Pharmacoeconomics 2008; 26 (5): 395-408

REVIEW ARTICLE 1170-7690/08/0005-0395/$48.00/0

© 2008 Adis Data Information BV. All rights reserved.

Economic Evaluations in
Rheumatoid Arthritis
A Critical Review of Measures Used to Define Health States
Nick Bansback,1 Roberta Ara,2 Jonathan Karnon3 and Aslam Anis1,4
1 Centre for Health Evaluation and Outcome Sciences, St Paul’s Hospital, Vancouver, British
Columbia, Canada
2 School of Health and Related Research, University of Sheffield, Sheffield, UK
3 Department of Public Health, School of Population Health and Clinical Practice, University of
Adelaide, Adelaide, South Australia, Australia
4 Department of Healthcare and Epidemiology, University of British Columbia,
Vancouver, Canada

Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
1. Review Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
2. Search Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
3. Health Assessment Questionnaire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
3.1 Health Utility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
3.2 Direct and Indirect Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
3.3 Mortality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
4. American College of Rheumatology Core Response Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
4.1 Health Utility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
4.2 Direct and Indirect Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
4.3 Mortality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
5. Disease Activity Score . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
5.1 Health Utility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
5.2 Direct and Indirect Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
5.3 Mortality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
6. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405

Abstract We reviewed the clinical measures used in rheumatoid arthritis (RA) economic
evaluations with respect to their relevance and sensitivity to changes in survival,
health-related quality of life (HR-QOL) and costs. We compared the measures
from the economic perspective and discussed the validity of methods used to
extrapolate beyond the trial data. Cost-effectiveness evaluations of disease-
modifying antirheumatic drugs in RA were identified by searching MEDLINE,
EMBASE, Econlit and NHS EED databases. Studies were retained if they
extrapolated beyond randomized controlled trial evidence using relationships
between clinical measures, costs and utilities.
396 Bansback et al.

In the 22 studies identified, clinical severity was measured using the Health
Assessment Questionnaire (HAQ) Disability Index, the American College of
Rheumatology (ACR) response criteria, the Disease Activity Score (DAS) or a
combination of the HAQ and DAS. The HAQ is correlated with mortality, costs
and HR-QOL instruments, and several studies used linear relationships to model
these associations. However, a polynomial relationship or discrete states may be
more appropriate for patients at the extremes of the disease spectrum, and
numerous HAQ health states may be required to capture differences in mortality
risk. While the ACR response criteria is a more comprehensive measure than the
HAQ, it is a relative measure, which creates difficulties when estimating absolute
changes in HR-QOL, costs and mortality risk. The evidence base linking DAS
scores with HR-QOL instruments, costs and mortality is less robust, possibly due
to the comparatively recent development of the measure and the limited number of
possible scores (mild/moderate/severe). While there is some evidence of a rela-
tionship between DAS scores and costs, the DAS does not capture all aspects of
HR-QOL, and no significant relationship has been established with mortality risk.
Evidence suggests the HAQ to be the primary clinical measure for use in
economic evaluations as it is measured in almost all clinical studies, and is closely
correlated to health utilities, mortality and costs. While new developments suggest
the sensitivity of health states may be improved by combining the HAQ with
measures such as the DAS, further research is required in this area. Further
research is also required to explore the advantages in using either continuous or
discrete health states.

Rheumatoid arthritis (RA) is a chronic, progres-
sive, inflammatory disease that affects approximate-
ly 0.8% of the adult population.[1] RA affects the
physical functioning and psychological and social
health of patients, and is associated with premature
mortality.[2-4] The disease also has a substantial ef-
fect on societal cost, in terms of healthcare resources
consumed and productivity lost.[5] Recent attention
to RA comes from the development of new biologi-
cal therapies that target cytokines, such as tumour
necrosis factor (TNF), T cells, B cells and interleu-
kin (IL)-1, and hold promise for the prevention of
joint destruction and maintenance of functional sta-
tus.[6] While effective, these new medications come
at a high cost. A recent analysis examining the costs
of RA in the US found that annual direct costs have
increased by 300% (from $US6164 to $US19 016,
year 2001 values) since the introduction of these
new medications,[7] and prescribing in RA alone has
placed considerable pressure on health budgets
across the world.[8,9] With many new biological
agents anticipated to enter the market soon,[10] cost-
effectiveness analyses are an important source of
information for policy decision makers charged with
making reimbursement decisions.[11]
In specifying the framework for economic evalu-
ation, a key issue concerns the choice of time hori-
zon for the evaluation.[12] Guidelines state that it
should be sufficiently long to reflect any differences
in costs or outcomes between the technologies being
compared.[13-15] As RA affects long-term function,
evaluations are encouraged to extrapolate the evi-
dence from clinical trials with relatively short fol-
low-up periods to estimate the lifetime costs and
consequences of alternative therapy options for RA.
Decision modelling describes the movement of pa-
tients through a discrete set of relevant health states
that represent events that have significant effects on
the costs and/or effects of treatments and/or the
disease process.[16] It is now an accepted tool for
synthesizing data from disparate sources to extrapo-
late trial results to a lifetime horizon,[13-15] and has

© 2008 Adis Data Information BV. All rights reserved. Pharmacoeconomics 2008; 26 (5)
Economic Evaluations in Rheumatoid Arthritis
been used extensively to evaluate therapy options
for RA.[17]
Unlike other disease areas where independently
developed decision models follow a very similar
structure,[18] in RA, several modelling strategies us-
ing different clinical measures to define health states
have been developed. This diversity is a result of the
wide-ranging effects of RA on health-related quality
of life (HR-QOL), which has led to the development
of multiple clinical measures, from which econo-
mists have differentially extrapolated to estimate
lifetime costs and effects.
This paper presents a review of the use of ex-
isting clinical measures of the effects of RA as the
basis for the economic evaluations of the new class
of biological interventions for RA. Evidence is pre-
sented of the relevance and sensitivity of the mea-
sures to the economic effects of RA, i.e. survival,
HR-QOL and costs (from a healthcare and societal
perspective). Based on the findings of the review,
the subsequent section summarizes the pros and
cons of the alternative measures from the economic
perspective. Issues around the optimal application of
the clinical measures are then discussed, including
the relevant representation of clinical pathways and
the validity of methods used to extrapolate the mea-
sures beyond trial settings.
1. Review Methodology
Cost-effectiveness studies published between
January 1990 and January 2007 were identified by
searching MEDLINE, EMBASE, Econlit and NHS
EED databases. Search terms for EMBASE are pro-
vided in appendix 1. Studies were included in the
review if they met the following criteria: (i) the
study population had RA; (ii) the study included a
disease-modifying antirheumatic drug (DMARD);
and (iii) the study extrapolated beyond the duration
of the randomized controlled trials (RCTs) using
relationships between clinical measures, costs and
utilities. Economic evaluations conducted alongside
RCTs and studies exploring the cost effectiveness of
nonbiological agents were excluded. Reference lists
were searched and we also included any other stud-
ies known to the authors.
© 2008 Adis Data Information BV. All rights reserved.
397
2. Search Results
A total of 22 evaluations were included in the
review (table I).[19-40] The global impact of the new
biological agents is demonstrated by the wide range
of settings used: Canada, Japan, the Netherlands,
Sweden, the UK and the US.
Other than Choi et al.,[29,30] who measured out-
come in terms of patients meeting different percent-
ages of the American College of Rheumatology
(ACR) response criteria, all other studies defined the
QALY as the primary outcome measure. The major-
ity (n = 15) used the health assessment questionnaire
disability index (HAQ-DI) alone to represent clin-
ical severity. Others used either the ACR response
criteria (n = 5), the disease activity score (DAS;
n = 1), or a combination of the HAQ and DAS
(n = 1). The following sections review the alterna-
tive clinical measures with respect to their relation-
ship with health utilities, direct and indirect costs
and mortality; i.e. how well do changes in the clin-
ical outcomes represent important changes in HR-
QOL, cost events, and mortality risk.
3. Health Assessment Questionnaire
The HAQ is a self-completed questionnaire that
is measured in almost every RA clinical study. It
was developed as a comprehensive measure of out-
come in patients with a wide variety of rheumatic
diseases, including RA, osteoarthritis, juvenile RA,
lupus, scleroderma, ankylosing spondylitis, fibro-
myalgia and psoriatic arthritis.[41] Although the
complete form includes an assessment of mortality,
disability, pain and symptom levels, drug adverse
effects and resource utilization, the majority of stud-
ies only use the physical disability scale, which
produces a disability index (HAQ-DI).[42] This scale
assesses upper and lower limb function in relation to
the degree of difficulty encountered in performing
daily living tasks. These tasks include walking,
dressing, bathing and shopping. The HAQ-DI con-
tains 20 items or components distributed across
eight categories. Each component has four possible
responses ranging from 0 (without any difficulty) to
3 (unable to do). The highest score on any item
within one category represents the category score.
Pharmacoeconomics 2008; 26 (5)
 Table I. A description of studies included in the review

