Full text of DOI:10.2165/00019053-200018040-00004

Pharmacoeconomics 2000 Oct; 18 (4): 355-368
1170-7690/00/0010-0355/$20.00/0
ORIGINAL RESEARCH ARTICLE
© Adis International Limited. All rights reserved.
Cost Effectiveness of Human
Immunodeficiency Virus Postexposure
Prophylaxis for Healthcare Workers
Dewey C. Scheid, Robert M. Hamm and Kevin W. Stevens
Clinical Decision Making Program, Department of Family and Preventive Medicine, College of
Medicine, University of Oklahoma Health Sciences Center, Oklahoma City, Oklahoma, USA
Objective: The United States Public Health Service (USPHS) published recom-
mendations for human immunodeficiency virus (HIV) postexposure prophylaxis
(PEP) of healthcare workers in May 1998. The aim of this study was to analyse
the cost effectiveness of the USPHS PEP guidelines.
Abstract
Design and setting: This was a modelling study in the setting ofthe US healthcare
system in 1998. The analysis was performed from the societal perspective; how-
ever, only HIV healthcare costs were considered and health-related losses of
productivity were not included.
Methods: A decision tree incorporating a Markov model was created for 4 PEP
strategies: the current USPHS recommendations, triple drug therapy, zidovudine
monotherapy or no prophylaxis. Aprobabilistic sensitivity analysis using a Monte
Carlo simulation was performed. Confidence intervals (CIs) around cost-effec-
tiveness estimates were estimated by a bootstrapping method.
Results: The costs (in 1997 US dollars) per quality-adjusted life-year (QALY)
saved by each strategy were as follows: monotherapy $US688 (95% CI: $US624
to $US750); USPHS recommendations $US5211 (95% CI: $US5126 to $US5293);
and triple drug therapy $US8827 (95% CI: $US8715 to $US8940). The marginal
cost per year of life saved was: USPHS recommendations $US81 987 (95% CI:
$US80 437 to $US83 689); triple drug therapy $US970 451 (95% CI: $US924 786
to $US1 014 429). Sensitivity testing showed that estimates of the probability of
seroconversion for each category of exposure were most influential, but did not
change the order of strategies in the baseline analysis. With the prolonged HIV
stage durations and increased costs associated with recent innovations in HIV
therapy, the marginal cost effectiveness of the USPHS PEP strategy was de-
creased to $US62 497/QALY saved. All 3 intervention strategies were cost effec-
tive compared with no postexposure prophylaxis.
Conclusions: Current USPHS PEP recommendations are marginally cost effec-
tive compared with monotherapy, but the additional efficacy of triple drug therapy
for all risk categories is rewarded by only a small reduction in HIV infections at
great expense. For the foreseeable future, assuming innovations in therapy that
employ expensive drug combinations earlier in the HIV disease course to extend
life expectancy and the increasing prevalence of HIV drug resistance, our model
supports the use of the USPHS PEP guidelines.

356
Scheid et al.
It has been estimated that hospital-based health-
been 2 previous economic analyses of zidovudine
chemoprophylaxis.[10,11] Assuming 75% efficacy of
zidovudine and using both a societal and third-party
payer perspective, Ramsey and Nettleman[10] re-
ported a cost per life-year gained of $US3184 and
$US15 788, respectively. The lower cost per year
of life gained in the societal perspective was due to
the conservation of healthcare worker productivity.
Allen et al.,[11] in an analysis using a societal per-
spective and assuming 75% efficacy of zidovudine,
determined that prophylaxis yielded a net saving of
$US6 244 000 per case prevented. There has been
1 published cost-effectiveness analysis of triple drug
PEP by Pinkerton et al.[12] This analysis found that
triple drug therapy was favourable at $US37 148
per year of quality-adjusted life saved, assuming a
79% treatment efficacy and a seroconversion rate
of 0.32%.
In the absence of a randomised controlled trial
that compares different PEP regimens, we attempt-
ed to answer the question of what is the most cost-
effective strategy for healthcare workers exposed
to HIV with a cost-effectiveness analysis of USPHS
PEP recommendations and 3 other PEP alternatives:
no therapy, monotherapy with zidovudine and tri-
ple drug therapy.
care workers experience at least 500 000 percuta-
neous blood exposures annually in the US alone.[1]
Approximately 5000 of these exposures involve
blood that is infected with human immunodefici-
ency virus (HIV). Although the average risk of trans-
mission after percutaneous exposure has been esti-
mated to be 0.3%, the risk is thought to be much
higher with higher blood volume and viral load.
AUnited States Public Health Service (USPHS)
interagency working group updated the previous
USPHS statement on management of occupational
exposure to HIV and recommendations for post-
exposure prophylaxis (PEP) in June 1996, prompted
by the rapid accumulation of information suggest-
ing the benefits of zidovudine prophylaxis, the po-
tency of combinations of antiretroviral drugs, and
the ability to stratify risk by type of exposure.[2]
These provisional recommendations included a ba-
sic regimen of zidovudine prophylaxis and a double
drug therapy for the highest risk group. All prev-
ious recommendations were again updated and con-
solidated in May 1998.[3] The latest USPHS PEP rec-
ommendations include a basic 4-week antiretroviral
prophylaxis regimen using 2 drugs and an expanded
regimen that includes the addition of a protease in-
hibitor, using an algorithm that varies the recommen-
dations according to risk of transmission by type of
exposure and the HIV status of the source.
Methods
Decision Analytic Model
Although PEP with zidovudine has been widely
used after healthcare worker exposure to reduce the
risk of transmission,[4-6] the only randomised con-
trolled trial of postexposure zidovudine was termi-
nated because an insufficient number of exposed
workers were recruited.[7] The use of postexposure
zidovudine is supported by an international case-
control study of HIV seroconversion in healthcare
workers after percutaneous exposure, in which use
of zidovudine was associated with a 79% risk re-
duction.[8] Further evidence of benefit is based on
a prospective trial of zidovudine administered to
HIV-infected pregnant women that demonstrated a
67% reduction of perinatal HIV transmission.[9]
The cost effectiveness of the current USPHS guide-
lines for PEP has never been established. There have
A decision tree incorporating a Markov model
was developed to compare 4 different PEP strate-
gies for exposures in which the HIV status of the
source was known:
• current USPHS recommendations
• triple therapy with zidovudine, lamivudine and
indinavir for all exposures
• monotherapy with zidovudine for all exposures
• no PEP.
