“…We assessed articles according to their way of generating comparable groups and found several randomized controlled trials 24,47,54,56,57 and 6 2-phase cohort trials 3,22,35,48,51,58 to have a high risk of bias. Because of a significant percentage of patients lost to follow-up, we judged 14 articles 3,15,21,25,30,34,37,[44][45][46]52,54,58,59 to have a high risk of incomplete data. Twenty articles did not describe the level of education 6,15,17,20-22,24, 27,29,35,39,44,46,48,52,53,55,58-60 and 9 articles did not mention whether there was a language barrier 12,15,23,25,35,43,44,51,60 and therefore probably had selection bias.…”