32 Chapter 2 user groups, but may not necessarily reflect the experience of all GHB users. It therefore remains to be elucidated whether the identified GHB-using populations in the current literature are indeed specific sub-populations of people using GHB. As all studies are cross sectional, it remains unclear to what extent people using GHB shift from one group to the other over time, and who might be more resilient or vulnerable for a transition from recreational GHB use to GUD. Third, our aim was to provide an overview of available studies on people using GHB. Studies from 10 years ago (45%) could be less relevant for today’s policy. However, except from a shift in focus to GUD after 2010, we did not find substantial differences in the GHB literature over time. Our selection criteria of English articles resulted in a possible overrep- resentation of studies carried out in the US (23%), Australia (18%), and parts of Western Europe (51% in particular the UK and the Netherlands). We did not specify ethnicity within studies, as most participants were white/European and none of the included studies made comparisons between different ethnicity. Both reduces the generalizability due to a risk of bias towards specific countries and sub-populations (e.g. Spanish-language countries). Various studies about MSM using GHB were not included in this review, as those studies did not report sufficient sociodemographic data, or GHB use was not distinguished from other drugs. For future research longitudinal studies should provide better insight in patterns and changes over time in GHB use, co-substance use, experienced comas, reasons to use, place of use, dependence diagnoses, psychiatric co-morbidity and social situation. Furthermore, population-based probability sampling strategies are advised, selecting predefined target groups (e.g. people with certain frequency of GHB use, sexual minorities, ethnic groups, specific age groups, or those with low/high social economic status), to allow for clear generalizability to both the target population and its sociodemographic subpopulations. Population-based probability sampling is still prohibitively costly and la- bor-intensive, but less compare to probability sampling without stratification and or clustering (Bornstein, Jager, & Putnick, 2013). When researchers are limited to convenience samples, homogeneous convenience samples are advised, e.g. with respect to one or more sociodemographic groups, as an alternative to conventional convenience samples (Jager et al., 2017). This limits ‘noise’ related to variation in subsamples (Bornstein et al., 2013). As a meta-analysis on existing data was not feasible due to different definitions and lack of sociodemographic information, we recommend the development of an international ‘standard’ protocol proposing standardized definitions related to GHB use, which will allow comparing data in the future. Furthermore, we would like to encourage researchers to make results from non-English speaking countries available.