The global impact of scaling up HIV/AIDS prevention programs in low- and middle-income countries.

Abstract

A strong, global commitment to expanded prevention programs targeted at sexual transmission and transmission among injecting drug users, started now, could avert 28 million new HIV infections between 2005 and 2015. This figure is more than half of the new infections that might otherwise occur during that period in 125 low- and middle-income countries. Although preventing these new infections would require investing about U.S.$122 billion over this period, it would reduce future needs for treatment and care. Our analysis suggests that it will cost about U.S.$3900 to prevent each new infection, but that this will produce a savings of U.S.$4700 in forgone treatment and care costs. Thus, greater spending on prevention now would not only prevent more than half the new infections that would occur from 2005 to 2015 but would actually produce a net financial saving as future costs for treatment and care are averted.

Full text

Page 1

ginally stable or folding-defective proteins in the
genetic background of conformational diseases
as potent extrinsic factors that modify aggrega-
tion and toxicity. Given the prevalence of poly-
morphisms in the human genome (30), they
could contribute to variability of disease onset
and progression (31). This interpretation also
provides a mechanistic basis to the notion that
the late onset of protein misfolding diseases may
be due to gradual accumulation of damaged pro-
teins (32), resulting in a compromise in folding
capacity. Indeed, in a screen for regulators of
polyQ aggregation in C. elegans, we identified
nearly 200 genes whose diverse functions have
the potential to affect protein homeostasis (21).
Cellular degeneration in diseases of protein
conformation is unlikely to be due to a single
defect. Thus, the many toxic effects on various
cellular processes attributed to misfolded pro-
teins (6–15) could in fact be an integral part of
the global disruption of protein homeostasis
identified in this work.
References and Notes
1. S. W. Davies et al., Cell 90, 537 (1997).
2. M. DiFiglia et al., Science 277, 1990 (1997).
3. R. Kayed et al., Science 300, 486 (2003).
4. M. Tanaka, P. Chien, N. Naber, R. Cooke, J. S. Weissman,
Nature 428, 323 (2004).
5. J. F. Gusella, M. E. MacDonald, Nat. Rev. Neurosci. 1, 109
(2000).
6. M. K. Perez et al., J. Cell Biol. 143, 1457 (1998).
7. A. McCampbell et al., Hum. Mol. Genet. 9, 2197 (2000).
8. P. J. Muchowski et al., Proc. Natl. Acad. Sci. U.S.A. 97,
7841 (2000).
9. S. Kim, E. A. A. Nollen, K. Kitagawa, V. P. Bindokas,
R. I. Morimoto, Nat. Cell Biol. 4, 826 (2002).
10. K. Ii, H. Ito, K. Tanaka, A. Hirano, J. Neuropathol. Exp.
Neurol. 56, 125 (1997).
11. C. J. Cummings et al., Nat. Genet. 19, 148 (1998).
12. N. F. Bence, R. M. Sampat, R. R. Kopito, Science 292,
1552 (2001).
13. C. I. Holmberg, K. E. Staniszewski, K. N. Mensah,
A. Matouschek, R. I. Morimoto, EMBO J. 23, 4307 (2004).
14. M. Gu et al., Ann. Neurol. 39, 385 (1996).
15. V. L. Gabai, A. B. Meriin, J. A. Yaglom, V. Z. Volloch,
M. Y. Sherman, FEBS Lett. 438, 1 (1998).
16. S. L. Rutherford, S. Lindquist, Nature 396, 336 (1998).
17. T. K. Van Dyk, A. A. Gatenby, R. A. LaRossa, Nature 342,
451 (1989).
18. C. R. Brown, L. Q. Hong-Brown,W. J. Welch, J. Clin. Invest.
99, 1432 (1997).
19. C. B. Pedersen et al., J. Biol. Chem. 278, 47449 (2003).
20. J. F. Morley, H. R. Brignull, J. J. Weyers, R. I. Morimoto,
Proc. Natl. Acad. Sci. U.S.A. 99, 10417 (2002).
21. E. A. Nollen et al., Proc. Natl. Acad. Sci. U.S.A. 101,
6403 (2004).
22. H. R. Brignull, S. Tang, R. I. Morimoto, in preparation.
23. K. Gengyo-Ando, H. Kagawa, J. Mol. Biol. 219, 429 (1991).
24. Gidalevitz et al., unpublished data.
25. S. G. Clark, D. L. Shurland, E. M. Meyerowitz,
C. I. Bargmann, A. M. van der Bliek, Proc. Natl. Acad. Sci.
U.S.A. 94, 10438 (1997).
26. Materials and methods are available as supporting