398
Study (year of Country Interventions Time horizon Type Model type ICER (year of pricing)
publication)
Brennan et UK DMARD sequence with and without TNFα Lifetime CUA Individual Incorporating TNFα inhibitors resulted in
al.[24] (2007) inhibitors sampling model an ICER of £23 882 per QALY vs
DMARDs (2004)
Marra[36] Canada Infliximab plus methotrexate vs 10 y CUA Markov ICERs varied from $Can32 000–70 000 per
(2007) methotrexate QALY depending on the utility measure
used (2004)
Wailoo et US Strategies including adalimumab, Lifetime CUA Individual Etanercept and adalimumab dominated
al.[23] (2007) etanercept and anakinra compared with sampling model infliximab, but all three drugs were more
infliximab cost effective than anakinra (2005)
Chen et al.[27] UK Sequences of DMARDs with and without Lifetime CUA Individual The ICERs for etanercept, adalimumab
(2006) etanercept, adalimumab and infliximab sampling model and infliximab when used after multiple
DMARD failure were approx. £24 000,

© 2008 Adis Data Information BV. All rights reserved.

£30 000 and £38 000 per QALY,
respectively (2004)
Coyle et al.[35] Canada Infliximab and etanercept vs DMARDs 5y CUA Markov Both infliximab and etanercept strategies
(2006) had ICERs >$US100 000 per QALY
compared with DMARD treatments (2003)
Spalding and US Etanercept, infliximab and adalimumab vs Lifetime CUA Markov Compared with methotrexate, ICERs
Hay[37] (2006) methotrexate ranged from $US63 769 per QALY for
adalimumab to $US409 000 per QALY for
infliximab (2004)
Tanno et Japan DMARD sequence with and without Lifetime CUA Markov Inclusion of etanercept vs no etanercept
al.[34] (2006) etanercept resulted in an ICER of ¥2.5 million per
QALY (2005)
Bansback et Sweden Adalimumab, etanercept and infliximab Lifetime CUA Individual Etanercept + methotrexate and
al.[22] (2005) with and without methotrexate compared sampling model adalimumab have ICERs ranging from
with DMARDs €35 000 to €52 000 per QALY vs
DMARDs (2001)
Barbieri et UK Sequence of methotrexate and DMARDs Lifetime CUA Markov The addition of infliximab gave an ICER of
al.[31] (2005) with and without infliximab £23 916 per QALY (2000)
Kobelt et Sweden Etanercept plus methotrexate vs 10 y CUA Markov A strategy of etanercept vs methotrexate
al.[19] (2005) methotrexate resulted in an ICER of €46 494 per QALY
(2004)
Schädlich et Germany Sequence of DMARDs 3y CUA/CEA Markov Incorporating leflunomide in the sequence
al.[38] (2005) was associated with an ICER of €8301
per QALY, or €5000 per ACR20RYG
(2001)

Continued next page

Pharmacoeconomics 2008; 26 (5)

Bansback et al.
 Table I. Contd
Study (year of Country Interventions Time horizon Type Model type ICER (year of pricing)
publication)
Brennan et UK DMARD sequence with and without Lifetime CUA Individual A strategy including etanercept vs no
al.[21] (2004) etanercept sampling model etanercept resulted in an ICER of £16 330
per QALY (2001)
Chiou et al.[33] US A comparison of etanercept, infliximab, 3y CUA Decision tree The ICER of etanercept + methotrexate vs
(2004) adalimumab and anakinra anakinra was $US7925 per QALY while
adalimumab and infliximab were dominated
by etanercept (2003)
Clark et al.[25] UK Sequences of DMARDs with Lifetime CUA Individual The ICER of anakinra vs DMARDs ranged
(2004) and without anakinra sampling model from £106 000 to £604 000 per QALY
(2002)
Welsing et Netherlands DMARD sequence with and without 5y CUA/CEA Markov Post-DMARD failure was found to be the

© 2008 Adis Data Information BV. All rights reserved.

al.[28] (2004) etanercept most cost-effective use of etanercept, with
an ICER of €163 556 per QALY (2002)
Economic Evaluations in Rheumatoid Arthritis

Kobelt et UK (Swedena) Infliximab vs methotrexate 10 y CUA Markov 2 y of infliximab treatment with 10-y follow-
al.[20] (2003) up produced an ICER of £29 900 per
QALY (2000)
Choi et al.[29] US Etanercept compared with 6 mo CEA Decision tree Compared with methotrexate, the ICER
(2002) various DMARDs was $US41 900 per additional ACR20
responder (1999)
Jobanputra et UK Sequences of DMARDs with and without Lifetime CUA Individual The ICER of infliximab or etanercept vs
al.[26] (2002) etanercept and infliximab sampling model DMARDs ranged from £47 000 to
£169 000 per QALY (2000)
Wong et al.[32] US Sequence of methotrexate and DMARDs 10 y CUA Markov The addition of infliximab gave an ICER of
(2002) with and without infliximab $US30 500 per QALY (1998)
Kobelt et UK Leflunomide compared with methotrexate 10 y CUA Markov Depending on which trial was used,
al.[39] (2002) or sulfasalazine leflunomide either dominated or was
dominated by methotrexate. Leflunomide
dominated sulfasalazine (2000)
Maetzel et Canada Adding leflunomide to a sequence of 5y CUA/CEA Markov The ICER of adding leflunomide was
al.[40] (2002) DMARDs between $US54 229 and $US71 988
depending on the utility values used. The
cost per ACR20 was $US13 096 (1998)
Choi et al.[30] US Etanercept with and without methotrexate 6 mo CEA Decision tree Compared with triple DMARD therapy, the
(2000) compared with various DMARDs ICER for etanercept was $US42 600 per
additional ACR20 responder (1999)

a Not shown.
ACR = American College of Rheumatology response criteria; CEA = cost-effectiveness analysis; CUA = cost-utility analysis; DMARDs = disease-modifying antirheumatic drugs;
ICER = incremental cost-effectiveness ratio; mo = month; RYG = response-years gained; TNF = tumour necrosis factor; y = year; ¥ = Japanese yen.

