European Heart Journal (1999) 20, 263–268
Article No. euhj.1999.1193, available online at http://www.idealibrary.com on
International economic analysis of primary prevention
of cardiovascular disease with pravastatin in
WOSCOPS
J. Caro, W. Klittich, A. McGuire, I. Ford, D. Pettitt, J. Norrie and J. Shepherd for
the WOSCOPS Economic Analysis Committee*
Aims The results of the West of Scotland Coronary
Prevention Study (WOSCOPS) demonstrated the clinical
benefit of using pravastatin for the primary prevention of
cardiovascular disease in hypercholesterolaemic men. To
inform decision makers, who must also consider costs,
this study assesses the economic efficiency of such an
intervention in a broad range of countries.
Methods and Results A generalized model of cardiovascular disease prevention was used to estimate the costeffectiveness of primary prevention with pravastatin
compared to diet alone. This model follows a cohort of
hypercholesterolaemic men over a given period quantifying
the effect in terms of the avoidance of cardiovascular
disease based on treatment-specific risks derived from
WOSCOPS data and extensive record-linkage data on
disease-specific survival. Country-specific costs are
accounted for by expressing all such parameters in terms of
Introduction
The West of Scotland Coronary Prevention Study
(WOSCOPS)[1] demonstrated, for the first time, that
intervention in hypercholesterolaemic middle-aged men
with no evidence of previous myocardial infarction can
substantially lower the risk of cardiovascular disease
with no increase in mortality from other causes[2].
In today’s cost-conscious health care environment,
however, such impressive clinical results are not
sufficient — decision makers must also consider the
economic implications of this intervention.
In a previous analysis[3], we estimated the
economic efficiency of preventing cardiovascular
disease with pravastatin instead of diet alone from the
perspective of the National Health Service in the United
Kingdom. For those results to provide guidance to those
Revision submitted 9 June 1998, and accepted 10 June 1998.
*See Appendix for Committee members.
Correspondence: Dr J. Caro, 336 Baker Avenue, Concord, MA
01742, U.S.A.
0195-668X/99/040263+06 $18.00/0
the ratio of monthly treatment to that of managing a
myocardial infarction. Over a broad range of inputs the
cost-effectiveness ratios remain below $25 000 per life years
gained, regardless of country. Subgroups with even better
economic efficacy can be defined on the basis of higher
baseline risk.
Conclusions In contrast to some previous reports, this
analysis based on trial data demonstrates that pravastatin
provides not only an effective means of primary cardiovascular disease prevention, but also an efficient one.
(Eur Heart J 1999; 20: 263–268)
Key Words: Lipids, cost-effectiveness, cholesterollowering, coronary heart disease, prevention, general
economic model.
See page 245 for the Editorial comment on this article
who set policy in other jurisdictions, they must account
for differences among health care systems. Thus, this
study was undertaken to generalize the results of
WOSCOPS to the perspective of any national health
service or other organization responsible for societal
health care costs. To do so, a formula was derived using
the United Kingdom model as the basis. In this paper,
the derivation of the cost-effectiveness formula is presented, together with its validation by comparing the
results it provides with those of the original United
Kingdom model and a separate Canadian model. In
addition, sample calculations are made for three new
countries, Sweden, Belgium and South Africa. The general model thus created can be used to derive the specific
economic consequences for any given health jurisdiction.
Methods
United Kingdom model
The United Kingdom cost-effectiveness analysis used a
Markov model expressly created for economic analysis
1999 The European Society of Cardiology
264
J. Caro et al.
Table 1
Definition of variables in the formula used to calculate country-specific cost-effectiveness ratios
Item
Description
Relative risk reduction
Baseline risk
Benefit value
Drug usage
Scaling factor
Discounting factor
Treatment cost
Cost ratio
Abbreviation
Value
P
B
V
U
22%
15·6%
7·724†
39·26
s
88·54%
x
m
0·893
*
R
*
Proportion of cardiovascular disease events prevented
Risk of developing cardiovascular disease when only dietary advice is given
Value of preventing a transition to cardiovascular disease (e.g. life years gained)
Number of months of drug use over the model time horizon of 60 months
accounting for compliance and discounting
Factor that relates the cost of a myocardial infarction to the average cost of
cardiovascular events in the model
Correction factor required when costs are discounted and benefits are not
Local cost of one month’s supply of pravastatin 40 mg per day including
monitoring
Ratio of m to the cost of a myocardial infarction
*Depends on country.