material on Science Online.
27. D. M. Eisenmann, S. K. Kim, Genetics 146, 553 (1997).
28. S. A. Goff, A. L. Goldberg, Cell 41, 587 (1985).
29. R. I. Morimoto, Genes Dev. 12, 3788 (1998).
30. International SNP Map Working Group, Nature 409, 928
(2001).
31. U.S.-Venezuela Collaborative Research Project,
N. S. Wexler, Proc. Natl. Acad. Sci. U.S.A. 101, 3498
(2004).
32. C. N. Oliver, B. W. Ahn, E. J. Moerman, S. Goldstein,
E. R. Stadtman, J. Biol. Chem. 262, 5488 (1987).
33. T.G. was supported by NIH Training Grant T32 HL076139
and by an Individual NRSA F32 GM075583-01; A.B.-Z.
was supported by an EMBO Long-Term Fellowship,
the Parkinson Foundation, and the Hereditary Disease
Foundation; and H.R.B. was supported by National
Institute of General Medical Sciences (NIGMS) Molecular
Biology of Disease Training Grant T32 GM08061.
R.I.M. was supported by grants from the NIH (NIGMS,
National Institute of Neurological Disorders and
Stroke, and National Institute on Aging), the
Huntington Disease Society of America Coalition for
the Cure, the ALS Association, and the Daniel F. and
Ada L. Rice Foundation. Some nematode strains used in
this work were provided by the Caenorhabditis Genetics
Center, which is funded by the NIH National Center
for Research Resources (NCRR). We thank J. West and
members of the Morimoto laboratory for their
discussion and comments on the manuscript and
F. Moore for reagents.
Supporting Online Material
www.sciencemag.org/cgi/content/full/1124514/DC1
Materials and Methods
Fig. S1
References
3 January 2006; accepted 16 January 2006
Published online 9 February 2006;
10.1126/science.1124514
Include this information when citing this paper.
The Global Impact of Scaling Up
HIV/AIDS Prevention Programs in
Low- and Middle-Income Countries
John Stover,1Stefano Bertozzi,2Juan-Pablo Gutierrez,2Neff Walker,3Karen A. Stanecki,4
Robert Greener,4Eleanor Gouws,4Catherine Hankins,4Geoff P. Garnett,5Joshua A. Salomon,6
J. Ties Boerma,7Paul De Lay,4Peter D. Ghys4*
A strong, global commitment to expanded prevention programs targeted at sexual transmission
and transmission among injecting drug users, started now, could avert 28 million new HIV
infections between 2005 and 2015. This figure is more than half of the new infections that
might otherwise occur during that period in 125 low- and middle-income countries. Although
preventing these new infections would require investing about U.S.$122 billion over this period,
it would reduce future needs for treatment and care. Our analysis suggests that it will cost about
U.S.$3900 to prevent each new infection, but that this will produce a savings of U.S.$4700 in
forgone treatment and care costs. Thus, greater spending on prevention now would not only
prevent more than half the new infections that would occur from 2005 to 2015 but would
actually produce a net financial saving as future costs for treatment and care are averted.
M
and care in the developing world was limited
by costs, by the complexity of early treatment
uch has changed in the global re-
sponse to the AIDS epidemic since
the late 1990s. Access to treatment
regimens, and by a perceived lack of capacity
to implement treatment programs even if drug
costs were greatly reduced. The pioneering
work of Brazil in providing broad access to
antiretroviral therapy proved that, with political
Fig. 3. Progressive disruption of cellular folding capacity by misfolded proteins. (A to D) Fluorescent
images of representative L2 larvae at the permissive temperature of heterozygous (A) or homozygous
(B) Q40m, ras(ts)þQ40m (C), and paramyosin(ts)þQ40m (D) strains. (E) Number of visible aggregates
in L2 larvae expressing indicated proteins. ras(ts)þQ40m in (C) and (E) denotes the fluorescent progeny
of an F2 ras(ts) animal expressing Q40m; these progeny could be either homozygous or heterozygous
for Q40m.
REPORTS
10 MARCH 2006 VOL 311SCIENCE www.sciencemag.org
1474

