Full text of DOI:10.2165/00019053-200119030-00007

Pharmacoeconomics 2001; 19 (3): 293-301
1170-7690/01/0003-0293/$22.00/0
ORIGINAL RESEARCH ARTICLE
©
Adis International Limited. All rights reserved.
Cost-Effectiveness Analysis of Inhaled
Zanamivir in the Treatment of
Influenza A and B in High-Risk Patients
1
1
2
Adrian D. Griffin, Andrew S. Perry and Douglas M. Fleming
1
2
GlaxoWellcome Research and Development, Greenford, UK
Northfield Health Centre, Birmingham, UK
Abstract
Objective: To evaluate the cost effectiveness of zanamivir 10mg twice daily for
days in the treatment of influenza in high-risk patients.
5
Design: Bootstrap cost-effectiveness analysis incorporating within-trial analysis
of pooled patient-level cost and effect data.
Setting: UK unit costs and utilities applied to high-risk patients drawn from 6
multinational clinical trials.
Patients: A total of 154 zanamivir and 167 placebo high-risk patients were in-
cluded in the analysis.
Main outcome measures: Cost per day of normal activities; cost per symptom-
free day; cost per complication averted; cost per quality-adjusted life-year (QALY).
Results: The mean benefit was estimated to be 2.5 days [95% confidence interval
(
CI): 0.68 to 4.27] of normal activities gained; 2.0 (95% CI: 0.56 to 3.51) symp-
tom-free days; and a 9% reduction in complications (95% CI: 0 to 18%). Exclud-
ing the effect of rare hospitalisation costs, the cost (1999 values) of gaining a day
of normal activities was £9.50 (95% CI: £5 to £39); cost per symptom-free day
was £11.56 (95% CI: £6 to £43); cost per complication averted was £262 (95%
CI: £90 to £1574). Influenza was estimated to reduce utility by 0.883 per day,
demonstrating the debilitating effect of the disease. Extrapolating a day of normal
activities to a standard utility measure resulted in a cost per QALY of £3900
excluding inpatient costs (£7490 including inpatient costs). Cost-effectiveness
acceptability curves demonstrated 90% certainty that zanamivir would be cost
effective at £8000 per QALY.
Conclusions: Significant health benefits can be obtained with zanamivir treat-
ment in high-risk patients. The cost per QALY for zanamivir in these patients
compares well with that of other commonly used pharmacological interventions.
Influenza outbreaks occur almost every winter
lost.[1] Influenza is associated with additional mor-
bidity or mortality in patients who are elderly and
in individuals who have an underlying disease such
as diabetes mellitus, asthma, chronic obstructive
pulmonary disease (COPD) or cardiovascular dis-
ease. During the 1989/1990 influenza epidemic, over
and result in high levels of morbidity and mortality.
Results from a recent systematic review of the im-
pact of influenza suggested that the average patient
with influenza had a 3% chance of being hospital-
ised for 8 days, and resulted in over 5 working days

2
94
Griffin et al.
2
5 000 excess deaths were estimated to have oc-
and uses a bootstrapping technique to generate 95%
confidence intervals (CIs) around the resulting cost-
effectiveness point estimates.
curred in England and Wales, many of whom would
[
2]
have been aged over 65 years.
Zanamivir, the first of a new class of antiviral
agents, was developed as a highly potent and selec-
tive inhibitor of all known influenza A and B virus
neuraminidases.[ Delivered directly to the primary
site of viral replication in the respiratory tract, zan-
amivir is an orally inhaled agent indicated for adults
and adolescents presenting within 48 hours after
onset of typical influenza symptoms when influ-
enza is circulating in the community. Other neur-
aminidase inhibitors are in development but have
yet to be licensed in Europe (oseltamivir is avail-
able in the US).
Methods
3]
Patients
The cost-effectiveness analysis incorporated all
studies in the main international clinical develop-
ment programme up to the 1998/1999 season in
which the licensed dose of zanamivir had been
investigated and in which high-risk patients were
recruited. This resulted in the inclusion of 5 in-
fluenza treatment studies and the treated cases
from 1 mixed influenza treatment and prophylaxis
This economic evaluation analysed the cost ef-
fectiveness of zanamivir in the treatment of influ-
enza in patients at ‘high risk’of complications from
the illness. International clinical and resource use
data were coupled with UK unit costs and utility
estimates. Previous zanamivir evaluations have re-
lied on modelling outcomes for high-risk patients
based upon the limited data available from individ-
ual clinical trials.[ This method allows for a de-
gree of freedom with sensitivity analysis but does
not provide a measure of uncertainty around the
cost-effectiveness ratio. The present evaluation was
based on the results of a pooled analysis of 6 large
[
6]
[7]
[8]
study [NAIB2007, NAIB3001, NAIA3002,
NAIB3002,[ NAI30009, and NAI30010 (index
9]
[10]
[11]
cases only, not prophylaxis)]. Table I describes
the studies and the recruitment of high-risk patients
in each.
