Systematic reviews and their roles in promoting evidence-based medicine in renal disease

Abstract

How to obtain definite information from incomplete information A systematic review of treatment proposed for a specific disease should be based on an exhaustive literature search, which should identify all randomized clinical trials (RCTs) relevant to the selected topic. Controlled non randomized studies may also be included in situations where random allocation of treatment is deemed to be difficult or impossible, or when no such trials exist. If enough trials are found and these report conflicting results, a meta-analysis is one way of assessing the efficacy of the treatment. Meta-analysis is a clinical research tool allowing the pooling of results from different clinical trials evaluating comparable treatments for a given condition. This is particularly useful when the number of patients is low, or the observed event rate is low. Meta-analyses can reduce the size of the confidence interval from the point estimate for the treatment effect. Thus if the treatment effect in each of three small trials shows a non-significant trend to efficacy, the pooling of the trials may result in a significant trend to efficacy. The selection of trials to be pooled, however, is an important step in a process that includes a rigorous statistical analysis (i.e. Mantel and Haenszel method). The choice of the outcome to be analysed should be based on strong epidemiological data or related to a clinically relevant question. Outcomes should be 'hard' objective events that are easily observed by non-specialists, such