Introduction Population size estimates for most at risk populations are essential for modelling and projections of the HIV epidemic and planning and prioritising prevention interventions. WHO/UNAIDS recommends countries produce multiple estimates given the large variance and biases inherent in available methods.

Methods Between January and May, 2012, 431 female sex workers (FSW) ages 18 and older recruited by time-location sampling participated in a HIV bio-behavioural surveillance survey (BSS) in the greater metropolitan area of Asuncion. Prior to recruitment, sex work venues and public places were mapped and FSW enumerated through key actor interviews. Makeup kits were then distributed to FSW present during site visits to a random sample of mapped venues. BSS question items assessed the percent of study participants that received the kit (the “unique object” method). Additional BSS question items assessed the percent of FSW that had accessed public HIV testing services from July to December, 2011 and services data were obtained on the number of FSW tested at the same sites in the same period (the services multiplier method). Estimated percentages from the BSS were weighted for differential probability of venue selection and incorporated Huber-White adjustments for clustering by venue.

Results Mapping identified 425 FSW at 72 sex work locations. Unique objects were distributed to 293 FSW. Of BSS participants, 22.7% (95% confidence interval [CI], 17.8%–28.4%) reported receiving the object, corresponding to a size estimate of 1292 (IC, 1031–1644) FSW. 53 FSW accessed HIV testing sites during the period and 6.0% (IC, 3.8%–9.4%) of BSS participants reported testing, for a size estimate of 880 (IC, 564–1392) FSW.

Conclusion Size estimates obtained by unique object and services multipliers were not significantly different, while mapping produced a significantly lower estimate, most likely reflecting changes in the population of FSW over time.