Full text of DOI:10.2165/00019053-200119110-00003

Pharmacoeconomics 2001; 19 (11): 1091-1102
1170-7690/01/0011-1091/$22.00/0
ORIGINAL RESEARCH ARTICLE
© Adis International Limited. All rights reserved.
Cost Effectiveness of Treatment
Options in Advanced Breast Cancer
in the UK
Ruth E. Brown,1 John Hutton1 and Anita Burrell2
1
2
MEDTAP® International, Inc., London, UK
Aventis, Bridgewater, New Jersey, USA
Objective: To compare clinical and economic study data for docetaxel, paclitaxel
and vinorelbine in the treatment of anthracycline-resistant advanced breast can-
cer.
Abstract
Study design and methods: A Markov decision-analysis model to simulate the
clinical course of a ‘typical’ patient with advanced breast cancer during salvage
chemotherapy was updated with response rates and adverse effect rates from
phase III clinical trial data for docetaxel, paclitaxel and vinorelbine. Costs were
taken from UK national databases and hospitals. Utilities were estimated from
30 oncology nurses in the UK using the standard gamble method.
Perspective: National Health Service.
Results: When compared with other chemotherapeutic agents, docetaxel has been
shown to increase response rate, time to progression and survival in patients with
advanced breast cancer. In the base-case analysis, the incremental cost-utility
ratio for docetaxel versus paclitaxel was £1995 per quality-adjusted life year
(QALY) gained (1998 values). The incremental cost-utility ratio for docetaxel
versus vinorelbine was £14 055 per QALY gained. In the comparison with vino-
relbine, docetaxel provided the equivalent of an additional 92 days of perfect
health. Sensitivity analyses confirmed the robustness ofthe model and the validity
of the base-case analysis results. Even in the worst case scenarios, docetaxel
remained cost effective compared with paclitaxel and vinorelbine.
Conclusions: These findings support the use of the taxoids, notably docetaxel,
in the management of advanced breast cancer.
There is currently no standard salvage chemo-
therapies have provided limited benefit and prog-
nosis has remained poor.[2] As a consequence, the
search for new drugs and treatment strategies con-
tinues in the hope of improving antitumour activ-
ity, time to disease progression, tolerability and
survival.
More recently, promising agents have become
available, namely the taxoids paclitaxel and doce-
taxel, and the vinca alkaloid, vinorelbine. The effi-
cacy and acceptable tolerability of these agents
therapy regimen for use in patients with locally ad-
vanced or metastatic breast cancer after therapy with
anthracyclines and alkylators has failed.[1] Treat-
ment options have included cyclophosphamide,
methotrexate and fluorouracil; continuous-infu-
sion fluorouracil; methotrexate and fluorouracil/
leucovorin; doxorubicin; mitomycin and vinblas-
tine either as single agents or in combination; and
platinum combinations. However, these salvage

1092
Brown et al.
have been confirmed in randomised phase III clin-
ical trials.[1,3-6] Docetaxel, paclitaxel and vinorelb-
ine are licensed for use in advanced breast cancer
within the UK, and are the three major salvage che-
motherapy options for patients with disease pro-
gression following failure of adjuvant or first-line
chemotherapy with anthracyclines. In addition,
docetaxel can also be used after failure with al-
kylators.
An increased emphasis on evidence-based med-
icine and justification for the use of healthcare in-
terventions on the basis of economic as well as clin-
ical factors has led to a broader evaluation of new
therapies. Advanced breast cancer is regarded as
incurable and, despite treatment, total survival time
is often less than a year.[7] For these patients, palli-
ation of symptoms, minimising adverse effects of
therapy and maximising quality of life are the ma-
jor issues in selecting a treatment strategy. The re-
strictions of healthcare budgets and the importance
of quality of life in terminally ill patients have led
to the development of new methods for assessing
the cost effectiveness of therapies. Since the thrust
of drug development in oncology is in establishing
improved efficacy and tolerability of novel agents,
the collection of empirical cost-effectiveness data
tends to be neglected. However, this information
may be derived using decision-analysis models that
make use of available evidence and expert opinion.
