Study design, study population and inclusion by respondent-driven sampling (RDS)
This was a cross-sectional, community-based, quantitative study conducted among women selling sex in Lahore, Pakistan. We defined a woman selling sex as "A female engaged in selling sex, part time or full time, as a means of making a living".
The sample size was calculated by using the computer software EPI info version 6.04D, on the basis of an expected frequency of sexually transmitted infection of 12% ± 3 percent units, and a confidence interval of 95%, resulting in a total required sample size of 726.
The participants (n = 730) were recruited by respondent-driven sampling (RDS) from three areas, mapped in 2004, where they resided as well as sold sex (http://nacp.gov.pk). RDS involves chain referral sampling in a manner that allows it to be qualified as a probability sampling method. RDS has several advantages in recruiting hard-to-access populations, such as broad inclusion, larger recruitment chains, time efficiency and low cost. RDS also statistically adjusts for the biases inherent in how individuals of similar characteristics are networked and are likely to know and recruit each other. Furthermore, RDS works on networks of participants and each wave of recruitment adjusts for inclusion probability [10–13].
Due to the sensitivity of the issues discussed and the possibility of increasing the discrimination against women selling sex, the investigated areas are referred to as "Area A", "Area B" and "Area C". Briefly regarding the RDS, one woman selling sex from each area was randomly selected as a seed. She was given three coupons to bring in three more women selling sex, who were neither pregnant nor menstruating at the time of recruitment. This created what is defined as wave one. The three women recruited through the seeds were given three coupons each to recruit three more women selling sex. This continued until wave five. The chains were followed carefully to ensure that the same woman did not return and also that all recruited women were engaged in selling sex. A confirmation of this was provided by the staff of the clinic who had been working in the areas with women selling sex for the last 4 years and by noting personal identification marks. Our study was conducted among "Home-based" and "Kothikhana-based" women selling sex [7]. In "Area A", women selling sex were "Home-Based". In areas "B" and "C", women selling sex were "Kothikhana"-based. The survey participants received a free clinical examination, HIV and STI testing, and a double financial incentive (for participating as well as bringing in three more women) to compensate for their time and travelling costs.
Data collection
Data were collected from September to November 2007. A questionnaire with mainly closed-ended questions was developed, which was largely based on an earlier qualitative study conducted in the same areas [8]. The questionnaire, administered by face to face interviews, addressed demographic information, sexual life history, sexual contacts, HIV/STI prevention knowledge and practices. Knowledge was defined as "correct" if the participants could cite two correct modes of transmission, one of which had to be "sexual", and two methods of prevention of HIV/AIDS, one of which had to be "use of condoms" (http://nacp.gov.pk).
The questionnaire was developed in English, translated into Urdu and pre-tested. After revision, the questionnaire was tested again, and further revised. The data collection team was trained in data collection techniques including standardized clinical examination, biological specimen collection, labelling, storage, transportation and how to conduct the face-to face interview using the questionnaire. Clinical examination and sampling of biological specimens were performed using the guidelines of the National AIDS Control Program of Pakistan (http://nacp.gov.pk) and following the phlebotomy procedures as defined by the College of American Pathologists [14].
Laboratory diagnostics
Serological screening for HIV and T pallidum was performed using AxSYM HIV 1/2 gO kit (Abbott Laboratories, Wiesbaden, Germany) and Immunotrep RPR as a qualitative test (Omega Diagnostics Limited, Alva, Scotland, UK), respectively, in accordance with the instructions of the manufacturer. Screening positive HIV and T pallidum samples were confirmed using Capillus HIV 1/HIV 2 (Trinity Biotech, Wicklow, Ireland) and TPHA (Randex, Antrim, UK), respectively. For diagnosis of N gonorrhoeae and C trachomatis, endocervical specimens were analysed using AMPLICOR CT/NG PCR (Roche Diagnostics, Indianapolis, USA), including AMPLICOR Neisseria gonorrhoeae Detection Kit and AMPLICOR Chlamydia trachomatis Detection Kit, in accordance with the instructions of the manufacturer. For diagnosis of T vaginalis, high vaginal swabs were used to collect samples, immediately inoculated into InPouch TV Culture System (BIOMED DIAGNOSTICS, Oregon, USA), incubated and subsequently interpreted in accordance with the instructions of the manufacturer.
Data analysis
Epi info software version 6.04D was used for data entry. Data were processed and analysed using SPSS 13 (SPSS Inc., Chicago, Illinois, USA). Frequencies, proportions, mean, minimum and maximum were used for the analysis. Pearson's chi-square analysis was performed to test associations between potential risk factors and specified diagnosed infections. Multivariable logistic regression analysis was employed to indicate influencing factors on HIV/STI prevalence. These included age, education, marital status, religion, place of work, condom use, duration of selling sex, number of sexual partners per day, injecting drug use and blood transfusion. In addition RDS Analysis Tool v5.3 (http://www.respondentdrivensampling.org) was used to estimate the transition probability (probability of one group recruiting the other), adjusted mean network size based on the median age, Homophily (a measure of preference for connections to one's own group, which varies between -1 (completely heterophilous) and +1 (completely homophilous), and the confidence intervals.
Ethics
Study protocols, questionnaires and consent forms were approved by the Ethics Committee of the Pakistan Medical Research Council, Islamabad, Pakistan (2006) and Regional Ethics Committee, Stockholm, Sweden (2006). The participants were informed and gave their consent before inclusion in the study. Participation was linked confidential and participants were given access to their biological results with the appropriate counselling and treatment. HIV positive participants were referred to the Voluntary Counselling and Testing Centre in Lahore for monitoring of the HIV infection, clinical management and social rehabilitation.