Each of the studies defined a group of patients
as being ‘high risk’ if they had chronic respiratory
disease (including asthma and COPD); significant
cardiovascular disease (excluding patients with hy-
pertension as the sole diagnosis); were immuno-
compromised; or were ≥65 years of age with or
without underlying medical conditions. In addition,
NAIB3001 and NAIB2007 included patients with
diabetes mellitus. This definition is similar to that
4]
[
5]
randomised, placebo-controlled clinical studies
Table I. Studies included in the analysis
Study
Description
Length of diary All patientsa
High-risk patientsa
card (days)
placebo
zanamivir
placebo
zanamivir
(
censored)
228 (101)
365 (103)
182 (71)
247 (51)
(censored)
(censored)
(censored)
37 (12)
49 (21)
13 (2)
NAIB3001
NAIA3002
NAIB3002
NAI30009
Phase III, Southern hemisphere, 1997
Phase III, North America, 1997-1998
Phase III, Europe, 1997-1998
Phase III paediatric study, North
America/Europe/Israel, 1999
14
28
28
28
227 (61)
412 (100)
174 (45)
224 (26)
39 (24)
60 (22)
19 (9)
14 (4)
22 (3)
NAI30010
Phase III family prophylaxis study with
treatment for index cases, North
America/Europe, 1998-1999 (only index
cases included)
28
160 (14)
161 (17)
11 (1)
10 (2)
NAIB2007
Phase II, Southern hemisphere, 1995-1996
5
183 (150)
188 (132)
24 (20)
23 (12)
Total
1365 (490)
1386 (381)
167 (80)
154 (52)
a
Number censored includes those with censored data for either ‘time to return to normal activities’ or ‘time to alleviation of clinically signifi-
cant symptoms’.
©
Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (3)

Zanamivir in High-Risk Patients
295
used to identify high-risk individuals appropriate
for vaccination in the UK.
Extrapolation of Censored Data
To enable CIs to be determined through boot-
strap techniques (see ‘Statistical Analysis’ section),
data were required for all patients, including those
who were censored. It was therefore necessary to
estimate the time to alleviation and time to return
to normal activities for all censored data.
The evaluation was performed on the intent-to-
treat (ITT) population, defined as all randomised
patients regardless of whether study drug was re-
ceived or whether the patient completed the planned
duration of treatment.
End-Points
Mean survival times were estimated using stan-
dard Kaplan-Meier techniques, while adjusting for
censoring. A randomly generated estimate of alle-
viation time was made for each censored value using
survival analysis based on the distribution obser-
ved from the standard Kaplan-Meier plot of all
patients in the integrated trial database. No extrap-
olation was made beyond the duration of the final
censoring point in the longest trial (26.5 days).
Kaplan-Meier methods are accepted as standard for
The end-points used in the evaluation were ‘time
to return to "normal" activities’ as defined by the
patient; ‘time to alleviation of clinically significant
symptoms’; and ‘incidence of complications requi-
[
5]
ring use of antibacterials’. Alleviation of symp-
toms was considered to have occurred when patients
reported absence of fever (temperature <37.8°C and
no feverishness), and recorded a symptom score
of ‘none’ or ‘mild’ (or ‘absent/minimal’ in studies
NAI30009 and NAI30010) for headache, cough,
myalgia and sore throat. These conditions had to
be maintained for a further 24 hours.
[12]
[13]
both outcomes and costing.
Cost data were not extrapolated, as resource use
data were collected by the investigator in the CRF
for both censored and uncensored patients.
The most common complications requiring anti-
bacterial use were acute bronchitis, acute sinusitis,
pharyngitis, ear infection, pneumonia and worsen-
ing of influenza symptoms.
Utility Data
Utility values were not collected directly during
the clinical trial programme. However, for the pur-
poses of creating a cost per quality-adjusted life-
year (QALY), utility estimates were collected from
2 sources.