In such circumstances economic evaluation
requires an outcome measure which takes into
account both survival and quality of life. This is
provided by the ‘utility’ which measures an indi-
vidual’s preference for time spent in different
health states with varying quality of life. The most
commonly used measure of utility is the quality-
adjusted life-year (QALY), which accumulates the
patient’s survival time, weighted by the utility
associated with each different health state experi-
enced by the patient. Utilities are measured on a 0
to 1 scale, where 1 is the best imaginable health
state. Thus, unless the patient is maintained in per-
fect health the number of QALYs will always be
less than the simple survival time. An advantage of
a generalised outcome measure like the QALY is
that it allows comparison of interventions within
and across therapeutic areas in terms of cost per
QALY gained.[8]
A computer-based decision-analysis model has
estimated the cost utility of docetaxel versus
Table I. Outcome estimates for metastatic breast cancer treatment used in the base-case analyses
Parameter
Docetaxela
Paclitaxelb
Vinorelbinec
Overall response rate (%)
41.7
41.7
23.8
24
28
28
21
21
46
4
16
Evaluable patient response rate (%)
Progressive disease (%)
16
38.4
Time to progression in weeks
Median survival in weeks
12
56
36
Infection with hospitalisation and/or intravenous antibacterials (%)
Febrile neutropenia with hospitalisation (%)
Neutropenia without hospitalisation (%)
Death associated with infection or febrile neutropenia (%)
Severe neurotoxicity (%)
8.8
7.3
10.7
1.2
5
10
0d
0d
0d
7
10
0d
9
2
Severe oedema (%)
4
0d
0d
0d
9
0d
0d
0d
Severe skin condition (%)
3
Severe nail condition (%)
3.4
9
Time to response in weeks
6
Courses of chemotherapy
6 (at 3-week intervals) 6 (at 3-week intervals)
12 (weekly)
a
b
c
d
Data were derived from randomised phase III clinical trials.[1,4,6]
Efficacy data were derived from a randomised phase III clinical trial;[3] safety data were derived from Phase II clinical trials.[11-16]
Data were derived from a randomised phase III clinical trial.[5]
Data not reported in an appropriate trial – for the purposes of the model, a value of zero was assumed.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (11)

Cost Effectiveness of Therapy for Advanced Breast Cancer
1093
Table II. Health state utilities for metastatic breast cancer
paclitaxel based on phase II clinical trial data for
these agents.[9] Since phase III clinical trial data are
now available for docetaxel and paclitaxel,[1,3,4,6]
an updated analysis incorporating these new data
has been conducted. In addition, the aforemen-
tioned describes a cost-utility analysis of docetaxel
versus vinorelbine using published phase III clini-
cal trial data for these agents.[1,4-6]
Health state
UK utility (SD)
Start of second-line therapy
Partial/complete response
Stable disease
0.64 (0.15)
0.84 (0.12)
0.62 (0.22)
0.33 (0.24)
0.13 (0.12)
0.62 (0.16)
Progressive disease
Terminal disease
Peripheral neuropathy with partial/complete
response
Severe oedema with partial/complete
response
0.78 (0.15)
0.56
The objectives of this study were to determine
the incremental cost-utility ratios of docetaxel ver-
sus paclitaxel and docetaxel versus vinorelbine in
the treatment of patients with advanced breast can-
cer within the UK-licensed indications. The first
analysis compared the only two taxoids currently
licensed in the UK for the treatment of advanced
breast cancer. This comparison aimed to reflect a
choice faced by oncologists practising in this area
and other healthcare decision makers. The second
analysis compared docetaxel, for which the most
promising clinical data are available, with vino-
relbine, another recently introduced agent. Despite
poorer response and survival rates, vinorelbine
tends to be used because its unit cost is less than
that of the taxoids.
Severe skin condition with partial/complete
responsea
Febrile neutropenia and hospitalised
Infection without hospitalisation
Death
0.24 (0.12)
0.48
0
a
Estimated from other toxicities.
SD = standard deviation.
breast cancer within the UK licensed indication has
been published.[3] This trial examined both anthra-
cycline-resistant (those with tumour progression
in spite of anthracycline treatment) and anthracy-
cline-naïve patients; this latter group falls outside
the confines of the UK licence. While the report of
this phase III clinical trial presented the overall
response rates for the subgroups separately thus
allowing use in the construction of the base-case
scenario for paclitaxel, the safety data were aggre-
gated and inappropriate for inclusion in the model.
In the absence of phase III safety data for single-
agent paclitaxel in anthracycline-resistant patients,
weighted averages of the safety data from phase II
clinical trials were calculated for the purposes of
this study.[11-16] Weights were calculated based on
the number of patients enrolled in each study as
described above for docetaxel.
Methods
Data Sources and Management
The efficacy and safety profiles of docetaxel,
paclitaxel and vinorelbine, as treatments for ad-
vanced breast cancer, have recently been com-
pared.[10] These findings have been used in the cur-
rent analysis.
Docetaxel versus Paclitaxel
Three randomised phase III clinical trials on the
use of docetaxel in patients with advanced breast
cancer were used to construct the base-case scen-
ario for docetaxel.[1,4,6] For each trial, the response
rate (complete and partial) and adverse effect rates
were weighted by the number of patients random-
ised. These weighted values were then used to cal-
culate weighted average efficacy (overall response
rate) and adverse effect rates for docetaxel.
Only one randomised phase III clinical trial on
the use of paclitaxel in patients with advanced
Docetaxel versus Vinorelbine
The weighted average efficacy and adverse
effect rates for docetaxel for use in the base-case
analysis were those derived from the three random-
ised phase III clinical trials as described above.
Weighted average efficacy and adverse effect
rates for construction of the base-case scenario for
vinorelbine were calculated using data from the
only published randomised phase III clinical trial
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (11)

1094
Brown et al.
designed to evaluate the use of single-agent vino-
relbine in anthracycline-resistant advanced breast
cancer.[5]
preference. Nurses were unaware of the chemo-
therapies being compared and were not paid for
participating in the study. The UK nurse mean re-
sponses were used in the basic analysis and sen-
sitivity analyses were conducted on the range of
responses and the mean responses from the 150
Western European nurses. As the patients in the
model experience the different health states, the
utilities increase or decrease as appropriate.