Pharmacoeconomics 2008; 26 (5)

399
400
The respondent also indicates whether he or she uses
aids or devices (14 items) or help from other people
(eight items). The scores for each dimension are
corrected for the use of aids or devices, summed and
transformed to give an overall disability index be-
tween 0 and 3. A score of 0 represents no disability
and 3 represents very severe, high-dependency disa-
bility.[41] Research has established that the minimal
clinically important difference (the size of change in
the instrument that corresponds to a patient’s per-
ceived change in health) for the HAQ-DI is between
0.2 and 0.25.[43-45] Of the 16 studies that used the
HAQ-DI to define health states, the 0–3 scale has
been separated into between 4[31,32] and 25[36] dis-
crete health states.
3.1 Health Utility
The HAQ is frequently used as the marker for
determining the construct validity and sensitivity of
HR-QOL instruments because of its validity in mea-
suring patient function. It is strongly correlated with
the EQ-5D (r = 0.61), the SF-6D (r = 0.73) and the
Health Utilities Index (HUI)-3 (r = 0.76).[43] A linear
relationship between the HAQ and utility has been
modelled in several of the studies.[21-23,25-27,34,37]
However, the relationship may not be linear at the
boundaries of the disease measure[46] and therefore
polynomial relationships or discrete states could be
more appropriate for analyses of healthy or very sick
patients.
A potential criticism of using the HAQ as a proxy
for health utility is that it does not measure all
aspects of HR-QOL valued in some preference-
based instruments, e.g. psychological impact, pain
or fatigue.[47] While this could potentially lead to an
underestimate of utility gain, a recent mapping exer-
cise between the HAQ and two utility measures
(EQ-5D and SF-6D) found differences between pre-
dicted and actual utilities to be small (root mean
square error [RMSE] < 0.1).[43]
3.2 Direct and Indirect Costs
Studies have found that severity in HAQ disabili-
ty is closely related to drug costs, physician visits
and inpatient stays.[7,48-52] The cost of treating drug-
© 2008 Adis Data Information BV. All rights reserved.
Bansback et al.
related adverse effects is more closely related to the
type of medication, and is therefore not directly
linked to disease severity. The hospital costs are
dominated by joint replacements,[53] and lead to
highly skewed total direct costs.[48] Yelin and
Wanke[48] found that people with RA in the worst
quartile of function (HAQ > 1.75) experienced total
annual direct costs and total hospital costs that were
2.55- and 6.97-fold higher than those in the best
quartile (HAQ < 0.62). After adjustment for other
patient characteristics, Michaud et al.[7] found that a
1-unit difference in the HAQ was equivalent to a
difference of $US1041 (direct costs, year 2001
value) per 6 months.
Indirect costs resulting from reduced productivi-
ty and early retirement have also been reported to be
closely related to the HAQ, and that these costs are
even higher than direct costs of medical care.[5,54,55]
Most studies in the literature have focused on
changes in employment status and absenteeism,[56]
and a worsening HAQ has been found to have a
close relationship with these.[50-52]
3.3 Mortality
The ability of the HAQ to predict mortality is
well established.[4,57-59] For example, Wolfe et al.[4]
found that the risk of mortality would be reduced by
50% if patients in the worst quartile were switched
to the best quartile. Models with discrete states of
HAQ will only capture this association if the HAQ
health states are sufficiently small.
4. American College of Rheumatology
Core Response Criteria
Since their inception in the early 1990s, the ACR
response criteria have been the primary outcome
measure in almost all RA trials.[60] The ACR20 is
defined as a reduction by ≥20% in the number of
tender and swollen joints plus a 20% improvement
in at least three of the following five measures: pain,
patient global assessment, physician global assess-
ment, self-assessed physical disability measured by
the HAQ, and an inflammatory marker in the blood
such as C-Reactive Protein (CRP). The ACR50 and
ACR70 (improvements of at least 50% and 70%) are
Pharmacoeconomics 2008; 26 (5)
Economic Evaluations in Rheumatoid Arthritis
frequently reported, and are sometimes deemed to
be more clinically relevant than the ACR20.[61] Only
five of the included studies used the ACR as the sole
description of health states.[29,30,33,35,40] Chiou et
al.[33] and Coyle et al.[35] split states into ACR20/50/
70 responses while Choi et al.[29,30] analyzed the
ACR20 along with the ACR70. Maetzel et al.[40]
split responders using just the ACR20 criteria. A
number of the other studies have used the ACR
scores to determine the pathway of patients on treat-
ment, including whether patients would remain on
treatment or not.[21,22,34,35,38]
4.1 Health Utility
As the ACR encompasses pain, joint counts and a
marker of inflammation, in addition to the HAQ, it
can be considered a more comprehensive measure
than the HAQ alone. A limitation of using the ACR
to define health states is that it is a relative measure,
and thus does not determine an absolute level of
disease. As a consequence, individuals classified as
ACR responders could have very different charac-
teristics of the disease.[62,63] For example, the HR-
QOL change associated with an ACR70 response
for an individual with very severe disease is likely to
be very different to the HR-QOL change associated
with an ACR70 response in an individual who is
only mildly affected by the disease.
One study has attempted to associate all ACR
response levels with health utilities[64] by eliciting
valuations of health states based on the ACR20, 50
and 70 responses directly from patients with RA.
Using a mixture of the time trade-off method and a
visual analogue scale (VAS) to elicit values, they
found utilities to range from 0.53 (ACR20) to 0.84
(ACR70). However, correlations of the resulting
values with the HAQ, pain and joint counts were
low, ranging from 0.2 to 0.3.
4.2 Direct and Indirect Costs
The relationship between direct costs and ACR
response criteria is also limited by the relative nature
of the outcome measure. The analysis by Chiou et
al.[33] included the costs of drugs, and associated
monitoring and drug adverse event costs, but was
© 2008 Adis Data Information BV. All rights reserved.
401
not able to incorporate the costs saved through re-
duced resource utilization such as joint replace-
ments and physician visits. Increased productivity
has been shown to be related to patients who have
achieved higher ACR response results.[65]
4.3 Mortality
While many components of the ACR response
criteria have been identified to have univariate cor-
relations with mortality (e.g. the inflammatory
marker CRP[66]), Wolfe et al.[4] found the HAQ
alone to be the strongest predictor of mortality.
Therefore, it would not appear that using the ACR in
conjunction with the HAQ would add to the rele-
vance of disease states for predicting mortality.
5. Disease Activity Score
The DAS is a combined index that assesses the
number of swollen and tender joints, the patient’s
general health and a biochemical marker such as the
CRP test.[67] The DAS28 provides a number be-
tween 0 and 10, which describes how active the
disease is based on a count of 28 specific joints.[68] A
derivative of the DAS28 is the European League
Against Rheumatism (EULAR) response, which is
becoming a popular alternative to the ACR crite-
ria.[69] The EULAR response combines both the
level of DAS28 improvement and the absolute level
of DAS28 reached.
Only the study by Welsing et al.[28] defined health
states entirely on DAS. It incorporated four states:
remission, mild, moderate and severe. Kobelt et
al.[19] used DAS scores (high or low) in combination
with the HAQ to describe the health states (table II).
5.1 Health Utility
Health utilities, as measured by the EQ-5D, have
been found to be independently related to DAS
scores (mild/moderate/severe).[28] Welsing et al.[28]
applied these weights to each health state (table II).
However, the relative contribution of DAS in ex-
plaining health utility is quite small when examined
in a multivariate analysis along with the HAQ.[51,70]
This is because joint counts alone, which are a major
Pharmacoeconomics 2008; 26 (5)
 Table II. Associations between clinical measures and economic outcomes in selected studies

