In a cross-sectional study of 926 subjects from 10 drug treatment programs conducted in 1984 in New Jersey, the seroprevalence of human immunodeficiency virus (HIV) was 35% overall; 30% in whites, 33% in hispanics, and 46% in blacks (p = 0.01 for comparison of blacks to non-blacks). Univariate analysis showed the seroprevalence of HIV was not associated with age or gender, but did correlate with frequency of cocaine or heroin injection (p trend less than 0.001); frequency of needle sharing (p trend = 0.007); and inversely with levels of education (p = 0.05). The prevalence of HIV was also inversely related to the distance of the treatment center from lower Manhattan; being highest for distances of less than 5 miles from lower Manhattan and lowest for distances of 80 miles, with intermediate rates for the intervening distances (p trend less than 0.001). In multivariate analyses, HIV seropositivity was consistently associated with the frequency of needle sharing (p = 0.02) and less than 12 years (high school level) of education (p = 0.02), but not with black race. However, blacks who shared needles less than once a month had a relative risk of 3.2 (95% CI 1.2, 7.7) while non-blacks who shared less than once a month had a relative risk of only 0.9 (95% CI 0.3, 2.4). The risk in non-blacks increased to more than twofold with more frequent needle sharing. When the analysis was stratified by gender and adjusted for needle sharing and geography, a significant twofold increased risk was observed for female (but not male) subjects who had two or more heterosexual partners compared with those who had one partner.(ABSTRACT TRUNCATED AT 250 WORDS)