Page 2

will, middle-income countries could overcome
these limitations (1). In 2001 when govern-
ments convened a special session of the United
Nations General Assembly, they adopted a new
set of commitments on AIDS that included
access to effective care and treatment. In prep-
aration for the special session, estimates were
made of the resources required for an expanded
response to the epidemic, including costs for
core prevention and care and treatment activ-
ities (2). The impact of these interventions, at
the anticipated costs, on the number of new
adult infections was estimated subsequently (3).
In response, significant financial resources have
been mobilized by both donors and the affected
governments and communities (4). A recent
analysis has explored the interactions of pre-
vention and treatment and care in sub-Saharan
Africa (5), and an update and extension of the
UNAIDS resource needs estimates was com-
pleted in 2005 (4). Increases in funding have
produced expansions in prevention services
and treatment and care services, but much of
the recent advocacy and press attention has
focused specifically on the need for antiret-
roviral therapy. Others have argued that, in the
absence of curative therapy, effective preven-
tion is the best way to prevent the premature
death of the millions of people being infected
with HIV each year (6). Here, we estimate the
global net cost of prevention activities in the
context of the new global commitment to
provide care and treatment for adults and
children.
We estimate the costs of an expanded pre-
vention program, including 15 specific inter-
ventions, in 125 low- and middle-income
countries, calculate the impact in terms of
infections averted, and estimate the gross and
netcostperinfection averted during2005–2015.
Briefly, the total resource needs and cor-
responding targets for scaling up a range of
prevention interventions targeting sexual trans-
mission and transmission among injecting drug
users are based on the resource needs estimates
prepared by the Joint United Nations Pro-
gramme on HIV/AIDS (UNAIDS) during 2005
(4). To calculate averted infections, we first
project HIV prevalence in each country under
current prevention efforts based on country-
specific models used by UNAIDS and the
World Health Organization (WHO) to develop
estimates for 2005. Treatment scale-up is added
to these baseline scenarios, rising from current
levels to 80% coverage of those in need for
antiretroviral treatment (ART) for adults and
children and for cotrimoxazole prophylaxis for
children, as well as to 80% of pregnant women
attending antenatal clinics for prevention of
mother-to-child transmission (PMTCT) inter-
ventions in all regions by 2010, and with this
coverage remaining at 80% through 2015. To
estimate the impact of prevention programs, the
changes in behavior resulting from exposure
to prevention interventions are estimated
using impact data from intervention studies,
the impact of these behavior changes on in-
cidence is modeled using two different simu-
lation models, and the consequences of the
incidence changes are projected using the
Spectrum software package. The average cost
per case averted is calculated by dividing the
total prevention costs (expressed in constant
2004 prices) over the period 2005–2015 by
the number of cases averted over the same
period.
The forgone treatment and care costs per
averted infection are then calculated. The
costs of averted treatment and care are con-
sistent with the 2005 UNAIDS resource needs
estimates (4). Estimated costs of care and
treatment are based on these estimates (a
detailed description of the methodology and
tools is provided in the supporting online
material). Lifetime costs for each AIDS case
are calculated on the assumption that the
median number of years of ART treatment is
7.5 years in all countries. The net present
value of the expected lifetime treatment costs
is calculated for the year in which the in-
fection is prevented, on the basis of a discount
rate of 5% and an average of 7 years from
infection to onset of ART treatment. Sensitiv-
ity analyses are conducted for both the cost of
averting an infection (by lowering the effec-
tiveness of interventions) and the lifetime cost
of treatment and care of these infections (by
modifying the average survival to 5 and 10
years, respectively).
At current levels of implementation of
prevention programs, the annual number of
global infections is expected to increase from
4.8 million in 2005 to 5.9 million in 2015. Over
this period, the total number of new infections is
estimated to be 62.3 million, 7.9 million in
children and 54.4 million in adults (fig. S1).
An estimated 31.1 million new infections (or
50%) between 2005 and 2015 would be averted
by implementing the comprehensive prevention
package examined here, 3.1 million in children
and 28 million in adults. The number of averted
infections would vary across regions, with 19.5
million in sub-Saharan Africa, 8.9 million in
Asia, 0.7 million in North Africa and the Middle
East, 0.7 in Eastern Europe, and 1.3 million in
Latin America and the Caribbean (Table 1). The
total cost of implementing this package during
2005–2015 would be U.S.$122 billion (see
fig. S4 for the distribution of resources by
region). UNAIDS has projected that about
U.S.$27 billion is already programmed or
promised for the period 2005 to 2007. About
one-third of this amount is expected to come
from local governments and out-of-pocket ex-
penditures, with most of the rest from inter-
1Futures Group/Constella, Glastonbury, CT 06033, USA.
2National Institute of Public Health (INSP), Cuernavaca
62508, and Centro de Investigacio ´n y Docencia Econo ´m-
icas (CIDE), Mexico City 01210, Mexico.3United Nations
Children’s Fund (UNICEF), New York, NY 10017, USA.
4Joint United Nations Programme on HIV/AIDS (UNAIDS),
Geneva CH-1211, Switzerland.5Imperial College, London
SW7 2AZ, UK.
USA.
Switzerland.
6Harvard University, Boston, MA 02138,
7World Health Organization, 1211 Geneva 27,
*To whom correspondence should be addressed. E-mail:
ghysp@unaids.org
Table 1. Number of new infections and infections averted during 2005–2015 by applying the full
prevention and treatment and care package, by region.
103number of infections
103Infections averted
Current effortExpanded prevention
Adults Children Adults ChildrenAdultsChildren
Sub-Saharan Africa
Asia
North Africa/Middle East
Eastern Europe
Latin America and Caribbean
Total
31,726
16,637
1,160
2,510
2,397
54,430
6,889
688
114
47
144
7,882
15,069
8,237
475
1,807
861
26,449
4,181
383
68
32
65
4,729
16,657
8,400
685
703
1,536
27,981
2,708
305
46
15
79
3,153
Table 2. Costs for prevention and averted future treatment and care, 2005–2015, by region.
Cost for prevention
per infection
averted (U.S.$)
Lifetime treatment
cost (net present
value U.S.$)
Savings per
infection averted
(U.S.$)
Sub-Saharan Africa
Asia
North Africa/Middle East
Eastern Europe
Latin America and Caribbean
Global
2,109
7,417
2,756
9,148
5,045
3,923
3,469
5,602
6,822
11,203
12,330
4,707
1,360
–1,815
4,066
2,055
7,285
784
REPORTS
www.sciencemag.orgSCIENCE VOL 31110 MARCH 2006
1475

