H I V / A I D SB R I E F R E P O R T
The Effect of Expanded
Antiretroviral Treatment Strategies
on the HIV Epidemic among Men
Who Have Sex with Men in San
Edwin D. Charlebois,1,2Moupali Das,1,3Travis C. Porco,4and Diane
1HIV/AIDS Division, Department of Medicine, San Francisco General Hospital, and
2Center for AIDS Prevention Studies, Department of Medicine, University of
California, and3San Francisco Department of Public Health, and4Francis I. Proctor
Foundation for Research in Ophthalmology, Department of Epidemiology and
Biostatistics, University of California, San Francisco, California
(See editorial commentary by DeGruttola and Schooley on pages
Modeling of expanding antiretroviral treatment to all HIV-
infected adults already in care in San Francisco predicts re-
ductions in new HIV infection at 5 years of 59% among men
who have sex with men. Addition of annual HIV testing for
men who have sex with men to universal treatment decreases
new infections by 76%.
A model of annual testing and immediate initiation of anti-
retroviral therapy (ART) in South Africa predicted a dramatic
reduction in incidence of HIV infection . In response, critics
highlighted the practical applicability of such an approach in
South Africa, and the debate about the optimum timing of ART
initiation continues. In San Francisco, which has a generalized
epidemic among men who have sex with men (MSM; 24%
in South Africa. More than 72% of MSM report annual testing,
and .85% of MSM are aware of their HIV status. Linkage to
primary care is high (88%) even among individuals who receive
a diagnosis at a public health clinic . Of most significance, the
San Francisco Department of Public Health estimates that 78%
of all known HIV-infected persons were receiving care in 2008
wide availability ofantiretroviralmedications, includingHealthy
San Francisco, a city-wide public health insurance safety net,
high population-level rates of virologic suppression are achiev-
able. In the context of this setting and early 2009 guidelines for
starting ART at CD4 cell counts ,350, we sought to determine
the potential impact on incident HIV infection in the MSM
population of offering ART to all patients in care—a strategy
that maximizes the individualand public health benefit for those
already receiving care without requiring additional investment
in outreach and expanded HIV testing.
We developed an ordinary differential equation simula-
tion model for HIV testing and treatment among MSM in San
Francisco extending previous models [1, 4, 5]. We tested 3 ART
expansion strategies: (1) treatment of all individuals currently
receiving HIV care with CD4 cell counts ,500 cells/mm3; (2)
treatment of all individuals receiving care; and (3) intensified
annual HIV testing combined with treatment of all HIV-in-
fected persons (the full test-and-treat strategy). Inputs to the
model were based on comprehensive surveillance information
on prevalence and incidence of HIV infection, testing rates, and
treatment outcomes data available for San Francisco from the
local health department and electronic patient databases of the
San Francisco General Hospital outpatient HIV treatment
clinics that contain information on 95% of individuals known
to be HIV infected in San Francisco . We stratified the
population of MSM according to risk groups, HIV status, CD4
cell count group, and treatment, as follows. We classified in-
dividuals as either uninfected or infected. Infected men, in turn,
are classified according to the untreated nadir CD4 cell count
into 1 of 4 stages: CD4 cell count .500 cells/mm3, CD4 cell
count of 350–500 cells/mm3, CD4 cell count of 200–350 cells/
mm3, and CD4 cell count ,200 cells/mm3. Individuals in the
model may (1) have unknown serostatus and not be receiving
treatment; (2) have known HIV infection but not receiving
ART; (3) be receiving ART but not yet achieved suppression; or
(4) be receiving long-term antiretroviral therapy. Eighty per-
cent of individuals receiving long-term therapy were assumed
to have achieved durable virological suppression. Untreated
Received 8 October 2010; accepted 12 December 2010.
Correspondence: Edwin D. Charlebois, MPH, PhD, Center for AIDS Prevention Studies,
University of California, San Francisco, 50 Beale St, 13th Fl, San Francisco, CA 94105
Clinical Infectious Diseases
? The Author 2011. Published by Oxford University Press on behalf of the Infectious Diseases
Society of America. All rights reserved. For Permissions, please email: journals.permissions@
oup.com. This is an Open Access article distributed under the terms of the Creative Commons
Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/2.5/), which
permits unrestricted non-commercial use, distribution, and reproduction in any medium,
provided the original work is properly cited.
