The "evidence" in EBM must be of high quality in order to be useful, but this is not always the case. Even the "gold standard" of evidence-based medicine, the randomised clinical trial, is bedevilled by low inclusion rates and potentially important recruitment biases. "Real world" trials often do not give the same results as these highly artificial controlled clinical studies. Meta-analysis, the next most important level of evidence in EBM, may be unreliable, sometimes giving different results to subsequent large randomised trials. There is a bias in the hypotheses tested in large clinical trials, as the costs involved are usually covered by commercially interested companies. For this reason, trials of non-patentable compounds or therapies of no commercial interest may not be performed. The process of journal review and publication is capricious, slow and may have a selection bias towards positive studies, meaning that communication channels for the "evidence" are often unsatisfactory. For many rarer conditions and situations, there is simply no "high level" evidence, such as in paediatrics and subspecialty surgery.