From science to practice: meta-analyses using individual patient data are needed

The differences in results of meta-analysis using individual patient data and data available from literature should not cause people to discount the merits of meta-analysis, but results should be evaluated carefully. Publication bias should be considered along with the beneficial outcomes that seem too good to be true. Treatments that show promising results are more likely to be published than those showing negative results. Meta-analysis may be more prone to statistical errors when based on small trials.Inflated results of treatments are often reported long before contradictory information. Clinical decisions are best made when based on a whole body of data.