Assessing risk/benefit for trials using preclinical evidence: a proposal