Empirical Evidence of Study Design Biases in Randomized Trials: Systematic Review of Meta-Epidemiological Studies

Certain characteristics of randomized trials may exaggerate intervention effect estimates. The average bias appears to be greatest in trials of subjective outcomes. More research on several characteristics, particularly attrition and selective reporting, is needed.