Cost variables for use in the model were devel-
oped in collaboration with leading oncologists so
that they accurately reflected current treatment
practices within the UK. For each health state spec-
ified in the model, the direct healthcare resource
use was defined by one oncologist and reviewed by
four other oncologists. The treatment pattern and
resource use for each health state was defined by
the overall views of the five oncologists. The treat-
ment algorithms for the different health states were
designed to reflect resource use during a 3-week
period.
Probabilities, Costs and Utilities
Probabilities of the occurrence of outcome vari-
ables in the base-case analysis were derived from
clinical trial data and are shown in table I.
In healthcare, utilities are a measure of the pa-
tient’s preferences for health states.[17] Prospective
patient utilities were not collected during the trials.
Consequently, the best option available was the use
of proxy respondents. Proxy utility values were
obtained from 30 oncology nurses from specialist
cancer centres in the UK and from 150 nurses in
four other Western European countries. The stand-
ard gamble method was used to obtain rankings of
the health states based on utilities.[18] The utilities
are presented in table II and are unique to the health
states regardless of the treatment. The health states
were defined without reference to particular che-
motherapy treatment to enable blind estimates of
Table III. Resource costs used in the base-case analyses for metastatic breast cancer
Resource
Unit costa (£ sterling;
1999 values for drug
costs, 1997/1998 values
for other resources)
Docetaxel per 3-week course
1150
Paclitaxel per 3-week course
1122
Vinorelbine per 3-week course
441
Chemotherapy administration per visit (outpatient clinic)
Hospital day (i.e. 1 day in hospital) [regular]
62
270
Hospital day (i.e. 1 day in hospital) [intensive care]
GP consultation
919
14
Specialist consultation
50
Combined resources
State cost (sterling) b
Early progressive diseasec (palliative medicines, GP visit, 50% with laboratory tests, x-ray, ultrasound)
Late progressive diseasec (palliative medicines, weekly GP visit, 80% with nurse home visits twice weekly)
Stable diseasec (palliative medicines, GP visit, 50% with laboratory tests and x-ray, 60% with outpatient visit)
154
267
126
372
Terminal diseasec (palliative medicine, twice-weekly GP visit, 80% with nurse home visits twice weekly,
nutritional supplements daily)
Neutropenia [7 days hospital; IV outpatient; oral antibacterial (depending on severity of neutropenia)]
Severe neurotoxicityc (GP visit, 50% with laboratory tests, outpatient visit)
Severe oedemac [GP visit, 50% with hospital day (i.e. 1 day in hospital), specialist]
2470; 907; 42
124
284
a
b
c
£1 = $US1.55.
Refers to the total costs for the related Markov state.
Cost per 3-week period.
GP = general practitioner; IV = intravenous.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (11)

Cost Effectiveness of Therapy for Advanced Breast Cancer
1095
Unit prices were applied to convert the resource
use patterns into costs for a 3-week period. Where
possible, unit costs were taken from national data-
bases; laboratory costs and costs of chest x-rays
were provided by specific hospitals (Hospital Trust
in Central England, personal communication,
1997). The National Health Service Hospital and
Community Inflation Index was used to convert
these costs to 1997/1998 prices.[19] Drug costs
were calculated using prices taken from the
Monthly Index for Medical Specialties[20] (1999
values) and were based on a nominal body surface
area of 1.75m2 and recommended dosages of 100
mg/m2 for docetaxel, 175 mg/m2 for paclitaxel and
30 mg/m2 for vinorelbine. The updated costs of the
health states in advanced breast cancer have re-
cently been published.[7] The resource costs used
in the model are shown in table III; indirect pro-
ductivity costs were not included in the model.
uations conducted in the UK following the Guid-
ance on Good Practice in the Conduct of Economic
Evaluations of Medicines.[21] Docetaxel and
paclitaxel are each administered as a single dose
every 3 weeks for a total of 6 chemotherapy cycles
and vinorelbine is administered weekly for a total
of 12 chemotherapy cycles. The model assumes
that all patients achieving response receive the full
drug dose. Treatment with paclitaxel involves a 3-
hour dose infusion period while docetaxel is ad-
ministered as a 1-hour infusion. Costs arising from
differences in the administration schedules, which
would increase the costs associated with paclitaxel
more than those associated with docetaxel, were
not included in the analyses. Indirect productivity
costs such as patient time lost from working or in
normal activities, were not included in the analyses
except insofar as they were incorporated into the
utilities.
Disease Process and Outcomes of Treatment
The representative patient considered in the
model had been diagnosed with advanced breast
cancer and had disease progression (increasing
tumour size) and metastases following first-line
chemotherapy with anthracyclines. The patient en-
tered the model and received chemotherapy with
docetaxel, paclitaxel or vinorelbine according to
the licensed dosage and administration schedules.