Page 3

national donors through bilateral and multi-
lateral mechanisms (4).
In addition to the health benefits described
above, the implementation of a comprehensive
global prevention approach would significantly
reduce the number of people requiring antiret-
roviral therapy in the future. The weighted
global average (considering country-specific
costs, intervention coverage, treatment package,
and epidemiology) is a cost for prevention ac-
tivities of U.S.$3900 per infection averted
(adults and children combined), which avoids
an expenditure for treatment and care with a net
presentvalueofU.S.$4700andresultsinsavings
of U.S.$780 per infection averted. The costs by
region are presented in Table 2. Sensitivity
analyses assuming lower prevention effective-
ness indicate that there continue to be net
savings in all regions except Asia and Eastern
Europe (table S7). Sensitivity analyses as-
suming a life expectancy of 5 years on treat-
ment indicate that there would still be net
savings in all regions except in Asia and
Eastern Europe. If one assumes a life expect-
ancy of 10 years after treatment initiation, an
even larger savings would accrue in all regions
compared with the baseline estimates.
Compared with estimates of adult infections
thatwouldhavebeenavertedinadultsfor2001–
2010 if implementation had ramped up in 2001
(3), the current analyses indicate that an
important window of opportunity was missed
by not adequately scaling up prevention ser-
vices in 2001–2005. Indeed, the latest evidence
regarding coverage of prevention services
shows only limited implementation of pre-
vention services by 2003 (7). Previous analy-
ses have shown that in sub-Saharan Africa,
the number of people in need of treatment in
future years could be drastically reduced if
comprehensive prevention programs were im-
plemented (5). Our estimates suggest that
scale-up of prevention programs would not
only be highly cost effective, but would even
be cost saving in most regions. Because of the
commitment to provision of universal access to
antiretroviral care, averting future treatment
needs could free resources to prevent more
infections. The unit costs of prevention and
treatment vary significantly by region, as do the
dynamics of the epidemic, which leads to
different net costs for different regions. Even
in those regions where net costs are positive,
the cost per disability-adjusted life year is well
below average gross domestic product (GDP)
per capita; highly cost effective by almost any
standard (8).
Estimates of cost of prevention interventions
per infection averted in this analysis at about
$3900 are considerably higher than estimates in
the earlier analysis at around $1000 (3). Part of
the explanation is that time has been lost by not
scaling up as called for in the United Nations
General Assembly Special Session (UNGASS)
Declaration of Commitment on HIV/AIDS,
while the number of new infections was grow-
ing rapidly between 2001 and 2010. That
growth is predicted to be less steep in 2005–
2015. Also, baseline estimates are lower than
before because of newly available surveillance
data. In addition, estimates of the unit costs of
prevention are higher than they were in 2001 as
a result of extensive consultations with country
experts. There may be more cost-effective
combinations of preventive interventions than
the ones considered here, especially those
interventions that are targeted at the major
modes of transmission (9). Although the in-
terventions and targets defined in the 2005
global resource needs estimation exercise used
in our analyses are ambitious, individual coun-
tries are encouraged to set targets that are
consistent with past scale-up rates, taking into
account current infrastructure, capacity, and re-
sources, as well as foreseeable financing. New
preventive technologies may become available
in 2005–2015, including male circumcision
(10), microbicides, or vaccines. These would
lead to additional reductions in the number of
new infections.
Our estimates of future cost savings are
conservative, given that they do not include
savings in programming and infrastructure and
human capacity costs, related to the reduced
treatment needs. They also exclude savings in
lost productivity due to illness and premature
death,alongwithcostsoforphanhood,including
orphan support costs and reduced investment in
their human capital. Finally, they also do not
include an increase in prevention effectiveness
as treatment programs scale-up (5).
As in any modeling exercise, there are un-
certainties around various model inputs. The
greatest degree of uncertainty in the estimates
reported here regards the likely effectiveness of
the array of prevention interventions when
implemented at scale. These uncertainties are
addressedonlypartiallythroughsensitivityanal-
ysis.Forveryfewinterventions,suchasPMTCT,
do reliable data exist on prevention effectiveness
at full scale. For most of the interventions, we
made use of the limited data available with
highlyvariableestimatesofeffectiveness—some
of which include zero. Currently available data
do not indicate whether there will be economies
of scale and scope, or diseconomies as inter-
ventions are extended to populations that are
more difficult to reach.
The scale-up rates were set by a Policy
Steering Committee organized by UNAIDS and
composedofrepresentativesofcivilsociety,UN
organizations, major donors, and technical ex-
perts (4). As with all international goals, these
are intended to be ambitious but achievable. A
slower scale-up of prevention would provide
fewer total benefits, although the cost per
infection averted would be about the same. A
slower scale-up in treatment coverage would
postpone fewer deaths and reduce the savings
derived from prevention.
Given the magnitude of the problem and the
high level of uncertainly regarding comparative
effectiveness of interventions, it is imperative
that roll-out of large-scale prevention programs
incorporate rigorous prospective evaluations of
their effectiveness. Further limitations are that
the model does not estimate the effects of
expansion of reproductive health services for
women that would reduce fertility. However,
previous models have shown synergistic effects
when prevention of pregnancy in HIV-positive
women is introduced (11).
Even with these limitations, the insights
from the current analyses are important to
inform the necessary long-term planning of
investments for health and development in
low- and middle income countries. Resources
fortreatmentandcarewilllargelybefundedby
the same sources as resources for prevention
programs, i.e., mostly by national health min-
istries and international donors. Our analyses
suggest that both national governments and
donor countries would be well advised to en-
sure that prevention programs are scaled up as
soon as possible, because early investment in
prevention will both prevent a greater propor-
tion of future infections andreduce future costs
for treatment and care by more than the cost of
the prevention programs.
References and Notes
1. J. R. P. Marins et al., AIDS 17, 1675 (2003).
2. B. Schwartla ¨nder et al., Science 292, 2434 (2001).
3. J. Stover et al., Lancet 360, 73 (2002).
4. UNAIDS, ‘‘Resource needs for an expanded response to
AIDS in low- and middle-income countries’’ (UNAIDS,
Geneva, August 2005).
5. J. A. Salomon et al., PLoS Med. 2, e16 (2005).
6. E. Marseille, P. B. Hofmann, J. G. Kahn, Lancet 359, 1851
(2002).
7. USAID, UNAIDS, WHO, UNICEF, POLICY Project,
‘‘Coverage of selected services for HIV/AIDS prevention,
care, and support in low and middle income countries
in 2003’’ (POLICY Project, Washington, DC, June 2004).
8. S. Bertozzi et al., in Disease Control Priorities in Developing
Countries (Oxford Univ. Press, ed. 2, Oxford, in press).
9. E. Pisani et al., BMJ 326, 1384 (2003).
10. B. Auvert et al., PLoS Med. 2, e298 (2005).
11. J. Stover, N. Fuchs, D. Halperin, A. Gibbons, D. Gillespie,
‘‘Adding family planning to PMTCT sites increases
the benefits of PMTCT’’ (USAID Issue Brief, Population and
Reproductive Health, Agency for International Develop-
ment, Washington, DC, October 2003); (www.info.usaid.
gov/our_work/global_health/pop/publications/docs/
familypmtct.html).
12. We thank D. Evans, T. Adams, T. Tan-Torres, B. Johns, and
P. Zurn for their contributions to the resource needs
model. We also thank the many individuals and
organizations that have supported the 2005 UNAIDS
resource needs estimation exercise.
Supporting Online Material
www.sciencemag.org/cgi/content/full/1121176/DC1
Materials and Methods
Figs. S1 to S4
Tables S1 to S7
References
11 October 2005; accepted 20 January 2006
Published online 2 February 2006;
10.1126/science.1121176
Include this information when citing this paper.
REPORTS
10 MARCH 2006 VOL 311SCIENCE www.sciencemag.org
1476

