Quantifying unmet prevention needs among MSM in Europe through a multi-site bio-behavioural survey

Introduction The HIV epidemic represents an important public health issue in Europe particularly among men who have sex with men (MSM). Global AIDS Monitoring indicators (GAM) have been widely and jointly promoted as a set of crucial standardised items to be adopted for monitoring and responding to the epidemic. Methods The Sialon II study, implemented in 13 European cities (2013-14), was a complex multi-centre integrated bio-behavioural cross-sectional survey targeted at MSM, with a concomitant collection of behavioural and biological (oral fluid or blood specimens) data. Rigorous sampling approaches for hard-to-reach populations were used (time-location sampling and respondent-driven sampling) and GAM indicators were calculated; sampling frames were adapted to allow weighted estimates of GAM indicators. Results 4,901 MSM were enrolled. HIV prevalence estimates ranged from 2.4% in Stockholm to 18.0% in Bucharest. When exploring city-level correlations between GAM indicators, prevention campaigns significantly correlated with levels of condom use and level of HIV testing among MSM. Conclusion The Sialon II project has made an important contribution to the monitoring and evaluation of the HIV epidemic across Europe, integrating the use of GAM indicators within a second generation HIV surveillance systems approach and in participatory collaboration with MSM communities. It influenced the harmonisation of European data collection procedures and indicators via GAM country reporting and contributed essential knowledge informing the development and implementation of strategic, evidence-based HIV prevention campaigns for MSM.


Introduction
The HIV epidemic represents an important public health issue in Europe, particularly among men who have sex with men (MSM). For example, across the entire European Region, one new HIV diagnosis case in every four is attributable to MSM. Moreover, in 15 The picture is arguably even more problematic in some Eastern European and Central Asian countries where, although the HIV epidemic among this population is often similar, the existence of relatively more stigmatising environment(s) are probably less conducive to the reporting of data that could potentially deepen the understanding of mechanisms of HIV transmission among MSM [5].
In order to target prevention strategies effectively and to monitor their impact at a local/regional and country level, a better understanding of the epidemiological patterns and identification of the most affected subpopulations, are key enabling factors in tackling the multifaceted HIV epidemic. In Europe this particularly relates to MSM and there is a clear need for a harmonised collection of reliable and comparable data on epidemiology and coverage of prevention measures in this population.
Consequently, international agencies (namely, the Joint United Nations Programme on HIV/AIDS (UNAIDS), WHO and ECDC) have called for countries to use robust surveillance and monitoring systems that adopt common and standardised indicators [6]. A key part of this international effort for harmonisation is the promotion and implementation of second generation HIV surveillance systems (SGSS), which collect and link biological and behavioural data [6]. The Global AIDS Monitoring indicators (GAM) are part of the SGSS and comprise a set of standardised items widely and jointly promoted by the WHO and UNAIDS [7,8].
However, despite concerted efforts by these international agencies, the implementation of such a methodological approach is patchy and requires both adoption and strengthening across countries [8]. Although country reporting to the UNAIDS GAM has improved consistently over the years, to date there remains considerable variability in response rates [8].
A recent review of the GAM reporting process has highlighted a lack of data relating to key populations. For example, in 2012, GAM data on key populations were reported in approximately only 30% of cases [9].
A standardised set of GAM indicators is crucial given their role in providing specific data and information to monitor the implementation of the Sustainable Development Goals (SDG) and the UNAIDS 90-90-90 strategy, recently endorsed by the European Commission (EC) Communication on 'Next steps for a sustainable European future' [10]. The EC has cofunded several projects in the area of HIV/AIDS, two of which have aimed specifically to implement a joint survey across different EU/EEA countries adopting the main principles of SGSS and the GAM approach as cornerstones. The two projects are the Sialon project [11] and the more recent Sialon II project [12,13]. In particular, in the Sialon II project a total of 13 countries with very different cultural and social environments with 30 institutions including public health institutions and non-governmental organisations (NGOs) were involved. The meaningful participation of Lesbian-Gay-Bisexual-Trans-plus (LGBT +) communities in all participating countries has been key in designing and implementing the study.
The value of the Sialon II project lays in the sampling approach (Time-Location Sampling (TLS) and Respondent-Driven Sampling (RDS)) and the use of GAM indicators. The methodology adopted allowed the weighted estimation of GAM indicators, presented in this manuscript. To our knowledge, this is the first paper delivering weighted estimates for GAM indicators in a large number of European cities set within the framework of a SGSS specifically targeting MSM.
The objectives of this paper are to present weighted estimates of the GAM indicators among MSM based on the Sialon II bio-behavioural survey implemented in 13 European cities  and to discuss the usefulness of these GAM indicators in monitoring the HIV epidemic and responses across EU countries.

