**3.2.3 Statistical analysis**

Mean, median, standard deviation, quartiles and inter-quartiles were calculated and proportions with 95% confidence intervals (CI) were adopted for all variables and indicators. With regard to the UNGASS indicators calculation, specific reference to the UNGASS-WHO publications was made. Finally, STATA 11 survey commands suite was used.

The SIALON Project: Report on HIV Prevalence

test according to HIV prevalence are presented in Table 1.

**4.1.2 Questionnaires and oral fluid samples collected** 

Table 1. PPV and NPV according to prevalence

**4.1 General description of the sample** 

**4.1.1 Numbers of subjects recruited** 

**4. Results** 

data collection.

and Risk Behaviour Among MSM in Six European Cities 213

study of Bio-Rad OF testing comparing serological testing involving 37 HIV positive patients and 35 controls per country was carried out according to Commission decision of 7 May 2002 on common technical specifications for in vitro medical devices. EIA on oral fluid samples from 259 of the 263 HIV positive subjects were positive, giving a sensitivity of 98.5% (CI 96.2- 99.6). All 233 controls were found negative for HIV in oral fluid and no false positives were detected (100% specificity; CI 98.4-100). The positive and negative predictive values of the OF

> Prevalence 5 % 15 % PPV 100 % 100 % NPV 99.9 % 99.7 %

In this section, the description of the sample is provided in terms of number of MSM recruited, questionnaires and oral fluid samples collected, and characteristics of the sample.

A total of 2,592 subjects were recruited as follows: 408 MSM in Prague, 185 in Athens, 405 in Verona, 398 in Bucharest, 394 in Bratislava, 401 in Ljubljana and 401 in Barcelona. This study does not include the analysis of data from Athens as the Greek partner did not achieve the

Table 2 shows the number of questionnaires and the number of oral fluid samples gathered during the data collection period for each country. Both the number of valid and invalid questionnaires and OF samples are given. From a general point of view, the percentage of invalid questionnaires is low, with the highest percentage in Ljubljana (0.7%). In three cases the percentage is 0% (Athens, Bucharest and Bratislava). With regards to the OF samples, in

Table 2. Number of questionnaires and OF samples collected and percentage of valid samples