Page 4

Revised 10 March 2006, for details, see page 7.


www.sciencemag.org/cgi/content/full/1121176/DC1





Supporting Online Material for
The Global Impact of Scaling-Up HIV/AIDS Prevention Programs in Low-
and Middle-Income Countries
John Stover, Stefano Bertozzi, Juan-Pablo Gutierrez, Neff Walker, Karen A. Stanecki,
Robert Greener, Eleanor Gouws, Catherine Hankins, Geoff P. Garnett, Joshua A
Salomon, J. Ties Boerma, Paul De Lay, Peter D. Ghys*

*To whom correspondence should be addressed. E-mail: ghysp@unaids.org

Published 2 February 2006 on Science Express
DOI: 10.1126/science.1121176

This PDF file includes

Materials and Methods
Figs. S1 to S4
Tables S1 to S7
References

Page 5

Science Supporting Online Material
Stover, p. 1
Science Supporting Online Material

1 Materials and Methods
1.1 The Costs and Impact of Comprehensive Prevention
1.1.1 Cost of Implementing Comprehensive Prevention
The total resource needs for prevention are based on the estimates prepared during
2005 by UNAIDS (S1). These were calculated from country-specific estimations,
grouped into three basic epidemic types, i.e., low-level, concentrated and generalized
(S2). The most recent evidence of scale-up rates (S3) (Tables S1a and S1b) implies
that by 2010 all countries could reach the expected coverage rates for a package of
interventions, depending on the type of epidemic that prevails. The coverage rates by
2010 for each type of epidemic are presented in Table S2. For our analyses we have
adopted the package of interventions and the targets as defined in the Resource Needs
estimates work, since these had been adopted by a steering committee with broad
representation of major institutions and civil society involved in the fight against
HIV/AIDS. Global estimates were built from country level estimates that are based on
the type of epidemic, their population sizes and local costs. The allocation between
interventions differs in each country. Unit costs for prevention programs were further
informed by experts from 78 countries that participated in regional workshops. As the
targets for the Resource Needs work were only defined up to 2010, we have extended
the calculation of resource needs through 2015, by assuming that the coverage of
prevention services remains constant from 2010 to 2015 at the 2010 levels. The costs
for blood safety, universal precautions, and safe injections were not included in the
calculation of resource needs for prevention for the current analysis, as this analysis is
focused only on sexual and IDU transmission of HIV, as these are the main global
modes of transmission (S4). Scaling up to the above coverage levels in a linear way
would require US$21 billion during 2006–2008 (S1). The analysis assumes constant
marginal costs during future scale-up. Although there may be economies of scale or
scope during this process, there may also be diseconomies as interventions are
extended to populations more difficult to reach. It was not felt that there is sufficient
data at present to support an assumption of varying marginal costs during the
projection.
These costs do not include the program costs for management, administration,
monitoring and evaluation (M&E), research, advocacy, and training that add about
7% to the total costs of prevention and treatment. They were excluded because they
are not purely prevention costs, but support the entire program of prevention,
treatment, care, and orphan support. Including them would raise the costs of
prevention by about 5%.
1.1.2 Calculation of Averted Infections
a) HIV projections
Projected HIV prevalence curves were based on current country-specific models (S5)
and assumptions about future trends in prevalence in the 125 low- and middle-income
countries with a population of more than 1 million.