After a course of the specified treatment, the pa-
tient had one of the following outcomes (classi-
fied according to World Health Organization cri-
teria used in the clinical trials):[22] complete
response, partial response, stable disease, or pro-
gressive disease.
As complete response is rare in patients with
advanced breast cancer, complete and partial re-
sponses were combined for the purposes of the
model to provide the overall response rate. Since
patients with advanced breast cancer also have a
poor prognosis, those with a response or stable dis-
ease eventually develop progressive disease lead-
ing to death.
The Decision-Analysis Model
Model Design
A previously reported decision-analysis model
was used for the cost-utility analyses in this
study.[9] The model was validated by a number of
practising oncologists at leading cancer treatment
centres in the UK to ensure that it accurately rep-
resented the care of patients with advanced breast
cancer. The model used the Markov process to sim-
ulate the clinical course of a representative patient
with advanced breast cancer through discrete dis-
ease states and toxicities. A ‘typical’ patient is dif-
ficult to define because of the highly complex na-
ture of the disease and its treatment. Expert clinical
advice was therefore sought to produce the defini-
tion and validate the disease states and toxicities
that might be encountered in successive cycles of
the model. A probability, a cost and a utility were
determined for each disease state and toxicity.
The cost-utility analyses considered only direct
medical costs over a time frame of 3 years from the
initiation of salvage therapy. A discount rate of 6%
was applied to all costs after the first year. This
discount rate was based on the UK Treasury rate,
and is the value commonly used in economic eval-
The model also took into account various toxic-
ities that might result from chemotherapy. The tox-
icities associated with chemotherapeutic agents are
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (11)

1096
Brown et al.
Withdraw/die
Continue
Intercurrent toxicity
Salvage
treatment
Die
No intercurrent toxicity
Continue
Withdraw/die
Continue
Cumulative toxicity
Respond
Die
No cumulative toxicity
Continue
Withdraw/die
Continue
Cumulative toxicity
Continue
salvage
treatment
SD
PD
Die
No cumulative toxicity
Discontinue treatment
Continue
Die
Remain in response
Advance to PD
Respone
SD
Remain in SD
Advance to PD
Post
treatment
Die
PD
Fig. 1. Model structure. PD = progressive disease; SD = stable disease.
usually classified as immediate (occurring during
or immediately after treatment), intercurrent (oc-
curring between courses of treatment) or cumula-
tive (occurring after several courses of treatment).
Pre-medication is often used to reduce the inci-
dence and severity of adverse effects of chemother-
apy, and pre-medication was included in the treat-
ment patterns used in the model to reflect current
practice.
Immediate toxicities, such as nausea and vom-
iting, are assumed to discontinue after completion
of each course of chemotherapy. Neutropenia is the
most common intercurrent toxicity. Patients with
neutropenia are at greater risk of succumbing to
infection, which may culminate in sepsis and an
increased risk of mortality. Patients who develop
severe neutropenia with fever, indicating a risk of
infection, may therefore be withdrawn from che-
motherapy. The model assumed that the immediate
toxicities occurred during cycles two and three.
Cumulative toxicities that could be encountered
with salvage chemotherapy and included in the
model were severe fluid retention, severe periph-
eral neuropathy, arthralgia, myalgia and skin con-
ditions. The model assumed that these conditions
persisted for 9 weeks. Patients may be withdrawn
from chemotherapy if these toxicities are severe.
Structure of the Model
Diagrammatic representation of the model is
shown in figure 1. Cycle length of the model was
3 weeks. This equated to the interval between
courses of treatment for docetaxel and paclitaxel;
a total of six courses of treatment was administered
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (11)

Cost Effectiveness of Therapy for Advanced Breast Cancer
1097
Table IV. Sensitivity analyses: docetaxel (D) versus paclitaxel (P)
Analysis (costs discounted by 6% unless otherwise
indicated)
Cost
(£)/QALY gained
Incremental cost
(D-P)
Incremental utility cost
(D-P)
Base case
1995
2070
4550
3220
3666
6055
172
172
251
172
316
522
0.0862
0.0831
0.0551
0.0534
0.0862
0.0862
Base case, benefits and costs discounted
Paclitaxel survival same as docetaxel
Utility values for disease states pooled from 6 countries
Cost of stable disease reduced to £55 per 3-week period
Cost of progressive disease reduced to £100 per 3-week
period
Cost of progressive disease increased to £300 per 3-week Docetaxel dominatesa 251
0.0862
period
a
Docetaxel is more effective and less costly
QALY = quality-adjusted life year.