Data Collection
Patients rated their symptoms and ability to per-
form normal activities in a diary card. The duration
of diary card collection in the 6 studies varied from
(i) Typical general population patients (aged ≥18
years, n = 21) who experienced influenza infection
(confirmed by virus culture) fewer than 3 months
earlier during the 1999/2000 season were identi-
fied from a UK general practitioner (GP) database
and asked to retrospectively complete one EQ-5D
(EuroQoL instrument) health questionnaire[ de-
scribing the entire period of their influenza illness
and another EQ-5D for the most severe day within
that period. The value describing the entire influ-
enza period is used in the analysis.
5
to 28 days. Since patients had to maintain their
alleviated state for at least 24 hours (3 consecutive
diary card entries), the latest recorded time to alle-
viation or return to normal activities was 1.5 days
prior to end of follow-up. This meant that data from
patients who were not alleviated during the course
of the study were described as censored.
14]
Resource use data were collected as part of the
case record form (CRF). The investigator completed
the CRF from patient records, irrespective of pa-
tient outcome. Resource use data were therefore
not subject to censoring. Table I lists the length of
follow-up in each study and the number of patients
whose symptoms were not alleviated during follow-
up.
(ii) As a check on the patient-generated values,
a UK GP panel (n = 8) used the EQ-5D question-
naire to rate the health-state for a hypothetical in-
fluenza high-risk patient (these values were not used
in the analysis).
©
Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (3)

2
96
Griffin et al.
In both cases, baseline (norm) health-state util-
of the CRF. Unscheduled contacts were defined as
those not required as part of the study protocol. In
addition, all prescription or over-the-counter (OTC)
medications started on or after the day of random-
isation were recorded.
Data on hospitalisations were not collected rou-
tinely as part of the CRF. Data concerning hospi-
talisations related to influenza were therefore gath-
ered from the serious adverse event (SAE) forms
reported by investigators (since hospitalisation for
any reason is classified as a SAE for clinical trials).
ities were also obtained. Utility values were ob-
tained by inputting each patient’s scores for the
[
15]
5
dimensions into a published algorithm.
This
method of obtaining utilities has been established
[
16]
in other disease areas, such as multiple sclerosis.
The QALY is calculated from the following
equation:[
17]
QALY = quality adjustment (Δutility) × time in state
For the QALY calculation, the time spent with-
out normal activities was valued as the health-state
with influenza. The rest of the year was valued at
the norm. The difference in QALYs between the
placebo and zanamivir groups was then calculated.
Costs
Standard UK unit costs were applied to the un-
scheduled healthcare contacts and antibacterials used
in the trials (table II). All costs are 1999 values.
Discounting was not considered appropriate as all
costs and benefits considered occur within 1 year.
Healthcare Resource Use
For each patient in the studies identified, data
were collected on all unscheduled healthcare con-
tacts (GP, nurse, practice or home visit, etc.) as part
Table II. Resource use and unit costs
Resource item
Resource use per patient
Unit cost (£;
Source (year)
1999 values)
placebo
0
zanamivir
1
Zanamivir (treatment course)
GP clinic visit
GP home visit
GP telephone contact
Nurse clinic visit
Nurse home visit
Nurse telephone contact
Outpatient visit
Other (as GP clinic visit)
Day in hospital
Penicillins (average course)
Macrolides (average course)
Cephalosporins (average course)
Tetracyclines (average course)
24.00
18.00
54.00
21.00
9.14
12.00
9.14
63.00
18.00
GlaxoWellcome UK
Netten et al.[ (1999)
Netten et al.[ (1999)
Netten et al.[ (1999)
Netten et al.[ (1999)
Netten et al.[ (1999)
18]
0.341
0.048
0.042
0.012
0.006
0.024
0.018
0.012
0.066 days
0.084
0.072
0.018
0.036
0.247
0.019
0.045
0.000
0.006
0.013
0.065
0.000
0.162 (0.058 ) days
0.058
0.078
0.045
0.000
18]
18]
18]
18]
[
18]
Netten et al. (1999)
Netten et al.[ (1999)
Netten et al.[ (1999)
18]
18]
a
[18]
222.00
1.70
4.25
3.82
1.92
0.38
1.50
Netten et al. (1999)
b
b
b
b
b
[19]
[20]
[20]
[20]
[20]
[20]
Medical Data Index and BNF
Medical Data Index and BNF
Medical Data Index and BNF
Medical Data Index and BNF
Medical Data Index and BNF
Estimate from GlaxoWellcome market
research
[19]
[19]
[19]
[19]
Other antibacterial (average course) 0.042
Over-the-counter medication
0.032
0.419
0.506
Total cost per patient
£27.21 (£12.59c) £72.15 (£36.11c)
a
b
Excluding inpatient hospitalisation costs for 1 outlying patient who was hospitalised for pneumonia within 24 hours of recruitment.
