Abstract
Evidence-based medicine rests on the assumption of a hierarchy of evidence. The highest level of evidence is assumed to come from meta-analyses of randomised clinical trials.
Five illustrative recent meta-analyses of psycho-social treatment for substance dependence are discussed (Prendergast et al. 2002; Irvin et al. 1999; SBU 2001; Griffith et al. 2000; Pearson & Lipton 1999). In all cases the methodological strengths and weaknesses of the study are discussed, as well as the clinical and theoretical implications of the analysis. Studies that attempt to address very broad questions such as “does treatment work” produce conclusions that appear inconsistent and may be flawed, and the clinical and theoretical implications of the findings of such meta-analyses are unclear. In contrast, analyses of specific treatments produce findings that are clinically and theoretically meaningful. In cases where the populations or settings studied are diverse, but the type of treatment is similar, well-conducted meta-analyses may produce significant data on the degree to which the apparent effectiveness of a specific treatment can be generalised to new populations or settings.
It is concluded that the efficacy of specific classes of interventions for drug or alcohol use disorders may be a relevant target for meta-analyses, whereas meta-analyses that attempt to do too much by asking very global questions are unproductive, and that although meta-analyses are generally effective at addressing the question of whether a given treatment has support in the existing research, they are not able to answer questions about what should be the means and ends of drug policy and drug treatment at a more general level.