Scaling Up Antiretroviral Therapy in South Africa:
The Impact of Speed on Survival
Rochelle P. Walensky,1,2,3,5Robin Wood,10Milton C. Weinstein,6Neil A. Martinson,8,11Elena Losina,4,5,7
Mariam O. Fofana,2Sue J. Goldie,6Nomita Divi,2Yazdan Yazdanpanah,12,13,14Bingxia Wang,2,5A. David Paltiel,9
and Kenneth A. Freedberg1,2,5,6,7for the CEPAC-International Investigatorsa
The Divisions of1Infectious Disease and2General Medicine, Department of Medicine, Massachusetts General Hospital,3Division of Infectious
Disease, Department of Medicine and4Department of Orthopedic Surgery, Brigham and Women’s Hospital, and the5Center for AIDS Research,
Harvard Medical School,6Department of Health Policy and Management, Harvard School of Public Health, and the7Departments of Epidemiology
and Biostatistics, Boston University School of Public Health, Boston, Massachusetts;8School of Medicine, Johns Hopkins University, Baltimore,
Maryland;9Yale School of Medicine, New Haven, Connecticut;10Desmond Tutu Research Center, Institute of Infectious Disease and Molecular
Medicine, University of Cape Town, Cape Town, and11Perinatal HIV Research Unit, Wits Health Consortium, Johannesburg, South Africa;
12Service Universitaire des Maladies Infectieuses et du Voyageur,13Centre Hospitalier de Tourcoing, EA 2694, et Faculté de Médecine de Lille,
14Laboratoire de Recherches Économiques et Sociales, CNRS URA 362, Lille, France
(See the editorial commentary by Hirschhorn and Skolnick, on pages 1223–5.)
Only 33% of eligible human immunodeficiency virus (HIV)–infected patients in South Africa
receive antiretroviral therapy (ART). We sought to estimate the impact of alternative ART scale-up scenarios on
patient outcomes from 2007–2012.
Using a simulation model of HIV infection with South African data, we projected HIV-associated
mortality with and without effective ART for an adult cohort in need of therapy (2007) and for adults who became
eligible for treatment (2008–2012). We compared 5 scale-up scenarios: (1) zero growth, with a total of 100,000 new
treatment slots; (2) constant growth, with 600,000; (3) moderate growth, with 2.1 million; (4) rapid growth, with 2.4
million); and (5) full capacity, with 3.2 million.
Our projections showed that by 2011, the rapid growth scenario fully met the South African need for
and 52% of the need, respectively. The latter scenarios resulted in 364,000 and 831,000 people alive and on ART in
2012. From 2007 to 2012, cumulative deaths in South Africa ranged from 2.5 million under the zero growth scenario
to 1.2 million under the rapid growth scenario.
Alternative ART scale-up scenarios in South Africa will lead to differences in the death rate that
amount to more than 1.2 million deaths by 2012. More rapid scale-up remains critically important.
South Africa has one of the largest burdens of HIV dis-
ease in the world, with an estimated 4.9–6.1 million
people infected and a reported prevalence of 18.8% in
adults 15–49 years old . In 2007, the World Health
gramme on HIV/AIDS (UNAIDS) estimated that
1,000,000 people in South Africa required antiretroviral
therapy (ART) . The number of patients with access
to ART is steadily increasing, largely as a result of fund-
ing from the South African government itself, as well as
strategic assistance from the US President’s Emergency
Plan for AIDS Relief (PEPFAR) and the Global Fund to
Received 21 June 2007; accepted 4 December 2007; electronically published 26
Presented in part: 14th Conference of Retroviruses and Opportunistic Infections
(CROI), 25–28 February 2007, Los Angeles, CA (abstract 549); and 2007 HIV/AIDS
Implementers’ Meeting, 16–19 June 2007, Kigali, Rwanda (abstract 1755).
Potential conflicts of interest: N.M. reports that he manages a President’s
Emergency Plan for AIDS Relief grant providing antiretroviral treatment. Y.Y.
reports no personal funding; he reports that he has served as an investigator for
trials with Tibotec Pharmaceutical, and has received travel grants to attend
scientific meetings from GlaxoSmithKline, Roche, Boehringer, Bristol-Myers
Squibb, Pfizer, Abbot, and Gilead. All other authors report no relevant conflicts of
The Journal of Infectious Diseases
© 2008 by the Infectious Diseases Society of America. All rights reserved.
Financial support: National Institute of Allergy and Infectious Diseases (R01
AI058736 to K.F., K24 AI062476 to K.F., K25 AI50436 to E.L., and P30 AI060354 to
Bruce Walker); Doris Duke Charitable Foundation (Clinical Scientist Development
Award 2005075 to R.W.). The funding sources had no input in study design,
collection, analysis, and interpretation of data, writing of the report, or the
decision to submit the paper for publication.
aMembers of the CEPAC-International Investigators team are listed after the
Reprints or correspondence: Rochelle P. Walensky, MD, MPH, Division of
General Medicine, Massachusetts General Hospital, 50 Staniford Street, 9th Floor,
Boston, Massachusetts 02114 (firstname.lastname@example.org).
M A J O R A R T I C L E
● JID 2008:197 (1 May)
● Walensky et al.
by guest on May 11, 2011
Fight AIDS, Tuberculosis and Malaria (GFATM) [3–5]. Treat-
infrastructure, and establishment of treatment guidelines .
eligible patients in South Africa .
