DEVELOPMENT OF A COST-UTILITY MODEL FOR COMPARING GINA STEP 3

ASTHMA MEDICATIONS BASED ON UTILITY AND SAFETY DATA DERIVED

FROM A LARGE HEAD-TO-HEAD TRIAL OF MONTELUKAST AND SALMETEROL
SORONCZ-SZAB T
1
, TAYLOR SD
2
, NAGY L
1
, SAZONOV KOCEVAR V
2
1
MSD Hungary Ltd., Budapest, Hungary;
2
Outcomes Research and Pricing Support, Merck & Co., Inc., West Point, PA, USA
BACKGROUND AND OBJECTIVES
Bronchial asthma is a chronic condition defined as an inflammation of the airways, whichif a reliever med-
ication (therapeutic step 1) fails to provide an acceptable symptomatic controlrequires continuous antiin-
flammatory therapy according to the international treatment guidelines, such as that of the Global Initiative for
Asthma (GINA).
1
Inhaled corticosteroids (ICS) are generally used as the first-line controller therapy (step 2),
which can be combined, as needed, (step 3) with the complementary antiinflammatory agents leukotriene re-
ceptor blockers, e.g., montelukast (MON), or the long-acting brochodilator
2
-ago nists (LABA), e.g., salme-
terol (SAL).
1
Previous economic analyses based on randomized comparisons of MON and SAL used as step
3 agents
2,3
disregard health utility outcomes, side effect costing, and real-life compliance. Other authors com-
pared the economic impact of MON and SAL prescribed to patient populations which differed in some im-
portant characteristics obtained from pharmacy claims data.
4,5
Using such databases provide the advantages
of, among others, a reflection of real-world practice and compliance, but the examination of causal relation-
ships is a particular challenge [] because subjects are not randomized to treatments.
6
We set therefore the objective to design a cost-utility model based on a randomized trial, which takes com-
pliance into account, and includes all relevant, locally imputable costs (of drugs, asthma attacks, and side ef-
fects, covered by the insurer).
METHODS
A spreadsheet-based model was developed using Microsoft Excel (Microsoft Corp., Redmond, WA, USA).
1. Selection of salmeterol as the alternative strategy
LABAs are the most widespread agents used worldwide in step 3, of which SAL is the most commonly pre-
scribed, packed in various inhaler devices, and reimbursed in several countries. If MON has a favorable cost-
utility profile vs. SAL, it could be considered for a similar reimbursement.
2. Selection of the Improved Asthma Control Trial (IMPACT) as the source of the models efficacy and safety
input variables
Four relevant comparison trials are cited in the GINA guidelines 2006 update (Table 1);
1
IMPACT was the
largest randomized, multinational, double-blind, double-dummy comparison (48 weeks, n=1490),
7
account-
ing for more than three times as many patient-years as the other three, 12-week-long trials combined.
8,9,10
These latter ones are probably less suitable for the asessment of treatment benefits in chronic asthma, which
generally requires drug therapy for years, especially when considering the tolerance against
2
agonists,
11,12
and the increased risk of severe exacerbations with LABAs.
13,14,15
3. Outcome (efficacy and utility) variables
Effects of MON and SAL on asthma attacks and asthma-specific quality of life (QoL) (based on the Asthma-
specific Quality of Life Questionnaire [AQLQ]) were obtained from the IMPACT trial.
7,16
The five-dimensional EuroQol (EQ-5D) weights were calculated from AQLQ results based on the best-fitting
multivariable model developed and validated by authors at ScHARR (School of Health and Related Research,
University of Sheffield) as follows:
17
Q = a +
i
.
2
A
i
.
2
+
i
.
3
A
i
.
3
+
i
.
7
A
i
.
7
+
n
.
7
A
n
.
7
where:
Q is the preference-based EQ-5D index (range: -0.591.00).
a is the intercept of the regression line.
i indicates the number of the AQLQ items included in te model (specifically: items No. 1, 3, 5, 6, 25, 26,
27, 29, 31, and 32)

