Was the HIV infection burden in female sex workers in China overestimated? – Authors' reply

Abstract

Stefan Baral and colleagues described the burden of HIV infection among female sex workers in low-income and middle-income countries, on the basis of a systematic review and meta-analysis of published studies. In their analysis, the authors estimated a pooled HIV prevalence of 3·0% (95% CI 2·8–3·3) among female sex workers in China. However, we believe that this prevalence is overestimated, partly because of selection and publication biases. HIV prevalence estimates from population-based surveys are vulnerable to selection bias. Many studies, yielding a wide range of HIV prevalence in female sex workers, have been published in both international English journals and in national Chinese journals. However, almost all studies included in the analysis came from English-language journals. The severity of the HIV epidemic varies widely across China. Six provinces with the most reported cases (Yunnan, Guangxi, Henan, Sichuan, Xinjiang, and Guangdong) are classifi ed as highprevalence areas, and account for 70–80% of total reports nationwide. The regions that include the Honghe and Dehong prefectures in Yunnan Province are HIV epidemic hotspots. In Baral and colleagues’ meta-analysis, six of the 11 cited studies were undertaken in the Honghe Prefecture, accounting for 96·7% (503/520) of all HIV infections in the analysis. Moreover, 2·8–16·3% of the female sex workers recruited from Honghe Prefecture in studies included in the systematic review and meta-analysis self-reported drug or injection drug use within the past 3 months. Such selection bias could result in poorly representative and over-generalised estimates. Publication bias might also be a concern because the literature search strategies used in the analysis include only published studies. Such searches are unlikely to provide a representative sample because studies with low or even zero prevalence of HIV infection in female sex workers might not be submitted fully to international or national journals. A literature review based on 15 studies showed a median HIV prevalence of 0·6% (range 0–10·3%), with the highest prevalence in Yunnan Province (8·3–10·3%). So far, the national HIV sentinel surveillance system in China has included more than 500 sentinel sites for female sex workers across 31 provinces. Data from the system in 2011 suggested that the HIV prevalence among female sex workers was consistently low in most regions, and the province-specifi c HIV prevalence rates higher than 1% were clustered in Yunnan, Xinjiang, Guangxi, Sichuan, and Guizhou provinces, where injection drug use was common among female sex workers. Many studies have shown a substantial diff erence in HIV prevalence between female sex workers recruited from high-tier (karaoke bars and hotels), middle-tier (beauty or hair salons, barber shops, massage parlours, and foot-bathing shops), and low-tier sex venues (the street and other outdoor public places), as the table shows. However, female sex workers from low-tier or middle-tier sex establishments are less well represented in the present surveillance programme in China than are those from hightier establishments. We believe that some factors, at least the geographical variations in HIV background prevalence and injection drug use behaviours, and the sociobehavioural variations in female sex workers from diff erent types of sex venues, should be considered carefully in estimations of the HIV burden among female sex workers in China.