Introduction: In some areas, such as is the case with interventions for drug and alcohol dependency, there are not enough RCTs and CCTs to conduct meta-analysis. In these cases it may be important to consider observational studies for inclusion in systematic reviews, although there is always the risk of introducing bias.

Objectives: The aim of this paper is to present and discuss guidelines for selection and inclusion of observational studies within a review.

Results: Several non-experimental designs are of interest for the Drug and Alcohol Group: i) cohort studies, where outcomes of interventions are measured at the individual level after follow-up; ii) controlled before after studies, where interventions are offered at a group level and the outcomes are measured at the individual level, and iii) interrupted time series, where both the allocation of the interventions and the outcome measures are carried out at a group level. Case-control studies are not included because they are rarely conducted in the drug and alcohol field. For each of these study designs a check list is proposed for quality appraisal based on the classification of major risks of bias.

Discussion: An instrument is presented to help reviewers in areas where RCTs are not possible or very difficult to be conducted (e.g. primary prevention) to include non-experimental studies which limits the risk of producing biased results.