Background:  The long-term effect of ART on sexual HIV transmission risk in Africa is unknown. We assessed changes in sexual behavior and estimated HIV transmission from HIV-infected adults during 3 years of ART in Uganda.

Methods:  Between May 2003 and May 2007, we enrolled and followed ART-naive HIV-infected adults in a home-based AIDS care program that included voluntary counseling and testing (VCT) for cohabitating partners at enrollment and annually. Participants made transmission risk reduction plans and were provided condoms. Cohabitating discordant couples were invited to 4 prevention support group sessions. At baseline and every 6 months, we assessed participants' HIV plasma viral load and partner-specific sexual behaviors. We defined risky sex as intercourse with inconsistent or no condom use with partners of HIV^– or unknown serostatus in the previous 3 months. We compared rates using a Poisson regression model and estimated transmission risk per partner based on established viral load-specific transmission estimates. We determined HIV seroconversion rates among cohabitating HIV-discordant partnerships.

Results: Of 928 participants enrolled, 755 (81%) had data at 36 months, 94 (10%) died before 36 months, and 79 (9%) had missing data. Sexual activity increased from 28% at baseline to 41% at 36 months (p <0.001) (women:  21 to 31%; men:  47 to 73%). The proportion of sexually active participants reporting risky sex changed from 22% at baseline to 8% at 6 months (p <0.001) to 14% at 36 months (p = 0.018). Median viral load among those reporting risky sex was 122,500 (IQR 45,100 to 353,000) copies/mL at baseline and undetectable at follow-up. One sero-conversion occurred within 62 cohabitating sero-discordant partners in year one; none in years two or three. At 36 months, consistent condom use was higher with discordant (74%) or unknown (55%) HIV status partners than with concordant partners (46%). There was a 92% reduction in estimated risk of HIV transmission, from 45.7 to 4.2/1000 person-years.

Conclusions: Despite increased sexual activity and a small waning of behavioral risk reduction among HIV-infected Ugandans over 3 years of therapy, risky sex remained lower than baseline levels and ART provision was associated with reduced estimated HIV transmission risk. Integrated ART and prevention programs may reduce HIV transmission in Africa. Programs should consider on-going support for prevention interventions.