i
.
x
are coefficients of categorical dummy variables (A
i
.
x
) determining the slope of the regression line.
A
i
.
x
are dummy (indicator) variables that have the value of 1 when the level of an AQLQ item i is x. Since
each AQLQ item can be answered on a scale of 1 to 7, six dummies per items (i.2 to i.7) is sufficient to
describe the item (1 is the baseline level, and has therefore no A
i.1
coefficient).
Changes adjusted with baseline EQ-5D were calculated using analysis of covariance (factors: treatment group
and study center; covariate: baseline value).
4. Effectiveness = efficacy compliance
IMPACT was an efficacy study conducted among the ideal conditions of a controlled trial.
7
Effectiveness
can be estimated from efficacy results through adjustment for the real-life compliance.
18
The existence of
a relationship between compliance and real-life effectiveness on asthma control and QoL is clearly docu-
mented.
1,19,20
In the systematic reviews of interventions to improve adherence to self-administered therapy,
asthma outcomes were improved in trials where an improvement of patient adherence was achieved, and
vice versa.
19,20
We assume this relationship is linear. Although there are data to support this,
21
the exact re-
lationship remains to be determined.
In the present model, drug effects on attacks and QoL as well as drug costs are adjustable for relative
compliance of SAL vs. MON. In database studies of the sponsors of each of MON
22
and SAL,
23
an approx.
2/3 compliance with SAL vs. MON was found. The compliance with a fixed-dose ICS+LABA combination
could be different vs. the separately used components, but the British Thoracic Societys guidelines con-
clude that Combination inhalers have not been shown to improve compliance in the medium to long
term.
24
The choice of 0.8 as the base-case value is therefore conservative, but widely variable in the model
(range: 0.5-0.95).
5. Costs used for the validation of the model
The model was validated with costs (2006) covered by the Swedish (see the analysis published separately
25
)
and the Hungarian (input data not shown) national insurance agencies.
Cost of attacks
Costs of asthma attacks are imputable according to definition categories used in the trial,
7
with the asthma-
related hospitalizations accounting for the majority of costs covered by insurance agencies.
Cost of side effects
SAL+ICS was associated with more drug-related adverse events in most of the body system categories
(Fig. 1). Costing of the whole, or even only the drug-related, side-effect profile of the two regimens is im-
practicable or implausible. We therefore looked for a simplified side-effect costing model approximating
both the absolute and the relative risks of the overall drug-related side-effects, focusing on events associ-
ated with the SAL+ICS combinations (based on the prescribing information of the drugs).
Cost of drugs
All available MON+ICS and SAL+ICS combinations are entered in the model, of which three pricing scena -
rios are used for calculations (Table 2).
RESULTS
As reported earlier, numerically less asthma attacks but more asthma-related hospitalizations and signifi-
cantly more serious and drug-related adverse events occurred with SAL compared to MON.
7
Among patients with a baseline and at least one post-baseline AQLQ measurement (n=1162), both treatments
resulted in a similar increase in AQLQ and the mapped EQ-5D (Table 3). Due to the gradual improvements in
AQLQ results over time,
16
approximately 75% of the study-end QALY gain applies over the 48 treatment weeks
in the IMPACT trial:
MON: 0.0803 (48/52) years 75% = 0.0556 QALY gain/patient
SAL: 0.0812 (48/52) years 75% = 0.0562 QALY gain/patient
To adjust for the real world, QALY results (and costs) of SAL are multiplied with the relative compliance (SAL
vs. MON ~0.8 [0.5-0.95]).
Of the various types of adverse events, oral infections (grouped into 2 homogeneous cost categories in Hun-
gary) and cardiovascular events (6 cost categories) were selected into the simplified side-effect costing model
(Fig. 2) to be causally. Although these events were not always considered drug-related by the investigator, they
are thought to be causally associated with SAL+ICS. The model matched well with the overall drug-related ad-
verse event profile in the treatment groups in terms of both absolute and relative risks (Fig. 2).
The increased costs of asthma attacks and side-effects with SAL were offset by the treatment costs with MON.
When drug costs of SAL are also reduced with 0.8 compliance, the incremental cost of 1 QALY gain from the Hun-
garian insurers perspective with MON+ICS is between 7 to 23 thousand with MON+ICS versus SAL+ICS using
product combinations of various prices (Fig. 3) (results from the Swedish health care are similar
25
).
CONCLUSIONS
The presented cost-utility model avoids well-known limitations of post hoc analyses from prescription
claims databases, especially the selection bias arising from non-randomized comparisons, and, at the same
time, allows for the calculation of incremental cost-effectiveness (based on incidence of asthma attacks)
and cost-utility ratios (based on QALYs mapped from AQLQ results), which are crucial for reimbursement
decisions, through an approximation of the real-world setting (using a compliance factor).
The model has been tested with costs taken from two very different European health-care systems in patients
with chronic bronchial asthma whose symptoms are not adequately controlled on ICS alone. The results in-
dicate that MON provides (in a real-world setting) greater QALY gains at a favorable incremental cost from
both the Hungartian and Swedish insurers perspective, due to
a better compliance
22,23
(subsequently, more prevented attacks and more QALYs gained in the real-
world setting);
less asthma-related hospitalizations
7
(in accordance with other LABA trials
14,15
);
less serious and drug-related advese events
7
(costed with a simplified side-effect model).
ACKNOWLEDGEMENTS AND CONFLICTS OF INTEREST
We are grateful to Aki Tsuchiya and Ron Akehurst (ScHARR) for the coefficients of the AQLQEQ5D transformation.
For further information please visit http://www.shef.ac.uk/scharr/sections/heds/mvh/aqlq.html.
Authors are employees of Merck & Co., Inc. (or its fully owned subsidiary), manufacturer of MON. The work presented
here served as a basis for the pharmacoeconomic evaluation of SINGULAIR