Methods
Detailed descriptions of the study procedures and methods have been published elsewhere [12]. Here we present a short overview of the main methodological aspects.

Study design
The Sialon II study was a complex multi-centre integrated bio-behavioural cross-sectional survey with a concomitant collection of behavioural data and biological data (oral fluid or blood specimens).

Setting
The survey was implemented in 13 European cities. The decision to use TLS or RDS in each study site was based on preliminary formative research and organisational issues. TLS was adopted in nine cities: Brussels (Belgium), Sofia (Bulgaria), Hamburg (Germany), Warsaw (Poland), Lisbon (Portugal), Ljubljana (Slovenia), Barcelona (Spain), Stockholm (Sweden) and Brighton (United Kingdom (UK)). The setting for data collection included social and/or commercial venues and cruising settings preliminarily identified through formative research [14] and then selected randomly for data collection sampling calendars. RDS was used in four cities: Verona (Italy), Vilnius (Lithuania), Bucharest (Romania) and Bratislava (Slovakia). Regarding the latter, enrolment was based on the individuals' social network and for the data collection locally accredited healthcare facilities (e.g. a hospital) were used. Data collection for all sites took place from April 2013 to November 2014.

Sample size
The sample size estimation was carried out based on the results from the former Sialon I project and other available studies [11]. Based on assumptions of HIV prevalence in the target population of at most 15%, a precision of 5%, a significance level of 95% and a design effect of 2.0 provided a random clustered sample size calculation of 392 MSM per study site. Taking into account the possibility of invalid samples, a final target of 408 MSM per city for TLS and 400 for RDS was planned.

Participants
Inclusion criteria were having had any kind of sex with another man during the previous year before enrolment, providing informed consent and agreeing to donate either an oral fluid (TLS) or blood specimen (RDS).
The exclusion criteria were being younger than the legal age of consent (18 years old) or having already participated in the study.

Data sources/measurement
Behavioural data Behavioural data were collected through a pen-andpaper self-administered questionnaire. Core items were developed in line with the GAM indicators [7]. To allow for sampling weight calculations, additional items were included in the questionnaires on the venues attendance (TLS) or on network size (RDS).

Biological data
Biological specimens were obtained from participants of both study arms (TLS/RDS). In cities where TLS was used, specimens were tested for HIV antibodies using Genscreen HIV 1/2 version 2, Bio-Rad (Marne la Roquette, France). A total IgG antibodies ELISA test Human IgG ELISA Kit 1x96, Quantitative/Immunology Consultants Laboratory was also used for oral fluid (OF) sample testing suitability and quality control. All HIVreactive samples were re-tested with Vironostika HIV Ag/Ab, Biomerieux (Marcy-l'Étoile, France). Samples reactive to the first ELISA HIV test, but negative to the second, were classified as negative.
In cities where RDS was used, blood specimens were collected and processed for serum extraction according to the respective national guidelines for safety and quality assurance. Serum samples were tested with a HIV fourth generation ELISA/CLIA screening test. A Western Blot test was used to confirm the positive cases.

Variables
The variables used for the present analysis included the GAM indicators suggested for MSM target population [15]. All proposed items included in the GAM guidelines for MSM were used in the survey questionnaire. Numerators and denominators were defined as seen in Table 1.

Statistical methods
The analysis was carried out according to the GAM indicator guidelines [15]. For all indicators, estimates were carried out with the following age disaggregation: < 25 years old and ≥ 25 years old. Age was calculated on the basis of the self-reported year of birth. Analyses were conducted using STATA Version 14.1 (StataCorp, College Station, Texas, United States). To allow calculation of the sampling weights, a specific procedure was devised on the basis of previous publications and methodological guidelines [12,16,17].
For the TLS survey, individual weights were assigned as the inverse of the product of the following: (i) the probability of the participant being at the sampled venue given he was at the sampled venue type (number of visits to sampled venue/number of visits to all types of venues); (ii) the length of the sampling time (out of all Venue-Day-Time units on the particular day) and (iii) the proportion of sampled individuals during the event in relation to the estimated number of visitors during the sampling event, a modification of the method proposed by Karon and Wejnert [17].
For the RDS survey, and in line with a RDS approach [18], weighted estimates were calculated using RDS Analyst (www.hpmrg.org), a suite of R commands developed by Handcock and colleagues (2015 RDS Analyst: Software for the Analysis of Respondent-Driven, Sampling Data, Version 0.52). Gile's Sequential Sampler approach was used for calculating the sampling weights. This approach is based on the inclusion probabilities of members of the sample which are based on reported network sizes [16]. This method is recommended when the sample is a significant fraction of the target population. Therefore, in order to use this method, population size estimates were carried out for each city. The calculation was based on the total number of inhabitants for the city area and the expected percentage of MSM (according to the consensus among the project's scientists and according to the scientific literature, given there is currently no comprehensive and precise agreement among experts on MSM population size estimations within the general population) [19,20]. All point estimates were reported with their respective sample size, 95% Confidence Intervals (CI), and estimated design effect.