†3·930 if the benefits are discounted.
of interventions of this type. Full details are provided
elsewhere[3]. Briefly, primary prevention of cardiovascular disease using pravastatin treatment was compared to
no primary intervention, each on top of normal dietary
advice. This analytic framework computes the number
of initial cardiovascular events (‘transitions’ from health
to illness) in a population defined by a set of risk factors
that include hypercholesterolaemia. The model runs in
one month cycles for whatever time period is being
considered. In each month, those who suffer a noncardiovascular death are removed from the cohort
before applying the relevant cardiovascular disease risk.
This risk determines the number of individuals who will
manifest cardiovascular disease (coronary or cardiovascular deaths, definite non-fatal myocardial infarction,
unrecognized or silent myocardial infarction, angioplasty, bypass grafting, angiography, hospitalized
angina, non-fatal stroke, and transient ischaemic attack)
for the first time during that month. These men are then
also removed from the cohort for subsequent months.
The process was carried out over a period of 60 months
(corresponding to the approximate average follow-up
time for a randomized subject in the trial) and the net
consequence of not using pravastatin was estimated as
the cumulative difference in the number of transitions.
Generalized formula
The original model was generalized to allow countryspecific assessments by deriving a formula that reexpresses the cost-effectiveness ratio, C/E in terms of the
elements that might be country-specific and those that
ought to be more general.
The coast-effectiveness ratio can be reformulated
as:
C
E
=
treatment cost − offsetting cost
effectiveness
=
m
V
F G
U
PB
−
xs
R
Table 1 lists the variables used in this formula, their
meanings, and baseline values.
Eur Heart J, Vol. 20, issue 4, February 1999
Most investigators would accept that the relative
effect, the proportion of events prevented, labelled P in
the formula above, is generally applicable. The baseline
risk, B, however, can be affected by differences in the
distribution of risk factors in different populations. With
this model, one can calculate the cost-effectiveness for
any given baseline risk.
The major factor that clearly characterizes the
country is the cost structure, but full accounting of these
details would require separate models. To avoid this, we
expressed the cost elements (cost of treatment and
offsetting cost due to events prevented) in terms of m,
the local cost of one year’s supply of pravastatin 40 mg
per day including any required monitoring. This approach required introducing some additional terms into
the equation. Since many patients who start treatment
do not continue it, either because they stop complying or
die, expressing the treatment cost in terms of m requires
the factor U. This factor reflects not only the usage of
the drug, but also discounting of costs.
To express the offsetting costs in terms of m
required two other new factors. First, we created a ratio,
labelled R, of the treatment cost to the average cost of
managing a myocardial infarction in each country
(chosen because it is probably easiest to obtain locally).
Second, a scaling factor, s, was defined to relate the
average cost of all cardiovascular events included in the
model to the cost of managing a myocardial infarction.
An additional factor is required: the local value,
V, placed on preventing one transition to cardiovascular
disease. If measured in terms of years of life gained, it
reflects the gains in life expectancy.
General parameters
To estimate P, the proportion of transitions from health
to cardiovascular disease prevented by pravastatin use,
the WOSCOPS data were re-analysed. The type of event
indicating the transition was considered by deriving the
corresponding proportion that each represents of the
Prevention of cardiovascular disease with pravastatin
265
Table 2 Country-specific cost factors used to calculate cost-effectiveness ratio in the
generalized model and rate of exchange used to convert all costs to like monetary units
(US$)
Country-specific cost factors
Average cost of managing MI (CMI)
Annual cost of treatment (m)**
R/(m/CMI)
Exchange rate (US$/local $)
Cost (US$)
U.K.
