MODELING INCIDENCE RATE RATIO AND RATE DIFFERENCE - ADDITIVITY OR MULTIPLICATIVITY OF HUMAN-IMMUNODEFICIENCY-VIRUS PARENTERAL AND SEXUAL TRANSMISSION AMONG INTRAVENOUS-DRUG-USERS

The analysis concerns data from the Northern Italian Seronegative Drug Addicts Study, a multicenter longitudinal study about the incidence of human immunodeficiency virus infection in intravenous drug users from Milan and other areas of northern Italy between 1987 and 1991. Different measures of parenteral and heterosexual exposure effects were estimated by fitting multiplicative models for rate ratio and additive models for both rate ratio and rate difference into a Poisson regression model for grouped cohort data. In areas of high human immunodeficiency virus prevalence among intravenous drug users, the adjusted rate ratio under a multiplicative structure was 6.2 (95% likelihood-based confidence interval (LCI) 2.9-14.4) for parenteral and 2.9 (95% LCI 1.3-6.1) for sexual transmission. Under the additive model, the rate ratio was 7.8 (95% LCI 3.4-20.2) for parenteral and 9.2 (95% LCI 2.2-29.7) for sexual transmission, and the rate difference per 100 person-years was 9.8 (95% LCI 5.3-15.6) for parenteral and 10.5 (95% LCI 1.8-24.2) for sexual transmission (controlled for each other). Because of the small sample size, a clear discrimination between models could not be reached. However, in spite of the greater risk associated with parenteral transmission under a multiplicative model, the additive model suggests that the relative impact of measures aimed at inducing condom use is similar to that which would be obtained by measures aimed at stopping syringe sharing.