Two cross sectional, anonymous, surveys of IDUs were undertaken simultaneously in Belgrade and Podgorica in September and October 2005. Both surveys used identical sampling and interview methods. Inclusion criteria were: injecting drugs in the previous four weeks; living or working in Belgrade or Podgorica; aged 18 years or older; willing to give a dried blood spot (DBS) for antibody testing; and not having been interviewed for the study previously.
Participants were recruited using RDS, and were interviewed at fixed sites located in the centre of each city. The Belgrade site was housed within a discrete building used by the International AID Network (IAN), an NGO working with vulnerable people in South Eastern Europe. The Podgorica site was housed within some unmarked rooms in a mini shopping mall near the centre of the city, and comprised part of the premises of a local NGO called "Juventas", which provides youth services in the city. Survey staff members in both cities included outreach workers and former injectors whom had extensive experience working with vulnerable populations in Serbia and Montenegro. Potential participants were screened using a list of questions about injecting drug preparation, type and cost and by looking at injection stigmata. Pre-survey training in both cities included RDS methodology, processing potential participants through screening for inclusion criteria, gathering RDS specific data for analysis, providing assistance to participants using ACASI, and managing incentive payments. Recruitment began with a set of initial recruits ("seeds") who met the inclusion criteria and who were diverse in terms of key characteristics of the drug injecting community (eg. age, gender, area of residence, duration of injecting).
After giving written informed consent, all participants were interviewed using ACASI, taking approximately 30 minutes to complete. The questionnaire included sections on socio-demographics, drug use, injecting and sexual risk behaviour, sex work, contact with the police, and HIV testing history. It was developed from other survey questionnaires[2, 27], underwent piloting, and was informed by a linked qualitative study.
Upon completion of the ACASI, all participants gave at least two DBS specimens for unlinked anonymous testing for anti-HIV and anti-HCV. Participants were given specific information detailing the unlinked anonymous nature of DBS testing in the Information Sheet for the study, and had the option to withdraw consent and opt out of the study. No identifiers were collected and so it was not possible to trace participants to report to them their test results. However all participants in Belgrade were offered voluntary anonymous rapid HIV testing and results, along with full pre- and post-test counselling. HIV test results were available within approximately 30 minutes after receiving pre-test counselling. Those testing HIV positive were referred to the Belgrade AIDS centre for appropriate confirmatory testing. HIV/AIDS educational materials, condoms and trained VCT counsellors were available throughout the duration of the study. In Podgorica, instead of rapid testing, participants were referred to a local, newly opened, free of charge HIV VCT centre.
IDUs received "primary" and "secondary" cash reimbursements of €10 each for participation in the survey and €5 for each of their recruited peers (maximum three) who met the inclusion criteria and enrolled in the survey.
Laboratory methods – unlinked anonymous testing
Capillary blood was collected onto absorbent paper (Whatman 903, Astron, Huntingdon, UK) by finger prick using single use disposable lancets, and left to dry, thus becoming a DBS. Specimens were stored in plastic pouches with desiccant in on-site freezers, and were later transferred to the Clinic for Infectious and Tropical Diseases, Belgrade where they were stored at -25°C until testing.
DBS specimens were tested at the Clinic for Infectious and Tropical Diseases, Belgrade, under the supervision of staff of the Health Protection Agency Virus Reference Department. Six mm diameter disks were punched and each was placed in a designated well of a 96-well microplate. Plasma proteins, including antibodies, were eluted from the disks by overnight immersion in 200 μL of PBS/tween buffer.