Antibacterial costs based on prescriptions written for influenza and pneumonia [International Classification of Diseases (ICD-10) codes
J10 to J18].
c
Excluding all inpatient hospitalisation costs.
BNF = British National Formulary; GP = general practitioner.
©
Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (3)

Zanamivir in High-Risk Patients
297
Table III. Mean incremental costs (1999 values) and effects for zanamivir in the treatment of influenza in high-risk patients. Values in
parentheses are 95% confidence intervals
Effectiveness measure Placebo
Zanamivir
Incremental effect
Incremental cost
ICER
Base case (including rare inpatient hospitalisations)
Time to return to
normal activities
11.96 days (10.64,
13.28)
9.48 days (8.26,
10.70)
2.48 days (0.68, 4.27) £44.94 (–£8, £98)
£18.14 (£1, £85)
£22.08 (£0, £97)
Time to alleviation of
symptoms
9.09 days (7.96, 10.21) 7.05 days (6.09, 8.01) 2.03 days (0.56, 3.51) £44.94 (–£8, £98)
Incidence of
complications
25% (18%, 31%) 16% (10%, 21%) 9% (0%, 18%) £44.94 (–£8, £98)
£501 (–£94,
£3234)
Sensitivity analysis excluding SAE hospitalisations
Time to return to
normal activities
11.96 days (10.64,
13.28)
9.48 days (8.26, 10.7) 2.48 days (0.68, 4.27) £23.52 (£18, £29)
£9.50 (£5, £39)
£11.56 (£6, £43)
£262 (£90, £1574)
Time to alleviation of
symptoms
Incidence of
9.09 days (7.96, 10.21) 7.05 days (6.09, 8.01) 2.03 days (0.56, 3.51) £23.52 (£18, £29)
25% (18%, 31%) 16% (10%, 21%) 9% (0%, 18%) £23.52 (£18, £29)
complications
Sensitivity analysis excluding 1 hospitalised patient
Time to return to
11.96 days (10.64,
9.44 days (8.22,
2.52 days (0.72, 4.32) £21.67 (–£6, £49)
£8.61 (£3, £43)
normal activities
13.28)
10.67)
Time to alleviation of
symptoms
Incidence of
9.09 days (7.96, 10.21) 7.02 days (6.06, 7.99) 2.06 days (0.58, 3.55) £21.67 (–£6, £49)
25% (18%, 31%) 16% (10%, 21%) 9% (0%, 18%) £21.67 (–£6, £49)
£10.50 (–£3, £49)
£245 (£84, £1736)
complications
ICER = incremental cost-effectiveness ratio; SAE = serious adverse event.
Statistical Analysis
The curves are constructed by varying the maxi-
mum cost-effectiveness threshold and calculating
the proportion of bootstrap resamples that would
fall under each threshold.
All results are reported as means in incremental
costs and effects with 95% CIs generated through a
bootstrap method. The incremental cost-effective-
ness ratio (ICER) is calculated from the difference
in mean costs divided by the difference in mean
effects. The nonparametric bootstrap method in-
volved creating new ‘trial populations’ by resam-
pling (with replacement) cost and effect pairs from
the control and treatment samples and then calcu-
lating the cost and effect averages for each boot-
strap sample. ICERs for each control and treatment
The cost-effectiveness acceptability curve was
also generated through the parametric ‘net benefit’
approach[ in order to corroborate the nonpara-
metric bootstrap method.
25]
Results
A total of 2751 patients were recruited into the
6
studies and received zanamivir 10mg twice daily
or placebo. Of these, 321 high-risk patients were
included in the analysis (154 zanamivir, 167 pla-
cebo). Detailed results of the integrated analysis
have been published elsewhere.[ 159 patients (50%)
were male. Mean age was 40.1 years with 76 (24%)
aged 65 years or over. 227 (71%) were confirmed
by laboratory methods as having influenza (105
zanamivir, 122 placebo). 74 (23%) of the 321 pa-
sample could then be calculated. Bootstrap sam-
pling was repeated 5000 times with the 2.5th and
7.5th percentiles interpreted as 95% CIs around
9
[
21]
the mean ICER from the original trial data.
