Why statistical inference from clinical trials is likely to generate false and and irreproducible results