Figure Funnel plot for all 22 publications. The two lines are parallel however the distribution is subjectively asymmetric, suggesting possible publication bias, p=0.11. Generated by code from Deeks et al (36)