5]
Cost-effectiveness acceptability curves were
generated to allow decision-makers to see the prob-
ability of zanamivir being cost effective (y-axis)
in relation to varying amounts they might be pre-
[
22-24]
pared to pay to secure the benefits (x-axis).
©
Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (3)

2
98
Griffin et al.
4
3
3
0
5
0
ysis was therefore conducted which excluded the
cost of hospitalisation. One patient in the zanami-
vir group was hospitalised for pneumonia within
24 hours of recruitment and withdrawn from the
study. The underlying pathology behind this pa-
tient’s pneumonia must therefore have been devel-
oping prior to commencing zanamivir and therefore
zanamivir would not be expected to avert the need for
hospitalisation. Since this patient remained hospital-
ised for 16 days and the number of hospital events
was small, a sensitivity analysis excluding this one
patient was performed to demonstrate the impact of
one outlier on the cost-effectiveness ratio.
2
2
5
0
1
5
0
5
0
1
2
1
0
1
2
3
4
5
6
Incremental days of normal activities
Fig. 1. Incremental cost-effectiveness plane. Plot of 5000 boot-
strap incremental cost (1999 values) and effect (days of normal
activities) resample means.
Cost Effectiveness
tients had been vaccinated against influenza (40
zanamivir, 34 placebo).
Table III reports the mean incremental costs and
effects per patient. The incremental cost per patient
treated with zanamivir was £44.94 including all hos-
pitalisations. Excluding the hospitalisation costs for
the one patient withdrawn from the study reduced
the incremental cost to £21.67. Excluding all inpa-
tient hospitalisation costs, treatment with zanami-
vir increased the cost of treatment per patient by
Extrapolation of Censored Data
The mean time to return to normal activities for
all recruited patients (including high-risk) using the
Kaplan-Meier product limit estimator was 8.8 days
in the observed placebo group and 8.9 days after
extrapolation. In the zanamivir group, the mean time
to return to normal activities was 7.7 days both in
the observed group and after extrapolation. The dis-
tribution containing extrapolated data closely over-
laid the distribution based on observed data when
plotted on a Kaplan-Meier graph.
£
23.52. This suggests that the cost of zanamivir
(
£24.00) is the main driver behind the incremental
cost, with offsets coming from the levels of addi-
tional healthcare contacts and antibacterial use.
The mean benefit was estimated to be 2.5 days
in terms of normal activities gained, 2.0 symptom-
free days and a 9% reduction in complications re-
quiring antibacterial use. The ICER in the high-risk
group (excluding hospitalisation costs) is therefore
Resource Use
£
9.50 per day of normal activities gained, £11.56
Table II shows the level of healthcare contacts,
hospitalisations, antibacterial use and OTC medi-
cations for the placebo and zanamivir groups.
The primary cost driver in the resource use cal-
culations was the cost of hospitalisation. In the to-
tal trial population, there were 6 influenza-related
hospitalisations in the zanamivir group and 4 in the
placebo group. In the high-risk subpopulation, 4
hospitalisations occurred in the zanamivir group (2
pneumonia, 2 dehydration) and 2 occurred in the
placebo group (1 pneumonia, 1 exacerbation of lower
respiratory tract symptoms). Neither of these dif-
ferences is statistically significant. Asensitivity anal-
per symptom-free day gained and £262 per compli-
cation averted. Depending on whether the 1 outlying
patient is included, the results showed ICERs ranging
from £9 to £18 per day of normal activities gained.
Figure 1 plots the results of the 5000 bootstrap re-
samples for time to return to normal activities (ex-
cluding inpatient hospitalisation costs).
Cost Utility
Results from the patient and GP assessments of
baseline (non-influenza) and influenza health-states
are displayed in table IV.
©
Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (3)

Zanamivir in High-Risk Patients
299
These agree well with the reported UK popula-
for managing influenza. Zanamivir is intended as
an adjunct in vaccinated patients since the decision
to administer a vaccine is based on patient risk fac-
tors and is made while the patient is asymptomatic
before the influenza season begins. During the in-
fluenza season, some of those who have been vac-
cinated will develop influenza symptoms, as will
some who have not been vaccinated. Since the con-
sequences of influenza infection are more severe
in those who have not been vaccinated, the associ-
ated costs would be expected to be higher in this
group.