[montelukast, MSD] in Hungary and Sweden.

Table 1. Randomized trials of MON+ICS vs. SAL+ICS (GINA, 2006)
1
Table 3. AQLQ results from the IMPACT trial transformed into the generic EuroQoL quotient
Figure 1. Relative risk of drug-related adverse events (AE) by body system reported in the IMPACT trial
Sponsor
Publication of...
N
Treatment
duration
(weeks)
patient
years
design/
baseline
results
economic
eval.
GSK Nelson 2000 OConnor 2004 447 12 103
GSK Fish 2001 948 12 219
GSK Ringdal 2003 Pieters 2005 725 12 167
Total of the above three: 489
MSD Bjermer 2000 Bjermer 2003 This analysis 1490 48 1375
Table 2. Pricing scenarios used in the model
Systemic
Cardiovascular
Skin and appendages
Gastrointestinal
Metabolic
Musculoskeletal
Respiratory
Psychiatric
Neurologic
Ear-nose-throat
Hematological
Hepatobiliary
Immunological
Endocrine
Urogenital
Patients with at least 1
drug-related AE
0.50 (0.15; 1.64)
0.50 (0.09; 2.71)
0.50 (0.17; 1.45)
0.55 (0.26; 1.13)
0.33 (0.03; 3.18)
0.37 (0.10; 1.40)
0.71 (0.32; 1.59)
0.33 (0.01; 8.13)
1.05 (0.55; 2.03)
1.08 (0.49; 2.35)
(none reported)
(none reported)
(none reported)
(none reported)
(none reported)
Risk ratio (95% CI)
MON+ICS vs. SAL+ICS
0.63 (0.44; 0.90)
SAL+ICS
more harmful
MON+ICS
more harmful
0.05 0.5 1 1.5 5
Relative risk
* Patients with at least one post-baseline AQLQ measurement