Canada
Sweden
Belgium
S. Africa
3077
1043
0·34
1·63
4362
701
0·16
0·72
2885*
635
0·22
0·13
7713†
1202
0·16
36·5
6698‡
896
0·13
5·1
MI=myocardial infarction.
*From [8].
†From [14].
‡Prof. H. Seftal (Southern Health Care JV) personal communication.
**Obtained from the local Bristol–Myers Squibb affiliate.
total number of transitions. All calculations were carried
out on an ‘intention-to-treat’ basis.
Although the baseline risk, B, may vary across
countries, it is more intelligible if it is considered in
terms of various patient subgroups. Thus, B, was
recalculated according to various risk profiles for cardiovascular disease, as detailed elsewhere[4]. Results are
presented over a range of 5 year risk of 0–30%.
One common expression of the value of preventing a transition, V, is the increased life expectancy
measured in life years gained. This was measured as the
difference between the survival curve after an event, and
the age-specific curve for males from the life table[5]. The
event-specific curves were estimated from data on more
than 460 000 cardiovascular disease events registered in
the Scottish Record Linkage System[6] from 1981 to 1994.
It was assumed that the change in life expectancy derived
from Scottish data would also apply to other countries.
Sensitivity analyses considered alternative possibilities, as
well as discount rates other than 5% per annum.
The factor U, representing the average number
of months of use over the 5 years, was derived by
considering the proportion of people starting on drug
who would continue to receive drug (one 40 mg tablet
daily) each month. This monthly value was estimated
as the proportion of men not yet transitioning to
cardiovascular disease, adjusted by the proportion in
WOSCOPS who had tablets dispensed at each 3-month
visit. U is the sum of these monthly proportions. As this
is a cost-factor, it was also discounted. The scaling
factor, s, was estimated initially on the basis of United
Kingdom data as 88·54%. It was found to change little
when it was recalculated for Canada. Indeed, upon
reflection, it is clear that it would require substantial
modification only for a country where management of
myocardial infarction is exceptionally different in relation to treatment of other cardiovascular disease events.
Validation
These calculations are carried out and presented over
broad ranges of variables that may be local and also
for five specific countries: United Kingdom, Canada,
Sweden, Belgium and South Africa. The values were
then compared to the results derived from countryspecific models developed for the United Kingdom and
Canada (personal communication, WOSCOPS Economic Analysis Committee, 10 November 1997). This was
done to validate the generalized equation methodology.
For the two validation countries, the annual cost of
using 40 mg of pravastatin per day, m, was based upon
the tablet price and the cost of monitoring, and included
a lipid profile and a visit to the general practitioner every
6 months. As there was no evidence in WOSCOPS of
significant side-effects due to pravastatin, no costs for
their management were included. Consistent with recent
labelling changes, liver function tests were also not
included.
For the United Kingdom, the cost of managing a
myocardial infarction was derived by combining estimates of average cost taken from extra-contractual
tariffs from a sample of over 200 Trusts and eventspecific average lengths of stay calculated from
WOSCOPS data (Table 2). For Canada, the cost was
based on data obtained for the Ontario Case Cost
Project[7], covering 13 hospitals. Neither costs of
management subsequent to the initial hospitalization
nor pre-admission costs were considered.
To aid comparisons across countries, all costs are
presented in US dollars based on currency exchange
rates published in the Wall Street Journal on 23 April
1997.
