Publication bias and clinical trials