Page 6

Science Supporting Online Material
Stover, p. 2
For countries with generalized epidemics (mostly in sub-Saharan Africa), current
epidemic trends as produced by the Estimates and Projection Package (EPP) (S6)
were projected through 2015. The model variables were estimated from sentinel site
prevalence data. In most cases the epidemics remained stable.
For countries with epidemics concentrated in groups with higher-risk behaviors—
e.g., men who have sex with men (MSMs), injecting drug users (IDUs), sex workers
(SWs) and their clients— the Workbook method was used to produce prevalence
estimates for 2004 (S7). Projections of these epidemics to 2015 were based on
assumptions about the saturation prevalence for each of the groups at high risk, the
time to saturation prevalence, and spread from groups with higher risk behavior to
populations at lower risk over time. For Eastern Europe, Asia, and North Africa and
the Middle East, the workbook projection spreadsheets used in the 2001 analysis by
Stover et al. (S8) were updated for 2004 using rules for HIV saturation levels among
various subpopulations and year of saturation. These saturation levels included: IDUs
at 25% by 2010, SWs at 15% by 2015, MSMs at 20% by 2015, and clients of sex
workers at 7.5% by 2020. The exception was the saturation levels for IDUs in Russia
and Ukraine and for some of the Caucasian countries where saturation levels were
assumed to reach 50%. In addition we used a ratio of low to high risk of .1 in 2005,
.15 in 2010, .2 in 2020 and .3 in 2030 for all countries with low-level or concentrated
epidemics. However, for some countries current evidence required some adjustment
to this ratio. For Latin America and the Caribbean, epidemic curves developed for the
2004 round of estimates in Spectrum were used assuming stable prevalence into the
future.
Spectrum software was used to calculate the number of new infections in adults and
children (Fig. S1) (S9). Baseline adult antiretroviral treatment (ART) coverage
estimates started from the coverage estimates published by WHO/UNAIDS in
December 2004 (S10) with 5%, 9%, 5–10%, and 65% coverage in Africa, Asia,
Eastern Europe, and Latin America and the Caribbean, respectively. For coverage of
child ART and co-trimoxazole (CTX), regional starting points based on the 2003
coverage survey (S3) were used unless country-specific data were available from the
survey; 0% for Africa and Asia, 3% for Latin America and the Caribbean (LAC), and
32% for Eastern Europe. For prevention of mother-to-child transmission (PMTCT)
programs, regional coverage levels were based on the 2003 coverage survey unless
country-specific values were available (S3). The regional starting points for PMTCT
in 2003 were 5% for Africa and Asia, 34% for LAC, and 37% for Eastern Europe.
b) Treatment scale-up
The Spectrum model (S9) was used to apply the treatment scale-up to 125 countries
and each of the treatment scenarios to project the impact on adult (age 15 to 49 years)
and child (age 0–14 years) epidemic trends over time, in terms of the total number of
adults in need of ART, adult incidence, adult prevalence and mortality, and child
treatment needs. ART and CTX treatment are assumed to reach 80% coverage of
those in need and PMTCT programs, 80% of pregnant women attending antenatal
clinics by 2010 in all regions. As provision of ART extends life, it will increase
prevalence for a constant incidence. Current ART treatment was assumed to have a
minimal effect on current prevalence in countries in Africa, Asia, and Eastern Europe
since ART coverage is low in these regions, whereas in countries in Latin America
and the Caribbean, ART treatment was assumed to have a significant effect on current
prevalence.

Page 7

Science Supporting Online Material
Stover, p. 3
Early detection of HIV infection among children based on PCR was assumed to be
available only in countries in Eastern Europe and not in the other regions. The
assumed PMTCT treatment protocol was Nevirapine and AZT with exclusive
breastfeeding. Regional medians of duration of breastfeeding were applied.
c) Estimating the number of infections and deaths
averted
Using the above described projections as baseline, the effect of prevention and
treatment programs on the number of new infections and deaths among adults and
children was estimated. To estimate the impact of prevention programs, we first
estimated the changes in behavior resulting from exposure to prevention
interventions, modeled the impact of these behavior changes on incidence, and then
projected the consequences of the incidence changes. To estimate behavior change,
we used an impact matrix (Table S3) that describes the effect of exposure to each of
the interventions listed in Table S2 on sexual behavior (condom use, number of
partners, age at first sex), sexually transmitted infection (STI) treatment–seeking
behavior and injecting drug use (prevalence of drug injection, needle sharing,
injection frequency, and number of sharing partners) (S11). The impact matrix has
been constructed from a detailed literature review first conducted in 2002 and updated
periodically since then. It now contains results from 230 impact studies (S12).
Different values describe the impacts of prevention on those engaging in commercial
sex, casual sex, marital sex, sex between men, and injecting drug use. For each cell in
the impact matrix, describing the effect of a particular intervention on a behavior in a
particular risk group, the impact is estimated as the mean result of the available
studies. To explore the sensitivity of the results to the uncertainty in the impact
estimates, an alternative version of the impact matrix was prepared that uses the lower
quartile of the study results instead of the mean. The resulting behavior patterns were
used in models of sexual transmission and injecting drug transmission, as described
previously, to determine the impact on adult incidence (S8).
In addition, for countries where transmission among IDUs is dominant, a separate
model of the transmission dynamics of HIV in a population divided into risk groups
was constructed. A model of the transmission dynamics of HIV in a population
divided into risk groups was compared with the results of projections based on
epidemiological data for sample countries. The model divides the population into
homosexual MSMs; bisexual MSMs; male IDUs; male clients of sex workers; other
sexually active men; women IDUs; female sex workers (FSWs); IDU FSWs; and
other sexually active women. The potential transmission routes for infection between
these groups are illustrated in Fig. S2. In calculating the incidence of HIV, the number
of sexual, same and different sex, partners, sex acts per partner, presence of another
STI, and condom use are included along with rate of sharing injection equipment,
both the numbers of individuals sharing with and number of times, and use of bleach.
These proximate determinants of risk can subsequently be altered by interventions.
The risk of infection is then dependent on the prevalence of infection and
infectiousness within the pool of contacts. The population is divided into those
susceptible; and infected with HIV, with four stages representing early infection with
a high transmission probability; the long low transmission probability incubation
period; a period prior to AIDS; and AIDS itself. In addition, the model allows both
successful treatment and treatment failure, but it is assumed this has little short-term
impact. These categories are all further separated according to whether they have been
included in prevention interventions.

