2.1. Study design

We utilised the Irish dataset from the European MSM Internet Survey (EMIS) 2017. EMIS-2017 was an online, anonymous, internationally-promoted survey among gbMSM across 50 countries in Europe, and was adapted from a previous survey, EMIS-2010 [32]. Novel questions were added to EMIS-2017, and were based on topical developments in certain areas which affect the overall health and wellbeing of gbMSM. For example, questions relating to the availability and use of pre-exposure prophylaxis (PrEP) were included in EMIS-2017 [33]. The survey, which was available in any one of 33 different languages, began on 18^th October 2017 and ended on 31^st January 2018. Overall, 409 items were included in the survey, and a broad range of areas were explored, including sociodemographic characteristics, mental health, sexual health, and substance use [34].

2.2. Study population and sampling

In order to meet inclusion criteria for this study, participants had to initially give informed (written) consent. Respondents were required to identify as either a man and/or a trans man, and had to be sexually attracted to, or sexually active with, men. For inclusion in the Irish dataset specifically, candidates had to be resident in the Republic of Ireland, regardless of their native country (gbMSM residing in Northern Ireland were excluded). In addition, the study population were required to be over the legal age of sexual consent in their respective countries, i.e. at least 17 years of age in Ireland. However, we restricted our study sample to those aged 18+ years in accordance with the Social Research Ethics Committee (SREC) policies at University College Cork (UCC).

In order to recruit participants, a multi-media strategy was utilised [35], including:

1. Statutory and non-governmental organisation (NGO) websites for gbMSM.
2. Geospatial networking/contact applications, including Scruff, Gaydar, and GROWLr.
3. Multiple social media platforms, such as Instagram, Twitter, and Facebook.

There was also widespread promotion of EMIS-2017 through lesbian, gay, bisexual, and transgender+ (LGBT+) community organisations, along with the issuing of a press release for print media. Finally, business cards and posters were distributed at gay social/community venues.

2.3. Measurements

2.4. Statistical analysis

Descriptive statistics were used to compare baseline characteristics of the study sample. Absolute numbers and percentages were reported for categorical variables. Pearson’s chi-square (χ²) was utilised for non-parametric tests of bivariate associations. Fisher’s exact test was employed for categorical covariates with a count of <5 in >20% of cells (i.e. Table 2) [43].

Single-predictor logistic regression analysis was initially conducted to calculate crude measures of association between independent variables and dichotomous outcomes. Subsequently, multivariable-adjusted logistic regression analysis was used in order to identify which independent variables were robust to adjustment with an array of potentially confounding covariates. All independent covariates included in these models were adjusted for one another. As we were dealing with cross-sectional data, our intention was to explore associations, without determining any causal relationships. Multicollinearity between explanatory variables was assessed using Spearman’s rank correlation coefficient, wherein all values were below the often-cited threshold of 0.40, indicating that multicollinearity was unlikely to be an issue [44]. Furthermore, when missing data for a particular categorical variable are small (i.e. <5%), biases and loss of statistical power are likely to be relatively inconsequential [45]. Thus, where 5% or more data were missing for an independent categorical variable, indicator variables were synthesised and included in the regression models.

Our analysis was stratified based on HIV status, with a separate sensitivity analysis of HIV-positive respondents using descriptive statistics and regression analyses for RDU and SDU outcomes. Variables pertaining to HIV testing history and current pre-exposure prophylaxis (PrEP) use were not included in these models, as all men were diagnosed HIV-positive, which obviated the requirement for their inclusion. This approach was determined a priori in conjunction with multi-agency groups who work to promote the overall health and wellbeing of gbMSM.

All analysis was undertaken using Stata version 17.0 (StataCorp., TX, USA). Two-sided significance tests were utilised throughout, and all p-values were tested against a pre-set alpha level of significance which was set at 5%.

2.5. Ethical considerations

The Observational Research Ethics Committee at the London School of Hygiene and Tropical Medicine (LSHTM) originally granted ethical approval for EMIS-2017. For this study specifically, the Social Research Ethics Committee (SREC) in University College Cork (UCC) granted approval, wherein LSHTM and UCC both approved a Data Transfer Agreement (S2 Appendix).