Ethics
All procedures adopted in the present study were in line with the 1964 Helsinki declaration and its amendments. Survey protocols were approved by the appropriate ethics committee in each participating city as well as by both the WHO Research Project Review Panel (RP2) and the WHO Research Ethics Review Committee (ERC). The name or any other identifier of the MSM enrolled in the study was not collected. All respondents were entitled to collect their test result at a nominated centre indicated to the participant during study enrolment. In case of a positive result, further testing, counselling, clinical follow up and ARV treatment were provided in line with the respective national guidelines.

Results
A total of 4,901 MSM were enrolled across the 13 participating cities. In TLS study sites, 3,596 participants were enrolled, while in RDS sites a total of 1,305 participants were enrolled. Participants enrolled per enrolment method/city, age mean, Min-Max, age group, are shown in Table 2. A detailed description of the sample is available in the Sialon II project report [13].

GAM 1.11 (Prevention programmes)
In eight of the nine cities where TLS was implemented, more than half of respondents answered positively to both questions ( ). It should be noted that for some cities, the number of participants in the younger age group was low and therefore the precision of the estimates is reduced.

GAM 1.13 (HIV testing)
The level of HIV testing as per GAM guidelines is reported in

City-level correlations between GAM indicators
When exploring city-level correlations between GAM indicators (Figure) a significant correlation was found between prevention programmes indicator (defined as condom availability and testing site knowledge) and HIV testing (correlation coefficient 0.52, p-value = 0.006).
In addition, a correlation was found between preventive programmes indicator and condom use (correlation coefficient 0.45, p-value = 0.022). Data suggest that the higher the prevention programmes indicator the higher the level of condom use and testing among MSM, in both age groups.

Discussion
HIV remains a public health priority in the EU and the data from the Sialon II study has made an important contribution to the monitoring and evaluation of the HIV epidemic across Europe, by integrating the use of GAM indicators within a second generation HIV surveillance systems approach and in participatory collaboration with MSM communities. It has influenced the harmonisation of European data collection procedures and indicators via GAM country reporting and contributed essential knowledge informing the development and implementation of strategic, evidence-based HIV prevention campaigns for MSM.

Table 3b
GAM indicators (weighted estimates) among MSM, design effect and number of participants, by city and age group, European Union cities (n = 13) run by an NGO before and during the data collection period), younger MSM showed the lowest level. These estimates are consistent with the literature and suggest the need for targeted prevention programmes tailored to locations and communities that can also accommodate for the needs of sub-populations, such as young MSM, MSM who are tourists and bisexuals [21,22].
Data focusing on the GAM 1.12 (Condom use) indicate that in the majority of the surveyed cities, condom use with any kind of partner was lower for younger men. However, the sample sizes for younger men were small for several cities and the association between age and condom use was not consistent across all study cities. In another publication based on the same Sialon-IIdataset, authors looked at condom use from another perspective e.g. any anal intercourse without a condom during the previous 6 months and not just from the last sexual encounter. The authors reported that almost half of the HIV-uninfected individuals reported condomless anal intercourse (CLAI). This was reported slightly more often by men living in Central European study cities and more frequently with steady partners compared with non-steady partners [23].
Based on the estimates of this GAM indicator there is a clear need to either increase condom use among younger men or to complement prevention strategies by providing meaningful access to other similarly effective HIV prevention tools such as HIV pre-exposure prophylaxis. Despite the fact that condom and lubricant distribution is often considered a simple and somewhat naïve approach to facilitate condom use, it is commonly acknowledged in the literature that fear of disapproval and discrimination by healthcare providers can deter gay, bisexual and other MSM from accessing mainstream health services [18]. Indeed, this limitation, could reduce access to free condom distribution, as well as low threshold HIV and STIs testing in healthcare settings [24].
A study by Marcus et al. [25] modelling the relationship between unprotected anal intercourse (UAI) and HIV disclosure with the same study dataset, found that among those respondents being aware of being HIV positive, condom use with steady partners was higher than among HIV negative men. However, condom use with non-steady partners was also lower. Men unaware of being infected with HIV reported the lowest condom use with non-steady partners [23,25]. In several cities there were differences between the two age groups with regards to receiving an HIV test within the last 12 months and knowing the results. For instance, participants from Brighton, Brussels, Lisbon, and Warsaw all reported differences of greater than 20% for the two age groups; however, considering the time span is limited to the last 12 months, the differences cannot be directly attributed to the age effect where, for example, older individuals may have had an increased opportunity to be tested over time. Furthermore, older participants were tested more frequently and within a shorter timeframe than the younger men, suggesting that increasing access to culturally sensitive HIV counselling, testing and antiretroviral therapy for MSM, found to have HIV, is an urgent health priority particularly for the younger generations. The current levels of HIV testing are insufficient to link gay, bisexual and other MSM with appropriate healthcare support shortly after acquiring HIV infection. Therefore, currently, testing frequencies can remain insufficient to effectively reduce the period of infectiousness of people who newly acquire HIV.
Alternative approaches such as the use of point of care tests (PoCTs) for HIV and STIs in low threshold community testing and LGBT venue-based testing, home collection testing, and HIV self-testing may represent effective approaches to increase diagnosis and linkage to care. The problem of late diagnosis reflects a lack of access to and uptake of HIV testing and counselling services in many countries [26]. A late diagnosis also means that a person has remained unaware of their HIV status for an indeterminate length of time, thus increasing the risk of transmitting the virus. The most recent surveillance data showed that despite significant efforts dedicated to the prevention and control of HIV, the rate of new HIV diagnoses has not substantially declined in the EU/EEA, but it has increased substantially over the last decade in the European Region.
Although HIV prevalence as an epidemic indicator is not a good parameter of the HIV infection spread dynamic, it can be helpful to provide an indirect picture of the epidemic history and patterns for some cities. HIV prevalence estimates (GAM 1.14), as measured by testing biological specimens ( [26], but also may be due to an increase in active offering of HIV and STI's testing over the years. A higher treatment coverage and higher percentage of HIV-positive MSM with undetectable viral load was attained in some Western European cities (Brussels, Hamburg, Brighton and Verona), indicating that when the service provision is proactive and the treatment widely available, the link between testing and treating can effectively influence the HIV epidemic.