Page 8

Science Supporting Online Material
Stover, p. 4
Interventions are linearly scaled up from 2005 to 2010 until coverage of 80% is
achieved and maintained. It is assumed that partner numbers are decreased by 25%
and condom use is increased by 51% for sex workers and the clients; MSMs reduce
partner numbers 8% and increase condom use by 31%; IDUs reduce the number of
people they share needles with by 40% and decrease "unclean" needle use by 50%.
Models used to project the spread of HIV in concentrated epidemics use the
relationship between the numbers with risk behaviors and the observed prevalence of
infection. However, to predict the rate and scale of spread through these populations it
is necessary to assume heterogeneity in risk behaviors among those within the groups.
Thus, a model generating the numbers infected over time has to make assumptions
about the fraction with high-risk behavior and the magnitude of this risk. The pattern
of growth and decline in the epidemic in the very highest risk population and the
relative importance of numbers of contacts and exposures to each contact are
underspecified in the available data and will further influence the potential impact of
interventions. The numbers of infections within the different population groups in the
model predicted for 2010 was used to generate a set of parameter values for this
model. The model was fitted to estimates for the Ukraine (Fig. S3), China,
Kyrgyzstan, and Uzbekistan. The epidemic is projected with and without the
interventions and the percentage reduction in the number of new infections in each
year calculated.
Simulations were conducted for typical countries in 16 epidemic categories defined
by the level of current adult HIV prevalence (very low, low, medium, and high), the
rate of growth in prevalence (low or high) and whether or not the epidemic is driven
by injecting drug use (Table S4). The patterns of incidence reductions resulting from
these simulations were used for all countries in those categories. For each country the
incidence in the base projection was modified according to the appropriate incidence
reduction pattern and a new projection was made using the Spectrum model to assess
the impact of prevention interventions on key indicators. Calculations were done for
the 46 countries with the largest epidemics which accounted for 95% of new
infections, and the results scaled to 100% of new infections.
1.1.3 Prevention Cost per Case Averted
The average cost per case averted was calculated by dividing the total prevention
costs (expressed in constant 2004 prices) over the period 2005–2015 by the number of
cases averted over the same period. This represents an average taken across changing
levels of coverage of interventions, as coverage of prevention interventions is
assumed to increase linearly to optimum coverage levels in 2010 and remain constant
thereafter. The point of comparison for the analysis was between this average cost of
preventing a case, and the net present value of the treatment and care costs that would
have been incurred had the case not been prevented.
1.2 The Costs of Averted Treatment and Care
1.2.1 Costs of Treatment and Care
The costs of averted treatment and care are also consistent with the 2005 UNAIDS
resource need estimates (S1) and include provider-initiated testing (diagnostic and
routine offer of testing), treatment and prophylaxis for opportunistic infections (OIs),
antiretroviral therapy (ART), including nutritional support, laboratory testing, and

