BACKGROUND HIV transmission between discordant couples has become an important source of new infections in China. To describe the seroconversion rate among serodiscordant couples and to identify salient behavioral and clinical risk factors including ART that affect heterosexual HIV transmission risk among couples in rural China. METHODS Longitudinal follow-up of an open cohort of HIV serodiscordant couples took place between 2007 and 2011 in Zhumadian, a city in southern Henan province in China, where blood plasma selling in 1990s led to a regional HIV epidemic. Annual follow-up included separate face-to-face interviews of husbands and wives, and HIV antibody testing for non-index partners. Cox proportional-hazard modeling was used to assess the relationship between HIV seroconversion and covariates of interest. RESULTS By the end of 2011, 4499 HIV serodiscordant couples had been enrolled in at least two follow-up interviews; 100 non-index partners seroconverted during the entire observation period for an incidence rate of 0.82 per 100 person-years (95% CI: 0.66-0.99). The incidence rates by the end of 2008, 2009, and 2010 were 2.14, 1.51, and 0.90 per 100 personyears respectively. Always using condoms in the past year of sex, gender of the index partner, frequency of sex, and ART exposure were all significant predictors of HIV seroconversion in the negative spouse. ART was highly protective against seroconversion whether the index partner was actively receiving treatment at the last follow-up (RR = 0.05, 95% CI, 0.01-0.16) or if the index partner had ever received ART (RR = 0.01, 95% CI, 0.00, 0.12). The risk of seroconversion in the nonindex spouse also decreased the longer the duration of the index partner's exposure to ART. CONCLUSIONS ART exposure and always using condom were highly protective against HIV seroconversion in the negative spouse. HIV incidence in serodiscordant couples has been decreasing over time, associated with ART treatment time within 7 years in the index partner. Gender of the index spouse and frequency of sex were also important predictors. Treatment as part of a combination prevention package may be a feasible method of HIV control in this population.