Abstract
Randomized controlled trial (RCT) can provide high grade evidence, which might determine the benefit of a therapy for patients. Recently, the number of published RCT is increasing in critical care medicine. Occasionally, more than one RCT are conducted on the same therapy and show different results (e.g. positive and negative). A possible explanation for the conflicting results is that a positive trial is conducted and assessed in an environment with a higher chance of false positive results. To read RCT critically, one need to understand three important factors that make the probability of false positive results increased. The first is subgroup analysis, which increases the number of comparisons and leads to selection biases. The results from subgroup analysis should be confirmed with another RCT in this subgroup. The second is single center open label studies, which cause the Hawthorne effect and result in less generalizability. The results from such trials should be confirmed by multicenter RCT. The third is early termination for benefit, which makes the chance of α error increased. We should take a careful criticism of the results, when a RCT contains at least one of these three factors.