For development of health policy, it is useful to quantify the
potential consequences of disease outcomes under different
treatment and practice scenarios . To this end, our objective
was to project alternative ART rollout scenarios over the next
ber of lives lost while awaiting needed therapy, to estimate the
number of people both in and out of care, and to project when
and whether HIV treatment needs would be fully met. These
estimates can be used to inform decisions regarding the life-
saving value of alternative treatment expansion scenarios in
South Africa, as well as in other developing countries.
We conduct this analysis in 4 steps. First, we parameterized a
detailed computer-based simulation model of HIV disease with
data from South Africa. Second, we conducted a series of analy-
ses that assessed the outcomes associated with different ART
strategies in selected cohorts of patients. Outcomes included
mean CD4 count and HIV RNA levels of the surviving cohort
each year. Treatment strategies included no ART and ART that
of sequential ART regimens available). Each strategy included
co-trimoxazole prophylaxis in all scenarios, provided according
to WHO guidelines . Cohorts were selected to reflect the fact
that in 2007 (the year of the most up-to-date WHO estimates)
, patients who met ART eligibility criteria were distributed
across HIV disease stages, as well as the fact that in subsequent
to meet ART criteria. Third, we used the information generated
from the model simulations to assess outcomes associated with
different scenarios intended to represent potential population-
based strategies for ART scale-up (table 1). Outcomes were ex-
modeled, the number and efficacy of available ART regimens
[10, 11], and the availability of CD4 count monitoring .
CEPAC-International Model and Parameterization
The CEPAC-International Model is a computer-
based simulation model of the natural history of HIV infection
in different settings; it has been used to estimate the clinical and
economic consequences of different strategies for opportunistic
infection prophylaxis and ART [13, 14]. We provide a focused
version) [13, 14]. Disease progression was portrayed as a se-
quence of monthly transitions between health states defined to
capture key elements of disease and prognosis. Health states re-
flected chronic HIV infection, acute illness related to HIV (e.g.,
opportunistic infection), or death. Health states were stratified
by HIV RNA level (?30,000 copies/mL; 10,001–30,000 copies/
mL; 3,001–10,000 copies/mL; 501–3,000 copies/mL and ?500
copies/mL), which informs the monthly rate of decline in CD4
350 cells/?L, 351–500 cells/?L, and ?500 cells/?L) informs the
monthly risk of opportunistic infection and death (table 2).
Clinical decisions, such as the initiation of ART, were based on
clinical findings and, when available, CD4 count and HIV RNA
results. Patients on effective ART experience monthly increases
death. After virologic failure, CD4 counts decline, with a con-
who remain on ART despite virologic failure continue to have
Table 1.Alternative scale-up scenarios for antiretroviral therapy analyzed in the study.
Scale-up scenarioDescriptionNew treatment slots available in year(t?1)
There are no new treatment slotsa
A fixed number of new slots open each year
Each year, 100,000 additional new slots open,
compared with the prior year
Each year, the number of new slots doubles,
compared with the prior year
Each year, there are slots available to treat
everyone in need
New slots(t?1)? 0
New slots(t?1)? new slots(t)
New slots(t?1)? new slots(t)? 100,000
Rapid growthNew slots(t?1)? min ([2 ? new slots(t)],
patients in need(t?1))b
New slots(t?1)? Patients in need(t?1)? existing
slots freed up by deaths(t)
aThis reflects newly available treatment slots for antiretroviral therapy (ART) and does not reflect treatment slots that became
available due to deaths of patients who received ART in the previous year.
bThe no. of new slots is equal to whichever of these 2 values is smaller.
Scaling Up ART in South Africa
● JID 2008:197 (1 May)
by guest on May 11, 2011
as the complement of the corresponding yearly probability of sur-
Patients receiving ART.
We obtained the probabilities of
survival for patients receiving ART by simulating the course of
HIV disease using the CEPAC-International model . Be-
cause baseline clinical characteristics differed between the prev-
alent and incident cohorts, mortality rates were derived sepa-
characteristics of patients who received ART depended on the
duration of time spent awaiting ART, several sets of yearly mor-
tality rates were derived (table A3).
We obtained the baseline characteristics of patients who had
awaited ART for varying durations of time (0–5 years) by sim-
ulating the course of disease for patients in the prevalent and
incident cohorts while receiving co-trimoxazole prophylaxis
only, and recording population characteristics at the end of ev-
ery year (table A2).
We then simulated the course of disease while receiving ART
table A3. The first column indicates the patients’ cohort and the
number of years that patients had been receiving ART since the
initiation of the analysis in 2007. Columns 2 through 7 list the
the number of calendar years they waited for treatment.
Figure A1 illustrates the population alive, as a percentage of the
total population eligible for ART in each year. As an additional
cohort), the cumulative percentage of patients alive depends on
the speed of treatment scale-up. The change in the slope (from
downward to upward) in the rapid-growth and moderate-
growth curves reflects improved survival rates over time, com-
pared with the increasing total denominator of patients in need
(an additional 551,000 patients yearly). By 2012, the rapid-
growth and moderate-growth scenarios result in 67% and 61%
of the population alive, respectively, compared with the cumu-
lative eligible number (3,755,000). The constant-growth and
zero-growth scenarios result in worse cumulative survival rates,
with 42% and 34% of the population alive, respectively.
year, for each modeled scenario. The cumulative total of all eligible patients is noted below the graph, by year. The changing denominator and rapid
increase in access to ART account for the change in slope of the moderate-growth and rapid-growth curves. Nonprioritized cases are indicated with
open symbols, and prioritized cases are indicated with solid symbols.
Percentage of total patients eligible for antiretroviral therapy (ART) (both those awaiting and those receiving ART) who were alive by
by guest on May 11, 2011