Data for this analysis were from the Partners in Prevention HSV/HIV Transmission Study, which was a randomized placebo-controlled trial of herpes simplex virus type-2 (HSV-2) suppression with acyclovir among 3381 heterosexual HSV-2/HIV-1 co-infected individuals and their HIV-1 seronegative partners from 7 countries in southern and east Africa (Botswana, Kenya, Rwanda, South Africa, Tanzania, Uganda, and Zambia). At enrollment, all HIV-1 infected participants had CD4 counts ≥250 cells/µL and were not yet on ART. The median CD4 count was 462 cells/µL (interquartile range 348–631) and 68% were women. The primary outcome of the trial was virologically-linked HIV-1 transmission within the partnerships; acyclovir did not decrease the risk of HIV-1 transmission. The design, methods, and primary outcomes have been described previously 
For the present analysis, we used cross-sectional data on sexual behavior, CD4 count, and plasma viral load collected at the time of enrollment into the Partners in Prevention HSV/HIV Transmission Study for HIV-1 infected partners with CD4 counts >350 cells/µL. Sexual behavior data were from a standardized questionnaire that asked the number of sex acts with the HIV-1 uninfected partner with and without condoms in the prior month. We used enrollment data only, because our goal was to describe a population presenting to clinical care for consideration of early ART (i.e., at CD4 >350) and thus to minimize the potential impact on sexual behavior reporting as a result of participation in the clinical trial, which included monthly risk-reduction counseling.
We used these cross-sectional data to estimate the number of transmissions that would occur in one year and the number that could be averted with ART in a hypothetical population of 100,000 HIV-1 infected persons in serodiscordant partnerships with CD4 counts and viral load similar to the study participants. To estimate expected transmissions, we first assigned each HIV-1 infected participant per-act infectivity estimates for sex acts with and without condoms, based on our prior analysis of infectivity in the same study population 
. Notably, in this prior analysis, neither gender nor CD4 count predicted per-act HIV-1 transmission after accounting for reported condom use and plasma viral load. Accordingly, infectivity for the present analysis was estimated solely as a function of reported condom use and viral load. For unprotected acts, this was: λ1
plasma viral load –4)]}, and for acts protected with a condom: λ0
plasma viral load –4)]}. The estimated annual numbers of sex acts with and without condoms were calculated from reported behaviors in the prior month and multiplied by 12. Next, each participant's probability of transmitting to his or her partner within the next year was estimated with the following equation: 1– [(1–λ0
)annual #acts with condom
) annual #acts without condom
]. Because reported condom use was higher in this clinical trial cohort than in previous observational studies of HIV-1 serodiscordant couples, potentially reflecting risk-reduction counseling during the pre-enrollment study screening period, we also calculated each participant's probability of transmitting with reduced condom use (half the amount reported and none). Each participant was assigned a weight of 100,000 divided by the sample size.
Participants were stratified according to plasma viral load (<10,000; 10,000–49,999; ≥50,000 copies/mL) and CD4 count (351–499; ≥500 cells/µL). Several strategies for early ART initiation were considered: CD4 count <500 cells/µL; plasma viral load ≥50,000 or ≥10,000 copies/mL; and universal coverage. The CD4 threshold was based on current international guidelines; viral load thresholds were based on the low risk of transmission with plasma viral loads <10,000 copies/mL 
, and the high risk of transmission at plasma viral loads ≥50,000 copies/mL 
. Within each early ART scenario, we calculated the number of transmissions potentially averted, assuming a 96% reduction in risk as a result of ART 
. In a sensitivity analysis, we assumed an 80% reduction in HIV-1 transmission risk, to reflect potential ART prevention benefits outside of a clinical trial 
. Finally, we estimated the proportion eligible to initiate treatment and the proportion of expected transmissions averted for each potential guideline, and summarized these with a ratio of proportion averted to proportion treated. To visualize the continuous relationship between increasing the proportion initiating treatment and the proportion of infections averted, we further stratified the population by deciles of increasing viral load and by deciles of decreasing CD4 count, and estimated the proportion of infections potentially averted within each decile. We plotted each decile of the population (potentially initiating ART) against the proportion of expected transmissions averted to compare this relationship when ART initiation is based on increasing viral load to the relationship based on decreasing CD4 count.
The clinical trial protocol and informed consent documents were reviewed and approved by human subjects research committees at the University of Washington and all local study site and affiliated institutions. All clinical trial participants provided written informed consent for participating in this research.