Figure 1

Classification of RCT results based on 95% confidence interval and minimum important difference from the trial sample size calculation (d). 1. Explanation of the labels: The label given to each result are the labels used by Djulbegovic et al [8]. The labels assigned to the two inconclusive categories have been slightly amended from that used by Djulbeovic.

2. Explanation of -d and d: Djulbegovic's method used -d = 0.8 and d = 1.2 for all trials reviewed because all outcomes they evaluated were binary and expressed as a hazard ratio, odds ratio or relative risk. The value used for -d and d in this review was the minimum important difference specified in the sample size calculation from each trial. For example if the sample size calculation for a trial indicated an increase of 5 points was the minimum important difference for the primary outcome then -d was equal to -5 and d was equal to 5.

3. If the primary outcome result was expressed as a Hazard Ratio, Relative Risk or Odds Ratio then zero in this figure is replaced with 1