The model was adapted from a previous model
of the 1996 provisional USPHS PEP guidelines.[13]
The analysis was performed from the societal per-
spective, i.e. all costs and all health benefits were
accounted for regardless of who incurred them.[14]
However, only HIV healthcare costs were consid-
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)

HIV Postexposure Prophylaxis
357
Alive
Alive
Dead
Male
M
M
HIV−
Dead
Female
[+]
CD4 >500
CD4 >200
CD4
>500
Choose
no PEP
CD4 >200
CD4 <200
CD4
>200
CD4 <200
AIDS
CD4
<200
HIV+
M
AIDS
Dead
Expanded
regimen
AIDS
Dead
HIV−
Major
[+]
[+]
adverse
USPHS
recommendations
Resistant
HIV
HIV+
[+]
Minor
adverse
Choose
PEP
No
adverse
Nonresistant
HIV
[+]
[+]
Basic
regimen
[+]
[+]
[+]
PEP?
Consider
basic regimen
PEP not
warranted
Triple
therapy
[+]
[+]
[+]
Monotherapy
No PEP
Fig. 1. Decision tree for HIV postexposure prophylaxis. [+] indicates truncated tree branches with identical subtrees, for example
female [+] = male. Nodes labelled with M begin Markovsubtrees. CD4 = CD4+ count (cells/μl); HIV–HIV-negative; HIV+ = HIV-positive;
PEP = postexposure prophylaxis; USPHS = US Public Health Service.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)

358
Scheid et al.
ered and health-related losses of productivity were
not included. The cost effectiveness with respect to
3 outcomes was evaluated: HIV infections, years of
life saved and quality-adjusted life-years (QALYs).
Monetary values are reported in 1997 US dollars.
Years of life, QALYs and cost outcomes were dis-
counted at 0, 3 and 5% for all strategies. Sensitivity
analyses were performed to determine the robust-
ness of the model.
A commercially available software product
(DATA™ 3.0; TreeAge Software, Inc., Williams-
town, MA, USA) was used to generate the model
and analyse cost effectiveness. A partially expan-
ded decision tree is presented in figure 1. Summa-
ries of model input parameters, including baseline
estimates, ranges and references, are included in
tables I and II.
cous membrane or skin exposure with higher titre
HIV status (EC1 and HIV SC2). PEP is not war-
ranted with no known risk exposures, i.e. small vol-
ume mucous membrane or skin exposure with
lower titre HIV status (EC1 and HIV SC1).
The probability of an exposure falling into one
of these classification categories was estimated
using data reported from large surveillance stud-
ies.[6,7,15,16] For example, we modelled the probab-
ility of EC3 by combining the independent probabil-
ities of percutaneous exposure, penetrating exposure
and visible blood on the device, such that:
EC3 = 0.661 × 0.800 × 0.309 = 0.164
(see table I for other probabilities and calculations).
The exposure classification groups were then com-
bined, as above, to determine the probability of an
exposure for each of the USPHS PEP risk catego-
ries.
Exposure Categories
USPHS PEP guidelines were used to determine
4 categories of exposure and corresponding recom-
mended PEP regimens:[3]
• expanded
• basic
• consider basic
• not warranted.
Transmission Probabilities
We could find no published data that reported
the probability of transmission for groups similar
to the USPHS exposure categories. Based on the
calculated probability of falling into each of the
exposure classification categories and the relative
odds of seroconversion after exposures with spe-
cific risk factors reported in an international case
control study, we estimated the probability of trans-
mission for each of the classification categories.[17]
Risks associated with exposure categories and the
HIV status of the exposure source were combined
to determine the risk of transmission for the 4 USPHS
PEP risk categories. There is evidence to suggest
that exposure risk and the risk associated with the
HIV status of the source operate independently and
hence the odds ratios are multiplicative.[1]
The USPHS PEP guidelines classify exposures
according to the type of exposure (exposure code;
EC) and HIV status of the exposure source (HIV
status code; HIV SC). For instance, the most severe
exposure code (EC3) includes exposures that are
percutaneous and involve a hollow needle or visi-
ble blood on the device. We used a source with
terminal AIDS as a surrogate for higher titre HIV
status (HIV SC2). The expanded regimen is recom-
mended for increased risk exposures, i.e. either the
most severe exposure (EC3) regardless of HIV sta-
tus code or less severe exposure combined with
higher titre HIV status (EC2 and HIV SC2). The
basic regimen is recommended for no increased
risk exposures, i.e. less severe percutaneous expo-
sure or large volume mucous membrane or skin ex-
posure with lower titre HIV status (EC2 and HIV
SC1). The basic regimen should be considered for
negligible risk exposures, i.e. small volume mu-
The reduction of HIV transmission following
PEPwith zidovudine was 79%.[8] Because there are
no studies of the efficacy of multidrug PEP, the
baseline efficacy of PEP therapy was assumed to
be 90% for 3-drug therapy, i.e.:
efficacy of triple therapy = efficacy of monotherapy +
dt × (1 – efficacy of monotherapy)
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)

HIV Postexposure Prophylaxis
359
Table I. Summary of exposure, transmission and treatment parameters
Parameter
Baseline
estimate
Range for
sensitivity analysis
Reference or derivation
Demographic variables
mean age of exposed healthcare workers (y)
female healthcare workers (%)
33.5
78.0
18-65
75.1-80.5
5,6,15,16
5,6,15,16
Exposure characteristics (%)
percutaneous exposure (A)
penetrating exposure (B)
blood visible on device (C)
mucocutaneous, large volume (D)
source patient with terminal AIDS (E)
66.1
80.0
30.9
11.7
18.3
63.7-68.4
76.6-83.1
27.5-34.5
7.2-17.8
5,6,15,16
5,6,15,16
5,6,15,16
5,6,15,16
8
14.6-22.7
Exposure classificationa (%)
EC3 (more severe percutaneous)
EC2P (less severe percutaneous)
EC2M (large volume mucocutaneous)
EC1 (small volume mucocutaneous)
SC1 (lower titre exposure)
16.4
49.7
4.0
29.9
81.7
18.3
Calculated as A × B × C
Calculated as A × (1 - B × C)
Calculated as (1 - A) × D
Calculated as (1 - A) × (1 - D)
Calculated as 1 - E
SC2 (higher titre exposure)
Calculated as E
Relative odds of exposure classifications
SC2 vs SC1
EC3 vs EC2P
EC2M vs EC1
5.6
6.2
6.2
2.0-16
2.2-21
2.2-21
17
17
Assumption
Transmission estimates (%)
percutaneous exposure
0.32
0.026
0.72
0.08
0.05
0.009
0.12-0.47
0.006-0.09
18
1,16
Calculatedb
Calculatedb
Calculatedb
Calculatedb
mucocutaneous exposure
expanded regimen eligiblea
basic regimen eligiblea
consider basic regimena
no regimen recommendeda
Efficacy of therapy (% seroconversion)
monotherapy
double therapy
21
15.5
10
6-57
3-57
1-57
8
Calculatedc
Calculatedc
triple therapy
Acceptance of therapy by regimen (%)
expanded regimena
76
76
21
69
58-86
58-86
14-31.1
62.7-74.8
3-28
6,7,19
4,7,19
4,7,19
4,7,19,20
4,7,19,20
21-24
basic regimena
consider basic regimena
Completion of therapy (%)
Partial completion (days)
Zidovudine resistance rate (%)
Major adverse effects (%)
8
17.5
0.001
0.0001-0.01
4,6,10,19,20
a
b
c
From Public Health Service PEP guidelines.[3]
Calculated using relative odds, exposure probability and transmission estimates for percutaneous or mucocutaneous exposure.
