Epidemiological and HIV prevention studies in sub-Saharan Africa have almost exclusively focussed on vaginal transmission of HIV, the primary mode of transmission in the region. Little is known about the prevalence of heterosexual anal intercourse (AI), its correlates and its role in the spread of HIV. Prevention messaging seldom, if ever, includes AI. Sexual and other risk behaviours (including frequency of AI) were assessed in two cross-sectional surveys of female sex workers (FSW) in Kigali, Rwanda (n=800) and Mombasa, Kenya (n=820). In addition, a subset of FSW surveyed in Kigali attended seven focus group discussions and four in-depth interviews. AI was reported by 5.5% and 4.3% of FSW in the cross-sectional surveys, in Kigali and Mombasa, respectively. FSW practising AI reported multiple risk factors for HIV transmission: inconsistent condom use (odds ratio (OR) Kigali 5.9 (95% CI 1.4-24.7); OR Mombasa 2.1 (1.1-4.2)); more than five sexual partners in the past week (OR Kigali 4.3 (1.5-12.4); OR Mombasa 2.2 (1.1-4.3)); alcohol use before sex (OR Kigali 2.8 (1.4-5.8)); more than 5 years of female sex work (OR Mombasa 2.4 (1.2-4.9)); and history of genital symptoms in the past year (OR Mombasa 3.6 (1.7-7.9)). AI was, however, not associated with HIV prevalence (OR Kigali 0.9 (0.5-1.9); OR Mombasa 0.5 (0.2-1.2)). Negative connotations and stigma associated with AI were expressed during qualitative interviews. AI was associated with several indicators of sexual risk behaviour. Prevalence of AI was probably underreported due to social desirability bias. Stigma associated with AI poses methodological challenges in obtaining valid data.