for these agents. Twelve courses of treatment of
vinorelbine were administered during four cycles
of the model. For each chemotherapeutic agent, the
model estimated the quality-adjusted time a patient
would spend in the discrete disease states (com-
plete or partial response, stable disease, progres-
sive disease) and toxicities during each cycle and
the associated resource costs. Costs accumulated
in the model as the patient progressed through the
health states. The response rates and adverse effect
profiles for each agent could then be assessed in
terms of the costs and benefits that they imply. For
a disease in which length of survival is not signif-
icantly increased by treatment, it is reasonable to
assume that patients who respond to treatment in-
cur a lower cost than patients who experience dis-
ease progression, since disease progression often
involves additional therapy to alleviate symptoms
and control disease. Similarly, patients who re-
spond are assumed to have a better quality of life
than patients who progress, since patients who re-
spond spend longer in better states of health. Pa-
tients who experience adverse effects from therapy
are assumed to have a reduced quality of life at that
time and to incur additional costs related to clinical
management of the adverse effect. Costs accumu-
late in the model as the patient progresses through
the health states.
gressed at cycle 2 were withdrawn from therapy
and developed early progressive disease. Those with
a response or stable disease proceeded through
cycle three to cycle four when the response was
confirmed. Patients who were not withdrawn from
therapy because of severe cumulative toxicities
completed the courses of treatment. These patients
remained in the response or stable disease states for
the median duration of response and then devel-
oped progressive disease. The model took into ac-
count the median survival times for docetaxel,
paclitaxel and vinorelbine, which were derived
from clinical trial data (table I). Following treat-
ment with docetaxel, paclitaxel or vinorelbine, half
of the patients had died by 56, 46 or 36 weeks,
respectively. The remainder of the patients died af-
ter these times.
Analyses
The probabilities, utilities and costs presented
in tables I, II and III, respectively, were used in the
base-case analyses. The base-case scenarios repre-
sented the most likely events to result from treat-
ment with docetaxel, paclitaxel and vinorelbine.
The output from this type of economic model
is incremental cost effectiveness. This output pro-
vides an assessment of the extra costs and benefits
of an intervention when compared with another
treatment option. It incorporates the notions of
choice and sacrifice faced in real-life clinical situ-
The model assumed that treatment response
would be evident by the second cycle and con-
firmed by the fourth cycle. Patients who pro-
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (11)

1098
Brown et al.
ations, namely that if one therapy is chosen then
another will not be used.
ease, and stable disease. If the model costs were
overestimated, the cost-effectiveness ratio would
be higher. However, if the model costs were under-
estimated, the ratio would be lower, and docetaxel
might become less expensive (dominate paclitaxel
by being less costly and more effective).
The results of the analyses consistently showed
that docetaxel resulted in higher health gain (in-
creased QALYs) than paclitaxel. In most cases,
docetaxel also resulted in greater healthcare costs.
However, using the base-case scenario, the in-
creased utility benefit was substantial and the ad-
ditional cost per QALY gained was small, illustrat-
ing that docetaxel is a cost-effective therapy for use
in patients with advanced breast cancer when com-
pared with paclitaxel. Sensitivity analyses with the
upper and lower bounds of utilities did not change
the advantage of docetaxel over paclitaxel. Assum-
ing the same survival for both agents results in
higher cost per QALY for docetaxel.
The evidence used in the model was derived
from three randomised phase III trials for doce-
taxel[1,4,6] and one randomised phase III trial for
paclitaxel.[3] Although the agents were not com-
pared head to head in these trials, models are dis-
igned to overcome this problem. Criteria for mak-
ing evidence-based decisions in the UK[23] suggest
that randomised trials provide the highest quality
evidence and, while adapting the criteria to cost-
effectiveness studies may be open to debate, under
their construct the model data are category I or II
(table V). Docetaxel would be ‘strongly recom-
Sensitivity analyses form an important part of
economic evaluations using models, especially
where input variables have been estimated. To test
the robustness of the model and the validity of the
base-case cost-utility results, sensitivity analyses
were performed using the following scenarios: dis-
counting benefits (clinical outcomes) as well as
costs at a rate of 6%; varying the utilities assumed
for each of the health states; increasing and de-
creasing the costs applied to the disease states and
toxicities experienced.
Results
Docetaxel versus Paclitaxel
In the base-case analysis, the total per-patient
cost of using docetaxel was £7817. This compared
with a total per-patient cost of £7645 for paclitaxel.
The QALY values for the base-case scenario were
0.7347 for docetaxel and 0.6485 for paclitaxel.
When compared with costs and benefits for pac-
litaxel, treatment with docetaxel resulted in an ad-
ditional total cost of £172 per patient for an addi-
tional 0.0862 QALY, which is equivalent to an
extra 31 days of perfect health. The incremental
cost-utility ratio for docetaxel versus paclitaxel
was therefore £172/0.0862 QALY or £1995 per
QALY gained.
The results of the sensitivity analyses are pre-
sented in table IV. The model was sensitive to the
costs associated with treatment, progressive dis-
Table V. Recommendations based on cost-utility estimates and the quality of evidence
Incremental cost per quality-adjusted life-year gained (£ × 1000)
Quality of evidencea
<3
3-20
>20
negative
I
Strongly recommended
Strongly recommended
Supported
Strongly recommended
Supported
Limited support
Limited support
Limited support
Not proven
Not supported
Not supported
Not supported
Not proven
II
III
IV
Limited support
Not proven
Not proven
a
Categories are based on the quality of data from clinical trials and other clinical evidence (adapted from Stevens et al.[23]).