Efficacy of triple therapy = efficacy of monotherapy + dt × (1 – efficacy of monotherapy), where dt = 0.5, range 0 to 1. Efficacy of double
therapy = efficacy of monotherapy + dd × (efficacy of triple therapy – efficacy of monotherapy), where dd = 0.5, range 0 to 1.
dd = multiplier for improved double drug efficacy; dt = multiplier for improved triple drug efficacy; EC = exposure code; PEP = postexposure
prophylaxis; SC = HIV status code.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)

360
Scheid et al.
where dt, the multiplier for improved triple drug
efficacy, was set at 0.5, and 84% for 2-drug therapy,
i.e.:
Costs
The cost of each PEP strategy was based on costs
of postexposure counselling, costs of PEP, average
healthcare costs of HIV-seropositive adults and the
average healthcare costs of all adults. The cost of
postexposure counselling included the cost of pe-
riodic office visits and HIV testing according to the
recommended schedule. The cost of PEP included
the average wholesale costs of antiviral drugs,
periodic office visits, and toxicity monitoring as
outlined in a current multicentre clinical trial of
PEP.[18,25] Information regarding the toxicity of anti-
retroviral drugs for PEP is available only for zi-
dovudine.[7,8,26] With the use of multidrug PEP, the
potential for toxicity, subsequent need for monitoring
and associated costs are substantially increased.[3]
The probabilities and the costs of major adverse
effects were included in the model. A summary of
all PEP cost estimates is included in table II.
For healthcare workers who became HIV posi-
tive, we estimated direct HIV healthcare costs and
life expectancy by using a staged Markov model[37,38]
with 5 states:
efficacy of double therapy = efficacy of monotherapy +
dd × (efficacy of triple therapy – efficacy of monother-
apy)
where dd, the multiplier for improved double drug
efficacy, was set at 0.5.
Treatment Efficacy
Previous studies have shown a differential rate
of acceptance and completion of PEP.[6,7,19] In our
baseline analysis, 76% of healthcare workers with
either an exposure classified as increased risk or
no increased risk accepted PEP, whereas 21% of
healthcare workers with a negligible risk exposure
chose PEP. Although serious short term toxicity does
not appear to be a major concern, up to 31% of
healthcare workers do not complete PEP because
of adverse effects.[6,7,19,20] For the healthcare work-
ers who partially completed therapy, the efficacy
of PEP was decreased as a function of the propor-
tion of the therapeutic regimen they did not com-
plete. This was modelled by multiplying the PEP
efficacy by a logarithmic function, adjusted to vary
from 0 to 1 as the days of therapy ranged from 0 to
28 using the formula log (10 × days/28), with zero
effect if taken less than 3 days. For example, if a
healthcare worker took the drugs for 8 days (the
mean duration for those who did not complete the
regimen[4,7,19,20]), the efficacy of PEP was 46% com-
pared with 79% for monotherapy for those who
completed the regimen.
• HIV stage with CD4+ cell count >500/μl
• HIV stage with CD4+ cell count >200/μl
• HIV stage with CD4+ cell count <200/μl
• AIDS
• death.
A summary of the cost of HIV healthcare by
stage obtained from 3 previously published cost
studies is included in table II.[29-31]
Life Expectancies
We obtained estimates of the durations of each
HIV stage from the same cost studies.[29-31] The
Markov model was simplified by allowing transi-
tion only through a progression of HIV stages to
death. We used quality-of-life adjustments for HIV
disease based on a survey of healthcare workers
using a time trade-off method.[34] The median qual-
ity adjustments for the relevant states reported in
this study were used. Quality-of-life adjustments
for healthcare workers who did not seroconvert var-
ied by age according to the community-based
The prevalence of zidovudine resistance is in-
creasing.[21-23] Failures of zidovudine prophylaxis
associated with zidovudine resistance have been re-
ported.[24] The model accounted for encountering
zidovudine resistance from the HIV exposure source:
if the source had zidovudine resistance, the number
of antiretroviral drugs would be increased by 1 (i.e.
