A Systematic Review of Medical Research
* Decide which studies to include in the review. The first and most important decision is to decide what research question will be answered. This is expressed as an aim, for example: 'To assess the effects of a new drug for a particular health problem in certain types of people.'
* Decide which studies to include in the review. This will be decided in part by the tightly defined research question, but there are further 'eligibility criteria' that must be decided in advance for the inclusion or exclusion of studies. Typically, all the studies need to be of a rigorous design, such as a randomized controlled trial.
* Search for the studies. The sources consulted, and the search terms used to trawl them are outlined for attack - in the case of Cochrane reviews, by specially trained search coordinators. Attempts should also be made in the search to reveal unpublished studies.
* Select the studies and collect data from them. Data are extracted from the studies that meet the eligibility criteria set out in advance. The data relevant to the research question being investigated may have to be extracted from varied formats.
* Assess the risk of bias in the included studies. This is a process to ensure that all the studies reviewed are relevant and reliable. For example, was the randomization in the trial double-blinded? Or was there a risk of bias over which participants were selected for either treatment or comparison? Some studies may be of a lower quality, but they can still be included, so long as this assessment of bias is taken into account.
* Analyze the data and undertake meta-analyses. This is the core process of a systematic review, the main step towards synthesizing conclusions. Researchers are warned not to jump straight into this process of data analysis without first defining the aims and taking the precautions in the above steps.
* Address any publication bias. One of the disadvantages of systematic reviews is publication bias, which effectively means study results being cherry-picked. The risk is that a systematic review misrepresents the true effects of treatment, so this step is designed to address that concern.
* Present the final results of the review.