The analysis was performed on integrated pa-
tient-level data in order to compile as large a pool
as possible of high-risk patients. This meant com-
bining data from studies conducted across different
countries and in different influenza seasons. Al-
though a formal meta-analysis could estimate mean
effects, the integration of patient level data was
necessary to provide the cost and effect pairs for
the bootstrap process. Extrapolation was necessary
for those effect data that were censored. The extra-
polation was limited to the length of the longest study
[
26]
tion norm of 0.825. As the influenza health-state
valuations were similar between the patients and
GPs, the more conservative patient valuation has
been used in the QALY calculations, in which the
difference in utility between baseline and influenza
states (0.883) is combined with the mean decrease
in days to return to normal activities (2.48) and the
incremental cost (£23.52). The result is a cost per
QALY of £3920 (£7490 including hospitalisation
costs). Since the cost per QALY is partly deter-
mined by the utility rate, a decrease in the utility
difference between the influenza state and full health
would lead to a proportional increase in the cost
per QALY.
Cost-effectiveness acceptability curves for the
cost per QALY are presented in figure 2. The re-
sults demonstrate that there is a 90% certainty that
zanamivir would be cost effective (excluding in-
patient costs) if the ceiling ratio for a QALY was
£
8000 and a 95% certainty that zanamivir would
be cost effective if the ceiling ratio was £11 500
per QALY. The shape of the acceptability curves
was replicated in each case using the parametric
net benefit method, demonstrating that the distri-
bution of ICERs was jointly normal.
(
28 days). However, since there were more placebo
patients than zanamivir patients whose symptoms
were not alleviated at the end of studies, this as-
sumption undervalues the effect of zanamivir.
Hospitalisations were not collected prospectively
in the trials but were included as a sensitivity anal-
ysis based on SAE forms. Although there were more
patients hospitalised in the high-risk zanamivir group,
it would be expected that in a large enough popu-
lation the significant reductions in complications
for patients receiving zanamivir would in turn re-
duce the incidence of hospitalisations.
Discussion
The comparator for zanamivir in this evaluation
was placebo. Although amantadine is registered for
prophylaxis and treatment of influenza A, it offers
neither the scope for treatment (being only effec-
tive against influenza A) nor the widespread accep-
tance for it to be considered the standard of care.
This is particularly so in acutely ill elderly patients
whose renal function may not be known at the time
of prescription. Vaccination is the primary method
It is recognised that estimating utility for acute
conditions has both measurement and evaluation
Table IV. Utility estimates from EQ-5D (EuroQoL instrument)[14] health-state questionnaire
Data source
n
Baseline
Influenza health-state Difference (95%
Influenza health-state
(worst day)
(
considering whole
influenza illness)
confidence interval)
0.883 (0.697, 1.069)
0.982 (0.875, 1.090)
Patients with influenza during past
influenza season
General practitioner assessment of
21
8
0.817
0.720
–0.066
–0.342
–0.263
Not collected
hypothetical high-risk patient with influenza
©
Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (3)

3
00
Griffin et al.
Excluding inpatient costs
Base case (including inpatient costs)
Including inpatient costs (except for 1 outlier)
der clinical trial conditions. Bootstrapping method-
ology was used to quantify the uncertainty around the
point estimate of cost effectiveness. The effective-
ness measures were based upon an observed diag-
nostic positivity rate of 71%. Additionally, the trials
recruited only those patients who attended within
1
0
0
0
0
0
0
0
0
0
.0
.9
.8
.7
.6
.5
.4
.3
.2
.1
0
4
8 hours of developing symptoms. When imple-
menting zanamivir treatment in practice, high di-
agnostic accuracy rates will depend upon the ability
of doctors to detect true influenza. Clinical diagno-
sis of influenza in these studies has been shown to
be 71% accurate. This relies solely on knowledge
of common symptoms and signs of influenza, along
with knowledge of local circulation of influenza
virus. The impact on GPs of patients with influenza
presenting outside 48 hours and patients with other
respiratory infections is not evaluated in this anal-
ysis. The use of bootstrapping and cost-effective-
ness acceptability curves address uncertainties in
the clinical data and in decision-makers’ willing-
ness to pay for the benefits obtained. Uncertainties
around the diagnostic rate and consultation patterns
can only be reliably estimated once zanamivir is in
routine use.