d CID 2011:52 (15 April)
individuals progress forward through successive HIV stages
according to rates estimated from previous studies . In-
dividuals who have achieved virologic suppression proceed
through nadir CD4 cell count stages at a slower rate (zero in the
baseline scenario). Individuals in either the short-term in-
complete suppression classes have ongoing HIV progression
through stages at a rate intermediate between the untreated and
completely suppressed groups. The mean time between initia-
tion of therapy and achievement of full suppression was 3
Individuals with complete or incomplete suppression may stop
therapy at rates estimated from observed fractions of therapy
therapy) depends on the HIV stage. Individuals with a CD4 cell
count ,200 are assumed to start therapy with a mean waiting
time of 3 months. The time to initiation of therapy for individuals
in other stages is assumed to be 3 months when a test-and-treat
program has been initiated and is longer otherwise. The rate of
stopping therapy was higher for individuals with incomplete
suppression than for individuals with complete virological sup-
per 3 years in the absence of a treatment program (and, thus,
transition from unknown status to known status) at this rate.
We assumed 2 risk groups (high and low) distinguished by
more unprotected contacts per unit time. The transmission
probability per uninfected contact was assumed to be 0.1.
Eighty percent of the population was assumed to be in the low-
risk group. The transmission probability was assumed to be
lower ifthe infectedpersonwasreceiving treatment;weassumed
20% relative transmission for infected persons with incomplete
virological suppression and 1% for persons with complete vi-
rological suppression . We assumed nonrandom mixing be-
tween the high- and low-risk groups, with a mixing parameter
of 0.6 ( 0 for completely random mixing to 1 for completely
positive assortative [like only with like] mixing). The number
of contacts perunit time andthe mixing parameterwere chosen to
give a prevalence of 0.24 in 2007, in agreement with data. Finally,
sexually active residence time to be 30 years .
We performed sensitivity analyses on key variable inputs
to assess model results variability by repeating model runs
with inflated and deflated inputs for single variables chosen
from the range of likely values as estimated from the litera-
ture and clinic databases. Change in model results were
evaluated by calculating percent change in reduction in new
All 3 expanded ART strategies resulted in reduction of pre-
dicted prevalence of HIV infection and new HIV infections,
compared with the 2008–2009 standard of care of starting ART
for HIV-infected persons with CD4 cell counts ,350 cells/mm3
(see Table 1). Significant reductions in new HIV infections
were predicted in as early as 5 years from the implementation
of expanded ART for all 3 strategies. The test-and-treat all
strategy had the largest effect of the 3 expanded ART strategies,
eventually doubling the number of new HIV infections averted,
compared with the least ambitious strategy. However, signifi-
cant gains in averting new HIV infections were seen for both
of the expanded ART strategies (,500 and treat all) that in-
volved changing ART initiation practices without implement-
ing additional HIV testing and linkage to care programs. The
model does not predict elimination of the HIV epidemic
among MSM in San Francisco (reduction of prevalence and
incidence of HIV infection to negligible or zero levels). How-
ever, at 20 years, the test-and-treat strategy predicts reduction
in prevalence of HIV infection among MSM in San Francisco
from 26.2% to 12.8%, greater than reducing the prevalence of
HIV infection by half in the absence of changes in 2008–2009
The model was most sensitive to assumptions of the effec-
tiveness of ART in reducing HIV transmission probabilities. A
reduction in the effectiveness of ART in reducing HIV trans-
mission from 99% to 90% resulted in a .30% reduction in the
percentage of new HIV infections averted over 20 years. The
model was less sensitive to changes in the proportion of in-
dividuals who are able to achieve virological suppression, with
a 10% decrease in the proportion achieving suppression re-
sulting in a ,10% decrease in the percentage of new infections
averted. Other model input parameters evaluated in sensitivity
analyses (treatment cessation rates, mortality rates, in-
fectiousness with incomplete viral suppression, testing rates, and
the proportion in the high-risk group) demonstrated similar
relative effects between the strategies over the range of likely
San Francisco is one of the original epicenters of the HIV
epidemic and remains the site of one of the largest concen-
trated epidemics of HIV infection in the United States, with
a 24% prevalence among MSM, exceeding levels reported in
other concentrated epidemics, such as MSM in New York City
and African-American men in Washington, DC . Despite
advancements in treatment and expanded outreach efforts,
?600 incident HIV infections occurred in San Francisco in
2008 . New strategies of combination prevention that
include both behavioral and biomedical interventions are thus
Our model demonstrates that expansion of ART to those
already in care with CD4 cell counts ,500 cells/mm3or at any
d CID 2011:52 (15 April)
CD4 cell count is likely to significantly reduce the number of
future HIV infections. Further reductions could be gained by
the addition of routine annual testing and linkage to care.