Results
Validation
Validation of the generalized formula for the United
Kingdom yields a cost-effectiveness ratio of $13 273
(£8297) per life years gained, which is the same as the
published estimate based on the detailed model. The
results for Canada are also quite accurate: $8876 per life
years gained for the generalized formula compared to
Eur Heart J, Vol. 20, issue 4, February 1999
266
J. Caro et al.
25
Cost effectiveness (US$ 1000/LYG)
Assumptions
5% discount rate
Life expectancy as in Scotland
age 50–55
inpatient costs only (1996 US$)
20
15
Drug price
($/month)
B
U
75
A
10
C
100
S
75
5
50
0
10
12
14
16
18
20
22
24
5 year cumulative risk (%)
26
28
30
Figure 1 Cost-effectiveness as a function of the untreated 5 year cumulative risk for various
monthly treatment costs. The width of the treatment cost bands is determined by R, ranging
from 0·1 to 0·4. The three lower bands do not have the benefit discounted, the uppermost band
discount benefits at 5% per annum. For illustration, the country-specific results are shown for
patients with the same baseline risk as those in WOSCOPS. B=Belgium; U=U.K.; A=South
Africa; C=Canada; S=Sweden; LYG=life year gained.
$7669 per life years gained for the independent model.
This concordance holds when assumptions, such as
discounting, are varied.
Sample countries
To illustrate the method we used the formula to
calculate the cost-effectiveness for Sweden, Belgium and
South Africa. We obtained estimates for the costs of
treatment and of managing a myocardial infarction
(Table 2). Using the Swedish estimates an R ratio of 0·22
was derived and combined with the values in Table 1.
We thus estimate a cost-effectiveness ratio of $8150 per
life year gained for Sweden. For Belgium, the R ratio is
0·16. This yields a cost-effectiveness ratio of $14 773/lifeyear gained. In South Africa, the cost effectiveness ratio
is $10 999/life year gained, based on an R ratio of 0·13.
Sensitivity analysis
The economic efficiency of preventing cardiovascular
disease is according to baseline risk, B. The costs of this
treatment and the savings due to preventing events is
shown in Fig. 1. Assuming that the factors that increase
cardiovascular disease risk do not adversely affect V, the
cost-effectiveness ratios improve with higher cardiovascular disease risk. R has limited effect as can be seen by
Eur Heart J, Vol. 20, issue 4, February 1999
the width of the bands (0·01 to 0·4). Moreover, as
treatment cost decreases, the impact of R lessens further.
As expected, at any given risk, higher treatment costs
imply less favourable cost-effectiveness ratios, but within
the range examined in this analysis, the values remain
below $25 000 per life years gained at a risk of cardiovascular disease as low as 9·5% over 5 years (undiscounted benefits). If the benefits are discounted, then a
5-year risk of at least 19·5% is required to stay below
$25 000 per life years gained at the highest treatment and
lowest event costs.
Apart from discounting, another factor that can
have some impact on the results is the gain in life years
ascribed to prevention of an event. This depends both on
the expected survival without events and the post-event
life expectancy. Both of these quantities are difficult to
estimate precisely. Nevertheless, some idea can be obtained. For example, if Canadian instead of Scottish life
tables are used, V increases by 0·81 to 4·74 (discounted),
or by 2·46 to 10·13 (undiscounted); with corresponding
cost-effectiveness ratios of $6768 per life years gained
(undiscounted benefits) and $16 199 per life years gained
(discounted). Thus, the change in values does not significantly alter the results. Indeed, when the V estimated
by the Scottish data is increased by 25%, the range of
resulting cost-effectiveness ratios is still only $6520 per
life years gained to $10 850 per life years gained (undiscounted) or $14 350 per life years gained to $23 881
per life years gained (discounted). Decreasing V by 25%
Prevention of cardiovascular disease with pravastatin
produces a range of $10 867 per life years gained to
$18 084 per life years gained (undiscounted) or $23 917
per life years gained to $39 801 per life years gained
(discounted).
Other factors and assumptions affect the results
relatively little. Variation in the scaling factor, s, over the
range observed in the selected countries has minimal
impact. This together with the limited effect of R indicates that the offsetting costs do not materially affect the
economic efficiency. Although they were estimated very
conservatively by considering only the first hospital
admission, it seems that expansion to include additional
elements within a reasonable range, would not alter the
results in a decision relevant way.