Page 9

Science Supporting Online Material
Stover, p. 5
palliative care. The estimation methodology for all these services followed the
previously described approach (S13). The unit costs of selected interventions by
region are included in Tables S5 and S6.
For the purposes of estimating averted care costs, it was assumed that the great
majority of people who would have been infected during 2005–2010 in the absence of
scaled-up prevention will not require treatment until after 2010, by which time ART
coverage is assumed to have increased to 80% in all low- and middle-income
countries. Thus, the cost estimates assume 80% coverage for all averted infections.
It is possible that this method yields an underestimate of the true costs of treatment
and care. If the projected treatment scale up is not accompanied by a simultaneous
injection of resources to strengthen health systems and build human capacity and
infrastructure in the affected countries, then it is likely that there will be additional
“shadow costs” resulting from resources diverted from other purposes.
Estimated costs of care and treatment are based on 2004 prices and informed by
country-specific estimates that adjust costs based on local unit costs when available
and on imputed local drug and salary costs when local unit costs are not available.
Palliative care, treatment of OIs, and testing were all estimated as lifetime costs for a
general package of services and drugs, regardless of when the services were provided.
Prophylaxis of OIs was estimated as an annual cost, and for those people on ART
40% of patients were estimated to discontinue prophylaxis after six months on ART;
the remainder continue until death. Treatment for children was estimated as 100% of
adult costs. All unit cost data were validated through regional workshops, organized
by UNAIDS, with over 155 experts drawn from 78 affected countries.
Unit costs for antiretroviral drugs and laboratory monitoring used prices negotiated
with providers of drugs and diagnostics. The Web site of the Global Fund to fight
AIDS, Tuberculosis, and Malaria provided antiretroviral drug costs for middle-
income countries, using originator company drugs (S14). Generic company drug
prices were used for low-income countries (S15). The prices of second-line drugs
were based on recent estimates by Médecins sans Frontières (S16). Within each
national income range, all of the drug prices were adjusted for individual countries
based on purchasing power parity (PPP) adjusted gross national income per capita
(S17) and were assumed to prevail to 2015. We validated the resulting country cost
for treatment and care against estimates generated by experts from 78 countries that
participated in regional workshops and also against additional sources of country-
level data. For the very few countries which, either due to the size of the epidemic or
their negotiating power/ability had been able to negotiate prices below the estimated
prices, the lower prices were used. Average per-patient year on first line was
estimated about US$1087 (min US$332, max US$2532) and for second line about
$2813 (min $2058, max $4249).
These costs do not include the program costs for management, administration,
M&E, research, advocacy, and training that add about 7% to the total costs of
prevention and treatment. They were excluded because they are not purely prevention
costs, but support the entire program of prevention, treatment, care, and orphan
support. Including them would raise the costs of treatment and care by about 8%.
1.2.2 Lifetime Costs of Care and Treatment per Patient
Lifetime costs for each AIDS case were calculated assuming that the median number
of years of ART treatment is 7.5 years in all countries. Since the potential cost of

Page 10

Science Supporting Online Material
Stover, p. 6
averted treatment is being calculated in order to calculate a net cost of prevention, the
net present value (NPV) of the expected lifetime treatment costs is calculated for the
year in which the infection is prevented. A discount rate of 5%, ART coverage of
80%, and an average of 7 years from infection to onset of ART treatment were used in
the NPV calculations. The country-level data were aggregated to form weighted
regional averages.
Calculations of care and treatment costs considered only direct costs of care and did
not consider competing causes of death because of the low mortality among young
adults when the great majority of infections occur. Nor did the calculations include
costs related to loss of productivity or costs related to hospitalization. The costs of
palliative care and OIs were distributed uniformly across all years of antiretroviral
treatment for purposes of discounting.
1.3 Sensitivity Analyses
Sensitivity analyses were conducted for both the cost of prevention of averting an
infection during 2005–2015, and the lifetime cost of treatment and care of these
infections. The results are shown in Table S7.
To explore the sensitivity of prevention costs, the effectiveness of interventions was
lowered by using estimates of impact based on the lower quartile of the studies
available for each cell of the impact matrices.
To explore the sensitivity of lifetime cost of treatment and care, the default average
survival of 7.5 years was modified to 5 and 10 years, respectively, as upper and lower
bounds.
The results are also sensitive to the assumption about the target coverage of ART.
If target coverage rate were only 70% instead of 80%, then the net savings per
infection averted would be $200. If the target coverage were 90%, the net savings per
infection averted would be $1400.

Page 11

Science Supporting Online Material
Stover, p. 7
This file was updated 23 February 2006 for clarification of abbreviated terms.

Page 12

Science Supporting Online Material
Stover, p. 8
2

Figures and Legends

Figure S1. Global HIV Incidence with and without
comprehensive prevention package
0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
7,000,000
2005 200620072008 20092010 20112012 201320142015
Numbers of new adult and child HIV infections per year
Current prevention effort
Comprehensive prevention


Legend.

The blue line represents the annual number of new infections according to the
baseline projections (assuming continuation of the current prevention effort).
The pink line represents the annual number of new infections according to the
scenario assuming scale-up of a comprehensive prevention and care effort.

Page 13

Science Supporting Online Material
Stover, p. 9
Figure S2



























Model risk groups and routes of transmission: Model risk groups and routes of transmission:
MSMMSM
Bisexual Bisexual IDUIDU ClientsClientsLow Low
IDU FSWIDU FSW IDUIDU FSWFSWLow Low
MSM sexMSM sex
‘needle’
sharing sharing
Sex work Sex work
Heterosexual
sexsex
‘needle’
Heterosexual

Page 14

Science Supporting Online Material
Stover, p. 10
Figure S3
Reduction in incidence each year (Ukraine)
0
10
20
30
40
50
60
70
80
90
100
2000
2002
2004
2006
2008
2010
2012
2014
2016
2018
Year
2020
2022
2024
2026
2028
2030
2032
2034
Residual incidence (%)

Page 15

Science Supporting Online Material
Stover, p. 11
Fig. S4. Distribution by region of resources needed for comprehensive
prevention. NAME, North Africa and the Middle East.

Africa, 29%
Sub-Saharan Africa, 29%
S/SE Asia, 21%
E Asia/Pacific, 26%
LA, 9%
E Europe, 12%
NA/NE, 3%
NAME, 3%




LAC, 9%