These findings extend other models of San Francisco and
Vancouver by anchoring the analysis to empirical data re-
garding engagement in care and the new
Health & Human Services (DHHS) guidelines for ART initi-
It is instructive to contrast our findings to those of a recent
modeling analysis of theHIV epidemic in Washington, DC. Our
model predicts greater reductions in new HIV infections,
compared with analysis from Washington, DC. Although the
deterministic model with its inherent limitations differed from
the stochastic simulation model used by Walensky et al may
have contributed to these differences, the critical difference is
that only 50% of persons in the DC model are linked to care
after a positive HIV test result . Indeed, these authors and
other modeling reports from South Africa emphasize the
importance of linkage to care as a limiting factor in ART
expansion strategies reducing incident HIV infections .
When the DC model includes nearly complete linkage to care,
results of the models converge. Testing and linkage to care are
appropriately one of the highest public health priorities in
populations with high prevalence worldwide. In San Fran-
cisco, with the high rates of HIV testing and linkage to care,
we have the opportunity to reduce HIV transmission by ex-
panding ART to those already receiving care while we
determine the most cost-effective approaches to further ex-
Of importance, realization of the prevention benefits of ex-
panded ART strategies depend on a number of components.
Retention in care and expanded financial, clinical, social, and
structural adherence and support measures for treatment, in-
cluding specific support for psychiatric and substance use co-
morbid conditions, and addressing homelessness and marginal
housing all need to be components of the strategy in a city such
as San Francisco. Also of concern is the potential for changes in
behavior among MSM leading to increased transmission risk,
thereby offsetting any potential gains and the specter of drug-
resistant strains of HIV. However, these obstacles are not in-
surmountable, as evidenced by the recent public health com-
mitment and movement to universal ART and a comprehensive,
multi-level HIV prevention strategy in San Francisco . It is
possible that such strategies are comparatively cost-effective
and should be evaluated as data become available. Of note,
there is little observed evidence of significant increases in HIV
transmission risk behavior , and contrary to recent mod-
eling exercises, transmitted drug resistant HIV strains have
remained stable or even decreased in San Francisco and in
similar cities, such as Vancouver, where ART has rapidly ex-
Modeling of complex phenomena, such as a local HIV
epidemic and making cogent predictions about the future, are
subject to numerous limitations. In essence, modeling is
Table 1.Model Results
Prevalence of HIV infection,%
Baseline CD4 cell
count ,350 cells/mm3
ART initiation, CD4
cell count ,500 cells/mm3
Treat all in care
200924.7 24.7 24.724.7
201425.1 22.9 21.920.9
New HIV infections since 2009Baseline (CD4,350)ART start CD4,500Treat all in care Test-and-treat all
2014 370321491534 893
2019 7446 43442896 1406
202914,960 10,0206739 2771
HIV Infections Averted*ART start CD4,500Treat all in care Test-and-treat all
2014 Reference1554 2169 2810
2029Reference 4940 822112,189
Percent reduction in new HIV
ReferenceART start CD4,500 Treat all in careTest-and-treat all
2029 Reference3355 81
aHIV infections averted and percent reduction in new infections are relative to 2009 model estimates. Expansion of ART treatment strategies are assumed to
start in 2009.
d CID 2011:52 (15 April)
a thought experiment, informed by data, but subject to the
validity and completeness of the model’s internal structures.
The model presented here while able to recapitulate the
observed HIV epidemic in San Francisco from its beginning
with high fidelity does not specifically model the impact
of acutely HIV-infected persons, which may be a sig-
nificant driver of new HIV infections or account for pertur-
bations arising from HIV drug resistance which may affect
virological suppression rates; however, the sensitivity analyses
performed indirectly address these issues by varying the reach
of HIV testing and ART effectiveness in producing viral
In conclusion, our projections suggest that ensuring HIV-
infected patients in San Francisco already receiving care are
offered ART would significantly reduce the incidence of HIV
infection in San Francisco. These predictions are supported by
recent surveillance data from San Francisco and Vancouver that
show decreasing rates of new HIV infections that correlate with
ART expansion and lower community viral load [20, 21]. With
new national and San Francisco Health Department ART
guidelines supporting treatment for most HIV-infected persons
for their individual health, secondary benefits of reducing HIV
transmission could be realized if adequate support for care de-
livery are in place.
The authors would like to acknowledge Drs Grant Colfax, Brad Hare,
Mitch Katz, Willi McFarland, and Susan Scheer for their assistance with
surveillance and clinic data and advice. This work was presented in part at
the 17th Conference on Retroviruses and Opportunistic Infections
(CROI), San Francisco, 2010, paper #996.