Discussion
The generalized formula described in this paper provides
a ready means for local adaptations of the WOSCOPS
United Kingdom economic model. Based on the validation carried out, the results appear sufficiently accurate
for health policy decision. Two major findings emerged
from this study. First, the resulting cost-effectiveness
ratios place this practice within bounds typically considered as ‘moderate to strong evidence for adoption
and appropriate utilization’[9]. Second, although the
precise estimate does depend on the specifics of the
country, it is notable that these variations do not have
much of an impact on the treatment decision that this
study supports.
One important issue in this analysis is the value
placed on preventing one transition. As already noted,
in order to quantify this transition in terms of life
years gained, projections of its impact on cumulative
survival were made beyond the end of the trial. This is
fraught with difficulty due to the scarcity of information
on the implications of cardiovascular disease for life
expectancy[10]. Nevertheless, to properly inform decision
makers these projections must be made and we have
done so on the basis of extensive contemporary data.
Not to do so, would imply the untenable assumption
that cardiovascular disease that does not result in
immediate death will not shorten life[11,12]. In any case,
over a fairly broad range of V, the positive conclusions implied by the cost-effectiveness ratios remain
consistent.
Other questions about the applicability of these
results may arise. This analysis did not include the costs
of detecting elevated cholesterol. It focused on patients
whose condition has already come to the attention of the
physician. Another concern may be that randomized
trial results do not reflect clinical practice. The compliance observed in WOSCOPS, however, was obtained
without extensive measures to ensure adherence to
the prescribed regimen — an average clinical practice
should be able to replicate them and organized programmes may even be able to exceed them. A third issue
with applicability has to do with the nature of the
267
WOSCOPS population. At its most strict interpretation,
the results apply to middle-aged Scottish males. Most
would extend these results to similar males in other
locations. Extension to females or to broader definitions
of the male population is more problematic, however, as
the efficacy of pravastatin has not been documented for
them. Nevertheless, so long as the hypercholesterolaemic
baseline risk is within the range explained here, the
economic efficiency should be similar.
The implications of this analysis using the
generalized model are consistent over a broad range of
factors that might be country-specific. This range covers
health care systems similar to those prevailing in most
industrialized nations. In countries where either the
pravastatin intervention costs are much higher than
those in the United Kingdom or the life expectancy is
much worse, the conclusions might not hold. However,
it would take an extremely unlikely treatment cost more
than seven-fold higher to do so (exceed $100 000 per life
years gained, discounted); or an equally implausible
scenario of gaining only one year of life by preventing
a transition. Needless to say, if very precise estimates
of the cost-effectiveness ratio, or its components, are
required for decision-making in a given country, then a
detailed analysis specific to that country may be
required. This would be the case if an aggregate analysis
aimed at estimating the total cost of primary prevention,
rather than just its economic efficiency, were to be
carried out.
The impact of baseline risk on cost-effectiveness
is somewhat unclear. While the number of events prevented increases as risk increases, the relation of V to
baseline risk remains uncertain. It is possible that the
same factors that increase risk diminish V when it is
quantified in terms of life expectancy. In this situation,
an increase in risk would also be associated with a
reduced V, perhaps even cancelling each other out. In
any case, such projections are unpredictable. A more
in-depth analysis which considers this relation is necessary to provide guidance for policy that envisions
restricting primary prevention according to risk.
In some countries, it may be felt that the implications of given risk factors for baseline risk would be
different from those observed in Scotland. If this were
the case, the risk equations derived in WOSCOPS may
not be applicable but the relationship of economic
efficiency to baseline risk would remain valid as long as
the relative efficacy of pravastatin, P, were considered
applicable. Thus, this model would still provide the
cost-effectiveness ratio for any population identified by a
given baseline risk.