3.1. Multi-predictor logistic regression analysis

3.1.1. RDU among gbMSM without a diagnosis of HIV.

In our initial multivariable model (Table 3), the odds of engaging in RDU were higher in younger age groups (vs. aged 40+ years) (18–24 years; AOR 2.96, 95% CI 2.05–4.28, 25–39 years; AOR 1.66, 95% CI 1.27–2.16), and in gbMSM who self-identified as bisexual (vs. gay/homosexual) (AOR 1.74, 95% CI 1.22–2.48). Increased odds of RDU were observed among men who disclosed their sexual orientation to a larger number of family/close friends (vs. none) (less than half; AOR 1.72, 95% CI 1.02–2.91, more than half; AOR 2.29, 95% CI 1.31–4.00, all or almost all; AOR 3.20, 95% CI 1.88–5.46), and among those who engaged in CAI with non-steady male partners in the previous 12 months (vs. none) (1–2 partners; AOR 1.79, 95% CI 1.34–2.38, 6+ partners; AOR 1.79, 95% CI 1.18–2.71). Any RDU in the previous year was significantly more likely among gbMSM who self-reported a bacterial STI diagnosis within the same timeframe (vs. none) (AOR 1.53, 95% CI 1.12–2.10), lived in Dublin (vs. outside Dublin) (AOR 1.47, 95% CI 1.17–1.83), and screened positive for a mild degree of anxiety/depression (vs. none) (AOR 1.45, 95% CI 1.14–1.84).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Factors associated with engagement in recreational drug use (excluding poppers) in the previous 12 months among gbMSM without a diagnosis of HIV who took part in EMIS-2017 and were included in the study sample (n = 1,898)*.

https://doi.org/10.1371/journal.pone.0288171.t003

3.1.2. SDU among gbMSM without a diagnosis of HIV.

As shown in Table 4, gbMSM who engaged in SDU in the previous year were more likely to self-identify as bisexual (vs. gay/homosexual) (AOR 1.75, 95% CI 1.05–2.92), report any number of non-steady CAI partners in the previous year (vs. none) (1–2 partners; AOR 3.16, 95% CI 2.05–4.88, 3–5 partners; AOR 2.50, 95% CI 1.47–4.26, 6+ partners; AOR 3.79, 95% CI 2.23–6.43), self-report a bacterial STI diagnosis within the same timeframe (vs. none) (AOR 2.15, 95% CI 1.48–3.12), and live in Dublin (vs. outside Dublin) (AOR 1.50, 95% CI 1.07–2.10). Furthermore, the odds of SDU were significantly higher among gbMSM who reported group sex in their last sexual encounter (vs. none) (AOR 1.73, 95% CI 1.15–2.60), and in gbMSM who had attended for HIV testing but had a negative test result (vs. never tested) (AOR 2.02, 95% CI 1.20–3.39). GbMSM who were students were less likely to report SDU (vs. employed gbMSM) (AOR 0.50, 95% CI 0.29–0.86).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Factors associated with engagement in sexualised drug use in the previous 12 months among gbMSM without a diagnosis of HIV who took part in EMIS-2017 and were included in the study sample (n = 1,898).

https://doi.org/10.1371/journal.pone.0288171.t004

3.1.3. RDU among gbMSM with diagnosis of HIV.

In Table 5, higher odds of RDU were found among HIV-diagnosed gbMSM who screened positive for severe anxiety/depression (vs. none) (AOR 5.86, 95% CI 1.21–28.37), and among those who disclosed their sexual orientation to all or almost all of their family/close friends (vs. less than half) (AOR 5.69, 95% CI 1.38–23.56).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Factors associated with engagement in recreational drug use (excluding poppers) in the previous 12 months among gbMSM with a diagnosis of HIV who took part in EMIS-2017 and were included in the study sample (n = 141).