402
Study (year of pricing) Clinical States Utility Relationship between clinical measure and economic outcome
measure measure related utilities direct costs direct and indirect costs
(per annum) (per annum)
Marra[36] (2004) HAQ 0 to 3 in increments of EQ-5D –0.20 × HAQ $e0.39 × HAQ $US6079 × HAQa
0.125 (24 states) SF-6D –0.13 × HAQ
HUI2 –0.17 × HAQ
HUI3 –0.29 × HAQ
Chen et al.[27] (2006) HAQ Continuous EQ-5D –0.33 × HAQ NA NA
Spalding and Hay[37] (2004) HAQ Continuous HUI3 –0.28 × HAQ $US874 × HAQ NA
Tanno et al.[34] (2005) HAQ Continuous EQ-5D –0.17 × HAQ ¥227 337 × HAQ NA
Bansback et al.[22] (2001) HAQ Continuous HUI –0.28 × HAQ SEK1416 × HAQ NA
Kobelt et al.[19] (2004) HAQ (DAS) <0.6 (<3.2, >3.2) EQ-5D 0.77 €773 €4242
0.6 < 1.1 (<3.2, >3.2) 0.65 €1590 €8950
1.1 < 1.6 (<3.2, >3.2) 0.54

© 2008 Adis Data Information BV. All rights reserved.

€2456 €12 179
1.6 < 2.1 (<3.2, >3.2) 0.49 €3496 €13 494
>2.1 (<3.2, >3.2) 0.24 €8890 €18 595
Barbieri et al.[31] HAQ 0 VASb 0.87 £2314 NA
Wong et al.[32] (2000) 0.1 0.74 £3270
1.1 0.55 £4848
2.1 0.41 £10 618
Brennan et al.[21] (2004) HAQ Continuous EQ-5D –0.20 × HAQ £860 × HAQ £3434 × HAQ
Chiou et al.[33] ACR <ACR20 VASc 0.53 NA NA
ACR20 0.68
ACR50 0.80
ACR70 0.84
Clark et al.[25] HAQ Continuous EQ-5D –0.33 × HAQ NA NA
Welsing et al.[28] (2002) DAS <1.6 EQ-5D 0.75 €94 €183
1.6–2.4 0.71 €181 €373
2.4–3.7 0.64 €274 €402
>3.7 0.56 €369 €1118
Kobelt et al.[20] (2000) HAQ <0.6 EQ-5D 0.74 £1228 £1376 [2000]
0.6 < 1.1 0.65 £3152 £5676
1.1 < 1.6 0.47 £2091 £5565
1.6 < 2.1 0.44 £3087 £8387
2.1 < 2.6 0.26 £3401 £11 471
>2.6 0.25 £2697 £11 104
Maetzel et al.[40] ACR20 <ACR20 RS/SG 0.71/0.82 NA NA
ACR20 0.77/0.88
Jobanputra et al.[26] HAQ Continuous EQ-5D –0.33 × HAQ NA NA
a Values reported for females.
b Values reported for DMARDs.
c Values reported for no adverse effects.
ACR = American College of Rheumatology; DAS = disease activity score; DMARDs = disease-modifying antirheumatic drugs; HAQ = health assessment questionnaire; HUI =
Health Utilities Index; NA = not available; RS = rating scale; SEK = Swedish krona; SG = standard gamble; VAS = visual analogue scale; ¥ = Japanese yen.

Pharmacoeconomics 2008; 26 (5)