Caution should be used when attempting to
extrapolate the clinical benefits used within this analysis
to other HMG-CoA reductase inhibitors, as no clinical
trial has shown that these benefits would be similar for
other statins. The analyses of the WOSCOPS data have
produced risk equations that suggest that even when
cholesterol lowering and other known determinants of
disease are controlled for, pravastatin use still lowers the
risk of clinical events further[13]. This additional effect
Eur Heart J, Vol. 20, issue 4, February 1999
268
J. Caro et al.
emphasizes the importance of basing analyses on direct
measurement of outcomes as opposed to using proxies
such as lipid values.
Health policy makers increasingly face the kind
of decision posed by WOSCOPS — a proven highly
effective means of primary prevention of cardiovascular
disease, which is perceived to be a serious strain on
constrained budgets. This analysis suggests that the
cost-effectiveness of primary prevention with pravastatin
should fit within bounds set by accepted therapies in
most countries.
This work was supported by grants from Bristol–Myers Squibb.
References
[1] The West of Scotland Coronary Prevention Study Group. A
coronary primary prevention study of Scottish men aged
45–64 years: trial design. J Clin Epidemiol 1992; 45: 849–60.
[2] Shepherd J, Cobbe SM, Ford I et al. Prevention of
coronary heart disease with pravastatin in men with hypercholesterolemia. N Engl J Med 1995; 333: 1301–7.
[3] Caro J, Klittich W, McGuire A et al. West of Scotland
Coronary Prevention Study: economic benefit analysis of
primary prevention with pravastatin. Br Med J 1997; 7122:
1577–82.
[4] West of Scotland Coronary Prevention Group. Baseline risk
factors and their association with outcomes in WOSCOPS.
Am J Cardiol 1997; 79: 756–62.
[5] Registrar General Scotland. Annual Report 1994. Her
Majesty’s Stationery Office, Edinburgh, 1995.
[6] Kendrick S, Clarke J. The Scottish Record Linkage System.
Health Bull (Edinb) 1993; 51: 72–9.
[7] Ontario Hospital Association. Ontario Guide to Case
Costing. OHA publication #238.
[8] Jönsson B, Johannesson M, Kjekshus J, Olsson AG, Pedersen
TR, Wedel H. Cost-effectiveness of cholesterol lowering.
Results from the Scandinavian Survival Study (4S). Eur Heart
J 1996; 17: 1001–7.
Eur Heart J, Vol. 20, issue 4, February 1999
[9] Laupacis A, Feeny D, Detsky AS, Tugwell PX. How
attractive does a new technology have to be to warrant
adoption and utilization? Tentative guidelines for using
clinical and economic valuations. Can Med Assoc J 1992; 146:
473–81.
[10] Caro JJ, Klittich WS for the WOSCOPS Economic
Committee. Estimating the life years gained from preventing
cardiovascular disease. American College of Cardiology 47th
Annual Scientific Session, Atlanta, March, 1998. (Poster)
[11] Pharoah PDP, Hollngworth W. Cost effectiveness of lowering
cholesterol concentration with statins in patients with and
without pre-existing coronary heart disease: life table method
applied to health authority population. Br Med J 1996; 312:
1443–8.
[12] Caro JJ for the WOSCOPS Group. Letter of Response. Br
Med J 1998; 316: 1241–2.
[13] West of Scotland Coronary Prevention Study Group.
Influence of pravastatin and plasma lipids on clinical events
in the West of Scotland Coronary Prevention Study
(WOSCOPS). Circulation 1998; 97: 1440–5.
[14] Muls E, Van Gense E, Closon MC. Cost-effectiveness of
pravastatin in secondary prevention of coronary heart disease:
comparison between Belgium and the United States of a
projected risk model. Atherosclerosis 1998; 137: S111–6.
Appendix
WOSCOPS Economic Analysis Committee
Jaime Caro, Ian Ford, Wendy Klittich, Alistair
McGuire, John McMurray, John Norrie, Daniel Pettitt,
James Shepherd
WOSCOPS Executive Committee
J. Shepherd (chairman), S. M. Cobbe, A. R. Lorimer,
J. H. McKillop, I. Ford, C. J. Packard, P. W.
Macfarlane, C. G. Isles