Testing for anti-HCV was carried out using a modified protocol for the Ortho HCV 3.0 SAVe ELISA (product number 940982, Ortho Diagnostics, Amersham). The method employs a cut-off (CO) optical density (OD) of 0.090, and any DBS sample giving an OD value between 0.080 and 2.500 was re-eluted from a new punch and re-tested in duplicate using the same assay. Those that were repeatedly reactive, as well as those whose initial screening OD value exceeded the maximum for the plate reader (2.500), were considered to be anti-HCV positive. It was decided not to apply a second line assay with the intent to confirm the presence of anti-HCV primarily because the method employed is highly reproducible and specific. Even for a low HCV prevalence in IDUs of, for example, 30% the positive predictive value is 1.0. Particularly with the added safeguard of duplicate re-testing of any specimen with an OD in the range 0.08 – 2.5 the application of a second test would not have added significantly to testing accuracy.
For Anti-HIV testing an in-house version of the Abbott/Murex GACELISA HIV 1+2 enzyme immunoassay was employed. The commercial GACELISA HIV 1+2 test was established in several studies to provide sensitivity and specificity close to 100%  and good sensitivity shortly after anti-HIV seroconversion. Development of the commercial kit was led by one of the authors (JVP), but ceased to be manufactured in 2003. Consequently, an in-house version of the GACELISA was developed and validated at the HPA Centre for Infections. In brief, the assay employs a 96-well U-bottomed microplate solid phase (Nunc Maxisorp) coated with anti-human IgG antibodies (A0423, Dako UK Ltd), negative and positive controls (HIV-1 × 2 and HIV-2 × 1), specimen diluent (PBS with 10% fœtal bovine serum and 0.1% tween 20), each prepared in bulk at HPA, along with the conjugate (HRPO-tagged HIV-1 and HIV-2 antigens) and substrate (TMB), with their associated buffers scavenged from the Abbott/Murex HIV-1.2.0 GE95 EIA kit. The test is processed employing standard EIA equipment and absorbances are measured at 450 nm employing 630 nm as a reference. The cut-off is calculated as the mean OD of 4 negative controls + 0.2, and performance limits are applied that were established when the assay was developed and form part of routine QC monitoring. Validation of the assay was performed employing a panel of oral fluid specimens whose HIV status had been determined employing oral fluid specimens first characterised with the commercial Abbott Murex Diagnostics' GACELISA HIV 1+2 EIA (VK61). The validation panel comprised 194 anti-HIV-1 positive oral fluid specimens collected in the course of surveillance activities during 2001–2002 and 653 anti-HIV negative specimens collected during 2001 and 2003. The in-house GACELISA method detected 191/194 anti-HIV positive specimens, equivalent to a sensitivity of 99.5% (95% CI 96.7–100%). The falsely negative specimen was retested and found to be weakly reactive (OD/CO 1.21). Of 654 anti-HIV negative specimens all but 3 were non-reactive, a specificity of 99.5% (95% CI 98.5–99.9%). While all anti-HIV positive specimens were, by definition, detected by the commercial VK61 GACELISA HIV 1+2 EIA, it generated 7 false positive reactions, a specificity of 98.9% (95% CI 97.7–99.5%). Specimens that were anti-HIV reactive on initial testing were re-tested using the GACELISA. All repeatedly reactive samples and several of those that were reactive only on initial testing were examined by a modified Western Blot procedure (HIV Blot 2.2, Abbott Diagnostics, Maidenhead, UK).
For Anti-HIV testing an in-house version of the GACELISA HIV 1+2 enzyme immunoassay was employed  as the commercial kit is no longer manufactured. In local validation studies it had been shown to have an accuracy equivalent to that of the commercial GACELISA. Specimens that were anti-HIV reactive on initial testing were re-tested using the GACELISA. All repeatedly reactive samples and several of those that were reactive only on initial testing were examined by a modified Western Blot procedure (HIV Blot 2.2, Abbott Diagnostics, Maidenhead, UK).
Rapid HIV testing
Rapid HIV testing was conducted using a parallel rapid testing algorithm. Capillary blood samples collected using single use disposable lancets were dropped into the sample ports of the rapid test devices according to the manufacturers' instructions. An additional DBS was taken from each participant for anti-HIV confirmatory testing (protocol as above) at a later date to estimate the diagnostic accuracy of the algorithm.
