In 1998, leflunomide (LEF) was the first disease modifying drug
approved in over a decade by the US Food and Drug Administration (FDA), soluble TNFa
receptor and monoclonal anti-TNFa antibodies followed. However, compared to existing
therapies, which may cost as low as USS300 per year, these new agents cost between US$3,500
and US$13,000 per year. Thus, their therapeutic potential needs to be measured not only by their
ability to slow or halt disease progression, but also by their share in decreasing the healthcare
costs incurred by patients with RA. Reimbursement approval for these new agents will likely
depend on their demonstrated cost-effectiveness.
The costs of illness associated with RA are compounded by a disease that often starts
early in life, that can lead to severe disability and frequently requires hospitalization and intense
medical and surgical treatment. The yearly costs of RA vary, but have been estimated to be
approximately US$8,416 per patient in 1996 dollan. Direct medical and non-medical costs and
indirect costs, such as losses in productivity, share equal parts of this sum. In Canada, the annual
costs incurred by RA patients were calculated to average CAD$6,253 (1 994 dollars) (22), with
direct costs responsible for 74% ($4,656) of the total and prescription drugs for - 20% of the
direct costs. Extrapolating these costs to the 0.5% to 1% of the Canadian population estimated to
be affected by RA would leads to an estimate of the total healthcare costs of RA patients in
Canada between 0.9 and 1.75 billion dollars (1994 Cdn.).
The relationship between costs and effectiveness of old or new anti-rheumatic therapies
has been examined in only very few economic evaluations (73-79, and all of these compared one
therapy to another over a short time-period. There have been two attempts to model the disease
over longer time horizons (84;85) but without the required efficacy data to study the costs and
benefits within a therapeutic framework. The present study is an attempt to examine the
incremental cost-effectiveness and cost-utility of LEF, a novel disease modifying antirheumatic
drug, within a realistic sequence of DMARDs modelled over a 5-year time horizon in patients
with active RA, from a Canadian payer's perspective.
5.3 Methods
Decision Analysis Model
A decision analysis model was developed to represent the events that occur as a
consequence of taking DMARDs (Figure 5.1 - following page). Several models were available
from the literature (73;74), but none of the models allowed for a combination of data from
observational studies - which primarily report on the proportion of patients maintaining therapy
and randomized controlled trials (RCTs) - which primarily report on the degree of treatment
response, e.g. the American College of Rheumatology composite criteria for 20% improvement
(ACR20) (2).
Adverse Event -.: Sty tnew DhMRD
I tack o f Efficacy .'. Stan new DMARD
Minor Adverse Event Clinical Response I --. : Continue same D W R D (ACUO)
i 1 1 1 No Adverse Event
- ' Continue same D W R D !
Minor Adverse Event 1 No Clinical Response ; - - ' Continue same DMARD I (ACR2O) P
'4 No Advcne Event : Continue same D M R D
6 Months - _ _ _ _ _ ___--__. . _ _ _ _ _ _ .___
Figure 5.1: Events occurring during a 6-month period as a consequence of taking a DMARD. Patients who stop one DMARD start a new one and repeat the cycle with the new DMARD.
To conform with available data from observational studies and RCTs, the decision tree is
designed so that patients may experience three types of events during a 6-month treatment period
(Figure 5.1): 1) they continue therapy, 2) they stop therapy because of adverse events or 3) they
stop therapy because it is lacking or losing efficacy. Furthermore, continuing patients may be
classified clinically with regard to the quality of the response to treatment - here defined as
meeting the A C E 0 criteria - and whether they experience minor adverse events. Because this
analysis is limited by the type of data available from the literature, a more detailed classification
of adverse events by type and severity could not be utilized.
Overall, two DMARD strategies were compared, one including LEF and the other
excluding it (Figure 5.2). Both strategies try to emulate treatment as it happens in real life, which
means that patients cycle through different treatment regimens when they encounter toxicity or
lack or loss of efficacy. A new treatment regimen is chosen for those patients who fail the
previous regimen.