double therapy instead of monotherapy) without
increasing the efficacy.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)

HIV Postexposure Prophylaxis
361
Table II. Summary of cost and utility parameters. Costs are expressed in 1997 US dollars
Parameter
Baseline
estimate
Range for sensitivity Reference or derivation
analysis
Cost of HIV PEP
Cost of office visits (CPT 99212) [$US]
28.89
17.77-40
27
Number of office visits
3
18,25
Cost of toxicity monitoring laboratory tests ($US)
complete blood count
blood chemistry
urinalysis (indinavir)
amylase (lamivudine)
15.97
23.35
8.75
14.60
3
7.93-24
15.69-31
4.49-13
9.20-20
27
27
27
27
number of times healthcare worker is tested
18,25
Cost of antiretroviral drugs ($US)
zidovudine
lamivudine
indinavir
257
206
391
221-284
178-223
333-420
28
28
28
Cost of PEP strategy ($US)
monotherapy
double therapy
triple therapy
Cost of major adverse effects
609
892
1331
490
Calculated
Calculated
Calculated
9
Healthcare costs for HIV disease by stage ($US)
CD4+ >500 cells/μl
CD4+ >200 cells/μl
CD4+ <200 cells/μl
AIDS
3384
5160
11 880
33 168
1934-3528a
29-31
29-31
29-31
29-31
5160-7584a
9031-15 720a
25 239-67 713a
Duration of HIV stages (y)
CD4+ >500 cells/μl
CD4+ >200 cells/μl
CD4+ <200 cells/μl
AIDS
5.6
5.5-7
29,31-33
29,31-33
29,31-33
29,31-33
3.67
1.03
2.08
3.67-5.8
1.03-1.7
1.3-6.2
Quality of life for HIV by stage (%)
CD4+ >500 cells/μl
CD4+ >200 cells/μl
CD4+ <200 cells/μl
AIDS
83
83-91
62-88
42-74
17-80
34,35
34,35
34,35
34,35
83
42
17
Inflation factors (medical consumer price index)
1995-1997
1.07
1.264
36
36
1992-1997
a
Univariate and multiway sensitivity analyses were also performed using 95% confidence intervals, not shown.[29]
PEP = postexposure prophylaxis.
Beaver Dam Health Outcomes Study.[39] To assess
the impact of new treatmentsonthecostofHIVhealth-
care, the model was adjusted by adding the costs
obtained from Hellinger:[29] the net cost of triple
drug therapy applied to all HIV stages where the
CD4+ cell count was <500/μl and to the proportion
(38%) of those with CD4+ cell counts >500/μl
who would have HIV RNA viral copies >10 000/ml.
For healthcare workers who did not seroconvert
following exposure, we estimated life expectancy
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)

362
Scheid et al.
using a simple Markov model with 2 states: alive
and dead. Healthcare workers entered the model
with a mean age at time of exposure of 33.5 years.
Transition probabilities were obtained from separ-
ate life tables for male and female healthcare work-
ers at the start of each cycle year.[40]
calculated parameters varied the parameters over
their entire range. For instance, the efficacy of tri-
ple drug therapy was calculated with the equation:
efficacy of triple therapy = efficacy of monotherapy +
dt × (1 – efficacy of monotherapy)
For sensitivity analyses, the efficacy of mono-
therapy was varied according to published 95% con-
fidence intervals. In addition, dt was varied from 0
to 1 so that the efficacy of triple drug therapy varied
from the same as that of monotherapy up to 100%.
To further test model uncertainty, we performed
a multiway probabilistic sensitivity analysis using
a Monte Carlo simulation as described by Doubilet
et al.[41] Normal distributions for healthcare work-
ers’ age at exposure and costs by HIV stage from
the Hellinger[29] cost model were used. Costs for
PEP were sampled from triangular distributions.
Probabilities were sampled from a logistic normal
distribution for exposure category, transmission, PEP
efficacy, the probability of acceptance of PEP and
completion of PEP. The Monte Carlo simulation
used 10 000 trials to evaluate the model, producing
a distribution of cost and QALY values. To model
sample uncertainty, which may be of concern to
Analysis
The cost effectiveness of each strategy with re-
spect to 3 outcomes was evaluated: HIV infections
prevented, years of life saved and QALYs saved.
The marginal cost of an HIV infection prevented,
the marginal cost per year of life saved, and the
marginal cost per QALY saved were calculated for
each strategy. This was repeated using each of the
3 cost models.
One-way sensitivity analysis was performed for
transmission rates of each exposure category, effi-
cacy of double and triple therapy, the probability
of acceptance of PEP, the cost of antiviral drugs,
quality of life and the cost of healthcare for HIV-
seropositive healthcare workers. Sensitivity analysis
bounds were the upper and lower 95% confidence
intervals when known. The sensitivity analyses of
Table III. Cost effectiveness of HIV postexposure prophylaxis strategies based on the cost model of Hellinger.[29] Costs are expressed in
1997 US dollars
Strategy
Costa ($US per 1000
healthcare workers)
Effectivenessb (per 1000
healthcare workers)
Cost effectiveness ($US
per unit of effectiveness)
Marginal cost
effectivenessc ($US per
unit of effectiveness)
Infections prevented
Monotherapy
USPHS PEP
Triple therapy
22 000
181 000
307 000
2.08
2.21
2.22
10 577
81 900
138 288
1 304 717
12 273 996
Life-years saved
Monotherapy
USPHS PEP
22 000
181 000
307 000
31.2
33.0
33.2
713
5 485
9 247
87 139
819 747
Triple therapy
Quality-adjusted life-years saved
Monotherapy
USPHS PEP
Triple therapy
22 000
181 000
307 000
33.6
35.6
35.8
654
5 087
8 587
80 865
760 730
a
b
c
Discounted 3% annually.
Compared with no PEP, discounted 3% annually for life-years saved and quality-adjusted life-years saved.
Compared with the previous strategy.
PEP = postexposure prophylaxis; USPHS = US Public Health Service.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)

HIV Postexposure Prophylaxis
363
Table IV. Cost effectiveness of HIV postexposure prophylaxis strategies by cost model; parentheses indicate cost saving. Costs are expressed
in 1997 US dollars
Strategy
Costa ($US per 1000
healthcare workers)
Effectiveness (QALYs
Cost effectiveness
Marginal
saved per 1000 healthcare ($US/QALY saved)
cost-effectivenessc
($US/QALY saved)
workers)b
Hellinger cost model[29]
Monotherapy
USPHS PEP
Triple therapy
22 000
181 000
307 000
33.6
35.6
35.8
654
5 087
8 587
80 865
760 730
Gable et al. cost model[30]
Monotherapy
USPHS PEP
98 000
261 000
388 000
33.6
35.6
35.8
2 915
7 335
10 853
83 107
762 972
Triple therapy
Hurley et al. cost model[31]
Monotherapy
USPHS PEP
(16 000)
140 000
267 000
32.0
33.8
34.0
(501)
4 138
7 855
83 836
798 880
Triple therapy
a
b
c
Discounted 3% annually.