https://doi.org/10.1371/journal.pone.0288171.t005

3.1.4. SDU among gbMSM with a diagnosis of HIV.

Finally, in Table 6, greater odds of SDU were found among gbMSM with a positive diagnosis of HIV who resided in Dublin (vs. outside Dublin) (AOR 5.00, 95% CI 1.13–22.12), and among those who reported 6+ non-steady CAI partners in the previous year (vs. none) (AOR 17.42, 95% CI 1.37–221.78).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Factors associated with engagement in sexualised drug use in the previous 12 months among gbMSM with a diagnosis of HIV who took part in EMIS-2017 and were included in the study sample (n = 141).

https://doi.org/10.1371/journal.pone.0288171.t006

4.1. Principal findings

GbMSM in our sample reported a high prevalence of past-year RDU (51.8% and 40.9% among those with and without a diagnosis of HIV, respectively). These estimates are in stark contrast to the 2019/2020 prevalence estimates of RDU among the general male population in Ireland (12.3%) [7]. Notably, our analysis was stratified based on HIV status because previous literature has suggested that HIV-positive gbMSM can systematically differ from HIV-negative men in certain aspects. For example, HIV-positive gbMSM are typically more likely to engage in substance use [19, 46]. They also tend to have a high level of formal engagement with healthcare services due to their diagnosis which, by default, presents opportunities for health promotion and preventive messaging from trusted healthcare professionals [47]. Finally, current PrEP use (an independent covariate included in our study) was specifically used by non-diagnosed positive gbMSM only. Consequently, stratification by HIV status was required to prevent the introduction of systematic error in our multivariable models.

Among gbMSM without a prior diagnosis of HIV, those who were younger were more likely to engage in RDU in the previous year. GbMSM who resided in Dublin were almost one and a half times more likely to report RDU. These findings are consistent with the results of prior studies [30, 39]. Among the general male population in Ireland, 15–24 year olds represent the age category with the highest burden of past-year drug use [6, 7], and our study revealed a similar pattern (albeit amongst 18–24 year olds). Conversely, while previous research suggests that SDU (specifically chemsex) is most common in the late twenties, thirties, and early forties among gbMSM [21, 37], our study did not elucidate any significant relationship between age and SDU. However, EMIS-2017 omitted GHB/GBL from its definition of SDU, and included a number of other drugs, such as cocaine and MDMA. Both of these drugs are common at the population-level, particularly among those aged 15–24 years [6], and perhaps their inclusion may have attenuated our findings towards null. Furthermore, gbMSM residing in Dublin had a greater tendency to engage in SDU, irrespective of HIV status, which is consistent with the most recent national study of Irish gbMSM, which analysed data from 2015 [30]. SDU often occurs in private homes, and can be facilitated through smart-phone geospatial networking applications. Therefore, in urban centres such as Dublin, large groups of gbMSM may congregate at relatively short notice [25]. This could potentially account for the higher odds of SDU among respondents who were residing in the capital city.

We observed increased odds of both RDU and SDU among non-diagnosed positive gbMSM who reported CAI encounters with numerous non-steady male partners in the previous 12 months. Non-diagnosed positive gbMSM who self-reported a bacterial STI diagnosis within the same timeframe were 53% and 115% more likely to engage in recent RDU and SDU, respectively. In addition, those who reported that their last sexual encounter involved group sex had a 73% increased odds of engaging in SDU within the previous 12 months. In Ireland, gbMSM account for less than 10% of the overall male population [48], yet it is estimated that they make up 86% and 65% of all incident diagnoses of syphilis and gonorrhoea, respectively [49, 50]. Non-diagnosed gbMSM who presented for HIV testing (and tested negative) were approximately twice as likely to engage in SDU in the previous year. These findings are consistent with previous studies [19, 22, 30, 51, 52]. Indeed, it is beneficial that these men are more proactive about their sexual health by engaging with healthcare services. Furthermore, among HIV-positive gbMSM in our study, 96% had an undetectable viral load due to treatment with ART. This is an important finding for these individuals in relation to their overall health and wellbeing, and it also reduces the likelihood of onward transmission to HIV-negative gbMSM with whom they may engage in sexual activities (i.e. undetectable = untransmittable [U = U]). However, our study found no significant association between current PrEP use and engagement in RDU and/or SDU among gbMSM without a diagnosis of HIV, although only 78 participants reported current PrEP use. Importantly, the provision of PrEP has increased dramatically in Ireland due to the national “HIV PrEP” programme which was rolled-out in November 2019 (i.e. after EMIS-2017 was conducted) [53]. Therefore, further research is warranted in Ireland as PrEP has become more widely available, as it may play a key role in curtailing the incidence of HIV among both gbMSM and the general population. Of note, gbMSM account for over 50% of all incident HIV diagnoses annually in Ireland [54].