Bansback et al.
Economic Evaluations in Rheumatoid Arthritis
component of the DAS, do not provide a complete
picture of a patient’s health status. In addition, the
DAS does not describe the part of the body that is
affected or the severity of the associated pain, espe-
cially with respect to the activities that the patient
performs in their daily living.
5.2 Direct and Indirect Costs
While Welsing et al.[28] found patients with a
high DAS (DAS > 3.7) consumed twice the health-
care resources of patients with a low DAS (DAS <
2.4), Kobelt et al. found that only HAQ and not DAS
was a significant predictor of resource utilisation in
their study.[51] However, both studies found indirect
costs were closely related to DAS.[28,51] Welsing et
al.[28] reported that patients with severe disease ac-
tivity have indirect costs >4-fold higher than those
for patients with low disease activity.
5.3 Mortality
The smaller evidence base for linking DAS to
mortality must be partly attributable to the more
recent development of the measure, and the length
of data collection necessary to find such associa-
tions. One study[71] did find a univariate link be-
tween DAS and mortality in a cohort of early RA
patients, but this did not remain significant in the
multivariate analysis. Consequently, mortality relat-
ed to disease was not incorporated into the model by
Welsing et al.[28]
6. Discussion
This paper has reviewed the evidence concerning
alternative clinical measures of RA with respect to
their relevance as the basis for the economic evalua-
tion of pharmacological interventions for the treat-
ment of RA. The three key components of a full
economic evaluation are estimates of the differential
costs, survival and QOL effects associated with
alternative interventions. A comprehensive evalua-
tion of the relationship between each clinical mea-
sure and survival, QOL and costs is difficult with the
sparse reporting of comparable statistical informa-
tion within each of the studies reviewed.
© 2008 Adis Data Information BV. All rights reserved.
403
Most commonly, goodness-of-fit measures are
presented to assess how well a model explains the
relationship between two variables, such as a clin-
ical measure and survival. If used, the preferred
measure is the coefficient of determination (R2),
which is the proportion of variability in a data set
that is accounted for by a statistical model. To
account for over-fitted models, the adjusted R2
should be reported. A maximum R2 of 1 indicates
that the fitted model explains all the variability in the
dependent variable, while a value such as R2 = 0.7
indicates that 30% of the variation in the model can
be explained by unknown variables or inherent vari-
ability. However, the preferred approach to assess
the accuracy of a mapping exercise would be to test
the predictive ability of the model, using statistics
such as the mean absolute error (MAE) or the
RMSE. These statistics describe the difference be-
tween the predicted and observed values, with a
value of zero indicating a perfect fit and a large
MAE indicating an inaccurate estimation. The
RMSE is more sensitive than the MAE to large
deviations from the observed values, and the extent
to which the RMSE exceeds the MAE provides an
indication of the extent of the outliers.
Correlation coefficients (rho) are also reported.
These indicate the strength and direction of a linear
relationship between two variables. The strength of
the association has been defined as follows: 0–0.19
as very weak, 0.2–0.39 as weak, 0.40–0.59 as mod-
erate, 0.6–0.79 as strong and 0.8–1 as very strong.
However, these are rather arbitrary limits, and the
context of the results should be considered. For
example, a correlation of 0.9 may be very low if a
physical law is being verified using high-quality
instruments, but may be regarded as very high in the
social sciences where there may be a greater contri-
bution from complicating factors.
Of the three clinical measures that have been
used to inform economic evaluations, it is clear that
the ACR response criteria has limited applicability
because of its focus on relative improvements. The
other measures, the HAQ and DAS, have different
individual properties. The HAQ is derived from
detailed assessments of the impact of RA on pa-
Pharmacoeconomics 2008; 26 (5)
404
tients’ health status, whilst the DAS is based on a
combination of clinical measures as well as an as-
sessment of general health using a VAS. The nature
of the measures means that the HAQ is more repre-
sentative of the impact of RA on HR-QOL, and is
likely to provide a better format for mapping to
utility weights. The HAQ also appears to have at
least as strong relationships with costs (both direct
and indirect) and mortality, although the develop-
ment of models combining the HAQ and DAS may
prove to be a useful future application.
Joint erosions are common features of RA, and
early prevention of the irreversible process can have
long-term consequences on function and thus on
subsequent health utilities and associated costs.[72,73]
Emerging evidence[6] suggests the new biological
DMARDs may prevent the onset of joint erosions.
As neither the HAQ nor DAS incorporate informa-
tion describing the underlying process in patients’
joints, none of the existing models can capture the
full benefits of these treatments. Kobelt et al.[19] and
Brennan et al.[24] attempt to incorporate joint ero-
sions indirectly, through the HAQ. However, the
relationship between joint erosions and function is
complex, and patients can sometimes develop se-
vere erosions without any initial impact on their pain
or function.[74] While health states that incorporate
erosions in addition to function and disease activity
will likely improve the accuracy of cost-effective-
ness models, this would require a standardization of
scoring erosions and the presentation of sufficient
longitudinal data.
The choice of clinical measure used to define
health states in economic models appears to have a
significant effect on the resulting incremental cost-
effectiveness ratio (ICER). In table I, ICERs based
solely on the DAS are higher than those using other
measures. However, there is still a wide variation in
the ICERs from studies that use the HAQ, The
remainder of the discussion focuses on the optimal
application of the HAQ as a framework for econom-
ic evaluations of interventions for RA.
The observed heterogeneity in results from the
HAQ-based models may be partly explained by
alternative frameworks, such as the different utility
© 2008 Adis Data Information BV. All rights reserved.
Bansback et al.
valuation approaches, the number and classification
of health states included, the data and methods used
to estimate rates of progression between health
states, and how pathways have been extrapolated
beyond clinical trial settings.
Most HAQ-based studies incorporated societal
utilities through the use of preference-based instru-
ments. Of these, the EQ-5D has been the most
commonly used measure. However, the HUI3 and
the SF-6D are gaining greater acceptance as evi-
dence demonstrates their ability to describe and
measure the condition. As the SF-36 is a common
measure in RA clinical trials, the SF-6D also has
practical appeal. Marra[36] found the choice of utility
instrument had a large effect on the ICER. In a
comparison of infliximab in combination with
methotrexate versus methotrexate alone, the incre-
mental QALYs ranged from 1.95 when using the
HUI3, to 0.89 when using the SF-6D; this resulted in
ICERs of $US32 018 and $US69 826 per QALY,
respectively (year 2004 values). Consequently, the
choice of utility instrument might be even more
influential than the choice of clinical measure to
describe a health state.[75]
With respect to describing HAQ progression, two
general methods were used: discrete and continuous.
In assigning HAQ states, the primary motivation is
to distinguish between states that have differential
cost and/or utility implications. The discrete ap-
proach requires the specification of an exhaustive
set of health states that cover the range of possible
HAQ values. A continuous representation provides
the most disaggregated specification of HAQ pro-
gression. However, the implementation of the pre-
ferred method is dependent on the availability of
satisfactory data to populate the model. These data
include continuous estimates of disease progression
and continuous estimates of costs and utility weights
as a function of HAQ responses.
The choice to model discrete or continuous HAQ
states would appear to be less dependent on the
relative sensitivity with utilities, direct or indirect
costs, or mortality, but instead on the method to
represent conditional transition probabilities be-
tween states. In RA, patients follow a sequence of
Pharmacoeconomics 2008; 26 (5)
Economic Evaluations in Rheumatoid Arthritis
therapies, and as each differentially affects HAQ
progression and joint erosions, the magnitude of
initial improvement on a new therapy and subse-
quent progression is related to previous treatments,
and levels of HAQ attained. Incorporating this in a
state transition approach requires a series of tunnel
states to represent the time of withdrawal from each
treatment, and conditional probabilities associated
to each subsequent transition. Whilst feasible, this
approach leads to models with enormous numbers of
health states, and requires difficult analyses from
large data sources to populate parameter estimates.
The alternative approach, as described by Barton et
al.,[76] is to utilize the individual sampling approach
where transitions are calculated for a hypothetical
patient, one at a time. In this, the ‘history’ of the
patient is more simply tracked, and conditional tran-
sitions easily populated and calculated. While this
does not preclude using discrete states of HAQ, this
is no longer a requirement, and continuous relation-
ships between HAQ and economic endpoints can be
utilized.
A circumstance where a difference between the
two approaches is clearly demonstrated is in the
representation of disease progression following the
discontinuation of a therapy before a patient is
switched to the next therapy in the sequence. With
very limited evidence to populate such transitions,
the most realistic approach is that on withdrawal the
patient’s pain and function rebounds to a level cal-
culated using the size of initial improvement at-
tained on the treatment, and the amount of progres-
sion realized whilst on that particular treatment.[21,22]
The subsequent improvement and progression for
the next treatment in the sequence is then calculated
based on this level. As this is difficult to implement
within a Markov framework, analysts have em-
ployed a variety of assumptions to model rebound.
The majority have optimistically assumed no re-
bound, thus the benefits of a treatment are realised
long beyond its withdrawal.[20]
Common to most of the models is the assumption
of linearity when using changes in clinical measures
to estimate corresponding changes in HR-QOL,
costs or mortality risks. In particular, in conditions
© 2008 Adis Data Information BV. All rights reserved.
405
such as RA where hospitalization costs are involved,
the direct healthcare costs can increase rapidly for
patients with severe disease, and the distribution of
costs is likely to be positively skewed. The impact of
this is that a slight improvement for a patient with
severe RA could be associated with a greater cost
offset than a similar improvement in a patient at the
other end of the disease spectrum where the poten-
tial to save costs is smaller. Similarly, an absolute
change in a clinical measure may have a very differ-
ent effect on mortality risk or HR-QOL, depending
on the baseline measurement. The use of a linear
relationship across the full disease range could un-
derestimate the benefits of treatments in patients
with severe RA or overestimate the benefits of treat-
ments in patients with less severe disease. An alter-
native could be to derive different relationships for
different levels of the disease.
It should be noted that the exact workings of
some of the individual sampling models is often not
well described, and sometimes the inner workings
are only described in large appendices (e.g. Wailoo
et al.[23]). It should be noted that there has been no
attempt to validate the results of any individual
sampling model with an external data source. In
fact, the only model included in our review to do so
was the DAS-based model by Welsing et al.[77]
7. Conclusions
Since the development of the effective yet expen-
sive biological agents for RA, a plethora of econom-
ic evaluations have been published. The hetero-
geneity in results of each study can be linked to the
differences in the types of health states used to
model progression of disease, and the utility values
and costs attributed to each state. Evidence suggests
the HAQ to be the primary clinical measure, since it
is measured in nearly all clinical studies, and is
closely correlated to health utilities, mortality and
both direct and indirect costs. New developments
suggesting combining the HAQ with other outcome
measures such as the DAS and/or joint erosions
might further improve the sensitivity of health
states. However, until robust evidence is available,
this would increase the uncertainty in the results.
Pharmacoeconomics 2008; 26 (5)
406 Bansback et al.