Compared with no PEP, discounted 3% annually.
Compared with the previous strategy.
PEP = postexposure prophylaxis; QALY = quality-adjusted life-year; USPHS = US Public Health Service.
policy-makers, bootstrap samples of 5000 and 500
trials were chosen to represent 2 different health-
care worker population sizes for which PEP strat-
egies could be implemented.[42] 1000 bootstrap sam-
ples were drawn by sampling with replacement from
the distribution produced by the Monte Carlo sim-
ulation to determine the mean cost, QALYs saved,
cost effectiveness and marginal cost effectiveness
of each strategy.[42,43] Confidence intervals were
determined by the 2.5th and 97.5th percentiles of
the distribution of the bootstrapped replicants.[43]
cost effectiveness of USPHS PEP compared with
monotherapy was $US1 304 717 per infection pre-
vented, $US87 139 per year of life saved and
$US80 865 per QALY saved. The marginal cost
effectiveness of triple therapy compared with
USPHS PEP was $US12 273 996 per infection
prevented, $US819 747 per year of life saved and
$US760 730 per QALY saved.
The baseline cost, effectiveness, cost effective-
ness and marginal cost effectiveness for each strat-
egy for all 3 cost models are shown in table IV. Ef-
fectiveness was the same for the Hellinger and Gable
et al. cost models because they used the same source
to estimate HIV stage durations.[29,30] The order of
the estimates of cost effectiveness was stable across
the 3 cost models. The most favourable estimates
were obtained using costs and HIV stage durations
from the Hurley et al. cost model, because base-
line costs for no PEP were the highest of the 3 cost
models. The costs of treating HIV infection were high
enough that monotherapy saved $US501 per QALY
saved and $US7692 per infection prevented. The
marginal cost effectiveness of the USPHS PEP and
triple therapy strategies for all 3 cost models were
Results
Baseline Analysis
The baseline cost, effectiveness, cost effective-
ness and marginal cost effectiveness per infection
prevented, per year of life saved and per QALY
saved, based on the Hellinger[29] cost model and
expressed in 1997 US dollars with an annual dis-
count rate of 3%, are shown in table III. Monother-
apy was the most cost-effective strategy: $US10 577
per infection prevented, $US713 per year of life
saved and $US654 per QALY saved. The marginal
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)

364
Scheid et al.
Seroconversion
Cost of CD4 >500
Cost of AIDS
Cost of CD4 >200
Cost of CD4 <200
7000 6000 5000 4000 3000 2000 1000
0
−1000
Cost effectiveness of monotherapy ($US/QALY saved)
Fig. 2. Tornado diagram summarising 1-way sensitivity analyses of the cost effectiveness of postexposure prophylaxis of HIV with
zidovudine monotherapy. Parameters were varied as follows: seroconversion after percutaneous exposure 0.18 to 0.46%; cost of
CD4+ count >500 cells/μl stage $US214 to $US349; cost of AIDS stage $US2607 to $US2921; cost of CD4+ count >200 cells/μl
stage $US351 to $US508; cost of CD4+ count <200 cells/μl stage $US839 to $US1141 (cost ranges are the 95% confidence intervals
around the mean monthly costs[29]). Negative values for cost effectiveness indicate that monotherapy is cost saving. Costs are
expressed in 1997 US dollars. CD4 = CD4+ count (cells/μl); QALY = quality-adjusted life-year.
quite similar. The marginal cost effectiveness of the
USPHS PEP guidelines was $US80 865 to $US83
836 per QALY saved across all 3 cost models.
saving. Costs for HIV stages were also influential,
but did not change the order of strategies from the
baseline analysis. All other variables in the model
had substantially less effect on model uncertainty.
As the probability of zidovudine resistance increa-
sed, the cost effectiveness of USPHS PEP guide-
lines approached that of triple therapy. However,
even when zidovudine resistance was 100%, the
marginal cost effectiveness of triple therapy com-
pared with USPHS PEP was still $US160 256 per
QALY saved. The model was insensitive to differ-
ences in the efficacy of double or triple drug ther-
apy. When the efficacies of double and triple ther-
apies were equal to that of monotherapy, the cost
effectivenesses of the 3 strategies were monotherapy
$US654 per QALY saved, USPHS PEP $US5855
per QALY saved, and triple therapy $US8969 per
QALY saved. When the efficacies of double and
triple therapy were both 90%, the marginal cost
effectiveness of the USPHS PEPand the triple ther-
apy for all exposed healthcare worker strategies were
$US72 525 per QALY saved and $US767 439 per
Sensitivity Analysis
The parameters with the greatest impact on cost-
effectiveness are shown in figure 2, a tornado dia-
gram that displays a summary of 1-way sensitivity
analyses using the Hellinger[29] cost model. Tor-
nado diagrams allow comparison of a set of 1-way
sensitivity analyses in a single graph. The horizon-
tal bar for each variable represents the range of cost
effectiveness generated by varying the variable. A
wide bar indicates that the variable had a large po-
tential influence on the uncertainty of cost-effec-
tiveness estimates. The graph is called a tornado
diagram because the horizontal bars are arranged
in order with the widest bar at the top.
The probability of seroconversion after percu-
taneous exposure had the greatest effect on cost
effectiveness. At a probability of seroconversion
greater than 0.348%, monotherapy became cost-
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)

HIV Postexposure Prophylaxis
365
QALY saved, respectively. The model was also in-
sensitive to varying the cost of major adverse ef-
fects from $US0 to $US980, varying the probability
of major adverse effects from 0.0001 to 0.001%,
or varying the quality-of-life adjustments for any
of the HIV disease stages.
The results of the multiway probabilistic sensi-
tivity analysis based on sample sizes of 5000 and
500 trials drawn from a Monte Carlo simulation of
10 000 trials are shown in table V. For both sample
sizes, the 95% confidence intervals for cost of the
3 strategies did not overlap. However, for effective-
ness, expressed in QALYs saved per 1000 health-
care workers, USPHS PEP and triple therapy had
considerable overlap when using a sample size of
5000: 95% CIs for USPHS PEP were 34.1 to 34.8
and for triple therapy were 34.2 to 34.9. Using a
sample size of 500, the effectiveness of all 3 strat-
egies overlapped: 95% CIs for monotherapy were
31.4 to 33.5, for USPHS PEP were 33.4 to 35.5 and
for triple therapy were 33.5 to 35.6.