Lack of Egicocy +
Methotrexate
Strategy with /"\ Strategy without leflunomide ,I leflunomide
MTXfHCQtSSZ $266.39 $266.76 (S 159.04 - $374.47) $642.85
LEF $ 140.50 $171.33 ($125.16-5217.50) $1,926.36
Gold 562.68 $419.24 ($384.06-5454-41) $518.62
Ut if it ies
Of 364 patients in the active treatment arms in the North American study of LEF, there
were 329 patients who participated at the utility interviews; 135 were defined as treatment
success, and 195 as not a treatment success by ACR20 criteria. SG utilities for success were
8 1-62 * 22.79 (mean * standard deviation) and 82.37 * 27.99 for patients considered not a
success (difference of 0.75 no statistically significant).
Cost-effectiveness and cost -141 ility rest clts including sensitivity analysis
Results of the cost-effectiveness analysis over a 5 year period, i.e. ten 6-month cycles,
show that the strategy including LEF would cost approximately 37,675 dollars per 5-year period
and a strategy excluding LEF would cost approximately $6,6 10, a difference of 5 1,065 over the 5
year period (Table 5.4). Patients in the sequence of DMARDs that includes LEF would, on
average, be in a state of response for 3.04 years over the 5-year period, gaining an extra 36.5 days
compared to patients in the sequence that excludes LEF, who would be in a state of response for
2.94 years. These findings translate into a cost-effectiveness ratio of $10,682 for each additional
year of response. Confidence intervals for the cost-effectiveness ratio were calculated from the
output of the simulations as ratios of the costs and effects plus/minus one standard deviation.
Adding or subtracting one standard deviation provided a range of $3,380 to $21,018 per year of
response. This cost-utility analysis shows that patients in the strategy including LEF gain 0.058
quality adjusted life-years (3.636 QALYs vs. 3.578 QALYs) which with the above mentioned
costs per strategy would provide a cost-utility ratio of S 18,474 per QALY gained (Confidence
interval i I standard deviation: 54,178 - $83,6 14).
Table 5.4: Expected costs, numbers of cycles patients respond to therapy, time spent on active therapy and cost per responder. Table shows point estimates and mean +/- one standard deviation as determined by Monte Carlo simulations with 5000 runs.
-- - -- - - - - - - - - -- - . -
Mean SD Minus SD Plus SD -- -- - -- - - - - - - -- -
Costs
Inc. Lef
Excl. Lef
Years in ACR2O response
Inc. Lef Excl. Lef
QALYs
Inc. Lef.
Exc. Lef
Cost per responder
fnc. vs. ExcI.
Cost per QALY
Inc. vs. Excl.
5.5 Discussion
In this investigation we compared, from the perspective of a Canadian health insurance
plan, the cost-effectiveness and cost-utility of adding LEF to a realistic treatment strategy in the
management of patients with active RA, as supported by data gathered through a systematic
review of the literature and surveys of Canadian rheumatologists. Evaluation over a 5-year period
shows that adding LEF as a therapeutic option increases the management costs by S 1,085
compared to the strategy without LEF with a cost-effectiveness ratio of $10,682 per additional
year of response to treatment gained, and a cost-utility ratio of $18,474 per QALY gained.
The model-based cost effectiveness approach was selected to demonstrate the added
value of LEF when introduced in a management approach adopted by Canadian rheumatologists
in their usual management of patients with active RA. This approach emulates the management
of EL4 patients in real life where they cycIe through different treatment regimens upon
experiencing toxicity or lack of efficacy. There are several advantages to this type of analysis.
The modelling approach allows estimation of the expected performance of the drug in the real
world and theoretically provides a better estimate of the cost-effectiveness of new interventions.
Furthermore, the impact of LEF can be assessed over a time horizon that exceeds the limited
time-horizons adopted in clinical trials. Within that more realistic framework, the addition of
LEF as a new treatment alternative for patients requiring a new therapy extends the time patients
may benefit from DMARDs.
