With the present emphasis on animal safety data, human benefit and risk may be inadequately quantified at the clinical trial stage and the assessment of risk from adverse reaction reporting is also based on methods that lack epidemiological precision. Under these conditions there must be serious doubts about the accuracy of assessment of some of the new drugs that come under suspicion. Suggestions have been made for improving the situation. A closer scrutiny of human data at the clinical trial stage, while involving some increase in the requirements for clinical trial data, would be advantageous for the pharmaceutical industry as well as for other interested parties if it were part of a deliberate change of emphasis from animal to human evidence of safety and from absolute to relative efficacy data in the assessment of product licenses.