USPHS PEP and triple therapy were $US62 497
and $US564 475 per QALY saved, respectively.
Discussion
In our model, all 3 intervention strategies were
cost effective compared with no intervention. Multi-
way probabilistic sensitivity analysis suggested that
decision-makers could not expect to find any dis-
cernible differences in the effectiveness of triple
therapy, USPHS PEP guidelines and monotherapy
in a small population (500 exposures). In a larger
population (5000 exposures), which would be sim-
ilar to a nationwide implementation, monotherapy
would be less effective than the other 2 alterna-
tives, but USPHS PEP guidelines and triple drug
therapy would be indistinguishable. A randomised
controlled trial is under way to determine the effi-
cacy of 2-drug and selective 3-drug PEP.[18,25] Given
the low transmission rates following exposure of
healthcare workers, demonstration of a clinically
significant difference in the efficacy of monother-
apy versus standard 2-drug with selective 3-drug
PEP will require a large number of participants.
The marginal cost effectiveness of PEP USPHS
guidelines compared with monotherapy falls within
the range of many routinely provided healthcare
interventions.[44] However, because of the low rates
of transmission from patients to healthcare workers,
We estimated the impact on cost effectiveness
of improved survival[32,33] and increased costs as-
sociated with recent innovations in HIV chemo-
therapy. With prolonged HIV stage durations and
increased costs associated with early HIV stage
triple drug therapy when HIV RNA copies/ml
were >10 000, the marginal cost effectivenesses of
Table V. Multiway sensitivity analysis of cost effectiveness of HIV postexposure prophylaxis, based on a Monte Carlo simulation of 10 000
trials. Costs are expressed in 1997 US dollars
Strategy
Costa ($US per 1000 healthcare workers) Effectivenessb (QALYs saved per Cost effectiveness
1000 healthcare workers)
mean
($US/QALY saved)
mean
mean
95% CIc
95% CIc
95% CIc
Sample size of 5000 trials
Monotherapy
USPHS PEP
22 380
179 490
305 240
20 370-24 470
177 190-181 690
302 790-307 630
32.5
34.4
34.6
32.2-32.8
34.1-34.8
34.2-34.9
688
5211
8827
624-750
5126-5293
8715-8940
Triple therapy
Sample size of 500 trials
Monotherapy
USPHS PEP
22 350
179 420
304 990
16 430-29 270
172 950-186 750
298 050-313 090
32.5
34.5
34.6
31.5-33.5
33.4-35.5
33.5-35.6
687
5209
8827
502-892
4980-5490
9508-9185
Triple therapy
a
b
c
Discounted 3% annually.
Compared with no PEP, discounted 3% annually.
Based on 2.5th and 97.5th percentiles of a distribution of 1000 bootstrapped replicants.
CI = confidence interval; PEP = postexposure prophylaxis; QALY = quality-adjusted life-year; USPHS = US Public Health Service.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)

366
Scheid et al.
the additional efficacy of triple drug PEP would be
rewarded by only a small reduction of HIV infec-
tions at great expense. The US healthcare system
would have to spend on average an additional
$US123 for each of 100 000 exposed healthcare
workers to prevent 1 additional case of HIV. If about
10 000 healthcare workers were exposed to HIV
each year, using triple drug therapy for all, instead
of the USPHS recommendations, would cost an ad-
ditional $US1 230 000 per year for 10 years to pre-
vent 1 case of HIV.
The model has several limitations. Only expo-
sure of a healthcare worker to a source for which
the HIV status was known was considered. We did
not have access to adequate information to model
the clinical situation of exposure to a source in which
the HIV status was not known. We would anticipate
that application of the strategy of triple drug ther-
apy for all exposures in which the HIV status was
unknown would yield an extremely unfavourable
marginal cost effectiveness. However, the safeguards
of exposure classification and treatment according
to risk in the USPHS PEP strategy would mitigate
this problem. The model is stable despite consider-
able underlying parameter uncertainty. In fact, the
parameter that has had the greatest scrutiny, the
probability of seroconversion following percutane-
ous exposure, had the greatest impact on model
uncertainty. Because the probability of seroconver-
sion is so small, other parameters that modify this
risk for which little information is available, such
as efficacy of multidrug therapy or partial PEP, have
minimal effect on model uncertainty.
proved survival. Despite this concern, the model
proved to be insensitive to varying quality-of-life
adjustment values for any of the HIV stages.
Although our estimates of cost effectiveness dif-
fered by a factor of 4 from the results of previous
researchers, this is due to different assumptions.
The cost effectiveness of triple therapy in the analysis
of Pinkerton et al.[12] was approximately $US37 000/
QALY saved compared with $US8500/QALY saved
in our analysis. Pinkerton et al.[12] estimated a life-
time cost of HIV medical care of $US98 000 and
7.83 QALYs saved per infection prevented. In our
analysis, the lifetime cost of HIV medical care was
estimated at $US138 288 and 16.1 QALYs saved
per infection prevented. Pinkerton et al.[12] trun-
cated the analysis at age 65 years and used a 5%
annual discount rate for cost and effectiveness. We
used a 3% annual discount rate and did not truncate
the analysis at age 65 years. Adjusting our discount
rate to 5% decreased the lifetime cost of HIV care
to $US96 741 and the QALYs saved per infection
prevented to 10.2. Additionally, Pinkerton et al.[12]
assumed in the base case that triple drug therapy
had an efficacy equivalent to that reported for zi-
dovudine (79%) and that partial PEP was ineffec-
tive, whereas we assumed that partial PEP was par-
tially effective. Using a 5% annual discount rate for
cost and effectiveness, triple drug therapy efficacy
of 79% and ineffective partial PEP in our model
yielded a cost-effectiveness estimate for triple ther-
apy of $US17 492/QALY saved compared with no
PEP therapy.
