The design of more than 90% of around 2000 randomised controlled trials published in Chinese medical journals was flawed, concludes a review.

The researchers searched the China national infrastructure electronic database for reports of randomised controlled trials on 20 common diseases published between 1994 and 2005. The original authors of 2235 of these trials were then interviewed about the methods used to randomise participants. Only 207 of the studies used accepted randomisation methods.

The reviewers considered that a randomisation sequence generated from a random number table, calculator, or computerised random number generator was authentic but that tossing a coin, drawing straws, or allocating a participant according to date of …