Model validity and risk of bias in randomised placebo-controlled trials of individualised homeopathic treatment