Ramsey and Nettleman[10] reported the net cost
of prophylactic zidovudine at $US3184 per life year
gained compared with $US713 in our analysis. In
their analysis, the cost of zidovudine prophylaxis
was $US1108 compared with $US420 in our anal-
ysis because of reductions in the cost of zidovudine
from 1992 to 1997. Using the same cost of zidovu-
dine as Ramsey and Nettleman,[10] and not discount-
ing life expectancy, yielded a cost-effectiveness es-
timate for monotherapy of $US2773 per life year
gained in our model.
Our measure of quality of life for HIV stages
was estimated on the basis of a small survey of
healthcare providers.[34] Holtgrave and Pinkerton[35]
recently reviewed the literature regarding HIV and
quality of life. The median estimate of the quality-
of-life adjustment for the AIDS stage from 6 studies
was 0.62, which contrasts with the 0.17 we used.
Although our estimate may be more representative
of the utilities of healthcare workers, the much lower
quality-of-life adjustment values could potentially
decrease the estimate of cost per QALY, particu-
larly in the analysis of the impact of recently im-
It is difficult to compare the results of Allen
et al.,[11] in which zidovudine prophylaxis saved
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)

HIV Postexposure Prophylaxis
367
$Can6 243 955 (1992 Canadian dollars) per infec-
tion prevented, because they incorporated indirect
benefits and the cost of initially treating all expo-
sures for 3 days pending results of HIV testing of
the source. In addition, their estimation of the cost
of medical care for HIV ($Can44 416 per infec-
tion) is much smaller than the estimates used in our
analysis.
With allowances for minor differences in the
models, our findings for the cost effectiveness of
monotherapy and triple therapy are consistent with
these previous HIV PEP cost-effectiveness analy-
ses. Since the cost effectiveness of USPHS PEP
recommendations falls between those of monothe-
rapy and triple drug therapy, this increases our con-
fidence in the accuracy of the estimate.
The treatment of HIV disease is undergoing rapid
and exciting change. Triple drug therapy with zido-
vudine, lamivudine and a protease inhibitor has
demonstrated a more sustained potent antiretrovi-
ral activity and reduced mortality.[45] Life expec-
tancies for those infected with HIV are dramati-
cally improving,[46] but so are costs associated with
these improvements. When our decision model in-
corporated adjustments for extended life expect-
ancy associated with aggressive multidrug therapy,
the marginal cost effectiveness of USPHS PEP com-
pared with zidovudine monotherapy was compara-
ble to that of many medical interventions in com-
mon use.[44]
alence of HIV drug resistance, our model supports
the use of the USPHS PEP guidelines.
References
1. Bell DM. Occupational risk of human immunodeficiency virus
infection in health care workers: an overview. Am J Med 1997;
102: 9-15
2. Centers for Disease Control and Prevention. Update: provision-
al public health service recommendations for chemoprophy-
laxis after occupational exposure to HIV. MMWR Morb Mortal
Wkly Rep 1996; 45: 468-72
3. Centers for Disease Control and Prevention. Public health ser-
vice guidelines for the management of health-care worker
exposures to HIV and recommendations for postexposure pro-
phylaxis. MMWR Morb Mortal Wkly Rep 1998; 47: 1-33
4. Henderson DK, Gerberding JL. Prophylactic zidovudine after
occupational exposure to the human immunodeficiency vi-
rus: an interim analysis. J Infect Dis 1989; 160: 321-7
5. Centers for Disease Control and Prevention. Public health ser-
vice statement on the management of occupational exposure
to human immunodeficiency virus, including considerations
regarding zidovudine postexposure use. MMWR Morb Mor-
tal Wkly Rep 1990; 39: 1-14
6. Puro V, Ippolito G, Guzzanti E, et al. Zidovudine prophylaxis
after accidental exposure to HIV. The Italian study group on
occupational risk of HIV infection. AIDS 1992; 6: 963-9
7. Tokars JI, Marcus R, Culver DH, et al. Surveillance of HIV
infection and zidovudine use among health care workers after
occupational exposure to HIV-infected blood. Ann Intern Med
1993; 118: 913-9
8. Centers for Disease Control and Prevention. Case-control study
of HIV seroconversion in health-care workers after percuta-
neous exposure to HIV-infected blood – France, United King-
dom, and United States, January 1988 - August 1994. MMWR
Morb Mortal Wkly Rep 1995; 44: 929-33
9. Connor EM, Sperling RS, Gelber R, et al. Reduction of maternal-
infant transmission of human immunodeficiency virus type 1
with zidovudine treatment. N Engl J Med 1994; 331: 1173-80
10. Ramsey SD, Nettleman MD. Cost-effectiveness of prophylactic
AZT following needlestick injury in health care workers. Med
Decis Making 1992; 12: 142-8
11. Allen UD, Read S, Gafni A. Zidovudine for chemoprophylaxis
after occupational exposure to HIV-infected blood: an eco-
nomic evaluation. Clin Infect Dis 1992; 14: 822-30
12. Pinkerton SD, Holtgrave DR, Pinkerton HJ. Cost-effectiveness
of chemoprophylaxis after occupational exposure to HIV. Arch
Intern Med 1997; 157: 1972-80
Conclusions
Current USPHS PEPrecommendations are mar-
ginally cost effective compared with monotherapy,
but the additional efficacy of triple drug therapy
for all risk categories is rewarded by only a small
reduction in HIV infections at great expense. Our
analysis was sensitive to the probability of sero-
conversion and to the costs of treatment for HIV
infections, but these did not affect the order of pre-
ferred strategies. For the foreseeable future, assu-
ming innovations in therapy that employ expensive
drug combinations earlier in the HIV disease course
to extend life expectancy and the increasing prev-
13. Scheid DC, Stevens K, Hamm RM. Cost-effectiveness of hu-
man immunodeficiency virus post-exposure prophylaxis. Med
Decis Making 1997; 17: 525
14. Gold MR, Siegel JE, Russel LB, et al. Cost-effectiveness in
health and medicine. New York: Oxford University Press, 1996
15. McCray E. Occupational risk of the acquired immunodefici-
ency syndrome among health care workers. N Engl J Med
1986; 314: 1127-32
16. Ippolito G, Puro V, De Carli G, et al. The risk of occupational
human immunodeficiency virus infection in health care work-
ers. Arch Intern Med 1993; 153: 1451-8
17. Cardo DM, Culver DH, Ciesielski CA, et al. A case-control
study of HIV seroconversion in health care workers after per-
cutaneous exposure. N Engl J Med 1997; 337: 1485-90
18. Gerberding JL. Prophylaxis for occupational exposure to HIV.
Ann Intern Med 1996; 125: 497-501
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)

368
Scheid et al.