Two rapid HIV tests, Uni-Gold Recombinant HIV-1/2 (Trinity Biotech, Bray, Ireland) and Determine HIV-1/2/O (Abbott Laboratories, Abbott Park, IL, USA) were employed simultaneously to avoid the need to take a second blood sample should the first test be reactive. These tests have undergone evaluation by the World Health Organization; both had 100% sensitivity, with the Determine test having 99.4% and Uni-Gold 100% specificity. A detailed protocol was followed which incorporated the manufacturers' instructions for use. Testing was conducted in a small temporary laboratory established within the Belgrade study site, and each test result was read by two trained technicians.
Additional file 1 describes the rapid testing algorithm. Reactivity of at least one rapid test result (two red bars, one in the "Test" and one in the "Control" window) indicated the possibility of HIV infection and the participant was referred to the Belgrade AIDS centre for appropriate confirmatory testing. Blood samples that were concordantly non-reactive, or where one test was non-reactive and the other invalid, were interpreted as indicating no evidence of HIV infection, and in which case no referral was made. In addition, the intensity of each reaction was recorded on a scale of one to four (no reaction; weak reaction; moderate reaction; strong reaction), as a quality check.
Characteristics of the two samples of IDUs were compared, and the Respondent Driven Sampling Analysis Tool (RDSAT) v. 5.0.1 http://www.respondentdrivensampling.org was used to weight the sample to control for differences in network size, homophily (the tendency for people to preferentially recruit others similar to themselves) and recruitment patterns, to provide population based estimates and 95% confidence intervals (CI) of IDU characteristics in each city. Correct implementation of RDS methodology (eg collection of participant's network sizes, and tracking who recruited whom) and analysis with RDSAT, provides representative estimates of the population sampled.
We then evaluated predictors of anti-HCV positivity using univariable and multivariable logistic regression, considering the following variables: age group, sex and education; injecting drug use characteristics (duration of injecting, frequency of injecting, main drug injected, frequency of use of new needles/syringes, main source of new needles/syringes); injecting risk behaviours and ever having engaged in sex work; history of arrest, imprisonment, drug treatment and HCV testing. As a result of incorrect programming of ACASI, which allowed some questions to be skipped without a value being entered, several variables had ≥ 5% missing values: duration of injection (20%); frequency of injection (9%); main drug injected (9%); number of times using a needle/syringe before disposal (21%); main source of new needles/syringes (9%); ever injected with used needles/syringes (12%); injected with used needles/syringes in the last four weeks (19%); shared injecting paraphernalia (9%); ever engaged in sex work (16%); and ever detained or arrested by the police (36%). Therefore multiple imputations were used in multivariable models assuming the data were missing at random to avoid excluding different injectors from different analyses. This involved creating or imputing plausible values for missing observations reflecting the uncertainty about the non-response model. Fifteen imputations were run on the missing data. The imputed datasets were combined and analysed together to allow for the uncertainty in the imputation model to be taken into account. Multivariable models were adjusted for robust standard errors as there was evidence of clustering by recruitment network. Statistical significance was assessed using Wald tests. All analyses were conducted using Stata 10 (Stata Corp, College Station, Texas).
Multivariable models were also adjusted for population weights on key demographic indicators to compensate for recruitment biases. Population weights adjusting for recruitment biases by education and sex were used for Belgrade, and age and education for Podgorica. Population weights were calculated using RDSAT v. 5.0.1 with anti-HCV status as the outcome variable.
Ethics approval was granted in the United Kingdom by the Charing Cross Research Ethics Committee. In Belgrade, the Ethics Committee of the Medical Faculty at the University of Belgrade approved the study and, in the absence of a suitable research ethics committee in Podgorica, approval for the study was granted by the Montenegrin Ministry of Health. In addition the Serbian Ministry of Health granted specific approval for the collection of biological specimens at study sites, and the rapid testing algorithm.