This longitudinal study of a homogeneously urban African-American community cohort drawn from the metropolitan New York area (USA) analyzed three-year change in HIV-1 antibody prevalence and self-reported risk practices separately for men and women.

Specifically, risk practices reported on personal interviews and results from blood sero-assay in 1989-1990 were compared to similar interview reports of risk practices and HIV assay from saliva samples in 1993.

Results showed :

(1) women's infection rates increased more than men's, resulting in a male to female infection ratio of 1.5 :

• 1 ;

• (2) heterosexual transmission accounted for most of women's new infection ;

• (3) in multivariate analysis, multiple partners was a significant heterosexual risk indicator for both men and women, increasing infection odds five-fold ;

• and (4) buying sex also increased heterosexual men's infection odds approximately five-fold.

Results document increasing rates of African-American HIV infection linked largely to heterosexual exposure risk, that the African-American male-female infection ratio resembles the global rather than North American one, and also the utility of an expanded classification of heterosexual exposure risk.