Faulty assumptions in Atkinson's criteria for clinical trial design