Increased odds of RDU were more likely among gbMSM who reported anxiety/depression. Specifically, non-diagnosed positive gbMSM who screened positive for a mild level of adverse mental health were approximately 50% more likely to engage in RDU, while HIV-diagnosed gbMSM who reported a severe level of adverse mental health were almost five times more likely to report SDU in the previous year. The exact reasons which underpin this disparity are likely multifactorial and complex. In 2022, a systematic review explored the determinants of depression among gbMSM living with HIV [55], and the authors concluded that HIV-related stigma likely plays a prominent role in the manifestation of poorer mental health outcomes among HIV-diagnosed gbMSM. Separately, in our study, those who were more open about their sexuality were also more likely to report RDU, wherein a gradient effect was observed (among gbMSM without a diagnosis of HIV). Although prior literature has highlighted similar findings in relation to drug use and adverse mental health outcomes [56, 57], greater disclosure of sexual orientation among gbMSM is usually an indication of self-acceptance and improved psychological adjustment [58]. It is usually the “rejection reactions” that are typically associated with drug use [59], and this may have contributed to the association we observe between greater “openness” and drug use among gbMSM in our sample. Conversely, another possible explanation for this association could be the long cultural history of drug use in the LGBT+ community, which may be motivated by the desire to experience pleasure, increase empathy and caring, and enhance community spirit [60]. Thus, the more “out” one is, the more likely they are to socialise in gay bars and immerse themselves in gay culture.

Finally, although we may draw comparisons between the prevalence of SDU among our study sample and in previous studies, it is important to consider the exact definition of SDU used, i.e. which specific set of drugs were included, and in what context. A common definition of SDU (specifically chemsex) refers to the consumption of certain drugs (typically crystal methamphetamine, mephedrone, GHB/GBL, and/or ketamine) by gbMSM, before and/or during sex [20–22]. However, an increasing number of sources now qualify cocaine, other stimulants (e.g. MDMA/ecstasy and amphetamine/speed), and new psychoactive substances (e.g. tryptamines) as “chemsex drugs” [19, 31, 61, 62]. Barrett et al. analysed the latest available data (2015) among a nationally-representative sample of Irish gbMSM, and found that 7% of gbMSM used one or more drugs associated with chemsex in the previous year [30]. However, a number of methodological differences from our study must be noted. Firstly, this study qualified crystal methamphetamine, mephedrone, GHB/GBL, and/or ketamine as drugs associated with chemsex. Additionally, this study only asked about the use of these “chemsex drugs” by gbMSM in general, as opposed to their specific consumption by gbMSM before and/or during sexual encounters. By contrast, our study refers to the consumption of MDMA/ecstasy, cocaine, amphetamine/speed, crystal methamphetamine, mephedrone, and/or ketamine in the sexual setting specifically. Therefore, caution should be exercised when comparing these prevalence estimates. Notably, using data from 2016, Frankis et al. estimated that among gbMSM who reside in Scotland, Wales, Northern Ireland, or the Republic of Ireland, 27% of those who consume one or more “chemsex drugs” (crystal methamphetamine, mephedrone, GHB/GBL, and/or ketamine) each year do not use these drugs in sexual encounters [21].