(MON+ICS) (SAL+ICS) difference in with at least LS (least squares) mean changes from baseline to the last measurements based on an ANCOVA model
(The present poster covers the models methods and concepts only, and not the results of a specific analysis using this model. The results above are presented only to illustrate the models output.)
Figure 3. Three-way sensitivity analysis: effect of compliance and price difference on the incremental cost-utility ratio (ICUR)
5 000
10 000
15 000
20 000
25 000
30 000
35 000
40 000
45 000
50 000
0
0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
References:
1
Global Initiative for Asthma (GINA): Global Strategy for Asthma Management and Prevention. Revised 2006, www.ginasthma.org
2
O'Connor, Pharmacoeconomics 2004;22:815-25.
3
Pieters, Treatments Respir Med 2005;4(2):129-38.
4
Stempel, J Allergy Clin Immunol 2002;109(3):433-9.
5
Ollendorf, Chest 2000;118(4 Suppl):185s.
6
Motheral,
Value in Health 2003;6(2):90-7.
7
Bjermer, BMJ 2003;327:891-6.
8
Ringdal, Respir Med 2003;97:234-41.
9
Nelson, J Allergy Clin Immunol 2000;106:1088-95.
10
Fish, Chest 2001;120(2):423-30.
11
Lee, Thorax 2004;59(8):662-7.
12
Lipworth, Br J Clin Pharmacol 2002;54:231-45.
13
McIvor, Am J Respir Crit Care Med 1998;158:924-30.
14
FDA Public
Health Advisory (updated 2006/5), http://www.fda.gov/cder/drug/advisory/LABA.htm
15
Salpeter, Ann Intern Med 2006;144:904-12.
16
Data on file, Merck & Co., Inc., Whitehouse Station, NJ, USA.
17
Tsuchiya, Discussion Paper Series, Sheffield Health Economics Group, May 2002, Ref: 02/1
18
Berger, (Eds.) ISPOR Book of Terms. ISPOR 2003:37.
19
Haynes, Cochrane Database Syst Rev 2005;(4):CD00011.
20
Kripalani, Arch Intern Med 2007;167:540-50
21
Schaffer, Clinical Nursing Research 2004;13:69-89.
22
Jones, J Asthma 2003;40(1): 103-5.
23
Carranza Rosenzweig, American Thoracic Society 99th International Conference, May 16-21, 2003, Seattle, Washington, USA, Abstract C039, Poster
C3.
24
British Thoracic Society, Thorax 2003;58:(suppl 1):194.
25
Pettersson, ISPOR 10th Annual European Congress, Oct. 20-23, 2007, Dublin, Ireland, Poster PAA11.
ICUR
MON vs. SAL
(Euro/QALY gain)
Comparisons using the best case pricing scenario
Drug effect and drug cost decreased with rel. compliance
Only drug effect decreased with rel. compliance
Comparisons using the worst case pricing scenario
Drug effect and drug cost decreased with rel. compliance
Only drug effect decreased with rel. compliance
ISPOR 10th Annual European Congress, Oct. 20-23, 2007, Dublin, Ireland
Risk Ratio (95% CI)
% MON+ICS vs. % SAL+ICS
0.80 (0.48; 1.35)
0.50 (0.26; 0.94)
0.66 (0.44; 0.97)
5.2% vs. 7.9%
0.63 (0.44; 0.90)
6.3% vs. 10.0%
Cardiovascular
(6 cost categories)
Oral infections
(2 cost categories)
Side effect costing model
(two above combined)
Overall drug-related
adverse event profile
SAL+ICS
more harmful
MON+ICS
more harmful
Relative risk
0.2 0.5 1 1.5
Figure 2. Components of the simplified side-effect costing model and its match with the overall drug-related adverse event profile
MON+ICS
(baseline: n=610 change: n=581*)
SAL+ICS
(baseline: n=602 change: n=581*)
Adjusted difference

(95%CI)
AQLQ index
Baseline
mean (SD)
4.70 (1.01) 4.74 (1.02)
Last measurement
mean (SD)
5.54 (1.01) 5.60 (1.08)
Adjusted improvement
LS mean (SE)
0.71 (0.04) 0.76 (0.04)
-0.05 (-0.150.06)
P=0.38
EQ-5D quotient (transformed from AQLQ)
Baseline
mean (SD)
0.721 (0.127) 0.725 (0.125)
Last measurement
mean (SD)
0.815 (0.119) 0.818 (0.124)
Change from baseline
Mean (SE)
0.0922 (0.143) 0.0907 (0.138) 0.00157 (0.0146-0.0178)
Adjusted improvement
LS mean (SE)
0.0803 (0.00544) 0.0812 (0.00541)
-0.000953 (-0.01360.0117)
P=0.88
MON+ICS SAL+ICS
Best case scenario Cheapest Most expensive
Worst case scenario Most expensive Cheapest
IMPACT scenario Price of the drugs used in IMPACT
Relative compliance (SAL versus MON)
PAA21