5000 10 000 15 000 20 000 25 000 30 000
Ceiling value for cost per QALY ratio (£)
Fig.2.Cost-effectivenessacceptabilitycurve(nonparametricboot-
strap method): cost (1999 values) per quality-adjusted life-year
(
QALY).
problems. However, Bala and Zarkin[27] maintain
that these difficulties are more likely to bias against
acute diseases. Since influenza is a self-limiting
illness for the majority of patients, a patient’s util-
ity will vary during the course of the disease, being
more severely impacted during the early days and
then increasing as the patient begins to recover. To
address this, the use of the utility estimate consid-
ering the entire period of the illness was used in the
calculation rather than the value for the worst day.
Our results estimate the cost per QALY of zana-
mivir in the high-risk group to be £3920 (excluding
inpatient costs). When compared with a recently
compiled list of costs per QALY, the cost effective-
ness of zanamivir compared favourably with that of
Conclusion
This economic evaluation reports a comprehen-
sive analysis of the cost effectiveness of zanamivir
in high-risk patients within the zanamivir clinical
trial programme. The evaluation demonstrates that
significant health benefits can be obtained with za-
namivir therapy in high-risk patients, with cost-
effectiveness point estimates (excluding inpatient
costs) of £9.50 per day of normal activities gained,
£
11.56 per symptom-free day, £262 per complica-
[
28]
other common pharmacological interventions.
tion averted and approximately £3900 per QALY.
Most economic evaluations present point esti-
mates of cost effectiveness and may include some
measure of CIs. The merits of these point estimates
depend not only upon the uncertainty surrounding
the estimates, but also decision-makers’ willing-
ness to pay for the benefits obtained. The cost-
effectiveness acceptability curves presented ad-
dress both these issues. We estimate that there is
a 90% chance that the true cost effectiveness of
A preliminary cost-effectiveness analysis based
upon the first completed zanamivir phase III study
estimated the cost effectiveness of zanamivir in high-
risk patients to be 11 715 Australian dollars (around
[
4]
£
4600; 1998 values). The present evaluation uses
more comprehensive clinical trial data and a differ-
ent methodology but shows a similar result.
This analysis was undertaken to estimate the cost
effectiveness of zanamivir in high-risk patients un-
©
Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (3)

Zanamivir in High-Risk Patients
301
1
0. Hendrick JA, Barzilai A, Behre U, et al. Zanamivir for treatment
of symptomatic influenza A and B infection in children five
to twelve years of age: a randomized controlled trial. Pediatr
Infect Dis J 2000; 19 (5): 410-6
zanamivir lies below £8000 per QALY. The cost
effectiveness of zanamivir in high-risk patientsthere-
fore compares well with the cost per QALY of other
accepted pharmacological interventions.
1
1. Hayden F, Gubareva L, Klein T, et al. Inhaled zanamivir for
preventing transmission of influenza in families. 39th Inter-
science Conference on Antimicrobial Agents and Chemother-
apy (ICAAC); 1999 Sep 26-29; San Francisco (CA), 13
2. Bland JM, Altman DG. Survival probabilities (the Kaplan-Me-
ier method). BMJ 1998; 317: 1572-80
3. Etzioni RD, Feuer EJ, Sullivan SD, et al. On the use of survival
analysis techniques to estimate medical care costs. J Health
Econ 1999; 18 (3): 365-80
Acknowledgements
1
1
We thank Andrew Briggs (Institute of Health Sciences,
University of Oxford); Neil Dobson; Rebecca Warren; and
Sophie Hall (GlaxoWellcome UK, Stockley Park, UK). A.D.
Griffin and A.S. Perry conceived, designed and performed
the economic evaluation. D.M. Fleming conducted the utility
measurement in influenza patients and contributed to inter-
pretation of clinical data. A.D. Griffin and A.S. Perry drafted
the manuscript and D.M. Fleming critically reviewed it. All
authors had final approval of the manuscript. D.M. Fleming
serves as a consultant advisor to GlaxoWellcome in relation
to their clinical trial programme for zanamivir. He has also
received honoraria in respect of lecture presentations on
influenza-related issues and for service on advisory boards
to the pharmaceutical industry covering influenza vaccina-
tion and treatment.
