Figure 2.

The risk of bias for individual trials. D1: Bias arising from the randomization process. D1a (for cluster-randomized designs): Bias arising from the randomization process. D1b (for cluster-randomized designs): Bias arising from the timing of identification or recruitment of participants. D2: Bias due to deviations from intended intervention. D3: Bias due to missing outcome data. D4: Bias in measurement of the outcome. D5: Bias in selection of the reported result. (X) high risk of bias; (−) some concerns; (+) low risk of bias.