I = well-designed randomised controlled trials; II = controlled trials with pseudo-randomisation or no randomisation or prospective cohort
studies with concurrent or historic controls or retrospective cohort studies with concurrent controls or case-control retrospective study; III =
large differences from comparisons between times and/or places with and without intervention; IV = opinion of respected authorities based
on clinical experience, descriptive studies and reports of expert committees.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (11)

Cost Effectiveness of Therapy for Advanced Breast Cancer
1099
Table VI. Sensitivity analyses: docetaxel (D) versus vinorelbine (V)
Analysis (costs discounted by 6% unless otherwise indicated)
Cost
(£)/QALY gained
Incremental cost
(D-V)
Incremental utility
cost (D-V)
0.2525
0.2398
0.2440
0.2351
0.2525
0.2525
0.2525
Base case
14 055
14 800
14 287
15 095
14 186
15 030
12 970
3549
3549
3486
3549
3582
3795
3274
Base case, benefits and costs discounted
Vinorelbine survival same as docetaxel
Utility values for disease states pooled from 6 countries
Cost of stable disease reduced to £55 per 3-week period
Cost of progressive disease reduced to £100 per 3-week period
Cost of progressive disease increased to £300 per 3-week period
£ = pounds sterling.
mended’ or ‘supported’ based on the cost-effec-
tiveness ratio.
Discussion
This work updates the cost-utility analysis of
docetaxel versus paclitaxel that has previously
been published.[9] At the time of this earlier study,
there were few published reports of clinical trials
of docetaxel and paclitaxel, and data from phase II
clinical trials were used to provide estimates of
treatment effects. Since the publication of that
study, phase III clinical trial data for docetaxel and
paclitaxel have become available.[1,3,4,6] These
randomised phase III trials recruited large numbers
of patients and, as such, are considered to provide
the strongest clinical evidence available on which
to base an analysis of docetaxel versus paclitaxel.
The original cost-utility analysis was therefore
repeated using estimates of clinical outcomes
based on the published phase III clinical trial data
for docetaxel and paclitaxel. In addition, this re-
analysis enabled resource costs to be updated ac-
cording to current prices and extend the original
analysis to vinorelbine. The inclusion of vinorelb-
ine in this study was considered important for the
analyses to be comprehensive and reflect the treat-
ment choices for advanced breast cancer.
Docetaxel versus Vinorelbine
In the base-case analysis, the total per-patient
costs of using docetaxel and vinorelbine were
£7817 and £4268, respectively. The QALY values
for the base-case scenario were 0.7347 for doce-
taxel and 0.4822 for vinorelbine. When compared
with costs and benefits for vinorelbine, treatment
with docetaxel produced an additional 0.2525
QALY (equivalent to 92 days of perfect health) for
an extra cost of £3549. The incremental cost-utility
ratio for docetaxel versus vinorelbine was there-
fore £3549/0.2525 QALY or £14 055 per QALY
gained.
The results of the sensitivity analyses are pre-
sented in table VI. The incremental cost-utility ra-
tio for docetaxel versus vinorelbine did not vary
greatly in any of the scenarios, thus confirming the
robustness of the model and the validity of the
results of the base-case analysis.
Docetaxel consistently produced higher health
gains (increased QALYs) and higher healthcare
costs than vinorelbine in the base-case analysis and
in the sensitivity analyses. The results demon-
strated that docetaxel is a cost-effective therapy for
use in patients with advanced breast cancer when
compared with vinorelbine. Once again, adopt-
ing the criteria on level of evidence in decision-
making,[23] docetaxel would be ‘strongly recom-
mended’ or ‘supported’ (table V).
Decision-analysis modelling is a useful tool in
situations where data are disparate and time con-
straints and financial restrictions preclude the pro-
spective collection of cost-effectiveness data. Al-
though the decision-analysis model employed in
this study may appear complex, it represents a con-
siderable simplification of the issues relating to the
treatment of advanced breast cancer. The model
necessitates the definition of a ‘typical’ patient,
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (11)

1100
Brown et al.
and such patients rarely exist. The results of anal-
ysis models should be interpreted with some cau-
tion since they often depend upon limited clinical
data and they rely upon estimates for many of the
parameters involved. Although we attempted to
use comparable published trial data, the quality of
data is limited by the absence of head to head com-
parisons of docetaxel and paclitaxel or vinorelbine.
Patients may differ across trials and clinical out-
comes may be reported differently. In addition, we
had to rely upon nurses to estimate patient utilities.
Using patient-derived utilities would strengthen
the model.
