On the scientific inference from clinical trials