Introduction
As HIV is very effectively acquired during condomless receptive anal intercourse (AI) with serodiscordant and viremic partners, the practice could contribute to the high prevalence among female sex workers (FSW) in eSwatini (formerly known as Swaziland). We aim to estimate the proportion reporting AI (AI prevalence) among Swazi FSW and to identify the correlates of AI practice in order to better inform HIV prevention interventions among this population.

Methods
Using respondent-driven sampling (RDS), 325 Swazi FSW were recruited in 2011. We estimated the prevalence of AI and AI with inconsistent condom use in the past month with any partner type, and inconsistent condom use during AI and vaginal intercourse (VI) by partner type. Univariate and multivariable logistic regression models were used to identify behavioural and structural correlates associated with AI and AI with inconsistent condom use.

Results
RDS-adjusted prevalence of AI and AI with inconsistent condom use was high, at 44%[95% confidence interval (95%CI):35–53%]) and 34%[95%CI:26–42%], respectively and did not vary by partner type. HIV prevalence was high in this sample of FSW (70%), but knowledge that AI increases HIV acquisition risk low, with only 10% identifying AI as the riskiest sex act. Those who reported AI were more likely to be better educated (adjusted odds ratio(aOR) = 1.92[95%CI:1.03–3.57]), to have grown up in rural areas (aOR = 1.90[95%CI:1.09–3.32]), have fewer new clients in the past month (aOR = 0.33[95%CI:0.16–0.68]), and for last sex with clients to be condomless (aOR = 2.09[95%CI:1.07–4.08]). Although FSW reporting AI in past month were more likely to have been raped (aOR = 1.95[95%CI:1.05–3.65]) and harassed because of being a sex worker (aOR = 2.09[95%CI:1.16–3.74]), they were also less likely to have ever been blackmailed (aOR = 0.50[95%CI:0.25–0.98]) or been afraid to walk in public places (aOR = 0.46[95%CI:0.25–0.87]). Correlates of AI with inconsistent condom use were similar to those of AI.

Conclusions
AI is commonly practised and condom use is inconsistent among Swazi FSW. Sex act data are needed to determine how frequently AI is practiced. Interventions to address barriers to condom use are needed, as are biomedical interventions that reduce acquisition risk during AI.