Models are used in an attempt to overcome the
absence of head to head comparisons and when
carefully validated and supported by expert opin-
ion, they can contribute to our understanding of
the socioeconomic factors surrounding the treat-
ment of diseases. The present study looked specif-
ically at the treatment of anthracycline-resistant
advanced breast cancer. It is important to note that
the model did not take into account treatments re-
ceived and healthcare costs incurred early in the
course of the disease. The model also does not take
into account the indirect costs associated with ad-
vanced breast cancer. Productivity losses incurred
by patients with advanced breast cancer are likely
to be high and may not be dependent to the chemo-
therapeutic agent used or the response achieved. In
practice, even patients who respond to treatment
experience long term adverse effects, such as pe-
ripheral neuropathy and bone pain. Consequently,
their work productivity could decline regardless of
chemotherapy or response to treatment.
is the fact that docetaxel is the only chemothera-
peutic agent to have produced a significantly higher
response rate than doxorubicin in a randomised
trial and the only agent to show a survival advan-
tage.[4] Until these data were published, doxo-
rubicin was widely considered to be the most active
single chemotherapeutic agent for treatment of pa-
tients with advanced breast cancer.
Paclitaxel was the first taxoid to be licensed in
the UK for advanced breast cancer. However, the
results of the original cost-utility analysis[9] and
of this updated analysis suggest that docetaxel is
a more cost-effective alternative. In the original
study, the incremental cost-utility ratio for doce-
taxel versus paclitaxel in the base-case analysis
was calculated as £2431 per QALY gained. How-
ever, there was no discounting of costs in that
study. In the present study, the incremental cost-
utility for docetaxel versus paclitaxel in the base-
case analysis was similar at a value of £1995 per
QALY gained. The use of docetaxel is further sup-
ported by the results of the comparison with vino-
relbine in the present study. Although the cost per
QALY gained was greater for docetaxel in the com-
parison with vinorelbine than in the comparison
with paclitaxel, the value still supported the use of
docetaxel. When compared with vinorelbine, doce-
taxel provided the equivalent of an additional 92
days of perfect health.
More recently, anecdotal comments from the
UK-based National Institute for Clinical Excel-
lence indicate that between £20 000 and £30 000 is
an acceptable range for cost per QALY within the
UK context.[24]
The quality of the phase II and III clinical trial
data for docetaxel in patients with advanced breast
cancer exceeds that for other medicines licensed
for this indication in the UK, such as paclitaxel and
vinorelbine. Phase III clinical trials have compared
docetaxel with mitomycin plus vinblastine, doxo-
rubicin, and methotrexate plus 5-fluorouracil.[1,4,6]
Docetaxel has been shown to increase response
rate, time to progression, and survival in patients
with advanced breast cancer when compared with
other chemotherapeutic agents. Of particular note
Sensitivity analyses confirmed the robustness of
the model. In the comparison of docetaxel and pac-
litaxel, increasing model treatment cost increased
the cost-effectiveness ratio, while decreasing costs
reduced the ratio so that docetaxel could dominate
paclitaxel. In the comparison of docetaxel and
vinorelbine, the cost-effectiveness ratio did not dif-
fer greatly in any of the scenarios. The use of
pooled utility values from oncology nurses in six
countries instead of the UK-derived values or the
upper and lower bounds of utility values increased
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (11)

Cost Effectiveness of Therapy for Advanced Breast Cancer
1101
tine in patients with metastatic breast cancer progressing de-
spite previous anthracycline-containing chemotherapy. J Clin
Oncol 1999; 17: 1413-24
the cost per QALY, but not sufficiently to change
the acceptability of the cost-effectiveness ratios.
The utility values clearly have an important impact
and using actual patient data rather than nurses as
proxies would strengthen the analysis.
2. Sedlacek SM. Salvage therapy for metastatic disease. Semin
Oncol 1990; 17 Suppl. 7: 45-9
3. Nabholtz J-M, Gelmon K, Bontenbal M, et al. Multicentre,
randomised comparative study of two doses of paclitaxel in
patients with metastatic breast cancer. J Clin Oncol 1996; 14:
1858-67
There are similarities between this study and
two other studies using modelling techniques to
examine the cost utilities of docetaxel, paclitaxel
and vinorelbine in the treatment of anthracycline-
resistant advanced breast cancer.[25,26] In the French
study that compared docetaxel with vinorelbine,
vinorelbine resulted in higher costs than docetaxel,
principally because of high rates of hospitalisation
of patients who were nonresponsive to therapy.[25]
Paclitaxel was intermediate in terms of cost and
benefit. Increasing the costs for progressive dis-
ease management in the UK results in docetaxel
being less costly than vinorelbine and providing
higher utilities. Similar results in support of doce-
taxel were obtained from a US comparison of doce-
taxel and paclitaxel,[26] that used data from the
phase III trial comparing docetaxel with doxorub-
icin.[4] Docetaxel provided substantially greater
utility benefits than paclitaxel at a small extra cost
per QALY gained. In contrast to these findings, a
Canadian study has reported that the most cost-
effective agent was vinorelbine and the least cost-
effective agent was docetaxel.[27] However, the
number of patients considered in this study was
small, and there was no significant difference in
quality-adjusted progression-free survival days.
