CONTEXT: Within the community-randomised ANRS 12249 Treatment-as-Prevention (TasP) trial conducted in rural South Africa, we analysed sexual behaviours stratified by sex over time, comparing immediate ART irrespective of CD4 count vs CD4-guided ART (start at CD4>350 then >500) arms. METHODS: As part of the 6-monthly home-based trial rounds, a sexual behaviour questionnaire (IQ) was administered to all residents >/=16 years. We considered seven indicators: sexual intercourse in the past month; at least one regular sexual partner in the past six months; at least one casual sexual partner in the past six months and more than one sexual partner in the past six months; condom use at last sex (CLS) with regular partner, CLS with casual partner, and point prevalence estimate of concurrency. We conducted repeated cross-sectional analyses, stratified by sex. GEE models were used, including trial arm, trial time, calendar time and interaction between trial arm and trial time. RESULTS: CLS with regular partner varied between 29%-51% and 23%- 46% for men and women, respectively, with significantly lower odds among women in the control versus intervention arm by trial end (p < 0.001). CLS with casual partner among men showed a significant interaction between arm and trial round, with no consistent pattern. Women declared more than one partner in the past 6 months in less that 1% of IQs; among men, rates varied between 5%-12%, and odds significantly and continuously declined between calendar rounds 1 and 7 (OR = 4.2 [3.24-5.45]). CONCLUSION: Universal Test and Treat was not associated with increased sexual risk behaviours.< Réduire