“…Initial assessment of the uniformity of the included studies using meta analytic software (RevMan) (Collaboration, 2003) revealed significant heterogeneity between a sub-set of four eligible case-control studies (Gjerde et al, 2013;Kuypers et al, 2012;Laumon et al, 2005;Movig et al, 2004) (Chi 2 =12.00, df=3 (p=0.007); I 2 =75%). Assessment of heterogeneity of the remaining four case-control studies (Bernhoft et al, 2012;Bogstrand et al, 2015;Gadegbeku et al, 2011;Gjerde et al, 2011) was not performed due to missing information (such as specific exposure rates between cases and controls for amphetamines), and assessment of study heterogeneity of the n=1 cohort study was not appropriate (Poulsen et al, 2014). As a result, we decided to apply a 'best evidence synthesis' approach to all case-control and cohort studies, which comprised a modified version of that previously outlined by Lievense and colleagues (see Table 1) (Lievense et al, 2001).…”