The question

There are many types of intervention that may be the subject of a systematic review, such as:

• therapy for a specific disease (e.g. aspirin to prevent stroke, surgery for coronary artery disease or cognitive therapy for depression);

• a change in a risk factor (e.g. blood pressure-lowering to prevent stroke, immunization to prevent hepatitis or publicity campaigns to reduce teenage smoking); or

• screening for earlier diagnosis (e.g. mammographic screening for breast cancer, antenatal screening for silent urinary tract infections or screening for cholesterol).

The defining feature is that some specific activity is undertaken with the aim of improving or preventing adverse health outcomes.

Study design

Because of their unique ability to control for confounders, known or unknown, randomized controlled trials (RCTs) generally provide the best evidence of efficacy for interventions. This section therefore focuses on systematic reviews of controlled trials; other study types for intervention will be discussed Section 9.

However, in interpreting RCTs for policy making and applying them to individuals, nontrial evidence will often be appropriate. For example, surveillance data may provide the best information on rare adverse effects, and cohort studies may provide the best information on the prognostic factors needed to predict the pretreatment risk of an individual.

Study design

Finding existing systematic reviews

Appendix A gives information on finding existing systematic reviews. A check should be made of the Cochrane Database of Systematic Reviews (CDSR; Cochrane Library) and DARE databases for Cochrane and nonCochrane reviews respectively. Even if the review is not considered to be completely appropriate, its reference list will provide a useful starting point.