19. Henderson DK. Postexposure chemoprophylaxis for occupa-
tional exposure to human immunodeficiency virus type 1:
current status and prospects for the future. Am J Med 1991;
91: 312-9
20. Ippolito G, Puro V, Italian registry of antiviral prophylaxis. Zi-
dovudine toxicity in uninfected healthcare workers. Am J Med
1997; 102: 58-62
EH, Brandeau ML, editors. Modeling the AIDS epidemic:
planning, policy, and prediction. New York (NY): Raven Press,
1994: 153-77
35. Holtgrave DR, Pinkerton SD. Updates of cost of illness and
quality of life estimates for use in economic evaluations of
HIV prevention programs. J Acquir Immune Defic Syndr Hum
Retrovirol 1996; 12: 413-20
36. Consumer Price Index – all urban consumers, medical services.
Series ID: CUUR0000SAM. Washington, DC: US Bureau of
Labor Statistics [online]. Available from: URL: http://146.
142.4.24/cgi-bin/surveymost?cu [Accessed 2000 Aug 14]
37. Beck JR, Pauker SG. The Markov process in medical prognosis.
Med Decis Making 1983; 3: 419-58
38. Sonnenberg FA, Beck JR. Markov models in medical decision
making: a practical guide. Med Decis Making 1993; 13: 322-38
39. Fryback DG, Dasbach EJ, Klein R, et al. The Beaver Dam Health
Outcomes Study: initial catalog of health-state quality factors.
Med Decis Making 1993; 13: 89-102
40. National Center for Health Statistics. U.S. decennial life tables
for 1989-1991. Hyattsville (MD): National Center for Health
Statistics, 1997
41. Doubilet P, Begg CB, Weinstein MC, et al. Probabilistic sensi-
tivity analysis using Monte Carlo simulation: a practical ap-
proach. Med Decis Making 1985; 5: 157-77
42. Hunink MGM, Bult JR, De Vries J, et al. Uncertainty in decision
models analyzing cost-effectiveness: the joint distribution of
incremental costs and effectiveness evaluated with a nonpar-
ametric bootstrap method. Med Decis Making 1998; 18: 337-48
43. Efron B, Tibshirani RJ. An introduction to the bootstrap. New
York (NY): Chapman and Hall, 1993
44. Laupacis A, Feeney D, Detsey AS, et al. How attractive does a
new technology have to be to warrant adoption and utiliza-
tion? Tentative guidelines for using clinical and economic
evaluations. Can Med Assoc J 1992; 146: 473-80
21. Wainberg MA, Friedland G. Public health implications of anti-
retroviral therapy and HIV drug resistance. JAMA1998; 279:
1977-83
22. Hirsch MS, Conway B, D’Aquilla RT, International AIDS So-
ciety – USAPanel, et al. Antiretroviral drug resistance testing
in adults with HIV infection. JAMA 1988; 279: 1984-91
23. Mayers DL. Prevalence and incidence of resistance to zidovud-
ine and other antiretroviral drugs. Am J Med 1997; 102: 70-5
24. Frankel LM, Wagner II LE, Demeter LM, et al. Effects of zido-
vudine use during pregnancy on resistance and vertical trans-
mission of human immunodeficiency virus type I. Clin Infect
Dis 1995; 20: 1321-6
25. Gerberding JL. Postexposure prophylaxis for human immuno-
deficiency virus at San Francisco General Hospital. Am J Med
1997; 102: 85-9
26. Fahrner R, Meekman SE, Koziol DE, et al. Safety of zidovudine
(ZDV) administered as post-exposure prophylaxis to health
care workers (HCW) sustaining HIV-related occupational ex-
posures (OE). In: Program and abstracts – Interscience Con-
ference on Antimicrobial Agents and Chemotherapy: 1994
Oct 4-7; Orlando (FL). Washington, DC: American Society
for Microbiology, 1994: 133
27. Locality fee schedule report. Oklahoma City (OK): Health Care
Financing Administration, 1997
28. Average wholesale price. Indianapolis (IN): Medispan Inc., 1997
29. Hellinger FJ. The lifetime cost of treating a person with HIV.
JAMA 1993; 270: 474-8
45. Hammer SM, Squires KE, Hughes MD, et al. A controlled trial
of two nucleoside analogues plus indinavir in persons with
human immunodeficiency virus infection and CD4 cell counts
of 200 per cubic millimeter or less. N Engl J Med 1997; 337:
725-33
46. Palella FJ, Delaney KM, Moorman AC, et al. Declining mor-
bidity and mortality among patients with advanced human
immunodeficiency virus infection. N Engl J Med 1998; 338:
853-60
30. Gable CB, Tierce JC, Simison D, et al. Cost of HIV+/AIDS at
CD4+ counts disease stages based on treatment protocols. J
Acquir Immune Defic Syndr Hum Retrovirol 1996; 12: 413-20
31. Hurley SF, Kaldor JM, Gardiner S, et al. Lifetime cost of human
immunodeficiency virus-related health care. J Acquir Immune
Defic Syndr Hum Retrovirol 1996; 12: 371-8
32. Longini IMJ, Clark WS, Karon JM. Effect of routine use of
therapy in slowing the course of human immunodeficiency
virus (HIV) infection in a population-based cohort. Am J Epi-
demiol 1993; 137: 1229-40
33. CAESAR Group. Randomized trial of addition of lamivudine
or lamivudine plus loviride to zidovudine-containing regim-
ens for patients with HIV-1 infection: the CAESAR trial. Lan-
cet 1997; 349: 1413-4
34. Owens DK, Nease RFJ. Transmission of HIV infection between
provider and patient: quantitative analysis of risk. In: Kaplan
Correspondence and offprints:Dr Dewey Scheid, Family Me-
dicine Center, 900Northeast10thStreet, OklahomaCity, OK
73104, USA.
E-mail: dewey-scheid@ouhsc.edu
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2000 Oct; 18 (4)