Using secondary data in statistical analysis

This chapter discusses the general concepts and potential limitations of meta-analysis. A meta-analysis is performed by applying this technique to studies examining the use of anti-inflammatory therapies in sepsis. Although these methods may be used for both observational and experimental studies, the discussion is restricted to the analysis of clinical trials. Meta-analysis has evolved as a technique useful for summarizing a large number of clinical trials and for resolving discrepancies raised by these trials. There are many similarities between randomized controlled trials and meta-analyses. Both randomized trials and meta-analyses are designed to answer a scientifically valid question. Likewise, both techniques require that the patients or studies included, the data collected, and the analysis performed be prospectively planned, and that the results obtained be analyzed for factors that may potentially interact with treatment effect. Finally, both techniques deal with populations, not with single individuals. Thus, clinicians must use discretion when applying the conclusions derived from both these techniques to the individual patient.