Generalisability and limitations
To our knowledge, this is the first paper presenting weighted estimates produced as a result of a standardised collection of GAM indicators for MSM in a large number of European cities adopting a common SGSS approach. The use of this approach (SGSS and GAM Indicators) and the active participation of key LGBT community stakeholders in the project's design and implementation, represent an asset that provides potentially usable data for both public health authorities and NGOs in each partner country.
The use of TLS and RDS methodologies within the context of a bio-behavioural survey using a participatory approach have allowed hidden and different subgroups of MSM to be reached; groups that are usually more difficult to access through surveillance studies [28]. Nevertheless, there are some limitations that should be taken into account when interpreting the results.
First, data can only be generalised to the particular MSM attending the gay venues in each study site (in the case of TLS survey) and only to those MSM socially linked to the gay community for each specific site (in case of the RDS survey). It has also been shown in other studies that these two sampling methods can result in different sample characteristics and the differences may persist even after applying weighting corrections [28].
Second, the generalisability of the findings may also be limited by contextual factors not measured in this survey (e.g. legislation and social norms) and an ecological fallacy cannot be excluded.
Third, an additional source of bias limiting generalisability relates to the self-reported behavioural data. This is of course an issue common to all surveys covering the self-reporting of sensitive information. However, the anonymity of the data collected and the self-administration of the questionnaire, with the careful design of the questionnaire items developed and validated in all cities with the involvement of the LGBT community, may well have reduced any social desirability effect [29]. It is also worth noting that no difficulties or limitations were reported by either respondents or data collectors with regards to the use of the GAM indicators (both TLS and RDS) in terms of interpretation and utility of the items and related indicators.
Fourth, the percentage of participants who reported ever having injected drugs ranged from 1.2% in Bratislava to 19.3% in Bucharest. In the latter, estimates might be difficult to be generalised due to this sub-population of injecting drug users (IDU)-MSM and to the fact that the target sample was not reached.
Fifth, the precision of the estimates were in some cases not optimal, particularly for the group of younger MSM, due to relatively small sample sizes (< 50) and the potential for sampling bias in some cities.
Finally, an isolated limitation to the validity of the survey relates to Sofia (Bulgaria) where, due to an incorrect translation in the items questionnaire related to GAM 1.13, the indicator could not be estimated.

Conclusions
The Sialon II project and the data generated through its implementation represent a collaborative and scientifically robust contribution to the monitoring and evaluation of the HIV epidemic across Europe, integrating the use of GAM indicators within a SGSS approach with active community involvement. The data collected provide new evidence for appropriate planning of HIV prevention campaigns among MSM, clearly responding to the urgent need of concerted use of common indicators, with a particular focus on most at risk populations such as MSM [9]. The project has actively contributed to: (i) common procedures piloting (including most-advanced sampling methods, standards research algorithms, advanced laboratory diagnostics); (ii) harmonised data collection, involving experts from different institutions and with different backgrounds; and (iii) GAM country reporting, with specific reference to MSM.