Financial Support.E.D.C. received support from NIH/NIMH center
grant P30MH062246 (Stephen F. Morin. PhD. PI).
Potential Conflicts of interest.
funded studies in Uganda that were provided antiretroviral drugs by
Abbott. All other authors: no conflicts.
D.V.H. has conducted two NIH-
1. Granich RM, Gilks CF, Dye C, De Cock KM, Williams BG. Universal
voluntary HIV testing with immediate antiretroviral therapy as
a strategy for elimination of HIV transmission: a mathematical model.
Lancet 2009; 373:48–57.
2. Zetola NM, Bernstein K, Ahrens K, et al. Using surveillance data to
monitor entry into care of newly diagnosed HIV-infected persons: San
Francisco, 2006–2007. BMC Public Health 2009; 9:17.
3. HIV/AIDS Statistics and Epidemiology Section. HIV/AIDS Epidemi-
ology Annual Report 2008. San Francisco: San Francisco Department
of Public Health, 2009.
4. Blower SM, Gershengorn HB, Grant RM. A tale of two futures: HIV and
antiretroviral therapy in San Francisco. Science 2000; 287:650–4.
5. Dodd PJ, Garnett GP, Hallett TB. Examining the promise of HIV
elimination by ‘test and treat’ in hyperendemic settings. AIDS 2010;
6. HIV/AIDS Statistics and Epidemiology Section. HIV/AIDS Epidemi-
ology Annual Report 2008. San Francisco: San Francisco Department
of Public Health, 2009.
7. Schechter MT, LeN,CraibKJ, LeTN, O’ShaughnessyMV, MontanerJS.
Use of the Markov model to estimate the waiting times in a modified
WHO staging system for HIV infection. J Acquir Immune Defic Syndr
Hum Retrovirol 1995; 8:474–9.
8. Grant RM, Wiley JA, Winkelstein W. Infectivity of the human im-
munodeficiency virus: estimates from a prospective study of homo-
sexual men. J Infect Dis 1987; 156:189–93.
9. Donnell D, Baeten JM, Kiarie J, et al. Heterosexual HIV-1 transmission
after initiation of antiretroviral therapy: a prospective cohort analysis.
Lancet 2010; 375:2092–8.
10. El-Sadr WM, Mayer KH, Hodder SL. AIDS in America–forgotten but
not gone. N Engl J Med 2010; 362:967–70.
11. Scheer S, Chin CS, Buckman A, McFarland W. Estimation of HIV
incidence in San Francisco. AIDS 2009; 23:533–4.
12. Coates TJ, Richter L, Caceres C. Behavioural strategies to reduce HIV
transmission: how to make them work better. Lancet 2008;
13. Department of Health & Human Services (DHHS). Panel on anti-
retroviral guidelines for adults and adolescents. Guidelines for the use of
antiretroviral agents in HIV-1-infected adults and adolescents. Available
pdf. Accessed 2010:1–161.
14. Walensky RP, Paltiel AD, Losina E, et al. Test and Treat DC: Fore-
casting the Impact of a Comprehensive HIV Strategy in Washington
DC. Clin Infect Dis 2010; 51:392–400.
15. Charlebois ED, Havlir DV. ‘‘A bird in the hand’’: a commentary on the
test and treat approach for HIV. Arch Intern Med 2010; 170:1354–6.
16. Okie S. Fighting H.I.V., a community at a time. New York Times, 2009.
October 26 2009.
17. Blower S, Volberding P. What can modeling tell us about the threat
of antiviral drug resistance? Curr Opin Infect Dis 2002; 15:609–14.
18. Jain V, Pilcher C, Deeks S, et al. Increasing prevalence of NNRTI-
associated drug-resistance mutations in patients with acute, early HIV
in San Francisco. 16th Conference on Retroviruses and Opportunistic
infections. Boston: Massachusetts, 2009.
19. Gill VS, Lima VD, Zhang W, et al. Improved virological outcomes in
British Columbia concomitant with decreasing incidence of HIV type 1
drug resistance detection. Clin Infect Dis 2010; 50:98–105.
20. Das M, Chu PL, Santos GM, et al. Decreases in community viral load
are accompanied by reductions in new HIV infections in San Francisco.
PLoS One 2010; 5:e11068.
21. Montaner JS, Lima VD, Barrios R, et al. Association of highly active
antiretroviral therapy coverage, population viral load, and yearly new
HIV diagnoses in British Columbia, Canada: a population-based study.
Lancet 2010; 376:532–9.
d CID 2011:52 (15 April)