Aims
To investigate putative links between alcohol use, and unsafe sex and incident HIV infection in sub-Saharan Africa.

Methods
A cohort of 400 HIV-negative female sex workers was established in Mombasa, Kenya. Associations between categories of the Alcohol Use Disorders Identification Test (AUDIT) and the incidence at one year of unsafe sex, HIV and pregnancy were assessed using Cox proportional hazards models. Violence or STIs other than HIV measured at one year was compared across AUDIT categories using multivariate logistic regression.

Results
Participants had high levels of hazardous (17.3%, 69/399) and harmful drinking (9.5%, 38/399), while 36.1% abstained from alcohol. Hazardous and harmful drinkers had more unprotected sex and higher partner numbers than abstainers. Sex while feeling drunk was frequent and associated with lower condom use. Occurrence of condom accidents rose step-wise with each increase in AUDIT category. Compared with non-drinkers, women with harmful drinking had 4.1-fold higher sexual violence (95% CI adjusted odds ratio [AOR] = 1.9-8.9) and 8.4 higher odds of physical violence (95% CI AOR = 3.9-18.0), while hazardous drinkers had 3.1-fold higher physical violence (95% CI AOR = 1.7-5.6). No association was detected between AUDIT category and pregnancy, or infection with Syphilis or Trichomonas vaginalis. The adjusted hazard ratio of HIV incidence was 9.6 comparing women with hazardous drinking to non-drinkers (95% CI = 1.1-87.9).

Conclusions
Unsafe sex, partner violence and HIV incidence were higher in women with alcohol use disorders. This prospective study, using validated alcohol measures, indicates that harmful or hazardous alcohol can influence sexual behaviour. Possible mechanisms include increased unprotected sex, condom accidents and exposure to sexual violence. Experimental evidence is required demonstrating that interventions to reduce alcohol use can avert unsafe sex.