Sexual Risk Behavior, HIV, and Sexually Transmitted Infections in a Cohort of Kenyan Female Sex Workers, 1993 – 2007

This thesis comprises a detailed analysis of sexual risk behavior among female sex workers participating in a prospective cohort study in Mombasa, Kenya between 1993 and 2007. To determine whether high-risk behavior has decreased over time, I examined trends in and associations with condom use and partner numbers at enrolment and over follow-up using multinomial logistic regression. While condom use increased among women enrolling into the cohort, women reduced partner numbers, rather than increasing condom use, over cohort visits. Workplace, charge for sex, duration of sex work, alcohol use, pregnancy and illness were all predictors of condom use. To evaluate the extent to which HIV risk estimates were affected by loss to follow-up, I investigated associations between sexual risk behavior, loss to follow-up, and HIV acquisition, using competing risks regression. Women reporting unprotected sex with multiple partners had the highest risk for HIV infection, and were also most likely to remain in the cohort. Finally, I used Andersen-Gill modeling to assess the impact of sexual risk behaviors on acquisition of sexually transmitted infections (STI) including gonorrhea, non-specific cervicitis, and trichomoniasis. While incident gonorrhea was closely associated with recent sexual risk behavior, incident trichomoniasis was not. Both conditions had high hazards for recurring in a 90-day window after a prior diagnosis. Non-specific cervicitis was demonstrated to be a chronic, relapsing condition associated with protected sex with multiple partners (possibly due to more frequent condom use) and with known biologic risk factors (i.e., pregnancy, hormonal contraceptive use, cervical ectopy, and genital ulcer disease). Overall, these analyses have led to a better understanding of how different sexual behavior patterns are associated with adverse outcomes, including HIV and STI acquisition, and identified specific factors associated with high-risk sexual behavior that may be amenable to intervention.Ph