Consequently, future studies should compare the 3. Pincus T, Callahan LF. The ’side effects’ of rheumatoid arthri-
tis: joint destruction, disability and early mortality. Br J Rheu-
differences between evaluations using the HAQ matol 1993 Mar; 32 Suppl. 1: 28-37
alone, and HAQ in combination with these other 4. Wolfe F, Michaud K, Gefeller O, et al. Predicting mortality in
patients with rheumatoid arthritis. Arthritis Rheum 2003 Jun;
measures. Finally, the use of continuous representa- 48 (6): 1530-42
tions of the HAQ versus discrete states in economic 5. Kvien TK. Epidemiology and burden of illness of rheumatoid
evaluations requires further examination. In the ab- arthritis. Pharmacoeconomics 2004; 22 (2 Suppl.): 1-12
sence of relevant evidence, a suitable approach 6. O’Dell JR. Therapeutic strategies for rheumatoid arthritis.
N Engl J Med 2004 Jun 17; 350 (25): 2591-602
might be to include alternative extrapolation as- 7. Michaud K, Messer J, Choi HK, et al. Direct medical costs and
sumptions in primary structural sensitivity analyses. their predictors in patients with rheumatoid arthritis: a three-
year study of 7,527 patients. Arthritis Rheum 2003 Oct; 48
(10): 2750-62
Acknowledgements 8. Sorensen J, Andersen LS. The case of tumour necrosis factor-
alpha inhibitors in the treatment of rheumatoid arthritis: a
Jonathan Karnon has received an honorarium for speaking budget impact analysis. Pharmacoeconomics 2005; 23 (3):
289-98
at a clinical meeting sponsored by Wyeth Pharmaceuticals.
9. Access to anti-TNF alpha therapies for adults with inflammatory
The remaining authors have no conflicts of interest to declare. arthritis: a report by the British Society for Rheumatology and
No sources of funding were used to assist in the preparation the Arthritis & Musculoskeletal Alliance, 2005 [online].
of this review. Available from URL: http://www.rheumatology.org.uk/public
_affairs/armabsrtnfsurvey [Accessed 2008 19 Mar]
10. Moreland LW. Biologic therapies on the horizon for rheumatoid
Appendix 1: EMBASE Search arthritis. J Clin Rheumatol 2004 Jun; 10 (3 Suppl.): S32-9
11. Drummond MF. Methods for the economic evaluation of health
1. rheumatoid arthritis/ care programmes. 3rd ed. Oxford (NY): Oxford University
Press, 2005
2. tum?r necrosis factor.mp. 12. Karnon J, Brennan A, Akehurst R. A critique and impact ana-
3. exp receptors tumor necros lysis of decision modeling assumption. Med Decis Making
2007; 27 (4): 491-9
4. anti tnf.mp.
13. Canadian Coordinating Office for Health Technology Assess-
5. disease modifying anti rheumatic drugs.mp ment. Guidelines for economic evaluation of pharmaceuticals:
6. or/2-5 Canada. 3rd ed. Ottawa: The Office, 2006
14. Australian Government, Department of Health and Ageing.
7. 1 and 6 Guidelines for preparing submissions to the Pharmaceutical
8. cost benefit analysis/ Benefits Advisory Committee [PBAC] (version 4.2) [online].
Available from URL: http://www.health.gov.au/internet/main/
9. cost effectiveness analysis/ publishing.nsf/content/pbacguidelines-index [Accessed 2008
10. cost minimization analysis/ Apr 11]
11. cost utility analysis/ 15. National Institute for Clinical Excellence (NICE). Guide to the
methods of technology appraisal. London: NICE, 2004
12. economic evaluation/ 16. Briggs AH. Decision modelling for health economic evaluation.
13. (cost or costs).tw Oxford (UK): Oxford University Press, 2006
14. (economic$ or pharmacoeconomic$ or price or 17. Bansback NJ, Regier DA, Ara R, et al. An overview of econom-
ic evaluations for drugs used in rheumatoid arthritis: focus on
pricing).tw. tumour necrosis factor-alpha antagonists. Drugs 2005; 65 (4):
15. or/8-14 473-96
18. Karnon J. Cost-effectiveness of letrozole, anastrozole, and ex-
16. 7 and 15 emestane for early breast cancer. Expert Rev Pharmacoeco-
17. limit 16 to yr=1990–2007 nomics Outcomes 2007; 7 (2): 143-53
19. Kobelt G, Lindgren P, Singh A, et al. Cost effectiveness of
etanercept (Enbrel) in combination with methotrexate in the
References treatment of active rheumatoid arthritis based on the TEMPO
1. Symmons D, Turner G, Webb R, et al. The prevalence of trial. Ann Rheum Dis 2005 Aug; 64 (8): 1174-9
rheumatoid arthritis in the United Kingdom: new estimates for 20. Kobelt G, Jonsson L, Young A, et al. The cost-effectiveness of
a new century. Rheumatology (Oxford) 2002 Jul; 41 (7): infliximab (Remicade) in the treatment of rheumatoid arthritis
793-800 in Sweden and the United Kingdom based on the ATTRACT
2. Yelin E, Callahan LF. The economic cost and social and psy- study. Rheumatology (Oxford) 2003 Feb; 42 (2): 326-35
chological impact of musculoskeletal conditions. National Ar- 21. Brennan A, Bansback N, Reynolds A, et al. Modelling the cost-
thritis Data Work Groups. Arthritis Rheum 1995 Oct; 38 (10): effectiveness of etanercept in adults with rheumatoid arthritis
1351-62 in the UK. Rheumatology (Oxford) 2004 Jan; 43 (1): 62-72