4.2. Strengths and limitations

This study has a number of strengths. Firstly, participants could elect to take the survey in any one of 33 different languages, which increased participation rates from non-English speaking respondents, thereby reducing selection bias. Latin American gbMSM (who may not speak English) are disproportionately impacted by HIV diagnoses in Ireland in comparison to their Irish contemporaries [54]. However, we could not extract meaningful data pertaining to participants’ ethnicities, apart from whether they were born in Ireland or not. This prevented us from assessing specific ethnic subgroups of interest. Secondly, EMIS-2017 was entirely anonymous, which may have increased the internal validity of the study, particularly in relation to sensitive questions (e.g. drug use, HIV/STI diagnoses, and CAI encounters). Previous studies have concluded that online anonymous surveys are the optimal method for assessing drug use [63, 64], and the reduction in social desirability bias in our study likely boosts the validity of our outcome data. Thirdly, we had relatively complete data on study respondents. Most of the variables in our regression models had <1% missing data. Only three variables had ≥5% missing data, and we synthesised indicator variables to control for this. Fourthly, approximately 70% of participants in the overall European dataset were recruited through social networking apps such as PlanetRomeo, Grindr, and Hornet [34]. In Ireland, this respective figure was only about 40% [35]. This was likely due to large community-led efforts with a concerted focus on recruiting gbMSM through gay bars or cafés, gay saunas, gay community centres, and other social venues [35]. Arguably, this may render our findings more generalisable to the wider gbMSM community in Ireland if it encompasses a wider selection of participants, rather than those who engage in the use of dating/networking apps, who tend to be younger, more technologically literature, and have a greater degree of disclose about their sexual orientation [65, 66]. Lastly, the large number of respondents increased our study’s statistical power. However, the smaller quantity of HIV-positive gbMSM in our sample likely introduced type 2 statistical error, which may have prevented us from detecting numerous significant associations [67].

Several limitations of this study merit discussion. Firstly, deciphering the exact temporal sequence of events is not possible, given our study’s cross-sectional design, which hinders our ability to draw any causal conclusions. As a corollary to this limitation, reverse causality must be considered, particularly for certain associations, such as that between drug use and depression/anxiety. Secondly, all data were self-reported, which may have led to information bias. Thirdly, convenience sampling was utilised in order to boost the participation rate which, in comparison to probability-based sampling, tends to enlist gbMSM with higher levels of drug use [68]. This may have inflated our prevalence estimates. Another significant limitation to our study is the exclusion of GHB/GBL in our definition of SDU. Future surveys (including EMIS) should endeavour to include GHB/GBL in questions which specifically relate to SDU/chemsex. Furthermore, we collapsed categories for certain covariates to circumvent the requirement for a Bonferroni correction, while still aiming to decrease the rate of family-wise error [69]. However, this may have reduced the overall variability within our dataset [70]. Finally, although EMIS data are the latest available data on gbMSM across Europe, the survey was undertaken prior to the global COVID-19 pandemic. Perhaps many socio-environmental changes have occurred in the intervening period, which may warrant further research.

4.3. Public health implications

Although gbMSM would likely derive benefit from population-based interventions to curtail the harms associated with drug use, these approaches could be complemented by more targeted interventions, and our study has identified potential subgroups of gbMSM who may be at higher odds of engaging in these drug use practices. For example, those who reside in Dublin appear to be at greater odds of RDU and SDU. Peer outreach workers may play an important role in providing direct information to under-reached individuals about safer use and recovery supports, or may signpost these men to available support services [71].

Additional targeted measures, such as brief intervention training for non-specialist healthcare professionals who work closely with the gbMSM community, may also be warranted. Brief interventions are an integral part of the Irish Health Service’s Making Every Contact Count (MECC) initiative, and can easily be conducted in many different clinical settings [72]. Harm reduction campaigns may also be of benefit, whereby a concerted focus is placed on reducing the adverse consequences that can be associated with drug use. For example, men who had a recent diagnosis of a bacterial STI and/or who engage in CAI appear to be at higher odds of RDU and SDU. Although it is beneficial that these men have presented for testing, perhaps more efforts should be made to support healthier drug use choices among this cohort (such as emphasising the importance of condom utilisation), given that these men are already engaging with healthcare services. This approach may be particularly effective in reducing drug-related harms amongst those who would not otherwise seek out substance use services [73].