Introduction: There is considerable debate in the Literature, both scientific and lay, about the overall risk:benefit ratio of benzodiazepines, commonly used medications for insomnia and anxiety.

Objective: We evaluated the feasibility of meta-analysis of the literature on therapeutic use, harm, and withdrawal of benzodiazepines (bzd).

Methods: A complete English language MEDLINE search using the terms "exp benzodiazepine" or individual bzd name (all drugs on Canadian market); "drug therapy (sh)"; and "randomized controlled trial" (RCT) or "all random:'. Reference lists were reviewed and pharmaceutical companies were solicited for relevant citations. Data extraction and quality assessment were done according to standard McMaster criteria.

Results: Over 545 randomized controlled trials on the 16 bzd available in Canada were retrieved for many indications including anxiety (n=12,036 patients) and insomnia (n=3,928). Significant problems in the reporting of individual trials were noted, including a) inadequate documentation of psychiatric diagnosis(es), b) exclusion of key groups, c) dose inequivalence of compared therapies, d) incomplete reporting of results with frequently missing measure of variability of results, e) the use of a variety of subjective endpoints as outcome measures, f) extremely short durations of therapy with infrequent reporting of withdrawal effects, and g) imprecise attribution of adverse effects. "One-armed" meta-analysis has been carried out to summarize occurrence of adverse effects and the impact of withdrawal from benzodiazepines.

Discussion: Despite a large number of relatively small RCTs on benzodiazepine therapy, the poor quality of reporting of these trials severely restricts the application of meta-analysis. We wish to discuss the relative merit of various salvage techniques.