The strategy modelled in this investigation was identified by means of a survey of
Canadian rheumatologists, which showed that MTX discontinuation is followed by prescriptions
of MTX combinations, gold, or immunosuppressants such as CYA (9) These treatment choices
were therefore incorporated into the model-based cost effectiveness analysis. The choice of
sequence was limited, on one hand by technical considerations, imposed by the type of model
used, and on the other hand by the availability of information which needed to be retrieved from
a systematic review of the literature. Even though it is clear that different rheumatologists will
opt for different treatment sequences, only the sequence thought to be most representative of
rheumatologists' treatment choices for patients with more aggressive RA was chosen for the
comparison. In addition, the choice of sequence was limited by the availability of emcacy data,
i.e. ACR.20 response, in the literature. For example, the survey of Canadian rheumatologists
identified the combination of MTX and HCQ to be more frequently used than that of MTX and
SSZ. However, the combination of MTX and HCQ could not be modeled, as there is no study
reporting ACWO response rates for this treatment strategy. Because the MTX-HCQ combination
is perceived in practice to be equally efficacious to the methotraxate-sulphasaiazine combination,
the substitution of one for the other in the model was not expected to change the conclusions of
the analysis.
The advantage of choosing a sequence of DMARDs was that the results could be
presented in a setting that is more realistic than single-agent comparisons of LEF to other
DMARDs or DMARD-combinations. Furthermore, the incremental cost-effectiveness of single
agent comparisons, for example LEF vs. CYA or gold would be lower than the cost-effectiveness
resulting from a comparison of treatment sequences. This is because the costs and efEects of each
treatment sequence were proportionally influenced by the time patients spend on the drugs in
each sequence. Indeed, given the relatively lower discontinuation rate, MTX and MTX
combinations were contributing more to the costs and effects than drugs that appear later in the
sequence, such as gold and CYA.
The analysis was conducted over a five-year time-horizon based on the 2-year
information available for LEF and information available in the literature for the other DMARDs.
Having to extrapolate from the 2-year data is a limitation of the model that does not Favor LEF.
For example, long-time experience with MTX showed that approximately 14% of patients
discontinue during a 6-month period. When referring to the North American trial comparing LEF
to MTX, the clinical trial setting points to a discontinuation rate of 32% with no difference in
discontinuation rates between the two agents (76).
Other outcomes, such as improvement in ACR criteria by 50%, would also have been
plausible choices for the model. This is especially important as the newer DMARDs may be
more potent and probably qualitatively superior to existing therapies, however, very few studies
in the literature report on ACRSO outcomes. Although here is little reason to suspect qualitatively
superior responses among the comparators used in the present study, this cannot be entirely
excluded, and would likely bias against leflunornide.
While it is likely that, with more exposure to the drug, discontinuation rates in real life
will fall to the level observed For MTX, the model used a 32% discontinuation rate for LEF
because of lack of other values. This also implies an underestimation of the time patients spend
on LEF and thus a possible overestimation of the cost-effectiveness ratio (i.e. overestimation of
the cost per additional year of response gained and cost per QALY) given the higher costs and
lower response rates of the following DMARDs.
Direct medical costs estimated for the purpose of the model-based comparison are only an
approximation of the costs likely to be encountered in real life. It is conceivable, that costs will
be lower in real life. First, monitoring costs were derived from monitoring recommendations
provided in the product monographs. In real life, physicians will likely monitor less, the more
they gain experience with each drug. Similarly, costs of managing adverse events were derived
kom a literature-based survey of 5 Toronto rheumatologists. Costing of the adverse events was
based on the answers provided in the swey , which may deviate from physicians' behaviour in
real practice. Physicians responding to the questionnaire may have adopted a more cautious
approach, not knowing the precise clinical circumstances accompanying the adverse event.
The majority of data supporting the model-based comparison were derived from the
literature. The literature, however, is sparse for much of the data needed. For example, only 35%
of the observational studies studying maintenance of DMARDs provide data needed for the
model. Similarly, only approximately 30% of RCTs or observational studies provide information
on the number of patients suffering one or more adverse events. Explicit response criteria have
only recently been used in RCTs and are seldom used in observational studies. For these reasons,
the data used in our model do not represent all studies that were conducted with the respective
drugs.
Research into the economics of DMARDs is complicated by the chronic nature of RA.
There is not one drug to treat RA, but a multitude and all of them are failing with time. This
makes modeling more complicated, in particular if combinations of drugs are used. Of Further
importance for modeling RA is the severity and stage of the disease, as a differential response
may be expected from patients with mild RA or those who try MTX as first drug or after
previous failure of other drugs, such as SSZ or HCQ. These details were impossible to consider
in the present modeling framework.
