Effective delivery of treatments requires clear procedural details of the essential elements of treatment. Hence, if a systematic review finds that a class of interventions is effective, then the users of the review will want to know: ?Which version of the intervention should I use?? Current methods to guide selection or synthesis from the variations of a treatment used across trials in a systematic review are poorly developed, and absent from most instructions on systematic review methods. We identify three basic approaches: (i) single-trial-based choice, where criteria such as feasibility, cost, effectiveness, or familiarity guide which trial s treatment to adopt; (ii) common components hybrid, which extracts then combines?based on frequency and importance?components of several trials; and (iii) model-guided synthesis, where a model of the mechanisms of effect is used to code and assess the importance of components for the version(s) recommended. Whichever method is used, we suggest review authors provide an ?intervention options table?, which describes the pros and cons of some intervention alternatives used in an individual trial or set of trials. If clinicians and policymakers are to be expected to base their practices on the results of systematic reviews in practice, these three approaches will need to be more widely adopted.