4. Chan S, Friedrichs K, Noel D, et al. Prospective randomised
trial of docetaxel versus doxorubicin in patients with meta-
static breast cancer. J Clin Oncol 1999; 17: 2341-54
5. Jones S, Winer E, Vogel C, et al. Randomised comparison of
vinorelbine and melphalan in anthracycline-refractory ad-
vanced breast cancer. J Clin Oncol 1995; 13: 2567-74
6. Sjöström J, Blomqvist C, Mouridsen H, et al. Docetaxel com-
pared with sequential methotrexate and 5-fluorouracil in pa-
tients with advanced breast cancer after anthracycline failure:
a randomised Phase III study with crossover on progression
by the Scandinavian Breast Group. Eur J Cancer 1999; 35:
1194-201
7. Brown RE, Burrell A. The economic advantages of response in
the treatment of advanced breast cancer. J Med Econ 1999; 2:
57-63
8. Revicki DA, Kaplan RM. Relationship between psychometric
and utility-based approaches to the measurement of health-
related quality of life. Qual Life Res 1993; 2: 477-87
9. Hutton J, Brown R, Borowitz M, et al. A new decision model
for cost-utility comparisons of chemotherapy in recurrent
metastatic breast cancer. Pharmacoeconomics 1996; 9 Suppl.
2: 8-22
10. Leonard R, Jowell A. A systematic review of docetaxel, pac-
litaxel and vinorelbine in the treatment of advanced breast
cancer. Adv Breast Cancer 2000; 2: 1-7
11. Dieras V, Marty M, Tubiana N, et al. Phase II randomised study
of paclitaxel versus mitomycin in advanced breast cancer.
Semin Oncol 1995; 22 Suppl. 8: 33-9
12. Seidman AD, Tiersten A, Hudis C, et al. Phase II trial of
paclitaxel by 3-hour infusion as initial and salvage chemo-
therapy for metastatic breast cancer. J Clin Oncol 1995; 13:
2575-81
13. Fountzilas G, Athanassiades A, Giannakakis T, et al. Aphase II
study of paclitaxel in advanced breast cancer resistant to an-
thracyclines. Eur J Cancer 1996; 32A: 47-51
14. Michael M, Bishop JF, Levi JA, et al. Australian multicentre
phase II trial of paclitaxel in women with metastatic breast
cancer and prior chemotherapy. Med J Aust 1997; 166: 520-3
15. Vici P, DiLauro L, Conti F, et al. Paclitaxel activity in anthra-
cycline refractory breast cancer. Tumori 1997; 83: 661-4
16. Pugliese P, Brugnatelli S, Giordano M, et al. Paclitaxel in an-
thracycline-treated breast cancer patients. Oncol Rep 1998;
5: 915-8
17. Torrance GW. Utility approach to measuring health-related
quality of life. J Chron Dis 1987; 40: 593
18. Furlong W, Feeny D, Torrance GW. Guide to design and devel-
opment of health state utility instrumentation. Hamilton
(ON): McMaster University, 1990
Conclusions
Docetaxel provided greater utility benefits than
paclitaxel or vinorelbine at slightly higher addi-
tional costs. The results of our study support the
use of docetaxel in the management of advanced
breast cancer.
Acknowledgements
19. Netten A, Dennet J, Knight J. Unit costs of health and social
care. Kent: Personal Services Research Unit at Kent Univer-
sity, 1998
20. MIMS. Monthly Index of Medical Specialties. London:
Haymarket Publishing Services Ltd, 1999 Aug
Model development supported by a grant from Rhone-
Poulenc Rorer, now Aventis.
References
21. Joint Government/Pharmaceutical Industry Working Party. UK
guidance on good practice in the conduct of economic eval-
uations of medicines. BMJ Econ 1994; 7: 63-4
1. Nabholtz J-M, Senn HJ, Bezwoda WR, et al. Prospective
randomised trial of docetaxel versus mitomycin plus vinblas-
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (11)

1102
Brown et al.
22. Miller AB, Hoogstraten B, Staquet M, et al. Reporting results
of cancer treatment. Cancer 1981; 47: 207-14
23. Stevens A, Colin-Jones D, Gabbay J. ‘Quick and clean’: author-
itative health technology assessment for local health care con-
tracting. Health Trends 1995; 27: 37-42
24. Anonymous. AQALY of 30,000 BPS is increasingly being seen
as the cut-off point for a positive NICE submission. Script
2001 Oct 19: 2
26. Brown RE, Hutton J. Cost-utility model comparing docetaxel
and paclitaxel in advanced breast cancer patients. Anticancer
Drugs 1998; 9: 899-907
27. Leung PP, Tannock IF, Oza AM, et al. Cost-utility analysis of
chemotherapy using paclitaxel, docetaxel, or vinorelbine for
patients with anthracycline-resistant breast cancer. J Clin
Oncol 1999; 17: 3082-90
25. Launois RJ, Reboul-Marty JM, Bonneterre J. Acost-utility anal-
ysis of second line chemotherapy in metastatic breast cancer:
docetaxel versus paclitaxel versus vinorelbine. Pharmaco-
economics 1996; 10: 504-21
Correspondence and offprints: Dr Ruth E. Brown, MEDTAP
International Inc., 20 Bloomsbury Square, London WC1A
2NS, UK.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (11)