“…Nine of the included studies (31.0%) described the appropriate method of random sequence generation such as a random number table [25,26,32] or computer-generated randomisation sequence [29,35,41,42,46,50] and only six studies (20.7%) adequately concealed group allocation [29,34,41,45,46,50]. For blinding of participants and personnel, 15 studies which tested herbal medicine as an adjunct to conventional medication against conventional medication alone were considered as having a high risk of bias [25, 27, 30, 32, 33, 36-40, 43, 44, 46, 49, 52] while all the placebo-controlled studies were given a low risk of bias except one study using a placebo not identical to verum [47].…”