The results of this study confirm that leflunomide has a place in the management of
patients with RA and that its integration into the therapeutic armamentarium comes at reasonable
cost. The exact therapeutic value of leflunomide in the routine management of patients with RA
needs to be Further established. As rheumatologists gain more experience with leflunomide, the
drug's effectiveness will likely improve and further enhance its economic profile.
CHAPTER 6
Epilogue
Pharmacoeconomic evaluations of pharmaceuticals often depend on information that needs to be
collected in routine care settings, and often country-specific. Data from randomized controlled
trials, observational studies and administrative databases, and those collected from experts need
to be integrated in an appropriate way to adequately estimate the pharmacoeconomic
characteristics of a given drug in a country-specific setting. The methods of economic evaluation
are inexact and often leave sufficient room for specific decisions or assumptions to influence the
final findings. Consequently, results of economic evaluations are preferably seen as supporting
reimbursement decisions rather than guiding them.
This thesis presents a comprehensive assessment of the pharmacoeconomic properties of
leflunomide, a novel disease-modifying anti-rheumatic drug. Several studies, surveys and rneta-
analyses were undertaken to put this economic evaluation on a more stable foundation. The
physician survey was necessary to understand treatment preferences and their sequences in
patients who are similar to those receiving leflunomide. Furthennore, to not just provide head to
head comparisons to methotrexate or other DMARDs, an understanding of the long-term
behaviour of patients on these DMARDs was required. The meta-analysis of DMARDs provided
this quantitative information from published studies of these DMARDs under clinical trial and
routine care conditions. These data made it possible to model the management of patients with
RA in the long-term and to align treatments in a sequence that is close to what physicians would
do in their routine care of patients with RA. Review of these published studies also provided
sufficient information about the incidence and severity of adverse events under routine care
conditions.
However, the information derived fkom observational studies needed to be supplemented by data
collected in randomized controlled trial settings. For example the standard gamble utilities
measured in the randomized controlled trial of leflunomide, methotrexate and placebo were
required to assess the cost-effectiveness of leflunomide in terms of cost per quality adjusted life-
year gained. Equally important were the ACR response rates. Therefore, this combination of trial
data, data derived fiom systematic reviews and separately conducted surveys corroborated the
treatment model underlying this economic evaluation fiom various perspectives, that of the
routine care setting and that of a clinical trial setting.
The separately conducted analysis of the immediate economic consequences of methotrexate,
leflunomide and placebo, as collected alongside the randomized controlled trial, provided a
further understanding of the pharmaco-economic properties of leflunomide. The results showed
that decision makers can count on economic consequences that are similar to methotrexate and
therefore assist decision maken in gaining a better understanding about the costs of leflunomide,
should they opt for Full or restricted reimbursement.
Leflunomide is a much needed DMARD that enables patients to delay the need for much more
aggressive therapies or even surgery. Its overall profile, including efficacy and adverse events, is
equal to that of rnethotrexate, at present the standard treatment for patients with rheumatoid
arthritis. It comes at an extra costs that decision makers need to assess in their particular
circumstances. The results of this thesis may provide helpful assistance.
References
[I] Amen FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF, Cooper NS et al. The
American Rheumatism Association 1987 revised criteria for the classification of rheumatoid
arkitis. Arthritis Rheum 1988; 3 1(3):3 15-324.
[2] Felson DT, Anderson JJ, Boers M, Bombardier C, Furst D, Goldsmith C et al. American
College of Rheumatology. Preliminary definition of improvement in rheumatoid arthritis.
Arthritis Rheum 1995; 38(6):727-735.
[3] Felson DT, Anderson JJ, Lange ML, Wells G, LaValley MP. Should improvement in
rheumatoid arthritis clinical trials be defined as fifty percent or seventy percent improvement in
core set measures, rather than twenty percent? Arthritis Rheum 1998; 4 l(9): 1564- 1570.
[4] Fries JF. The hierarchy olquality-of-life assessment, the Health Assessment Questionnaire
(HAQ), and issues mandating development of a toxicity index. Controlled Clin Trials 199 1 ; l2(4
Suppl): 106s- 1 17s.
[5] Pincus T, Summey JA, Soraci SA, Jr., Wallston KA, Hurnmon NP. Assessment of patient
satisfaction in activities of daily living using a modified Stanford Health Assessment