© 2008 Adis Data Information BV. All rights reserved. Pharmacoeconomics 2008; 26 (5)
Economic Evaluations in Rheumatoid Arthritis
22. Bansback NJ, Brennan A, Ghatnekar O. Cost effectiveness of
adalimumab in the treatment of patients with moderate to
severe rheumatoid arthritis in Sweden. Ann Rheum Dis 2005
Jul; 64 (7): 995-1002
23. Wailoo AJ, Bansback N, Brennan A, et al. Biologic drugs for
rheumatoid arthritis in the Medicare program: a cost-effective-
ness analysis. Arthritis Rheum 2008 Mar 27; 58 (4): 939-46
24. Brennan A, Bansback NJ, Nixon RM, et al. Modelling the cost
effectiveness of TNF alpha antagonists in the management of
rheumatoid arthritis: results from the British Society for Rheu-
matology Biologics Registry. Rheumatology (Oxford) 2007;
16 (8): 1345-54
25. Clark W, Jobanputra P, Barton P, et al. The clinical and cost-
effectiveness of anakinra for the treatment of rheumatoid ar-
thritis in adults: a systematic review and economic analysis.
Health Technol Assess 2004 May; 8 (18): iii-iv, ix-x, 1-105
26. Jobanputra P, Barton P, Bryan S, et al. The effectiveness of
infliximab and etanercept for the treatment of rheumatoid
arthritis: a systematic review and economic evaluation. Health
Technol Assess 2002; 6 (21): 1-110
27. Chen YF, Jobanputra P, Barton P, et al. A systematic review of
the effectiveness of adalimumab, etanercept and infliximab for
the treatment of rheumatoid arthritis in adults and an economic
evaluation of their cost-effectiveness. Health Technol Assess
2006 Nov; 10 (42): iii-iv, xi-xiii, 1-229
28. Welsing PM, Severens JL, Hartman M, et al. Modeling the
5-year cost effectiveness of treatment strategies including tu-
mor necrosis factor-blocking agents and leflunomide for treat-
ing rheumatoid arthritis in the Netherlands. Arthritis Rheum
2004 Dec 15; 51 (6): 964-73
29. Choi HK, Seeger JD, Kuntz KM. A cost effectiveness analysis
of treatment options for methotrexate-naive rheumatoid arthri-
tis. J Rheumatol 2002 Jun; 29 (6): 1156-65
30. Choi HK, Seeger JD, Kuntz KM. A cost-effectiveness analysis
of treatment options for patients with methotrexate-resistant
rheumatoid arthritis. Arthritis Rheum 2000 Oct; 43 (10):
2316-27
31. Barbieri M, Wong JB, Drummond M. The cost effectiveness of
infliximab for severe treatment-resistant rheumatoid arthritis
in the UK. Pharmacoeconomics 2005; 23 (6): 607-18
32. Wong JB, Singh G, Kavanaugh A. Estimating the cost-effec-
tiveness of 54 weeks of infliximab for rheumatoid arthritis.
Am J Med 2002 Oct 1; 113 (5): 400-8
33. Chiou CF, Choi J, Reyes C. Cost-effectiveness of biological
treatments for rheumatoid arthritis. Expert Rev Pharmacoeco-
nomics Outcomes 2004; 4 (3): 307-15
34. Tanno M, Nakamura I, Ito K, et al. Modeling and cost-effective-
ness analysis of etanercept in adults with rheumatoid arthritis
in Japan: a preliminary analysis. Mod Rheumatol 2006; 16 (2):
77-84
35. Coyle D, Judd M, Blumenauer B, et al. Infliximab and etaner-
cept in patients with rheumatoid arthritis: a systematic review
and economic evaluation [technology report no. 64]. Ottawa:
Canadian Coordinating Office for Health Technology Assess-
ment, 2006
36. Marra CA. Not all ‘quality-adjusted life years’ are equal. J Clin
Epidemiol 2007; 60 (6): 616-24
37. Spalding JR, Hay J. Cost effectiveness of tumour necrosis
factor-alpha inhibitors as first-line agents in rheumatoid arthri-
tis. Pharmacoeconomics 2006; 24 (12): 1221-32
38. Schädlich PK, Zeidler H, Zink A, et al. Modelling cost effec-
tiveness and cost utility of sequential DMARD therapy includ-
ing leflunomide for rheumatoid arthritis in Germany: II. The
© 2008 Adis Data Information BV. All rights reserved.
407
contribution of leflunomide to efficiency. Pharmacoeconomics
2005; 23 (4): 395-420
39. Kobelt G, Lindgren P, Young A. Modelling the costs and effects
of leflunomide in rheumatoid arthritis. Eur J Health Econ
2002; 3 (3): 180-7
40. Maetzel A, Strand V, Tugwell P, et al. Cost effectiveness of
adding leflunomide to a 5-year strategy of conventional dis-
ease-modifying antirheumatic drugs in patients with rheuma-
toid arthritis. Arthritis Rheum 2002; 47 (6): 655-61
41. Fries JF, Spitz PW, Young DY. The dimensions of health
outcomes: the health assessment questionnaire, disability and
pain scales. J Rheumatol 1982 Sep-Oct; 9 (5): 789-93
42. Bruce B, Fries JF. The Stanford Health Assessment Question-
naire: a review of its history, issues, progress, and documenta-
tion. J Rheumatol 2003 Jan; 30 (1): 167-78
43. Marra CA, Woolcott JC, Kopec JA, et al. A comparison of
generic, indirect utility measures (the HUI2, HUI3, SF-6D, and
the EQ-5D) and disease-specific instruments (the RAQoL and
the HAQ) in rheumatoid arthritis. Soc Sci Med 2005 Apr; 60
(7): 1571-82
44. Wells GA, Tugwell P, Kraag GR, et al. Minimum important
difference between patients with rheumatoid arthritis: the pa-
tient’s perspective. J Rheumatol 1993 Mar; 20 (3): 557-60
45. Kosinski M, Zhao SZ, Dedhiya S, et al. Determining minimally
important changes in generic and disease-specific health-relat-
ed quality of life questionnaires in clinical trials of rheumatoid
arthritis. Arthritis Rheum 2000 Jul; 43 (7): 1478-87
46. Bansback NJ, Marra CA, Tsuchiya A, et al. Using the Health
Assessment Questionnaire to estimate preference-based single
indices in patients with rheumatoid arthritis. Arthritis Rheum
2007; 57 (6): 2963-7
47. Wolfe F, Michaud K. HAQ-based utilities and SF6D systemati-
cally overvalue quality of life (QOL) in RA patients with
severe RA, pain and psychological distress [abstract]. Arthritis
Rheum 2003; 48 (9): 982
48. Yelin E, Wanke LA. An assessment of the annual and long-term
direct costs of rheumatoid arthritis: the impact of poor function
and functional decline. Arthritis Rheum 1999 Jun; 42 (6):
1209-18
49. Leardini G, Salaffi F, Montanelli R, et al. A multicenter cost-of-
illness study on rheumatoid arthritis in Italy. Clin Exp Rheu-
matol 2002 Jul-Aug; 20 (4): 505-15