14. EuroQol Group. EuroQol – a new facility for the measurement
of health-related quality of life. HealthPolicy1990; 16:199-208
1
1
1
1
5. Dolan P, Gudex C, Kind P, et al. A social tariff for EuroQol:
results from a UK General Population Study. University of
York: Centre for Health Economics Discussion Paper 138, 1995
6. Forbes RB, Lees A, Waugh N, et al. Population based cost utility
study of interferon beta-1b in secondary progressive multiple
sclerosis. BMJ 1999; 319: 1529-33
7. Drummond MF, O’Brien B, Stoddart GL, et al. Methods for the
economic evaluation of health care programmes. 2nd ed. Ox-
ford: Oxford University Press, 1997
8. Netten A, Dennett J, Knight J. Unit costs of health and social
care 1999. Canterbury: Personal Social Services Research Unit,
University of Kent, 1999
9. Authors compilation of data from Medical Data Index, IMS
Health Incorporated, Westport (CT), USA
0. British National Formulary. 38th ed. London: British Medical
Association and Royal Pharmaceutical Society of Great Brit-
ain, 1999
1. Briggs AH, Wonderling DE, Mooney CZ. Pulling cost-effec-
tiveness analysis up by its bootstraps: a non-parametric ap-
proach to confidence interval estimation. Health Econ 1997;
6 (4): 327-40
1
2
References
1
2
3
4
. Jefferson T, Demichelli V. Socieoeconomics of influenza. In:
Nicholson KG, Webster RG, Hay AJ, editors. Textbook of
influenza. London: Blackwell Science, 1998: 541-7
2
. Fleming DM. The contribution of influenza to combined acute
respiratory infections, hospital admissions and deaths in win-
ter. Commun Dis Public Health 2000; 3 (1): 32-8
. von Itzstein M, Wu WY, Kok GB, et al. Rational design of
potent sialidase-based inhibitors of influenza virus replica-
tion. Nature 1993; 363: 418-23
22. Van Hout BA, Al MJ, Gordon GS, et al. Costs, effects and
C/E-ratios alongside a clinical trial. Health Econ 1994; 3 (5):
309-19
23. UK Prospective Diabetes Study Group. Cost effectiveness anal-
ysis of improved blood pressure control in hypertensive pati-
ents with type 2 diabetes: UKPDS 40. BMJ 1998; 317: 720-6
24. Löthgren M, Zethraeus N. Definition, interpretation and calcu-
lation of cost-effectiveness acceptability curves. Health Econ
2000; 9: 623-30
25. Stinnet AA, Mullahy J. Net health benefits: a new framework
for the analysis of uncertainty in cost-effectiveness analysis.
Med Decis Making 1998; 18 (2 Suppl.): S65-S80
26. Kind P, Dolan P, Gudex C, et al. Variations in population health
status: results from a United Kingdom national questionnaire
survey. BMJ 1998; 316: 736-41
27. Bala MV, Zarkin GA. Are QALYs an appropriate measure for
valuing morbidity in acute diseases? Health Econ 2000; 9:
177-80
. Mauskopf JA, Cates SC, Griffin AD, et al. Cost effectiveness
of zanamivir for the treatment of influenza in a high risk
population in Australia. Pharmacoeconomics 2000; 17 (6):
6
11-20
5
6
7
8
9
. Lalezari J, Campion K, Keene O, et al. Zanamivir for the treat-
ment of influenza A and B infection in high-risk patients: a
pooled analysis of randomized controlled trials. Arch Intern
Med 2001; 161: 212-7
. Monto AS, Webster A, Keene O. Randomized, placebo-control-
led studies of inhaled zanamivir in the treatment of influenza
A and B: pooled efficacy analysis. J Antimicrob Chemother
1
999; 44: 23-9
. MIST (Management of Influenza in the Southern Hemisphere
Trialists) Study Group. Randomised trial of efficacy and
safety of inhaled zanamivir in treatment of influenza A and B
virus infections. Lancet 1998; 352: 1877-81
28. Briggs AH, Gray AM. Handling uncertainty when performing
economic evaluation of healthcare interventions. Health Tech-
nol Assess 1999; 3(2): 1-134
. Lalezari J, Klein T, Stapleton J, et al. The efficacy and safety of
inhaled zanamivir in the treatment of influenza in otherwise
healthy and ‘high risk’individuals in North America [abstract
no. P8]. J Antimicrob Chemother 1999; 44 Suppl. A: 42
. Makela MJ, Pauksens K, Rostila T, et al. Efficacy and safety of
the orally inhaled neuraminidase inhibitor, zanamivir, in the
treatment of influenza: a randomized, double-blind, placebo-
controlled European study. J Infect 2000; 40: 42-8
Correspondence and offprints: Andrew Perry, GlaxoWellcome
R&D, Greenford Road, Greenford, Middlesex, UB6 0HE, UK.
E-mail: asp90059@glaxowellcome.co.uk
©
Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (3)