50. Kobelt G, Eberhardt K, Jonsson L, et al. Economic conse-
quences of the progression of rheumatoid arthritis in Sweden.
Arthritis Rheum 1999 Feb; 42 (2): 347-56
51. Kobelt G, Lindgren P, Lindroth Y, et al. Modelling the effect of
function and disease activity on costs and quality of life in
rheumatoid arthritis. Rheumatology (Oxford) 2005 Sep; 44
(9): 1169-75
52. Clarke AE, Zowall H, Levinton C, et al. Direct and indirect
medical costs incurred by Canadian patients with rheumatoid
arthritis: a 12 year study. J Rheumatol 1997 Jun; 24 (6):
1051-60
53. Wolfe F, Zwillich SH. The long-term outcomes of rheumatoid
arthritis: a 23-year prospective, longitudinal study of total joint
replacement and its predictors in 1600 patients with rheuma-
toid arthritis. Arthritis Rheum 1998 Jun; 41 (6): 1072-82
54. Pugner KM, Scott DI, Holmes JW, et al. The costs of rheuma-
toid arthritis: an international long-term view. Semin Arthritis
Rheum 2000 Apr; 29 (5): 305-20
55. Rat AC, Boissier MC. Rheumatoid arthritis: direct and indirect
costs. Joint Bone Spine 2004 Nov; 71 (6): 518-24
Pharmacoeconomics 2008; 26 (5)
408
56. Burton W, Morrison A, Maclean R, et al. Systematic review of
studies of productivity loss due to rheumatoid arthritis. Occup
Med (Lond) 2006 Jan; 56 (1): 18-27
57. Farragher TM, Lunt M, Bunn DK, et al. Early functional disabil-
ity predicts both all-cause and cardiovascular mortality in
people with inflammatory polyarthritis: results from the Nor-
folk Arthritis Register. Ann Rheum Dis 2007 Apr; 66 (4):
486-92
58. Sokka T, Hakkinen A, Krishnan E, et al. Similar prediction of
mortality by the health assessment questionnaire in patients
with rheumatoid arthritis and the general population. Ann
Rheum Dis 2004 May; 63 (5): 494-7
59. Yelin E, Trupin L, Wong B, et al. The impact of functional
status and change in functional status on mortality over 18
years among persons with rheumatoid arthritis. J Rheumatol
2002 Sep; 29 (9): 1851-7
60. Felson DT, Anderson JJ, Boers M, et al. The American College
of Rheumatology preliminary core set of disease activity mea-
sures for rheumatoid arthritis clinical trials. The Committee on
Outcome Measures in Rheumatoid Arthritis Clinical Trials.
Arthritis Rheum 1993 Jun; 36 (6): 729-40
61. Felson DT, Anderson JJ, Lange ML, et al. 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 Sep; 41 (9):
1564-70
62. O’Dell JR. The horseless carriage: moving forward with the
hybrid ACR. Arthritis Rheum 2007 Feb 28; 57 (2): 189-90
63. Verhoeven AC, Boers M, van Der Linden S. Responsiveness of
the core set, response criteria, and utilities in early rheumatoid
arthritis. Ann Rheum Dis 2000 Dec; 59 (12): 966-74
64. Chiou CF, Weisman M, Sherbourne CD, et al. Measuring pref-
erence weights for American college of rheumatology res-
ponse criteria for patients with rheumatoid arthritis. J Rheu-
matol 2005 Dec; 32 (12): 2326-9
65. Yelin E, Trupin L, Katz P, et al. Association between etanercept
use and employment outcomes among patients with rheuma-
toid arthritis. Arthritis Rheum 2003 Nov; 48 (11): 3046-54
66. Cook NR, Buring JE, Ridker PM. The effect of including C-
reactive protein in cardiovascular risk prediction models for
women. Ann Intern Med 2006 Jul 4; 145 (1): 21-9
67. van der Heijde DM, van’t Hof MA, van Riel PL, et al. Judging
disease activity in clinical practice in rheumatoid arthritis: first
step in the development of a disease activity score. Ann Rheum
Dis 1990 Nov; 49 (11): 916-20
68. Prevoo ML, van’t Hof MA, Kuper HH, et al. Modified disease
activity scores that include twenty-eight-joint counts: develop-
© 2008 Adis Data Information BV. All rights reserved.
Bansback et al.
ment and validation in a prospective longitudinal study of
patients with rheumatoid arthritis. Arthritis Rheum 1995 Jan;
38 (1): 44-8
69. van Gestel AM, Prevoo ML, van’t Hof MA, et al. Development
and validation of the European League Against Rheumatism
response criteria for rheumatoid arthritis: comparison with the
preliminary American College of Rheumatology and the
World Health Organization/International League Against
Rheumatism criteria. Arthritis Rheum 1996 Jan; 39 (1): 34-40
70. Kobelt G, Eberhardt K, Geborek P. TNF inhibitors in the
treatment of rheumatoid arthritis in clinical practice: costs and
outcomes in a follow up study of patients with RA treated with
etanercept or infliximab in southern Sweden. Ann Rheum Dis
2004 Jan; 63 (1): 4-10
71. Young A, Koduri G, Batley M, et al. Mortality in rheumatoid
arthritis: increased in the early course of disease, in ischaemic
heart disease and in pulmonary fibrosis. Rheumatology (Ox-
ford) 2007 Feb; 46 (2): 350-7
72. Aletaha D, Smolen J, Ward MM. Measuring function in rheu-
matoid arthritis: identifying reversible and irreversible com-
ponents. Arthritis Rheum 2006 Sep; 54 (9): 2784-92
73. Kirwan JR. Links between radiological change, disability, and
pathology in rheumatoid arthritis. J Rheumatol 2001 Apr; 28
(4): 881-6
74. Scott DL, Smith C, Kingsley G. Joint damage and disability in
rheumatoid arthritis: an updated systematic review. Clin Exp
Rheumatol 2003 Sep-Oct; 21 (5 Suppl. 31): S20-7
75. Drummond MF, Barbieri M, Wong JB. Analytic choices in
economic models of treatments for rheumatoid arthritis: what
makes a difference? Med Decis Making 2005 Sep-Oct; 25 (5):
520-33
76. Barton P, Jobanputra P, Wilson J, et al. The use of modelling to
evaluate new drugs for patients with a chronic condition: the
case of antibodies against tumour necrosis factor in rheumatoid
arthritis. Health Technol Assess 2004 Mar; 8 (11): iii, 1-91
77. Welsing PM, Severens JL, Hartman M, et al. The initial valida-
tion of a Markov model for the economic evaluation of (new)
treatments for rheumatoid arthritis. Pharmacoeconomics 2006;
24 (10): 1011-20
Correspondence: Nick Bansback, Centre for Health Evalua-
tion and Outcomes Sciences, St Paul’s Hospital, 620B-1081
Burrard Street, Vancouver, BC V6Z 1Y6, Canada.
E-mail: nbansback@cheos.ubc.ca
Pharmacoeconomics 2008; 26 (5)